WorksPhysics and Astronomy
A New GUT
Abstract
We present a new class of 4-dimensional Grand Unified Theories where the gauge sectors of the standard model have a novel embedding in the unified group that does not factor through an subgroup. The smallest example has gauge algebra and an anomaly-free spectrum that produces the three standard model generations with no additional chiral matter. The embedding predicts the axion-photon anomaly ratio and the weak mixing angle at the unification scale. Gauge-mediated proton decay proceeds through flipped--like operators and the minimally charged smooth monopole carries two Dirac charge units. The global gauge group can be or , with the former allowing color singlets of half-integer electric charge. We develop a GUT bootstrap to constrain embeddings with three generations and no additional chiral matter. For unified fermions in up to two-index representations, we prove that all new embeddings arise from the family, including flavor unified models for .
Grand Unification offers a compelling origin story for the Standard Model (SM). Unification explains why the SM gauge couplings are all of similar order and the pattern of charges of SM fermions, in particular the quantization of electric charge in integers [1, 2, 3, 4].
The classic example of a grand unified theory (GUT) is the GUT [1]. All the simple GUTs1 in the literature contain or branch through [5, 6, 7, 8, 9], e.g.
For all these cases, the SM gauge group is [10]. Both and are embedded at equal levels, the hypercharge normalization factor relative to QCD is fixed to be the standard . The QCD axion couples to the photon with a coupling [11, 12].
We present a new GUT which departs from this paradigm. The unified gauge group is (or depending on the matter content) and has QCD embedded at level , with , . This particle content appears in the literature [13, 14], but to our knowledge our work is the first GUT construction with a non-canonical SM embedding specified, c.f. [9, 15, 16]. Our embedding has immediate consequences: the weak mixing angle at the GUT scale is and the QCD axion-photon coupling is set by .
The example is the first member of an infinite family of embeddings into , with . They share the ratio but have , with or (the latter for ).
Explicit anomaly-free fermion spectra with three SM generations are given in Sec. S2 of the Supplemental Material. Remarkably, for we find that the theories do not need a 3-fold replication of UV fermions, but the minimal spectrum reproduces 3 SM generations, providing interesting candidates for flavor-unified GUTs.
Interestingly, we find that for with and any other compact simple Lie groups, are the only two possibilities, irrespective of the dimension of the representations! This result is achieved by considering general constraints on the SM embedding in the GUT, a kind of GUT bootstrap that can scan over large classes of GUTs and SM embeddings. We provide further details in the appendix S3 and a catalog of results as an application.
We comment on the novelty of our construction relative to the literature. Ref. [16] scanned all SM embeddings in for and, for , those that go through an subgroup. Besides the standard embedding, the latter include for , with , which corresponds to our family II (Sec. S4) of the Supplemental Material. All embeddings found there give . Our embedding does not go through and is not found in the scan of Ref. [16]. QCD at level 3 also appears in the color-flavor unified models of Refs. [17, 18], where under , and in the models of Ref. [19] (see also the classification in [20]). In these models the weak group is not unified, so neither nor is predicted. Family unification using spinors of the groups [21, 22], with , provides an even number of SM generations. The additional standard and mirror families must be removed from the spectrum [23, 24].
In this paper we present the simplest version of the model, emphasizing the chiral content. There are a few interesting phenomenological consequences of this theory. If the gauge group is , the SM gauge group is and the usual electric charge quantization follows [10]. But if the group is , we get a different global structure for the SM gauge group, , and vector-like fermions with charges are allowed.2 For both and the smooth monopole carries two Dirac charge units as a consequence of the embedding, and there are no stable defects of higher dimension (strings or domain walls) associated to GUT symmetry breaking. is violated via dimension-6 operators similar to those in flipped- models and leads to proton decay that can be within current bounds [26] and could be distinguished from other GUTs. In contrast to GUTs based on , there is no symmetry.
The axion-photon coupling is one of the most distinctive phenomenological consequences and is what ultimately can distinguish this theory from standard GUTs. The theory contains fractional instantons that can have a non-trivial impact in axion physics, too.
There are many future model building directions – characterizing the spectrum for different options for gauge coupling unification, the Higgs sector of the GUT (including, e.g. doublet-triplet splitting), flavor phenomenology (specially in the flavor unified models of App. S2), neutrino mass generation, and embedding in string theory. We note that the model can be made supersymmetric, but in this case the coupling runs strongly and leads to a Landau pole close to the unification scale.
The model. The GUT gauge algebra is , and the embedding of the SM can be shown by the branching of the fundamental representation of :
The indices of embedding are and can be obtained from the normalization of the generators [9, 28]
We will use the indices to label the 5 SM representations in the branching of the 12 in eq. (2).
The fermion content organizes into three copies of the antisymmetric and (conjugate) symmetric two-index representations, . Each of these contains exactly one SM family together with vector-like representations (denoted , see App. S1 for details),
The contains fermions that all admit SM-invariant masses, so the only chiral states below are the three SM families.
The and representations have even 12-ality, which allows this model to have gauge group or . Representations with even 12-ality have only SM-like charges, with integer electric charges. However, if the gauge group is it allows for fermions in the fundamental 12 representation. In that case, the low energy gauge group is
and every color-singlet state from the fundamental has an electric charge which is an integer multiple of . This is therefore a GUT that would survive the discovery of exotic particles with fractional charges.
Predictions. The standard GUT predictions are modified due to the non-standard embedding of QCD. Below we briefly discuss the most salient features.
The weak mixing angle. —The gauge coupling matching condition is
The weak mixing angle at the GUT scale can also be written in terms of the levels of embedding as [12, 29]
This is to be compared with for standard GUT embeddings. This requires states below the unification scale to modify the running from the SM (or MSSM) predictions, see Figure 1. We give some representative examples of how this can be achieved below.

