PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

Nonperturbative definition of the standard models

1,2,*
Juven Wang and Xiao-Gang Wen3
1
School of Natural Sciences, Institute for Advanced Study, Princeton, New Jersey 08540, USA
2
Center of Mathematical Sciences and Applications, Harvard University, Cambridge, Massachusetts 02138, USA
3
Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA

(Received 14 June 2019; accepted 21 May 2020; published 17 June 2020)

The standard models contain chiral fermions coupled to gauge theories. It has been a longstanding problem to
give such gauged chiral fermion theories a quantum nonperturbative definition. By classification of quantum
anomalies (including perturbative local anomalies and nonperturbative global anomalies) and symmetric
interacting invertible topological orders via a mathematical cobordism theorem for differentiable and triangulable
manifolds, and by the existence of a symmetric gapped boundary (designed for the mirror sector) on the trivial
symmetric invertible topological orders, we propose that Spin(10) chiral fermion theories with Weyl fermions
in 16-dimensional spinor representations can be defined on a 3 + 1D lattice without fermion doubling, and
subsequently dynamically gauged to be a Spin(10) chiral gauge theory. As a result, the standard models from the
16n-chiral fermion SO(10) grand unification can be defined nonperturbatively via a 3 + 1D local lattice model
of bosons or qubits. Furthermore, we propose that standard models from the 15n-chiral fermion SU(5) Grand
Unification can be regularized by a 3 + 1D local lattice model of fermions.

DOI: 10.1103/PhysRevResearch.2.023356

I. INTRODUCTION AND DEFINITIONS to the longstanding gauged chiral fermion problem: how to
define a chiral fermion theory, with the parity violation [7],
The standard models [1–3], gauge theories with the Lie
coupled to the gauge field, nonperturbatively and in the same
algebra u(1) × su(2) × su(3) in 3 + 1D, coupled to fermions
dimension, as a well-defined quantum theory with a finite-
and bosons, are believed to describe elementary particles.1
dimensional Hilbert space for a finite-size system (for de-
In the standard standard model, there are 15 two-component
tails, see Appendix A 1), but without suffering from fermion
complex Weyl fermions per family. The SU(5) grand unifi-
doublings [8]. In this work, we use the term gauged chiral
cation [5] has 15 complex Weyl fermions per family. There
fermion theory to mean chiral fermion theory coupled to
are also nonstandard standard models, such as the one from
a nondynamical background gauge field. In fact, the gauge
the SO(10) grand unification [6] which has 16 complex Weyl
theories focused on in this article are mostly nondynamical,
fermions per family. But for a long time, the standard models
unless mentioned otherwise.
were only defined via a perturbative expansion, which is
There were many previous pioneer attempts, such as a lat-
known not to converge. So the standard models were not yet
known to be well-defined quantum theories. This is related tice gauge approach [9], Ginsparg-Wilson fermion approach
[10], domain-wall fermion approach [11,12], and overlap-
fermion approach [13,14]. In the Ginsparg-Wilson fermion
approach, the to-be-gauged symmetry is not strictly an on-site
1
Elementary particles include fermions from quarks and leptons, symmetry, but only a quasilocal symmetry (see Definition 2
and bosons from gauge mediators and Higgs particle. Gravitons are and [15–18]; the quasilocal symmetry is still a non-on-site
not yet discovered experimentally. In addition, in our work, we do symmetry), and thus it is very challenging to gauge. (The
not consider any dynamical gravity; we only consider anomalies of Abelian chiral gauge theory is achieved by Ref. [14]; however,
gauge or gravitational nondynamical background fields. The local the non-Abelian case is still an open question). In the domain-
Lie algebra of standard standard models is u(1) × su(2) × su(3), wall fermion approach, we have an extra dimension, where the
but the global structure Lie group can be U(1)×SU(2)×SU(3)
Zq
, where dynamical gauge fields can propagate. The overlap-fermion
q = 1, 2, 3, 6; see a recent overview [4] on this issue. In fact, as we approach is a reformulation of the domain-wall fermion ap-
will show later, for SO(10) and SU(5) grand unifications, it is more proach. The above approaches normally start with a spacetime
natural to study the case q = 6. Also, we denote the d-dimensional Euclidean lattice path integral and implement the Ginsparg-
space and one-dimensional time as d + 1D. Wilson fermion.
*
Corresponding author: [email protected]
In contrast, in our work, we do not formulate a spacetime
lattice path integral or Ginsparg-Wilson fermion. Instead,
Published by the American Physical Society under the terms of the we consider a discretized spatial lattice Hamiltonian with a
Creative Commons Attribution 4.0 International license. Further continuous time, with additional criteria (see Definition 1): (1)
distribution of this work must maintain attribution to the author(s) with a tensor product Hilbert space, (2) with all interaction
and the published article’s title, journal citation, and DOI. terms bounded by a finite range of lattice spacings (called

2643-1564/2020/2(2)/023356(21) 023356-1 Published by the American Physical Society

JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

short-range interactions), and (3) we only discuss on-site Recently, Ref. [23] conjectured a different gapping condi-
symmetries (see Definition 2). Below we refer to our setup tion for the mirror sector:
as a local lattice model.2 Proposition i. Consider a mirror sector in d + 1D with
In this work, we aim to show nontrivial evidence that a symmetry group G. The mirror sector can be gapped out
the gauged chiral fermion problem in both the 16n-fermion without breaking the symmetry G if (1) there exist (possibly
and the 15n-fermion standard models can be solved via a G-symmetry breaking) mass terms that make all the fermions
generalized lattice gauge approach under local lattice model massive, and (2) πn (G/Ggrnd ) = 0 for n  d + 2, where Ggrnd
(Definition 1). In the standard lattice gauge approach, the is the unbroken symmetry group.
fermions do not interact directly. The generalized lattice gauge The above Ref. [23]’s claim is obtained based on the
approach simply adds an extra direct fermion interaction following assumption (not rigorously proven so far):
or an indirect fermion interaction via some Higgs fields. A Proposition ii. A d + 1D G-symmetric nonlinear σ model
generalized lattice gauge approach, called the mirror fermion with topologically trivial target space M = G/Ggrnd [i.e.,
approach, was proposed in 1986 [19,20]. In such an approach, πn (M ) = 0 for n  d + 2] allows a gapped G-symmetric
one starts with a lattice model containing chiral fermions ground state.
(named the chiral sector or the normal sector) and a chiral con- Applying the above propositions, Ref. [23] claimed that
jugated mirror sector (the mirror sector), with a to-be-gauged 3 + 1D Spin(10) chiral fermion theory with Weyl fermions
symmetry acting as an on-site symmetry. Then, one includes in a 16-dimensional spinor representation can be defined via
a proper fermion interaction [21,22] in such a local lattice an interacting local lattice model with a Spin(10) on-site
model, attempting to gap out the mirror sector completely, symmetry which can be gauged.5 The 16-fermion standard
without breaking the on-site symmetry and without affecting model [i.e., SO(10) grand unification or SO(10) grand unified
the low-energy properties of the normal sector. This is the key theory ≡ SO(10) (GUT)] can then be obtained from a 3 + 1D
step, which will be referred to as gapping out the mirror sector Spin(10) chiral gauge theory, coupled to Spin(10) chiral Weyl
without breaking the (to-be-gauged) symmetry. Lastly, one can fermions in the 16-dimensional representation of Spin(10).
gauge the on-site symmetry to obtain a gauged chiral fermion Purpose of our present work. The homotopy group argu-
theory, regularized by a local lattice model.3 ment in Ref. [23] only proposed a sufficient condition. There
Reference [19] proposed a way to gap out the mirror sector are mirror sectors (thus also normal sectors) that do not satisfy
without breaking the symmetry, by introducing composite the condition, but that can still be gapped out without breaking
fermion fields formed by mirror fermion fields, and by adding the symmetry and without altering low-energy physics in the
symmetric mass terms between composite fermion fields and normal sector. In this work, we are going to prescribe a more
the mirror fermion fields to make all those fermion fields general condition, to capture the cases missed by Ref. [23]:
formally massive. However, such a proposal cannot work in Proposition 1. Consider a continuum field theory in d +
general. Even we can make all the fermion fields formally 1D with an internal symmetry group G f .6 If the following
massive; it does not imply we can fully gap out the mirror two conditions hold: (i) if the field theory can be regularized
sector. This is because, even for some models with a pertur- as the low-energy effective theory of a boundary of a gapped
bative local anomaly,4 one can find composite fermion fields local lattice model in one higher dimension d + 2D with a
formed by mirror fermion fields and to make all those fermion bulk on-site symmetry G f , and (ii) if the gapped ground
fields formally massive (see the Appendix of Ref. [23], arXiv state of the bulk lattice model represents a trivial cobordism
version). Also, the extensive studies of the previous mir- invariant in d + 2D, then, the d + 1D field theory can be
ror fermion proposal [24–27] had difficulties to demonstrate regularized as the low-energy effective theory of a local lattice
that interactions can fully gap out the mirror sector without model in the same dimension d + 1D with an on-site internal
breaking the symmetry and without modifying the low-energy symmetry G f .
dynamics of the normal sector. It was pointed out in Ref. [28] Our above statement used the following assumption:
that “attempts to decouple lattice fermion doubles by the Proposition 2. A gapped local lattice model with an on-site
method of Swift and Smit cannot succeed.” Consequently, internal symmetry G f in d + 2D must exist a G f -symmetric
many people gave up the mirror fermion approach. gapped d + 1D boundary (that does not break the G f sym-
metry), if its gapped bulk ground state in d + 2D represents a
trivial cobordism invariant in d + 2D.
2 The existence of a symmetric gapped boundary is based on
For a concrete lattice model, we mostly focus on a spatial lattice
the belief that the bulk with a trivial cobordism invariant can
Hamiltonian. However, our arguments and Propositions 1, 2, and 3
are more general than a Hamiltonian picture; they are also applicable
be smoothly deformed into a symmetric product state without
to quantum field theory and spacetime path integral approaches. closing the gap. The symmetric product state always have a
3
Colloquially, we refer to a lattice regularization the same as a symmetric gapped boundary. See Appendices A and D for
lattice realization. When we say a field theory can be regularized further details.
(on the boundary of a lattice in one higher dimension), we also mean
a field theory can be realized.
4 5
We overview the concepts of anomalies, including perturbative In this work, a local lattice model is a lattice model of bosons
local anomaly and nonperturbative global anomaly in Appendix A. and/or fermions with short-range interactions and a tensor-product
In the main text, however, we purposefully reduce the information structured Hilbert space; see Definition 1.
6
on anomalies, but focus on the mathematically very well-defined An internal symmetry may or may not be an on-site symmetry.
concepts called the cobordism theory. But an on-site symmetry must be an internal symmetry.

