Introduction
The journey through these pages begins with a simple idea. Upon going once around a circle of radius R R R , a field need not return to itself, but may return to its image,
Φ ( y + 2 π R ) = U Φ ( y ) , (1) \Phi(y + 2\pi R) = U\Phi(y), \tag*{(1)} Φ ( y + 2 π R ) = U Φ ( y ) , (1)
under a transformation U U U . Applied in increasingly elaborate theories, this simple idea has many ramifications, which have taken decades to explore.
Before turning to the story, let us first set the stage and introduce its main characters. The first ingredient is extra dimensions. Their use in attempts to relate gravity to other interactions goes back to the earliest developments of Kaluza–Klein theory [1, 2, 3, 4] (see also Ref. [5] for a historical discussion). Many of the explicit examples considered in this review will involve the simplest internal spaces: the circle and the interval. They are sufficient to illustrate many of the main ideas discussed below, while remaining simple enough that several calculations can be carried through explicitly.
The second ingredient is supersymmetry. Supersymmetric theories have an undeniable appeal to a high-energy theorist. Nolens volens , the question of supersymmetry breaking then imposes itself upon us: first, if such theories are ever to describe our world, and second, out of intellectual curiosity about what becomes of their remarkable structures once supersymmetry is broken.
The third ingredient is string theory, our best-developed framework for a quantum theory of gravity, in which both extra dimensions and supersymmetry arise naturally. Familiar as these ingredients have become in theoretical physics, neither has so far received experimental support. Our discussion will therefore remain deliberately theoretical. In particular, we shall largely refrain from phenomenological considerations whose details are inevitably more sensitive to the passage of time.
If U U U acts differently on bosons and fermions, then Eq. (1) may break supersymmetry. But can one choose U U U arbitrarily? What properties must it satisfy? How does supersymmetry breaking manifest itself, and in what sense is it spontaneous? What scalar potential does it induce? What can be said about the gravitino mass? Do all states feel supersymmetry breaking already at the classical level? If not, how is the breaking transmitted to them by radiative corrections? And is there any obstruction to embedding the whole construction into string theory?
These are among the questions this review tries to answer. The novice reader will already have guessed, from these questions and from the title of the review, that the mechanism we shall be discussing is what is commonly called Scherk–Schwarz compactification [6, 7]. In their original construction, symmetries of the higher-dimensional theory were used to generate masses through a non-trivial dimensional reduction and, when the transformation acts differently on the members of a supersymmetry multiplet, to break supersymmetry spontaneously. Early field-theoretic implementations in higher-dimensional supersymmetric theories, including matter and gauge multiplets and the resulting supersymmetry-breaking mass spectra, were developed in detail by Fayet [8, 9].
String theory adds another layer to the problem. If supersymmetry is unbroken at tree level, ordinary perturbative radiative corrections do not provide a mechanism for breaking it, as follows from the non-renormalization properties of supersymmetric string vacua [10]. At the same time, classical supersymmetric and non-supersymmetric string vacua cannot in general be connected by an arbitrarily small deformation at finite distance in moduli space [11, 12]. The alternatives are therefore sharply constrained: supersymmetry breaking must already be present in the classical string background, or arise from genuinely non-perturbative physics.
This made Scherk–Schwarz breaking particularly attractive in perturbative string theory, since it provides a tree-level and calculable realization of supersymmetry breaking through the global compactification data. Starting with the early string constructions [13, 14, 15, 16], this possibility was explored in increasingly explicit settings. At the same time, string theory made clear that the apparent freedom of the field-theory construction is reduced by microscopic consistency conditions. In particular, a small gravitino mass in perturbative string constructions is tied to a tower of light states and to a decompactification limit rather than to an arbitrary small deformation of a compact vacuum [17]. These observations will reappear later in the review in a form that separates the rigid monodromy from the physical scale associated with it.
How this review is organized
There are many ways of writing a review on a subject such as this one. I have chosen not to follow the most traditional route, in which a sequence of chapters surveys and summarizes the literature on the successive developments of the subject. Instead, I have adopted a more personal approach.
I went back to several of my own papers and asked myself which questions had led me to them, and what questions a reader would have to understand in order to grasp them in depth. I then tried to write the corresponding questions and answers in the form in which I would have liked to find them in the literature. My hope is that the result will be useful to a newcomer entering any of the subjects discussed here.
One consequence of this approach is that the review is, inevitably, rather personal. A few of my own works appear repeatedly along the way. They are not, however, the subject of the review, and the reader will quickly see that they constitute only a small fraction of the material covered. For me they have served mainly as guides: to understand them, a student has to acquire a number of notions, both explicit and implicit, and the discussion has been developed outward from there.
This choice has drawbacks. The treatment of the bibliography is necessarily partial and is biased by the way in which each question is posed. Many detours that would have made the bibliography more complete have deliberately been avoided because they would also have changed the nature of the review.
The scope of the review is similarly limited. It is neither a historical account nor an exhaustive survey of Scherk–Schwarz constructions. In particular, I shall not attempt a systematic catalogue of heterotic, type II, or orientifold constructions, nor a systematic review of flux compactifications or gauged supergravities in their own right, nor a review of supersymmetry phenomenology. Such constructions will enter only when they illuminate one of the questions addressed here. The aim is to understand the mechanism and the structures underlying it, rather than to cover every context in which it has been used.
For the same reason, the discussion is not organized around a catalogue of currently open problems. This is largely intentional. I have tried to limit discussions whose content is particularly likely to become obsolete as the subject develops, and to concentrate instead on questions and structures that I hope will remain useful for longer.
The review starts at a deliberately elementary level, but the technical detail increases as the discussion progresses and becomes, in a few places, rather specialized. Roughly the first two thirds can be read entirely within quantum field theory and supergravity, while the final third brings string theory increasingly to the foreground.
Although the questions are arranged in a definite progression, individual questions, or small groups of questions dealing with closely related issues, have been written so as to remain reasonably self-contained. A reader interested in a particular aspect of the subject should therefore be able to enter the review at several different points without having to read everything that precedes it. This partial independence is also one reason why some notation and arguments are occasionally recalled.
A certain discipline has guided the questions and answers. Every notion should first be accessible at an intuitive level. The review is intended for physicists with rather different backgrounds, although familiarity with quantum field theory, supergravity, and the basic ingredients of string theory is assumed. Concepts that are more specific to the subject are introduced and defined when they first become necessary.
Whenever possible, an abstract statement is accompanied by an explicit example or calculation. Quite often, however, the answer admits a more general formulation once the appropriate mathematical language is introduced.1 The purpose is not to pursue an elegant formalism for its own sake. The hope is rather that the formalism provides a broader view of the question and makes it easier for a newcomer to recognize the same structures in related subjects and in parts of the literature not discussed explicitly here. This choice sometimes makes the exposition heavier and introduces a certain amount of repetition; I hope that the price is worth paying.
The discussion follows a corresponding progression. We begin with fields on a circle and an interval, where the kinematics of the twist can be made fully explicit. We then ask how the same data act on supersymmetry, how they enter supergravity, and how the resulting breaking is transmitted through quantum corrections. The last part of the review asks which of these effective-field-theory constructions survive the additional consistency conditions of string theory and how the same ideas are realized when a simple classical internal geometry is no longer the most useful description.
The primary purpose of a review is not to present new calculations or new results in the same way as a research paper. I have nevertheless tried to avoid producing a catalogue of summaries and, whenever possible, to bring something original to the discussion. Sometimes this lies mainly in the way the material is organized or explained; sometimes it lies in a different way of looking at familiar physical results. Some formulations may be less familiar, and a few may not have appeared in quite this form in the literature. I make no claim of priority for these observations; their purpose is to clarify the underlying structures.
Let me end this introduction by trying to answer a more personal question: why write this review now?
The first answer is very simple: I was invited to write a review, and if one is going to write one, why not this one? I first worked on Scherk–Schwarz compactifications about thirty years ago, and in a sense I began writing this review more than twenty years ago. On several occasions I intended to write a review or a broader account, with the scope oscillating between a general introduction to theories with extra dimensions and a more focused discussion of Scherk–Schwarz compactification. None of these projects was ever completed.
Over the years, however, a number of ideas, questions, and remarks accumulated and remained, so to speak, in the drawers of my mind. Some of them resurfaced when I prepared a workshop talk in Corfu; more recently, others found their way into a paper. At that point I thought that I had enough notes, and perhaps a sufficiently broad view of the subject, to write the review. I had underestimated the challenge.
What happens to fields in higher-dimensional spaces when they move along a circle?
Essential idea
On a circle, the points labelled by y y y and y + 2 π R y+2\pi R y + 2 π R are not distinct: they represent the same point of the compact space. Requiring all fields to be strictly periodic is the simplest way to implement this identification, but it is not the only one. A field may instead return after one circuit of the circle to its image under a transformation U U U acting on the field space. This gluing transformation is the monodromy and provides the basic geometric ingredient of a Scherk–Schwarz compactification.
Periodicity and compactification
Let us consider a field theory on a spacetime
M d × S 1 , (2) \mathcal{M}_d \times S^1, \tag*{(2)} M d × S 1 , (2)
where M d \mathcal{M}_d M d is d d d -dimensional Minkowski space and S 1 S^1 S 1 is a one-dimensional compact manifold. Let y y y denote a coordinate on the covering space R \mathbb{R} R , with the identification
y ∼ y + 2 π R . (3) y \sim y+2\pi R. \tag*{(3)} y ∼ y + 2 π R . (3)
The simplest possibility is to regard a field Φ \Phi Φ as an ordinary single-valued function on S 1 S^1 S 1 , taking values in a fixed vector space. Its lift to the covering line then satisfies
Φ ( x , y + 2 π R ) = Φ ( x , y ) . (4) \Phi(x,y+2\pi R)=\Phi(x,y). \tag*{(4)} Φ ( x , y + 2 π R ) = Φ ( x , y ) . (4)
For a scalar field this gives the Fourier expansion
Φ ( x , y ) = ∑ n ∈ Z Φ n ( x ) e i n y / R , (5) \Phi(x,y)=\sum_{n\in\mathbb{Z}}\Phi_n(x)e^{iny/R}, \tag*{(5)} Φ ( x , y ) = n ∈ Z ∑ Φ n ( x ) e in y / R , (5)
and hence the familiar tower of compact momenta
p y = n R , n ∈ Z . (6) p_y=\frac{n}{R},\qquad n\in\mathbb{Z}. \tag*{(6)} p y = R n , n ∈ Z . (6)
Historically, Kaluza–Klein theories are distinguished by the emergence of lower-dimensional gauge interactions from higher-dimensional gravity [1, 2, 3, 4]; see also Ref. [5] for a historical discussion. Following standard usage in high-energy physics, we shall nevertheless refer loosely to the modes generated by compactification as KK states, even when gravity plays no role. Strict periodicity gives the simplest KK spectrum, with momenta separated by integer units of 1 / R 1/R 1/ R . This result, however, already contains an additional assumption: the fibre above y y y has been identified with the fibre above y + 2 π R y+2\pi R y + 2 π R by the identity transformation. Compactification identifies the points of the base space, but it does not by itself require this particular identification of the field fibres.
For visualization, the value of Φ \Phi Φ in each fibre will be represented in Figure 1 by an arrow. This should not be understood as implying that Φ \Phi Φ is a vector field: the arrow is merely a graphical representation of an element of the fibre, on which the gluing transformation U U U acts.
Figure 1. Compactifying on a circle identifies y y y and y + 2 π R y + 2\pi R y + 2 π R ; the interval is only a convenient way of drawing S 1 S^1 S 1 (insets). The arrows representing Φ \Phi Φ are schematic elements of the fibre, not spacetime vectors. In (a) the fibre is glued to itself by the identity, U = 1 U = 1 U = 1 , whereas in (b) it is glued by a non-trivial monodromy U U U .
The more general global identification
A more general description regards a field as a section of a fibre bundle over the compact space. Imagine cutting the circle at one point and representing it as the interval 0 ≤ y ≤ 2 π R 0 \le y \le2\pi R 0 ≤ y ≤ 2 π R . Reconstructing the circle requires us to identify the two endpoints, but the corresponding fibres need not be glued by the identity map. They may instead be related by a transformation U U U ,
Φ ( x , 2 π R ) = U Φ ( x , 0 ) . (7) \Phi(x, 2\pi R) = U\Phi(x, 0). \tag*{(7)} Φ ( x , 2 π R ) = U Φ ( x , 0 ) . (7)
As we shall discuss in the next question, consistency requires U U U to be an admissible symmetry transformation of the higher-dimensional theory. Unrolling the circle onto its covering space gives the equivalent condition
Φ ( x , y + 2 π R ) = U Φ ( x , y ) . (8) \Phi(x, y + 2\pi R) = U\Phi(x, y). \tag*{(8)} Φ ( x , y + 2 π R ) = U Φ ( x , y ) . (8)
For U = 1 U = 1 U = 1 one recovers strict periodicity. For U ≠ 1 U \ne1 U = 1 the lifted field on R \mathbb{R} R is no longer periodic in the naive sense, even though it defines a perfectly good global field on the compact space once the fibres are glued by U U U . The relevant global datum is therefore not the periodicity of a representative on the covering line, but the monodromy specifying how the field is identified after one circuit of the circle.
This distinction between the identification of points in the base space and the gluing of the fibres above them is the elementary geometric ingredient of a Scherk–Schwarz compactification [6, 7]. Much of what follows in this review will consist of determining which transformations U U U are allowed, understanding how the same U U U acts in different representations, and tracing the physical consequences of that action.
Monodromy and the fundamental group
The same statement can be formulated intrinsically in terms of the fundamental group. The group π 1 ( S 1 ) \pi_1(S^1) π 1 ( S 1 ) consists of equivalence classes of closed loops in S 1 S^1 S 1 , two loops being equivalent when one can be continuously deformed into the other while remaining on the circle. Composition is defined by concatenation of loops, and for the circle
π 1 ( S 1 ) ≃ Z , (9) \pi_1(S^1) \simeq\mathbb{Z}, \tag*{(9)} π 1 ( S 1 ) ≃ Z , (9)
where the integer labels the algebraic winding number around the circle. There is therefore a single independent generator, and specifying the monodromy associated with this generator determines the gluing around every loop.
More generally, for a compact space K K K , flat monodromy data can be described by a homomorphism
ρ : π 1 ( K ) ⟶ G , (10) \rho: \pi_1(K) \longrightarrow G, \tag*{(10)} ρ : π 1 ( K ) ⟶ G , (10)
where G G G acts on the fibres in the representation appropriate to the field under consideration. The circle is the simplest possible example: π 1 ( S 1 ) \pi_1(S^1) π 1 ( S 1 ) has one generator, so the entire global datum is encoded in one transformation U = ρ ( 1 ) U = \rho(1) U = ρ ( 1 ) . Which such homomorphisms define physically distinct backgrounds, and in particular the role played by conjugation in G G G , will be discussed in the next question.
A familiar example of non-trivial global boundary conditions is provided by spinors. On S 1 S^1 S 1 there are two spin structures, corresponding to
ψ ( y + 2 π R ) = + ψ ( y ) , (11) \psi(y + 2\pi R) = +\psi(y), \tag*{(11)} ψ ( y + 2 π R ) = + ψ ( y ) , (11)
and
ψ ( y + 2 π R ) = − ψ ( y ) . (12) \psi(y + 2\pi R) = -\psi(y). \tag*{(12)} ψ ( y + 2 π R ) = − ψ ( y ) . (12)
This elementary example already shows that the geometry of the base space does not uniquely determine the global identification of every field. The precise relations between spin structures, spacetime fermion parity ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s , and genuine R R R -symmetry twists are subtler and will be discussed later in the review.
The higher-dimensional origins of Kaluza–Klein compactification are represented here by the original works of Kaluza, Klein, and Einstein–Bergmann [1, 2, 3, 4]; Ref. [5] provides a useful modern discussion of part of this early history. The generalized dimensional-reduction construction based on a non-trivial gluing by a symmetry transformation was developed by Scherk and Schwarz [6, 7]. For the mathematical language of fibre bundles, transition functions, and gluing data, a standard reference is Husemoller [18], while spin structures and spin geometry are treated systematically by Lawson and Michelsohn [19]. We shall use these mathematical notions only to the extent needed to formulate the physical questions that arise in the compactification.
Take-home message
Fields on a compact space are naturally regarded as sections of fibre bundles. Identifying points of the base space does not by itself require the fibres to be identified trivially.
On a circle, strict periodicity corresponds to trivial gluing and gives the compact momenta p y = n / R p_y = n/R p y = n / R .
More generally, a field may return to itself up to a monodromy U U U after going around a non-contractible cycle.
For a general compact space, flat monodromy data are encoded by the action of its fundamental group on the relevant field fibres.
This naturally leads to the next question:
Why must the twist be associated with a symmetry of the higher-dimensional theory?
Why must the twist be associated with a symmetry of the higher-dimensional theory?
Essential idea
Compactification identifies different representatives of the same point of the internal space. The gluing transformation must therefore assign the same local physics to all such representatives. In the standard field-theoretic Scherk–Schwarz construction, this requires the twist to act as a symmetry of the higher-dimensional theory.
Why is this condition necessary?
As discussed in the previous question, fields need not be strictly periodic on the covering space: one may impose
Φ ( x , y + 2 π R ) = U Φ ( x , y ) . (13) \Phi(x,y+2\pi R)=U\Phi(x,y). \tag*{(13)} Φ ( x , y + 2 π R ) = U Φ ( x , y ) . (13)
This freedom is not unlimited. The coordinates y y y and y + 2 π R y+2\pi R y + 2 π R represent the same point of the compact space, and the two representatives cannot be assigned different local dynamics. The restriction becomes particularly transparent if we deliberately try to violate it.
Consider, for example, a real scalar field with Lagrangian
L = 1 2 ( ∂ ϕ ) 2 − a [ 1 − cos ( ϕ f ) ] . (14) \mathcal{L}=\frac{1}{2}(\partial\phi)^2-a\left[1-\cos\left(\frac{\phi}{f}\right)\right]. \tag*{(14)} L = 2 1 ( ∂ ϕ ) 2 − a [ 1 − cos ( f ϕ ) ] . (14)
where 2 π f 2\pi f 2 π f is the periodicity of the scalar potential. Suppose that we try to impose the boundary condition
ϕ ( y + 2 π R ) = ϕ ( y ) + Δ . (15) \phi(y+2\pi R)=\phi(y)+\Delta. \tag*{(15)} ϕ ( y + 2 π R ) = ϕ ( y ) + Δ. (15)
The kinetic term is invariant under a constant shift, but the potential transforms according to
cos ( ϕ + Δ f ) = cos ( ϕ f + Δ f ) . (16) \cos\left(\frac{\phi+\Delta}{f}\right)=\cos\left(\frac{\phi}{f}+\frac{\Delta}{f}\right). \tag*{(16)} cos ( f ϕ + Δ ) = cos ( f ϕ + f Δ ) . (16)
For a generic value of Δ \Delta Δ , this is not the original potential. The two representatives of the same point would therefore be assigned different local dynamics. The gluing is consistent only when
Δ = 2 π k f , k ∈ Z , (17) \Delta=2\pi kf,\qquad k\in\mathbb{Z}, \tag*{(17)} Δ = 2 π k f , k ∈ Z , (17)
for which the shift is an actual symmetry of the theory. The example is elementary, but it displays the general principle: the compact identification may act non-trivially on the fields, but it cannot change the local theory as one goes around the circle.
The field-theoretic criterion
For an ordinary internal symmetry in a theory admitting a Lagrangian description, the simplest form of the condition is
L ( U Φ ) = L ( Φ ) . (18) \mathcal{L}(U\Phi)=\mathcal{L}(\Phi). \tag*{(18)} L ( U Φ ) = L ( Φ ) . (18)
More generally, it is sufficient that the action be invariant; the Lagrangian density may change by a total derivative, provided that this change is itself compatible with the global identification and gives no contribution to the action. Equivalently, one may formulate the requirement directly in terms of the equations of motion and the physical observables: U U U must map the higher-dimensional theory into itself. The boundary condition (2.1) can then identify two representatives of the same point without creating any local inconsistency.
At the quantum level, the same statement must hold for the full quantum theory. A classical transformation spoiled by an anomaly, for example, cannot in general be used as an exact Scherk–Schwarz twist. Thus the twist is not an arbitrary choice of quasi-periodic boundary condition: around a non-contractible one-cycle it is a monodromy valued in the appropriate symmetry group of the higher-dimensional theory. Which elements of that group act non-trivially on the supercharges is a separate question, to which we shall turn next.
Geometric interpretation
The geometric formulation expresses the same requirement in the language introduced in the previous question. A field is a section of a fibre bundle, and the transition functions specify how local descriptions of the fibre are related on overlapping patches. For a circle, the transition function associated with going once around the compact direction is precisely the monodromy U U U . A mathematically well-defined bundle can therefore be constructed with non-trivial gluing.
Mathematical consistency of the bundle alone, however, is not enough for a physical compactification. Its transition functions must also preserve the structures that define the higher-dimensional theory: its action, equations of motion, gauge structure, and, at the quantum level, its exact operator content. In the standard Scherk–Schwarz construction this is the physical statement that the monodromy takes values in an exact symmetry group G G G . On a circle one may therefore write
ρ : π 1 ( S 1 ) ≃ Z ⟶ G , (19) \rho:\pi_1(S^1)\simeq\mathbb{Z}\longrightarrow G, \tag*{(19)} ρ : π 1 ( S 1 ) ≃ Z ⟶ G , (19)
with U = ρ ( 1 ) U=\rho(1) U = ρ ( 1 ) . The abstract homomorphism introduced in the previous question thus acquires a direct physical meaning: it specifies how an allowed symmetry of the higher-dimensional theory is used to glue the fields globally around the compact direction.
Beyond ordinary symmetries
The preceding formulation can be extended beyond transformations acting locally on a conventional set of fields. What is fundamentally required is an exact automorphism of the complete theory. This distinction becomes important in string theory, where two apparently different local descriptions may be related by an exact duality, and such a duality can itself be used as part of the gluing. This leads to duality-twisted and non-geometric compactifications, in which no globally defined higher-dimensional field configuration of the conventional kind need exist.
Those possibilities will be encountered only much later in the review. For the field-theoretic Scherk–Schwarz construction developed in the next questions, the relevant twists will be ordinary symmetries of the higher-dimensional theory. Keeping this simpler setting for the moment will allow us to separate two logically distinct issues: first, whether a twist is admissible at all, and second, whether an admissible twist actually breaks supersymmetry.
The requirement that the non-trivial dependence on the compact coordinates be controlled by a symmetry of the higher-dimensional theory is already explicit in the original construction of Scherk and Schwarz [6, 7]. Their generalized dimensional reduction uses precisely such a symmetry to factor the dependence on the internal coordinates and to generate masses in the reduced theory. For the underlying bundle language and the role of transition functions, we refer again to the standard treatment in Ref. [18]. More general duality-valued gluings will be discussed, with their appropriate literature, when we turn to string compactifications.
Take-home message
A monodromy is not automatically an admissible Scherk–Schwarz twist. Since compactification identifies different representatives of the same spacetime point, the gluing must preserve the local higher-dimensional physics.
In the standard field-theoretic construction, the twist therefore acts as a symmetry of the higher-dimensional theory. For a Lagrangian invariant without boundary terms, this reduces to L ( U Φ ) = L ( Φ ) \mathcal{L}(U\Phi)=\mathcal{L}(\Phi) L ( U Φ ) = L ( Φ ) .
The symmetry must be exact at the level at which the compactification is defined. In particular, a classical symmetry obstructed quantum mechanically cannot in general define an exact quantum twist.
More generally, the gluing may be an exact automorphism of the complete theory, including suitable dualities in string theory.
This naturally leads to the next question:
Among the admissible twists, which ones break supersymmetry?
Among the admissible twists, which ones break supersymmetry?
Essential Idea
A twisted compactification preserves precisely those supersymmetry transformations whose parameters are globally compatible with the monodromy. For the flat compactifications considered here, the surviving supersymmetries are therefore the invariant subspace of the monodromy induced on the supersymmetry-parameter space.
This criterion tells us which supersymmetries survive the compactification. It does not determine whether their absence should be interpreted as spontaneous breaking. That is a separate dynamical question.
An admissible twist need not preserve all the original symmetries
The previous question determined when a twist is admissible: it must act as a symmetry, or more generally as an exact automorphism, of the higher-dimensional theory. This requirement does not imply that every symmetry of the parent theory survives after compactification. In particular, an admissible monodromy may act non-trivially on the supersymmetry generators. Admissibility and preservation are therefore two distinct questions: the first asks whether the twist can consistently be used to define the compactification, while the second asks which symmetries remain globally well defined once that twist has been imposed.
The supercharge invariant-subspace criterion
Consider a supersymmetric higher-dimensional theory with supersymmetry generators Q I Q^{I} Q I and corresponding spinorial transformation parameters ϵ I \epsilon^{I} ϵ I . Schematically, an infinitesimal supersymmetry transformation acts as
δ ϵ Φ ∼ ϵ ˉ I [ Q I , Φ ] , (20) \delta_{\epsilon}\Phi\sim\bar{\epsilon}_{I} [Q^{I},\Phi], \tag*{(20)} δ ϵ Φ ∼ ϵ ˉ I [ Q I , Φ ] , (20)
where the bracket is understood as graded whenever appropriate. Let S \mathcal{S} S denote the vector space of supersymmetry parameters. If the fields obey the twisted identification
Φ ( x , y + 2 π R ) = U Φ ( x , y ) , (21) \Phi(x,y+2\pi R)=U\Phi(x,y), \tag*{(21)} Φ ( x , y + 2 π R ) = U Φ ( x , y ) , (21)
the same transformation induces an action on S \mathcal{S} S , which we denote by R ϵ ( U ) R_{\epsilon}(U) R ϵ ( U ) .
For the flat compactification considered here, a supersymmetry of the lower-dimensional theory is generated by a parameter that can be chosen independent of the compact coordinate. Compatibility with the global identification then requires
R ϵ ( U ) ϵ = ϵ . (22) R_{\epsilon}(U)\,\epsilon=\epsilon. \tag*{(22)} R ϵ ( U ) ϵ = ϵ . (22)
The preserved supersymmetries are consequently
S p r e s e r v e d = ker ( R ϵ ( U ) − 1 ) . (23) S_{\mathrm{preserved}} = \ker(R_{\epsilon}(U)-1). \tag*{(23)} S preserved = ker ( R ϵ ( U ) − 1 ) . (23)
With several supersymmetry parameters, R ϵ ( U ) R_{\epsilon}(U) R ϵ ( U ) is simply a matrix acting on the full parameter space, and partial supersymmetry breaking corresponds to a non-zero invariant subspace of dimension smaller than dim S \dim S dim S .
This criterion isolates the global part of the problem. In a more general background, preserved supersymmetries must also satisfy the appropriate local Killing-spinor equations, and the monodromy condition alone need not be sufficient. For the flat backgrounds and flat twists that will occupy us first, however, the invariant-subspace criterion contains the relevant information.
When does a twist break supersymmetry?
If the twist acts trivially on all supersymmetry generators, then
R ϵ ( U ) = 1 , (24) R_{\epsilon}(U)=1, \tag*{(24)} R ϵ ( U ) = 1 , (24)
and all supersymmetries survive. Notice that this does not require the twist itself to be trivial: U U U may act non-trivially on other quantum numbers and modify the Kaluza–Klein spectrum while leaving every supercharge invariant. The existence of a non-trivial monodromy and the breaking of supersymmetry are therefore logically distinct statements.
An R R R -symmetry is distinguished precisely by its non-trivial action on the supersymmetry generators. For a continuous R R R -symmetry twist U R ( ω ) U_R(\omega) U R ( ω ) , let the induced action on the supersymmetry parameters be
R ϵ ( U R ( ω ) ) = exp ( 2 π i ω R ϵ ) , (25) R_{\epsilon}(U_R(\omega))=\exp(2\pi i\omega R_{\epsilon}), \tag*{(25)} R ϵ ( U R ( ω )) = exp ( 2 π iω R ϵ ) , (25)
where R ϵ R_{\epsilon} R ϵ denotes the R R R -symmetry generator in the representation carried by S S S . In a basis in which this generator is diagonal,
R ϵ ( U R ( ω ) ) ϵ A = e 2 π i ω q A ϵ A , (26) R_{\epsilon}(U_R(\omega))\,\epsilon_A=e^{2\pi i\omega q_A}\epsilon_A, \tag*{(26)} R ϵ ( U R ( ω )) ϵ A = e 2 π iω q A ϵ A , (26)
with q A q_A q A the corresponding R R R -charge. The supersymmetry generated by ϵ A \epsilon_A ϵ A survives if and only if
e 2 π i ω q A = 1. (27) e^{2\pi i\omega q_A}=1. \tag*{(27)} e 2 π iω q A = 1. (27)
Depending on the spectrum of R R R -charges and on the value of the twist, the compactification may therefore preserve all, part, or none of the supersymmetry of the parent theory.
Several cycles
Nothing in the invariant-subspace argument is specific to a single circle. For a compact space K K K , let
ρ ϵ : π 1 ( K ) ⟶ G L ( S ) (28) \rho_{\epsilon}:\pi_1(K)\longrightarrow\mathrm{GL}(S) \tag*{(28)} ρ ϵ : π 1 ( K ) ⟶ GL ( S ) (28)
denote the monodromy representation induced on the supersymmetry parameters. In the absence of additional local Killing-spinor constraints, the globally preserved supersymmetries are
S p r e s e r v e d = { ϵ ∈ S ∣ ρ ϵ ( γ ) ϵ = ϵ for every γ ∈ π 1 ( K ) } . (29) S_{\mathrm{preserved}}=\{\epsilon\in S\mid\rho_{\epsilon}(\gamma)\epsilon=\epsilon\ \text{for every }\gamma\in\pi_1(K)\}. \tag*{(29)} S preserved = { ϵ ∈ S ∣ ρ ϵ ( γ ) ϵ = ϵ for every γ ∈ π 1 ( K )} . (29)
A supersymmetry parameter must thus belong to the common invariant subspace of all monodromies. For a torus, the generators of π 1 ( T n ) ≃ Z n \pi_1(T^n) \simeq\mathbb{Z}^n π 1 ( T n ) ≃ Z n commute, and so must their monodromies. For a more general compact space, the monodromies instead obey the relations of its fundamental group.
What this criterion does not tell us
The invariant-subspace criterion is kinematical. If S p r e s e r v e d = { 0 } S_{\mathrm{preserved}} = \{0\} S preserved = { 0 } , no global supersymmetry survives, but this alone does not determine whether the breaking is spontaneous, whether a Goldstino exists, or whether a super-Higgs mechanism occurs. Those questions require additional dynamical information and will be addressed later.
The use of a higher-dimensional symmetry that acts non-trivially on the supersymmetry generators is already central to the original construction of Scherk and Schwarz [6, 7]. Their first explicit supergravity example uses a chiral symmetry of the higher-dimensional theory to give mass to the gravitino while leaving the graviton massless. Applications in higher-dimensional supersymmetric theories with matter and gauge multiplets, including explicit R R R -symmetry twists and the corresponding patterns of partial or complete supersymmetry breaking, were developed further by Fayet [8, 9]. We shall return to these constructions when discussing the mass spectrum, the gravitino, and the dynamical interpretation of Scherk–Schwarz breaking.
Take-home message
Admissibility and supersymmetry preservation are distinct questions: an allowed twist need not preserve all the symmetries of the higher-dimensional theory.
For the flat compactifications considered here, the preserved supersymmetries form the invariant subspace of the monodromy acting on the supersymmetry-parameter space.
A non-trivial twist may preserve all supersymmetries if it acts trivially on the supercharges, while an R R R -symmetry twist may preserve all, part, or none of them. With several cycles, the relevant subspace is the common invariant subspace of all monodromies.
This criterion is kinematical: the absence of invariant supersymmetry parameters does not by itself establish spontaneous supersymmetry breaking or a super-Higgs mechanism.
This naturally leads to the next question:
How does a monodromy modify the Kaluza–Klein spectrum?
How does a monodromy modify the Kaluza–Klein spectrum?
Essential idea
A monodromy changes the global spectral problem. Fields become sections of the flat bundle defined by the monodromy, and the Kaluza–Klein masses are determined by the spectrum of the corresponding twisted internal operator. The familiar shift ( n + ω ) / R (n+\omega)/R ( n + ω ) / R is the special result obtained for a phase monodromy on a flat circle.
The twisted spectral problem
Consider a higher-dimensional spacetime locally of the form M d × K \mathcal{M}_d \times K M d × K , with K K K compact. For a scalar field of higher-dimensional mass m 0 m_0 m 0 , the Klein–Gordon operator separates into a d d d -dimensional d’Alembertian and an internal Laplacian Δ K \Delta_K Δ K . Writing
− Δ K ψ α = λ α ψ α (30) -\Delta_K\psi_\alpha=\lambda_\alpha\psi_\alpha \tag*{(30)} − Δ K ψ α = λ α ψ α (30)
the corresponding lower-dimensional masses are
m α 2 = m 0 2 + λ α . (31) m_\alpha^2=m_0^2+\lambda_\alpha. \tag*{(31)} m α 2 = m 0 2 + λ α . (31)
This is the ordinary Kaluza–Klein spectral problem.
Now equip the compactification with monodromy data
ρ : π 1 ( K ) ⟶ G (32) \rho:\pi_1(K)\longrightarrow G \tag*{(32)} ρ : π 1 ( K ) ⟶ G (32)
where G G G acts in the representation carried by the field. The field is then a section of the associated flat bundle E ρ E_\rho E ρ , and the internal spectral problem becomes
− Δ ρ ψ α = λ α ( ρ ) ψ α (33) -\Delta_\rho\psi_\alpha=\lambda_\alpha(\rho)\psi_\alpha \tag*{(33)} − Δ ρ ψ α = λ α ( ρ ) ψ α (33)
so that
m α 2 = m 0 2 + λ α ( ρ ) . (34) m_\alpha^2=m_0^2+\lambda_\alpha(\rho). \tag*{(34)} m α 2 = m 0 2 + λ α ( ρ ) . (34)
Locally, on a contractible patch, the twisted and untwisted operators may have exactly the same differential expression. What changes is their global domain: only sections obeying the prescribed monodromy conditions are admissible. A twist can therefore modify the spectrum without modifying the local field equations. This distinction will recur repeatedly, for example when supersymmetry can be defined locally along a compact cycle although no supersymmetry parameter is globally compatible with its monodromy.
The familiar example of a circle
For a flat circle S R 1 S^{1}_{R} S R 1 with
y ∼ y + 2 π R , (35) y \sim y + 2\pi R, \tag*{(35)} y ∼ y + 2 π R , (35)
consider a field satisfying the phase-twisted boundary condition
Φ ( x , y + 2 π R ) = e 2 π i ω Φ ( x , y ) . (36) \Phi(x,y+2\pi R)=e^{2\pi i\omega}\Phi(x,y). \tag*{(36)} Φ ( x , y + 2 π R ) = e 2 π iω Φ ( x , y ) . (36)
An internal plane wave ψ p ( y ) = e i p y \psi_p(y)=e^{ipy} ψ p ( y ) = e i p y is allowed only if
e 2 π i R p = e 2 π i ω , (37) e^{2\pi i R p}=e^{2\pi i\omega}, \tag*{(37)} e 2 π i R p = e 2 π iω , (37)
which gives
p n = n + ω R , n ∈ Z . (38) p_n=\frac{n+\omega}{R}, \qquad n\in\mathbb{Z}. \tag*{(38)} p n = R n + ω , n ∈ Z . (38)
The internal eigenvalues are therefore
λ n ( ω ) = ( n + ω ) 2 R 2 , (39) \lambda_n(\omega)=\frac{(n+\omega)^2}{R^2}, \tag*{(39)} λ n ( ω ) = R 2 ( n + ω ) 2 , (39)
and the lower-dimensional masses become
m n 2 = m 0 2 + ( n + ω ) 2 R 2 . (40) m_n^2=m_0^2+\frac{(n+\omega)^2}{R^2}. \tag*{(40)} m n 2 = m 0 2 + R 2 ( n + ω ) 2 . (40)
Thus the familiar replacement n → n + ω n\to n+\omega n → n + ω is simply the spectrum required by a phase-twisted boundary condition on a flat circle. Figure 2 displays this special case. The uniform translation of the momentum lattice visible there should not be mistaken for a universal property of twisted compactifications: it is the particularly simple realization of the more general spectral problem described above.
Figure 2. Kaluza–Klein levels for a phase monodromy on a flat circle S R 1 S^1_R S R 1 . In the untwisted case the allowed momenta are p n = n / R p_n = n/R p n = n / R , whereas Φ ( y + 2 π R ) = e 2 π i ω Φ ( y ) \Phi(y + 2\pi R) = e^{2\pi i\omega}\Phi(y) Φ ( y + 2 π R ) = e 2 π iω Φ ( y ) shifts them to p n = ( n + ω ) / R p_n = (n + \omega)/R p n = ( n + ω ) / R . The tower is therefore translated in momentum space. This rigid shift is special to the flat-circle problem; on a general compact space the spectrum is determined by the appropriate twisted internal operator.
Twisted fields or periodic fields
The same compactification can be described using the periodic field
Φ ~ ( x , y ) = e − i ω y / R Φ ( x , y ) , (41) \widetilde{\Phi}(x,y)=e^{-i\omega y/R}\Phi(x,y), \tag*{(41)} Φ ( x , y ) = e − iω y / R Φ ( x , y ) , (41)
for which
Φ ~ ( x , y + 2 π R ) = Φ ~ ( x , y ) . (42) \widetilde{\Phi}(x,y+2\pi R)=\widetilde{\Phi}(x,y). \tag*{(42)} Φ ( x , y + 2 π R ) = Φ ( x , y ) . (42)
The price of making the field periodic is that the internal derivative acts as
∂ y ⟶ ∂ y + i ω R . (43) \partial_y\longrightarrow\partial_y+\frac{i\omega}{R}. \tag*{(43)} ∂ y ⟶ ∂ y + R iω . (43)
The same global information may therefore be encoded either in twisted boundary conditions or in a flat connection. This equivalence will become particularly useful later when Scherk–Schwarz breaking is described in terms of an R R R -symmetry connection or Wilson line.
Several field components and several cycles
If the field takes values in a representation space V V V of the symmetry group, a single monodromy need not act in the same way on every component. For
Φ ( y + 2 π R ) = U Φ ( y ) , (44) \Phi(y + 2\pi R) = U\Phi(y), \tag*{(44)} Φ ( y + 2 π R ) = U Φ ( y ) , (44)
suppose that U U U is diagonalizable and choose a basis { v α } \{v_\alpha\} { v α } satisfying
U v α = e 2 π i ω α v α . (45) Uv_\alpha= e^{2\pi i\omega_\alpha}v_\alpha. \tag*{(45)} U v α = e 2 π i ω α v α . (45)
Writing
Φ ( x , y ) = ∑ α Φ α ( x , y ) v α , (46) \Phi(x,y) = \sum_\alpha\Phi_\alpha(x,y)v_\alpha, \tag*{(46)} Φ ( x , y ) = α ∑ Φ α ( x , y ) v α , (46)
each component obeys its own phase-twisted boundary condition and has momenta
p n , α = n + ω α R , n ∈ Z . (47) p_{n,\alpha} = \frac{n+\omega_\alpha}{R}, \qquad n \in\mathbb{Z}. \tag*{(47)} p n , α = R n + ω α , n ∈ Z . (47)
Different components of the same higher-dimensional multiplet may therefore generate differently shifted Kaluza–Klein towers. In particular, a zero mode with vanishing internal momentum can survive only in a monodromy-invariant component.
The same reasoning applies when the compact space contains several independent cycles. For a rectangular r r r -torus,
y i ∼ y i + 2 π R i , i = 1 , … , r , (48) y^i \sim y^i + 2\pi R_i,\qquad i=1,\ldots,r, \tag*{(48)} y i ∼ y i + 2 π R i , i = 1 , … , r , (48)
let U i U_i U i denote the monodromy around the i i i th cycle. Since π 1 ( T r ) ≃ Z r \pi_1(T^r) \simeq\mathbb{Z}^r π 1 ( T r ) ≃ Z r , these monodromies commute. When they can be simultaneously diagonalized, common eigenvectors v α v_\alpha v α may be chosen such that
U i v α = e 2 π i ω i , α v α . (49) U_i v_\alpha= e^{2\pi i\omega_{i,\alpha}}v_\alpha. \tag*{(49)} U i v α = e 2 π i ω i , α v α . (49)
The allowed internal momenta are then
p i , n i , α = n i + ω i , α R i , n i ∈ Z , (50) p_{i,n_i,\alpha} = \frac{n_i+\omega_{i,\alpha}}{R_i},\qquad n_i\in\mathbb{Z}, \tag*{(50)} p i , n i , α = R i n i + ω i , α , n i ∈ Z , (50)
and for a rectangular flat metric
m n , α 2 = m 0 2 + ∑ i = 1 r ( n i + ω i , α ) 2 R i 2 . (51) m^2_{n,\alpha}=m_0^2+\sum_{i=1}^{r}\frac{(n_i+\omega_{i,\alpha})^2}{R_i^2}. \tag*{(51)} m n , α 2 = m 0 2 + i = 1 ∑ r R i 2 ( n i + ω i , α ) 2 . (51)
Geometrically, each common monodromy eigenspace sees a translated momentum lattice. If Λ ∗ \Lambda^* Λ ∗ denotes the dual lattice of the torus and
δ α = ( ω 1 , α R 1 , … , ω r , α R r ) , (52) \delta_\alpha=\left(\frac{\omega_{1,\alpha}}{R_1},\ldots,\frac{\omega_{r,\alpha}}{R_r}\right), \tag*{(52)} δ α = ( R 1 ω 1 , α , … , R r ω r , α ) , (52)
then the allowed momenta lie in the affine lattice
p α ∈ Λ ∗ + δ α . (53) p_\alpha\in\Lambda^*+\delta_\alpha. \tag*{(53)} p α ∈ Λ ∗ + δ α . (53)
Even on a flat torus there is therefore no unique shifted tower: the spectrum and its degeneracies depend on the geometry, the field representation, and the complete set of monodromy eigenvalues.
Beyond flat compactifications
On a general internal space the Kaluza–Klein spectrum is determined by the appropriate twisted internal operator,
Spec K K ( ρ ) = Spec ( O K , ρ ) , (54) \operatorname{Spec}_{\mathrm{KK}}(\rho)=\operatorname{Spec}(\mathcal{O}_{K,\rho}), \tag*{(54)} Spec KK ( ρ ) = Spec ( O K , ρ ) , (54)
whose eigenvalues need not be equally spaced. For scalars, O K , ρ \mathcal{O}_{K,\rho} O K , ρ is typically a covariant Laplacian; for spinors, the corresponding twisted Dirac operator; and for gauge or higher-spin fields, the appropriate gauge-fixed kinetic operator or system. A monodromy may lift zero modes, split degeneracies, and change multiplicities, but on a general compact manifold there is no universal analogue of the rigid ( n + ω ) / R (n+\omega)/R ( n + ω ) / R translation.
The discussion above concerns monodromies that act on fields transported around ordinary one-cycles. This should be distinguished from another type of global datum that will appear later in the review: holonomies of higher-form gauge fields over higher-dimensional cycles and the associated wrapped extended objects. These involve a different geometric structure and will be treated separately when we discuss higher cycles.
The shifted Kaluza–Klein spectra generated by non-trivial higher-dimensional boundary conditions are already part of the original Scherk–Schwarz construction [6, 7]. Their formulation in terms of sections of flat bundles and monodromy representations is the standard geometric language reviewed, for example, in Ref. [18]. The equivalence between twisted boundary conditions and a periodic description with a flat connection is one of the recurring forms of the Scherk–Schwarz mechanism; its interpretation in supergravity and in terms of Wilson lines will be developed in later questions.
Take-home message
A monodromy changes the global spectral problem even when the local differential equations are unchanged.
The Kaluza–Klein masses are eigenvalues of the appropriate twisted internal operator acting on sections of the bundle defined by the monodromy.
On a flat circle, a phase monodromy gives p n = ( n + ω ) / R p_n = (n + \omega)/R p n = ( n + ω ) / R . Equivalently, one may use periodic fields and a flat connection in the internal derivative.
For matrix-valued monodromies, different field components may acquire different shifts; on a flat torus each common eigenspace sees an affine momentum lattice.
On a general compact manifold there is no universal ( n + ω ) / R (n + \omega)/R ( n + ω ) / R formula.
This naturally leads to the next question:
In what sense is Scherk–Schwarz supersymmetry breaking spontaneous?
In what sense is Scherk–Schwarz supersymmetry breaking spontaneous?
Essential idea
Spontaneous supersymmetry breaking means that supersymmetry remains a symmetry of the theory but is not preserved by the chosen background. Explicit breaking means instead that the corresponding supersymmetry is not a symmetry of the theory itself.
In a Scherk–Schwarz compactification, the absence of an invariant supercharge establishes that supersymmetry is broken. It does not by itself determine the nature of that breaking. In continuous R R R -symmetry reductions of supergravity, the spontaneous interpretation can be made explicit: the broken local supersymmetries remain part of the gauge structure, and the Goldstino and super-Higgs mechanism provide their dynamical realization. Other constructions must be examined separately.
Symmetry of the theory and symmetry of the state
For an ordinary global symmetry, the distinction between explicit and spontaneous breaking is straightforward. In spontaneous breaking, the symmetry remains a symmetry of the action but is not preserved by the vacuum. For rigid supersymmetry one may write schematically
δ ϵ S = 0 , Q A ∣ 0 ⟩ ≠ 0 (55) \delta_{\epsilon} S = 0,\qquad Q_A |0\rangle\ne0 \tag*{(55)} δ ϵ S = 0 , Q A ∣0 ⟩ = 0 (55)
for at least one supercharge Q A Q_A Q A . Explicit breaking means instead that the corresponding supersymmetry is not a symmetry of the theory in the first place.
For local supersymmetry the language is slightly different because supersymmetry is a gauge symmetry. A bosonic background is supersymmetric when there exists a non-zero supersymmetry parameter for which all fermionic variations vanish. If no such parameter exists, the background breaks supersymmetry. When the broken local supersymmetry nevertheless remains part of the gauge structure of the theory, the breaking is spontaneous: a fermionic direction has an inhomogeneous supersymmetry variation and provides the Goldstino, which is absorbed by the gravitino through the super-Higgs mechanism.
Scherk–Schwarz compactification makes this distinction less immediate. A non-trivial monodromy may eliminate every globally defined supersymmetry parameter, so there is no ambiguity about whether supersymmetry survives: it does not. The more subtle question is what the word spontaneous means in this situation. The answer depends on the physical status of the compactification data. One must determine whether the missing supersymmetries remain symmetries of the same underlying dynamical theory and fail only to be preserved by the chosen configuration, or whether the global definition of the compactified theory has removed them from its symmetry structure. The spectrum alone cannot decide between these possibilities.
The continuous Scherk–Schwarz case
Consider a continuous R R R -symmetry monodromy
U R ( ω ) = exp ( 2 π i R ) , (56) U_R(\omega)=\exp(2\pi i\mathcal{R}), \tag*{(56)} U R ( ω ) = exp ( 2 π i R ) , (56)
where R \mathcal{R} R denotes a generator of the R R R -symmetry group. Its action on the supersymmetry-parameter space is generated by R ϵ \mathcal{R}_{\epsilon} R ϵ , in the notation of Question 3. Supersymmetry is restored whenever the induced monodromy acts trivially on the supercharges. In the simplest normalization, the branch connected to the untwisted point contains
ω = 0 , (57) \omega=0, \tag*{(57)} ω = 0 , (57)
up to the identifications of the twist parameter.
On a flat circle the characteristic Scherk–Schwarz splittings behave as
m S S ∼ ∣ ω ∣ R , (58) m_{\mathrm{SS}}\sim\frac{|\omega|}{R}, \tag*{(58)} m SS ∼ R ∣ ω ∣ , (58)
up to R R R -charges and normalization conventions. Supersymmetry is then restored continuously as the monodromy becomes trivial on the supercharges.
Continuity of the parameter, however, is not sufficient to establish spontaneous breaking. A continuous boundary-condition parameter need not be the expectation value of a dynamical field, and a continuous family of theories with different explicit breaking parameters is not the same thing as a continuous family of configurations of one supersymmetric dynamical theory. This distinction is particularly important in deciding how the Scherk–Schwarz parameter should be interpreted.
In supergravity, there are formulations in which the Scherk–Schwarz twist is represented by an R R R -symmetry connection or holonomy belonging to the locally supersymmetric theory itself. In appropriate off-shell descriptions, the same information may be encoded in the radion sector. In such realizations the monodromy is not introduced as an explicit supersymmetry-breaking operator: different values correspond to different configurations of a theory whose local supersymmetry remains part of its gauge structure. The accompanying Goldstino and super-Higgs mechanism then provide the dynamical content required for the spontaneous interpretation.
Two ways of restoring supersymmetry
Equation (58) also shows that the Scherk–Schwarz splitting vanishes in the decompactification limit,
m S S ⟶ 0 ( R → ∞ ) . (59) m_{\mathrm{SS}}\longrightarrow0 \qquad(R\to\infty). \tag*{(59)} m SS ⟶ 0 ( R → ∞ ) . (59)
There are therefore two conceptually different ways in which the splitting may disappear: the monodromy can become trivial on the supercharges, ω → 0 \omega\to0 ω → 0 , or the compactification scale itself can vanish, R − 1 → 0 R^{-1}\to0 R − 1 → 0 , while the rigid monodromy remains non-trivial. The two limits should not be identified. In the latter case the effective dimensionality changes, and when R R R is a modulus the decompactification limit is typically at infinite distance in moduli space. Supersymmetry restoration as R → ∞ R\to\infty R → ∞ therefore shows that the breaking scale is geometrical, but does not by itself establish that the finite-radius breaking is spontaneous.
A more decisive diagnostic is the realization of the broken symmetry in the low-energy theory. For global supersymmetry, spontaneous breaking requires a Goldstino, equivalently a fermionic direction with an inhomogeneous supersymmetry variation on the vacuum. For local supersymmetry, the corresponding realization is the super-Higgs mechanism. The next two questions will identify the Goldstino in the Scherk–Schwarz compactification and explain the role of the full Kaluza–Klein gravitino tower. In continuous supergravity reductions one indeed finds the fermionic shifts and gravitino–Goldstino mixing expected from spontaneous breaking.
What each diagnostic tells us
Several frequently invoked properties answer different logical questions and should not be used interchangeably.
Table 1.
Why discrete twists require separate treatment
Discreteness by itself does not decide whether supersymmetry breaking is spontaneous. An ordinary orbifold or orientifold projection retains only the invariant physical states and supercharges. A supercharge genuinely projected out of the physical Hilbert space is no longer a symmetry acting on the projected theory, so its absence should not be interpreted as spontaneous breaking of that theory. Discrete Scherk–Schwarz constructions need not have this interpretation: freely acting orbifolds, for example, can provide realizations of spontaneous supersymmetry breaking despite the discreteness of the quotient. What matters is not whether the twist parameter is continuous or discrete, but how the quotient acts on the physical state space and whether the missing supersymmetry remains realized in the underlying theory.
Nor does the reappearance of supersymmetry in a decompactification limit settle the issue by itself. It may simply mean that the global distinction responsible for the breaking becomes invisible when the compact cycle disappears. Discrete twists, spin structures, freely acting orbifolds, and related constructions must therefore be analysed at the level of their physical state spaces and supercharges. Their relation to ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s and to continuous R R R -symmetry twists will be discussed later in the review: sharing the name Scherk–Schwarz does not by itself imply sharing the same dynamical mechanism.
Working criterion
In this review, we shall use the term spontaneous Scherk–Schwarz breaking when the supersymmetry not preserved by the compactification remains a symmetry of the same underlying dynamical theory rather than being removed by the global definition or projection, and when its broken realization exhibits the corresponding Goldstino or super-Higgs structure. Continuous R R R -symmetry Scherk–Schwarz reductions of supergravity provide the canonical example: the monodromy can be represented through gauge or dynamical data of the supersymmetric theory, while the broken local supersymmetry remains part of its gauge structure and is realized through the super-Higgs mechanism.
By contrast, a continuous parameter, a radius-dependent mass splitting, the absence of a local explicit breaking term, or the restoration of supersymmetry in a limiting regime are useful characteristics of particular constructions, but none of them alone determines whether the breaking is spontaneous. The criterion adopted here deliberately separates these useful diagnostics from the dynamical statement that we ultimately wish to make.
The spontaneous interpretation of generalized dimensional reduction is already explicit in the original papers of Scherk and Schwarz [6, 7], where local supersymmetry is retained while gravitino masses are generated by the compactification. Explicit higher-dimensional realizations based on R R R -symmetry, including the resulting fermion and scalar mass splittings, were developed further by Fayet [8, 9]. The distinction between supersymmetry restoration by varying a compactification datum and restoration in a decompactification limit becomes especially important in string theory; we shall return to the corresponding constraints and their early analyses when the microscopic string construction is introduced. The Goldstino and super-Higgs aspects relevant to the field-theory discussion will instead be developed directly in the next questions.
Take-home message
The absence of an invariant supercharge establishes that supersymmetry is broken, but does not by itself determine whether the breaking is spontaneous or explicit.
The central question is whether the missing supercharges remain symmetries of the same underlying dynamical theory, or are removed by the global definition or projection of the compactified theory.
A continuous path to a supersymmetric point, or the disappearance of mass splittings as R → ∞ R \to\infty R → ∞ , are useful diagnostics but are not by themselves sufficient to establish spontaneous supersymmetry breaking.
In continuous supergravity Scherk–Schwarz reductions, the monodromy can be realized through gauge or dynamical data of the supersymmetric theory, while the Goldstino and super-Higgs mechanism provide the characteristic dynamical realization of spontaneous breaking.
Discrete twists and projections require a separate analysis: what matters is not discreteness itself, but whether the missing supersymmetry remains realized in the underlying physical theory.
This naturally leads to the next question:
Where is the Goldstino in a Scherk–Schwarz compactification?
Where is the Goldstino in a Scherk–Schwarz compactification?
Essential idea
The fermionic Stückelberg structure of a Kaluza–Klein gravitino tower is already present before any Scherk–Schwarz twist is introduced. For a five-dimensional gravitino, the internal component transforms at non-zero Kaluza–Klein level as
δ ψ y ( n ) ∼ i n R ϵ ( n ) . \delta\psi_y^{(n)} \sim\frac{in}{R}\epsilon^{(n)} . δ ψ y ( n ) ∼ R in ϵ ( n ) .
A continuous Scherk–Schwarz twist does not create this structure. It shifts the internal operator,
i n R ⟶ i n R + M , \frac{in}{R} \longrightarrow\frac{in}{R}+\mathcal{M}, R in ⟶ R in + M ,
where M \mathcal{M} M acts on the relevant R R R -symmetry indices. The internal gravitino components selected by this shifted operator provide the corresponding would-be Goldstino directions.
Their absorption by the gravitini, and the resulting spin-3 2 \frac{3}{2} 2 3 mass matrix, are properties of the reduced Rarita–Schwinger Lagrangian and will be established in the next question.
Fermion shifts and Goldstino directions
On a bosonic background, lower-dimensional spin-1 2 \frac{1}{2} 2 1 fermions transform schematically as
δ ϵ χ a = F A a ϵ A + ⋯ , (60) \delta_\epsilon\chi^a=\mathcal{F}^{a}_{A}\epsilon^{A}+\cdots, \tag*{(60)} δ ϵ χ a = F A a ϵ A + ⋯ , (60)
where F A a \mathcal{F}^{a}_{A} F A a is the fermion-shift matrix. For a broken supersymmetry direction, the non-vanishing vector F A a ϵ A \mathcal{F}^{a}_{A}\epsilon^{A} F A a ϵ A identifies the corresponding would-be Goldstino direction in fermion space. The Goldstino subspace is therefore Im F \operatorname{Im}\mathcal{F} Im F . The physical normalized Goldstino combination also depends on the fermionic kinetic terms, a point that becomes relevant when several fermions contribute.
Parameters belonging to ker F \operatorname{ker}\mathcal{F} ker F are candidates for preserved supersymmetries, provided that the remaining Killing-spinor equations are satisfied. For the flat-circle Scherk–Schwarz compactifications considered below, this reduces to the monodromy criterion discussed in the previous questions.
The Stückelberg structure of the gravitino KK tower
Consider a five-dimensional gravitino compactified on a circle of radius R R R , with Ψ M → ( Ψ μ , Ψ y ) \Psi_M\to(\Psi_\mu,\Psi_y) Ψ M → ( Ψ μ , Ψ y ) . Its local supersymmetry transformation contains
δ ϵ Ψ M = D M ϵ + ⋯ , δ ϵ Ψ y = D y ϵ + ⋯ . (61) \delta_\epsilon\Psi_M=D_M\epsilon+\cdots,\qquad\delta_\epsilon\Psi_y=D_y\epsilon+\cdots. \tag*{(61)} δ ϵ Ψ M = D M ϵ + ⋯ , δ ϵ Ψ y = D y ϵ + ⋯ . (61)
The internal gravitino is therefore directly sensitive to the operator governing the dependence of the supersymmetry parameter on the compact coordinate, which explains its special role in the Goldstino sector.
Before introducing a twist, expand
ε ( x , y ) = ∑ n ∈ Z ε ( n ) ( x ) e i n y / R , Ψ y ( x , y ) = ∑ n ∈ Z ψ y ( n ) ( x ) e i n y / R . (62) \varepsilon(x,y)=\sum_{n\in\mathbb{Z}}\varepsilon^{(n)}(x)e^{iny/R},\qquad\Psi_y(x,y)=\sum_{n\in\mathbb{Z}}\psi_y^{(n)}(x)e^{iny/R}. \tag*{(62)} ε ( x , y ) = n ∈ Z ∑ ε ( n ) ( x ) e in y / R , Ψ y ( x , y ) = n ∈ Z ∑ ψ y ( n ) ( x ) e in y / R . (62)
The Fourier expansion then gives
δ ψ y ( n ) = i n R ε ( n ) + ⋯ . (63) \delta\psi_y^{(n)}=\frac{in}{R}\varepsilon^{(n)}+\cdots. \tag*{(63)} δ ψ y ( n ) = R in ε ( n ) + ⋯ . (63)
For n ≠ 0 n\ne0 n = 0 , this is the fermionic analogue of the ordinary Kaluza–Klein Stückelberg transformation. The internal gravitino modes supply the spin-1 2 \frac{1}{2} 2 1 Stückelberg degrees of freedom associated with the massive spin-3 2 \frac{3}{2} 2 3 Kaluza–Klein tower.
Question 4 showed that a Scherk–Schwarz monodromy shifts the compact momentum. Its effect on the gravitino is therefore naturally expected to deform this pre-existing Stückelberg operator rather than to generate an entirely new structure.
Adding the Scherk–Schwarz twist
Write the continuous R R R -symmetry monodromy acting on the supersymmetry parameters as
U R , ε = exp ( 2 π R M ) = exp ( 2 π i ω R ε ) , M = i ω R R ε . (64) U_{R,\varepsilon}=\exp(2\pi R\mathcal{M})=\exp(2\pi i\omega\mathcal{R}_{\varepsilon}),\qquad\mathcal{M}=\frac{i\omega}{R}\mathcal{R}_{\varepsilon}. \tag*{(64)} U R , ε = exp ( 2 π R M ) = exp ( 2 π iω R ε ) , M = R iω R ε . (64)
Here R ε \mathcal{R}_{\varepsilon} R ε is the action of the R R R -symmetry generator R \mathcal{R} R on the supersymmetry-parameter space S \mathcal{S} S , in the notation of Questions 3 and 5.
Using the periodic-field description introduced in Question 4,
ε ( x , y ) = e y M ε ~ ( x , y ) , Ψ M ( x , y ) = e y M Ψ ~ M ( x , y ) , (65) \varepsilon(x,y)=e^{y\mathcal{M}}\widetilde{\varepsilon}(x,y),\qquad\Psi_M(x,y)=e^{y\mathcal{M}}\widetilde{\Psi}_M(x,y), \tag*{(65)} ε ( x , y ) = e y M ε ( x , y ) , Ψ M ( x , y ) = e y M Ψ M ( x , y ) , (65)
the internal derivative becomes ∂ y → ∂ y + M \partial_y\to\partial_y+\mathcal{M} ∂ y → ∂ y + M . The internal gravitino transformation is consequently
δ ε ~ Ψ ~ y = ( ∂ y + M ) ε ~ + ⋯ , (66) \delta_{\widetilde{\varepsilon}}\widetilde{\Psi}_y=(\partial_y+\mathcal{M})\widetilde{\varepsilon}+\cdots, \tag*{(66)} δ ε Ψ y = ( ∂ y + M ) ε + ⋯ , (66)
and at Kaluza–Klein level n n n ,
δ ψ y ( n ) = D n ε ( n ) + ⋯ , D n ≡ i n R + M . (67) \delta\psi_y^{(n)}=\mathcal{D}_n\varepsilon^{(n)}+\cdots,\qquad\mathcal{D}_n\equiv\frac{in}{R}+\mathcal{M}. \tag*{(67)} δ ψ y ( n ) = D n ε ( n ) + ⋯ , D n ≡ R in + M . (67)
For the mode represented by n = 0 n=0 n = 0 in this periodic frame,
δ ψ y ( 0 ) = M ε ( 0 ) + ⋯ . (68) \delta\psi_y^{(0)}=\mathcal{M}\varepsilon^{(0)}+\cdots. \tag*{(68)} δ ψ y ( 0 ) = M ε ( 0 ) + ⋯ . (68)
The components of ψ y ( 0 ) \psi_y^{(0)} ψ y ( 0 ) lying in Im M \operatorname{Im}\mathcal{M} Im M therefore provide the corresponding would-be Goldstino directions. What distinguished the n = 0 n=0 n = 0 sector in the untwisted theory was precisely the absence of the Stückelberg shift; the Scherk–Schwarz deformation extends the same structure into that sector.
Figure 3. Kaluza–Klein Stückelberg structure and its Scherk–Schwarz deformation. The upper panel recalls the common higher-dimensional origin of massive Kaluza–Klein fields of spin 1, 3/2, and 2: at non-zero Kaluza–Klein level, internal field components provide the Stückelberg degrees of freedom required by the corresponding massive lower-dimensional fields. For the gravitino tower, the untwisted fermionic transformation is governed by i n / R in/R in / R . A continuous Scherk–Schwarz twist shifts this operator to i n / R + M in/R+\mathcal{M} in / R + M . The twist therefore deforms a Stückelberg structure already present in the higher-dimensional theory. The integer Kaluza–Klein label depends on the representative chosen for the twist, whereas the complete shifted tower is invariant under the corresponding relabelling.
For extended supersymmetry, M \mathcal{M} M is a matrix acting on R R R -symmetry indices. In minimal five-dimensional supergravity, circle compactification gives N = 2 \mathcal{N}=2 N = 2 supersymmetry in four dimensions, and the supersymmetry parameters form an S U ( 2 ) R SU(2)_R S U ( 2 ) R doublet. Thus M \mathcal{M} M acts as a 2 × 2 2 \times2 2 × 2 matrix. More supersymmetric parent theories are treated analogously: only combinations selected by the image of the shift operator transform inhomogeneously.
The Kaluza–Klein label is not invariant
There is an important subtlety in calling the preceding field the “zero-mode Goldstino”. For an eigenstate of R R R -charge q q q , the spectrum depends on the combination n + ω q n+\omega q n + ω q . An integer shift can be compensated by relabelling the Kaluza–Klein tower,
ω q ⟶ ω q + k , n ⟶ n − k , k ∈ Z . (69) \omega q \longrightarrow\omega q+k,\qquad n \longrightarrow n-k,\qquad k \in\mathbb{Z}. \tag*{(69)} ω q ⟶ ω q + k , n ⟶ n − k , k ∈ Z . (69)
Consequently, the statement that the Scherk–Schwarz super-Higgs mechanism involves the n = 0 n=0 n = 0 gravitino is not invariant by itself. A representative of the monodromy class may be chosen in which the light gravitino carries that label, but another representative can assign a different integer label to the same physical state.
The invariant datum is the complete shifted operator D n \mathcal{D}_n D n , or equivalently the monodromy together with the full Kaluza–Klein tower. The fundamental domain of the twist parameter and the classification of inequivalent values of ω \omega ω will be discussed later.
From the Goldstino direction to the super-Higgs mechanism
The Fourier components ϵ ( n ) \epsilon^{(n)} ϵ ( n ) are all modes of the same higher-dimensional local supersymmetry parameter; there is no change of gauge status between n = 0 n=0 n = 0 and n ≠ 0 n\ne0 n = 0 . What changes is the action of the internal operator. Without the twist it is i n / R in/R in / R , whereas the Scherk–Schwarz deformation replaces it by
D n = i n R + M . (70) \mathcal{D}_n=\frac{in}{R}+\mathcal{M}. \tag*{(70)} D n = R in + M . (70)
The same operator that produces the inhomogeneous transformation of the internal gravitino will reappear in the reduced Rarita–Schwinger Lagrangian, both in the spin-3 2 \frac{3}{2} 2 3 mass matrix and in the gravitino–Goldstino mixing.
The transformation law therefore identifies the would-be Goldstino direction, but does not yet prove that it is absorbed or determine the physical gravitino masses. These are dynamical statements about the reduced action and are the subject of the next question.
For ∣ ω ∣ ≪ 1 |\omega|\ll1 ∣ ω ∣ ≪ 1 , the gravitino that becomes massless at the supersymmetric point may have
m 3 / 2 ∼ ∣ ω ∣ R ≪ 1 R . (71) m_{3/2}\sim\frac{|\omega|}{R}\ll\frac{1}{R}. \tag*{(71)} m 3/2 ∼ R ∣ ω ∣ ≪ R 1 . (71)
up to the relevant R R R -charge and normalization. At energies E ≪ 1 / R E\ll1/R E ≪ 1/ R , the heavier Kaluza–Klein gravitini can then be integrated out, leaving a four-dimensional supergravity description of the light sector. This low-energy separation does not make the label n = 0 n=0 n = 0 physical: a change of representative can still relabel the complete tower while leaving its spectrum unchanged.
Additional fermion shifts
The internal gravitino need not exhaust the Goldstino sector. In a general compactification, additional spin-1 2 \frac{1}{2} 2 1 fermions may have non-zero shifts,
δ χ a = F A a ϵ A + ⋯ . (72) \delta\chi^{a}=\mathcal{F}^{a}_{A}\epsilon^{A}+\cdots. \tag*{(72)} δ χ a = F A a ϵ A + ⋯ . (72)
The physical Goldstino is then the linear combination selected by the complete fermion-shift matrix together with the fermionic kinetic terms. Modulini, gaugini, hyperini, or localized fermions may therefore modify the physical combination without changing the basic role of the twisted internal gravitino derivative.
The discussion here assumes a continuous supergravity realization in which the broken supersymmetries remain part of the higher-dimensional local gauge symmetry. It should not automatically be applied to projections that remove the corresponding supercharges from the physical theory. Orbifolds, localized breaking sectors, spin structures, and ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s projections will be treated when those ingredients are introduced.
The appearance of the internal gravitino component as the fermionic degree of freedom associated with Scherk–Schwarz supersymmetry breaking is already visible in the original supergravity construction of Scherk and Schwarz [6, 7]. Higher-dimensional supersymmetric realizations and their fermionic mass spectra were developed further by Fayet [8, 9]. The relation between the fermion shift, the gravitino mass matrix, and the super-Higgs mechanism will be established directly from the reduced Rarita–Schwinger action in the next question.
Take-home message
The untwisted Kaluza–Klein gravitino tower already has a fermionic Stückelberg structure, δ ψ y ( n ) ∼ ( i n / R ) ϵ ( n ) \delta\psi_y^{(n)} \sim(in/R)\epsilon^{(n)} δ ψ y ( n ) ∼ ( in / R ) ϵ ( n ) .
A continuous Scherk–Schwarz twist replaces i n / R in/R in / R by D n = i n / R + M \mathcal{D}_n = in/R + \mathcal{M} D n = in / R + M . The internal gravitino components in its image provide the corresponding would-be Goldstino directions.
The designation of a particular gravitino as the n = 0 n = 0 n = 0 mode depends on the representative of the twist. The invariant datum is the complete shifted tower, not an individual KK label.
A fermion shift identifies the would-be Goldstino direction; its absorption and the physical spin-3 2 \frac{3}{2} 2 3 masses follow from the reduced Rarita–Schwinger Lagrangian.
If other fermions also carry shifts, the physical Goldstino is selected by the complete fermion-shift and kinetic structure.
This naturally leads to the next question:
How does the super-Higgs mechanism work in a Scherk–Schwarz compactification?
How does the super-Higgs mechanism work in a Scherk–Schwarz compactification?
Essential idea
For a five-dimensional gravitino, the ordinary Kaluza–Klein tower already possesses the fermionic Stückelberg structure required by massive spin-3 2 \frac{3}{2} 2 3 fields. A Scherk–Schwarz twist does not create this structure: it deforms the internal operator that controls it.
The same deformed operator appears in the gravitino–Goldstino mixing and in the spin-3 2 \frac{3}{2} 2 3 spectrum. In an R ξ R_{\xi} R ξ gauge the mixing can be cancelled explicitly; in the unitary-gauge limit the would-be Goldstinos cease to propagate independently and provide the longitudinal helicities of the massive gravitini.
This super-Higgs mechanism is a statement about the full fermionic system and does not require the gravitino mass to be parametrically below the Kaluza–Klein scale. The high-energy equivalence theorem is a separate statement whose practical usefulness depends on the hierarchy of scales.
Specializing to five dimensions
The preceding questions were formulated, whenever possible, for a compactification from d + 1 d+1 d + 1 to d d d dimensions. We now specialize to five-dimensional supergravity compactified to four dimensions. Accordingly, we write the compact coordinate as x 5 ≡ y x^5 \equiv y x 5 ≡ y and denote the internal gravitino components by ψ 5 I \psi_{5I} ψ 5 I . This specialization will allow us to use explicitly four-dimensional Weyl, Majorana, and helicity language.
Consider minimal five-dimensional supergravity, in which the gravitini form an S U ( 2 ) R SU(2)_{R} S U ( 2 ) R doublet. Their four-dimensional and internal components will be denoted
ψ μ I , ψ 5 I , I = 1 , 2. (73) \psi_{\mu I}, \qquad\psi_{5I}, \qquad I=1,2. \tag*{(73)} ψ μ I , ψ 5 I , I = 1 , 2. (73)
As already anticipated in Question 6, the internal components behave as spin-1 2 \frac{1}{2} 2 1 fermions from the four-dimensional point of view and provide the Stückelberg directions associated with the massive gravitini.
From Kaluza–Klein Stückelberg to super-Higgs
After the field redefinitions required to put the spin-1 2 \frac{1}{2} 2 1 kinetic terms in canonical form, the relevant bulk bilinears contain schematically
L b i l i n e a r ⊃ ψ μ 1 σ μ ν ∂ 5 ψ ν 2 − ψ μ 2 σ μ ν ∂ 5 ψ ν 1 − ω R ( ψ μ 1 σ μ ν ψ ν 1 + ψ μ 2 σ μ ν ψ ν 2 ) − i 6 2 [ ∂ 5 ψ 51 σ μ ψ μ 1 + ∂ 5 ψ 52 σ μ ψ μ 2 + ω R ( ψ 52 σ μ ψ μ 1 − ψ 51 σ μ ψ μ 2 ) ] + ⋯ (74) \mathcal{L}_{\mathrm{bilinear}} \supset\psi_{\mu1}\sigma^{\mu\nu}\partial_{5}\psi_{\nu2}-\psi_{\mu2}\sigma^{\mu\nu}\partial_{5}\psi_{\nu1}
-\frac{\omega}{R}\left(\psi_{\mu1}\sigma^{\mu\nu}\psi_{\nu1}+\psi_{\mu2}\sigma^{\mu\nu}\psi_{\nu2}\right)
-\frac{i\sqrt{6}}{2}\left[\partial_{5}\psi_{51}\sigma^{\mu}\psi_{\mu1}+\partial_{5}\psi_{52}\sigma^{\mu}\psi_{\mu2}
+\frac{\omega}{R}\left(\psi_{52}\sigma^{\mu}\psi_{\mu1}-\psi_{51}\sigma^{\mu}\psi_{\mu2}\right)\right]+\cdots \tag*{(74)} L bilinear ⊃ ψ μ 1 σ μν ∂ 5 ψ ν 2 − ψ μ 2 σ μν ∂ 5 ψ ν 1 − R ω ( ψ μ 1 σ μν ψ ν 1 + ψ μ 2 σ μν ψ ν 2 ) − 2 i 6 [ ∂ 5 ψ 51 σ μ ψ μ 1 + ∂ 5 ψ 52 σ μ ψ μ 2 + R ω ( ψ 52 σ μ ψ μ 1 − ψ 51 σ μ ψ μ 2 ) ] + ⋯ (74)
The numerical coefficients depend on normalization and spinor conventions. The structural point is that the compact derivative and the Scherk–Schwarz deformation always appear in the same S U ( 2 ) R SU(2)_{R} S U ( 2 ) R -covariant combinations.
In the periodic-field representation these combinations are generated by
D 5 = ∂ 5 + ω R ( 0 1 − 1 0 ) , (75) \mathcal{D}_{5}=\partial_{5}+\frac{\omega}{R}
\begin{pmatrix}
0 & 1 \\
-1 & 0
\end{pmatrix}, \tag*{(75)} D 5 = ∂ 5 + R ω ( 0 − 1 1 0 ) , (75)
in the normalization adopted here. Acting on the internal gravitino doublet gives
g 1 = ∂ 5 ψ 51 + ω R ψ 52 , (76) g_{1}=\partial_{5}\psi_{51}+\frac{\omega}{R}\psi_{52}, \tag*{(76)} g 1 = ∂ 5 ψ 51 + R ω ψ 52 , (76)
g 2 = ∂ 5 ψ 52 − ω R ψ 51 . (77) g_{2}=\partial_{5}\psi_{52}-\frac{\omega}{R}\psi_{51}. \tag*{(77)} g 2 = ∂ 5 ψ 52 − R ω ψ 51 . (77)
These are precisely the would-be Goldstino combinations entering the gravitino–Goldstino mixing. The same operator that generated the inhomogeneous fermion shifts in Question 6 therefore appears directly in the reduced action.
The R ξ R_{\xi} R ξ gauge
The mixing between ψ μ I \psi_{\mu I} ψ μ I and the Stückelberg fermions can be removed without eliminating either field. Introduce
h 1 = σ μ ψ μ 1 − 6 2 ξ σ μ ∂ μ ∂ 2 g 1 , (78) h_{1}=\sigma^{\mu}\psi_{\mu1}-\frac{\sqrt{6}}{2}\xi\frac{\sigma^{\mu}\partial_{\mu}}{\partial^{2}}g_{1}, \tag*{(78)} h 1 = σ μ ψ μ 1 − 2 6 ξ ∂ 2 σ μ ∂ μ g 1 , (78)
h 2 = σ μ ψ μ 2 − 6 2 ξ σ μ ∂ μ ∂ 2 g 2 . (79) h_{2}=\sigma^{\mu}\psi_{\mu2}-\frac{\sqrt{6}}{2}\xi\frac{\sigma^{\mu}\partial_{\mu}}{\partial^{2}}g_{2}. \tag*{(79)} h 2 = σ μ ψ μ 2 − 2 6 ξ ∂ 2 σ μ ∂ μ g 2 . (79)
and add the fermionic gauge-fixing term
L G F = − i 2 ξ ( h 1 σ μ ∂ μ h 1 + h 2 σ μ ∂ μ h 2 ) + h . c . (80) \mathcal{L}_{\mathrm{GF}}=-\frac{i}{2\xi}\left(h_{1}\sigma^{\mu}\partial_{\mu}h_{1}+h_{2}\sigma^{\mu}\partial_{\mu}h_{2}\right)+\mathrm{h.c.} \tag*{(80)} L GF = − 2 ξ i ( h 1 σ μ ∂ μ h 1 + h 2 σ μ ∂ μ h 2 ) + h.c. (80)
The cross terms generated by Eq. (7.6) cancel the mixing in Eq. (7.2). The separation into a spin-3 2 \frac{3}{2} 2 3 field and a Stückelberg spinor is gauge dependent, while physical poles and gauge-invariant observables are not.
This form of the gauge fixing also makes clear why the discussion is not intrinsically tied to one Kaluza–Klein level. The functions g I g_I g I are five-dimensional objects. After Fourier expansion or, more generally, after diagonalization of the complete fermionic mass matrix, they provide the Goldstino combination associated with each massive gravitino eigenstate.
Unitary gauge and the super-Higgs mechanism
In the convention of Eq. (80) , unitary gauge is obtained for ξ → ∞ \xi\to\infty ξ → ∞ . The would-be Goldstino combinations then obey
g 1 = g 2 = 0 , (81) g_1 = g_2 = 0, \tag*{(81)} g 1 = g 2 = 0 , (81)
or, for the pure bulk Scherk–Schwarz system,
∂ 5 ψ 51 + ω R ψ 52 = 0 , ∂ 5 ψ 52 − ω R ψ 51 = 0. (82) \begin{aligned}
\partial_5 \psi_{51} + \frac{\omega}{R}\psi_{52} &= 0,\\
\partial_5 \psi_{52} - \frac{\omega}{R}\psi_{51} &= 0. \tag*{(82)}
\end{aligned} ∂ 5 ψ 51 + R ω ψ 52 ∂ 5 ψ 52 − R ω ψ 51 = 0 , = 0. (82)
The would-be Goldstinos no longer propagate as independent fields. Their degrees of freedom have become the helicity-1 2 \frac{1}{2} 2 1 components of the massive gravitini.
This is the super-Higgs mechanism in the present compactification. It should be distinguished from an approximation in which only a light four-dimensional gravitino is retained. Even at ω = 0 \omega= 0 ω = 0 , every non-zero Kaluza–Klein gravitino already undergoes its ordinary fermionic Stückelberg mechanism. Turning on ω \omega ω simply extends the same structure to the direction that contained the massless gravitino at the supersymmetric point.
The gravitino spectrum
In unitary gauge, the bulk equations for a four-dimensional mass eigenstate take the form
∂ 5 ψ μ 2 + ( m − ω R ) ψ μ 1 = 0 , ∂ 5 ψ μ 1 − ( m − ω R ) ψ μ 2 = 0. (83) \begin{aligned}
\partial_5 \psi_{\mu2} + \left(m - \frac{\omega}{R}\right)\psi_{\mu1} &= 0,\\
\partial_5 \psi_{\mu1} - \left(m - \frac{\omega}{R}\right)\psi_{\mu2} &= 0. \tag*{(83)}
\end{aligned} ∂ 5 ψ μ 2 + ( m − R ω ) ψ μ 1 ∂ 5 ψ μ 1 − ( m − R ω ) ψ μ 2 = 0 , = 0. (83)
Periodic solutions give the signed eigenvalues
m n e i g e n = n + ω R , n ∈ Z , (84) m_n^{\mathrm{eigen}} = \frac{n+\omega}{R}, \qquad n \in\mathbb{Z}, \tag*{(84)} m n eigen = R n + ω , n ∈ Z , (84)
up to the sign convention for the S U ( 2 ) R SU(2)_R S U ( 2 ) R generator, and hence the physical masses
m n = ∣ n + ω ∣ R . (85) m_n = \frac{\lvert n+\omega\rvert}{R}. \tag*{(85)} m n = R ∣ n + ω ∣ . (85)
The mass spectrum and the Goldstino shifts are therefore governed by the same deformation. At ω = 0 \omega= 0 ω = 0 , the non-zero Kaluza–Klein levels are already massive spin-3 2 \frac{3}{2} 2 3 states with their Stückelberg fermions, while the supersymmetric zero-mode gravitino is massless. Switching on the twist produces
n R ⟶ n + ω R , (86) \frac{n}{R} \longrightarrow\frac{n+\omega}{R}, \tag*{(86)} R n ⟶ R n + ω , (86)
and the direction that was massless at the supersymmetric point now participates in the same super-Higgs structure.
As emphasized in Question 6, the integer label is not itself invariant. For an R R R -charge eigenstate, the physical combination is n + ω q n+\omega q n + ω q , and a change of representative may shift ω q \omega q ω q by an integer while relabelling the entire tower. The super-Higgs mechanism belongs to the shifted spectrum, not to a privileged label called n = 0 n=0 n = 0 .
What the equivalence theorem does, and does not, say
The super-Higgs mechanism above is a statement about the complete quadratic fermionic system. No high-energy limit was used. The high-energy equivalence theorem is a different statement.
After diagonalizing the fermionic system, let ψ ^ μ a \widehat{\psi}_{\mu a} ψ μ a be a gravitino mass eigenstate of physical mass m a m_a m a , and let η ^ a \widehat{\eta}_a η a denote the corresponding Goldstino eigenstate. For an on-shell amplitude containing external helicity-1 2 \frac{1}{2} 2 1 gravitini, the equivalence theorem gives schematically
A ( ψ ^ a 1 ± 1 / 2 , … , ψ ^ a N ± 1 / 2 , X ) = A ( η ^ a 1 , … , η ^ a N , X ) + O ( max i m a i E i ) (87) \mathcal{A}\left(\widehat{\psi}^{\pm1/2}_{a_1},\ldots,\widehat{\psi}^{\pm1/2}_{a_N},X\right)=\mathcal{A}\left(\widehat{\eta}_{a_1},\ldots,\widehat{\eta}_{a_N},X\right)+\mathcal{O}\left(\max_i\frac{m_{a_i}}{E_i}\right) \tag*{(87)} A ( ψ a 1 ± 1/2 , … , ψ a N ± 1/2 , X ) = A ( η a 1 , … , η a N , X ) + O ( i max E i m a i ) (87)
up to conventional phases and normalizations. Here the hard kinematic scales E i E_i E i must be large compared with the masses of the external longitudinal gravitini involved in the process. The theorem does not require the energy to exceed the masses of the entire Kaluza–Klein tower.
It is important not to read more into Eq. (87) than it says. The theorem replaces external longitudinal polarizations by the corresponding Goldstino eigenstates. It does not integrate out the rest of the Kaluza–Klein tower, nor does it turn a genuinely higher-dimensional high-energy process into a theory containing only a finite number of four-dimensional fields.