Figure 1. Two-loop running of , with for with one added pair. Solid: with the vector-like states and (gray lines mark their masses); dashed: SM content only. Inputs at from Ref. [35]. Above (black) the coupling grows and reaches a formal pole at ; the gray band lies above the species scale [34].
The axion-photon coupling. —The ratio of the anomaly coefficients, which sets the axion-photon coupling, can be written in terms of the levels of embedding [12]
The axion coupling to photons is given by [27]
so that for a fixed decay constant, gives a coupling 1.7 times larger than the standard GUT value (see Figure 2).
Figure 2. Axion-photon coupling versus mass. Black: standard GUTs, . Solid red: this model, , with a QCD-dominated axion mass. Dashed and dotted red: the same with an aligned contribution from small instantons, , for and . Hatched: not allowed in this model. As in standard GUTs axion-like particles above the QCD line are forbidden [12]. Yellow: the QCD axion band, [30]. Shaded regions and dashed curves: limits and projections from Ref. [31].
This provides a well-motivated target for experimental precision. The power in a haloscope experiment scales as . If a DM axion is detected, and its mass is measured accurately, the two embeddings differ by a factor in the power. Distinguishing the non-standard GUT studied here from canonical unified models requires reaching sufficient precision, usually limited by the determination of the local DM abundance. The Gaia experiment may reduce it to 10% in a near future [32, 33].
Gauge-coupling unification
The prediction (7) is far from the SM or MSSM extrapolation. The running is sensitive to the details of the spectrum. Matching the gauge couplings requires to decrease faster than the SM predicts, requiring light states with hypercharge, as well as new colored states that slow down the evolution of the QCD coupling.
We find two representative options. With the minimal content, , a one-loop bound allowing arbitrary partial multiplets gives only for , and – requires –, that is, a GUT scale close to the Planck scale. Adding one vector-like pair allows unification at with at two loops. For this benchmark there exist new degrees of freedom at low energies, 4 Dirac pairs of and one Dirac at 1 TeV, 4 Dirac at and 5 Weyl at (see Figure 1). In both cases the remaining states are at . Above the large matter content makes the gauge coupling grow fast, and the theory must be completed within a factor of a few of . This is comparable to the species scale [34].
Unlike standard GUTs where the weak mixing angle is a successful prediction of unification, here it is an aspect that requires light vectorlike exotic states, some of which lie around the TeV scale and provide well-motivated collider benchmarks.
The scalar sector
A minimal scalar sector that suffices to break the GUT down to the SM contains a real adjoint , a complex , and a complex singlet (responsible for PQ breaking in models with an axion). is broken to the SM by the vacuum expectation value (vev)
and . The indices label the SM representations that make up the of in eq. (2). A more detailed analysis of the possible scalar sectors and the relations between fermion masses that result is beyond the scope of this letter.
and each contain one doublet which we call and . One combination is eaten by the gauge boson, and the orthogonal one is the candidate for the SM Higgs,
The eaten combination is fixed by the broken generator in the direction acting on the vevs. We find for the vevs in Eq. (10). More generally, for and , . Only has renormalizable Yukawa couplings. Giving all vectorlike spectator masses requires additional scalar representations or higher-dimensional operators; the physical SM Yukawas then depend on the resulting Higgs couplings and light-fermion mixing. The mass of is protected and must be tuned, as in standard GUTs. It would be nice to explore if this class of theories allows for new solutions to the analog of the doublet-triplet splitting problems of standard GUTs, and explore cases where the doublet arises as a pseudo-Goldstone.
Topological defects. — An element of that commutes with is block diagonal, . Invariance of and requires and , invariance of requires , and fixes . The unbroken subgroup of is therefore exactly
for , , hypercharge the abelian factor of , and a direct factor. is connected and . The breaking therefore produces monopoles, , but no topologically stable strings or domain walls, . The same holds for , where is the SM group with its quotient. In this respect the model resembles rather than the , as in the latter monopoles can be attached by strings and domain walls can be bounded by strings [36, 37].
The generator of is the loop , , with magnetic generator
where on a color triplet. It winds one third of the hypercharge circle , twice the minimal loop of the SM group (with ). The lightest ’t Hooft–Polyakov monopole thus carries two Dirac units, , together with color-magnetic charge and no weak charge: the long-range field of the doubly charged monopole of . There is no singly charged soliton, so it is stable within the field theory.
The global form determines the remaining one-form symmetry [10, 38, 39]. For the minimal charge is , and the doubly charged monopole saturates Dirac quantization. For , Dirac quantization allows , but there is no smooth field configuration with that charge. Since smooth monopoles have there is a magnetic one-form symmetry in this case. Completeness of the spectrum [40] suggests that this symmetry is broken in the full gravitational theory and singly charged monopoles exist in the UV.
Proton decay channels. — Gauge bosons live in the adjoint representation. From the branching of the fundamental we see that the adjoint of will contain a but no . This differs from standard and GUTs [1, 5, 6] and resembles flipped [41, 42, 43]. Tree-level gauge boson exchange then generates the same operators as in the latter [44, 45], that is but no operator with . The dominant proton decay modes are and , with , depending on lepton mixing, see [44, 45]. In addition to the gauge-mediated processes there will be (model-dependent) scalar-mediated processes from the representation that breaks down to the SM.
Fractional instantons and the QCD axion. — Because the unified coupling is stronger than usual, GUT instantons are less suppressed than in normal GUTs [46, 47]. Their tree level action is
and might spoil the axion solution to the strong CP problem [48, 49, 50].
