WorksPhysics and Astronomy
Lean-Verified No-Go Theorems for -Symmetries and Modular Symmetries in 4D GUTs
Abstract
-symmetries are instrumental in supersymmetric model building. However, a no-go theorem restricts us from using them in a large class of four-dimensional (4D) grand unified theories (GUTs). Specifically, it has been shown that a supersymmetric 4D GUT based on SU(5), with finitely many chiral multiplets, cannot be broken spontaneously to the exact minimal supersymmetric standard model (MSSM) while preserving a discrete , with , or continuous U(1) -symmetry. We use Lean 4 to formalize this proof. In the process, we identify and close a gap in the general case of the published argument. Further, we discuss the implications for GUT models with modular symmetries. We point out that modular symmetries are inevitably -symmetries in a supersymmetric framework that also describes gravity. At certain critical points in moduli space, discrete -symmetries are linearly realized. In the context of 4D GUTs, the no-go theorem then implies that exotic states become massless at these points, provided the GUT-breaking vacuum preserves these symmetries. In the vicinity of these critical points, these exotics are parametrically light, which may spoil gauge coupling unification.
Introduction
Lean has revolutionized mathematics [1, 2, 3, 4]. This is because Lean allows us to formalize proofs, and check them automatically. Modern artificial intelligence (AI) agents are fluent in the Lean 4 syntax, and can even develop proofs. When applied to physics, arguably this development has not yet fully taken off. On the other hand, checking proofs is a very relevant task since theorems in the physics literature have been found to be not entirely correct when their proofs were formalized [5]. These results were obtained using PhysLib [6, 7]. The main purpose of this work is to use Lean 4 to check, and extend, the proof of a no-go theorem [8] in the physics literature.
This theorem concerns -symmetries and grand unified theories (GUTs). -symmetries are instrumental in supersymmetric model building. For instance, they can be argued to be essential to address the problem [9, 10, 11, 12, 13, 14] in the minimal supersymmetric standard model (MSSM). four-dimensional (4D) supersymmetry (SUSY) GUTs (see e.g. [15] for a review) are popular not only because they are motivated by the fact that gauge couplings unify in the MSSM [16] but also because matter furnishes complete SU(5) representations. However, there is an obstruction that limits the use of -symmetries in 4D GUTs. In [8] it was shown that a 4D GUT with a simple gauge group cannot yield the exact MSSM together with an unbroken or symmetry. This means that,if the GUT model has an -symmetry, there are certain exotic states the mass of which is tied to the scale of -symmetry breakdown.
This paper has two purposes. The first is to make the theorem fully rigorous by formalizing it in the Lean 4 proof assistant on top of Mathlib [17]. Even though PhysLib, which was used in [5], has extended group-theory capabilities [18], we work directly with Lean 4 [19]. As we shall see in § 2 and appendix B, this leads to a somewhat shorter proof than the one provided in [8]. It also closes a gap in the general case of the published argument, see § 3.1.
The second purpose is to discuss the implications for modular symmetries, which are -symmetries in the supergravity (SUGRA) framework [20, 21]. At certain critical points in moduli space, discrete -symmetries are linearly realized. The no-go theorem then implies that, in the context of 4D GUTs, at these points exotic states beyond the MSSM become massless.
The outline of the rest of the paper is as follows. In § 2, we state the abstract no-go theorem, and we prove it. In Section 3, we apply the no-go theorem to symmetries in SU(5) GUTs, and we compare the formalized proof with the original one. Section 4 is devoted to the extension of the results to larger gauge groups. In § 5, we explain how higher-dimensional models of grand unification evade the no-go theorem. § 6 discusses modular symmetries, which are -symmetries in SUGRA. There we identify the modular -symmetries that are linearly realized, and we discuss modular zeros, which can forbid mass terms even if no such symmetry is present. Finally, § 7 contains our concluding remarks. Details of how we use Lean are deferred to two appendices. In Appendix § 8, we summarize what has been formalized in Lean, and what has been assumed. Appendix § 9 contains the central parts of the Lean code.
No-Go Theorem
In this section, we develop a more general version of the no-go theorem as given in [8], using some key observations from [22]. This general proof, verified in Lean 4, is then used to show that a 4D GUT with gauge group SU(5) cannot have both an unbroken discrete -symmetry, , with and a massless spectrum consisting only of the MSSM field content plus gauge singlets. We then comment on the extension of the no-go proof to larger GUT gauge groups, and on how higher-dimensional GUTs avoid the no-go theorem.
Setting, Notation, and Hypotheses
Consider a 4D GUT with gauge group , where is a compact connected Lie group. Let be the closed connected subgroup left unbroken by the vacuum. We will denote the maximal torus of as , and we denote the set of roots of , i.e. the non-zero -weights of , as . Note in particular that we do not assume rank , which will be crucial in extending our results beyond SU(5) GUT gauge groups later.
Let be the group of linear symmetries of the microscopic theory that commute with , which includes -symmetries that act on the superspace coordinates . The vacuum in general is not preserved by all of . We define as the subgroup of that leaves the vacuum invariant. The theorem we develop will apply to any unbroken symmetry element , whose action on generically multiplies by a phase . We assume that since the transformation that sends all spinors and to their negatives is a symmetry of any relativistic quantum field theory (QFT), and not a nontrivial -symmetry, see also Footnote 4. We further define as a closed connected subgroup of , with , that commutes with . Since and commute, they will share a simultaneous eigenspace, which will prove useful later.
The centralizer and normalizer of in are given as
respectively. In words, the centralizer of in consists of the elements that commute with all of . The normalizer of in consists of the elements that map back to itself. That is, the normalizer permutes the -weights, while elements of commute with all of and leave every weight fixed. Two normalizer elements that differ by a centralizer element then permute the weights in the same way. Therefore, the part of the Weyl group that preserves the torus of is given by
Here, represents as a coset representative1, is a -weight, and . This reduces to the ordinary Weyl group of when is the maximal torus2 of and . Since is connected,3 we can give a more concrete expression for in terms of a maximal torus of containing as
This just says that in practice we can take the ordinary Weyl group of , keep only the elements that map the Lie algebra of to itself, and then restrict them to .
We now split the massless fermions in the spectrum based on the action of . We take to act linearly and unitarily on all fields. If acts on the scalar of a chiral multiplet by , it acts on the fermion in the same multiplet by . On the gauginos of sitting in vector multiplets, acts by . Define as the -eigenspaces of on fermions. Then, consists of fermions whose scalar partner is -invariant, and consists of those whose scalar partner has eigenvalue , together with the gauginos. For the case that is a pure element with , the fermion has -charge . then consists of the fermions of -charge 0 multiplets, and consists of the fermions of -charge 2 multiplets plus all gauginos.
We now state the following hypotheses we will make use of throughout our argument:
(A1) is compact and connected, with , , and as defined above.
(A2) is a microscopic symmetry, i.e. it acts on the full field content, not just light fields, and is linear, unitary, unbroken, and satisfies .
(A3) commutes with a closed connected subgroup with . The direct product case where commutes with is . Commuting with excludes outer-automorphism actions.
(A4) By definition, the vacuum is invariant under , with and . Therefore, the fermion mass bilinear , a symmetric bilinear form on the fermion space , is - and -invariant.
(A5) There are finitely many fields, so is a finite-dimensional unitary -representation, i.e. chiral fermions in -multiplets plus gauginos in .
Intermediate Results
To proceed to a statement of the no-go theorem and its proof, we first note a few intermediate results.
Here, we make use of our earlier observation that since and commute, they share a simultaneous eigenspace. The fermion space splits into -eigenspaces, which we label as . Since is unitary and commutes with , each is also a -representation. It is useful to also construct the spaces in the following manner. Recall that since , contains the maximal torus of , i.e. . We can then take the spaces , which are necessarily -representations, and split them further into -weight spaces. We denote the space with -weight as . That is, acts on states in by multiplying by , while acts on states by the phase .