Its simplest use occurs when the gravitino is parametrically lighter than the Kaluza–Klein scale. For a pure Scherk–Schwarz circle,
m 3 / 2 ≃ ∣ ω ∣ R (88) m_{3/2}\simeq\frac{|\omega|}{R} \tag*{(88)} m 3/2 ≃ R ∣ ω ∣ (88)
so that for ∣ ω ∣ ≪ 1 |\omega|\ll1 ∣ ω ∣ ≪ 1 there is a parametrically large window
m 3 / 2 ≪ E ≪ 1 R (89) m_{3/2}\ll E\ll\frac{1}{R} \tag*{(89)} m 3/2 ≪ E ≪ R 1 (89)
In this regime the excited Kaluza–Klein states are genuinely decoupled and the equivalence theorem reduces to the familiar four-dimensional supergravity statement. From the low-energy gravitino and its eaten Goldstino alone, one need not be able to reconstruct whether the microscopic supersymmetry breaking originated from a bulk Scherk–Schwarz background or from a localized dynamical sector: different higher-dimensional realizations can match onto the same light four-dimensional super-Higgs system.
The situation changes as ∣ ω ∣ |\omega| ∣ ω ∣ becomes of order one. The gravitino mass then approaches the compactification scale and the window (89) disappears. The super-Higgs mechanism itself remains perfectly well defined: it can still be followed through the Kaluza–Klein gravitino and Goldstino system. What is lost is the parametrically controlled interpretation in terms of one isolated four-dimensional gravitino.
A finite truncation of the tower can still display the Stückelberg and super-Higgs structure of the modes that have been retained. Such a truncation should not, however, be used by itself to infer the high-energy behaviour of the complete compactified theory. Once the energy probes the Kaluza–Klein scale, additional states become dynamically relevant, and cancellations or sum rules involving the tower can matter for questions such as perturbative unitarity. The equivalence theorem does not remove this requirement.
This point becomes particularly sharp near the half-twist ω = 1 / 2 \omega= 1/2 ω = 1/2 . There is then no parametrically isolated light gravitino: the two lightest Majorana states are already degenerate at 1 / ( 2 R ) 1/(2R) 1/ ( 2 R ) . Their individual super-Higgs structure may still be followed in the Majorana description, while their recombination into a Dirac gravitino is a separate symmetry statement that will be established later in the review.
Localized breaking and additional Goldstino directions
Boundaries can modify this picture in two logically distinct ways. A localized gravitino bilinear may simply represent another distribution of the same R R R -symmetry rotation between bulk and boundary data. In the elementary interval problem, localized masses M 0 M_0 M 0 and M π M_\pi M π give
m n e i g e n = 1 R [ n + ω + 1 π ( arctan M 0 + arctan M π ) ] , (90) m_n^{\mathrm{eigen}} = \frac{1}{R}\left[n+\omega+\frac{1}{\pi}\left(\arctan M_0+\arctan M_\pi\right)\right], \tag*{(90)} m n eigen = R 1 [ n + ω + π 1 ( arctan M 0 + arctan M π ) ] , (90)
in the normalization adopted here. A cancellation among the bulk and boundary contributions can therefore produce a gravitino much lighter than 1 / R 1/R 1/ R , again leaving a genuine four-dimensional window in which the equivalence theorem is useful.
A genuine localized supersymmetry-breaking sector contains more information than such a boundary rotation. Its fermions may themselves have inhomogeneous supersymmetry transformations and therefore add independent would-be Goldstino directions. The physical Goldstino eaten by a given gravitino is then a combination of bulk and localized fermions. If the number of independent would-be Goldstino directions exceeds the number of broken local supersymmetries, orthogonal combinations remain in the physical spectrum as pseudo-Goldstinos.
Consequently, two compactifications can have the same light gravitino mass while differing in their low-energy fermionic content. In particular, pseudo-Goldstinos can retain information about several independent supersymmetry-breaking sectors that is invisible in the gravitino mass alone. The distinction between localized gravitino rotations and genuine localized supersymmetry breaking will be treated systematically when boundary terms are discussed.
Beyond the flat circle
The explicit formulae above exploit the simplicity of flat five-dimensional supergravity on a circle or interval. On a more general compact space there need be no Fourier label n n n and no universal shift ( n + ω ) / R (n+\omega)/R ( n + ω ) / R . The structural relation survives: the internal supersymmetry operator determines the fermion shifts and enters the spin-3 / 2 3/2 3/2 spectral problem, while the reduced Rarita–Schwinger action determines how the corresponding Stückelberg directions are absorbed.
The flat circle is therefore not special because the super-Higgs mechanism exists only there, but because the entire construction can be followed analytically from the higher-dimensional gauge transformation to the complete gravitino spectrum.
The Goldstino associated with spontaneous supersymmetry breaking and its absorption by the gravitino were developed in the early literature on supersymmetry and supergravity [20, 21, 22, 23, 24]. The high-energy equivalence theorem for the longitudinal gravitino was formulated and established in Refs. [25, 26].
The role of the complete Kaluza–Klein tower in the super-Higgs mechanism, including the possibility that the physical gravitino and Goldstino are mixtures of many Fourier modes, was analysed in Ref. [27]. The equivalence theorem for Kaluza–Klein gravitino mass eigenstates, together with the corresponding four- and five-dimensional R ξ R_{\xi} R ξ gauges, was developed in Ref. [28]. An explicit treatment including both bulk Scherk–Schwarz breaking and localized supersymmetry-breaking F F F -terms, in general R ξ R_{\xi} R ξ and unitary gauges, as well as the resulting pseudo-Goldstinos, was given in Ref. [29].
Generalized Scherk–Schwarz boundary conditions and localized gravitino masses in five-dimensional supergravity were developed in Refs. [30, 31, 32, 33]. We shall return to this literature more systematically when discussing orbifolds, intervals, and localized boundary terms.
Take-home message
The super-Higgs mechanism in a Scherk–Schwarz compactification is the deformation of the fermionic Stückelberg structure already present in the ordinary Kaluza–Klein gravitino tower.
The same twisted internal operator controls the Goldstino shifts, the gravitino–Goldstino mixing, and the spin-3 2 \frac{3}{2} 2 3 spectrum. An R ξ R_{\xi} R ξ gauge separates the two sectors explicitly, while unitary gauge removes the would-be Goldstinos.
On a flat circle the gravitino spectrum is m n = ∣ n + ω ∣ / R m_n = \lvert n + \omega\rvert/R m n = ∣ n + ω ∣ / R , up to R R R -charge normalization and Kaluza–Klein relabelling.
The super-Higgs mechanism does not require m 3 / 2 ≪ 1 / R m_{3/2} \ll1/R m 3/2 ≪ 1/ R . The latter hierarchy is instead what permits a controlled four-dimensional truncation to an isolated light gravitino.
When m 3 / 2 ≪ 1 / R m_{3/2} \ll1/R m 3/2 ≪ 1/ R , the high-energy equivalence theorem can be used inside the four-dimensional window m 3 / 2 ≪ E ≪ 1 / R m_{3/2} \ll E \ll1/R m 3/2 ≪ E ≪ 1/ R . For an order-one twist this window disappears: the theorem does not justify replacing the full Kaluza–Klein system by a finite four-dimensional truncation.
Localized gravitino rotations can modify the effective spectral shift, whereas genuine localized supersymmetry-breaking sectors can introduce additional Goldstino directions. Orthogonal uneaten combinations may survive as pseudo-Goldstinos.
This naturally leads to the next question:
How does Scherk–Schwarz compactification work on orbifolds and intervals?
How does Scherk–Schwarz compactification work on orbifolds and intervals?
Essential idea
The orbifold S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 is most cleanly described from its covering space R \mathbb{R} R . Its orbifold group is generated by the two reflections
r 0 : y ↦ − y , r π : y ↦ 2 π R − y , r_0 : y \mapsto-y,\qquad r_\pi: y \mapsto2\pi R-y, r 0 : y ↦ − y , r π : y ↦ 2 π R − y ,
whose product is the translation by one circumference,
t = r π r 0 , t ( y ) = y + 2 π R . t=r_\pi r_0,\qquad t(y)=y+2\pi R. t = r π r 0 , t ( y ) = y + 2 π R .
For a chosen equivariant lift of the orbifold action to a field Φ \Phi Φ , let
P 0 ( Φ ) = ρ Φ ( r 0 ) , P π ( Φ ) = ρ Φ ( r π ) . P_0^{(\Phi)}=\rho_\Phi(r_0),\qquad P_\pi^{(\Phi)}=\rho_\Phi(r_\pi). P 0 ( Φ ) = ρ Φ ( r 0 ) , P π ( Φ ) = ρ Φ ( r π ) .
Then
( P 0 ( Φ ) ) 2 = ( P π ( Φ ) ) 2 = 1 , U c o v ( Φ ) = P π ( Φ ) P 0 ( Φ ) , \left(P_0^{(\Phi)}\right)^2=\left(P_\pi^{(\Phi)}\right)^2=1,\qquad U_{\mathrm{cov}}^{(\Phi)}=P_\pi^{(\Phi)}P_0^{(\Phi)}, ( P 0 ( Φ ) ) 2 = ( P π ( Φ ) ) 2 = 1 , U cov ( Φ ) = P π ( Φ ) P 0 ( Φ ) ,
and therefore
P 0 ( Φ ) U c o v ( Φ ) P 0 ( Φ ) = ( U c o v ( Φ ) ) − 1 . P_0^{(\Phi)}U_{\mathrm{cov}}^{(\Phi)}P_0^{(\Phi)}=\left(U_{\mathrm{cov}}^{(\Phi)}\right)^{-1}. P 0 ( Φ ) U cov ( Φ ) P 0 ( Φ ) = ( U cov ( Φ ) ) − 1 .
These relations follow from the orbifold group. They constrain how a Scherk–Schwarz deformation can be represented, but they do not fix its value. A continuous twist is additional compactification data. It may be represented by an ω \omega ω -dependent relative orientation of the endpoint involutions or, in another trivialization, by a flat bulk connection.
The quotient has the interval 0 ≤ y ≤ π R 0\leq y\leq\pi R 0 ≤ y ≤ π R as its underlying topological space. In the interval description, the relevant global information is carried jointly by the endpoint projections and by the parallel transport between them. Supersymmetry survives only when these data admit a common global Killing spinor.
Why orbifolds?
The circle is the natural starting point for a pedagogical discussion of Scherk–Schwarz monodromy. It is smooth, has a single non-contractible direction, and makes the basic identification particularly transparent. For phenomenological applications, however, it is often too simple. Starting, for example, from five-dimensional supergravity, compactification on a circle does not by itself produce a chiral four-dimensional fermion spectrum. One therefore wants a compactification that can project out part of the higher-dimensional field content and, at the same time, reduce the amount of supersymmetry seen in the lower-dimensional theory.
Orbifolds provide the simplest way of doing this. Their appearance in high-energy physics came most prominently through string compactification, although the mathematical notion is older. Satake introduced what he called V V V -manifolds in the 1950s, namely spaces that are locally quotients of smooth manifolds by finite groups. Thurston later developed this point of view extensively and popularized the term “orbifold”. In string theory, the construction acquired a much more concrete physical meaning in the work of Dixon, Harvey, Vafa and Witten: strings were quantized on quotients of smooth compactification spaces by discrete symmetries, with the quotient supplemented by the twisted sectors required for a consistent worldsheet theory.
This last point is worth keeping in mind. An orbifold in string theory is not merely a singular quotient space on which one decides to put fields. The quotient is part of a microscopic construction, and worldsheet consistency supplies additional sectors that have no analogue in a naive field theory obtained by simply deleting points or imposing boundary conditions. In a low-energy field-theory or supergravity treatment, by contrast, one can work directly with the quotient or with a fundamental domain and specify the corresponding projection, equivariance and localized data explicitly. The present question adopts this latter viewpoint. Later, when we return to string theory, the additional microscopic requirements will become important again.
Our basic example is S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 , whose underlying topological space is an interval. Relative to the circle, however, a bulk field must now be assigned not only a monodromy around the covering circle but also transformation laws under the two reflections whose fixed loci become the endpoints. These data cannot be chosen independently: together they must define a representation of the full orbifold group.
This is where Scherk–Schwarz compactification on an orbifold differs from simply taking the circle formula and cutting the circle in half. The twist, the two endpoint actions and the covering-space translation are different descriptions of related group data, and their compatibility must be imposed before discussing the spectrum or supersymmetry.
We will first describe the orbifold as a quotient of its covering line, then lift the group action to the fields, and finally ask how local endpoint projections can coexist with, or obstruct, a globally defined supersymmetry. Localized dynamics at the fixed points is a separate issue and will be left to the next question.
The orbifold as a quotient of the covering line
Begin with the real line R \mathbb{R} R , with coordinate y y y . Define the two reflections
r 0 ( y ) = − y , r π ( y ) = 2 π R − y . r_0(y)=-y,\qquad r_\pi(y)=2\pi R-y. r 0 ( y ) = − y , r π ( y ) = 2 π R − y .
Both are involutions, r 0 2 = r π 2 = id R r_0^2=r_\pi^2=\operatorname{id}_{\mathbb{R}} r 0 2 = r π 2 = id R , and their product is the translation
t ≡ r π r 0 , t ( y ) = y + 2 π R . (91) t\equiv r_\pi r_0,\qquad t(y)=y+2\pi R. \tag*{(91)} t ≡ r π r 0 , t ( y ) = y + 2 π R . (91)
The group generated by the two reflections is the infinite dihedral group,
D ∞ = ⟨ r 0 , r π ∣ r 0 2 = r π 2 = 1 ⟩ . (92) D_\infty=\langle r_0,r_\pi\mid r_0^2=r_\pi^2=1\rangle. \tag*{(92)} D ∞ = ⟨ r 0 , r π ∣ r 0 2 = r π 2 = 1 ⟩ . (92)
Using t = r π r 0 t=r_{\pi}r_0 t = r π r 0 , the same group can equivalently be presented as
D ∞ = ⟨ t , r 0 ∣ r 0 2 = 1 , r 0 t r 0 = t − 1 ⟩ . D_{\infty}=\langle t,r_0\mid r_0^2=1,\quad r_0tr_0=t^{-1}\rangle. D ∞ = ⟨ t , r 0 ∣ r 0 2 = 1 , r 0 t r 0 = t − 1 ⟩ .
The translation subgroup ⟨ t ⟩ ≃ Z \langle t\rangle\simeq\mathbb{Z} ⟨ t ⟩ ≃ Z produces the circle,
S 1 = R / ( 2 π R Z ) , S^1=\mathbb{R}/(2\pi R\mathbb{Z}), S 1 = R / ( 2 π R Z ) ,
and quotienting in addition by a reflection gives
S 1 / Z 2 ≃ R / D ∞ . (93) S^1/\mathbb{Z}_2\simeq\mathbb{R}/D_{\infty}. \tag*{(93)} S 1 / Z 2 ≃ R / D ∞ . (93)
Its underlying topological space is the interval [ 0 , π R ] [0,\pi R] [ 0 , π R ] . What distinguishes the orbifold description from a bare interval is not the point-set topology, nor necessarily the field content, but the isotropy and equivariance data inherited from the covering-space action.
The endpoints y = 0 y=0 y = 0 and y = π R y=\pi R y = π R are fixed loci with stabilizer Z 2 \mathbb{Z}_2 Z 2 . Bulk fields carry compatible lifts of both reflections. A degree of freedom intrinsically localized at one endpoint, on the other hand, only has to transform consistently under the stabilizer of that endpoint; it need not carry the complete bulk monodromy around the covering circle.
An ordinary field theory formulated directly on an interval can therefore reproduce the same physics if the corresponding endpoint actions, bulk identifications and consistency conditions are supplied by hand. The orbifold description packages these data into the group action from which they originate.
Fields as equivariant sections
Let V Φ V_{\Phi} V Φ denote the local representation space in which the components of a field Φ \Phi Φ transform. For a scalar multiplet this may simply be an internal multiplet space; for a spinor, V Φ V_{\Phi} V Φ contains the corresponding spin representation, possibly together with gauge or R R R -symmetry indices. Globally, Φ \Phi Φ is a section of the appropriate associated bundle, but V Φ V_{\Phi} V Φ is sufficient for describing the orbifold action locally.
An element g ∈ D ∞ g\in D_{\infty} g ∈ D ∞ sends y y y to g ⋅ y g\cdot y g ⋅ y , while the field components at the two points are related by an allowed transformation. Thus one imposes
Φ ( g ⋅ y ) = ρ Φ ( g ) Φ ( y ) , (94) \Phi(g\cdot y)=\rho_{\Phi}(g)\Phi(y), \tag*{(94)} Φ ( g ⋅ y ) = ρ Φ ( g ) Φ ( y ) , (94)
rather than requiring the field to be invariant under every geometrical operation.
For a single real scalar, ρ Φ ( g ) \rho_{\Phi}(g) ρ Φ ( g ) may reduce to a sign, giving the familiar even or odd parity. For a multiplet it may rotate, reflect or permute components, for example through a gauge or R R R -symmetry transformation. Since successive geometrical operations must be represented consistently on the fields,
ρ Φ : D ∞ ⟶ G L ( V Φ ) \rho_{\Phi}:D_{\infty}\longrightarrow GL(V_{\Phi}) ρ Φ : D ∞ ⟶ G L ( V Φ )
is a representation whose image is built from transformations allowed by the parent theory.
For the two reflections define
P 0 ( Φ ) = ρ Φ ( r 0 ) , P π ( Φ ) = ρ Φ ( r π ) . P_0^{(\Phi)}=\rho_{\Phi}(r_0),\qquad P_{\pi}^{(\Phi)}=\rho_{\Phi}(r_{\pi}). P 0 ( Φ ) = ρ Φ ( r 0 ) , P π ( Φ ) = ρ Φ ( r π ) .
Their involutive character gives
( P 0 ( Φ ) ) 2 = ( P π ( Φ ) ) 2 = 1. \left(P_{0}^{(\Phi)}\right)^{2}=\left(P_{\pi}^{(\Phi)}\right)^{2}=1. ( P 0 ( Φ ) ) 2 = ( P π ( Φ ) ) 2 = 1.
Since t = r π r 0 t=r_{\pi}r_{0} t = r π r 0 , the lift of the translation is not independent:
U c o v ( Φ ) = ρ Φ ( t ) = P π ( Φ ) P 0 ( Φ ) . (95) U_{\mathrm{cov}}^{(\Phi)}=\rho_{\Phi}(t)=P_{\pi}^{(\Phi)}P_{0}^{(\Phi)}. \tag*{(95)} U cov ( Φ ) = ρ Φ ( t ) = P π ( Φ ) P 0 ( Φ ) . (95)
Hence
Φ ( y + 2 π R ) = U c o v ( Φ ) Φ ( y ) . \Phi(y+2\pi R)=U_{\mathrm{cov}}^{(\Phi)}\Phi(y). Φ ( y + 2 π R ) = U cov ( Φ ) Φ ( y ) .
The two endpoint involutions therefore determine the monodromy seen after one complete turn around the covering circle. In this trivialization, the Scherk–Schwarz twist is encoded in their relative action rather than introduced as an unrelated third operation.
Covering-space monodromy and equivalent trivializations
The relation
U c o v ( Φ ) = P π ( Φ ) P 0 ( Φ ) U_{\mathrm{cov}}^{(\Phi)}=P_{\pi}^{(\Phi)}P_{0}^{(\Phi)} U cov ( Φ ) = P π ( Φ ) P 0 ( Φ )
does not mean that the geometry of S 1 / Z 2 S^{1}/\mathbb{Z}_{2} S 1 / Z 2 chooses a Scherk–Schwarz parameter. The geometry fixes the orbifold group relations; how those relations are represented on the fields is part of the compactification data.
For a continuous Scherk–Schwarz deformation, one possible description places the parameter in the relative orientation of the two endpoint involutions, so that U c o v ( Φ ) = U S S ( Φ ) ( ω ) U_{\mathrm{cov}}^{(\Phi)}=U_{\mathrm{SS}}^{(\Phi)}(\omega) U cov ( Φ ) = U SS ( Φ ) ( ω ) .
The same compactification may instead be described with fixed endpoint involutions and with the Scherk–Schwarz dependence moved into the compact derivative by a y y y -dependent change of field basis,
D y = ∂ y + A y , A y ∝ ω R R Φ . (96) D_y=\partial_y+\mathcal{A}_y,\qquad\mathcal{A}_y\propto\frac{\omega}{R}\mathcal{R}_{\Phi}. \tag*{(96)} D y = ∂ y + A y , A y ∝ R ω R Φ . (96)
At this level A y \mathcal{A}_y A y records the same monodromy in a periodic field basis. When the twisting symmetry is realized locally in supergravity, the same object can be the internal component of the corresponding symmetry connection, and the description becomes a Wilson-line background in the usual gauge-theory sense.
For the backgrounds considered here the connection is flat, so its local curvature vanishes even though its parallel transport can carry non-trivial global information. The two descriptions differ only in the choice of field basis used to compare fibres at different points: the same data can be moved between the relative endpoint orientation and a flat bulk connection without changing the compactification.
On the interval this is particularly transparent. A flat connection can be removed from the bulk by a y y y -dependent transformation, but that transformation changes the relative field bases at y = 0 y=0 y = 0 and y = π R y=\pi R y = π R , and hence changes the endpoint data. On the covering circle the corresponding transformation need not be single-valued; its failure to be single-valued retains the global monodromy.
The orbifold group law also restricts the admissible endpoint data. Suppressing the label ( Φ ) (\Phi) ( Φ ) , one has P π = U c o v P 0 P_{\pi}=U_{\mathrm{cov}}P_{0} P π = U cov P 0 . Together with P 0 2 = P π 2 = 1 P_{0}^{2}=P_{\pi}^{2}=1 P 0 2 = P π 2 = 1 , this implies
P 0 U c o v P 0 = U c o v − 1 . (97) P_{0}U_{\mathrm{cov}}P_{0}=U_{\mathrm{cov}}^{-1}. \tag*{(97)} P 0 U cov P 0 = U cov − 1 . (97)
This is simply the representation on V Φ V_{\Phi} V Φ of r 0 t r 0 = t − 1 r_{0}t r_{0}=t^{-1} r 0 t r 0 = t − 1 .
The continuous Scherk–Schwarz freedom does not arise by relaxing this relation. It arises because a fixed group relation can admit continuous families of representations on the field space.
Continuous twists and the orbifold compatibility condition
Take a continuous Scherk–Schwarz deformation generated by a one-parameter subgroup of an allowed symmetry of the parent theory. For a field Φ \Phi Φ , choose the convention
U S S ( Φ ) ( ω ) = exp ( 2 π i ω R Φ ) , (98) U_{\mathrm{SS}}^{(\Phi)}(\omega)=\exp(2\pi i\omega R_{\Phi}), \tag*{(98)} U SS ( Φ ) ( ω ) = exp ( 2 π iω R Φ ) , (98)
where R Φ R_{\Phi} R Φ represents the chosen generator on V Φ V_{\Phi} V Φ .
Work in the trivialization in which the flat bulk connection has been gauged away. The Scherk–Schwarz data then reside in the relative endpoint involutions:
U c o v ( Φ ) = U S S ( Φ ) ( ω ) , P π ( Φ ) ( ω ) = U S S ( Φ ) ( ω ) P 0 ( Φ ) . U_{\mathrm{cov}}^{(\Phi)}=U_{\mathrm{SS}}^{(\Phi)}(\omega),\qquad P_{\pi}^{(\Phi)}(\omega)=U_{\mathrm{SS}}^{(\Phi)}(\omega)P_{0}^{(\Phi)}. U cov ( Φ ) = U SS ( Φ ) ( ω ) , P π ( Φ ) ( ω ) = U SS ( Φ ) ( ω ) P 0 ( Φ ) .
Suppose P 0 ( Φ ) P_{0}^{(\Phi)} P 0 ( Φ ) is fixed and this whole one-parameter family satisfies the orbifold compatibility condition near the identity. Equation (8.7) gives
P 0 ( Φ ) e 2 π i ω R Φ P 0 ( Φ ) = e − 2 π i ω R Φ . P_{0}^{(\Phi)}e^{2\pi i\omega R_{\Phi}}P_{0}^{(\Phi)}=e^{-2\pi i\omega R_{\Phi}}. P 0 ( Φ ) e 2 π iω R Φ P 0 ( Φ ) = e − 2 π iω R Φ .
Differentiating at ω = 0 \omega=0 ω = 0 yields
P 0 ( Φ ) R Φ P 0 ( Φ ) = − R Φ , or equivalently { P 0 ( Φ ) , R Φ } = 0. (99) P_{0}^{(\Phi)}R_{\Phi}P_{0}^{(\Phi)}=-R_{\Phi},\qquad\text{or equivalently}\qquad\{P_{0}^{(\Phi)},R_{\Phi}\}=0. \tag*{(99)} P 0 ( Φ ) R Φ P 0 ( Φ ) = − R Φ , or equivalently { P 0 ( Φ ) , R Φ } = 0. (99)
This statement concerns the generator of a continuous family, not an isolated group element. A special value of U S S U_{\mathrm{SS}} U SS can satisfy the orbifold relation even when an arbitrarily chosen logarithm of that element does not obey Eq. (8.9).
The orbifold geometry therefore constrains which continuous deformations are compatible with a chosen endpoint involution, but it does not determine ω \omega ω . The latter is additional compactification data.
A simple S U ( 2 ) SU(2) S U ( 2 ) example
Consider an S U ( 2 ) SU(2) S U ( 2 ) doublet and choose P 0 = σ 3 P_{0}=\sigma_{3} P 0 = σ 3 . Let the continuous deformation lie along the σ 2 \sigma_{2} σ 2 direction,
U S S ( ω ) = e 2 π i ω σ 2 . (100) U_{\mathrm{SS}}(\omega)=e^{2\pi i\omega\sigma_{2}}. \tag*{(100)} U SS ( ω ) = e 2 π iω σ 2 . (100)
With this normalization ω ∼ ω + 1 \omega\sim\omega+1 ω ∼ ω + 1 , since the eigenvalues of σ 2 \sigma_{2} σ 2 are ± 1 \pm1 ± 1 .
Because σ 3 σ 2 σ 3 = − σ 2 \sigma_{3}\sigma_{2}\sigma_{3}=-\sigma_{2} σ 3 σ 2 σ 3 = − σ 2 , one immediately obtains P 0 U S S P 0 = U S S − 1 P_{0}U_{\mathrm{SS}}P_{0}=U_{\mathrm{SS}}^{-1} P 0 U SS P 0 = U SS − 1 . The orbifold relation is therefore satisfied for every ω \omega ω , and the second endpoint involution is
P π ( ω ) = U S S ( ω ) P 0 = e 2 π i ω σ 2 σ 3 . (101) P_{\pi}(\omega)=U_{\mathrm{SS}}(\omega)P_{0}=e^{2\pi i\omega\sigma_{2}}\sigma_{3}. \tag*{(101)} P π ( ω ) = U SS ( ω ) P 0 = e 2 π iω σ 2 σ 3 . (101)
Indeed,
P π ( ω ) 2 = U S S P 0 U S S P 0 = 1. P_{\pi}(\omega)^{2}=U_{\mathrm{SS}}P_{0}U_{\mathrm{SS}}P_{0}=1. P π ( ω ) 2 = U SS P 0 U SS P 0 = 1.
The same matrix can be written as
P π ( ω ) = e i π ω σ 2 P 0 e − i π ω σ 2 . (102) P_{\pi}(\omega) = e^{i\pi\omega\sigma_2}P_0e^{-i\pi\omega\sigma_2}. \tag*{(102)} P π ( ω ) = e iπ ω σ 2 P 0 e − iπ ω σ 2 . (102)
The two endpoint involutions are therefore conjugate, and their + 1 +1 + 1 eigenspaces are related by the corresponding S U ( 2 ) SU(2) S U ( 2 ) rotation. The parameter ω \omega ω measures their relative orientation.
Nothing in the orbifold group changes as ω \omega ω varies; only its representation on the doublet changes. At the half-twist, ω = 1 2 \omega= \frac{1}{2} ω = 2 1 , one has U S S ( 1 2 ) = − 1 U_{\mathrm{SS}}\left(\frac{1}{2}\right) = -1 U SS ( 2 1 ) = − 1 . We shall return to this special point in Question 11.
Local projections and global supersymmetry
The orbifold action must also be specified on the supersymmetry parameter. Let P 0 \mathcal{P}_0 P 0 and P π \mathcal{P}_{\pi} P π denote the full involutions acting on the supersymmetry-parameter space S S S :
ϵ ( − y ) = P 0 ϵ ( y ) , ϵ ( 2 π R − y ) = P π ϵ ( y ) . (103) \begin{aligned}
\epsilon(-y) &= \mathcal{P}_0\epsilon(y),\\
\epsilon(2\pi R-y) &= \mathcal{P}_{\pi}\epsilon(y). \tag*{(103)}
\end{aligned} ϵ ( − y ) ϵ ( 2 π R − y ) = P 0 ϵ ( y ) , = P π ϵ ( y ) . (103)
They obey P 0 2 = P π 2 = 1 \mathcal{P}_0^2 = \mathcal{P}_{\pi}^2 = 1 P 0 2 = P π 2 = 1 .
The notation P b \mathcal{P}_b P b distinguishes the complete action on the supersymmetry parameter from the internal matrices that may act on other fields. For a spinorial parameter, P b \mathcal{P}_b P b may combine the spinorial action of the reflection with an internal R R R -symmetry action, depending on the parent theory.
The supersymmetries locally compatible with the two fixed points are the respective + 1 +1 + 1 eigenspaces
S 0 = ker ( P 0 − 1 ) , S π = ker ( P π − 1 ) . (104) S_0 = \operatorname{ker}(\mathcal{P}_0 - 1),\qquad S_{\pi} = \operatorname{ker}(\mathcal{P}_{\pi} - 1). \tag*{(104)} S 0 = ker ( P 0 − 1 ) , S π = ker ( P π − 1 ) . (104)
First choose a trivialization in which the flat Scherk–Schwarz connection has been removed from the bulk. The Killing-spinor equation along the interval then reduces to ∂ y ϵ = 0 \partial_y\epsilon= 0 ∂ y ϵ = 0 , while the relative endpoint involutions carry the Scherk–Schwarz information. A globally preserved supersymmetry must satisfy both endpoint conditions, so
S g l o b a l = S 0 ∩ S π . (105) S_{\mathrm{global}} = S_0 \cap S_{\pi}. \tag*{(105)} S global = S 0 ∩ S π . (105)
It is perfectly possible for both dim S 0 \dim S_0 dim S 0 and dim S π \dim S_{\pi} dim S π to be non-zero while S 0 ∩ S π = { 0 } S_0 \cap S_{\pi} = \{0\} S 0 ∩ S π = { 0 } . Each endpoint then admits a local supersymmetry subspace, but no common supersymmetry extends over the whole interval.
In another trivialization the endpoint involutions may be kept fixed while the same Scherk–Schwarz deformation is placed in a flat bulk connection. The Killing-spinor equation becomes
D 5 ϵ = ( ∂ 5 + A 5 ( y ) ) ϵ = 0. (106) D_5\epsilon= \left(\partial_5 + \mathcal{A}_5(y)\right)\epsilon= 0. \tag*{(106)} D 5 ϵ = ( ∂ 5 + A 5 ( y ) ) ϵ = 0. (106)
For an initial value ϵ 0 = ϵ ( 0 ) \epsilon_0 = \epsilon(0) ϵ 0 = ϵ ( 0 ) , its solution is
ϵ ( y ) = U ( y , 0 ) ϵ 0 , U ( y , 0 ) = P exp [ − ∫ 0 y d s A 5 ( s ) ] . (107) \epsilon(y) = \mathcal{U}(y,0)\epsilon_0,\qquad\mathcal{U}(y,0) = \mathcal{P}\exp\left[-\int_0^y ds\,\mathcal{A}_5(s)\right]. \tag*{(107)} ϵ ( y ) = U ( y , 0 ) ϵ 0 , U ( y , 0 ) = P exp [ − ∫ 0 y d s A 5 ( s ) ] . (107)
The endpoint conditions then require
ϵ 0 ∈ S 0 , U ( π R , 0 ) ϵ 0 ∈ S π . \epsilon_0 \in S_0,\qquad\mathcal{U}(\pi R,0)\epsilon_0 \in S_{\pi}. ϵ 0 ∈ S 0 , U ( π R , 0 ) ϵ 0 ∈ S π .
Subject to any remaining Killing-spinor equations imposed by the background, global supersymmetry exists precisely when
S 0 ∩ U ( π R , 0 ) − 1 S π ≠ { 0 } . (108) \mathcal{S}_0 \cap\mathcal{U}(\pi R,0)^{-1}\mathcal{S}_{\pi} \ne\{0\}. \tag*{(108)} S 0 ∩ U ( π R , 0 ) − 1 S π = { 0 } . (108)
Equivalently, U ( π R , 0 ) S 0 ∩ S π ≠ { 0 } \mathcal{U}(\pi R,0)\mathcal{S}_0 \cap\mathcal{S}_{\pi} \ne\{0\} U ( π R , 0 ) S 0 ∩ S π = { 0 } .
One may instead compare both endpoint conditions in the fibre over y = 0 y=0 y = 0 . Transporting the second endpoint involution back to that fibre gives
P ^ π = U ( π R , 0 ) − 1 P π U ( π R , 0 ) , \widehat{\mathcal{P}}_{\pi}=\mathcal{U}(\pi R,0)^{-1}\mathcal{P}_{\pi}\mathcal{U}(\pi R,0), P π = U ( π R , 0 ) − 1 P π U ( π R , 0 ) ,
so that
S g l o b a l = ker ( P 0 − 1 ) ∩ ker ( P ^ π − 1 ) , (109) \mathcal{S}_{\mathrm{global}}=\ker(\mathcal{P}_0-1)\cap\ker(\widehat{\mathcal{P}}_{\pi}-1), \tag*{(109)} S global = ker ( P 0 − 1 ) ∩ ker ( P π − 1 ) , (109)
again subject to the other Killing-spinor equations of the background.
The endpoint involutions and the bulk transport must always be written in the same trivialization. A change of trivialization can move the same Scherk–Schwarz datum between relative endpoint orientation and flat bulk connection. Combining the two descriptions as independent ingredients would count the twist twice.
From orbifold parity to boundary conditions
For completeness, consider a smooth scalar near y = 0 y=0 y = 0 satisfying
ϕ ( − y ) = p ϕ ( y ) , p = ± 1. \phi(-y)=p\phi(y),\qquad p=\pm1. ϕ ( − y ) = pϕ ( y ) , p = ± 1.
For odd parity, evaluation at the fixed point gives ϕ ( 0 ) = 0 \phi(0)=0 ϕ ( 0 ) = 0 . For even parity, differentiation of the parity relation gives ∂ y ϕ ( 0 ) = 0 \partial_y\phi(0)=0 ∂ y ϕ ( 0 ) = 0 .
Thus, for a smooth scalar in the absence of localized interactions, odd parity gives Dirichlet data and even parity gives Neumann data. The same reasoning applies at y = π R y=\pi R y = π R in the local coordinate y − π R y-\pi R y − π R .
These conditions follow from smooth equivariance; they are not the general definition of boundary conditions on an orbifold. Localized operators modify the variational problem and can produce Robin, jump, or more general matching conditions. This is the subject of the next question.
The mathematical notion underlying orbifolds can be traced to Satake’s V V V -manifolds and was developed further in the work of Thurston. Orbifold compactifications in string theory are described in Refs. [34, 35].
The ingredients needed for Scherk–Schwarz supersymmetry breaking already appear in the original generalized reductions Refs. [6, 7], where compact-coordinate dependence generated by a symmetry is used to produce fermion masses. The appearance of the compact component of a local internal-symmetry connection in the fermionic mass matrix can be seen explicitly in Ref. [36].
For S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 supergravity, descriptions in which the Scherk–Schwarz data are distributed between endpoint actions, localized terms and a flat R R R -symmetry connection can be found in Refs. [30, 37, 38, 39].
Take-home message
P 0 2 = P π 2 = 1 , U c o v = P π P 0 , P 0 U c o v P 0 = U c o v − 1 . P_0^2=P_\pi^2=1,\qquad U_{\mathrm{cov}}=P_\pi P_0,\qquad P_0U_{\mathrm{cov}}P_0=U_{\mathrm{cov}}^{-1}. P 0 2 = P π 2 = 1 , U cov = P π P 0 , P 0 U cov P 0 = U cov − 1 .
These relations follow from the orbifold group and do not by themselves fix a Scherk–Schwarz parameter.
P 0 ( Φ ) ( R Φ P 0 ( Φ ) = − R Φ . P_0^{(\Phi)}\mathop{\mathstrut R}\nolimits_{\Phi}P_0^{(\Phi)}=-R_{\Phi}. P 0 ( Φ ) ( R Φ P 0 ( Φ ) = − R Φ .
S 0 ∩ U ( π R , 0 ) − 1 S π ≠ { 0 } , S_0\cap U(\pi R,0)^{-1}S_\pi\ne\{0\}, S 0 ∩ U ( π R , 0 ) − 1 S π = { 0 } ,
with all quantities expressed in one common trivialization.
For smooth bulk scalars without localized interactions, odd and even orbifold parities imply Dirichlet and Neumann data, respectively. Localized terms can modify these dynamical boundary or matching conditions.
This naturally leads to the next question:
How do localized boundary terms modify Scherk–Schwarz breaking?
How do localized boundary terms modify Scherk–Schwarz breaking?
Essential idea
Localized operators do not modify the orbifold group relations established in Question 8. They modify the dynamical and spectral problem obeyed by the fields.
A localized potential, a boundary mass, a boundary kinetic term, or a fermion bilinear changes the variational problem and can therefore modify the boundary or matching conditions, the Kaluza–Klein eigenvalues, the wave functions and their normalization, and the couplings of bulk modes to localized sectors.
For a second-order bulk field, a localized mass produces a Robin condition. A kinetic term containing derivatives tangent to the boundary produces an eigenvalue-dependent boundary condition and changes the inner product with respect to which the Kaluza–Klein modes are normalized. The resulting problem is naturally a generalized Sturm–Liouville problem.
For first-order fermionic systems, gravitini among them, localized bilinears instead give matching conditions for the fields themselves. A pure localized mass can rotate the allowed fermionic subspace at an endpoint by an energy-independent angle, whereas a localized kinetic term generically makes this rotation depend on the four-dimensional eigenvalue.
This distinction limits the familiar statement that a boundary term can be traded for a Scherk–Schwarz twist. A boundary interaction can be represented by a constant effective twist only when its complete effect is an allowed, energy-independent symmetry rotation. Generic localized kinetic terms reshape the tower in a way that cannot be summarized by one universal value of ω e f f \omega_{\mathrm{eff}} ω eff .
Localized terms are also part of the quantum effective theory rather than optional decorations. Fixed points and boundaries break translation invariance in the compact direction, and ultraviolet divergences can be supported on them. Local boundary renormalization must, however, be distinguished from the finite contribution that requires propagation over a non-zero compact distance. The latter is non-local and has a natural Casimir interpretation: the boundaries determine how fields are reflected, while the compact geometry and Scherk–Schwarz data determine their transport between successive reflections.
Question 8 distinguished the orbifold equivariance data from the dynamical boundary conditions following from the action. Localized operators are where this distinction becomes unavoidable.
Let L = π R L=\pi R L = π R and consider a real scalar field on the interval 0 ≤ y ≤ L 0\leq y\leq L 0 ≤ y ≤ L . For this first part we return to a general compactification from d + 1 d+1 d + 1 to d d d dimensions, with
x M = ( x μ , y ) , M , N = 0 , … , d , μ , ν = 0 , … , d − 1 , x^M=(x^\mu,y),\qquad M,N=0,\ldots,d,\qquad\mu,\nu=0,\ldots,d-1, x M = ( x μ , y ) , M , N = 0 , … , d , μ , ν = 0 , … , d − 1 ,
and use the mostly-plus metric
η M N = diag ( − , + , … , + ) . \eta_{MN}=\operatorname{diag}(-,+,\ldots,+). η M N = diag ( − , + , … , + ) .
Expand
ϕ ( x , y ) = ∑ n φ n ( x ) f n ( y ) , (110) \phi(x,y)=\sum_n\varphi_n(x)f_n(y), \tag*{(110)} ϕ ( x , y ) = n ∑ φ n ( x ) f n ( y ) , (110)
where the lower-dimensional fields are defined by
□ d φ n = m n 2 φ n , □ d ≡ η μ ν ∂ μ ∂ ν . (111) \Box_d\varphi_n=m_n^2\varphi_n,\qquad\Box_d\equiv\eta^{\mu\nu}\partial_\mu\partial_\nu. \tag*{(111)} □ d φ n = m n 2 φ n , □ d ≡ η μν ∂ μ ∂ ν . (111)
Consider the quadratic action
S = − 1 2 ∫ d d x ∫ 0 L d y [ ( ∂ M ϕ ) ( ∂ M ϕ ) + M 2 ϕ 2 ] − 1 2 ∑ b = 0 , L ∫ d d x [ r b ∂ μ ϕ ∂ μ ϕ + λ b ϕ 2 ] y = b . (112) S=-\frac{1}{2}\int d^d x\int_0^Ldy\left[(\partial_M\phi)(\partial^M\phi)+M^2\phi^2\right]
-\frac{1}{2}\sum_{b=0,L}\int d^d x\left[r_b\partial_\mu\phi\,\partial^\mu\phi+\lambda_b\phi^2\right]_{y=b}. \tag*{(112)} S = − 2 1 ∫ d d x ∫ 0 L d y [ ( ∂ M ϕ ) ( ∂ M ϕ ) + M 2 ϕ 2 ] − 2 1 b = 0 , L ∑ ∫ d d x [ r b ∂ μ ϕ ∂ μ ϕ + λ b ϕ 2 ] y = b . (112)
The coefficient λ b \lambda_b λ b is a localized quadratic potential, whereas r b r_b r b multiplies a kinetic term involving derivatives tangent to the boundary. We restrict here to such tangential kinetic terms. Localized operators containing normal derivatives require additional care in the thin-boundary limit and will not be included in the simple spectral problem below.
Away from the endpoints,
( − ∂ y 2 + M 2 ) f n = m n 2 f n . (113) \left(-\partial_y^2+M^2\right)f_n=m_n^2f_n. \tag*{(113)} ( − ∂ y 2 + M 2 ) f n = m n 2 f n . (113)
Let n 0 y = − 1 n_0^y=-1 n 0 y = − 1 and n L y = + 1 n_L^y=+1 n L y = + 1 be the outward normals at the two endpoints. The boundary part of the variation is
δ S ∣ y b = ∫ d d x [ − n b y ∂ y ϕ + r b □ d ϕ − λ b ϕ ] δ ϕ ∣ y b . (114) \delta S\big|_{y_b}=\int d^d x\left[-n_b^y\partial_y\phi+r_b\Box_d\phi-\lambda_b\phi\right]\delta\phi\big|_{y_b}. \tag*{(114)} δ S y b = ∫ d d x [ − n b y ∂ y ϕ + r b □ d ϕ − λ b ϕ ] δ ϕ y b . (114)
Using (111) , stationarity of the action gives
n b y ∂ y f n ( y b ) + ( λ b − r b m n 2 ) f n ( y b ) = 0. (115) n_b^y\partial_yf_n(y_b)+\left(\lambda_b-r_bm_n^2\right)f_n(y_b)=0. \tag*{(115)} n b y ∂ y f n ( y b ) + ( λ b − r b m n 2 ) f n ( y b ) = 0. (115)
Thus
f n ′ ( 0 ) = ( λ 0 − r 0 m n 2 ) f n ( 0 ) (116) f_n'(0)=\left(\lambda_0-r_0m_n^2\right)f_n(0) \tag*{(116)} f n ′ ( 0 ) = ( λ 0 − r 0 m n 2 ) f n ( 0 ) (116)
f n ′ ( L ) = − ( λ L − r L m n 2 ) f n ( L ) . (117) f_n'(L)=-\left(\lambda_L-r_Lm_n^2\right)f_n(L). \tag*{(117)} f n ′ ( L ) = − ( λ L − r L m n 2 ) f n ( L ) . (117)
The localized mass and the localized kinetic term therefore enter the same boundary condition in physically different ways. The former gives an eigenvalue-independent Robin coefficient. The latter multiplies the spectral parameter itself.
A more general localized potential V b ( ϕ ) V_b(\phi) V b ( ϕ ) can also change the background profile. Its boundary equation is
n b y ∂ y ϕ + V b ′ ( ϕ ) = 0. (118) n_b^y \partial_y\phi+ V'_b(\phi)=0. \tag*{(118)} n b y ∂ y ϕ + V b ′ ( ϕ ) = 0. (118)
For fluctuations ϕ = ϕ ˉ + δ ϕ \phi=\bar{\phi}+\delta\phi ϕ = ϕ ˉ + δ ϕ around a background ϕ ˉ \bar{\phi} ϕ ˉ , the coefficient entering the quadratic boundary condition is simply
λ b = V b ′ ′ ( ϕ ˉ ( y b ) ) . \lambda_b=V''_b(\bar{\phi}(y_b)). λ b = V b ′′ ( ϕ ˉ ( y b )) .
Localized potentials can therefore affect both the background and the spectrum of fluctuations around it.
A generalized Sturm–Liouville problem
It is useful first to recall the standard Sturm–Liouville problem,
− d d y ( p ( y ) d f d y ) + q ( y ) f = λ w ( y ) f , (119) -\frac{d}{dy}\left(p(y)\frac{df}{dy}\right)+q(y)f=\lambda w(y)f, \tag*{(119)} − d y d ( p ( y ) d y df ) + q ( y ) f = λ w ( y ) f , (119)
where w ( y ) w(y) w ( y ) is the weight and λ \lambda λ the spectral parameter. The ordinary Kaluza–Klein equation is already of this form, with λ = m n 2 \lambda=m_n^2 λ = m n 2 . Localized masses add singular contributions to the quadratic operator, while localized kinetic terms add singular contributions to the weight. The latter also make the boundary conditions depend explicitly on m n 2 m_n^2 m n 2 .
There is a useful variational formulation that collects these facts into a single generalized eigenvalue problem. Let g ( y ) g(y) g ( y ) be an arbitrary sufficiently regular test profile. More precisely, g g g belongs to the form domain of the problem, for example H 1 ( [ 0 , L ] ) H^1([0,L]) H 1 ([ 0 , L ]) in the scalar example considered here, subject only to kinematic constraints that are independent of the eigenvalue.
Define
Q [ f , g ] = ∫ 0 L d y [ ∂ y f ∗ ∂ y g + M 2 f ∗ g ] + ∑ b = 0 , L λ b f ∗ ( y b ) g ( y b ) , (120) Q[f,g]=\int_0^L dy\left[\partial_y f^*\partial_y g+M^2f^*g\right]+\sum_{b=0,L}\lambda_b f^*(y_b)g(y_b), \tag*{(120)} Q [ f , g ] = ∫ 0 L d y [ ∂ y f ∗ ∂ y g + M 2 f ∗ g ] + b = 0 , L ∑ λ b f ∗ ( y b ) g ( y b ) , (120)
N [ f , g ] = ∫ 0 L d y f ∗ g + ∑ b = 0 , L r b f ∗ ( y b ) g ( y b ) . (121) N[f,g]=\int_0^L dy\,f^*g+\sum_{b=0,L}r_b f^*(y_b)g(y_b). \tag*{(121)} N [ f , g ] = ∫ 0 L d y f ∗ g + b = 0 , L ∑ r b f ∗ ( y b ) g ( y b ) . (121)
For real M 2 M^2 M 2 , λ b \lambda_b λ b , and r b r_b r b , both forms are Hermitian.
Multiplying Eq. (113) by g ∗ g^* g ∗ , integrating by parts, and using the boundary conditions (115) , one obtains
Q [ g , f n ] = m n 2 N [ g , f n ] (122) Q[g,f_n]=m_n^2N[g,f_n] \tag*{(122)} Q [ g , f n ] = m n 2 N [ g , f n ] (122)
for every admissible test profile g g g . This is not an additional assumption: it is simply the bulk equation and the boundary conditions written in weak form.
If N N N is positive definite, setting g = f n g=f_n g = f n gives the Rayleigh quotient
m n 2 = Q [ f n , f n ] N [ f n , f n ] . (123) m_n^2=\frac{Q[f_n,f_n]}{N[f_n,f_n]}. \tag*{(123)} m n 2 = N [ f n , f n ] Q [ f n , f n ] . (123)
Both numerator and denominator are real, so the eigenvalues are real. Hermiticity also implies that two modes with distinct eigenvalues are orthogonal with respect to the kinetic form,
N [ f m , f n ] = 0 , m m 2 ≠ m n 2 . (124) N[f_m,f_n]=0,\qquad m_m^2\ne m_n^2. \tag*{(124)} N [ f m , f n ] = 0 , m m 2 = m n 2 . (124)
This is the important point: localized kinetic terms do not merely perturb the potential of an ordinary Sturm–Liouville problem. They change the kinetic metric with respect to which the problem is Hermitian and the modes are normalized.
For r b ≥ 0 r_b \ge0 r b ≥ 0 , the localized pieces make non-negative contributions to this norm. Negative coefficients require more care. If sufficiently large in magnitude, they can make the kinetic form indefinite and generate zero- or negative-norm states, signalling that the effective description is not physically acceptable without additional structure.
In distributional notation the same problem is often summarized as
[ − ∂ y 2 + M 2 + ∑ b λ b δ b ] f n = m n 2 [ 1 + ∑ b r b δ b ] f n (125) \left[-\partial_y^2+M^2+\sum_b\lambda_b\delta_b\right]f_n=m_n^2\left[1+\sum_b r_b\delta_b\right]f_n \tag*{(125)} [ − ∂ y 2 + M 2 + b ∑ λ b δ b ] f n = m n 2 [ 1 + b ∑ r b δ b ] f n (125)
where the endpoint distributions are normalized so as to reproduce Eqs. (9.11) and (9.12).
Attraction, repulsion, and normalization
The Rayleigh quotient gives a precise version of the intuitive statement that a localized interaction can push a wave function away from a boundary.
For λ 0 > 0 \lambda_0>0 λ 0 > 0 , a profile with a large value of ∣ f ( 0 ) ∣ |f(0)| ∣ f ( 0 ) ∣ pays the positive contribution λ 0 ∣ f ( 0 ) ∣ 2 \lambda_0|f(0)|^2 λ 0 ∣ f ( 0 ) ∣ 2 in the numerator of (123) . Low-lying modes therefore tend to reduce their support at y = 0 y=0 y = 0 . In the strong-coupling limit λ 0 → + ∞ \lambda_0\to+\infty λ 0 → + ∞ , the Robin condition approaches f n ( 0 ) = 0 f_n(0)=0 f n ( 0 ) = 0 and the endpoint becomes effectively Dirichlet.
A negative localized mass acts in the opposite direction. It can enhance a wave function near the boundary and can generate a boundary-localized light state. If the attraction is sufficiently strong, the lowest value of m n 2 m_n^2 m n 2 can become negative, signalling an instability of the assumed background.
A positive localized kinetic term has a different interpretation. Since r b r_b r b appears in N \mathcal{N} N , it changes the normalization of the mode rather than adding a static potential. It also enters the boundary condition through r b m n 2 r_bm_n^2 r b m n 2 , so its influence depends on the eigenvalue.
For a massive mode, a large positive r b m n 2 r_bm_n^2 r b m n 2 can substantially distort the profile and suppress its canonically normalized overlap with the boundary. A zero mode behaves differently. If M = λ 0 = λ L = 0 M=\lambda_0=\lambda_L=0 M = λ 0 = λ L = 0 , the constant profile remains a solution,
f 0 ( y ) = 1 L + r 0 + r L (126) f_0(y)=\frac{1}{\sqrt{L+r_0+r_L}} \tag*{(126)} f 0 ( y ) = L + r 0 + r L 1 (126)
provided the norm is positive. Its shape remains flat: the localized kinetic terms have changed its normalization, and hence its effective couplings, rather than repelling it spatially.
Indeed, if a mode couples to an operator localized at y b y_b y b , its effective coupling is proportional to its normalized boundary value,
g n , b e f f ∝ f n ( y b ) . (127) g_{n,b}^{\mathrm{eff}}\propto f_n(y_b). \tag*{(127)} g n , b eff ∝ f n ( y b ) . (127)
Boundary kinetic terms can consequently affect physical couplings even when the light-mode profile itself looks almost unchanged.
Matching conditions on the covering space
The same localized action can be described on the covering space by delta-function sources. For a second-order scalar equation, integration across a fixed point gives a jump in the first derivative. Denoting by hatted quantities the coefficients in a chosen covering-space delta-function convention,
f n ′ ( 0 + ) − f n ′ ( 0 − ) = ( λ ^ 0 − r ^ 0 m n 2 ) f n ( 0 ) . (128) f_n'(0^+) - f_n'(0^-) = \left(\widehat{\lambda}_0 - \widehat{r}_0 m_n^2\right) f_n(0). \tag*{(128)} f n ′ ( 0 + ) − f n ′ ( 0 − ) = ( λ 0 − r 0 m n 2 ) f n ( 0 ) . (128)
The relation between hatted and interval coefficients depends on the normalization adopted for the orbifold delta distribution.
For an even profile, f n ′ ( 0 − ) = − f n ′ ( 0 + ) f_n'(0^-)=-f_n'(0^+) f n ′ ( 0 − ) = − f n ′ ( 0 + ) , and the jump condition reduces to the interval Robin condition with the corresponding normalization. Thus the interval boundary condition and the covering-space jump condition are two descriptions of the same variational problem.
The distinction becomes even more important for first-order systems. There the field itself can have a finite jump. Localized fermion bilinears are consequently more naturally described by matching matrices, or equivalently by rotated allowed boundary subspaces, than by the usual scalar terminology of Neumann and Dirichlet conditions.
Boundary terms are generated by renormalization
Localized operators should not be regarded as optional decorations of an orbifold effective theory. A fixed point breaks translation invariance in the compact direction. Short-distance singularities can approach the fixed locus and generate ultraviolet divergences supported there.
Bulk interactions may consequently renormalize boundary masses, boundary kinetic terms, gauge kinetic terms, superpotentials, and other localized operators allowed by the symmetries preserved at the fixed point. An effective theory must contain the corresponding counterterm basis. Setting one of these coefficients to zero is generically a renormalization condition at a chosen scale, rather than a prediction of the orbifold geometry. Particular coefficients may of course remain zero when protected by an additional symmetry or when the relevant contributions cancel.
This local renormalization must be distinguished from finite dependence on the complete compactification. Schematically,
Γ 1 − l o o p = Γ b u l k l o c a l + Γ 0 l o c a l + Γ L l o c a l + Γ n o n l o c a l ( L ) . (129) \Gamma_{\mathrm{1-loop}} = \Gamma_{\mathrm{bulk}}^{\mathrm{local}} + \Gamma_{0}^{\mathrm{local}} + \Gamma_{L}^{\mathrm{local}} + \Gamma_{\mathrm{nonlocal}}(L). \tag*{(129)} Γ 1 − loop = Γ bulk local + Γ 0 local + Γ L local + Γ nonlocal ( L ) . (129)
The first three terms contain local bulk and boundary counterterms, together with scheme-dependent local finite pieces. The last term requires propagation over a non-zero distance and can depend on the relative boundary data, the compactification length, and a Scherk–Schwarz holonomy.
We are not yet attempting a complete calculation of the Scherk–Schwarz one-loop potential. The purpose of the next two subsections is instead to make this local/non-local distinction physically transparent and to identify how boundary data and bulk transport enter the non-local part.
The first Casimir interpretation
(129) is the first place where the analogy with the Casimir effect becomes particularly useful.
A loop that explores only an arbitrarily small neighbourhood of one boundary is local. It can renormalize operators already present there, such as a boundary mass or kinetic term. These contributions are ultraviolet sensitive, do not require the second boundary, and do not know its distance.
A qualitatively different contribution is obtained when the field propagates from one boundary to the other and back. Such a process cannot be assigned to either endpoint independently. It knows the distance L L L , the reflection law at both boundaries, and, when the field has internal indices, any transformation accumulated while propagating between them. This is the genuinely Casimir-like piece.
Consider first one real scalar with no internal indices. Wick rotate the non-compact momentum to Euclidean space and define
κ = k E 2 + M 2 . (130) \kappa= \sqrt{k_E^2 + M^2}. \tag*{(130)} κ = k E 2 + M 2 . (130)
At fixed k E k_E k E , propagation through a distance L L L gives a factor e − κ L e^{-\kappa L} e − κ L ; a complete trip from 0 to L L L and back therefore carries e − 2 κ L e^{-2\kappa L} e − 2 κ L .
Let R 0 ( κ ) \mathcal{R}_0(\kappa) R 0 ( κ ) and R L ( κ ) \mathcal{R}_L(\kappa) R L ( κ ) denote the corresponding Euclidean reflection amplitudes. A localized mass changes the Robin coefficient and hence changes R b \mathcal{R}_b R b . A localized kinetic term makes the boundary condition momentum dependent and therefore generally makes R b \mathcal{R}_b R b momentum dependent as well.
One complete round trip consequently carries
R 0 ( κ ) R L ( κ ) e − 2 κ L . (131) \mathcal{R}_0(\kappa)\mathcal{R}_L(\kappa)e^{-2\kappa L}. \tag*{(131)} R 0 ( κ ) R L ( κ ) e − 2 κ L . (131)
Repeated reflections generate paths that cross the interval two, four, six, and progressively more times. The functional determinant sums all of them.
For one real bosonic degree of freedom with Euclidean quadratic operator Δ \Delta Δ ,
Γ 1-loop = 1 2 Tr log Δ . (132) \Gamma_{\text{1-loop}} = \frac{1}{2}\operatorname{Tr}\log\Delta. \tag*{(132)} Γ 1-loop = 2 1 Tr log Δ. (132)
Dividing by the volume V d = ∫ d d x E \mathcal{V}_d = \int d^d x_E V d = ∫ d d x E of the d d d non-compact Euclidean dimensions turns the corresponding part of the trace into
1 V d Tr R d ⟶ ∫ d d k E ( 2 π ) d . \frac{1}{\mathcal{V}_d}\operatorname{Tr}_{\mathbb{R}^d} \longrightarrow\int\frac{d^d k_E}{(2\pi)^d}. V d 1 Tr R d ⟶ ∫ ( 2 π ) d d d k E .
After the pieces that are local in the bulk or at either boundary have been separated and renormalized, the part involving both boundaries has the structural form
V nonlocal ( L ) = 1 2 ∫ d d k E ( 2 π ) d log [ 1 − R 0 ( κ ) R L ( κ ) e − 2 κ L ] . (133) V_{\text{nonlocal}}(L) = \frac{1}{2}\int\frac{d^d k_E}{(2\pi)^d}\log\left[1-\mathcal{R}_0(\kappa)\mathcal{R}_L(\kappa)e^{-2\kappa L}\right]. \tag*{(133)} V nonlocal ( L ) = 2 1 ∫ ( 2 π ) d d d k E log [ 1 − R 0 ( κ ) R L ( κ ) e − 2 κ L ] . (133)
The precise reflection coefficients, and possible extra isolated contributions such as boundary bound states, depend on the spectral problem under consideration. (133) is meant to display the universal round-trip structure.
Its dimensional content is already informative. The argument of the logarithm is dimensionless, so [ V n o n l o c a l ] = ( m a s s ) d [V_{\mathrm{nonlocal}}] = (\mathrm{mass})^d [ V nonlocal ] = ( mass ) d . If L L L is the only physical length, defining q E = L k E q_E = L k_E q E = L k E makes the scaling explicit. For a massless field,
V n o n l o c a l ( L ) = 1 2 L d ∫ d d q E ( 2 π ) d log [ 1 − R 0 ( q E ) R L ( q E ) e − 2 ∣ q E ∣ ] , (134) V_{\mathrm{nonlocal}}(L) = \frac{1}{2L^d}\int\frac{d^d q_E}{(2\pi)^d}\log\left[1-\mathcal{R}_0(q_E)\mathcal{R}_L(q_E)e^{-2|q_E|}\right], \tag*{(134)} V nonlocal ( L ) = 2 L d 1 ∫ ( 2 π ) d d d q E log [ 1 − R 0 ( q E ) R L ( q E ) e − 2∣ q E ∣ ] , (134)
with any further dimensionless dependence contained in the reflection coefficients. Thus the scaling is V n o n l o c a l ( L ) ∝ L − d V_{\mathrm{nonlocal}}(L) \propto L^{-d} V nonlocal ( L ) ∝ L − d .
With further scales present,
V n o n l o c a l ( L ) = 1 L d F ( M L , λ 0 L , λ L L , r 0 L , r L L , … ) , (135) V_{\mathrm{nonlocal}}(L) = \frac{1}{L^d}F\left(ML,\lambda_0L,\lambda_LL,\frac{r_0}{L},\frac{r_L}{L},\ldots\right), \tag*{(135)} V nonlocal ( L ) = L d 1 F ( M L , λ 0 L , λ L L , L r 0 , L r L , … ) , (135)
where F F F is dimensionless.
The logarithm gives an equally useful physical interpretation. With
X ( κ ) = R 0 ( κ ) R L ( κ ) e − 2 κ L , X(\kappa)=\mathcal{R}_0(\kappa)\mathcal{R}_L(\kappa)e^{-2\kappa L}, X ( κ ) = R 0 ( κ ) R L ( κ ) e − 2 κ L ,
one has
log ( 1 − X ) = − ∑ ℓ = 1 ∞ X ℓ ℓ . (136) \log(1-X)=-\sum_{\ell=1}^{\infty}\frac{X^\ell}{\ell}. \tag*{(136)} log ( 1 − X ) = − ℓ = 1 ∑ ∞ ℓ X ℓ . (136)
where
X ℓ = [ R 0 ( κ ) R L ( κ ) ] ℓ e − 2 ℓ κ L . X^\ell=\left[\mathcal{R}_0(\kappa)\mathcal{R}_L(\kappa)\right]^\ell e^{-2\ell\kappa L}. X ℓ = [ R 0 ( κ ) R L ( κ ) ] ℓ e − 2 ℓ κ L .
The integer ℓ \ell ℓ counts complete round trips, while the factor 1 / ℓ 1/\ell 1/ ℓ is the familiar combinatorial weight generated by the logarithm of the determinant.
This expansion also makes the ultraviolet behaviour transparent. Every contribution that probes both boundaries contains at least e − 2 κ L e^{-2\kappa L} e − 2 κ L and is therefore exponentially suppressed when κ → ∞ \kappa\to\infty κ → ∞ . High-momentum fluctuations can renormalize the bulk or an individual boundary locally, but they cannot probe two separated boundaries without paying the finite propagation distance. Once the allowed local counterterms have been included, the genuinely non-local contribution is ultraviolet finite.
Internal indices and Scherk–Schwarz transport
For a multiplet, reflections need not act diagonally on the internal components. The reflection amplitudes become matrices R 0 ( κ ) \mathcal{R}_0(\kappa) R 0 ( κ ) and R L ( κ ) \mathcal{R}_L(\kappa) R L ( κ ) , and propagation between the endpoints may itself rotate the internal state.
Let
U ( L , 0 ) = P exp [ − ∫ 0 L d y A y ] (137) \mathcal{U}(L,0)=\mathcal{P}\exp\left[-\int_0^L dy\,\mathcal{A}_y\right] \tag*{(137)} U ( L , 0 ) = P exp [ − ∫ 0 L d y A y ] (137)
denote parallel transport in the relevant internal space. For a flat Scherk–Schwarz connection, A y \mathcal{A}_y A y has vanishing local curvature but can nevertheless produce non-trivial transport between the endpoints.
Combining ordinary Euclidean propagation with this internal transport gives
T 0 L ( κ ) = e − κ L U ( L , 0 ) , (138) \mathcal{T}_{0L}(\kappa)=e^{-\kappa L}\mathcal{U}(L,0), \tag*{(138)} T 0 L ( κ ) = e − κ L U ( L , 0 ) , (138)
T L 0 ( κ ) = e − κ L U ( 0 , L ) . (139) \mathcal{T}_{L0}(\kappa)=e^{-\kappa L}\mathcal{U}(0,L). \tag*{(139)} T L 0 ( κ ) = e − κ L U ( 0 , L ) . (139)
For ordinary parallel transport by a non-singular connection, U ( 0 , L ) = U ( L , 0 ) − 1 \mathcal{U}(0,L)=\mathcal{U}(L,0)^{-1} U ( 0 , L ) = U ( L , 0 ) − 1 .
Choose as reference state a wave immediately after reflection at y = 0 y=0 y = 0 , propagating toward L L L . One complete round trip is then
0 + → T 0 L L − → R L L − → T L 0 0 + → R 0 0 + . 0^{+} \xrightarrow{\mathcal{T}_{0L}} L^{-} \xrightarrow{\mathcal{R}_{L}} L^{-} \xrightarrow{\mathcal{T}_{L0}} 0^{+} \xrightarrow{\mathcal{R}_{0}} 0^{+}. 0 + T 0 L L − R L L − T L 0 0 + R 0 0 + .
The last reflection is essential: arrival back at y = 0 y=0 y = 0 has reversed the propagation direction, so the state has not yet returned to the chosen reference state.
The corresponding round-trip operator is
T r t ( κ ) = R 0 ( κ ) T L 0 ( κ ) R L ( κ ) T 0 L ( κ ) . (140) \mathcal{T}_{\mathrm{rt}}(\kappa)=\mathcal{R}_{0}(\kappa)\mathcal{T}_{L0}(\kappa)\mathcal{R}_{L}(\kappa)\mathcal{T}_{0L}(\kappa). \tag*{(140)} T rt ( κ ) = R 0 ( κ ) T L 0 ( κ ) R L ( κ ) T 0 L ( κ ) . (140)
The order of these matrices matters. In particular, U ( 0 , L ) = U ( L , 0 ) − 1 \mathcal{U}(0,L)=\mathcal{U}(L,0)^{-1} U ( 0 , L ) = U ( L , 0 ) − 1 does not make the Scherk–Schwarz information disappear because the two transport operators are separated by reflections that need not commute with them.
The scalar expression (133) is accordingly replaced by the structural matrix formula
V n o n l o c a l ( L ) = η Φ 2 ∫ d d k E ( 2 π ) d log det [ 1 − T r t ( κ ) ] . (141) V_{\mathrm{nonlocal}}(L)=\frac{\eta_{\Phi}}{2}\int\frac{d^{d}k_{E}}{(2\pi)^{d}}\log\det\left[1-\mathcal{T}_{\mathrm{rt}}(\kappa)\right]. \tag*{(141)} V nonlocal ( L ) = 2 η Φ ∫ ( 2 π ) d d d k E log det [ 1 − T rt ( κ ) ] . (141)
Here the determinant acts on the internal space. The overall sign, multiplicity, and precise power of the determinant depend on the statistics, reality properties, spin and gauge fixing of the field; η Φ \eta_{\Phi} η Φ is only a compact reminder of this bookkeeping.
The physical organization is nevertheless universal:
localized boundary dynamics ⟶ \longrightarrow ⟶ reflection matrices,
bulk propagation and Scherk–Schwarz data ⟶ \longrightarrow ⟶ transport matrices,
complete round trips ⟶ \longrightarrow ⟶ non-local one-loop contribution.
A localized mass modifies the reflection law. A localized kinetic term makes it momentum dependent. A Scherk–Schwarz deformation enters the ordered transport between the endpoints. The non-local determinant depends on all of these data together.
Where the Casimir analogy stops
The analogy is useful precisely because it separates local from non-local physics; it should not be pushed beyond that point.
An orbifold fixed point is not a material plate. Localized masses, kinetic terms, superpotentials and other couplings are parameters of the effective field theory. Their coefficients are renormalized by local ultraviolet physics and are not predicted by a Casimir calculation.
The Casimir-like quantity is instead the piece that cannot be assigned to either endpoint separately because it requires propagation across a non-zero distance and back. Local boundary terms enter that calculation as input through their reflection matrices.
This distinction remains true when boundaries are idealized as infinitely thin defects. The idealization can itself produce surface divergences, but those divergences are localized and are absorbed by local counterterms. They do not contradict the ultraviolet finiteness of the part that genuinely probes the finite separation L L L .
The same logic will reappear later after Poisson resummation of Kaluza–Klein sums. The zero-winding sector can contain contributions local in the higher-dimensional theory, whereas non-zero winding sectors necessarily probe the compact geometry globally. The scattering and winding descriptions are two ways of recognizing the same separation.
Localized terms for a fermionic doublet
We now turn from second-order fields to first-order fermionic systems. Consider a two-component fermion
Ψ = ( ψ 1 ψ 2 ) \Psi= \begin{pmatrix} \psi_1 \\ \psi_2 \end{pmatrix} Ψ = ( ψ 1 ψ 2 )
with components of opposite orbifold parity.
Without localized interactions, the orbifold projection selects a one-dimensional subspace of this doublet at each endpoint. With localized interactions, it is useful to formulate the resulting conditions in terms of one-sided limits from within the physical interval:
ψ 2 ( 0 + ) = β 0 ( m n ) ψ 1 ( 0 + ) , (142) \psi_2(0^+) = \beta_0(m_n)\,\psi_1(0^+), \tag*{(142)} ψ 2 ( 0 + ) = β 0 ( m n ) ψ 1 ( 0 + ) , (142)
ψ 2 ( L − ) = − β L ( m n ) ψ 1 ( L − ) . (143) \psi_2(L^-) = -\beta_L(m_n)\,\psi_1(L^-). \tag*{(143)} ψ 2 ( L − ) = − β L ( m n ) ψ 1 ( L − ) . (143)
The relative sign at the second endpoint depends on the orientation and parity conventions. The values on the opposite sides of the fixed points are related by orbifold equivariance and are not independent boundary data.
For a pure localized mass, the matching coefficient is independent of the KK eigenvalue, β b ( m n ) = M b \beta_b(m_n) = M_b β b ( m n ) = M b . The first endpoint then selects the line proportional to ( 1 , M 0 ) T (1,M_0)^T ( 1 , M 0 ) T , while the second selects, with the convention above, the line proportional to ( 1 , − M L ) T (1,-M_L)^T ( 1 , − M L ) T .
Introducing the angle Θ b = arctan M b \Theta_b = \arctan M_b Θ b = arctan M b makes the geometrical meaning transparent: a pure localized mass rotates the one-dimensional fermionic subspace allowed at the boundary by a fixed angle.
A localized kinetic term changes this conclusion. Since it contains derivatives tangent to the boundary, its contribution to the matching condition depends on the lower-dimensional momentum and hence, on shell, on the eigenvalue. Schematically,
β b ( m n ) = M b + r b ( ψ ) m n + ⋯ , (144) \beta_b(m_n) = M_b + r_b^{(\psi)}m_n + \cdots, \tag*{(144)} β b ( m n ) = M b + r b ( ψ ) m n + ⋯ , (144)
where the precise dependence depends on the fermion system and the localized operator.
Different KK levels can see different allowed boundary directions. This simple observation will be crucial below: a fixed boundary rotation may be equivalent to Scherk–Schwarz data, whereas an eigenvalue-dependent one generally is not.
Localized gravitino masses and the supersymmetry parameter
We now specialize again to minimal five-dimensional supergravity compactified on S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 . The four-dimensional gravitini form an S U ( 2 ) R SU(2)_R S U ( 2 ) R doublet ψ μ I \psi_\mu^I ψ μ I , I = 1 , 2 I = 1, 2 I = 1 , 2 , and the supersymmetry parameter carries the corresponding doublet index.
At y = 0 y = 0 y = 0 , take ψ μ 1 \psi_{\mu1} ψ μ 1 to be even and ψ μ 2 \psi_{\mu2} ψ μ 2 odd. At y = L = π R y = L = \pi R y = L = π R , introduce
ψ μ + = cos ( π ω ) ψ μ 1 − sin ( π ω ) ψ μ 2 , ψ μ − = sin ( π ω ) ψ μ 1 + cos ( π ω ) ψ μ 2 , (145) \begin{aligned}
\psi_{\mu+} &= \cos(\pi\omega)\psi_{\mu1} - \sin(\pi\omega)\psi_{\mu2},\\
\psi_{\mu-} &= \sin(\pi\omega)\psi_{\mu1} + \cos(\pi\omega)\psi_{\mu2}, \tag*{(145)}
\end{aligned} ψ μ + ψ μ − = cos ( π ω ) ψ μ 1 − sin ( π ω ) ψ μ 2 , = sin ( π ω ) ψ μ 1 + cos ( π ω ) ψ μ 2 , (145)
and analogously
ϵ + = cos ( π ω ) ϵ 1 − sin ( π ω ) ϵ 2 , ϵ − = sin ( π ω ) ϵ 1 + cos ( π ω ) ϵ 2 . (146) \begin{aligned}
\epsilon_+ &= \cos(\pi\omega)\epsilon_1 - \sin(\pi\omega)\epsilon_2,\\
\epsilon_- &= \sin(\pi\omega)\epsilon_1 + \cos(\pi\omega)\epsilon_2. \tag*{(146)}
\end{aligned} ϵ + ϵ − = cos ( π ω ) ϵ 1 − sin ( π ω ) ϵ 2 , = sin ( π ω ) ϵ 1 + cos ( π ω ) ϵ 2 . (146)
Without localized interactions, ψ μ 1 \psi_{\mu1} ψ μ 1 and ψ μ + \psi_{\mu+} ψ μ + are the even gravitino components at the two fixed points. A localized gravitino bilinear changes the matching of the odd component. In the normalization used here,
ψ μ 2 ( 0 + ) = M 0 ψ μ 1 ( 0 + ) , ψ μ − ( L − ) = − M π ψ μ + ( L − ) . (147) \begin{aligned}
\psi_{\mu2}(0^+) &= M_0\psi_{\mu1}(0^+),\\
\psi_{\mu-}(L^-) &= -M_\pi\psi_{\mu+}(L^-). \tag*{(147)}
\end{aligned} ψ μ 2 ( 0 + ) ψ μ − ( L − ) = M 0 ψ μ 1 ( 0 + ) , = − M π ψ μ + ( L − ) . (147)
The corresponding restrictions on the supersymmetry parameter are not independent assumptions. Local supersymmetry relates ψ μ I \psi_{\mu I} ψ μ I to ϵ I \epsilon_I ϵ I , schematically through
δ ψ μ I = 2 D μ ϵ I + ⋯ . (148) \delta\psi_{\mu I} = 2D_\mu\epsilon_I + \cdots. \tag*{(148)} δ ψ μ I = 2 D μ ϵ I + ⋯ . (148)
For a supersymmetric boundary coupling, the space of allowed gravitino boundary values must be stable under the allowed supersymmetry transformations. The same matching therefore appears for the parameter,
ϵ 2 ( 0 + ) = M 0 ϵ 1 ( 0 + ) , ϵ − ( L − ) = − M π ϵ + ( L − ) . (149) \begin{aligned}
\epsilon_2(0^+) &= M_0\epsilon_1(0^+),\\
\epsilon_-(L^-) &= -M_\pi\epsilon_+(L^-). \tag*{(149)}
\end{aligned} ϵ 2 ( 0 + ) ϵ − ( L − ) = M 0 ϵ 1 ( 0 + ) , = − M π ϵ + ( L − ) . (149)
These conditions also ensure the cancellation of the distributional terms generated when derivatives act on discontinuous odd supersymmetry parameters.
The two localized masses define the angles Θ 0 = arctan M 0 \Theta_0 = \arctan M_0 Θ 0 = arctan M 0 and Θ π = arctan M π \Theta_\pi= \arctan M_\pi Θ π = arctan M π . Thus the spin-3 2 \frac{3}{2} 2 3 boundary problem is geometrically a problem of comparing two one-dimensional directions in the S U ( 2 ) R SU(2)_R S U ( 2 ) R doublet space.
These directions describe only the restrictions coming from the gravitino sector. If localized matter fermions are present, their supersymmetry variations can impose additional conditions.
What the lower-dimensional theory sees
The physical four-dimensional gravitini are obtained by solving the complete compact-direction eigenvalue problem. Write
ψ μ ( x , y ) = ∑ n f n I ( y ) ψ μ , n ( x ) . (150) \psi_\mu(x,y)=\sum_n f_{nI}(y)\,\psi_{\mu,n}(x). \tag*{(150)} ψ μ ( x , y ) = n ∑ f n I ( y ) ψ μ , n ( x ) . (150)
The two-component object f n I ( y ) f_{nI}(y) f n I ( y ) is the internal profile, while ψ μ , n ( x ) \psi_{\mu,n}(x) ψ μ , n ( x ) is the corresponding four-dimensional gravitino field.
A massless gravitino exists only when the global spectral problem admits a non-zero solution with m n = 0 m_n=0 m n = 0 satisfying the bulk equations and both endpoint matching conditions.
If localized kinetic terms are present, its normalization is also defined by the kinetic metric obtained from the reduced action, schematically
∫ 0 L d y f m † ( y ) W ( y ) f n ( y ) = δ m n , (151) \int_0^L dy\,f_m^\dagger(y)\,W(y)\,f_n(y)=\delta_{mn}, \tag*{(151)} ∫ 0 L d y f m † ( y ) W ( y ) f n ( y ) = δ mn , (151)
where W W W includes any localized kinetic contributions.
A profile whose direction in S U ( 2 ) R SU(2)_R S U ( 2 ) R space changes along the interval can therefore represent one perfectly ordinary four-dimensional gravitino. There is no requirement that its separate higher-dimensional components be constant or have independent four-dimensional interpretations.
For the flat locally supersymmetric system considered below, the massless-gravitino equation and the gravitino Killing-spinor equation are closely related. The existence of a zero-mode gravitino is an important test of unbroken supersymmetry, although, as we shall see, it is not the complete test when localized matter fermions also have non-zero shifts.
The zero-mode condition in twisted variables
First use the description in which the Scherk–Schwarz datum resides in the twisted identification and in the relative endpoint actions. There is then no explicit flat connection in the open interval.
For a candidate massless gravitino, the bulk profile is constant,
f 0 ( y ) = ( v 1 v 2 ) . f_0(y)=\begin{pmatrix}v_1\\v_2\end{pmatrix}. f 0 ( y ) = ( v 1 v 2 ) .
The first boundary condition gives
v 2 = M 0 v 1 . (152) v_2=M_0v_1. \tag*{(152)} v 2 = M 0 v 1 . (152)
At the second endpoint, define
v + = cos ( π ω ) v 1 − sin ( π ω ) v 2 , v − = sin ( π ω ) v 1 + cos ( π ω ) v 2 . v_+=\cos(\pi\omega)v_1-\sin(\pi\omega)v_2,\qquad v_-=\sin(\pi\omega)v_1+\cos(\pi\omega)v_2. v + = cos ( π ω ) v 1 − sin ( π ω ) v 2 , v − = sin ( π ω ) v 1 + cos ( π ω ) v 2 .
The matching condition requires
v − = − M π v + . (153) v_-=-M_\pi v_+. \tag*{(153)} v − = − M π v + . (153)
Substituting Eq. (152) gives
[ ( 1 − M 0 M π ) sin ( π ω ) + ( M 0 + M π ) cos ( π ω ) ] v 1 = 0. (154) \left[(1-M_0M_\pi)\sin(\pi\omega)+(M_0+M_\pi)\cos(\pi\omega)\right]v_1=0. \tag*{(154)} [ ( 1 − M 0 M π ) sin ( π ω ) + ( M 0 + M π ) cos ( π ω ) ] v 1 = 0. (154)
For a non-zero profile this is equivalently
sin ( π ω + Θ 0 + Θ π ) = 0 , \sin(\pi\omega+\Theta_0+\Theta_\pi)=0, sin ( π ω + Θ 0 + Θ π ) = 0 ,
or
π ω + Θ 0 + Θ π = n π , n ∈ Z . (155) \pi\omega+\Theta_0+\Theta_\pi=n\pi,\qquad n\in\mathbb{Z}. \tag*{(155)} π ω + Θ 0 + Θ π = nπ , n ∈ Z . (155)
The half-twist gives a particularly transparent example. For ω = 1 / 2 \omega=1/2 ω = 1/2 and M 0 = M π = 0 M_0=M_\pi=0 M 0 = M π = 0 , the first endpoint selects span ( 1 , 0 ) T \operatorname{span}(1,0)^{T} span ( 1 , 0 ) T , whereas the second selects the orthogonal line span ( 0 , 1 ) T \operatorname{span}(0,1)^{T} span ( 0 , 1 ) T . No non-zero constant profile belongs to both, so no massless four-dimensional gravitino exists.
Localized masses can compensate the misalignment. At the half-twist, a zero mode is restored when
Θ 0 + Θ π = − π 2 ( m o d π ) . \Theta_0+\Theta_\pi=-\frac{\pi}{2}\pmod{\pi}. Θ 0 + Θ π = − 2 π ( mod π ) .
The same compactification in periodic variables
The same physics can be described with periodic fields and a flat bulk connection. Let
R [ α ] = ( cos α sin α − sin α cos α ) R[\alpha]=
\begin{pmatrix}
\cos\alpha& \sin\alpha\\
-\sin\alpha& \cos\alpha
\end{pmatrix} R [ α ] = ( cos α − sin α sin α cos α )
and write
\begin{pmatrix}
\psi_{\mu1}\\
\psi_{\mu2}
\end{pmatrix}
=R[\alpha(y)]
\begin{pmatrix}
\widetilde{\psi}_{\mu1}\\
\widetilde{\psi}_{\mu2}
\end{pmatrix},
\qquad
\alpha(y+2\pi R)=\alpha(y)+2\pi\omega. \tag*{(156)}\end{pmatrix}
The periodic fields acquire the connection
∂ y ψ μ = R [ α ( y ) ] [ ∂ y + α ′ ( y ) J ] ψ ~ μ , J = ( 0 1 − 1 0 ) . (157) \partial_y\psi_\mu=R[\alpha(y)]\left[\partial_y+\alpha'(y)J\right]\widetilde{\psi}_\mu,
\qquad
J=
\begin{pmatrix}
0 & 1\\
-1 & 0
\end{pmatrix}. \tag*{(157)} ∂ y ψ μ = R [ α ( y )] [ ∂ y + α ′ ( y ) J ] ψ μ , J = ( 0 − 1 1 0 ) . (157)
The supersymmetry parameter, endpoint projectors and localized action must all be transformed by the same change of variables.
The pure half-twist makes the equivalence especially easy to see. In the twisted-field description,
A y = 0 , L 0 = span ( Q 1 ) , L π = span ( Q 2 ) , \mathcal{A}_y=0,\qquad
\mathcal{L}_0=\operatorname{span}(Q_1),\qquad
\mathcal{L}_\pi=\operatorname{span}(Q_2), A y = 0 , L 0 = span ( Q 1 ) , L π = span ( Q 2 ) ,
so a constant zero-mode profile cannot satisfy both endpoint conditions.