For the theory, the instanton has an action . Because QCD sits in at level 3, a unit instanton carries one third of the QCD topological charge. Fractional instantons near the GUT scale contribute to the axion potential as
Here is a model-dependent CP-violating phase indicating that the two contributions are not necessarily aligned. For , requires . The scale follows from the constrained instanton density [51, 52, 53],
where is a calculable constant. If the theory is not asymptotically free above , the integral may be dominated by the smallest instantons, a few . The factor accounts for the fermion zero modes of the ’t Hooft vertex, which must be lifted by fermion masses or Yukawa couplings. Since the content above carries many zero modes, , and hence , is very sensitive to the heavy vector-like spectrum. For , the axion is heavier by and the QCD line moves towards smaller .
Towards flavor unification. — In the model the three families come from three copies of , as in standard GUTs. This is no longer required for larger members of the family (see S2). Anomaly cancellation fixes the fermion content to antisymmetric and conjugate symmetric representations, and for some embeddings this minimal spectrum already contains exactly three SM families (see S2). The first example is with . Its embedding indices are , which give and . In these theories the number of generations is fixed by the embedding rather than by replication, making them promising candidates for gauge and flavor unification.
Conclusions. —We presented the first 4D GUT in which the SM group does not factor through . The embedding is characterized by the integers which control the main predictions. The SM spectrum arises from of and contains only three chiral families and vector-like matter.
The theory is characterized by the prediction and the absence of gauge bosons. Proton decay occurs in a manner similar to flipped models. Minimal models achieve gauge coupling unification provided that there are vector-like states near the TeV scale. In these models, there are fermions transforming as within the reach of HL-LHC and other future colliders, as shown in Figure 1. In the case the gauge group is instead of , the fundamental representation is allowed, and color-singlet states carry half of the electron charge. A particle of charge would then be consistent with this class of models, but charge would rule it out. The monopole spectrum contains only doubly charged smooth monopoles.
A robust prediction is the axion-photon coupling, determined by the ratio , rather than the prediction which holds for GUTs which branch through . We have shown that is the only possible alternative to the standard for GUTs with no chiral exotica, up to models with . In these theories, is a sharper test of unification than proton decay, as a value outside of would rule out GUT models without chiral exotica and . The axion-photon coupling is a factor 1.7 larger if relative to , which is potentially measurable in haloscope experiments in the event of an axion discovery and an improved estimate of the local dark matter density. As in standard GUTs an axion above the QCD line, would exclude it, too [12].
The non-standard embedding also has implications for the strong CP problem. Because , small instantons are less suppressed than standard GUTs. Their contribution depends strongly on the heavy spectrum, and how fermionic zero modes are saturated. These fractional instantons can spoil the solution of the strong CP problem unless they are subdominant or aligned with QCD. They can also easily make the QCD axion much heavier than standard predictions.
The theory that we studied in detail is the first member of an infinite family based on . We have shown that using up to two-index representations, we can reduce all embeddings to the canonical embedding or our new embedding. We have also developed a simple GUT bootstrap, which explores the space possible GUTs and SM embeddings using general constraints, independent of representation. It is exciting to see that some of these, as the theory we briefly discuss in S2, generate three SM families from minimal UV spectra without flavor replication.
Several interesting directions remain. A detailed study of the symmetry breaking pattern classifying intermediate phases, a complete calculation of the impact of small instantons on the QCD axion, as well as neutrino mass generation via the singlets or flavor predictions (e.g. unification) are interesting follow-up directions.
The catalog in the appendix was built with extensive help of ChatGPT Astra, including generating the set of constraints for the bootstrap and the proof for the two-index theorem. The driving question was a group theory classification for the possible QCD axion-photon couplings. The novel embedding was found by the LLM as a counterexample to an theorem. All explicit results and constraints have been checked by the authors. ChatGPT and Claude were used to help edit and check the manuscript; however it is almost entirely written by — the authors.
The research of PA is supported by the U.S. Department of Energy under the grant DE-SC0011702.
Acknowledgments
References
Branching and particle content for
In this appendix we provide the detailed spectrum for the branching in the main text. We label the familiar SM representations of left-handed Weyl fermions as , , , , . In Table S1 we show the quantum numbers of the SM fermions, the spectator multiplets, and other multiplets that appear in the defining blocks used below. Bars denote conjugate representations (e.g. ).
| SM fermions | Spectator multiplets | Defining blocks | |||
| Label | Representation | Label | Representation | Label | Representation |
Table S1. Notation for representations of . The spectator representations , , are real.
The defining embedding in Eq. (2) and its antisymmetric and conjugate symmetric two-index branchings are shown in Table S2. Here and are arbitrary traceless Hermitian matrices in the fundamental representations of and , respectively; is the identity, and a real coefficient multiplying hypercharge.
Table S2. Defining embedding (top) and schematic tensor block decompositions. and are the antisymmetric and symmetric matrices built linearly from the components of in the relevant block bases. Thus the triplet–triplet block contains . Other entries label the SM representation content; repeated labels in different blocks denote independent field copies. Stars are fixed by and , where label the five defining blocks.
For the spectrum , the net spectrum per UV family branches to
This corresponds to 210 = 15 + 195 Weyl components per family, or 630 = 45 + 585 in total. All spectators in , including , admit SM-preserving masses.
Five Families of SU(n) embeddings
The example in the text is a special case of a family of SM embeddings in for , . We will show a proof in section that there are only five families of embedding up to two-index representations. Two of these are the canonical embeddings, and the other three families are based on and are closely related to the embedding in the text. The representations of interest are and . All fermions below are left-handed Weyl fields. The construction continues to arbitrarily large .