Invariance of the vacuum under , Assumption (A4), requires , i.e. for states and . Equality of both sides implies that pairs only with . By Assumption (A2), , so vanishes on , and pairs only with . In particular, gaugino Majorana masses, which would live in , are forbidden. On the mass matrix is given by
where denotes the dual space to . This sector decouples from the other sectors, and its zero modes are with .
Restricted to , and split into -weight spaces as discussed earlier. -invariance of says that unless . Therefore, decomposes in blocks as with
We define , i.e. the number of massless -eigenspace fermions at weight , , i.e. the number of massless -eigenspace fermions at weight , and
For an -symmetry such as or with superpotential -charge 2, counts the massless fermions of weight from -charge 0 multiplets, whereas counts the massless fermions of weight from -charge 2 multiplets together with the massless gauginos of weight . By rank-nullity, with we have
Note that this corresponds to the cancellation of massive pairs in the arguments of [22].
Now let be a finite-dimensional -representation, , and . Then, for ,
where the second equality uses that for , by definition. The third equality rewrites the conjugation of by as the action of an element on the weight . We can then conclude that . Repeating the argument for shows , so that and . Applying this to and using (7), we find
Note that itself is only -equivariant, and the invariance above comes from (7), which involves only the dimensions of -representations.
The No-Go Theorem and Its Proof
Theorem 1 (Abstract no-go theorem). Assume (A1)–(A5) within a 4D GUT. Let and with (so , ). Suppose that for
(a) , i.e. no massless -eigenspace fermion has weight or ,
(b) , i.e. the massless -eigenspace fermions at and are exactly the gauginos there, so no massless chiral -eigenspace fermion, and no massless broken-generator gaugino,
then the index identity (11) is violated, i.e. such a massless spectrum cannot occur. Equivalently, there must be additional massless fermions at weight or .
Proof. Since and root spaces are one-dimensional, , and because , as and . We also have that . Then,
contradicting (11). □
The Lean formalization of this proof is summarized in Appendix § 8, with the relevant code presented in Appendix § 9.
Implications of the Theorem
In the following sections, we will apply this theorem to 4D GUT models. In this process, we will first identify the data , , , , , , and . Then, we verify that the hypotheses (A1)–(A5) are satisfied. Next, we identify the relevant sectors, and determine whether there are and that we can use to derive the contradiction. Such a pair exists if and only if is not stable under . In words, this means there exists some element of the Weyl group that preserves the torus of and maps an unbroken root to a weight that is a broken root. In practice, we can construct using Equation (5) and then test the stability of under . We will see that the theorem of [8] follows. We will also discuss, using key insights from [22], how higher-dimensional models of grand unification evade this theorem.
Application of the No-Go Theorem to Symmetries in SU(5) GUTs
Let us consider a 4D supersymmetric SU(5) GUT. We assume that there are finitely many multiplets, cf. Assumption (A5). Suppose that the vacuum leaves unbroken, and also an element of an unbroken symmetry, where the action of on the superspace coordinates is the phase . In particular, , so .4 The inputs to the abstract no-go theorem are and . is the diagonal torus and is maximal in both and , so , the Weyl group of SU(5). Let us further denote the fundamental weights by . Roots are then given as differences of weights, . are the roots of , which with color indices and weak indices . corresponds to the roots of SU(5) where both are in or . There are 8 unbroken roots, and the other 12 correspond to the roots for the heavy leptoquark gauge bosons of SU(5). By assumption, the is unbroken by the vacuum, so the vacuum is invariant under the action of both and . One can then verify that all of this input data satisfies the assumptions (A1)–(A5). Now we identify a pair of roots, , that will lead to a contradiction. The element of sends the gluon root to the root , with and . Next, we input a light chiral spectrum and check if such a spectrum is inconsistent with the existence of the unbroken in the GUT. Let us assume the light spectrum is the MSSM, possibly with additional gauge singlets. Recall that with denotes the number of massless fermions of weight and -charge . The gauginos transform under in the same way as the superspace coordinates, so they transform with and are not in the -eigenspace. There are no other massless chiral fermions in the spectrum that can sit at a root, so regardless of the -charge assignments to the MSSM fields and singlets. Similarly, for the only massless fermions that appear at a root in the -eigenspace are the gauginos of . At , there is a gluino and . At , there is no massless fermion since the gauge bosons are heavy, so . This then provides the contradiction,
This means that the spectrum cannot be the exact MSSM with additional gauge singlets.
The extension to non-standard embeddings is straightforward. The choice of embedding changes the position of in the Cartan of , which fixes . The no-go theorem can then be applied if is not stable under . This fails, for example, for flipped , where shrinks to , under which is stable.
Comparison with the Original Proof
The general case in [8 Section 2.1.2] proceeds in two steps. First, all pairs of representations that admit an -invariant mass term are removed from the fields of -charge 0 and -charge 2. Such a pair consists of a representation of -charge 0 and its conjugate of -charge 2, for which is an allowed mass term. The removal is decided by group theory alone. It asks whether a mass term is allowed, not whether the fields are actually massive in the vacuum. Second, the highest weights of the remaining representations are compared. After the removal, the two sectors no longer share an irreducible representation, so their highest weights differ. At the larger of the two, say the highest weight of the fields of -charge 0, there are states with no partner of -charge 2 at the opposite weight. These states, and their Weyl images, therefore remain massless. The case analysis of [8] then shows that they cannot all be harmless, and give rise to light exotics. The case is analogous. This comparison uses that the remaining fields of -charge 0 still contain the vacuum, since otherwise the heavy gauge bosons could not acquire masses. However, the first step may remove the multiplet that carries the vacuum. Consider, for instance, fields of -charge 0 and fields of -charge 2 that both transform as . All of these representations are real, so all of them can be removed in pairs. In this case, no highest weights remain, and the argument does not apply. The conclusion of [8] is nevertheless correct. Removing invariant pairs does not change the formal difference of the two sectors. By this we mean, for every weight , the number of states of -charge 0 at minus the number of states of -charge 2 at . A removed pair contributes equally to both numbers. It is therefore legitimate for the comparison of highest weights. It is not legitimate for the vacuum and its Goldstone directions, which depend on the full field content. Where the VEV sits, which directions are eaten, and which states are actually massless depend on the superpotential of the full theory. Once the VEV is switched on, the removed fields mix with the field that carries it, with the Goldstone modes, and with the gauginos.
On the other hand, the proof of Theorem 1 does not remove any pairs. The index (7) is computed for the full field content, and the vacuum enters only through the gaugino count (b). Hence the example above is covered by the result (11). In particular, in this example the two sectors have the same content. Therefore, the chiral multiplets cancel in (7) and only the gaugino remains. The index is therefore at all twenty roots. An exact MSSM would instead require at the unbroken roots and 0 at the broken ones. The special case in which only -plets are present is treated separately in [8 Section 2.1.1], and its proof is correct.
Symmetries with Larger Gauge Groups
Now we turn to the application of the theorem to GUTs with larger gauge groups broken to .
Theorem 2. Let be compact and connected, and let be a subgroup with embedded in the standard way, i.e. is the block subgroup and hypercharge . We assume (A1)–(A5), with . Then, if an unbroken symmetry element with commutes with , the massless spectrum cannot be the MSSM plus gauge singlets.
Proof. Choose . Since is closed, connected, contains , and commutes with by assumption, this satisfies Assumption (A3). is the maximal torus of , and with the standard embedding is also the maximal torus of . As a result, . By Assumption (A5), the fermion space is a -representation, which when restricted to must be an -representation as every -representation branches to an -representation. Furthermore, commutes with , so each is also an -representation. The index relation Equation (11) then holds for , and as before we can take and so that . Now we verify the requirements (a) and (b) of Theorem 1.
We first verify (a), namely for . As in the discussion for the case, the MSSM contains no massless fields sitting at a root. The singlets sit at weight , which is not a root, and the gauginos are in , not . Hence, no massless fermion sits at or , and we conclude .