In periodic variables the endpoint lines may instead be aligned,
L ~ 0 = L ~ π = span ( Q 1 ) , \widetilde{\mathcal{L}}_0=\widetilde{\mathcal{L}}_\pi=\operatorname{span}(Q_1), L 0 = L π = span ( Q 1 ) ,
but the flat connection now transports the initial direction according to U ( L , 0 ) Q 1 ∝ Q 2 U(L,0)Q_1\propto Q_2 U ( L , 0 ) Q 1 ∝ Q 2 . The transported zero-mode candidate again fails the second endpoint condition.
The two descriptions therefore agree. A spurious zero mode appears only if one combines the endpoint line from the twisted description with the non-trivial transport from the periodic description. That would insert the same Scherk–Schwarz datum twice. Endpoint data and bulk transport must always belong to one common trivialization.
The Killing-spinor equations give the same condition
The equivalence can also be seen directly from the Killing-spinor equations. In periodic variables let α ( y ) \alpha(y) α ( y ) denote the rotation angle introduced above. If it contains localized jumps, one may understand it as the thin-boundary limit of a smooth regularization.
For the flat background and the conventions used here, the relevant equations take the form
∂ y ϵ ~ 1 + α ′ ( y ) ϵ ~ 2 = 0 , \partial_y\widetilde{\epsilon}_1+\alpha'(y)\widetilde{\epsilon}_2=0, ∂ y ϵ 1 + α ′ ( y ) ϵ 2 = 0 ,
∂ y ϵ ~ 2 − α ′ ( y ) ϵ ~ 1 = 2 M 0 ϵ ~ 1 δ ( y ) + 2 M π ϵ ~ 1 δ ( y − L ) . (158) \partial_y\widetilde{\epsilon}_2-\alpha'(y)\widetilde{\epsilon}_1=2M_0\widetilde{\epsilon}_1\delta(y)+2M_\pi\widetilde{\epsilon}_1\delta(y-L). \tag*{(158)} ∂ y ϵ 2 − α ′ ( y ) ϵ 1 = 2 M 0 ϵ 1 δ ( y ) + 2 M π ϵ 1 δ ( y − L ) . (158)
Integration across the two fixed points reproduces
ϵ ~ 2 ( 0 + ) = M 0 ϵ ~ 1 ( 0 ) , ϵ ~ 2 ( L − ) = − M π ϵ ~ 1 ( L ) . (159) \widetilde{\epsilon}_2(0^+)=M_0\widetilde{\epsilon}_1(0),\qquad\widetilde{\epsilon}_2(L^-)=-M_\pi\widetilde{\epsilon}_1(L). \tag*{(159)} ϵ 2 ( 0 + ) = M 0 ϵ 1 ( 0 ) , ϵ 2 ( L − ) = − M π ϵ 1 ( L ) . (159)
Solving between the boundaries then gives once again
π ω + Θ 0 + Θ π = n π . (160) \pi\omega+\Theta_0+\Theta_\pi=n\pi. \tag*{(160)} π ω + Θ 0 + Θ π = nπ . (160)
This is an important consistency check, but at this stage it concerns only the gravitino/Killing-spinor sector. Variations of additional localized fermions have not yet been imposed.
The gravitino spectrum gives the same answer
The same condition follows directly from the spin-3 2 \frac{3}{2} 2 3 spectrum. Solving the first-order bulk equations for a signed four-dimensional mass m m m and imposing the endpoint matching conditions gives
m n s i g n e d = 1 R [ n + ω + Θ 0 + Θ π π ] , n ∈ Z . (161) m_n^{\mathrm{signed}}=\frac{1}{R}\left[n+\omega+\frac{\Theta_0+\Theta_\pi}{\pi}\right],\qquad n\in\mathbb{Z}. \tag*{(161)} m n signed = R 1 [ n + ω + π Θ 0 + Θ π ] , n ∈ Z . (161)
The physical masses are simply m n p h y s = ∣ m n s i g n e d ∣ m_n^{\mathrm{phys}}=\lvert m_n^{\mathrm{signed}}\rvert m n phys = ∣ m n signed ∣ .
A zero appears precisely when
n + ω + Θ 0 + Θ π π = 0 n+\omega+\frac{\Theta_0+\Theta_\pi}{\pi}=0 n + ω + π Θ 0 + Θ π = 0
for some n ∈ Z n\in\mathbb{Z} n ∈ Z , which is equivalent to Eq. (160) . For this restricted class of energy-independent boundary rotations, one may therefore define
ω e f f = ω + Θ 0 + Θ π π ( m o d 1 ) . (162) \omega_{\mathrm{eff}}=\omega+\frac{\Theta_0+\Theta_\pi}{\pi}\pmod{1}. \tag*{(162)} ω eff = ω + π Θ 0 + Θ π ( mod 1 ) . (162)
The spin-3 2 \frac{3}{2} 2 3 tower remains equally spaced and depends only on ω e f f \omega_{\mathrm{eff}} ω eff . The agreement between the spectral zero and the Killing-spinor condition is a consequence of the locally supersymmetric flat bulk-plus-boundary gravitino system. It should not yet be promoted to a statement about every fermion in a more general boundary sector.
The invariant geometric statement
The calculation can now be summarized without referring to a particular field basis.
In one fixed trivialization, let B 0 ( 3 / 2 ) \mathcal{B}_0^{(3/2)} B 0 ( 3/2 ) and B π ( 3 / 2 ) \mathcal{B}_{\pi}^{(3/2)} B π ( 3/2 ) denote the one-dimensional subspaces of allowed gravitino-doublet boundary values at y = 0 y=0 y = 0 and y = L y=L y = L . Let U ( L , 0 ) \mathcal{U}(L,0) U ( L , 0 ) denote the transport operator obtained from the bulk zero-mode equation, or equivalently the parallel transport generated by the connection A y \mathcal{A}_y A y in a trivialization where that connection is explicit.
A gravitino zero mode exists precisely when
U ( L , 0 ) B 0 ( 3 / 2 ) ∩ B π ( 3 / 2 ) ≠ { 0 } . (163) \mathcal{U}(L,0)\mathcal{B}_0^{(3/2)} \cap\mathcal{B}_{\pi}^{(3/2)} \ne\{0\}. \tag*{(163)} U ( L , 0 ) B 0 ( 3/2 ) ∩ B π ( 3/2 ) = { 0 } . (163)
This merely states that an allowed boundary value at one end can be transported by the bulk equation into an allowed boundary value at the other.
Under a change of basis ψ ′ ( y ) = G ( y ) − 1 ψ ( y ) \psi'(y)=G(y)^{-1}\psi(y) ψ ′ ( y ) = G ( y ) − 1 ψ ( y ) , the three ingredients transform together,
B 0 ′ ( 3 / 2 ) = G ( 0 ) − 1 B 0 ( 3 / 2 ) , B π ′ ( 3 / 2 ) = G ( L ) − 1 B π ( 3 / 2 ) , U ′ ( L , 0 ) = G ( L ) − 1 U ( L , 0 ) G ( 0 ) . (164) \begin{aligned}
\mathcal{B}_0'^{(3/2)} &= G(0)^{-1}\mathcal{B}_0^{(3/2)},\\
\mathcal{B}_{\pi}'^{(3/2)} &= G(L)^{-1}\mathcal{B}_{\pi}^{(3/2)},\\
\mathcal{U}'(L,0) &= G(L)^{-1}\mathcal{U}(L,0)G(0). \tag*{(164)}
\end{aligned} B 0 ′ ( 3/2 ) B π ′ ( 3/2 ) U ′ ( L , 0 ) = G ( 0 ) − 1 B 0 ( 3/2 ) , = G ( L ) − 1 B π ( 3/2 ) , = G ( L ) − 1 U ( L , 0 ) G ( 0 ) . (164)
The dimension of the intersection in Eq. (163) is therefore invariant.
This is the cleanest way of stating the double-counting warning: changing trivialization moves information between transport and endpoint data, but cannot create or remove a zero mode.
Localized kinetic terms do not define one universal twist
The simple quantity ω e f f \omega_{\mathrm{eff}} ω eff exists because the two boundary rotations above are independent of the KK eigenvalue.
With a localized fermion kinetic term, the matching coefficient becomes a function β b ( m ) \beta_b(m) β b ( m ) and the associated angle Θ b ( m ) = arctan β b ( m ) \Theta_b(m)=\arctan\beta_b(m) Θ b ( m ) = arctan β b ( m ) depends on the state being quantized. The spectral condition then has the implicit form
m n R = n + ω + Θ 0 ( m n ) + Θ π ( m n ) π , (165) m_nR=n+\omega+\frac{\Theta_0(m_n)+\Theta_{\pi}(m_n)}{\pi}, \tag*{(165)} m n R = n + ω + π Θ 0 ( m n ) + Θ π ( m n ) , (165)
up to the conventions used for the signed tower.
The eigenvalue now appears on both sides. Different KK states generically see different boundary directions, their spacing need not remain uniform, and no single number ω e f f \omega_{\mathrm{eff}} ω eff can encode the whole spectrum.
A field redefinition can redistribute an energy-independent symmetry rotation between the bulk and the endpoints. It cannot eliminate the eigenvalue-dependent dynamics of a genuine localized kinetic term.
What does the gravitino zero-mode condition test?
The condition derived above refers to the gravitino boundary-value problem defined by the bulk equations and by the localized gravitino bilinears included in that problem. For the flat locally supersymmetric system considered here, a zero of the spin-3 2 \frac{3}{2} 2 3 spectrum coincides with a solution of the corresponding gravitino Killing-spinor equations.
This is not yet a statement about a more general localized sector. If additional boundary fermions are present, their supersymmetry variations must be imposed as well. In particular, a localized fermion may have an inhomogeneous transformation of the form
δ χ b i = F b i ϵ b + ⋯ . (166) \delta\chi_b^i = F_b^i \epsilon_b + \cdots. \tag*{(166)} δ χ b i = F b i ϵ b + ⋯ . (166)
A supersymmetry parameter preserved by the complete configuration must satisfy these conditions together with the bulk and gravitino Killing-spinor equations.
Thus Eq. (155) is complete for the gravitino system from which it was derived. Adding an independent localized breaking sector enlarges the fermionic problem.
When is a localized gravitino mass just a Scherk–Schwarz rotation?
The spectrum Eq. (161) shows that an energy-independent gravitino boundary mass enters through the angles Θ 0 \Theta_0 Θ 0 and Θ π \Theta_\pi Θ π in exactly the same way as a shift of the Scherk–Schwarz parameter. This is not accidental.
A y y y -dependent S U ( 2 ) R SU(2)_R S U ( 2 ) R change of variables can redistribute a fixed total monodromy between a smooth bulk connection and localized rotations at the two endpoints. If the corresponding endpoint angles are Ω 0 \Omega_0 Ω 0 and Ω π \Omega_\pi Ω π , one may write schematically
π ω = π ω B + Ω 0 + Ω π . (167) \pi\omega= \pi\omega_B + \Omega_0 + \Omega_\pi. \tag*{(167)} π ω = π ω B + Ω 0 + Ω π . (167)
The bulk and boundary descriptions then differ only by the trivialization used to represent the same global Scherk–Schwarz datum.
This equivalence requires more than agreement of the gravitino mass spectrum: the boundary operation must represent the same allowed symmetry transformation on the complete system.
A genuine localized supersymmetry-breaking sector can instead contain additional fermions with independent shifts and hence additional would-be Goldstino directions. This information is not encoded in the shifted spin-3 2 \frac{3}{2} 2 3 spectrum.
Discontinuous representatives and one-sided limits
The gravitino matching condition derived above provides a simple example of how an orbifold field can become discontinuous at a fixed point. Consider y = 0 y=0 y = 0 , with ψ μ 1 \psi_{\mu1} ψ μ 1 even and ψ μ 2 \psi_{\mu2} ψ μ 2 odd under y → − y y\to-y y → − y .
For a four-dimensional gravitino eigenstate of signed mass m m m , the bulk equations are first order in the compact coordinate. In the presence of a localized gravitino mass, one of them has, in the conventions used here, the schematic form
∂ y ψ μ 2 + m ψ μ 1 = 2 M 0 δ ( y ) ψ μ 1 , (168) \partial_y\psi_{\mu2}+m\psi_{\mu1}=2M_0\delta(y)\psi_{\mu1}, \tag*{(168)} ∂ y ψ μ 2 + m ψ μ 1 = 2 M 0 δ ( y ) ψ μ 1 , (168)
while the companion equation contains no delta-function source at y = 0 y = 0 y = 0 . The precise normalization of M 0 M_0 M 0 depends on the convention for the orbifold delta function; the argument below is independent of this choice once the convention is fixed.
Integrating Eq. (168) across an infinitesimal neighbourhood of the fixed point gives
ψ μ 2 ( 0 + ) − ψ μ 2 ( 0 − ) + m ∫ − ε + ε d y ψ μ 1 ( y ) = 2 M 0 ψ μ 1 ( 0 ) . (169) \psi_{\mu2}(0^+) - \psi_{\mu2}(0^-) + m \int_{-\varepsilon}^{+\varepsilon} \mathrm{d}y \, \psi_{\mu1}(y)
= 2M_0 \psi_{\mu1}(0). \tag*{(169)} ψ μ 2 ( 0 + ) − ψ μ 2 ( 0 − ) + m ∫ − ε + ε d y ψ μ 1 ( y ) = 2 M 0 ψ μ 1 ( 0 ) . (169)
The second term vanishes as ε → 0 \varepsilon\to0 ε → 0 because it is the integral of a regular bulk contribution. The localized source therefore fixes the finite jump,
ψ μ 2 ( 0 + ) − ψ μ 2 ( 0 − ) = 2 M 0 ψ μ 1 ( 0 ) . (170) \psi_{\mu2}(0^+) - \psi_{\mu2}(0^-) = 2M_0 \psi_{\mu1}(0). \tag*{(170)} ψ μ 2 ( 0 + ) − ψ μ 2 ( 0 − ) = 2 M 0 ψ μ 1 ( 0 ) . (170)
Orbifold equivariance relates the two one-sided limits. Since ψ μ 1 \psi_{\mu1} ψ μ 1 is even and ψ μ 2 \psi_{\mu2} ψ μ 2 odd,
ψ μ 1 ( 0 − ) = ψ μ 1 ( 0 + ) , ψ μ 2 ( 0 − ) = − ψ μ 2 ( 0 + ) . (171) \psi_{\mu1}(0^-) = \psi_{\mu1}(0^+), \qquad\psi_{\mu2}(0^-) = -\psi_{\mu2}(0^+). \tag*{(171)} ψ μ 1 ( 0 − ) = ψ μ 1 ( 0 + ) , ψ μ 2 ( 0 − ) = − ψ μ 2 ( 0 + ) . (171)
The even component is continuous, so we may identify ψ μ 1 ( 0 ) \psi_{\mu1}(0) ψ μ 1 ( 0 ) with either one-sided limit. Substituting the parity relations into Eq. (170) then gives
ψ μ 2 ( 0 + ) = M 0 ψ μ 1 ( 0 + ) , (172) \psi_{\mu2}(0^+) = M_0 \psi_{\mu1}(0^+), \tag*{(172)} ψ μ 2 ( 0 + ) = M 0 ψ μ 1 ( 0 + ) , (172)
which is precisely the matching condition used above.
This calculation also explains an apparent paradox of orbifold language. An odd field is often said to vanish at a fixed point. That statement assumes continuity. Here odd parity instead requires
ψ μ 2 ( 0 − ) = − ψ μ 2 ( 0 + ) , \psi_{\mu2}(0^-) = -\psi_{\mu2}(0^+), ψ μ 2 ( 0 − ) = − ψ μ 2 ( 0 + ) ,
and the two limits need not vanish separately. The localized first-order interaction has replaced continuity by the matching condition (9.63).
The discontinuity itself is representation dependent. To see this, collect the two components into
Ψ μ = ( ψ μ 1 ψ μ 2 ) \Psi_\mu=
\begin{pmatrix}
\psi_{\mu1}\\
\psi_{\mu2}
\end{pmatrix} Ψ μ = ( ψ μ 1 ψ μ 2 )
and perform the piecewise rotation
Ψ μ ( y ) = R [ − Θ 0 ε ( y ) ] Ψ ~ μ ( y ) , tan Θ 0 = M 0 , (173) \Psi_\mu(y) = R[-\Theta_0 \varepsilon(y)] \widetilde{\Psi}_\mu(y), \qquad\tan\Theta_0 = M_0, \tag*{(173)} Ψ μ ( y ) = R [ − Θ 0 ε ( y )] Ψ μ ( y ) , tan Θ 0 = M 0 , (173)
where
R [ α ] = ( cos α sin α − sin α cos α ) , ε ( y ) = { + 1 , y > 0 , − 1 , y < 0. R[\alpha] =
\begin{pmatrix}
\cos\alpha& \sin\alpha\\
-\sin\alpha& \cos\alpha
\end{pmatrix},
\qquad
\varepsilon(y) =
\begin{cases}
+1, & y > 0,\\
-1, & y < 0.
\end{cases} R [ α ] = ( cos α − sin α sin α cos α ) , ε ( y ) = { + 1 , − 1 , y > 0 , y < 0.
Take ψ ~ μ 2 \widetilde{\psi}_{\mu2} ψ μ 2 to be continuous and odd, so that its one-sided limits vanish at the fixed point. (173) then gives
ψ μ 2 ( 0 + ) ψ μ 1 ( 0 + ) = tan Θ 0 = M 0 , ψ μ 2 ( 0 − ) ψ μ 1 ( 0 − ) = − tan Θ 0 , \frac{\psi_{\mu2}(0^+)}{\psi_{\mu1}(0^+)} = \tan\Theta_0 = M_0,
\qquad
\frac{\psi_{\mu2}(0^-)}{\psi_{\mu1}(0^-)} = -\tan\Theta_0, ψ μ 1 ( 0 + ) ψ μ 2 ( 0 + ) = tan Θ 0 = M 0 , ψ μ 1 ( 0 − ) ψ μ 2 ( 0 − ) = − tan Θ 0 ,
and reproduces exactly the discontinuous representative obtained from the matching condition.
The information has not disappeared: it has moved into the differential operator. Since ∂ y ε ( y ) = 2 δ ( y ) \partial_y\varepsilon(y) = 2\delta(y) ∂ y ε ( y ) = 2 δ ( y ) , differentiating the y y y -dependent rotation generates a localized contribution, schematically
R − 1 ∂ y R ∝ Θ 0 δ ( y ) . (174) R^{-1}\partial_y R \propto\Theta_0\,\delta(y). \tag*{(174)} R − 1 ∂ y R ∝ Θ 0 δ ( y ) . (174)
Thus a discontinuous field with a simple bulk operator and a continuous field with a singular localized operator can describe the same matching problem.
What is invariant is neither the value assigned to an odd field exactly at the fixed point nor the continuity of a chosen field representative. The physical information is contained in the one-sided limits and in the matching relation that connects them.
Putting the pieces together
Localized boundary operators affect Scherk–Schwarz compactifications at several logically distinct levels.
At the classical level, they modify the global spectral problem. For second-order fields this appears through Robin or eigenvalue-dependent boundary conditions and through a modified kinetic inner product. For first-order fermionic systems it appears as matching conditions between components.
For energy-independent gravitino boundary rotations, the effect on the spin-3 2 \frac{3}{2} 2 3 spectrum can be summarized by
ω e f f = ω + arctan M 0 + arctan M π π ( m o d 1 ) . (175) \omega_{\mathrm{eff}} = \omega+ \frac{\arctan M_0 + \arctan M_\pi}{\pi} \pmod{1}. \tag*{(175)} ω eff = ω + π arctan M 0 + arctan M π ( mod 1 ) . (175)
This description ceases to be universal when the boundary conditions depend on the eigenvalue, as happens generically in the presence of localized kinetic terms.
At the supersymmetric level, one must distinguish a redistribution of the same Scherk–Schwarz rotation from an independent localized breaking sector. The latter can introduce additional fermion shifts and Goldstino directions and therefore requires analysis of the complete fermionic system.
At the quantum level, bulk and boundary operators require local renormalization. This local ultraviolet dependence is distinct from the finite part that requires propagation over the compact interval. The latter depends on the relative boundary data and on the Scherk–Schwarz transport and admits the Casimir interpretation developed above.
Localized gravitino masses, discontinuous orbifold representatives, and their relation to generalized Scherk–Schwarz descriptions were studied in Refs. [30, 37]. The Wilson-line description of Scherk–Schwarz breaking on orbifolds was discussed in Ref. [38], and related brane-assisted constructions in Ref. [39]. The interplay between bulk Scherk–Schwarz breaking, localized supersymmetry-breaking sectors, the gravitino tower, and pseudo-Goldstinos was analysed explicitly in Ref. [29]. Related treatments of localized supersymmetry breaking in five-dimensional supergravity include Refs. [33, 40]. Boundary-localized counterterms and kinetic operators are generic ingredients of orbifold effective field theories. Their renormalization and the complementary finite non-local contributions will reappear when we discuss quantum effects.
Take-home message
Localized operators modify the boundary-value problem, not the orbifold group relations. They can change the Kaluza–Klein spectrum, wave functions, normalization, and localized couplings.
Localized masses and localized kinetic terms play different roles. The former modify the boundary potential; the latter also modify the kinetic norm and generally make the boundary conditions eigenvalue dependent.
For first-order fermionic systems, localized terms are naturally described by matching conditions or rotated boundary subspaces.
Energy-independent gravitino boundary rotations can be absorbed into an effective Scherk–Schwarz angle. Eigenvalue-dependent boundary conditions cannot in general be represented by one universal ω e f f \omega_{\mathrm{eff}} ω eff .
A genuine localized supersymmetry-breaking sector can contain additional fermion shifts and Goldstino directions. Agreement of the gravitino spectrum alone does not establish equivalence to a Scherk–Schwarz rotation.
Boundary ultraviolet divergences are local and renormalize boundary operators. The finite contribution requiring propagation between the two boundaries is non-local and has the Casimir interpretation described above.
This naturally leads to the next question:
How is a Scherk–Schwarz twist related to an R R R -symmetry Wilson line and to the radion F F F -term?
Essential idea
A continuous Scherk–Schwarz twist can be moved from the boundary condition into the compact derivative by a y y y -dependent change of field basis. Periodic fields then see a flat connection whose holonomy contains the same global information as the original monodromy.
When the twisting symmetry acts on the supercharges, it is an R R R -symmetry. In supergravity the corresponding connection enters the transport of the supersymmetry parameter and the gravitino sector.
In a lower-dimensional effective supergravity obtained from a more complete compactification, the same breaking may appear through auxiliary fields of several compactification moduli. In a five-dimensional description adapted to a single radial modulus, the relevant auxiliary direction is packaged into the radion F F F -term. The relation between that F F F -term and the Scherk–Schwarz parameter depends on the choice of fields and normalization.
From the Scherk–Schwarz ansatz to a flat connection
Consider a field Φ \Phi Φ on a circle of radius R R R , with
Φ t w ( y + 2 π R ) = U S S ( Φ ) ( ω ) Φ t w ( y ) , U S S ( Φ ) ( ω ) = exp ( 2 π i ω R Φ ) . (176) \Phi_{\mathrm{tw}}(y+2\pi R)=U_{\mathrm{SS}}^{(\Phi)}(\omega)\Phi_{\mathrm{tw}}(y),\qquad U_{\mathrm{SS}}^{(\Phi)}(\omega)=\exp(2\pi i\omega R_{\Phi}). \tag*{(176)} Φ tw ( y + 2 π R ) = U SS ( Φ ) ( ω ) Φ tw ( y ) , U SS ( Φ ) ( ω ) = exp ( 2 π iω R Φ ) . (176)
Here R Φ R_{\Phi} R Φ denotes the generator of the twisting symmetry in the representation carried by Φ \Phi Φ . This equation fixes our normalization of ω \omega ω .
Introduce the periodic representative
Φ p e r ( y ) = exp ( − i ω y R R Φ ) Φ t w ( y ) . (177) \Phi_{\mathrm{per}}(y)=\exp\left(-i\frac{\omega y}{R}R_{\Phi}\right)\Phi_{\mathrm{tw}}(y). \tag*{(177)} Φ per ( y ) = exp ( − i R ω y R Φ ) Φ tw ( y ) . (177)
It satisfies
Φ p e r ( y + 2 π R ) = Φ p e r ( y ) . \Phi_{\mathrm{per}}(y+2\pi R)=\Phi_{\mathrm{per}}(y). Φ per ( y + 2 π R ) = Φ per ( y ) .
Using the inverse transformation gives
∂ y Φ t w = e i ω y R Φ / R ( ∂ y + i ω R R Φ ) Φ p e r . (178) \partial_y\Phi_{\mathrm{tw}}=e^{i\omega yR_{\Phi}/R}\left(\partial_y+i\frac{\omega}{R}R_{\Phi}\right)\Phi_{\mathrm{per}}. \tag*{(178)} ∂ y Φ tw = e iω y R Φ / R ( ∂ y + i R ω R Φ ) Φ per . (178)
The compact derivative acting on periodic fields therefore takes the form
D y = ∂ y + A y ( Φ ) , A y ( Φ ) = i ω R R Φ . (179) D_y=\partial_y+\mathcal{A}_y^{(\Phi)},\qquad\mathcal{A}_y^{(\Phi)}=\frac{i\omega}{R}R_{\Phi}. \tag*{(179)} D y = ∂ y + A y ( Φ ) , A y ( Φ ) = R iω R Φ . (179)
For a constant twist this connection is flat.
The matrix obtained by going once around the positively oriented circle is
H Φ = P exp [ ∮ d y A y ( Φ ) ] = e 2 π i ω R Φ = U S S ( Φ ) ( ω ) . (180) \mathcal{H}_{\Phi}=\mathcal{P}\exp\left[\oint dy\,\mathcal{A}^{(\Phi)}_{y}\right]=e^{2\pi i\omega\mathcal{R}_{\Phi}}= \mathcal{U}^{(\Phi)}_{\mathrm{SS}}(\omega). \tag*{(180)} H Φ = P exp [ ∮ d y A y ( Φ ) ] = e 2 π iω R Φ = U SS ( Φ ) ( ω ) . (180)
We denote the holonomy by H Φ \mathcal{H}_{\Phi} H Φ , reserving W W W for the four-dimensional superpotential.
The transport convention used in Question 9 was
U ( y 2 , y 1 ) = P exp [ − ∫ y 1 y 2 d y A y ] . \mathcal{U}(y_{2},y_{1})=\mathcal{P}\exp\left[-\int_{y_{1}}^{y_{2}}dy\,\mathcal{A}_{y}\right]. U ( y 2 , y 1 ) = P exp [ − ∫ y 1 y 2 d y A y ] .
For the constant connection considered here, the transport operator along the same oriented path is therefore the inverse of the holonomy matrix,
U ( 2 π R , 0 ) = H Φ − 1 . \mathcal{U}(2\pi R,0)=\mathcal{H}_{\Phi}^{-1}. U ( 2 π R , 0 ) = H Φ − 1 .
For a general non-Abelian, y y y -dependent connection, the relation between a path-ordered exponential and its inverse also involves reversing the path ordering.
When the twisting symmetry acts on supersymmetry
The preceding construction applies to any continuous symmetry admitted by the higher-dimensional theory. If the symmetry acts on the supersymmetry generators, its action on the supersymmetry parameter is described by
D y ϵ = ( ∂ y + i ω R R ϵ ) ϵ . (181) D_y\epsilon=\left(\partial_y+i\frac{\omega}{R}\mathcal{R}_{\epsilon}\right)\epsilon. \tag*{(181)} D y ϵ = ( ∂ y + i R ω R ϵ ) ϵ . (181)
The corresponding holonomy is
H ϵ = exp ( 2 π i ω R ϵ ) . \mathcal{H}_{\epsilon}=\exp(2\pi i\omega\mathcal{R}_{\epsilon}). H ϵ = exp ( 2 π iω R ϵ ) .
Subject to the remaining Killing-spinor equations, the supersymmetries preserved on the flat circle obey
S p r e s e r v e d = ker ( H ϵ − 1 ) . (182) S_{\mathrm{preserved}}=\ker(\mathcal{H}_{\epsilon}-1). \tag*{(182)} S preserved = ker ( H ϵ − 1 ) . (182)
This is the invariant-subspace condition encountered earlier, now written in the connection basis.
The ingredients needed for this interpretation are already visible in the original generalized reductions [6, 7]. In the supergravity example of Ref. [6], the compact-coordinate dependence acts chirally on the gravitino while leaving the vielbein untwisted. A formulation in which the compact component of a local internal connection enters the fermionic mass terms can be found in Ref. [36].
When the symmetry is realized locally in the supergravity formulation, A y ( R ) \mathcal{A}^{(R)}_{y} A y ( R ) is an R R R -symmetry connection and the flat background may be described as an R R R -symmetry Wilson line. A detailed off-shell realization on an orbifold can be found in Ref. [38].
The same connection appears in the local supersymmetry transformation of the internal gravitino,
δ ψ y = 2 D y ϵ + ⋯ . (183) \delta\psi_y=2D_y\epsilon+\cdots. \tag*{(183)} δ ψ y = 2 D y ϵ + ⋯ . (183)
With the notation introduced in Question 6, the constant part of the internal operator is
M = i ω R R ϵ . (184) \mathcal{M}=i\frac{\omega}{R}\mathcal{R}_{\epsilon}. \tag*{(184)} M = i R ω R ϵ . (184)
This identifies explicitly the matrix M \mathcal{M} M used there with the action of the Scherk–Schwarz generator on the supersymmetry parameters.
On S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 , the underlying topological space is an interval and contains no non-contractible closed loop. The circle holonomy should therefore be understood on the covering space. In an interval description the same compactification data are distributed between the endpoint actions and the open transport between them.
Using the convention of Question 9, the latter is simply
U ( π R , 0 ) = P exp [ − ∫ 0 π R d y A y ] . (185) \mathcal{U}(\pi R,0)=\mathcal{P}\exp\left[-\int_{0}^{\pi R}dy\,\mathcal{A}_y\right]. \tag*{(185)} U ( π R , 0 ) = P exp [ − ∫ 0 π R d y A y ] . (185)
Under a change of basis G ( y ) G(y) G ( y ) ,
U ( π R , 0 ) ⟶ G ( π R ) − 1 U ( π R , 0 ) G ( 0 ) , P b ⟶ G ( y b ) − 1 P b G ( y b ) , y b = 0 , π R . (186) \begin{aligned}
\mathcal{U}(\pi R,0)&\longrightarrow G(\pi R)^{-1}\mathcal{U}(\pi R,0)G(0),\\
P_b&\longrightarrow G(y_b)^{-1}P_bG(y_b),\qquad y_b=0,\pi R. \tag*{(186)}
\end{aligned} U ( π R , 0 ) P b ⟶ G ( π R ) − 1 U ( π R , 0 ) G ( 0 ) , ⟶ G ( y b ) − 1 P b G ( y b ) , y b = 0 , π R . (186)
Neither the open transport nor the endpoint matrices should be interpreted separately. This is the change-of-trivialization freedom already used in Questions 8 and 9.
Compactification moduli and the radial direction
The effective theory obtained from a complete compactification need not contain a field whose name and normalization coincide with the radion of a simple five-dimensional truncation. In four-dimensional N = 1 N=1 N = 1 supergravity, let
G = K + ln ∣ W ∣ 2 . G=K+\ln\lvert W\rvert^2. G = K + ln ∣ W ∣ 2 .
On a bosonic background the chiral auxiliary fields are proportional, up to standard convention-dependent factors, to
F I ∝ e G / 2 G I , G I = G I J ˉ G J ˉ . F^I\propto e^{G/2}G^I,\qquad G^I=G^{I\bar J}G_{\bar J}. F I ∝ e G /2 G I , G I = G I J ˉ G J ˉ .
Supersymmetry breaking can therefore point along several coordinates of the compactification-moduli space.
A string-derived effective-supergravity description of coordinate-dependent compactification can be found in Ref. [ref-41 ]. In one of the examples discussed there, the supersymmetry-breaking direction satisfies
G S G S = G T ′ G T ′ = G U ′ G U ′ = 1. (187) G^S G_S=G^{T'}G_{T'}=G^{U'}G_{U'}=1. \tag*{(187)} G S G S = G T ′ G T ′ = G U ′ G U ′ = 1. (187)
while
m 3 / 2 2 = k 2 ( S + S ˉ ) ( T ′ + T ˉ ′ ) ( U ′ + U ˉ ′ ) = ( e 1 + e 2 ) 2 g U 2 2 R 2 , (188) m_{3/2}^2=\frac{k^2}{(S+\bar S)(T'+\bar T')(U'+\bar U')}=\frac{(e_1+e_2)^2g_U^2}{2R^2}, \tag*{(188)} m 3/2 2 = ( S + S ˉ ) ( T ′ + T ˉ ′ ) ( U ′ + U ˉ ′ ) k 2 = 2 R 2 ( e 1 + e 2 ) 2 g U 2 , (188)
with
R = 2 ( T ′ + T ˉ ′ ) ( U ′ + U ˉ ′ ) . (189) R=2\sqrt{(T'+\bar T')(U'+\bar U')}. \tag*{(189)} R = 2 ( T ′ + T ˉ ′ ) ( U ′ + U ˉ ′ ) . (189)
In the conventions of Ref. [41], k = ( e 1 + e 2 ) / 2 k=(e_1+e_2)/2 k = ( e 1 + e 2 ) /2 is the constant term of the superpotential in Planck units, e 1 , e 2 e_1,e_2 e 1 , e 2 are quantized charges, and g U g_U g U is the unified gauge coupling. These quantities belong to that string-derived example, rather than to the general Scherk–Schwarz circle. The effective supergravity describes the Scherk–Schwarz background through non-vanishing auxiliary directions in the compactification-moduli sector and gives the characteristic scaling m 3 / 2 ∝ R − 1 m_{3/2}\propto R^{-1} m 3/2 ∝ R − 1 .
The variables S , T ′ , U ′ S,T',U' S , T ′ , U ′ are coordinates adapted to that four-dimensional effective description. They need not coincide with fields adapted to a higher-dimensional Einstein-frame parametrization. The physical radius in Eq. (189) is already a combination of the T ′ T' T ′ - and U ′ U' U ′ -type moduli, and the relation between the higher-dimensional metric and the four-dimensional Einstein frame also involves the dilaton sector.
If the radial degree of freedom can be isolated locally as a chiral coordinate
T r a d = T r a d ( Z I ) , T_{\mathrm{rad}}=T_{\mathrm{rad}}(Z^I), T rad = T rad ( Z I ) ,
its auxiliary field transforms under a holomorphic change of coordinates as
F T r a d = ∂ T r a d ∂ Z I F I . (190) F^{T_{\mathrm{rad}}}=\frac{\partial T_{\mathrm{rad}}}{\partial Z^I}F^I. \tag*{(190)} F T rad = ∂ Z I ∂ T rad F I . (190)
Several non-zero F I F^I F I in one parametrization can therefore contribute to a single auxiliary direction in another. Determining the precise combination requires the map between the four-dimensional moduli and the higher-dimensional radion. We shall not identify any one of the auxiliary fields appearing in Eq. (187) with F T F_T F T by itself.
An adapted five-dimensional radion description
For a flat five-dimensional theory in which the radial degree of freedom has already been isolated, the relation becomes particularly simple. A four-dimensional N = 1 N=1 N = 1 superfield formulation with dimensionless coordinate y ^ ∼ y ^ + 2 π \hat{y}\sim\hat{y}+2\pi y ^ ∼ y ^ + 2 π and
d s 2 = η μ ν d x μ d x ν + R 2 d y ^ 2 ds^2=\eta_{\mu\nu}dx^\mu dx^\nu+R^2d\hat{y}^2 d s 2 = η μν d x μ d x ν + R 2 d y ^ 2
can be written in terms of the radion multiplet
T r a d = R + i B 5 + θ Ψ R 5 + θ 2 F T . (191) T_{\mathrm{rad}}=R+iB_5+\theta\Psi_R^5+\theta^2F_T. \tag*{(191)} T rad = R + i B 5 + θ Ψ R 5 + θ 2 F T . (191)
A realization of this form can be found in Ref. [42].
For a real constant F T F_T F T , the non-zero Kaluza–Klein levels of the five-dimensional gaugino doublet have mass matrix
M n = 1 R ( F T / 2 n n F T / 2 ) , n ≥ 1. (192) \mathcal{M}_n=\frac{1}{R}
\begin{pmatrix}
F_T/2 & n \\
n & F_T/2
\end{pmatrix},
\qquad n\ge1. \tag*{(192)} M n = R 1 ( F T /2 n n F T /2 ) , n ≥ 1. (192)
Its eigenvalues give
m n , ± = ∣ n ± F T / 2 ∣ R , (193) m_{n,\pm}=\frac{\lvert n\pm F_T/2\rvert}{R}, \tag*{(193)} m n , ± = R ∣ n ± F T /2 ∣ , (193)
while the zero-mode gaugino has mass
m λ , 0 = ∣ F T ∣ 2 R . m_{\lambda,0}=\frac{\lvert F_T\rvert}{2R}. m λ , 0 = 2 R ∣ F T ∣ .
In the normalization used in that formulation, the Scherk–Schwarz shift is
q R = F T 2 . (194) q_R = \frac{F_T}{2}. \tag*{(194)} q R = 2 F T . (194)
The relation can also be displayed directly by a field redefinition. For an R R R -charge eigenstate q = ± 1 q = \pm1 q = ± 1 , consider
D q = 1 R ( − i ∂ y ^ + q F T 2 ) . (195) \mathcal{D}_q = \frac{1}{R}\left(-i\partial_{\hat{y}} + q\frac{F_T}{2}\right). \tag*{(195)} D q = R 1 ( − i ∂ y ^ + q 2 F T ) . (195)
Define
λ q p e r ( y ^ ) = e − i q F T y ^ / 2 λ q t w ( y ^ ) . (196) \lambda_q^{\mathrm{per}}(\hat{y}) = e^{-iqF_T\hat{y}/2}\lambda_q^{\mathrm{tw}}(\hat{y}). \tag*{(196)} λ q per ( y ^ ) = e − i q F T y ^ /2 λ q tw ( y ^ ) . (196)
The constant term in Eq. (195) is removed, while the periodicity of λ q p e r \lambda_q^{\mathrm{per}} λ q per gives
λ q t w ( y ^ + 2 π ) = e i π q F T λ q t w ( y ^ ) . (197) \lambda_q^{\mathrm{tw}}(\hat{y} + 2\pi) = e^{i\pi qF_T}\lambda_q^{\mathrm{tw}}(\hat{y}). \tag*{(197)} λ q tw ( y ^ + 2 π ) = e iπ q F T λ q tw ( y ^ ) . (197)
Comparing with Eq. (176) , for a generator with eigenvalues q = ± 1 q = \pm1 q = ± 1 ,
ω = F T 2 (198) \omega= \frac{F_T}{2} \tag*{(198)} ω = 2 F T (198)
in these conventions. A component treatment of the corresponding field redefinition can be found in Ref. [43].
The numerical relation in Eq. (198) is not formulation independent. Rescaling the radion superfield or the R R R -symmetry generator changes it. The invariant statement is the agreement of the monodromy, or of the corresponding shifted compact operator, after the fields and generators have been normalized consistently.
The essential ingredients of generalized dimensional reduction, including a compact-coordinate dependence generated by a symmetry acting on the gravitino, can already be found in Refs. [6, 7]. A description in which the compact component of a local internal-symmetry connection contributes to the fermion mass matrix can be found in Ref. [36].
Descriptions of coordinate-dependent compactifications in string-derived four-dimensional effective supergravities, with supersymmetry breaking along compactification-modulus directions, can be found in Refs. [14, 15, 41]. A five-dimensional off-shell formulation making the R R R -symmetry-connection description explicit on orbifolds can be found in Ref. [38]. A superfield formulation in which the radial multiplet is isolated and its auxiliary component produces the shifted Kaluza–Klein spectrum can be found in Ref. [42]. The relation to twisted boundary conditions through a field redefinition is discussed in Ref. [43].
Figure 4. Three descriptions used in this question. In the twisted basis, the compactification is specified by the monodromy U S S U_{\mathrm{SS}} U SS . A y y y -dependent change of field basis gives periodic fields with a flat R R R -symmetry connection. In a formulation adapted to the radial multiplet, the same Scherk–Schwarz shift can be encoded in a non-zero auxiliary field F T F_T F T . The numerical relation between F T F_T F T and the twist parameter depends on the normalization of the radion and of the R R R -symmetry generator.
Take-home message
This naturally leads to the next question:
What is special about the half-twist, and why does a Dirac gravitino appear?
What is special about the half-twist, and why does a Dirac gravitino appear?
Essential idea
At the half-twist, ω = 1 / 2 \omega= 1/2 ω = 1/2 , the Kaluza–Klein masses of the five-dimensional gravitino occur in equal-mass pairs. The two ends of the orbifold preserve different supersymmetries, so none survives on the whole interval. Yet a U ( 1 ) R U(1)_{R} U ( 1 ) R symmetry remains. It allows the two gravitini to form a Dirac field, with mass 1 / ( 2 R ) 1/(2R) 1/ ( 2 R ) for the lightest pair, and forbids separate Majorana masses.
The Dirac mass is already present in the local five-dimensional Rarita–Schwinger action. Its effective description in four-dimensional superspace is a separate question, to which we turn next.
The physical picture
There are two ways of thinking about the half-twist. The first is to follow a field around the circle. With ordinary periodic boundary conditions its compact momentum is an integer multiple of 1 / R 1/R 1/ R . An anti-periodic fermion picks up a minus sign on a complete turn; its momenta are instead half-integer multiples of 1 / R 1/R 1/ R . The first possible mass is 1 / ( 2 R ) 1/(2R) 1/ ( 2 R ) , followed by 3 / ( 2 R ) 3/(2R) 3/ ( 2 R ) , 5 / ( 2 R ) 5/(2R) 5/ ( 2 R ) , and so on. There is no massless gravitino.
The second picture uses the interval 0 ≤ y ≤ π R 0 \le y \le\pi R 0 ≤ y ≤ π R . Each end admits one direction in the doublet of supersymmetry parameters. At the ordinary untwisted point the directions agree. At the half-twist they are orthogonal. A spinor that is constant along the interval cannot belong to both. This explains why the compactified background has no unbroken supersymmetry, although the original five-dimensional theory is locally supersymmetric.
A minus sign around the circle might seem to destroy every continuous R R R -symmetry. It does not. On the gravitino doublet, the half-twist is the central matrix − 1 -1 − 1 , so it commutes with every S U ( 2 ) R SU(2)_{R} S U ( 2 ) R rotation. The orbifold projections restrict these rotations, but leave a U ( 1 ) R U(1)_{R} U ( 1 ) R . The two gravitini carry opposite charges under this U ( 1 ) R U(1)_{R} U ( 1 ) R . Their mixed mass has zero charge, while either Majorana mass has a non-zero charge. This is the physical distinction between an accidental degeneracy and a Dirac gravitino.
The twist acts on the supersymmetry doublet, not on the four-dimensional graviton. The graviton retains its massless zero mode. Its partner gravitino does not. When the two gravitini are written as N = 1 N = 1 N = 1 superfields, their different multiplet assignments make the Dirac mass operator less familiar than the spectrum itself.
The pairing of the signed Kaluza–Klein masses and the U ( 1 ) R U(1)_{R} U ( 1 ) R selection rule can both be derived without four-dimensional superspace.
A symmetry of the signed spectrum
We saw the shifted spectrum in Question 4: for a phase twist on the circle, the signed masses are
m n ( ω ) = n + ω R , n ∈ Z . m_n(\omega) = \frac{n+\omega}{R}, \qquad n \in\mathbb{Z}. m n ( ω ) = R n + ω , n ∈ Z .
The sign is useful here, even though the physical mass of a given mode is M n = ∣ m n ∣ M_n = |m_n| M n = ∣ m n ∣ . Let D ω D_\omega D ω act on the Fourier labels by D ω ∣ n ⟩ = m n ( ω ) ∣ n ⟩ D_\omega|n\rangle= m_n(\omega)|n\rangle D ω ∣ n ⟩ = m n ( ω ) ∣ n ⟩ , and introduce the reflection J ∣ n ⟩ = ∣ − n − 1 ⟩ J|n\rangle= |-n-1\rangle J ∣ n ⟩ = ∣ − n − 1 ⟩ . A short calculation gives
J D ω J − 1 = − D 1 − ω . (199) J D_\omega J^{-1} = -D_{1-\omega}. \tag*{(199)} J D ω J − 1 = − D 1 − ω . (199)
In particular,
{ J , D 1 / 2 } = 0 , m − n − 1 ( 1 2 ) = − m n ( 1 2 ) . (200) \{J,D_{1/2}\} = 0, \qquad m_{-n-1}\left(\frac{1}{2}\right) = -m_n\left(\frac{1}{2}\right). \tag*{(200)} { J , D 1/2 } = 0 , m − n − 1 ( 2 1 ) = − m n ( 2 1 ) . (200)
There is no integer fixed by n ↦ − n − 1 n \mapsto-n-1 n ↦ − n − 1 . The lowest pair has signed masses + 1 / ( 2 R ) +1/(2R) + 1/ ( 2 R ) and − 1 / ( 2 R ) -1/(2R) − 1/ ( 2 R ) , followed by ± 3 / ( 2 R ) \pm3/(2R) ± 3/ ( 2 R ) , ± 5 / ( 2 R ) \pm5/(2R) ± 5/ ( 2 R ) , and so forth.
The equality of masses by itself does not guarantee a Dirac particle. Two Majorana fermions may be degenerate without carrying a conserved charge. The labels n n n and − n − 1 -n-1 − n − 1 can be used for the two Weyl vector-spinors ψ μ k \psi_{\mu k} ψ μ k and ρ μ k \rho_{\mu k} ρ μ k of one orbifold level k ≥ 0 k \ge0 k ≥ 0 . This is one Dirac gravitino, not two copies of it. Pairing alone does not protect it: that requires a symmetry of the full boundary problem.
What the two boundaries preserve
Take the gravitino S U ( 2 ) R SU(2)_R S U ( 2 ) R doublet in the orbifold basis and set
P 0 = τ 3 , U ( ω ) = e 2 π i ω τ 2 , P π ( ω ) = U ( ω ) P 0 . P_0 = \tau_3, \qquad U(\omega) = e^{2\pi i\omega\tau_2}, \qquad P_\pi(\omega) = U(\omega)P_0. P 0 = τ 3 , U ( ω ) = e 2 π iω τ 2 , P π ( ω ) = U ( ω ) P 0 .
The τ a \tau_a τ a act on the internal doublet, not on the Lorentz spinor indices. The consistency condition P 0 U P 0 = U − 1 P_0 U P_0 = U^{-1} P 0 U P 0 = U − 1 is automatic. At the half-twist,
U ( 1 2 ) = − 1 , P π = − P 0 = − τ 3 . (201) U\left(\frac{1}{2}\right) = -1, \qquad P_\pi= -P_0 = -\tau_3. \tag*{(201)} U ( 2 1 ) = − 1 , P π = − P 0 = − τ 3 . (201)
We use the twisted-boundary frame: the bulk derivative is the ordinary ∂ y \partial_y ∂ y , while the doublet obeys ϵ ( y + 2 π R ) = U ( ω ) ϵ ( y ) \epsilon(y+2\pi R) = U(\omega)\epsilon(y) ϵ ( y + 2 π R ) = U ( ω ) ϵ ( y ) . A massless global supersymmetry parameter would satisfy ∂ y ϵ = 0 \partial_y\epsilon= 0 ∂ y ϵ = 0 and both endpoint projections. At ω = 1 / 2 \omega= 1/2 ω = 1/2 these projectors select ( 1 , 0 ) T (1,0)^T ( 1 , 0 ) T at y = 0 y=0 y = 0 and ( 0 , 1 ) T (0,1)^T ( 0 , 1 ) T at y = π R y=\pi R y = π R . Thus
S 0 ∩ S π = { 0 } . (202) S_0 \cap S_\pi= \{0\}. \tag*{(202)} S 0 ∩ S π = { 0 } . (202)
Each endpoint still admits a local supersymmetry direction, but no constant spinor belongs to both. The Goldstino and super-Higgs aspects were discussed in Questions 5, 6 and 7.
On the gravitino doublet, the central holonomy U ( 1 / 2 ) = − 1 U(1/2)=-1 U ( 1/2 ) = − 1 has the same action as ( − 1 ) F (-1)^F ( − 1 ) F . The orbifold projections select a smaller surviving group, namely
G b c ( ω ) = Z S U ( 2 ) R ( P 0 ) ∩ Z S U ( 2 ) R ( P π ( ω ) ) . (203) G_{\mathrm{bc}}(\omega) = Z_{SU(2)_R}(P_0) \cap Z_{SU(2)_R}(P_\pi(\omega)). \tag*{(203)} G bc ( ω ) = Z S U ( 2 ) R ( P 0 ) ∩ Z S U ( 2 ) R ( P π ( ω )) . (203)
For generic ω \omega ω , the two projection axes are different, and G b c G_{\mathrm{bc}} G bc contains only { ± 1 } \{\pm1\} { ± 1 } . At ω = 1 / 2 \omega= 1/2 ω = 1/2 the axes coincide, up to a sign; the common centralizer is therefore
G b c ( 1 2 ) = U ( 1 ) τ 3 . (204) G_{\mathrm{bc}}\left(\frac{1}{2}\right)=U(1)_{\tau_3}. \tag*{(204)} G bc ( 2 1 ) = U ( 1 ) τ 3 . (204)
At ω = 0 \omega=0 ω = 0 this group is also U ( 1 ) U(1) U ( 1 ) , but there P π = P 0 P_{\pi}=P_0 P π = P 0 and one supersymmetry survives. The sign of the second projection changes the physics.
The mass matrix takes different forms in the two familiar bases. In the orbifold basis Q R = τ 3 Q_R=\tau_3 Q R = τ 3 , and an allowed mass matrix is proportional to τ 1 \tau_1 τ 1 . With the change of basis
ψ M a j = V ψ o r b , V = 1 2 ( 1 1 − i i ) , V τ 3 V − 1 = − σ 2 , ( V − 1 ) T τ 1 V − 1 = 1. \psi_{\mathrm{Maj}}=V\psi_{\mathrm{orb}},\qquad V=\frac{1}{\sqrt{2}}\begin{pmatrix}1&1\\-i&i\end{pmatrix},\qquad V\tau_3V^{-1}=-\sigma_2,\qquad(V^{-1})^T\tau_1V^{-1}=1. ψ Maj = V ψ orb , V = 2 1 ( 1 − i 1 i ) , V τ 3 V − 1 = − σ 2 , ( V − 1 ) T τ 1 V − 1 = 1.
the mass matrix becomes proportional to 1 1 1 and the charge generator becomes − σ 2 -\sigma_2 − σ 2 . The two equal Majorana masses are therefore invariant under real S O ( 2 ) SO(2) S O ( 2 ) rotations: for O ( α ) = e − i α σ 2 O(\alpha)=e^{-i\alpha\sigma_2} O ( α ) = e − i α σ 2 , one has O T O = 1 O^TO=1 O T O = 1 . In the charged basis the same invariance is the off-diagonal Dirac selection rule.
Indeed, for two Weyl gravitini with charges + 1 +1 + 1 and − 1 -1 − 1 , a symmetric mass matrix obeys
Q R T M 3 / 2 + M 3 / 2 Q R = 0 , Q R = ( 1 0 0 − 1 ) . (205) Q_R^T\mathcal{M}_{3/2}+\mathcal{M}_{3/2}Q_R=0,\qquad Q_R=\begin{pmatrix}1&0\\0&-1\end{pmatrix}. \tag*{(205)} Q R T M 3/2 + M 3/2 Q R = 0 , Q R = ( 1 0 0 − 1 ) . (205)
The diagonal terms must vanish, leaving
M 3 / 2 = ( 0 m D m D 0 ) . (206) \mathcal{M}_{3/2}=\begin{pmatrix}0&m_D\\m_D&0\end{pmatrix}. \tag*{(206)} M 3/2 = ( 0 m D m D 0 ) . (206)
The eigenvalues of the mass matrix in Eq. (206) are + m D +m_D + m D and − m D -m_D − m D . They are the signed masses of the lowest pair in Eq. (200) , with m D = 1 / ( 2 R ) m_D=1/(2R) m D = 1/ ( 2 R ) . For canonical Rarita–Schwinger kinetic terms, the mass bilinear in the charged Weyl basis is
L m a s s = − 2 m D ψ + μ σ μ ν ψ − ν + h . c . (207) \mathcal{L}_{\mathrm{mass}}=-2m_D\psi_+^\mu\sigma_{\mu\nu}\psi_-^\nu+\mathrm{h.c.} \tag*{(207)} L mass = − 2 m D ψ + μ σ μν ψ − ν + h.c. (207)
The two off-diagonal entries carry one mass m D m_D m D and give the coefficient 2 m D 2m_D 2 m D . The canonical normalization is checked in Question 12. The action of the lowest pair is
L p h y s = L R S [ ψ + ] + L R S [ ψ − ] − 2 m D ψ + μ σ μ ν ψ − ν + h . c . , m D = 1 2 R . (208) \mathcal{L}_{\mathrm{phys}}=\mathcal{L}_{\mathrm{RS}}[\psi_+]+\mathcal{L}_{\mathrm{RS}}[\psi_-]-2m_D\psi_+^\mu\sigma_{\mu\nu}\psi_-^\nu+\mathrm{h.c.},\qquad m_D=\frac{1}{2R}. \tag*{(208)} L phys = L RS [ ψ + ] + L RS [ ψ − ] − 2 m D ψ + μ σ μν ψ − ν + h.c. , m D = 2 R 1 . (208)
In this equation L R S [ ψ ] \mathcal{L}_{\mathrm{RS}}[\psi] L RS [ ψ ] is the canonically normalized Weyl Rarita–Schwinger kinetic term.
The U ( 1 ) R U(1)_R U ( 1 ) R is a symmetry of the pure half-twisted bulk and its boundary conditions. If localized masses or matter interactions break it, the diagonal Majorana terms are no longer forbidden.
From the fifth derivative to a Dirac mass
The mode profiles give a useful check that does not require any four-dimensional superspace. Write ℓ = π R \ell=\pi R ℓ = π R and m k = ( k + 1 2 ) / R m_k=(k+\frac{1}{2})/R m k = ( k + 2 1 ) / R . In a twisted-boundary frame, the two components of the five-dimensional gravitino have profiles proportional to
Ψ μ 1 ( x , y ) = 2 ℓ ∑ k ≥ 0 cos ( m k y ) ψ μ k ( x ) , Ψ μ 2 ( x , y ) = − 2 ℓ ∑ k ≥ 0 sin ( m k y ) ρ μ k ( x ) . (209) \Psi^1_\mu(x,y)=\sqrt{\frac{2}{\ell}}\sum_{k\geq0}\cos(m_k y)\,\psi_{\mu k}(x),
\qquad
\Psi^2_\mu(x,y)=-\sqrt{\frac{2}{\ell}}\sum_{k\geq0}\sin(m_k y)\,\rho_{\mu k}(x). \tag*{(209)} Ψ μ 1 ( x , y ) = ℓ 2 k ≥ 0 ∑ cos ( m k y ) ψ μ k ( x ) , Ψ μ 2 ( x , y ) = − ℓ 2 k ≥ 0 ∑ sin ( m k y ) ρ μ k ( x ) . (209)
The fields ψ μ k , α \psi_{\mu k,\alpha} ψ μ k , α and ρ μ k , α \rho_{\mu k,\alpha} ρ μ k , α are the left-handed four-dimensional Weyl components of the five-dimensional symplectic-Majorana doublet. Their dotted conjugates give the corresponding right-handed components; the symplectic-Majorana condition avoids an independent doubling of the states.
The profiles in Eq. (209) obey the endpoint projections directly:
Ψ μ 2 ( x , 0 ) = 0 , Ψ μ 1 ( x , ℓ ) = 0 , sin ( m k ℓ ) = ( − 1 ) k . \Psi^2_\mu(x,0)=0,\qquad\Psi^1_\mu(x,\ell)=0,\qquad\sin(m_k\ell)=(-1)^k. Ψ μ 2 ( x , 0 ) = 0 , Ψ μ 1 ( x , ℓ ) = 0 , sin ( m k ℓ ) = ( − 1 ) k .
Thus P 0 = τ 3 P_0=\tau_3 P 0 = τ 3 selects the first component at y = 0 y=0 y = 0 , whereas P π = − τ 3 P_\pi=-\tau_3 P π = − τ 3 selects the second at y = ℓ y=\ell y = ℓ .
The fifth-dimensional Rarita–Schwinger term contains an internal first-derivative coupling between the two Weyl components. Its relevant overlap is
2 ℓ ∫ 0 ℓ d y cos ( m k y ) ∂ y sin ( m k ′ y ) = m k ′ δ k k ′ . (210) \frac{2}{\ell}\int_0^\ell dy\,\cos(m_k y)\partial_y\sin(m_{k'}y)=m_{k'}\delta_{kk'}. \tag*{(210)} ℓ 2 ∫ 0 ℓ d y cos ( m k y ) ∂ y sin ( m k ′ y ) = m k ′ δ k k ′ . (210)
The term with the fifth derivative acting on the other component gives the same overlap after integration by parts. With the canonical Rarita–Schwinger normalization and the sine-profile sign chosen in Eq. (209) , the two terms give
L ∂ y ⊃ − 2 ∑ k ≥ 0 m k ψ μ k σ μ ν ρ ν k + h . c . (211) \mathcal{L}_{\partial_y}\supset-2\sum_{k\geq0}m_k\,\psi_{\mu k}\sigma^{\mu\nu}\rho_{\nu k}+\mathrm{h.c.} \tag*{(211)} L ∂ y ⊃ − 2 k ≥ 0 ∑ m k ψ μ k σ μν ρ ν k + h.c. (211)
The two Weyl vector-spinors of each level form one Dirac gravitino. The first one has mass 1 / ( 2 R ) 1/(2R) 1/ ( 2 R ) ; there is no massless gravitino at the half-twist. The spectrum, boundary projections and U ( 1 ) R U(1)_R U ( 1 ) R selection rule account for the Dirac gravitino without introducing auxiliary fields or four-dimensional superspace.
Moving away from the half-twist
Finally let ω = 1 2 − ε \omega=\frac{1}{2}-\varepsilon ω = 2 1 − ε , with ∣ ε ∣ ≪ 1 |\varepsilon|\ll1 ∣ ε ∣ ≪ 1 . The two lightest physical masses become
M 0 = 1 2 − ε R , M − 1 = 1 2 + ε R , Δ M = 2 ∣ ε ∣ R . M_0=\frac{\frac{1}{2}-\varepsilon}{R},\qquad M_{-1}=\frac{\frac{1}{2}+\varepsilon}{R},\qquad\Delta M=\frac{2|\varepsilon|}{R}. M 0 = R 2 1 − ε , M − 1 = R 2 1 + ε , Δ M = R 2∣ ε ∣ .
with the labels exchanged when ε < 0 \varepsilon<0 ε < 0 . The endpoint projections no longer have the same axis. Their common centralizer is reduced to Z 2 \mathbb{Z}_2 Z 2 , and diagonal Majorana terms are no longer forbidden by the half-twist U ( 1 ) R U(1)_R U ( 1 ) R . The two nearly degenerate states then form a pseudo-Dirac pair.
A localized rotation can give the same leading spectral splitting. For energy-independent rotations of the gravitino boundary data, Question 9, Eq. (161) , gives the shifted masses in terms of ω + ( Θ 0 + Θ π ) / π \omega+(\Theta_0+\Theta_\pi)/\pi ω + ( Θ 0 + Θ π ) / π . This is an equality of the gravitino spectra for that restricted boundary problem. It says nothing by itself about localized fermions or their Goldstino directions.
To describe the mass in N = 1 N=1 N = 1 superspace, we must now do more than identify its eigenvalues. We need a gauge-invariant operator with the correct kinetic normalization, together with the momentum dependence of the full Kaluza–Klein tower. This is the subject of the next question.
The pairing of the gravitini at the half-twist and its U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R selection rule are discussed in (Pseudo-)Dirac Gravitinos [44]. The mass spectrum follows from the five-dimensional Scherk–Schwarz reduction, while the relation between boundary rotations and the spectral shift was derived in Question 9. The four-dimensional superspace operator is treated separately in the next question.
Take-home message
At ω = 1 / 2 \omega= 1/2 ω = 1/2 the signed KK masses are paired by n ↔ − n − 1 n \leftrightarrow-n - 1 n ↔ − n − 1 . Each pair gives one Dirac gravitino. The first has mass m D = 1 / ( 2 R ) m_{\mathrm{D}} = 1/(2R) m D = 1/ ( 2 R ) .
The two orbifold boundaries select orthogonal supersymmetry directions, P π = − P 0 P_{\pi} = -P_{0} P π = − P 0 . No global Killing spinor survives, but their common U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R protects the Dirac mass.
Two degenerate Majorana gravitini do not automatically form a symmetry-protected Dirac field. The residual U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R permits the off-diagonal mass and excludes independent Majorana masses.
The mass follows directly from the fifth derivative in the local five-dimensional theory. Rotating the boundary data or moving away from ω = 1 / 2 \omega= 1/2 ω = 1/2 can split the pair and produce a pseudo-Dirac gravitino.
This naturally leads to the next question:
How can the Dirac gravitino mass be described in effective superspace?
How can the Dirac gravitino mass be described in effective superspace?
Essential idea
At the half-twist, the fifth derivative gives a Dirac mass to the two four-dimensional gravitini. The first belongs to the gravity multiplet, together with the graviton. The second belongs to a companion spin-3 2 \frac{3}{2} 2 3 multiplet containing a vector. We would like to see their mass in N = 1 N=1 N = 1 superspace, without choosing a gravitino gauge.
A direct coupling between the two prepotentials fails. Its unprojected form is not gauge invariant, while its transverse projection vanishes identically. The invariant curvatures allow a non-zero transverse mass, represented by an operator containing an inverse d’Alembertian. The Kaluza–Klein tower fixes its momentum dependence.
The five-dimensional calculation also gives the diagonal kinetic response. Two superspace invariants, one Hermitian and one chiral, reproduce the fermionic curvature kernel after the boundary sources have been properly identified. This is a quadratic, fixed-background result. The full bosonic and auxiliary completion is a separate task.
The question in four dimensions
Question 11 gave the physical mass m D = 1 / ( 2 R ) m_{\mathrm{D}} = 1/(2R) m D = 1/ ( 2 R ) and its U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R protection. That answer follows from a local five-dimensional action. In four-dimensional N = 1 N=1 N = 1 language, however, the two gravitini cannot be treated as two copies of the ordinary gravitational prepotential. We must respect the independent gauge transformations of the two multiplets. We shall first test the obvious bilinear, then construct the curvature operator and match its fermionic part to the full KK response. The underlying five-dimensional calculation and its component reductions are developed in Ref. [45].
Why the direct H Ψ H\Psi H Ψ term fails
Let H α α ˙ H_{\alpha\dot{\alpha}} H α α ˙ be the real gravitational prepotential and Ψ α \Psi_{\alpha} Ψ α the spinor prepotential of the companion gravitino multiplet. At the linearized level,
δ H α α ˙ = D ˉ α ˙ L α − D α L ˉ α ˙ , δ Ψ α = D α Ω + η α , D ˉ α ˙ η β = 0. (212) \delta H_{\alpha\dot{\alpha}} = \bar{D}_{\dot{\alpha}}L_{\alpha} - D_{\alpha}\bar{L}_{\dot{\alpha}},
\qquad
\delta\Psi_{\alpha} = D_{\alpha}\Omega+ \eta_{\alpha},
\qquad
\bar{D}_{\dot{\alpha}}\eta_{\beta} = 0. \tag*{(212)} δ H α α ˙ = D ˉ α ˙ L α − D α L ˉ α ˙ , δ Ψ α = D α Ω + η α , D ˉ α ˙ η β = 0. (212)
The parameters L α L_{\alpha} L α and Ω \Omega Ω are unconstrained, and η α \eta_{\alpha} η α is chiral. The complete off-shell multiplets also contain compensators. First consider a bilinear made only from these two prepotentials.
An obvious current is
J α α ˙ ( Ψ ) = D ˉ α ˙ Ψ α − D α Ψ ˉ α ˙ , J_{\alpha\dot{\alpha}}(\Psi)=\bar{D}_{\dot{\alpha}}\Psi_{\alpha}-D_{\alpha}\bar{\Psi}_{\dot{\alpha}}, J α α ˙ ( Ψ ) = D ˉ α ˙ Ψ α − D α Ψ ˉ α ˙ ,
S H Ψ = μ ∫ d 8 z H α α ˙ J α α ˙ ( Ψ ) , d 8 z = d 4 x d 2 θ d 2 θ ˉ . (213) S_{H\Psi}=\mu\int d^{8}z\,H^{\alpha\dot{\alpha}}J_{\alpha\dot{\alpha}}(\Psi),\qquad d^{8}z=d^{4}x\,d^{2}\theta\,d^{2}\bar{\theta}. \tag*{(213)} S H Ψ = μ ∫ d 8 z H α α ˙ J α α ˙ ( Ψ ) , d 8 z = d 4 x d 2 θ d 2 θ ˉ . (213)
Vary H H H while keeping Ψ \Psi Ψ fixed. The spinor gauge parameters are Grassmann odd. Thus integration by parts uses ∫ d 8 z ( D α L ) J = + ∫ d 8 z L ( D α J ) \int d^{8}z\,(D_{\alpha}L)J=+\int d^{8}z\,L(D_{\alpha}J) ∫ d 8 z ( D α L ) J = + ∫ d 8 z L ( D α J ) , and similarly for D ˉ \bar{D} D ˉ . With this ordering and no surface terms,
δ H S H Ψ = μ ∫ d 8 z ( L α D ˉ α ˙ J α α ˙ − L ˉ α ˙ D α J α α ˙ ) . (214) \delta_{H}S_{H\Psi}=\mu\int d^{8}z\left(L^{\alpha}\bar{D}_{\dot{\alpha}}J^{\alpha\dot{\alpha}}-\bar{L}_{\dot{\alpha}}D_{\alpha}J^{\alpha\dot{\alpha}}\right). \tag*{(214)} δ H S H Ψ = μ ∫ d 8 z ( L α D ˉ α ˙ J α α ˙ − L ˉ α ˙ D α J α α ˙ ) . (214)
For arbitrary L α L_{\alpha} L α and L ˉ α ˙ \bar{L}_{\dot{\alpha}} L ˉ α ˙ , we would need
D ˉ α ˙ J α α ˙ = 0 , D α J α α ˙ = 0. (215) \bar{D}^{\dot{\alpha}}J_{\alpha\dot{\alpha}}=0,\qquad D^{\alpha}J_{\alpha\dot{\alpha}}=0. \tag*{(215)} D ˉ α ˙ J α α ˙ = 0 , D α J α α ˙ = 0. (215)
The current above satisfies neither condition for an unconstrained Ψ \Psi Ψ . It also changes under the companion gauge transformation: δ Ω J α α ˙ = D ˉ α ˙ D α Ω − D α D ˉ α ˙ Ω ˉ \delta_{\Omega}J_{\alpha\dot{\alpha}}=\bar{D}_{\dot{\alpha}}D_{\alpha}\Omega-D_{\alpha}\bar{D}_{\dot{\alpha}}\bar{\Omega} δ Ω J α α ˙ = D ˉ α ˙ D α Ω − D α D ˉ α ˙ Ω ˉ , which is generally non-zero. The unprojected coupling is therefore not a gauge-invariant mass term.
One may instead try the transverse superspin-3 2 \frac{3}{2} 2 3 projector Π 3 / 2 T \Pi^{\mathrm{T}}_{3/2} Π 3/2 T . On unconstrained prepotentials it is non-local, with inverse powers of □ \square □ (up to □ − 2 \square^{-2} □ − 2 ), and would give
S m i x T = μ ∫ d 8 z H α α ˙ ( Π 3 / 2 T J ) α α ˙ . (216) S^{\mathrm{T}}_{\mathrm{mix}}=\mu\int d^{8}z\,H^{\alpha\dot{\alpha}}\left(\Pi^{\mathrm{T}}_{3/2}J\right)_{\alpha\dot{\alpha}}. \tag*{(216)} S mix T = μ ∫ d 8 z H α α ˙ ( Π 3/2 T J ) α α ˙ . (216)
But the current is already a gauge variation of H H H :
J α α ˙ ( Ψ ) = δ H α α ˙ ∣ L = Ψ . (217) J_{\alpha\dot{\alpha}}(\Psi)=\left.\delta H_{\alpha\dot{\alpha}}\right|_{L=\Psi}. \tag*{(217)} J α α ˙ ( Ψ ) = δ H α α ˙ ∣ L = Ψ . (217)
The projector removes gauge directions. In particular,
Π 3 / 2 T ( D ˉ α ˙ L α − D α L ˉ α ˙ ) = 0. (218) \Pi^{\mathrm{T}}_{3/2}\left(\bar{D}_{\dot{\alpha}}L_{\alpha}-D_{\alpha}\bar{L}_{\dot{\alpha}}\right)=0. \tag*{(218)} Π 3/2 T ( D ˉ α ˙ L α − D α L ˉ α ˙ ) = 0. (218)
Consequently,
Π 3 / 2 T J ( Ψ ) = 0 , S m i x T = 0. (219) \Pi^{\mathrm{T}}_{3/2}J(\Psi)=0,\qquad S^{\mathrm{T}}_{\mathrm{mix}}=0. \tag*{(219)} Π 3/2 T J ( Ψ ) = 0 , S mix T = 0. (219)
This is an algebraic identity, without a field equation or gauge fixing.
The cancellation can also be checked without knowing the projector. The two chiral curvatures used below are
W α β γ ( V ) = − 1 4 D ˉ 2 D ( α ∂ β δ ˙ V γ ) δ ˙ , F α β ( Ψ ) = − 1 4 D ˉ 2 D ( α Ψ β ) . (220) \begin{aligned}
W_{\alpha\beta\gamma}(V)&=-\frac{1}{4}\bar{D}^{2}D_{(\alpha}\partial_{\beta}^{\ \dot{\delta}}V_{\gamma)\dot{\delta}},\\
F_{\alpha\beta}(\Psi)&=-\frac{1}{4}\bar{D}^{2}D_{(\alpha}\Psi_{\beta)}. \tag*{(220)}
\end{aligned} W α β γ ( V ) F α β ( Ψ ) = − 4 1 D ˉ 2 D ( α ∂ β δ ˙ V γ ) δ ˙ , = − 4 1 D ˉ 2 D ( α Ψ β ) . (220)
Apply the first one to V = J ( Ψ ) V=J(\Psi) V = J ( Ψ ) , using { D α , D ˉ α ˙ } = − 2 i ∂ α α ˙ \{D_{\alpha},\bar{D}_{\dot{\alpha}}\}=-2i\partial_{\alpha\dot{\alpha}} { D α , D ˉ α ˙ } = − 2 i ∂ α α ˙ . The two pieces give
D ˉ 2 D ( α ∂ β ) δ ˙ D ˉ δ ˙ Ψ γ = − 2 i ∂ ( β δ ˙ ∂ α ) δ ˙ D ˉ 2 Ψ γ = 0 , (221) \bar{D}^{2}D_{(\alpha}\partial_{\beta)}^{\ \dot{\delta}}\bar{D}_{\dot{\delta}}\Psi_{\gamma}=-2i\,\partial_{(\beta}^{\ \dot{\delta}}\partial_{\alpha)\dot{\delta}}\bar{D}^{2}\Psi_{\gamma}=0, \tag*{(221)} D ˉ 2 D ( α ∂ β ) δ ˙ D ˉ δ ˙ Ψ γ = − 2 i ∂ ( β δ ˙ ∂ α ) δ ˙ D ˉ 2 Ψ γ = 0 , (221)
D ˉ 2 D ( α ∂ β ) δ ˙ D γ Ψ ˉ δ ˙ = 0. (222) \bar{D}^{2}D_{(\alpha}\partial_{\beta)}^{\ \dot{\delta}}D_{\gamma}\bar{\Psi}_{\dot{\delta}}=0. \tag*{(222)} D ˉ 2 D ( α ∂ β ) δ ˙ D γ Ψ ˉ δ ˙ = 0. (222)
The first zero follows because ∂ β δ ˙ ∂ α δ ˙ ∝ ε α β □ \partial_{\beta\dot{\delta}}\partial_{\alpha}^{\dot{\delta}} \propto\varepsilon_{\alpha\beta}\Box ∂ β δ ˙ ∂ α δ ˙ ∝ ε α β □ is antisymmetric in α \alpha α , β \beta β . The second follows from D ( α D γ ) = 0 D_{(\alpha}D_{\gamma)}=0 D ( α D γ ) = 0 . We have used D ˉ 2 D ˉ δ ˙ = 0 \bar{D}^{2}\bar{D}_{\dot{\delta}}=0 D ˉ 2 D ˉ δ ˙ = 0 in the first line. Equations (221) and (222) therefore give W α β γ ( J ) = 0 \mathcal{W}_{\alpha\beta\gamma}(J)=0 W α β γ ( J ) = 0 .
The same argument covers the real current with an arbitrary complex phase,
J α α ˙ ( a ) = a D ˉ α ˙ Ψ α − a ˉ D α Ψ ˉ α ˙ = δ H α α ˙ ∣ L = a Ψ , a ∈ C . (223) J_{\alpha\dot{\alpha}}^{(a)}=a\bar{D}_{\dot{\alpha}}\Psi_{\alpha}-\bar{a}D_{\alpha}\bar{\Psi}_{\dot{\alpha}}=\delta H_{\alpha\dot{\alpha}}\big|_{L=a\Psi},\qquad a\in\mathbb{C}. \tag*{(223)} J α α ˙ ( a ) = a D ˉ α ˙ Ψ α − a ˉ D α Ψ ˉ α ˙ = δ H α α ˙ L = a Ψ , a ∈ C . (223)
Every gauge-invariant observable of the gravitational prepotential that vanishes at H = 0 H=0 H = 0 vanishes on H = J ( a ) H=J^{(a)} H = J ( a ) as well.
The conclusion concerns the specific current J ( Ψ ) J(\Psi) J ( Ψ ) . It is not a no-go theorem for massive gravitini: the local five-dimensional action already supplies such a mass, and a different four-dimensional field content may include additional Stückelberg or compensator superfields.
The curvatures W ( H ) \mathcal{W}(H) W ( H ) and F ( Ψ ) \mathcal{F}(\Psi) F ( Ψ ) in Eq. (220) are both chiral and invariant under the transformations in Eq. (212) . The lowest component of W ( H ) \mathcal{W}(H) W ( H ) contains the first gravitino curvature, and a spinor derivative of F ( Ψ ) \mathcal{F}(\Psi) F ( Ψ ) contains the second. Their coupling need not vanish.
The curvature operator and its component mass
Take a real Abelian prepotential V ′ V' V ′ and its chiral field strength
W α ′ = − 1 4 D ˉ 2 D α V ′ . \mathcal{W}'_{\alpha}=-\frac{1}{4}\bar{D}^{2}D_{\alpha}V'. W α ′ = − 4 1 D ˉ 2 D α V ′ .
Its homogeneous auxiliary background is
⟨ W ′ α ⟩ = θ α D 0 , ⟨ F μ ν ′ ⟩ = 0. \langle\mathcal{W}'^{\alpha}\rangle=\theta^{\alpha}D_{0},\qquad\langle F'_{\mu\nu}\rangle=0. ⟨ W ′ α ⟩ = θ α D 0 , ⟨ F μν ′ ⟩ = 0.
The auxiliary value D 0 D_{0} D 0 supplies one spinor derivative in the chiral integral and allows a fermionic curvature to couple to the spinor descendant of the other. With a scalar momentum kernel K \mathcal{K} K we may write
S D ′ [ K ] = i c V 2 Λ m e d ∫ d 4 x d 2 θ W ′ α W α β γ ( H ) K ( Q ) F β γ ( Ψ ) + h . c . (224) S_{D'}[\mathcal{K}]=\frac{i c_{V}}{2\Lambda_{\mathrm{med}}}\int d^{4}x\,d^{2}\theta\,\mathcal{W}'^{\alpha}\mathcal{W}_{\alpha\beta\gamma}(H)\mathcal{K}(Q)\mathcal{F}^{\beta\gamma}(\Psi)+\mathrm{h.c.} \tag*{(224)} S D ′ [ K ] = 2 Λ med i c V ∫ d 4 x d 2 θ W ′ α W α β γ ( H ) K ( Q ) F β γ ( Ψ ) + h.c. (224)
The coefficient c V c_{V} c V is dimensionless, while Λ m e d \Lambda_{\mathrm{med}} Λ med is the suppression scale, and Q Q Q is the spectral operator with Euclidean eigenvalue q = p E 2 q=p_{E}^{2} q = p E 2 . Throughout this question we use η μ ν = diag ( − , + , + , + ) \eta_{\mu\nu}=\operatorname{diag}(-,+,+,+) η μν = diag ( − , + , + , + ) and the Wess–Bagger two-component spinor conventions. Thus, for a Fourier mode e i p M ⋅ x e^{ip_{M}\cdot x} e i p M ⋅ x , □ ↦ − p M 2 \Box\mapsto-p_{M}^{2} □ ↦ − p M 2 ; we take Q = − □ Q=-\Box Q = − □ and write its Euclidean eigenvalue as q = p E 2 q=p_{E}^{2} q = p E 2 . The continuation back to the mostly-plus Minkowski metric is q = p M 2 q=p_{M}^{2} q = p M 2 , so physical mass poles lie at q = − M 2 q=-M^{2} q = − M 2 . The kernel K = 1 / Q \mathcal{K}=1/Q K = 1/ Q has mass dimension − 2 -2 − 2 . With R ( H ) = 0 R(H)=0 R ( H ) = 0 , R ( Ψ ) = − 1 R(\Psi)=-1 R ( Ψ ) = − 1 and R ( W ′ ) = + 1 R(\mathcal{W}')=+1 R ( W ′ ) = + 1 , the chiral integrand has the required R R R -charge + 2 +2 + 2 .
Why the inverse d’Alembertian? The answer is already contained in the components of the curvatures. With the conventions of the spinor contractions, we have
W α β γ ( H ) ∣ = − C α β γ ( ψ ) , D ( α F β γ ) ( Ψ ) ∣ = i C α β γ ( ρ ) , (225) \mathcal{W}_{\alpha\beta\gamma}(H)\big|=-C_{\alpha\beta\gamma}(\psi),\qquad D_{(\alpha}\mathcal{F}_{\beta\gamma)}(\Psi)\big|=iC_{\alpha\beta\gamma}(\rho), \tag*{(225)} W α β γ ( H ) = − C α β γ ( ψ ) , D ( α F β γ ) ( Ψ ) = i C α β γ ( ρ ) , (225)
where C α β γ ( ψ ) = − 2 ( σ μ ν ) ( α β ∂ μ ψ ν γ ) C_{\alpha\beta\gamma}(\psi)=-2(\sigma^{\mu\nu})_{(\alpha\beta}\partial_{\mu}\psi_{\nu\gamma)} C α β γ ( ψ ) = − 2 ( σ μν ) ( α β ∂ μ ψ ν γ ) . Thus each curvature carries one spacetime derivative. For opposite Fourier momenta ( − p M , + p M ) (-p_{M},+p_{M}) ( − p M , + p M ) , the Wess–Bagger raising convention used here gives, on the transverse and gamma-traceless subspace,
C α β γ ( ψ ; − p M ) C α β γ ( ρ ; + p M ) = − 4 p M 2 ψ μ T ( − p M ) σ μ ν ρ ν T ( + p M ) , q = p M 2 . (226) C_{\alpha\beta\gamma}(\psi;-p_M)C^{\alpha\beta\gamma}(\rho;+p_M)=-4p_M^2\psi^\mathrm{T}_{\mu}(-p_M)\sigma^{\mu\nu}\rho^\mathrm{T}_{\nu}(+p_M),\qquad q=p_M^2. \tag*{(226)} C α β γ ( ψ ; − p M ) C α β γ ( ρ ; + p M ) = − 4 p M 2 ψ μ T ( − p M ) σ μν ρ ν T ( + p M ) , q = p M 2 . (226)
The minus sign belongs to the displayed lower–upper contraction, with Fourier momenta ( − p M , + p M ) (-p_M,+p_M) ( − p M , + p M ) kept in that order. Raising the first spinor triple instead of the second reverses the sign:
C α β γ ( ψ ; − p M ) C α β γ ( ρ ; + p M ) = + 4 p M 2 ψ μ T ( − p M ) σ μ ν ρ ν T ( + p M ) . C^{\alpha\beta\gamma}(\psi;-p_M)C_{\alpha\beta\gamma}(\rho;+p_M)=+4p_M^2\psi^\mathrm{T}_{\mu}(-p_M)\sigma^{\mu\nu}\rho^\mathrm{T}_{\nu}(+p_M). C α β γ ( ψ ; − p M ) C α β γ ( ρ ; + p M ) = + 4 p M 2 ψ μ T ( − p M ) σ μν ρ ν T ( + p M ) .
The calculation in Ref. [45] uses the opposite Lorentzian metric signature, for which q = − p M 2 q=-p_M^2 q = − p M 2 , there. With the mostly-plus metric used here, q = p M 2 q=p_M^2 q = p M 2 . In both conventions the lower–upper contraction is − 4 q ψ σ ρ -4q\,\psi\sigma\rho − 4 q ψ σ ρ and the upper–lower contraction is + 4 q ψ σ ρ +4q\,\psi\sigma\rho + 4 q ψ σ ρ . Thus the signature change does not alter the q q q -dependent component kernel. These identities hold on the transverse, gamma-traceless sector at non-zero momentum, without imposing the massive Rarita–Schwinger equation.
With K = 1 / Q \mathcal{K}=1/Q K = 1/ Q , the two derivatives cancel the factor 1 / q 1/q 1/ q . Rephase the companion field, if necessary, so that the mass bilinear has a positive coefficient. The operator then gives
L D ′ ∣ 3 / 2 = c V D 0 Λ m e d ψ μ T σ μ ν ρ ν T + h . c . (227) \mathcal{L}_{D'}\big|_{3/2}=\frac{c_VD_0}{\Lambda_{\mathrm{med}}}\psi^\mathrm{T}_{\mu}\sigma^{\mu\nu}\rho^\mathrm{T}_{\nu}+\mathrm{h.c.} \tag*{(227)} L D ′ 3/2 = Λ med c V D 0 ψ μ T σ μν ρ ν T + h.c. (227)
For two canonically normalized Weyl gravitini the mass term is − 2 m D ψ σ μ ν ρ ν + h . c . -2m_D\psi\sigma^\mu{}_\nu\rho_\nu+\mathrm{h.c.} − 2 m D ψ σ μ ν ρ ν + h.c. . Thus m D = ∣ c V D 0 ∣ / ( 2 Λ m e d ) m_D=|c_VD_0|/(2\Lambda_{\mathrm{med}}) m D = ∣ c V D 0 ∣/ ( 2 Λ med ) . For diagonal kinetic coefficients Z ψ Z_\psi Z ψ and Z ρ Z_\rho Z ρ , canonical normalization gives
m D = ∣ c V D 0 ∣ 2 Λ m e d Z ψ Z ρ . m_D=\frac{|c_VD_0|}{2\Lambda_{\mathrm{med}}\sqrt{Z_\psi Z_\rho}}. m D = 2 Λ med Z ψ Z ρ ∣ c V D 0 ∣ .