Embedding and predictions. The family of embedding is defined by 3 classes labeled by . Setting , , with , the defining representation branches as
Thus counts weak doublets and is the triplet–antitriplet imbalance in . The sum of dimensions of the SM representations dimension is and . The physical hypercharge is defined as and nonabelian fundamental normalization . The embedding indices are
For a PQ symmetry commuting with the unified group and having a nonzero mixed gauge anomaly, and with tree-level gauge matching at ,
Three-generation fermion spectra. The fermion content for the GUT is
The three choices above give nonnegative integral multiplicities. Along with the embeddings, we get five families, shown in Table S3.
| Family | |||||
| I | 5 | ||||
| II | 15 | ||||
| III | |||||
| IV | |||||
| V |
Table S3. Five families of two-index embeddings in , with defining branching and UV Weyl-fermion spectrum . The defining blocks are listed in Table S1. Family II corresponds to the higher-level-embedding found in [16]. Families III, IV and V have , and , and correspond to , respectively. The table displays cores without neutral defining singlets.
The last two rows use different SM embeddings despite identical GUT fermion representations. It is worth emphasizing that families IV and V do not have a 3-fold replication, with the three SM generations coming from one copy of in the UV theory. The smallest anomaly-free spectrum with with this embedding naturally produces 3 generations of the SM; hence, these constitute a model of flavor unification as well! The smallest group for which this works is .
Anomaly cancellation: Normalizing the cubic anomaly of to one, the anomaly for these reps are and , so satisfies anomaly cancellation,
Chirality The branching of and to the SM can be calculated by using
The same formula holds for the symmetric square upon replacing by on both sides. The branchings can be written in terms of the SM representations defined in appendix S1. The expression is,
It is straightforward to check that the net chiral spectrum gives the SM,
where is one SM generation. All other complex representations pair with their conjugates. Therefore those, as well as the real spectators , and all admit a SM-preserving mass. Since for three SM families, we see that can only be embedded at or .
Examples. We list selected examples in this family below.
| Group | Fermion content | ||||
Table 1.
For , we have two inequivalent embeddings, corresponding to and , respectively. These are explicit anomaly-free fermion constructions with the required SM chirality.
A 3-index family with color sextets. Beyond the two-index family, there are new embeddings that use higher UV representations. We present an example of such a family (but we do not have a proof that this is the only possibility). The matter content of these theories is very large and it will probably prove challenging to build realistic models out of these embeddings.
The family of GUT groups is with , , and defining branching , where is the case of Eq. (19). Here is a charged real representation containing a color sextet–antisextet pair. Its indices and predictions, under the same PQ and matching assumptions as above, are
Writing , , and define the irreducible mixed tensors by and . An explicit left-handed fermion spectrum is
It obeys and the exact character identity . The remaining complex representations pair, and pseudoreal spectators have even multiplicity, so all spectators admit SM-preserving masses. The first new member is , with ; the sequence continues through , , . These are not asymptotically free for any . We leave three-index families for future exploration.
A GUT bootstrap
The search for a consistent embedding of the SM into a unified gauge group is a classification task that is subject to several mathematical constraints. In this respect, it resembles a bootstrap program, where we can use these general constraints to explore the GUT landscape, and islands evading these constraints point to interesting possible new structures. Indeed, the case in the text is exactly such an example. In this appendix we derive the relevant constraints on embeddings that contain the SM, and scan over embeddings to find viable GUT models.
Assumptions and the odd-trace constraint
The criteria for the search are the following: we seek a compact, connected, simple gauge group to contain the Standard Model (SM) and admit an anomaly-free, finite Weyl-fermion spectrum with exactly three SM generations and no additional chiral matter. The spectators must admit SM-preserving fermion bilinear masses, and hypercharge must generate a closed circle. For the axion prediction we assume a Peccei–Quinn symmetry commuting with , with a nonzero mixed gauge anomaly.
Complex spectators are paired, real spectators admit Majorana masses, and pseudoreal spectators must occur with even multiplicity. The odd traces for these representations cancel. For an SM Cartan element , the UV spectrum must therefore satisfy
where is one SM generation. In particular, , but .
The fifth trace already gives a strong constraint on the embedding. The nonzero fifth SM trace requires a degree-five invariant of the UV algebra. The invariant-degree list [54], together with the SM rank, leaves only , and . For the latter two, the degree-seven invariant is the degree-five invariant times the unique quadratic form. Matching the SM fifth and seventh traces therefore fixes the relative color, weak and hypercharge normalizations to the canonical ones, giving . For , additional independent invariants prevent this immediate conclusion; the remaining bootstrap is necessary.
The SU(n) bootstrap
In this section we describe a scan over the possible embeddings of the SM in , and general constraints that such an embedding must satisfy. A key feature of the bootstrap is that the constraints do not require enumerating the fermion spectrum , and therefore can rule out broad classes of GUTs.
For , an embedding is specified by the branching of the fundamental representation into color and weak irreducible blocks, with a given hypercharge. This parametrizes both standard and more exotic embeddings:
Here is a color representation, labels the representation with dimension , a positive integer multiplicity, and its hypercharge. Both nonabelian factors must act nontrivially. The two sums impose the dimension budget and traceless hypercharge. may be defined only on the simply connected cover, while should be consistent with the chosen global group. The embedding indices are defined as:
We use and fundamental normalization . Under the stated PQ assumption, the anomalous axion-photon coupling parameter is
The defining traces are determined by this branching alone:
Every trace in an arbitrary UV representation is an invariant polynomial of , and can be expressed in terms of . For example, is a linear combination of and , the only monomials of total degree five. Summing over any UV fermion spectrum and using Eq. (30) gives the constrain equation
The integers , encode the unknown fermion multiplicities. If no choice of these constants makes the polynomial identity hold for every , the embedding is excluded, independently of the UV representation sizes.
