Introduction
The remarkable success of the Standard Model (SM), culminating in the discovery of a Higgs boson with properties consistent with its predictions, has so far not been accompanied by direct evidence for new particles at the electroweak scale. If new degrees of freedom are sufficiently heavier than the energies probed experimentally, their low-energy effects can be systematically described within an effective field theory (EFT). The Standard Model Effective Field Theory (SMEFT) provides a model-independent framework for this purpose, extending the SM Lagrangian by higher-dimensional operators constructed from SM fields and invariant under the SM gauge symmetries. The corresponding Wilson coefficients encode the effects of heavy dynamics, while the operators are suppressed by powers of a characteristic new-physics scale Λ \Lambda Λ [1 , 2 , 3 , 4 , 5 ] . In the absence of compelling evidence for a particular ultraviolet (UV) completion, the SMEFT has become a standard framework for interpreting electroweak, flavor, Higgs, and collider measurements in searches for physics beyond the SM.
The generality of the SMEFT also presents a major challenge. Already at dimension six, the number of independent Wilson coefficients becomes very large once the full flavor structure is retained [6 , 7 ] . Experimental constraints must therefore be interpreted in a multidimensional parameter space, where correlations among operators, approximate flat directions, and assumptions about flavor can substantially affect the results. In this context, theoretical constraints derived from general properties of scattering amplitudes provide valuable complementary information. Two particularly useful approaches are perturbative unitarity and analyticity-based dispersion relations [8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 ] . While both are rooted in fundamental properties of the scattering matrix, they provide distinct yet complementary constraints on the effective theory.
Perturbative unitarity provides a quantitative criterion for assessing the validity of the EFT description at a given energy. Higher-dimensional operators typically generate scattering amplitudes that grow with the center-of-mass energy, eventually challenging the perturbative expansion. Partial-wave unitarity translates this behavior into bounds on combinations of Wilson coefficients [10 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 ] . These bounds should not be interpreted as absolute exclusions of EFT parameter space: rather, their saturation indicates the energy scale at which the perturbative description ceases to be reliable and additional dynamics must become relevant. This issue is particularly important for collider measurements probing energies potentially comparable to the scale of new physics [40 ] .
While traditional partial-wave analyses are generally formulated in terms of 2 → 2 2 \to2 2 → 2 scattering, a complete treatment of higher-dimensional interactions may require processes with larger numbers of external particles. In particular, some SMEFT operators do not generate the relevant energy-growing contributions in 2 → 2 2 \to2 2 → 2 amplitudes, making higher-multiplicity scattering essential. The on-shell formalism developed in Ref. [36 ] extends partial-wave unitarity to generic N → M N \to M N → M amplitudes, allowing scattering processes with different particle multiplicities to be treated within a common framework; see also Ref. [38 ] . Moreover, the strongest bounds are obtained by imposing unitarity on the eigenvalues of the full coupled-channel partial-wave matrices, which include all states with the same conserved quantum numbers. The resulting constraints generally involve several Wilson coefficients simultaneously, rather than restricting each operator independently.
Previous studies have investigated partial-wave unitarity in various sectors of the dimension-six SMEFT [18 , 19 , 20 , 21 ] , focusing predominantly on bosonic interactions and selected operators containing fermions. In Ref. [37 ] , we performed a complete coupled-channel reassessment of perturbative unitarity in the baryon-number-conserving dimension-six SMEFT, extending the analysis to the full four-fermion basis, gluonic operators, ψ 2 φ 3 \psi^2\varphi^3 ψ 2 φ 3 interactions, and the purely scalar operator ( φ † φ ) 3 (\varphi^\dagger\varphi)^3 ( φ † φ ) 3 . An important aspect of this approach is the treatment of correlations among Wilson coefficients: bounds on individual coefficients are obtained by marginalizing over the remaining interactions, yielding conservative projections of the multidimensional unitarity region rather than conventional single-operator limits. In the present work, we provide a detailed account of this analysis, including explicit examples of the coupled-channel construction and the role of different angular-momentum and particle-multiplicity sectors.
Perturbative unitarity primarily constrains the magnitude of effective interactions, but generally provides limited information about the quantum numbers of the heavy states responsible for them. Additional insights can be obtained from dispersion relations derived from the analyticity, crossing symmetry, and unitarity of the scattering matrix. In particular, spinning sum rules relate suitable combinations of dimension-six four-fermion Wilson coefficients to the angular momentum of intermediate states in the UV [11 , 12 , 13 , 14 , 16 , 17 ] . Under appropriate assumptions, these relations lead to sign constraints that can distinguish scalar- from vector-dominated heavy dynamics. Therefore, they offer a complementary perspective on the possible microscopic origin of effective four-fermion interactions.
The interpretation of spinning sum rules, however, requires care. Unlike perturbative unitarity bounds, their sign constraints are not universal consistency conditions on arbitrary SMEFT coefficients. In the form considered here, they rely on assumptions about the UV dynamics, including a tree-level origin of the relevant interactions, dominance by states of definite spin, and the absence of boundary terms in the dispersion relations [12 , 13 , 14 ] . These conditions need not hold in general. Loop-induced interactions, contributions from states of different spins, or non-vanishing boundary terms associated with t t t -channel vector exchange can modify or invalidate the resulting inequalities. Consequently, a violation of a scalar- or vector-dominated sum-rule condition does not necessarily imply an inconsistent UV completion. Rather, the applicability and limitations of these relations must be assessed in conjunction with the underlying dynamics. In this work, we formulate the spinning sum rules in the Warsaw basis and investigate their interpretation through representative UV completions.
A central objective of our analysis is to extend these theoretical constraints beyond the single-flavor limit. This is particularly relevant for four-fermion operators, whose Wilson coefficients carry four flavor indices and whose number of independent components grows rapidly when all three SM generations are included [6 , 7 ] . At the same time, flavor non-universality is a characteristic feature of many well-motivated new-physics scenarios, especially those involving enhanced couplings to third-generation fermions [41 , 42 , 43 ] . A completely general treatment of the flavor indices is therefore difficult to implement and interpret, whereas the single-flavor limit misses important correlations among generations. We address this problem by considering two theoretically motivated flavor hypotheses. The first is Minimal Flavor Violation (MFV), based on the U ( 3 ) 5 U(3)^5 U ( 3 ) 5 flavor symmetry of the SM gauge interactions, broken only by the Yukawa matrices, treated as spurions [44 , 45 ] . In our analysis, we retain the flavor-symmetric structures and the leading corrections involving the up-type Yukawa matrix, which capture the dominant effects associated with the top quark. The second hypothesis is an approximate U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, under which the first two generations transform as doublets while the third generation is a singlet. This structure provides a natural framework for describing non-universal interactions involving third-generation fermions and is particularly relevant to a broad class of TeV-scale new-physics scenarios [41 , 42 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 ] . Both hypotheses reduce the number of independent four-fermion coefficients while retaining non-trivial information about their flavor structure.
Within these flavor frameworks, we derive perturbative unitarity bounds and spinning sum-rule constraints on the independent four-fermion Wilson coefficients. The resulting conditions involve correlated combinations of flavor structures and can reveal restrictions that are absent in the single-flavor limit. In particular, the allowed regions need not be described by independent intervals for each coefficient, and their geometry can reflect non-trivial relations among flavor components. We further investigate the interpretation of these constraints by matching representative UV completions onto the corresponding effective operators, illustrating both the information encoded in spinning sum rules and the circumstances under which their conventional spin interpretation can fail.
The paper is organized as follows. In § 2 , we review the generalized partial-wave formalism and present the perturbative unitarity bounds for the complete baryon-number-conserving dimension-six SMEFT, including illustrative coupled-channel examples. In § 3 , we formulate spinning sum rules for four-fermion operators in the Warsaw basis, derive their implications in flavor space, and discuss their assumptions and limitations. In § 4 , we extend both analyses to the three-generation four-fermion sector under MFV and U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, and examine representative UV completions. We summarize our findings in § 5 . Mathematical tools used in the analysis are collected in App. § 7 , while the complete multiflavor results are presented in App. § 8 .
Partial-wave unitarity bounds
In this section we derive perturbative partial-wave unitarity bounds for the baryon-number-conserving dimension-six SMEFT. We first review the on-shell formalism used to decompose generic N → M N \to M N → M scattering amplitudes into states of definite total angular momentum. We then illustrate its implementation through two representative examples. The operator O φ = ( φ † φ ) 3 O_{\varphi} = (\varphi^{\dagger}\varphi)^3 O φ = ( φ † φ ) 3 highlights the role of higher-multiplicity processes, since its dimension-six 2 → 2 2 \to2 2 → 2 contribution does not grow with energy, whereas O φ e O_{\varphi e} O φ e provides a simple example in which different partial waves compete in setting the strongest constraint. We then systematically apply the formalism to the complete baryon-number-conserving dimension-six Warsaw basis, performing a fully coupled-channel analysis over all relevant scattering processes and retaining the correlations among Wilson coefficients that contribute to the same partial-wave amplitudes.
Review of the method
We adopt the on-shell partial-wave formalism developed in Ref. [36 ] , which generalizes the usual 2 → 2 2 \to2 2 → 2 partial-wave decomposition to scattering amplitudes with arbitrary multiplicity. For a generic transition from an initial state i i i to a final state f f f , the amplitude can be regarded as a vector in the corresponding kinematic space and decomposed onto a basis of states with definite total angular momentum J J J ,
∣ A i → f ⟩ = ∑ J a i → f J ∣ B i → f J ⟩ . (1) \lvert\mathcal{A}_{i\to f} \rangle= \sum_J a^J_{i\to f} \lvert\mathcal{B}^J_{i\to f} \rangle. \tag*{(1)} ∣ A i → f ⟩ = J ∑ a i → f J ∣ B i → f J ⟩ . (1)
where a i → f J a^J_{i\to f} a i → f J denotes the partial-wave coefficient. The angular-momentum basis is chosen to satisfy
⟨ B i → f J ∣ B i → f J ′ ⟩ = ∫ d Φ i d Φ f B i → f J ′ ( B i → f J ) ∗ = ( 2 J + 1 ) δ J J ′ . (2) \langle\mathcal{B}^J_{i\to f} \lvert\mathcal{B}^{J'}_{i\to f} \rangle= \int d\Phi_i\,d\Phi_f\,\mathcal{B}^{J'}_{i\to f}\left(\mathcal{B}^J_{i\to f}\right)^* = (2J+1)\delta^{JJ'} . \tag*{(2)} ⟨ B i → f J ∣ B i → f J ′ ⟩ = ∫ d Φ i d Φ f B i → f J ′ ( B i → f J ) ∗ = ( 2 J + 1 ) δ J J ′ . (2)
with d Φ i d\Phi_i d Φ i and d Φ f d\Phi_f d Φ f the Lorentz-invariant phase-space measures associated with the initial and final states. For massless n n n -particle states of distinguishable particles, the phase-space volume is normalized as
Φ n = 1 8 π ( n − 1 ) ! ( n − 2 ) ! ( s 4 π ) 2 n − 4 . (3) \Phi_n = \frac{1}{8\pi(n-1)!(n-2)!}\left(\frac{\sqrt{s}}{4\pi}\right)^{2n-4}. \tag*{(3)} Φ n = 8 π ( n − 1 )! ( n − 2 )! 1 ( 4 π s ) 2 n − 4 . (3)
With this normalization, unitarity of the S S S matrix implies
∣ Re a i → i J ∣ ≤ 1 , 0 ≤ Im a i → i J ≤ 2 , ∣ a i → f J ∣ ≤ 1 ( i ≠ f ) . (4) \left|\operatorname{Re}a^J_{i\to i}\right|\le1,\qquad0\le\operatorname{Im}a^J_{i\to i}\le2,\qquad\left|a^J_{i\to f}\right|\le1\quad(i\ne f). \tag*{(4)} Re a i → i J ≤ 1 , 0 ≤ Im a i → i J ≤ 2 , a i → f J ≤ 1 ( i = f ) . (4)
For 2 → 2 2\to2 2 → 2 scattering, the basis elements ∣ B i → f J ⟩ \lvert\mathcal{B}^J_{i\to f}\rangle ∣ B i → f J ⟩ reduce, up to normalization, to the familiar Wigner d d d -functions [63 , 64 ] . The advantage of the present formulation is that the same construction continues to apply when either the initial or final state contains more than two particles, without requiring a separate partial-wave parametrization for each multiplicity.
For a fixed value of J J J , the strongest bounds are obtained by considering simultaneously all states carrying the same conserved quantum numbers. The corresponding partial-wave coefficients therefore assemble into a coupled-channel matrix in the space of particle species, helicities, gauge indices, flavor indices, and, when relevant, different particle multiplicities. Perturbative unitarity is then imposed on the eigenvalues of this matrix, rather than separately on individual scattering amplitudes. This step is essential whenever several processes communicate through the same interaction, since coherent combinations of channels can strengthen the resulting constraint.
In practice, the construction proceeds directly from on-shell amplitudes. For each operator, we generate all scattering processes that display the leading energy growth at a given order in the EFT expansion, determine the allowed values of J J J from their helicity structure, and project the amplitudes onto the corresponding angular-momentum basis [36 , 65 , 66 ] . All independent gauge configurations and, whenever retained, flavor configurations are included in the coupled-channel matrix, together with the appropriate normalization factors for identical particles.
When more than one angular momentum contributes, the unitarity condition is imposed independently in each J J J sector and the strongest resulting constraint is retained. As we shall see explicitly below, the lowest allowed angular momentum does not necessarily provide the strongest bound once the full multiplicity of internal quantum numbers is taken into account.
Throughout this section we work in the high-energy limit, where particle masses can be neglected, and retain the leading energy-growing contributions induced by the dimension-six interactions. Purely SM terms, which do not exhibit the energy growth relevant for the unitarity bounds derived here, are therefore neglected unless explicitly stated otherwise. Besides isolating the UV behavior of interest, this prescription avoids infrared singularities associated with possible massless t t t -channel exchanges in the SM amplitudes, whose partial-wave projection would require a separate infrared treatment; see, e.g., Refs. [67 , 68 , 69 , 70 , 71 , 72 , 73 , 74 , 75 , 76 , 77 , 78 , 79 ] . An exception is provided by the X 3 X^3 X 3 operators, for which the role of the SM contribution is discussed separately in § 2.2.1 .
The resulting bounds should therefore be interpreted as perturbative consistency conditions on the SMEFT at a given center-of-mass energy, rather than as sharp exclusions of points in Wilson-coefficient space. Their saturation identifies the scale at which the perturbative EFT description ceases to be reliable and additional dynamics must become relevant.
For later convenience, we label an external particle by N I N_I N I , where N = 1 , 2 , … N=1,2,\ldots N = 1 , 2 , … denotes its momentum label and I I I collectively denotes its gauge quantum numbers. An antiparticle is indicated by placing a bar over the corresponding momentum label. This notation will be used in the explicit examples below.
Example: O φ O_\varphi O φ operator
A particularly instructive application of the generalized N → M N\to M N → M formalism is provided by the purely Higgs operator
O φ = ( φ † φ ) 3 . (5) O_\varphi=(\varphi^\dagger\varphi)^3 . \tag*{(5)} O φ = ( φ † φ ) 3 . (5)
Unlike the derivative dimension-six operators, its 2 → 2 2\to2 2 → 2 contribution does not display the energy growth relevant for the high-energy unitarity bounds. The relevant unitarity constraint instead arises from processes involving six external scalar fields, so that one must consider 3 → 3 3\to3 3 → 3 and 2 → 4 2\to4 2 → 4 scattering. This makes O φ O_\varphi O φ a simple example in which the extension beyond the standard 2 → 2 2\to2 2 → 2 partial-wave decomposition is essential.
We first consider the 3 → 3 3\to3 3 → 3 channels. Up to charge conjugation, the independent amplitudes can be written as
A ( 1 a , 2 b , 3 c → 4 d , 5 e , 6 f ) = A ( 1 ˉ a , 2 ˉ b , 3 ˉ c → 4 ˉ d , 5 ˉ e , 6 ˉ f ) = 36 C φ Λ 2 δ a d δ b e δ c f , (6) \mathcal{A}(1_a,2_b,3_c\to4_d,5_e,6_f)=\mathcal{A}(\bar{1}_a,\bar{2}_b,\bar{3}_c\to\bar{4}_d,\bar{5}_e,\bar{6}_f)=36\frac{C_\varphi}{\Lambda^2}\delta_a^d\delta_b^e\delta_c^f , \tag*{(6)} A ( 1 a , 2 b , 3 c → 4 d , 5 e , 6 f ) = A ( 1 ˉ a , 2 ˉ b , 3 ˉ c → 4 ˉ d , 5 ˉ e , 6 ˉ f ) = 36 Λ 2 C φ δ a d δ b e δ c f , (6)
A ( 1 a , 2 b , 3 ˉ c → 4 ˉ d , 5 ˉ e , 6 ˉ f ) = A ( 1 ˉ a , 2 ˉ b , 3 c → 4 ˉ d , 5 ˉ e , 6 f ) = 36 C φ Λ 2 δ a d δ b e δ f c . (7) \mathcal{A}(1_a,2_b,\bar{3}_c\to\bar{4}_d,\bar{5}_e,\bar{6}_f)=\mathcal{A}(\bar{1}_a,\bar{2}_b,3_c\to\bar{4}_d,\bar{5}_e,6_f)=36\frac{C_\varphi}{\Lambda^2}\delta_a^d\delta_b^e\delta_f^c . \tag*{(7)} A ( 1 a , 2 b , 3 ˉ c → 4 ˉ d , 5 ˉ e , 6 ˉ f ) = A ( 1 ˉ a , 2 ˉ b , 3 c → 4 ˉ d , 5 ˉ e , 6 f ) = 36 Λ 2 C φ δ a d δ b e δ f c . (7)
where a , b , c , d , e , f = 1 , 2 a,b,c,d,e,f=1,2 a , b , c , d , e , f = 1 , 2 are S U ( 2 ) L \mathrm{SU}(2)_L SU ( 2 ) L indices and we have defined the symmetrized tensor
δ a ( d δ b e δ c f ) = 1 6 [ δ a d δ b e δ c f + δ a d δ b f δ c e + δ a e δ b d δ c f + δ a e δ b f δ c d + δ a f δ b d δ c e + δ a f δ b e δ c d ] . (8) \delta_a^{(d}\delta_b^e\delta_c^{f)}=\frac{1}{6}\left[\delta_a^d\delta_b^e\delta_c^f+\delta_a^d\delta_b^f\delta_c^e+\delta_a^e\delta_b^d\delta_c^f+\delta_a^e\delta_b^f\delta_c^d+\delta_a^f\delta_b^d\delta_c^e+\delta_a^f\delta_b^e\delta_c^d\right]. \tag*{(8)} δ a ( d δ b e δ c f ) = 6 1 [ δ a d δ b e δ c f + δ a d δ b f δ c e + δ a e δ b d δ c f + δ a e δ b f δ c d + δ a f δ b d δ c e + δ a f δ b e δ c d ] . (8)
The corresponding 2 → 4 2\to4 2 → 4 amplitudes, together with their crossed 4 → 2 4\to2 4 → 2 processes, are
A ( 1 a , 2 b → 3 c , 4 d , 5 e , 6 ˉ f ) = A ( 1 ˉ a , 2 ˉ b → 3 ˉ c , 4 ˉ d , 5 ˉ e , 6 f ) = 36 C φ Λ 2 δ a c δ b d δ f e , (9) \mathcal{A}(1_a,2_b\to3_c,4_d,5_e,\bar{6}_f)=\mathcal{A}(\bar{1}_a,\bar{2}_b\to\bar{3}_c,\bar{4}_d,\bar{5}_e,6_f)=36\frac{C_\varphi}{\Lambda^2}\delta_a^c\delta_b^d\delta_f^e, \tag*{(9)} A ( 1 a , 2 b → 3 c , 4 d , 5 e , 6 ˉ f ) = A ( 1 ˉ a , 2 ˉ b → 3 ˉ c , 4 ˉ d , 5 ˉ e , 6 f ) = 36 Λ 2 C φ δ a c δ b d δ f e , (9)
A ( 1 a , 2 ˉ b → 3 c , 4 d , 5 ˉ e , 6 f ) = 36 C φ Λ 2 δ a b δ c d δ e f . (10) \mathcal{A}(1_a,\bar{2}_b\to3_c,4_d,\bar{5}_e,6_f)=36\frac{C_\varphi}{\Lambda^2}\delta_a^b\delta_c^d\delta_e^f. \tag*{(10)} A ( 1 a , 2 ˉ b → 3 c , 4 d , 5 ˉ e , 6 f ) = 36 Λ 2 C φ δ a b δ c d δ e f . (10)
Since all external particles are scalars and the interaction contains no derivatives, these amplitudes contribute entirely to the J = 0 J=0 J = 0 partial wave. The phase-space integrations are therefore particularly simple, and the relevant normalized basis elements are
B 3 → 3 0 = 256 π 3 s , B 2 → 4 0 = B 4 → 2 0 = 256 π 3 3 s . (11) \mathcal{B}^0_{3\to3}=\frac{256\pi^3}{s},\qquad\mathcal{B}^0_{2\to4}=\mathcal{B}^0_{4\to2}=\frac{256\pi^3\sqrt{3}}{s}. \tag*{(11)} B 3 → 3 0 = s 256 π 3 , B 2 → 4 0 = B 4 → 2 0 = s 256 π 3 3 . (11)
In particular, B 2 → 4 0 = 1 / Φ 2 Φ 4 \mathcal{B}^0_{2\to4}=1/\sqrt{\Phi_2\Phi_4} B 2 → 4 0 = 1/ Φ 2 Φ 4 , consistently with the normalization introduced in § 2.1 . When a state contains identical particles the symmetry factor S = ∏ k n k ! S=\prod_k n_k! S = ∏ k n k ! of its phase-space measure is accounted for by a factor 1 / S i S f 1/\sqrt{S_iS_f} 1/ S i S f in the partial-wave coefficient. The full J = 0 J=0 J = 0 scattering matrix decomposes into independent blocks according to the quantum numbers and field content of the states connected by O φ O_\varphi O φ .