The factor 2 in the definition of m D m_D m D is essential. Later we shall meet a crossed coefficient for collective boundary fields. It is a coefficient in a non-local quadratic functional, not the pole mass of a single gravitino.
The bosonic interaction
The same superspace operator also contains a bosonic term. Define
f β γ = F β γ ( Ψ ) ∣ , t β γ = D α W α β γ ( H ) ∣ . f_{\beta\gamma}=\mathcal{F}_{\beta\gamma}(\Psi)|,\qquad t_{\beta\gamma}=D^\alpha W_{\alpha\beta\gamma}(H)|. f β γ = F β γ ( Ψ ) ∣ , t β γ = D α W α β γ ( H ) ∣.
On the homogeneous D 0 D_0 D 0 background, the component expansion gives
L D ′ ∣ b o s = − i c V D 0 4 Λ m e d t β γ K ( Q ) f β γ + h . c . (228) \mathcal{L}_{D'}\big|_{\mathrm{bos}}=-\frac{i c_VD_0}{4\Lambda_{\mathrm{med}}}t_{\beta\gamma}\mathcal{K}(Q)f^{\beta\gamma}+\mathrm{h.c.} \tag*{(228)} L D ′ bos = − 4 Λ med i c V D 0 t β γ K ( Q ) f β γ + h.c. (228)
This is a mixing between the trace descendant of the gravitational curvature and the vector or auxiliary part of the companion multiplet. In a conformal-supergravity component description, t β γ t_{\beta\gamma} t β γ contains the curl of an auxiliary U ( 1 ) R U(1)_R U ( 1 ) R connection. Its coefficient depends on the normalization of that auxiliary field; no numerical identification is needed to establish the mixing in Eq. (228) .
One can see why the gravitational Weyl tensor does not enter. Decompose
D δ W α β γ ( H ) ∣ = C δ α β γ + 3 4 ε δ ( α t β γ ) , C δ α β γ = D ( δ W α β γ ) ( H ) ∣ . (229) D_\delta W_{\alpha\beta\gamma}(H)|=C_{\delta\alpha\beta\gamma}+\frac{3}{4}\varepsilon_{\delta(\alpha}t_{\beta\gamma)},\qquad C_{\delta\alpha\beta\gamma}=D_{(\delta}W_{\alpha\beta\gamma)}(H)|. \tag*{(229)} D δ W α β γ ( H ) ∣ = C δ α β γ + 4 3 ε δ ( α t β γ ) , C δ α β γ = D ( δ W α β γ ) ( H ) ∣. (229)
The factor 3 / 4 3/4 3/4 makes the epsilon trace equal to t t t . The completely symmetric spinor C δ α β γ C_{\delta\alpha\beta\gamma} C δ α β γ contains the linearized graviton Weyl curvature. The D D D -spurion supplies one θ \theta θ , and the gravitational supercurvature supplies another. Since the two Grassmann coordinates are antisymmetric,
θ α θ δ C δ α β γ = 0. (230) \theta^\alpha\theta^\delta C_{\delta\alpha\beta\gamma}=0. \tag*{(230)} θ α θ δ C δ α β γ = 0. (230)
The trace term in Eq. (229) is the contribution to Eq. (228) .
The operator is bilinear in H H H and Ψ \Psi Ψ , so it has no h h hh hh term in the first place. The useful result here is the bosonic mixing : it involves the auxiliary U ( 1 ) R U(1)_{R} U ( 1 ) R connection, not the graviton Weyl curvature. The masses and kinetic terms of the remaining bosons require the complete quadratic action and elimination of its auxiliary fields.
The first result can be read on its own. The prepotential coupling is not gauge invariant and its transverse projection is zero, whereas the curvature operator gives a non-zero Dirac mass. The remaining calculations establish its normalization in the five-dimensional theory. They are not needed for a first reading.
The complete KK matching
The endpoint response contains the diagonal kinetic term as well as the crossed bilinear. Their common normalization is needed to compare the collective boundary variables with individual KK mass eigenstates.
We start with one canonically normalized spin-3 2 \frac{3}{2} 2 3 pair. For four-component Majorana vector-spinors χ μ A \chi_{\mu A} χ μ A , we use
L R S [ χ A ] = − 1 2 χ ˉ μ A γ μ ν ρ ∂ ν χ ρ A , L m a s s = − 1 2 M A B χ ˉ μ A γ μ ν χ ν B , γ μ ν = 1 2 [ γ μ , γ ν ] . (231) \mathcal{L}_{\mathrm{RS}}[\chi_A] = -\frac{1}{2}\bar{\chi}_{\mu A}\gamma^{\mu\nu\rho}\partial_\nu\chi_{\rho A},
\qquad
\mathcal{L}_{\mathrm{mass}} = -\frac{1}{2}M_{AB}\bar{\chi}_{\mu A}\gamma^{\mu\nu}\chi_{\nu B},
\qquad
\gamma^{\mu\nu} = \frac{1}{2}[\gamma^\mu,\gamma^\nu]. \tag*{(231)} L RS [ χ A ] = − 2 1 χ ˉ μ A γ μν ρ ∂ ν χ ρ A , L mass = − 2 1 M A B χ ˉ μ A γ μν χ ν B , γ μν = 2 1 [ γ μ , γ ν ] . (231)
The indices A A A , B = 1 , 2 B=1,2 B = 1 , 2 are summed, and γ μ ν ρ \gamma^{\mu\nu\rho} γ μν ρ denotes the normalized antisymmetrized product of three gamma matrices. We use { γ μ , γ ν } = 2 η μ ν \{\gamma^\mu,\gamma^\nu\}=2\eta^{\mu\nu} { γ μ , γ ν } = 2 η μν and η = diag ( − , + , + , + ) \eta=\operatorname{diag}(-,+,+,+) η = diag ( − , + , + , + ) . The factor 1 / 2 1/2 1/2 is the Majorana normalization. We follow the standard free-field normalization discussed by Freedman and Van Proeyen.2 For the two-component fields we use σ μ ν = ( σ μ σ ˉ ν − σ ν σ ˉ μ ) / 4 \sigma^{\mu\nu}=(\sigma^\mu\bar{\sigma}^\nu-\sigma^\nu\bar{\sigma}^\mu)/4 σ μν = ( σ μ σ ˉ ν − σ ν σ ˉ μ ) /4 with Wess–Bagger index conventions.
The real Majorana basis has the mass matrix m D 1 m_D1 m D 1 . The change of basis of Question 11 turns it into the off-diagonal matrix of Eq. (206) . Translating to Weyl spinors gives two equal crossed terms, hence the coefficient 2 m D 2m_D 2 m D in Eq. (207) .
On a transverse, gamma-traceless vector-spinor, γ μ ν ρ p ν χ ρ = p̸ χ μ \gamma^{\mu\nu\rho}p_\nu\chi_\rho=\not{p}\,\chi^\mu γ μν ρ p ν χ ρ = p χ μ and γ μ ν χ ν = − χ μ \gamma^{\mu\nu}\chi_\nu=-\chi^\mu γ μν χ ν = − χ μ . The first-order Euclidean operator for a single pair is
O 1 ( p E ) = i ̸ p E σ 2 + m σ 1 , O 1 − 1 ( p E ) = − i ̸ p E σ 2 + m σ 1 p E 2 + m 2 . (232) O_1(p_E)=i\not{p}_E\sigma_2+m\sigma_1,
\qquad
O_1^{-1}(p_E)=\frac{-i\not{p}_E\sigma_2+m\sigma_1}{p_E^2+m^2}. \tag*{(232)} O 1 ( p E ) = i p E σ 2 + m σ 1 , O 1 − 1 ( p E ) = p E 2 + m 2 − i p E σ 2 + m σ 1 . (232)
The pole is at p E 2 = − m 2 p_E^2=-m^2 p E 2 = − m 2 . The off-diagonal entry of this first-order operator is m m m , although the associated crossed Weyl bilinear has coefficient 2 m 2m 2 m . The signed eigenvalues of m σ 1 m\sigma_1 m σ 1 are ± m \pm m ± m ; for the lightest half-twisted level they agree with Eq. (200) .
The complete tower gives more information: the response of the bulk field at the two ends of the interval.
Let p = ∣ p E ∣ p = |p_E| p = ∣ p E ∣ , m n = ( n + 1 2 ) / R m_n = (n + \frac{1}{2})/R m n = ( n + 2 1 ) / R and ℓ = π R \ell= \pi R ℓ = π R . At the half-twist, two sums are
∑ n ∈ Z 1 p 2 + m n 2 = ℓ p tanh ( ℓ p ) , ∑ n ∈ Z ( − 1 ) n m n p 2 + m n 2 = ℓ sech ( ℓ p ) . (233) \sum_{n\in\mathbb{Z}} \frac{1}{p^2 + m_n^2} = \frac{\ell}{p}\tanh(\ell p), \qquad\sum_{n\in\mathbb{Z}} \frac{(-1)^n m_n}{p^2 + m_n^2} = \ell\,\operatorname{sech}(\ell p). \tag*{(233)} n ∈ Z ∑ p 2 + m n 2 1 = p ℓ tanh ( ℓ p ) , n ∈ Z ∑ p 2 + m n 2 ( − 1 ) n m n = ℓ sech ( ℓ p ) . (233)
The first sum is absolutely convergent. The second is conditionally convergent and is understood by pairing n n n with − n − 1 -n-1 − n − 1 , or by the corresponding Abel limit. The unweighted mass-odd diagonal sum ∑ n m n / ( p 2 + m n 2 ) \sum_n m_n/(p^2+m_n^2) ∑ n m n / ( p 2 + m n 2 ) vanishes under symmetric principal-value summation. The factor ( − 1 ) n (-1)^n ( − 1 ) n in the crossed sum comes from the relative endpoint wave-function signs.
For the transverse fermionic channel the unit-weight endpoint Green matrix is
G_{\partial}(p_E) = \begin{pmatrix} S & A \\ A & S \end{pmatrix}, \qquad S = -i\ell\frac{\slashed p_E}{p}\tanh(\ell p), \qquad A = \ell\operatorname{sech}(\ell p), \tag*{(234)}
where Euclidean gamma matrices obey \slashed p_E^2 = p^2 . It follows that S 2 − A 2 = − ℓ 2 S^2-A^2=-\ell^2 S 2 − A 2 = − ℓ 2 , and hence
K_{\partial}=G_{\partial}^{-1}=\frac{1}{\ell}\left[i\frac{\slashed p_E}{p}\tanh(\ell p)\,1_2+\operatorname{sech}(\ell p)\sigma_1^{(\partial)}\right], \tag*{(235)}
with σ 1 ( ∂ ) \sigma_1^{(\partial)} σ 1 ( ∂ ) interchanging the two endpoints. The crossed entry is
K 0 ℓ ( p ) = sech ( ℓ p ) ℓ . (236) K_{0\ell}(p)=\frac{\operatorname{sech}(\ell p)}{\ell}. \tag*{(236)} K 0 ℓ ( p ) = ℓ sech ( ℓ p ) . (236)
The diagonal and crossed terms must be kept together when interpreting this answer. In the first-order endpoint convention, write
K_{\partial}(p_E)=i\slashed p_E Z_{\partial}(q)\,1_2+\mu_{\partial}(q)\sigma_1^{(\partial)}, \qquad Z_{\partial}(q)=\frac{\tanh(\ell p)}{\ell p}, \qquad\mu_{\partial}(q)=\frac{\operatorname{sech}(\ell p)}{\ell}. \tag*{(237)}
In particular μ ∂ = K 0 ℓ \mu_{\partial}=K_{0\ell} μ ∂ = K 0 ℓ , while Z ∂ ( 0 ) = 1 Z_{\partial}(0)=1 Z ∂ ( 0 ) = 1 . For one canonically normalized Rarita–Schwinger pair, (232) identified the off-diagonal first-order entry as the mass parameter. Applying exactly that convention to the two entries in (237) gives the momentum-dependent ratio
m ∂ ( q ) : = μ ∂ ( q ) Z ∂ ( q ) = p sinh ( ℓ p ) . (238) m_{\partial}(q):=\frac{\mu_{\partial}(q)}{Z_{\partial}(q)}=\frac{p}{\sinh(\ell p)}. \tag*{(238)} m ∂ ( q ) := Z ∂ ( q ) μ ∂ ( q ) = sinh ( ℓ p ) p . (238)
At q = 0 q=0 q = 0 , m ∂ ( 0 ) = 1 / ℓ m_{\partial}(0)=1/\ell m ∂ ( 0 ) = 1/ ℓ . This is the leading coefficient of a derivative expansion of a collective response. It is not a pole mass. The ratio has a finite analytic continuation at a KK pole, p = i M k p=iM_k p = i M k with M k = ( k + 1 2 ) / R M_k=(k+\frac{1}{2})/R M k = ( k + 2 1 ) / R , and gives
m ∂ ( − M k 2 ) = M k 2 sin ( ℓ M k ) = ( − 1 ) k M k . (239) m_{\partial}(-M_k^2)=\frac{M_k^2}{\sin(\ell M_k)}=(-1)^kM_k. \tag*{(239)} m ∂ ( − M k 2 ) = sin ( ℓ M k ) M k 2 = ( − 1 ) k M k . (239)
The sign alternates with the KK level and may be rephased. The magnitude is the correct canonical mass. This tests the normalization against the diagonal kinetic term, rather than against the crossed coefficient alone.
The ratio in (239) has a simple interpretation: in the endpoint variables, the kinetic and mixed terms share the same poles. Their ratio agrees with the signed mass at each pole. The superfield calculation must reproduce both terms, not just the crossed one.
For ∣ ℓ p ∣ < π / 2 |\ell p| < \pi/2 ∣ ℓ p ∣ < π /2 one may expand
K 0 ℓ ( p ) = 1 ℓ [ 1 − ( ℓ p ) 2 2 + 5 ( ℓ p ) 4 24 + O ( ( ℓ p ) 6 ) ] . K_{0\ell}(p) = \frac{1}{\ell}\left[1-\frac{(\ell p)^2}{2}+\frac{5(\ell p)^4}{24}+O((\ell p)^6)\right]. K 0 ℓ ( p ) = ℓ 1 [ 1 − 2 ( ℓ p ) 2 + 24 5 ( ℓ p ) 4 + O (( ℓ p ) 6 ) ] .
After p → i M p \to iM p → i M the exact function has poles at M = ( k + 1 2 ) / R M=(k+\frac{1}{2})/R M = ( k + 2 1 ) / R for all k ≥ 0 k\geq0 k ≥ 0 , with non-zero residues. A finite rational KK sum cannot reproduce it exactly.
The two boundary values originate from one five-dimensional field. Two different linear spaces enter the calculation. The fermionic component Hessian for one generic massive KK harmonic has 76 = 32 + 8 + 32 + 4 coefficients, of which 36 are gauge directions and 40 are non-gauge coefficients. This is the space on which the component operator and its generalized inverse are constructed.
The boundary polarization uses a separate, redundant set of 80 candidate source values at the two ends of the interval. The source count is not obtained by equating these values with the 76 bulk-field coefficients. On the left, the Ward condition is L ( p ) T j 0 = 0 L(p)^T{}_{j0}=0 L ( p ) T j 0 = 0 and its kernel has dimension 12. On the right, B Ω ( p ) T j ℓ = 0 B_{\Omega}(p)^T{}_{j\ell}=0 B Ω ( p ) T j ℓ = 0 leaves 36 directions. The unequal numbers reflect the different endpoint source maps in this polarization, including the larger, initially non-chiral right curvature representative; they do not count physical gravitino states. Four directions at each end have zero pairing with all admissible sources. Consequently,
dim J a d m = 12 + 36 = 48 , dim rad G = 8 , 48 − 8 = 40 = 16 + 24. (240) \dim\mathcal{J}_{\mathrm{adm}}=12+36=48,\qquad\dim\operatorname{rad}\mathcal{G}=8,\qquad48-8=40=16+24. \tag*{(240)} dim J adm = 12 + 36 = 48 , dim rad G = 8 , 48 − 8 = 40 = 16 + 24. (240)
The 16 curvature coordinates consist of eight at each end; the remaining 24 are local-contact directions. The calculation first uses the generalized inverse of the 76-component bulk Hessian, then imposes the endpoint Ward constraints and takes the source quotient. This order is what removes the gauge-null response without discarding the contact terms.
We already met the two functions generated by the KK tower. It is convenient to name them
f ( q ) = tanh ( ℓ q ) q , h ( q ) = sech ( ℓ q ) , q f ( q ) 2 + h ( q ) 2 = 1. (241) f(q)=\frac{\tanh(\ell\sqrt{q})}{\sqrt{q}},\qquad h(q)=\operatorname{sech}(\ell\sqrt{q}),\qquad qf(q)^2+h(q)^2=1. \tag*{(241)} f ( q ) = q tanh ( ℓ q ) , h ( q ) = sech ( ℓ q ) , q f ( q ) 2 + h ( q ) 2 = 1. (241)
The last relation is simply tanh 2 x + sech 2 x = 1 \tanh^2 x+\operatorname{sech}^2 x=1 tanh 2 x + sech 2 x = 1 . The diagonal and crossed curvature pairings are matrices C ^ d ( p ) \widehat{C}_{\mathrm{d}}(p) C d ( p ) and C ^ x ( p ) \widehat{C}_{\mathrm{x}}(p) C x ( p ) , acting on the eight curvature components at an endpoint. With the source phases and right-boundary orientation fixed as in the component reduction, their two-point matrix is
C c u r v ( p ) = ( f ( q ) C ^ d ( p ) h ( q ) C ^ x ( p ) h ( q ) C ^ x ( p ) f ( q ) C ^ d ( p ) ) . (242) C_{\mathrm{curv}}(p)=
\begin{pmatrix}
f(q)\widehat{C}_{\mathrm{d}}(p) & h(q)\widehat{C}_{\mathrm{x}}(p)\\
h(q)\widehat{C}_{\mathrm{x}}(p) & f(q)\widehat{C}_{\mathrm{d}}(p)
\end{pmatrix}. \tag*{(242)} C curv ( p ) = ( f ( q ) C d ( p ) h ( q ) C x ( p ) h ( q ) C x ( p ) f ( q ) C d ( p ) ) . (242)
The contact-source contributions cancel out of these four blocks. At generic q ≠ 0 q\ne0 q = 0 the component curvature matrices satisfy
C ^ x − 1 C ^ d = − q C ^ d − 1 C ^ x . (243) \widehat{C}_{\mathrm{x}}^{-1}\widehat{C}_{\mathrm{d}}=-q\widehat{C}_{\mathrm{d}}^{-1}\widehat{C}_{\mathrm{x}}. \tag*{(243)} C x − 1 C d = − q C d − 1 C x . (243)
Combining this identity with (241) , we obtain
K c u r v ( p ) : = C c u r v ( p ) − 1 = ( q f ( q ) C ^ d − 1 h ( q ) C ^ x − 1 h ( q ) C ^ x − 1 q f ( q ) C ^ d − 1 ) . (244) \mathcal{K}_{\mathrm{curv}}(p):=C_{\mathrm{curv}}(p)^{-1}=
\begin{pmatrix}
qf(q)\widehat{C}_{\mathrm{d}}^{-1} & h(q)\widehat{C}_{\mathrm{x}}^{-1}\\
h(q)\widehat{C}_{\mathrm{x}}^{-1} & qf(q)\widehat{C}_{\mathrm{d}}^{-1}
\end{pmatrix}. \tag*{(244)} K curv ( p ) := C curv ( p ) − 1 = ( q f ( q ) C d − 1 h ( q ) C x − 1 h ( q ) C x − 1 q f ( q ) C d − 1 ) . (244)
This is an equality of 16 × 16 16 \times16 16 × 16 matrices, not a replacement of the full off-shell Hessian. The inverse is taken on the curvature source quotient; the physical poles are found by continuation of its two-point function.
The two superspace invariants
We now look for N = 1 N=1 N = 1 expressions whose fermionic components give Eq. (244) . Both the diagonal term and the crossed term must be present. The diagonal one is Hermitian and involves a curvature and its conjugate. The crossed one is holomorphic.
The left curvature is the ordinary chiral super-Weyl field. The right curvature comes from a superfield R ℓ \mathcal{R}_{\ell} R ℓ of the five-dimensional theory that is not, in general, chiral in this N = 1 N=1 N = 1 parametrization. On the fixed radion background let
X S S = θ 2 F E , Ξ = X S S F E , ⟨ Ξ ⟩ = θ 2 , F E ≠ 0. (245) X_{\mathrm{SS}}=\theta^{2}F_{E},\qquad\Xi=\frac{X_{\mathrm{SS}}}{F_{E}},\qquad\langle\Xi\rangle=\theta^{2},\qquad F_{E}\ne0. \tag*{(245)} X SS = θ 2 F E , Ξ = F E X SS , ⟨ Ξ ⟩ = θ 2 , F E = 0. (245)
For a curvature representative R ℓ ( g ) \mathcal{R}_{\ell}^{(g)} R ℓ ( g ) in the geometric normalization we define the chiral projection
W ℓ α β γ ( g ) = − 1 4 D ˉ 2 ( Ξ ˉ R ℓ α β γ ( g ) ) , D ˉ α ˙ W ℓ ( g ) = 0 , W ℓ ( g ) ∣ = R ℓ ( g ) ∣ . (246) \mathcal{W}_{\ell\alpha\beta\gamma}^{(g)}=-\frac{1}{4}\bar{D}^{2}\left(\bar{\Xi}\mathcal{R}_{\ell\alpha\beta\gamma}^{(g)}\right),\qquad\bar{D}_{\dot{\alpha}}\mathcal{W}_{\ell}^{(g)}=0,\qquad\left.\mathcal{W}_{\ell}^{(g)}\right|=\left.\mathcal{R}_{\ell}^{(g)}\right|. \tag*{(246)} W ℓ α β γ ( g ) = − 4 1 D ˉ 2 ( Ξ ˉ R ℓ α β γ ( g ) ) , D ˉ α ˙ W ℓ ( g ) = 0 , W ℓ ( g ) = R ℓ ( g ) . (246)
The final equality uses − D ˉ 2 θ ˉ 2 / 4 = 1 -\bar{D}^{2}\bar{\theta}^{2}/4=1 − D ˉ 2 θ ˉ 2 /4 = 1 and holds on this fixed background. It does not turn the original right superfield into an unconstrained second chiral curvature.
We distinguish the geometric curvature variables from the variables used to couple boundary sources. Let
R 0 ( s ) = W 0 ( s ) = c 0 W 0 ( g ) , \mathcal{R}_{0}^{(s)}=\mathcal{W}_{0}^{(s)}=c_{0}\mathcal{W}_{0}^{(g)}, R 0 ( s ) = W 0 ( s ) = c 0 W 0 ( g ) ,
R ℓ ( s ) = − c ℓ R ℓ ( g ) , W ‾ ℓ ( s ) = − 1 4 D ˉ 2 ( Ξ ˉ R ℓ ( s ) ) = − c ℓ W ℓ ( g ) . (247) \mathcal{R}_{\ell}^{(s)}=-c_{\ell}\mathcal{R}_{\ell}^{(g)},\qquad\overline{\mathcal{W}}_{\ell}^{(s)}=-\frac{1}{4}\bar{D}^{2}\left(\bar{\Xi}\mathcal{R}_{\ell}^{(s)}\right)=-c_{\ell}\mathcal{W}_{\ell}^{(g)}. \tag*{(247)} R ℓ ( s ) = − c ℓ R ℓ ( g ) , W ℓ ( s ) = − 4 1 D ˉ 2 ( Ξ ˉ R ℓ ( s ) ) = − c ℓ W ℓ ( g ) . (247)
The superscripts ( g ) (g) ( g ) and ( s ) (s) ( s ) mean geometric and source normalization. In particular, the unnormalized chiral projection W ℓ ( g ) \mathcal{W}_{\ell}^{(g)} W ℓ ( g ) and the normalized W ‾ ℓ ( s ) \overline{\mathcal{W}}_{\ell}^{(s)} W ℓ ( s ) are different objects. The real minus sign is the chosen right-end source orientation: the odd sine profiles give Ψ μ 2 ( x , ℓ ) = − 2 / ℓ ∑ k ( − 1 ) k ρ μ k ( x ) \Psi_{\mu}^{2}(x,\ell)=-\sqrt{2/\ell}\sum_{k}(-1)^{k}\rho_{\mu k}(x) Ψ μ 2 ( x , ℓ ) = − 2/ ℓ ∑ k ( − 1 ) k ρ μ k ( x ) . The component-to-superfield phase is fixed separately by their kinetic and fifth-derivative terms.
Write d 8 z = d 4 x d 2 θ d 2 θ ˉ d^{8}z=d^{4}x\,d^{2}\theta\,d^{2}\bar{\theta} d 8 z = d 4 x d 2 θ d 2 θ ˉ and d 6 z = d 4 x d 2 θ d^{6}z=d^{4}x\,d^{2}\theta d 6 z = d 4 x d 2 θ . The two invariants are
S A m a t c h = 2 i ∑ b ∈ { 0 , ℓ } ∫ d 8 z Ξ Ξ ˉ R b α β γ ( s ) f ( q ) q 2 ∂ α α ˙ ∂ β β ˙ ∂ γ γ ˙ R ‾ b α ˙ β ˙ γ ˙ ( s ) . (248) S_{A}^{\mathrm{match}}=2i\sum_{b\in\{0,\ell\}}\int d^{8}z\,\Xi\bar{\Xi}\mathcal{R}_{b\alpha\beta\gamma}^{(s)}\frac{f(q)}{q^{2}}\partial^{\alpha\dot{\alpha}}\partial^{\beta\dot{\beta}}\partial^{\gamma\dot{\gamma}}\overline{\mathcal{R}}_{b\dot{\alpha}\dot{\beta}\dot{\gamma}}^{(s)}. \tag*{(248)} S A match = 2 i b ∈ { 0 , ℓ } ∑ ∫ d 8 z Ξ Ξ ˉ R b α β γ ( s ) q 2 f ( q ) ∂ α α ˙ ∂ β β ˙ ∂ γ γ ˙ R b α ˙ β ˙ γ ˙ ( s ) . (248)
S B m a t c h = 2 ∫ d 6 z Ξ W 0 α β γ ( s ) h ( q ) q W ‾ ℓ ( s ) α β γ + h . c . (249) S_{B}^{\mathrm{match}}=2\int d^{6}z\,\Xi\mathcal{W}_{0\alpha\beta\gamma}^{(s)}\frac{h(q)}{q}\overline{\mathcal{W}}_{\ell}^{(s)\alpha\beta\gamma}+\mathrm{h.c.} \tag*{(249)} S B match = 2 ∫ d 6 z Ξ W 0 α β γ ( s ) q h ( q ) W ℓ ( s ) α β γ + h.c. (249)
In S A S_{A} S A the three derivatives act on the conjugate curvature; this is the Hermitian same-boundary kinetic term. The spurion Ξ Ξ ˉ \Xi\bar{\Xi} Ξ Ξ ˉ selects its lowest curvature components. The term S B S_{B} S B connects the two ends. Its spinor triples have the indicated lowered and raised indices, in that order. The two normalizations are fixed by the component calculation, not by choosing an arbitrary coefficient for the mass term.
The spinor algebra behind this matching is worth recording. For symmetric triples ordered as (111, 112, 122, 222), the contraction of three antisymmetric epsilon tensors gives
E 3 = ( 0 0 0 1 0 0 − 3 0 0 3 0 0 − 1 0 0 0 ) . (250) E_{3} =
\begin{pmatrix}
0 & 0 & 0 & 1 \\
0 & 0 & -3 & 0 \\
0 & 3 & 0 & 0 \\
-1 & 0 & 0 & 0
\end{pmatrix}. \tag*{(250)} E 3 = 0 0 0 − 1 0 0 3 0 0 − 3 0 0 1 0 0 0 . (250)
The factors of 3 count the ordered occurrences of the symmetric mixed-index components. Replacing the epsilon contractions by the appropriate momentum contractions produces a cubic matrix P 3 ( p ) P_{3}(p) P 3 ( p ) for the diagonal term. After retaining the opposite Fourier momenta, the Grassmann signs, and the factors 2 i 2i 2 i and 2 in Eqs. (248) and (249) , their fermionic numerators are
D A ( p ) = q 3 diag ( C ^ d − 1 , C ^ d − 1 ) , D_{A}(p) = q^{3}\operatorname{diag}(\widehat{C}_{d}^{-1},\widehat{C}_{d}^{-1}), D A ( p ) = q 3 diag ( C d − 1 , C d − 1 ) ,
D B ( p ) = q ( 0 C ^ x − 1 C ^ x − 1 0 ) . (251) D_{B}(p) = q
\begin{pmatrix}
0 & \widehat{C}_{x}^{-1} \\
\widehat{C}_{x}^{-1} & 0
\end{pmatrix}. \tag*{(251)} D B ( p ) = q ( 0 C x − 1 C x − 1 0 ) . (251)
These are the polynomial numerators obtained from the graded spinor contractions. In the diagonal term, the three displayed spacetime derivatives give the symmetric-cube momentum map P 3 ( p ) P_{3}(p) P 3 ( p ) , which is cubic in p α α ˙ p_{\alpha\dot{\alpha}} p α α ˙ . In the curvature basis C ^ d \widehat{C}_{d} C d is itself cubic in momentum. Thus C ^ d − 1 \widehat{C}_{d}^{-1} C d − 1 has degree − 3 -3 − 3 and q 3 C ^ d − 1 q^{3}\widehat{C}_{d}^{-1} q 3 C d − 1 has degree + 3 +3 + 3 , exactly that of P 3 P_{3} P 3 . The same source-matrix conventions give C ^ x ∼ q \widehat{C}_{x}\sim q C x ∼ q and hence q C ^ x − 1 q\widehat{C}_{x}^{-1} q C x − 1 of degree zero for the crossed contraction. The component reduction fixes the coefficients as well as these momentum degrees. The underlying matrices and Grassmann contractions are given in the subsection Component reduction of the superspace invariants in the appendix Endpoint Polarization and Fermionic Superspace Matching of Ref. [45]. Multiplying by the kernels in the two invariants gives
K A + B F ( p ) = f ( q ) q 2 D A ( p ) + h ( q ) q D B ( p ) = ( q f ( q ) C ^ d − 1 h ( q ) C ^ x − 1 h ( q ) C ^ x − 1 q f ( q ) C ^ d − 1 ) = K c u r v ( p ) . (252) \begin{aligned}
\mathcal{K}^{F}_{A+B}(p) = \frac{f(q)}{q^{2}}D_{A}(p) + \frac{h(q)}{q}D_{B}(p) \\
&= \begin{pmatrix}
qf(q)\widehat{C}_{d}^{-1} & h(q)\widehat{C}_{x}^{-1} \\
h(q)\widehat{C}_{x}^{-1} & qf(q)\widehat{C}_{d}^{-1}
\end{pmatrix}
= \mathcal{K}_{\mathrm{curv}}(p). \tag*{(252)}
\end{aligned} K A + B F ( p ) = q 2 f ( q ) D A ( p ) + q h ( q ) D B ( p ) = ( q f ( q ) C d − 1 h ( q ) C x − 1 h ( q ) C x − 1 q f ( q ) C d − 1 ) = K curv ( p ) . (252)
This is the fermionic 16 × 16 16 \times16 16 × 16 matching obtained from the local five-dimensional component system. The graded contractions leading to Eq. (251) are lengthy; Eq. (252) displays their result and its comparison with the independently calculated curvature kernel. Changing 2 i 2i 2 i in S A S_{A} S A to 4 i 4i 4 i , for example, spoils the diagonal blocks, and reversing the sign of S B S_{B} S B spoils the crossed blocks. The relative coefficient and sign are therefore fixed within this two-invariant ansatz.
From curvature sources to canonical gravitini
The endpoint matrix in Eq. (235) was written with unit weights for the signed KK modes. The curvature-source action uses c 0 c_{0} c 0 and c ℓ c_{\ell} c ℓ . We now compare them with the canonical Rarita–Schwinger normalization.
Let a 0 ℓ ( q ) a_{0\ell}(q) a 0 ℓ ( q ) be the coefficient of the crossed Weyl bilinear, and Z b ( q ) Z_{b}(q) Z b ( q ) the coefficient of the diagonal kinetic action for b ∈ { 0 , ℓ } b \in\{0,\ell\} b ∈ { 0 , ℓ } , in the same quadratic functional. Eq. (207) then gives
m c u r v ( q ) = ∣ a 0 ℓ ( q ) ∣ 2 Z 0 ( q ) Z ℓ ( q ) . (253) m_{\mathrm{curv}}(q) = \frac{\lvert a_{0\ell}(q)\rvert}{2\sqrt{Z_{0}(q)Z_{\ell}(q)}}. \tag*{(253)} m curv ( q ) = 2 Z 0 ( q ) Z ℓ ( q ) ∣ a 0 ℓ ( q )∣ . (253)
For real Euclidean q ≥ 0 q \ge0 q ≥ 0 , and in the source conventions of Eqs. (248) and (249) , the component reductions in Eq. (252) give the transverse quadratic action in the form
L ∂ , T T ( 2 ) = ∑ b = 0 , ℓ Z b ( q ) L R S ( 0 ) [ ψ b ] + [ a 0 ℓ ( q ) ψ 0 μ σ μ ν ψ ℓ ν + h . c . ] , \mathcal{L}^{(2)}_{\partial,\mathrm{TT}}=\sum_{b=0,\ell}Z_b(q)\mathcal{L}_{\mathrm{RS}}^{(0)}[\psi_b]+\left[a_{0\ell}(q)\psi_0^\mu\sigma_{\mu\nu}\psi_\ell^\nu+\mathrm{h.c.}\right], L ∂ , TT ( 2 ) = b = 0 , ℓ ∑ Z b ( q ) L RS ( 0 ) [ ψ b ] + [ a 0 ℓ ( q ) ψ 0 μ σ μν ψ ℓ ν + h.c. ] ,
where L R S ( 0 ) \mathcal{L}_{\mathrm{RS}}^{(0)} L RS ( 0 ) denotes the unit-coefficient kinetic term. The computed coefficients are
Z b ( q ) = 4 ∣ c b ∣ 2 f ( q ) , b ∈ { 0 , ℓ } , ∣ a 0 ℓ ( q ) ∣ = 8 ∣ c 0 c ℓ ∣ h ( q ) , m c u r v ( q ) = h ( q ) f ( q ) = m ∂ ( q ) . (254) Z_b(q)=4|c_b|^2f(q),\qquad b\in\{0,\ell\},\qquad|a_{0\ell}(q)|=8|c_0c_\ell|h(q),\qquad m_{\mathrm{curv}}(q)=\frac{h(q)}{f(q)}=m_{\partial}(q). \tag*{(254)} Z b ( q ) = 4∣ c b ∣ 2 f ( q ) , b ∈ { 0 , ℓ } , ∣ a 0 ℓ ( q ) ∣ = 8∣ c 0 c ℓ ∣ h ( q ) , m curv ( q ) = f ( q ) h ( q ) = m ∂ ( q ) . (254)
The factor 4 comes from the curvature-to-Rarita–Schwinger spinor contraction in the chosen normalization; the crossed Weyl bilinear includes both symmetric off-diagonal entries, giving the factor 8. Both coefficients use the same curvature-source maps. The absolute values in Eq. (254) refer to the Euclidean normalization; the alternating signs at the KK poles follow from analytic continuation of the signed ratio in Eq. (238) . They are not obtained by declaring the crossed entry of the Green-function inverse to be a physical mass. The matrix calculation establishes the two fermionic blocks; the conversion to the Weyl fields uses the same source maps and the canonical Rarita–Schwinger normalization. This is a fermionic result on the specified radion background, rather than an off-shell reconstruction of the full multiplets.
For the geometric operator below the source normalization has
∣ c 0 c ℓ ∣ = F E 8 π = 1 8 ℓ , F E = 1 R . (255) |c_0c_\ell|=\frac{F_E}{8\pi}=\frac{1}{8\ell},\qquad F_E=\frac{1}{R}. \tag*{(255)} ∣ c 0 c ℓ ∣ = 8 π F E = 8 ℓ 1 , F E = R 1 . (255)
At q = 0 q=0 q = 0 , since f ( 0 ) = ℓ f(0)=\ell f ( 0 ) = ℓ and h ( 0 ) = 1 h(0)=1 h ( 0 ) = 1 , Eq. (254) gives
∣ a 0 ℓ ( 0 ) ∣ = 1 ℓ , Z 0 ( 0 ) Z ℓ ( 0 ) = 1 2 , ∣ a 0 ℓ c a n ( 0 ) ∣ = 2 ℓ . (256) |a_{0\ell}(0)|=\frac{1}{\ell},\qquad\sqrt{Z_0(0)Z_\ell(0)}=\frac{1}{2},\qquad|a_{0\ell}^{\mathrm{can}}(0)|=\frac{2}{\ell}. \tag*{(256)} ∣ a 0 ℓ ( 0 ) ∣ = ℓ 1 , Z 0 ( 0 ) Z ℓ ( 0 ) = 2 1 , ∣ a 0 ℓ can ( 0 ) ∣ = ℓ 2 . (256)
The geometric mean depends only on ∣ c 0 c ℓ ∣ |c_0c_\ell| ∣ c 0 c ℓ ∣ . The individual equalities Z 0 ( 0 ) = Z ℓ ( 0 ) = 1 / 2 Z_0(0)=Z_\ell(0)=1/2 Z 0 ( 0 ) = Z ℓ ( 0 ) = 1/2 would also require ∣ c 0 ∣ = ∣ c ℓ ∣ |c_0|=|c_\ell| ∣ c 0 ∣ = ∣ c ℓ ∣ . The coefficient 2 / ℓ 2/\ell 2/ ℓ is the crossed Weyl coefficient in canonical endpoint kinetic variables at zero momentum. Its corresponding first-order entry is 1 / ℓ 1/\ell 1/ ℓ . Neither is the mass of one isolated mode. The quantities
K 0 ℓ ( 0 ) = 1 π R , m l i g h t = 1 2 R K_{0\ell}(0)=\frac{1}{\pi R},\qquad m_{\mathrm{light}}=\frac{1}{2R} K 0 ℓ ( 0 ) = π R 1 , m light = 2 R 1
refer, respectively, to the zero-momentum endpoint response and to the first KK pole. The relation among them is checked independently by Eq. (239) .
The Scherk–Schwarz curvature operator
The geometric operator is more compact than the pair of matched invariants. The normalized background is X S S = θ 2 F E X_{\mathrm{SS}}=\theta^2F_E X SS = θ 2 F E with F E = 1 / R F_E=1/R F E = 1/ R at the half-twist. This normalization is fixed by the component coupling to the five-dimensional fifth derivative; it is not a new mass parameter. With the geometric curvatures defined above, the transverse crossed term is represented by
S S S ( 3 / 2 ) = − 1 4 π ∫ d 4 x d 2 θ X S S W 0 ( g ) α β γ sech ( ℓ Q ) Q W ℓ α β γ ( g ) + h . c . (257) S_{\mathrm{SS}}^{(3/2)}=-\frac{1}{4\pi}\int d^4x\,d^2\theta\,X_{\mathrm{SS}}\,W_0^{(g)\alpha\beta\gamma}\frac{\operatorname{sech}(\ell\sqrt{Q})}{Q}W_{\ell\alpha\beta\gamma}^{(g)}+\mathrm{h.c.} \tag*{(257)} S SS ( 3/2 ) = − 4 π 1 ∫ d 4 x d 2 θ X SS W 0 ( g ) α β γ Q sech ( ℓ Q ) W ℓ α β γ ( g ) + h.c. (257)
The 1 / Q 1/Q 1/ Q cancels the two derivatives in the transverse gravitino curvatures, while sech is fixed by the entire KK tower. The sign depends on the sine-profile orientation and on the position of the three raised spinor indices. In the conventions used here, the latter gives the exact identity
W 0 α β γ ( g ) W ℓ ( g ) α β γ = ( − 1 ) 3 W 0 ( g ) α β γ W ℓ α β γ ( g ) = − W 0 ( g ) α β γ W ℓ α β γ ( g ) . (258) W^{(g)}_{0\alpha\beta\gamma}W^{(g)\alpha\beta\gamma}_{\ell}=(-1)^3W^{(g)\alpha\beta\gamma}_{0}W^{(g)}_{\ell\alpha\beta\gamma}=-W^{(g)\alpha\beta\gamma}_{0}W^{(g)}_{\ell\alpha\beta\gamma}. \tag*{(258)} W 0 α β γ ( g ) W ℓ ( g ) α β γ = ( − 1 ) 3 W 0 ( g ) α β γ W ℓ α β γ ( g ) = − W 0 ( g ) α β γ W ℓ α β γ ( g ) . (258)
Consequently, the source coefficient − 2 c 0 c ℓ = + F E / ( 4 π ) -2c_0c_\ell=+F_E/(4\pi) − 2 c 0 c ℓ = + F E / ( 4 π ) in the first contraction becomes − F E / ( 4 π ) -F_E/(4\pi) − F E / ( 4 π ) in the geometric contraction of Eq. (257) . The matching condition is
c 0 c ℓ = − F E 8 π . (259) c_0c_\ell=-\frac{F_E}{8\pi}. \tag*{(259)} c 0 c ℓ = − 8 π F E . (259)
The sign of c 0 c ℓ c_0c_\ell c 0 c ℓ follows from the source rescaling and the odd number of raised spinor indices in Eq. (258) ; it is independent of the choice of Lorentzian signature. To check the component sign in our mostly-plus convention, the upper–lower contraction of Eq. (226) is + 4 q +4q + 4 q . The factor 1 / q 1/q 1/ q in Eq. (257) therefore gives a crossed Weyl coefficient − F E h ( q ) / π -F_Eh(q)/\pi − F E h ( q ) / π for the geometric right-end field ψ ℓ , c u r v \psi_{\ell,\mathrm{curv}} ψ ℓ , curv . The endpoint-source phase is ψ ℓ , b d r y = − ψ ℓ , c u r v \psi_{\ell,\mathrm{bdry}}=-\psi_{\ell,\mathrm{curv}} ψ ℓ , bdry = − ψ ℓ , curv . In that basis the coefficient is + F E h ( q ) / π = h ( q ) / ℓ +F_Eh(q)/\pi=h(q)/\ell + F E h ( q ) / π = h ( q ) / ℓ , as required by Eq. (236) . A rephasing changes the sign of an off-diagonal source correlator, but not the pole positions or the absolute values of its residues.
The transverse component contraction of Eq. (257) gives
∣ a 0 ℓ ( q ) ∣ = F E π sech ( ℓ q ) = h ( q ) ℓ = K 0 ℓ ( q ) , F E = 1 R . (260) \lvert a_{0\ell}(q)\rvert=\frac{F_E}{\pi}\operatorname{sech}(\ell\sqrt{q})=\frac{h(q)}{\ell}=K_{0\ell}(q),\qquad F_E=\frac{1}{R}. \tag*{(260)} ∣ a 0 ℓ ( q )∣ = π F E sech ( ℓ q ) = ℓ h ( q ) = K 0 ℓ ( q ) , F E = R 1 . (260)
It agrees with the crossed endpoint coefficient in the chosen source convention. Equation (12.37) provides the diagonal invariant needed to compare this with the canonical kinetic normalization. The full curvature matching is Eq. (252) .
There is also a check relating the vector-background operator of Eq. (224) to the geometric curvature operator. On fixed backgrounds W ′ α = θ α D 0 W^{\prime\alpha}=\theta^\alpha D_0 W ′ α = θ α D 0 and X = θ 2 F E X=\theta^2F_E X = θ 2 F E , let I D [ K ] I_D[\mathcal{K}] I D [ K ] denote the chiral integral in Eq. (224) without its prefactor. Define
I X [ K ] = ∫ d 4 x d 2 θ X W 0 ( g ) α β γ K W ℓ α β γ ( g ) , I_X[\mathcal{K}]=\int d^4x\,d^2\theta\,X\,W^{(g)\alpha\beta\gamma}_0\mathcal{K}W^{(g)}_{\ell\alpha\beta\gamma}, I X [ K ] = ∫ d 4 x d 2 θ X W 0 ( g ) α β γ K W ℓ α β γ ( g ) ,
B [ K ] = ∫ d 4 x ( D α W 0 α β γ ( g ) ) ∣ K F β γ ∣ . B[\mathcal{K}]=\int d^4x\,(D^\alpha W^{(g)}_{0\alpha\beta\gamma})\vert\mathcal{K}\mathcal{F}^{\beta\gamma}\vert. B [ K ] = ∫ d 4 x ( D α W 0 α β γ ( g ) ) ∣ K F β γ ∣.
The first operation is a graded integration by parts:
I D [ K ] = D 0 2 ∫ d 4 x [ W 0 α β γ ( g ) ∣ K D α F β γ ∣ − ( D α W 0 α β γ ( g ) ) ∣ K F β γ ∣ ] . I_D[\mathcal{K}]=\frac{D_0}{2}\int d^4x\left[W^{(g)}_{0\alpha\beta\gamma}\vert\mathcal{K}D^\alpha\mathcal{F}^{\beta\gamma}\vert-(D^\alpha W^{(g)}_{0\alpha\beta\gamma})\vert\mathcal{K}\mathcal{F}^{\beta\gamma}\vert\right]. I D [ K ] = 2 D 0 ∫ d 4 x [ W 0 α β γ ( g ) ∣ K D α F β γ ∣ − ( D α W 0 α β γ ( g ) ) ∣ K F β γ ∣ ] .
The second uses the component identification, including the right-end phase,
D ( α F β γ ) ∣ = i C α β γ ( ρ ) = − i W ℓ α β γ ( g ) ∣ . D_{(\alpha}\mathcal{F}_{\beta\gamma)}\vert=iC_{\alpha\beta\gamma}(\rho)=-iW^{(g)}_{\ell\alpha\beta\gamma}\vert. D ( α F β γ ) ∣ = i C α β γ ( ρ ) = − i W ℓ α β γ ( g ) ∣.
Only after this identification do we obtain
I D [ K ] = i D 0 2 F E I X [ K ] − D 0 2 B [ K ] . (261) I_D[\mathcal{K}]=\frac{iD_0}{2F_E}I_X[\mathcal{K}]-\frac{D_0}{2}B[\mathcal{K}]. \tag*{(261)} I D [ K ] = 2 F E i D 0 I X [ K ] − 2 D 0 B [ K ] . (261)
The remainder B B B has bosonic lowest components. Therefore the two routes yield the same quadratic fermionic bilinear when their coefficients and phases are matched. The equation is a fixed-background component identity, not an equality of unconstrained off-shell superspace functionals.
What the fermionic matching leaves undetermined
In the radion-spurion convention of this five-dimensional reduction, the fifth-derivative coupling is parametrized by F E = 2 ω / R F_E = 2\omega/R F E = 2 ω / R . At ω = 1 2 \omega= \frac{1}{2} ω = 2 1 this gives F E = 1 / R = 2 m l i g h t F_E = 1/R = 2m_{\mathrm{light}} F E = 1/ R = 2 m light . Once the radion-spurion convention is fixed, this value is set by the KK mass spectrum; the separate normalizations c 0 c_0 c 0 and c ℓ c_{\ell} c ℓ of the boundary curvature sources were fixed by the matching above. Determining the dynamical radion kinetic term requires the independent radion–compensator reduction and is not part of the transverse fermionic curvature matching.
It is possible to change the bosonic action without changing the 16 × 16 16 \times16 16 × 16 fermionic curvature kernel. For example, the bosonic supercurvature
B α β γ δ = D ( α W β γ δ ) ( H ) \mathcal{B}_{\alpha\beta\gamma\delta}=D_{(\alpha}\mathcal{W}_{\beta\gamma\delta)}(H) B α β γ δ = D ( α W β γ δ ) ( H )
allows the invariant
Δ S = λ ∫ d 8 z Ξ Ξ ˉ B α β γ δ B α β γ δ + h . c . (262) \Delta S=\lambda\int d^8z\,\Xi\bar{\Xi}\,\mathcal{B}_{\alpha\beta\gamma\delta}\mathcal{B}^{\alpha\beta\gamma\delta}+\mathrm{h.c.} \tag*{(262)} Δ S = λ ∫ d 8 z Ξ Ξ ˉ B α β γ δ B α β γ δ + h.c. (262)
Its selected lowest component is bosonic. On the fixed background, its component projection leaves Eq. (252) unchanged while changing the bosonic Hessian. Thus the fermionic matching is exact on its stated curvature sector, but it does not uniquely specify the bosonic and auxiliary completion.
The 24 contact directions in Eq. (240) provide another part of the reconstruction problem. The complete quadratic boundary functional separates into curvature and contact coordinates only after the appropriate mixed terms are included. Its local contact subtraction depends on the chosen prescription; the nonlocal curvature response does not. An off-shell completion must reproduce the full bosonic, auxiliary and contact responses, including the longitudinal constraints of the parent theory.
The spectral pairing and the U ( 1 ) R U(1)_R U ( 1 ) R protection of the Dirac mass are discussed by Benakli and Goudeau [44] in *(Pseudo-)*Dirac Gravitinos. The local five-dimensional reduction, the curvature-source matrices and the superspace matching discussed here are developed in Ref. [45], Anti-periodic Gravitino and Dirac Effective Operators (arXiv:2611.16047). The two invariants S A m a t c h S_A^{\mathrm{match}} S A match and S B m a t c h S_B^{\mathrm{match}} S B match reproduce its complete quadratic fermionic curvature kernel on the fixed radion background. Their fermionic matching does not determine all bosonic, auxiliary and contact couplings. The linearized supergravity superfield formulation and its five-dimensional extensions were developed in Refs. [46, 47]. These provide formal background for the superspace construction considered here.
Table 2.
tures.
This naturally leads to the next question:
How does Scherk–Schwarz compactification appear as a gauging in the lower-dimensional supergravity?
How does Scherk–Schwarz compactification appear as a gauging in the lower-dimensional supergravity?
Essential idea
A Scherk–Schwarz reduction does more than shift Kaluza–Klein masses. It tells a gauge symmetry inherited from the compactification how to act on the lower-dimensional fields.
For an ordinary circle reduction, local translations of the compact coordinate are higher-dimensional diffeomorphisms. In the lower-dimensional theory they are gauged by the Kaluza–Klein vector A μ A^\mu A μ . If the Scherk–Schwarz ansatz contains a generator M \mathcal{M} M of a global symmetry of the parent theory, the same Kaluza–Klein gauge transformation acts on the reduced fields through M \mathcal{M} M . This is the simplest rank-one gauging.
More generally, the embedding tensor specifies which global-symmetry generator is gauged by each vector field,
Θ : E v e c ⟶ g . \Theta: \mathcal{E}_{\mathrm{vec}} \longrightarrow\mathfrak{g}. Θ : E vec ⟶ g .
Its image is initially only a vector subspace of the global symmetry algebra g \mathfrak{g} g . A quadratic constraint requires the chosen generators to close consistently under gauge transformations, while a separate linear constraint keeps only the components of Θ \Theta Θ allowed by supersymmetry.
The constant embedding tensor is converted into the scalar-dependent quantities entering the supergravity transformations by dressing it with the scalar fields. The resulting T T T -tensor determines the fermion shifts, mass matrices and, through supersymmetry Ward identities, the scalar potential.
The same language extends beyond the elementary circle. A geometric Scherk–Schwarz reduction can be organized by a constant Lie algebra of internal frames. In double and exceptional field theory the ordinary frame is replaced by a generalized frame, and the resulting constant algebra is naturally Leibniz rather than Lie. Its coefficients encode the lower-dimensional gauging.
A consistent gauging does not explicitly destroy local supersymmetry. Supersymmetry breaking is a property of a particular bosonic background and is decided by the simultaneous vanishing of all fermionic supersymmetry variations.
The circle reduction as a rank-one gauging
The simplest circle reduction already contains the essential mechanism. The compact coordinate can be shifted locally in the non-compact spacetime, and the mixed metric component becomes the corresponding gauge field. A Scherk–Schwarz ansatz then determines how this gauge transformation acts on the other reduced fields.
Let the compact coordinate be dimensionless, θ ∼ θ + 2 π \theta\sim\theta+ 2\pi θ ∼ θ + 2 π , and let G G G be a continuous symmetry group of the higher-dimensional theory, with g = Lie ( G ) \mathfrak{g} = \operatorname{Lie}(G) g = Lie ( G ) . Choose M ∈ g \mathcal{M} \in\mathfrak{g} M ∈ g and take
Φ ( x , θ ) = e θ M ϕ ( x ) . (263) \Phi(x,\theta) = e^{\theta\mathcal{M}}\phi(x). \tag*{(263)} Φ ( x , θ ) = e θ M ϕ ( x ) . (263)
Here ϕ ( x ) \phi(x) ϕ ( x ) is independent of the compact coordinate. We retain only this field for clarity; the complete Kaluza–Klein tower is obtained by multiplying the ansatz by the periodic factors e i n θ e^{in\theta} e in θ , n ∈ Z n \in\mathbb{Z} n ∈ Z , without changing the gauging mechanism.
Going once around the circle gives
Φ ( x , θ + 2 π ) = U S S Φ ( x , θ ) , U S S = e 2 π M . (264) \Phi(x,\theta+ 2\pi) = U_{\mathrm{SS}}\Phi(x,\theta), \qquad U_{\mathrm{SS}} = e^{2\pi\mathcal{M}}. \tag*{(264)} Φ ( x , θ + 2 π ) = U SS Φ ( x , θ ) , U SS = e 2 π M . (264)
For the one-parameter convention used in the preceding questions, M = i ω R Φ \mathcal{M} = i\omega R_{\Phi} M = iω R Φ , so U S S = e 2 π i ω R Φ U_{\mathrm{SS}} = e^{2\pi i\omega R_{\Phi}} U SS = e 2 π iω R Φ .
The gauging itself comes from the gravitational sector. Ignoring the radion-dependent Weyl factors, write
d s 2 = g μ ν ( x ) d x μ d x ν + R 2 ( d θ + A μ 0 ( x ) d x μ ) 2 . (265) ds^2 = g_{\mu\nu}(x)\,dx^\mu dx^\nu+ R^2\left(d\theta+ A_\mu^0(x)dx^\mu\right)^2. \tag*{(265)} d s 2 = g μν ( x ) d x μ d x ν + R 2 ( d θ + A μ 0 ( x ) d x μ ) 2 . (265)
The mixed metric component gives the Kaluza–Klein vector A μ 0 A_\mu^0 A μ 0 . Under the higher-dimensional diffeomorphism θ → θ + λ ( x ) \theta\to\theta+ \lambda(x) θ → θ + λ ( x ) , one has d θ → d θ + d λ d\theta\to d\theta+ d\lambda d θ → d θ + d λ , so the metric keeps the form Eq. (265) provided
A μ 0 ⟶ A μ 0 − ∂ μ λ . (266) A_\mu^0 \longrightarrow A_\mu^0 - \partial_\mu\lambda. \tag*{(266)} A μ 0 ⟶ A μ 0 − ∂ μ λ . (266)
The Scherk–Schwarz ansatz now tells us how the same transformation acts on ϕ \phi ϕ . Since
e θ M ϕ ( x ) = e ( θ + λ ) M e − λ M ϕ ( x ) , e^{\theta\mathcal{M}}\phi(x) = e^{(\theta+\lambda)\mathcal{M}}e^{-\lambda\mathcal{M}}\phi(x), e θ M ϕ ( x ) = e ( θ + λ ) M e − λ M ϕ ( x ) ,
the lower-dimensional field transforms as
ϕ ( x ) ⟶ e − λ ( x ) M ϕ ( x ) . (267) \phi(x) \longrightarrow e^{-\lambda(x)\mathcal{M}}\phi(x). \tag*{(267)} ϕ ( x ) ⟶ e − λ ( x ) M ϕ ( x ) . (267)
Its covariant derivative is therefore
D μ ϕ = ( ∂ μ − A μ 0 M ) ϕ , (268) D_\mu\phi= \left(\partial_\mu- A_\mu^0\mathcal{M}\right)\phi, \tag*{(268)} D μ ϕ = ( ∂ μ − A μ 0 M ) ϕ , (268)
which indeed transforms in the same way as ϕ \phi ϕ .
If t K K t_{\mathrm{KK}} t KK denotes the Abelian generator associated with circle translations, the reduction has therefore defined the map
t K K ⟼ M ∈ g . (269) t_{\mathrm{KK}} \longmapsto\mathcal{M} \in\mathfrak{g}. \tag*{(269)} t KK ⟼ M ∈ g . (269)
Only one independent direction of g \mathfrak{g} g participates. This is the simplest rank-one Scherk–Schwarz gauging.
The embedding tensor as a linear map
The circle example suggests the general question. A lower-dimensional supergravity can contain several vector fields, and each of them may gauge a different generator, or a different linear combination of generators, of the global symmetry group. The embedding tensor is the object that records this assignment.
Let
A μ M , M = 1 , … , n v , A^M_\mu,\qquad M=1,\ldots,n_{\mathrm v}, A μ M , M = 1 , … , n v ,
be the vector gauge potentials of the ungauged theory. Depending on the compactification, they can include Kaluza–Klein vectors, vectors descending from higher-dimensional gauge fields, and vectors already belonging to lower-dimensional supergravity multiplets. The label M M M enumerates these fields. Associate with them the finite-dimensional vector space
E v e c = span { e M } , \mathcal{E}_{\mathrm{vec}}=\operatorname{span}\{e_M\}, E vec = span { e M } ,
where the abstract basis element e M e_M e M corresponds to A μ M A^M_\mu A μ M .
Let G G G now denote the global symmetry group of the ungauged lower-dimensional theory and g = Lie ( G ) \mathfrak{g}=\operatorname{Lie}(G) g = Lie ( G ) . Choose a basis t α t_\alpha t α of g \mathfrak{g} g , with
[ t α , t β ] = f α β γ t γ . (270) [t_\alpha,t_\beta]=f_{\alpha\beta}{}^\gamma t_\gamma. \tag*{(270)} [ t α , t β ] = f α β γ t γ . (270)
A gauging specifies which element of g \mathfrak{g} g is gauged by each vector A μ M A^M_\mu A μ M . This information is collected in the linear map
Θ : E v e c ⟶ g , (271) \Theta:\mathcal{E}_{\mathrm{vec}}\longrightarrow\mathfrak{g}, \tag*{(271)} Θ : E vec ⟶ g , (271)
defined by
Θ ( e M ) = X M = Θ M α t α . (272) \Theta(e_M)=X_M=\Theta_M{}^\alpha t_\alpha. \tag*{(272)} Θ ( e M ) = X M = Θ M α t α . (272)
Thus X M X_M X M is the global-symmetry generator gauged by A μ M A^M_\mu A μ M , while Θ M α \Theta_M{}^\alpha Θ M α are the components of the embedding tensor.
Before any consistency condition is imposed,
Im Θ = span { X M } ⊆ g \operatorname{Im}\Theta=\operatorname{span}\{X_M\}\subseteq\mathfrak{g} Im Θ = span { X M } ⊆ g
is only a vector subspace. There is no reason yet for it to be closed under commutators, nor for an arbitrary choice of Θ \Theta Θ to be compatible with supersymmetry.
When the gauging is consistent, the gauge-covariant derivative takes the form
D μ = ∂ μ − g g a u A μ M X M . (273) D_\mu=\partial_\mu-g_{\mathrm{gau}}A^M_\mu X_M. \tag*{(273)} D μ = ∂ μ − g gau A μ M X M . (273)
where g g a u g_{\mathrm{gau}} g gau may alternatively be absorbed into the normalization of Θ \Theta Θ . For the circle reduction above, only the Kaluza–Klein vector participates. If it is labelled by M = 0 M=0 M = 0 , one can choose the normalization so that
g g a u X 0 = M , X M ^ = 0 g_{\mathrm{gau}}X_0=\mathcal{M},\qquad X_{\widehat M}=0 g gau X 0 = M , X M = 0
for all remaining vector directions M ^ \widehat M M . Hence rank Θ = 1 \operatorname{rank}\Theta=1 rank Θ = 1 .
The quadratic constraint: closure of the gauged generators
The basic consistency question is simple. Once a set of generators has been chosen to be gauged, a gauge transformation generated by one of them must not take us outside that same set. The embedding tensor must therefore be invariant under the gauge transformations that it itself defines.
The global symmetry group can act on the vector fields. Let ( t α ) N P (t_\alpha)_{N}^{\ P} ( t α ) N P denote the matrix representing t α t_\alpha t α on E v e c \mathcal{E}_{\mathrm{vec}} E vec ,
t α : e N ⟼ ( t α ) N P e P . (274) t_\alpha: e_N \longmapsto(t_\alpha)_{N}^{\ P} e_P. \tag*{(274)} t α : e N ⟼ ( t α ) N P e P . (274)
The generator X M = Θ M α t α X_M = \Theta_M^{\ \alpha} t_\alpha X M = Θ M α t α then acts on the vector representation through
X M N P ≡ Θ M α ( t α ) N P . (275) X_{MN}^{\ \ P} \equiv\Theta_M^{\ \alpha}(t_\alpha)_{N}^{\ \ P}. \tag*{(275)} X M N P ≡ Θ M α ( t α ) N P . (275)
The coefficient X M N P X_{MN}^{\ \ P} X M N P tells us how the gauge transformation associated with A μ M A_\mu^M A μ M acts on the vector direction labelled by N N N .
To impose gauge invariance, temporarily regard Θ \Theta Θ as a tensor transforming under G G G . Since Θ N α \Theta_N^{\ \alpha} Θ N α carries both a vector index N N N and an adjoint index α \alpha α , the generator X M X_M X M acts on both. With the conventions used here,
δ M Θ N α = X M N P Θ P α + Θ M β Θ N γ f β γ α . (276) \delta_M\Theta_N^{\ \alpha}=X_{MN}^{\ \ P}\Theta_P^{\ \alpha}+\Theta_M^{\ \beta}\Theta_N^{\ \gamma}f_{\beta\gamma}^{\ \ \alpha}. \tag*{(276)} δ M Θ N α = X M N P Θ P α + Θ M β Θ N γ f β γ α . (276)
The first term is the action on the vector index, while the second is the adjoint action. Gauge invariance requires δ M Θ N α = 0 \delta_M\Theta_N^{\ \alpha}=0 δ M Θ N α = 0 , hence
Θ M β Θ N γ f β γ α + X M N P Θ P α = 0. (277) \Theta_M^{\ \beta}\Theta_N^{\ \gamma}f_{\beta\gamma}^{\ \ \alpha}+X_{MN}^{\ \ P}\Theta_P^{\ \alpha}=0. \tag*{(277)} Θ M β Θ N γ f β γ α + X M N P Θ P α = 0. (277)
This is the quadratic constraint. Both terms are quadratic in Θ \Theta Θ , since X M N P X_{MN}^{\ \ P} X M N P is itself linear in the embedding tensor. The relative signs depend on the conventions chosen for the generators and the covariant derivative; here they match Eq. (273) . Multiplying Eq. (277) by t α t_\alpha t α gives
[ X M , X N ] = − X M N P X P . (278) [X_M,X_N]=-X_{MN}^{\ \ P}X_P. \tag*{(278)} [ X M , X N ] = − X M N P X P . (278)
Thus
[ Im Θ , Im Θ ] ⊆ Im Θ . (279) [\operatorname{Im}\Theta,\operatorname{Im}\Theta]\subseteq\operatorname{Im}\Theta. \tag*{(279)} [ Im Θ , Im Θ ] ⊆ Im Θ. (279)
Only after this condition is imposed can one identify
g g a u g e = Im Θ ⊆ g \mathfrak{g}_{\mathrm{gauge}}=\operatorname{Im}\Theta\subseteq\mathfrak{g} g gauge = Im Θ ⊆ g
as the gauge subalgebra.
There is one subtlety. The coefficients X M N P X_{MN}^{\ \ P} X M N P need not be antisymmetric in M , N M,N M , N , even though Eq. (278) is a commutator. Symmetrizing gives
X ( M N ) P X P = 0 , (280) X_{(MN)}^{\ \ \ P}X_P=0, \tag*{(280)} X ( M N ) P X P = 0 , (280)
where X ( M N ) P = 1 2 ( X M N P + X N M P ) X_{(MN)}^{\ \ \ P}=\frac{1}{2}(X_{MN}^{\ \ P}+X_{NM}^{\ \ P}) X ( M N ) P = 2 1 ( X M N P + X N M P ) . This does not imply X ( M N ) P = 0 X_{(MN)}^{\ \ \ P}=0 X ( M N ) P = 0 , because the generators X P X_P X P need not be linearly independent. Thus X M N P X_{MN}^{\ \ P} X M N P should not in general be identified directly with the structure constants of the gauge algebra. The symmetric part will reappear naturally in the generalized Scherk–Schwarz description below.
The linear constraint: compatibility with supersymmetry
Closure is not enough in a supergravity theory. Even a perfectly good Lie subalgebra may fail to define an admissible supersymmetric gauging because the corresponding couplings do not fit into the allowed supergravity multiplets. The linear constraint selects precisely the representation content that supersymmetry permits.
The embedding tensor is a linear map
Θ : E v e c ⟶ g . \Theta:\mathcal{E}_{\mathrm{vec}}\longrightarrow\mathfrak{g}. Θ : E vec ⟶ g .
It therefore belongs to the vector space of all such linear maps,
Θ ∈ Hom ( E v e c , g ) , \Theta\in\operatorname{Hom}(\mathcal{E}_{\mathrm{vec}},\mathfrak{g}), Θ ∈ Hom ( E vec , g ) ,
where Hom ( V , W ) \operatorname{Hom}(V,W) Hom ( V , W ) denotes the vector space of linear maps from V V V to W W W . For finite-dimensional vector spaces,
Hom ( E v e c , g ) ≃ E v e c ∗ ⊗ g . \operatorname{Hom}(\mathcal{E}_{\mathrm{vec}},\mathfrak{g})\simeq\mathcal{E}_{\mathrm{vec}}^{*}\otimes\mathfrak{g}. Hom ( E vec , g ) ≃ E vec ∗ ⊗ g .
Hence
Θ ∈ E v e c ∗ ⊗ g . (281) \Theta\in\mathcal{E}_{\mathrm{vec}}^{*}\otimes\mathfrak{g}. \tag*{(281)} Θ ∈ E vec ∗ ⊗ g . (281)
In components, this is simply the statement that Θ M α \Theta_{M}{}^{\alpha} Θ M α carries one index in the dual vector representation and one in the adjoint representation of g \mathfrak{g} g .
The tensor product E v e c ∗ ⊗ g \mathcal{E}_{\mathrm{vec}}^{*}\otimes\mathfrak{g} E vec ∗ ⊗ g decomposes into irreducible representations of G G G . Supersymmetry allows only particular components of this decomposition. Denote their sum by
R Θ ⊂ E v e c ∗ ⊗ g . \mathcal{R}_{\Theta}\subset\mathcal{E}_{\mathrm{vec}}^{*}\otimes\mathfrak{g}. R Θ ⊂ E vec ∗ ⊗ g .
The linear constraint is
Θ ∈ R Θ , (282) \Theta\in\mathcal{R}_{\Theta}, \tag*{(282)} Θ ∈ R Θ , (282)
or, if P f o r b i d d e n \mathbb{P}_{\mathrm{forbidden}} P forbidden projects onto the excluded representations,
P f o r b i d d e n Θ = 0. (283) \mathbb{P}_{\mathrm{forbidden}}\Theta=0. \tag*{(283)} P forbidden Θ = 0. (283)
A standard example is maximal supergravity in four dimensions. Its global symmetry group is E 7 ( 7 ) E_{7(7)} E 7 ( 7 ) , the vectors transform in the 56 56 56 , and the generators lie in the adjoint 133 133 133 . Before imposing supersymmetry,
56 ⊗ 133 = 56 ⊕ 912 ⊕ 6480. 56\otimes133=56\oplus912\oplus6480. 56 ⊗ 133 = 56 ⊕ 912 ⊕ 6480.
The linear constraint removes the 56 56 56 and 6480 6480 6480 , leaving
Θ ∈ 912. \Theta\in912. Θ ∈ 912.
This example makes the distinction between the two constraints explicit: the quadratic constraint asks whether the selected generators form a consistent gauging, while the linear constraint asks whether that gauging has the representation content permitted by the supersymmetric theory. In formulations containing both electric and magnetic potentials, additional locality conditions can also arise.
Dressing the gauging: the T T T -tensor
The embedding tensor specifies the gauging in a G G G -covariant way and is independent of the scalar fields. The fermion shifts, however, depend on the scalar configuration and are naturally written in representations of the local isotropy group. The T T T -tensor is the same gauging expressed in this scalar-dependent local frame.
Suppose, for definiteness, that the scalar manifold is locally a coset
M s c a l a r = G / H , \mathcal{M}_{\mathrm{scalar}} = G/H, M scalar = G / H ,
where H H H is the isotropy subgroup of G G G , and let ϕ i \phi^i ϕ i , i = 1 , … , n s i=1,\ldots,n_s i = 1 , … , n s , be local scalar coordinates. A configuration ϕ \phi ϕ specifies a point of G / H G/H G / H . To represent this point by an element transforming under G G G , choose a coset representative
V ( ϕ ) ∈ G . \mathcal{V}(\phi) \in G. V ( ϕ ) ∈ G .
This is not an additional scalar degree of freedom. It is a choice of element of G G G projecting to the point ϕ ∈ G / H \phi\in G/H ϕ ∈ G / H .
The representative is not unique. With our convention,
V ( ϕ ) ⟶ V ( ϕ ) h ( ϕ ) , h ( ϕ ) ∈ H , (284) \mathcal{V}(\phi) \longrightarrow\mathcal{V}(\phi)h(\phi), \qquad h(\phi) \in H, \tag*{(284)} V ( ϕ ) ⟶ V ( ϕ ) h ( ϕ ) , h ( ϕ ) ∈ H , (284)
describes the same scalar point. This is the local H H H redundancy of the coset description.
The usefulness of V ( ϕ ) \mathcal{V}(\phi) V ( ϕ ) is that it converts between two natural descriptions of the theory. The embedding tensor carries indices transforming under the global group G G G , whereas the fermions and their supersymmetry transformations are organized in representations of the local group H H H . Dressing the G G G -indices of Θ \Theta Θ with the coset representative gives the conventional T T T -tensor. To avoid confusion with the radion superfield T T T of Question 10, we denote it by
T ( ϕ ) = V ( ϕ ) − 1 ⋅ Θ , (285) \mathcal{T}(\phi) = \mathcal{V}(\phi)^{-1}\cdot\Theta, \tag*{(285)} T ( ϕ ) = V ( ϕ ) − 1 ⋅ Θ , (285)
where the dot denotes the induced action on all relevant G G G -indices. No new gauging data are introduced. The tensor Θ \Theta Θ is constant, while T ( ϕ ) \mathcal{T}(\phi) T ( ϕ ) is the same gauging expressed in the H H H -covariant frame appropriate to the scalar point ϕ \phi ϕ . Its scalar dependence comes entirely from V ( ϕ ) \mathcal{V}(\phi) V ( ϕ ) .
The T T T -tensor can be decomposed into irreducible representations of H H H . Particular components give the fermionic shift matrices, schematically
S A B ( ϕ ) = P S T ( ϕ ) , N A I ( ϕ ) = P N T ( ϕ ) , (286) S_{AB}(\phi) = \mathbb{P}_{S}\mathcal{T}(\phi), \qquad N^{I}_{A}(\phi) = \mathbb{P}_{N}\mathcal{T}(\phi), \tag*{(286)} S A B ( ϕ ) = P S T ( ϕ ) , N A I ( ϕ ) = P N T ( ϕ ) , (286)
where P S \mathbb{P}_{S} P S and P N \mathbb{P}_{N} P N are the theory-dependent projections appropriate to the gravitini and spin-1 2 \frac{1}{2} 2 1 fermions.
These quantities are called fermion shifts because they appear without derivatives of the supersymmetry parameter:
δ ψ μ A = D μ ϵ A + g g a u S A B ( ϕ ) γ μ ϵ B + ⋯ , \delta\psi_{\mu A} = \mathcal{D}_{\mu}\epsilon_{A} + g_{\mathrm{gau}}S_{AB}(\phi)\gamma_{\mu}\epsilon^{B}+\cdots, δ ψ μ A = D μ ϵ A + g gau S A B ( ϕ ) γ μ ϵ B + ⋯ ,
δ χ I = ⋯ + g g a u N A I ( ϕ ) ϵ A . (287) \delta\chi^{I} = \cdots+ g_{\mathrm{gau}}N^{I}_{A}(\phi)\epsilon^{A}. \tag*{(287)} δ χ I = ⋯ + g gau N A I ( ϕ ) ϵ A . (287)
The group-theoretic role of the T T T -tensor is thus to turn the constant gauging data into scalar-dependent quantities entering the fermionic transformations.
The scalar potential and the supersymmetry Ward identity
Once the gauging has generated fermion shifts, local supersymmetry does not allow the scalar potential to be chosen independently. The variation of the shift and fermion-mass terms must be cancelled by bosonic terms in the action, and this fixes the potential in terms of the same gauging data.
The resulting supersymmetry Ward identity has the schematic form
V g a u ( ϕ ) δ A B = g g a u 2 [ ∑ I c I N A I ( ϕ ) N ‾ I B ( ϕ ) − c ψ S A C ( ϕ ) S ‾ B C ( ϕ ) + ⋯ ] . (288) V_{\mathrm{gau}}(\phi)\delta_A^B=g_{\mathrm{gau}}^2\left[\sum_I c_I N_A^I(\phi)\overline{N}_I^B(\phi)-c_\psi S_{AC}(\phi)\overline{S}^{BC}(\phi)+\cdots\right]. \tag*{(288)} V gau ( ϕ ) δ A B = g gau 2 [ I ∑ c I N A I ( ϕ ) N I B ( ϕ ) − c ψ S A C ( ϕ ) S B C ( ϕ ) + ⋯ ] . (288)
The coefficients c I c_I c I , c ψ c_\psi c ψ , and the possible additional shift matrices depend on the supergravity theory and on conventions. The important structural feature is the relative sign between the spin-1 2 \frac{1}{2} 2 1 and gravitino contributions.
Since the shift matrices are linear in the dressed gauging data, the potential is quadratic in the embedding tensor. If the gauge coupling is absorbed into Θ \Theta Θ , the explicit factor g g a u 2 g_{\mathrm{gau}}^2 g gau 2 in Eq. (288) disappears.
In four-dimensional N = 1 \mathcal{N}=1 N = 1 supergravity the familiar expression
V = F i g i j ˉ F ‾ j ˉ + 1 2 D a D a − 3 ∣ m 3 / 2 ∣ 2 (289) V=F^i g_{i\bar{j}}\overline{F}^{\bar{j}}+\frac{1}{2}D^aD_a-3\lvert m_{3/2}\rvert^2 \tag*{(289)} V = F i g i j ˉ F j ˉ + 2 1 D a D a − 3 ∣ m 3/2 ∣ 2 (289)
provides a simple realization of the same structure. The F F F - and D D D -terms provide the positive contributions, while the gravitino shift gives the negative term. This correlation between the gravitino mass, the remaining fermion shifts and the scalar potential will be central to the no-scale discussion in the next question.
Geometric Scherk–Schwarz reduction and frame algebras
The circle is the simplest geometric example because its invariant one-form d θ d\theta d θ is closed. A non-Abelian group manifold gives the next natural step: its invariant one-forms rotate into one another. For instance, on S U ( 2 ) ≃ S 3 SU(2)\simeq S^3 S U ( 2 ) ≃ S 3 , the left-invariant one-forms satisfy Maurer–Cartan relations with constant coefficients.
More generally, suppose that the internal space admits a globally defined coframe η a \eta^a η a , a = 1 , … , n i n t a=1,\ldots,n_{\mathrm{int}} a = 1 , … , n int , obeying
d η a = − 1 2 f b c a η b ∧ η c , f b c a = constant . (290) d\eta^a=-\frac{1}{2}f_{bc}{}^a\eta^b\wedge\eta^c,\qquad f_{bc}{}^a=\text{constant}. \tag*{(290)} d η a = − 2 1 f b c a η b ∧ η c , f b c a = constant . (290)
The constants f b c a f_{bc}{}^a f b c a measure the failure of the coframe to be closed and encode how the internal frame rotates over the compact space. The circle corresponds to the trivial case f b c a = 0 f_{bc}{}^a=0 f b c a = 0 .
Their constancy is essential: it allows the reduced theory to inherit constant algebraic data rather than arbitrary functions of the internal coordinates. Applying d d d once more and using d 2 = 0 d^2=0 d 2 = 0 gives
f [ a b d f c ] d e = 0 , (291) f_{[ab}{}^d f_{c]d}{}^e=0, \tag*{(291)} f [ ab d f c ] d e = 0 , (291)
which is the Jacobi identity. If E a E_a E a denotes the dual frame, η a ( E b ) = δ b a \eta^a(E_b)=\delta^a_b η a ( E b ) = δ b a , then
[ E a , E b ] = f a b c E c . [E_a,E_b]=f_{ab}{}^cE_c. [ E a , E b ] = f ab c E c .
The internal frame therefore defines a Lie algebra, and the constants f a b c f_{ab}^{c} f ab c descend as part of the lower-dimensional gauging data.
This is only the local algebraic part of the compactification. Let G i n t G_{\mathrm{int}} G int be the Lie group integrating the frame algebra. A compact space may be obtained by a quotient
K = Γ \ G i n t , K = \Gamma\backslash G_{\mathrm{int}}, K = Γ\ G int ,
with Γ \Gamma Γ a suitable discrete subgroup, but such a cocompact lattice need not exist. In particular, its existence requires unimodularity,
f a b b = 0 , (292) f_{ab}^{b} = 0, \tag*{(292)} f ab b = 0 , (292)
which is necessary but not sufficient in general. The Jacobi identity therefore guarantees consistency of the local frame algebra, not the existence of a compact global realization.
Generalized Scherk–Schwarz reduction
The ordinary frame construction only sees the internal tangent bundle. This becomes too restrictive once internal diffeomorphisms and gauge transformations of higher-dimensional form fields mix under dualities. Generalized geometry combines these transformations into a single enlarged frame. In the NS–NS sector, for example, metric twisting and H H H -flux can then appear as different components of the same generalized frame algebra.
In double and exceptional field theory the ordinary frame E a E_a E a is replaced by a generalized frame E A ( Y ) E_A(Y) E A ( Y ) . The variables Y Y Y are generalized internal coordinates: in double field theory the ordinary coordinates are supplemented by dual coordinates to make T-duality manifest, while exceptional field theory uses the corresponding coordinate representation of the relevant U-duality group.
A generalized Scherk–Schwarz reduction is based on a generalized parallelization satisfying
L E A E B = − X A B C E C , X A B C = constant , (293) \mathcal{L}_{E_A}E_B = -X_{AB}^{C}E_C,\qquad X_{AB}^{C}=\text{constant}, \tag*{(293)} L E A E B = − X A B C E C , X A B C = constant , (293)
together with the appropriate section and consistency conditions. The generalized frame contains both tangent and form-gauge directions, so X A B C X_{AB}^{C} X A B C can package geometric twisting and form-field flux in one algebraic object. In toroidal NS–NS examples, the same language also accommodates the duality-related quantities usually denoted Q Q Q - and R R R -flux.
The algebraic structure is different from the ordinary frame case. The generalized Lie derivative is not antisymmetric in its two arguments. Defining
E A ∘ E B ≡ L E A E B , E_A \circ E_B \equiv\mathcal{L}_{E_A}E_B, E A ∘ E B ≡ L E A E B ,
the product need not obey E A ∘ E B = − E B ∘ E A E_A\circ E_B=-E_B\circ E_A E A ∘ E B = − E B ∘ E A . Instead, subject to the section and consistency conditions, it satisfies
E A ∘ ( E B ∘ E C ) = ( E A ∘ E B ) ∘ E C + E B ∘ ( E A ∘ E C ) . (294) E_A\circ(E_B\circ E_C)=(E_A\circ E_B)\circ E_C+E_B\circ(E_A\circ E_C). \tag*{(294)} E A ∘ ( E B ∘ E C ) = ( E A ∘ E B ) ∘ E C + E B ∘ ( E A ∘ E C ) . (294)
This is the Leibniz identity. The constants X A B C X_{AB}^{C} X A B C therefore define, in general, a Leibniz algebra rather than an ordinary Lie algebra.
This is the generalized-geometric counterpart of Eq. (280) . The symmetric directions of X A B C X_{AB}^{C} X A B C do not define additional independent Lie generators, but they cannot simply be discarded from the gaugecovariant formulation. Their realization involves higher-rank potentials and generalized Stückelberg transformations. Vectors can mix with two-forms, two-forms with three-forms, and so on; the resulting sequence of fields and gauge transformations is the tensor hierarchy. When the generalized frame, its density factors and its coordinate dependence obey the required conditions, the constants X A B C X_{AB}{}^{C} X A B C reproduce the embedding-tensor data of the lower-dimensional gauged supergravity.
The gauged theory remains locally supersymmetric
The algebraic machinery above does not by itself break local supersymmetry. A consistent gauging modifies the covariant derivatives, fermionic transformations, mass terms and scalar potential in a correlated way such that the resulting supergravity remains locally supersymmetric. Supersymmetry breaking is instead a property of a bosonic solution.
On a given bosonic background, all fermionic variations must vanish for the same non-zero supersymmetry parameter. It is convenient to collect them into the Killing-spinor operator
K : S ⟶ E f e r m , K ( ϵ ) = ( δ ϵ ψ μ A , δ ϵ χ I , … ) , \mathcal{K}: S \longrightarrow\mathcal{E}_{\mathrm{ferm}}, \qquad\mathcal{K}(\epsilon)=\left(\delta_{\epsilon}\psi_{\mu A},\delta_{\epsilon}\chi^{I},\ldots\right), K : S ⟶ E ferm , K ( ϵ ) = ( δ ϵ ψ μ A , δ ϵ χ I , … ) ,
where S S S is the space of supersymmetry parameters and E f e r m \mathcal{E}_{\mathrm{ferm}} E ferm denotes the collection of fermionic variations evaluated on the chosen background.