We first allow arbitrary rational coefficients to obtain a necessary test; the additional integer constraints imposed by a physical UV spectrum are applied below. More generally, one must be able to write
The sum runs over distinct products of total degree , with and all product lengths .
Additional constraints
In this subsection we enumerate some of the restrictions on GUT completions of the SM that arise from the constraint equations derived in the previous section. We do not give full derivations of all of our results, instead summarizing the constraints. A detailed derivation of these constraints, together with the code used for the scan, will be presented in a forthcoming publication.
Integer constraints. Gauge-anomaly cancellation implies that the coefficients in Eq. (35) must be and , strengthening the degree-five matching condition to:
These integers parametrize the UV spectrum, so are additional data on the embedding. Matching coefficients gives further constraints, including that must be odd. There are seventh- and ninth-degree integral tests that can further restrict the embedding.
Charge realizability. For each proposed branching, the trace constraints must admit solutions with rational hypercharges for the specified blocks.
Character compatibility. The defining branching also determines the full set of eigenvalue of . Here is the matrix by which an element of a cover of the SM group, , acts on . Write , where is the character of . For a given fermion spectrum, is the sum of over its representations, with multiplicity. On the common cover of the SM where is defined, any UV completion must obey
Conjugate matrices have the same character in every UV representation, so violation of Eq. (38) rejects the embedding independently of the UV representation. Indeed, the character of any representation at depends only on the eigenvalues of , which conjugation by transformations does not change. Decomposed under the SM, consists of three generations plus spectators with vanishing (vectorlike pairs, real and pseudoreal SM representations), so . Conjugate and thus give Eq. (38). The scan combines these tests with integral information through degree nine, enumerating nonabelian blocks at fixed while keeping charges symbolic. The tests are necessary but not sufficient to establish an anomaly free UV spectrum without chiral exotics. Embeddings which pass the tests highlight interesting candidates, leading to families of viable models, as we will see below.
Explicit checks used in the scan. At zero hypercharge, the integer constraint can be evaluated using three coefficients computed from the defining branching:
where and is a pure color Cartan element. For the classified triplet embeddings, equals the imbalance . The candidate is rejected unless simultaneously satisfy
In particular, are odd, and forces odd. For the branching in Eq. (2),
Thus . The spectrum realizes these coefficients. Its full branching in Sec. S1 confirms a possible UV construction.
Results and coverage
Theorem (two-index classification). For every , the admissible embeddings and net chiral spectra with UV fermions carrying at most two indices in total, fundamental or antifundamental, consist of the two canonical and three noncanonical branches listed in Table S3 and proved in Sec. S4. These families of admit simple extensions to with additional singlets in the defining block. Together, this is an exhaustive classification. All five branches have exactly three chiral SM generations and vectorlike spectators. The canonical branches (families I & II) have , with for or for [16]. With odd, the three noncanonical branches project to
Here counts neutral singlets added to the charged core.
The family of Eqs. (28)–(29) instead gives
The displayed spectrum uses three-index fermions, while Eq. (42) and the canonical branches exclude any two-index completion. These are constructions, not an exhaustive three-index classification.
| UV tensors | Domain | Established result |
| Up to two indices | All | Embeddings and net chiral spectra classified |
| Up to three indices | , | New family; two-index completion excluded |
| Unrestricted | or | |
| Unrestricted | Triplet-only colored blocks, all |
Table 2.
In the last row, colored defining blocks are weak-singlet triplets or antitriplets; colorless blocks are arbitrary. The unrestricted finite-rank and low-index results include certified finite calculations. They fix the ratio if a completion exists, rather than classify all spectra. The unrestricted rank-two-value theorem remains open.
Beyond the two-index cutoff, UV-chiral sectors can be SM-nonchiral and anomaly-free [16], so classifying embeddings need not classify all UV spectra. The plot therefore distinguishes a complete low-degree classification from particular higher-degree constructions.

Figure S1. Bootstrap coverage and explicit families for . Each cell labels , with the background colors showing the status: ruled out, only , only , both ratios among surviving candidates, or unresolved. Ratio colors are necessary conditions. Solid teal and dashed blue horizontal lines show the exhaustive two-index constructions, starting at . The canonical GUTs are and . Diamonds mark the known three-index family , , admitting a three-index completion but no two-index completion. The vertical dashed line marks the catalogue boundary. Beyond it, only all-rank results are colored; white cells are outside coverage. We note that due to the matching condition in Eq. (6), models with correspond to strongly coupled theories . All theories here contain three chiral SM families and vectorlike matter.
Classification with two-index fermions
In this section we outline the proof for the two-index classification given in Table S3 for every , (up to trivial extensions such as adding singlets to the branching of ). The two-index restriction means that the UV fermions may occupy the fundamental , its antisymmetric square and symmetric square , their conjugates, and the real singlet and adjoint, with arbitrary multiplicities. We require cancellation of the UV gauge anomaly and exactly three SM generations of net chirality, with all other fermions admitting SM-preserving vectorlike masses. The branching of , which specifies the embedding, is initially unrestricted.
The argument first bounds the weak content using a single group element and a weighted anomaly. Ordinary spin addition then excludes defining blocks carrying both color and weak charge. Finally, nonnegative charge multiplicities fix hypercharge. The result is the two canonical and three noncanonical families of Sec. S2.
Weighted Anomaly Matching
We show that a form of weighted anomaly matching can put a sharp restriction on the allowed embedding. Up to additional real and vectorlike components the GUT fermion spectrum can be written as
where , , are signed integers, with a negative coefficient denoting the conjugate representation. It will be useful to define . The anomaly coefficients in these representations are , , , so anomaly cancellation implies
We can get additional constraints using a weighted anomaly equation. For a representation branching to SM,
define
where is the anomaly coefficient and the trace for some element . We will choose the weak group element
Its and eigenspaces in have equal dimension , since and are conjugate in . Thus
Every nontrivial weak multiplet contributes at least one eigenvector to each of these eigenspaces. A defining block with color dimension therefore contributes at least to if it is a weak nonsinglet. In particular, is a positive integer.