Now we verify (b), that . The above argument for (a) establishes that there are no massless chiral -eigenspace fermions at and . At , the gluino is massless since is unbroken, and it lies in as a gaugino. At , there is a massive gaugino since every gauge generator of at this weight lies outside of , and is broken. Then the proof proceeds as in the case, where there is a massless gluino at weight , and a massive gaugino outside of at weight . As a result, , and . This provides the contradiction that , so the spectrum cannot be the exact MSSM with additional gauge singlets. □
How Higher-Dimensional Models of Grand Unification Evade the No-Go Theorem
In order to understand how higher-dimensional models evade the no-go theorem, let us first recall, at a qualitative level, why the no-go theorem applies. The basic argument can be illustrated in an model with an arbitrary -symmetry. Assume that a GUT Higgs transforming in the adjoint representation breaks to . Under , the adjoint decomposes as
In order to preserve the -symmetry, has to carry -charge . This implies that any superpotential term built from alone has -charge and is, therefore, forbidden. The components are eaten by the heavy gauge bosons of . On the other hand, the remaining components , and remain massless. These are light, exotic states that go beyond the MSSM.
One may now try to introduce mass partners for the exotics. In [8], it has been shown that these mass partners have to be in the same congruency class as the adjoint. For instance, introducing an additional -plet with -charge does not help. The mass term is allowed, and allows us to pair up the , and components of and . However, the components of have no partners since the corresponding components of are eaten. We are thus, again, left with massless exotics.
More generally, let us consider adjoints with -charge and adjoints with -charge . The absence of massless states requires , while the absence of massless states requires . These conditions are incompatible with each other for any finite numbers and . This counting is the special case of (11) in which only adjoints are present.
The basic argument can be generalized to arbitrary SU(5) representations, to larger GUT groups and to singlet extensions of the MSSM [8]. The general version rests on three assumptions. First, the GUT is four-dimensional and based on a gauge group . Second, the GUT symmetry is broken spontaneously. Third, there are only finitely many fields. Under these assumptions it is impossible to obtain a low-energy effective theory that has both the exact MSSM field content and a residual -symmetry. See § 3.1 for additional details. In § 2, we have turned these arguments into a precise, formalized theorem.
The argument, and the theorem, no longer hold if there are infinitely many 24-plets, or other appropriate representations. As a consequence, higher-dimensional models of grand unification evade the no-go theorem. This is because, from a 4D perspective, they contain an infinite number of multiplets [22], thus violating Assumption (A5). In particular, scenarios in which the GUT symmetry gets broken by compactification evade the no-go theorem. These scenarios include the so-called orbifold GUTs and orbifold compactifications of the heterotic string [25, 26] (see e.g. [27, 28] for more references).
Modular Symmetries
Modular Symmetries Are -Symmetries
We start by pointing out that in a supersymmetric theory that also describes gravity, modular symmetries are -symmetries. Under a modular transformation , the modulus , chiral superfields and superpotential transform as
Here, we introduced the shorthand , and denotes the modulus and the chiral superfields of the theory. and stand for the modular weights of the chiral superfields and the superpotential, respectively. With this sign convention, a coupling requires . Here, denotes the space of holomorphic modular forms of weight (regular at the cusp) for the full modular group, unless specified otherwise [29]. The modular group is generated by
To understand why modular symmetries are necessarily -symmetries, recall that the unique supersymmetric framework that incorporates gravity is SUGRA. As is well known, in SUGRA the so-called -function is given by
A typical Kähler potential is given by
where denotes the number of tori parametrizes. Since transforms nontrivially under (15), the superpotential has to undergo the transformation (17) to ensure modular invariance of the -function (20). Indeed, , hence . Invariance of then requires , up to a possible multiplier. Specifically, the Kähler potential (21) one has
for the modular transformations (13). In particular, in the presence of gravity, modular invariance and SUSY imply that modular symmetries are -symmetries.
In heterotic orbifolds, discrete -symmetries have traditionally been linked to the H-momentum selection rule [30, 31, 32, 33, 34]. However, as is well known, in string compactifications -symmetries really originate from modular symmetries [20, 21]. This is, of course, in agreement with our earlier statement that in a supersymmetric framework that describes gravity, modular symmetries are -symmetries. It is, therefore, natural to ask what the implications of the no-go theorems for -symmetries are for modular symmetries.
R-Symmetries as Stabilizers
There are certain critical values of , , which are invariant under a subgroup of . The modular transformations with form the so-called stabilizer subgroup of . For one then has
where now . Provided that , (18) is an -symmetry. In Table 1 we list the nontrivial stabilizers of . Note that different modular symmetries such as congruence subgroups of can have different stabilizers [35].
| generator | superpotential phase | odd | even | |||
| generic | — | |||||
| i | S | for ; — for | ||||
| for ; for | for ; — for | |||||
| T | 1 | 1 | — |
Table 1. Stabilizers of the vacuum value in and the corresponding superpotential phases. “—” indicates that there is no linearly realized symmetry. The last three columns refer to the full stabilizer, which always contains the center .
The stabilizer symmetries at , and in Table 1 act as symmetries unless takes special values. In particular, in the arguably most interesting case of a single torus, i.e. , there are linearly realized , and symmetries at generic , at and at , respectively. The superpotential has -charge . We introduced the factor 2 here to make sure that the superspace coordinates have integer -charge.6
The Center of SL(2,Z) and a Modular Z4R Symmetry
At a generic point in moduli space, the stabilizer group consists of the center only. The center acts on chiral multiplets and the superpotential as
This center symmetry holds independently of . We call center-even if and center-odd if . Note that the fact that can only take the values follows from Schur’s lemma and the fact that the symmetry is central.
For odd and an appropriate multiplier7, the center symmetry has order 4 since it acts on the superspace coordinates as
The sign ambiguity is the same as the one commented on in Footnote 4. In particular, every modular symmetry with and an appropriate multiplier leads to a symmetry that is linearly realized, independently of . Note that the center symmetry, , has been dubbed modular spin in [37].
Modular Stabilizer Symmetries in 4D GUTs
Modular symmetries provide us with a compelling way to address the flavor puzzle within the framework of modular flavor symmetries [38], see also the reviews [39, 40, 41, 42]. It is, therefore, natural to ask to which extent modular symmetries can be combined with 4D GUTs.
Given that the stabilizer symmetries of § 6.2 are linearly realized -symmetries, our no-go theorem of § 2 applies. This implies, in particular, that at the critical points in moduli space, some exotic states become massless. Away from the fixed point, portions of the stabilizer symmetries are spontaneously broken, and the exotics can pick up masses. However, as stressed in Section 6.3, the center symmetry remains unbroken.
A No-Go Corollary for Modular Symmetries
The abstract no-go theorem, Theorem 1, applies with and , where leaves the GUT-breaking vacuum invariant. Here, denotes the stabilizer group defined above Equation (18). Then, consists of the fermions whose scalar partners are invariant under . Likewise, consists of fermions whose scalar partners pick up the phase , together with the gauginos. Mass terms within and within are forbidden if and only if , where denotes the order of (Modular.stabilizer_phases). In this case, the mass matrix has the block off-diagonal form (6). If generates the full stabilizer, , 4 and 6 at generic , at and at , respectively.
Corollary 1. The light spectrum of a 4D SU(5) GUT with unbroken cannot be the exact MSSM with additional gauge singlets if an element of fixes the GUT-breaking vacuum and multiplies the superpotential by a nontrivial phase. More generally, this holds for any linearly and unitarily realized symmetry , commuting with SU(5), with , and .
Note that Corollary 1 covers , , non-Abelian -symmetries [43], and the discrete -symmetries of heterotic orbifolds [44, 34, 37].
Modular Z4R Symmetry in 4D GUTs
As we have seen in Section 6.3, for odd superpotential weights there is a linearly realized symmetry at a generic point in moduli space. This leads to another corollary.