Since these variations are linear in ϵ \epsilon ϵ , the preserved supersymmetries form the kernel
S p r e s e r v e d = ker K . (295) S_{\mathrm{preserved}}=\ker\mathcal{K}. \tag*{(295)} S preserved = ker K . (295)
A non-zero gravitino shift S A B ≠ 0 S_{AB}\ne0 S A B = 0 is not by itself a criterion for supersymmetry breaking. Supersymmetric anti-de Sitter vacua provide the standard counterexample: a non-vanishing gravitino shift can be required by the Killing-spinor equation.
For a consistent supersymmetric truncation, a lower-dimensional Killing spinor uplifts through the same reduction ansatz. The converse holds only for higher-dimensional Killing spinors retained by the truncation, so an uplift can preserve additional supersymmetries invisible in the reduced theory.
What the lower-dimensional gauging remembers
The embedding tensor captures the local gauging data of the lower-dimensional supergravity. Together with the scalar geometry, it determines the gauge generators, covariant derivatives, fermion shifts, mass matrices and scalar potential. It does not reconstruct the complete higher-dimensional compactification.
The geometric reductions above already provide an example. The constants f a b c f_{ab}{}^{c} f ab c determine the local frame algebra, but not the discrete subgroup Γ \Gamma Γ used to form a compact quotient K = Γ \ G i n t K=\Gamma\backslash G_{\mathrm{int}} K = Γ\ G int . Different global compactifications can therefore share the same local gauging data. Topology, lattices and discrete identifications are not contained in Θ \Theta Θ .
The converse problem also occurs. An embedding tensor may satisfy the lower-dimensional linear and quadratic constraints without admitting an uplift within a specified higher-dimensional construction. In a generalized Scherk–Schwarz reduction, for example, the generalized frame must also satisfy the section and consistency conditions of double or exceptional field theory. These are higher-dimensional requirements that are not encoded solely in Θ \Theta Θ .
The embedding tensor is the local algebraic image of the compactification data in the reduced theory. For an ordinary circle reduction this information comes from the single twist generator M \mathcal{M} M ; for geometric reductions it includes f a b c f_{ab}{}^c f ab c ; for generalized reductions it is encoded in the constant generalized-frame data X A B C X_{AB}{}^C X A B C .
Embedding-tensor formulations of gauged supergravity, including the linear and quadratic constraints and the associated tensor hierarchy, can be found in Refs. [48, 49, 50, 51, 52]. Geometric reductions in which constant internal frame data descend to lower-dimensional gaugings are discussed in Refs. [53, 54]. Generalized Scherk–Schwarz reductions in double and exceptional field theory, in which generalized frame data reproduce the embedding tensor of the reduced gauged supergravity, can be found in Refs. [55, 56, 57, 58].
Take-home message
On a circle, local translations of the compact coordinate are gauged by the Kaluza–Klein vector, while the Scherk–Schwarz ansatz makes this symmetry act through the twist generator M \mathcal{M} M .
The embedding tensor assigns a global-symmetry generator X M X_M X M to each vector field A μ M A_\mu{}^M A μ M . The elementary circle reduction is the rank-one case.
The quadratic constraint expresses gauge covariance and closure of the selected generators. The linear constraint independently restricts the embedding tensor to representations compatible with supersymmetry.
Dressing the embedding tensor with the scalar fields gives the T T T -tensor. Its components determine the fermion shifts and, through the supersymmetry Ward identity, the scalar potential.
Geometric Scherk–Schwarz reductions are governed by constant Lie algebras of internal frames. Generalized reductions replace them by generalized frame algebras that are naturally Leibniz and are accompanied by the tensor hierarchy.
A consistent gauging remains locally supersymmetric. Supersymmetry breaking is determined only after choosing a bosonic background and solving the complete fermionic Killing-spinor system.
The embedding tensor remembers the local lower-dimensional gauging, not the complete global compactification that produced it.
This naturally leads to the next question:
Why does Scherk–Schwarz breaking often lead to a no-scale structure?
Why does Scherk–Schwarz breaking often lead to a no-scale structure?
Essential idea
In many classic Scherk–Schwarz compactifications, the supersymmetry-breaking scale is set by a modulus while the classical potential remains flat along the same direction. The gravitino mass can therefore vary continuously along a family of Minkowski vacua.
The origin of this structure becomes particularly transparent in a four-dimensional N = 1 N = 1 N = 1 description. Suppose the relevant compactification moduli enter through
K = − log Y ( r ) , r a = T a + T ˉ a ˉ K = -\log Y(r), \qquad r^{a} = T^{a} + \bar{T}^{\bar{a}} K = − log Y ( r ) , r a = T a + T ˉ a ˉ
with Y Y Y homogeneous of degree p p p . Homogeneity implies
K a b ˉ K b ˉ = − r a , K a b ˉ K a K b ˉ = p . K^{a\bar{b}}K_{\bar{b}} = -r^{a}, \qquad K^{a\bar{b}}K_{a}K_{\bar{b}} = p. K a b ˉ K b ˉ = − r a , K a b ˉ K a K b ˉ = p .
If the superpotential is independent of these moduli, the second identity gives the no-scale cancellation for p = 3 p = 3 p = 3 . For constant non-zero W W W , the first gives F a ∝ r a F^{a} \propto r^{a} F a ∝ r a , so supersymmetry is broken along the Euler direction that rescales the compactification.
A Scherk–Schwarz spectrum naturally depends on this same scale modulus. When
m i ( r , ω ) = M K K ( r ) μ i ( ω ) , m_{i}(r,\omega) = M_{\mathrm{KK}}(r)\,\mu_{i}(\omega), m i ( r , ω ) = M KK ( r ) μ i ( ω ) ,
motion along the no-scale direction changes the gravitino and Kaluza–Klein masses without costing classical potential energy. This is the origin of the characteristic sliding spectrum of Scherk–Schwarz no-scale models.
For the homogeneous geometry above, the Goldstino direction also has a universal holomorphic sectional curvature,
κ f = − 2 p \kappa_{f} = -\frac{2}{p} κ f = − p 2
in the curvature convention adopted below, and hence κ f = − 2 / 3 \kappa_{f} = -2/3 κ f = − 2/3 at the no-scale value p = 3 p = 3 p = 3 .
Additional superpotential dependence, further gaugings, localized interactions or quantum corrections can lift this structure. The discussion below identifies the assumptions under which the standard Scherk–Schwarz no-scale behaviour emerges.
Why the no-scale question arises
A Scherk–Schwarz compactification can give the gravitino a non-zero mass while leaving the modulus that sets the compactification scale classically undetermined. This is initially surprising. As discussed in Question 13, the fermion shifts generated by a gauging are tied by local supersymmetry to the scalar potential. One might therefore expect the modulus controlling the supersymmetry-breaking scale to acquire a potential as well.
The no-scale situation is the case in which the two effects separate. The gravitino mass varies with the modulus, but the classical scalar potential remains flat along the same direction. The Scherk–Schwarz spectrum explains why the masses depend on that modulus; the scalar geometry explains why the corresponding potential can nevertheless cancel.
A sliding scale along a vacuum family
Before specializing to supergravity, it is useful to state precisely what is meant by a classically flat family. The elementary potential V ( x , y ) = x 2 V(x,y) = x^2 V ( x , y ) = x 2 has minima along N = { x = 0 } \mathcal{N} = \{x = 0\} N = { x = 0 } . Along this family V = 0 V = 0 V = 0 , and d V = 2 x d x dV = 2x\,dx d V = 2 x d x vanishes at every point. Thus y y y is a genuine flat direction of a continuous family of stationary vacua.
More generally, let M s c a l \mathcal{M}_{\mathrm{scal}} M scal be the scalar manifold of the lower-dimensional theory and let N ⊂ M s c a l \mathcal{N} \subset\mathcal{M}_{\mathrm{scal}} N ⊂ M scal be a connected submanifold satisfying
V ∣ N = 0 , d V ∣ N = 0. (296) V|_{\mathcal{N}} = 0,\qquad dV|_{\mathcal{N}} = 0. \tag*{(296)} V ∣ N = 0 , d V ∣ N = 0. (296)
We shall call N \mathcal{N} N a stationary Minkowski family . No assumption is made here about stability in directions transverse to N \mathcal{N} N .
The existence of such a family already implies massless scalar directions. If q ∈ N q \in\mathcal{N} q ∈ N and X ∈ T q N X \in T_q\mathcal{N} X ∈ T q N , then d V dV d V vanishes all along N \mathcal{N} N , so for any Y ∈ T q M s c a l Y \in T_q\mathcal{M}_{\mathrm{scal}} Y ∈ T q M scal ,
( ∇ 2 V ) q ( X , Y ) = ( ∇ X d V ) q ( Y ) = 0. (\nabla^2 V)_q(X,Y) = (\nabla_X dV)_q(Y) = 0. ( ∇ 2 V ) q ( X , Y ) = ( ∇ X d V ) q ( Y ) = 0.
Hence
T q N ⊆ ker ( ∇ 2 V ) q . (297) T_q\mathcal{N} \subseteq\ker(\nabla^2 V)_q. \tag*{(297)} T q N ⊆ ker ( ∇ 2 V ) q . (297)
The characteristic no-scale situation is obtained when supersymmetry is broken on this family and the breaking scale itself changes along it. At a generic point there is then a tangent direction X ∈ T q N X \in T_q\mathcal{N} X ∈ T q N for which
X ( m 3 / 2 ) ≠ 0. (298) X(m_{3/2}) \ne0. \tag*{(298)} X ( m 3/2 ) = 0. (298)
The potential remains unchanged while the gravitino mass slides.
The Scherk–Schwarz origin of the sliding spectrum
The simplest example is again the circle. For a field of Scherk–Schwarz charge q q q ,
m n , q ( R , ω ) = n + q ω R , n ∈ Z , (299) m_{n,q}(R,\omega) = \frac{n+q\omega}{R},\qquad n \in\mathbb{Z}, \tag*{(299)} m n , q ( R , ω ) = R n + q ω , n ∈ Z , (299)
up to the charge and normalization conventions used in the preceding questions. The twist fixes the dimensionless displacement of the tower, while R R R fixes its physical scale.
At the half-twist discussed in Question 11, the light Dirac gravitino has m 3 / 2 = 1 / ( 2 R ) m_{3/2}=1/(2R) m 3/2 = 1/ ( 2 R ) . If R R R parametrizes a flat vacuum family, its mass varies continuously along that family. The same is true for the rest of the shifted Kaluza–Klein spectrum.
The mechanism generalizes whenever the internal operator factorizes as
D i n t ( r , ω ) = M K K ( r ) D ^ ω (300) \mathcal{D}_{\mathrm{int}}(r,\omega)=M_{\mathrm{KK}}(r)\widehat{\mathcal{D}}_{\omega} \tag*{(300)} D int ( r , ω ) = M KK ( r ) D ω (300)
with D ^ ω \widehat{\mathcal{D}}_{\omega} D ω dimensionless. If D ^ ω u i = μ i ( ω ) u i \widehat{\mathcal{D}}_{\omega}u_i=\mu_i(\omega)u_i D ω u i = μ i ( ω ) u i , then
m i ( r , ω ) = M K K ( r ) μ i ( ω ) . (301) m_i(r,\omega)=M_{\mathrm{KK}}(r)\mu_i(\omega). \tag*{(301)} m i ( r , ω ) = M KK ( r ) μ i ( ω ) . (301)
For the circle, M K K = 1 / R M_{\mathrm{KK}}=1/R M KK = 1/ R and μ n , q = n + q ω \mu_{n,q}=n+q\omega μ n , q = n + q ω . Once the scale modulus is flat, the whole shifted spectrum moves with it. What remains to be explained is why the scalar potential can leave that scale modulus unfixed.
The homogeneous no-scale geometry
The answer is particularly transparent in an important class of four-dimensional N = 1 N=1 N = 1 supergravities. Consider moduli T a T^a T a , define r a = T a + T ˉ a ˉ r^a=T^a+\bar{T}^{\bar{a}} r a = T a + T ˉ a ˉ , and take
K = − log Y ( r ) , (302) K=-\log Y(r), \tag*{(302)} K = − log Y ( r ) , (302)
where Y > 0 Y>0 Y > 0 is homogeneous of degree p p p ,
Y ( λ r ) = λ p Y ( r ) . (303) Y(\lambda r)=\lambda^pY(r). \tag*{(303)} Y ( λ r ) = λ p Y ( r ) . (303)
The familiar one-modulus example,
Y ( r ) = r 3 , r = T + T ˉ , Y(r)=r^3,\qquad r=T+\bar{T}, Y ( r ) = r 3 , r = T + T ˉ ,
gives
K = − 3 log ( T + T ˉ ) . K=-3\log(T+\bar{T}). K = − 3 log ( T + T ˉ ) .
It is simply the case p = 3 p=3 p = 3 of the more general geometry.
Let Y a = ∂ Y / ∂ r a Y_a=\partial Y/\partial r^a Y a = ∂ Y / ∂ r a and Y a b = ∂ a ∂ b Y Y_{ab}=\partial_a\partial_bY Y ab = ∂ a ∂ b Y . Euler’s theorem and its derivative give
r a Y a = p Y , r a Y a b = ( p − 1 ) Y b . (304) r^aY_a=pY,\qquad r^aY_{ab}=(p-1)Y_b. \tag*{(304)} r a Y a = p Y , r a Y ab = ( p − 1 ) Y b . (304)
Since K K K depends on the complex moduli only through r a r^a r a ,
K a = − Y a Y , K a b ˉ = − Y a b Y + Y a Y b Y 2 . K_a=-\frac{Y_a}{Y},\qquad K_{a\bar{b}}=-\frac{Y_{ab}}{Y}+\frac{Y_aY_b}{Y^2}. K a = − Y Y a , K a b ˉ = − Y Y ab + Y 2 Y a Y b .
Contracting the metric with r b r^b r b gives K a b ˉ r b = − K a K_{a\bar{b}}r^b=-K_a K a b ˉ r b = − K a , and hence
K a b ˉ K b ˉ = − r a . (305) K^{a\bar{b}}K_{\bar{b}}=-r^a. \tag*{(305)} K a b ˉ K b ˉ = − r a . (305)
A further contraction gives
K a b ˉ K a K b ˉ = p . (306) K^{a\bar b}K_aK_{\bar b}=p. \tag*{(306)} K a b ˉ K a K b ˉ = p . (306)
The two identities follow from the same homogeneity property, but they encode different physics.
Suppose that, in the Kähler frame under consideration, the superpotential is independent of these moduli,
∂ a W = 0. (307) \partial_a W=0. \tag*{(307)} ∂ a W = 0. (307)
Then D a W = K a W D_aW=K_aW D a W = K a W . If this modulus sector contains the non-vanishing F -terms relevant to the vacuum family, and no additional D -term or other tree-level contribution is present, the F -term potential is
V F = e K ( K a b ˉ D a W D b W ‾ − 3 ∣ W ∣ 2 ) = e K ( p − 3 ) ∣ W ∣ 2 . (308) V_F=e^K\left(K^{a\bar b}D_aW\overline{D_bW}-3|W|^2\right)
=e^K(p-3)|W|^2. \tag*{(308)} V F = e K ( K a b ˉ D a W D b W − 3∣ W ∣ 2 ) = e K ( p − 3 ) ∣ W ∣ 2 . (308)
For p = 3 p=3 p = 3 , the positive moduli F -term contribution therefore cancels the universal negative gravitino contribution. The potential vanishes along these moduli even though supersymmetry can remain broken.
This is the familiar no-scale cancellation. The crucial point is not that the gravitino mass vanishes, but precisely that it need not: m 3 / 2 = e K / 2 ∣ W ∣ m_{3/2}=e^{K/2}|W| m 3/2 = e K /2 ∣ W ∣ can vary as the moduli move while V F V_F V F remains zero.
The Goldstino points along the scale direction
The first identity, Eq. (305) , determines the direction of supersymmetry breaking.
For a constant non-zero superpotential W = W 0 W=W_0 W = W 0 ,
F a = − e K / 2 K a b ˉ D b ˉ W ‾ . F^a=-e^{K/2}K^{a\bar b}D_{\bar b}\overline{W}. F a = − e K /2 K a b ˉ D b ˉ W .
Since D b ˉ W ‾ = K b ˉ W ‾ 0 D_{\bar b}\overline{W}=K_{\bar b}\overline{W}_0 D b ˉ W = K b ˉ W 0 , one obtains
F a = e K / 2 W ‾ 0 r a . (309) F^a=e^{K/2}\overline{W}_0r^a. \tag*{(309)} F a = e K /2 W 0 r a . (309)
The Goldstino direction is therefore parallel to the Euler vector E = r a ∂ / ∂ r a E=r^a\partial/\partial r^a E = r a ∂ / ∂ r a , which generates a common rescaling of the homogeneous coordinates. In the compactification interpretation, this is precisely the scale direction.
This gives a direct geometric interpretation of the radion F -term discussed in Question 10: in the homogeneous benchmark, the auxiliary direction responsible for supersymmetry breaking is aligned with the modulus that rescales the compactification.
The alignment itself does not require p = 3 p=3 p = 3 . It follows from homogeneity. The special value p = 3 p=3 p = 3 is what makes the scalar potential cancel.
The curvature seen by the Goldstino
The same homogeneous structure also fixes the intrinsic Kähler geometry along the supersymmetry-breaking direction.
Let g a b ˉ = K a b ˉ g_{a\bar b}=K_{a\bar b} g a b ˉ = K a b ˉ . Since K = − log Y K=-\log Y K = − log Y and Y Y Y has degree p p p , the metric has degree − 2 -2 − 2 ,
r c ∂ c g a b ˉ = − 2 g a b ˉ . (310) r^{c}\partial_{c}g_{a\bar b}=-2g_{a\bar b}. \tag*{(310)} r c ∂ c g a b ˉ = − 2 g a b ˉ . (310)
Because K K K depends on T a T^{a} T a and T ˉ a ˉ \bar T^{\bar a} T ˉ a ˉ only through r a r^{a} r a , this implies
K a b ˉ c r c = − 2 K a b ˉ , K a c b ˉ r a r c = 2 K b ˉ , K_{a\bar b c}r^{c}=-2K_{a\bar b},\qquad K_{a c\bar b}r^{a}r^{c}=2K_{\bar b}, K a b ˉ c r c = − 2 K a b ˉ , K a c b ˉ r a r c = 2 K b ˉ ,
together with the complex-conjugate relation. Repeating the differentiation and contraction gives
K a b ˉ c d ˉ r a r b ˉ r c r d ˉ = 6 p , (311) K_{a\bar b c\bar d}r^{a}r^{\bar b}r^{c}r^{\bar d}=6p, \tag*{(311)} K a b ˉ c d ˉ r a r b ˉ r c r d ˉ = 6 p , (311)
while K a b ˉ r a r b ˉ = p K_{a\bar b}r^{a}r^{\bar b}=p K a b ˉ r a r b ˉ = p .
We adopt the curvature convention
R a b ˉ c d ˉ = − ∂ a ∂ b ˉ g c d ˉ + g e f ˉ ( ∂ a g c f ˉ ) ( ∂ b ˉ g e d ˉ ) . (312) R_{a\bar b c\bar d}=-\partial_{a}\partial_{\bar b}g_{c\bar d}+g^{e\bar f}\left(\partial_{a}g_{c\bar f}\right)\left(\partial_{\bar b}g_{e\bar d}\right). \tag*{(312)} R a b ˉ c d ˉ = − ∂ a ∂ b ˉ g c d ˉ + g e f ˉ ( ∂ a g c f ˉ ) ( ∂ b ˉ g e d ˉ ) . (312)
Contracting with four Euler vectors gives
R a b ˉ c d ˉ r a r b ˉ r c r d ˉ = − 6 p + 4 K e K e f ˉ K f ˉ = − 2 p . (313) \begin{aligned}
R_{a\bar b c\bar d}r^{a}r^{\bar b}r^{c}r^{\bar d}=-6p+4K_{e}K^{e\bar f}K_{\bar f} \\
&=-2p. \tag*{(313)}
\end{aligned} R a b ˉ c d ˉ r a r b ˉ r c r d ˉ = − 6 p + 4 K e K e f ˉ K f ˉ = − 2 p . (313)
Since the Euler direction has norm p p p , define the corresponding unit vector by f a = r a / p f^{a}=r^{a}/\sqrt{p} f a = r a / p . Its holomorphic sectional curvature is
κ f ≡ R a b ˉ c d ˉ f a f ˉ b ˉ f c f ˉ d ˉ = − 2 p . (314) \kappa_{f}\equiv R_{a\bar b c\bar d}f^{a}\bar f^{\bar b}f^{c}\bar f^{\bar d}=-\frac{2}{p}. \tag*{(314)} κ f ≡ R a b ˉ c d ˉ f a f ˉ b ˉ f c f ˉ d ˉ = − p 2 . (314)
For the no-scale value p = 3 p=3 p = 3 , this gives κ f = − 2 / 3 \kappa_{f}=-2/3 κ f = − 2/3 .
The sign of this quantity depends on the convention used for the Riemann tensor. References employing the opposite convention must therefore be translated before sectional-curvature bounds are compared.
The relevance of κ f \kappa_{f} κ f is that the Kähler curvature along the Goldstino direction enters the scalar mass matrix and the sgoldstino sector. General analyses of non-supersymmetric Minkowski vacua make this dependence explicit [59]. Here, however, the curvature is not being used to establish the flatness. For the homogeneous p = 3 p=3 p = 3 benchmark that flatness already follows directly from Eq. (308) . The curvature calculation gives an additional intrinsic property of the same supersymmetry-breaking direction.
A comparison with O’Raifeartaigh pseudomoduli
There is a useful analogy with global supersymmetry. In renormalizable Wess–Zumino models with canonical Kähler metric, a stable tree-level supersymmetry-breaking configuration possesses, under the usual assumptions, a pseudomodulus aligned with the supersymmetry-breaking direction [60, 61].
The resemblance is that the Goldstino direction is again associated with a classically flat scalar direction. The origin of the flatness is different. In the O’Raifeartaigh case it follows from the structure of the renormalizable global-supersymmetry model. In supergravity there is no general theorem requiring the scalar partner of the Goldstino to remain massless: its behaviour depends on the Kähler geometry and on the other couplings.
In the homogeneous no-scale model, the flat direction instead follows from the cancellation derived above. The R R R -symmetry used for the Scherk–Schwarz twist does not by itself protect the radial modulus. The relation between R R R -symmetry and generic F F F -term breaking in global supersymmetry is a separate issue [62].
How the Scherk–Schwarz and no-scale structures meet
We can now see why the two structures occur together so naturally in the standard examples.
The Scherk–Schwarz monodromy provides the dimensionless deformation μ i ( ω ) \mu_i(\omega) μ i ( ω ) of the Kaluza–Klein spectrum. The compactification modulus supplies its overall scale M K K ( r ) M_{\mathrm{KK}}(r) M KK ( r ) . The homogeneous scalar geometry aligns the Goldstino with the Euler direction, so the same modulus also controls the supersymmetry-breaking auxiliary field. At degree p = 3 p = 3 p = 3 , the supersymmetry Ward identity becomes the no-scale cancellation and leaves this direction classically flat.
The result is a family of Minkowski vacua in which the supersymmetry-breaking scale can slide:
V = 0 , m 3 / 2 = m 3 / 2 ( r ) , m i = M K K ( r ) μ i ( ω ) . V = 0,\qquad m_{3/2} = m_{3/2}(r),\qquad m_i = M_{\mathrm{KK}}(r)\mu_i(\omega). V = 0 , m 3/2 = m 3/2 ( r ) , m i = M KK ( r ) μ i ( ω ) .
The same scale direction controls the Kaluza–Klein spectrum, the gravitino mass and the Goldstino direction, while costing no classical potential energy.
When the no-scale structure is lifted
The preceding result relies on a definite set of assumptions. The internal spectrum need not depend on a single overall scale; several radii, warping or additional dimensionful parameters may enter. The scalar geometry need not have the homogeneous form used above, or its degree need not be p = 3 p = 3 p = 3 .
The superpotential may also depend on the scale modulus. Additional F F F -terms, D D D -terms, fluxes, internal curvature, localized interactions or further gaugings can then contribute to the scalar potential. Localized terms of the type discussed in Question 9 provide one possible source of such lifting. Quantum corrections can likewise generate a potential for a direction that is flat at tree level and can modify the effective Kähler geometry.
Thus the no-scale behaviour discussed here is characteristic of an important class of Scherk–Schwarz compactifications rather than a kinematical consequence of the twist alone. The mechanisms that lift the flat direction and determine the supersymmetry-breaking scale are the subject of the next question.
The no-scale structure of supergravity and its relation to supersymmetry breaking are discussed in the early literature on no-scale models. Homogeneous formulations and sliding-gravitino supergravities provide a particularly useful language for the Scherk–Schwarz setting; the structures used above are discussed in Refs. [41, 63, 64].
The relation between the Goldstino direction and Kähler curvature in non-supersymmetric Minkowski vacua is discussed in Ref. [59]. The comparison with tree-level pseudomoduli in global supersymmetry is related to Refs. [60, 61, 62].
Take-home message
m i ( r , ω ) = M K K ( r ) μ i ( ω ) . m_i(r,\omega) = M_{\mathrm{KK}}(r)\mu_i(\omega). m i ( r , ω ) = M KK ( r ) μ i ( ω ) .
K = − log Y ( r ) , Y ( λ r ) = λ p Y ( r ) , K = -\log Y(r), \qquad Y(\lambda r) = \lambda^p Y(r), K = − log Y ( r ) , Y ( λ r ) = λ p Y ( r ) ,
one has
K a b ˉ K b ˉ = − r a , K a b ˉ K a K b ˉ = p . K^{a\bar{b}}K_{\bar{b}} = -r^a, \qquad K^{a\bar{b}}K_aK_{\bar{b}} = p. K a b ˉ K b ˉ = − r a , K a b ˉ K a K b ˉ = p .
κ f = − 2 p , \kappa_f = -\frac{2}{p}, κ f = − p 2 ,
and hence κ f = − 2 / 3 \kappa_f=-2/3 κ f = − 2/3 for p = 3 p=3 p = 3 .
This naturally leads to the next question:
What can lift the no-scale direction and determine the supersymmetry-breaking scale?
What can lift the no-scale direction and determine the supersymmetry-breaking scale?
Essential idea
A no-scale theory contains a classically flat vacuum family. The question is not simply whether a correction is present, but whether it produces a non-constant potential along the former flat directions.
A useful picture is a flat valley. A perturbation can tilt the floor of the valley directly, or it can first displace the minimum in the transverse directions. These two effects generally occur at different orders.
Let the unperturbed scalar potential be V 0 V_0 V 0 , with no-scale vacuum manifold N \mathcal{N} N , and perturb it as
V ε = V 0 + ε V 1 + ε 2 V 2 + ⋯ V_{\varepsilon} = V_0 + \varepsilon V_1 + \varepsilon^2 V_2 + \cdots V ε = V 0 + ε V 1 + ε 2 V 2 + ⋯
If the Hessian of V 0 V_0 V 0 is non-degenerate transverse to N \mathcal{N} N , the transverse fields can be solved for perturbatively. Their first displacement is
y ⋆ A = − ε ( H ⊥ − 1 ) A B ∂ B V 1 ∣ N + O ( ε 2 ) , y_{\star}^{A} = -\varepsilon(H_{\perp}^{-1})^{AB} \partial_B V_1\big|_{\mathcal{N}} + O(\varepsilon^2), y ⋆ A = − ε ( H ⊥ − 1 ) A B ∂ B V 1 N + O ( ε 2 ) ,
while the induced potential along the former flat family is
V e f f ( N ) = ε V 1 ∣ N + ε 2 [ V 2 ∣ N − 1 2 ( ∂ A V 1 ) ( H ⊥ − 1 ) A B ( ∂ B V 1 ) ] + O ( ε 3 ) . V_{\mathrm{eff}}^{(\mathcal{N})} = \varepsilon V_1\big|_{\mathcal{N}} + \varepsilon^2 \left[V_2\big|_{\mathcal{N}} - \frac{1}{2}(\partial_A V_1)(H_{\perp}^{-1})^{AB}(\partial_B V_1)\right] + O(\varepsilon^3). V eff ( N ) = ε V 1 N + ε 2 [ V 2 N − 2 1 ( ∂ A V 1 ) ( H ⊥ − 1 ) A B ( ∂ B V 1 ) ] + O ( ε 3 ) .
The first-order lifting is therefore controlled by the restriction V 1 ∣ N V_1|_{\mathcal{N}} V 1 ∣ N , not by the mere presence of V 1 V_1 V 1 in the full scalar theory. If this restriction is constant, the degeneracy survives at first order, but transverse backreaction can still generate a non-constant potential at the next order.
Lifting does not yet mean stabilization. A lifted direction may run away, possess only unstable extrema, or admit a stable minimum. Moreover, fixing the compactification scale need not fix an effective Scherk–Schwarz phase, and neither statement by itself determines the amount of preserved supersymmetry or the vacuum energy.
The aim of this question is therefore not to review particular models of modulus stabilization, but to isolate a general criterion that can be applied to any classical or quantum deformation of a no-scale Scherk–Schwarz compactification.
A flat valley and what a perturbation can do
Consider two real scalar fields t t t and y y y , with
V 0 ( t , y ) = 1 2 M 2 y 2 , M 2 > 0. (315) V_0(t,y) = \frac{1}{2}M^2y^2,\qquad M^2 > 0. \tag*{(315)} V 0 ( t , y ) = 2 1 M 2 y 2 , M 2 > 0. (315)
The minima form the continuous family N = { y = 0 } \mathcal{N} = \{y = 0\} N = { y = 0 } . The coordinate t t t parametrizes the flat direction, while y y y is a stable transverse field of mass M M M .
Now perturb the potential by
V ε ( t , y ) = 1 2 M 2 y 2 + ε [ U ( t ) + J ( t ) y ] + O ( ε 2 ) , (316) V_\varepsilon(t,y) = \frac{1}{2}M^2y^2 + \varepsilon\left[U(t) + J(t)y\right] + O(\varepsilon^2), \tag*{(316)} V ε ( t , y ) = 2 1 M 2 y 2 + ε [ U ( t ) + J ( t ) y ] + O ( ε 2 ) , (316)
where U ( t ) U(t) U ( t ) and J ( t ) J(t) J ( t ) are arbitrary smooth functions.
Two effects are immediately visible. Restricting the perturbation to the original flat family gives V 1 ∣ N = U ( t ) V_1|_{\mathcal{N}} = U(t) V 1 ∣ N = U ( t ) . If U ( t ) U(t) U ( t ) is non-constant, the flat direction is lifted directly at order ε \varepsilon ε .
The transverse minimum also moves. The y y y -equation gives
M 2 y + ε J ( t ) + O ( ε 2 ) = 0 , M^2y + \varepsilon J(t) + O(\varepsilon^2) = 0, M 2 y + ε J ( t ) + O ( ε 2 ) = 0 ,
and hence
y ⋆ ( t ) = − ε J ( t ) M 2 + O ( ε 2 ) . (317) y_\star(t) = -\varepsilon\frac{J(t)}{M^2} + O(\varepsilon^2). \tag*{(317)} y ⋆ ( t ) = − ε M 2 J ( t ) + O ( ε 2 ) . (317)
Substituting back into the potential gives
V e f f ( t ) = ε U ( t ) + 1 2 M 2 y ⋆ 2 + ε J ( t ) y ⋆ + O ( ε 3 ) = ε U ( t ) − ε 2 2 M 2 J ( t ) 2 + O ( ε 3 ) . (318) \begin{aligned}
V_{\mathrm{eff}}(t) = \varepsilon U(t) + \frac{1}{2}M^2y_\star^2 + \varepsilon J(t)y_\star+ O(\varepsilon^3) \\
&= \varepsilon U(t) - \frac{\varepsilon^2}{2M^2}J(t)^2 + O(\varepsilon^3). \tag*{(318)}
\end{aligned} V eff ( t ) = ε U ( t ) + 2 1 M 2 y ⋆ 2 + ε J ( t ) y ⋆ + O ( ε 3 ) = ε U ( t ) − 2 M 2 ε 2 J ( t ) 2 + O ( ε 3 ) . (318)
This elementary example already contains the two mechanisms that will appear in the general result. If U ( t ) U(t) U ( t ) is non-constant, the valley is tilted directly at first order. If U ( t ) U(t) U ( t ) is constant but J ( t ) J(t) J ( t ) is not, the first-order degeneracy survives, while the transverse readjustment generates a t t t -dependent potential at second order.
The negative sign of the second contribution has a simple meaning: allowing a stable transverse field to relax lowers the energy relative to holding it fixed at its unperturbed value.
Nothing here guarantees stabilization. A non-constant U ( t ) U(t) U ( t ) , or the induced term − J ( t ) 2 / ( 2 M 2 ) -J(t)^2/(2M^2) − J ( t ) 2 / ( 2 M 2 ) , may instead produce a runaway or an unstable extremum. The same distinction will remain important in the general case.
Lifting a no-scale vacuum manifold
Let V 0 V_0 V 0 possess a no-scale stationary Minkowski family in the sense of Question 14,
V 0 ∣ N = 0 , d V 0 ∣ N = 0 , (319) V_0|_{\mathcal{N}} = 0,\qquad dV_0|_{\mathcal{N}} = 0, \tag*{(319)} V 0 ∣ N = 0 , d V 0 ∣ N = 0 , (319)
with N ⊂ M s c a l \mathcal{N} \subset\mathcal{M}_{\mathrm{scal}} N ⊂ M scal .
As shown there,
T q N ⊆ ker ( ∇ 2 V 0 ) q T_q\mathcal{N} \subseteq\operatorname{ker}(\nabla^2 V_0)_q T q N ⊆ ker ( ∇ 2 V 0 ) q
for every q ∈ N q \in\mathcal{N} q ∈ N . We now assume that these tangential zero modes are the only zero modes of the Hessian.
Choose local coordinates ( t a , y A ) (t^a,y^A) ( t a , y A ) adapted to the vacuum family, so that N = { y A = 0 } \mathcal{N}=\{y^A=0\} N = { y A = 0 } . The t a t^a t a parametrize the vacuum manifold, while the y A y^A y A describe transverse directions. Define the transverse Hessian by
( H ⊥ ) A B ( t ) = ∂ A ∂ B V 0 ( t , 0 ) , (320) (H_\perp)_{AB}(t)=\partial_A\partial_B V_0(t,0), \tag*{(320)} ( H ⊥ ) A B ( t ) = ∂ A ∂ B V 0 ( t , 0 ) , (320)
and assume that it is invertible.
This is the Morse–Bott non-degeneracy condition appropriate to the problem: the critical set is the smooth manifold N \mathcal{N} N , and the Hessian has no zero modes transverse to it [65]. Since V 0 V_0 V 0 and its first derivatives vanish on N \mathcal{N} N , the transverse gradient starts linearly,
∂ A V 0 ( t , y ) = ( H ⊥ ) A B ( t ) y B + O ( y 2 ) . \partial_A V_0(t,y)=(H_\perp)_{AB}(t)y^B+O(y^2). ∂ A V 0 ( t , y ) = ( H ⊥ ) A B ( t ) y B + O ( y 2 ) .
Now perturb the theory,
V ε = V 0 + ε V 1 + ε 2 V 2 + O ( ε 3 ) . (321) V_\varepsilon=V_0+\varepsilon V_1+\varepsilon^2 V_2+O(\varepsilon^3). \tag*{(321)} V ε = V 0 + ε V 1 + ε 2 V 2 + O ( ε 3 ) . (321)
The transverse fields readjust according to ∂ A V ε = 0 \partial_A V_\varepsilon=0 ∂ A V ε = 0 . At first order this gives
( H ⊥ ) A B y B + ε ∂ A V 1 ∣ N + O ( ε 2 ) = 0 , (H_\perp)_{AB}y^B+\varepsilon\partial_A V_1\big|_{\mathcal{N}}+O(\varepsilon^2)=0, ( H ⊥ ) A B y B + ε ∂ A V 1 N + O ( ε 2 ) = 0 ,
and therefore
y ⋆ A ( t ) = − ε ( H ⊥ − 1 ) A B ∂ B V 1 ∣ N + O ( ε 2 ) . (322) y_\star^A(t)=-\varepsilon(H_\perp^{-1})^{AB}\partial_B V_1\big|_{\mathcal{N}}+O(\varepsilon^2). \tag*{(322)} y ⋆ A ( t ) = − ε ( H ⊥ − 1 ) A B ∂ B V 1 N + O ( ε 2 ) . (322)
The transverse fields can thus move already at first order even when the former flat directions are not yet lifted at that order.
Substituting y = y ⋆ ( t ) y=y_\star(t) y = y ⋆ ( t ) into the full potential gives
V e f f ( N ) ( t ) = ε V 1 ∣ N + ε 2 [ V 2 ∣ N − 1 2 ( ∂ A V 1 ) ( H ⊥ − 1 ) A B ( ∂ B V 1 ) ] + O ( ε 3 ) , (323) V_{\mathrm{eff}}^{(\mathcal{N})}(t)=\varepsilon V_1\big|_{\mathcal{N}}+\varepsilon^2\left[V_2\big|_{\mathcal{N}}-\frac{1}{2}(\partial_A V_1)(H_\perp^{-1})^{AB}(\partial_B V_1)\right]+O(\varepsilon^3), \tag*{(323)} V eff ( N ) ( t ) = ε V 1 N + ε 2 [ V 2 N − 2 1 ( ∂ A V 1 ) ( H ⊥ − 1 ) A B ( ∂ B V 1 ) ] + O ( ε 3 ) , (323)
with all quantities on the right-hand side evaluated on N \mathcal{N} N .
The factor 1 / 2 1/2 1/2 follows from the competition between the quadratic energy cost associated with V 0 V_0 V 0 and the linear gain from V 1 V_1 V 1 , exactly as in the two-field example.
The result can be written intrinsically. The scalar metric defines the normal space
N q N = ( T q N ) ⊥ , N_q\mathcal{N}=(T_q\mathcal{N})^\perp, N q N = ( T q N ) ⊥ ,
and the transverse Hessian is the restriction of ∇ 2 V 0 \nabla^2V_0 ∇ 2 V 0 to this space. If d ⊥ V 1 d_\perp V_1 d ⊥ V 1 denotes the restriction of d V 1 dV_1 d V 1 to the normal directions, the backreaction term is
− 1 2 ⟨ d ⊥ V 1 , H ⊥ − 1 d ⊥ V 1 ⟩ . (324) -\frac{1}{2}\langle d_{\perp}V_1,H_{\perp}^{-1}d_{\perp}V_1\rangle. \tag*{(324)} − 2 1 ⟨ d ⊥ V 1 , H ⊥ − 1 d ⊥ V 1 ⟩ . (324)
If H ⊥ H_{\perp} H ⊥ is positive definite, this term is non-positive. Integrating out a stable transverse sector therefore lowers, or leaves unchanged, the induced potential at this order. No such sign statement follows when the transverse Hessian is indefinite.
The first non-trivial lifting order
(323) gives a direct criterion for lifting. At first order, the induced potential is simply
V 1 = V 1 ∣ N . \mathcal{V}_1=V_1|_{\mathcal{N}}. V 1 = V 1 ∣ N .
The no-scale family is therefore lifted at order ε \varepsilon ε only when this restriction is non-constant along N \mathcal{N} N , or equivalently when
d N ( V 1 ∣ N ) ≠ 0 , (325) d_{\mathcal{N}}\left(V_1|_{\mathcal{N}}\right)\ne0, \tag*{(325)} d N ( V 1 ∣ N ) = 0 , (325)
where d N d_{\mathcal{N}} d N denotes the intrinsic differential along the vacuum manifold.
A deformation may thus be non-zero throughout the full scalar manifold and still leave the no-scale degeneracy intact at first order. A constant V 1 ∣ N V_1|_{\mathcal{N}} V 1 ∣ N can shift the vacuum energy without selecting any point on the former flat family.
More generally, write
V e f f ( N ) = ∑ k ≥ 1 ε k V k . (326) V_{\mathrm{eff}}^{(\mathcal{N})}=\sum_{k\ge1}\varepsilon^k\mathcal{V}_k. \tag*{(326)} V eff ( N ) = k ≥ 1 ∑ ε k V k . (326)
The first two coefficients are
V 1 = V 1 ∣ N , \mathcal{V}_1=V_1|_{\mathcal{N}}, V 1 = V 1 ∣ N ,
and
V 2 = V 2 ∣ N − 1 2 ⟨ d ⊥ V 1 , H ⊥ − 1 d ⊥ V 1 ⟩ . \mathcal{V}_2=V_2|_{\mathcal{N}}-\frac{1}{2}\langle d_{\perp}V_1,H_{\perp}^{-1}d_{\perp}V_1\rangle. V 2 = V 2 ∣ N − 2 1 ⟨ d ⊥ V 1 , H ⊥ − 1 d ⊥ V 1 ⟩ .
If the degeneracy is first lifted at some finite order, define
k ⋆ = min { k ≥ 1 ∣ d N V k ≠ 0 } . (327) k_{\star}=\min\{k\ge1\mid d_{\mathcal{N}}\mathcal{V}_k\ne0\}. \tag*{(327)} k ⋆ = min { k ≥ 1 ∣ d N V k = 0 } . (327)
The leading lifting potential is then
V l i f t = ε k ⋆ V k ⋆ + O ( ε k ⋆ + 1 ) . (328) V_{\mathrm{lift}}=\varepsilon^{k_{\star}}\mathcal{V}_{k_{\star}}+O(\varepsilon^{k_{\star}+1}). \tag*{(328)} V lift = ε k ⋆ V k ⋆ + O ( ε k ⋆ + 1 ) . (328)
The relevant perturbative order is not the first order at which the microscopic action changes. It is the first order at which the induced potential on the former vacuum manifold becomes non-constant.
Lifting is not yet stabilization
Once the former flat direction acquires a non-constant potential, the problem is not finished. The resulting potential may be monotonic, in which case the modulus runs toward the boundary of field space.
A finite scale can be selected only if the induced potential has a critical point t ⋆ t_\star t ⋆ ,
d N V e f f ( N ) ∣ t ⋆ = 0. (329) \left.d_{\mathcal{N}}V_{\mathrm{eff}}^{(\mathcal{N})}\right|_{t_\star}=0. \tag*{(329)} d N V eff ( N ) t ⋆ = 0. (329)
Local stabilization additionally requires the Hessian along the relevant former flat directions to be positive.
The transverse spectrum must also remain acceptable. If H ⊥ > 0 H_\perp>0 H ⊥ > 0 initially, sufficiently small perturbations preserve transverse positivity by continuity. If the unperturbed family already has a transverse instability, the perturbative construction still computes the induced tangential potential, but it does not cure that instability.
Thus a non-constant potential can produce a runaway, a critical point can be a saddle, and a locally stable minimum can have positive, zero or negative vacuum energy. Lifting, scale selection, stability and vacuum characterization are separate questions.
What can generate the lifting potential?
The general criterion allows very different microscopic effects to be treated in the same way.
At tree level, additional gauging data can deform the Scherk–Schwarz embedding tensor, schematically Θ = Θ S S + Δ Θ \Theta=\Theta_{\mathrm{SS}}+\Delta\Theta Θ = Θ SS + ΔΘ . The dressed T T T -tensor, fermion shifts and scalar potential introduced in Question 13 then change accordingly. The resulting gauged supergravity can remain locally supersymmetric while its no-scale family is lifted.
Internal curvature and background fluxes can generate modulus-dependent terms after dimensional reduction and Weyl rescaling. Localized interactions of the type discussed in Question 9 can contribute directly to the scalar potential or modify fields whose expectation values enter it.
There is also a possibility intrinsic to the Scherk–Schwarz construction. If the twist generator M \mathcal{M} M acts non-trivially on a bosonic background field, M Φ 0 ≠ 0 \mathcal{M}\Phi_0\ne0 M Φ 0 = 0 , then the corresponding covariant internal derivative does not vanish in the periodic-field description. Internal kinetic energy can then contribute to the reduced scalar potential.
The microscopic origins of these effects differ, but the test for lifting is always the same: compute the induced potential on N \mathcal{N} N and determine whether it varies along the former flat directions.
The homogeneous N = 1 N=1 N = 1 construction of Question 14 provides a useful laboratory for applying the general criterion.
Let
K = K 0 + ε Δ K , W = W 0 + ε Δ W , (330) K=K_0+\varepsilon\Delta K,\qquad W=W_0+\varepsilon\Delta W, \tag*{(330)} K = K 0 + ε Δ K , W = W 0 + ε Δ W , (330)
with
K 0 = − log Y , Y ( λ r ) = λ 3 Y ( r ) , ∂ a W 0 = 0. K_0=-\log Y,\qquad Y(\lambda r)=\lambda^3Y(r),\qquad\partial_a W_0=0. K 0 = − log Y , Y ( λ r ) = λ 3 Y ( r ) , ∂ a W 0 = 0.
The unperturbed modulus sector obeys the no-scale identity, while the deformed F F F -term potential remains
V F = e K ( K a b ˉ D a W D b W ‾ − 3 ∣ W ∣ 2 ) . (331) V_F=e^K\left(K^{a\bar b}D_aW\overline{D_bW}-3|W|^2\right). \tag*{(331)} V F = e K ( K a b ˉ D a W D b W − 3∣ W ∣ 2 ) . (331)
Expanding it as
V F = V 0 + ε V 1 + ε 2 V 2 + ⋯ V_F=V_0+\varepsilon V_1+\varepsilon^2V_2+\cdots V F = V 0 + ε V 1 + ε 2 V 2 + ⋯
reduces the problem to the general analysis above. At first order, the no-scale family is lifted only if
d N ( V 1 ∣ N ) ≠ 0. (332) d_{\mathcal N}\left(V_1\big|_{\mathcal N}\right)\ne0. \tag*{(332)} d N ( V 1 N ) = 0. (332)
A non-zero Δ K \Delta K Δ K or Δ W \Delta W Δ W therefore need not generate a force along the original no-scale family at the same perturbative order. What matters is the corresponding coefficient of the scalar potential after restriction to N \mathcal N N . If the first-order restriction is constant, one proceeds to the next order, including the transverse backreaction derived above.
This N = 1 \mathcal N=1 N = 1 sector is used only as a convenient supergravity realization of the criterion. In string compactifications, modulus stabilization is generally a coupled problem involving several fields, flux and quantization conditions, localized sources and global consistency constraints.
Rigid monodromy versus a dynamical spectral phase
There is another way in which a supersymmetry-breaking scale can become dynamical, logically distinct from stabilizing the compactification radius.
Question 9 showed that localized data can modify the phase entering a Kaluza–Klein spectrum. For the commuting boundary problem considered there, one may write
ω e f f = ω r i g i d + Ω 0 ( φ ) + Ω π ( φ ) π ( m o d 1 ) . (333) \omega_{\mathrm{eff}}=\omega_{\mathrm{rigid}}+\frac{\Omega_0(\varphi)+\Omega_\pi(\varphi)}{\pi}\pmod{1}. \tag*{(333)} ω eff = ω rigid + π Ω 0 ( φ ) + Ω π ( φ ) ( mod 1 ) . (333)
Here ω r i g i d \omega_{\mathrm{rigid}} ω rigid is the microscopic Scherk–Schwarz monodromy parameter, whereas the boundary angles depend on localized dynamical fields φ I \varphi^I φ I .
The microscopic parameter is fixed, d ω r i g i d = 0 d\omega_{\mathrm{rigid}}=0 d ω rigid = 0 , while d ω e f f d\omega_{\mathrm{eff}} d ω eff can be non-zero when these localized fields vary. The microscopic monodromy can therefore remain unchanged even though the phase seen by a particular Kaluza–Klein spectrum becomes dynamical.
Determining the compactification scale M K K ( r ) M_{\mathrm{KK}}(r) M KK ( r ) and determining the fields entering ω e f f ( φ ) \omega_{\mathrm{eff}}(\varphi) ω eff ( φ ) are distinct dynamical problems. They need not be solved at the same order or by the same mechanism.
A measured spectral shift constrains ω e f f \omega_{\mathrm{eff}} ω eff , not necessarily the microscopic ω r i g i d \omega_{\mathrm{rigid}} ω rigid . As emphasized in Question 9, compactifications with the same effective spectral phase can still differ in their localized dynamics and in their complete supersymmetry-breaking structure.
Quantum scale selection
The no-scale family may survive all classical effects retained in a given effective theory. The first non-trivial potential can then arise quantum mechanically.
Question 14 showed that the classical spectrum can take the form
m i ( r , ω ) = M K K ( r ) μ i ( ω ) . m_i(r,\omega) = M_{\mathrm{KK}}(r)\,\mu_i(\omega). m i ( r , ω ) = M KK ( r ) μ i ( ω ) .
These same masses enter the quantum functional determinant. Schematically,
V 1 -loop ( r , ω ) ∼ 1 2 Str ∑ n ∫ d d p ( 2 π ) d log [ p 2 + m n 2 ( r , ω ) ] . (334) V_{1\text{-loop}}(r,\omega) \sim\frac{1}{2}\operatorname{Str}\sum_n \int\frac{\mathrm{d}^d p}{(2\pi)^d}\log\left[p^2 + m_n^2(r,\omega)\right]. \tag*{(334)} V 1 -loop ( r , ω ) ∼ 2 1 Str n ∑ ∫ ( 2 π ) d d d p log [ p 2 + m n 2 ( r , ω ) ] . (334)
The spectrum whose scale slides classically can therefore generate a quantum potential for the modulus controlling that scale.
The local and non-local pieces of this quantum potential, their ultraviolet sensitivity, and the relevant supertraces are separate questions. They will be treated in the quantum part of the review.
What has actually been fixed?
Even after a stable critical point has been found, one must still identify which quantities have actually been selected dynamically. The compactification scale may be fixed while an effective spectral phase remains dynamical, or conversely; other flat directions may also survive.
The amount of preserved supersymmetry and the vacuum energy are independent questions. A fixed non-zero gravitino mass does not determine the former: as in Question 13, preserved supersymmetry is found from the kernel of the complete fermionic Killing system. The value of the potential and the full scalar Hessian are likewise required to characterize the resulting vacuum as de Sitter, Minkowski or anti-de Sitter and to determine its local stability.
The no-scale supergravity framework and the relation between flat moduli, vanishing classical vacuum energy and a sliding gravitino mass are discussed in the early literature on no-scale models and Scherk–Schwarz supergravity; see, for example, Refs. [41, 63, 64].
The perturbative treatment used here is formulated as a deformation problem for a Morse–Bott [65] vacuum manifold. The leading lifting potential is obtained by restricting the perturbation to the unperturbed vacuum family, while integrating out non-degenerate transverse directions generates the second-order backreaction term in Eq. (323) .
Corrections capable of lifting no-scale directions arise in many forms in effective supergravity and string compactifications, including additional gaugings, flux and curvature effects, localized interactions and quantum corrections. Their detailed realization is strongly model dependent and will not be reviewed here.
Take-home message
V e f f ( N ) = ε V 1 ∣ N + O ( ε 2 ) , V_{\mathrm{eff}}^{(\mathcal{N})}=\left.\varepsilon V_1\right|_{\mathcal{N}}+O(\varepsilon^2), V eff ( N ) = ε V 1 ∣ N + O ( ε 2 ) ,
so the no-scale direction is lifted at first order only if this restriction is non-constant.
y ⋆ A = − ε ( H − 1 ) A B ∂ B V 1 ∣ N + O ( ε 2 ) , y_\star^A=-\varepsilon\left(H^{-1}\right)^{AB}\left.\partial_B V_1\right|_{\mathcal{N}}+O(\varepsilon^2), y ⋆ A = − ε ( H − 1 ) A B ∂ B V 1 ∣ N + O ( ε 2 ) ,
and their backreaction contributes to the tangential potential one order later.
Additional gaugings, curvature, fluxes, localized interactions, bosonic Scherk–Schwarz charges and quantum corrections can all contribute to the induced potential. In an N = 1 N=1 N = 1 description, the same criterion applies to corrections of K K K and W W W .
This naturally leads to the next question:
How are inequivalent Scherk–Schwarz twists classified, and how much supersymmetry can they preserve?
How are inequivalent Scherk–Schwarz twists classified, and how much supersymmetry can they preserve?
Essential idea
A Scherk–Schwarz twist is not fundamentally a number. It is a representation of the global identification group of the compactification,
ρ : Γ ⟶ G , \rho:\Gamma\longrightarrow G, ρ : Γ ⟶ G ,
where Γ \Gamma Γ contains the identifications of the internal space and G G G is the group of admissible automorphisms of the parent theory.
Two representations related by conjugation describe the same gluing data written in different field bases. The first classical classification problem is therefore
X ( Γ , G ) = Hom ( Γ , G ) / G , \mathcal{X}(\Gamma,G)=\operatorname{Hom}(\Gamma,G)/G, X ( Γ , G ) = Hom ( Γ , G ) / G ,
with the usual mathematical refinements understood when necessary.
Two different questions can then be asked at each point of this representation space. The residual internal symmetry is
G r e s i d u a l ( ρ ) = Z G ( ρ ( Γ ) ) , G_{\mathrm{residual}}(\rho)=Z_G\bigl(\rho(\Gamma)\bigr), G residual ( ρ ) = Z G ( ρ ( Γ ) ) ,
the centralizer of the complete image of ρ \rho ρ , while the preserved supersymmetries form
S p r e s e r v e d ( ρ ) = ⋂ γ ∈ Γ ker [ ρ S ( γ ) − 1 ] , S_{\mathrm{preserved}}(\rho)=\bigcap_{\gamma\in\Gamma}\ker\bigl[\rho_S(\gamma)-1\bigr], S preserved ( ρ ) = γ ∈ Γ ⋂ ker [ ρ S ( γ ) − 1 ] ,
where ρ S \rho_S ρ S is the action of the same gluing data on the supersymmetry parameters.
These two quantities probe the same global data in different representations. Residual symmetry is controlled by the adjoint action on the internal symmetry algebra, while preserved supersymmetry is controlled by the representation carried by the supercharges. Their enhancement loci therefore need not coincide.
The representation space of twists
Let Γ \Gamma Γ be the group generated by the global identifications of the compactification. Familiar examples are Γ = Z \Gamma=\mathbb{Z} Γ = Z for a circle, Γ = Z r \Gamma=\mathbb{Z}^r Γ = Z r for an r r r -torus, and Γ = D ∞ \Gamma=D_\infty Γ = D ∞ for the orbifold S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 discussed in Question 8.
Let G G G denote the group of transformations of the parent theory that may consistently be used as gluing data, in the sense explained in Question 2. A rigid Scherk–Schwarz twist is then a homomorphism
ρ : Γ ⟶ G . (335) \rho:\Gamma\longrightarrow G. \tag*{(335)} ρ : Γ ⟶ G . (335)
For each γ ∈ Γ \gamma\in\Gamma γ ∈ Γ , the element ρ ( γ ) \rho(\gamma) ρ ( γ ) tells us how the fields are reidentified under that global identification, while the homomorphism condition ρ ( γ 1 γ 2 ) = ρ ( γ 1 ) ρ ( γ 2 ) \rho(\gamma_1\gamma_2)=\rho(\gamma_1)\rho(\gamma_2) ρ ( γ 1 γ 2 ) = ρ ( γ 1 ) ρ ( γ 2 ) ensures compatibility with composition.
Two representations related by a global change of field basis,
ρ ′ ( γ ) = g ρ ( γ ) g − 1 , g ∈ G , (336) \rho'(\gamma)=g\rho(\gamma)g^{-1}, \qquad g\in G, \tag*{(336)} ρ ′ ( γ ) = g ρ ( γ ) g − 1 , g ∈ G , (336)
describe the same rigid gluing data. The corresponding classical representation space is therefore
X ( Γ , G ) = Hom ( Γ , G ) / G . (337) \mathcal{X}(\Gamma,G)=\operatorname{Hom}(\Gamma,G)/G. \tag*{(337)} X ( Γ , G ) = Hom ( Γ , G ) / G . (337)
For compact G G G , this orbit-space description is sufficient for the semisimple holonomies considered below. We shall use X ( Γ , G ) \mathcal{X}(\Gamma,G) X ( Γ , G ) without dwelling on refinements needed in more general representation varieties.
Two qualifications are useful. First, Eq. (337) classifies rigid global gluing data. A dynamically generated effective phase ω e f f \omega_{\mathrm{eff}} ω eff , of the type discussed in Question 15, is not a new point of this space unless the rigid representation ρ \rho ρ itself changes.
Second, conjugation is only the minimal equivalence relation. Mapping-class transformations and exact dualities of the complete compactification may identify further points of X ( Γ , G ) \mathcal{X}(\Gamma,G) X ( Γ , G ) .
The circle: conjugacy classes and alcoves
For a circle, Γ = Z \Gamma=\mathbb{Z} Γ = Z , so a homomorphism is completely determined by the image U ∈ G U\in G U ∈ G of one generator. Hence
X ( Z , G ) = G / Ad G , \mathcal{X}(\mathbb{Z},G)=G/\operatorname{Ad}G, X ( Z , G ) = G / Ad G ,
the space of conjugacy classes of G G G .
If G G G is compact and connected, every element is conjugate to an element of a maximal torus T T T , and two elements of T T T are conjugate precisely when they are related by the Weyl group W \mathcal{W} W . Thus
X ( Z , G ) ≃ T / W . (338) \mathcal{X}(\mathbb{Z},G)\simeq T/\mathcal{W}. \tag*{(338)} X ( Z , G ) ≃ T / W . (338)
Choose generators H \mathbf{H} H of a Cartan subalgebra and write
U = exp ( 2 π i ω ⋅ H ) . (339) U=\exp(2\pi i\,\omega\cdot\mathbf{H}). \tag*{(339)} U = exp ( 2 π i ω ⋅ H ) . (339)
The coordinate ω \omega ω is not unique. The exponential map identifies ω ∼ ω + λ \omega\sim\omega+\lambda ω ∼ ω + λ with λ ∈ Λ c o c h a r \lambda\in\Lambda_{\mathrm{cochar}} λ ∈ Λ cochar , where Λ c o c h a r \Lambda_{\mathrm{cochar}} Λ cochar is the cocharacter lattice of G G G , while the Weyl group further identifies ω ∼ w ( ω ) \omega\sim w(\omega) ω ∼ w ( ω ) , with w ∈ W w\in\mathcal{W} w ∈ W . These two identifications generate the extended affine Weyl action,
W e x t = W ⋉ Λ c o c h a r . (340) \mathcal{W}_{\mathrm{ext}}=\mathcal{W}\ltimes\Lambda_{\mathrm{cochar}}. \tag*{(340)} W ext = W ⋉ Λ cochar . (340)
For a simply connected semisimple group, the familiar affine Weyl group is obtained using the coroot lattice. The cocharacter lattice is the appropriate object when the global form of G G G is kept explicit.
A fundamental domain of this action provides an alcove description of the inequivalent circle twists. The lattice identification is the group-theoretic origin of the Kaluza–Klein relabelling freedom encountered in Questions 6 and 7.
For G = S U ( 2 ) G = SU(2) G = S U ( 2 ) , with
U ( ω ) = e 2 π i ω σ 3 , U(\omega) = e^{2\pi i\omega\sigma_3}, U ( ω ) = e 2 π iω σ 3 ,
one has ω ∼ ω + 1 \omega\sim\omega+ 1 ω ∼ ω + 1 and ω ∼ − ω \omega\sim-\omega ω ∼ − ω . A fundamental alcove is therefore
0 ≤ ω ≤ 1 2 . (341) 0 \leq\omega\leq\frac{1}{2}. \tag*{(341)} 0 ≤ ω ≤ 2 1 . (341)
Its endpoints are the untwisted point and the half-twist.
Several cycles and commuting tuples
For an r r r -torus, Γ = Z r \Gamma= \mathbb{Z}^r Γ = Z r , so a homomorphism is specified by r r r commuting elements of G G G :
Hom ( Z r , G ) = { ( U 1 , … , U r ) ∣ U i U j = U j U i } . (342) \operatorname{Hom}(\mathbb{Z}^r,G) = \left\{(U_1,\ldots,U_r)\mid U_iU_j = U_jU_i\right\}. \tag*{(342)} Hom ( Z r , G ) = { ( U 1 , … , U r ) ∣ U i U j = U j U i } . (342)
The quotient by G G G acts by simultaneous conjugation,
( U 1 , … , U r ) ∼ ( g U 1 g − 1 , … , g U r g − 1 ) . (343) (U_1,\ldots,U_r) \sim(gU_1g^{-1},\ldots,gU_rg^{-1}). \tag*{(343)} ( U 1 , … , U r ) ∼ ( g U 1 g − 1 , … , g U r g − 1 ) . (343)
In the component where all monodromies can be conjugated into a common maximal torus, one may write
U i = exp ( 2 π i ω i ⋅ H ) . U_i = \exp(2\pi i\,\omega_i\cdot H). U i = exp ( 2 π i ω i ⋅ H ) .
Each ω i \omega_i ω i has its own cocharacter-lattice identifications, while one common Weyl transformation w ∈ W w \in W w ∈ W acts simultaneously on the whole tuple.
The toral component T r / W T^r/W T r / W need not exhaust Hom ( Z r , G ) / G \operatorname{Hom}(\mathbb{Z}^r,G)/G Hom ( Z r , G ) / G . For compact groups with non-trivial global structure, non-toral commuting tuples can occur, and the representation space may possess additional components. This richer structure will matter only when the global form of the symmetry group and several independent cycles are simultaneously relevant.
The compactification also carries a marking of its cycles. A change of primitive basis Λ ∈ G L ( r , Z ) \Lambda\in GL(r,\mathbb{Z}) Λ ∈ G L ( r , Z ) acts on π 1 ( T r ) = Z r \pi_1(T^r) = \mathbb{Z}^r π 1 ( T r ) = Z r , and hence on the monodromy tuple. Such a transformation is a physical redundancy only when the complete torus data, including its lattice, metric and background fields, are transformed as well. Acting only on the monodromies while keeping the rest of the compactification fixed need not define an equivalence.
Residual symmetry as a centralizer
Once a rigid twist ρ \rho ρ has been chosen, an internal transformation survives the compactification only if it is compatible with every gluing operation. The residual symmetry group is therefore
G r e s i d u a l ( ρ ) = Z G ( ρ ( Γ ) ) , (344) G_{\mathrm{residual}}(\rho) = Z_G(\rho(\Gamma)), \tag*{(344)} G residual ( ρ ) = Z G ( ρ ( Γ )) , (344)
with
Z G ( ρ ( Γ ) ) = { g ∈ G ∣ g ρ ( γ ) = ρ ( γ ) g ∀ γ ∈ Γ } . Z_G(\rho(\Gamma)) = \{g \in G \mid g\rho(\gamma) = \rho(\gamma)g \ \forall\gamma\in\Gamma\}. Z G ( ρ ( Γ )) = { g ∈ G ∣ g ρ ( γ ) = ρ ( γ ) g ∀ γ ∈ Γ } .
The important point is that this is the centralizer of the complete image ρ ( Γ ) \rho(\Gamma) ρ ( Γ ) , not necessarily the centralizer of one derived element. For a circle there is only one independent generator, so the distinction is invisible. A regular element in the interior of a Weyl alcove has centralizer equal to a maximal torus, while at special points the centralizer can enlarge.
On orbifolds the distinction becomes essential. A translation monodromy can have a large centralizer even when the complete set of orbifold generators has only a smaller common centralizer. Residual-symmetry enhancement is therefore naturally described as a jump in the conjugacy type of Z G ( ρ ( Γ ) ) Z_G(\rho(\Gamma)) Z G ( ρ ( Γ )) .
Preserved supersymmetry as a fixed subspace
Supersymmetry is tested on the same global gluing data, but in a different representation. Let S S S be the vector space of supersymmetry parameters and let
ρ S : Γ ⟶ G L ( S ) \rho_S:\Gamma\longrightarrow GL(S) ρ S : Γ ⟶ G L ( S )
be the representation induced by ρ \rho ρ on that space.
For a flat background in which the global gluing is the only non-trivial obstruction to a Killing spinor, a supersymmetry parameter survives precisely when it is invariant under every element of Γ \Gamma Γ :
S p r e s e r v e d ( ρ ) = S ρ S ( Γ ) = ⋂ γ ∈ Γ ker [ ρ S ( γ ) − 1 ] . (345) S_{\mathrm{preserved}}(\rho)=S^{\rho_S(\Gamma)}=\bigcap_{\gamma\in\Gamma}\ker[\rho_S(\gamma)-1]. \tag*{(345)} S preserved ( ρ ) = S ρ S ( Γ ) = γ ∈ Γ ⋂ ker [ ρ S ( γ ) − 1 ] . (345)
This is the criterion of Question 3, now written for an arbitrary identification group. Partial breaking corresponds to 0 < dim S p r e s e r v e d < dim S 0<\dim S_{\mathrm{preserved}}<\dim S 0 < dim S preserved < dim S .
For commuting monodromies lying in a common maximal torus, let λ A \lambda_A λ A be the weight of a supercharge Q A Q_A Q A . Then
U i Q A = e 2 π i ⟨ λ A , ω i ⟩ Q A . U_iQ_A=e^{2\pi i\langle\lambda_A,\omega_i\rangle}Q_A. U i Q A = e 2 π i ⟨ λ A , ω i ⟩ Q A .
The supercharge survives precisely when
e 2 π i ⟨ λ A , ω i ⟩ = 1 for every independent cycle i . (346) e^{2\pi i\langle\lambda_A,\omega_i\rangle}=1 \qquad\text{for every independent cycle }i. \tag*{(346)} e 2 π i ⟨ λ A , ω i ⟩ = 1 for every independent cycle i . (346)
Several cycles can therefore reduce the preserved supersymmetry because the final invariant subspace is the intersection of the fixed subspaces of all monodromies.
Enhancement loci
At each point [ ρ ] ∈ X ( Γ , G ) [\rho]\in\mathcal{X}(\Gamma,G) [ ρ ] ∈ X ( Γ , G ) , two quantities are naturally associated with the same gluing data:
[ ρ ] ⟼ ( Z G ( ρ ( Γ ) ) , S ρ S ( Γ ) ) . (347) [\rho]\longmapsto\left(Z_G(\rho(\Gamma)),S^{\rho_S(\Gamma)}\right). \tag*{(347)} [ ρ ] ⟼ ( Z G ( ρ ( Γ )) , S ρ S ( Γ ) ) . (347)
The first probes the adjoint action on the internal symmetry algebra; the second probes the representation carried by the supercharges. The two enhancement loci can differ.
Consider first
G = S U ( 2 ) R , U ( ω ) = e 2 π i ω σ 3 , G=SU(2)_R,\qquad U(\omega)=e^{2\pi i\omega\sigma_3}, G = S U ( 2 ) R , U ( ω ) = e 2 π iω σ 3 ,
acting on a doublet of supercharges.
For 0 < ω < 1 / 2 0 < \omega< 1/2 0 < ω < 1/2 , the element is regular and Z S U ( 2 ) ( U ) = U ( 1 ) Z_{SU(2)}(U) = U(1) Z S U ( 2 ) ( U ) = U ( 1 ) . Its eigenvalues on the doublet are e ± 2 π i ω e^{\pm2\pi i\omega} e ± 2 π iω , so no supercharge survives for generic ω \omega ω in this interval.
At ω = 0 \omega= 0 ω = 0 , U = 1 U = 1 U = 1 : the centralizer becomes all of S U ( 2 ) SU(2) S U ( 2 ) , and the complete supercharge doublet is invariant.
At the other endpoint, ω = 1 / 2 \omega= 1/2 ω = 1/2 , one has U = − 1 U = -1 U = − 1 . Since this element lies in the centre,
Z S U ( 2 ) ( U ) = S U ( 2 ) , Z_{SU(2)}(U) = SU(2), Z S U ( 2 ) ( U ) = S U ( 2 ) ,
but U − 1 = − 21 U - 1 = -2\mathbf{1} U − 1 = − 2 1 is invertible on the supercharge doublet. Thus no supercharge survives. Maximal enhancement of the residual internal symmetry can coexist with complete supersymmetry breaking.
This circle result should not be confused with the half-twist orbifold of Question 11. On S 1 / Z 2 S^{1}/\mathbb{Z}_{2} S 1 / Z 2 , the rigid datum is the full representation of D ∞ D_{\infty} D ∞ , equivalently the pair ( P 0 , P π ) (P_{0}, P_{\pi}) ( P 0 , P π ) . The translation monodromy U = P π P 0 U = P_{\pi}P_{0} U = P π P 0 is only a derived element. At the half-twist, U = − 1 U = -1 U = − 1 still has centralizer S U ( 2 ) R SU(2)_{R} S U ( 2 ) R , whereas the symmetry preserving the complete orbifold gluing data is
Z S U ( 2 ) R ( P 0 ) ∩ Z S U ( 2 ) R ( P π ) = U ( 1 ) R . (348) Z_{SU(2)_{R}}(P_{0}) \cap Z_{SU(2)_{R}}(P_{\pi}) = U(1)_{R}. \tag*{(348)} Z S U ( 2 ) R ( P 0 ) ∩ Z S U ( 2 ) R ( P π ) = U ( 1 ) R . (348)
This is the common centralizer found in Question 11, and it is this U ( 1 ) R U(1)_{R} U ( 1 ) R that protects the Dirac gravitino.
The converse phenomenon can also occur. Suppose the relevant R R R -symmetry contains S U ( 4 ) R SU(4)_{R} S U ( 4 ) R , with the supercharges in the fundamental representation, and consider
U = diag ( 1 , e i α , e i β , e − i ( α + β ) ) . (349) U = \operatorname{diag}\left(1, e^{i\alpha}, e^{i\beta}, e^{-i(\alpha+\beta)}\right). \tag*{(349)} U = diag ( 1 , e i α , e i β , e − i ( α + β ) ) . (349)
For generic α \alpha α and β \beta β such that all four eigenvalues are distinct, U U U is regular and
Z S U ( 4 ) ( U ) = U ( 1 ) 3 . Z_{SU(4)}(U) = U(1)^{3}. Z S U ( 4 ) ( U ) = U ( 1 ) 3 .
Nevertheless, the first component of the fundamental has eigenvalue one, so one supercharge survives. Partial supersymmetry can therefore occur without any enhancement of the residual internal symmetry.
The four possibilities can be summarized as
Table 3.
Twist space is stratified independently by centralizer type and by the dimension of the preserved-supercharge subspace.
Beyond the flat monodromy problem
The fixed-subspace formula Eq. (345) is exact only when the global gluing is the sole non-trivial obstruction to a Killing spinor.
In a curved or fluxed compactification, the supersymmetry parameters must also satisfy the complete local Killing system, schematically
D m ϵ = 0 , A ( ϕ , F , … ) ϵ = 0 , (350) \mathcal{D}_m\epsilon= 0,\qquad\mathcal{A}(\phi,F,\ldots)\epsilon= 0, \tag*{(350)} D m ϵ = 0 , A ( ϕ , F , … ) ϵ = 0 , (350)
where D m \mathcal{D}_m D m is the relevant supercovariant derivative and the second equation denotes collectively the algebraic fermionic supersymmetry conditions generated by background scalars, fluxes and other bosonic fields.
The preserved supersymmetries are then the global solutions of this complete system. Equation (345) is the flat special case in which the only non-trivial global information is the representation ρ S \rho_S ρ S .
Rigid orbifold gluing fits naturally into the same language. For S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 , Γ = D ∞ \Gamma= D_\infty Γ = D ∞ , with ρ ( r 0 ) = P 0 \rho(r_0) = P_0 ρ ( r 0 ) = P 0 and ρ ( r π ) = P π \rho(r_\pi) = P_\pi ρ ( r π ) = P π . The relations
U = P π P 0 , P 0 U P 0 = U − 1 U = P_\pi P_0,\qquad P_0 U P_0 = U^{-1} U = P π P 0 , P 0 U P 0 = U − 1
follow directly from the defining relations of D ∞ D_\infty D ∞ .
Localized boundary dynamics of the type discussed in Question 9 adds further matching conditions and need not be encoded by the rigid representation ρ \rho ρ alone.
Physical equivalences beyond conjugation
The quotient by conjugation in Eq. (337) is the minimal identification associated with a global change of field basis. It need not be the complete physical quotient.
Schematically, the physical parameter space takes the form
M p h y s i c a l = { consistent rigid compactification and gluing data } { physical equivalences of the complete theory } . (351) \mathcal{M}_{\mathrm{physical}} =
\frac{\{\text{consistent rigid compactification and gluing data}\}}
{\{\text{physical equivalences of the complete theory}\}}. \tag*{(351)} M physical = { physical equivalences of the complete theory } { consistent rigid compactification and gluing data } . (351)
The denominator may contain mapping-class transformations, further global identifications and exact quantum dualities.
Care is required not to quotient twice by the same redundancy. A gauge transformation at a base point already acts on holonomies by simultaneous conjugation and is therefore contained in Hom ( Γ , G ) / G \operatorname{Hom}(\Gamma,G)/G Hom ( Γ , G ) / G . Likewise, in a Cartan description, the cocharacter-lattice identifications are already part of the extended affine Weyl action.
A further transformation should be imposed only when it represents an additional physical equivalence of the complete compactification. This is theory dependent: a symmetry of the classical lower-dimensional supergravity may fail to survive in the ultraviolet completion, whereas an exact string duality may identify backgrounds that appear different in a particular effective description.
What a spectrum can and cannot determine
A Kaluza–Klein spectrum probes the twist only in the representation carried by the states whose masses are being measured.
For a sector transforming in a representation R \mathcal{R} R , the spectral data contain information such as
Spec R ( ρ ( γ ) ) . \operatorname{Spec}_{\mathcal{R}}\bigl(\rho(\gamma)\bigr). Spec R ( ρ ( γ ) ) .
This need not determine the full class [ ρ ] ∈ X ( Γ , G ) [\rho] \in\mathcal{X}(\Gamma,G) [ ρ ] ∈ X ( Γ , G ) .
Two twists can, for example, have identical eigenvalues on one multiplet while acting differently on other fields. That multiplet then has the same Kaluza–Klein shifts in the two compactifications, even though the complete representations ρ : Γ → G \rho:\Gamma\to G ρ : Γ → G are inequivalent.
A spectrum measured in one representation therefore cannot in general reconstruct the complete twist class, still less the global compactification, its localized sectors or its ultraviolet equivalences.
There is an additional subtlety from Question 15: the phase extracted from a spectrum may be a dynamical ω e f f \omega_{\mathrm{eff}} ω eff , rather than the rigid parameter ω r i g i d \omega_{\mathrm{rigid}} ω rigid entering ρ \rho ρ .
The spectral problem and the classification problem are therefore closely related, but they contain different amounts of information.
The classification of flat holonomies on circles and higher tori is closely related to standard results on compact Lie groups and flat bundles. For a circle, conjugacy classes in a compact connected group are described by a maximal torus modulo the Weyl group, with the global form of the group encoded in the relevant cocharacter lattice; see, for example, Refs. [66, 67].
Representation spaces Hom ( Γ , G ) / G \operatorname{Hom}(\Gamma,G)/G Hom ( Γ , G ) / G and moduli spaces of flat connections are discussed, for example, in Refs. [68, 69]. For several commuting holonomies, the representation space can contain components not captured by the simple toral description; related structures are discussed in Refs. [70, 71].
In Scherk–Schwarz compactification, the same mathematics classifies rigid global gluing data. Preserved supersymmetry is obtained by applying the gluing representation to the supersymmetry parameters and taking the common invariant subspace, while the residual internal symmetry is the centralizer of the full image of the identification group. Partial supersymmetry breaking by non-trivial string monodromies is discussed, for example, in Ref. [72]. The distinction between the translation monodromy and the complete orbifold data was illustrated explicitly in Questions 8 and 11.
For compactifications with several cycles or non-trivial global forms of the symmetry group, the topology of commuting-tuple spaces can be richer than the toral description T r / W T^r/W T r / W . Exact mapping classes and ultraviolet dualities may further identify points of the classical representation space, but such identifications depend on the complete theory.
Take-home message
ρ : Γ → G , X ( Γ , G ) = Hom ( Γ , G ) / G , \rho: \Gamma\to G,\qquad\mathcal{X}(\Gamma,G) = \operatorname{Hom}(\Gamma,G)/G, ρ : Γ → G , X ( Γ , G ) = Hom ( Γ , G ) / G ,
where the quotient implements global changes of field basis.
X ( Z , G ) ≃ T / W . \mathcal{X}(\mathbb{Z},G) \simeq T/W. X ( Z , G ) ≃ T / W .
In Cartan coordinates, the global form of G G G enters through the cocharacter lattice and the corresponding affine or extended affine Weyl identifications.
G r e s i d u a l = Z G ( ρ ( Γ ) ) , S p r e s e r v e d = ⋂ γ ∈ Γ ker [ ρ S ( γ ) − 1 ] . G_{\mathrm{residual}} = Z_G\bigl(\rho(\Gamma)\bigr),\qquad S_{\mathrm{preserved}} = \bigcap_{\gamma\in\Gamma}\ker\bigl[\rho_S(\gamma)-1\bigr]. G residual = Z G ( ρ ( Γ ) ) , S preserved = γ ∈ Γ ⋂ ker [ ρ S ( γ ) − 1 ] .
They probe different representations of the same gluing data and their enhancement loci need not coincide.
Conjugation is only the minimal equivalence relation, and a spectrum in one representation does not in general reconstruct the complete twist class. Mapping classes, exact dualities and dynamical effective phases must be treated separately when present.
This naturally leads to the next question:
What do supertrace sum rules reveal about the ultraviolet softness of Scherk–Schwarz breaking?
What do supertrace sum rules reveal about the ultraviolet softness of Scherk–Schwarz breaking?
Essential idea
Supertraces organize the ultraviolet sensitivity of a supersymmetry-breaking spectrum, but spontaneous supersymmetry breaking alone does not imply
Str M 2 = 0. \operatorname{Str} M^2 = 0. Str M 2 = 0.
A finite lower-dimensional supergravity can therefore display ordinary quadratic sensitivity.
A genuine Scherk–Schwarz compactification contains additional information: the complete Kaluza–Klein tower is correlated by one global monodromy. This correlation is not captured by an arbitrary finite mass matrix.
The distinction becomes manifest after Poisson resummation. The Kaluza–Klein sum is rewritten as a sum over winding sectors. The zero-winding sector contains the local short-distance contribution and is independent of the rigid Scherk–Schwarz phase. All dependence on the twist lies in non-zero winding sectors, which require propagation around the compact direction and are exponentially suppressed at short proper time.
The twist-dependent one-loop contribution is therefore finite and non-local, of Casimir type. For D D D non-compact dimensions,
V D ( ω ) = 1 R D f D ( ω ) , f D ( ω ) ∝ ∑ w = 1 ∞ 1 − cos ( 2 π w ω ) w D + 1 . V_D(\omega) = \frac{1}{R^D} f_D(\omega), \qquad f_D(\omega) \propto\sum_{w=1}^{\infty} \frac{1-\cos(2\pi w\omega)}{w^{D+1}}. V D ( ω ) = R D 1 f D ( ω ) , f D ( ω ) ∝ w = 1 ∑ ∞ w D + 1 1 − cos ( 2 π w ω ) .
For a fixed order-one twist this scales parametrically as m 3 / 2 D m_{3/2}^{D} m 3/2 D when m 3 / 2 ∼ ω / R m_{3/2} \sim\omega/R m 3/2 ∼ ω / R , although the full ω \omega ω -dependence must be retained when the twist is small.
The ultraviolet softness of Scherk–Schwarz breaking is thus a property of the complete higher-dimensional spectral problem and of the locality of the breaking datum. It does not imply the absence of ordinary local counterterms in the parent theory or, on orbifolds, at fixed loci.
Why begin with supertraces
For a four-dimensional theory with field-dependent masses m s ( ϕ ) m_s(\phi) m s ( ϕ ) , a cutoff representation of the one-loop effective potential has schematically the form
V 1 -loop = c 0 Λ 4 Str M 0 + c 2 Λ 2 Str M 2 + c 4 Str [ M 4 log M 2 Λ 2 ] + ⋯ (352) V_{1\text{-loop}} = c_0 \Lambda^4 \operatorname{Str} M^0 + c_2 \Lambda^2 \operatorname{Str} M^2 + c_4 \operatorname{Str}\left[M^4 \log\frac{M^2}{\Lambda^2}\right] + \cdots \tag*{(352)} V 1 -loop = c 0 Λ 4 Str M 0 + c 2 Λ 2 Str M 2 + c 4 Str [ M 4 log Λ 2 M 2 ] + ⋯ (352)
The numerical coefficients depend on the regulator and normalization conventions, but the supertraces organize which combinations of the spectrum can accompany the various ultraviolet sensitivities.
For states labelled by s s s , with spin J s J_s J s and mass m s m_s m s , we define
Str M 2 k = ∑ s ( − 1 ) 2 J s ( 2 J s + 1 ) m s 2 k . (353) \operatorname{Str} M^{2k} = \sum_s (-1)^{2J_s}(2J_s+1)m_s^{2k}. \tag*{(353)} Str M 2 k = s ∑ ( − 1 ) 2 J s ( 2 J s + 1 ) m s 2 k . (353)
The sign distinguishes bosons from fermions, while 2 J s + 1 2J_s+1 2 J s + 1 counts the physical polarizations in the usual four-dimensional massive-state notation.
An exactly supersymmetric massive spectrum has equal bosonic and fermionic contributions within complete supermultiplets. In particular, Str M 0 = 0 \operatorname{Str} M^0=0 Str M 0 = 0 . Once supersymmetry is broken, however, the fate of higher supertraces depends on the theory and on which spectrum is included in the trace.
A generic broken supergravity need not have Str M 2 = 0 \operatorname{Str} M^2=0 Str M 2 = 0
Spontaneous supersymmetry breaking by itself does not force the quadratic supertrace to vanish. In a four-dimensional N = 1 N=1 N = 1 supergravity with broken supersymmetry and vanishing classical vacuum energy, one may write schematically
Str M 2 = 2 Q m 3 / 2 2 , (354) \operatorname{Str} M^2 = 2Qm_{3/2}^2, \tag*{(354)} Str M 2 = 2 Q m 3/2 2 , (354)
where Q Q Q is a dimensionless quantity determined by the scalar and gauge kinetic geometry and by the supersymmetry-breaking direction. There is no general reason for Q Q Q to vanish.
Particular no-scale or string-derived supergravities can satisfy additional relations for which Q = 0 Q=0 Q = 0 , and hence Str M 2 = 0 \operatorname{Str} M^2=0 Str M 2 = 0 . Examples of this type occur in the string-effective supergravities discussed in Ref. [41].
This condition should not be confused with the no-scale cancellation of Question 14. The latter is a statement about the classical scalar potential; the former is a statement about a mass spectrum and its ultraviolet supertrace.