Importantly, depends only on the chiral spectrum (with contributions from vectorlike matter vanishing identically), so can be matched with the SM,
where is the fermion content in one generation of SM fermions. The quark doublet contributes and the antiquarks to the weighted anomaly.
Define . With , the color anomaly coefficients and the weak characters at the chosen are integers, so . To extract the weighted anomaly for the two-index representations, consider , with a color generator normalized so . Then , , and . Using the identity,
and differentiating the expression three times, we obtain ( and are related by conjugation),
These weighted anomaly formulae are the usual anomaly formulae with replaced by . Matching the weighted anomaly with the SM gives
Consequently or , independently of .
Restricting the defining blocks
Given the bound on , we can derive constraints on the branching of the fundamental representation.
No simultaneous color and weak charge. In the branching of the fundamental representation such a block contributes at least three to , therefore the bound permits just one copy with color or with all other defining blocks being weak singlets. Such a branching cannot reproduce the SM chiral spectrum. More explicitly, let the weak representation of this block be labelled by its spin, . Then its square in the cross-product contained in or has a rep has weak spin and colour charge, while all other representations have weak spin or . Thus, there is an exotic chiral fermion in the rep left over if representations in the branching of carry both color and weak charges.
Match quark doublets and antiquarks. With hypercharge omitted for the time being, write , where contains the weak singlets and the colorless weak nonsinglets. All colored weak nonsinglets in the branching of the two-index representations arise from the cross product . It appears once in each two-index representation, so its signed coefficient is . Since when hypercharge is omitted, matching net chirality three quark doublets requires
The right-hand side contains only weak doublets, so must contain only doublets. Each contributes one to the eigenspace dimension . Since all colored blocks are weak singlets, is now the ordinary color anomaly. Matching the color representations and multiplicities therefore gives
Choosing the convention , and . We can thus conclude from eq. (32) that gives the weak embedding level, .
We can also fix the unsigned color content as well. Define , which is well-defined since . Some algebra with symmetric and antisymmetric squares gives,
Using Eqs. (45) and (S38), this simplifies to,
Thus, every higher color irrep is also excluded. We can restrict the branching to,
The integers , are odd, since and . By Eq. (57) the only colored blocks in are the triplets and antitriplets of Eq. (58), all of them weak singlets, and since each contributes to the color index in Eq. (32), we have .
Hypercharge and the surviving families
We now assign hypercharges to the defining blocks already found. The strategy is to use sextets that appear in the product representations and require they do not appear in the chiral spectrum, and to use the color and weak singlets.
Cancel the color sextets. Two defining (anti)triplet blocks can produce a (anti)sextet in a two-index representation. In the fermion spectrum, all sextets must be paired with antisextets of opposite hypercharge. Consider the defining branching of the fundamental representation. Each of the triplets and antitriplets can carry different hypercharges. Cancellation of chiral sextets forces all triplets to carry the same hypercharge and antitriplets to carry .
Explicitly, let be the multiplicity for each triplet with hypercharge , so that . Similarly, define as the multiplicity of antitriplets with hypercharge . Choose to be the highest hypercharge for which . Such an exists since . If there is a triplet with hypercharge in the spectrum, by definition of , . Products of these hypercharged triplets/antitriplets with the blocks at then leave unequal numbers of sextets at and antisextets at , which cannot be the case as we do not allow chiral exotica. Therefore, is the highest hypercharge for triplets in the defining branching.
This implies that sextets at and antisextets at arise from the triplets at and antitriplets at . Counting their pairs in the two-index representations gives
where . This means , the total number of colored defining blocks. Every defining triplet therefore has hypercharge , and every defining antitriplet has charge .
The SM quark doublet comes from a triplet–doublet product with total hypercharge , so the defining doublet that pairs with a triplet of charge has charge . The only hypercharges for the defining weak doublets are ; any other hypercharge will leave behind chiral exotica. For , their multiplicities are fixed by asking for 3 quark doublets and cancelling the unwanted representations.
If , the two contributions combine into hypercharge neutral doublets.
Use nonnegative singlet multiplicities. The final constraint comes from the hypercharges of the color and weak singlets. Let be the hypercharges of the defining color-and-weak singlets, and denote such a singlet by . Matching the antiquark charges and , and adding the charge-conjugate equation, gives
The left side counts actual singlets and their conjugates, so all multiplicities are nonnegative. Each negative coefficient on the right must be cancelled by a positive coefficient at the same hypercharge. The only possible positive contributions are at 0 or . Consequently
Summing squared charges in Eq. (61) fixes the singlet contribution to the hypercharge index. Including the colored blocks and weak doublets gives
Since and , a commuting PQ symmetry with nonzero mixed anomaly therefore has
Five families
The constraints above leave behind five families of GUTs, each of which can be consistently realized at the level of fermion content. In the branchings, . The anomaly condition fixes the UV multiplicities in Table S3 from the values of and already determined above. The table lists the gauge group, defining branching , and UV Weyl fermion spectrum together for each family. It displays the cores without neutral singlets in ; their common neutral-singlet extension is given below. The only free integer in the list is for families III–V, with . Of course, the matter content is defined up to vectorlike and real representations.
Families I and II: , . The doublet charges in Eq. (60) are . Their squares give net weak-triplet multiplicity at charge . Its absence requires . Equation Eq. (61) then leaves only neutral singlets. The conditions and give or . These are families I and II in Table S3.