Corollary 2 (Modular.center_bipartite). Let be odd, so that the superpotential is odd under the center with trivial multiplier. Further, let the GUT-breaking vacuum be center-even. Then the conclusion of Corollary 1 holds for every value of .
Hence, to avoid massless exotics, the GUT-breaking vacuum has to break the modular , for instance through a center-odd GUT Higgs. Indeed, for and a single center-odd 24-plet of weight , the and components acquire masses at generic . This breaking occurs at and may affect the predictions of modular flavor models.
Note that one cannot restore the modular by combining it with an element of the center of . This is because vacua of vanishing 5-ality are invariant under the center of . Hence, if the light spectrum is the exact MSSM, no remnant of the modular survives as an -symmetry at low energies. In particular, the modular cannot be used to forbid the term.
Parametrically Light Exotics near Critical Points
At the critical points and with , the respective superpotential phases are and . Our no-go theorem applies at , provided that the full GUT-breaking vacuum preserves the stabilizer. Near it provides us with additional valuable insights. To see this, define [45, 46]
which measures the departure of from its critical value. Under the stabilizer transformations (18), , i.e. transforms like a field of modular weight . Hence carries -charge, which we list in Table 2.
| odd | even | ||
| under | under | under for ; — for | |
| under | under for ; under for ; under for | under for ; under for ; — for |
Table 2. -charge of the expansion parameter under the linearly realized symmetries of Table 1. The -charges are normalized such that has -charge 1 and has -charge 2. “—” indicates that there is no linearly realized -symmetry.
By holomorphy, only nonnegative powers of appear in the superpotential. Consider now a mass term whose residual charge differs from that of the superpotential by . In order to be allowed, it must be multiplied by , where is the least nonnegative integer with in . Here, denotes the -charge of under the linearly realized as given in Table 2. Since the GUT-breaking vacuum preserves the residual symmetry at , the GUT Higgs has residual charge 0. For the mass term , the charge mismatch is therefore .
As in the Froggatt–Nielsen [47] (FN) picture, acts as a charged FN spurion, so for a GUT Higgs neutral under the residual symmetry, the leading allowed term is then determined by charge conservation. That is, , with the smallest nonnegative integer satisfying this condition. More specifically, we have
For odd and , no exists. A center-even vacuum leaves the exotics massless for every by Corollary 2. A center-odd vacuum, on the other hand, breaks the center. Every nontrivial subgroup of the stabilizer $\langle S\rangle\cong\mathbb{Z}_4 of contains the center. Hence a center-odd vacuum breaks the stabilizer completely. At , however, generates a subgroup that does not contain the center. At level one, a center-odd vacuum preserves this if . For , the multiplies the superpotential by . Then, the leading allowed term has for , and for , so the exotics again have masses . For even , this rule reproduces the values for given above.
It is instructive to consider a simple toy model (Modular.toy_model). Take and a single of weight . The superpotential is then given by
We impose that the -term of the GUT Higgs vanishes, . For , the common factor cancels in the vacuum value , which stays finite as . The masses of and , however, are proportional to , which has a simple zero at . Numerically, for , . In particular, the masses of the exotics vanish exactly at .
What are the phenomenological consequences of these results? In modular flavor models, the proximity of the modulus to a fixed point is often used to generate fermion mass hierarchies [45, 46, 48]. In fact, the expansion parameter of Equation (29) was originally used to parametrize such hierarchies. Our above discussion, therefore, implies that when such scenarios get combined with 4D GUTs, there will be parametrically light adjoint-type exotics. The suppression of their masses goes like , just like the fermion masses. Generically, the presence of such states well below the GUT scale is expected to impact gauge coupling unification, one of the very motivations of GUTs.
Let us also note that in string models winding modes become massless when attains a critical value [49, 50, 51, 52]. However, in contrast to the above-mentioned exotics, these states do not carry standard model (SM) quantum numbers, and therefore do not affect gauge coupling unification.
Modular Zeros beyond Symmetries
There is a corner case in which no linearly realized -symmetry exists, but the no-go theorem still applies. It arises away from the fixed points, for even and a multiplier with , e.g. the trivial one. In this case, the center acts trivially on the superpotential, so it is not an -symmetry. Hence Corollary 1 does not apply. However, the proof of Theorem 1 uses the -symmetry only to obtain the block off-diagonal form (6). As we shall see, so-called modular zeros [37] can enforce this form even without a symmetry.
Modular zeros fall into three classes, see Figure 1. Trivial zeros arise if the required weight has the wrong sign, e.g. . They are similar to the holomorphic zeros of FN-type models [53, 54], which arise if the required charge has the wrong sign. Structural zeros are due to modular spin, i.e. the center. These zeros can explain string selection rules that cannot be understood in terms of ordinary, linearly realized symmetries [37, 55]. Sporadic zeros follow no obvious pattern, and in general they, too, cannot be explained by linear symmetries.

Figure 1. Classification of modular zeros.
In particular, modular zeros can forbid all mass terms among the fermions of the fields that attain VEVs, as well as all mass terms among their potential partners. This is sufficient for (6) to hold.
Theorem 3 (Modular.gap_bipartite). Consider level one with trivial multipliers and . Suppose all chiral multiplets have modular weight or , and only fields of weight attain VEVs. Then, for every , the mass matrix has the block off-diagonal form (6). Here consists of the fermions of weight , and of the fermions of weight and the gauginos. Hence the conclusion of Theorem 1 holds, although the center acts trivially on the superpotential.
Modular zeros can thus enforce (6). In the example of Theorem 3, an ordinary , under which the weight- fields carry the same charge as , does the same. However, the two scenarios differ in their implications for other couplings. For a multiplet of weight , the forbids , whereas modular invariance allows . More generally, no Abelian symmetry under which the vacuum is invariant allows , and but forbids . This means that the modular zero scenario has quite unusual properties when compared to settings with ordinary symmetries.
Summary of the Modular Case
Let us summarize our discussion of modular symmetries in 4D GUTs. The different cases are illustrated in Figure 2.

Figure 2. Consequences of the abstract no-go theorem, Theorem 1, for modular-invariant 4D GUTs. Shaded boxes mark the cases in which the exact MSSM is excluded, and white boxes the cases in which no unbroken symmetry forbids it. denotes the GUT-breaking Higgs and the local coordinate around the fixed point , see § 6.4.3.
The no-go theorem applies to every linearly realized -symmetry that leaves the vacuum invariant and multiplies the superpotential by a nontrivial phase (Corollary 1). For modular symmetries, the stabilizer of falls into this class. At generic , only the center remains. Therefore, the symmetry-based no-go theorem holds for all if and only if is center-odd and the vacuum center-even (Corollary 2).
Near a fixed point the no-go theorem turns into a prediction of light exotics. Their masses are suppressed by the powers , which are determined by the residual charges (§ 6.4.3).
In the absence of linearly realized symmetries, there can still be modular zeros which prevent us from giving exotics masses (Theorem 3).
Concluding Remarks
The combination of Lean and AI has revolutionized mathematics. In this work, we used Lean to formalize no-go theorems in high-energy physics. This has allowed us to double-check the validity of these theorems and extend their scope. In the process, we identified and closed a gap in an earlier proof.
We also extended the analysis to modular symmetries, which we have argued to be inevitably -symmetries in any supersymmetric framework that also describes gravity. The no-go theorems then led to interesting corollaries for modular symmetries. In the context of 4D GUTs, modular symmetries imply the existence of exotic states that become massless at the critical points. This is because modular -symmetries are linearly realized at these points. In the framework of modular flavor symmetries, the modulus is often required to settle close to some critical point in order to accommodate fermion mass hierarchies. The no-go theorems then imply the existence of exotic states that become parametrically light. These states appear in incomplete GUT multiplets, and thus have the potential of ruining gauge coupling unification, one of the main motivations for GUTs.