023356-2
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

To obtain Proposition 1, we have to apply Proposition is a one-to-one correspondence (isomorphism “∼ =”) between
2. We first regularize the field theory as a boundary (also the following two well-defined “mathematical objects” (these
referred as the normal sector) of the gapped lattice model “objects” turn out to form the Abelian group structures):
in one higher dimension, then assume the lattice model has ⎧ ⎫
⎪
⎪ Deformation classes of the reflection positive ⎪ ⎪
a finite thickness, and make the boundary on the other side ⎪
⎨ ⎪
⎬
(also referred as the mirror sector) to be the symmetric gapped D-dimensional extended invertible
boundary ensured by Proposition 2. ⎪
⎪ topological field theories (iTQFTs) with ⎪
⎪
⎪
⎩ ⎪
⎭
Using the above statements, we will show that a 3 + 1D G
symmetry group G = spacetime
×G f
Nshared
Spin(10) chiral fermion theory with Weyl fermions in a
16-dimensional spinor representation can be defined via an ∼
= [MT (G),  D+1 IZ]tors . (2)
interacting local lattice model with a Spin(10) on-site internal
The MT (G) is the Madsen-Tillmann spectrum [43] of the
symmetry which can be gauged. In addition, we will show that
G group, the  is the suspension, the IZ is the Anderson
a 3 + 1D SU(5) chiral fermion theory with Weyl fermions in
dual spectrum, and  D+1 IZ is the D + 1-th suspension of
five-dimensional and 10-dimensional representations can be
the spectrum. The tors means taking only the finite group
defined via an interacting local lattice model with an SU(5)
sector (i.e., the torsion group). The right-hand side is the
on-site symmetry which can be gauged.
torsion subgroup of the homotopy classes of maps from a
Lastly, we remark that to fully characterize the global
Thom-Madsen-Tillmann spectrum [43,44] to a shift of the
symmetry in a fermion system, we need to specify the full
Anderson dual to the sphere spectrum. The spacetime sym-
internal global symmetry group G f and how the fermion
metry Gspacetime and the internal symmetry G f , mod out the
number parity Z2f is embedded in G f . So we can denote the shared common normal subgroup Nshared , is combined to a G
fermion symmetry as G f ⊃ Z2f . In our case, the full internal structure:
symmetry is actually G f = Spin(10), while SO(10) is the Gspacetime × G f
quotient group Spin(10)/Z2f = SO(10). So, in this work, we G= . (3)
Nshared
use the following name: a Spin(10) chiral fermion model
[rather than an SO(10) chiral fermion model which was This also means the pertinent iTQFTs of (2) are defined on
sometimes used by others]. manifolds with G structure.
In condensed-matter physics, this roughly means the fol-
lowing:
II. COBORDISM THEORY AND SYMMETRIC
Proposition 3. There is a one-to-one correspondence [29]
GAPPED BOUNDARY
between “the invertible gapped states with an internal sym-
Let us first explain the cobordism theory used in Proposi- metry G f (including higher symmetries [30,33])” that can
tions 1 and 2. Based on a theorem of Freed-Hopkin [29] and be regularized on a lattice with G f realized as an on-site
an extended generalization [30–32] (including higher symme- symmetry [45] in its own dimensions and “the group elements
tries [33–40]), there is a one-to-one correspondence between as the corresponding generators in a cobordism group for the
“the deformation class of invertible topological quantum field internal symmetry G f ,” at least in lower dimensions.9
theories (iTQFTs7 ) [41,42] with symmetry (including higher
symmetries)” and “a cobordism group.”8 More precisely, there
symmetric iTQFT classified by the cobordant properties of smooth
manifolds have a triangulation (thus a lattice regularization) on a sim-
7
It is called an invertible TQFT because its partition function plicial complex [thus an ultraviolet (UV) competition on a lattice].
Z(M D ) on any closed manifold M D must have its absolute value This implies a correspondence between “the symmetric iTQFTs (on
|Z(M D )| = 1, namely, Z(M D ) = eiθ can only be a complex phase. On smooth manifolds)” and “the symmetric invertible topological orders
a closed spatial manifold M D−1 , it always has a single ground state (on triangulable manifolds)” for D  4. This leads to our application
Z(M D−1 × S 1 ) = 1 with no topological ground-state degeneracy. of this mathematical fact on the lattice regularization of symmetric
Thus, Z(M ) = eiθ has an inverted phase Z† (M ) = e−iθ that can be iTQFTs and symmetric invertible topological orders for various
defined as its complex conjugated iTQFT. The combined iTQFT standard models of particle physics. In this work, we only focus
Z(M ) · Z† (M ) = 1 is the trivial iTQFT (i.e., the trivial gapped vac- on the smooth differentiable (DIFF) manifolds and their associated
uum). all possible iTQFTs. The tools we use in either case would be a
8
By all symmetric iTQFTs, their classifications and characteriza- certain version of cobordism theory suitable for a specific category
tions depend on the category of manifolds that can detect them. of manifolds.
9
The categories of manifolds can be TOP (topological manifolds), We clarify that before gauging, the G f symmetry discussed in our
PL (piecewise linear manifolds), or DIFF (differentiable thus equiva- setup must be an on-site internal symmetry of the lattice model (see
lently smooth manifolds), etc. These categories are different and they Appendix A). Certainly, this does not exclude the possibility that the
are related by the following inclusions: lattice model may have a larger symmetry. We stress that the G f in
the cobordism calculations is the on-site internal symmetry for
TOP ⊇ PL ⊇ DIFF. (1)
the G f -symmetric deformation class of the Hamiltonians. (4)
In contrast, triangulable manifolds are smooth manifolds at least
for dimensions up to D = 4 (i.e., the “if and only if” statement Thus, we consider many G f -symmetric Hamiltonians under the G f -
is true below D  4). The concept of piecewise linear (PL) and symmetric preserving deformations. The on-site G f always needs to
smooth DIFF structures are equivalent in dimensions D  6. Thus all be preserved in order to be gauged later. For example, we choose

023356-3
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

There is a logic gap here to establish Proposition 3 since, by invertible topological order with symmetry.”10 In other words,
(2), we only know that there is a one-to-one correspondence there exists a symmetry-preserving local unitary transforma-
between “the iTQFTs with symmetry” and “the cobordism tion that deforms the “trivial lattice invertible topological
invariants from a cobordism group.” We do not yet know order with symmetry” into a “trivial tensor product state with
if there is a one-to-one correspondence between “the lattice symmetry” [54], where its gapped symmetric boundary can
invertible topological order with symmetry” and “the iTQFTs always be constructed. (We provide more steps along with
with symmetry.” In particular, we do not mathematically these logical arguments in Appendix A 2). Crucially, this is
and rigorously prove how to construct a lattice Hamiltonian precisely why the cobordism approach allows us to obtain the
realization for each iTQFT with symmetry classified by a gapping condition for the mirror sector.
cobordism group. (We remark that some of the “lattice in- Proposition 1 can be obtained from Proposition 2, if we
vertible topological orders with symmetry on a lattice” are can show that the normal sector or the mirror sector can be
also called the symmetry-protected topological/trivial states regularized as some boundary states of a gapped local lattice
(SPTs) [15,16,46,47], if they can deform to a trivial tensor model. We will provide such a local lattice model construction
product state under local unitary transformations after ex- for the Spin(10) chiral fermion theory in Sec. III and in
plicitly breaking the symmetry). Regardless of a logic gap in Appendix B as an example.
the rigorous mathematical sense, the broad literature suggests Although we propose Propositions 1, 2, and 3, we do
strong physical evidence of the following: not require the complete versions of all these propositions
(a) The classification of iTQFT [29,30,48–50] so far to establish our claim of a local lattice model with a chiral
matches with the classification of lattice invertible topological fermion low-energy spectrum. We only require the weaker
orders and lattice SPTs [47,51–53]. Many such iTQFTs can Proposition 1. Let us clarify:
thus be constructed on the lattice Hamiltonian. (i) Proposition 1’s “the existence of a fully gapped bound-
(b) Moreover, in (2), we only focus on iTQFTs definable ary” is a static statement. On the other hand, “the gapless sec-
on differentiable and triangulable manifolds, and thus those tor can be fully gapped out without breaking the symmetry”
iTQFTs may be regularized by the same lattice from the is a dynamic statement, more challenging than Proposition 1.
simplicial complex of triangulable manifolds. But the two statements are related; their detailed relations are
In summary, based on the support of (a) and (b), below given in Appendices A 2 and D, based on physical intuitions
we propose and assume that a refined and rigorous version of of phase boundaries and quantum phase transitions. In fact,
Proposition 3 is true. we only require the weaker static statement in Proposition
Proposition 2 can be obtained from Proposition 3. There 1’s “the existence of a fully gapped boundary” in order to
can be two kinds of gapped fermion systems on a lattice: establish the gapped mirror sector.
those with topological excitations (which may be fractional- (ii) To use Proposition 3, we only require a local lat-
ized) and those without topological excitations (i.e., all the tice construction for the cobordism class whose boundary
excitations correspond to the original fermions or bosons). gives rise to the normal sector or mirror sector. To establish
By definition, the gapped states with topological excitations Spin(10), Spin(18), and SU(5) chiral fermion theories, we
are the lattice noninvertible topological orders. The gapped only require a local lattice construction of the trivial cobor-
states without topological excitations are the “lattice invertible dism class (the identity element 0 in the cobordism group).
topological orders with symmetry.” According to Proposition They happen to be a trivial bulk gapped insulator in which we
3, if a “lattice invertible topological order with symmetry” has certainly can construct their local lattice model with a gapless
a trivial cobordism invariant, then it must be a “trivial lattice normal sector on the boundary (Sec. III).
In the following sections, we also provide the physics
interpretations of the classifications of all 4 + 1D iTQFTs
whose boundaries are associated with the 3 + 1D Spin(10)
an internal symmetry group G f = Spin(10) for SO(10) GUTs. How- and Spin(18) chiral fermion theories [for SO(10) and SO(18)
ever, we point out that the full symmetry group G used in cobordism GUTs] in Sec. III,11 and the 3 + 1D SU(5) chiral fermion
calculations also includes the emergent spacetime symmetry at low- theories [for SU(5) GUTs] in Sec. IV. We relegate the
energy infrared (IR) as Gspacetime = Spin(D) (for a D-dimensional
Euclidean spacetime). Thus, for the Spin(10) fermion model, in
Eq. (3), we have N = Z2f , so G = ( (Spin(D)×Spin(10))
f ). Given a lattice
Z2
model, there can be a larger symmetry Gonsite ⊃ G f . Such as some 10
A “trivial lattice invertible topological order with symmetry”
lattice models in Sec. III and in Appendix B, we have means the “trivial gapped vacuum with symmetry” in quantum field
theory, or the “symmetric gapped direct product state” in condensed
Gonsite ⊃ U(16N f ) ⊃ U(16) ⊃ G f = Spin(10), matter.
11
for some flavor number N f . There are also other space-group symme- To be precise, in order to embed the standard standard-model-like
G ×G f
tries on a lattice, say Glattice,space , while Glattice,space is typically smaller spacetime-and-internal symmetry group with this G = spacetime N shared
Spin(D)×Spin(10)
than the emergent Gspacetime , so usually structure (3) to the SO(10) grand unification’s f , it
Z2

Glattice,space ⊂ Gspacetime . is natural to consider an alternative standard standard model

spacetime-and-internal group Spin(D) × U(1)×SU(2)×SU(3)
Z6
, while their
Overall, all these “symmetries” are not crucial to our discussion; the gauge Lie algebra is still u(1) × su(2) × su(3). Here, Spin(10) and
U(1)×SU(2)×SU(3)
only key symmetries are the on-site internal G f symmetry, and the Z6
are their gauge groups, respectively. See more dis-
overall G. cussions in footnote 18.