Families III–V: , . Equation (S44) fixes the total number of singlets at charges , and traceless hypercharge fixes their difference. Thus
Nonnegative multiplicities require and . Together with Eq. (55), these give , respectively. Substituting their fixed multiplicities and setting gives families III–V in Table S3, all with group . All displayed multiplicities are nonnegative. Families III–V start at , and , respectively.
Extensions. Every listed can be trivially extended to for any integer : add to the defining branching, interpret , , in the spectrum as representations of the larger group, and add in families I and III, or in families II, IV and V. This gives , and , respectively. The extension preserves UV anomaly cancellation and adds only vectorlike or neutral SM fermions, so the net SM chirality and are unchanged. The general allowed UV spectrum can also contain arbitrary conjugate pairs of the allowed UV representations and arbitrary singlets and adjoints to .
Chirality, masses, and scope Write for net chirality. For the canonical families, . For the noncanonical ones, expanding the two-index representations gives
The bracket is precisely one SM generation.
Complex spectators occur in conjugate pairs. All self-conjugate spectators are real: half-integer weak-spin states have nonzero hypercharge or complex color. Adding neutral defining singlets to adds and neutral singlets to the positive fermion spectrum. Thus pseudoreal spectators occur in even multiplicities as well. All spectators admit SM-preserving mass bilinears. Arbitrary UV-vectorlike pairs and real UV representations within the cutoff can of course also be added, since they already admit UV-invariant masses.
References
- [1]H. Georgi and S. L. Glashow, Unity of All Elementary Particle Forces, Phys. Rev. Lett. 32, 438 (1974).DOI
- [2]J. C. Pati and A. Salam, Lepton Number as the Fourth Color, Phys. Rev. D 10, 275 (1974).
- [3]H. Georgi, H. R. Quinn and S. Weinberg, Hierarchy of Interactions in Unified Gauge Theories, Phys. Rev. Lett. 33, 451 (1974).
- [4]P. Langacker, Grand Unified Theories and Proton Decay, Phys. Rept. 72, 185 (1981).DOI
- [5]H. Georgi, The State of the Art—Gauge Theories, AIP Conf. Proc. 23, 575 (1975).DOI
- [6]H. Fritzsch and P. Minkowski, Unified Interactions of Leptons and Hadrons, Annals Phys. 93, 193 (1975).
- [7]F. Gürsey, P. Ramond and P. Sikivie, A Universal Gauge Theory Model Based on E₆, Phys. Lett. B 60, 177 (1976).DOI
- [8]H. Georgi, Towards a Grand Unified Theory of Flavor, Nucl. Phys. B 156, 126 (1979).DOI
- [9]R. Slansky, Group Theory for Unified Model Building, Phys. Rept. 79, 1 (1981).DOI
- [10]D. Tong, Line Operators in the Standard Model, JHEP 07, 104 (2017), arXiv:1705.01853.DOI
- [11]M. Srednicki, Axion Couplings to Matter. 1. CP Conserving Parts, Nucl. Phys. B 260, 689 (1985).
- [12]P. Agrawal, M. Nee and M. Reig, Axion Couplings in Grand Unified Theories, JHEP 10, 141 (2022), arXiv:2206.07053.DOI
- [13]G. Cacciapaglia, A. Deandrea, K. Kollias and F. Sannino, Grand-unification Theory Atlas: Standard Model and Beyond, arXiv:2507.06368.
- [14]M. M. Anber and S. Y. L. Chan, 2-index chiral gauge theories, arXiv:2308.08052.arxiv.org/abs/2308.08052
- [15]M. Gell-Mann, P. Ramond and R. Slansky, Color Embeddings, Charge Assignments, and Proton Stability in Unified Gauge Theories, Rev. Mod. Phys. 50, 721 (1978).
- [16]R. M. Fonseca, On the chirality of the SM and the fermion content of GUTs, Nucl. Phys. B 897, 757 (2015), arXiv:1504.03695.DOI
- [17]C. Córdova, S. Hong and S. Koren, Noninvertible Peccei- Quinn symmetry and the massless quark solution to the strong CP problem, Phys. Rev. X 15, 031011 (2025), arXiv:2402.12453.
- [18]A. Delgado and S. Koren, Quark-lepton color-flavor unification, Phys. Rev. D 114, 075021 (2026), arXiv:2605.30413.arxiv.org/abs/2605.30413
- [19]T. Li, A. Mansha and R. Sun, Generalized Supersymmetric Pati-Salam Models from Intersecting D6-branes, arXiv:1912.11633.arxiv.org/abs/1912.11633
- [20]B. C. Allanach, B. Gripaios and J. Tooby-Smith, Semisimple extensions of the Standard Model gauge algebra, Phys. Rev. D 104, 035035 (2021) [Erratum: Phys. Rev. D 106, 019901 (2022)], arXiv:2104.14555.DOI
- [21]F. Wilczek and A. Zee, Families from Spinors, Phys. Rev. D 25, 553 (1982).DOI
- [22]J. Bagger and S. Dimopoulos, O(18) Revived: Splitting the Spinor, Nucl. Phys. B 244, 247 (1984).DOI
- [23]G. F. Giudice and S. Raby, A New paradigm for the revival of technicolor theories, Nucl. Phys. B 368, 221 (1992).DOI
- [24]M. Reig, J. W. F. Valle, C. A. Vaquera-Araujo and F. Wilczek, A Model of Comprehensive Unification, Phys. Lett. B 774, 667 (2017), arXiv:1706.03116.DOI
- [25]S. Koren and A. Martin, Technically charged particles at the energy frontier: The SM gauge group and one-form global symmetry, SciPost Phys. 18, 004 (2025), arXiv:2406.17850.DOI
- [26]A. Takenaka et al. [Super-Kamiokande Collaboration], Search for proton decay via p → e+π0 and p → μ+π0 with an enlarged fiducial volume in the Super-Kamiokande I-IV, Phys. Rev. D 102, 112011 (2020), arXiv:2010.16098.DOI
- [27]G. Grilli di Cortona, E. Hardy, J. Pardo Vega and G. Villadoro, The QCD axion, precisely, JHEP 01, 034 (2016), arXiv:1511.02867.DOI
- [28]E. B. Dynkin, Semisimple subalgebras of semisimple Lie algebras, Mat. Sbornik 30, 349 (1952) [Amer. Math. Soc. Transl. Ser. 2 6, 111 (1957)].