We have also seen that higher-dimensional models of grand unification evade the no-go theorems. This means that an MSSM model with - and modular symmetries is compatible with unification provided that the GUT symmetry gets broken by compactification. In particular, string theory can give rise to such scenarios.
Acknowledgments
X.-G. L. was supported by the Universidad Nacional Autónoma de México Postdoctoral Program (POSDOC). A.S. was supported by the National Science Foundation Graduate Research Fellowship Program. This material is based upon work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-2235784. We acknowledge help by Anthropic’s Claude to generate Lean code, and to fix typographical and grammatical errors.
What Has Been Formalized
We have formalized the abstract no-go theorem, Theorem 1, and its application to SU(5) in the proof assistant Lean 4, using the mathematics library Mathlib. Lean accepts a proof only if every step can be checked mechanically against the definitions. The formalization therefore tests whether the argument of § 2 really follows from the stated hypotheses. It also makes these hypotheses explicit. The no-go theorem is formalized in the file RNoGoAbstract.lean, which is provided as supplementary material and described in Appendix § 9. The modular selection rules of § 6 are formalized in a separate file, Modular.lean. The command #print axioms lists every axiom a theorem ultimately relies on. For all main results, it reports only the three standard axioms of Lean’s logic, propext, Classical.choice and Quot.sound. Table 3 lists the main statements along with the corresponding Lean theorems.
| Statement | Lean theorem | Status |
| Selection rule behind (6) | selection_rule | proved |
| Counting identity (9) | RFermionSystem.index_eq | proved |
| Weyl invariance (11) | finrank_weightSpace_act, RFermionSystem.index_act | proved |
| Theorem 1 | no_go, no_go_of_Lplus_ne | proved |
| SU(5) contradiction (13) | SU5.no_exact_MSSM | proved |
| Selection rule of Corollary 1 | Modular.stabilizer_phases | proved |
| Selection rule of Corollary 2 | Modular.center_bipartite | proved |
| Selection rule of Theorem 3 | Modular.gap_bipartite | proved |
| Selection rules of the toy model | Modular.toy_model | proved |
| Equation (3), Section 4, light exotics (Section 6.4.3) | — | not formalized |
Table 3. Main statements of the paper and the corresponding Lean theorems in RNoGoAbstract.lean and, for the theorems with the prefix Modular, in Modular.lean.
What Is Proved
Let us go through the argument of § 2. The selection rule is proved. If , then also and . Hence an -invariant mass term can only pair with . The counting identity (9) is proved from rank-nullity, separately for every weight block . The Weyl invariance (11) is proved as in the text. A representative of maps the weight space onto . Hence the two weight spaces have the same dimension. Theorem 1 then follows, because a Weyl-invariant function cannot take the value at and 0 at . Finally, the SU(5) contradiction (13) is proved as a special case. Here the weights are elements of , permutes them, and is the set of the eight roots of . 5-ality shows that no MSSM field sits at a root. A singlet weight is a multiple of and hence never a root. What remains are three statements about the massless spectrum. First, the massless fermions in are MSSM fields or singlets. Second, there is exactly one massless fermion in at the weight , namely a gluino. Third, there is no massless fermion in at the weight . The Lean theorem shows that these three statements are contradictory.
Modular Selection Rules
The file Modular.lean formalizes the selection rules of § 6 at level one with trivial multiplier. A coupling is defined to be allowed if the space of modular forms of the required weight is nonzero. Which of these spaces vanish is taken from Mathlib. On this basis, the file proves the selection rules behind Corollaries 1 and 2 and Theorem 3, as well as those of the toy model of § 6.4.3. These results establish the block off-diagonal form (6). They have not yet been combined with the Lean version of Theorem 1 into a single theorem.
What Is Assumed
Mathlib does not yet contain the structure theory of compact Lie groups, such as maximal tori and the Weyl groups . The corresponding input therefore enters the Lean theorems as explicit hypotheses, never as axioms. These hypotheses are the fields of the structure RFermionSystem and the hypotheses of the theorem no_go. There are three kinds of input.
First, and are finite-dimensional representations of . Further, for every there are invertible maps on and on that intertwine the action of with that of , where represents . It is then proved that these maps send the weight space of onto that of . This replaces Assumptions (A3) and (A5). Indeed, by Assumption (A5) is a finite-dimensional -representation. Since commutes with , and are separately -representations, and representatives of in provide the required maps.
Second, the mass operator is given directly as the family of maps . That is, the block form (6) and the -invariance of Assumption (A4) are built into the setup. They are not derived from a supersymmetric Lagrangean.
Third, the Weyl group acts on the weights by additive maps, as . In Lean this is the hypothesis CharCompat of no_go. It is the only fact from Lie theory that the proof of Theorem 1 uses. No further property of the roots is needed. Indeed, enters only through and . For SU(5) the weights, the action of and the torus are defined explicitly, and CharCompat is proved. Hence the SU(5) result rests only on the first two kinds of input and on the three statements about the massless spectrum.
The conditions (a) and (b) of Theorem 1 are hypotheses of the Lean theorem. In (b), the gaugino count is written as 1 on and 0 elsewhere, which is equivalent because the root spaces of are one-dimensional.
What Is Not Formalized
The expression (5) for and the connectedness of are not formalized, and neither is the derivation of the fermion mass matrix at a supersymmetric vacuum. For a general group the existence of a pair is not formalized. For SU(5) the pair , is checked explicitly. The application to larger gauge groups in Section 4 is proved on paper only, as is the discussion of non-standard embeddings. In the modular case, the zero and the scaling of the masses of the light exotics in § 6.4.3 are not formalized.
Lessons
Two lessons from the formalization deserve mention. First, the order of the -symmetry is used in a single place, through in the selection rule. The argument is therefore the same for every with . It also applies to . Second, the vacuum enters in only three ways. It enters through its symmetries, through the requirement that the light chiral spectrum be the MSSM plus singlets, and through the gaugino count (b). The gaugino count means one massless gluino at the weight of an unbroken root, and no massless fermion at the weight of its broken Weyl image. Neither - and -flatness nor the counting of Goldstone modes is needed.
Lean Formalization of the Abstract No-Go Theorem
In this appendix we show the central parts of the Lean file, so that the reader can see how the argument of § 2 reads in Lean. The complete file is provided as supplementary material, RNoGoAbstract.lean. It was checked with Lean 4.33.0 and Mathlib commit db584cd6d46c92f209a44c0f1c829460d327499d. Each box below reproduces a range of lines of this file without any change, and the title of the box gives the range.
Reading the Lean Code. A few words on the syntax are in order. A theorem lists its hypotheses as named arguments in parentheses, then a colon and the conclusion, and then the proof after :=. Most of our theorems have the conclusion False. That is, they state that their hypotheses cannot all hold, which is how a no-go theorem looks in Lean. A structure collects data together with the conditions they satisfy, and an argument such as (S : RFermionSystem act chr wt) reads “let be a fermion system.” Text between /- and -/ is documentation and is not checked by Lean.
We follow the notation of § 2 as far as possible. is the set of -weights, is the group Weyl, and act w χ stands for . The fermion spaces and are called Fm and Fp. In the comments of the file they are labeled by the fermion -charges and that they have for a symmetry. The letter K denotes the field of numbers, for instance , and not a group. One convention needs care. Lminus χ is , the number of massless fermions in at weight . Lplus χ, on the other hand, is the dimension of the cokernel of , i.e. , the number of massless fermions in at weight .
The Fermion System. Recall that the argument of § 2 uses three ingredients, the two fermion sectors with the action of , the action of the Weyl group, and the mass blocks . The structure RFermionSystem collects exactly these. Its fields are the finite-dimensional spaces Fm and Fp, cf. Assumption (A5), the torus actions and , the invertible maps nm w and np w by which a representative of acts, together with the condition that they intertwine the torus action, and the mass blocks m. The maps nm and np encode that the fermions come in -multiplets and that commutes with , Assumption (A3). Note that is a map from to the dual of by construction. That is, the block form (6) and the -invariance of Assumption (A4) are built into the definition. The algebraic step behind (6), that implies , is proved separately as selection_rule.