The Scherk–Schwarz problem adds another ingredient: the spectrum is an infinite correlated Kaluza–Klein tower rather than an arbitrary finite collection of split multiplets.
Consider the elementary circle spectrum with N N N real bosonic degrees of freedom and the same number of real fermionic degrees of freedom,
m B , n 2 = n 2 R 2 , m F , n 2 = ( n + ω ) 2 R 2 , n ∈ Z . (355) m_{B,n}^2=\frac{n^2}{R^2},\qquad m_{F,n}^2=\frac{(n+\omega)^2}{R^2},\qquad n\in\mathbb{Z}. \tag*{(355)} m B , n 2 = R 2 n 2 , m F , n 2 = R 2 ( n + ω ) 2 , n ∈ Z . (355)
The masses at different Kaluza–Klein levels are not independent: the complete fermionic tower is obtained from the bosonic one by the single spectral displacement n ↦ n + ω n\mapsto n+\omega n ↦ n + ω .
One may formally write
Str M 2 k = N ∑ n ∈ Z [ ( n R ) 2 k − ( n + ω R ) 2 k ] . (356) \operatorname{Str} M^{2k}=N\sum_{n\in\mathbb{Z}}\left[\left(\frac{n}{R}\right)^{2k}-\left(\frac{n+\omega}{R}\right)^{2k}\right]. \tag*{(356)} Str M 2 k = N n ∈ Z ∑ [ ( R n ) 2 k − ( R n + ω ) 2 k ] . (356)
This is not an ordinary convergent sum, so its ultraviolet meaning depends on how the infinite tower is regulated. A cutoff imposed directly on the integer label n n n , for example, need not respect the spectral-flow relation between the shifted and unshifted towers.
The appropriate ultraviolet statement is instead obtained from the complete one-loop spectral determinant.
The one-loop determinant and Poisson resummation
Begin in four non-compact dimensions and subtract the supersymmetric value at ω = 0 \omega= 0 ω = 0 :
V 1 ( ω ) = N 2 ∑ n ∈ Z ∫ d 4 p ( 2 π ) 4 [ log ( p 2 + n 2 R 2 ) − log ( p 2 + ( n + ω ) 2 R 2 ) ] . (357) V_1(\omega) = \frac{N}{2}\sum_{n\in\mathbb{Z}}\int\frac{d^4p}{(2\pi)^4}\left[\log\left(p^2+\frac{n^2}{R^2}\right)-\log\left(p^2+\frac{(n+\omega)^2}{R^2}\right)\right]. \tag*{(357)} V 1 ( ω ) = 2 N n ∈ Z ∑ ∫ ( 2 π ) 4 d 4 p [ log ( p 2 + R 2 n 2 ) − log ( p 2 + R 2 ( n + ω ) 2 ) ] . (357)
Here p p p is Euclidean momentum and the factor 1 / 2 1/2 1/2 is the usual one-loop factor for a real degree of freedom.
Using
log A = − ∫ 0 ∞ d t t e − t A , ∫ d 4 p ( 2 π ) 4 e − t p 2 = 1 16 π 2 t 2 , \log A=-\int_0^\infty\frac{dt}{t}e^{-tA},\qquad\int\frac{d^4p}{(2\pi)^4}e^{-tp^2}=\frac{1}{16\pi^2t^2}, log A = − ∫ 0 ∞ t d t e − t A , ∫ ( 2 π ) 4 d 4 p e − t p 2 = 16 π 2 t 2 1 ,
gives
V 1 ( ω ) = − N 32 π 2 ∫ 0 ∞ d t t 3 ∑ n ∈ Z [ e − t n 2 / R 2 − e − t ( n + ω ) 2 / R 2 ] . (358) V_1(\omega)=-\frac{N}{32\pi^2}\int_0^\infty\frac{dt}{t^3}\sum_{n\in\mathbb{Z}}\left[e^{-tn^2/R^2}-e^{-t(n+\omega)^2/R^2}\right]. \tag*{(358)} V 1 ( ω ) = − 32 π 2 N ∫ 0 ∞ t 3 d t n ∈ Z ∑ [ e − t n 2 / R 2 − e − t ( n + ω ) 2 / R 2 ] . (358)
The Schwinger parameter t t t has the usual interpretation: the short-proper-time region t → 0 t\to0 t → 0 probes ultraviolet physics.
Now use Poisson resummation,
∑ n ∈ Z e − t ( n + ω ) 2 / R 2 = R π t ∑ w ∈ Z e − π 2 R 2 w 2 / t e 2 π i w ω . (359) \sum_{n\in\mathbb{Z}}e^{-t(n+\omega)^2/R^2}=R\sqrt{\frac{\pi}{t}}\sum_{w\in\mathbb{Z}}e^{-\pi^2R^2w^2/t}e^{2\pi i w\omega}. \tag*{(359)} n ∈ Z ∑ e − t ( n + ω ) 2 / R 2 = R t π w ∈ Z ∑ e − π 2 R 2 w 2 / t e 2 π i w ω . (359)
The integer w w w is the winding number in the resummed description. Substitution gives
V 1 ( ω ) = − N R π 32 π 2 ∑ w ∈ Z ( 1 − e 2 π i w ω ) ∫ 0 ∞ d t t − 7 / 2 e − π 2 R 2 w 2 / t . (360) V_1(\omega)=-\frac{NR\sqrt{\pi}}{32\pi^2}\sum_{w\in\mathbb{Z}}\left(1-e^{2\pi i w\omega}\right)\int_0^\infty dt\,t^{-7/2}e^{-\pi^2R^2w^2/t}. \tag*{(360)} V 1 ( ω ) = − 32 π 2 N R π w ∈ Z ∑ ( 1 − e 2 π i w ω ) ∫ 0 ∞ d t t − 7/2 e − π 2 R 2 w 2 / t . (360)
Why the twist-dependent ultraviolet part vanishes
The zero-winding sector is special. For w = 0 w=0 w = 0 , the factor multiplying the proper-time integral is exactly
1 − e 2 π i w ω = 0. (361) 1-e^{2\pi i w\omega}=0. \tag*{(361)} 1 − e 2 π i w ω = 0. (361)
The complete zero-winding contribution to the ω \omega ω -dependent difference therefore vanishes before the proper-time integral is performed. There is no limiting 0 × ∞ 0\times\infty 0 × ∞ ambiguity.
This has a direct geometric interpretation. A zero-winding path does not traverse the non-contractible circle. It probes only a local neighbourhood of the higher-dimensional spacetime and cannot detect a global Scherk–Schwarz holonomy.
For w ≠ 0 w \ne0 w = 0 , the integrand instead contains
exp ( − π 2 R 2 w 2 t ) \exp\left(-\frac{\pi^2 R^2 w^2}{t}\right) exp ( − t π 2 R 2 w 2 )
which is exponentially suppressed as t → 0 t \to0 t → 0 . Every non-zero winding sector is ultraviolet finite.
The conclusion is precise: the divergent local part of the one-loop effective action is insensitive to the rigid Scherk–Schwarz phase, while the phase-dependent contribution comes entirely from finite non-zero winding sectors. This does not say that the parent higher-dimensional theory has no ultraviolet divergences. It says that its local divergent coefficients cannot depend on a global flat monodromy.
The general-D D D Casimir law
The same argument works in any number D D D of non-compact dimensions. The Euclidean Gaussian integral is
∫ d D p ( 2 π ) D e − t p 2 = 1 ( 4 π t ) D / 2 . (362) \int\frac{d^D p}{(2\pi)^D} e^{-t p^2} = \frac{1}{(4\pi t)^{D/2}}. \tag*{(362)} ∫ ( 2 π ) D d D p e − t p 2 = ( 4 π t ) D /2 1 . (362)
After Poisson resummation, a non-zero winding sector contains
∫ 0 ∞ d t t − ( D + 3 ) / 2 e − π 2 R 2 w 2 / t = Γ ( D + 1 2 ) π D + 1 R D + 1 ∣ w ∣ D + 1 . (363) \int_0^\infty dt\, t^{-(D+3)/2} e^{-\pi^2 R^2 w^2/t}
= \frac{\Gamma\left(\frac{D+1}{2}\right)}{\pi^{D+1}R^{D+1}|w|^{D+1}}. \tag*{(363)} ∫ 0 ∞ d t t − ( D + 3 ) /2 e − π 2 R 2 w 2 / t = π D + 1 R D + 1 ∣ w ∣ D + 1 Γ ( 2 D + 1 ) . (363)
Combining this result with the factor produced by Poisson resummation and pairing w w w with − w -w − w gives
V D ( ω ) = − N Γ ( D + 1 2 ) ( 4 π ) D / 2 π D + 1 / 2 1 R D ∑ w = 1 ∞ 1 − cos ( 2 π w ω ) w D + 1 . (364) V_D(\omega) = -\frac{N\Gamma\left(\frac{D+1}{2}\right)}{(4\pi)^{D/2}\pi^{D+1/2}}\frac{1}{R^D}\sum_{w=1}^{\infty}\frac{1-\cos(2\pi w\omega)}{w^{D+1}}. \tag*{(364)} V D ( ω ) = − ( 4 π ) D /2 π D + 1/2 N Γ ( 2 D + 1 ) R D 1 w = 1 ∑ ∞ w D + 1 1 − cos ( 2 π w ω ) . (364)
The sign corresponds to the convention used above, with the bosonic tower unshifted and the fermionic tower shifted.
The structurally important result is
V D ( ω ) = 1 R D f D ( ω ) , (365) V_D(\omega) = \frac{1}{R^D} f_D(\omega), \tag*{(365)} V D ( ω ) = R D 1 f D ( ω ) , (365)
with f D f_D f D a finite periodic dimensionless function. For a fixed non-zero twist of order unity, m 3 / 2 ∼ ω / R m_{3/2} \sim\omega/R m 3/2 ∼ ω / R , and therefore V D ∼ m 3 / 2 D V_D \sim m_{3/2}^D V D ∼ m 3/2 D parametrically.
This last statement is not an identity in ω \omega ω . For small ω \omega ω and D > 2 D>2 D > 2 ,
1 − cos ( 2 π w ω ) = 2 π 2 w 2 ω 2 + O ( ω 4 ) , 1-\cos(2\pi w\omega)=2\pi^2w^2\omega^2+O(\omega^4), 1 − cos ( 2 π w ω ) = 2 π 2 w 2 ω 2 + O ( ω 4 ) ,
so that
V D ( ω ) ∼ ω 2 R D ∑ w = 1 ∞ 1 w D − 1 ∼ m 3 / 2 2 R D − 2 . (366) V_D(\omega)\sim\frac{\omega^2}{R^D}\sum_{w=1}^{\infty}\frac{1}{w^{D-1}}\sim\frac{m_{3/2}^2}{R^{D-2}}. \tag*{(366)} V D ( ω ) ∼ R D ω 2 w = 1 ∑ ∞ w D − 1 1 ∼ R D − 2 m 3/2 2 . (366)
Thus R − D f D ( ω ) R^{-D}f_D(\omega) R − D f D ( ω ) , rather than a pure power of the gravitino mass, is the invariant statement.
The explicit four-dimensional result
For D = 4 D=4 D = 4 , Eq. (364) reduces to
V 4 ( ω ) = − 3 N 64 π 6 R 4 ∑ w = 1 ∞ 1 − cos ( 2 π w ω ) w 5 . (367) V_4(\omega)=-\frac{3N}{64\pi^6 R^4}\sum_{w=1}^{\infty}\frac{1-\cos(2\pi w\omega)}{w^5}. \tag*{(367)} V 4 ( ω ) = − 64 π 6 R 4 3 N w = 1 ∑ ∞ w 5 1 − cos ( 2 π w ω ) . (367)
The result is finite, periodic under ω → ω + 1 \omega\to\omega+1 ω → ω + 1 , vanishes at the supersymmetric point, and has the expected R − 4 R^{-4} R − 4 Casimir scaling.
If desired, the winding sum can be expressed in terms of polylogarithms:
V 4 ( ω ) = − 3 N 64 π 6 R 4 [ ζ ( 5 ) − 1 2 Li 5 ( e 2 π i ω ) − 1 2 Li 5 ( e − 2 π i ω ) ] . (368) V_4(\omega)=-\frac{3N}{64\pi^6R^4}\left[\zeta(5)-\frac{1}{2}\operatorname{Li}_5\left(e^{2\pi i\omega}\right)-\frac{1}{2}\operatorname{Li}_5\left(e^{-2\pi i\omega}\right)\right]. \tag*{(368)} V 4 ( ω ) = − 64 π 6 R 4 3 N [ ζ ( 5 ) − 2 1 Li 5 ( e 2 π iω ) − 2 1 Li 5 ( e − 2 π iω ) ] . (368)
This form is useful for analytic manipulations, but the winding sum makes the locality properties more transparent.
The Casimir interpretation
Poisson resummation has reorganized the determinant according to the distance explored in the compact direction. The zero-winding sector is local: it can be represented by the usual short-time heat-kernel coefficients of the higher-dimensional differential operator. Non-zero winding sectors probe the global compact geometry.
This is the same physical separation encountered from the boundary scattering picture in Question 9. There the non-local part arose from propagation over a finite compact distance and back; here it appears as non-zero winding around the circle. The two descriptions are different organizations of the same Casimir physics.
The periodic-field description of Question 10 provides a complementary view. A rigid Scherk–Schwarz monodromy can be represented by a flat connection: its local curvature vanishes, while its gauge-invariant information is carried by the holonomy around the circle. Local ultraviolet counterterms can depend on local field strengths, couplings and background invariants, but they cannot intrinsically depend on a flat holonomy that can only be measured by transporting a state around a non-contractible cycle.
Thus the winding, Casimir and Wilson-line descriptions give three ways of recognizing the same locality property.
What a finite Kaluza–Klein truncation misses
The need for the complete tower can be seen directly. Retain only the levels ∣ n ∣ ≤ N max |n|\leq N_{\max} ∣ n ∣ ≤ N m a x . Then
Str ∣ n ∣ ≤ N max M 2 = N R 2 ∑ n = − N max N max [ n 2 − ( n + ω ) 2 ] = − N ( 2 N max + 1 ) ω 2 R 2 . (369) \operatorname{Str}_{|n|\leq N_{\max}}M^2=\frac{N}{R^2}\sum_{n=-N_{\max}}^{N_{\max}}\left[n^2-(n+\omega)^2\right]
=-\frac{N(2N_{\max}+1)\omega^2}{R^2}. \tag*{(369)} Str ∣ n ∣ ≤ N m a x M 2 = R 2 N n = − N m a x ∑ N m a x [ n 2 − ( n + ω ) 2 ] = − R 2 N ( 2 N m a x + 1 ) ω 2 . (369)
This diverges linearly as $N_{\max}\to\infty.
The divergence does not reproduce the ultraviolet behaviour of the complete higher-dimensional determinant. A symmetric cutoff on the integer label truncates the shifted and unshifted towers at the same values of n n n , thereby breaking the spectral-flow relation precisely at the edges of the sum.
A finite Kaluza–Klein truncation can therefore have supertraces that look ultraviolet sensitive even though the twist-dependent part of the complete spectral determinant is finite.
This also explains why the coefficient Q Q Q of a finite lower-dimensional supergravity and the finite Scherk–Schwarz Casimir potential answer different questions. The former is a local supertrace in the truncated effective theory; the latter is obtained from the complete correlated Kaluza–Klein spectrum.
Rigid monodromy and effective spectral phases
Question 15 distinguished a rigid microscopic monodromy ω r i g i d \omega_{\mathrm{rigid}} ω rigid from an effective phase ω e f f ( φ ) \omega_{\mathrm{eff}}(\varphi) ω eff ( φ ) that can also depend on localized dynamical fields.
The ultraviolet argument above applies directly to the rigid global part: a contractible short-distance process cannot detect it. If the effective phase receives contributions from localized dynamical sectors, however, those sectors are local microscopic degrees of freedom and can renormalize local operators in the usual way.
Two theories can consequently display the same ω e f f \omega_{\mathrm{eff}} ω eff in a particular spectrum while having different ultraviolet structures. Equality of the spectral shift does not imply equality of the underlying local dynamics.
From the smooth circle to boundaries
A smooth circle has no distinguished local submanifold on which new counterterms can be supported. Orbifolds and intervals do.
A loop confined to an arbitrarily small neighbourhood of a fixed locus can detect the local projection, the local field content and the symmetries preserved there. It cannot, however, detect a distant boundary or the complete global Scherk–Schwarz holonomy.
The one-loop effective action can therefore be organized schematically as
Γ 1 -loop = Γ b u l k l o c a l + ∑ f Γ f l o c a l + Γ f i n i t e n o n l o c a l , (370) \Gamma_{1\text{-loop}} = \Gamma_{\mathrm{bulk}}^{\mathrm{local}} + \sum_f \Gamma_f^{\mathrm{local}} + \Gamma_{\mathrm{finite}}^{\mathrm{nonlocal}}, \tag*{(370)} Γ 1 -loop = Γ bulk local + f ∑ Γ f local + Γ finite nonlocal , (370)
where f f f labels fixed loci. The precise local operators depend on the field content, interactions and boundary conditions; the final term contains the genuinely non-local dependence on the complete compactification.
The ultraviolet softness of the global Scherk–Schwarz effect therefore does not eliminate local bulk or fixed-locus counterterms. It tells us instead that the divergent coefficients of such local operators cannot depend on global information that is invisible in an arbitrarily small neighbourhood. The allowed local divergences are the subject of the next question.
Supertrace sum rules have long been used to characterize ultraviolet sensitivity in spontaneously broken supersymmetric theories and supergravities. Additional relations in no-scale and string-derived effective supergravities can lead to cancellations such as Str M 2 = 0 \operatorname{Str} M^2 = 0 Str M 2 = 0 ; an example relevant to the present discussion can be found in [41].
For Scherk–Schwarz compactifications, proper-time methods and Poisson resummation make the ultraviolet softness of the twist-dependent contribution manifest: the local zero-winding sector is separated from the finite non-zero winding sectors that probe the global compactification. On orbifolds and intervals, additional local counterterms may be supported on fixed loci; these will be discussed in the next question.
Take-home message
Supertraces diagnose the ultraviolet sensitivity of a specified spectrum, but spontaneous supersymmetry breaking alone does not imply Str M 2 = 0 \operatorname{Str} M^2 = 0 Str M 2 = 0 . A finite lower-dimensional supergravity and the complete Scherk–Schwarz Kaluza–Klein tower need not have the same supertrace behaviour.
After Poisson resummation, the zero-winding sector contains the local short-distance contribution. Its twist-dependent part vanishes identically before the proper-time integral is performed.
All dependence on a rigid Scherk–Schwarz phase resides in non-zero winding sectors. They are exponentially suppressed at short proper time and give a finite non-local Casimir contribution.
For D D D non-compact dimensions,
V D ( ω ) = 1 R D f D ( ω ) . V_D(\omega) = \frac{1}{R^D} f_D(\omega). V D ( ω ) = R D 1 f D ( ω ) .
For a fixed order-one twist this scales parametrically as m 3 / 2 D m_{3/2}^D m 3/2 D , while the small-ω \omega ω behaviour must be extracted from the full periodic function f D f_D f D .
A cutoff on a finite set of Kaluza–Klein levels can destroy the spectral-flow organization of the tower and generate divergent supertraces that do not reproduce the ultraviolet behaviour of the complete higher-dimensional determinant.
The finiteness concerns the global twist-dependent contribution, not all ultraviolet divergences. Local bulk counterterms remain allowed, and orbifolds or intervals can additionally generate local fixed-locus counterterms.
This naturally leads to the next question:
Which ultraviolet divergences and local counterterms remain allowed?
Which ultraviolet divergences and local counterterms remain allowed?
Essential idea
The ultraviolet softness established in Question 17 concerns the genuinely non-local, Scherk–Schwarz-dependent part of the effective action. It does not imply that the compactified theory has no local ultraviolet divergences.
The correct organization is by support. At one loop, the short-proper-time expansion of the heat kernel separates local contributions supported in the regular bulk from those supported on fixed loci. Bulk coefficients renormalize bulk operators, while fixed-locus coefficients renormalize operators localized on the corresponding strata.
The global Scherk–Schwarz information belongs to a different part of the effective action. In an image or worldline representation, a group element can contribute a local ultraviolet term only where it has a fixed set. If it moves every point by a finite distance, its short-proper-time contribution is exponentially suppressed as exp [ − ℓ 2 / ( 4 t ) ] \exp[-\ell^2/(4t)] exp [ − ℓ 2 / ( 4 t )] and is ultraviolet finite.
For S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 , the identity therefore produces bulk local counterterms, the endpoint reflections produce fixed-point local counterterms, and non-zero translations produce finite winding or Casimir terms.
A divergent functional localized at one endpoint may depend on the local boundary operator there, but not jointly on independently variable data at a distant endpoint. Mixed dependence on disjoint fixed loci is consequently non-local and ultraviolet finite.
Schematically,
Γ ( 1 ) = Γ b u l k l o c a l + ∑ f Γ f l o c a l + Γ C a s i m i r n o n l o c a l . \Gamma^{(1)}=\Gamma_{\mathrm{bulk}}^{\mathrm{local}}+\sum_f\Gamma_f^{\mathrm{local}}+\Gamma_{\mathrm{Casimir}}^{\mathrm{nonlocal}}. Γ ( 1 ) = Γ bulk local + f ∑ Γ f local + Γ Casimir nonlocal .
The first terms may contain divergences and scheme-dependent finite pieces. The last contains the dependence on relative boundary data, transported holonomy and compactification size that cannot be altered by local counterterms.
The one-loop problem as a heat-kernel problem
Work in Euclidean signature on X = M D × K \mathcal{X}=\mathcal{M}_D\times K X = M D × K , where M D \mathcal{M}_D M D is the D D D -dimensional non-compact spacetime and K K K the compact internal space.
After gauge fixing, the quadratic fluctuation operators can be organized into Laplace-type blocks
Δ = − ( g M N ∇ M ∇ N + E ) (371) \Delta=-\left(g^{MN}\nabla_M\nabla_N+\mathcal{E}\right) \tag*{(371)} Δ = − ( g M N ∇ M ∇ N + E ) (371)
where ∇ M \nabla_M ∇ M contains the relevant spin and gauge connections and E E E is a local endomorphism of the field bundle. First-order fermionic operators may instead be treated through the corresponding squared Laplace-type operator, with ghosts, statistics and multiplicities included in the supertrace.
The one-loop effective action is formally
Γ ( 1 ) = − 1 2 ∫ 0 ∞ d t t Str e − t Δ . (372) \Gamma^{(1)}=-\frac{1}{2}\int_0^\infty\frac{dt}{t}\operatorname{Str}e^{-t\Delta}. \tag*{(372)} Γ ( 1 ) = − 2 1 ∫ 0 ∞ t d t Str e − t Δ . (372)
The ultraviolet region is the short-proper-time limit t → 0 t\to0 t → 0 .
Let the internal space be stratified into smooth pieces Σ \Sigma Σ , including the regular bulk and any fixed or singular loci. The heat trace then has the asymptotic expansion
Str e − t Δ ∼ ∑ Σ 1 ( 4 π t ) ( D + dim Σ ) / 2 ∑ n ≥ 0 t n / 2 A n , Σ . (373) \operatorname{Str}e^{-t\Delta}\sim\sum_{\Sigma}\frac{1}{(4\pi t)^{(D+\dim\Sigma)/2}}\sum_{n\geq0}t^{n/2}A_{n,\Sigma}. \tag*{(373)} Str e − t Δ ∼ Σ ∑ ( 4 π t ) ( D + d i m Σ ) /2 1 n ≥ 0 ∑ t n /2 A n , Σ . (373)
where each A n , Σ A_{n,\Sigma} A n , Σ is an integral over Σ ^ = M D × Σ \widehat{\Sigma}=M_D\times\Sigma Σ = M D × Σ of a local density constructed from the fluctuation operator, the background geometry and fields, and the local boundary or matching data.
These are the standard heat-kernel coefficients for manifolds with boundaries or singular strata [73, 74]. Hence the divergent part is a sum of local functionals supported on the individual strata,
Γ d i v ( 1 ) = ∑ Σ Γ Σ , d i v l o c a l . (374) \Gamma_{\mathrm{div}}^{(1)}=\sum_{\Sigma}\Gamma_{\Sigma,\mathrm{div}}^{\mathrm{local}}. \tag*{(374)} Γ div ( 1 ) = Σ ∑ Γ Σ , div local . (374)
Ultraviolet divergences are therefore stratum local .
The fixed-set principle
The geometric origin of this localization is especially transparent when K = K ~ / Γ K=\widetilde{K}/\Gamma K = K /Γ is a quotient of a covering space by a discrete identification group. Let ρ : Γ → G \rho:\Gamma\to G ρ : Γ → G denote the corresponding action on the field fibres, as in Question 16.
In an image representation, the internal heat trace has the form
Tr K , ρ e − t Δ = ∑ γ ∈ Γ ∫ F d d y tr [ ρ ( γ ) K K ~ ( t ; y , γ y ) ] , (375) \operatorname{Tr}_{K,\rho}e^{-t\Delta}=\sum_{\gamma\in\Gamma}\int_{\mathcal{F}}d^d y\ \operatorname{tr}\left[\rho(\gamma)K_{\widetilde{K}}(t;y,\gamma y)\right], \tag*{(375)} Tr K , ρ e − t Δ = γ ∈ Γ ∑ ∫ F d d y tr [ ρ ( γ ) K K ( t ; y , γ y ) ] , (375)
where F \mathcal{F} F is a fundamental domain.
At short proper time, the off-diagonal heat kernel behaves as
K K ~ ( t ; y , γ y ) ∼ e − d ( y , γ y ) 2 / ( 4 t ) ( 4 π t ) d / 2 ∑ k ≥ 0 t k a k ( y , γ y ) . (376) K_{\widetilde{K}}(t;y,\gamma y)\sim\frac{e^{-d(y,\gamma y)^2/(4t)}}{(4\pi t)^{d/2}}\sum_{k\geq0}t^k a_k(y,\gamma y). \tag*{(376)} K K ( t ; y , γ y ) ∼ ( 4 π t ) d /2 e − d ( y , γ y ) 2 / ( 4 t ) k ≥ 0 ∑ t k a k ( y , γ y ) . (376)
If γ \gamma γ has no fixed point and the action is properly discontinuous, its displacement has a strictly positive minimum ℓ γ \ell_\gamma ℓ γ on the fundamental domain. Its contribution is then suppressed by e − ℓ γ 2 / ( 4 t ) e^{-\ell_\gamma^2/(4t)} e − ℓ γ 2 / ( 4 t ) as t → 0 t\to0 t → 0 , and is ultraviolet finite.
If γ \gamma γ has a fixed set, the displacement vanishes there. The short-time integral is no longer exponentially suppressed, and expansion in the normal directions produces local heat-kernel coefficients supported on Fix ( γ ) \operatorname{Fix}(\gamma) Fix ( γ ) .
Thus fixed-point-free images contribute only to the finite non-local part, whereas images with fixed sets can generate local ultraviolet terms on those sets. This is the fixed-set form of the local/non-local Casimir decomposition.
The image decomposition on S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2
For the infinite dihedral group introduced in Question 8, take on the covering line t : y ↦ y + 2 π R t:y\mapsto y+2\pi R t : y ↦ y + 2 π R and r 0 : y ↦ − y r_0:y\mapsto-y r 0 : y ↦ − y , with ρ ( t ) = U \rho(t)=U ρ ( t ) = U and ρ ( r 0 ) = P 0 \rho(r_0)=P_0 ρ ( r 0 ) = P 0 . The second endpoint involution is P π = U P 0 P_\pi=UP_0 P π = U P 0 .
Every element of D ∞ D_\infty D ∞ is either t w t^w t w or t w r 0 t^wr_0 t w r 0 , with w ∈ Z w\in\mathbb{Z} w ∈ Z . For a flat scalar Laplacian, the image kernel on 0 ≤ y ≤ π R 0\leq y\leq\pi R 0 ≤ y ≤ π R can be written as
K ρ ( t ; y , y ′ ) = ∑ w ∈ Z [ U w K R ( t ; y , y ′ + 2 π R w ) + U w P 0 K R ( t ; y , − y ′ + 2 π R w ) ] . (377) K_\rho(t;y,y')=\sum_{w\in\mathbb{Z}}\left[U^wK_{\mathbb{R}}(t;y,y'+2\pi Rw)+U^wP_0K_{\mathbb{R}}(t;y,-y'+2\pi Rw)\right]. \tag*{(377)} K ρ ( t ; y , y ′ ) = w ∈ Z ∑ [ U w K R ( t ; y , y ′ + 2 π R w ) + U w P 0 K R ( t ; y , − y ′ + 2 π R w ) ] . (377)
On the diagonal, the two image displacements are
d ( y , t w y ) = 2 π R ∣ w ∣ , (378) d(y,t^wy)=2\pi R|w|, \tag*{(378)} d ( y , t w y ) = 2 π R ∣ w ∣ , (378)
d ( y , t w r 0 y ) = 2 ∣ y − π R w ∣ . (379) d(y,t^wr_0y)=2|y-\pi Rw|. \tag*{(379)} d ( y , t w r 0 y ) = 2∣ y − π R w ∣. (379)
The identity translation fixes every point and produces the bulk coefficients. Every non-zero translation has no fixed point and is suppressed by exp [ − π 2 R 2 w 2 / t ] \exp[-\pi^2R^2w^2/t] exp [ − π 2 R 2 w 2 / t ] ; these sectors contain the finite winding dependence on U w U^w U w .
The reflection r 0 r_0 r 0 fixes y = 0 y=0 y = 0 and carries P 0 P_0 P 0 , while r π = t r 0 r_\pi=tr_0 r π = t r 0 fixes y = π R y=\pi R y = π R and carries P π P_\pi P π . The bulk, fixed-point and winding contributions therefore arise directly from the three geometrically distinct classes of images.
For the flat scalar block, the leading short-time trace is
Tr I , ρ e t ∂ y 2 = π R 4 π t tr 1 + 1 4 tr ( P 0 + P π ) + O ( e − π 2 R 2 / t ) . (380) \operatorname{Tr}_{I,\rho}e^{t\partial_y^2}=\frac{\pi R}{\sqrt{4\pi t}}\operatorname{tr}1+\frac{1}{4}\operatorname{tr}(P_0+P_\pi)+O\left(e^{-\pi^2R^2/t}\right). \tag*{(380)} Tr I , ρ e t ∂ y 2 = 4 π t π R tr 1 + 4 1 tr ( P 0 + P π ) + O ( e − π 2 R 2 / t ) . (380)
For one scalar component, Neumann–Neumann conditions give the constant term + 1 / 2 +1/2 + 1/2 , Dirichlet–Dirichlet gives − 1 / 2 -1/2 − 1/2 , and mixed Dirichlet–Neumann gives zero. The first term is a bulk invariant, the constant term contains local endpoint data, and dependence requiring propagation over a finite distance lies beyond the local short-time expansion.
Local boundary data
The involutions P b P_b P b are only the simplest local boundary data. A standard mixed boundary problem for a Laplace-type operator is
Π − ( b ) Φ ∣ b = 0 , (381) \left.\Pi_-^{(b)}\Phi\right|_b=0, \tag*{(381)} Π − ( b ) Φ b = 0 , (381)
( ∇ n + S b ) Π + ( b ) Φ ∣ b = 0 , (382) \left.(\nabla_n+S_b)\Pi_+^{(b)}\Phi\right|_b=0, \tag*{(382)} ( ∇ n + S b ) Π + ( b ) Φ b = 0 , (382)
where ∇ n \nabla_n ∇ n is the outward normal derivative and S b S_b S b a local endomorphism. The projectors are complementary; for a pure orbifold parity,
Π ± ( b ) = 1 2 ( 1 ± P b ) . \Pi_{\pm}^{(b)} = \frac{1}{2}(1 \pm P_b). Π ± ( b ) = 2 1 ( 1 ± P b ) .
The elementary even Neumann condition corresponds to S b = 0 S_b = 0 S b = 0 .
Localized masses and interactions can modify S b S_b S b or produce more general matching conditions, as discussed in Question 9. We denote collectively by B b = ( Π − ( b ) , Π + ( b ) , S b , … ) \mathcal{B}_b = (\Pi_{-}^{(b)}, \Pi_{+}^{(b)}, S_b, \ldots) B b = ( Π − ( b ) , Π + ( b ) , S b , … ) the local data defining the endpoint problem.
Boundary heat-kernel coefficients are local polynomials in the germ of ( Δ , B b ) (\Delta, \mathcal{B}_b) ( Δ , B b ) . They may involve the induced metric, intrinsic and extrinsic curvature, E E E , the curvature Ω M N = [ ∇ M , ∇ N ] \Omega_{MN} = [\nabla_M, \nabla_N] Ω M N = [ ∇ M , ∇ N ] of the bundle connection, the projectors, S b S_b S b , and their local covariant derivatives.
They cannot involve independent data located at a disjoint endpoint.
The support principle
Let Σ \Sigma Σ be the support of a local counterterm. Its divergent coefficient is determined by a finite jet of the fluctuation operator and of the local matching data on that support. Schematically,
Γ Σ , d i v l o c a l = ∫ Σ L Σ , d i v [ j N ( Δ , B Σ ) ] . (383) \Gamma_{\Sigma,\mathrm{div}}^{\mathrm{local}} = \int_{\Sigma} \mathcal{L}_{\Sigma,\mathrm{div}}\left[j^N(\Delta,\mathcal{B}_{\Sigma})\right]. \tag*{(383)} Γ Σ , div local = ∫ Σ L Σ , div [ j N ( Δ , B Σ ) ] . (383)
For a bulk counterterm, only bulk germs enter. For a counterterm at y = 0 y = 0 y = 0 , the coefficient may depend on B 0 \mathcal{B}_0 B 0 , but not on an independently variable boundary operator B π \mathcal{B}_{\pi} B π at the other endpoint. A radion field may of course enter if it is itself part of the local field content. What cannot enter a divergent coefficient is information whose definition requires comparing separated regions, such as a transported product of independent endpoint data.
A disjoint-support finiteness criterion
Let B 0 \mathcal{B}_0 B 0 and B π \mathcal{B}_{\pi} B π denote independently variable data at the two endpoints. Locality implies
Γ d i v = Γ b u l k l o c a l + Γ 0 l o c a l [ B 0 ] + Γ π l o c a l [ B π ] . (384) \Gamma_{\mathrm{div}} = \Gamma_{\mathrm{bulk}}^{\mathrm{local}} + \Gamma_0^{\mathrm{local}}[\mathcal{B}_0] + \Gamma_{\pi}^{\mathrm{local}}[\mathcal{B}_{\pi}]. \tag*{(384)} Γ div = Γ bulk local + Γ 0 local [ B 0 ] + Γ π local [ B π ] . (384)
There can therefore be no connected divergent term depending jointly on the two disjoint supports:
δ 2 Γ d i v δ B 0 δ B π = 0. (385) \frac{\delta^2 \Gamma_{\mathrm{div}}}{\delta\mathcal{B}_0 \, \delta\mathcal{B}_{\pi}} = 0. \tag*{(385)} δ B 0 δ B π δ 2 Γ div = 0. (385)
The second functional derivative of a local functional is supported on the diagonal, whereas the two variations above have disjoint supports. Consequently,
δ 2 Γ ( 1 ) δ B 0 δ B π = δ 2 Γ n o n l o c a l ( 1 ) δ B 0 δ B π (386) \frac{\delta^2 \Gamma^{(1)}}{\delta\mathcal{B}_0 \, \delta\mathcal{B}_{\pi}} = \frac{\delta^2 \Gamma_{\mathrm{nonlocal}}^{(1)}}{\delta\mathcal{B}_0 \, \delta\mathcal{B}_{\pi}} \tag*{(386)} δ B 0 δ B π δ 2 Γ ( 1 ) = δ B 0 δ B π δ 2 Γ nonlocal ( 1 ) (386)
is ultraviolet finite, apart from possible infrared singularities.
This is the analogue of the interaction part of a Casimir energy: individual surface self-energies may be divergent and scheme-dependent, while a response that necessarily correlates separated supports belongs to the finite non-local sector.
A related effect appears explicitly in heterotic orbifold amplitudes. Twisted states localized at fixed points renormalize amplitudes involving Kaluza–Klein exchange differently from purely untwisted contributions because localized interactions need not conserve internal KK momentum [75]. This provides an early concrete example of localized radiative effects; the general support argument above does not rely on the model-specific resummation used there.
Local symmetries on each stratum
The support principle also determines which symmetries constrain a local counterterm. Let B b \mathcal{B}_b B b be the complete local boundary operator at an endpoint b b b . Its local stabilizer in the parent internal symmetry group is
G b l o c = Stab G ( B b ) . (387) G_b^{\mathrm{loc}} = \operatorname{Stab}_G(\mathcal{B}_b). \tag*{(387)} G b loc = Stab G ( B b ) . (387)
For a pure orbifold boundary condition this reduces to G b l o c = Z G ( P b ) G_b^{\mathrm{loc}} = Z_G(P_b) G b loc = Z G ( P b ) , the centralizer encountered in Questions 11 and 16. Additional localized masses, kinetic terms or interactions can reduce this stabilizer further.
Let S b l o c S_b^{\mathrm{loc}} S b loc denote the supersymmetry parameters satisfying the local boundary projector and all local algebraic boundary Killing conditions at b b b . Assuming these symmetries are non-anomalous and the regulator preserves them, a divergent density supported at b b b must respect the corresponding local internal and supersymmetry constraints.
The local supersymmetry spaces at the two endpoints need not agree. Global supersymmetry may vanish because they fail to match after transport,
S g l o b a l = S 0 l o c ∩ U ( π R , 0 ) − 1 S π l o c = { 0 } . (388) S_{\mathrm{global}} = S_0^{\mathrm{loc}} \cap U(\pi R,0)^{-1}S_\pi^{\mathrm{loc}} = \{0\}. \tag*{(388)} S global = S 0 loc ∩ U ( π R , 0 ) − 1 S π loc = { 0 } . (388)
A divergence at y = 0 y=0 y = 0 is nevertheless constrained by the supersymmetry locally realized there. It cannot detect its misalignment with the supersymmetry preserved at y = π R y=\pi R y = π R , because that comparison requires finite propagation between the two endpoints.
This is the ultraviolet counterpart of the distinction between local endpoint supersymmetry and global Killing-spinor compatibility made in Questions 8 and 9.
When can a boundary divergence depend on the twist?
The answer must be phrased in terms of local boundary data rather than in terms of a coordinate called ω \omega ω .
In one description, a continuous family of compactifications may be represented by a rotating endpoint involution P π = P π ( ω ) P_\pi= P_\pi(\omega) P π = P π ( ω ) . A counterterm localized at y = π R y=\pi R y = π R may then depend on P π ( ω ) P_\pi(\omega) P π ( ω ) , because the local boundary problem itself has changed. This is ordinary sensitivity to a local boundary condition, not ultraviolet sensitivity to a distant holonomy.
In another trivialization, part of the same compactification data may be represented by a flat bulk connection together with transformed endpoint operators. The local heat-kernel coefficients depend on the local equivalence class of ( Δ , B b ) (\Delta,\mathcal{B}_b) ( Δ , B b ) , not on how the same physical data are distributed between a bulk connection and an endpoint matrix.
Thus, if ω \omega ω varies while the local germ at a support remains fixed and only the global flat transport changes, the divergent coefficient there cannot depend on ω \omega ω . If varying ω \omega ω changes the local boundary operator itself, the local counterterms can change accordingly.
Boundary rotations and the non-local determinant
For the commuting quadratic boundary problem of Question 9, the spectrum can take the form
m n = n + ω e f f R , ω e f f = ω r i g i d + Θ 0 + Θ π π . (389) m_n = \frac{n+\omega_{\mathrm{eff}}}{R}, \qquad\omega_{\mathrm{eff}} = \omega_{\mathrm{rigid}}+\frac{\Theta_0+\Theta_\pi}{\pi}. \tag*{(389)} m n = R n + ω eff , ω eff = ω rigid + π Θ 0 + Θ π . (389)
If the boundary data affect the quadratic problem only through this uniform shift, without adding extra propagating fields, momentum-dependent form factors or additional eigenvalue branches, the non-local determinant is obtained from the smooth-circle result of Question 17 by replacing ω \omega ω with ω e f f \omega_{\mathrm{eff}} ω eff .
This statement concerns only the non-local spectral determinant. The local counterterms still depend separately on the microscopic operators B 0 \mathcal{B}_0 B 0 and B π \mathcal{B}_\pi B π . Two theories with the same ω e f f \omega_{\mathrm{eff}} ω eff can share the same quadratic Casimir dependence while having different local renormalizations, localized degrees of freedom and Goldstino sectors.
Counterterms versus radiatively transmitted soft terms
The same lower-dimensional operator can receive both a locally renormalized contribution and a finite non-local Scherk–Schwarz contribution. For an operator O b \mathcal{O}_b O b supported on a boundary, write schematically
c O b p h y s = c O b l o c a l ( μ ) + c O b S S ( R , U , B 0 , B π ) . (390) c^{\mathrm{phys}}_{\mathcal{O}_b}=c^{\mathrm{local}}_{\mathcal{O}_b}(\mu)+c^{\mathrm{SS}}_{\mathcal{O}_b}(R,U,\mathcal{B}_0,\mathcal{B}_\pi). \tag*{(390)} c O b phys = c O b local ( μ ) + c O b SS ( R , U , B 0 , B π ) . (390)
The first term contains ultraviolet matching information and depends on the renormalization prescription. The second is generated by propagation that detects the global Scherk–Schwarz configuration and is ultraviolet finite. A calculation of the finite term c O b S S c^{\mathrm{SS}}_{\mathcal{O}_b} c O b SS does not determine the full physical coefficient whenever the corresponding local counterterm is allowed by the exact local symmetries.
A Scherk–Schwarz-induced soft term is therefore predicted without additional ultraviolet input only when the corresponding local counterterm is forbidden, related to other couplings by an exact local symmetry, or fixed by an independent renormalization condition.
If the local term is forbidden while the global compactification breaks the protecting symmetry non-locally, the leading contribution can be a finite transmitted soft term. This is the mechanism studied in the next question. If instead the local counterterm is allowed, the finite Scherk–Schwarz contribution remains a calculable threshold, but the total coefficient is not a UV-independent prediction of the compactification alone.
Heat-kernel methods on manifolds with boundaries and singular strata, including mixed Dirichlet–Robin conditions and local boundary invariants, are reviewed in Refs. [73, 74].
The generation of counterterms localized at orbifold fixed planes by bulk quantum loops was exhibited explicitly in Ref. [76]. Related one-loop bulk and brane renormalization effects in gauge theories on S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 were analysed in Ref. [77].
An early heterotic-orbifold example in which fixed-point localized twisted states affect the renormalization of Kaluza–Klein exchange amplitudes, reflecting the loss of internal KK-momentum conservation at localized interactions, can be found in Ref. [75].
Take-home message
For S 1 / Z 2 S^{1}/\mathbb{Z}_{2} S 1 / Z 2 , the identity generates bulk terms, the two endpoint reflections generate fixed-point terms, and non-zero translations generate finite winding/Casimir contributions.
δ 2 Γ d i v δ B 0 δ B π = 0 , \frac{\delta^{2}\Gamma_{\mathrm{div}}}{\delta\mathcal{B}_{0}\,\delta\mathcal{B}_{\pi}}=0, δ B 0 δ B π δ 2 Γ div = 0 ,
so a response that necessarily correlates two disjoint endpoints is ultraviolet finite.
A finite Scherk–Schwarz contribution and a local counterterm are different pieces of the effective theory. Their sum is a UV-independent prediction only when the local term is forbidden, symmetry related, or fixed independently.
This naturally leads to the next question:
How is Scherk–Schwarz supersymmetry breaking radiatively transmitted to boundary fields?
How is Scherk–Schwarz supersymmetry breaking radiatively transmitted to boundary fields?
Essential idea
A field localized on a boundary has no Kaluza–Klein tower of its own and does not wind around the compact dimension. It can nevertheless feel Scherk–Schwarz supersymmetry breaking through bulk propagators whose spectrum and boundary wave functions depend on the global compactification data.
The cleanest benchmark is the pure anti-periodic point on S 1 / Z 2 S^1/\mathbb{Z}_2 S 1 / Z 2 , with ω = 1 2 \omega= \frac{1}{2} ω = 2 1 and M 0 = M π = 0 M_0 = M_\pi= 0 M 0 = M π = 0 . The two light spin-3 2 \frac{3}{2} 2 3 states then form the Dirac gravitino of Question 11, with m 3 / 2 = 1 / ( 2 R ) m_{3/2} = 1/(2R) m 3/2 = 1/ ( 2 R ) , and the boundary-value problem has the residual U ( 1 ) R U(1)_R U ( 1 ) R discussed there.
Radiative transmission is controlled by three separate pieces of information: whether a boundary operator is allowed by the local symmetries, whether boundary matter couples to the required bulk channel, and whether the corresponding boundary-restricted propagator contains a local short-distance contribution or probes the global compactification.
These distinctions already separate scalar and fermion masses. The ordinary gravitino couples universally to the boundary supercurrent, and a boundary scalar receives a finite Scherk–Schwarz gravitational threshold. A Majorana mass for an R R R -charged boundary fermion is forbidden at the exact half-twist. A Dirac mass can be R R R -neutral, but requires an additional matter current coupled to the companion spin-3 2 \frac{3}{2} 2 3 channel; its existence and normalization are not fixed by the Scherk–Schwarz geometry alone.
Near the half-twist, the protecting U ( 1 ) R U(1)_R U ( 1 ) R is weakly broken. Scalar thresholds are even under reflection about ω = 1 2 \omega= \frac{1}{2} ω = 2 1 , whereas Majorana responses are odd. The former therefore change quadratically in the displacement from the half-twist, while the latter can appear linearly.
Let the compact interval be I = [ 0 , π R ] I = [0,\pi R] I = [ 0 , π R ] , and denote by Δ ω \Delta_\omega Δ ω the Euclidean quadratic bulk operator with the Scherk–Schwarz boundary conditions labelled by ω \omega ω . For fixed non-compact Euclidean momentum p p p , define
G ω ( p ) = ( p 2 + Δ ω ) − 1 . (391) \mathcal{G}_\omega(p) = \left(p^2 + \Delta_\omega\right)^{-1}. \tag*{(391)} G ω ( p ) = ( p 2 + Δ ω ) − 1 . (391)
The field value itself is not always the appropriate datum at an orbifold boundary. We therefore let T b \mathcal{T}_b T b denote the boundary restriction map at the endpoint b = 0 , π b = 0,\pi b = 0 , π , including the relevant parity projectors and normal derivatives.
The boundary-to-boundary kernels are then
G b b ′ ( p ; ω ) = T b G ω ( p ) T b ′ † . (392) \mathcal{G}_{bb'}(p;\omega)=\mathcal{T}_b\mathcal{G}_\omega(p)\mathcal{T}_{b'}^\dagger. \tag*{(392)} G b b ′ ( p ; ω ) = T b G ω ( p ) T b ′ † . (392)
If boundary currents J b J_b J b couple linearly to the corresponding restricted bulk fields, integrating out the bulk fluctuations gives, at quadratic order,
Γ i n d ( 2 ) = − 1 2 ∑ b , b ′ ∫ d 4 p ( 2 π ) 4 J b ( − p ) G b b ′ ( p ; ω ) J b ′ ( p ) , (393) \Gamma_{\mathrm{ind}}^{(2)}=-\frac{1}{2}\sum_{b,b'}\int\frac{d^4p}{(2\pi)^4}J_b(-p)\mathcal{G}_{bb'}(p;\omega)J_{b'}(p), \tag*{(393)} Γ ind ( 2 ) = − 2 1 b , b ′ ∑ ∫ ( 2 π ) 4 d 4 p J b ( − p ) G b b ′ ( p ; ω ) J b ′ ( p ) , (393)
up to statistics and index contractions.
The same restricted resolvents enter loop corrections to boundary masses. The compactification fixes the kernels G b b ′ \mathcal{G}_{bb'} G b b ′ , while the microscopic boundary theory fixes which currents exist and how strongly they couple. An operator allowed by symmetry is therefore not necessarily generated.
The pure half-twist benchmark
We now set ω = 1 2 \omega=\frac{1}{2} ω = 2 1 and M 0 = M π = 0 M_0=M_\pi=0 M 0 = M π = 0 . The breaking is then entirely in the Scherk–Schwarz gluing data, with no independent localized gravitino masses.
On 0 ≤ y ≤ π R 0\le y\le\pi R 0 ≤ y ≤ π R , the two half-integer towers relevant to the spin-3 2 \frac{3}{2} 2 3 system may be represented by
f n ( + ) ( y ) = 2 π R cos [ n + 1 2 R y ] , f_n^{(+)}(y)=\sqrt{\frac{2}{\pi R}}\cos\left[\frac{n+\frac{1}{2}}{R}y\right], f n ( + ) ( y ) = π R 2 cos [ R n + 2 1 y ] ,
f n ( − ) ( y ) = 2 π R sin [ n + 1 2 R y ] , n ≥ 0. (394) f_n^{(-)}(y)=\sqrt{\frac{2}{\pi R}}\sin\left[\frac{n+\frac{1}{2}}{R}y\right],\qquad n\ge0. \tag*{(394)} f n ( − ) ( y ) = π R 2 sin [ R n + 2 1 y ] , n ≥ 0. (394)
At y = 0 y=0 y = 0 , the + + + profiles have non-zero boundary values, whereas the − - − profiles vanish but have non-zero normal derivatives. Thus the even channel couples directly to ordinary boundary currents, while the odd channel can enter through derivative or bulk-mediated couplings. This kinematic fact does not determine whether a suitable matter current exists for the odd channel.
For p = ∣ p E ∣ p=\lvert p_E\rvert p = ∣ p E ∣ , the corresponding Green functions are
G + ( p ; y , y ′ ) = cosh ( p y < ) sinh [ p ( π R − y > ) ] p cosh ( π p R ) , G_+(p;y,y')=\frac{\cosh(py_<)\sinh\left[p(\pi R-y_>)\right]}{p\cosh(\pi pR)}, G + ( p ; y , y ′ ) = p cosh ( π pR ) cosh ( p y < ) sinh [ p ( π R − y > ) ] ,
G − ( p ; y , y ′ ) = sinh ( p y < ) cosh [ p ( π R − y > ) ] p cosh ( π p R ) , (395) G_-(p;y,y')=\frac{\sinh(py_<)\cosh\left[p(\pi R-y_>)\right]}{p\cosh(\pi pR)}, \tag*{(395)} G − ( p ; y , y ′ ) = p cosh ( π pR ) sinh ( p y < ) cosh [ p ( π R − y > ) ] , (395)
where y < y_< y < and y > y_> y > denote respectively the smaller and larger of y , y ′ y,y' y , y ′ .
The even same-boundary propagator at y = 0 y=0 y = 0 is tanh ( π p R ) / p \tanh(\pi pR)/p tanh ( π pR ) / p at the half-twist, while in the unbroken ω = 0 \omega=0 ω = 0 problem it is coth ( π p R ) / p \coth(\pi pR)/p coth ( π pR ) / p . Their Scherk–Schwarz-dependent difference is therefore
Δ G 00 S S ( p ) = G + ( 1 / 2 ) ( p ; 0 , 0 ) − G + ( 0 ) ( p ; 0 , 0 ) = − 2 p sinh ( 2 π p R ) . (396) \Delta G_{00}^{\mathrm{SS}}(p)=G_+^{(1/2)}(p;0,0)-G_+^{(0)}(p;0,0)=-\frac{2}{p\sinh(2\pi pR)}. \tag*{(396)} Δ G 00 SS ( p ) = G + ( 1/2 ) ( p ; 0 , 0 ) − G + ( 0 ) ( p ; 0 , 0 ) = − p sinh ( 2 π pR ) 2 . (396)
At large momentum this behaves as Δ G 00 S S ( p ) ∼ − 4 e − 2 π p R / p \Delta G_{00}^{\mathrm{SS}}(p)\sim-4e^{-2\pi pR}/p Δ G 00 SS ( p ) ∼ − 4 e − 2 π pR / p .
This provides an instructive qualification to the locality discussion of Question 18. A same-boundary amplitude need not be ultraviolet divergent: here the local short-distance singularity is identical in the twisted and untwisted theories and cancels in their difference. By contrast, a contribution that necessarily propagates between opposite endpoints contains factors such as 1 / cosh ( π p R ) 1/\cosh(\pi pR) 1/ cosh ( π pR ) , and is exponentially suppressed because the two supports are separated by the finite distance π R \pi R π R .
The universal scalar threshold
Consider a canonically normalized chiral multiplet Q = ( ϕ , ψ , F ) Q=(\phi,\psi,F) Q = ( ϕ , ψ , F ) localized at one endpoint. The ordinary gravitino couples universally to the local supersymmetry current,
L ψ J = − 1 2 M 4 ψ μ J S U S Y μ + h . c . , J S U S Y μ ⊃ 2 ( ∂ ν ϕ ˉ ) σ ν σ ˉ μ ψ + ⋯ . (397) \mathcal{L}_{\psi J}=-\frac{1}{2M_{4}}\psi_{\mu}J_{\mathrm{SUSY}}^{\mu}+\mathrm{h.c.},\qquad J_{\mathrm{SUSY}}^{\mu}\supset\sqrt{2}(\partial_{\nu}\bar{\phi})\sigma^{\nu}\bar{\sigma}^{\mu}\psi+\cdots. \tag*{(397)} L ψ J = − 2 M 4 1 ψ μ J SUSY μ + h.c. , J SUSY μ ⊃ 2 ( ∂ ν ϕ ˉ ) σ ν σ ˉ μ ψ + ⋯ . (397)
The scalar threshold is not obtained from one isolated gravitino-exchange diagram. Local supersymmetry relates the cubic supercurrent vertices to quartic contact, or seagull, interactions, and the complete locally supersymmetric set must be combined before subtracting the unbroken reference background.
For the minimal boundary-coupled five-dimensional supergravity benchmark, the resulting matching coefficient is
C A = 2 π R 3 M 4 2 = 2 3 M 5 3 , M 4 2 = π R M 5 3 . (398) C_{A}=\frac{2\pi R}{3M_{4}^{2}}=\frac{2}{3M_{5}^{3}},\qquad M_{4}^{2}=\pi R M_{5}^{3}. \tag*{(398)} C A = 3 M 4 2 2 π R = 3 M 5 3 2 , M 4 2 = π R M 5 3 . (398)
The Scherk–Schwarz contribution to the scalar mass is then
δ m ϕ 2 = C A ∫ d 4 p E ( 2 π ) 4 p E 2 Δ G 00 S S ( p ) . (399) \delta m_{\phi}^{2}=C_{A}\int\frac{\mathrm{d}^{4}p_{E}}{(2\pi)^{4}}p_{E}^{2}\Delta G_{00}^{\mathrm{SS}}(p). \tag*{(399)} δ m ϕ 2 = C A ∫ ( 2 π ) 4 d 4 p E p E 2 Δ G 00 SS ( p ) . (399)
Using Eq. (396) and performing the angular integral gives
δ m ϕ 2 = − C A 4 π 2 ∫ 0 ∞ d p p 4 sinh ( 2 π p R ) . (400) \delta m_{\phi}^{2}=-\frac{C_{A}}{4\pi^{2}}\int_{0}^{\infty}\mathrm{d}p\,\frac{p^{4}}{\sinh(2\pi pR)}. \tag*{(400)} δ m ϕ 2 = − 4 π 2 C A ∫ 0 ∞ d p sinh ( 2 π pR ) p 4 . (400)
With x = 2 π R p x=2\pi Rp x = 2 π R p and
∫ 0 ∞ d x x 4 sinh x = 93 2 ζ ( 5 ) , (401) \int_{0}^{\infty}\mathrm{d}x\,\frac{x^{4}}{\sinh x}=\frac{93}{2}\zeta(5), \tag*{(401)} ∫ 0 ∞ d x sinh x x 4 = 2 93 ζ ( 5 ) , (401)
one obtains
δ m ϕ 2 = − 31 ζ ( 5 ) 128 π 6 1 R 4 M 4 2 = − 31 ζ ( 5 ) 128 π 7 1 R 5 M 5 3 . (402) \delta m_{\phi}^{2}=-\frac{31\zeta(5)}{128\pi^{6}}\frac{1}{R^{4}M_{4}^{2}}=-\frac{31\zeta(5)}{128\pi^{7}}\frac{1}{R^{5}M_{5}^{3}}. \tag*{(402)} δ m ϕ 2 = − 128 π 6 31 ζ ( 5 ) R 4 M 4 2 1 = − 128 π 7 31 ζ ( 5 ) R 5 M 5 3 1 . (402)
The sign is negative in this minimal anti-periodic benchmark. This is a statement about one finite Scherk–Schwarz threshold, not about the sign of the complete physical scalar mass. Local renormalized terms, radion effects and additional interactions can contribute to the full mass matrix, as emphasized in Question 18.
The gravitational scalar threshold in this class of Scherk–Schwarz compactifications has been analysed in Refs. [32, 33, 78, 79].
Fermion masses and the residual U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R
At the exact half-twist, the residual U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R constrains the fermionic operators that can be generated on the boundary.
For a localized Weyl fermion ψ \psi ψ of R R R -charge q ψ q_{\psi} q ψ , a Majorana bilinear ψ ψ \psi\psi ψ ψ carries charge 2 q ψ 2q_{\psi} 2 q ψ . An exact U ( 1 ) R U(1)_{\mathrm{R}} U ( 1 ) R therefore forbids such a mass whenever q ψ ≠ 0 q_{\psi} \ne0 q ψ = 0 .
A Dirac bilinear ψ ξ \psi\xi ψ ξ , by contrast, is compatible with the symmetry when q ψ + q ξ = 0 q_{\psi}+q_{\xi}=0 q ψ + q ξ = 0 . This condition is necessary but not sufficient. The existence of the Dirac gravitino of Question 11 does not by itself imply that both of its spin-3 2 \frac{3}{2} 2 3 components couple to the boundary matter sector in the way required to generate a Dirac mass.
Denoting the two gravitini by ψ μ \psi_{\mu} ψ μ and χ μ \chi_{\mu} χ μ , the relevant couplings have schematically the form
L m a t t e r ⊃ − 1 2 M 4 ψ μ J μ − χ μ j χ μ + h . c . (403) \mathcal{L}_{\mathrm{matter}} \supset-\frac{1}{2M_{4}}\psi_{\mu}J^{\mu}-\chi_{\mu}j_{\chi}^{\mu}+\mathrm{h.c.} \tag*{(403)} L matter ⊃ − 2 M 4 1 ψ μ J μ − χ μ j χ μ + h.c. (403)
The ordinary supercurrent J μ J^{\mu} J μ is fixed by local supersymmetry. The existence and normalization of a companion current j χ μ j_{\chi}^{\mu} j χ μ depend instead on how the boundary matter sector is embedded into the higher-dimensional or extended-supergravity theory.
A radiative Dirac mass therefore requires more than the spin-3 2 \frac{3}{2} 2 3 spectrum: an oppositely charged fermion must be present, a companion matter current must couple to χ μ \chi_{\mu} χ μ , and the loop must connect the two matter channels. These ingredients are not fixed by the Scherk–Schwarz geometry alone.
The companion-channel kernels
Once the companion current is specified, its propagation is fixed by the half-twist boundary problem. The mixed even–odd propagator with one bulk Dirac-mass insertion can be written
G + − ( 1 ) ( p ; y , y ′ ) = m 3 / 2 ∫ 0 π R d z G + ( p ; y , z ) G − ( p ; z , y ′ ) . (404) G_{+-}^{(1)}(p;y,y')=m_{3/2}\int_{0}^{\pi R} dz\,G_{+}(p;y,z)G_{-}(p;z,y'). \tag*{(404)} G +− ( 1 ) ( p ; y , y ′ ) = m 3/2 ∫ 0 π R d z G + ( p ; y , z ) G − ( p ; z , y ′ ) . (404)
For two currents on the same boundary,
∂ y ′ G + − ( 1 ) ( p ; 0 , y ′ ) ∣ y ′ = 0 = m 3 / 2 2 p 2 tanh 2 ( π p R ) . (405) \left.\partial_{y'}G_{+-}^{(1)}(p;0,y')\right|_{y'=0}=\frac{m_{3/2}}{2p^{2}}\tanh^{2}(\pi pR). \tag*{(405)} ∂ y ′ G +− ( 1 ) ( p ; 0 , y ′ ) y ′ = 0 = 2 p 2 m 3/2 tanh 2 ( π pR ) . (405)
The kernel approaches a power-law local term at large momentum. Accordingly, its complete coefficient requires a boundary counterterm. After that local piece is subtracted, the finite Scherk–Schwarz remainder contains
∫ 0 ∞ d p p [ tanh 2 ( π p R ) − 1 ] = − ln 2 π 2 R 2 . (406) \int_{0}^{\infty}dp\,p\left[\tanh^{2}(\pi pR)-1\right]=-\frac{\ln2}{\pi^{2}R^{2}}. \tag*{(406)} ∫ 0 ∞ d p p [ tanh 2 ( π pR ) − 1 ] = − π 2 R 2 ln 2 . (406)
For currents on opposite boundaries one finds instead
G + − ( 1 ) ( p ; 0 , π R ) = m 3 / 2 2 p 3 cosh ( π p R ) [ π p R − tanh ( π p R ) ] . (407) G_{+-}^{(1)}(p;0,\pi R)=\frac{m_{3/2}}{2p^{3}\cosh(\pi pR)}\left[\pi pR-\tanh(\pi pR)\right]. \tag*{(407)} G +− ( 1 ) ( p ; 0 , π R ) = 2 p 3 cosh ( π pR ) m 3/2 [ π pR − tanh ( π pR ) ] . (407)
This kernel is exponentially suppressed at large momentum and is intrinsically ultraviolet finite. The corresponding dimensionless integral is
∫ 0 ∞ d x x − tanh x cosh x = 2 G C a t − 1 , (408) \int_{0}^{\infty} dx\,\frac{x-\tanh x}{\cosh x}=2G_{\mathrm{Cat}}-1, \tag*{(408)} ∫ 0 ∞ d x cosh x x − tanh x = 2 G Cat − 1 , (408)
where G C a t G_{\mathrm{Cat}} G Cat is Catalan’s constant.
To separate these geometrical results from the microscopic matter couplings, introduce coefficients A s a m e \mathcal{A}_{\mathrm{same}} A same and A o p p \mathcal{A}_{\mathrm{opp}} A opp , of mass dimensions − 2 -2 − 2 and − 1 -1 − 1 , respectively. They contain the companion-current normalization, tensor contractions and the factors required to close the matter loop. The finite same-boundary part and the opposite-boundary contribution then take the form
m D , s a m e f i n = − A s a m e ln 2 4 π 4 m 3 / 2 3 , m D , o p p = A o p p ( 2 G C a t − 1 ) 8 π 3 m 3 / 2 2 . (409) m^{\mathrm{fin}}_{D,\mathrm{same}}=-\frac{\mathcal{A}_{\mathrm{same}}\ln2}{4\pi^{4}}m_{3/2}^{3},
\qquad
m_{D,\mathrm{opp}}=\frac{\mathcal{A}_{\mathrm{opp}}(2G_{\mathrm{Cat}}-1)}{8\pi^{3}}m_{3/2}^{2}. \tag*{(409)} m D , same fin = − 4 π 4 A same ln 2 m 3/2 3 , m D , opp = 8 π 3 A opp ( 2 G Cat − 1 ) m 3/2 2 . (409)
The first must be supplemented by an independently renormalized local term, m D , s a m e p h y s ( μ ) = m D , l o c a l ( μ ) + m D , s a m e f i n m^{\mathrm{phys}}_{D,\mathrm{same}}(\mu)=m_{D,\mathrm{local}}(\mu)+m^{\mathrm{fin}}_{D,\mathrm{same}} m D , same phys ( μ ) = m D , local ( μ ) + m D , same fin .
For a conservative supergravity normalization one may write A s a m e = a s a m e / M 4 2 \mathcal{A}_{\mathrm{same}}=a_{\mathrm{same}}/M_{4}^{2} A same = a same / M 4 2 and A o p p = a o p p m 3 / 2 / M 4 2 \mathcal{A}_{\mathrm{opp}}=a_{\mathrm{opp}}m_{3/2}/M_{4}^{2} A opp = a opp m 3/2 / M 4 2 , with dimensionless microscopic coefficients. Both topologies then give parametrically
m D m a t t e r ∼ 1 16 π 2 m 3 / 2 3 M 4 2 , (410) m_{D}^{\mathrm{matter}}\sim\frac{1}{16\pi^{2}}\frac{m_{3/2}^{3}}{M_{4}^{2}}, \tag*{(410)} m D matter ∼ 16 π 2 1 M 4 2 m 3/2 3 , (410)
up to the model-dependent coefficients and the numerical factors displayed above. A parametrically larger result requires a stronger normalization of the companion current and is an additional effective theory assumption, not a consequence of the Scherk–Schwarz boundary conditions.
Small departures from the half-twist
Consider now an effective spectral phase ω e f f = 1 2 − ϵ ω \omega_{\mathrm{eff}}=\frac{1}{2}-\epsilon_{\omega} ω eff = 2 1 − ϵ ω , with ∣ ϵ ω ∣ ≪ 1 |\epsilon_{\omega}|\ll1 ∣ ϵ ω ∣ ≪ 1 . The two light spin-3 2 \frac{3}{2} 2 3 masses are ( 1 2 + ϵ ω ) / R (\frac{1}{2}+\epsilon_{\omega})/R ( 2 1 + ϵ ω ) / R , so their splitting is 2 ∣ ϵ ω ∣ / R 2|\epsilon_{\omega}|/R 2∣ ϵ ω ∣/ R . The residual U ( 1 ) R U(1)_{R} U ( 1 ) R of the exact half-twist is then weakly broken and Majorana operators are no longer forbidden.
The expansion around the half-twist is nevertheless constrained by the reflection ω ↔ 1 − ω \omega\leftrightarrow1-\omega ω ↔ 1 − ω . Scalar thresholds are even under this reflection, whereas the Majorana response is odd:
m s c a l a r 2 ( 1 2 − ϵ ω ) = m s c a l a r 2 ( 1 2 ) + O ( ϵ ω 2 ) , m M a j o r a n a ( 1 2 − ϵ ω ) = O ( ϵ ω ) . (411) m_{\mathrm{scalar}}^{2}\left(\frac{1}{2}-\epsilon_{\omega}\right)=m_{\mathrm{scalar}}^{2}\left(\frac{1}{2}\right)+O(\epsilon_{\omega}^{2}),
\qquad
m_{\mathrm{Majorana}}\left(\frac{1}{2}-\epsilon_{\omega}\right)=O(\epsilon_{\omega}). \tag*{(411)} m scalar 2 ( 2 1 − ϵ ω ) = m scalar 2 ( 2 1 ) + O ( ϵ ω 2 ) , m Majorana ( 2 1 − ϵ ω ) = O ( ϵ ω ) . (411)
For the gravitationally transmitted Majorana benchmark, the relevant spectral function may be written
f 1 / 2 ( ω ) = 3 i 8 π 3 [ Li 4 ( e − 2 π i ω ) − Li 4 ( e 2 π i ω ) ] , (412) f_{1/2}(\omega)=\frac{3i}{8\pi^{3}}\left[\operatorname{Li}_{4}\left(e^{-2\pi i\omega}\right)-\operatorname{Li}_{4}\left(e^{2\pi i\omega}\right)\right], \tag*{(412)} f 1/2 ( ω ) = 8 π 3 3 i [ Li 4 ( e − 2 π iω ) − Li 4 ( e 2 π iω ) ] , (412)
which gives
f 1 / 2 ( 1 2 − ϵ ω ) = 9 ζ ( 3 ) 8 π 2 ϵ ω + O ( ϵ ω 3 ) . (413) f_{1/2}\left(\frac{1}{2}-\epsilon_{\omega}\right)=\frac{9\zeta(3)}{8\pi^{2}}\epsilon_{\omega}+O(\epsilon_{\omega}^{3}). \tag*{(413)} f 1/2 ( 2 1 − ϵ ω ) = 8 π 2 9 ζ ( 3 ) ϵ ω + O ( ϵ ω 3 ) . (413)
With gravitational-strength mediation this corresponds parametrically to
M 1 / 2 g r a v ∼ ϵ ω 1 16 π 2 m 3 / 2 3 M 4 2 . (414) M_{1/2}^{\mathrm{grav}} \sim\epsilon_{\omega}\frac{1}{16\pi^{2}}\frac{m_{3/2}^{3}}{M_{4}^{2}}. \tag*{(414)} M 1/2 grav ∼ ϵ ω 16 π 2 1 M 4 2 m 3/2 3 . (414)
The exact half-twist is a useful expansion point: the scalar threshold is already non-zero there but has no linear displacement, whereas the Majorana response starts at first order. The Dirac/pseudo-Dirac organization and the corresponding matter selection rules are discussed further in Ref. [44].
What the spin-3 2 \frac{3}{2} 2 3 spectrum does not fix
The distinction between rigid monodromy and an effective spectral phase made in Questions 9, 15 and 18 remains important here.
Two compactifications can have the same spin-3 2 \frac{3}{2} 2 3 eigenvalues and the same ω e f f \omega_{\mathrm{eff}} ω eff while differing in their localized Goldstino sectors, boundary counterterms, companion currents, or direct local R R R -breaking interactions. The spin-3 2 \frac{3}{2} 2 3 spectrum fixes the propagation problem only after the complete quadratic boundary conditions have been specified; it does not fix the microscopic boundary vertices.
The pure half-twist geometry therefore determines the even and odd bulk wave functions, the corresponding restricted kernels and the residual U ( 1 ) R U(1)_{R} U ( 1 ) R . Local supersymmetry also fixes the ordinary supercurrent coupling and hence the universal scalar threshold. Companion-current normalizations and allowed local counterterms remain additional information.
Finite radiative transmission of supersymmetry breaking through an extra dimension, including applications to localized matter and soft terms, was studied in Refs. [80, 81, 82, 83, 84]. Related Scherk–Schwarz orbifold calculations can be found in Ref. [31].
The Dirac and pseudo-Dirac organization near the half-twist, together with the companion-current analysis used above, is discussed in Ref. [44].
Take-home message
Boundary fields can feel Scherk–Schwarz breaking through boundary-restricted bulk propagators even though they have no Kaluza–Klein momentum of their own.
Same-boundary and opposite-boundary effects can both be ultraviolet finite for different reasons. In the former, the local singularity can cancel after subtraction of the supersymmetric reference theory; in the latter, finite separation gives exponential suppression.
For the minimal pure half-twist benchmark, the locally supersymmetric gravitational
itational coupling gives
δ m ϕ 2 = − 31 ζ ( 5 ) 128 π 6 1 R 4 M 4 2 . \delta m_{\phi}^{2}=-\frac{31\zeta(5)}{128\pi^{6}}\frac{1}{R^{4}M_{4}^{2}}. δ m ϕ 2 = − 128 π 6 31 ζ ( 5 ) R 4 M 4 2 1 .
This is one finite contribution to the scalar mass matrix, not a prediction of its final sign.
At the exact half-twist, the residual U ( 1 ) R U(1)_{R} U ( 1 ) R forbids a Majorana mass for an R R R -charged boundary fermion. An R R R -invariant Dirac mass is not generated merely because the bulk gravitino is Dirac: the second spin-3 2 \frac{3}{2} 2 3 component must couple to an appropriate companion matter current so that a non-vanishing loop can actually be closed. The existence and normalization of this current are additional microscopic data, not consequences of the Scherk–Schwarz spectrum.
Once such a companion current exists, the half-twist fixes the propagation kernels connecting it to the ordinary matter current. Same-boundary diagrams contain a local piece that must be renormalized, leaving a finite remainder proportional to ln 2 \ln2 ln 2 ; opposite-boundary diagrams necessarily propagate across the interval and are intrinsically ultraviolet finite, with the corresponding kernel giving the constant 2 G C a t − 1 2G_{\mathrm{Cat}}-1 2 G Cat − 1 .
Near the half-twist, scalar thresholds are even in the displacement from ω = 1 2 \omega=\frac{1}{2} ω = 2 1 , while Majorana responses are odd. The same spin-3 2 \frac{3}{2} 2 3 spectrum can nevertheless correspond to different boundary soft spectra when the localized currents and counterterms differ.
This naturally leads to the next question:
What quantum consistency conditions must a Scherk–Schwarz twist satisfy?
What quantum consistency conditions must a Scherk–Schwarz twist satisfy?
Essential idea
Classically, a Scherk–Schwarz compactification is specified by telling the fields how they are reidentified under the global identifications of the compact space. Question 16 organized these data into a homomorphism ρ : Γ → G \rho: \Gamma\to G ρ : Γ → G , where Γ \Gamma Γ is the identification group and G G G an admissible symmetry group of the parent theory.
Quantum mechanically, two further questions arise.
The first concerns fermions. Two transformations that act identically on all bosonic operators can still differ by fermion parity ( − 1 ) F (-1)^{F} ( − 1 ) F . The same bosonic gluing can therefore admit different fermionic lifts, or fail to admit one at all.
The second concerns anomalies. A transformation may preserve the classical action while changing the phase of the quantum functional integral. Such a transformation can still be used as a prescribed background, but it cannot automatically define an autonomous quotient or orbifold.
The classical twist must therefore first be lifted to the full fermionic symmetry, ρ ~ : Γ → G f \widetilde{\rho} : \Gamma\to G_{\mathrm{f}} ρ : Γ → G f , and the anomaly of the parent theory must then be restricted to this lifted action. If the parent anomaly is denoted by A \mathfrak{A} A , the relevant anomaly is
A ρ ~ = ( B ρ ~ ) ∗ A . \mathfrak{A}_{\widetilde{\rho}} = (B\widetilde{\rho})^{*}\mathfrak{A}. A ρ = ( B ρ ) ∗ A .
An autonomous quantum compactification requires a coherent trivialization of this anomaly. At this level of description, the quantum refinement of the classical twist is specified by the fermionic lift together with such a trivialization.
A second layer appears when the twist is varied continuously. The fermionic spectrum changes, and particular Kaluza–Klein branches can cross zero. These resonance walls depend on the representation being examined. The adjoint representation detects residual-symmetry enhancement, the supercharge representation detects supersymmetry enhancement, and physical fermion representations detect their own zero-mode walls. These loci need not coincide.
Spectral flow and the eta invariant describe the passage through these walls. Globally, the fermion determinants form an anomaly line over twist space. Its curvature contains the perturbative anomaly, while its holonomy detects global anomalies.
Why classical gluing is not yet quantum gluing
The classical twist of Question 16 tells us how fields are identified when one moves around the compact space. At the quantum level, however, two extra consistency checks appear before one can regard the compactification as a stand-alone theory.
The first check is already visible on a circle. Suppose all bosonic fields are periodic. A fermion can nevertheless satisfy either ψ ( y + 2 π R ) = + ψ ( y ) \psi(y+2\pi R)=+\psi(y) ψ ( y + 2 π R ) = + ψ ( y ) or ψ ( y + 2 π R ) = − ψ ( y ) \psi(y+2\pi R)=-\psi(y) ψ ( y + 2 π R ) = − ψ ( y ) . The bosonic gluing is identical, but the fermionic theories are different: these are the two spin structures on the circle.
The reason is fermion parity. The transformation ( − 1 ) F (-1)^{F} ( − 1 ) F acts trivially on every bosonic operator and by a minus sign on every fermionic operator. Bosonic data alone therefore cannot distinguish a transformation g g g from ( − 1 ) F g (-1)^{F}g ( − 1 ) F g .
The first quantum question is consequently whether the classical gluing admits a globally consistent action on the fermionic theory.
A second question remains even after such an action has been found. A transformation can leave the classical action invariant but change the phase of the functional integral. The symmetry is then anomalous. Using it as a fixed external background remains meaningful, but quotienting or gauging it requires additional anomaly-cancellation data.
The two issues are logically distinct. A twist may fail before anomalies are considered because it cannot be lifted consistently to the fermions, or the fermionic lift may exist but be anomalous.
Lifting the twist to the fermionic symmetry
Let G b G_{\mathrm b} G b denote the symmetry group seen by bosonic operators. The full symmetry acting on fermions can be a non-trivial central extension
1 ⟶ Z 2 F ⟶ G f → π G b ⟶ 1 , (415) 1 \longrightarrow\mathbb{Z}^{F}_{2} \longrightarrow G_{\mathrm f} \xrightarrow{\pi} G_{\mathrm b} \longrightarrow1, \tag*{(415)} 1 ⟶ Z 2 F ⟶ G f π G b ⟶ 1 , (415)
where Z 2 F = { 1 , ( − 1 ) F } \mathbb{Z}^{F}_{2}=\{1,(-1)^{F}\} Z 2 F = { 1 , ( − 1 ) F } .
Suppose the classical bosonic gluing is ρ b : Γ → G b \rho_{\mathrm b}:\Gamma\to G_{\mathrm b} ρ b : Γ → G b . A fermionic lift is a homomorphism ρ ~ : Γ → G f \widetilde{\rho}:\Gamma\to G_{\mathrm f} ρ : Γ → G f satisfying
π ∘ ρ ~ = ρ b . (416) \pi\circ\widetilde{\rho}=\rho_{\mathrm b}. \tag*{(416)} π ∘ ρ = ρ b . (416)
The central extension Eq. (415) is characterized by a class e ∈ H 2 ( B G b , Z 2 ) e\in H^{2}(BG_{\mathrm b},\mathbb{Z}_{2}) e ∈ H 2 ( B G b , Z 2 ) , where B G b BG_{\mathrm b} B G b is the classifying space of G b G_{\mathrm b} G b . The bosonic twist induces a map B ρ b : B Γ → B G b B\rho_{\mathrm b}:B\Gamma\to BG_{\mathrm b} B ρ b : B Γ → B G b , and hence the pulled-back class
( B ρ b ) ∗ e ∈ H 2 ( B Γ , Z 2 ) . (417) (B\rho_{\mathrm b})^{*}e\in H^{2}(B\Gamma,\mathbb{Z}_{2}). \tag*{(417)} ( B ρ b ) ∗ e ∈ H 2 ( B Γ , Z 2 ) . (417)
A strict fermionic lift exists precisely when the pulled-back extension splits,
( B ρ b ) ∗ e = 0. (418) (B\rho_{\mathrm b})^{*}e=0. \tag*{(418)} ( B ρ b ) ∗ e = 0. (418)
This is the abstract statement that all fermionic signs assigned to the generators of Γ \Gamma Γ must be compatible with the relations among those generators.
Even when a lift exists, it need not be unique. For a central extension on which Γ \Gamma Γ acts trivially on Z 2 F \mathbb{Z}^{F}_{2} Z 2 F , the possible lifts form a torsor under
H 1 ( B Γ , Z 2 ) ≃ Hom ( Γ , Z 2 ) . (419) H^{1}(B\Gamma,\mathbb{Z}_{2}) \simeq\operatorname{Hom}(\Gamma,\mathbb{Z}_{2}). \tag*{(419)} H 1 ( B Γ , Z 2 ) ≃ Hom ( Γ , Z 2 ) . (419)
A torsor has the same relative structure as a group but no preferred origin: two lifts have a well-defined difference in H 1 ( B Γ , Z 2 ) H^{1}(B\Gamma,\mathbb{Z}_{2}) H 1 ( B Γ , Z 2 ) , without one lift being singled out as canonical.
The circle as the elementary lifting problem
For a circle, Γ = Z \Gamma= \mathbb{Z} Γ = Z and B Z ≃ S 1 B\mathbb{Z} \simeq S^{1} B Z ≃ S 1 . Consequently,
H 2 ( B Z , Z 2 ) = H 2 ( S 1 , Z 2 ) = 0. H^{2}(B\mathbb{Z},\mathbb{Z}_{2}) = H^{2}(S^{1},\mathbb{Z}_{2}) = 0. H 2 ( B Z , Z 2 ) = H 2 ( S 1 , Z 2 ) = 0.
There is therefore no obstruction of the type Eq. (418) : once the bosonic image of the generator has been chosen, it can be lifted to the fermionic symmetry.
The ambiguity is instead measured by
H 1 ( B Z , Z 2 ) = H 1 ( S 1 , Z 2 ) ≃ Z 2 . H^{1}(B\mathbb{Z},\mathbb{Z}_{2}) = H^{1}(S^{1},\mathbb{Z}_{2}) \simeq\mathbb{Z}_{2}. H 1 ( B Z , Z 2 ) = H 1 ( S 1 , Z 2 ) ≃ Z 2 .
Let 1 ∈ Z 1 \in\mathbb{Z} 1 ∈ Z denote the generator and let g ∈ G f g \in G_{\mathrm{f}} g ∈ G f be one lift of its bosonic image. The other is ( − 1 ) F g (-1)^{F}g ( − 1 ) F g . On a fermion,
ρ ~ + ( 1 ) = g ⟹ ψ ( y + 2 π R ) = g ψ ( y ) , \widetilde{\rho}_{+}(1) = g \qquad\Longrightarrow\qquad\psi(y + 2\pi R) = g\psi(y), ρ + ( 1 ) = g ⟹ ψ ( y + 2 π R ) = g ψ ( y ) ,
ρ ~ − ( 1 ) = ( − 1 ) F g ⟹ ψ ( y + 2 π R ) = − g ψ ( y ) . (420) \widetilde{\rho}_{-}(1) = (-1)^{F}g \qquad\Longrightarrow\qquad\psi(y + 2\pi R) = -g\psi(y). \tag*{(420)} ρ − ( 1 ) = ( − 1 ) F g ⟹ ψ ( y + 2 π R ) = − g ψ ( y ) . (420)
For trivial internal bosonic gluing, g = 1 g = 1 g = 1 , these are simply the periodic and anti-periodic spin structures. The elementary sign choice encountered at the beginning of the review is the simplest example of the general statement that an unobstructed bosonic twist can possess inequivalent fermionic lifts.
The fermionic lift must also be compatible with the geometric spin structure. For an ordinary orientation-preserving unitary symmetry, the combined local structure is schematically
G ^ d = Spin ( d ) × Z 2 F G f , (421) \widehat{G}_{d} = \operatorname{Spin}(d) \mathop{\times}_{\mathbb{Z}_{2}^{F}} G_{\mathrm{f}}, \tag*{(421)} G d = Spin ( d ) × Z 2 F G f , (421)
where the central element − 1 ∈ Spin ( d ) -1 \in\operatorname{Spin}(d) − 1 ∈ Spin ( d ) is identified with fermion parity. If the compactification includes orientation reversal or an antiunitary operation, the Spin structure is replaced by the appropriate Pin or more general twisted tangential structure.
The first quantum consistency condition is therefore already non-trivial: the classical gluing must lift coherently to the complete fermionic and geometric symmetry data.