- [29]P. H. Ginsparg, Gauge and Gravitational Couplings in Four-Dimensional String Theories, Phys. Lett. B 197, 139 (1987).DOI
- [30]L. Di Luzio, F. Mescia and E. Nardi, Redefining the Axion Window, Phys. Rev. Lett. 118, 031801 (2017), arXiv:1610.07593.DOI
- [31]C. O’Hare, cajohare/AxionLimits: AxionLimits, https://cajohare.github.io/AxionLimits/ (2020), doi:10.5281/zenodo.3932430.DOI
- [32]S. H. Lim, E. Putney, M. R. Buckley and D. Shih, Mapping Dark Matter in the Milky Way using Normalizing Flows and Gaia DR3, arXiv:2305.13358.DOI
- [33]E. Putney, D. Shih, S. H. Lim and M. R. Buckley, Clear-Potential: Revealing Local Dark Matter in Three Dimensions, arXiv:2512.09898.
- [34]G. D. Dvali, Black Holes and Large N Species Solution to the Hierarchy Problem, Fortsch. Phys. 58, 528 (2010), arXiv:0706.2050.DOI
- [35]S. Navas et al. [Particle Data Group], Review of Particle Physics, Phys. Rev. D 110, 030001 (2024).
- [36]D. I. Dunsky, A. Ghoshal, H. Murayama, Y. Sakaki-hara and G. White, GUTs, hybrid topological defects, and gravitational waves, Phys. Rev. D 106, 075030 (2022), arXiv:2111.08750.
- [37]W. Buchmuller, V. Domcke and K. Schmitz, Metastable cosmic strings, JCAP 11, 020 (2023), arXiv:2307.04691.arxiv.org/abs/2307.04691
- [38]D. Gaiotto, A. Kapustin, N. Seiberg and B. Willett, Generalized Global Symmetries, JHEP 02, 172 (2015), arXiv:1412.5148.DOI
- [39]O. Aharony, N. Seiberg and Y. Tachikawa, Reading between the lines of four-dimensional gauge theories, JHEP 08, 115 (2013), arXiv:1305.0318.DOI
- [40]J. Polchinski, Monopoles, duality, and string theory, Int. J. Mod. Phys. A 19S1, 145 (2004), arXiv:hep-th/0304042.DOI
- [41]S. M. Barr, A New Symmetry Breaking Pattern for SO(10) and Proton Decay, Phys. Lett. B 112, 219 (1982).DOI
- [42]J. P. Derendinger, J. E. Kim and D. V. Nanopoulos, Anti-SU(5), Phys. Lett. B 139, 170 (1984).DOI
- [43]I. Antoniadis, J. R. Ellis, J. S. Hagelin and D. V. Nanopoulos, Supersymmetric Flipped SU(5) Re-vitalized, Phys. Lett. B 194, 231 (1987).
- [44]I. Dorsner and P. Fileviez Perez, Distinguishing between SU(5) and flipped SU(5), Phys. Lett. B 605, 391 (2005), arXiv:hep-ph/0409095.DOI
- [45]J. Ellis, M. A. G. Garcia, N. Nagata, D. V. Nanopoulos and K. A. Olive, Proton Decay: Flipped vs Unflipped SU(5), JHEP 05, 021 (2020), arXiv:2003.03285.DOI
- [46]B. Holdom and M. E. Peskin, Raising the Axion Mass, Nucl. Phys. B 208, 397 (1982).DOI
- [47]P. Agrawal and K. Howe, Factoring the Strong CP Problem, JHEP 12, 029 (2018), arXiv:1710.04213.DOI
- [48]M. Kamionkowski and J. March-Russell, Planck scale physics and the Peccei-Quinn mechanism, Phys. Lett. B 282, 137 (1992), arXiv:hep-th/9202003.DOI
- [49]R. Holman, S. D. H. Hsu, T. W. Kephart, E. W. Kolb, R. Watkins and L. M. Widrow, Solutions to the strong CP problem in a world with gravity, Phys. Lett. B 282, 132 (1992), arXiv:hep-ph/9203206.DOI
- [50]S. M. Barr and D. Seckel, Planck scale corrections to axion models, Phys. Rev. D 46, 539 (1992).pubmed.ncbi.nlm.nih.gov/10014967
- [51]I. Affleck, On Constrained Instantons, Nucl. Phys. B 191, 429 (1981).
- [52]J. Fuentes-Martín, M. Reig and A. Vicente, Strong CP problem with low-energy emergent QCD: The 4321 case, Phys. Rev. D 100, 115028 (2019), arXiv:1907.02550.
- [53]C. Csáki, M. Ruhdorfer and Y. Shirman, UV Sensitivity of the Axion Mass from Instantons in Partially Broken Gauge Groups, JHEP 04, 031 (2020), arXiv:1912.02197.arxiv.org/abs/1912.02197
- [54]A. M. Perelomov, A simple construction of basic polynomials invariant under the Weyl group of the simple finite-dimensional Lie algebra, Commun. Math. 28, 301 (2020), doi:10.2478/cm-2020-0024.DOI