As a first example, consider the φ φ φ † → φ φ φ † \varphi\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger φφ φ † → φφ φ † sector. The corresponding partial-wave matrix is
a φ φ φ † → φ φ φ † 0 = s C φ 256 π 3 Λ 2 ( 18 0 0 6 2 0 0 0 6 0 0 0 0 0 0 12 0 0 6 2 6 2 0 0 12 0 0 0 0 0 0 6 0 0 0 6 2 0 0 18 ) . (12) a^0_{\varphi\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger}=\frac{sC_\varphi}{256\pi^3\Lambda^2}
\begin{pmatrix}
18&0&0&6\sqrt{2}&0&0\\
0&6&0&0&0&0\\
0&0&12&0&0&6\sqrt{2}\\
6\sqrt{2}&0&0&12&0&0\\
0&0&0&0&6&0\\
0&0&6\sqrt{2}&0&0&18
\end{pmatrix}. \tag*{(12)} a φφ φ † → φφ φ † 0 = 256 π 3 Λ 2 s C φ 18 0 0 6 2 0 0 0 6 0 0 0 0 0 0 12 0 0 6 2 6 2 0 0 12 0 0 0 0 0 0 6 0 0 0 6 2 0 0 18 . (12)
The matrix is 6 × 6 6\times6 6 × 6 , corresponding to the six independent S U ( 2 ) L \mathrm{SU}(2)_L SU ( 2 ) L configurations
( a , b , c ) ∈ { ( 1 , 1 ) , ( 1 , 2 ) , ( 2 , 2 ) } ⊗ { 1 , 2 } , (13) (a,b,c)\in\{(1,1),(1,2),(2,2)\}\otimes\{1,2\}, \tag*{(13)} ( a , b , c ) ∈ {( 1 , 1 ) , ( 1 , 2 ) , ( 2 , 2 )} ⊗ { 1 , 2 } , (13)
where a a a and b b b label the two φ \varphi φ fields and c c c labels φ † \varphi^\dagger φ † . Imposing unitarity on the eigenvalue of largest modulus yields
3 s 32 π 3 Λ 2 ∣ C φ ∣ ≤ 1. (14) \frac{3s}{32\pi^3\Lambda^2}|C_\varphi|\leq1. \tag*{(14)} 32 π 3 Λ 2 3 s ∣ C φ ∣ ≤ 1. (14)
A qualitatively different block couples the two- and four-particle states
i , f ∈ { φ φ † , φ φ φ † φ † } . (15) i,f\in\left\{\varphi\varphi^\dagger,\varphi\varphi\varphi^\dagger\varphi^\dagger\right\}. \tag*{(15)} i , f ∈ { φ φ † , φφ φ † φ † } . (15)
Since O φ O_\varphi O φ does not connect the two-particle sector to itself at the energy-growing order considered here, nor the four-particle sector to itself through the same six-field contact vertex, the partial-wave matrix takes the off-diagonal form
a i → f 0 = ( 0 M φ φ † → φ φ φ † φ † M φ φ † → φ φ φ † φ † † 0 ) , (16) a^0_{i\to f} =
\begin{pmatrix}
0 & M_{\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger\varphi^\dagger} \\
M^\dagger_{\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger\varphi^\dagger} & 0
\end{pmatrix}, \tag*{(16)} a i → f 0 = ( 0 M φ φ † → φφ φ † φ † † M φ φ † → φφ φ † φ † 0 ) , (16)
with
M φ φ † → φ φ φ † φ † = s C φ 256 π 3 Λ 2 ( 6 3 0 0 0 4 3 0 0 0 2 3 0 2 6 0 0 0 2 6 0 0 0 0 0 0 2 6 0 0 0 2 6 0 2 3 0 0 0 4 3 0 0 0 6 3 ) . (17) M_{\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger\varphi^\dagger}
=
\frac{s C_\varphi}{256\pi^3\Lambda^2}
\begin{pmatrix}
6\sqrt{3} & 0 & 0 & 0 & 4\sqrt{3} & 0 & 0 & 0 & 2\sqrt{3} \\
0 & 2\sqrt{6} & 0 & 0 & 0 & 2\sqrt{6} & 0 & 0 & 0 \\
0 & 0 & 0 & 2\sqrt{6} & 0 & 0 & 0 & 2\sqrt{6} & 0 \\
2\sqrt{3} & 0 & 0 & 0 & 4\sqrt{3} & 0 & 0 & 0 & 6\sqrt{3}
\end{pmatrix}. \tag*{(17)} M φ φ † → φφ φ † φ † = 256 π 3 Λ 2 s C φ 6 3 0 0 2 3 0 2 6 0 0 0 0 0 0 0 0 2 6 0 4 3 0 0 4 3 0 2 6 0 0 0 0 0 0 0 0 2 6 0 2 3 0 0 6 3 . (17)
Here the 4 × 9 4 \times9 4 × 9 rectangular block follows from the independent gauge configurations
( a , b ) ∈ { 1 , 2 } ⊗ { 1 , 2 } , ( c , d , e , f ) ∈ { ( 1 , 1 ) , ( 1 , 2 ) , ( 2 , 2 ) } ⊗ { ( 1 , 1 ) , ( 1 , 2 ) , ( 2 , 2 ) } . (18) (a,b) \in\{1,2\} \otimes\{1,2\}, \qquad(c,d,e,f) \in\{(1,1),(1,2),(2,2)\} \otimes\{(1,1),(1,2),(2,2)\}. \tag*{(18)} ( a , b ) ∈ { 1 , 2 } ⊗ { 1 , 2 } , ( c , d , e , f ) ∈ {( 1 , 1 ) , ( 1 , 2 ) , ( 2 , 2 )} ⊗ {( 1 , 1 ) , ( 1 , 2 ) , ( 2 , 2 )} . (18)
The non-vanishing eigenvalues of the complete Hermitian matrix are given by the singular values of M M M with both signs (see App. § 7 .§ 2 ). The largest one leads to
3 2 s 64 π 3 Λ 2 ∣ C φ ∣ ≤ 1. (19) \frac{3\sqrt{2}\,s}{64\pi^3\Lambda^2}|C_\varphi| \le1. \tag*{(19)} 64 π 3 Λ 2 3 2 s ∣ C φ ∣ ≤ 1. (19)
The remaining independent blocks can be analyzed in the same way. Their bounds are collected in Table 1 . The strongest constraint is obtained from the φ φ φ † → φ φ φ † \varphi\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger φφ φ † → φφ φ † sector,
Table 1. Partial-wave unitarity bounds on the modulus of C φ C_\varphi C φ obtained from the independent J = 0 J=0 J = 0 scattering sectors generated by O φ O_\varphi O φ . Channels related by CP conjugation, such as φ φ † φ † → φ φ † φ † \varphi\varphi^\dagger\varphi^\dagger\to\varphi\varphi^\dagger\varphi^\dagger φ φ † φ † → φ φ † φ † , yield identical bounds and are not listed separately. The factors s / Λ 2 s/\Lambda^2 s / Λ 2 are left implicit.
s Λ 2 ∣ C φ ∣ ≤ 32 π 3 3 . (20) \frac{s}{\Lambda^2}|C_\varphi| \le\frac{32\pi^3}{3}. \tag*{(20)} Λ 2 s ∣ C φ ∣ ≤ 3 32 π 3 . (20)
This example illustrates two features of the generalized partial-wave formalism. First, higher-multiplicity channels are indispensable for operators whose energy-growing contribution is not captured by 2 → 2 2 \to2 2 → 2 scattering. Second, channels with different initial- and final-state multiplicities can lead to parametrically comparable constraints once the normalization of the multiparticle phase space and the symmetry factors of identical states are properly included.
Example: O φ e O_{\varphi e} O φ e operator
A second instructive example is provided by the operator1
O φ e = ( φ † i D μ ↔ φ ) ( e ˉ p γ μ e r ) . (21) O_{\varphi e} = \left(\varphi^\dagger i\overleftrightarrow{D_\mu}\varphi\right)\left(\bar e_p\gamma^\mu e_r\right). \tag*{(21)} O φ e = ( φ † i D μ φ ) ( e ˉ p γ μ e r ) . (21)
Unlike the purely scalar example discussed above, this interaction illustrates the non-trivial interplay between partial waves with different total angular momentum. At the 2 → 2 2 \to2 2 → 2 level, the relevant amplitudes are
A ( 1 p + , 2 ˉ r − → 3 a , 4 b ) = 2 Λ 2 [ C φ e ] r p δ b a [ 23 ] ⟨ 31 ⟩ , (22) \mathcal{A}\left(1_p^+,\bar{2}_r^- \to3_a,4_b\right) = \frac{2}{\Lambda^2}[C_{\varphi e}]_{rp}\delta_b^a[23]\langle31\rangle, \tag*{(22)} A ( 1 p + , 2 ˉ r − → 3 a , 4 b ) = Λ 2 2 [ C φ e ] r p δ b a [ 23 ] ⟨ 31 ⟩ , (22)
A ( 1 p + , 2 a → 3 r + , 4 b ) = 2 Λ 2 [ C φ e ] r p δ a b [ 23 ] ⟨ 21 ⟩ , (23) \mathcal{A}\left(1_p^+,2_a \to3_r^+,4_b\right) = \frac{2}{\Lambda^2}[C_{\varphi e}]_{rp}\delta_a^b[23]\langle21\rangle, \tag*{(23)} A ( 1 p + , 2 a → 3 r + , 4 b ) = Λ 2 2 [ C φ e ] r p δ a b [ 23 ] ⟨ 21 ⟩ , (23)
where p , r = 1 , 2 , 3 p,r=1,2,3 p , r = 1 , 2 , 3 denote flavor indices and a , b = 1 , 2 a,b=1,2 a , b = 1 , 2 are S U ( 2 ) L \mathrm{SU}(2)_L SU ( 2 ) L indices. For a right-handed fermion the physical helicity is h = + 1 / 2 h=+1/2 h = + 1/2 . The helicity differences therefore imply that the annihilation channel e e ˉ → φ φ † e\bar e \to\varphi\varphi^\dagger e e ˉ → φ φ † starts at J = 1 J=1 J = 1 , whereas the scattering channel e φ → e φ e\varphi\to e\varphi e φ → e φ admits J = 1 / 2 J=1/2 J = 1/2 .
To illustrate the relative strength of the two sectors, we work in the single-flavor limit. Projecting the annihilation amplitude onto the J = 1 J=1 J = 1 partial wave gives
a i → f 1 = ( 0 M M † 0 ) , M = − 2 s 24 π Λ 2 C φ e ( 1 0 0 1 ) , (24) a^1_{i\to f}=\begin{pmatrix} 0 & M \\ M^\dagger& 0 \end{pmatrix}, \qquad M=-\frac{\sqrt{2}s}{24\pi\Lambda^2}C_{\varphi e}\begin{pmatrix}1&0&0&1\end{pmatrix}, \tag*{(24)} a i → f 1 = ( 0 M † M 0 ) , M = − 24 π Λ 2 2 s C φ e ( 1 0 0 1 ) , (24)
with
i , f ∈ { e e ˉ , φ φ † } . (25) i,f\in\left\{e\bar e,\varphi\varphi^\dagger\right\}. \tag*{(25)} i , f ∈ { e e ˉ , φ φ † } . (25)
The rectangular block is 1 × 4 1\times4 1 × 4 , corresponding to the four independent weak-isospin configurations
( a , b ) ∈ { 1 , 2 } ⊗ { 1 , 2 } . (26) (a,b)\in\{1,2\}\otimes\{1,2\}. \tag*{(26)} ( a , b ) ∈ { 1 , 2 } ⊗ { 1 , 2 } . (26)
The non-vanishing eigenvalues of the full Hermitian matrix are again given, up to a sign, by the singular values of M M M (see App. § 7 .§ 2 ). Since M M M has rank one, its only non-vanishing singular value, 2 × 2 s ∣ C φ e ∣ / ( 24 π Λ 2 ) \sqrt{2}\times\sqrt{2}s|C_{\varphi e}|/(24\pi\Lambda^2) 2 × 2 s ∣ C φ e ∣/ ( 24 π Λ 2 ) , leads to
s 12 π Λ 2 ∣ C φ e ∣ ≤ 1. (27) \frac{s}{12\pi\Lambda^2}|C_{\varphi e}|\leq1. \tag*{(27)} 12 π Λ 2 s ∣ C φ e ∣ ≤ 1. (27)
The crossed channel e φ → e φ e\varphi\to e\varphi e φ → e φ instead contributes already at J = 1 / 2 J=1/2 J = 1/2 . Its partial-wave matrix is
a e φ → e φ 1 / 2 = s 8 π Λ 2 C φ e ( 1 0 0 1 ) , (28) a^{1/2}_{e\varphi\to e\varphi}=\frac{s}{8\pi\Lambda^2}C_{\varphi e}\begin{pmatrix}1&0\\0&1\end{pmatrix}, \tag*{(28)} a e φ → e φ 1/2 = 8 π Λ 2 s C φ e ( 1 0 0 1 ) , (28)
and therefore yields
s 8 π Λ 2 ∣ C φ e ∣ ≤ 1. (29) \frac{s}{8\pi\Lambda^2}|C_{\varphi e}|\leq1. \tag*{(29)} 8 π Λ 2 s ∣ C φ e ∣ ≤ 1. (29)
For O φ e O_{\varphi e} O φ e , the J = 1 / 2 J=1/2 J = 1/2 channel is thus the most constraining one.
This conclusion should not, however, be interpreted as a general rule that the smallest kinematically allowed angular momentum always gives the strongest unitarity bound. The strength of a coupled-channel constraint depends not only on the partial wave itself, but also on the number and structure of the internal gauge configurations that contribute coherently to the scattering matrix. For operators involving fermions with non-trivial weak-isospin or color quantum numbers, the enhancement associated with these additional degrees of freedom can make a higher-J J J sector more restrictive. As discussed in § 2.2.2 , this occurs for several operators of the class ψ 2 φ 2 D \psi^2\varphi^2D ψ 2 φ 2 D , for which the J = 1 J=1 J = 1 bound can supersede the corresponding J = 1 / 2 J=1/2 J = 1/2 constraint.
Results
Using the framework described above, we systematically compute the partial-wave unitarity bounds for the complete set of baryon-number-conserving dimension-six SMEFT operators in the Warsaw basis [2 ] , collected in Table 2 . For each operator class, we include all scattering channels displaying the relevant high-energy behavior and account for all allowed helicity, gauge, and, whenever applicable, flavor configurations. The resulting marginalized bounds are summarized in Table 3 . For operators involving fermions, the results presented in this section are obtained in the single-flavor limit; the extension of the four-fermion sector to three generations will be discussed in § 4 .
Table 2. Dimension-six operators of the SMEFT in the Warsaw basis [2 ] that conserve baryon number. Non-Hermitian operators are denoted with ‡ O \ddagger O ‡ O and the corresponding Hermitian conjugates must be included in the Lagrangian. Here p p p , r r r , s s s , t t t are flavor indices, j j j , k k k are SU(2)L _L L indices, ϵ j k \epsilon_{jk} ϵ j k is the antisymmetric tensor with ϵ 12 = + 1 \epsilon_{12}=+1 ϵ 12 = + 1 , σ I \sigma^I σ I are the Pauli matrices, T A T^A T A are the SU(3)C _C C generators, φ ~ = i σ 2 φ ∗ \widetilde{\varphi}=i\sigma_2\varphi^* φ = i σ 2 φ ∗ , and X ~ μ ν = 1 2 ϵ μ ν ρ σ X ρ σ \widetilde{X}_{\mu\nu}=\frac{1}{2}\epsilon_{\mu\nu\rho\sigma}X^{\rho\sigma} X μν = 2 1 ϵ μν ρ σ X ρ σ with ϵ 0123 = + 1 \epsilon_{0123}=+1 ϵ 0123 = + 1 .
Table 3. Table 3. Unitarity bounds on the modulus of the baryon-number-conserving dimension-six SMEFT Wilson coefficients in the Warsaw basis [2 ] (Table 2 ), obtained by marginalizing over all other coefficients. The factors s / Λ 2 s/\Lambda^{2} s / Λ 2 are left implicit, i.e., the reported values constrain the combination s ∣ C ∣ / Λ 2 s|C|/\Lambda^{2} s ∣ C ∣/ Λ 2 . Fermionic operators are considered in the single-flavor limit.
An important feature of our analysis is that the bound on a given Wilson coefficient is not obtained by switching off all the remaining ones. Instead, coefficients contributing to the same partial-wave amplitudes are treated simultaneously, and the limit on each coefficient is obtained by marginalizing over the others. Since the partial-wave amplitudes depend linearly on the Wilson coefficients, the unitarity conditions define a convex allowed region in coefficient space, whose projection onto a given coefficient axis yields the corresponding marginalized bound. Since the complete set of scattering channels is included, this region is bounded in all independent directions of coefficient space. The bounds reported in Table 3 should therefore be interpreted as conservative one-dimensional projections of the full coupled-channel unitarity region, rather than as single-operator bounds.
This distinction becomes particularly relevant whenever several Wilson coefficients contribute to the same scattering amplitudes. This occurs, for instance, for the pair ( O φ □ , O φ D ) (O_{\varphi\Box}, O_{\varphi D}) ( O φ □ , O φ D ) , for operators involving a field-strength tensor and its dual, and for fermionic operators that differ only in the singlet or non-singlet SU(2)L _{L} L or SU(3)C _{C} C contraction of their bilinears, such as ( O ℓ q ( 1 ) , O ℓ q ( 3 ) ) (O_{\ell q}^{(1)}, O_{\ell q}^{(3)}) ( O ℓ q ( 1 ) , O ℓ q ( 3 ) ) or ( O q u ( 1 ) , O q u ( 8 ) ) (O_{qu}^{(1)}, O_{qu}^{(8)}) ( O q u ( 1 ) , O q u ( 8 ) ) . Keeping these correlations is essential for obtaining bounds that are independent of arbitrary one-operator-at-a-time assumptions.
To fully exploit the coupled-channel construction, for each operator class we consider all possible initial and final states in the allowed helicity and gauge configurations. At dimension six, the numerically leading constraints frequently arise from the lowest-multiplicity amplitudes. Nevertheless, the general N → M N \to M N → M formalism of Ref. [36 ] is essential for treating the complete basis in a uniform way. In particular, higher-multiplicity amplitudes are required for O φ O_{\varphi} O φ and for the ψ 2 φ 3 \psi^{2}\varphi^{3} ψ 2 φ 3 class, as discussed in § 2.1.1 and § 2.2.2 , respectively.
Compared with previous SMEFT analyses [19 , 20 , 21 ] , our calculation covers the complete set of four-fermion interactions and all operators involving gluons. The fully coupled-channel treatment can also strengthen the bounds whenever several internal gauge configurations contribute coherently to the same partial-wave matrix. In the following we discuss separately the purely bosonic operators, operators containing a fermionic bilinear, and four-fermion interactions.
Purely bosonic operators
We begin with the purely bosonic sector. For the operator classes whose leading high-energy behavior is captured by 2 → 2 2 \to2 2 → 2 scattering, a representative set of channels contributing to the J = 0 J=0 J = 0 partial wave is
φ i φ j ( + ) → φ k φ ℓ ( + ) , X I X J → φ i φ j ( + ) , X I X J → X K X L , (30) \varphi_i\varphi_j^{(+)} \to\varphi_k\varphi_\ell^{(+)}, \qquad X^I X^J \to\varphi_i\varphi_j^{(+)}, \qquad X^I X^J \to X^K X^L, \tag*{(30)} φ i φ j ( + ) → φ k φ ℓ ( + ) , X I X J → φ i φ j ( + ) , X I X J → X K X L , (30)
where i , j , k , ℓ i,j,k,\ell i , j , k , ℓ denote S U ( 2 ) L \mathrm{SU}(2)_L SU ( 2 ) L fundamental or antifundamental indices, while I , J , K , L I,J,K,L I , J , K , L denote adjoint gauge indices and X = B , W , G X=B,W,G X = B , W , G . All states with identical conserved quantum numbers are included simultaneously in the corresponding coupled-channel matrices.
For illustration, with the X 3 X^3 X 3 coefficients set to zero, the weak-isospin- and color-singlet and the weak-isospin-triplet J = 0 J=0 J = 0 sectors that contain the state φ φ † \varphi\varphi^\dagger φ φ † lead, respectively, to the constraints
s 2 Λ 4 ∑ X = B , W , G n X ∣ C φ X + i C φ X ~ ∣ 2 ≤ π ( 8 π − 3 s Λ 2 ∣ C φ □ ∣ ) , (31) \frac{s^2}{\Lambda^4}\sum_{X=B,W,G} n_X\left|\mathcal{C}_{\varphi X}+i\mathcal{C}_{\widetilde{\varphi X}}\right|^2 \leq\pi\left(8\pi-\frac{3s}{\Lambda^2}\left|\mathcal{C}_{\varphi\Box}\right|\right), \tag*{(31)} Λ 4 s 2 X = B , W , G ∑ n X C φX + i C φX 2 ≤ π ( 8 π − Λ 2 3 s ∣ C φ □ ∣ ) , (31)
s 2 Λ 4 ∣ C φ W B + i C φ W B ~ ∣ 2 ≤ 2 π ( 8 π − s Λ 2 ∣ C φ D − C φ □ ∣ ) . (32) \frac{s^2}{\Lambda^4}\left|\mathcal{C}_{\varphi WB}+i\mathcal{C}_{\widetilde{\varphi WB}}\right|^2 \leq2\pi\left(8\pi-\frac{s}{\Lambda^2}\left|\mathcal{C}_{\varphi D}-\mathcal{C}_{\varphi\Box}\right|\right). \tag*{(32)} Λ 4 s 2 C φ W B + i C φ W B 2 ≤ 2 π ( 8 π − Λ 2 s ∣ C φ D − C φ □ ∣ ) . (32)
where n B = 1 n_B=1 n B = 1 , n W = 3 n_W=3 n W = 3 , and n G = 8 n_G=8 n G = 8 . With the X 3 X^3 X 3 coefficients switched on, the singlet matrix also contains the entries X ± X ± → X ∓ X ∓ X^\pm X^\pm\to X^\mp X^\mp X ± X ± → X ∓ X ∓ , and Eq. (31) generalizes to
s 2 Λ 4 ∑ X = B , W , G n X ∣ C φ X ∣ 2 + ρ X Re ( C φ X C X ∗ ) 1 − ρ X 2 ∣ C X ∣ 2 ≤ π ( 8 π ± 3 s Λ 2 C φ □ ) , (33) \frac{s^2}{\Lambda^4}\sum_{X=B,W,G} \frac{n_X\left|\mathcal{C}_{\varphi X}\right|^2+\rho_X\operatorname{Re}\left(\mathcal{C}_{\varphi X}\mathcal{C}_X^*\right)}{1-\rho_X^2\left|\mathcal{C}_X\right|^2} \leq\pi\left(8\pi\pm\frac{3s}{\Lambda^2}\mathcal{C}_{\varphi\Box}\right), \tag*{(33)} Λ 4 s 2 X = B , W , G ∑ 1 − ρ X 2 ∣ C X ∣ 2 n X ∣ C φX ∣ 2 + ρ X Re ( C φX C X ∗ ) ≤ π ( 8 π ± Λ 2 3 s C φ □ ) , (33)
where C φ X ≡ C φ X + i C φ X ~ \mathcal{C}_{\varphi X}\equiv\mathcal{C}_{\varphi X}+i\mathcal{C}_{\widetilde{\varphi X}} C φX ≡ C φX + i C φX , C X ≡ C X + i C X ~ \mathcal{C}_X\equiv\mathcal{C}_X+i\mathcal{C}_{\widetilde X} C X ≡ C X + i C X , ρ B = 0 \rho_B=0 ρ B = 0 , ρ W = 3 g S / ( 2 π Λ 2 ) \rho_W=3g_S/(2\pi\Lambda^2) ρ W = 3 g S / ( 2 π Λ 2 ) , and ρ G = 9 g S / ( 4 π Λ 2 ) \rho_G=9g_S/(4\pi\Lambda^2) ρ G = 9 g S / ( 4 π Λ 2 ) , with ρ W ∣ C W ∣ < 1 \rho_W|\mathcal{C}_W|<1 ρ W ∣ C W ∣ < 1 and ρ G ∣ C G ∣ < 1 \rho_G|\mathcal{C}_G|<1 ρ G ∣ C G ∣ < 1 . Marginalizing these and the remaining independent coupled-channel conditions over the other Wilson coefficients leads to the limits reported in Table 3 . The corresponding J = 1 , 2 J=1,2 J = 1 , 2 sectors have also been included in the analysis, but, for the purely bosonic operators, they yield weaker constraints than those arising from J = 0 J=0 J = 0 .
The operators O φ □ O_{\varphi\Box} O φ □ and O φ D O_{\varphi D} O φ D provide a simple illustration of the importance of retaining correlations among Wilson coefficients. Both contribute to common Higgs scattering amplitudes, and unitarity therefore constrains a two-dimensional region in the ( C φ □ , C φ D ) (\mathcal{C}_{\varphi\Box},\mathcal{C}_{\varphi D}) ( C φ □ , C φ D ) plane rather than two independent intervals. The corresponding allowed region is shown in Figure 1 (left) and the unitarity conditions are listed in Table 4 . The marginalized limits quoted in Table 3 are obtained by projecting this region onto the two coefficient axes.
Figure 1. Parameter space allowed by partial-wave unitarity in the planes ( C φ □ , C φ D ) (C_{\varphi\Box}, C_{\varphi D}) ( C φ □ , C φ D ) (left) and ( C φ □ , ∣ C φ X + i C φ X ~ ∣ ) (C_{\varphi\Box}, \lvert C_{\varphi X} + i C_{\varphi\widetilde X}\rvert) ( C φ □ , ∣ C φX + i C φ X ∣) for X = B , W , G X = B, W, G X = B , W , G (right). The left region is bounded by the weak-isospin singlet and triplet conditions of the J = 0 J = 0 J = 0 channel φ φ † → φ φ † \varphi\varphi^\dagger\to\varphi\varphi^\dagger φ φ † → φ φ † listed in Table 4 . The right regions follow from the J = 0 J = 0 J = 0 weak-isospin- and color-singlet condition of Eq. (31) , for one X X X at a time and with all other Wilson coefficients set to zero. In both panels, the extent of each region along each axis gives the marginalized bounds of Table 3 .
Table 4. Partial-wave unitarity conditions on correlated pairs of Wilson coefficients of the classes φ 4 D 2 \varphi^4D^2 φ 4 D 2 and ψ 2 φ 2 D \psi^2\varphi^2D ψ 2 φ 2 D in the single-flavor limit, for each scattering channel and total angular momentum J J J . The corresponding allowed regions are shown in Figure 1 (left) and 2. The factors s / Λ 2 s/\Lambda^2 s / Λ 2 are left implicit.
Consequently, imposing C φ □ = 0 C_{\varphi\Box} = 0 C φ □ = 0 from the outset would lead to an artificially stronger constraint on C φ D C_{\varphi D} C φ D .