023356-4
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

mathematical calculation details on algebraic topology in trivial cobordism class (in 1, 2, and 3) [29,30], which belongs
Appendix E. (See, also, Ref. [55]). to the zeroth class in the classification.
If so, the 4 + 1D hopping model can have a symmet-
III. Spin(N) CHIRAL FERMION THEORY, AND SO(10) AND ric gapped boundary, and the 3 + 1D Spin(10) chiral Weyl
SO(18) GRAND UNIFICATION fermions (5) for the mirror sector can be gapped by inter-
actions without breaking the symmetry by Propositions 1
We now construct a local lattice Hamiltonian model. A 3 + and 2.
1D two-component Weyl fermion described by a Hamiltonian To show that the 4 + 1D hopping model gives rise to a
(Model 1 defined in Appendix A 1), trivial Spin(10) SPT state in interacting fermion systems, we
H = ψ † i σ i ∂i ψ, σ 1,2,3 are Pauli matrices, (5) use a recent conjectured complete classification of interacting
fermionic invertible topological orders [29,30,49,50,61–63]
can be regularized on the boundary of a fermion hopping with on-site symmetry, via a twisted version of the spin
model on a 4D spatial cubic lattice with a Hamiltonian op- cobordism theory of Freed-Hopkin [29]. This classification
erator [23], includes all known interacting fermionic SPT states and all

 known interacting fermionic invertible topological orders on
Ĥhop = †
tiabj ĉa,i ĉb, j + H.c. , (6) a lattice [52,59,60,64].
ij We first note that for fermions with the full symmetry
which has four fermion orbitals (a, b = 1, . . . , 4) per site (i, j G f ⊃ Z2f in the D spacetime dimensions, they transform
Spin(D)×G f
for sites). The H.c. contains the Hermitian conjugate term. The as G = f under the combined spacetime symmetry
Z2
4 × 4 hopping matrices ti j are given by
Gspacetime = Spin(D) rotation and the internal G f transforma-
H4D (k1 , k2 , k3 , k4 ) tion, where a double-counted fermion parity symmetry Z2f is
= 2[ 1 sin(k1 ) +  2 sin(k2 ) +  3 sin(k3 ) +  4 sin(k4 )] mod out. This shared normal subgroup Z2f is due to the fact
that rotating a fermion by 2π in the spacetime (namely, the
+ 2 5 [cos(k1 ) + cos(k2 ) + cos(k3 ) + cos(k4 ) − 3] (7) spin statistics) gives rise to the same fermion parity minus sign
in the momentum k space, where  1 = σ 1 ⊗ σ 3 ,  2 = σ 2 ⊗ for the fermion operator ψ → −ψ.
σ 3 ,  3 = σ 3 ⊗ σ 1 ,  4 = σ 0 ⊗ σ 2 , and  5 = σ 0 ⊗ σ 3 , which To classify the iTQFT whose boundary can have a 3 + 1D
obey { i ,  j } = 2δi j . If the 4D lattice is formed by two Spin(10) chiral fermion theory, we focus on the following
layers of 3D cubic lattices, the one-body Hamiltonian in the cobordism group:
(k1 , k2 , k3 ) space is given by the following 8 × 8 matrix:  
 [Spin(D = 5) × Spin(10)]
M1 M2
[Spin(D=5)×Spin(10)] ≡ TPD=5
D=5
.
H3D (k1 , k2 , k3 ) = , where Z
f Z2f
M2† M1 2
(9)
M1 = 2[ 1 sin(k1 ) +  2 sin(k2 ) +  3 sin(k3 )] More generally, we find that 4 + 1D fermionic invertible
topological orders with G f = Spin(N ) ⊃ Z2f on-site global
+ 2 5 [cos(k1 ) + cos(k2 ) + cos(k3 ) − 3],
symmetry for N  7 are classified by the fifth cobordism
M2 = −i  4 +  5 . (8) group [30]:
One can directly check that the above 3D fermion hopping
D=5
(Spin(D=5)×Spin(N )) = Z2 , N  7. (10)
model gives rise to a two-component massless complex Weyl Z
2
f

fermion on each of the two 3D surfaces of the 4D lattice. The

Weyl fermion on one boundary is a left-hand Weyl fermion Note that we define the cobordism group, classifying symmet-
and on the other boundary is a right-hand Weyl fermion. We ric fermionic invertible topological orders, as
have a similar result when the 4D lattice is formed by many
D
G ≡
Gspacetime ×G f
D
layers of 3D cubic lattices. ( Nshared )
The 16 copies of the local lattice model (7) give rise ≡ TPD (G) ≡ [MT (G),  D+1 IZ], (11)
to the 3 + 1D Weyl fermions in the 16-dimensional spinor
representation of the Spin(10) on the lattice boundary’s low- which stands for the homotopy classes of maps from the
energy spectrum. The ground state of the 4 + 1D hopping Thom-Madsen-Tillmann spectrum [43,44] MT (G) to the
model is D + 1-th suspension of the Anderson dual spectrum  D+1 IZ.
(a) a “lattice invertible topological state (invertible topolog- Our notations follow Refs. [29,30,50] and [65]: TP abbrevi-
ical order whose low energy is an iTQFT) with a Spin(10) ⊃ ates “topological phases” classifying the symmetric invert-
Z2f symmetry,” since it has no nontrivial topological excita- ible topological orders (or invertible topological quantum
tions; field theories), and Nshared is the shared normal subgroup of
(b) a lattice nontrivial 4 + 1D Spin(10) noninteracting free Gspacetime and G f .
fermionic SPT state [56–58], which belongs to the 16-th class The cobordism group of topological phases (TP) defined in
(or the 16n-th) in the Z classification. [29] as TPD (G) classifies the deformation classes of reflection
But such a state may correspond to positive invertible d-dimensional extended topological field
(c) a trivial state for Spin(10) SPT state in the lattice theories with symmetry group GD . The cobordism group
interacting fermionic SPT systems [49,52,59–61] and to a TPD (G) ≡
D G and the bordism group
D are related by a
G

023356-5
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

short exact sequence, one possible candidate of the 3 + 1D anomaly for interacting

 fermion systems with a Spin(N ) symmetry (N  7), which
0 → Ext
GD , Z → TPD (G)
1
is a nonperturbative global mixed gauge-gravity (i.e., gauge-

G 
G → Hom
D+1 , Z → 0,
≡
D (12) diffeomorphism) anomaly characterized by (14). Thus, M 5
may not be a spin manifold [note that a spin manifold requires
with Ext denotes the extension functor; see Appendix E. w2 (T M ) = 0], generally called a nonspin manifold.
In contrast, we do not define the cobordism group as the We can detect the 4 + 1D cobordism invariant

usual definition of Pontryagin dual of the torsion subgroup (≡ eiπ M 5 w2 (T M )w3 (T M ) for the 4 + 1D invertible fermionic
tors) of the bordism group
G D as the homomorphism (Hom) topological order by study its boundary state. In particular,
map to U(1): if the 4 + 1D state has a boundary described by 3 + 1D
  Spin(N ) chiral Weyl fermion theory, then we can detect the
Hom
G,tors
D , U(1) , (13)
4 + 1D cobordism invariant via the Spin(N ) representation
although the torsion (i.e., finite group) sectors of (11) and of the chiral Weyl fermions on the boundary. Here we use
(13) are equivalent. Mathematical details for the above result a fact that the 4 + 1D cobordism invariant can be detected
are presented in Refs. [30,50,55].12 We classify the deforma- by restricting to a SU(2) = Spin(3) subgroup of Spin(N )
tion classes of invertible topological quantum field theories [67]: Let n j be the number of isospin- j representations of
(more precisely, the reflection positive invertible extended SU(2) = Spin(3) ⊆ Spin(N ) for 3 + 1D boundary chiral
topological field theories) via
D G , by classifying the cobor- Weyl fermions; then the 4 + 1D cobordism invariant
dant differentiable and triangulable manifolds with a stable G eiπ M 5 w2 (T M )w3 (T M ) is absent if
structure, via associating them to the homotopy groups of the ∞ ∞

Thom-Madsen-Tillmann spectra [43,44], thanks to a theorem n2r+ 1 ∈ Zeven , n4r+ 3 ∈ Zeven . (15)
2 2
in Ref. [29]. r=0 r=0
To be precise, here the spin cobordism theory is believed
To see how the representation of Spin(N ) reduces to the
to completely classify all the fermionic iTQFTs. By applying
representations of SU(2) = Spin(3), let us describe the repre-
this spin cobordism theory, we classify 4 + 1D Spin(N ) SPT
sentation of Spin(N ) [the spinor representation of Spin(N )],
states and 4 + 1D Spin(N ) symmetric invertible fermionic
assuming N = even. We first introduce γ -matrices γa , a =
topological orders. In fact, in this context, the 3 + 1D Spin(N )
1, . . . , N:
fermion theories already include all possible 3 + 1D Spin(N )
chiral fermion theories that we need. To this end, we will es- γ2k−1 = σ
 ⊗ ·
0
· · ⊗ σ 0 ⊗σ 1 ⊗ σ
 ⊗ ·
3
· · ⊗ σ 3 ,
pecially focus on the 3 + 1D Spin(10) chiral fermion theories N
−k σ 0 ’s k−1 σ 3 ’s
2
with Weyl fermions in a 16-dimensional spinor representation. (16)
The above Z2 classification in (10) implies that there is γ2k = σ
 ⊗ ·
0
· · ⊗ σ 0 ⊗σ 2 ⊗ σ
 ⊗ ·
3
· · ⊗ σ 3 ,
only one nontrivial 4 + 1D invertible fermionic topological N
−k σ 0 ’s k−1 σ 3 ’s
order with a Spin(N ) on-site symmetry. We find that such 2

a topological phase is characterized by a five-dimensional k = 1, . . . , N2 , which satisfy {γa , γb } = 2δab and γa† = γa .
topological invariant [30] written in terms of a bulk partition Here, σ 0 is the 2 × 2 identity matrix and σ l with l = 1, 2, 3
function on a 5-manifold M 5 , are the Pauli matrices. The N (N−1) Hermitian matrices γab =
 2
Z = eiπ M5 w2 (T M )∪w3 (T M )
, (14)
i
2 a
[γ , γb ] = iγa γb , a <b, generate a 2N/2 -dimensional repre-
sentation of Spin(N ). The above 2N/2 -dimensional represen-
where wn (T M ) is the nth-Stiefel-Whitney class for the tan- tation is reducible. To obtain an irreducible representation, we
gent bundle of 4 + 1D spacetime manifold M 5 , and the ∪ is introduce
the cup product (which we may omit writing ∪) [68].
We note that on M 5 , we have a Spin(D=5)×Spin(N f
) γFIVE = (−i)N/2 γ1 · · · γN = σ 3 ⊗ ·
· · ⊗ σ 3 . (17)
Z2
connection—a mixed gravitational and gauge connection, N
2 σ 3 ’s
rather than a pure gravitational Spin(D = 5) connection,
We have (γFIVE )2 = 1, its trace Tr(γFIVE ) = 0, and
such that w2 (T M ) = w2 (VSO(N ) ) and w3 (T M ) = w3 (VSO(N ) ),
{γFIVE , γa } = [γFIVE , γab ] = 0. This allows us to obtain
where wn (VSO(N ) ) is the nth-Stiefel-Whitney class for an
two 2N/2−1 -dimensional irreducible representations: one for
SO(N ) gauge bundle.13 Namely, we find that there is only
γFIVE = 1 and the other for γFIVE = −1.
Now, let us consider an SU(2) = Spin(3) subgroup
of Spin(N ), generated by γ12 = I ⊗ σ 0 ⊗ σ 3 , γ23 = I ⊗
12
In contrast, Ref. [66] computes a different bordism group σ 1 ⊗ σ 1 , and γ31 = I ⊗ σ 1 ⊗ σ 2 . We see that the 2N/2−1 -
[Spin(D=5)×Spin(10)]

D=5 = 0. Instead, we study the bordism group dimensional representation of Spin(N ) becomes 2N/2−2
f
[Spin(D=5)×Spin(10)]/Z isospin-1/2 representations of SU(2).

D=5 2
= Z2 , whose manifold generator can detect the
new SU(2) anomaly [67]. Summarizing the above results, we see that the 2N/2−1
13
In the context of anomalies (see Appendix A for details), the copies of 4 + 1D hopping model (7) formed by many (but
boundary of this 4 + 1D Spin(N )-SPT state may have a mixed finite) layers of 3D cubic lattices has a 3 + 1D boundary
anomaly of SO(N )-gauge bundle and spacetime geometry/gravity, chiral Weyl fermion in the 2N/2−1 -dimensional representation
and we can use the this 3 + 1D anomaly on the boundary to detect of Spin(N ) on one boundary (the normal sector), and a conju-
the bulk invertible topological order. gate 3 + 1D chiral Weyl fermion on the other boundary (the

023356-6
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

mirror sector). For an even N  8, the 3 + 1D boundary chiral We remark that, in fact, for N = 3, the 4 + 1D fermionic
Weyl fermions only reduce to an even number of isospin-1/2 invertible topological orders with a Spin(3)=SU(2) internal
representations,
 and, according to (15), the 4 + 1D cobordism global symmetry are classified by the cobordism group of
invariant eiπ M 5 w2 (T M )w3 (T M ) is absent. Thus the corresponding [Spin(D = 5) × Spin(N = 3)]/Z2f [30,67]:
4 + 1D bulk state is a trivial fermionic iTQFT (i.e., the
identity element as the trivial cobordism class in Propositions [Spin(D=5)×Spin(3)] =
[Spin(D=5)×SU(2)] = (Z2 ) .