/-- Fermion system (𝓕₋, 𝓕₊) with ‘T₀‘-action and Weyl intertwiners. The type of ‘M‘ is modelling
input: bipartite from ‘ζ ^2 ≠ 1‘ (‘selection_rule‘), weight-by-weight from ‘T₀‘-invariance. -/
structure RFermionSystem {K : Type u} [Field K] {X : Type uw} [AddCommGroup X]
{W : Type x} [Group W] (act : W →* AddMonoid.End X) {T : Type y} [Group T]
(chr : X → T → K) (wt : W →* MulAut T) where
/-- 𝓕₋: fermions with r-phase ζ⁻¹ (f = -1 for Z_M^R). -/
Fm : Type v
/-- 𝓕₊: fermions with r-phase ζ, incl. all gauginos (f = +1 for Z_M^R). -/
Fp : Type v
[addCommGroupFm : AddCommGroup Fm]
[moduleFm : Module K Fm]
[finiteFm : FiniteDimensional K Fm]
[addCommGroupFp : AddCommGroup Fp]
[moduleFp : Module K Fp]
[finiteFp : FiniteDimensional K Fp]
/-- Torus action on ‘Fm‘. -/
ρm : T →* (Fm →ₗ[K] Fm)
/-- Torus action on ‘Fp‘. -/
ρp : T →* (Fp →ₗ[K] Fp)
/-- Weyl intertwiner on ‘Fm‘: representative of ‘w‘ in ‘N_B(T₀)‘. -/
nm : W → (Fm ≃ₗ[K] Fm)
/-- ‘nm w‘ intertwines the torus action on ‘Fm‘. -/
hnm : ∀ w, IsIntertwiner wt ρm w (nm w)
/-- Weyl intertwiner on ‘Fp‘: representative of ‘w‘ in ‘N_B(T₀)‘. -/
np : W → (Fp ≃ₗ[K] Fp)
/-- ‘np w‘ intertwines the torus action on ‘Fp‘. -/
hnp : ∀ w, IsIntertwiner wt ρp w (np w)
/-- Mass block at ‘T₀‘-weight ‘χ‘: ‘Fm(χ) → (Fp(-χ))^∨‘. -/
M : ∀ χ : X, weightSpace chr ρm χ →ₗ[K] Module.Dual K (weightSpace chr ρp (-χ))
The Index and Its Weyl Invariance. Next, the file defines Lminus, Lplus and the index ind 𝔐χ = . The theorem index_eq is the counting identity (9). Its proof applies rank-nullity to 𝔐χ, uses that a space and its dual have the same dimension, and finishes with omega, a tactic that solves linear equations between integers. The theorem index_act is the Weyl invariance (11). It rewrites both sides with index_eq and then uses that weight multiplicities are Weyl invariant, finrank_weightSpace_act (lines 115–121). The latter is proved exactly as in Section 2. A representative of maps onto . The hypothesis hc : CharCompat act chr wt is the formula that relates the action of on weights to its action on the torus.
/-- ‘Lminus χ = dim ker (M χ)‘: number of massless ‘Fm‘ fermions at weight ‘χ‘. -/
noncomputable def Lminus (χ : X) : N := Module.finrank K (LinearMap.ker (S.M χ))
/-- ‘Lplus χ = dim coker (M χ)‘ = L₊(-χ): number of massless ‘Fp‘ fermions at weight ‘-χ‘. -/
noncomputable def Lplus (χ : X) : N :=
Module.finrank K (weightSpace chr S.ρp (-χ) / LinearMap.range (S.M χ))
/-- Index at ‘χ‘: massless ‘Fm‘ fermions at ‘χ‘ minus massless ‘Fp‘ fermions at ‘-χ‘. -/
noncomputable def index (χ : X) : Z := (S.Lminus χ : Z) - (S.Lplus χ : Z)
/-- Rank-nullity: ‘index χ = dim Fm(χ) - dim Fp(-χ)‘, independent of the mass matrix. -/
theorem index_eq (χ : X) :
S.index χ = (Module.finrank K (weightSpace chr S.ρm χ) : Z)
- Module.finrank K (weightSpace chr S.ρp (-χ)) := by
have h1 := LinearMap.finrank_range_add_finrank_ker (S.M χ) -- range + ker = dim Fm(χ)
have h2 := Submodule.finrank_quotient_add_finrank (LinearMap.range (S.M χ))
-- coker + range = dim of the dual
have h3 := Subspace.dual_finrank_eq (K := K) (V := weightSpace chr S.ρp (-χ))
unfold index Lminus Lplus
omega
/-- Weyl invariance of the index: ‘index (w·χ) = index χ‘. -/
theorem index_act (hc : CharCompat act chr wt) {w : W} {χ : X} :
S.index (act w χ) = S.index χ := by
rw [S.index_eq, S.index_eq, finrank_weightSpace_act hc (S.hnm w) χ, ← act_neg,
finrank_weightSpace_act hc (S.hnp w) (-𝜒)]
The No-Go Theorem. The proof of Theorem 1 has two steps. The first is the elementary statement orbit_contradiction (lines 23–27). It states that a function that is invariant under a group action cannot take different values at and . The theorem no_go_of_Lplus_ne applies it to the index. If vanishes at and at , the index equals Lplus at both weights, so Lplus must take the same value there.
/-- No massless ‘Fm‘ at ‘β‘, ‘w·β‘ forces ‘Lplus β = Lplus (w·β)‘ (index invariance). -/
theorem no_go_of_Lplus_ne (S : RFermionSystem.{u, uw, x, y, v} act chr wt)
(hc : CharCompat act chr wt) (β : X) (w : W)
(ha : S.Lminus β = 0 ∧ S.Lminus (act w β) = 0)
(hne : S.Lplus β ≠ S.Lplus (act w β)) : False := by
refine orbit_contradiction (fun g x => act g x) S.index
(fun g x => S.index_act hc (w := g) (χ := x)) w β ?_
intro h
apply hne
simp only [RFermionSystem.index, ha.1, ha.2] at h
omega
The second step is the theorem no_go, which is Theorem 1 itself. Read from top to bottom, it says the following. Let be a fermion system whose Weyl action satisfies CharCompat, let be a set of weights, let , and let be an element of the Weyl group with . Assume condition (a), i.e. Lminus vanishes at and at , and condition (b) in the form Lplus for and 0 otherwise, for . Then we arrive at a contradiction. Since and root spaces are one-dimensional, this form of (b) is condition (b) of Theorem 1. The input from Lie theory enters only through CharCompat and the intertwiners of the fermion system, and no property of is needed beyond and . The proof evaluates (b) at the two weights, which gives Lplus and Lplus , and applies no_go_of_Lplus_ne. The line open Classical in just above the box (line 227) allows the case distinction in (b) for an arbitrary set .