The X 2 φ 2 X^2\varphi^2 X 2 φ 2 operators provide another example in which the coupled-channel structure is important. Coefficients multiplying a field-strength tensor and its dual enter the helicity amplitudes through the combinations C φ X ± i C φ X ~ C_{\varphi X} \pm iC_{\varphi\widetilde X} C φX ± i C φ X (and C φ W B ± i C φ W ~ B C_{\varphi WB} \pm iC_{\varphi\widetilde{W}B} C φ W B ± i C φ W B ), the sign depending on the helicities of the gauge bosons. The corresponding CP-even and CP-odd interactions must therefore be treated simultaneously when deriving their marginalized bounds. Figure 1 (right) shows the regions allowed by Eq. (31) in the planes (C_{\varphi\Box}, |C_{\varvarphi X} + i C_{\varphi\widetilde{X}}|) : the singlet channel correlates the φ 4 D 2 \varphi^4D^2 φ 4 D 2 and X 2 φ 2 X^2\varphi^2 X 2 φ 2 classes, and the bounds on C φ X C_{\varphi X} C φX , attained at C φ □ = 0 C_{\varphi\Box}=0 C φ □ = 0 , scale as 1 / n X 1/\sqrt{n_X} 1/ n X .
A separate remark concerns the X 3 X^3 X 3 operators. In this case the relevant unitarity bounds are computed from amplitudes containing the SM contribution and at most one insertion of a dimension-six interaction, thereby truncating the amplitude consistently at O ( 1 / Λ 2 ) O(1/\Lambda^2) O ( 1/ Λ 2 ) . Terms containing two insertions of dimension-six operators would enter at O ( 1 / Λ 4 ) O(1/\Lambda^4) O ( 1/ Λ 4 ) and could, in principle, modify the numerical constraint, especially when the corresponding gauge coupling is small. Including such contributions while neglecting dimension-eight operators, however, would not constitute a consistent EFT truncation. Indeed, positivity arguments imply that the presence of a dimension-six X 3 X^3 X 3 interaction is accompanied by dimension-eight X 4 X^4 X 4 operators, which would contribute at the same order in the EFT expansion [80 ] .
The remaining purely scalar operator, O φ = ( φ † φ ) 3 O_\varphi=(\varphi^\dagger\varphi)^3 O φ = ( φ † φ ) 3 , deserves special treatment. Its dimension-six 2 → 2 2\to2 2 → 2 contribution does not exhibit the energy growth relevant for the present analysis, so that the corresponding unitarity constraint arises from higher-multiplicity scattering. As discussed explicitly in § 2.1.1 , up to CP conjugation the relevant J = 0 J=0 J = 0 channels are
φ i φ j φ k ( † ) → φ ℓ φ m φ n ( † ) , φ i φ j ( † ) → φ k φ ℓ φ m ( † ) φ n † (34) \varphi_i\varphi_j\varphi_k^{(\dagger)} \to\varphi_\ell\varphi_m\varphi_n^{(\dagger)}, \qquad\varphi_i\varphi_j^{(\dagger)} \to\varphi_k\varphi_\ell\varphi_m^{(\dagger)}\varphi_n^{\dagger} \tag*{(34)} φ i φ j φ k ( † ) → φ ℓ φ m φ n ( † ) , φ i φ j ( † ) → φ k φ ℓ φ m ( † ) φ n † (34)
and the strongest of these, φ φ φ † → φ φ φ † \varphi\varphi\varphi^\dagger\to\varphi\varphi\varphi^\dagger φφ φ † → φφ φ † , gives
s Λ 2 ∣ C φ ∣ ≤ 32 π 3 3 . (35) \frac{s}{\Lambda^2}|C_\varphi| \le\frac{32\pi^3}{3}. \tag*{(35)} Λ 2 s ∣ C φ ∣ ≤ 3 32 π 3 . (35)
This provides a concrete example of why restricting the analysis to 2 → 2 2\to2 2 → 2 scattering would fail to probe the complete dimension-six SMEFT basis.
Finally, our bounds on the purely bosonic operators considered in Ref. [21 ] agree with those reported there, up to the factor-of-two difference associated with the normalization of the partial-wave unitarity condition. In Ref. [21 ] the convention ∣ a i → f J ∣ ≤ 2 |a^J_{i\to f}|\le2 ∣ a i → f J ∣ ≤ 2 was adopted, whereas throughout this work we use Eq. (4) .
Operators with a fermionic bilinear
We next consider dimension-six operators containing a single fermionic bilinear. Depending on their field content, their leading unitarity constraints arise either from higher-multiplicity scattering or from 2 → 2 2\to2 2 → 2 processes involving different allowed angular-momentum sectors. We therefore discuss separately the classes ψ 2 φ 3 \psi^2\varphi^3 ψ 2 φ 3 , ψ 2 X φ \psi^2X\varphi ψ 2 X φ , and ψ 2 φ 2 D \psi^2\varphi^2D ψ 2 φ 2 D .
ψ 2 φ 3 \psi^2\varphi^3 ψ 2 φ 3 operators. The operators of the class ψ 2 φ 3 \psi^2\varphi^3 ψ 2 φ 3 provide a second application of the generalized N → M N\to M N → M formalism, first discussed in Ref. [36 ] . In the Warsaw basis they are
O e φ = ( φ † φ ) ( ℓ ˉ p e r φ ) , O d φ = ( φ † φ ) ( q ˉ p d r φ ) , O u φ = ( φ † φ ) ( q ˉ p u r φ ~ ) (36) O_{e\varphi}=(\varphi^\dagger\varphi)(\bar{\ell}_p e_r\varphi), \qquad O_{d\varphi}=(\varphi^\dagger\varphi)(\bar{q}_p d_r\varphi), \qquad O_{u\varphi}=(\varphi^\dagger\varphi)(\bar{q}_p u_r\widetilde{\varphi}) \tag*{(36)} O e φ = ( φ † φ ) ( ℓ ˉ p e r φ ) , O d φ = ( φ † φ ) ( q ˉ p d r φ ) , O u φ = ( φ † φ ) ( q ˉ p u r φ ) (36)
together with their Hermitian conjugates. After electroweak symmetry breaking these interactions modify the SM Yukawa couplings, but in the unbroken high-energy theory they also generate genuine five-point contact amplitudes. Their leading unitarity constraints are therefore obtained from 2 → 3 2 \to3 2 → 3 and crossed 3 → 2 3 \to2 3 → 2 processes rather than from the conventional 2 → 2 2 \to2 2 → 2 partial-wave analysis.
To make the construction explicit, it is convenient to decompose the Higgs doublet into four real scalar components,
φ = 1 2 ( φ 1 + i φ 3 φ 2 + i φ 4 ) . (37) \varphi= \frac{1}{\sqrt{2}} \begin{pmatrix} \varphi_1 + i\varphi_3 \\ \varphi_2 + i\varphi_4 \end{pmatrix}. \tag*{(37)} φ = 2 1 ( φ 1 + i φ 3 φ 2 + i φ 4 ) . (37)
The J = 0 J=0 J = 0 coupled-channel analysis then contains all contact amplitudes of the schematic form
ϕ a ϕ b → ϕ c ψ i ψ j , ψ i ψ j → ϕ a ϕ b ϕ c , (38) \phi_a\phi_b \to\phi_c\psi_i\psi_j,\qquad\psi_i\psi_j \to\phi_a\phi_b\phi_c, \tag*{(38)} ϕ a ϕ b → ϕ c ψ i ψ j , ψ i ψ j → ϕ a ϕ b ϕ c , (38)
with a , b , c = 1 , … , 4 a,b,c=1,\ldots,4 a , b , c = 1 , … , 4 and with all allowed fermion species, gauge configurations, and flavor indices included.
As a representative example, consider the contribution of O e φ O_{e\varphi} O e φ . For a suitable choice of external Higgs components, one of the non-vanishing amplitudes is
A ( 1 ϕ 1 , 2 ϕ 1 → 3 ϕ 1 , 4 ν p + , 5 e r + ) = 3 2 [ C e φ ] p r ∗ Λ 2 [ 54 ] . (39) \mathcal{A}(1_{\phi_1},2_{\phi_1}\to3_{\phi_1},4_{\nu_p}^{+},5_{e_r}^{+})=\frac{3}{\sqrt{2}}\frac{[C_{e\varphi}]_{pr}^{*}}{\Lambda^2}[54]. \tag*{(39)} A ( 1 ϕ 1 , 2 ϕ 1 → 3 ϕ 1 , 4 ν p + , 5 e r + ) = 2 3 Λ 2 [ C e φ ] p r ∗ [ 54 ] . (39)
The relevant normalized J = 0 J=0 J = 0 basis element for this helicity configuration is
B 1 ϕ 1 , 2 ϕ 1 → 3 ϕ 1 , 4 ν p + , 5 e r + 0 = 32 6 π 2 s [ 54 ] . (40) \mathcal{B}^{0}_{1_{\phi_1},2_{\phi_1}\to3_{\phi_1},4_{\nu_p}^{+},5_{e_r}^{+}}=\frac{32\sqrt{6}\,\pi^2}{s}[54]. \tag*{(40)} B 1 ϕ 1 , 2 ϕ 1 → 3 ϕ 1 , 4 ν p + , 5 e r + 0 = s 32 6 π 2 [ 54 ] . (40)
Projecting Eq. (39) onto Eq. (40) , and including the appropriate 1 / 2 1/\sqrt{2} 1/ 2 normalization factor for the two identical scalars in the initial state, gives
a ϕ 1 ϕ 1 → ϕ 1 ν ˉ p e r + 0 = 3 s 64 2 π 2 Λ 2 [ C e φ ] p r ∗ . (41) a^{0}_{\phi_1\phi_1\to\phi_1\bar{\nu}_p e_r^{+}}=\frac{\sqrt{3}\,s}{64\sqrt{2}\,\pi^2\Lambda^2}[C_{e\varphi}]_{pr}^{*}. \tag*{(41)} a ϕ 1 ϕ 1 → ϕ 1 ν ˉ p e r + 0 = 64 2 π 2 Λ 2 3 s [ C e φ ] p r ∗ . (41)
This illustrates explicitly how a five-point contact interaction is projected onto a definite-angular-momentum partial wave without reducing the process to an effective 2 → 2 2 \to2 2 → 2 description.
The strongest constraint follows from the complete J = 0 J=0 J = 0 coupled-channel matrix containing all ψ 2 φ 3 \psi^2\varphi^3 ψ 2 φ 3 interactions. It can be written compactly as
s Λ 2 Tr [ 3 C u φ † C u φ + 3 C d φ † C d φ + C e φ † C e φ ] ≤ 32 π 2 3 . (42) \frac{s}{\Lambda^2}\sqrt{\operatorname{Tr}\left[3C_{u\varphi}^{\dagger}C_{u\varphi}+3C_{d\varphi}^{\dagger}C_{d\varphi}+C_{e\varphi}^{\dagger}C_{e\varphi}\right]}\leq\frac{32\pi^2}{\sqrt{3}}. \tag*{(42)} Λ 2 s Tr [ 3 C u φ † C u φ + 3 C d φ † C d φ + C e φ † C e φ ] ≤ 3 32 π 2 . (42)
The trace runs over flavor space, while the factors of three multiplying the quark coefficients originate from the color multiplicity. Equation (42) makes explicit that the different Yukawa-like SMEFT interactions participate in the same coupled-channel unitarity condition.
In the single-flavor limit, the marginalization over the remaining ψ 2 φ 3 \psi^{2}\varphi^{3} ψ 2 φ 3 coefficients gives
s Λ 2 ∣ C e φ ∣ ≤ 32 π 2 3 , s Λ 2 ∣ C u φ ∣ ≤ 32 π 2 3 , s Λ 2 ∣ C d φ ∣ ≤ 32 π 2 3 , (43) \frac{s}{\Lambda^{2}}|C_{e\varphi}| \leq\frac{32\pi^{2}}{\sqrt{3}}, \qquad\frac{s}{\Lambda^{2}}|C_{u\varphi}| \leq\frac{32\pi^{2}}{3}, \qquad\frac{s}{\Lambda^{2}}|C_{d\varphi}| \leq\frac{32\pi^{2}}{3}, \tag*{(43)} Λ 2 s ∣ C e φ ∣ ≤ 3 32 π 2 , Λ 2 s ∣ C u φ ∣ ≤ 3 32 π 2 , Λ 2 s ∣ C d φ ∣ ≤ 3 32 π 2 , (43)
in agreement with the results reported in Table 3 .
The class ψ 2 φ 3 \psi^{2}\varphi^{3} ψ 2 φ 3 therefore provides, together with O φ O_{\varphi} O φ , a direct illustration of why higher-multiplicity scattering is needed for a complete unitarity analysis of the dimension-six SMEFT. While O φ O_{\varphi} O φ requires six-point amplitudes and probes 3 → 3 3 \to3 3 → 3 and 2 → 4 2 \to4 2 → 4 channels, the Yukawa-like operators are instead controlled by five-point amplitudes and 2 → 3 2 \to3 2 → 3 scattering.
ψ 2 X φ \psi^{2}X\varphi ψ 2 X φ and ψ 2 φ 2 D \psi^{2}\varphi^{2}D ψ 2 φ 2 D operators. We now turn to the classes ψ 2 X φ \psi^{2}X\varphi ψ 2 X φ and ψ 2 φ 2 D \psi^{2}\varphi^{2}D ψ 2 φ 2 D , whose leading high-energy constraints are instead captured by 2 → 2 2 \to2 2 → 2 amplitudes with two fermionic and two bosonic external states. The allowed total angular momentum is determined by the helicity differences of the initial and final states. In particular, half-integer values of J J J arise when one fermion belongs to the initial state and the other to the final state, so that the lowest allowed partial wave is J = 1 / 2 J = 1/2 J = 1/2 . A representative set of such channels is
ψ R α , p φ j → ψ L i , β , r X I , ψ R α , p X I → ψ L i , β , r φ j † , (44) \psi_{R}^{\alpha,p}\varphi_{j} \to\psi_{L}^{i,\beta,r}X^{I}, \qquad\psi_{R}^{\alpha,p}X^{I} \to\psi_{L}^{i,\beta,r}\varphi_{j}^{\dagger}, \tag*{(44)} ψ R α , p φ j → ψ L i , β , r X I , ψ R α , p X I → ψ L i , β , r φ j † , (44)
ψ R α , p φ i ( † ) → ψ R β , r φ j ( † ) , ψ L k , α , p φ i ( † ) → ψ L ℓ , β , r φ j ( † ) . (45) \psi_{R}^{\alpha,p}\varphi_{i}^{(\dagger)} \to\psi_{R}^{\beta,r}\varphi_{j}^{(\dagger)}, \qquad\psi_{L}^{k,\alpha,p}\varphi_{i}^{(\dagger)} \to\psi_{L}^{\ell,\beta,r}\varphi_{j}^{(\dagger)}. \tag*{(45)} ψ R α , p φ i ( † ) → ψ R β , r φ j ( † ) , ψ L k , α , p φ i ( † ) → ψ L ℓ , β , r φ j ( † ) . (45)
Here p , r p,r p , r denote flavor indices, α , β \alpha,\beta α , β are S U ( 3 ) C \mathrm{SU}(3)_{C} SU ( 3 ) C fundamental or antifundamental indices whenever applicable, i , j , k , ℓ i,j,k,\ell i , j , k , ℓ denote S U ( 2 ) L \mathrm{SU}(2)_{L} SU ( 2 ) L indices, and L , R L,R L , R specify the fermion chirality. The first line contains chirality-changing amplitudes generated by operators of the class ψ 2 X φ \psi^{2}X\varphi ψ 2 X φ , while the second is representative of chirality-preserving interactions of the class ψ 2 φ 2 D \psi^{2}\varphi^{2}D ψ 2 φ 2 D . The remaining J = 1 / 2 J = 1/2 J = 1/2 channels are related to these by crossing and by discrete symmetries.
Integer angular momentum instead arises when the two fermions belong to the same side of the scattering process. The lowest such contribution is at J = 1 J = 1 J = 1 , with representative channels
ψ R α , p ψ ‾ L β , r → X I φ j † , ψ R α , p ψ ‾ R β , r → φ i φ j † , ψ L k , α , p ψ ‾ L ℓ , β , r → φ i φ j † . (46) \psi_{R}^{\alpha,p}\overline{\psi}_{L}^{\beta,r} \to X^{I}\varphi_{j}^{\dagger}, \qquad\psi_{R}^{\alpha,p}\overline{\psi}_{R}^{\beta,r} \to\varphi_{i}\varphi_{j}^{\dagger}, \qquad\psi_{L}^{k,\alpha,p}\overline{\psi}_{L}^{\ell,\beta,r} \to\varphi_{i}\varphi_{j}^{\dagger}. \tag*{(46)} ψ R α , p ψ L β , r → X I φ j † , ψ R α , p ψ R β , r → φ i φ j † , ψ L k , α , p ψ L ℓ , β , r → φ i φ j † . (46)
For these configurations, the minimum value J = 1 J = 1 J = 1 follows directly from the corresponding helicity differences.
We compute the coupled-channel unitarity bounds in both angular-momentum sectors. The results are summarized in Table 5 , while the operator O φ e O_{\varphi e} O φ e was discussed explicitly in Sec. 2.1.2. For operators of the class ψ 2 X φ \psi^{2}X\varphi ψ 2 X φ , the J = 1 / 2 J = 1/2 J = 1/2 constraints are always stronger than those obtained from J = 1 J = 1 J = 1 . For the class ψ 2 φ 2 D \psi^{2}\varphi^{2}D ψ 2 φ 2 D , instead, the situation is operator-dependent. Although one might naively expect the lowest allowed angular momentum to yield the strongest constraint, the enhancement associated with the multiplicity of internal gauge degrees of freedom can increase the largest eigenvalue of the J = 1 J = 1 J = 1 coupled-channel matrix and make the corresponding bound more restrictive. This occurs, in particular, for operators involving quark bilinears, where the color multiplicity can play an important role.
Table 5. Comparison of the partial-wave unitarity bounds on the modulus of the individual Wilson coefficients associated with the operator classes ψ 2 X φ \psi^2 X\varphi ψ 2 X φ and ψ 2 φ 2 D \psi^2\varphi^2D ψ 2 φ 2 D , obtained from the J = 1 / 2 J=1/2 J = 1/2 and J = 1 J=1 J = 1 sectors. The factors s / Λ 2 s/\Lambda^2 s / Λ 2 are left implicit.
For the operators in these classes, the correlations among Wilson coefficients do not modify the one-dimensional marginalized limits, which coincide with the strongest single-coefficient bounds of Table 5 . The correlations are nonetheless apparent in the two-dimensional allowed regions of the pairs ( O φ q ( 1 ) , O φ q ( 3 ) ) (O_{\varphi q}^{(1)}, O_{\varphi q}^{(3)}) ( O φq ( 1 ) , O φq ( 3 ) ) and ( O φ ℓ ( 1 ) , O φ ℓ ( 3 ) ) (O_{\varphi\ell}^{(1)}, O_{\varphi\ell}^{(3)}) ( O φ ℓ ( 1 ) , O φ ℓ ( 3 ) ) , whose coefficients enter common coupled-channel amplitudes. The corresponding allowed regions in the single-flavor limit are shown in Figure 2 .
Figure 2. Parameter space allowed by partial-wave unitarity for the coefficients ( C φ ℓ ( 1 ) , C φ ℓ ( 3 ) ) (C_{\varphi\ell}^{(1)}, C_{\varphi\ell}^{(3)}) ( C φ ℓ ( 1 ) , C φ ℓ ( 3 ) ) (blue) and ( C φ q ( 1 ) , C φ q ( 3 ) ) (C_{\varphi q}^{(1)}, C_{\varphi q}^{(3)}) ( C φq ( 1 ) , C φq ( 3 ) ) (red) in the single-flavor limit. Each region is the intersection of the J = 1 / 2 J=1/2 J = 1/2 and J = 1 J=1 J = 1 conditions listed in Table 4 and its extent along each axis gives the marginalized bounds of Table 3 .
Compared with previous SMEFT unitarity analyses [19 , 20 , 21 ] , the fermionic sector exhibits larger numerical differences than the purely bosonic one. These arise from the full coupled-channel treatment, in particular of the J = 1 J=1 J = 1 sectors, in which the color and weak-isospin multiplicities enhance the largest eigenvalue, while the marginalization plays no role. After accounting for the different normalization of the unitarity condition discussed in § 2.2.1 , the differences for the operators of the classes ψ 2 X φ \psi^2 X\varphi ψ 2 X φ and ψ 2 φ 2 D \psi^2\varphi^2D ψ 2 φ 2 D are operator-dependent and can reach a factor of two. The resulting bounds are generally more stringent than those previously reported, with the exception of O φ u d O_{\varphi ud} O φ u d , for which our limit is weaker by a factor of 2 \sqrt{2} 2 .
Four-fermion operators
We finally turn to four-fermion operators. In this case, the lowest angular momentum contributing to a given scattering channel is determined by the chiralities, and hence by the helicities, of the external fermions. For the channels in which the helicity differences of both the initial and final states vanish, the partial-wave expansion starts at J = 0 J=0 J = 0 . A representative set is
ψ R α , p ψ R β , r → ψ R γ , s ψ R δ , t , ψ L i , α , p ψ L j , β , r → ψ L k , γ , s ψ L ℓ , δ , t (47) \psi_R^{\alpha,p}\psi_R^{\beta,r}\to\psi_R^{\gamma,s}\psi_R^{\delta,t},\qquad\psi_L^{i,\alpha,p}\psi_L^{j,\beta,r}\to\psi_L^{k,\gamma,s}\psi_L^{\ell,\delta,t} \tag*{(47)} ψ R α , p ψ R β , r → ψ R γ , s ψ R δ , t , ψ L i , α , p ψ L j , β , r → ψ L k , γ , s ψ L ℓ , δ , t (47)
ψ R α , p ψ R β , r → ψ L i , γ , s ψ L j , δ , t , ψ R α , p ψ ‾ L i , β , r → ψ R γ , s ψ ‾ L j , δ , t . (48) \psi_R^{\alpha,p}\psi_R^{\beta,r}\to\psi_L^{i,\gamma,s}\psi_L^{j,\delta,t},\qquad\psi_R^{\alpha,p}\overline{\psi}_L^{i,\beta,r}\to\psi_R^{\gamma,s}\overline{\psi}_L^{j,\delta,t}. \tag*{(48)} ψ R α , p ψ R β , r → ψ L i , γ , s ψ L j , δ , t , ψ R α , p ψ L i , β , r → ψ R γ , s ψ L j , δ , t . (48)
Here p p p , r r r , s s s , t t t denote flavor indices, α \alpha α , β \beta β , γ \gamma γ , δ \delta δ denote color indices whenever applicable, and i i i , j j j , k k k , ℓ \ell ℓ are S U ( 2 ) L \mathrm{SU}(2)_L SU ( 2 ) L indices.
Other crossings of the same interactions instead start at J = 1 J=1 J = 1 . A representative set of such channels is
ψ R α , p ψ ‾ R β , r → ψ R γ , s ψ ‾ R δ , t , ψ L i , α , p ψ ‾ L j , β , r → ψ L k , γ , s ψ ‾ L ℓ , δ , t (49) \psi_R^{\alpha,p}\overline{\psi}_R^{\beta,r}\to\psi_R^{\gamma,s}\overline{\psi}_R^{\delta,t},\qquad\psi_L^{i,\alpha,p}\overline{\psi}_L^{j,\beta,r}\to\psi_L^{k,\gamma,s}\overline{\psi}_L^{\ell,\delta,t} \tag*{(49)} ψ R α , p ψ R β , r → ψ R γ , s ψ R δ , t , ψ L i , α , p ψ L j , β , r → ψ L k , γ , s ψ L ℓ , δ , t (49)
ψ R α , p ψ ‾ R β , r → ψ L i , γ , s ψ ‾ L j , δ , t , ψ R α , p ψ ‾ L j , β , r → ψ R γ , s ψ ‾ L j , δ , t . (50) \psi_R^{\alpha,p}\overline{\psi}_R^{\beta,r}\to\psi_L^{i,\gamma,s}\overline{\psi}_L^{j,\delta,t},\qquad\psi_R^{\alpha,p}\overline{\psi}_L^{j,\beta,r}\to\psi_R^{\gamma,s}\overline{\psi}_L^{j,\delta,t}. \tag*{(50)} ψ R α , p ψ R β , r → ψ L i , γ , s ψ L j , δ , t , ψ R α , p ψ L j , β , r → ψ R γ , s ψ L j , δ , t . (50)
The remaining channels are related to these by crossing and discrete symmetries and are included in the complete coupled-channel analysis.