A symmetry of the action can still be anomalous
Assume now that a fermionic lift ρ ~ : Γ → G f \widetilde{\rho}:\Gamma\to G_{\mathrm{f}} ρ : Γ → G f has been chosen. Classical symmetry means that the action is invariant. Quantum symmetry requires more: the functional measure must transform consistently.
Schematically, one can have S [ g Φ ] = S [ Φ ] S[g\Phi] = S[\Phi] S [ g Φ ] = S [ Φ ] while
D ( g Φ ) = e i α [ g , A ] D Φ , D(g\Phi) = e^{i\alpha[g,A]}D\Phi, D ( g Φ ) = e i α [ g , A ] D Φ ,
where the phase can depend on the transformation and on background fields A A A . When this phase cannot be removed consistently by local counterterms, the symmetry is anomalous.
This distinction matters directly for Scherk–Schwarz compactification. A transformation can be an admissible classical monodromy and nevertheless fail to define an autonomous quantum quotient.
There are several equivalent languages for anomalies. The one most useful for treating perturbative and global anomalies together is to regard the anomaly of a d d d -dimensional quantum field theory as an invertible ( d + 1 ) (d+1) ( d + 1 ) -dimensional field theory A G f \mathcal{A}_{G_f} A G f . The detailed categorical construction will not be needed here. Its immediate physical consequence is that, on a closed d d d -dimensional spacetime X X X with symmetry background A A A , the anomalous partition function is not naturally a complex number but an element of a one-dimensional complex vector space,
Z Q ( X , A ) ∈ L X , A . (422) Z_Q(X,A) \in\mathcal{L}_{X,A}. \tag*{(422)} Z Q ( X , A ) ∈ L X , A . (422)
The space L X , A \mathcal{L}_{X,A} L X , A is the anomaly line.
Locally in the space of backgrounds one may choose a phase convention and represent the fermion determinant by an ordinary complex function. On an overlap, two such choices need agree only up to a non-zero phase. These local functions therefore assemble naturally into a section of a complex line bundle rather than into one globally defined function.
A trivialization identifies these lines coherently with C \mathbb{C} C and turns the partition function into an ordinary number. Determinant lines and exponentiated eta invariants provide concrete realizations of this structure for fermionic theories [85, 86, 87].
Only the anomaly seen by the twist matters
The parent symmetry group can be anomalous even though the particular subgroup used in the Scherk–Schwarz compactification is not.
The lifted twist induces a map B ρ ~ : B Γ → B G f B\widetilde{\rho}: B\Gamma\to BG_f B ρ : B Γ → B G f . The anomaly relevant to the compactification is the anomaly of the parent theory restricted to this action,
A ρ ~ = ( B ρ ~ ) ∗ A G f . (423) \mathcal{A}_{\widetilde{\rho}} = (B\widetilde{\rho})^{*}\mathcal{A}_{G_f}. \tag*{(423)} A ρ = ( B ρ ) ∗ A G f . (423)
It is therefore neither necessary that the whole group G f G_f G f be anomaly free nor sufficient that one infinitesimal generator look harmless. An anomalous continuous symmetry can contain an anomaly-free discrete subgroup suitable for Scherk–Schwarz gluing. Conversely, a continuous generator can have vanishing perturbative anomaly while a corresponding discrete subgroup retains a torsion or global anomaly.
The relevant object is always the complete pulled-back anomaly A ρ ~ \mathcal{A}_{\widetilde{\rho}} A ρ .
Fixed backgrounds, compactifications, and orbifolds
An anomaly does not forbid every use of the corresponding symmetry. What is required depends on what one wants to do with the symmetry background.
If the Scherk–Schwarz configuration is treated as a prescribed external background, the anomalous theory can still be evaluated. Its partition function remains line valued, Z ( X , ρ ~ ) ∈ L X , ρ ~ Z(X,\tilde{\rho}) \in\mathcal{L}_{X,\tilde{\rho}} Z ( X , ρ ~ ) ∈ L X , ρ ~ .
A stand-alone compactification requires more. Its partition function must be an ordinary number, so the pulled-back anomaly must be trivialized. An orbifold or gauging is still more restrictive: different symmetry backgrounds must be summed over or identified, and the trivialization must be coherent over the corresponding family of Γ \Gamma Γ -bundles and compatible with cutting and gluing.
A trivialization is a natural isomorphism
τ : A ρ ~ → ∼ 1 , (424) \tau:\mathcal{A}_{\tilde{\rho}} \xrightarrow{\sim} 1, \tag*{(424)} τ : A ρ ~ ∼ 1 , (424)
where 1 1 1 denotes the trivial invertible theory.
At this level of description, the quantum Scherk–Schwarz data may therefore be summarized by
( ρ ~ , τ ) . (425) (\tilde{\rho},\tau). \tag*{(425)} ( ρ ~ , τ ) . (425)
The lift specifies how the gluing acts on fermions, while the trivialization supplies the anomaly-cancellation data needed to make that action an autonomous quantum compactification.
Obstruction and choice are different
The anomaly can affect the classical twist space in two distinct ways. For some lifted representations, Triv ( A ρ ~ ) \operatorname{Triv}(\mathcal{A}_{\tilde{\rho}}) Triv ( A ρ ~ ) is empty. The twist is then obstructed as an autonomous quantum compactification.
For other representations a trivialization exists but is not unique. Different choices can define distinct quantum theories even when the classical monodromy is identical.
The quantum refinement of the classical twist space can therefore be written schematically as
X q ( Γ , G f ; A ) = { ( ρ ~ , τ ) ∣ ρ ~ : Γ → G f , τ ∈ Triv ( ( B ρ ~ ) ∗ A ) } quantum equivalences . (426) \mathfrak{X}_{\mathrm{q}}(\Gamma,G_{\mathrm{f}};\mathcal{A}) =
\frac{\left\{(\tilde{\rho},\tau)\mid\tilde{\rho}:\Gamma\to G_{\mathrm{f}},\quad\tau\in\operatorname{Triv}\left((B\tilde{\rho})^{*}\mathcal{A}\right)\right\}}
{\text{quantum equivalences}}. \tag*{(426)} X q ( Γ , G f ; A ) = quantum equivalences { ( ρ ~ , τ ) ∣ ρ ~ : Γ → G f , τ ∈ Triv ( ( B ρ ~ ) ∗ A ) } . (426)
A simple cohomological model makes the distinction transparent. Suppose that a finite-group anomaly is represented by α ∈ Z d + 1 ( B Γ , U ( 1 ) ) \alpha\in Z^{d+1}(B\Gamma,U(1)) α ∈ Z d + 1 ( B Γ , U ( 1 )) . A trivialization is a d d d -cochain β \beta β satisfying
δ β = α . (427) \delta\beta=\alpha. \tag*{(427)} δ β = α . (427)
There is a direct physical interpretation of this equation. The anomalous transformation of the original quantum theory produces the phase represented by α \alpha α . Multiplying the path integral by the local topological phase represented by β \beta β changes its transformation law by δ β \delta\beta δ β . Equation (20.13) says precisely that the added phase cancels the anomaly.
If [ α ] ≠ 0 [\alpha]\ne0 [ α ] = 0 in H d + 1 ( B Γ , U ( 1 ) ) H^{d+1}(B\Gamma,U(1)) H d + 1 ( B Γ , U ( 1 )) , no such β \beta β exists and the anomaly is an obstruction.
If [ α ] = 0 [\alpha]=0 [ α ] = 0 , however, the trivialization need not be unique. Two solutions β \beta β and β ′ \beta' β ′ obey δ ( β ′ − β ) = 0 \delta(\beta'-\beta)=0 δ ( β ′ − β ) = 0 , so their difference is a d d d -cocycle. Modulo equivalence, the possible choices form a torsor under
H d ( B Γ , U ( 1 ) ) . (428) H^d(B\Gamma,U(1)). \tag*{(428)} H d ( B Γ , U ( 1 )) . (428)
Thus the same mathematics distinguishes two physically different questions: whether an anomaly can be cancelled, and which anomaly-cancelling phase has been chosen.
The distinction is familiar in two-dimensional finite-group orbifolds: H 3 ( B Γ , U ( 1 ) ) H^3(B\Gamma,U(1)) H 3 ( B Γ , U ( 1 )) contains the obstruction to gauging, while H 2 ( B Γ , U ( 1 ) ) H^2(B\Gamma,U(1)) H 2 ( B Γ , U ( 1 )) parametrizes discrete-torsion choices.
For general fermionic systems, ordinary group cohomology is replaced by the appropriate generalized cohomology or bordism theory. The logical separation between obstruction and choice remains the same.
Varying the twist: where do fermion zero modes appear?
Once a quantum-consistent twist has been identified, one can ask how the quantum theory changes when the twist is varied continuously.
Let a circle monodromy lie in a maximal torus,
U ( ω ) = exp ( 2 π i ω ⋅ H ) , U(\omega)=\exp(2\pi i\,\omega\cdot\mathbf{H}), U ( ω ) = exp ( 2 π i ω ⋅ H ) ,
and let R \mathcal{R} R be a representation with weight set Wt ( R ) \operatorname{Wt}(\mathcal{R}) Wt ( R ) . Let ν ∈ { 0 , 1 2 } \nu\in\{0,\frac{1}{2}\} ν ∈ { 0 , 2 1 } encode the periodic or anti-periodic spin structure.
For a state of weight μ \mu μ , the signed internal Dirac eigenvalues are
λ n , μ = 1 R ( n + ν + ⟨ μ , ω ⟩ ) , n ∈ Z . (429) \lambda_{n,\mu}=\frac{1}{R}\left(n+\nu+\langle\mu,\omega\rangle\right),\qquad n\in\mathbb{Z}. \tag*{(429)} λ n , μ = R 1 ( n + ν + ⟨ μ , ω ⟩ ) , n ∈ Z . (429)
A zero mode appears when
ν + ⟨ μ , ω ⟩ ∈ Z . (430) \nu+\langle\mu,\omega\rangle\in\mathbb{Z}. \tag*{(430)} ν + ⟨ μ , ω ⟩ ∈ Z . (430)
The union of these loci defines the resonance arrangement of the representation,
H R = ⋃ μ ∈ Wt ( R ) { ω ∣ ν + ⟨ μ , ω ⟩ ∈ Z } . (431) \mathcal{H}_{R}=\bigcup_{\mu\in\operatorname{Wt}(\mathcal{R})}\{\omega\mid\nu+\langle\mu,\omega\rangle\in\mathbb{Z}\}. \tag*{(431)} H R = μ ∈ Wt ( R ) ⋃ { ω ∣ ν + ⟨ μ , ω ⟩ ∈ Z } . (431)
Away from H R \mathcal{H}_{R} H R , the internal Dirac operator is invertible. On a resonance wall its kernel jumps and one or more Kaluza–Klein branches cross zero.
The S U ( 2 ) R SU(2)_{R} S U ( 2 ) R example revisited
Take
U ( ω ) = e 2 π i ω σ 3 , 0 ≤ ω ≤ 1 2 . (432) U(\omega)=\mathrm{e}^{2\pi i\omega\sigma_{3}},\qquad0\leq\omega\leq\frac{1}{2}. \tag*{(432)} U ( ω ) = e 2 π iω σ 3 , 0 ≤ ω ≤ 2 1 . (432)
In the adjoint representation of S U ( 2 ) R SU(2)_{R} S U ( 2 ) R , the non-zero weights are α = ± 2 \alpha=\pm2 α = ± 2 . For ν = 0 \nu=0 ν = 0 , the resonance condition is 2 ω ∈ Z 2\omega\in\mathbb{Z} 2 ω ∈ Z , so inside the fundamental alcove
H a d = { 0 , 1 2 } . (433) \mathcal{H}_{\mathrm{ad}} = \left\{0,\frac{1}{2}\right\}. \tag*{(433)} H ad = { 0 , 2 1 } . (433)
These are exactly the two points where the centralizer of the circle monodromy is enhanced.
The supercharge doublet instead has weights λ = ± 1 \lambda= \pm1 λ = ± 1 . For periodic supersymmetry parameters, the condition is ω ∈ Z \omega\in\mathbb{Z} ω ∈ Z , and hence
H S = { 0 } . (434) \mathcal{H}_{S} = \{0\}. \tag*{(434)} H S = { 0 } . (434)
The half-twist ω = 1 / 2 \omega= 1/2 ω = 1/2 is a symmetry-enhancement wall in the adjoint representation but not a supersymmetry-enhancement wall. A physical fermion can behave differently again, depending on its representation. An adjoint fermion has a resonance at ω = 1 / 2 \omega= 1/2 ω = 1/2 , while a doublet fermion with the same spin structure does not.
Thus
H a d , H S , H R f \mathcal{H}_{\mathrm{ad}}, \qquad\mathcal{H}_{S}, \qquad\mathcal{H}_{R_f} H ad , H S , H R f
are different representation-dependent kernel strata of the same family of internal Dirac operators.
This spectral statement reproduces the group-theoretic distinction found in Questions 11 and 16 from a different viewpoint.
Spectral flow and the eta invariant
A resonance wall is a point at which a fermionic eigenvalue crosses zero. The eta invariant keeps track of the resulting spectral asymmetry.
For one weight, write a = ν + ⟨ μ , ω ⟩ a = \nu+ \langle\mu,\omega\rangle a = ν + ⟨ μ , ω ⟩ . Away from an integer value of a a a , let { a } ∈ ( 0 , 1 ) \{a\} \in(0,1) { a } ∈ ( 0 , 1 ) denote its fractional part. The eta function of the signed circle spectrum is
η a ( s ) = ζ ( s , { a } ) − ζ ( s , 1 − { a } ) , (435) \eta_a(s) = \zeta(s,\{a\}) - \zeta(s,1-\{a\}), \tag*{(435)} η a ( s ) = ζ ( s , { a }) − ζ ( s , 1 − { a }) , (435)
where ζ ( s , q ) \zeta(s,q) ζ ( s , q ) is the Hurwitz zeta function.
Using ζ ( 0 , q ) = 1 2 − q \zeta(0,q) = \frac{1}{2} - q ζ ( 0 , q ) = 2 1 − q , one finds
η a ( 0 ) = 1 − 2 { a } . (436) \eta_a(0) = 1 - 2\{a\}. \tag*{(436)} η a ( 0 ) = 1 − 2 { a } . (436)
The eta invariant is piecewise linear in the twist and jumps by ± 2 \pm2 ± 2 when one eigenvalue crosses zero. The corresponding spectral flow is ± 1 \pm1 ± 1 .
For anomaly questions the relevant quantity is the reduced eta invariant
η ˉ = η + h 2 , h = dim ker D . (437) \bar{\eta} = \frac{\eta+ h}{2}, \qquad h = \dim\ker\mathcal{D}. \tag*{(437)} η ˉ = 2 η + h , h = dim ker D . (437)
At the crossing the kernel dimension changes, and the change of η ˉ \bar{\eta} η ˉ is integral. The exponentiated quantity
exp ( 2 π i η ˉ ) (438) \exp(2\pi i\bar{\eta}) \tag*{(438)} exp ( 2 π i η ˉ ) (438)
therefore remains continuous.
The resonance wall is not a singularity of the anomaly data. What changes is the local description in which the crossing mode has been integrated out.
From spectral flow to the anomaly line
Fix a spacetime X X X and consider a smooth family of lifted twists over the classical representation space X ( Γ , G f ) \mathcal{X}(\Gamma,G_{\mathrm f}) X ( Γ , G f ) .
Locally, away from a resonance wall, the fermion determinant can be represented as an ordinary complex function of the twist parameters. Across different patches these functions need not agree by equality; they can differ by phases. The determinants therefore assemble into a section of a line bundle
L S S ⟶ X ( Γ , G f ) , (439) \mathcal{L}_{\mathrm{SS}} \longrightarrow\mathcal{X}(\Gamma,G_{\mathrm f}), \tag*{(439)} L SS ⟶ X ( Γ , G f ) , (439)
and the partition function is a section
Z X ∈ Γ ( X ( Γ , G f ) , L S S ) . (440) Z_X \in\Gamma(\mathcal{X}(\Gamma,G_{\mathrm f}),\mathcal{L}_{\mathrm{SS}}). \tag*{(440)} Z X ∈ Γ ( X ( Γ , G f ) , L SS ) . (440)
The perturbative anomaly determines the curvature of the natural connection on this line. Global anomalies determine its holonomy around non-contractible loops.
Let ℓ : S 1 → X ( Γ , G f ) \ell:S^1\to\mathcal{X}(\Gamma,G_{\mathrm f}) ℓ : S 1 → X ( Γ , G f ) be such a loop. Transport the background continuously around ℓ \ell ℓ . To compare the initial and final theories, introduce an auxiliary interval parametrizing the path and identify its ends after the corresponding symmetry transformation. The resulting closed ( d + 1 ) (d+1) ( d + 1 ) -dimensional manifold is the mapping torus Y ℓ Y_\ell Y ℓ .
The holonomy of the anomaly line is
Hol ℓ ( L S S ) = A G f ( Y ℓ ) . (441) \operatorname{Hol}_{\ell}(\mathcal{L}_{\mathrm{SS}})=\mathcal{A}_{G_{\mathrm f}}(Y_\ell). \tag*{(441)} Hol ℓ ( L SS ) = A G f ( Y ℓ ) . (441)
For fermionic theories, this phase can be represented by the exponentiated reduced eta invariant together with local inflow terms, schematically
A ( Y ) ∼ exp ( 2 π i η ˉ ( D Y ) ) exp ( 2 π i S i n f l o w ( Y ) ) . (442) \mathcal{A}(Y)\sim\exp(2\pi i\bar{\eta}(D_Y))\exp(2\pi i S_{\mathrm{inflow}}(Y)). \tag*{(442)} A ( Y ) ∼ exp ( 2 π i η ˉ ( D Y )) exp ( 2 π i S inflow ( Y )) . (442)
The overall sign convention in the exponent depends on orientation and chirality conventions. The distinction between trivial and non-trivial holonomy does not.
The resonance walls are therefore zeros of particular determinant sections, not singularities of the anomaly line itself. The anomaly line tells us how the local descriptions on the different sides of those walls fit together globally.
Local and global anomalies are different tests
For chiral fermions in even dimension d d d , the perturbative anomaly is encoded by a degree-( d + 2 ) (d+2) ( d + 2 ) characteristic form. For a complex Weyl fermion in representation R \mathcal{R} R ,
I d + 2 = [ A ^ ( T X ) ch R ( F ) ] d + 2 , (443) I_{d+2}=\left[\widehat{A}(TX)\,\operatorname{ch}_{\mathcal{R}}(F)\right]_{d+2}, \tag*{(443)} I d + 2 = [ A ( T X ) ch R ( F ) ] d + 2 , (443)
with the appropriate chirality sign and normalization. This local index density determines the infinitesimal anomalous variation and the curvature of the anomaly line.
If the local anomaly polynomial cancels, the anomaly line is flat. Flat does not mean trivial: a flat line bundle may still carry non-trivial holonomy around a non-contractible loop.
A familiar example: the S U ( 2 ) SU(2) S U ( 2 ) global anomaly
A standard four-dimensional example makes the distinction concrete. Consider an odd number of left-handed Weyl fermions in the doublet of a gauged S U ( 2 ) SU(2) S U ( 2 ) .
There is no ordinary cubic S U ( 2 ) SU(2) S U ( 2 ) gauge anomaly. Group theoretically, the symmetric invariant d a b c d^{abc} d ab c vanishes for S U ( 2 ) SU(2) S U ( 2 ) , so the familiar perturbative triangle test gives zero.
Nevertheless,
π 4 ( S U ( 2 ) ) ≃ Z 2 . \pi_{4}(SU(2)) \simeq\mathbb{Z}_{2}. π 4 ( S U ( 2 )) ≃ Z 2 .
Under a gauge transformation in the non-trivial homotopy class, the fermion path integral changes sign when the number of Weyl doublets is odd. The theory is quantum mechanically inconsistent despite the complete absence of the perturbative cubic anomaly. This is the S U ( 2 ) SU(2) S U ( 2 ) global anomaly of Witten [88].
The example shows sharply what is meant by a global anomaly. Even after every ordinary local anomaly coefficient has vanished, the fermion determinant can retain a discrete phase under a topologically non-trivial transformation. No perturbative expansion around the identity can detect that phase.
The bordism formulation is the modern generalization of this phenomenon.
From global anomalies to bordism
After perturbative anomalies have been cancelled, the remaining global anomaly is characterized by a homomorphism
α ρ ~ : Ω d + 1 s ( B Γ ) t o r s ⟶ U ( 1 ) , (444) \alpha_{\tilde{\rho}} : \Omega^{\mathfrak{s}}_{d+1}(B\Gamma)_{\mathrm{tors}} \longrightarrow U(1), \tag*{(444)} α ρ ~ : Ω d + 1 s ( B Γ ) tors ⟶ U ( 1 ) , (444)
where s \mathfrak{s} s denotes the relevant Spin, Pin or other tangential structure.
The notation has a direct geometric meaning. A closed ( d + 1 ) (d+1) ( d + 1 ) -dimensional manifold equipped with the required tangential structure and a Γ \Gamma Γ -background defines a possible global anomaly probe. Two such backgrounds are bordant when their disjoint union is the boundary of a ( d + 2 ) (d+2) ( d + 2 ) -dimensional manifold carrying compatible structures. Bordism identifies probes that can be continuously filled in this generalized sense.
The torsion subgroup is the part not detected by integration of ordinary local characteristic forms. It is therefore precisely where purely global phases can survive after the perturbative anomaly polynomial has vanished.
Thus cancellation of I d + 2 I_{d+2} I d + 2 does not imply α ρ ~ = 1 \alpha_{\tilde{\rho}} = 1 α ρ ~ = 1 . The first statement says that the anomaly line is flat; the second says that its remaining global holonomies are trivial.
Checking one generator at a time is not enough
For a non-cyclic identification group, anomaly freedom of each generator separately does not guarantee anomaly freedom of the full twist.
Suppose
Γ = ⟨ γ 1 , … , γ r ∣ R a ( γ 1 , … , γ r ) = 1 ⟩ . \Gamma= \langle\gamma_1,\ldots,\gamma_r \mid R_a(\gamma_1,\ldots,\gamma_r)=1\rangle. Γ = ⟨ γ 1 , … , γ r ∣ R a ( γ 1 , … , γ r ) = 1 ⟩ .
Classically, the images ρ ^ ( γ i ) \widehat{\rho}(\gamma_i) ρ ( γ i ) must satisfy the same relations.
Quantum mechanically, mixed anomaly phases can involve several generators at once. It is therefore possible for every cyclic subgroup ⟨ γ i ⟩ \langle\gamma_i\rangle ⟨ γ i ⟩ to be individually anomaly free while the full Γ \Gamma Γ -action remains obstructed.
The simplest finite-dimensional analogue is a projective representation,
ρ ^ ( γ 1 ) ρ ^ ( γ 2 ) = c ( γ 1 , γ 2 ) ρ ^ ( γ 1 γ 2 ) , (445) \widehat{\rho}(\gamma_1)\widehat{\rho}(\gamma_2)=c(\gamma_1,\gamma_2)\widehat{\rho}(\gamma_1\gamma_2), \tag*{(445)} ρ ( γ 1 ) ρ ( γ 2 ) = c ( γ 1 , γ 2 ) ρ ( γ 1 γ 2 ) , (445)
where c ( γ 1 , γ 2 ) ∈ U ( 1 ) c(\gamma_1,\gamma_2)\in U(1) c ( γ 1 , γ 2 ) ∈ U ( 1 ) . If this phase cannot be removed by rephasing the operators, the group law is realized only projectively.
In field theory, the anomaly theory is the higher-dimensional generalization of this phenomenon. The trivialization τ \tau τ supplies the coherence required for all relations of Γ \Gamma Γ to hold quantum mechanically.
For Γ = Z r \Gamma=\mathbb{Z}^r Γ = Z r , this includes mixed anomalies among commuting monodromies. For Γ = D ∞ \Gamma=D_\infty Γ = D ∞ , it includes the compatibility of the two orbifold reflections, not merely the separate consistency of each endpoint involution.
Fixed loci and anomaly inflow
Orbifolds add another layer because anomalies can be localized on fixed loci.
Let Σ f \Sigma_f Σ f denote the fixed submanifolds. The anomalous variation of the effective action can have the schematic form
δ λ Γ e f f = 2 π i [ ∫ b u l k I b u l k ( 1 ) ( λ ) + ∑ f ∫ Σ f I f ( 1 ) ( λ ) ] . (446) \delta_\lambda\Gamma_{\mathrm{eff}}=2\pi i\left[\int_{\mathrm{bulk}}I_{\mathrm{bulk}}^{(1)}(\lambda)+\sum_f\int_{\Sigma_f}I_f^{(1)}(\lambda)\right]. \tag*{(446)} δ λ Γ eff = 2 π i ∫ bulk I bulk ( 1 ) ( λ ) + f ∑ ∫ Σ f I f ( 1 ) ( λ ) . (446)
Here I b u l k ( 1 ) I_{\mathrm{bulk}}^{(1)} I bulk ( 1 ) and I f ( 1 ) I_f^{(1)} I f ( 1 ) are descent forms and S i n f l o w S_{\mathrm{inflow}} S inflow denotes a bulk term whose variation can compensate the localized anomaly.
Quantum consistency requires the total variation to vanish,
δ λ ( Γ e f f + S i n f l o w ) = 0. (447) \delta_\lambda\left(\Gamma_{\mathrm{eff}}+S_{\mathrm{inflow}}\right)=0. \tag*{(447)} δ λ ( Γ eff + S inflow ) = 0. (447)
The anomaly therefore need not vanish separately in every microscopic sector. A fixed-locus anomaly can be cancelled by the variation of a bulk Chern–Simons, Green–Schwarz or other invertible inflow term.
Conversely, cancellation only after integrating over all fixed loci is not enough if gauge invariance fails locally before the inflow contribution is included.
The distribution of anomaly between bulk and fixed loci is partly scheme dependent. Local Bardeen counterterms or bulk Chern–Simons terms can move where the anomaly appears. The invariant information is the total relative anomaly class and its trivialization, not one particular assignment of anomaly density to individual fixed points.
This mirrors the support analysis of Question 18: local contributions can live on different strata, while consistency is a statement about the complete system.
The Kaluza–Klein tower participates in anomaly matching
Scherk–Schwarz masses do not remove an anomaly.
After compactification, one may separate the light fields from the massive Kaluza–Klein tower. The anomaly of the parent theory restricted to the Scherk–Schwarz background must be reproduced by the complete lower-dimensional effective description. Schematically,
A p a r e n t , ρ ~ ≃ A l i g h t ⊗ A m a s s i v e . (448) \mathcal{A}_{\mathrm{parent},\tilde{\rho}} \simeq\mathcal{A}_{\mathrm{light}} \otimes\mathcal{A}_{\mathrm{massive}}. \tag*{(448)} A parent , ρ ~ ≃ A light ⊗ A massive . (448)
The massive contribution need not appear as a light anomalous fermion. Integrating out Kaluza–Klein states can generate Chern–Simons, Wess–Zumino or eta-invariant terms whose variation supplies the missing anomaly.
A lower-dimensional truncation can look anomalous even when the complete compactification is consistent, because the omitted tower also contains the topological term required for anomaly matching.
The converse is equally important. Giving anomalous fermions Scherk–Schwarz masses does not make the anomaly disappear. Without the corresponding matching topological contribution, the quantum theory remains inconsistent.
Massive states can therefore remain essential to anomaly consistency even when they are absent from the low-energy particle spectrum.
Mapping tori are useful tests, but not the whole classification
A loop in twist space naturally produces a mapping torus, and Eq. (441) evaluates the anomaly on that manifold.
This makes mapping tori particularly useful for testing the transformations that actually appear in a Scherk–Schwarz compactification. They do not necessarily generate the entire relevant bordism group.
Once the perturbative anomaly has been cancelled, complete absence of global anomalies requires
α ρ ~ ( Y ) = 1 for every [ Y ] ∈ Ω d + 1 S p i n ( B Γ ) t o r s . (449) \alpha_{\tilde{\rho}}(Y) = 1 \qquad\text{for every } [Y] \in\Omega_{d+1}^{\mathrm{Spin}}(B\Gamma)_{\mathrm{tors}}. \tag*{(449)} α ρ ~ ( Y ) = 1 for every [ Y ] ∈ Ω d + 1 Spin ( B Γ ) tors . (449)
Passing the mapping-torus tests associated with a selected set of loops proves only that the anomaly vanishes on those particular bordism classes. The full bordism problem depends on the spacetime dimension, the fermionic symmetry extension, the global form of the gauge group and the relevant tangential structure.
The quantum consistency hierarchy
The quantum consistency problem can now be read as a sequence of increasingly refined questions.
One starts from the classical bosonic gluing ρ b : Γ → G b \rho_b:\Gamma\to G_b ρ b : Γ → G b . The first question is whether this gluing admits a coherent fermionic lift ρ ~ : Γ → G f \widetilde{\rho}:\Gamma\to G_f ρ : Γ → G f .
The anomaly of the parent quantum theory is then restricted to the lifted twist, A ρ ~ = ( B ρ ~ ) ∗ A \mathcal{A}_{\widetilde{\rho}}=(B\widetilde{\rho})^*\mathcal{A} A ρ = ( B ρ ) ∗ A , and one asks whether it admits a coherent trivialization τ : A ρ ~ ≃ 1 \tau:\mathcal{A}_{\widetilde{\rho}}\simeq1 τ : A ρ ≃ 1 .
In practice, this can require checking the perturbative anomaly polynomial, mixed anomalies among several identification generators, fixed-locus anomalies and inflow on orbifolds, and the remaining global anomaly on the relevant bordism group.
Only after these steps does the classical gluing define an autonomous quantum Scherk–Schwarz compactification. Physically equivalent choices of ( ρ ~ , τ ) (\widetilde{\rho},\tau) ( ρ , τ ) must finally be identified.
Quantum consistency therefore does not simply delete anomalous points from the classical twist space. It refines each classical point by asking whether a fermionic lift exists, whether its anomaly can be trivialized, and, when it can, which trivialization has been chosen.
The relation between spectral asymmetry, eta invariants and index theory originates in the Atiyah–Patodi–Singer framework [89]. The relation between eta invariants, determinant lines and fermionic anomalies was developed further in the Dai–Freed construction [85]. The interpretation of anomalies in terms of invertible field theories is discussed in Ref. [86], while a particle-physics oriented introduction to Dai–Freed anomalies and their applications can be found in Ref. [87].
The four-dimensional S U ( 2 ) SU(2) S U ( 2 ) example used above is the global gauge anomaly of Ref. [88]. It provides the standard illustration that cancellation of perturbative gauge anomalies need not eliminate global anomalies.
The resonance-wall construction used here combines this anomaly framework with the Scherk–Schwarz twist space of Question 16. The adjoint, supercharge and physical-fermion resonance loci are different representation-dependent kernel strata of the same family of internal Dirac operators. Spectral flow and the eta invariant then organize the fermion determinant as these strata are crossed.
As a concrete string-theory realization of the pullback-anomaly criterion, cyclic asymmetric heterotic orbifolds were analysed in [90]. It was shown that the pulled-back worldsheet Dai–Freed anomaly reproduces the familiar level-matching conditions and, for appropriate even-order actions, additional mixed mod-two constraints. This example illustrates the general framework but does not determine the anomaly theory appropriate to other dimensions, symmetry groups or tangential structures.
Take-home message
A classical bosonic twist must first lift to the full fermionic symmetry, including ( − 1 ) F (-1)^F ( − 1 ) F and the relevant geometric Spin or Pin structure. On a circle the lift is unobstructed, but the two choices differing by ( − 1 ) F (-1)^F ( − 1 ) F give the two spin structures.
A ρ = ( B ρ ) ∗ A . \mathfrak{A}_{\rho}=(B\rho)^*\mathfrak{A}. A ρ = ( B ρ ) ∗ A .
An autonomous quotient requires a coherent trivialization of this anomaly.
Obstruction and choice are different. A twist may admit no quantum completion, or several inequivalent ones. In the elementary cohomological description, δ β = α \delta\beta=\alpha δ β = α expresses cancellation of the anomalous phase, while closed shifts of β \beta β parametrize different trivializations.
ν + ⟨ μ , ω ⟩ ∈ Z . \nu+\langle\mu,\omega\rangle\in\mathbb{Z}. ν + ⟨ μ , ω ⟩ ∈ Z .
Thus residual-symmetry enhancement, supersymmetry enhancement and physical-fermion resonance walls need not coincide.
Crossing a resonance wall produces spectral flow. The eta invariant measures the spectral asymmetry, while its exponentiated reduced form fits continuously through a zero crossing. Globally, the different local determinants are organized as a section of the anomaly line over twist space.
Vanishing of the perturbative anomaly polynomial is not sufficient for quantum consistency. Global anomalies can survive as discrete holonomies, as illustrated by the four-dimensional S U ( 2 ) SU(2) S U ( 2 ) anomaly. Mixed anomalies, orbifold fixed-locus anomalies and inflow, the massive Kaluza–Klein tower, and the relevant bordism classes must all be included in the complete test.
This naturally leads to the next question:
What additional restrictions arise when Scherk–Schwarz breaking is embedded in string theory?
What additional restrictions arise when Scherk–Schwarz breaking is embedded in string theory?
Essential idea
A Scherk–Schwarz twist that is perfectly consistent in a higher-dimensional effective field theory need not define a consistent string background.
The reason is physical. In field theory one specifies how a set of fields is reidentified around a non-contractible cycle. Perturbative string theory must instead define a complete two-dimensional quantum theory. The twist must act consistently not only on the massless fields but also on the massive string states, momentum and winding charges, worldsheet fermions, spin structures, GSO projections, operator products and all twisted sectors required by the quotient.
If C C C denotes the parent worldsheet CFT and T E F T \mathcal{T}_{\mathrm{EFT}} T EFT its low-energy effective theory, a candidate action of a finite group G \mathcal{G} G on the light fields must first lift to an exact action on C C C . Such a microscopic lift can fail to exist, and inequivalent microscopic actions can induce the same low-energy Scherk–Schwarz twist.
If G \mathcal{G} G is actually quotiented, the twisted sectors are not optional. On a torus they are labelled by commuting pairs of holonomies, and modular transformations mix these sectors. Modular closure leads to level-matching conditions, while higher-genus consistency requires factorization and sewing.
String theory also turns many continuous symmetries of the low-energy equations into arithmetic symmetries of an integral charge lattice. A transformation that acts consistently on the massless supergravity fields can therefore fail to be an exact symmetry of the string spectrum.
Finally, the usual spontaneous Scherk–Schwarz interpretation requires more than a non-supersymmetric quotient. The breaking should belong to a dynamical family in which moduli control the supersymmetry-breaking scale and an appropriate limit restores the complete supersymmetric parent background.
Thus the string-theory problem is not merely to choose an admissible monodromy. One must construct its microscopic realization, its twisted sectors and amplitudes, and verify the global quantum consistency of the complete string background.
Why a string is more demanding than a point particle
The additional string conditions can already be understood from the difference between a particle loop and a string loop.
For a point particle moving on a target-space circle of radius R R R , a Scherk–Schwarz boundary condition may be written
Φ ( y + 2 π R ) = U Φ ( y ) , (450) \Phi(y + 2\pi R) = U\Phi(y), \tag*{(450)} Φ ( y + 2 π R ) = U Φ ( y ) , (450)
where U U U is an internal symmetry acting on the field Φ \Phi Φ . Once U U U is specified, the Kaluza–Klein spectral problem is defined.
A closed string propagating at one loop sweeps out a worldsheet torus. The torus itself has two independent non-contractible cycles. A string configuration can carry one orbifold transformation around one worldsheet cycle and another around the second.
This is why a condition such as ψ ( y + 2 π R ) = − ψ ( y ) \psi(y + 2\pi R) = -\psi(y) ψ ( y + 2 π R ) = − ψ ( y ) does not by itself define a string vacuum. It specifies how one spacetime fermion behaves around a target-space cycle, but says nothing about the required worldsheet twisted sectors, their modular transformations, their operator products or their factorization properties.
Thus specifying the anti-periodicity of a spacetime fermion fixes only one piece of the construction. A string background must also supply the worldsheet sectors and amplitudes required to realize that anti-periodicity consistently.
From an EFT symmetry to an exact CFT symmetry
Let C \mathcal{C} C be the complete worldsheet conformal field theory of the supersymmetric parent background, and let T E F T \mathcal{T}_{\mathrm{EFT}} T EFT denote the low-energy theory obtained from its light string states.
By an automorphism of C \mathcal{C} C we mean a transformation that preserves the structures required to define the CFT: the stress tensor, operator products, state space, charge assignments, spin-structure data and projections.
Restriction to the light sector defines schematically
Res I R : Aut ( C ) ⟶ Aut ( T E F T ) . (451) \operatorname{Res}_{\mathrm{IR}} : \operatorname{Aut}(\mathcal{C}) \longrightarrow\operatorname{Aut}(\mathcal{T}_{\mathrm{EFT}}). \tag*{(451)} Res IR : Aut ( C ) ⟶ Aut ( T EFT ) . (451)
Suppose that the low-energy theory admits a candidate action ρ E F T : G ⟶ Aut ( T E F T ) \rho_{\mathrm{EFT}} : G \longrightarrow\operatorname{Aut}(\mathcal{T}_{\mathrm{EFT}}) ρ EFT : G ⟶ Aut ( T EFT ) . A microscopic realization is an action ρ w s : G ⟶ Aut ( C ) \rho_{\mathrm{ws}} : G \longrightarrow\operatorname{Aut}(\mathcal{C}) ρ ws : G ⟶ Aut ( C ) satisfying
Res I R ∘ ρ w s = ρ E F T . (452) \operatorname{Res}_{\mathrm{IR}} \circ\rho_{\mathrm{ws}} = \rho_{\mathrm{EFT}}. \tag*{(452)} Res IR ∘ ρ ws = ρ EFT . (452)
It is useful to denote the inequivalent microscopic realizations by
Lift C ( ρ E F T ) = { ρ w s ∣ Res I R ∘ ρ w s = ρ E F T } / equivalence . (453) \operatorname{Lift}_{\mathcal{C}}(\rho_{\mathrm{EFT}}) = \{\rho_{\mathrm{ws}} \mid\operatorname{Res}_{\mathrm{IR}} \circ\rho_{\mathrm{ws}} = \rho_{\mathrm{EFT}}\}/\text{equivalence}. \tag*{(453)} Lift C ( ρ EFT ) = { ρ ws ∣ Res IR ∘ ρ ws = ρ EFT } / equivalence . (453)
This set can be empty, contain one element, or contain several.
If it is empty, the candidate EFT twist does not exist in the chosen string background. If several lifts exist, the same action on the light fields has more than one microscopic completion.
This possibility is physically natural. Two transformations can act identically on every massless state while acting differently on massive oscillators, momentum–winding states or internal gauge-lattice states. The corresponding string compactifications can therefore share the same leading Scherk–Schwarz EFT while differing in their massive spectra, twisted sectors and threshold corrections.
The low-energy twist retains only part of the microscopic information.
Twisted sectors are worldsheet gauge backgrounds
Assume now that the finite symmetry G \mathcal{G} G is actually used to form an orbifold.
A twisted sector on a connected closed worldsheet Σ \Sigma Σ can be described as a flat principal G \mathcal{G} G -bundle. After choosing a base point, its holonomies define a homomorphism
φ : π 1 ( Σ ) ⟶ G . (454) \varphi:\pi_1(\Sigma)\longrightarrow\mathcal{G}. \tag*{(454)} φ : π 1 ( Σ ) ⟶ G . (454)
Changing the trivialization at the base point conjugates all holonomies simultaneously. The topological sectors are therefore classified by
B G ( Σ ) = Hom ( π 1 ( Σ ) , G ) / G . (455) \mathcal{B}_{\mathcal{G}}(\Sigma)=\operatorname{Hom}(\pi_1(\Sigma),\mathcal{G})/\mathcal{G}. \tag*{(455)} B G ( Σ ) = Hom ( π 1 ( Σ ) , G ) / G . (455)
This representation problem should not be confused with the target-space Scherk–Schwarz representation discussed earlier. There, π 1 ( K ) \pi_1(K) π 1 ( K ) describes transport through the compact target space. Here, π 1 ( Σ ) \pi_1(\Sigma) π 1 ( Σ ) labels orbifold holonomies carried by the worldsheet.
If the action of G \mathcal{G} G on the parent CFT is not faithful, different elements of G \mathcal{G} G can act identically on the local CFT fields. It is nevertheless not always legitimate to quotient G \mathcal{G} G by this kernel. The apparently redundant elements can still label distinct twisted sectors or modify the global orbifold data. We therefore keep track of the microscopic orbifold group together with its specified action on the CFT.
For a closed oriented surface of genus g g g ,
π 1 ( Σ g ) = ⟨ a 1 , b 1 , … , a g , b g | ∏ i = 1 g [ a i , b i ] = 1 ⟩ . (456) \pi_1(\Sigma_g)=\left\langle a_1,b_1,\ldots,a_g,b_g\ \middle|\ \prod_{i=1}^{g}[a_i,b_i]=1\right\rangle. \tag*{(456)} π 1 ( Σ g ) = ⟨ a 1 , b 1 , … , a g , b g i = 1 ∏ g [ a i , b i ] = 1 ⟩ . (456)
A sector is specified by ( g 1 , h 1 , … , g g , h g ) ∈ G 2 g (g_1,h_1,\ldots,g_g,h_g)\in\mathcal{G}^{2g} ( g 1 , h 1 , … , g g , h g ) ∈ G 2 g satisfying
∏ i = 1 g [ g i , h i ] = 1 , (457) \prod_{i=1}^{g}[g_i,h_i]=1, \tag*{(457)} i = 1 ∏ g [ g i , h i ] = 1 , (457)
modulo simultaneous conjugation.
The torus explains why twisted sectors are not optional
At one loop the worldsheet is a torus, with π 1 ( T 2 ) = Z 2 \pi_1(T^2)=\mathbb{Z}^2 π 1 ( T 2 ) = Z 2 . Let g g g and h h h denote the holonomies around two oriented primitive cycles. They must commute.
We denote the corresponding torus amplitude by Z g , h ( τ , τ ˉ ) Z_{g,h}(\tau,\bar{\tau}) Z g , h ( τ , τ ˉ ) , where τ \tau τ is the complex structure modulus. The element g g g specifies the spatially twisted Hilbert space, while h h h is inserted in the trace.
For a finite orbifold group, the torus partition function has the schematic form
Z o r b = 1 ∣ G ∣ ∑ g , h ∈ G g h = h g ϵ ( g , h ) Z g , h , (458) Z_{\mathrm{orb}}=\frac{1}{|\mathcal{G}|}\sum_{\substack{g,h\in\mathcal{G}\\ gh=hg}}\epsilon(g,h)Z_{g,h}, \tag*{(458)} Z orb = ∣ G ∣ 1 g , h ∈ G g h = h g ∑ ϵ ( g , h ) Z g , h , (458)
where ϵ ( g , h ) \epsilon(g,h) ϵ ( g , h ) denotes any allowed relative phase.
The two torus cycles are not physically distinguished. Large orientation-preserving diffeomorphisms change their primitive basis and generate S L ( 2 , Z ) SL(2,\mathbb{Z}) S L ( 2 , Z ) . With the convention adopted here,
S : ( g , h ) ⟼ ( h , g − 1 ) , (459) S:\quad(g,h)\longmapsto(h,g^{-1}), \tag*{(459)} S : ( g , h ) ⟼ ( h , g − 1 ) , (459)
T : ( g , h ) ⟼ ( g , g h ) . (460) T:\quad(g,h)\longmapsto(g,gh). \tag*{(460)} T : ( g , h ) ⟼ ( g , g h ) . (460)
Other orientation conventions redistribute the inverses without changing the underlying physics.
A twisted sector cannot therefore be considered in isolation. Modular transformations generate an orbit of sectors, and a consistent orbifold must contain the complete orbit with mutually compatible amplitudes.
This is the first place where the string problem is visibly more restrictive than a field-theory monodromy. The latter specifies how the retained fields transform. The worldsheet construction must supply every additional sector required for modular closure [34, 35, 91].
Why modular invariance requires level matching
Consider an element g ∈ G g \in G g ∈ G of finite order N N N , and let H g \mathcal{H}_g H g denote its twisted Hilbert space.
A state in this sector has left- and right-moving conformal Hamiltonians L 0 L_0 L 0 and L ˉ 0 \bar{L}_0 L ˉ 0 . Their difference measures the worldsheet momentum around the closed string. Under T : τ ↦ τ + 1 T:\tau\mapsto\tau+1 T : τ ↦ τ + 1 , the state acquires the relative phase e 2 π i ( L 0 − L ˉ 0 ) e^{2\pi i(L_0-\bar{L}_0)} e 2 π i ( L 0 − L ˉ 0 ) .
After N N N applications of T T T , the topological sector returns to itself. In the ordinary cyclic case this gives
N ( L 0 − L ˉ 0 ) ∈ Z . (461) N(L_0-\bar{L}_0)\in\mathbb{Z}. \tag*{(461)} N ( L 0 − L ˉ 0 ) ∈ Z . (461)
For a twisted ground state of conformal weights h g , h ˉ g h_g,\bar{h}_g h g , h ˉ g , one similarly requires
N ( h g − h ˉ g ) ∈ Z . (462) N(h_g-\bar{h}_g)\in\mathbb{Z}. \tag*{(462)} N ( h g − h ˉ g ) ∈ Z . (462)
Fractional oscillator frequencies, zero-point energies, shifted momenta, winding contributions and internal gauge charges all enter L 0 − L ˉ 0 L_0-\bar{L}_0 L 0 − L ˉ 0 . Level matching is a condition on the complete twisted string spectrum rather than on its massless sector.
Its physical meaning is simple: it allows the modular T T T -orbit of the twisted sector to close consistently.
Modular invariance is only the first sewing test
A modularly invariant torus amplitude is necessary, but it is not by itself a definition of a consistent orbifold CFT.
One must first construct the twisted Hilbert spaces H g \mathcal{H}_g H g , specify how the centralizer of g g g acts in each sector, choose the allowed phases and define operator products involving twisted fields.
String amplitudes must also exist on worldsheets of arbitrary genus. When a surface degenerates or is cut along a cycle, the amplitude must factorize into amplitudes of the resulting pieces, with the correct sum over intermediate states.
Schematically, a consistent orbifold assigns an amplitude
( Σ , P ) ⟼ Z C [ Σ , P ] (463) (\Sigma,P)\longmapsto Z_{\mathcal{C}}[\Sigma,P] \tag*{(463)} ( Σ , P ) ⟼ Z C [ Σ , P ] (463)
to every allowed closed worldsheet equipped with a flat G \mathcal{G} G -bundle P P P . These amplitudes must transform consistently under mapping-class groups and satisfy the required sewing relations.
The torus partition function is an initial quantum test of the construction, not its complete definition.
For open or unoriented strings, further amplitudes enter. D-branes and orientifold planes require Klein-bottle, annulus and Möbius-strip amplitudes, together with consistent boundary and crosscap states. Their modular channels and factorization properties must be compatible with the Scherk–Schwarz action, and tadpole cancellation can impose additional constraints. We shall not develop those conditions here.
The charge lattice makes the symmetry problem arithmetic
String theory also restricts which transformations can be exact symmetries in the first place.
In a toroidal compactification, left- and right-moving charges belong to an integral lattice
Λ ⊂ R d L , d R . (464) \Lambda\subset\mathbb{R}^{d_L,d_R}. \tag*{(464)} Λ ⊂ R d L , d R . (464)
For a Narain compactification, this lattice is even and self-dual with respect to a bilinear form η \eta η of signature ( d L , d R ) (d_L,d_R) ( d L , d R ) .
The continuum equations admit O ( d L , d R ; R ) O(d_L,d_R;\mathbb{R}) O ( d L , d R ; R ) , but an exact string symmetry must preserve the spectrum of allowed charges. Hence it must satisfy g Λ = Λ g\Lambda=\Lambda g Λ = Λ and g T η g = η g^{T}\eta g=\eta g T η g = η . The exact symmetry group is therefore reduced to the lattice automorphism group
O ( Λ ) = { g ∈ O ( d L , d R ; R ) ∣ g Λ = Λ } . (465) O(\Lambda)=\{g\in O(d_L,d_R;\mathbb{R})\mid g\Lambda=\Lambda\}. \tag*{(465)} O ( Λ ) = { g ∈ O ( d L , d R ; R ) ∣ g Λ = Λ } . (465)
In an integral basis this becomes the appropriate arithmetic group usually denoted O ( d L , d R ; Z ) O(d_L,d_R;\mathbb{Z}) O ( d L , d R ; Z ) .
The physical origin of the restriction is immediate. A continuous transformation may act perfectly on every massless supergravity field while sending a massive momentum–winding state to a charge vector that does not belong to the string spectrum. Such a transformation is not an exact symmetry of the microscopic theory.
Lattice preservation is still not sufficient. The symmetry must also lift consistently to vertex operators and their cocycle phases [92, 93].
The role of the complete momentum–winding charge lattice, including its relation to Scherk–Schwarz shifts, is the subject of the next question. Discrete restrictions on rigid twists will be discussed thereafter.
Fermions, GSO projections, and asymmetric actions
The microscopic symmetry must also act consistently on the worldsheet fermionic sectors.
Worldsheet fermions possess different spin structures around the cycles of the worldsheet, and perturbative superstring theories sum over them with phases determined by the GSO projection. Spacetime supercharges arise from particular Ramond-sector states. A transformation acting on spacetime supersymmetry must have a consistent action on the worldsheet fermions, the Ramond and Neveu–Schwarz sectors, the GSO projection and any accompanying gauge or lattice degrees of freedom.
This is the microscopic reason why imposing an anti-periodic boundary condition on a spacetime fermion is not yet a string construction.
Exact string symmetries need not, however, be geometric. A toroidal CFT can admit an asymmetric action
( X L , X R ) ↦ ( g L X L , g R X R ) , g L ≠ g R , (466) (X_L,X_R) \mapsto(g_L X_L,g_R X_R), \qquad g_L \ne g_R, \tag*{(466)} ( X L , X R ) ↦ ( g L X L , g R X R ) , g L = g R , (466)
where X L X_L X L and X R X_R X R are the left- and right-moving components of an internal coordinate.
Such a transformation may have no interpretation as an isometry of a classical target-space manifold and yet be an exact automorphism of the full CFT. String symmetry is therefore broader than geometric symmetry. At the same time, it remains more restrictive than symmetry of the low-energy equations.
The anomaly condition of Question 20 reappears microscopically
Question 20 asked whether the symmetry used for Scherk–Schwarz gluing survives as a consistent quantum operation.
In a worldsheet orbifold this becomes concrete: quotienting by G \mathcal{G} G gauges the corresponding finite symmetry of the CFT. An anomalous action can obstruct the quotient even if each group element preserves the classical operator algebra.
Failures of modular covariance, level matching or higher-genus sewing can be manifestations of such a microscopic obstruction. In a class of cyclic asymmetric heterotic orbifolds, for example, a Dai–Freed analysis reproduces the familiar level-matching condition and, for even-order actions, additional mixed mod-two constraints [90].
This gives a concrete realization of the relation between the general anomaly language of Question 20 and worldsheet consistency. It does not imply that every string consistency condition is encoded in one anomaly invariant.
When does the string construction describe spontaneous Scherk–Schwarz breaking?
A consistent non-supersymmetric quotient is not automatically a spontaneously broken Scherk–Schwarz background.
The first question is whether the quotient acts non-trivially on the spacetime supercharges at all. If some supercharges remain invariant, the corresponding supersymmetry survives.
If all are removed, supersymmetry is broken, but the spontaneous interpretation requires additional dynamical information, as discussed in Question 5. In particular, restoration of supersymmetry in some limiting background is not by itself sufficient; the parameter controlling the breaking should belong to the dynamical moduli of the same underlying theory.
The standard perturbative string realization combines an internal action g R g_R g R with a translation δ \delta δ along a compact direction,
g S S = g R δ . (467) g_{\mathrm{SS}} = g_R \,\delta. \tag*{(467)} g SS = g R δ . (467)
The combined quotient is then freely acting.
If g R g_R g R acts non-trivially on all spacetime supercharges, the background breaks supersymmetry. When the translation radius R R R is a modulus, the breaking scale typically behaves as m S S ∼ 1 / R m_{\mathrm{SS}} \sim1/R m SS ∼ 1/ R . In the decompactification limit R → ∞ R \to\infty R → ∞ , finite-energy spectra and amplitudes approach those of the supersymmetric parent background.
Free action and supersymmetry restoration therefore play different roles. The free translation ensures that the breaking is not tied to a fixed locus, while the dynamical radius provides the modulus that controls the breaking scale. When supersymmetry is restored as this modulus is taken to the appropriate limit, the construction has the standard string-theoretic interpretation of spontaneous Scherk–Schwarz breaking [13, 16, 17].
What can the low-energy EFT test, and what can it miss?
A Wilsonian effective theory derived from a consistent string background retains information about heavy string states through higher-dimension operators, thresholds and topological terms. Schematically,
S W = S 2 ∂ + ∑ i c i M s Δ i − D ∫ d D x O i + S t o p + ⋯ . (468) S_{\mathrm{W}} = S_{2\partial} + \sum_i \frac{c_i}{M_{\mathrm{s}}^{\Delta_i-D}} \int d^D x\, O_i + S_{\mathrm{top}} + \cdots. \tag*{(468)} S W = S 2 ∂ + i ∑ M s Δ i − D c i ∫ d D x O i + S top + ⋯ . (468)
An exact Scherk–Schwarz symmetry must preserve the complete Wilsonian data. A transformation that is only a symmetry of the two-derivative action can still provide a useful approximate low-energy description, but does not thereby define an exact string quotient.
Some microscopic consistency conditions are visible in the EFT. Anomaly matching, quantized topological terms, charges of retained states and symmetry relations among Wilson coefficients can all rule out proposed twists. Other constraints require information that a finite truncation does not contain: the complete momentum–winding lattice, massive oscillator states, twisted Hilbert spaces, modular completion or higher-genus sewing data.
The logical asymmetry is therefore important. Failure of a low-energy consistency test can rule out a proposed microscopic realization. Passing every test visible in a finite EFT does not prove that a microscopic string lift exists. Likewise, when several string lifts induce the same leading Scherk–Schwarz EFT, they can still be distinguished by their heavy spectra, threshold corrections, topological terms and higher-dimension operators. The map from microscopic string backgrounds to finite low-energy truncations is many-to-one and does not admit a general inverse.
The construction of string orbifolds, including twisted sectors, modular covariance, level matching and discrete-torsion phases, was developed in the classic orbifold literature; see Refs. [34, 35, 91]. The description in terms of commuting worldsheet holonomies makes explicit why a single target-space monodromy does not define the complete string quotient. Toroidal compactifications provide the standard setting in which continuous symmetries of the low-energy equations are reduced to arithmetic automorphisms of an integral momentum–winding lattice [92, 93]. Freely acting orbifolds provide the standard perturbative string realizations of Scherk–Schwarz supersymmetry breaking and its restoration along a modulus-dependent family [13, 16, 17]. The anomaly framework of Question 20 gives a complementary description of some microscopic orbifold constraints. In suitable asymmetric constructions, the pulled-back worldsheet anomaly reproduces level matching and additional discrete consistency conditions [90].
Take-home message
Hom ( π 1 ( Σ ) , G ) / G . \operatorname{Hom}\left(\pi_{1}(\Sigma), \mathcal{G}\right)/\mathcal{G}. Hom ( π 1 ( Σ ) , G ) / G .
On the torus this reduces to commuting pairs ( g , h ) (g,h) ( g , h ) , which are mixed by modular transformations and therefore occur in complete modular orbits.
For a cyclic twist of order N N N , modular closure requires level matching, N ( L 0 − L ˉ 0 ) ∈ Z N(L_{0}-\bar{L}_{0}) \in\mathbb{Z} N ( L 0 − L ˉ 0 ) ∈ Z . Genus-one modular covariance is necessary but not sufficient: factorization and sewing must also hold at higher genus.
Exact toroidal string symmetries must preserve the integral charge lattice. Continuous symmetries of the low-energy equations are reduced to arithmetic automorphism groups, and lattice preservation must be supplemented by consistent vertex-operator, fermionic and GSO data.
A finite low-energy EFT can rule out a proposed string lift when one of its necessary consistency conditions fails, but passing every test visible below the cutoff cannot establish that the required charge lattice, twisted sectors and worldsheet amplitudes exist.
This naturally leads to the next question:
How is Scherk–Schwarz breaking encoded in the string charge lattice?
How is Scherk–Schwarz breaking encoded in the string charge lattice?
Essential idea
In field theory, a Scherk–Schwarz deformation is often summarized by the shift
m ⟶ m + ξ q , m \longrightarrow m+\xi q, m ⟶ m + ξ q ,
where m m m is a Kaluza–Klein momentum number and q q q an internal charge.
A string state carries more information. Momentum, winding, gauge charges and the charges of worldsheet currents belong to one Lorentzian charge system. A consistent deformation cannot shift one entry while leaving the others untouched, because the quadratic form controlling level matching and the string mass formula must be preserved.
An ordinary gauge Wilson line gives the simplest example. If a string of winding number n n n experiences the gauge shift P ⟶ P + n A P \longrightarrow P+nA P ⟶ P + n A , preservation of the charge-lattice norm forces
m ⟶ m − A ⋅ P − 1 2 A 2 n . m \longrightarrow m-A\cdot P-\frac{1}{2}A^{2}n. m ⟶ m − A ⋅ P − 2 1 A 2 n .
The linear and quadratic terms are therefore parts of one orthogonal transformation, rather than independent corrections to the mass formula.
The same charge-space geometry applies to a coordinate-dependent Scherk–Schwarz deformation once the charge acting on spacetime supersymmetry is included among the internal charges. The familiar field-theory momentum shift is then the zero-winding limit of a Lorentzian transformation of the complete string charge vector.
Gauge Wilson lines and Scherk–Schwarz deformations are thus described by the same type of parabolic isometry of an enlarged charge space, although they act on different internal directions and have different physical meanings.
Not every low-energy R R R -symmetry charge can be used in this way. The charge must arise from an admissible dimension-one worldsheet current, act non-trivially on the sector generating spacetime supersymmetry, and transform consistently under the orbifold. These worldsheet conditions are logically distinct from the integrality condition determining whether the real charge-space isometry preserves the exact string lattice.
This question develops the charge-space geometry. The additional conditions that make the deformation rigid or discrete are treated in the next question.
Why the field-theory shift is incomplete
Consider a field-theory tower whose masses depend on a Kaluza–Klein number m m m and an internal charge q q q through m K K ∼ ( m + ξ q ) / R m_{\mathrm{KK}} \sim(m+\xi q)/R m KK ∼ ( m + ξ q ) / R . At this level, it is natural to think that the Scherk–Schwarz deformation simply translates one momentum quantum number.
A closed string on the same circle also carries winding. In a heterotic compactification it may carry an internal gauge-lattice charge as well. The labels m m m , n n n and P P P are not independent decorations of a field-theory state. They are components of a single string charge vector.
The question consequently changes. One must ask not only how the Kaluza–Klein momentum is shifted, but whether the complete transformation preserves the quadratic form of the string charge lattice.
The charge vector behind a toroidal string state
For definiteness, consider the heterotic string compactified on a d d d -dimensional torus. A state carries integer Kaluza–Klein numbers m i m_i m i , integer winding numbers n i n^i n i , and a gauge charge P P P in the sixteendimensional gauge lattice.
The corresponding left- and right-moving momenta combine with P P P into a Narain charge vector P = ( P ; p L , p R ) \mathcal{P}=(P;p_L,p_R) P = ( P ; p L , p R ) . The allowed vectors form an even self-dual lattice Λ N ⊂ R d + 16 , d \Lambda_{\mathrm{N}}\subset\mathbb{R}^{d+16,d} Λ N ⊂ R d + 16 , d .
Evenness means that the norm of every lattice vector is an even integer in the conventional normalization. Self-duality means that every vector pairing integrally with the lattice already belongs to the lattice.
The Lorentzian quadratic form is
⟨ P , P ⟩ = P 2 + p L 2 − p R 2 . (469) \langle\mathcal{P},\mathcal{P}\rangle=P^2+p_L^2-p_R^2. \tag*{(469)} ⟨ P , P ⟩ = P 2 + p L 2 − p R 2 . (469)
It enters both the level-matching condition and the string mass formula.
For one compact circle, a basis adapted to momentum and winding reduces the relevant part of the norm to
P 2 + 2 m n , (470) P^2+2mn, \tag*{(470)} P 2 + 2 mn , (470)
up to the overall normalization convention.
This simple expression already shows why a shift of P P P or m m m cannot be considered in isolation.
A Wilson line shows why the quadratic term is necessary
Let A A A be a gauge Wilson line along the compact circle. A string with winding number n n n goes around the circle n n n times and experiences the gauge holonomy on each traversal. Its gauge charge is therefore shifted to
P A = P + n A . (471) P_A=P+nA. \tag*{(471)} P A = P + n A . (471)
If this were the only change, the gauge contribution to the lattice norm would become
( P + n A ) 2 = P 2 + 2 n A ⋅ P + n 2 A 2 , (P+nA)^2=P^2+2nA\cdot P+n^2A^2, ( P + n A ) 2 = P 2 + 2 n A ⋅ P + n 2 A 2 ,
and the original norm would not be preserved.
The compact momentum must change at the same time:
m A = m − A ⋅ P − 1 2 A 2 n . (472) m_A = m - A \cdot P - \frac{1}{2}A^2 n. \tag*{(472)} m A = m − A ⋅ P − 2 1 A 2 n . (472)
Indeed,
P A 2 + 2 n m A = ( P + n A ) 2 + 2 n ( m − A ⋅ P − 1 2 A 2 n ) = P 2 + 2 m n . (473) \begin{aligned}
P_A^2 + 2n m_A &= (P+nA)^2 + 2n\left(m-A\cdot P-\frac{1}{2}A^2n\right) \\
&= P^2 + 2mn. \tag*{(473)}
\end{aligned} P A 2 + 2 n m A = ( P + n A ) 2 + 2 n ( m − A ⋅ P − 2 1 A 2 n ) = P 2 + 2 mn . (473)
The coefficient of the quadratic term is fixed by this equality. It is not an optional string correction added to the linear field-theory shift.
In terms of p L p_L p L and p R p_R p R , the same combination m − A ⋅ P − 1 2 A 2 n m-A\cdot P-\frac{1}{2}A^2n m − A ⋅ P − 2 1 A 2 n appears in both momenta, with the winding contribution entering them with opposite signs.
The preceding transformation has a useful coordinate-independent form.
Let W W W be an internal charge space with bilinear form ( ⋅ , ⋅ ) W (\cdot,\cdot)_W ( ⋅ , ⋅ ) W . Add two null vectors e e e and f f f satisfying
e 2 = f 2 = 0 , e ⋅ f = 1 , e ⋅ W = f ⋅ W = 0. (474) e^2=f^2=0,\qquad e\cdot f=1,\qquad e\cdot W=f\cdot W=0. \tag*{(474)} e 2 = f 2 = 0 , e ⋅ f = 1 , e ⋅ W = f ⋅ W = 0. (474)
The pair ( e , f ) (e,f) ( e , f ) represents one momentum–winding plane, and the full real charge space is V = W ⊕ span R { e , f } V=W\oplus\operatorname{span}_{\mathbb{R}}\{e,f\} V = W ⊕ span R { e , f } .
For a ∈ W a\in W a ∈ W , define
T a ( e ) = e , T a ( x ) = x − ( a , x ) W e , x ∈ W , T a ( f ) = f + a − 1 2 ( a , a ) W e . (475) \begin{aligned}
T_a(e)&=e,\\
T_a(x)&=x-(a,x)_W e,\qquad x\in W,\\
T_a(f)&=f+a-\frac{1}{2}(a,a)_W e. \tag*{(475)}
\end{aligned} T a ( e ) T a ( x ) T a ( f ) = e , = x − ( a , x ) W e , x ∈ W , = f + a − 2 1 ( a , a ) W e . (475)
The coefficient of e e e in the last line is fixed by requiring the image of f f f to remain null:
T a ( f ) 2 = ( f + a − 1 2 a 2 e ) 2 = a 2 − 2 ( 1 2 a 2 ) ( f ⋅ e ) = 0. (476) \begin{aligned}
T_a(f)^2&=\left(f+a-\frac{1}{2}a^2e\right)^2\\
&=a^2-2\left(\frac{1}{2}a^2\right)(f\cdot e)\\
&=0. \tag*{(476)}
\end{aligned} T a ( f ) 2 = ( f + a − 2 1 a 2 e ) 2 = a 2 − 2 ( 2 1 a 2 ) ( f ⋅ e ) = 0. (476)
Here f 2 = e 2 = f ⋅ a = e ⋅ a = 0 f^2=e^2=f\cdot a=e\cdot a=0 f 2 = e 2 = f ⋅ a = e ⋅ a = 0 and f ⋅ e = 1 f\cdot e=1 f ⋅ e = 1 . The remaining scalar products are preserved directly, so T a ∈ O ( V ) T_a\in O(V) T a ∈ O ( V ) .
The transformation is a unipotent parabolic isometry fixing the null direction e e e , and may be viewed as an Eichler-transvection-type transformation associated with that null direction.
Writing a charge vector as v = P + m e + n f v = P + m e + n f v = P + m e + n f and taking a = A a = A a = A , Eq. (475) reproduces
P → P + n A , m → m − A ⋅ P − 1 2 A 2 n . P \to P + nA,\qquad m \to m - A \cdot P - \frac{1}{2}A^{2}n. P → P + n A , m → m − A ⋅ P − 2 1 A 2 n .
The component Wilson-line formula is therefore simply the coordinate expression of one orthogonal transformation.
Adding the charge that acts on spacetime supersymmetry
A Scherk–Schwarz deformation requires a charge that distinguishes states belonging to the same spacetime supermultiplet.
Let J S S ( z ) J_{\mathrm{SS}}(z) J SS ( z ) be a dimension-one worldsheet current and Q S S Q_{\mathrm{SS}} Q SS its zero-mode charge. The microscopic requirement is that this charge act non-trivially on the worldsheet sector from which the spacetime supercharges are constructed.
In the heterotic constructions considered below, the infinitesimal current charge therefore acts non-trivially on the relevant two-dimensional supercurrent G w s G_{\mathrm{ws}} G ws . At the level of the local generator this can be expressed as
[ Q S S , G w s ] ≠ 0. (477) [Q_{\mathrm{SS}},G_{\mathrm{ws}}] \ne0. \tag*{(477)} [ Q SS , G ws ] = 0. (477)
This does not mean that the finite quotient violates worldsheet supersymmetry. The finite transformation used in the compactification must preserve the super-Virasoro structure. When the local generator does not commute with G w s G_{\mathrm{ws}} G ws , only selected finite rotations can have this property, as discussed in Question 23.
The charge space must now be enlarged to include the eigenvalues of Q S S Q_{\mathrm{SS}} Q SS . Denote the resulting internal charge vector by Q Q Q , and let ξ \xi ξ specify the Scherk–Schwarz direction in this space.
For one compact circle, the complete transformation is
Q ′ = Q − n ξ , m ′ = m + ξ ⋅ Q − 1 2 ξ 2 n , n ′ = n . (478) \begin{aligned}
Q' &= Q - n\xi,\\
m' &= m + \xi\cdot Q - \frac{1}{2}\xi^{2}n,\\
n' &= n. \tag*{(478)}
\end{aligned} Q ′ m ′ n ′ = Q − n ξ , = m + ξ ⋅ Q − 2 1 ξ 2 n , = n . (478)
With the convention of Eq. (475) , this is T − ξ T_{-\xi} T − ξ . Setting a = − ξ a=-\xi a = − ξ gives Q ′ = Q + n a = Q − n ξ Q' = Q + na = Q - n\xi Q ′ = Q + na = Q − n ξ and m ′ = m − a ⋅ Q − 1 2 a 2 n m' = m - a \cdot Q - \frac{1}{2}a^{2}n m ′ = m − a ⋅ Q − 2 1 a 2 n .
The transformation preserves the extended quadratic form:
( Q − n ξ ) 2 + 2 n ( m + ξ ⋅ Q − 1 2 ξ 2 n ) = Q 2 + 2 m n . (479) (Q-n\xi)^{2} + 2n\left(m+\xi\cdot Q-\frac{1}{2}\xi^{2}n\right)=Q^{2}+2mn. \tag*{(479)} ( Q − n ξ ) 2 + 2 n ( m + ξ ⋅ Q − 2 1 ξ 2 n ) = Q 2 + 2 mn . (479)
In the zero-winding sector, n = 0 n=0 n = 0 , the charge is unchanged and the momentum becomes m ′ = m + ξ ⋅ Q m' = m+\xi\cdot Q m ′ = m + ξ ⋅ Q . For a single U ( 1 ) U(1) U ( 1 ) charge q q q , this is precisely the field-theory rule m → m + ξ q m \to m+\xi q m → m + ξ q .
The winding-dependent terms are the additional string information required by preservation of the full charge-space norm.
With the conventions used above, a gauge Wilson line is represented by T A T_A T A , while the Scherk–Schwarz deformation is represented by T − ξ T_{-\xi} T − ξ . Once gauge and Scherk–Schwarz charges are included in the same internal charge space W W W , both are parabolic transvections fixing the same momentum–winding null direction.
For any a , b ∈ W a,b \in W a , b ∈ W ,
T a T b = T a + b = T b T a . (480) T_aT_b=T_{a+b}=T_bT_a. \tag*{(480)} T a T b = T a + b = T b T a . (480)
No orthogonality assumption on a a a and b b b is required. The mixed term a ⋅ b a \cdot b a ⋅ b is already contained in − 1 2 ( a + b ) e 2 -\frac{1}{2}(a+b)^2_{\mathrm e} − 2 1 ( a + b ) e 2 . Thus T A T_A T A and T − ξ T_{-\xi} T − ξ commute even when their charge-space directions are not orthogonal.
Their commutativity in this setting is a property of the underlying charge-space transformations, not an accidental cancellation in the component mass formula.
This conclusion relies on both deformations belonging to the Abelian unipotent subgroup associated with one common null direction. More general backgrounds can involve several null directions or non-Abelian current algebras, for which the corresponding transformations need not commute.
Which worldsheet currents are admissible?
The existence of an R R R -symmetry generator in the low-energy supergravity is not sufficient to produce a string Scherk–Schwarz deformation.
The candidate charge must be represented by a dimension-one worldsheet current. It must act non-trivially on the sector from which spacetime supersymmetry is constructed, while the finite transformation used in the quotient must remain an exact operation of the worldsheet theory.
An orbifold imposes a further compatibility condition. Suppose an orbifold generator ϑ \vartheta ϑ acts on a compact coordinate direction i i i through a representation R i ( ϑ ) R_i(\vartheta) R i ( ϑ ) . A Scherk–Schwarz current attached to that direction must transform in the corresponding representation:
ϑ J S S ϑ − 1 = R i ( ϑ ) J S S . (481) \vartheta J_{\mathrm{SS}}\vartheta^{-1}=R_i(\vartheta)J_{\mathrm{SS}}. \tag*{(481)} ϑ J SS ϑ − 1 = R i ( ϑ ) J SS . (481)
This allows the coordinate-dependent deformation and the orbifold projection to be imposed simultaneously.
Current covariance and lattice integrality answer different questions. Equation (22.13) tests whether the candidate current fits the orbifold action. Requiring the associated transvection to preserve the integral string lattice tests whether it defines an exact microscopic charge transformation. The example below verifies the first condition, not the second.
A concrete Z 4 \mathbb{Z}_4 Z 4 eigen-current
Consider a heterotic Z 4 \mathbb{Z}_4 Z 4 orbifold with twist vector
ν = 1 4 ( 1 , 1 , − 2 ) . (482) \nu=\frac{1}{4}(1,1,-2). \tag*{(482)} ν = 4 1 ( 1 , 1 , − 2 ) . (482)
Since the three components sum to zero, the rotation lies in S U ( 3 ) SU(3) S U ( 3 ) , as required for the standard supersymmetric parent orbifold.
Let f 1 f_1 f 1 , f 2 f_2 f 2 , f 3 f_3 f 3 be three complex internal worldsheet fermions, and let ψ \psi ψ be a worldsheet fermion associated with a non-compact spacetime direction. The orbifold acts as
f 1 ↦ i f 1 , f 2 ↦ i f 2 , f 3 ↦ − f 3 , ψ ↦ ψ . (483) f_1 \mapsto i f_1,\qquad f_2 \mapsto i f_2,\qquad f_3 \mapsto-f_3,\qquad\psi\mapsto\psi. \tag*{(483)} f 1 ↦ i f 1 , f 2 ↦ i f 2 , f 3 ↦ − f 3 , ψ ↦ ψ . (483)
A current associated with the first compact complex direction is
J 1 = 1 2 [ : f 1 † Re f 3 : + : f 1 ψ : ] , (484) J^1 = \frac{1}{\sqrt{2}}\left[ : f_1^\dagger\operatorname{Re} f_3 : + : f_1\psi: \right], \tag*{(484)} J 1 = 2 1 [ : f 1 † Re f 3 : + : f 1 ψ : ] , (484)
where the colons denote normal ordering.
Each fermion has conformal weight 1 / 2 1/2 1/2 , so both bilinears have conformal weight one.
Under the orbifold, f 1 † ↦ − i f 1 † f_1^\dagger\mapsto-i f_1^\dagger f 1 † ↦ − i f 1 † and Re f 3 ↦ − Re f 3 \operatorname{Re} f_3 \mapsto-\operatorname{Re} f_3 Re f 3 ↦ − Re f 3 . Hence
: f 1 † Re f 3 : ↦ ( − i ) ( − 1 ) : f 1 † Re f 3 : = i : f 1 † Re f 3 : . (485) : f_1^\dagger\operatorname{Re} f_3 : \mapsto(-i)(-1) : f_1^\dagger\operatorname{Re} f_3 :
= i : f_1^\dagger\operatorname{Re} f_3 : . \tag*{(485)} : f 1 † Re f 3 :↦ ( − i ) ( − 1 ) : f 1 † Re f 3 := i : f 1 † Re f 3 : . (485)
The second bilinear transforms in the same way, : f 1 ψ : ↦ i : f 1 ψ : : f_1\psi: \mapsto i : f_1\psi: : f 1 ψ :↦ i : f 1 ψ : . Therefore
ϑ J 1 ϑ − 1 = i J 1 . (486) \vartheta J^1 \vartheta^{-1} = i J^1 . \tag*{(486)} ϑ J 1 ϑ − 1 = i J 1 . (486)
This is the phase carried by the first compact complex direction, e 2 π i v 1 = i e^{2\pi i v_1} = i e 2 π i v 1 = i . The current therefore has precisely the covariance required to associate the Scherk–Schwarz charge with that direction.
The current has been written in a complex eigenbasis. Its Hermitian conjugate has eigenvalue − i -i − i ; in a real basis the corresponding two current components are rotated into one another by the Z 4 \mathbb{Z}_4 Z 4 generator.
The occurrence of the spacetime fermion ψ \psi ψ is also important, because it allows the current charge to act on the sector involved in the construction of spacetime supersymmetry [94].
What the massive charge lattice can remember
Projection to the massless effective theory loses much of the charge-lattice information.
A Wilson line can remove a gauge-lattice state from the massless spectrum while leaving it at a mass of order 1 / R 1/R 1/ R when the corresponding compact direction is large. Its charge remains the original lattice vector P P P , and need not coincide with the gauge quantum numbers of any massless field.
For zero winding, its momentum is shifted by A ⋅ P A \cdot P A ⋅ P . After choosing the nearest integer Kaluza–Klein number, the lightest state in that charge sector has a mass of order
dist ( A ⋅ P , Z ) R . \frac{\operatorname{dist}(A \cdot P,\mathbb{Z})}{R}. R dist ( A ⋅ P , Z ) .
A fractional value of A ⋅ P A \cdot P A ⋅ P can therefore produce states below the first integer Kaluza–Klein excitation of the massless spectrum.
Knowledge of the massless gauge algebra is not sufficient to reconstruct these states. One must know how the Wilson line is embedded in the complete charge lattice.
The same loss of information occurs for supersymmetry. A four-dimensional orbifold may leave only N = 1 N = 1 N = 1 supersymmetry in the massless sector, while an untwisted subsector descends from a parent sector with N = 2 N = 2 N = 2 or N = 4 N = 4 N = 4 . Its Kaluza–Klein excitations can still assemble into extended supermultiplets before the Scherk–Schwarz splitting is applied.
The complete tower can retain cancellations and representation-theoretic information absent from the massless four-dimensional spectrum, providing a microscopic counterpart of the tower-level cancellations discussed in Question 17.
The affine charge transformation acts directly in sectors whose states are described by the momentum, winding and current charges included in the extended lattice.
In the symmetric Abelian heterotic orbifolds for which the construction above was developed, ordinary fixed-point twisted states do not carry the same internal momentum quantum numbers as the untwisted sector. They therefore do not receive the same tree-level Scherk–Schwarz shift [94].
This is a property of that class of models, not a theorem about every string orbifold. Freely acting orbifolds, asymmetric constructions and more general CFT realizations can assign different momentum, winding or fractional charges to their twisted sectors.
The action of the Scherk–Schwarz deformation must consequently be determined sector by sector from the complete worldsheet construction. The charge-lattice formula is exact in the sectors to which its charge vector applies; it should not be extended to other sectors by analogy alone.
From the real charge space to the exact string lattice
The analysis above identifies the real Lorentzian transformation behind the string Scherk–Schwarz shift. The linear momentum shift and its quadratic winding correction are not independent ingredients: they are fixed components of one parabolic isometry.
This is not yet sufficient to define an exact string operation. The isometry must preserve the integral lattice of physical charges, and the associated current must remain compatible with the orbifold, GSO projection and operator algebra.
These arithmetic and worldsheet conditions determine which continuous transformations of the real charge space survive as exact string operations. They are the subject of the next question.
The organization of momentum, winding and gauge charges into an even self-dual Lorentzian lattice is the basis of the Narain description of toroidal string compactification [92, 93].
Coordinate-dependent compactifications can be described in terms of transformations acting on enlarged charge and momentum data; related formulations can be found in Refs. [14, 15, 16, 95].
A treatment in which gauge Wilson lines and Scherk–Schwarz supersymmetry breaking act simultaneously on an extended heterotic charge vector, together with explicit orbifold restrictions on the supersymmetry-breaking current and the dependence of light Kaluza–Klein quantum numbers on the Wilson-line embedding, can be found in Ref. [94].
The same coordinate-dependent compactification language has also been used in later heterotic constructions, where the deformation acts on spacetime, internal R R R -symmetry and gauge charges and determines the shifted Kaluza–Klein spectrum [96].
Take-home message
Momentum, winding and internal charges form a single Lorentzian string charge vector. A Scherk–Schwarz deformation must therefore preserve the complete charge-space quadratic form, not only the Kaluza–Klein spectrum of the zero-winding sector.
P → P + n A , m → m − A ⋅ P − 1 2 A 2 n . P \to P+nA,\qquad m \to m-A\cdot P-\frac{1}{2}A^2n. P → P + n A , m → m − A ⋅ P − 2 1 A 2 n .
The linear and quadratic shifts are fixed components of one parabolic isometry of the momentum–winding charge space.
Q → Q − n ξ , m → m + ξ ⋅ Q − 1 2 ξ 2 n . Q \to Q-n\xi,\qquad m \to m+\xi\cdot Q-\frac{1}{2}\xi^2n. Q → Q − n ξ , m → m + ξ ⋅ Q − 2 1 ξ 2 n .
At zero winding this reduces to the field-theory shift m → m + ξ q m \to m+\xi q m → m + ξ q .
This naturally leads to the next question:
When is a rigid Scherk–Schwarz twist discrete, and how can the effective phase nevertheless vanish continuously?
When is a rigid Scherk–Schwarz twist discrete, and how can the effective phase nevertheless vanish continuously?
Essential idea
The mass scale remains continuous even for a quantized twist.
The half-twist gives the simplest example. If the microscopic identification acts as − 1 -1 − 1 on the gravitino, the Kaluza–Klein tower is
m n = ∣ n + 1 2 ∣ R , m 3 / 2 = 1 2 R . (487) m_n = \frac{\left|n+\frac{1}{2}\right|}{R}, \qquad m_{3/2} = \frac{1}{2R}. \tag*{(487)} m n = R n + 2 1 , m 3/2 = 2 R 1 . (487)
The twist remains exactly fixed as the radius varies, while m 3 / 2 → 0 m_{3/2} \to0 m 3/2 → 0 for R → ∞ R \to\infty R → ∞ .
The dimensionless twist and the dimensionful Kaluza–Klein scale are therefore different data. More generally, the physical Scherk–Schwarz gap seen by a given branch is
m S S = M K K dist ( ω e f f , Z ) , (488) m_{\mathrm{SS}} = M_{\mathrm{KK}}\,\operatorname{dist}\left(\omega_{\mathrm{eff}}, \mathbb{Z}\right), \tag*{(488)} m SS = M KK dist ( ω eff , Z ) , (488)
where M K K M_{\mathrm{KK}} M KK is the relevant Kaluza–Klein spacing and ω e f f ∈ R / Z \omega_{\mathrm{eff}} \in\mathbb{R}/\mathbb{Z} ω eff ∈ R / Z is the phase actually seen by that branch.
The microscopic twist ω r i g i d \omega_{\mathrm{rigid}} ω rigid can be quantized by the finite-order group action, the integral string charge lattice, the worldsheet operator algebra and modular consistency. The effective phase need not coincide with this rigid datum: continuous Wilson-line moduli, dynamical backgrounds or localized interactions can also contribute to the spectral boundary condition.
A small Scherk–Schwarz gap can arise in two different ways. The Kaluza–Klein scale may tend to zero while the effective phase remains of order one, as in decompactification. Alternatively, continuous dynamics can drive ω e f f \omega_{\mathrm{eff}} ω eff close to an integer while the rigid microscopic twist remains fixed.
These two mechanisms can have very different relations to the tower of states that limits the lower-dimensional description.
A fixed twist and a variable mass scale
The half-twist already resolves the most immediate apparent paradox. The microscopic identification remains U S S ∣ ψ 3 / 2 ⟩ = − ∣ ψ 3 / 2 ⟩ U_{\mathrm{SS}}|\psi_{3/2}\rangle= -|\psi_{3/2}\rangle U SS ∣ ψ 3/2 ⟩ = − ∣ ψ 3/2 ⟩ ; nothing is continuously dialled between U S S = − 1 U_{\mathrm{SS}}=-1 U SS = − 1 and the identity. The mass nevertheless depends continuously on the radius through Eq. (487) .