/-- **The no-go theorem.** An unbroken root ‘β ∈ Δ0‘ with Weyl image ‘w·β ∉ Δ0‘, no massless ‘Fm‘
at ‘β‘, ‘w·β‘, and ‘Lplus β = 1‘, ‘Lplus (w·β) = 0‘ (‘G0‘ gauginos) contradict invariance. -/
theorem no_go (S : RFermionSystem.{u, uw, x, y, v} act chr wt) (hc : CharCompat act chr wt)
(Δ0 : Set X) (β : X) (hβ : β ∈ Δ0) (w : W) (hw : act w β ∉ Δ0)
(ha : S.Lminus β = 0 ∧ S.Lminus (act w β) = 0)
(hb : ∀ χ ∈ ({β, act w β} : Set X), S.Lplus χ = if χ ∈ Δ0 then 1 else 0) : False := by
-- evaluate (b) at ‘β‘ (uses ‘hβ‘) and at ‘w·β‘ (uses ‘hw‘)
have h1 : S.Lplus β = 1 := by
rw [hb β (Set.mem_insert _ _)] ; simp [hβ]
have h2 : S.Lplus (act w β) = 0 := by
rw [hb (act w β) (Set.mem_insert_of_mem _ _)] ; simp [hw]
exact no_go_of_Lplus_ne S hc β w ha (by rw [h1, h2]; decide)
The SU(5) Case. Finally, let us see how the SU(5) result follows. The weights are the elements of , i.e. the characters of the diagonal torus of U(5). Weights that differ by a multiple of agree on the torus of SU(5), and these multiples are the singlet weights. W is acting by permutations, and is the set of the eight roots of G. The theorem SU5.no_exact_MSSM has three hypotheses. minus_content says that every massless fermion in has the weight of an MSSM field or of a singlet. plus_unbroken says that there is exactly one massless fermion in , the gluino, at a weight . plus_broken says that there is none at weight . The proof first shows that vanishes at both roots. No MSSM weight is a root because of 5-ality, and no singlet weight is a root because a multiple of never is. It then applies no_go with and . This is the contradiction (13). Here the action of on the weights and on the torus is defined explicitly and CharCompat is proved as SU5.su5_CharCompat, so no hypothesis about Lie theory remains.
/-- **No exact MSSM** in SU(5): MSSM-or-singlet massless ‘Fm‘ fermions, only the ‘SU(3)‘ gaugino
at ‘-(e0-e1)‘ and no massless ‘Fp‘ fermion at ‘-(e0-e3)‘ are contradictory. -/
theorem no_exact_MSSM {K : Type*} [Field K] (S : RFermionSystem (su5Act) (charOf (K := K)) su5Wt)
(minus_content : ∀ χ, 0 < S.Lminus χ → IsMSSMWeight χ ∨ IsSingletWeight χ)
(plus_unbroken : S.Lplus (root 0 1) = 1)
(plus_broken : S.Lplus (root 0 3) = 0) : False := by
-- (a) at a root ‘root i j‘: no light field there (5-ality), no singlet there
have ha : ∀ i j : Fin 5, i ≠ j → S.Lminus (root i j) = 0 := fun i j hij =>
Lminus_eq_zero_of_z’ (Light := {χ | IsMSSMWeight χ}) (Sing := {χ | IsSingletWeight χ})
(z := fivealityHom) S minus_content fivealityHom_mssm
(fivealityHom_rootSet _ ⟨i, j, hij, rfl⟩) (root_not_singlet hij)
-- the abstract no-go with ‘Δ0 = unbrokenSet‘, ‘β = e0-e1‘, ‘w = (1 3)‘
refine _root_.RNoGoAbstract.no_go S su5_charCompat unbrokenSet (root 0 1) root01_mem
(Equiv.swap 1 3)
(by rw [su5Act_apply, weylAct_swap_root]; exact root03_not_mem)
⟨ha 0 1 (by decide), by rw [su5Act_apply, weylAct_swap_root]; exact ha 0 3 (by decide)⟩ ?_
intro χ hχ
simp only [Set.mem_insert_iff, Set.mem_singleton_iff, su5Act_apply, weylAct_swap_root] at hχ
rcases hχ with rfl | rfl
· simp [plus_unbroken, root01_mem]
· simp [plus_broken, root03_not_mem]
References
References
- [1]D. Castelvecchi, “Mathematicians welcome computer-assisted proof in ‘grand unification’ theory,” Nature 595 no. 7865, (2021) 18–19.DOI
- [2]J. Commelin and A. Topaz, “Abstraction boundaries and spec driven development in pure mathematics,” Bull. Amer. Math. Soc. (N.S.) 61 no. 2, (2024) 241–255.arxiv.org/abs/2309.14870
- [3]K. Buzzard, “Mathematical reasoning and the computer,” Bull. Amer. Math. Soc. (N.S.) 61 no. 2, (2024) 211–224.arxiv.org/abs/2502.07850
- [4]J. Avigad, “Mathematics and the formal turn,” Bull. Amer. Math. Soc. (N.S.) 61 no. 2, (2024) 225–240.arxiv.org/abs/2311.00007
- [5]J. Tooby-Smith, “Formalizing the stability of the two Higgs doublet model potential into Lean: identifying an error in the literature,” arXiv:2603.08139 [hep-ph].
- [6]J. Tooby-Smith, “HepLean: Digitalising high energy physics,” Comput. Phys. Commun. 308 (2025) 109457, arXiv:2405.08863 [hep-ph].
- [7]T. P. community, “Physlib: The Lean Physics Library.” https://github.com/leanprover-community/physlib, 2024.
- [8]M. Fallbacher, M. Ratz, and P. K. S. Vaudrevange, “No-go theorems for R symmetries in four-dimensional GUTs,” Phys. Lett. B 705 (2011) 503–506, arXiv:1109.4797 [hep-ph].DOI
- [9]L. J. Hall, Y. Nomura, and A. Pierce, “R symmetry and the mu problem,” Phys. Lett. B 538 (2002) 359–365, arXiv:hep-ph/0204062.DOI
- [10]K. S. Babu, I. Gogoladze, and K. Wang, “Natural R parity, μ-term, and fermion mass hierarchy from discrete gauge symmetries,” Nucl. Phys. B 660 (2003) 322–342, arXiv:hep-ph/0212245.DOI
- [11]H. M. Lee, S. Raby, M. Ratz, G. G. Ross, R. Schieren, K. Schmidt-Hoberg, and P. K. Vaudrevange, “A unique Z₄ᴿ symmetry for the MSSM,” Phys. Lett. B694 (2011) 491–495, arXiv:1009.0905 [hep-ph].
- [12]H. M. Lee, S. Raby, M. Ratz, G. G. Ross, R. Schieren, K. Schmidt-Hoberg, and P. K. Vaudrevange, “Discrete R symmetries for the MSSM and its singlet extensions,” Nucl. Phys. B850 (2011) 1–30, arXiv:1102.3595 [hep-ph].DOI
- [13]H. Baer, V. Barger, and D. Sengupta, “Gravity safe, electroweak natural axionic solution to strong CP and SUSY μ problems,” Phys. Lett. B 790 (2019) 58–63, arXiv:1810.03713 [hep-ph].DOI
- [14]P. N. Bhattiprolu and S. P. Martin, “High-quality axions in solutions to the μ problem,” Phys. Rev. D 104 no. 5, (2021) 055014, arXiv:2106.14964 [hep-ph].DOI
- [15]S. Raby, Supersymmetric Grand Unified Theories: From Quarks to Strings via SUSY GUTs, vol. 939. Springer, 2017.DOI
- [16]S. Dimopoulos, S. Raby, and F. Wilczek, “Supersymmetry and the Scale of Unification,” Phys. Rev. D 24 (1981) 1681–1683.DOI
- [17]The mathlib Community, “The Lean mathematical library,” in Proceedings of the 9th ACM SIGPLAN International Conference on Certified Programs and Proofs (CPP 2020), pp. 367–381. ACM, New York, 2020. arXiv:1910.09336 [cs.LO].DOI
- [18]S. Krippendorf and J. Tooby-Smith, “Physics as Code: From Scans to Theorems with ITP APIs in SU(5) Model Building,” arXiv:2603.28406 [hep-th].
- [19]L. de Moura and S. Ullrich, “The Lean 4 Theorem Prover and Programming Language,” in Automated Deduction – CADE 28, A. Platzer and G. Sutcliffe, eds., vol. 12699 of Lecture Notes in Computer Science, pp. 625–635. Springer, Cham, 2021.DOI
- [20]L. J. Dixon, V. Kaplunovsky, and J. Louis, “On Effective Field Theories Describing (2,2) Vacua of the Heterotic String,” Nucl. Phys. B 329 (1990) 27–82.DOI
- [21]L. E. Ibáñez and D. Lüst, “A Comment on duality transformations and (discrete) gauge symmetries in four-dimensional strings,” Phys. Lett. B 302 (1993) 38–46, arXiv:hep-th/9212089.DOI
- [22]M. W. Goodman and E. Witten, “Global symmetries in four and higher dimensions,” Nucl. Phys. B 271 (1986) 21–52.DOI
- [23]P. Fayet and J. Iliopoulos, “Spontaneously Broken Supergauge Symmetries and Goldstone Spinors,” Phys. Lett. B 51 (1974) 461–464.DOI
- [24]S. Dimopoulos and H. Georgi, “Softly Broken Supersymmetry and SU(5),” Nucl. Phys. B 193 (1981) 150–162.DOI
- [25]L. J. Dixon, J. A. Harvey, C. Vafa, and E. Witten, “Strings on Orbifolds,” Nucl. Phys. B 261 (1985) 678–686.