To illustrate the relative importance of the two angular-momentum sectors, Table 6 compares the bounds obtained from J = 0 J=0 J = 0 and J = 1 J=1 J = 1 for one Wilson coefficient at a time. In most cases the J = 0 J=0 J = 0 partial wave gives the stronger constraint. There are, however, several exceptions, most notably among operators carrying singlet weak-isospin and color structures. As in the case of operators with a single fermionic bilinear, the multiplicity and coherent combination of internal gauge degrees of freedom can enhance a higher-J J J coupled-channel eigenvalue. Consequently, determining the strongest unitarity constraint requires examining all relevant angular-momentum sectors rather than stopping at the smallest kinematically allowed value of J J J .
Table 6. Comparison of the partial-wave unitarity bounds on the modulus of individual four-fermion Wilson coefficients obtained from the J = 0 J=0 J = 0 and J = 1 J=1 J = 1 sectors. For the purpose of this comparison, one coefficient at a time is kept non-vanishing. The factors s / Λ 2 s/\Lambda^2 s / Λ 2 are left implicit.
The four-fermion sector is characterized by a considerably richer pattern of correlations among Wilson coefficients than the operator classes discussed above. Accordingly, the one-coefficient-at-a-time limits in Table 6 should only be used to identify which angular-momentum sector is potentially dominant; they should not be confused with the marginalized bounds reported in Table 3 . The latter follow from the complete coupled-channel constraints, with all correlated coefficients allowed to vary simultaneously.
Representative examples of these correlations are shown in Figure 3 and the corresponding unitarity conditions are listed in Table 7 . In particular, operators with different weak-isospin or color contractions, such as ( O ℓ q ( 1 ) , O ℓ q ( 3 ) ) (O_{\ell q}^{(1)}, O_{\ell q}^{(3)}) ( O ℓ q ( 1 ) , O ℓ q ( 3 ) ) , ( O q q ( 1 ) , O q q ( 3 ) ) (O_{qq}^{(1)}, O_{qq}^{(3)}) ( O q q ( 1 ) , O q q ( 3 ) ) , ( O q u ( 1 ) , O q u ( 8 ) ) (O_{qu}^{(1)}, O_{qu}^{(8)}) ( O q u ( 1 ) , O q u ( 8 ) ) , and ( O u d ( 1 ) , O u d ( 8 ) ) (O_{ud}^{(1)}, O_{ud}^{(8)}) ( O u d ( 1 ) , O u d ( 8 ) ) , contribute simultaneously to common scattering channels and are therefore constrained jointly.
Table 7. Partial-wave unitarity conditions on correlated pairs of four-fermion Wilson coefficients in the single-flavor limit, for each scattering channel and total angular momentum J J J . The corresponding allowed regions are shown in Figure 3 and Figure 4 . Channels related by CP conjugation, and for O q u q d ( 1 , 8 ) O_{quqd}^{(1,8)} O q u q d ( 1 , 8 ) by u ↔ d u \leftrightarrow d u ↔ d , yield identical conditions and are not listed separately. The conditions for ( C q d ( 1 ) , C q d ( 8 ) ) (C_{qd}^{(1)}, C_{qd}^{(8)}) ( C q d ( 1 ) , C q d ( 8 ) ) follow from those for ( C q u ( 1 ) , C q u ( 8 ) ) (C_{qu}^{(1)}, C_{qu}^{(8)}) ( C q u ( 1 ) , C q u ( 8 ) ) with u → d u \to d u → d . The factors s / Λ 2 s/\Lambda^2 s / Λ 2 are left implicit.
Figure 3. Partial-wave unitarity constraints on correlated pairs of four-fermion Wilson coefficients in the single-flavor limit. Each shaded region is allowed by the scattering channel and total angular momentum J J J indicated in the legend, according to the conditions of Table 7 . The combined unitarity constraint is given by the intersection of all regions. Its extent along each axis reproduces the marginalized bounds of Table 3 . The allowed region for ( C q d ( 1 ) , C q d ( 8 ) ) (C_{qd}^{(1)}, C_{qd}^{(8)}) ( C q d ( 1 ) , C q d ( 8 ) ) is identical to that for ( C q u ( 1 ) , C q u ( 8 ) ) (C_{qu}^{(1)}, C_{qu}^{(8)}) ( C q u ( 1 ) , C q u ( 8 ) ) .
For non-Hermitian four-fermion operators, the relative complex phases of different Wilson coefficients provide an additional physical direction in parameter space. This is relevant, for example, for the pairs ( C q u q d ( 1 ) , C q u q d ( 8 ) ) (C_{quqd}^{(1)}, C_{quqd}^{(8)}) ( C q u q d ( 1 ) , C q u q d ( 8 ) ) and ( C ℓ e q u ( 1 ) , C ℓ e q u ( 3 ) ) (C_{\ell equ}^{(1)}, C_{\ell equ}^{(3)}) ( C ℓ e q u ( 1 ) , C ℓ e q u ( 3 ) ) . The corresponding unitarity regions depend on the relative phases arg ( C q u q d ( 1 ) / C q u q d ( 8 ) ) \arg(C_{quqd}^{(1)}/C_{quqd}^{(8)}) arg ( C q u q d ( 1 ) / C q u q d ( 8 ) ) and arg ( C ℓ e q u ( 1 ) / C ℓ e q u ( 3 ) ) \arg(C_{\ell equ}^{(1)}/C_{\ell equ}^{(3)}) arg ( C ℓ e q u ( 1 ) / C ℓ e q u ( 3 ) ) , as illustrated in Figure 4 and reported in Table 7 .
Figure 4. Regions allowed by partial-wave unitarity in the planes of the moduli (∣ C q u q d ( 1 ) ∣ |C_{quqd}^{(1)}| ∣ C q u q d ( 1 ) ∣ , ∣ C q u q d ( 8 ) ∣ |C_{quqd}^{(8)}| ∣ C q u q d ( 8 ) ∣ ) (left) and (∣ C ℓ e q u ( 1 ) ∣ |C_{\ell equ}^{(1)}| ∣ C ℓ e q u ( 1 ) ∣ , ∣ C ℓ e q u ( 3 ) ∣ |C_{\ell equ}^{(3)}| ∣ C ℓ e q u ( 3 ) ∣ ) (right), in the single-flavor limit, for the values of the relative complex phases arg ( C q u q d ( 1 ) / C q u q d ( 8 ) ) \arg(C_{quqd}^{(1)}/C_{quqd}^{(8)}) arg ( C q u q d ( 1 ) / C q u q d ( 8 ) ) and arg ( C ℓ e q u ( 1 ) / C ℓ e q u ( 3 ) ) \arg(C_{\ell equ}^{(1)}/C_{\ell equ}^{(3)}) arg ( C ℓ e q u ( 1 ) / C ℓ e q u ( 3 ) ) indicated in the legend. The corresponding conditions are listed in Table 7 .
Spinning sum rules
Perturbative unitarity constrains the magnitude of effective interactions, but by itself provides limited information on the nature of the heavy states responsible for them. Additional information can be obtained by combining unitarity with analyticity and crossing symmetry of the S S S matrix. Dispersion relations can then relate low-energy Wilson coefficients
to weighted sums over UV partial waves, thereby retaining information on the angular momentum of the states exchanged in the UV [11 , 12 , 13 , 14 ] . Under suitable assumptions, these so-called spinning sum rules correlate the sign of specific combinations of SMEFT Wilson coefficients with the spin of the heavy states that generate them.
In this section we formulate these constraints directly in the Warsaw basis, with particular emphasis on four-fermion interactions. We first review the relevant dispersion relations and identify the Wilson-coefficient combinations associated with scalar- and vectordominated UV dynamics. We then show how the resulting sign conditions can be imposed exactly in flavor space and, finally, discuss the assumptions under which this interpretation is valid and provide explicit UV examples illustrating how the corresponding inequalities can be evaded.
Spinning sum rules in the SMEFT
Following Refs. [11 , 12 , 13 , 14 ] , we consider the elastic amplitude A α β ( s , t ) A_{\alpha\beta}(s,t) A α β ( s , t ) for two fermionic states of opposite helicity. The labels α \alpha α and β \beta β collectively denote their internal quantum numbers, including flavor and gauge charges, while α ˉ \bar{\alpha} α ˉ denotes the crossed state, with conjugated quantum numbers and reversed helicity. The two crossed channels have different helicity structures: the α β \alpha\beta α β channel carries one unit of net helicity and therefore starts at J = 1 J=1 J = 1 , whereas the α ˉ β \bar{\alpha}\beta α ˉ β channel has vanishing net helicity and admits a J = 0 J=0 J = 0 partial wave.
Expanding the UV amplitudes in partial waves of definite total angular momentum, the forward dispersion relation can be written as2
lim s → 0 ∂ s A α β ( s , 0 ) + C α β ( s , ∞ ) = − lim s → 0 ∂ s A α ˉ β ˉ ( s , 0 ) − C α ˉ β ˉ ( s , ∞ ) = 8 ∫ s 0 ∞ d s s 2 [ − Im a α ˉ β ˉ 0 ( s ) + ∑ J = 1 ∞ ( 2 J + 1 ) ( Im a α β J ( s ) − Im a α ˉ β ˉ J ( s ) ) ] , (51) \lim_{s\to0}\partial_s A_{\alpha\beta}(s,0)+C_{\alpha\beta}^{(s,\infty)}=-\lim_{s\to0}\partial_s A_{\bar{\alpha}\bar{\beta}}(s,0)-C_{\bar{\alpha}\bar{\beta}}^{(s,\infty)}
=8\int_{s_0}^{\infty}\frac{\mathrm{d}s}{s^2}\left[-\operatorname{Im}a_{\bar{\alpha}\bar{\beta}}^0(s)+\sum_{J=1}^{\infty}(2J+1)\left(\operatorname{Im}a_{\alpha\beta}^J(s)-\operatorname{Im}a_{\bar{\alpha}\bar{\beta}}^J(s)\right)\right], \tag*{(51)} s → 0 lim ∂ s A α β ( s , 0 ) + C α β ( s , ∞ ) = − s → 0 lim ∂ s A α ˉ β ˉ ( s , 0 ) − C α ˉ β ˉ ( s , ∞ ) = 8 ∫ s 0 ∞ s 2 d s [ − Im a α ˉ β ˉ 0 ( s ) + J = 1 ∑ ∞ ( 2 J + 1 ) ( Im a α β J ( s ) − Im a α ˉ β ˉ J ( s ) ) ] , (51)
where s 0 s_0 s 0 denotes the characteristic mass-squared scale at which the UV degrees of freedom enter. The boundary contribution is
C α β ( s , ∞ ) = Res s = ∞ A α β ( s , 0 ) s 2 = − lim ∣ s ∣ → ∞ A α β ( s , 0 ) s . (52) C_{\alpha\beta}^{(s,\infty)}=\underset{s=\infty}{\operatorname{Res}}\frac{\mathcal{A}_{\alpha\beta}(s,0)}{s^2}=-\lim_{|s|\to\infty}\frac{\mathcal{A}_{\alpha\beta}(s,0)}{s}. \tag*{(52)} C α β ( s , ∞ ) = s = ∞ Res s 2 A α β ( s , 0 ) = − ∣ s ∣ → ∞ lim s A α β ( s , 0 ) . (52)
A second dispersion relation follows by applying the same contour argument to the coefficient of t t t in A α β ( 0 , t ) \mathcal{A}_{\alpha\beta}(0,t) A α β ( 0 , t ) . This gives a beyond-forward relation with an independent boundary contribution, which we denote by C α β ( t , ∞ ) C_{\alpha\beta}^{(t,\infty)} C α β ( t , ∞ ) . For a dimension-six interaction involving two opposite-helicity fermions, the local amplitude is proportional to the spinor structure [ 13 ] ⟨ 24 ⟩ [13]\langle24\rangle [ 13 ] ⟨ 24 ⟩ . As a consequence, an appropriate combination of the forward and beyond-forward dispersion relations has a vanishing low-energy side, up to boundary contributions, while all UV partial waves enter with non-negative weights.
If the relevant boundary terms vanish and the partial-wave expansion converges, positivity then requires the higher-spin contributions to vanish individually. In particular, the absorptive parts of the relevant channels with J ≥ 2 J \ge2 J ≥ 2 vanish, together with the crossed J = 1 J=1 J = 1 contribution. The remaining dispersion relation therefore involves only a scalar J = 0 J=0 J = 0 channel and a vector J = 1 J=1 J = 1 channel:
lim s → 0 ∂ s A ‾ α β ‾ ( s , 0 ) = − lim s → 0 ∂ s A α β ( s , 0 ) = 8 ∫ s 0 ∞ d s s 2 [ Im a α β ‾ 0 ( s ) − 3 Im a α β 1 ( s ) ] . (53) \lim_{s\to0}\partial_s\overline{\mathcal{A}}_{\alpha\overline{\beta}}(s,0)=-\lim_{s\to0}\partial_s\mathcal{A}_{\alpha\beta}(s,0)=8\int_{s_0}^{\infty}\frac{\mathrm{d}s}{s^2}\left[\operatorname{Im}a_{\alpha\overline{\beta}}^0(s)-3\operatorname{Im}a_{\alpha\beta}^1(s)\right]. \tag*{(53)} s → 0 lim ∂ s A α β ( s , 0 ) = − s → 0 lim ∂ s A α β ( s , 0 ) = 8 ∫ s 0 ∞ s 2 d s [ Im a α β 0 ( s ) − 3 Im a α β 1 ( s ) ] . (53)
By partial-wave unitarity, the absorptive parts appearing in Eq. (53) are non-negative. Consequently, if the UV dynamics is dominated by spin-0 states, the left-hand side is non-negative for any choice of the external-state quantum numbers. Conversely, dominance by spin-1 states implies a non-positive left-hand side. This is the central observation underlying the spinning sum rules.
For four-fermion SMEFT operators, the derivative of the low-energy amplitude is linear in the corresponding Wilson coefficients. Equation (53) can therefore be translated into sign conditions on specific combinations of their flavor-dependent entries. Since the basis adopted in Ref. [12] differs from the Warsaw basis, we collect in Table 8 the corresponding Warsaw-basis combinations. Denoting any such combination by b p r s t ψ b^\psi_{prst} b p r s t ψ , its contraction
Table 8. Combinations of four-fermion Wilson coefficients that, when contracted with α p α s ∗ β t β r ∗ \alpha_p\alpha_s^*\beta_t\beta_r^* α p α s ∗ β t β r ∗ , are positive for scalar-dominated UV dynamics and negative for vector-dominated UV dynamics, under the assumptions leading to Eq. (53) .
b p r s t ψ α p α s ∗ β r ∗ β t (54) b^\psi_{prst}\alpha_p\alpha_s^*\beta_r^*\beta_t \tag*{(54)} b p r s t ψ α p α s ∗ β r ∗ β t (54)
is positive for scalar-dominated UV dynamics and negative for vector-dominated UV dynamics, for arbitrary complex flavor vectors α \alpha α and β \beta β .
Spinning sum rules are not restricted in principle to four-fermion operators, but it is in this sector that they lead most directly to sign-definite constraints on Wilson coefficients. For purely bosonic operators of the class φ 4 D 2 \varphi^4 D^2 φ 4 D 2 , analogous relations can be derived only for restricted combinations of coefficients [13 ] . For operators of the classes X 2 φ 2 X^2\varphi^2 X 2 φ 2 and X 3 X^3 X 3 , the relevant elastic amplitudes vanish identically at dimension six, so that the corresponding Wilson coefficients are constrained by neither conventional nor spinning sum rules [11 , 13 ] . If the boundary terms vanish, however, the dispersion relations imply that the UV amplitudes mediating the elastic scattering of these states must vanish [13 ] .
The situation is also different for operators of the class φ 2 φ 2 D \varphi^2\varphi^2 D φ 2 φ 2 D . Since the corresponding fermion–Higgs amplitudes vanish at s = 0 s=0 s = 0 , the beyond-forward relation does not involve their Wilson coefficients, and no spinning sum rule can be constructed. The corresponding dispersion relations can still constrain the UV structure — in particular, vanishing boundary terms imply dominance by spin-1/2 states — but they do not provide the scalar/vector discrimination available for four-fermion interactions [13 ] . For this reason, in the remainder of this work we focus on the four-fermion sector, for which Eq. (3.4) can be turned into explicit inequalities in flavor space.
Application to flavor space
We now explain how the spinning sum rules derived above can be translated into explicit inequalities among flavor-dependent Wilson coefficients. For any of the four-fermion combinations b p r s t ψ b^\psi_{prst} b p r s t ψ collected in Table 8 , scalar dominance requires
b p r s t ψ α p α q ∗ β s ∗ β t ≥ 0 ∀ α , β , (55) b^\psi_{prst}\alpha_p\alpha_q^*\beta_s^*\beta_t \ge0 \qquad\forall\,\alpha,\beta, \tag*{(55)} b p r s t ψ α p α q ∗ β s ∗ β t ≥ 0 ∀ α , β , (55)
where α \alpha α and β \beta β are arbitrary complex vectors in flavor space. For vector-dominated UV dynamics the inequality is reversed. Since Eq. (55) is homogeneous in both flavor vectors, their overall normalization is irrelevant and they can be taken to have unit norm.
The condition in Eq. (55) is a biquadratic form in flavor space. For fixed β \beta β , we define the Hermitian matrix
[ M ψ ( β ) ] r p = b p r s t ψ β s ∗ β t . (56) \left[M^\psi(\beta)\right]_{rp}=b^\psi_{prst}\beta_s^*\beta_t. \tag*{(56)} [ M ψ ( β ) ] r p = b p r s t ψ β s ∗ β t . (56)
Equation (55) can then be written as
α † M ψ ( β ) α ≥ 0 ∀ α , β . (57) \alpha^\dagger M^\psi(\beta)\alpha\ge0 \qquad\forall\,\alpha,\beta. \tag*{(57)} α † M ψ ( β ) α ≥ 0 ∀ α , β . (57)
Scalar dominance is therefore equivalent to requiring M ψ ( β ) M^\psi(\beta) M ψ ( β ) to be positive semidefinite for every direction β \beta β in flavor space, or, equivalently,
λ i [ M ψ ( β ) ] ≥ 0 ∀ i , β , (58) \lambda_i\left[M^\psi(\beta)\right]\ge0 \qquad\forall\,i,\beta, \tag*{(58)} λ i [ M ψ ( β ) ] ≥ 0 ∀ i , β , (58)
where λ i \lambda_i λ i denote the eigenvalues of M ψ ( β ) M^\psi(\beta) M ψ ( β ) . For vector dominance, the same construction applies with M ψ ( β ) M^\psi(\beta) M ψ ( β ) required to be negative semidefinite.
For three generations, a convenient parametrization of a normalized flavor direction is
β = ( e i δ 1 β 1 e i δ 2 β 2 1 − β 1 2 − β 2 2 ) , β 1 2 + β 2 2 ≤ 1 , (59) \beta=
\begin{pmatrix}
e^{i\delta_1}\beta_1\\
e^{i\delta_2}\beta_2\\
\sqrt{1-\beta_1^2-\beta_2^2}
\end{pmatrix},
\qquad
\beta_1^2+\beta_2^2\le1, \tag*{(59)} β = e i δ 1 β 1 e i δ 2 β 2 1 − β 1 2 − β 2 2 , β 1 2 + β 2 2 ≤ 1 , (59)
where β 1 , 2 ≥ 0 \beta_{1,2}\ge0 β 1 , 2 ≥ 0 and δ 1 , 2 ∈ [ 0 , 2 π ) \delta_{1,2}\in[0,2\pi) δ 1 , 2 ∈ [ 0 , 2 π ) , and the third component has been taken real and non-negative by removing an overall phase. For all flavor structures encountered in our analysis, the non-trivial eigenvalues of M ψ ( β ) M^\psi(\beta) M ψ ( β ) can be reduced to functions of the form
f ± ( x ) = a + b x ± c 2 + d x + e x 2 , x ≡ β 1 2 + β 2 2 ∈ [ 0 , 1 ] , (60) f_\pm(x)=a+bx\pm\sqrt{c^2+dx+ex^2},
\qquad
x\equiv\beta_1^2+\beta_2^2\in[0,1], \tag*{(60)} f ± ( x ) = a + b x ± c 2 + d x + e x 2 , x ≡ β 1 2 + β 2 2 ∈ [ 0 , 1 ] , (60)
with a , b , c , d , e ∈ R a,b,c,d,e\in\mathbb{R} a , b , c , d , e ∈ R functions of the Wilson coefficients. The remaining eigenvalues, when present, are linear functions and are obtained as special cases of the same analysis. Since f + ( x ) ≥ f − ( x ) f_+(x)\ge f_-(x) f + ( x ) ≥ f − ( x ) , positive semidefiniteness is controlled by the lowest eigenvalue, and scalar dominance reduces to the condition
f − ( x ) ≥ 0 ∀ x ∈ [ 0 , 1 ] . (61) f_-(x)\ge0 \qquad\forall\,x\in[0,1]. \tag*{(61)} f − ( x ) ≥ 0 ∀ x ∈ [ 0 , 1 ] . (61)
A numerical scan over flavor directions would test Eq. (61) only on a finite set of points and would therefore provide, in general, only an approximate characterization of the allowed region. Instead, the requirement of positivity over the complete interval can be translated exactly into a finite set of algebraic conditions. For f − ( x ) f_-(x) f − ( x ) , assumed real for x ∈ [ 0 , 1 ] x \in[0,1] x ∈ [ 0 , 1 ] , one has f − ( x ) ≥ 0 f_{-}(x) \ge0 f − ( x ) ≥ 0 throughout the interval if and only if the system of inequalities
Δ ( a , b , c , d , e ) = a ≥ 0 ∧ a + b ≥ 0 ∧ a 2 − c 2 ≥ 0 ∧ ( a + b ) 2 − c 2 − d − e ≥ 0 ∧ [ e − b 2 ≥ 0 ∨ ( 2 a b − d ) ( 2 a b − d + 2 b 2 − 2 e ) ≥ 0 ∨ 4 ( b 2 − e ) ( a 2 − c 2 ) − ( 2 a b − d ) 2 ≥ 0 ] (62) \Delta(a,b,c,d,e) = a \ge0 \land a+b \ge0 \land a^{2}-c^{2} \ge0 \land(a+b)^{2}-c^{2}-d-e \ge0
\qquad\land\left[e-b^{2} \ge0 \lor(2ab-d)(2ab-d+2b^{2}-2e) \ge0 \right.
\qquad\left. \lor4(b^{2}-e)(a^{2}-c^{2})-(2ab-d)^{2} \ge0\right] \tag*{(62)} Δ ( a , b , c , d , e ) = a ≥ 0 ∧ a + b ≥ 0 ∧ a 2 − c 2 ≥ 0 ∧ ( a + b ) 2 − c 2 − d − e ≥ 0 ∧ [ e − b 2 ≥ 0 ∨ ( 2 ab − d ) ( 2 ab − d + 2 b 2 − 2 e ) ≥ 0 ∨ 4 ( b 2 − e ) ( a 2 − c 2 ) − ( 2 ab − d ) 2 ≥ 0 ] (62)
is true. A proof of Eq. (62) is given in App. § 7 .§ 1 . This converts the requirement that the sum rule hold for an infinite set of flavor directions into a finite system of inequalities among Wilson coefficients.
The vector-dominance conditions follow without repeating the analysis: one simply applies the scalar-dominance criterion to − M ψ ( β ) -M^{\psi}(\beta) − M ψ ( β ) , or equivalently to − b p r s t ψ -b^{\psi}_{prst} − b p r s t ψ . In this way, the two UV hypotheses define two different regions in the space of flavor-dependent Wilson coefficients.
In the single-flavor limit, the flavor vectors become trivial and the conditions above reduce directly to sign inequalities on the combinations listed in Table 8 . With several generations, instead, requiring the sum rule to hold for every possible flavor direction can generate non-trivial correlations among the independent coefficients and can be considerably more restrictive than testing only individual flavor entries. This observation will be particularly important in § 4 , where we apply the procedure to the MFV and U(2)⁵ flavor structures.
Limitations and UV examples
Unlike the perturbative unitarity bounds derived in § 2 , the spinning sum-rule inequalities do not follow from EFT consistency alone. Their interpretation in terms of the spin of the UV completion relies on a set of additional assumptions and can therefore fail whenever one of these assumptions is violated. In particular, the scalar- and vector-dominance conditions discussed above rest on three main ingredients [11 , 12 , 13 , 14 ] :
the effective interaction is generated at tree level;
the UV dynamics is dominated by states of a definite spin, so that the right-hand side of Eq. (53) has a definite sign;
the boundary contributions entering the relevant dispersion relations vanish.