D=5 D=5 2
(19)
f f
Z Z
1 and 2) with a Spin(N) symmetry. 2 2

In this case, the mirror sector can be chosen as a Spin(N )- The corresponding cobordism invariant is given by
symmetric gapped boundary (Proposition 1), or the mirror  
Arf 
sector can be gapped out without breaking the Spin(N ) sym- Z = eiαπ M5 w3 (T M ) iβπ
e M5 w2 (T M )w3 (T M )
. (20)
metry by introducing a proper symmetric fermion interaction Here, Arf is the Arf invariant [69], which characterizes the
on the boundary (Proposition 2). Since the 4 + 1D hopping 1 + 1D fermionic chain whose open ends host Majorana zero
model has many layers and the 3 + 1D boundary massless modes [70]. This 1 + 1D fermionic chain is also known as the
chiral Weyl fermion has no symmetric relevant deformation Kitaev chain [70], whose low-energy physics is governed by a
operators [in the renormalization group sense, e.g., there is 1 + 1D invertible fermionic topological order. The 
w3 (T M ) is
no Spin(N ) symmetric mass term], the symmetric interaction a twisted version of the third-Stiefel-Whitney class w3 (T M ).
in the mirror sector on one boundary will not affect the low- The above cobordism invariant can be detected by the SU(2)
energy dynamics of the 3 + 1D massless chiral Weyl fermion representations of 3 + 1D boundary chiral Weyl fermions, and
in the normal sector on the other boundary.14 α, β in (20) are given by [67]
Now we apply a well-known lattice method. Finite- ∞ ∞
width/layer lattice dimensional reduction:
α= n2r+ 1 mod 2, β= n4r+ 3 mod 2. (21)
2 2
An n + 1D lattice model with finite layers along one r=0 r=0

extra direction (a finite width w) can be dimensionally In this work, we only suggest that there exists a symmetric
reduced to an nD lattice model via absorbing the degrees short-range nonperturbative interaction that can fully gap out
of freedom along w to the orbital in nD. (18) the mirror sector without breaking the Spin(10) symmetry.16
Our approach only proves the symmetric gapped boundary
Thus, by (18), the 4 + 1D hopping model with finite layers exists (via Propositions 1 and 2), but does not provide a
can be viewed as a 3 + 1D lattice model with finite orbitals prescription to design such an interaction. The approach in
per site, and the 3 + 1D Spin(N ) chiral Weyl fermion theory Refs. [23,71] proposes a design: The interaction in the mirror
in the 2N/2−1 -dimensional representation can be regularized sector is given by the smooth orientation fluctuations of Higgs
by a lattice model in the same dimension without breaking the field (thus beyond the Higgs mechanism [18,72]), where a
Spin(N ) symmetry for even N  8. constant orientation will gap out all the mirror fermions. But
In particular, for N = 10, the Spin(10) chiral fermion the validity of the design requires confirmation by numerical
theory with Weyl fermions in a 16-dimensional spinor rep-
resentation [similarly, for Spin(18), chiral fermion theory in a
256-dimensional spinor representation] can be regularized by known as the hard gauge, such that the outcome gauge theory does
a local lattice fermion model in the same dimension. After not have a tensor product Hilbert space (Definition 1), and thus it is
regularizing the Spin(10) chiral fermion theory as a lattice not a local lattice model that we aim for. However, we can further
fermion model in the same dimension (3 + 1D) with an on- maintain a tensor product Hilbert space (Definition 1) by designing
site Spin(10) symmetry, we can gauge the on-site Spin(10) the soft gauge. We relegate the details of the soft gauge via a local
symmetry to obtain a gauged Spin(10) chiral fermion theory,15 lattice model in Appendix B. See discussions on hard gauge and soft
again regularized by a lattice model in the same dimension. gauge in Ref. [18].
16
There exists such a symmetric gapping interaction preserving
Spin(10). Moreover, for 16 chiral Weyl fermions at IR, there can be
14
In the previous paragraph, we had determined that the 4 + 1D a U(16) global symmetry, with U(16) ⊃ Spin(10). It is possible that
hopping model (7) without higher-order interactions,whose bound- additional constraints happen on what interactions we can engineer
U(16)
ary hosts a 3 + 1D chiral Weyl fermion in the 2N/2−1 -dimensional in the quotient space Spin(10) (also a homogeneous space) without
representation of Spin(N ), has a trivial cobordism class in the bulk. breaking Spin(10). We provide some further guidelines for designing
Readers may wonder whether the bulk’s cobordism class would interactions: (a) To gap the mirror sector by adding nonperturbative
change under the interactions that we required. The answer is no. To lattice-scale interactions on the boundary, we may need to look for a
recall, our setup follows: (i) Bulk does not include nonperturbative larger symmetry G+ than Spin(10), such that
interactions. Bulk only allows small perturbative interactions if any. Spin(10) ⊆ G+ ⊆ U(16).
So the bulk gap does not close, and can never be closed. (ii) Only on
the boundary, we can add “arbitrary Spin(N) preserving interactions” This G+ can be useful for constructing exactly solvable and inte-
(both small perturbative or large nonperturbative interactions). In grable models. (b) However, we can further weakly break G+ down
summary, since the bulk gap does not close, the bulk phase remains to Spin(10), by small perturbative interactions that are not merely
the same trivial cobordism class, which stays valid before and after irrelevant perturbations in the renormalization group (RG) and field
adding boundary interactions. theory sense, but also with small lattice-scale couplings. Thus we can
15
To gauge the on-site symmetry, one way is by inserting gauge break the redundant symmetry outside Spin(10) without changing the
variables on the one-dimensional links between local sites. This is quantum dynamics.

023356-7
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

simulations. A first step is taken in Ref. [71] for a 1 + 1D symmetry: introducing dynamical gauge link variables be-
system. In such a design, crucially, the mass of the mirror tween local sites (e.g., dynamically sum over gauge inequiv-
fermions induced by the Higgs field must be comparable with alent configurations in the partition function)—this is a hard-
the fermion bandwidth. Some other gapping-mirror-fermion gauge model but not a local lattice model; see footnote 15.
approaches have also been proposed recently [17,72–77]. We can further apply the soft-gauge method [18] to obtain
Many previous calculations [26,27] checking the mirror a local lattice model; see Appendix B. We emphasize if all
fermion approach choose an induced energy gap (i.e., an gauge-invariant operators are bosonic, the above dynamical
effective mass) to be much bigger than the bandwidth (i.e., lattice gauge theory coupled to fermions is actually a local
at the infinite coupling limit). The infinite coupling limit in lattice bosonic model in disguise, as one can see from the
the mirror sector generates a dead layer; a neighbor layer next slave-particle/parton approach [36,78–81].
to the mirror sector would become the new mirror sector with We remark that the dynamical Spin(10) chiral gauge theory
fermion doublings [8], which would fail to produce a chiral coupled to Weyl fermions in the 16-dimensional spinor rep-
fermion/gauge theory at low energies. resentation is a local bosonic theory, since all gauge-invariant
operators are bosonic.17 The lattice regularization that realizes
IV. SU(5) CHIRAL FERMION THEORY the dynamical Spin(10) chiral gauge theory is also a local
AND SU(5) GRAND UNIFICATION bosonic model (see Appendix B). In other words, the Spin(10)
dynamical chiral gauge theory with Weyl fermions in a 16-
Above we have discussed the lattice regularization of a dimensional representation, and the induced 16-fermion stan-
Spin(10) gauged chiral fermion theory. To consider a lattice dard model, can be regularized as the low-energy effective
regularization of a SU(5) gauged chiral fermion theory [with theory of a local lattice model of qubits (since any local
G f = Z2f × SU(5), but only SU(5) will be gauged], we clas- bosonic lattice model can be viewed as a lattice model of
sify the 4 + 1D invertible fermionic topological order with qubits). Based on the stability of cobordism group of Eq. (10)
G f = Z2f × SU(5) symmetry by a cobordism group defined for N  7, our result directly applies to a Spin(N = 18)
in Eq. (11) [note that Spin(D = 5) ⊃ Z2f ] [30,55]: chiral gauge theory [72,82], which is also a local bosonic
model. Thus our study implies that all elementary particles
Spin(D=5)×SU(5) ≡ TPD [Spin(D = 5) × SU(5)] = Z, (22)

D=5 (except the graviton) can be viewed as originated from qubits
where the topological invariant is given by the SU(5) Chern- [83–85]. It is a concrete realization of “it from qubit [86],”
Simons 5-form, associated with perturbative local anomalies representing an ultra unification of all gauge interactions and
captured by perturbative Feynman diagram calculations in matter fermions in term of quantum information (i.e., qubits).
3 + 1D. The statement that all elementary particles arise from
Again, such a cobordism invariant and the associated in- bosonic qubits has a falsifiable experimental prediction:
vertible topological order can be detected by the boundary all fermions and their fermionic bound states must carry
chiral fermions: if the 3 + 1D boundary SU(5) chiral fermion nontrivial gauge charge [36,87]. As a result, the “stan-
theory is free from any of the Z class of SU(5) perturbative dard model” from a lattice qubit model cannot just have a
U(1)×SU(2)×SU(3)
local anomaly, then the corresponding cobordism invariant Zq
gauge group, since such a standard model
and the 4 + 1D bulk invertible topological order are trivial. indeed has fermionic bound states that carry no gauge charge.
Thus, by Propositions 1 and 2, any SU(5) gauged chiral Thus, the “standard model” from a lattice qubit model must
fermion theory that can be regularized at the boundary of a have a larger gauge group, e.g., adding a new [Z2 ] gauge
4 + 1D gapped local lattice model can be regularized by a 3 + sector,18 where we gain a new cosmic string (whose spacetime
1D local lattice model via the method (18), provided that the
SU(5) gauge theory is free of the SU(5) perturbative anomalies
(see, also, Proposition 4 in Appendix A 1). In particular, the 17
In Appendix B, we provide the explicit slave-particle/parton con-
SU(5) grand unified theory [5] can be regularized by a lattice. struction for a 4 + 1D local bosonic lattice model, whose boundary
This implies that its induced 15-fermion standard model can can give rise to the dynamical Spin(10) chiral gauge theory coupled
be regularized by a lattice fermion model. to Weyl fermions (Model 3) in the 16-dimensional representation.
18
See Footnote 11; we can show that
V. IMPLICATIONS AND CONCLUSIONS Spin(D) × Spin(10)
⊃ Spin(D) × SU(5) ⊃ Spin(D)
In fact, an n + 1D G-symmetric iTQFT given by a cobor- Z q
dism class in Propositions 1 and 2 corresponds to an nD ’t U(1) × SU(2) × SU(3)
Hooft anomaly of G-symmetry (see footnote 4 and the details × ,
Z6
of anomalies in Appendix A). So a trivial cobordism class
where q = 1 or 2, while
in n + 1D for G-symmetry means all-’t Hooft-anomaly-free
in nD for the full G-symmetry. Namely, by far we only U(1) × SU(2) × SU(3)
SO(10) ⊃ SU(5) ⊃
show that anomaly-free gauged chiral fermion theories can Z6
be defined on a lattice with nondynamical background gauge and
fields (Models 1 and 2 in Appendix A 1), regularized with U(1) × SU(2) × SU(3)
Spin(10) ⊃ SU(5) ⊃ .
on-site symmetries in its own dimensions [via (18)]. How- Z6
ever, we can obtain a dynamical chiral gauge theory (Model For q = 2, when we gauge the [Spin(10)], we also require one to
3 in Appendix A 1) by dynamically gauging the on-site gauge the [Z2f × SU(5)] in the embedded smaller group Spin(D) ×