- [26]L. J. Dixon, J. A. Harvey, C. Vafa, and E. Witten, “Strings on Orbifolds. 2.,” Nucl. Phys. B 274 (1986) 285–314.
- [27]M. Quiros, “New ideas in symmetry breaking,” in Theoretical Advanced Study Institute in Elementary Particle Physics (TASI 2002): Particle Physics and Cosmology: The Quest for Physics Beyond the Standard Model(s), pp. 549–601. 2, 2003. arXiv:hep-ph/0302189.
- [28]S. Ramos-Sanchez and M. Ratz, Heterotic Orbifold Models. Springer, 2024. arXiv:2401.03125 [hep-th].
- [29]X.-G. Liu and G.-J. Ding, “Modular flavor symmetry and vector-valued modular forms,” JHEP 03 (2022) 123, arXiv:2112.14761 [hep-ph].DOI
- [30]S. Hamidi and C. Vafa, “Interactions on Orbifolds,” Nucl. Phys. B 279 (1987) 465–513.DOI
- [31]L. J. Dixon, D. Friedan, E. J. Martinec, and S. H. Shenker, “The Conformal Field Theory of Orbifolds,” Nucl. Phys. B 282 (1987) 13–73.
- [32]T. Kobayashi, S. Raby, and R.-J. Zhang, “Searching for realistic 4d string models with a Pati-Salam symmetry: Orbifold grand unified theories from heterotic string compactification on a Z(6) orbifold,” Nucl. Phys. B 704 (2005) 3–55, arXiv:hep-ph/0409098.DOI
- [33]N. G. Cabo Bizet, T. Kobayashi, D. K. Mayorga Pena, S. L. Parameswaran, M. Schmitz, and I. Zavala, “Discrete R-symmetries and Anomaly Universality in Heterotic Orbifolds,” JHEP 02 (2014) 098, arXiv:1308.5669 [hep-th].arxiv.org/abs/1308.5669
- [34]H. P. Nilles, S. Ramos-Sánchez, M. Ratz, and P. K. S. Vaudrevange, “A note on discrete R symmetries in Z₆-II orbifolds with Wilson lines,” Phys. Lett. B 726 (2013) 876–881, arXiv:1308.3435 [hep-th].DOI
- [35]F. Diamond, J. M. Shurman, and J. Shurman, A first course in modular forms, vol. 228. Springer, 2005.
- [36]L. E. Ibáñez and D. Lü̈st, “Duality anomaly cancellation, minimal string unification and the effective low-energy Lagrangian of 4-D strings,” Nucl. Phys. B 382 (1992) 305–361, arXiv:hep-th/9202046.DOI
- [37]X.-G. Liu and M. Ratz, “Modular Zeros,” Phys. Lett. B 875 (2026) 140372, arXiv:2512.21386 [hep-th].arxiv.org/abs/2512.21386
- [38]F. Feruglio, Are neutrino masses modular forms?, pp. 227–266. 2019. arXiv:1706.08749 [hep-ph].arxiv.org/abs/1706.08749
- [39]F. Feruglio and A. Romanino, “Lepton flavor symmetries,” Rev. Mod. Phys. 93 no. 1, (2021) 015007, arXiv:1912.06028 [hep-ph].
- [40]Y. Almumin, M.-C. Chen, M. Cheng, V. Knapp-Perez, Y. Li, A. Mondol, S. Ramos-Sanchez, M. Ratz, and S. Shukla, “Neutrino Flavor Model Building and the Origins of Flavor and CP Violation,” Universe 9 no. 12, (2023) 512, arXiv:2204.08668 [hep-ph].arxiv.org/abs/2204.08668
- [41]G.-J. Ding and S. F. King, “Neutrino mass and mixing with modular symmetry,” Rept. Prog. Phys. 87 no. 8, (2024) 084201, arXiv:2311.09282 [hep-ph].arxiv.org/abs/2311.09282
- [42]T. Kobayashi and M. Tanimoto, “Modular flavor symmetric models,” Int. J. Mod. Phys. A 39 no. 09n10, (2024) 2441012, arXiv:2307.03384 [hep-ph].arxiv.org/abs/2307.03384
- [43]M.-C. Chen, M. Ratz, and A. Trautner, “Non-Abelian discrete R symmetries,” JHEP 09 (2013) 096, arXiv:1306.5112 [hep-ph].DOI
- [44]R. Kappl, B. Petersen, S. Raby, M. Ratz, R. Schieren, and P. K. S. Vaudrevange, “String-Derived MSSM Vacua with Residual R Symmetries,” Nucl. Phys. B 847 (2011) 325–349, arXiv:1012.4574 [hep-th].DOI
- [45]F. Feruglio, V. Gherardi, A. Romanino, and A. Titov, “Modular invariant dynamics and fermion mass hierarchies around τ = i,” JHEP 05 (2021) 242, arXiv:2101.08718 [hep-ph].DOI
- [46]P. P. Novichkov, J. T. Penedo, and S. T. Petcov, “Fermion mass hierarchies, large lepton mixing and residual modular symmetries,” JHEP 04 (2021) 206, arXiv:2102.07488 [hep-ph].DOI
- [47]C. D. Froggatt and H. B. Nielsen, “Hierarchy of Quark Masses, Cabibbo Angles and CP Violation,” Nucl. Phys. B 147 (1979) 277–298.DOI
- [48]M.-C. Chen, X. Li, X.-G. Liu, and M. Ratz, “Modular flavor symmetries and fermion mass hierarchies,” JHEP 10 (2025) 033, arXiv:2506.23343 [hep-ph].arxiv.org/abs/2506.23343
- [49]L. E. Ibáñez, W. Lerche, D. Lü̈st, and S. Theisen, “Some Considerations About the Stringy Higgs Effect,” Nucl. Phys. B 352 (1991) 435–450.
- [50]R. Dijkgraaf, C. Vafa, E. P. Verlinde, and H. L. Verlinde, “The Operator Algebra of Orbifold Models,” Commun. Math. Phys. 123 (1989) 485.DOI
- [51]F. Beye, T. Kobayashi, and S. Kuwakino, “Gauge Origin of Discrete Flavor Symmetries in Heterotic Orbifolds,” Phys. Lett. B 736 (2014) 433–437, arXiv:1406.4660 [hep-th].DOI
- [52]X. Li, X.-G. Liu, H. P. Nilles, M. Ratz, and A. Stewart, “Flavor symmetries and winding modes,” JHEP 09 (2025) 026, arXiv:2506.12887 [hep-th].arxiv.org/abs/2506.12887
- [53]M. Leurer, Y. Nir, and N. Seiberg, “Mass matrix models,” Nucl. Phys. B 398 (1993) 319–342, arXiv:hep-ph/9212278.DOI
- [54]M. Leurer, Y. Nir, and N. Seiberg, “Mass matrix models: The Sequel,” Nucl. Phys. B 420 (1994) 468–504, arXiv:hep-ph/9310320.DOI
- [55]V. Knapp-Perez, X.-G. Liu, H. P. Nilles, and S. Ramos-Sanchez, “Demystifying stringy miracles with eclectic flavor symmetries,” Phys. Rev. D 113 no. 10, (2026) 106015, arXiv:2512.21382 [hep-th].