Violating any of these conditions can move an otherwise consistent UV completion outside the scalar- or vector-dominated regions derived in § 3.2 . The three possibilities are physically distinct and we discuss them in turn.
Loop-generated interactions provide the first source of violation. The simple sign classification derived above assumes a tree-level UV origin of the dimension-six four-fermion interactions and need not hold when the matching is generated at loop level. Explicit examples have been discussed in Refs. [12 , 14 ] , and we do not reconsider them here.
A second possibility is a non-trivial interplay between UV states of different spin. If both spin-0 and spin-1 exchanges contribute with comparable strength, the two terms on the right-hand side of Eq. (53) compete, and the resulting low-energy coefficient need not be sign definite for arbitrary flavor directions. As a simple illustration, consider the schematic interactions
L ⊃ ( c A μ μ ˉ R γ μ e R + h . c . ) + ( y ϕ τ ˉ R μ R c + h . c . ) , (63) \mathcal{L} \supset(c A_\mu\bar{\mu}_{R}\gamma^\mu e_{R}+\mathrm{h.c.})+(y\phi\,\bar{\tau}_{R}\mu_{R}^{c}+\mathrm{h.c.}), \tag*{(63)} L ⊃ ( c A μ μ ˉ R γ μ e R + h.c. ) + ( y ϕ τ ˉ R μ R c + h.c. ) , (63)
where A μ A_\mu A μ is a heavy complex vector and ϕ \phi ϕ a heavy complex scalar. The purpose of this example is only to illustrate the loss of sign-definiteness when both scalar and vector mediators contribute. Both exchanges contribute to the same four-lepton operator O e e O_{ee} O ee .
Indeed, integrating out the heavy mediators gives the low-energy Lagrangian
L E F T ⊃ ∣ y ∣ 2 2 M ϕ 2 ( μ ˉ R γ μ μ R ) ( τ ˉ R γ μ τ R ) − ∣ c ∣ 2 M A 2 ( μ ˉ R γ μ e R ) ( e ˉ R γ μ μ R ) . (64) \mathcal{L}_{\mathrm{EFT}} \supset\frac{\lvert y\rvert^2}{2M_\phi^2}(\bar{\mu}_{R}\gamma^\mu\mu_{R})(\bar{\tau}_{R}\gamma_\mu\tau_{R})-\frac{\lvert c\rvert^2}{M_A^2}(\bar{\mu}_{R}\gamma^\mu e_{R})(\bar{e}_{R}\gamma_\mu\mu_{R}). \tag*{(64)} L EFT ⊃ 2 M ϕ 2 ∣ y ∣ 2 ( μ ˉ R γ μ μ R ) ( τ ˉ R γ μ τ R ) − M A 2 ∣ c ∣ 2 ( μ ˉ R γ μ e R ) ( e ˉ R γ μ μ R ) . (64)
Using the combination entering the spinning sum rule in Table 8 , we obtain
b α β e ≡ 1 Λ 2 [ C e e ] p r s t α p α s ∗ β t ∗ β t = ∣ y ∣ 2 8 M ϕ 2 ∣ α 3 ∗ β 3 + α 3 ∗ β 2 ∣ 2 − ∣ c ∣ 2 4 M A 2 ∣ α 1 ∗ β 2 + α 2 ∗ β 1 ∣ 2 , (65) b_{\alpha\beta}^{e} \equiv\frac{1}{\Lambda^2}[C_{ee}]_{prst}\alpha_p\alpha_s^*\beta_t^*\beta_t = \frac{\lvert y\rvert^2}{8M_\phi^2}\lvert\alpha_3^*\beta_3+\alpha_3^*\beta_2\rvert^2-\frac{\lvert c\rvert^2}{4M_A^2}\lvert\alpha_1^*\beta_2+\alpha_2^*\beta_1\rvert^2, \tag*{(65)} b α β e ≡ Λ 2 1 [ C ee ] p r s t α p α s ∗ β t ∗ β t = 8 M ϕ 2 ∣ y ∣ 2 ∣ α 3 ∗ β 3 + α 3 ∗ β 2 ∣ 2 − 4 M A 2 ∣ c ∣ 2 ∣ α 1 ∗ β 2 + α 2 ∗ β 1 ∣ 2 , (65)
where the Fierz-symmetric convention [ C e e ] p r s t = [ C e e ] s t p r = [ C e e ] p t s r = [ C e e ] s r p t [C_{ee}]_{prst}=[C_{ee}]_{stpr}=[C_{ee}]_{ptsr}=[C_{ee}]_{srpt} [ C ee ] p r s t = [ C ee ] s tp r = [ C ee ] pt sr = [ C ee ] sr pt is used. For non-vanishing y y y and c c c , different choices of the flavor vectors α \alpha α and β \beta β can make b α β e b_{\alpha\beta}^{e} b α β e either positive or negative. Hence, the resulting flavor tensor [ C e e ] p r s t [C_{ee}]_{prst} [ C ee ] p r s t is in general neither positive nor negative semidefinite. Explicitly, α = ( 0 , 1 , 0 ) \alpha=(0,1,0) α = ( 0 , 1 , 0 ) with β = ( 0 , 0 , 1 ) \beta=(0,0,1) β = ( 0 , 0 , 1 ) or β = ( 1 , 0 , 0 ) \beta=(1,0,0) β = ( 1 , 0 , 0 ) isolates the positive or the negative contribution, respectively.
The corresponding EFT point therefore lies outside both the purely scalar- and purely vector-dominated regions, despite arising from a perfectly consistent UV completion. If one mediator dominates parametrically, the expected sign-definite behavior is recovered approximately.
The third possibility is the presence of non-vanishing boundary terms. This is particularly relevant for tree-level vector exchange in the t t t channel, for which the boundary contribution can carry the entire low-energy amplitude. Consider, for example, a heavy neutral vector B μ B_\mu B μ with interactions
L ⊃ − [ y e ] p r e ˉ R p γ μ e R r B μ − [ y u ] p r u ˉ R p γ μ u R r B μ . (66) \mathcal{L} \supset-[y^e]_{pr}\bar{e}_{R}^{p}\gamma^\mu e_{R}^{r}B_\mu-[y^u]_{pr}\bar{u}_{R}^{p}\gamma^\mu u_{R}^{r}B_\mu. \tag*{(66)} L ⊃ − [ y e ] p r e ˉ R p γ μ e R r B μ − [ y u ] p r u ˉ R p γ μ u R r B μ . (66)
Integrating out B μ B_\mu B μ generates
L E F T ⊃ 1 M B 2 [ C e u ] p r s t ( e ˉ R p γ μ e R r ) ( u ˉ R s γ μ u R t ) , (67) \mathcal{L}_{\mathrm{EFT}} \supset\frac{1}{M_B^2}[C_{eu}]_{prst}(\bar{e}_{R}^{p}\gamma^\mu e_{R}^{r})(\bar{u}_{R}^{s}\gamma_\mu u_{R}^{t}), \tag*{(67)} L EFT ⊃ M B 2 1 [ C e u ] p r s t ( e ˉ R p γ μ e R r ) ( u ˉ R s γ μ u R t ) , (67)
with
[ C e u ] p r s t = − [ y e ] p r [ y u ] s t . (68) [C_{eu}]_{prst}=-[y^e]_{pr}[y^u]_{st}. \tag*{(68)} [ C e u ] p r s t = − [ y e ] p r [ y u ] s t . (68)
For the corresponding opposite-helicity elastic amplitude one finds
lim s → 0 ∂ s A α β ( s , 0 ) = − 2 M B 2 [ C e u ] p r s t α p α r ∗ β s ∗ β t . (69) \lim_{s\to0}\partial_s\mathcal{A}_{\alpha\beta}(s,0)=-\frac{2}{M_B^2}[C_{eu}]_{prst}\alpha_p\alpha_r^*\beta_s^*\beta_t. \tag*{(69)} s → 0 lim ∂ s A α β ( s , 0 ) = − M B 2 2 [ C e u ] p r s t α p α r ∗ β s ∗ β t . (69)
In this case the same contribution is entirely reproduced by the boundary term,
C α β ( s , ∞ ) = − lim ∣ s ∣ → ∞ A α β ( s , 0 ) s = 2 M B 2 [ C e u ] p r s t α p α r ∗ β s ∗ β t , (70) C_{\alpha\beta}^{(s,\infty)}=-\lim_{|s|\to\infty}\frac{\mathcal{A}_{\alpha\beta}(s,0)}{s}=\frac{2}{M_B^2}[C_{eu}]^{prst}\alpha_p\alpha_r^*\beta_s^*\beta_t, \tag*{(70)} C α β ( s , ∞ ) = − ∣ s ∣ → ∞ lim s A α β ( s , 0 ) = M B 2 2 [ C e u ] p r s t α p α r ∗ β s ∗ β t , (70)
while the absorptive s s s -channel partial wave vanishes. The sign of the low-energy Wilson coefficient is therefore no longer tied to the spin-1 contribution appearing on the right-hand side of Eq. (53) , because the assumption of vanishing boundary terms is not satisfied.
This behavior is characteristic of tree-level vector exchange in the t t t channel and occurs generically in the corresponding four-fermion UV completions. It should therefore not be interpreted as a violation of analyticity or unitarity, but rather as a failure of the assumptions required for the simple spinning sum-rule classification. Conversely, scalar tree-level completions with vanishing boundary contributions satisfy the corresponding scalar-dominance inequalities, while vector mediators exchanged in the s s s channel are classified according to their physical spin.
Non-vanishing boundary terms do not necessarily imply that the resulting Wilson coefficients lie outside the nominal scalar- or vector-dominated regions. For example, the same neutral vector can generate O e e O_{ee} O ee through both crossed and direct channels. In that case, part of the low-energy amplitude is carried by the boundary term while another part is associated with a physical vector partial wave. Depending on the flavor structure of the couplings, the resulting coefficient may still satisfy the vector-dominance inequalities. Thus, satisfying the corresponding spinning sum-rule inequalities is consistent with a simple spin-dominated completion under the assumptions above, but does not uniquely identify the UV dynamics.
These examples make clear that the spinning sum rules should be interpreted as UV discriminants rather than as model-independent consistency conditions. Their violation does not signal an inconsistent EFT, but rather that the assumptions underlying the simple scalar/vector classification are not realized. In this sense, the failure of the sign constraints can itself provide information on the structure of the UV completion.
Flavorful four-fermion operators
The results presented so far were obtained either in the single-flavor limit or, in the case of the spinning sum rules, in terms of generic flavor tensors. We now turn to the full three-generation structure of the four-fermion sector. Once all flavor indices are retained, the number of independent Wilson coefficients becomes very large, making a completely unrestricted coupled-channel analysis impractical. At the same time, flavor non-universality is a central feature of many well-motivated new-physics scenarios, in particular those with enhanced couplings to third-generation fermions. It is therefore useful to impose theoretically motivated flavor symmetries that reduce the number of independent structures while preserving non-trivial flavor information.
In this section we consider two standard hypotheses: MFV and an exact U ( 2 ) 5 \mathrm{U}(2)^5 U ( 2 ) 5 flavor symmetry. We apply both the perturbative unitarity bounds of § 2 and the spinning sum rules of § 3 directly in the resulting flavor space. A representative example will be discussed in the main text, while the complete set of bounds for all current–current four-fermion operators is collected in App. § 8 .
Flavor symmetries
We first specify the flavor structures retained in the analysis.
Minimal Flavor Violation
MFV assumes that the flavor structure of new physics is governed by the same sources of flavor-symmetry breaking that appear in the SM. In the limit of vanishing Yukawa couplings, the SM fermion sector is invariant under the global flavor symmetry
G M F V = U ( 3 ) 5 = U ( 3 ) q × U ( 3 ) u × U ( 3 ) d × U ( 3 ) ℓ × U ( 3 ) e , (71) G_{\mathrm{MFV}} = \mathrm{U}(3)^5 = \mathrm{U}(3)_q \times\mathrm{U}(3)_u \times\mathrm{U}(3)_d \times\mathrm{U}(3)_\ell\times\mathrm{U}(3)_e, \tag*{(71)} G MFV = U ( 3 ) 5 = U ( 3 ) q × U ( 3 ) u × U ( 3 ) d × U ( 3 ) ℓ × U ( 3 ) e , (71)
under which the fermion multiplets transform as
q → U q q , u → U u u , d → U d d , ℓ → U ℓ ℓ , e → U e e . (72) q \to U_q q,\qquad u \to U_u u,\qquad d \to U_d d,\qquad\ell\to U_\ell\ell,\qquad e \to U_e e. \tag*{(72)} q → U q q , u → U u u , d → U d d , ℓ → U ℓ ℓ , e → U e e . (72)
The Yukawa matrices are promoted to spurions transforming as
Y u → U q Y u U u † , Y d → U q Y d U d † , Y e → U ℓ Y e U e † , (73) Y_u \to U_q Y_u U_u^\dagger,\qquad Y_d \to U_q Y_d U_d^\dagger,\qquad Y_e \to U_\ell Y_e U_e^\dagger, \tag*{(73)} Y u → U q Y u U u † , Y d → U q Y d U d † , Y e → U ℓ Y e U e † , (73)
so that the SM Yukawa interactions are formally invariant under G M F V G_{\mathrm{MFV}} G MFV .
The MFV hypothesis requires the SMEFT Wilson coefficients to transform in such a way that all flavor-symmetry breaking can be expressed through appropriate insertions of these spurions. Consequently, the four flavor indices carried by a generic four-fermion coefficient are not independent, but can be organized into a finite set of U ( 3 ) 5 \mathrm{U}(3)^5 U ( 3 ) 5 -invariant tensors constructed from Kronecker deltas and Yukawa combinations.
In the present analysis, we neglect structures involving insertions of Y d Y_d Y d and Y e Y_e Y e , and retain the exact U ( 3 ) 5 \mathrm{U}(3)^5 U ( 3 ) 5 structures together with the leading flavor-breaking terms up to O ( Y u 2 ) \mathcal{O}(Y_u^2) O ( Y u 2 ) . This captures the dominant source of flavor non-universality while keeping the number of independent structures tractable. In the fermion mass basis, these flavor-breaking terms are controlled predominantly by the top Yukawa coupling through
Y u Y u † ≃ y t 2 ( 0 0 0 0 0 0 0 0 1 ) , Y u † Y u ≃ y t 2 ( 0 0 0 0 0 0 0 0 1 ) , (74) Y_uY_u^\dagger\simeq y_t^2 \begin{pmatrix} 0 & 0 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 1 \end{pmatrix},\qquad
Y_u^\dagger Y_u \simeq y_t^2 \begin{pmatrix} 0 & 0 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & 1 \end{pmatrix}, \tag*{(74)} Y u Y u † ≃ y t 2 0 0 0 0 0 0 0 0 1 , Y u † Y u ≃ y t 2 0 0 0 0 0 0 0 0 1 , (74)
up to the appropriate flavor-basis rotations. Since y t y_t y t is of order one, the truncation at O ( Y u 2 ) \mathcal{O}(Y_u^2) O ( Y u 2 ) should be regarded as a definite choice of flavor expansion rather than as a parametric expansion in a small quantity. Higher powers of Y u Y_u Y u would progressively resolve the third generation from the first two and would effectively approach a U ( 2 ) q × U ( 2 ) u \mathrm{U}(2)_q \times\mathrm{U}(2)_u U ( 2 ) q × U ( 2 ) u flavor structure [6 , 7 ] .
U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry
As a second benchmark, we consider the flavor symmetry
G F ≡ U ( 2 ) 5 = U ( 2 ) q × U ( 2 ) u × U ( 2 ) d × U ( 2 ) ℓ × U ( 2 ) e , (75) G_F \equiv U(2)^5 = U(2)_q \times U(2)_u \times U(2)_d \times U(2)_\ell\times U(2)_e, \tag*{(75)} G F ≡ U ( 2 ) 5 = U ( 2 ) q × U ( 2 ) u × U ( 2 ) d × U ( 2 ) ℓ × U ( 2 ) e , (75)
under which the first two generations of each fermion species transform as a doublet, while the third generation is a singlet. Accordingly, we decompose
ψ = ( ψ a , ψ 3 ) , a = 1 , 2 , ψ = q , u , d , ℓ , e . (76) \psi= (\psi_a, \psi_3), \qquad a = 1, 2, \qquad\psi= q, u, d, \ell, e. \tag*{(76)} ψ = ( ψ a , ψ 3 ) , a = 1 , 2 , ψ = q , u , d , ℓ , e . (76)
This symmetry pattern is particularly well suited to new-physics scenarios in which the first two generations are approximately universal while the third generation is allowed to couple differently, see for instance Refs. [41 , 42 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 ] . The independent flavor structures of dimension-six SMEFT operators under U ( 2 ) 5 U(2)^5 U ( 2 ) 5 have been systematically classified in Refs. [6 , 7 ] . There is no unique or universally preferred choice of spurions describing the breaking of U ( 2 ) 5 U(2)^5 U ( 2 ) 5 . We therefore restrict our analysis to the exact-symmetry limit, without introducing symmetry-breaking spurion insertions. Our results are consequently independent of the specific realization of U ( 2 ) 5 U(2)^5 U ( 2 ) 5 breaking.
The resulting Wilson coefficients are therefore organized according to whether the fermion bilinears involve fields belonging to the light U ( 2 ) U(2) U ( 2 ) doublets or to the third-generation singlets. For example, a coefficient involving two different fermion species can contain independent structures corresponding schematically to
( ψ ˉ a ψ a ) ( χ ˉ b χ b ) , ( ψ ˉ 3 ψ 3 ) ( χ ˉ a χ a ) , ( ψ ˉ a ψ a ) ( χ ˉ 3 χ 3 ) , ( ψ ˉ 3 ψ 3 ) ( χ ˉ 3 χ 3 ) , (77) (\bar{\psi}_a\psi_a)(\bar{\chi}_b\chi_b), \qquad(\bar{\psi}_3\psi_3)(\bar{\chi}_a\chi_a), \qquad(\bar{\psi}_a\psi_a)(\bar{\chi}_3\chi_3), \qquad(\bar{\psi}_3\psi_3)(\bar{\chi}_3\chi_3), \tag*{(77)} ( ψ ˉ a ψ a ) ( χ ˉ b χ b ) , ( ψ ˉ 3 ψ 3 ) ( χ ˉ a χ a ) , ( ψ ˉ a ψ a ) ( χ ˉ 3 χ 3 ) , ( ψ ˉ 3 ψ 3 ) ( χ ˉ 3 χ 3 ) , (77)
with additional exchange structures whenever allowed by the fermion content of the operator. The precise decomposition is operator dependent and will be given where needed below, while the complete set of flavor structures is collected in App. § 8 .
Theoretical bounds in flavor space
We now apply the perturbative unitarity bounds derived in § 2 and the spinning sum rules of § 3 to the multiflavor structures introduced above. For each four-fermion operator, the MFV or U ( 2 ) 5 U(2)^5 U ( 2 ) 5 decomposition expresses the full flavor tensor in terms of a reduced set of independent coefficients. These coefficients are then inserted into the complete coupled-channel scattering matrices, retaining all flavor configurations allowed by the chosen symmetry hypothesis.
The unitarity constraints are imposed on the eigenvalues of the resulting partial-wave matrices exactly as in the single-flavor analysis. Since several independent flavor structures can contribute to the same scattering channels, the corresponding coefficients are treated simultaneously and the one-dimensional limits are obtained by marginalizing over the remaining coefficients. The resulting bounds should therefore be interpreted as projections of the full multiflavor allowed region, rather than as one-coefficient-at-a-time constraints.
The spinning sum rules are implemented in the same flavor space. For each relevant combination b p r s t ψ b^\psi_{prst} b p r s t ψ listed in Table 8 , the MFV or U ( 2 ) 5 U(2)^5 U ( 2 ) 5 decomposition is substituted into the biquadratic form
b p r s t ψ α p α r ∗ β s ∗ β t , (78) b^{\psi}_{prst}\alpha_p\alpha_r^*\beta_s^*\beta_t, \tag*{(78)} b p r s t ψ α p α r ∗ β s ∗ β t , (78)
and the scalar- or vector-dominance conditions are imposed for arbitrary flavor directions α \alpha α and β \beta β , following the procedure described in § 3.2 . The resulting inequalities therefore constrain correlated combinations of the independent flavor coefficients, rather than individual tensor entries.
To illustrate the resulting constraints, we focus in the following on the semileptonic operators O ℓ q ( 1 ) O_{\ell q}^{(1)} O ℓ q ( 1 ) and O ℓ q ( 3 ) O_{\ell q}^{(3)} O ℓ q ( 3 ) , while the complete results are collected in App. § 8 .
Representative example: O ℓ q ( 1 ) O_{\ell q}^{(1)} O ℓ q ( 1 ) and O ℓ q ( 3 ) O_{\ell q}^{(3)} O ℓ q ( 3 )
As a representative example of the multiflavor analysis, we consider the semileptonic operators
O ℓ q ( 1 ) = ( ℓ ˉ p γ μ ℓ r ) ( q ˉ s γ μ q t ) , O ℓ q ( 3 ) = ( ℓ ˉ p γ μ σ I ℓ r ) ( q ˉ s γ μ σ I q t ) . (79) O_{\ell q}^{(1)} = (\bar{\ell}_p\gamma_\mu\ell_r)(\bar{q}_s\gamma^\mu q_t), \qquad O_{\ell q}^{(3)} = (\bar{\ell}_p\gamma_\mu\sigma^I\ell_r)(\bar{q}_s\gamma^\mu\sigma^I q_t). \tag*{(79)} O ℓ q ( 1 ) = ( ℓ ˉ p γ μ ℓ r ) ( q ˉ s γ μ q t ) , O ℓ q ( 3 ) = ( ℓ ˉ p γ μ σ I ℓ r ) ( q ˉ s γ μ σ I q t ) . (79)
These operators provide a useful benchmark because their singlet and triplet Wilson coefficients contribute simultaneously to common scattering channels and are therefore correlated by both perturbative unitarity and the spinning sum rules.
Under MFV, retaining the exact flavor structures and terms up to O ( Y u 2 ) O(Y_u^2) O ( Y u 2 ) , each coefficient contains two independent structures:
[ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p r δ s t + C ℓ q ( i , 2 ) δ p r ( Y u Y u † ) s t , i = 1 , 3. (80) [C_{\ell q}^{(i)}]_{prst} = C_{\ell q}^{(i,1)}\delta_{pr}\delta_{st} + C_{\ell q}^{(i,2)}\delta_{pr}(Y_uY_u^\dagger)_{st}, \qquad i=1,3. \tag*{(80)} [ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p r δ s t + C ℓ q ( i , 2 ) δ p r ( Y u Y u † ) s t , i = 1 , 3. (80)
The first term is flavor universal, whereas the second introduces the leading non-universality associated with the up-type Yukawa spurion.
In the exact U ( 2 ) 5 \mathrm{U}(2)^5 U ( 2 ) 5 limit, four independent structures arise for each of the two operators:
[ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p a δ a r δ s b δ b t + C ℓ q ( i , 2 ) δ p 3 δ 3 r δ s a δ a t + C ℓ q ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ q ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 3 , (81) \begin{aligned}
[C_{\ell q}^{(i)}]_{prst} ={}& C_{\ell q}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{\ell q}^{(i,2)}\delta_{p3}\delta_{3r}\delta_{sa}\delta_{at} \\
&+ C_{\ell q}^{(i,3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{\ell q}^{(i,4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t}, \qquad i=1,3, \tag*{(81)}
\end{aligned} [ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p a δ a r δ s b δ b t + C ℓ q ( i , 2 ) δ p 3 δ 3 r δ s a δ a t + C ℓ q ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ q ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 3 , (81)
where a , b = 1 , 2 a,b=1,2 a , b = 1 , 2 are summed U ( 2 ) \mathrm{U}(2) U ( 2 ) indices. The four terms correspond, respectively, to light–light, third–light, light–third, and third–third fermion bilinears.
Substituting Eqs. (80) and (81) into the complete coupled-channel matrices gives the marginalized unitarity bounds reported in Table 9 . In the numerical MFV analysis, here and in App. § 8 , we set y t = 1 y_t=1 y t = 1 in Eq. (74) , so that the bounds on the coefficients of the Y u Y_u Y u -dependent structures, such as C ℓ q ( 1 , 2 ) C_{\ell q}^{(1,2)} C ℓ q ( 1 , 2 ) and C ℓ q ( 3 , 2 ) C_{\ell q}^{(3,2)} C ℓ q ( 3 , 2 ) , apply to y t 2 y_t^2 y t 2 times these coefficients.