023356-8
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

trajectory is a two-dimensional worldsheet)—the flux line of PHY-1606531. X.-G.W. is partially supported by NSF Grant
the new [Z2 ] gauge field [88]. No. DMS-1664412. This work is also supported by NSF
In contrast, the dynamical SU(5) chiral gauge theory Grants No. PHY-1306313, No. PHY-0937443, No. DMS-
coupled to Weyl fermions in the five- and 10-dimensional 1308244, No. DMS-0804454, and No. DMS-1159412, and
representations is a fermionic theory definable on spin mani- Center for Mathematical Sciences and Applications at Har-
folds, since some gauge-invariant operators are fermionic. The vard University.
lattice regularization that realizes the dynamical SU(5) chiral
gauge theory is also a local fermionic model (which is not a APPENDIX A: DEFINITIONS OF TERMINOLOGY AND
local lattice model of qubits). The standard model from local DISCUSSIONS BASED ON ANOMALIES
fermionic lattice models can have U(1)×SU(2)×SU(3)
Zq
as its gauge
group; see footnote 18. It does not require extra gauge groups. In the main text, we have described our results without
In our work, we have shown that a Spin(10) [or SU(5)] directly mentioning the quantum anomaly (footnote 4). How-
chiral gauge theory with 16 (or 15) Weyl fermions can be ever, in the literature, many people discuss the gauge chiral
regularized by a lattice, since the mirror sector can be fully fermion problem in terms of anomaly. In this section, we will
gapped by Spin(10) [or SU(5)] symmetric interactions with- discuss our approach using the concept of the anomaly. We
out spontaneously breaking the symmetry. However, it is pos- will carefully define several different anomalies. We will also
sible that the mirror sector can be fully gapped by interactions carefully define several concepts of chiral fermion field theory
with a larger symmetry G+ without spontaneously breaking and the concepts of lattice theory as a well-defined quantum
the symmetry G+ . In this case, after gauging Spin(10) [or theory.
SU(5)], the chiral gauge theory may have an exact global
symmetry GQ (on a lattice scale or a UV cutoff scale such as 1. Detailed definitions of some relevant concepts
an effective Planck scale) sitting as a quotient group satisfying We should clarify several related concepts of Spin(10)
the short exact sequences: chiral fermion field theories and models as follows:
1 → Spin(10) → G+ → GQ → 1, Model 1. Without gauging or before gauging Spin(10)
symmetry, the theory is a “Spin(10) chiral fermion theory”
or 1 → SU(5) → G+ → GQ → 1. (23) with the full internal global symmetry G f = Spin(10) ⊃ Z2f .
The reason GQ is still a global symmetry after gauging In this case, we call the anomaly associated with the global
Spin(10) [or SU(5)] is that because G+ can be chosen as an symmetry G f as the ’t Hooft anomaly of G f . We classify the
on-site symmetry on the UV cutoff scale and G+ is anomaly ’t Hooft anomaly [89] of G f in Eq. (10) and Eq. (11).
free on the d + 1D. Once a normal subgroup is gauged, the Model 2. We may twist the Spin(10) symmetry via a non-
GQ is still anomaly free and unbroken, and thus can still be dynamical background Spin(10) gauge field, known as the
made on-site in d + 1D. symmetry twist. We name such a theory as a “Spin(10) gauged
Lastly, we comment on the dynamics of these dynam- chiral fermion theory.” The anomaly of G classified in (10)
ical chiral gauge theories (Model 3, as highly long-range becomes the background gauge anomaly, which is the same
entangled states). At the low energy of these chiral gauge as the ’t Hooft anomaly in nature.
theories, there could be emergent symmetries (e.g., higher- Model 3. After dynamical gauging Spin(10) symmetry via
form symmetries [33] or higher symmetries in general [30]) dynamical weakly fluctuating Spin(10) gauge field, the theory
having new ’t Hooft anomalies. However, emergent new becomes a “Spin(10) chiral gauge theory.” In this case, it is
anomalies only mean the emergent symmetries cannot be a standard terminology to call the anomaly, descending from
strictly regularized locally on-site, on-link, on-n-simplex, etc., Model 1’s ’t Hooft anomaly and after gauging Spin(10), as
which, we emphasize, is a rather distinct issue deviated from the dynamical gauge anomaly. We will thus also study the
regularizing chiral fermion theories which we solved earlier. dynamical gauge anomaly of Spin(10), thanks to Eq. (10). If
After regularizing chiral fermion theories on a lattice and any model possesses any dynamical gauge anomaly, then this
after dynamically gauging, the emergent new anomalies only theory is inconsistent and thus ill defined.
constrain the dynamics of gauge theories (e.g., gapless near Reference [23] adopted a different viewpoint (or a different
a quantum critical fixed point, or emergent symmetry spon- definition) of anomalies proposed in Ref. [90] (see Definition
taneously broken, etc). We aim to address the dynamics of 7), for interacting quantum theories, which in turn leads to
gauge theories in future work. a classification of anomalies. Before we proceed, we should
clarify some conventions of terminology as the definitions:
Definition 1. Well-defined quantum theories are quantum
ACKNOWLEDGMENTS
theories defined with a finite-dimensional Hilbert space and a
We thank E. Witten for many helpful discussions and finite-dimensional Hamiltonian matrix for a finite-size system
comments, and for a collaboration on Ref. [67]. J.W. thanks in the real space. In this work, we only focus on this class of
Z. Wan for related collaborations. J.W. is supported by Corn- quantum theories.
ing Glass Works Foundation Fellowship and NSF Grant No. Local lattice models are interacting or noninteracting lattice
models whose many-body Hilbert space V has the following
tensor product decomposition:
SU(5) ⊂ Spin(D)×Spin(10) . The dynamically gauging [Z2f ] symmetry

Zq =2 V= Vi , (A1)
produces the new [Z2 ] gauge sector. i

023356-9
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

where Vi is a finite-dimensional Hilbert space for each lattice product state. This leads to a concept of all lattice obstruction
site. free: By saying a d + 1D QFT is all lattice obstruction
By interacting models, we mean the Hamiltonian contains free, we always mean a d + 1D well-defined quantum field
certain higher-order terms beyond the quadratic terms of theory in Definition 3, which can be regularized as the low-
fundamental lattice operators (such as quartic fermionic or energy effective boundary theory of a d + 2D gapped tensor
spin operator terms beyond the quadratic terms). product state (i.e., a gapped trivial vacuum) on a one-higher-
By noninteracting models (or the so-called free or dimensional lattice. Note that a tensor product state (i.e., a
quadratic models), we mean the Hamiltonian contains, at trivial state, with neither short-range nor long-range entan-
most, the quadratic terms (thus easily diagonalizable and glement of Definition 5) in a local lattice model is defined as
solvable) of fundamental lattice operators. 
Local interactions. By local interactions, we mean that all | = |ψi , |ψi  ∈ Vi , (A3)
interaction terms in the Hamiltonian (or a Lagrangian in the i

path integral) must be bounded by a finite range of lattice which can be gapped and decoupled from its boundary theory.
spacings. We call these types as local, finite-range, or short- In contrast, the generic state is more general and is not
range interactions. We do not allow infinite-range interactions necessarily a tensor product, such as
or the interactions with strength that exponentially decays to  
zero only at infinite. For any interaction term of our lattice 
| = c{i} |ψi  , |ψi  ∈ Vi , (A4)
model, it must be bounded by a finite spatial range; say, if the
{c{i} } i
operators act on any site i to j, then the locality means that
“|i − j|  a finite distance.” with generic complex normalizable coefficients c{i} .
We emphasize that conventional lattice gauge theories with Definition 5. Gapped system and entanglement. By a
dynamical gauge fields are usually not local lattice models: gapped tensor product state, we mean that the tensor product
Since there is a nonlocal gauge constraint, the tensor product state is a unique ground state (with an energy E0 ) of some
decomposition (A1) is violated. In this work, we do not lattice bulk Hamiltonian system whose energy spectrum has
use models of conventional lattice gauge theories, but limit a finite energy gap E = Eexcited − E0 > 0 separated from
ourselves to only local lattice models. all excited states Eexcited . Below the energy gap E , the
Definition 2. The on-site symmetry, for such local lattice system behaves as a gapped trivial vacuum (or a gapped trivial
models, is defined as a global internal symmetry, whose insulator in condensed matter) with no entanglement.
symmetry transformation operator has the following tensor On the other hand, general gapped systems (E =
product decomposition: Eexcited − E0 > 0) can generically possess short-range or long-
 range entanglements.
U = Ui , (A2) Short-range entangle states, short-range entanglements
i (SREs), and SPT states are defined as those gapped quantum
where Ui is a unitary operator acting on Vi . ground states which can be deformed via local unitary trans-
Definition 3. Well-defined quantum field theory (living on formations (LUTs) to a trivial tensor product state once we
the boundary of lattice model). When we mention a “well- remove part of or all of the internal global symmetries [16].
defined” quantum field theory (QFT), we always mean a lim- Namely, along the deformations to a trivial tensor product
ited class of QFTs which can be regularized (i.e., regularized) state, the LUTs may break some internal global symmetry of
as the low-energy effective boundary theory of a gapped local the state. Gapped SRE states are also named SPT states.
lattice model (see Definition 1) in one higher dimension (the Long-range entangle states, long-range entanglements
so-called bulk). The global symmetry, if any, is regularized (LREs), and topological orders are defined as those gapped
as an on-site symmetry (see Definition 2) for the full bulk- quantum ground states which cannot be deformed via local
boundary system. Such a QFT has, at most, the b-anomaly to unitary transformations (LUT) to a trivial tensor product state,
be defined later in Definition 7. A “well-defined” QFT cannot even if we remove all internal global symmetries. Gapped
have the r-anomaly that is defined later in Definition 8. LRE states are also named topological orders.
Definition 4. All lattice obstruction free (required to be reg- By this Definition 5, we can also rephrase Definition 4 as
ularizable in the same dimension). The above-defined QFTs a well-defined quantum field theory (in Definition 3) is all
include d + 1D QFTs that can be regularized by a lattice lattice obstruction-free (Definition 4) if it can be regularized
model in the same dimension d + 1D (with the symmetry, as the low-energy effective boundary theory of a gapped bulk
if any, regularized as an on-site symmetry or a local on-n- lattice system whose bulk has no LRE (i.e., no topological
simplex symmetry19 ) because the gapped bulk in one higher order) and no SRE (i.e., no SPT state), and thus the bulk has no
dimension (of Definition 3) can be a decoupled gapped tensor entanglement structure at all as a gapped trivial tensor product
state. Readers should be cautious that although this gapped
bulk alone has no entanglement, the boundary theory (such
19
Here we only focus on the well-defined G-symmetric QFTs with as an all-anomaly-free QFT) can be highly entangled and can
ordinary G-global symmetries (the 0-form symmetry in the sense of have gapless states.
generalized global symmetries [33]). If there is a generalized higher Definition 6. All anomaly free (i.e., here, free of all invert-
global symmetry [33], then we need to modify the “lattice on-site ible bosonic and fermionic b-anomalies). The recent devel-
symmetry realization” to the “lattice local on-n-simplex symmetry opment suggests that all anomaly-free conditions of d + 1D
realization.” G-symmetric QFT can be understood as the QFT can live