If R R R is regarded as an external parameter, the limit R → ∞ R \to\infty R → ∞ compares a family of compactification backgrounds. If the radius is a dynamical modulus, its expectation value controls the breaking scale within the lower-dimensional theory, as discussed in Questions 10 and 21.
The restoration limit also changes the range of validity of the lower-dimensional description. The Kaluza–Klein tower behaves as
m n K K = ∣ n ∣ R ⟶ 0 ( R → ∞ ) . (489) m_n^{\mathrm{KK}} = \frac{\lvert n\rvert}{R} \longrightarrow0 \qquad(R \to\infty). \tag*{(489)} m n KK = R ∣ n ∣ ⟶ 0 ( R → ∞ ) . (489)
A finite four-dimensional truncation therefore ceases to be an appropriate description near the limit. Supersymmetry is restored because the physical scale multiplying the fixed identification tends to zero, not because the microscopic twist approaches the identity.
Why finite-order string twists are discrete
Let U ξ U_{\xi} U ξ be the group element implementing the microscopic Scherk–Schwarz identification. In a quotient of order N N N , it must satisfy
U ξ N = 1. (490) U_{\xi}^{N} = 1. \tag*{(490)} U ξ N = 1. (490)
If a string state ∣ Ψ ⟩ \lvert\Psi\rangle ∣ Ψ ⟩ has eigenphase U ξ ∣ Ψ ⟩ = e 2 π i Ω Ψ ∣ Ψ ⟩ U_{\xi}\lvert\Psi\rangle= e^{2\pi i\Omega_{\Psi}}\lvert\Psi\rangle U ξ ∣ Ψ ⟩ = e 2 π i Ω Ψ ∣ Ψ ⟩ , then (490) requires
N Ω Ψ ∈ Z (491) N\Omega_{\Psi} \in\mathbb{Z} \tag*{(491)} N Ω Ψ ∈ Z (491)
for every state in the complete string Hilbert space.
For a one-parameter action with Ω Ψ = ω r i g i d q Ψ \Omega_{\Psi} = \omega_{\mathrm{rigid}}q_{\Psi} Ω Ψ = ω rigid q Ψ , this becomes
N ω r i g i d q Ψ ∈ Z for every charge q Ψ in the string spectrum . (492) N\omega_{\mathrm{rigid}}q_{\Psi} \in\mathbb{Z} \qquad\text{for every charge } q_{\Psi} \text{ in the string spectrum}. \tag*{(492)} N ω rigid q Ψ ∈ Z for every charge q Ψ in the string spectrum . (492)
The massive charge spectrum can therefore impose restrictions that are invisible in the massless effective theory.
The charge-by-charge condition above can be expressed intrinsically as a statement about the group element itself. The charges q Ψ q_{\Psi} q Ψ are the weights with which a maximal torus acts on the string Hilbert space, so requiring every eigenphase to have order dividing N N N is equivalent to requiring the corresponding point of the torus to be N N N -torsion.
Let T = t / Λ c o c h a r T = \mathfrak{t}/\Lambda_{\mathrm{cochar}} T = t / Λ cochar be a compact maximal torus, where t \mathfrak{t} t is its Lie algebra and Λ c o c h a r \Lambda_{\mathrm{cochar}} Λ cochar its cocharacter lattice. If
U S S = exp ( 2 π i ω r i g i d ⋅ H ) , U_{\mathrm{SS}} = \exp\left(2\pi i\,\omega_{\mathrm{rigid}}\cdot H\right), U SS = exp ( 2 π i ω rigid ⋅ H ) ,
then U S S U_{\mathrm{SS}} U SS has order dividing N N N precisely when
N ω r i g i d ∈ Λ c o c h a r . (493) N\omega_{\mathrm{rigid}} \in\Lambda_{\mathrm{cochar}}. \tag*{(493)} N ω rigid ∈ Λ cochar . (493)
For fixed N N N , the allowed points are the N N N -torsion points of the torus, T [ N ] = N − 1 Λ c o c h a r / Λ c o c h a r T[N] = N^{-1}\Lambda_{\mathrm{cochar}}/\Lambda_{\mathrm{cochar}} T [ N ] = N − 1 Λ cochar / Λ cochar .
This is only the first microscopic test. The transformation must also preserve the integral string charge lattice, lift consistently to the vertex operators and their cocycles, act consistently in every twisted sector, and satisfy the modular, sewing and anomaly conditions discussed in Questions 20 and 21.
Thus the set of twists is discrete for a fixed finite-order construction. Allowing the order N N N , the orbifold group or the parent CFT to change does not create a continuous modulus within one fixed theory.
A worldsheet current need not generate a continuous string symmetry
The same quantization can be understood directly on the worldsheet.
Suppose a dimension-one current has zero-mode charge Q S S Q_{\mathrm{SS}} Q SS . One can formally write e i α Q S S e^{i\alpha Q_{\mathrm{SS}}} e i α Q SS for any real α \alpha α . This does not imply that every α \alpha α defines an automorphism of the complete string theory.
For a Scherk–Schwarz charge, Q S S Q_{\mathrm{SS}} Q SS must act non-trivially on the sector from which spacetime supersymmetry is constructed. In the coordinate-dependent heterotic constructions discussed in Question 22, this can be expressed at the level of the local generator as
[ Q S S , G w s ] ≠ 0 (494) [Q_{\mathrm{SS}},G_{\mathrm{ws}}] \ne0 \tag*{(494)} [ Q SS , G ws ] = 0 (494)
where G w s G_{\mathrm{ws}} G ws is the relevant worldsheet supercurrent.
A generic value of α \alpha α therefore need not define an automorphism of the full worldsheet theory. Isolated finite values can nevertheless do so, because their action may close on the complete operator algebra and preserve the GSO and super-Virasoro structures.
In the simplest example, the local fields entering the worldsheet supercurrent have integer charge under Q S S Q_{\mathrm{SS}} Q SS , whereas the spin fields creating spacetime fermions have half-integer charge. For
U = e 2 π i Q S S , U=e^{2\pi iQ_{\mathrm{SS}}}, U = e 2 π i Q SS ,
a field of integer charge q q q acquires the phase e 2 π i q = 1 e^{2\pi iq}=1 e 2 π i q = 1 , while a field of half-integer charge acquires e 2 π i q = − 1 e^{2\pi iq}=-1 e 2 π i q = − 1 . The same finite transformation can therefore act trivially on the local worldsheet algebra and as − 1 -1 − 1 on spacetime fermions.
This produces the anti-periodic Scherk–Schwarz spectrum without requiring a continuous U ( 1 ) U(1) U ( 1 ) symmetry of the complete string theory.
The Z 4 \mathbb{Z}_4 Z 4 calculation in Question 22 tested a related but different condition: the candidate current had the required orbifold covariance. Current covariance, finite-order exponentiation and preservation of the integral charge lattice are distinct microscopic requirements.
Continuous Wilson-line moduli and discrete orbifold embeddings
The term “Wilson line” can describe two objects with different microscopic status.
Before orbifolding, a surviving Wilson-line zero mode can be a continuous Narain modulus. Its holonomy can be written
W ( θ ) = exp ( 2 π i θ ⋅ H ) , θ ∈ t / Λ c o c h a r , W(\theta)=\exp(2\pi i\,\theta\cdot H),\qquad\theta\in\mathfrak{t}/\Lambda_{\mathrm{cochar}}, W ( θ ) = exp ( 2 π i θ ⋅ H ) , θ ∈ t / Λ cochar ,
with a further Weyl identification when conjugate holonomies are identified. The parameter θ \theta θ is then the expectation value of a dynamical scalar field, and moving it changes the background continuously within the same moduli space.
A discrete Wilson line used in an orbifold embedding has a different status. If it is associated with an element of order N N N , it obeys a lattice condition of the form
N A ∈ Λ g a u g e , (495) NA\in\Lambda_{\mathrm{gauge}}, \tag*{(495)} N A ∈ Λ gauge , (495)
together with model-dependent orbifold and modular constraints. It is a rigid microscopic choice rather than a continuously variable zero mode.
A continuous Wilson-line modulus can nevertheless contribute to the phase seen by a particular state. Its contribution belongs to the continuous background data, not to the discrete label of the microscopic quotient.
The spectrum can see more than the rigid twist
The statement that a quantized ω r i g i d \omega_{\mathrm{rigid}} ω rigid admits no continuous small-ω r i g i d \omega_{\mathrm{rigid}} ω rigid limit is correct when the spectrum depends only on that rigid datum. It need not remain true for the effective phase entering the complete spectral problem.
Localized interactions give a concrete lower-dimensional example. In the conventions of Question 9, let M 0 M_0 M 0 and M π M_\pi M π be the dimensionless coefficients of localized gravitino bilinears at the two ends of an interval. The tower is
m n = 1 R ∣ n + ω r i g i d + 1 π [ arctan ( M 0 ) + arctan ( M π ) ] ∣ . (496) m_n = \frac{1}{R}\left|n+\omega_{\mathrm{rigid}}+\frac{1}{\pi}\left[\arctan(M_0)+\arctan(M_\pi)\right]\right|. \tag*{(496)} m n = R 1 n + ω rigid + π 1 [ arctan ( M 0 ) + arctan ( M π ) ] . (496)
The phase seen by this spectrum is therefore
ω e f f = ω r i g i d + 1 π [ arctan ( M 0 ) + arctan ( M π ) ] ( m o d 1 ) . (497) \omega_{\mathrm{eff}}=\omega_{\mathrm{rigid}}+\frac{1}{\pi}\left[\arctan(M_0)+\arctan(M_\pi)\right]\pmod{1}. \tag*{(497)} ω eff = ω rigid + π 1 [ arctan ( M 0 ) + arctan ( M π ) ] ( mod 1 ) . (497)
The rigid monodromy can remain fixed while the localized contributions vary through dynamical expectation values. There need not be a continuous small-ω r i g i d \omega_{\mathrm{rigid}} ω rigid limit. What can become continuously small is the spectral mismatch produced by the sum of rigid and dynamical contributions.
(497) concerns the quadratic spectrum. It does not imply that a rigid twist and a localized supersymmetry-breaking sector define the same microscopic theory. Their Goldstino content, local interactions and ultraviolet completion can remain different even when they give the same ω e f f \omega_{\mathrm{eff}} ω eff .
Approaching a cancellation locus non-perturbatively
Let Φ \Phi Φ denote the moduli and other dynamical fields controlling the additional contribution to the phase, and write
ω e f f ( Φ ) = ω r i g i d + ω d y n ( Φ ) ( m o d 1 ) . \omega_{\mathrm{eff}}(\Phi)=\omega_{\mathrm{rigid}}+\omega_{\mathrm{dyn}}(\Phi)\pmod{1}. ω eff ( Φ ) = ω rigid + ω dyn ( Φ ) ( mod 1 ) .
A supersymmetric spectral locus is reached when ω r i g i d + ω d y n ( Φ ) ∈ Z \omega_{\mathrm{rigid}}+\omega_{\mathrm{dyn}}(\Phi)\in\mathbb{Z} ω rigid + ω dyn ( Φ ) ∈ Z . Near such a locus, let k ∈ Z k\in\mathbb{Z} k ∈ Z denote the integer approached by ω e f f \omega_{\mathrm{eff}} ω eff , and write
ω d y n ( Φ ) = k − ω r i g i d + δ ω n p ( Φ ) . (498) \omega_{\mathrm{dyn}}(\Phi)=k-\omega_{\mathrm{rigid}}+\delta\omega_{\mathrm{np}}(\Phi). \tag*{(498)} ω dyn ( Φ ) = k − ω rigid + δ ω np ( Φ ) . (498)
The exponentially small term is not, in general, what cancels an order-one rigid twist. The perturbative or order-one dynamical contributions must first place the system close to the cancellation locus. The residual displacement can then be non-perturbatively small,
∣ δ ω n p ( Φ ) ∣ ∼ exp [ − Re S n p ( Φ ) ] . (499) \left|\delta\omega_{\mathrm{np}}(\Phi)\right|\sim\exp\left[-\operatorname{Re}S_{\mathrm{np}}(\Phi)\right]. \tag*{(499)} ∣ δ ω np ( Φ ) ∣ ∼ exp [ − Re S np ( Φ ) ] . (499)
A familiar source of exponentially small terms is non-perturbative gauge dynamics,
W n p ∼ A ( Φ ) e − a f ( Φ ) . (500) W_{\mathrm{np}} \sim A(\Phi)e^{-a f(\Phi)}. \tag*{(500)} W np ∼ A ( Φ ) e − a f ( Φ ) . (500)
The gauge kinetic function f ( Φ ) f(\Phi) f ( Φ ) can depend on the dilaton, geometric moduli or several fields. How this contribution feeds into a localized mass or an effective Scherk–Schwarz phase is model dependent, but whenever it does, the residual displacement from the cancellation locus can inherit the exponential suppression e − Re S n p e^{-\operatorname{Re} S_{\mathrm{np}}} e − Re S np . This suppression is therefore not universally controlled by the string coupling alone.
When ω r i g i d = 0 \omega_{\mathrm{rigid}}=0 ω rigid = 0 , a small non-perturbative contribution can instead generate the entire effective phase. Near a cancellation locus one then has parametrically
m S S ∼ M K K e − Re S n p ( Φ ) . (501) m_{\mathrm{SS}} \sim M_{\mathrm{KK}}e^{-\operatorname{Re} S_{\mathrm{np}}(\Phi)}. \tag*{(501)} m SS ∼ M KK e − Re S np ( Φ ) . (501)
Comparing the supersymmetry-breaking scale with the tower
To distinguish the possible restoration mechanisms without committing to a particular modulus, it is useful to compare the supersymmetry-breaking gap directly with the first infinite tower that limits the lower-dimensional description.
Let m t o w e r m_{\mathrm{tower}} m tower be the characteristic scale of that tower along the trajectory under consideration. Define
Ξ = m t o w e r m S S = m t o w e r M K K 1 dist ( ω e f f , Z ) . (502) \Xi= \frac{m_{\mathrm{tower}}}{m_{\mathrm{SS}}}
= \frac{m_{\mathrm{tower}}}{M_{\mathrm{KK}}}\frac{1}{\operatorname{dist}(\omega_{\mathrm{eff}},\mathcal{Z})}. \tag*{(502)} Ξ = m SS m tower = M KK m tower dist ( ω eff , Z ) 1 . (502)
For ordinary decompactification, the light tower is the Kaluza–Klein tower itself, with m t o w e r ∼ M K K ∼ 1 / R m_{\mathrm{tower}}\sim M_{\mathrm{KK}}\sim1/R m tower ∼ M KK ∼ 1/ R . If the dimensionless twist remains of order one, then
Ξ = O ( 1 ) . (503) \Xi= O(1). \tag*{(503)} Ξ = O ( 1 ) . (503)
The tower and the Scherk–Schwarz gap collapse at the same parametric rate.
At a finite-distance cancellation locus, the tower scale can remain finite while m S S → 0 m_{\mathrm{SS}}\to0 m SS → 0 , and Ξ \Xi Ξ diverges.
Along an asymptotic non-perturbative trajectory, (499) instead gives
Ξ ∼ m t o w e r M K K e Re S n p ( Φ ) . (504) \Xi\sim\frac{m_{\mathrm{tower}}}{M_{\mathrm{KK}}}e^{\operatorname{Re} S_{\mathrm{np}}(\Phi)}. \tag*{(504)} Ξ ∼ M KK m tower e Re S np ( Φ ) . (504)
If m t o w e r / M K K m_{\mathrm{tower}}/M_{\mathrm{KK}} m tower / M KK varies more slowly than the exponential, supersymmetry is restored parametrically faster than the tower descends.
The ratio above can become parametrically large in a concrete asymptotic regime. To illustrate this possibility, consider a weak-coupling trajectory in which the non-perturbative action is controlled by the dilaton. Suppose Re S n p ∼ a / g s p \operatorname{Re} S_{\mathrm{np}}\sim a/g_s^p Re S np ∼ a / g s p , with a , p > 0 a,p>0 a , p > 0 , while the dimensionless internal volume v = V 6 M s 6 v=V_6M_s^6 v = V 6 M s 6 is held fixed. Then
M P l 2 ∼ v g s 2 M s 2 , M s M P l ∼ g s v . (505) M_{\mathrm{Pl}}^2 \sim\frac{v}{g_s^2}M_s^2,\qquad
\frac{M_s}{M_{\mathrm{Pl}}}\sim\frac{g_s}{\sqrt{v}}. \tag*{(505)} M Pl 2 ∼ g s 2 v M s 2 , M Pl M s ∼ v g s . (505)
Along this trajectory, the perturbative string tower becomes light only as a power of g s g_s g s in Planck units, while the residual Scherk–Schwarz phase can be exponentially small. A parametrically large hierarchy can therefore open,
m S S ≪ m t o w e r ≪ M P l . (506) m_{\mathrm{SS}} \ll m_{\mathrm{tower}} \ll M_{\mathrm{Pl}}. \tag*{(506)} m SS ≪ m tower ≪ M Pl . (506)
This example is deliberately specific. If the non-perturbative action is controlled mainly by another modulus, the accompanying tower and its scaling in Planck units must be determined along that trajectory. The invariant statement is the comparison in Eq. (502) , not a universal identification of S n p S_{\mathrm{np}} S np with the dilaton.
What is and is not quantized
The finite-order group element defining a rigid string quotient can be quantized, while the moduli controlling the Kaluza–Klein scale, surviving Wilson lines or localized dynamical contributions remain continuous. The phase entering a particular spectrum can contain both discrete and continuous data.
A small Scherk–Schwarz gap can consequently arise from decompactification, from cancellation in the effective phase, or from both. These mechanisms are distinguished by the accompanying tower of states and by the microscopic origin of the dynamical contribution. If additional independent supersymmetry-breaking sectors are present, closing the Scherk–Schwarz gap alone does not restore the complete supersymmetry.
The distinction between continuous Narain deformations and discrete coordinate-dependent compactifications is discussed in early string implementations of Scherk–Schwarz breaking; related formulations can be found in Refs. [14, 15, 16, 95].
A treatment combining gauge Wilson lines and coordinate-dependent Scherk–Schwarz deformations in an enlarged heterotic charge lattice, including discrete restrictions imposed by Abelian orbifolds, can be found in Ref. [94].
Coordinate-dependent compactifications and discrete charge vectors also appear in later heterotic constructions. For example, Ref. [96] uses a CDC vector acting on spacetime, internal R R R -symmetry and gauge charges to determine the shifted Kaluza–Klein spectrum.
The distinction emphasized here between a quantized rigid twist and a modulus-dependent effective spectral phase is more general than these specific constructions. Localized masses provide one explicit lower-dimensional realization; their microscopic origin and possible non-perturbative dependence are model dependent.
Take-home message
A finite-order microscopic twist is discrete because its action must close on the complete string spectrum. For an order-N N N quotient, N ω r i g i d q Ψ ∈ Z N\omega_{\mathrm{rigid}}q_{\Psi} \in\mathbb{Z} N ω rigid q Ψ ∈ Z for every string charge q Ψ q_{\Psi} q Ψ , together with the lattice, worldsheet and modular consistency conditions.
A discrete twist does not imply a discrete supersymmetry-breaking scale. At the half-twist, m 3 / 2 = 1 / ( 2 R ) m_{3/2} = 1/(2R) m 3/2 = 1/ ( 2 R ) , so the mass vanishes in the decompactification limit while the microscopic twist remains fixed.
ω e f f = ω r i g i d + 1 π [ arctan ( M 0 ) + arctan ( M π ) ] ( m o d 1 ) . \omega_{\mathrm{eff}} = \omega_{\mathrm{rigid}} + \frac{1}{\pi}\left[\arctan(M_0) + \arctan(M_{\pi})\right] \pmod{1}. ω eff = ω rigid + π 1 [ arctan ( M 0 ) + arctan ( M π ) ] ( mod 1 ) .
Near a dynamical cancellation locus, the residual mismatch can be non-perturbatively small. For an order-one rigid twist, the exponential term controls the residual only after order-one contributions have already brought the system close to that locus.
The ratio m t o w e r / m S S m_{\mathrm{tower}}/m_{\mathrm{SS}} m tower / m SS distinguishes restoration by decompactification from restoration by cancellation. It is typically O ( 1 ) O(1) O ( 1 ) for a fixed order-one twist in decompactification, but can become parametrically large when the effective phase approaches an integer much faster than the relevant tower descends.
This naturally leads to the next question:
How does a freely acting orbifold realize stringy Scherk–Schwarz breaking?
How does a freely acting orbifold realize stringy Scherk–Schwarz breaking?
Essential idea
A freely acting orbifold gives an exact string realization of the Scherk–Schwarz mechanism by combining an internal transformation that acts on spacetime supercharges with a translation along a compact direction.
Take a covering circle of radius N R NR N R , with y ∼ y + 2 π N R y \sim y + 2\pi NR y ∼ y + 2 π N R , and let δ N : y ↦ y + 2 π R \delta_N : y \mapsto y + 2\pi R δ N : y ↦ y + 2 π R be its order-N N N translation. If g R g_R g R is an internal symmetry satisfying g R N = 1 g_R^N = 1 g R N = 1 , the combined generator is
g S S = g R δ N . (507) g_{\mathrm{SS}} = g_R \delta_N . \tag*{(507)} g SS = g R δ N . (507)
Every non-trivial power of the translation moves every point of the covering circle. In the elementary product construction, the Scherk–Schwarz subgroup therefore has no fixed points, even when g R g_R g R has fixed points on another factor of the compactification. The breaking is global rather than localized on a defect.
The quotient of the covering circle by δ N \delta_N δ N is a smooth circle of radius R R R . The internal action survives as a flat holonomy. If a field component obeys g R Φ = e 2 π i q / N Φ g_R\Phi= e^{2\pi i q/N}\Phi g R Φ = e 2 π i q / N Φ , then one circuit around the quotient circle gives
Φ ( y + 2 π R ) = e − 2 π i q / N Φ ( y ) , (508) \Phi(y + 2\pi R) = e^{-2\pi i q/N}\Phi(y), \tag*{(508)} Φ ( y + 2 π R ) = e − 2 π i q / N Φ ( y ) , (508)
and hence
p y , m = m − q / N R , m ∈ Z . (509) p_{y,m} = \frac{m-q/N}{R}, \qquad m \in\mathbb{Z}. \tag*{(509)} p y , m = R m − q / N , m ∈ Z . (509)
The Scherk–Schwarz shift is therefore not added after the orbifold has been constructed: it is the exact spectrum implied by the equivariance condition on the smooth quotient.
Freeness does not remove twisted sectors. It changes their geometric character: their non-trivial information is carried by winding along the translated direction rather than by localization at a fixed point.
At large radius the holonomy remains fixed while the Scherk–Schwarz mass splittings vanish. The supersymmetric parent is then recovered at finite energy even though the microscopic orbifold generator never approaches the identity.
Why adding a translation changes the orbifold
An ordinary orbifold transformation can leave points of the internal space fixed. Its twisted sectors may then contain states localized near those fixed loci, and local operators can be supported there.
A translation behaves differently. For 1 ≤ k < N 1 \le k < N 1 ≤ k < N , δ N k \delta_N^k δ N k shifts the covering coordinate by 2 π k R 2\pi kR 2 π k R . This displacement is non-zero modulo the covering circumference 2 π N R 2\pi NR 2 π N R , so
Fix ( g S S k ) = ∅ , 1 ≤ k < N . (510) \operatorname{Fix}\left(g_{\mathrm{SS}}^k\right)=\varnothing,\qquad1 \le k < N. \tag*{(510)} Fix ( g SS k ) = ∅ , 1 ≤ k < N . (510)
The statement concerns the complete action g R k δ N k g_R^k\delta_N^k g R k δ N k . The internal transformation g R k g_R^k g R k can have fixed points on another factor of the compactification, but no point of the product is fixed while its circle coordinate is translated.
The quotient therefore introduces no fixed-point singularity associated with the Scherk–Schwarz subgroup. It replaces the covering circle of radius N R NR N R by a circle of radius R R R , while retaining the internal action through a global boundary condition on the fields.
How the twist survives on the smooth quotient
Consider a component of a field in the g R g_R g R -eigenspace with eigenvalue e 2 π i q / N e^{2\pi iq/N} e 2 π i q / N . Invariance under the combined identification requires the translation to compensate this internal phase, giving Eq. (508) .
Transport the field once around the quotient circle. It returns to the same spacetime point but to a value differing by e − 2 π i q / N e^{-2\pi iq/N} e − 2 π i q / N .
Mathematically, the field is a section of a flat complex line bundle, or equivalently of the corresponding one-dimensional local system. Its holonomy is
ρ q ( 1 ) = e − 2 π i q / N , (511) \rho_q(1)=e^{-2\pi iq/N}, \tag*{(511)} ρ q ( 1 ) = e − 2 π i q / N , (511)
where 1 1 1 denotes one positively oriented circuit of the quotient circle.
The underlying complex line bundle is topologically trivial, since H 2 ( S 1 , Z ) = 0 H^2(S^1,\mathbb{Z})=0 H 2 ( S 1 , Z ) = 0 . Its flat connection can nevertheless have non-trivial holonomy, which cannot be removed by a single-valued field redefinition unless q = 0 q=0 q = 0 modulo N N N .
The compatible modes are
Φ m ( y ) = exp [ i m − q / N R y ] , m ∈ Z . (512) \Phi_m(y)=\exp\left[i\frac{m-q/N}{R}y\right],\qquad m\in\mathbb{Z}. \tag*{(512)} Φ m ( y ) = exp [ i R m − q / N y ] , m ∈ Z . (512)
Their shifted Kaluza–Klein spectrum is the direct spectral consequence of the flat holonomy.
The covering-space projection gives the same spectrum
The same result can be derived directly on the covering circle.
A covering-space momentum eigenmode is
exp ( i n N R y ) , n ∈ Z . \exp\left(i\frac{n}{NR}y\right),\qquad n\in\mathbb{Z}. exp ( i N R n y ) , n ∈ Z .
Under δ N \delta_N δ N , it acquires the phase e 2 π i n / N e^{2\pi i n/N} e 2 π in / N . Multiplication by the internal eigenphase gives the total phase e 2 π i ( n + q ) / N e^{2\pi i(n+q)/N} e 2 π i ( n + q ) / N .
Orbifold invariance requires
n + q = 0 ( m o d N ) . (513) n+q=0 \pmod{N}. \tag*{(513)} n + q = 0 ( mod N ) . (513)
Writing n = N m − q n=Nm-q n = N m − q gives
n N R = m − q / N R , \frac{n}{NR}=\frac{m-q/N}{R}, N R n = R m − q / N ,
which is exactly Eq. (509) .
The covering-space projection and the quotient-space holonomy are therefore two exact descriptions of the same orbifold spectrum.
The half-shift benchmark
The simplest supersymmetry-breaking example has N = 2 N=2 N = 2 . The covering circle has radius 2 R 2R 2 R , and the generator combines a half-period translation with spacetime fermion parity,
g S S = ( − 1 ) F s δ 2 , δ 2 : y ⟼ y + 2 π R . (514) g_{\mathrm{SS}}=(-1)^{F_s}\delta_2,\qquad\delta_2:y\longmapsto y+2\pi R. \tag*{(514)} g SS = ( − 1 ) F s δ 2 , δ 2 : y ⟼ y + 2 π R . (514)
Here F s F_s F s denotes spacetime fermion number. Its difference from a continuous R R R -symmetry twist is explained in Question 26. Only the resulting spectral shift is needed below.
Bosons have q = 0 q=0 q = 0 , so their momenta are p B , m = m / R p_{B,m}=m/R p B , m = m / R . Fermions have q = 1 q=1 q = 1 , giving p F , m = ( m − 1 2 ) / R p_{F,m}=(m-\frac{1}{2})/R p F , m = ( m − 2 1 ) / R . After an integer relabelling of m m m , this becomes
p F , m = m + 1 2 R . (515) p_{F,m}=\frac{m+\frac{1}{2}}{R}. \tag*{(515)} p F , m = R m + 2 1 . (515)
The lightest fermionic mass is therefore 1 / ( 2 R ) 1/(2R) 1/ ( 2 R ) , exactly as in the half-twist analysis of Question 11. The same spectrum is now understood as the difference between trivial holonomy for spacetime bosons and holonomy − 1 -1 − 1 for spacetime fermions on the same smooth quotient circle.
Twisted sectors carry fractional winding
A freely acting orbifold still has twisted sectors. What changes is their geometric interpretation.
In the sector twisted by g S S k g_{\mathrm{SS}}^k g SS k , with 1 ≤ k < N 1\leq k<N 1 ≤ k < N , the full worldsheet fields obey the corresponding g R k g_R^k g R k boundary condition. Along the translated circle,
y ( σ + 2 π ) = y ( σ ) + 2 π N R ( w + k N ) , w ∈ Z . (516) y(\sigma+2\pi)=y(\sigma)+2\pi NR\left(w+\frac{k}{N}\right),\qquad w\in\mathbb{Z}. \tag*{(516)} y ( σ + 2 π ) = y ( σ ) + 2 π N R ( w + N k ) , w ∈ Z . (516)
The winding is fractional when measured relative to the covering circle. On the quotient circle the same string winds N w + k Nw+k N w + k times, so the sector label k k k selects a congruence class of ordinary quotient windings.
The state is not localized on a fixed locus of the full Scherk–Schwarz element. It can still be localized in internal directions fixed by g R k g_R^k g R k , but it necessarily extends or winds along the translated circle.
The winding contribution to its mass scales as
M w i n d ∼ N R α ′ ∣ w + k N ∣ = R α ′ ∣ N w + k ∣ (517) M_{\mathrm{wind}} \sim\frac{NR}{\alpha'}\left|w+\frac{k}{N}\right|=\frac{R}{\alpha'}\left|Nw+k\right| \tag*{(517)} M wind ∼ α ′ N R w + N k = α ′ R ∣ N w + k ∣ (517)
up to oscillator, zero-point and additional lattice contributions. For non-trivial k k k , the minimum of ∣ N w + k ∣ \left|Nw+k\right| ∣ N w + k ∣ is non-zero, so these sectors are generally driven to infinite mass as R → ∞ R \to\infty R → ∞ .
Freeness therefore changes the nature of the twisted sectors, not their existence. Their non-trivial information is carried by winding along the translated direction rather than by localization at a fixed locus.
Why the twist-dependent ultraviolet part is non-local
Question 18 showed that local ultraviolet terms are associated with group elements that leave some point fixed. The freely acting construction provides a direct geometric realization of that principle.
For 1 ≤ k < N 1 \le k < N 1 ≤ k < N , the shortest displacement between a point and its image under δ N k \delta_N^k δ N k is
ℓ k = 2 π R min ( k , N − k ) > 0. (518) \ell_k=2\pi R\min(k,N-k)>0. \tag*{(518)} ℓ k = 2 π R min ( k , N − k ) > 0. (518)
In a target-space heat-kernel representation, the corresponding image contributes at short proper time through
exp ( − ℓ k 2 4 t ) , (519) \exp\left(-\frac{\ell_k^2}{4t}\right), \tag*{(519)} exp ( − 4 t ℓ k 2 ) , (519)
and is exponentially suppressed as t → 0 t \to0 t → 0 .
The identity sector still contains the ordinary local ultraviolet structure of the parent theory. What is absent is a local divergent coefficient depending on a non-trivial freely acting Scherk–Schwarz image.
After Poisson resummation, the same statement appears as the fact that the twist dependence is carried by non-zero winding sectors. The ultraviolet softness of the Scherk–Schwarz-dependent contribution is a direct consequence of the finite displacement.
Large-radius restoration at fixed holonomy
For a fixed sector q q q , the holonomy ρ q ( 1 ) = e − 2 π i q / N \rho_q(1)=e^{-2\pi i q/N} ρ q ( 1 ) = e − 2 π i q / N does not vary with the radius. The physical gap does:
Δ S S ( q ) = 1 R dist ( q N , Z ) . (520) \Delta_{\mathrm{SS}}^{(q)}=\frac{1}{R}\operatorname{dist}\left(\frac{q}{N},\mathbb{Z}\right). \tag*{(520)} Δ SS ( q ) = R 1 dist ( N q , Z ) . (520)
At fixed q q q and N N N , it vanishes as R → ∞ R \to\infty R → ∞ even though the microscopic holonomy remains non-trivial.
Different parts of the string spectrum behave differently in this limit. The ordinary Kaluza–Klein tower becomes light, m n K K ∼ ∣ n ∣ / R m_n^{\mathrm{KK}}\sim|n|/R m n KK ∼ ∣ n ∣/ R , while the non-trivial fractionally wound sectors of Eq. (517) are generally driven to infinite mass.
There is no requirement that every finite-radius state approach a finite-mass state of the supersymmetric parent. What matters is that, at any fixed physical energy scale, the spectrum and amplitudes of the states that remain accessible approach those of the supersymmetric higher-dimensional background.
If R R R is a dynamical modulus, this large-radius family provides the standard spontaneous Scherk–Schwarz interpretation adopted in this review. If R R R is only an external parameter, the same equations describe a family of backgrounds without by themselves establishing spontaneous breaking.
Scope and limitations
The elementary construction used a geometric circle and a cyclic group, but neither restriction is fundamental.
In an asymmetric orbifold, left- and right-moving worldsheet fields can transform differently. A classical quotient geometry may then be unavailable, while the shift, its twisted sectors and its modular completion remain well defined in the CFT. Whenever such a model has a large-radius field-theory limit, the same equivariance mechanism can still produce a Scherk–Schwarz spectrum.
For an orbifold with several generators, freeness must be checked for the complete group rather than only for a generating set. Two generators can each contain a non-trivial translation while their product cancels the translations and leaves a pure internal action with fixed loci.
Freeness also says nothing by itself about perturbative stability. It removes fixed-point localization associated with the freely acting subgroup, but the complete string mass formula still contains winding, oscillator, zero-point and internal-lattice contributions. Tachyons can therefore appear in other regions of moduli space, particularly when the translated circle becomes small.
These qualifications identify precisely what follows from free action and what must still be determined from the complete worldsheet spectrum.
Early string realizations of Scherk–Schwarz breaking through anti-periodic, coordinate-dependent or freely acting constructions can be found in Refs. [13, 14, 15, 16, 95].
A charge-lattice treatment of coordinate-dependent breaking in heterotic orbifolds, including momentum and winding sectors and the interplay with Wilson-line backgrounds, can be found in Ref. [94].
Freely acting orbifolds with open-string descendants were studied in Ref. [97]. In these constructions the orbifold twists are accompanied by shifts acting on momentum or winding states, and the resulting brane sectors can preserve different fractions of the supersymmetry of the parent theory. They provide explicit open-string realizations of the distinction between the global freely acting quotient and the supersymmetry seen by individual localized sectors.
The flat-holonomy description used above is the quotient-space reformulation of the same construction: an internal phase on the covering space becomes the holonomy of a flat local system on the smooth quotient. This description makes particularly transparent the distinction between the topological triviality of the underlying line bundle and the non-trivial global information carried by its flat connection.
Take-home message
A freely acting Scherk–Schwarz orbifold combines an internal action with a translation. On a covering circle of radius N R NR N R , g S S = g R δ N g_{\mathrm{SS}} = g_R\delta_N g SS = g R δ N , and every non-trivial element of the elementary cyclic Scherk–Schwarz subgroup moves the circle coordinate.
The quotient circle has radius R R R and is smooth, but the internal action survives as a flat holonomy. A component with g R g_R g R -eigenvalue e 2 π i q / N e^{2\pi i q/N} e 2 π i q / N has
p y , m = m − q / N R . p_{y,m} = \frac{m-q/N}{R}. p y , m = R m − q / N .
The same spectrum follows exactly from the covering-space orbifold projection.
For g S S = ( − 1 ) F S δ 2 g_{\mathrm{SS}} = (-1)^{F_S}\delta_2 g SS = ( − 1 ) F S δ 2 , bosons have integer momenta and fermions half-integer momenta, reproducing the half-twisted Scherk–Schwarz spectrum.
Free action does not remove twisted sectors. In the g S S k g_{\mathrm{SS}}^k g SS k -twisted sector the string carries fractional winding on the covering circle; these sectors are not localized at a fixed point and are generally driven to large mass as R → ∞ R \to\infty R → ∞ .
A non-trivial freely acting image has a finite displacement, so its short-proper-time heat kernel is exponentially suppressed. The Scherk–Schwarz-dependent ultraviolet contribution is non-local.
At large radius the holonomy remains fixed while the physical Scherk–Schwarz gap vanishes as 1 / R 1/R 1/ R . The Kaluza–Klein tower descends, whereas non-trivial fractionally wound sectors can leave the finite-energy spectrum.
For asymmetric or multi-generator orbifolds, freeness must be checked for the complete group action. Free action alone does not guarantee supersymmetry breaking or perturbative stability.
This naturally leads to the next question:
How does Scherk–Schwarz breaking act on D-branes and open strings?
How does Scherk–Schwarz breaking act on D-branes and open strings?
Essential idea
The main low-energy mechanisms discussed in this question are already visible in higher-dimensional supergravity with localized sectors. In the present review, the string-theory discussion is therefore used to show how these ingredients can arise from D-branes and orientifolds and which additional microscopic consistency conditions accompany them, rather than to survey the many possible string constructions of supersymmetry breaking.
The supergravity description of Question 9 already provides a useful starting point. For the simple interval problem, localized gravitino masses rotate the allowed supersymmetry subspaces at the boundaries. Writing
Θ b = arctan M b , \Theta_b = \arctan M_b, Θ b = arctan M b ,
the gravitino spectrum depends on the effective phase
ω e f f = ω + Θ 0 + Θ π π ( m o d 1 ) . (521) \omega_{\mathrm{eff}} = \omega+ \frac{\Theta_0+\Theta_\pi}{\pi} \pmod{1}. \tag*{(521)} ω eff = ω + π Θ 0 + Θ π ( mod 1 ) . (521)
The bulk Scherk–Schwarz twist and the localized sectors can contribute to the same spectral mismatch without having the same microscopic origin.
The limit M b → ∞ M_b \to\infty M b → ∞ gives
Θ b → π 2 , Θ b π → 1 2 , \Theta_b \to\frac{\pi}{2}, \qquad\frac{\Theta_b}{\pi} \to\frac{1}{2}, Θ b → 2 π , π Θ b → 2 1 ,
so a maximally rotated local projector produces a half-unit contribution to the spectral shift. This is a useful effective-supergravity description of projector misalignment, not an identification of a D-brane with a localized gravitino mass.
More generally, several localized sectors can preserve different local supersymmetries. The supersymmetry problem is then one of compatibility between bulk transport and the local projectors. String theory realizes such sectors microscopically by D-branes and orientifold planes and adds data that are absent from the supergravity description: brane images and orbits, Chan–Paton actions, RR and NS–NS charges, and the consistency of annulus, Möbius and Klein-bottle amplitudes.
A D-brane can consequently be locally supersymmetric at tree level while the complete compactification admits no exact global supercharge. Conversely, the brane itself can break part of the local parent supersymmetry and realize it non-linearly through a brane Goldstino. These are distinct mechanisms.
Open strings also need not feel the Scherk–Schwarz deformation directly at tree level. The relevant question is whether their boundary conditions allow them to carry the compact quantum number on which the Scherk–Schwarz action operates. When no direct tree-level splitting is present, finite bulk propagation can nevertheless detect the global monodromy and transmit supersymmetry breaking radiatively to the brane.
From localized masses to local projector rotations
Question 9 showed that a localized fermionic bilinear can rotate the subspace of fields allowed at a boundary. The supersymmetry parameter obeys the corresponding matching condition, so the same rotation acts on the locally allowed supersymmetry subspace.
For the flat S U ( 2 ) R SU(2)_{R} S U ( 2 ) R interval example, the signed gravitino eigenvalues can be written
m n e i g = 1 R [ n + ω + Θ 0 + Θ π π ] , Θ b = arctan M b . (522) m_{n}^{\mathrm{eig}}=\frac{1}{R}\left[n+\omega+\frac{\Theta_{0}+\Theta_{\pi}}{\pi}\right],\qquad\Theta_{b}=\arctan M_{b}. \tag*{(522)} m n eig = R 1 [ n + ω + π Θ 0 + Θ π ] , Θ b = arctan M b . (522)
Only the total relative rotation enters this spectrum.
This does not mean that all microscopic realizations of the same ω e f f \omega_{\mathrm{eff}} ω eff are equivalent. A genuine localized supersymmetry-breaking sector can contain fermions that transform inhomogeneously and provide additional Goldstino directions. The gravitino spectrum alone does not reconstruct these local degrees of freedom, their interactions or their ultraviolet counterterms.
The saturated limit
M b ⟶ ∞ , Θ b ⟶ π 2 , Θ b π ⟶ 1 2 (523) M_{b}\longrightarrow\infty,\qquad\Theta_{b}\longrightarrow\frac{\pi}{2},\qquad\frac{\Theta_{b}}{\pi}\longrightarrow\frac{1}{2} \tag*{(523)} M b ⟶ ∞ , Θ b ⟶ 2 π , π Θ b ⟶ 2 1 (523)
is nevertheless instructive. It shows explicitly how a maximal local projector rotation produces the same half-unit spectral displacement that appeared in the pure Scherk–Schwarz benchmark of Question 11.
Localized sectors need not sit only at the endpoints
The interval language can be generalized to localized sectors placed at arbitrary positions in the bulk. Let y n y_{n} y n denote a sequence of localized branes or defects, and let θ n \theta_{n} θ n characterize the local supersymmetry projector selected at the n n n -th position.
Between two neighboring localized sectors, the bulk transports the supersymmetry parameter, while crossing a localized mass produces an additional rotation. In the simple commuting S U ( 2 ) R SU(2)_{R} S U ( 2 ) R example, the condition for a common Killing spinor can be expressed as
θ n + 1 − θ n + arctan M n + arctan M n + 1 = k π , k ∈ Z , (524) \theta_{n+1}-\theta_{n}+\arctan M_{n}+\arctan M_{n+1}=k\pi,\qquad k\in\mathbb{Z}, \tag*{(524)} θ n + 1 − θ n + arctan M n + arctan M n + 1 = k π , k ∈ Z , (524)
up to the sign conventions used for the localized rotations.
The significance of this equation is more general than the particular parametrization. Several localized sectors do not impose independent supersymmetry conditions. Their projectors must be compared after transport through the bulk.
Thus the basic supergravity problem already has the structure that will reappear for D-branes:
bulk transport + local projector data ⟶ global supersymmetry . \text{bulk transport} \quad+ \quad\text{local projector data} \quad\longrightarrow\quad\text{global supersymmetry}. bulk transport + local projector data ⟶ global supersymmetry .
What the microscopic D-brane description adds
In string theory, the localized sectors above can be realized by D-branes and orientifold planes. The effective quantities M b M_b M b and the local projector angles summarize only part of the microscopic information.
A complete open-string construction must also specify how the Scherk–Schwarz quotient acts on the D-branes themselves. A brane can be invariant under the quotient, or it can be mapped to another brane, in which case the complete orbit of images must be included. The orbifold action must also be represented on the Chan–Paton labels of open-string endpoints.
When orientifold planes are present, further data include the orientifold projection and the RR and NS–NS charges carried by the localized objects. The one-loop vacuum amplitude has the schematic form
Z 1-loop = T + K + A + M , (525) \mathcal{Z}_{\text{1-loop}} = \mathcal{T} + \mathcal{K} + \mathcal{A} + \mathcal{M}, \tag*{(525)} Z 1-loop = T + K + A + M , (525)
where the four terms denote the torus, Klein-bottle, annulus and Möbius-strip contributions. Their transverse-channel factorization and the associated tadpole conditions constrain which brane images, Chan–Paton representations and charge assignments are allowed.
These requirements are genuinely microscopic. They restrict which localized supergravity boundary data admit a consistent string completion. They do not, by themselves, define a new low-energy mechanism of Scherk–Schwarz supersymmetry breaking.
What does the open-string spectrum feel at tree level?
Consider a Scherk–Schwarz circle y ∼ y + 2 π R y \sim y + 2\pi R y ∼ y + 2 π R . An open string with Neumann boundary conditions along this direction carries Kaluza–Klein momentum. If the Scherk–Schwarz action couples to a charge q q q , its momentum can take the shifted form
p y = n + ω q R , n ∈ Z . (526) p_y = \frac{n+\omega q}{R}, \qquad n \in\mathbb{Z}. \tag*{(526)} p y = R n + ω q , n ∈ Z . (526)
The resulting boson–fermion splitting is then present already at tree level.
For Dirichlet boundary conditions, the same open string does not carry this Kaluza–Klein momentum tower. A brane localized in the Scherk–Schwarz direction can therefore possess a massless open-string spectrum that is supersymmetric at tree level even when the full compactification breaks supersymmetry.
This should not be turned into a universal “parallel versus transverse” rule. A Dirichlet sector can still carry stretched-string or image-separation data, and after T-duality momentum can become winding or boundary-position information.
The invariant statement is that a direct tree-level Scherk–Schwarz splitting requires the open-string boundary conditions to admit a compact quantum number on which the Scherk–Schwarz action acts.
The brane orbit is part of the open-string data
Let B a B_a B a denote a D-brane boundary condition. The Scherk–Schwarz generator acts not only on closed-string states but also on this boundary condition,
g S S : B a ⟼ B a ′ . g_{\mathrm{SS}}: B_a \longmapsto B_{a'}. g SS : B a ⟼ B a ′ .
If B a B_a B a is invariant, the Scherk–Schwarz group acts directly on the corresponding open-string Hilbert space, including its Chan–Paton indices. If B a B_a B a is moved, the open-string theory must contain the complete orbit
B a , g S S B a , g S S 2 B a , … B_a,\quad g_{\mathrm{SS}}B_a,\quad g_{\mathrm{SS}}^2B_a,\quad\ldots B a , g SS B a , g SS 2 B a , …
Strings can then stretch between different members of this orbit, and their masses depend on the relative position, Wilson lines and the Scherk–Schwarz action.
The same closed-string background can admit inequivalent open-string completions. The closed-string monodromy does not, by itself, determine the brane orbit or the Chan–Paton embedding.
Local supersymmetry is not a global supercharge
The distinction between local and global supersymmetry becomes particularly transparent in the presence of localized branes.
Let S y S_y S y denote the space of supersymmetry parameters preserved locally by the bulk equations near y y y . After choosing a reference trivialization,
S y = U ( y , 0 ) S 0 . (527) S_y = U(y,0)S_0. \tag*{(527)} S y = U ( y , 0 ) S 0 . (527)
A globally defined supercharge must return to itself after transport around the compact direction. Thus
S g l o b a l = Fix ( U S S ) . S_{\mathrm{global}} = \operatorname{Fix}(U_{\mathrm{SS}}). S global = Fix ( U SS ) .
A D-brane localized at y b y_b y b imposes an additional local projector. Let S B S_B S B denote its preserved subspace, and S O S_O S O the subspace selected by an orientifold projection when one is present. The supersymmetry respected by the local tree-level brane action is
S B t r e e , l o c ( y b ) = S y b ∩ S B ∩ S O . (528) S_B^{\mathrm{tree,loc}}(y_b) = S_{y_b} \cap S_B \cap S_O. \tag*{(528)} S B tree , loc ( y b ) = S y b ∩ S B ∩ S O . (528)
This is only a local statement. To determine exact supersymmetry, all subspaces must first be expressed in the same local trivialization and the global monodromy condition must also be imposed.
One can therefore have
S B t r e e , l o c ( y b ) ≠ { 0 } , S B e x a c t = { 0 } . (529) S_B^{\mathrm{tree,loc}}(y_b) \ne\{0\},\qquad S_B^{\mathrm{exact}} = \{0\}. \tag*{(529)} S B tree , loc ( y b ) = { 0 } , S B exact = { 0 } . (529)
The brane then has supersymmetric local couplings at tree level, but there is no globally conserved supercharge protecting its complete quantum effective action.
Several branes require transported intersections
For two localized stacks at y a y_a y a and y b y_b y b , the supersymmetry conditions cannot be intersected before transporting them to the same fibre.
If S a t r e e , l o c S_a^{\mathrm{tree,loc}} S a tree , loc and S b t r e e , l o c S_b^{\mathrm{tree,loc}} S b tree , loc denote the respective local supersymmetry subspaces, then, after transporting the second to y a y_a y a , the supersymmetry compatible with both endpoints is
S a b t r e e = S a t r e e , l o c ∩ U ( y a , y b ) S b t r e e , l o c . (530) S_{ab}^{\mathrm{tree}} = S_a^{\mathrm{tree,loc}} \cap U(y_a,y_b)S_b^{\mathrm{tree,loc}} . \tag*{(530)} S ab tree = S a tree , loc ∩ U ( y a , y b ) S b tree , loc . (530)
When both projectors are already expressed in the same trivialization, this reduces to an ordinary intersection.
It is therefore possible for the a a aa aa and b b bb bb sectors to look supersymmetric separately while the strings stretching between the two stacks do not. Their relative projector orientation, separation, Wilson lines and Scherk–Schwarz transport all enter the same compatibility problem.
Non-linear supersymmetry is a different local statement
A second distinction is needed. A D-brane can itself break part of the supersymmetry available in its local parent background.
The supersymmetries broken specifically by the brane projector are then realized non-linearly on its worldvolume. The corresponding Goldstino has an inhomogeneous transformation of the form
δ λ G = f G ϵ b r o k e n + ⋯ . (531) \delta\lambda_G = f_G\epsilon_{\mathrm{broken}} + \cdots. \tag*{(531)} δ λ G = f G ϵ broken + ⋯ . (531)
This is different from a brane whose local projector preserves the available local supersymmetry while the Scherk–Schwarz monodromy removes all global supercharges.
The two situations can be summarized as follows. If the brane projector itself eliminates a locally available supersymmetry, the brane carries the corresponding non-linear Goldstino realization. If the projector is locally compatible with that supersymmetry but Fix ( U S S ) = { 0 } \operatorname{Fix}(U_{\mathrm{SS}})=\{0\} Fix ( U SS ) = { 0 } , the brane need not carry an independent Goldstino associated with the global Scherk–Schwarz obstruction.
String constructions in which these two mechanisms coexist provide a useful illustration: a local brane sector can realize one supersymmetry non-linearly while an independent Scherk–Schwarz deformation removes the globally defined supersymmetries of the bulk.
Goldstinos and pseudo-Goldstinos
When several localized sectors break supersymmetry independently, each sector can contain a local would-be Goldstino. The complete super-Higgs mechanism is nevertheless controlled by the broken local supersymmetries of the full compactification, not by the separate counting performed one sector at a time.
Only the combinations aligned with the true Goldstino directions are absorbed by the massive gravitini. Orthogonal combinations can remain as physical spin-1 2 \frac{1}{2} 2 1 states, conventionally called pseudo-Goldstinos.
This is another reason why the spectral parameter ω e f f \omega_{\mathrm{eff}} ω eff does not characterize the full localized supersymmetry-breaking system. Two configurations with the same gravitino spectrum can have different Goldstino and pseudo-Goldstino content.
The half-shift and T-duality
The standard half-shift provides a simple check of the preceding statements. Consider
g S S = ( − 1 ) F s δ 1 / 2 , g_{\mathrm{SS}} = (-1)^{F_s}\delta_{1/2}, g SS = ( − 1 ) F s δ 1/2 ,
where F s F_s F s denotes spacetime fermion number.
For a brane wrapping the Scherk–Schwarz circle, open strings carry the shifted momentum directly, so the fermionic tower differs from the bosonic one by half a Kaluza–Klein unit.
After T-duality along the circle, Neumann and Dirichlet boundary conditions are exchanged and momentum data are converted into winding or boundary-position data. The same half-shift can then appear as a half-period displacement or exchange of brane images.
The physical criterion is therefore not whether a brane is called parallel or transverse in one geometric frame. It is whether the boundary conditions and brane orbit allow the open sector to carry the data on which the Scherk–Schwarz action operates.
Why a locally supersymmetric brane can be broken radiatively
A brane localized at y b y_b y b can satisfy
S B t r e e , l o c ( y b ) ≠ { 0 } \mathcal{S}_{B}^{\mathrm{tree,loc}}(y_b) \ne\{0\} S B tree , loc ( y b ) = { 0 }
even when the compactification has no exact global supersymmetry.
An ultraviolet process confined to an arbitrarily small neighbourhood of the brane cannot detect the global Scherk–Schwarz monodromy. Its local divergent counterterms are constrained by the local symmetry structure described in Question 18.
Finite bulk propagation is different. A bulk line can leave the brane, propagate through regions in which the locally preserved supersymmetry has rotated, and return. Such a diagram can probe the full transport around the compact direction and therefore the monodromy U S S U_{\mathrm{SS}} U SS .
When no global supercharge survives, finite non-local diagrams can then generate supersymmetry-breaking operators on a brane that was supersymmetric at tree level:
m s o f t t r e e = 0 , m s o f t l o o p ≠ 0. (532) m_{\mathrm{soft}}^{\mathrm{tree}} = 0,\qquad m_{\mathrm{soft}}^{\mathrm{loop}} \ne0. \tag*{(532)} m soft tree = 0 , m soft loop = 0. (532)
Whether a particular operator is actually generated still depends on the local symmetries and on the existence of the required couplings. The mechanism is the microscopic counterpart of the boundary mediation discussed in Question 19.
Scherk–Schwarz breaking and intrinsic brane breaking
Open and unoriented string constructions can contain supersymmetry breaking that is independent of the Scherk–Schwarz monodromy.
A brane–orientifold configuration can fail to possess a common local supersymmetry projector already before the Scherk–Schwarz identification is imposed. This is an intrinsic local source of supersymmetry breaking. It must be distinguished from a brane that is locally supersymmetric at tree level and receives breaking only through the global Scherk–Schwarz background.
The two mechanisms can coexist. Schematically, a low-energy observable may receive contributions of the form
m s o f t = m S S t r e e + m S S l o o p + m b r a n e l o c a l + ⋯ . (533) m_{\mathrm{soft}} = m_{\mathrm{SS}}^{\mathrm{tree}} + m_{\mathrm{SS}}^{\mathrm{loop}} + m_{\mathrm{brane}}^{\mathrm{local}} + \cdots. \tag*{(533)} m soft = m SS tree + m SS loop + m brane local + ⋯ . (533)
The separation is physical because the terms need not share the same Goldstino structure, ultraviolet behaviour or restoration limit.
A finite-energy restoration test
A splitting generated purely by the Scherk–Schwarz mechanism must vanish in a Scherk–Schwarz restoration limit. The converse is not true: an independent local contribution may happen to vanish along the same trajectory. A splitting that remains finite therefore signals an additional source of supersymmetry breaking.
This statement concerns the finite-energy theory. Open-string sectors that become infinitely heavy along the restoration trajectory need not survive in the limiting spectrum. Tadpole cancellation and perturbative stability are separate consistency conditions of the complete string background.
Explicit open-string realizations of Scherk–Schwarz supersymmetry breaking were studied in Refs. [98, 99]. These constructions exhibit the different response of open-string sectors to a bulk Scherk–Schwarz deformation: sectors carrying the shifted compact momentum can be split already at tree level, whereas localized sectors can remain supersymmetric at tree level and feel the breaking only through the global compactification. Related boundary and orientifold realizations, including supersymmetric open-string sectors embedded in non-supersymmetric Scherk–Schwarz backgrounds, can be found in Refs. [100, 101].
A distinct mechanism, in which supersymmetry is broken intrinsically by the brane–orientifold configuration rather than by the Scherk–Schwarz monodromy, is discussed in Ref. [102]. We use this distinction here to separate local brane supersymmetry breaking, with its Goldstino realization, from global Scherk–Schwarz breaking that can be transmitted to a locally supersymmetric brane.
The purpose of the present discussion is not to review the broader class of non-supersymmetric D-brane and orientifold constructions. Rather, these examples identify which parts of the localized-supergravity description survive in string theory. A consistent open-string completion also requires brane orbits, Chan–Paton actions, orientifold projections and tadpole conditions.
Take-home message
ω e f f = ω + arctan M 0 + arctan M π π ( m o d 1 ) , \omega_{\mathrm{eff}} = \omega+ \frac{\arctan M_0 + \arctan M_\pi}{\pi} \pmod{1}, ω eff = ω + π arctan M 0 + arctan M π ( mod 1 ) ,
while M b → ∞ M_b \to\infty M b → ∞ gives a half-unit contribution to the spectral shift.
D-branes and orientifold planes give microscopic realizations of such localized sectors, but add independent data such as brane orbits, Chan–Paton embeddings, charges and tadpole constraints. These data restrict the supergravity description rather than defining a new low-energy Scherk–Schwarz mechanism.
This naturally leads to the next question:
What is the relation between spin structures, ( − 1 ) F (-1)^F ( − 1 ) F , and R R R -symmetry twists?
What is the relation between spin structures, spacetime fermion parity, and R R R -symmetry twists?
Essential Idea
A minus sign acquired by a fermion around a compact circle can have different origins. It may follow from the choice of spin structure, from a holonomy of spacetime fermion parity ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s , or from an R R R -symmetry transformation that acts as − 1 -1 − 1 on the supercharges.
These possibilities can give the same gravitino boundary condition without defining the same compactification. The comparison must be made on the complete set of physical states.
There is, however, one exact spacetime relation. Spin–statistics implies
( − 1 ) F s = ( − 1 ) 2 J (534) (-1)^{F_s} = (-1)^{2J} \tag*{(534)} ( − 1 ) F s = ( − 1 ) 2 J (534)
on every physical spacetime state. Changing the spin structure by a minus sign can therefore be compensated by inserting ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s .
A central R R R -symmetry element is different. It can act in the same way as fermion parity on the supercharges while acting differently on matter fields. A five-dimensional hypermultiplet will give an explicit example.
The distinction becomes even sharper in string theory. Two operations that are indistinguishable on the light fields of an effective theory can still act differently on massive string states, momentum or winding sectors, twisted sectors or vertex operators.
Three ways of producing the same fermionic sign
Consider a circle y ∼ y + 2 π R y \sim y + 2\pi R y ∼ y + 2 π R . A spacetime spinor may satisfy
ψ ( y + 2 π R ) = − ψ ( y ) . (535) \psi(y + 2\pi R) = -\psi(y). \tag*{(535)} ψ ( y + 2 π R ) = − ψ ( y ) . (535)
The sign can originate directly from the spin structure. Alternatively, one may choose the periodic spin structure and accompany one circuit of the circle by an internal transformation that acts as − 1 -1 − 1 on the spinor. Spacetime fermion parity ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s provides a universal example. A central element of an R R R -symmetry group can also act as − 1 -1 − 1 on the gravitino or supercharges.
Agreement on this one field is not sufficient to identify the three constructions. To decide whether two descriptions define the same background, their relative action must be tested on all physical representations.
Spin structure and spacetime fermion parity
A spin structure specifies the global gluing of local spinors. Once a reference spin structure has been chosen on a spin manifold X X X , the remaining choices form a torsor for H 1 ( X , Z 2 ) H^1(X,\mathbb{Z}_2) H 1 ( X , Z 2 ) . On a circle,
H 1 ( S 1 , Z 2 ) ≃ Z 2 , H^1(S^1,\mathbb{Z}_2) \simeq\mathbb{Z}_2, H 1 ( S 1 , Z 2 ) ≃ Z 2 ,
so there are periodic and anti-periodic choices.
Writing their holonomy as σ s p i n = ± 1 \sigma_{\mathrm{spin}}=\pm1 σ spin = ± 1 , a spinor obeys
ψ ( y + 2 π R ) = σ s p i n ψ ( y ) . (536) \psi(y+2\pi R)=\sigma_{\mathrm{spin}}\psi(y). \tag*{(536)} ψ ( y + 2 π R ) = σ spin ψ ( y ) . (536)
Its Kaluza–Klein momenta are
p y = n + ν s p i n R , ν s p i n = { 0 , σ s p i n = + 1 , 1 2 , σ s p i n = − 1. (537) p_y=\frac{n+\nu_{\mathrm{spin}}}{R},\qquad\nu_{\mathrm{spin}}=\begin{cases}0,&\sigma_{\mathrm{spin}}=+1,\\ \frac{1}{2},&\sigma_{\mathrm{spin}}=-1.\end{cases} \tag*{(537)} p y = R n + ν spin , ν spin = { 0 , 2 1 , σ spin = + 1 , σ spin = − 1. (537)
The spin-structure holonomy should not be confused with the central element − 1 S p i n -1_{\mathrm{Spin}} − 1 Spin of the spin group. On a state of spacetime spin J J J ,
ρ J ( − 1 S p i n ) = ( − 1 ) 2 J . \rho_J(-1_{\mathrm{Spin}})=(-1)^{2J}. ρ J ( − 1 Spin ) = ( − 1 ) 2 J .
Spacetime fermion parity is independently defined as the grading of the physical Hilbert space, + 1 +1 + 1 on bosons and − 1 -1 − 1 on fermions. (534) says that these two central actions coincide on physical spacetime states.
It follows that an anti-periodic spin structure with no internal twist and a periodic spin structure accompanied by a ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s holonomy have the same action on the complete spacetime field theory. Their relative transformation
( − 1 S p i n , ( − 1 ) F s ) \left(-1_{\mathrm{Spin}},(-1)^{F_s}\right) ( − 1 Spin , ( − 1 ) F s )
acts trivially on every physical spacetime state.
Relation to finite temperature
Antiperiodic spacetime fermions also occur in finite-temperature theory. The relation is not accidental: coordinate-dependent compactifications were used in string theory to formulate and analyse thermal backgrounds. The same compactification technology can therefore appear in both problems.
The physical interpretation is nevertheless different. At temperature T = β − 1 T=\beta^{-1} T = β − 1 , Euclidean time is compact with period β \beta β , and the ordinary thermal partition function is
Z ( β ) = Tr e − β H . (538) Z(\beta)=\operatorname{Tr}e^{-\beta H}. \tag*{(538)} Z ( β ) = Tr e − β H . (538)
Its Euclidean path integral has periodic bosons and anti-periodic spacetime fermions. Their half-integer frequencies are Matsubara frequencies.
For a spatial Scherk–Schwarz compactification, the compact coordinate is instead a spatial direction of the vacuum theory. The shifted momentum along it contributes to the physical mass spectrum of the lower-dimensional Lorentzian theory. Thus the same boundary condition
ψ ( y + 2 π R ) = − ψ ( y ) \psi(y + 2\pi R) = -\psi(y) ψ ( y + 2 π R ) = − ψ ( y )
can occur in both settings without making the two physical problems identical.
The operator formulation also fixes the role of fermion parity. Inserting ( − 1 ) F S (-1)^{F_{\mathrm{S}}} ( − 1 ) F S in a thermal trace gives
Tr [ ( − 1 ) F S e − β H ] , (539) \operatorname{Tr}\left[(-1)^{F_{\mathrm{S}}}e^{-\beta H}\right], \tag*{(539)} Tr [ ( − 1 ) F S e − β H ] , (539)
whose path-integral representation has periodic spacetime fermions around the Euclidean time circle. A fermion-parity holonomy on a spatial compactification and a fermion-parity insertion in a thermal trace therefore use closely related boundary-condition technology but compute different physical quantities.
Why a central R R R -twist can be different
A central R R R -symmetry element need not agree with fermion parity on every field. A five-dimensional hypermultiplet gives a simple counterexample.
The bulk R R R -symmetry is S U ( 2 ) R SU(2)_{R} S U ( 2 ) R . The supercharges Q i Q^{i} Q i form a doublet. The hypermultiplet contains a scalar doublet ϕ i \phi^{i} ϕ i and an S U ( 2 ) R SU(2)_{R} S U ( 2 ) R -singlet fermion ψ \psi ψ .
Let
z R = − 1 S U ( 2 ) R . z_{R}=-1_{SU(2)_{R}}. z R = − 1 S U ( 2 ) R .
The central spin rotation acts as − 1 -1 − 1 on Q i Q^{i} Q i and ψ \psi ψ , and as + 1 +1 + 1 on ϕ i \phi^{i} ϕ i . The central S U ( 2 ) R SU(2)_{R} S U ( 2 ) R element acts as − 1 -1 − 1 on Q i Q^{i} Q i and ϕ i \phi^{i} ϕ i , but as + 1 +1 + 1 on ψ \psi ψ . Their product
κ R = ( − 1 s p i n , − 1 S U ( 2 ) R ) (540) \kappa_{R}=\left(-1_{\mathrm{spin}},-1_{SU(2)_{R}}\right) \tag*{(540)} κ R = ( − 1 spin , − 1 S U ( 2 ) R ) (540)
therefore acts as
Q i ↦ Q i , ϕ i ↦ − ϕ i , ψ ↦ − ψ . (541) Q^{i}\mapsto Q^{i},\qquad\phi^{i}\mapsto-\phi^{i},\qquad\psi\mapsto-\psi. \tag*{(541)} Q i ↦ Q i , ϕ i ↦ − ϕ i , ψ ↦ − ψ . (541)
The supercharges cannot distinguish this relative transformation, but the hypermultiplet can.
The same difference is visible in the spectrum. With periodic spin structure and
U R = e 2 π i ω T R , U_{R}=e^{2\pi i\omega T_{R}}, U R = e 2 π iω T R ,
where the scalar doublet has weights ± 1 \pm1 ± 1 ,
m ϕ ± , n = ∣ n ± ω ∣ R , m ψ , n = ∣ n ∣ R . (542) m_{\phi_{\pm},n}=\frac{\lvert n\pm\omega\rvert}{R},\qquad m_{\psi,n}=\frac{\lvert n\rvert}{R}. \tag*{(542)} m ϕ ± , n = R ∣ n ± ω ∣ , m ψ , n = R ∣ n ∣ . (542)
At ω = 1 2 \omega=\frac{1}{2} ω = 2 1 , the scalars are half-shifted while the fermion is not. With anti-periodic spin structure and no R R R -twist, the fermion is half-shifted while the scalars remain periodic.
The two compactifications therefore agree on the supercharges but have different matter spectra.
Two different notions of triviality
The hypermultiplet example shows that there are two distinct questions one can ask about a relative central transformation.
First, does it act trivially on the parent supercharges? Second, does it act trivially on the complete physical theory?
To formulate this precisely, let G ^ R \widehat{G}_R G R denote the fermionic internal symmetry group relevant for the comparison. It contains the R R R -symmetry group G R G_R G R and, when necessary, the central grading ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s . Fermion parity need not belong to the connected R R R -symmetry group.
Consider
G ~ = Spin l o c × ^ G ^ R . (543) \widetilde{G} = \operatorname{Spin}_{\mathrm{loc}} \mathbin{\widehat{\times}} \widehat{G}_R. \tag*{(543)} G = Spin loc × G R . (543)
For central pairs κ = ( z s p i n , z R ) \kappa= (z_{\mathrm{spin}},z_R) κ = ( z spin , z R ) , define
Γ Q = { κ ∈ Z ( G ~ ) ∣ ρ Q ( κ ) = 1 on all parent supercharges } , (544) \Gamma_Q = \left\{ \kappa\in Z(\widetilde{G}) \mathrel{\big|} \rho_Q(\kappa) = 1 \text{ on all parent supercharges} \right\}, \tag*{(544)} Γ Q = { κ ∈ Z ( G ) ρ Q ( κ ) = 1 on all parent supercharges } , (544)
and
Γ t r i v = { κ ∈ Z ( G ~ ) ∣ κ acts trivially on the complete physical theory } . (545) \Gamma_{\mathrm{triv}} = \left\{ \kappa\in Z(\widetilde{G}) \mathrel{\big|} \kappa\text{ acts trivially on the complete physical theory} \right\}. \tag*{(545)} Γ triv = { κ ∈ Z ( G ) κ acts trivially on the complete physical theory } . (545)
Since the supercharges probe only part of the representation content,
Γ t r i v ⊆ Γ Q . (546) \Gamma_{\mathrm{triv}} \subseteq\Gamma_Q. \tag*{(546)} Γ triv ⊆ Γ Q . (546)
For the hypermultiplet,
κ R ∈ Γ Q , κ R ∉ Γ t r i v . \kappa_R \in\Gamma_Q,\qquad\kappa_R \notin\Gamma_{\mathrm{triv}}. κ R ∈ Γ Q , κ R ∈ / Γ triv .
Thus a relative transformation in Γ Q ∖ Γ t r i v \Gamma_Q \setminus\Gamma_{\mathrm{triv}} Γ Q ∖ Γ triv preserves exactly the same supercharges while changing the physical compactification.
The faithfully acting combined symmetry
A central transformation belonging to Γ t r i v \Gamma_{\mathrm{triv}} Γ triv is invisible on the complete theory. Descriptions differing only by such an element represent the same physical holonomy.
The faithfully acting combined group is therefore
G c o m b = Spin l o c × ^ G ^ R Γ t r i v . (547) G_{\mathrm{comb}} = \frac{\operatorname{Spin}_{\mathrm{loc}} \mathbin{\widehat{\times}} \widehat{G}_R}{\Gamma_{\mathrm{triv}}}. \tag*{(547)} G comb = Γ triv Spin loc × G R . (547)
A physical compactification holonomy takes values in G c o m b G_{\mathrm{comb}} G comb . Writing it separately as a spin-structure contribution and an internal contribution amounts to choosing a lift to G ~ \widetilde{G} G .
This explains why the decomposition into “spin structure” and “internal twist” need not itself be physical. Only the resulting element of the faithfully acting group is intrinsic.
A practical test for a half-R R R twist
Let U 1 / 2 ∈ Z ( G R ) U_{1/2} \in Z(G_R) U 1/2 ∈ Z ( G R ) be a central order-two element used as a half-R R R twist. To compare it with a spin-structure change, consider its relative action on a physical representation Φ \Phi Φ :
D 1 / 2 ( Φ ) = ( − 1 ) 2 J Φ ρ R Φ ( U 1 / 2 ) . (548) \mathcal{D}_{1/2}^{(\Phi)} = (-1)^{2J_\Phi}\rho_R^\Phi(U_{1/2}). \tag*{(548)} D 1/2 ( Φ ) = ( − 1 ) 2 J Φ ρ R Φ ( U 1/2 ) . (548)
If D 1 / 2 ( Q ) = 1 \mathcal{D}_{1/2}^{(Q)} = 1 D 1/2 ( Q ) = 1 for every supercharge, the two descriptions preserve the same supersymmetry. They describe the same physical compactification only if D 1 / 2 ( Φ ) = 1 \mathcal{D}_{1/2}^{(\Phi)} = 1 D 1/2 ( Φ ) = 1 for every physical representation.
For integer-normalized R R R -charges,
ρ R Φ ( U 1 / 2 ) = ( − 1 ) q Φ , \rho_R^\Phi(U_{1/2}) = (-1)^{q_\Phi}, ρ R Φ ( U 1/2 ) = ( − 1 ) q Φ ,
so complete equivalence requires
q Φ ≡ F s ( Φ ) ( m o d 2 ) for every physical representation Φ . (549) q_\Phi\equiv F_s(\Phi) \pmod{2}
\qquad\text{for every physical representation } \Phi. \tag*{(549)} q Φ ≡ F s ( Φ ) ( mod 2 ) for every physical representation Φ. (549)
The equality of an R R R -half-twist with fermion parity is a property of the complete representation content of the parent theory, not a consequence of their common action on the gravitino.
A global qualification: lifting the combined holonomy
On a single circle the preceding comparison can be made directly. With several non-contractible loops, one must also ask whether a combined physical holonomy can be decomposed globally into separate spin and internal pieces.
This is a lifting problem. Equation (26.14) defines the central extension
1 ⟶ Γ t r i v ⟶ G ~ ⟶ G c o m b ⟶ 1. (550) 1 \longrightarrow\Gamma_{\mathrm{triv}} \longrightarrow\widetilde{G} \longrightarrow G_{\mathrm{comb}} \longrightarrow1. \tag*{(550)} 1 ⟶ Γ triv ⟶ G ⟶ G comb ⟶ 1. (550)
Let
ϱ : π 1 ( X ) ⟶ G c o m b (551) \varrho: \pi_1(X) \longrightarrow G_{\mathrm{comb}} \tag*{(551)} ϱ : π 1 ( X ) ⟶ G comb (551)
be a flat combined holonomy. Pulling back the extension gives an obstruction
ϱ ∗ [ G ~ ] ∈ H 2 ( π 1 ( X ) , Γ t r i v ) . (552) \varrho^*[\widetilde{G}] \in H^2(\pi_1(X),\Gamma_{\mathrm{triv}}). \tag*{(552)} ϱ ∗ [ G ] ∈ H 2 ( π 1 ( X ) , Γ triv ) . (552)
A global lift to Spin l o c × G ^ R \operatorname{Spin}_{\mathrm{loc}}\times\widehat{G}_R Spin loc × G R exists when this class vanishes.
For X = S 1 X=S^1 X = S 1 , π 1 ( S 1 ) = Z \pi_1(S^1)=\mathbb{Z} π 1 ( S 1 ) = Z has cohomological dimension one, so this obstruction vanishes.
When a lift exists, it need not be unique. Choose a reference spin structure. A split flat description can then be written as
α ∈ H 1 ( X , Z 2 ) , h R : π 1 ( X ) ⟶ G ^ R , (553) \alpha\in H^1(X,\mathbb{Z}_2), \qquad h_R:\pi_1(X)\longrightarrow\widehat{G}_R, \tag*{(553)} α ∈ H 1 ( X , Z 2 ) , h R : π 1 ( X ) ⟶ G R , (553)
where α \alpha α measures the spin structure relative to the reference and h R h_R h R is the internal flat holonomy.
A second lift can differ by a homomorphism
c : π 1 ( X ) ⟶ Γ t r i v . (554) c:\pi_1(X)\longrightarrow\Gamma_{\mathrm{triv}}. \tag*{(554)} c : π 1 ( X ) ⟶ Γ triv . (554)
Writing
c ( γ ) = ( c s p i n ( γ ) , c R ( γ ) ) , c(\gamma) = \left(c_{\mathrm{spin}}(\gamma),c_R(\gamma)\right), c ( γ ) = ( c spin ( γ ) , c R ( γ ) ) ,
the change of lift acts as
α ⟼ α + [ c s p i n ] , (555) \alpha\longmapsto\alpha+ [c_{\mathrm{spin}}], \tag*{(555)} α ⟼ α + [ c spin ] , (555)
h R ( γ ) ⟼ c R ( γ ) h R ( γ ) . (556) h_R(\gamma) \longmapsto c_R(\gamma)\,h_R(\gamma). \tag*{(556)} h R ( γ ) ⟼ c R ( γ ) h R ( γ ) . (556)
Since c R ( γ ) c_R(\gamma) c R ( γ ) is central, the transformed h R h_R h R remains a homomorphism.
Thus the ambiguity of a lift is measured by
Hom ( π 1 ( X ) , Γ t r i v ) . \operatorname{Hom}\left(\pi_1(X),\Gamma_{\mathrm{triv}}\right). Hom ( π 1 ( X ) , Γ triv ) .
This still concerns ordinary holonomies around one-cycles. The next question asks whether the Scherk–Schwarz idea itself has a meaningful analogue for higher-dimensional cycles.
Microscopic string equivalence is stronger
The meaning of “acts trivially” depends on which theory is being tested.
In a low-energy effective theory, one tests only the fields and operators retained in the EFT. Denote the corresponding kernel by Γ t r i v E F T \Gamma_{\mathrm{triv}}^{\mathrm{EFT}} Γ triv EFT .
In a perturbative string background, the same transformation must be trivial on the complete physical string spectrum and operator algebra. Denote this stronger kernel by Γ t r i v C F T \Gamma_{\mathrm{triv}}^{\mathrm{CFT}} Γ triv CFT . For the same candidate central symmetry,
Γ t r i v C F T ⊆ Γ t r i v E F T . (557) \Gamma_{\mathrm{triv}}^{\mathrm{CFT}} \subseteq\Gamma_{\mathrm{triv}}^{\mathrm{EFT}}. \tag*{(557)} Γ triv CFT ⊆ Γ triv EFT . (557)
The inclusion can be strict. A transformation invisible on all light spacetime fields can still act on massive oscillator states, momentum or winding sectors, twisted sectors, vertex operators or their cocycles.
For example, the boundary condition
ψ ( y + 2 π R ) = − ψ ( y ) \psi(y+2\pi R)=-\psi(y) ψ ( y + 2 π R ) = − ψ ( y )
specifies only the monodromy of one spacetime field. A freely acting string construction such as
g S S = ( − 1 ) F s δ 1 / 2 (558) g_{\mathrm{SS}}=(-1)^{F_s}\delta_{1/2} \tag*{(558)} g SS = ( − 1 ) F s δ 1/2 (558)
also specifies the orbifold projection, momentum and winding sectors, twisted sectors and their modular completion.
Operators such as ( − 1 ) F L (-1)^{F_L} ( − 1 ) F L and ( − 1 ) F R (-1)^{F_R} ( − 1 ) F R require the same care. They act separately on left- and right-moving worldsheet sectors and should not be identified automatically with total spacetime fermion parity.
Three levels of comparison
The preceding discussion gives three increasingly strong tests.
Two descriptions may first be compared only on the parent supercharges. Equality at this level determines whether they preserve the same supersymmetry.
One may next compare their action on every physical state retained in the effective field theory. This determines whether they define the same EFT compactification.
Finally, in a string embedding, one must compare their action on the complete microscopic spectrum and operator algebra. Equality at this level is stronger than either of the previous two.
Thus microscopic equivalence implies EFT equivalence, and EFT equivalence implies the same preserved supersymmetry. Neither converse holds in general.
Does this tell us whether supersymmetry is broken spontaneously?
Not by itself. Knowing the monodromy of the supercharges, or even establishing that two descriptions are microscopically equivalent, does not determine how the missing supersymmetries are realized dynamically.
The relevant question is whether a given supercharge remains a symmetry of the physical theory. If it does but the background does not preserve it, the symmetry can be spontaneously broken and must be realized accordingly, through a Goldstino or, in supergravity, the super-Higgs mechanism. A supercharge removed by an orbifold or orientifold projection is instead absent from the symmetry algebra of the projected theory and is not a spontaneously broken generator of that theory.
The D-brane discussion of Question 25 gives a concrete example. An isolated D-brane is a solitonic object of the parent theory: it preserves part of the bulk supersymmetry and spontaneously breaks the complementary part, which is realized non-linearly on its worldvolume. After an orientifold projection, the locally available supersymmetry is instead determined by the common brane and orientifold projectors. A supersymmetry that would have been linearly realized on the isolated brane can therefore be projected out, while another can remain non-linearly realized with its Goldstino. Such local projection and spontaneous breaking can coexist [102].
This local effect is distinct from the global Scherk–Schwarz monodromy. A locally allowed supersymmetry must still return to itself after transport around the compact space. The two mechanisms can remove different parent supercharges in the same compactification.
This distinction is particularly relevant for the freely acting fermion-parity construction of Question 24. A generator
( − 1 ) F s δ \left(-1\right)^{F_s}\delta ( − 1 ) F s δ
can produce mass splittings that vanish when the translation radius becomes large. The existence of this supersymmetric decompactification limit does not by itself establish that every supersymmetry absent at finite radius is spontaneously broken in the microscopic theory.
This situation should be distinguished from the continuous R R R -symmetry realizations discussed in Question 10. There the relevant supersymmetry transformations remain part of the local gauge structure, and the Goldstino and super-Higgs descriptions provide the standard dynamical realization of spontaneous Scherk–Schwarz breaking.
A restoration limit, the existence of a Goldstino for a particular broken generator, and the statement that all missing parent supercharges are spontaneously broken symmetries of one microscopic theory are therefore distinct statements.
Periodic and anti-periodic boundary conditions, together with internal transformations acting on supersymmetry generators, occur already in the original Scherk–Schwarz constructions [6, 7]. Their implementation in string theory and the relation between coordinate-dependent compactifications and fermionic boundary conditions were developed further in early string constructions [13, 16, 95].
Coordinate-dependent compactification techniques were also used in the study of finite-temperature string theory. This historical connection underlies the close formal relation between thermal fermionic boundary conditions and Scherk–Schwarz shifts discussed above, while the thermal and spatial compactifications retain different physical interpretations.
Open-string and D-brane realizations make the distinction between global Scherk–Schwarz breaking and supersymmetry breaking intrinsic to localized brane/orientifold sectors particularly explicit [98, 99, 102]. More recent constructions also illustrate that the fermionic action of the twist can involve a non-trivial cover of the microscopic duality group, and that supersymmetric open-string sectors can coexist with a non-supersymmetric Scherk–Schwarz closed-string background [101]. The detailed classification of these constructions is not needed here; their role is to show how projections, local supersymmetry realizations and global monodromies can coexist in the same microscopic theory.
A complementary M-theory example is provided by opposite chirality conditions at the two E 8 E_8 E 8 boundaries, which combine into an anti-periodic gravitino spin structure on the covering circle [100].
Take-home message
( − 1 ) F s = ( − 1 ) 2 J (-1)^{F_s} = (-1)^{2J} ( − 1 ) F s = ( − 1 ) 2 J
on physical spacetime states. An anti-periodic spin structure and a periodic spin structure accompanied by a ( − 1 ) F s (-1)^{F_s} ( − 1 ) F s holonomy are therefore equivalent at the level of the complete spacetime field theory.
Γ t r i v ⊆ Γ Q . \Gamma_{\mathrm{triv}} \subseteq\Gamma_Q. Γ triv ⊆ Γ Q .
q Φ = F s ( Φ ) ( m o d 2 ) q_{\Phi} = F_s(\Phi) \pmod{2} q Φ = F s ( Φ ) ( mod 2 )
for every physical representation, not only for the gravitino.
G c o m b = S p i n l o c × G ^ R Γ t r i v . G_{\mathrm{comb}} = \frac{\mathrm{Spin}_{\mathrm{loc}} \times\widehat{G}_R}{\Gamma_{\mathrm{triv}}}. G comb = Γ triv Spin loc × G R .
Separate spin and internal holonomies are lifts of this combined datum, and on a general compact space such a lift can have a genuine global obstruction.
Γ t r i v C F T ⊆ Γ t r i v E F T . \Gamma_{\mathrm{triv}}^{\mathrm{CFT}} \subseteq\Gamma_{\mathrm{triv}}^{\mathrm{EFT}}. Γ triv CFT ⊆ Γ triv EFT .
Massive string states, winding sectors and twisted sectors can distinguish backgrounds that are identical in a low-energy truncation.
A supersymmetric restoration limit does not by itself prove spontaneous breaking. A supercharge projected out of the physical theory is not a spontaneously broken generator, whereas a supercharge that remains a symmetry and is broken dynamically has the associated Goldstino or super-Higgs realization.
This naturally leads to the next question:
Can Scherk–Schwarz data be associated with p p p -cycles for p > 1 p > 1 p > 1 ?