Table 9. Marginalized bounds on the independent coefficients of O ℓ q ( 1 ) O_{\ell q}^{(1)} O ℓ q ( 1 ) and O ℓ q ( 3 ) O_{\ell q}^{(3)} O ℓ q ( 3 ) under MFV and U ( 2 ) 5 \mathrm{U}(2)^5 U ( 2 ) 5 . The intervals [ a , b ] [a,b] [ a , b ] correspond to ( s / Λ 2 ) C ℓ q ( i , j ) ∈ [ a , b ] (s/\Lambda^2)C_{\ell q}^{(i,j)} \in[a,b] ( s / Λ 2 ) C ℓ q ( i , j ) ∈ [ a , b ] . The first column of bounds follows from perturbative unitarity alone, whereas the last two additionally impose the spinning sum rules under scalar or vector dominance, respectively.
For comparison, we also show the bounds obtained after imposing the spinning sum rules under scalar or vector dominance. As throughout this work, these latter regions should be interpreted under the assumptions discussed in § 3.3 .
The correlations among the independent coefficients are more transparent in the two-dimensional sections of the allowed regions shown in Figure 5 and Figure 6 . Each panel is a two-dimensional slice of the full multidimensional region, obtained by setting all independent coefficients other than the two displayed to zero. This should be distinguished from the one-dimensional intervals in Table 9 , which are obtained by marginalizing over the remaining coefficients.
Figure 5. Partial-wave unitarity bounds and spinning sum-rule constraints for O ℓ q ( 1 , 3 ) O_{\ell q}^{(1,3)} O ℓ q ( 1 , 3 ) under MFV, where [ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p r δ s t + C ℓ q ( i , 2 ) δ p r ( Y u Y u † ) s t [C_{\ell q}^{(i)}]_{prst}=C_{\ell q}^{(i,1)}\delta_{pr}\delta_{st}+C_{\ell q}^{(i,2)}\delta_{pr}(Y_uY_u^\dagger)_{st} [ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p r δ s t + C ℓ q ( i , 2 ) δ p r ( Y u Y u † ) s t , i = 1 , 3 i=1,3 i = 1 , 3 . Gray denotes the region allowed by perturbative unitarity, while blue and orange correspond to the additional spinning sum-rule constraints under scalar- and vector-dominated UV dynamics, respectively. The panels are two-dimensional slices of the full allowed region, obtained by setting all other independent coefficients to zero.
Figure 6. Partial-wave unitarity bounds and spinning sum-rule constraints for O ℓ q ( 1 , 3 ) O_{\ell q}^{(1,3)} O ℓ q ( 1 , 3 ) under U(2)5 ^{5} 5 flavor symmetry, where [ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p a δ a r δ s b δ b t + C ℓ q ( i , 2 ) δ p 3 δ 3 r δ s a δ a t + C ℓ q ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ q ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{\ell q}^{(i)}]_{prst} = C_{\ell q}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{\ell q}^{(i,2)}\delta_{p3}\delta_{3r}\delta_{sa}\delta_{at} + C_{\ell q}^{(i,3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{\ell q}^{(i,4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ℓ q ( i ) ] p r s t = C ℓ q ( i , 1 ) δ p a δ a r δ s b δ b t + C ℓ q ( i , 2 ) δ p 3 δ 3 r δ s a δ a t + C ℓ q ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ q ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 3 i=1,3 i = 1 , 3 . Gray denotes the region allowed by perturbative unitarity, while blue and orange correspond to the additional spinning sum-rule constraints under scalar- and vector-dominated UV dynamics, respectively. The panels are two-dimensional slices of the full allowed region, obtained by setting all other independent coefficients to zero.
The comparison illustrates the complementary information carried by perturbative unitarity and the spinning sum rules. Unitarity bounds the overall size of the interactions and generally produces regions symmetric under an overall sign reversal, whereas the sum rules select particular directions in coefficient space according to the assumed spin of the UV dynamics. In some two-dimensional slices, the flavor-space sum-rule conditions are sufficiently restrictive that setting the remaining coefficients to zero forces one or both displayed coefficients onto lower-dimensional loci, so that the allowed region collapses to a line or even to a point. In the multiflavor case, this behavior follows from the requirement that the positivity or negativity condition hold for every flavor direction, which correlates coefficients that would appear independent in a single-flavor analysis.
General features
The example above illustrates several features that are common to the multiflavor analysis of four-fermion operators. In particular:
Imposing a flavor symmetry does not reduce the problem to a set of independent single-flavor constraints. Different flavor structures generally contribute to the same coupled-channel amplitudes and are therefore constrained simultaneously. As a result, both the perturbative unitarity bounds and the spinning sum-rule conditions generate non-trivial correlations among the independent flavor coefficients.
The dimensionality of the scattering problem can remain sizable even after the flavor reduction. For instance, in the U(2)5 ^5 5 analysis of the q q → q q qq \to qq q q → q q sector, the coupled-channel matrix has dimension 324 × 324 324 \times324 324 × 324 , reflecting the combined multiplicity of color, weak isospin, and flavor degrees of freedom. This provides a representative example of the computational complexity of the full three-generation analysis.
Spinning sum rules become increasingly restrictive once their definiteness conditions are imposed over the complete flavor space. As seen in the O ℓ q ( 1 , 3 ) O_{\ell q}^{(1,3)} O ℓ q ( 1 , 3 ) example, correlations among the independent flavor coefficients can force some two-dimensional slices of the allowed region to collapse onto lower-dimensional loci. In the multiflavor case, this behavior arises from requiring the sum-rule inequalities to hold for arbitrary flavor directions.
Finally, the spinning sum rules act as discriminants of simple UV structures rather than as model-independent consistency conditions. Regions outside both the scalar- and vector-dominated domains would indicate that the assumptions underlying the simple spin-dominance classification are not realized, pointing to a more structured UV origin involving, for instance, non-vanishing boundary terms, mixed-spin dynamics, or loop-generated matching.
The same procedure has been applied systematically to all remaining current–current
four-fermion operators under both MFV and U(2)5 ^{5} 5 . For each operator, we determine the independent flavor structures, construct the corresponding coupled-channel matrices, derive the marginalized perturbative unitarity bounds, and impose the scalar- and vector-dominance sum-rule conditions whenever applicable.
Since the complete set of flavor decompositions, bounds, and allowed regions is extensive, we do not reproduce it in the main text. The full results are collected in App. B, where the bounds on each independent coefficient and, where more than one coefficient is present, the corresponding two-dimensional slices are reported for all current–current four-fermion operators. An accompanying Mathematica notebook [37 ] provides a machine-readable implementation of the flavor decompositions and theoretical constraints, together with the analytic and numerical results used to obtain the bounds reported in App. B.
Matching to UV completions
To illustrate more explicitly both the diagnostic power and the limitations of the theoretical constraints derived above, we now consider representative single-field UV completions that generate the semileptonic operators O ℓ q ( 1 ) O_{\ell q}^{(1)} O ℓ q ( 1 ) and O ℓ q ( 3 ) O_{\ell q}^{(3)} O ℓ q ( 3 ) at tree level. The complete tree-level dictionary of renormalizable heavy mediators generating dimension-six SMEFT operators was classified in Ref. [81 ] . Restricting to mediators that generate the left-handed semileptonic structures relevant here leads to the scalar leptoquarks ω 1 \omega_{1} ω 1 and ζ \zeta ζ , the vector leptoquarks U 2 U_{2} U 2 and X \mathcal{X} X , and the color-singlet vectors B \mathcal{B} B and W W W . Their gauge quantum numbers, relevant interactions, and tree-level matching onto [ O ℓ q ( 1 ) ] 3333 \left[O_{\ell q}^{(1)}\right]_{3333} [ O ℓ q ( 1 ) ] 3333 and [ O ℓ q ( 3 ) ] 3333 \left[O_{\ell q}^{(3)}\right]_{3333} [ O ℓ q ( 3 ) ] 3333 are summarized in Table 10 .
Table 10. Single-field tree-level UV completions generating [ O ℓ q ( 1 ) ] 3333 [O_{\ell q}^{(1)}]_{3333} [ O ℓ q ( 1 ) ] 3333 and [ O ℓ q ( 3 ) ] 3333 [O_{\ell q}^{(3)}]_{3333} [ O ℓ q ( 3 ) ] 3333 . We report the SM gauge representation, spin, relevant interaction, and tree-level matching onto the Warsaw-basis Wilson coefficients. The mediator classification follows the tree-level SMEFT dictionary of Ref. [81 ] .
For definiteness, we consider the exact U(2)5 ^{5} 5 limit and assume that the new mediators couple predominantly to third-generation fermions. The matching then populates only the components [ C ℓ q ( 1 ) ] 3333 \left[C_{\ell q}^{(1)}\right]_{3333} [ C ℓ q ( 1 ) ] 3333 and [ C ℓ q ( 3 ) ] 3333 \left[C_{\ell q}^{(3)}\right]_{3333} [ C ℓ q ( 3 ) ] 3333 , corresponding to the coefficients C ℓ q ( 1 , 4 ) C_{\ell q}^{(1,4)} C ℓ q ( 1 , 4 ) and C ℓ q ( 3 , 4 ) C_{\ell q}^{(3,4)} C ℓ q ( 3 , 4 ) in the decomposition of Eq. (81) . This choice allows the UV completions to be compared directly with the two-dimensional slice of Figure 6 spanned by C ℓ q ( 1 , 4 ) C_{\ell q}^{(1,4)} C ℓ q ( 1 , 4 ) and C ℓ q ( 3 , 4 ) C_{\ell q}^{(3,4)} C ℓ q ( 3 , 4 ) .
From the viewpoint of the dispersion relations, the six mediators fall into two classes. The scalar leptoquarks ω 1 \omega_{1} ω 1 and ζ \zeta ζ are exchanged in the s s s channel of ℓ q → ℓ q \ell q \to\ell q ℓ q → ℓ q , and the vector leptoquarks U 2 U_{2} U 2 and X \mathcal{X} X in the s s s channel of ℓ q ˉ → ℓ q ˉ \ell\bar q \to\ell\bar q ℓ q ˉ → ℓ q ˉ , so that, in the absence of boundary terms, the sum-rule sign tracks their spin. The color-singlet vectors B \mathcal{B} B and W W W instead contribute through t t t -channel exchange, for which non-vanishing boundary terms invalidate the simple scalar/vector classification [11 , 12 , 13 , 14 ] .
This can be seen directly from the two combinations entering the spinning sum rules,
[ C ℓ q ( 1 ) ] 3333 + [ C ℓ q ( 3 ) ] 3333 , [ C ℓ q ( 1 ) ] 3333 − [ C ℓ q ( 3 ) ] 3333 . (82) \left[C_{\ell q}^{(1)}\right]_{3333}+\left[C_{\ell q}^{(3)}\right]_{3333}, \qquad\left[C_{\ell q}^{(1)}\right]_{3333}-\left[C_{\ell q}^{(3)}\right]_{3333}. \tag*{(82)} [ C ℓ q ( 1 ) ] 3333 + [ C ℓ q ( 3 ) ] 3333 , [ C ℓ q ( 1 ) ] 3333 − [ C ℓ q ( 3 ) ] 3333 . (82)
For the scalar leptoquarks ω 1 \omega_{1} ω 1 and ζ \zeta ζ , both combinations are non-negative, whereas for the vector leptoquarks U 2 U_{2} U 2 and X \mathcal{X} X they are both non-positive. The scalar and vector leptoquarks are therefore placed precisely in the scalar- and vector-dominated regions predicted by the spinning sum rules.
The two t t t -channel vectors exhibit a qualitatively different behavior. For the singlet vector B \mathcal{B} B , [ C ℓ q ( 3 ) ] 3333 = 0 \left[C_{\ell q}^{(3)}\right]_{3333}=0 [ C ℓ q ( 3 ) ] 3333 = 0 , so that the two sum-rule combinations coincide. Depending on the sign of [ g b ℓ ] 33 [ g b q ] 33 [g_b^\ell]_{33}[g_b^q]_{33} [ g b ℓ ] 33 [ g b q ] 33 , the matched point can therefore lie either in the nominal vector-dominated region or in the scalar-dominated one. In the latter case, a spin-1 mediator would be incorrectly classified as scalar if the boundary contribution were neglected.
For the triplet vector W \mathcal{W} W , instead, [ C ℓ q ( 1 ) ] 3333 = 0 [C_{\ell q}^{(1)}]_{3333}=0 [ C ℓ q ( 1 ) ] 3333 = 0 . The two combinations C ℓ q ( 1 ) ± C ℓ q ( 3 ) C_{\ell q}^{(1)} \pm C_{\ell q}^{(3)} C ℓ q ( 1 ) ± C ℓ q ( 3 ) consequently have opposite signs for any non-vanishing coupling product. The corresponding UV trajectory therefore lies outside both the scalar- and vector-dominated regions, except at the origin. This does not signal an inconsistency of the UV theory: it is precisely the behavior expected when the non-vanishing boundary term associated with vector t t t -channel exchange invalidates the simple spin classification.
The perturbative unitarity constraints have a different status. Substituting the matching relations of Table 10 into the analytic bounds derived above gives the conditions listed in Table 11 . They constrain combinations of the form
Table 11. Perturbative unitarity bounds and spinning sum-rule conditions for the single-field UV completions of Table 10 , after substituting the corresponding tree-level matching relations. For the s s s -channel leptoquarks, the sum-rule conditions reduce to manifestly satisfied inequalities with the sign dictated by the mediator spin. For the t t t -channel vectors, the entries show the conditions that would follow from the naive spin classification if boundary contributions were neglected; the latter are non-vanishing and invalidate this interpretation.
∣ g U V ∣ 2 s ≲ O ( 4 π ) M U V 2 , (83) \lvert g_{\mathrm{UV}}\rvert^2 s \lesssim\mathcal{O}(4\pi) M_{\mathrm{UV}}^2, \tag*{(83)} ∣ g UV ∣ 2 s ≲ O ( 4 π ) M UV 2 , (83)
with representation-dependent numerical coefficients. Therefore, throughout the regime in which the mediator can consistently be integrated out, s ≪ M U V 2 s \ll M_{\mathrm{UV}}^2 s ≪ M UV 2 , perturbative couplings naturally satisfy the EFT unitarity bounds. Conversely, extrapolating the effective interaction toward s ∼ M U V 2 s \sim M_{\mathrm{UV}}^2 s ∼ M UV 2 translates the unitarity condition into an upper bound on the UV coupling of the expected perturbative size. At such energies, however, the EFT expansion itself approaches its breakdown scale and the explicit mediator should be restored.
Figure 7 summarizes these results geometrically. For a fixed single-field completion, the matching relation fixes a direction in the ( C ℓ q ( 1 , 4 ) , C ℓ q ( 3 , 4 ) ) (C_{\ell q}^{(1,4)}, C_{\ell q}^{(3,4)}) ( C ℓ q ( 1 , 4 ) , C ℓ q ( 3 , 4 ) ) plane, while the coupling-to-mass ratio sets the distance from the origin. The figure makes manifest the correct spin classification of the s s s -channel leptoquarks and its possible evasion by the t t t -channel vectors.
Figure 7. Single-field UV completions in the (s [ C ℓ q ( 1 ) ] 3333 / Λ 2 s[C_{\ell q}^{(1)}]_{3333}/\Lambda^2 s [ C ℓ q ( 1 ) ] 3333 / Λ 2 , s [ C ℓ q ( 3 ) ] 3333 / Λ 2 s[C_{\ell q}^{(3)}]_{3333}/\Lambda^2 s [ C ℓ q ( 3 ) ] 3333 / Λ 2 ) plane. For each mediator, the red line denotes the one-dimensional trajectory selected by the tree-level matching relation, the gray region is allowed by perturbative unitarity, and the regions shaded in blue and orange show the scalar- and vector-dominated domains implied by the spinning sum rules, respectively. The scalar leptoquarks ω 1 \omega_1 ω 1 and ζ \zeta ζ and the vector leptoquarks U 2 \mathcal{U}_2 U 2 and X \mathcal{X} X , exchanged in the s s s channel, are classified according to their physical spin. By contrast, the t t t -channel vectors B \mathcal{B} B and W \mathcal{W} W can evade this classification because of non-vanishing boundary contributions.
This exercise makes explicit the interpretation discussed in § 3.3 . A point outside both spin-dominated regions does not imply an inconsistent UV theory, but rather the failure of at least one assumption underlying the sign-definite sum rules. The t t t -channel vectors provide a concrete tree-level realization through non-vanishing boundary terms.
Conclusions
We have presented a systematic analysis of perturbative unitarity bounds and spinning sum rules in the baryon-number-conserving dimension-six SMEFT, with particular emphasis on the correlations among Wilson coefficients and their flavor structure. Our results provide complementary information on the range of validity of the effective description and, under suitable assumptions, on the possible ultraviolet origin of four-fermion interactions.
Using the on-shell partial-wave formalism for generic N → M N \to M N → M scattering, we have performed a fully coupled-channel analysis of the complete dimension-six operator basis. By accounting for all relevant helicity and gauge configurations, and retaining correlations among operators contributing to the same scattering amplitudes, we have obtained perturbative unitarity bounds on individual Wilson coefficients through marginalization over the remaining interactions. This procedure yields conservative one-dimensional limits
that can differ from conventional single-operator bounds. The generalized formalism also allows higher-multiplicity processes to be treated consistently, which is essential for the operator ( ϕ † ϕ ) 3 (\phi^\dagger\phi)^3 ( ϕ † ϕ ) 3 and the Yukawa-like ψ 2 ϕ 3 \psi^2\phi^3 ψ 2 ϕ 3 interactions. Our analysis includes the complete four-fermion basis and operators involving gluon fields, extending the scope of previous systematic studies.
We have complemented these results by expressing spinning sum rules for dimension-six four-fermion operators in the Warsaw basis. We have formulated their implications as definiteness conditions in flavor space and translated them into inequalities among Wilson coefficients. When the relevant ultraviolet assumptions are satisfied, these relations connect the signs of particular coefficient combinations to the spin of the exchanged heavy states, providing a means of distinguishing scalar- from vector-dominated dynamics beyond the information accessible through perturbative unitarity alone.
A central result of this work is the extension of both classes of constraints to the three-generation four-fermion sector under Minimal Flavor Violation and U(2)5 ^5 5 flavor symmetry. These hypotheses reduce the number of independent flavor structures while preserving non-trivial correlations among different generations. Within these frameworks, we have derived the corresponding unitarity and sum-rule constraints, identifying restrictions on combinations of flavor coefficients that cannot be inferred from a single-flavor analysis. The resulting allowed regions exhibit non-trivial correlations and, in certain sections of coefficient space, collapse onto lower-dimensional loci. Such features are particularly relevant for scenarios with non-universal new-physics couplings, including interactions preferentially involving third-generation fermions.
The interpretation of spinning sum rules nevertheless requires particular care. Unlike perturbative unitarity bounds, their sign constraints are not universal consistency conditions on the SMEFT. They rely on assumptions concerning the ultraviolet dynamics, including the relevant spin content and the absence of boundary contributions to the dispersion relations. Through explicit matching to representative ultraviolet completions, we have illustrated both the discriminating power and the limitations of the spinning sum rules. In particular, vector exchange in the t t t channel can generate non-vanishing boundary terms and invalidate a naive identification of the sign of a Wilson coefficient with the spin of the mediator. Departures from both the scalar- and vector-dominated sum-rule regions therefore need not signal an inconsistent ultraviolet completion, but can instead point to more general underlying dynamics.
Taken together, our results provide a systematic set of theoretical constraints that can complement experimental SMEFT analyses. The unitarity bounds offer quantitative criteria for assessing perturbative validity at the hard-scattering scale, while the spinning sum rules, where applicable, supply additional information about the structure of possible ultraviolet completions. Incorporating these constraints alongside theoretically motivated flavor hypotheses can help characterize the viable parameter space of new-physics scenarios and sharpen the interpretation of present and future high-energy measurements.
Acknowledgments
We thank Giulio Crisanti for valuable help with the diagonalization of the largest multiflavor partial-wave matrices of the four-fermion operators. This work received funding from the INFN Iniziativa Specifica APINE and from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreements n. 860881 — HIDDeN, n. 101086085 — ASYMMETRY. This work was also partially supported by the Italian MUR Departments of Excellence grant 2023-2027 “Quantum Frontiers” and by the European Union — Next Generation EU and by the Italian Ministry of University and Research (MUR) via the PRIN 2022 project n. 2022K4B5BX — AxionOrigins.
Mathematical tools
Logical expression
The function f ( x ) = a + b x − c 2 + d x + e x 2 f(x)=a+bx-\sqrt{c^2+dx+ex^2} f ( x ) = a + b x − c 2 + d x + e x 2 assumed real on [ 0 , 1 ] [0,1] [ 0 , 1 ] and with a , b , c , d , e ∈ R a,b,c,d,e\in\mathbb{R} a , b , c , d , e ∈ R — is such that f ( x ) ≥ 0 f(x)\geq0 f ( x ) ≥ 0 for every x ∈ [ 0 , 1 ] x\in[0,1] x ∈ [ 0 , 1 ] if and only if the system of inequalities
Δ ( a , b , c , d , e ) = a ≥ 0 ∧ a + b ≥ 0 ∧ a 2 − c 2 ≥ 0 ∧ ( a + b ) 2 − c 2 − d − e ≥ 0 ∧ [ e − b 2 ≥ 0 ∨ ( 2 a b − d ) ( 2 a b − d + 2 b 2 − 2 e ) ≥ 0 ] \begin{aligned}
\Delta(a,b,c,d,e) &= a\geq0\ \wedge\ a+b\geq0\ \wedge\ a^2-c^2\geq0\ \wedge\ (a+b)^2-c^2-d-e\geq0 \\
&\wedge\ [e-b^2\geq0\ \vee\ (2ab-d)(2ab-d+2b^2-2e)\geq0]
\end{aligned} Δ ( a , b , c , d , e ) = a ≥ 0 ∧ a + b ≥ 0 ∧ a 2 − c 2 ≥ 0 ∧ ( a + b ) 2 − c 2 − d − e ≥ 0 ∧ [ e − b 2 ≥ 0 ∨ ( 2 ab − d ) ( 2 ab − d + 2 b 2 − 2 e ) ≥ 0 ]
∨ 4 ( b 2 − e ) ( a 2 − c 2 ) − ( 2 a b − d ) 2 ≥ 0 (84) \vee\ 4(b^2-e)(a^2-c^2)-(2ab-d)^2\geq0 \tag*{(84)} ∨ 4 ( b 2 − e ) ( a 2 − c 2 ) − ( 2 ab − d ) 2 ≥ 0 (84)
is true.