023356-10
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

on the boundary of a trivial cobordism class of a trivial ungauged quantum theory may have a b-anomaly (’t Hooft
invertible topological quantum field theories (iTQFT) from a anomaly) instead of an r-anomaly. (See examples below).
corresponding cobordism group [29] or its higher-symmetry Definition 9. Invertible vs noninvertible anomalies. There
and higher-classifying space generalization [30–32]: are invertible anomalies that can be canceled by other anoma-
lies. (The anomalies discussed in the field theory literature

d+2
G . (A5) are mostly invertible anomalies). Invertible anomalies form
an Abelian group, such as an infinite integer group Z (i.e.,
Namely, the trivial iTQFT is the trivial element 0 in the
d+2
G . a perturbative local anomaly, captured by a Feynman diagram
Let us explain this development below. loop calculation) or a finite group Zn of some positive integer
The ’t Hooft anomaly is a property in which the global n (i.e., a nonperturbative global anomaly), or the product
symmetry of the theory cannot be made on-site on a lattice, groups of Z and Zn . The invertible anomaly labeled by an
and thus there is an obstruction to gauge the non-on-site Abelian group element g can be canceled by an inverted
symmetry, which is called the anomalous symmetry (Model anomaly labeled by an inverted Abelian group element g−1 .
1) [18,90]. Dynamical gauge anomaly is a property in which There are also noninvertible anomalies [41,90,95–98] that
its theory is ill defined (discussed in Model 3). How do cannot be canceled by any other anomalies.
we classify the property of non-on-site global symmetries or Definition 10. Bosonic vs fermionic anomalies. There are
seemly ill-defined theories? bosonic anomalies where the local operators in the corre-
The previous anomaly inflow picture relates the anomalous sponding anomalous theories are all bosonic [99,100]. There
noninteracting field theories or noninteracting lattice models are fermionic anomalies where some local operators in the
(Definition 1) to the boundary of one-higher-dimensional bulk corresponding anomalous theories are fermionic. For exam-
[91,92]. Reference [93] systematically described anomalies in ple, a Spin(10) chiral Weyl fermion theory has an internal
field theories in terms of topological invariants in one higher symmetry Spin(10) ⊃ Z2f containing the fermion parity, and
dimension (such as the index of a Dirac operator), which turn thus we will need to classify possible fermionic anomalies of
out to be cobordism invariants [94]. However, to construct an the interacting fermionic theory [later in Eq. (10)] in order to
interacting lattice regularization of a field theory, we need to classify all of its anomalies.20
classify anomalies in interacting field theories and interacting For more examples,
lattice models. Reference [90] attempts to classify anoma- (a) A 1 + 1D chiral complex Weyl fermion theory with
lies in interacting lattice models, via topological orders and a Hamiltonian, H = i ψ † ∂x ψ, and a one-component complex
symmetry-protected topological states (SPTs) of interacting Weyl spinor ψ, has
lattice models in one higher dimension.
Let us introduce a few different concepts of anomalies as a fermionic invertible b-anomaly.
the definitions of the terminology: It is invertible because the anomaly has a Z class as a group
Definition 7. b-anomaly (≡ boundary defined anomaly). classification.
There are anomalous theories that can be regularized as the (b) A 3 + 1D Weyl fermion doublet coupled to a probed
low-energy effective boundary theory of a gapped local inter- (thus nondynamical) SU(2) background gauge field has the
acting lattice model in one higher dimension, where the global Witten SU(2) anomaly [101] as a type of ’t Hooft anomaly of
symmetry, if any, is regularized as an on-site symmetry for the the SU(2) global symmetry, which is
whole bulk-boundary coupled system. However, the effective
symmetry, if any, on the effective boundary theory alone is a fermionic invertible b-anomaly.
non-on-site. There is an obstruction to gauge the non-on-site It is fermionic because the SU(2) ⊃ Z2f has the fermion parity
symmetry [18,90] because the standard gauging only works at its Z2 center. It is invertible because the anomaly has a Z2
for an on-site symmetry: there is no canonical way to input the class as a group classification.
gauge variables on the links between “nonlocal sites” where (c) A 3 + 1D Weyl fermion doublet coupled to a dynamical
the non-on-site symmetry acts. The obstruction of gauging is SU(2) gauge field has the Witten SU(2) anomaly [101],
the same phenomenon that happened in ’t Hooft anomalies. which is
We will call this kind of anomalies the b-anomalies, which
include the ’t Hooft anomalies (associated with some internal a bosonic invertible r-anomaly.
global symmetry), gravitational anomalies (associated with no
It is bosonic since all the local operators are gauge invari-
internal global symmetry), and their mixed anomalies.
Definition 8. r-anomalies (≡ radical anomaly). There are ant and bosonic. Namely, the SU(2) (⊃ Z2f ) is dynamically
also anomalous theories that cannot be regularized as the gauged, and thus the fermion parity Z2f is also gauged and
low-energy effective boundary theory of any gapped local the full theory is bosonic. It is an r-anomaly since the Weyl
lattice model in one higher dimension. We will call this kind fermion coupled to this SU(2) gauge theory cannot be regular-
of anomalies the r-anomalies, which include the dynamical ized as a boundary of any gapped local bosonic lattice model
gauge anomalies. A theory with an r-anomaly is simply an
ill-defined quantum theory.
However, for a dynamical gauge theory with an r-anomaly, 20
However, once the [Spin(10)] ⊃ Z2f is gauged and thus the
very often, we ungauge the theory to turn the dynamical gauge fermion parity Z2f is gauged in the Spin(10) chiral Weyl fermion
field (on the link or on n-simplex) into a global symmetry theory, it becomes a Spin(10) chiral gauge theory, where all local
transformation [on-site or on (n − 1)-simplex]. The resulting gauge-invariant operators are bosons.

023356-11
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

[67]. However, if we ungauge the SU(2) of this ill-defined as an on-site symmetry (or generalized local on-n-simplex
gauge theory, then its bosonic invertible r-anomaly (Definition symmetries) [90].22
8) becomes the previous fermionic invertible b-anomaly. This result can be used to solve the gauged chiral fermion
(d) A Z2 gauge theory in 2 + 1D or above with only Z2 problem via the mirror fermion approach [23]: Given a d +
charge excitations has 1D gauged chiral fermion theory with a gauge group G f ⊃
Z2f , we first ungauge, and obtain a d + 1D chiral fermion the-
a bosonic noninvertible b-anomaly,
ory with an internal global symmetry group G f ⊃ Z2f . Then,
regularized as a boundary theory of a one-higher-dimensional we find a gapped d + 2D lattice model with a symmetry G f ⊃
Z2 gauge theory, which is a topological quantum field Z2f whose boundary regularizes the ungauged d + 1D chiral
theory (TQFT). fermion theory (Model 1). The symmetry G f is regularized as
The classification in Ref. [90] is a classification of all an on-site symmetry of the d + 2D lattice model. Next, we
b-anomalies in terms of the topological orders [102] or determine whether or not the ground state of the bulk gapped
symmetry-protected topological (SPT) states [15,16,46] in d + 2D lattice model has a trivial topological order and a
local lattice models in one higher dimension. A b-anomaly is trivial SPT state. If the d + 2D ground state indeed has no
invertible if it is characterized by an SPT state or an invertible topological order and no SPT state (which is a trivial tensor
topological order [41,42,103,104] in one higher dimension. In product state by Definition 4), then the d + 1D ungauged
this work, we will only focus on the invertible b-anomalies chiral fermion theory can be regularized as the low-energy
and their classifications. effective theory of a d + 1D local lattice model. Also, the
From now on, by anomalous field theory, we will specifi- d + 1D gauged chiral fermion theory can be regularized as the
cally mean a well-defined quantum field theory (Definition 3) low-energy effective theory of a d + 1D local lattice model
with, at most, some invertible b-anomalies (defined in Defi- after gauging the on-site symmetry G f .
nition 7). In this work, we only study well-defined quantum To show the above claim, we can choose the d + 1D lattice
field theories (Definition 3) as the effective low-energy theory model to be a slab of the d + 2D lattice model with a finite
of the boundary of local lattice models (Definition 1). So, we number of layers in the extra dimension. In such a model,
exclude theories with the r-anomaly (defined in Definition 8) the normal sector (or the chiral fermion sector) lives on one
since they are not well-defined quantum theories (by the norm surface of the slab and the mirror fermion sector lives on the
of both Definitions 1 and 7, and the standard lore). other surface of the slab. If the normal sector is free of all
According to the above classification, an anomaly-free anomalies, it implies that the d + 2D bulk is actually a trivial
(Definition 6) well-defined quantum field theory (Definition gapped phase. If so, the mirror sector can be chosen to be
3)21 is simply a boundary theory of a gapped trivial state a symmetric gapped boundary and can be fully gapped out
(a tensor product state) on a one-higher-dimensional lattice, without breaking the on-site symmetry [15,16,18]. A detailed
which means all lattice obstruction free that can also be explanation is given in Appendix A 2. Since the d + 2D slab
regularizable in the same dimension (Definition 4). has only finite layers, the d + 2D slab is actually a d + 1D
The generalization of the anomaly inflow to a lattice model lattice model with finite orbitals per site. Lastly, we gauge the
with interactions is crucial to obtain this result since some of on-site symmetry to obtain a gauged chiral fermion theory.
the key concepts, such as the tensor product state and the on- Thus, the above understanding suggests the following:
site symmetry, require a lattice (providing the locality of sites)
to define.
With the above terminology definitions, we claim a propo- 22
See footnote 19 for the comment on the local symmetry realiza-
sition: tions on the lattice. Above we propose the following:
Proposition 4. For any well-defined quantum field theory
(Definition 3), if it is If “all anomaly free” → then “all lattice obstruction free.” (A6)
(Definition 6) (Definition 4)

all anomaly free (Definition 6) However, some well-defined QFTs (Definition 3) can be regularized
on the boundary of one-higher-dimensional lattice model, e.g., even
with a list of conditions in footnote 21, then it is if they have b-anomalies in Definition 7. Thus, there is a subtlety
about the converse statement. Only when we restrict the “all lattice
all lattice obstruction free (Definition 4), obstruction-free” requiring QFT to be regularizable in the same
which is required to be regularizable in the same dimension. dimension and all symmetries regularized strictly locally (Definition
Namely, any well-defined QFT that is all anomaly free can 4), is the converse statement also true:
be regularized by a local interacting lattice model in the If “all lattice obstruction free” → then “all anomaly free.” (A7)
(Definition 4) (Definition 6)
same dimension, where the global symmetry is regularized
In this work, when we classify invertible ’t Hooft anomalies of global
symmetries G, we use the cobordism group
Gd+2 in (A5) whose
category of manifolds are only smooth and differentiable manifolds.
21
Thus, here the all-anomaly-free condition for a well-defined Therefore, we can apply a known mathematical fact that all those
quantum field theory (Definition 3) specifically satisfies the follow- smooth and differentiable manifolds are triangulable manifolds, via
ing: (i) free of b-anomalies in Definition 7, (ii) free of all invertible the Morse theory. Thus the anomalies captured in
Gd+2 of smooth
anomalies in Definition 9, (iii) free of bosonic and fermionic anoma- and differentiable manifolds can be triangulated on a lattice of
lies in Definition 10. triangulable manifolds. See more comments in Sec. II.