To prove this, note that, since g ( x ) = c 2 + d x + e x 2 ≥ 0 g(x)=c^2+dx+ex^2\geq0 g ( x ) = c 2 + d x + e x 2 ≥ 0 on [ 0 , 1 ] [0,1] [ 0 , 1 ] by assumption, f ( x ) ≥ 0 f(x)\geq0 f ( x ) ≥ 0 on [ 0 , 1 ] [0,1] [ 0 , 1 ] is equivalent to ℓ ( x ) = a + b x ≥ 0 \ell(x)=a+bx\geq0 ℓ ( x ) = a + b x ≥ 0 and h ( x ) = ℓ ( x ) 2 − g ( x ) ≥ 0 h(x)=\ell(x)^2-g(x)\geq0 h ( x ) = ℓ ( x ) 2 − g ( x ) ≥ 0 on [ 0 , 1 ] [0,1] [ 0 , 1 ] . The latter reads h ( x ) = α x 2 + β x + γ h(x)=\alpha x^2+\beta x+\gamma h ( x ) = α x 2 + β x + γ , with α = b 2 − e \alpha=b^2-e α = b 2 − e , β = 2 a b − d \beta=2ab-d β = 2 ab − d , and γ = a 2 − c 2 \gamma=a^2-c^2 γ = a 2 − c 2 . Thus, it is sufficient to prove that p ( x ) = α x 2 + β x + γ ≥ 0 p(x)=\alpha x^2+\beta x+\gamma\geq0 p ( x ) = α x 2 + β x + γ ≥ 0 on [ 0 , 1 ] [0,1] [ 0 , 1 ] if and only if
γ ≥ 0 ∧ α + β + γ ≥ 0 ∧ [ α ≤ 0 ∨ β ( β + 2 α ) ≥ 0 ∨ 4 α γ − β 2 ≥ 0 ] . (85) \gamma\geq0\ \wedge\ \alpha+\beta+\gamma\geq0\ \wedge\ [\alpha\leq0\ \vee\ \beta(\beta+2\alpha)\geq0\ \vee\ 4\alpha\gamma-\beta^2\geq0]. \tag*{(85)} γ ≥ 0 ∧ α + β + γ ≥ 0 ∧ [ α ≤ 0 ∨ β ( β + 2 α ) ≥ 0 ∨ 4 α γ − β 2 ≥ 0 ] . (85)
Indeed, Δ \Delta Δ is the conjunction of Eq. (A.2) applied to ℓ \ell ℓ and h h h . The first two conditions in Eq. (A.2) are p ( 0 ) ≥ 0 p(0)\geq0 p ( 0 ) ≥ 0 and p ( 1 ) ≥ 0 p(1)\geq0 p ( 1 ) ≥ 0 , thus they are necessary; assume them in what follows. If α ≤ 0 \alpha\leq0 α ≤ 0 , then p p p is concave and has its minimum on [ 0 , 1 ] [0,1] [ 0 , 1 ] at an endpoint, so p ≥ 0 p\geq0 p ≥ 0 ; instead, if α > 0 \alpha>0 α > 0 , the vertex is at x ∗ = − β / ( 2 α ) x_*=-\beta/(2\alpha) x ∗ = − β / ( 2 α ) , and
x ∗ ≤ 0 ⟺ β ≥ 0 , x ∗ ≥ 1 ⟺ β + 2 α ≤ 0 , (86) x_*\leq0\ \Longleftrightarrow\ \beta\geq0,\qquad x_*\geq1\ \Longleftrightarrow\ \beta+2\alpha\leq0, \tag*{(86)} x ∗ ≤ 0 ⟺ β ≥ 0 , x ∗ ≥ 1 ⟺ β + 2 α ≤ 0 , (86)
so x ∗ ∉ ( 0 , 1 ) x_*\notin(0,1) x ∗ ∈ / ( 0 , 1 ) if and only if β ( β + 2 α ) ≥ 0 \beta(\beta+2\alpha)\geq0 β ( β + 2 α ) ≥ 0 , in which case the minimum is again at an endpoint and p ≥ 0 p\geq0 p ≥ 0 . Otherwise, x ∗ ∈ ( 0 , 1 ) x_*\in(0,1) x ∗ ∈ ( 0 , 1 ) and min [ 0 , 1 ] p = p ( x ∗ ) = ( 4 α γ − β 2 ) / ( 4 α ) \min_{[0,1]}p=p(x_*)=(4\alpha\gamma-\beta^2)/(4\alpha) min [ 0 , 1 ] p = p ( x ∗ ) = ( 4 α γ − β 2 ) / ( 4 α ) , which has the same sign as 4 α γ − β 2 4\alpha\gamma-\beta^2 4 α γ − β 2 . The last condition is then both necessary and sufficient.
Eigenvalues of Jordan–Wielandt matrices
Given the Hermitian Jordan–Wielandt matrix
M = ( O m × m A A † O n × n ) (87) M=\begin{pmatrix}
\mathbb{O}_{m\times m} & A\\
A^\dagger& \mathbb{O}_{n\times n}
\end{pmatrix} \tag*{(87)} M = ( O m × m A † A O n × n ) (87)
of size m + n m+n m + n , with A ∈ C m × n A\in\mathbb{C}^{m\times n} A ∈ C m × n , we have that its characteristic polynomial — with the convention p X ( x ) = det [ x 1 − X ] p_X(x)=\det[x\mathbb{1}-X] p X ( x ) = det [ x 1 − X ] — is
p M ( x ) = x n − m p A A † ( x 2 ) = x m − n p A † A ( x 2 ) , (88) p_M(x)=x^{n-m}p_{AA^\dagger}(x^2)=x^{m-n}p_{A^\dagger A}(x^2), \tag*{(88)} p M ( x ) = x n − m p A A † ( x 2 ) = x m − n p A † A ( x 2 ) , (88)
which means that the eigenvalues of M M M can be computed as the positive and negative square roots of the eigenvalues of A A † AA^\dagger A A † if m ≤ n m\leq n m ≤ n (of A † A A^\dagger A A † A if m > n m>n m > n ), together with additional ∣ n − m ∣ |n-m| ∣ n − m ∣ vanishing eigenvalues.
To prove this, consider a generic block matrix that takes the form
\begin{pmatrix}
P & Q\\
R & S
\end{pmatrix}. \tag*{(89)}\end{pmatrix}
If S S S is invertible, then
\begin{pmatrix}
P & Q\\
R & S
\end{pmatrix}
\begin{pmatrix}
\mathbb{1} & \mathbb{O}\\
-S^{-1}R & \mathbb{1}
\end{pmatrix}
=
\begin{pmatrix}
P-QS^{-1}R & Q\\
\mathbb{O} & S
\end{pmatrix}, \tag*{(90)}\end{pmatrix}
which implies that
det [ ( P Q R S ) ] = det [ P − Q S − 1 R ] det [ S ] . (91) \det\left[\begin{pmatrix} P & Q \\ R & S \end{pmatrix}\right] = \det\left[P-QS^{-1}R\right]\det[S]. \tag*{(91)} det [ ( P R Q S ) ] = det [ P − Q S − 1 R ] det [ S ] . (91)
By applying this to the matrix 1 ( m + n ) × ( m + n ) − M \mathbb{1}_{(m+n)\times(m+n)}-M 1 ( m + n ) × ( m + n ) − M , we find the result above.
Complete results for flavorful four-fermion operators
Here we collect the complete perturbative-unitarity and spinning sum-rule constraints on the current–current four-fermion SMEFT operators under the MFV and U(2)5 ^{5} 5 flavor hypotheses. Table 12 indexes the tables of marginalized bounds and the corresponding figures.
Table 12. References to the tables of marginalized bounds (from partial-wave unitarity alone and combined with the spinning sum rules) and to the figures showing the corresponding allowed regions, for each four-fermion operator under the MFV and U(2)5 ^{5} 5 flavor hypotheses. A dash indicates that the operator admits a single independent coefficient under MFV, so that the bound is one-dimensional and no region is displayed. Under U(2)5 ^{5} 5 , O d d O_{dd} O dd , O e d O_{ed} O e d , O ℓ d O_{\ell d} O ℓ d , and O q d ( 1 , 8 ) O_{qd}^{(1,8)} O q d ( 1 , 8 ) have the same decomposition and the same allowed regions as O u u O_{uu} O uu , O e u O_{eu} O e u , O ℓ u O_{\ell u} O ℓ u , and O q u ( 1 , 8 ) O_{qu}^{(1,8)} O q u ( 1 , 8 ) , respectively, and share the corresponding figure.
Operator: O ℓ ℓ = ( ℓ ˉ p γ μ ℓ r ) ( ℓ ˉ s γ μ ℓ t ) O_{\ell\ell} = (\bar{\ell}_p\gamma^\mu\ell_r)(\bar{\ell}_s\gamma_\mu\ell_t) O ℓℓ = ( ℓ ˉ p γ μ ℓ r ) ( ℓ ˉ s γ μ ℓ t )
MFV decomposition: [ C ℓ ℓ ] p r s t = C ℓ ℓ ( 1 ) δ p r δ s t + C ℓ ℓ ( 2 ) δ p t δ r s [C_{\ell\ell}]_{prst} = C_{\ell\ell}^{(1)}\delta_{pr}\delta_{st} + C_{\ell\ell}^{(2)}\delta_{pt}\delta_{rs} [ C ℓℓ ] p r s t = C ℓℓ ( 1 ) δ p r δ s t + C ℓℓ ( 2 ) δ pt δ r s
U(2)5 ^5 5 decomposition: [ C ℓ ℓ ] p r s t = C ℓ ℓ ( 1 ) δ p a δ a r δ s b δ b t + C ℓ ℓ ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C ℓ ℓ ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C ℓ ℓ ( 4 ) δ s a δ a r δ p b δ b t + C ℓ ℓ ( 5 ) ( δ p a δ a t δ s 3 δ 3 r + δ s a δ a r δ p 3 δ 3 t ) [C_{\ell\ell}]_{prst} = C_{\ell\ell}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{\ell\ell}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + \delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}) + C_{\ell\ell}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} + C_{\ell\ell}^{(4)}\delta_{sa}\delta_{ar}\delta_{pb}\delta_{bt} + C_{\ell\ell}^{(5)}(\delta_{pa}\delta_{at}\delta_{s3}\delta_{3r} + \delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t}) [ C ℓℓ ] p r s t = C ℓℓ ( 1 ) δ p a δ a r δ s b δ b t + C ℓℓ ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C ℓℓ ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C ℓℓ ( 4 ) δ s a δ a r δ p b δ b t + C ℓℓ ( 5 ) ( δ p a δ a t δ s 3 δ 3 r + δ s a δ a r δ p 3 δ 3 t )
Operators:
O q q ( 1 ) = ( q ˉ p γ μ q r ) ( q ˉ s γ μ q t ) , O q q ( 3 ) = ( q ˉ p γ μ σ I q r ) ( q ˉ s γ μ σ I q t ) O_{qq}^{(1)} = (\bar{q}_{p}\gamma^{\mu}q_{r})(\bar{q}_{s}\gamma_{\mu}q_{t}), \qquad O_{qq}^{(3)} = (\bar{q}_{p}\gamma^{\mu}\sigma^{I}q_{r})(\bar{q}_{s}\gamma_{\mu}\sigma^{I}q_{t}) O q q ( 1 ) = ( q ˉ p γ μ q r ) ( q ˉ s γ μ q t ) , O q q ( 3 ) = ( q ˉ p γ μ σ I q r ) ( q ˉ s γ μ σ I q t )
MFV decomposition:
[ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p r δ s t + C q q ( i , 2 ) δ p t δ s r + C q q ( i , 3 ) ( ( Y u Y u † ) s t δ p r + ( Y u Y u † ) p r δ s t ) + C q q ( i , 4 ) ( ( Y u Y u † ) s r δ p t + ( Y u Y u † ) p t δ s r ) , i = 1 , 3 [C_{qq}^{(i)}]_{prst} = C_{qq}^{(i,1)}\delta_{pr}\delta_{st} + C_{qq}^{(i,2)}\delta_{pt}\delta_{sr}
+ C_{qq}^{(i,3)}\left((Y_{u}Y_{u}^{\dagger})_{st}\delta_{pr} + (Y_{u}Y_{u}^{\dagger})_{pr}\delta_{st}\right)
+ C_{qq}^{(i,4)}\left((Y_{u}Y_{u}^{\dagger})_{sr}\delta_{pt} + (Y_{u}Y_{u}^{\dagger})_{pt}\delta_{sr}\right), \quad i = 1,3 [ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p r δ s t + C q q ( i , 2 ) δ pt δ sr + C q q ( i , 3 ) ( ( Y u Y u † ) s t δ p r + ( Y u Y u † ) p r δ s t ) + C q q ( i , 4 ) ( ( Y u Y u † ) sr δ pt + ( Y u Y u † ) pt δ sr ) , i = 1 , 3
U(2)5 ^{5} 5 decomposition:
[ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p a δ a r δ s b δ b t + C q q ( i , 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C q q ( i , 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C q q ( i , 4 ) δ p a δ a t δ s b δ b r + C q q ( i , 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) , i = 1 , 3 [C_{qq}^{(i)}]_{prst} = C_{qq}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{qq}^{(i,2)}\left(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + \delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}\right)
+ C_{qq}^{(i,3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} + C_{qq}^{(i,4)}\delta_{pa}\delta_{at}\delta_{sb}\delta_{br}
+ C_{qq}^{(i,5)}\left(\delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t} + \delta_{pa}\delta_{at}\delta_{s3}\delta_{3r}\right), \quad i = 1,3 [ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p a δ a r δ s b δ b t + C q q ( i , 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C q q ( i , 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C q q ( i , 4 ) δ p a δ a t δ s b δ b r + C q q ( i , 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) , i = 1 , 3
Table 14. Same as Table 13 but for O q q ( 1 , 3 ) O_{qq}^{(1,3)} O q q ( 1 , 3 ) .
Table 13. Marginalized partial-wave unitarity bounds for O ℓ ℓ O_{\ell\ell} O ℓℓ under MFV and U(2)5 ^5 5 flavor symmetry, with and without spinning sum rules under scalar or vector dominance in the UV. Bounds are intervals [ a , b ] [a,b] [ a , b ] with ( s / Λ 2 ) C ∈ [ a , b ] (s/\Lambda^2)C \in[a,b] ( s / Λ 2 ) C ∈ [ a , b ] , rounded to 3 significant digits.
Operator: O e e = ( e ˉ p γ μ e r ) ( e ˉ s γ μ e t ) O_{ee} = (\bar e_p\gamma^\mu e_r)(\bar e_s\gamma_\mu e_t) O ee = ( e ˉ p γ μ e r ) ( e ˉ s γ μ e t )
MFV decomposition: [ C e e ] p r s t = C e e ( 1 ) δ p r δ s t [C_{ee}]_{prst} = C_{ee}^{(1)}\delta_{pr}\delta_{st} [ C ee ] p r s t = C ee ( 1 ) δ p r δ s t
U(2)5 ^5 5 decomposition: [ C e e ] p r s t = C e e ( 1 ) δ p a δ a r δ s b δ b t + C e e ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C e e ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{ee}]_{prst} = C_{ee}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{ee}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + \delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}) + C_{ee}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ee ] p r s t = C ee ( 1 ) δ p a δ a r δ s b δ b t + C ee ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C ee ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 15. Same as Table 13 but for O e e O_{ee} O ee .
Operator: O u u = ( u ˉ p γ μ u r ) ( u ˉ s γ μ u t ) O_{uu} = (\bar u_p\gamma^\mu u_r)(\bar u_s\gamma_\mu u_t) O uu = ( u ˉ p γ μ u r ) ( u ˉ s γ μ u t )
MFV decomposition: [ C u u ] p r s t = C u u ( 1 ) δ p r δ s t + C u u ( 2 ) δ p t δ s r + C u u ( 3 ) ( ( Y u † Y u ) s t δ p r + ( Y u † Y u ) p r δ s t ) + C u u ( 4 ) ( ( Y u † Y u ) s r δ p t + ( Y u † Y u ) p t δ s r ) [C_{uu}]_{prst} = C_{uu}^{(1)}\delta_{pr}\delta_{st} + C_{uu}^{(2)}\delta_{pt}\delta_{sr} + C_{uu}^{(3)}((Y_u^\dagger Y_u)_{st}\delta_{pr} + (Y_u^\dagger Y_u)_{pr}\delta_{st}) + C_{uu}^{(4)}((Y_u^\dagger Y_u)_{sr}\delta_{pt} + (Y_u^\dagger Y_u)_{pt}\delta_{sr}) [ C uu ] p r s t = C uu ( 1 ) δ p r δ s t + C uu ( 2 ) δ pt δ sr + C uu ( 3 ) (( Y u † Y u ) s t δ p r + ( Y u † Y u ) p r δ s t ) + C uu ( 4 ) (( Y u † Y u ) sr δ pt + ( Y u † Y u ) pt δ sr )
U(2)5 ^5 5 decomposition: [ C u u ] p r s t = C u u ( 1 ) δ p a δ a r δ s b δ b t + C u u ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C u u ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C u u ( 4 ) δ p a δ a t δ s b δ b r + C u u ( 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) [C_{uu}]_{prst} = C_{uu}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{uu}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + \delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}) + C_{uu}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} + C_{uu}^{(4)}\delta_{pa}\delta_{at}\delta_{sb}\delta_{br} + C_{uu}^{(5)}(\delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t} + \delta_{pa}\delta_{at}\delta_{s3}\delta_{3r}) [ C uu ] p r s t = C uu ( 1 ) δ p a δ a r δ s b δ b t + C uu ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C uu ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C uu ( 4 ) δ p a δ a t δ s b δ b r + C uu ( 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r )
Table 16. Same as Table 13 but for O u u O_{uu} O uu .
Operator: O d d = ( d ˉ p γ μ d r ) ( d ˉ s γ μ d t ) O_{dd} = (\bar{d}_p\gamma^\mu d_r)(\bar{d}_s\gamma_\mu d_t) O dd = ( d ˉ p γ μ d r ) ( d ˉ s γ μ d t )
MFV decomposition: [ C d d ] p r s t = C d d ( 1 ) δ p r δ s t + C d d ( 2 ) δ p t δ s r [C_{dd}]_{prst} = C_{dd}^{(1)}\delta_{pr}\delta_{st} + C_{dd}^{(2)}\delta_{pt}\delta_{sr} [ C dd ] p r s t = C dd ( 1 ) δ p r δ s t + C dd ( 2 ) δ pt δ sr
U(2)5 ^5 5 decomposition: [ C d d ] p r s t = C d d ( 1 ) δ p a δ a r δ s b δ b t + C d d ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C d d ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C d d ( 4 ) δ p a δ a t δ s b δ b r + C d d ( 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) [C_{dd}]_{prst} = C_{dd}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{dd}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + \delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}) + C_{dd}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} + C_{dd}^{(4)}\delta_{pa}\delta_{at}\delta_{sb}\delta_{br} + C_{dd}^{(5)}(\delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t} + \delta_{pa}\delta_{at}\delta_{s3}\delta_{3r}) [ C dd ] p r s t = C dd ( 1 ) δ p a δ a r δ s b δ b t + C dd ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C dd ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C dd ( 4 ) δ p a δ a t δ s b δ b r + C dd ( 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r )
Table 17. Same as Table 13 but for O d d O_{dd} O dd .
Operator: O e u = ( e ˉ p γ μ e r ) ( u ˉ s γ μ u t ) O_{eu} = (\bar{e}_p\gamma^\mu e_r)(\bar{u}_s\gamma_\mu u_t) O e u = ( e ˉ p γ μ e r ) ( u ˉ s γ μ u t )
MFV decomposition: [ C e u ] p r s t = C e u ( 1 ) δ p r δ s t + C e u ( 2 ) δ p r ( Y u † Y u ) s t [C_{eu}]_{prst} = C_{eu}^{(1)}\delta_{pr}\delta_{st} + C_{eu}^{(2)}\delta_{pr}(Y_u^\dagger Y_u)_{st} [ C e u ] p r s t = C e u ( 1 ) δ p r δ s t + C e u ( 2 ) δ p r ( Y u † Y u ) s t
U(2)5 ^5 5 decomposition: [ C e u ] p r s t = C e u ( 1 ) δ p a δ a r δ s b δ b t + C e u ( 2 ) δ p 3 δ 3 r δ s a δ a t + C e u ( 3 ) δ s 3 δ 3 t δ p a δ a r + C e u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{eu}]_{prst} = C_{eu}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{eu}^{(2)}\delta_{p3}\delta_{3r}\delta_{sa}\delta_{at} + C_{eu}^{(3)}\delta_{s3}\delta_{3t}\delta_{pa}\delta_{ar} + C_{eu}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C e u ] p r s t = C e u ( 1 ) δ p a δ a r δ s b δ b t + C e u ( 2 ) δ p 3 δ 3 r δ s a δ a t + C e u ( 3 ) δ s 3 δ 3 t δ p a δ a r + C e u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 18. Same as Table 13 but for O e u O_{eu} O e u .
Operator: O e d = ( e ˉ p γ μ e r ) ( d ˉ s γ μ d t ) \boxed{O_{ed} = (\bar{e}_p\gamma^\mu e_r)(\bar{d}_s\gamma_\mu d_t)} O e d = ( e ˉ p γ μ e r ) ( d ˉ s γ μ d t )
MFV decomposition: [ C e d ] p r s t = C e d ( 1 ) δ p r δ s t [C_{ed}]_{prst} = C_{ed}^{(1)}\delta_{pr}\delta_{st} [ C e d ] p r s t = C e d ( 1 ) δ p r δ s t
U(2)5 ^5 5 decomposition: [ C e d ] p r s t = C e d ( 1 ) δ p a δ a r δ s b δ b t + C e d ( 2 ) δ p 3 δ 3 r δ s a δ a t + C e d ( 3 ) δ s 3 δ 3 t δ p a δ a r + C e d ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{ed}]_{prst} = C_{ed}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{ed}^{(2)}\delta_{p3}\delta_{3r}\delta_{sa}\delta_{at} + C_{ed}^{(3)}\delta_{s3}\delta_{3t}\delta_{pa}\delta_{ar} + C_{ed}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C e d ] p r s t = C e d ( 1 ) δ p a δ a r δ s b δ b t + C e d ( 2 ) δ p 3 δ 3 r δ s a δ a t + C e d ( 3 ) δ s 3 δ 3 t δ p a δ a r + C e d ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 19. Same as Table 13 but for O e d O_{ed} O e d .
Operators: O u d ( 1 ) = ( u ˉ p γ μ u r ) ( d ˉ s γ μ d t ) , O u d ( 8 ) = ( u ˉ p γ μ T A u r ) ( d ˉ s γ μ T A d t ) O_{ud}^{(1)} = (\bar{u}_p\gamma^\mu u_r)(\bar{d}_s\gamma_\mu d_t),\quad O_{ud}^{(8)} = (\bar{u}_p\gamma^\mu T^A u_r)(\bar{d}_s\gamma_\mu T^A d_t) O u d ( 1 ) = ( u ˉ p γ μ u r ) ( d ˉ s γ μ d t ) , O u d ( 8 ) = ( u ˉ p γ μ T A u r ) ( d ˉ s γ μ T A d t )
MFV decomposition: [ C u d ( i ) ] p r s t = C u d ( i , 1 ) δ p r δ s t + C u d ( i , 2 ) ( Y u † Y u ) p r δ s t , i = 1 , 8 [C_{ud}^{(i)}]_{prst} = C_{ud}^{(i,1)}\delta_{pr}\delta_{st} + C_{ud}^{(i,2)}(Y_u^\dagger Y_u)_{pr}\delta_{st},\quad i = 1,8 [ C u d ( i ) ] p r s t = C u d ( i , 1 ) δ p r δ s t + C u d ( i , 2 ) ( Y u † Y u ) p r δ s t , i = 1 , 8
U(2)5 ^5 5 decomposition: [ C u d ( i ) ] p r s t = C u d ( i , 1 ) δ p a δ a r δ s b δ b t + C u d ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C u d ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C u d ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8 [C_{ud}^{(i)}]_{prst} = C_{ud}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{ud}^{(i,2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + C_{ud}^{(i,3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{ud}^{(i,4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t},\quad i = 1,8 [ C u d ( i ) ] p r s t = C u d ( i , 1 ) δ p a δ a r δ s b δ b t + C u d ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C u d ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C u d ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8
Table 20. Same as Table 13 but for O u d ( 1 , 8 ) O_{ud}^{(1,8)} O u d ( 1 , 8 ) .
Operator: O ℓ e = ( ℓ ˉ p γ μ ℓ r ) ( e ˉ s γ μ e t ) O_{\ell e} = (\bar{\ell}_p\gamma^\mu\ell_r)(\bar{e}_s\gamma_\mu e_t) O ℓ e = ( ℓ ˉ p γ μ ℓ r ) ( e ˉ s γ μ e t )
MFV decomposition: [ C ℓ e ] p r s t = C ℓ e ( 1 ) δ p r δ s t [C_{\ell e}]_{prst} = C_{\ell e}^{(1)}\delta_{pr}\delta_{st} [ C ℓ e ] p r s t = C ℓ e ( 1 ) δ p r δ s t
U(2)5 ^5 5 decomposition: [ C ℓ e ] p r s t = C ℓ e ( 1 ) δ p a δ a r δ s b δ b t + C ℓ e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{\ell e}]_{prst} = C_{\ell e}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{\ell e}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + C_{\ell e}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{\ell e}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ℓ e ] p r s t = C ℓ e ( 1 ) δ p a δ a r δ s b δ b t + C ℓ e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 21. Same as Table 13 but for O ℓ e O_{\ell e} O ℓ e .
Operator: O ℓ u = ( ℓ ˉ p γ μ ℓ r ) ( u ˉ s γ μ u t ) O_{\ell u} = (\bar{\ell}_p\gamma^\mu\ell_r)(\bar{u}_s\gamma_\mu u_t) O ℓ u = ( ℓ ˉ p γ μ ℓ r ) ( u ˉ s γ μ u t )
MFV decomposition: [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p r δ s t + C ℓ u ( 2 ) δ p r ( Y u † Y u ) s t [C_{\ell u}]_{prst} = C_{\ell u}^{(1)}\delta_{pr}\delta_{st} + C_{\ell u}^{(2)}\delta_{pr}(Y_u^\dagger Y_u)_{st} [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p r δ s t + C ℓ u ( 2 ) δ p r ( Y u † Y u ) s t
U(2)5 ^5 5 decomposition: [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p a δ a r δ s b δ b t + C ℓ u ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ u ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{\ell u}]_{prst} = C_{\ell u}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{\ell u}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + C_{\ell u}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{\ell u}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p a δ a r δ s b δ b t + C ℓ u ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ u ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 22. Same as Table 13 but for O ℓ u O_{\ell u} O ℓ u .
Operator: O ℓ d = ( ℓ ˉ p γ μ ℓ r ) ( d ˉ s γ μ d t ) \mathcal{O}_{\ell d}=(\bar{\ell}_p\gamma^\mu\ell_r)(\bar{d}_s\gamma_\mu d_t) O ℓ d = ( ℓ ˉ p γ μ ℓ r ) ( d ˉ s γ μ d t )
MFV decomposition: [ C ℓ d ] p r s t = C ℓ d ( 1 ) δ p r δ s t [C_{\ell d}]_{prst}=C_{\ell d}^{(1)}\delta_{pr}\delta_{st} [ C ℓ d ] p r s t = C ℓ d ( 1 ) δ p r δ s t
U(2)5 ^5 5 decomposition: [ C ℓ d ] p r s t = C ℓ d ( 1 ) δ p a δ a r δ s b δ b t + C ℓ d ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ d ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ d ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{\ell d}]_{prst}=C_{\ell d}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{\ell d}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+C_{\ell d}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}+C_{\ell d}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ℓ d ] p r s t = C ℓ d ( 1 ) δ p a δ a r δ s b δ b t + C ℓ d ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ d ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ d ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 23. Same as Table 13 but for O ℓ d \mathcal{O}_{\ell d} O ℓ d .
Operator: O q e = ( q ˉ p γ μ q r ) ( e ˉ s γ μ e t ) \mathcal{O}_{qe}=(\bar{q}_p\gamma^\mu q_r)(\bar{e}_s\gamma_\mu e_t) O q e = ( q ˉ p γ μ q r ) ( e ˉ s γ μ e t )
MFV decomposition: [ C q e ] p r s t = C q e ( 1 ) δ p r δ s t + C q e ( 2 ) ( Y u Y u † ) p r δ s t [C_{qe}]_{prst}=C_{qe}^{(1)}\delta_{pr}\delta_{st}+C_{qe}^{(2)}(Y_uY_u^\dagger)_{pr}\delta_{st} [ C q e ] p r s t = C q e ( 1 ) δ p r δ s t + C q e ( 2 ) ( Y u Y u † ) p r δ s t
U(2)5 ^5 5 decomposition: [ C q e ] p r s t = C q e ( 1 ) δ p a δ a r δ s b δ b t + C q e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C q e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C q e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{qe}]_{prst}=C_{qe}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{qe}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+C_{qe}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}+C_{qe}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C q e ] p r s t = C q e ( 1 ) δ p a δ a r δ s b δ b t + C q e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C q e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C q e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t
Table 24. Same as Table 13 but for O q e \mathcal{O}_{qe} O q e .
Operators:
O q u ( 1 ) = ( q ˉ p γ μ q r ) ( u ˉ s γ μ u t ) , O q u ( 8 ) = ( q ˉ p γ μ T A q r ) ( u ˉ s γ μ T A u t ) O_{qu}^{(1)} = (\bar q_p \gamma^\mu q_r)(\bar u_s \gamma_\mu u_t), \qquad O_{qu}^{(8)} = (\bar q_p \gamma^\mu T^A q_r)(\bar u_s \gamma_\mu T^A u_t) O q u ( 1 ) = ( q ˉ p γ μ q r ) ( u ˉ s γ μ u t ) , O q u ( 8 ) = ( q ˉ p γ μ T A q r ) ( u ˉ s γ μ T A u t )
MFV decomposition:
[ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p r δ s t + C q u ( i , 2 ) ( Y u Y u † ) p r δ s t + C q u ( i , 3 ) δ p r ( Y u † Y u ) s t + C q u ( i , 4 ) ( Y u ) p t ( Y u † ) s r , i = 1 , 8 [C_{qu}^{(i)}]_{prst} = C_{qu}^{(i,1)}\delta_{pr}\delta_{st} + C_{qu}^{(i,2)}(Y_uY_u^\dagger)_{pr}\delta_{st}
+ C_{qu}^{(i,3)}\delta_{pr}(Y_u^\dagger Y_u)_{st} + C_{qu}^{(i,4)}(Y_u)_{pt}(Y_u^\dagger)_{sr}, \qquad i=1,8 [ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p r δ s t + C q u ( i , 2 ) ( Y u Y u † ) p r δ s t + C q u ( i , 3 ) δ p r ( Y u † Y u ) s t + C q u ( i , 4 ) ( Y u ) pt ( Y u † ) sr , i = 1 , 8
U(2)5 ^5 5 decomposition:
[ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p a δ a r δ s b δ b t + C q u ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C q u ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C q u ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8 [C_{qu}^{(i)}]_{prst} = C_{qu}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{qu}^{(i,2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}
+ C_{qu}^{(i,3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{qu}^{(i,4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t}, \qquad i=1,8 [ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p a δ a r δ s b δ b t + C q u ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C q u ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C q u ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8
Table 25. Same as Table 13 but for O q u ( 1 , 8 ) O_{qu}^{(1,8)} O q u ( 1 , 8 ) .
Operators:
O q d ( 1 ) = ( q ˉ p γ μ q r ) ( d ˉ s γ μ d t ) , O q d ( 8 ) = ( q ˉ p γ μ T A q r ) ( d ˉ s γ μ T A d t ) O_{qd}^{(1)} = (\bar{q}_p\gamma^\mu q_r)(\bar{d}_s\gamma_\mu d_t), \qquad O_{qd}^{(8)} = (\bar{q}_p\gamma^\mu T^A q_r)(\bar{d}_s\gamma_\mu T^A d_t) O q d ( 1 ) = ( q ˉ p γ μ q r ) ( d ˉ s γ μ d t ) , O q d ( 8 ) = ( q ˉ p γ μ T A q r ) ( d ˉ s γ μ T A d t )
MFV decomposition:
[ C q d ( i ) ] p r s t = C q d ( i , 1 ) δ p r δ s t + C q d ( i , 2 ) ( Y u Y u † ) p r δ s t , i = 1 , 8 [C_{qd}^{(i)}]_{prst} = C_{qd}^{(i,1)}\delta_{pr}\delta_{st} + C_{qd}^{(i,2)}(Y_uY_u^\dagger)_{pr}\delta_{st}, \qquad i=1,8 [ C q d ( i ) ] p r s t = C q d ( i , 1 ) δ p r δ s t + C q d ( i , 2 ) ( Y u Y u † ) p r δ s t , i = 1 , 8
U(2)5 ^5 5 decomposition:
[ C q d ( i ) ] p r s t = C q d ( i , 1 ) δ p a δ a r δ s b δ b t + C q d ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C q d ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C q d ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8 [C_{qd}^{(i)}]_{prst} = C_{qd}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{qd}^{(i,2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}
+ C_{qd}^{(i,3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{qd}^{(i,4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t}, \qquad i=1,8 [ C q d ( i ) ] p r s t = C q d ( i , 1 ) δ p a δ a r δ s b δ b t + C q d ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C q d ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C q d ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8
Table 26. Same as Table Table 13 but for O q d ( 1 , 8 ) O_{qd}^{(1,8)} O q d ( 1 , 8 ) .
Figure 8. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O ℓ ℓ O_{\ell\ell} O ℓℓ under MFV, where [ C ℓ ℓ ] p r s t = C ℓ ℓ ( 1 ) δ p r δ s t + C ℓ ℓ ( 2 ) δ p t δ s r [C_{\ell\ell}]_{prst}=C_{\ell\ell}^{(1)}\delta_{pr}\delta_{st}+C_{\ell\ell}^{(2)}\delta_{pt}\delta_{sr} [ C ℓℓ ] p r s t = C ℓℓ ( 1 ) δ p r δ s t + C ℓℓ ( 2 ) δ pt δ sr .
Figure 9. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O ℓ ℓ O_{\ell\ell} O ℓℓ under U ( 2 ) 5 \mathrm{U}(2)^5 U ( 2 ) 5 flavor symmetry, where [ C ℓ ℓ ] p r s t = C ℓ ℓ ( 1 ) δ p a δ a r δ s b δ b t + C ℓ ℓ ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C ℓ ℓ ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C ℓ ℓ ( 4 ) δ s a δ a r δ p b δ b t + C ℓ ℓ ( 5 ) ( δ p a δ a t δ s 3 δ 3 r + δ s a δ a r δ p 3 δ 3 t ) [C_{\ell\ell}]_{prst}=C_{\ell\ell}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{\ell\ell}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t})+C_{\ell\ell}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t}+C_{\ell\ell}^{(4)}\delta_{sa}\delta_{ar}\delta_{pb}\delta_{bt}+C_{\ell\ell}^{(5)}(\delta_{pa}\delta_{at}\delta_{s3}\delta_{3r}+\delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t}) [ C ℓℓ ] p r s t = C ℓℓ ( 1 ) δ p a δ a r δ s b δ b t + C ℓℓ ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C ℓℓ ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C ℓℓ ( 4 ) δ s a δ a r δ p b δ b t + C ℓℓ ( 5 ) ( δ p a δ a t δ s 3 δ 3 r + δ s a δ a r δ p 3 δ 3 t ) . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 10. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O q q ( 1 , 3 ) \mathcal{O}_{qq}^{(1,3)} O q q ( 1 , 3 ) under MFV, where [ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p r δ s t + C q q ( i , 2 ) δ p t δ s r + C q q ( i , 3 ) ( ( Y u Y u † ) s t δ p r + ( Y u Y u † ) p r δ s t ) + C q q ( i , 4 ) ( ( Y u Y u † ) s r δ p t + ( Y u Y u † ) p t δ s r ) [C_{qq}^{(i)}]_{prst}=C_{qq}^{(i,1)}\delta_{pr}\delta_{st}+C_{qq}^{(i,2)}\delta_{pt}\delta_{sr}+C_{qq}^{(i,3)}((Y_uY_u^\dagger)_{st}\delta_{pr}+(Y_uY_u^\dagger)_{pr}\delta_{st})+C_{qq}^{(i,4)}((Y_uY_u^\dagger)_{sr}\delta_{pt}+(Y_uY_u^\dagger)_{pt}\delta_{sr}) [ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p r δ s t + C q q ( i , 2 ) δ pt δ sr + C q q ( i , 3 ) (( Y u Y u † ) s t δ p r + ( Y u Y u † ) p r δ s t ) + C q q ( i , 4 ) (( Y u Y u † ) sr δ pt + ( Y u Y u † ) pt δ sr ) , i = 1 , 3 i=1,3 i = 1 , 3 . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 11. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O q q ( 1 , 3 ) O^{(1,3)}_{qq} O q q ( 1 , 3 ) under U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, where [ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p a δ a r δ s b δ b t + C q q ( i , 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C q q ( i , 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C q q ( i , 4 ) δ p a δ a t δ s b δ b r + C q q ( i , 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) [C^{(i)}_{qq}]_{prst}=C^{(i,1)}_{qq}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C^{(i,2)}_{qq}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t})+C^{(i,3)}_{qq}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t}+C^{(i,4)}_{qq}\delta_{pa}\delta_{at}\delta_{sb}\delta_{br}+C^{(i,5)}_{qq}(\delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t}+\delta_{pa}\delta_{at}\delta_{s3}\delta_{3r}) [ C q q ( i ) ] p r s t = C q q ( i , 1 ) δ p a δ a r δ s b δ b t + C q q ( i , 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C q q ( i , 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C q q ( i , 4 ) δ p a δ a t δ s b δ b r + C q q ( i , 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) , i = 1 , 3 i=1,3 i = 1 , 3 . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 12. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O e e O_{ee} O ee under U(2)5 ^5 5 flavor symmetry, where [ C e e ] p r s t = C e e ( 1 ) δ p a δ a r δ s b δ b t + C e e ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C e e ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{ee}]_{prst}=C_{ee}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{ee}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t})+C_{ee}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ee ] p r s t = C ee ( 1 ) δ p a δ a r δ s b δ b t + C ee ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C ee ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 13. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O u u O_{uu} O uu under MFV, where [ C u u ] p r s t = C u u ( 1 ) δ p r δ s t + C u u ( 2 ) δ p t δ s r + C u u ( 3 ) ( ( Y u † Y u ) s t δ p r + ( Y u † Y u ) p r δ s t ) + C u u ( 4 ) ( ( Y u † Y u ) s r δ p t + ( Y u † Y u ) p t δ s r ) [C_{uu}]_{prst}=C_{uu}^{(1)}\delta_{pr}\delta_{st}+C_{uu}^{(2)}\delta_{pt}\delta_{sr}+C_{uu}^{(3)}((Y_u^\dagger Y_u)_{st}\delta_{pr}+(Y_u^\dagger Y_u)_{pr}\delta_{st})+C_{uu}^{(4)}((Y_u^\dagger Y_u)_{sr}\delta_{pt}+(Y_u^\dagger Y_u)_{pt}\delta_{sr}) [ C uu ] p r s t = C uu ( 1 ) δ p r δ s t + C uu ( 2 ) δ pt δ sr + C uu ( 3 ) (( Y u † Y u ) s t δ p r + ( Y u † Y u ) p r δ s t ) + C uu ( 4 ) (( Y u † Y u ) sr δ pt + ( Y u † Y u ) pt δ sr ) . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 14. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O u u O_{uu} O uu under U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, where [ C u u ] p r s t = C u u ( 1 ) δ p a δ a r δ s b δ b t + C u u ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C u u ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C u u ( 4 ) δ p a δ a t δ s b δ b r + C u u ( 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) [C_{uu}]_{prst}=C_{uu}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{uu}^{(2)}(\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t})+C_{uu}^{(3)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t}+C_{uu}^{(4)}\delta_{pa}\delta_{at}\delta_{sb}\delta_{br}+C_{uu}^{(5)}(\delta_{sa}\delta_{ar}\delta_{p3}\delta_{3t}+\delta_{pa}\delta_{at}\delta_{s3}\delta_{3r}) [ C uu ] p r s t = C uu ( 1 ) δ p a δ a r δ s b δ b t + C uu ( 2 ) ( δ s a δ a t δ p 3 δ 3 r + δ p a δ a r δ s 3 δ 3 t ) + C uu ( 3 ) δ p 3 δ 3 r δ s 3 δ 3 t + C uu ( 4 ) δ p a δ a t δ s b δ b r + C uu ( 5 ) ( δ s a δ a r δ p 3 δ 3 t + δ p a δ a t δ s 3 δ 3 r ) . Same decomposition and plots are valid also for O d d O_{dd} O dd . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 15. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O d d O_{dd} O dd under MFV, where [ C d d ] p r s t = C d d ( 1 ) δ p r δ s t + C d d ( 2 ) δ p t δ s r [C_{dd}]_{prst}=C_{dd}^{(1)}\delta_{pr}\delta_{st}+C_{dd}^{(2)}\delta_{pt}\delta_{sr} [ C dd ] p r s t = C dd ( 1 ) δ p r δ s t + C dd ( 2 ) δ pt δ sr .
Figure 16. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O e u O_{eu} O e u under MFV, where [ C e u ] p r s t = C e u ( 1 ) δ p r δ s t + C e u ( 2 ) δ p r ( Y u † Y u ) s t [C_{eu}]_{prst}=C_{eu}^{(1)}\delta_{pr}\delta_{st}+C_{eu}^{(2)}\delta_{pr}(Y_u^\dagger Y_u)_{st} [ C e u ] p r s t = C e u ( 1 ) δ p r δ s t + C e u ( 2 ) δ p r ( Y u † Y u ) s t .
Figure 17. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O e u O_{eu} O e u under U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, where [ C e u ] p r s t = C e u ( 1 ) δ p a δ a r δ s b δ b t + C e u ( 2 ) δ p 3 δ 3 r δ s a δ a t + C e u ( 3 ) δ s 3 δ 3 t δ p a δ a r + C e u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{eu}]_{prst}=C_{eu}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{eu}^{(2)}\delta_{p3}\delta_{3r}\delta_{sa}\delta_{at}+C_{eu}^{(3)}\delta_{s3}\delta_{3t}\delta_{pa}\delta_{ar}+C_{eu}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C e u ] p r s t = C e u ( 1 ) δ p a δ a r δ s b δ b t + C e u ( 2 ) δ p 3 δ 3 r δ s a δ a t + C e u ( 3 ) δ s 3 δ 3 t δ p a δ a r + C e u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t . Same decomposition and plots are valid also for O e d O_{ed} O e d . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 18. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O u d ( 1 , 8 ) O_{ud}^{(1,8)} O u d ( 1 , 8 ) under MFV, where [ C u d ( i ) ] p r s t = C u d ( i , 1 ) δ p r δ s t + C u d ( i , 2 ) ( Y u † Y u ) p r δ s t [C_{ud}^{(i)}]_{prst}=C_{ud}^{(i,1)}\delta_{pr}\delta_{st}+C_{ud}^{(i,2)}(Y_u^\dagger Y_u)_{pr}\delta_{st} [ C u d ( i ) ] p r s t = C u d ( i , 1 ) δ p r δ s t + C u d ( i , 2 ) ( Y u † Y u ) p r δ s t , i = 1 , 8 i=1,8 i = 1 , 8 . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 19. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O u d ( 1 , 8 ) O_{ud}^{(1,8)} O u d ( 1 , 8 ) under U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, where [ C u d ] p r s t = C u d ( i , 1 ) δ p a δ a r δ s b δ b t + C u d ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C u d ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C u d ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{ud}]_{prst}=C_{ud}^{(i,1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{ud}^{(i,2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+C_{ud}^{(i,3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}+C_{ud}^{(i,4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C u d ] p r s t = C u d ( i , 1 ) δ p a δ a r δ s b δ b t + C u d ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C u d ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C u d ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8 i=1,8 i = 1 , 8 . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 20. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O ℓ e O_{\ell e} O ℓ e under U(2)5 ^5 5 flavor symmetry, where [ C ℓ e ] p r s t = C ℓ e ( 1 ) δ p a δ a r δ s b δ b t + C ℓ e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{\ell e}]_{prst}=C_{\ell e}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{\ell e}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+C_{\ell e}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}+C_{\ell e}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ℓ e ] p r s t = C ℓ e ( 1 ) δ p a δ a r δ s b δ b t + C ℓ e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 21. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O ℓ u O_{\ell u} O ℓ u under MFV, where [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p r δ s t + C ℓ u ( 2 ) δ p r ( Y u † Y u ) s t [C_{\ell u}]_{prst}=C_{\ell u}^{(1)}\delta_{pr}\delta_{st}+C_{\ell u}^{(2)}\delta_{pr}(Y_u^\dagger Y_u)_{st} [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p r δ s t + C ℓ u ( 2 ) δ p r ( Y u † Y u ) s t .
Figure 22. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O ℓ u O_{\ell u} O ℓ u under U ( 2 ) 5 \mathrm{U}(2)^5 U ( 2 ) 5 flavor symmetry, where [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p a δ a r δ s b δ b t + C ℓ u ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ u ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{\ell u}]_{prst}=C_{\ell u}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt}+C_{\ell u}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r}+C_{\ell u}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t}+C_{\ell u}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C ℓ u ] p r s t = C ℓ u ( 1 ) δ p a δ a r δ s b δ b t + C ℓ u ( 2 ) δ s a δ a t δ p 3 δ 3 r + C ℓ u ( 3 ) δ p a δ a r δ s 3 δ 3 t + C ℓ u ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t . Same decomposition and plots are valid also for O ℓ d O_{\ell d} O ℓ d . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 23. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O q e O_{qe} O q e under MFV, where [ C q e ] p r s t = C q e ( 1 ) δ p r δ s t + C q e ( 2 ) ( Y u Y u † ) p r δ s t [C_{qe}]_{prst}=C_{qe}^{(1)}\delta_{pr}\delta_{st}+C_{qe}^{(2)}(Y_uY_u^\dagger)_{pr}\delta_{st} [ C q e ] p r s t = C q e ( 1 ) δ p r δ s t + C q e ( 2 ) ( Y u Y u † ) p r δ s t .
Figure 24. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operator O q e O_{qe} O q e under U(2)5 ^{5} 5 flavor symmetry, where [ C q e ] p r s t = C q e ( 1 ) δ p a δ a r δ s b δ b t + C q e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C q e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C q e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{qe}]_{prst} = C_{qe}^{(1)}\delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{qe}^{(2)}\delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + C_{qe}^{(3)}\delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{qe}^{(4)}\delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C q e ] p r s t = C q e ( 1 ) δ p a δ a r δ s b δ b t + C q e ( 2 ) δ s a δ a t δ p 3 δ 3 r + C q e ( 3 ) δ p a δ a r δ s 3 δ 3 t + C q e ( 4 ) δ p 3 δ 3 r δ s 3 δ 3 t . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 25. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O q u ( 1 , 8 ) \mathcal{O}_{qu}^{(1,8)} O q u ( 1 , 8 ) under MFV, where [ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p r δ s t + C q u ( i , 2 ) ( Y u Y u † ) p r δ s t + C q u ( i , 3 ) δ p r ( Y u † Y u ) s t + C q u ( i , 4 ) ( Y u ) p t ( Y u † ) s r [C_{qu}^{(i)}]_{prst}=C_{qu}^{(i,1)}\delta_{pr}\delta_{st}+C_{qu}^{(i,2)}(Y_uY_u^\dagger)_{pr}\delta_{st}+C_{qu}^{(i,3)}\delta_{pr}(Y_u^\dagger Y_u)_{st}+C_{qu}^{(i,4)}(Y_u)_{pt}(Y_u^\dagger)_{sr} [ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p r δ s t + C q u ( i , 2 ) ( Y u Y u † ) p r δ s t + C q u ( i , 3 ) δ p r ( Y u † Y u ) s t + C q u ( i , 4 ) ( Y u ) pt ( Y u † ) sr , i = 1 , 8 i=1,8 i = 1 , 8 . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 26. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O q u ( 1 , 8 ) O_{qu}^{(1,8)} O q u ( 1 , 8 ) under U ( 2 ) 5 U(2)^5 U ( 2 ) 5 flavor symmetry, where [ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p a δ a r δ s b δ b t + C q u ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C q u ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C q u ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t [C_{qu}^{(i)}]_{prst} = C_{qu}^{(i,1)} \delta_{pa}\delta_{ar}\delta_{sb}\delta_{bt} + C_{qu}^{(i,2)} \delta_{sa}\delta_{at}\delta_{p3}\delta_{3r} + C_{qu}^{(i,3)} \delta_{pa}\delta_{ar}\delta_{s3}\delta_{3t} + C_{qu}^{(i,4)} \delta_{p3}\delta_{3r}\delta_{s3}\delta_{3t} [ C q u ( i ) ] p r s t = C q u ( i , 1 ) δ p a δ a r δ s b δ b t + C q u ( i , 2 ) δ s a δ a t δ p 3 δ 3 r + C q u ( i , 3 ) δ p a δ a r δ s 3 δ 3 t + C q u ( i , 4 ) δ p 3 δ 3 r δ s 3 δ 3 t , i = 1 , 8 i = 1, 8 i = 1 , 8 . Same decomposition and plots are valid also for O q d ( 1 , 8 ) O_{qd}^{(1,8)} O q d ( 1 , 8 ) . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.
Figure 27. Partial-wave unitarity bounds and sum-rule constraints (under scalar and vector dominance) for the operators O q d ( 1 , 8 ) O^{(1,8)}_{qd} O q d ( 1 , 8 ) under MFV, where [ C q d ( i ) ] p r s t = C q d ( i , 1 ) δ p r δ s t + C q d ( i , 2 ) ( Y u Y u † ) p r δ s t [C^{(i)}_{qd}]_{prst} = C^{(i,1)}_{qd} \delta_{pr}\delta_{st} + C^{(i,2)}_{qd}(Y_uY_u^\dagger)_{pr}\delta_{st} [ C q d ( i ) ] p r s t = C q d ( i , 1 ) δ p r δ s t + C q d ( i , 2 ) ( Y u Y u † ) p r δ s t , i = 1 , 8 i = 1, 8 i = 1 , 8 . The regions shown are two-dimensional slices of the full allowed region, obtained by setting to zero all the independent coefficients other than the two displayed.