023356-12
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

Proposition 5. Any d + 1D gauged chiral fermion theory a symmetric disordered phase. Some other related approaches
(Model 2) that can be regularized as the low-energy effective have also been proposed [17,72–74,77].
boundary theory of a d + 2D gapped local lattice model in one In this work, we do not require the proposed conditions of
higher dimension (Definition 3), can be regularized as the low- Ref. [23] above, nor do we need the assumption of new fluctu-
energy effective theory of a local lattice model in the same ating Higgs fields in Ref. [23]. Instead, we will independently
d + 1D dimension (Definition 4), as long as the theory is free and rigorously show that the above Spin(10) chiral fermion
of all anomalies (given by Definition 6 and footnote 21). theory (Model 1) is indeed free of all ’t Hooft anomalies
We remark that for a certain anomalous d + 1D chiral by a cobordism group approach [Eq. (10)], and thus it can
fermion theory with an internal symmetry group G f , their be defined on a 3 + 1D lattice; this can become a Spin(10)
corresponding d + 2D topological/SPT orders may have a gauged chiral fermion theory (Model 2) by coupling to a
gapped boundary that does not break the G f symmetry, but has Spin(10) background gauge field, or become a Spin(10) chiral
a nontrivial G f -symmetric anomalous boundary topological gauge theory (Model 3) by dynamically gauging Spin(10).
order [18,105]—the low-energy theory of topological order
may be a d + 1D G f -symmetric topological quantum field
theories (TQFT) canceling the same ’t Hooft anomaly of 2. Gapped boundary of a state with a trivial invertible
d + 1D chiral fermion theory. topological order with symmetry
For such an anomalous d + 1D chiral fermion theory, we In the following, we show that:
can have a lattice model in the same d + 1D dimension (1) there exists a 3 + 1D gapped boundary for the above
that exactly regularizes all the low-energy particles of the lattice model without breaking the Spin(10) symmetry at the
anomalous chiral fermion theories. However, the full low- low energy, and
energy effective theory of the lattice model will contain an (2) there exist nonperturbative interactions to gap the
extra gauge field for a finite gauge group Gextra , prescribing mirror world chiral fermions without breaking the Spin(10)
the nontrivial anomalous d + 1D topological order and TQFT. symmetry.
Thus, if we only concern ourselves with low-energy particles, The focus of this section is on showing the existence (in
even some anomalous gauged chiral fermion theories can be the mathematical sense), instead of proving the constructions
regularized by lattice models in the same dimension [23]. But (which may not be unique for the uniqueness in the mathemat-
the lattice models will also produce an extra d + 1D Gextra - ical sense). In Sec. III, we provide the 16 copies of the lattice
gauge theory with no additional low-energy particles, but may model (7) that give rise to the 3 + 1D Weyl fermions in the
give rise to additional extended objects such as string and 16-dimensional spinor representation of the Spin(10) on the
brane excitations from the TQFT. lattice boundary; see Fig. 1.
It is well known that a Spin(10) chiral fermion theory In order to show Proposition 4, which consequently also
(Model 1) is free of all perturbative ’t Hooft anomalies; includes Proposition 5, we break down this proposition into
similarly, it is also well known that a Spin(10) chiral gauge several related helpful subpropositions. For any well-defined
theory (Model 3) is free of all perturbative dynamical gauge d + 1D QFT (defined in Definition 3) that is all anomaly free
anomalies [106,107]. But it is not known before whether (defined in Definition 6) with an internal symmetry G f , which
or not the Spin(10) chiral fermion theory (Model 1) is free can live on the boundary of d + 2D bulk regularized lattice
of all other nonperturbative global anomalies (of ’t Hooft model, we aim to show the following (in which we focus on
anomalies). Thus, it was also not known in the past literature the spatial dimension d = 3):
whether or not the Spin(10) chiral gauge theory (Model 3) Proposition 6. There exists a symmetric gapped boundary
is free of all other nonperturbative global anomalies (as for the corresponding d + 2D bulk regularized lattice model.
This d + 1D symmetric gapped boundary does not break any
dynamical gauge anomalies).
internal symmetry G f of the whole bulk-boundary system,
Reference [23] provides an argument that the Spin(10)
and does not contribute any ground-state degeneracy (neither
chiral fermion theory is free of all anomalies, by proposing
symmetry-breaking degeneracy nor topological degeneracy
a sufficient condition: A gauged chiral fermion theory in a
[108,109]).
d + 1-dimensional spacetime with a gauge group G f is free
Proposition 7. There exist nonperturbative symmetric in-
of all anomalies if (0) it can be regularized as a low-energy
teractions to fully gap this well-defined all-anomaly-free d +
effective boundary theory of a gapped local lattice model
1D QFT, via deforming the QFT by adding any all-anomaly-
in one higher dimension (Definition 3), (1) there exists a
free gapless or gapped sectors, while still preserving the
nonzero Higgs field that makes all the fermions massive, full G f internal symmetry, without any symmetry breaking
and (2) πn (G f /Ggrnd ) = 0 for 0  n  d + 2, where Ggrnd is and without contributing any degeneracy (neither symmetry-
the unbroken gauge symmetry group for the nonzero Higgs breaking degeneracy nor topological degeneracy).
field. The chiral fermions satisfying the above conditions We will see that showing Proposition 6 is sufficient enough
can be gapped out by direct interactions or boson-induced to show that Proposition 4 is also true. In other words, we only
interactions without breaking the G f symmetry, even when need Proposition 6 but do not need to prove Proposition 7, in
the fermion mass term is forbidden by the symmetry. This order to prove Proposition 4.
mechanism to give fermions an effective energy gap (or an To show Proposition 6, we first note that by a symmetric
effective mass) is referred to as the “mass without mass term gapped boundary, we also mean that the ground-state energy
[72].” But the above statement is based on an assumption that E0 (of this whole bulk-boundary system) to its higher-energy
a smooth orientation fluctuation of Higgs field can give rise to excited states (at energy E1 , . . . , etc.) is separated by a

023356-13
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

E(k) E((k) E((k)

1 single 16 copies 16 copies

gapless complex of gapless complex of gapless complex
Weyl fermion Weyl fermions ················ Weyl fermions ················
k k k
in 3+1D with interactions
in 3+1D

(a) (b) (c)

FIG. 1. (a) A lattice construction of a single Weyl fermion is given in Sec. III; the subfigure shows the gapless energy spectrum E (k) of the
Brillouin zone in the schematic three-dimensional momentum k = (kx , ky , kz ) space with a linear dispersion |E (k)| ∝ c|k| for some effective
speed of light c. (b) The 16 copies of the same lattice model (7) give rise to the 3 + 1D Weyl fermions at the low energy in the 16-dimensional
spinor representation of the Spin(10) on the lattice boundary shown in Sec. III. The 16 gapless Weyl points (schematically the 16 dots •) may
be separated, but can be tuned to the same point on the k-space Brillouin zone. We show that this Spin(10) chiral Weyl fermion theory is
free from all ’t Hooft anomalies via a cobordism theory in Sec. II. (c) There are two ways to obtain the symmetric gapped boundary for the
bulk of the 16 copies of the lattice model: First, via Proposition 6, there exists a symmetric gapped boundary for the corresponding d + 2D
bulk regularized lattice model (without the need to access from gapping out the gapless theories from interactions. Second, via Proposition
7, there exist nonperturbative symmetric interactions to fully gap this well-defined all-anomaly-free Spin(10) chiral fermion theory with 16
Weyl fermions in 16-dimensional spinor representation of the Spin(10). (Schematic interactions are drawn in the shaded blue region). In this
work, we only prove Proposition 6, but we suggest some supportive evidence for Proposition 7 but without proving Proposition 7. However,
applying only Proposition 6 (but without requiring Proposition 7) is sufficient enough for us to construct the Spin(10) chiral fermion theory on
the lattice via Proposition 4.

finite-energy gap E = E1 − E0 > 0. Of course, by defining To proceed showing Proposition 6, if the bulk regularized
the energy gap E > 0 here, we should first set up a toy- lattice is in the gapped trivial phase (i.e., has a gapped trivial
model system with only such a d + 1D symmetric gapped tensor product ground state), we can make a boundary by
boundary and a fully gapped d + 2D bulk. (We either have first deforming the bulk ground state (by symmetry-preserving
only this gapped boundary without other boundaries, or other LUT in Definition 5) into a tensor product state. Such a
boundaries that are also fully gapped). deformation does not close the energy gap since the bulk state
If the gapped d + 2D bulk also has a symmetric gapless is already in the symmetric gapped trivial phase. The trivial
boundary (say, on a d+1D boundary A) other than the sym- tensor product state always can have a gapped boundary with
metric gapped boundary of Proposition 6 (say, on another respect to a trivial vacuum23 —by saying so, we mean that
d+1D boundary B), then the gapless boundary A contributes we set the energy scale of the trivial vacuum (normally to
to the low-energy spectrum at the infrared (IR) of a tiny energy below some energy scale such as a finite energy E > 0, or
subgap, below an infinite energy gap E → ∞) to be the same as
the energy scale of the gapped boundary (say, on B) E 
δE ,A  exp(−L/ξ ), (A8) E ,B > 0 in Eq. (A9). We note that the above deformation
respects the on-site symmetry (if any), and the resulting tensor
which scales exponentially over the linear system size L over
product state also respects the on-site symmetry. The gapped
the correlation length ξ ; the gapped boundary B contributes
boundary does not break the on-site symmetry, and thus has no
to the energy spectrum only at the higher energy at a deeper
symmetry-breaking degeneracy. Since the symmetric gapped
ultraviolet (UV) of a finite-energy gap,
boundary has no entanglements, it therefore has no topologi-
E ,B  E > 0, (A9) cal degeneracy (because topological degeneracy [108,109] is
due to LRE defined in Definition 5).
mentioned earlier. Then the whole bulk-boundary system The above completes our proof of Proposition 6.
would become gapless instead of being gapped.
The important issue is that when the d + 2D gapped bulk
has no entanglements (i.e., no LRE nor SRE by Definition 23
The trivial tensor product state always can have a gapped bound-
5), then “the d+1D symmetric gapless boundary A” and “the ary respect to a trivial vacuum, because the trivial tensor product
d+1D symmetric gapped boundary B” actually cannot affect state is itself the same phase indistinguishable as the trivial vacuum.
each other, and thus are isolated from each other. See more in Thus its gapped boundary simply is the trivial gapped domain wall
Appendix C. between the same phase [110,111].

023356-14
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

In Appendix D, we can also show Proposition 6 by a sec- possible degrees of freedom and their symmetry-preserving
ond viewpoint: a derivation from the classification of quantum local interactions.
phases of matter and their phase transitions. We do not have a direct proof of Proposition 7, but we have
This second viewpoint from the classification of quantum several supportive pieces of evidence to argue that Proposition
phases of matter shows that there is no need for an energy-gap 7 should be true:
closing phase transition. By maintaining a finite-energy gap (a) “The deformation classes of QFTs” advocated by
E between two phases, there must exist a symmetric gapped Seiberg [112]: Given a continuum QFT with some energy
boundary between two phases, and thus we have given an scale , with a given global symmetry and derivable ’t Hooft
alternative proof of Proposition 6. anomaly of global symmetry. We are allowed to add arbitrary
The slight conceptual difference between the first view- degrees of freedom and new fields preserving the symme-
point and the second viewpoint is that the first is about the tries (and selection rules) and with no additional anomaly
one-spatial-dimensional-lower phase boundary in d + 1D be- (without modifying the original ’t Hooft anomaly) at some
tween two d + 2D phases, while the second is about no need energies. The new degrees of freedom do not directly affect
for the phase transition in d + 2D between two d + 2D phases the dynamics at lower energies. Next, we can deform the
in a quantum phase diagram (at zero temperature T = 0) by parameters of this larger theory with the new degrees of
tuning a certain coupling g. freedom, by making the new degrees of freedom interacting
To show Proposition 4, we consider a d + 2D bulk regu- with the original QFT, which do affect the dynamics. This is
larized lattice model that regularizes the d + 1D QFT as its a much larger space of theories, which can land into different
boundary theory by Definition 3. We choose the bulk lattice new phases with different dynamics. Seiberg names all these
model to be a slab of finite thickness, such that one boundary possible deformations of QFT as a deformation class of
of the slab regularizes the QFT (Boundary A), and the other the QFT.
boundary is a symmetric gapped boundary (Boundary B) in Seiberg [112] conjectured that given two QFTs, say parti-
Proposition 6. We apply Proposition 6 that was proven earlier. tion functions Z1 and Z2 , in the same spacetime dimension
Here a slab of finite thickness is always achievable for this with:
system (especially for the gauged chiral fermion problem of (i) the same global symmetry (and selection rules), and
Model 1 and Model 2) because of the isolation between two (ii) the same ’t Hooft anomalies,
d + 1D boundaries A and B due to Appendix C’s remark (i) we can always add new degrees of freedom at short dis-
on the isolation of the energy scale and remark (ii) on the tances so that we can interpolate between two QFTs: The two
isolation of the mutual entanglement; see Appendix C for the QFTs, Z1 and Z2 , are in the same deformation class of QFT.
energy scale and mutual entanglement between gapless and In other words, this also means that the deformation class of
gapped boundaries. QFT can be determined and defined by the symmetries and
Thus the low-energy physics of the d + 2D slab is de- the ’t Hooft anomalies of QFT. Seiberg’s conjecture is in fact
scribed by this d + 1D QFT. A lattice model of this d + 2D shown to be true for many examples.
slab of a finite thickness can be constructed explicitly as a What we claim on Proposition 7 is indeed a special case of
lattice model in one lower dimension (d + 1D), by rewriting Seiberg’s proposal [112]: We consider the deformation class
the “quantum Hilbert space associated with different lattice of the anomaly-free well-defined QFT, containing the trivial
sites along the finite width thickness w (i.e., an extra small gapped phase [e.g., a symmetric gapped Spin(10) boundary]
dimension along w)” to “quantum Hilbert space associated and a gapless phase [e.g., a symmetric gapless Spin(10) chiral
with finite orbitals per site” in d + 1D. fermion theory], in which both have (i) the same global
This completes our proof of Proposition 4.24 symmetry (and selection rules) and (ii) no ’t Hooft anoma-
lies (that we will show via a cobordism theory in Sec. II).
If Seiberg’s proposal [112] is true, our proposal must also
3. A deformation class of all-anomaly-free well-defined QFTs
be true.
For Proposition 7, we again consider a bulk lattice model (b) We can start from the d + 1D symmetric gapless
that regularizes the QFT as its boundary theory by Definition all-anomaly-free theory, and adding new d + 1D symmet-
3. Since the same bulk model can also have a symmetric ric gapped all-anomaly-free sectors (this is analogous to
gapped boundary according to Proposition 6, we thus ask: Seiberg’s proposal [112]).
How can we modify the symmetric interactions in the QFT Moreover, we can also add additional gapless all-anomaly-
to make it into a fully symmetric gapped theory describing free sectors in the various possible representations (Rep) of
the symmetric gapped boundary? symmetry.25 The symmetry organizes the (both gapped and
One key ingredient is that there are more degrees of gapless) energy eigenstates in the energy spectrum into var-
freedom given by the full Hilbert space to help us reach the ious possible representations of the symmetry group, whose
goal. We may be able to access the symmetric gapped phase
not only within this specific well-defined all-anomaly-free
QFT, but also a higher energy spectrum by engineering all
25
For example, the trivial Rep of pairs of left- and right-moving
3 + 1D Weyl fermions ψL and ψR in the trivial Rep of Spin(10).
Then, adding their mass term, e.g., m(ψL† ψR + ψR† ψL ) (only for these
24
As we said earlier, we do not need Proposition 7 to show additional trivial Rep gapless sectors), does not break the Spin(10)
Proposition 4. symmetry.

023356-15
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

selection rules constrain the interactions and dynamics be- To have gapless Weyl fermions on the boundary, we add
tween states in the energy spectrum. the third term,
Based on (a) and (b), we propose that Proposition 7 is also

true. Ĥ3 = †
tiabj ĉα,a,i ĉβ,b, j χ̂iαβ
j + H.c. . (B4)
ij

APPENDIX B: CONSTRUCT A LOCAL BOSONIC LATTICE

Now, on each site, we have fermions ψ̂α,m and ĉα,a , but we
MODEL REALIZING A 3 + 1D Spin(10) GAUGED CHIRAL
still project into the subspace with an even fermion per site.
FERMION THEORY
The Ĥ1 + Ĥ2 + Ĥ3 acts within this subspace. So the model is
Below we will use the slave-particle/parton approach still a lattice bosonic model. When ti j is given by Eq. (7), the
[78–80] to explicitly construct a 4 + 1D local bosonic lat- model
tice model, whose boundary can give rise to the dynami-
cal Spin(10) chiral gauge theory coupled to Weyl fermions Ĥ1 + Ĥ2 + Ĥ3 (B5)
(Model 3) in the 16-dimensional representation. We start with
will give rise to emergent massless Weyl fermions on the
a fermionic model on a 4D cubic lattice. On each site, we
boundary coupled to the Spin(10) gauge field.
have 16N f complex fermions ψ̂α,m , with α = 1, . . . , 16, and
Consider a 4 + 1D slab of the local bosonic lattice model
m = 1, . . . , N f .26 So on each side, there are 216N f states. Now
described by Ĥ1 + Ĥ2 + Ĥ3 . In the main text, based on the
we project into the even fermion subspace on each site, and
complete classification of the ’t Hooft anomaly of the group
turn the fermionic model into a bosonic model with 216N f −1
states per site. The Hamiltonian for such a bosonic model is G = [Spin(5) × Spin(10)]/Z2f [in Eq. (10) and Eq. (14)], we
given by have shown that there exists a symmetric gapped boundary by

αβ † αβ Propositions 4 and 5 (or there exists a symmetric boundary
Ĥ1 = χ̂i j χ̂i j + (−)i n̂i , gapping interaction, called Ĥint.bdry.nonpert ), which allows us
i j αβ i to gap out the boundary Weyl fermions (without inducing
a boundary 3 + 1D topological order [18]) on one of the
χ̂iαβ
j = ψ̂α,m,i
†
ψ̂β,m, j , n̂i = ψ̂α,m,i
†
ψ̂α,m,i . (B1) surfaces of the slab.
m m,α In this case, by including such interactions Ĥint.bdry.nonpert
The above model has [U(16)]Nsite local symmetry. In the large into (B5), we propose a Hamiltonian
N f limit, χ̂iαβ
j is weakly fluctuating and can be replaced by Ĥ1 + Ĥ2 + Ĥ3 + Ĥint.bdry.nonpert . (B6)
χ (eiAi j )αβ = χ̂iαβ
j  expectation value and Ai j is a 16 × 16
Hermitian matrix to describe the U(16) gauge fluctuation. Then by applying Eq. (18)’s finite-width/layer lattice dimen-
This leads to the following emergent U(16) gauge theory (at a sional reduction, the 4 + 1D slab, with a finite width in the
mean-field level): extra dimension, indeed becomes a 3 + 1D local bosonic
lattice model that regularizes a 3 + 1D dynamical Spin(10)
Ĥ1mean = [ψ̂α,m,i
†
χ ∗ (e−iAi j )αβ ψ̂β,m, j + H.c.] gauge theory coupled to Weyl fermions in the 16-dimensional
i j αβ,m spinor representation.

+ (−)i n̂i . (B2)
i APPENDIX C: ENERGY SCALE AND MUTUAL
ENTANGLEMENT BETWEEN GAPLESS
The ground state is given by Ai j = 0. The emergent fermions AND GAPPED BOUNDARIES
are in a fully gapped product state and the bosonic model Ĥ1
gives rise to a U(16) gauge theory at low energies. The important issue is that when the d + 2D gapped bulk
Next, we reduce the U(16) gauge theory to the Spin(10) has no entanglements (i.e., no LRE or SRE by Definition 5),
gauge theory by adding a term, “the d+1D symmetric gapless boundary A” and “the d+1D
symmetric gapped boundary B” actually cannot affect each
Ĥ2 =  αβγ λ ψ̂α,m,i ψ̂β,m,i ψ̂γ ,m,i ψ̂λ,m,i + H.c. (B3) other, and thus are isolated from each other, in the following
i,m sense:
(i) Energy scale. Boundary A and Boundary B are decou-
to break the [U(16)]Nsite local symmetry to a [Spin(10)]Nsite
pled below the energy scale  E ,B . But when the energy is
local symmetry where the fermions ψα,m,i form the 16-
above the scale E ,B , the energy spectra of A, B, and the
dimensional spinor representation. Here,  αβγ λ is the anti-
bulk may affect and mix together with each other.
symmetric tensor which is invariant under the Spin(10) trans-
(ii) Mutual entanglement. Although the d+1D symmetric
formations. The 4 + 1D bosonic model Ĥ1 + Ĥ2 will give rise
gapless boundary A is highly entangled (due to the low-lying
to an emergent Spin(10) gauge theory with the fermions in a
massless chiral fermions as the energy gapless spectrum), and
fully gapped product state. Those fermions are also gapped on
the symmetric gapped boundary B is trivially gapped with
the boundary.
no entanglements as a tensor product state on Boundary B,
thanks to the trivial gapped bulk, Boundary A and Boundary
B on two sides have no entanglements in between. More
26
We introduce a new flavor parameter N f , so that we gain a benefit precisely, if we choose a d + 1D bipartite cut inside the
to do a large N f analysis for N f  1 or further, N f → ∞. d + 2D gapped bulk, we get a zero bipartite Von Neumann

023356-16
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

entanglement entropy, ground states, they must be in the same trivial SPT phase, and
thus the same trivial gapped vacuum, at least below the energy
SEE = S(ρA ) = − Tr[ρA ln ρA ] gap E of the bulk phase. This leads to a contradiction, and
= − Tr[ρB ln ρB ] = S(ρB ) = 0, (C1) thus we end the proof successfully.
“Constructing such a LUT deformation path.” This path
for the mutual entanglement between two sides. (This un- construction is basically what we had in the earlier proof.
derstanding is consistent with the entanglement structure dis- Since both phases are symmetric gapped trivial phases (both
cussed in Definition 5). a trivial SPT phase and a trivial gapped vacuum respect to
For the lattice regularization of the gauged chiral fermion the G f symmetry), the LUT deformation path is simply the
problem, we should emphasize that our statements in (i) and deformation to make both symmetric gapped trivial phases
(ii) apply to Model 1 (a chiral fermion theory) and Model become exactly the same symmetric gapped trivial tensor
2 (a gauged chiral fermion theory). However, we do not product states in a certain “canonical basis” with respect to
intend to apply our statements in (i) and (ii) to Model 3 (a the G f symmetry. (Normally, it is known as the symmetric
chiral gauge theory)—once we dynamically gauge the internal disordered phase, where the canonical basis is chosen to be
global symmetry for the bulk-boundary coupled system, then the dual variable of the symmetry-breaking basis).
“the bulk, Boundary A and Boundary B” form altogether
highly entangled quantum states (as a dynamical gauge the-
ory). The SEE of the dynamically gauged system (Model 3), APPENDIX E: COBORDISM THEORY AND
based on the previous bipartite cut in Eq. (C1), is generically CLASSIFICATION OF ALL POSSIBLE INVERTIBLE
nonzero. ANOMALIES RELATED TO SU(5) AND SO(10)
GRAND UNIFICATIONS

APPENDIX D: THE EXISTENCE OF SYMMETRIC Here we provide the cobordism group calculations clas-
GAPPED BOUNDARY VIA QUANTUM sifying all potential invertible ’t Hooft anomalies of SU(5),
PHASE TRANSITIONS Spin(10), and Spin(18) chiral fermion theories. Our calcu-
lations are crucial for showing all gauge anomaly-free con-
A derivation can also be obtained from the classification ditions for SU(5), SO(10), and SO(18) grand unifications.
of quantum phases of matter and their phase transitions. To Notice that other related work [66] computes
Spin×SU(5)
D
give a proof of Proposition 6, all we need to show is that
and
Spin×Spin(10)
D based on a different method, i.e., Atiyah-
there exists a LUT deformation path (Definition 5) between
Hirzebruch spectral sequence (AHSS), while our work focus
two bulk gapped quantum phases: Spin×Spin(10)/Z f
(1) Bulk phase: The d + 2D bulk regularized lattice model on
Spin×SU(5)
D and
D 2
, also based on a more
which has a symmetric gapped trivial tensor product ground powerful Adams spectral sequence. See, also, Ref. [55].
state, with a finite energy gap E .
(2) Trivial gapped vacuum phase (mentioned above),
1. Adams spectral sequence
such that this LUT deformation path satisfies the following
criteria: The Adams spectral sequence shows the following:
(a) it does not close the energy gaps between the bulk phase
∗ ∧
and the trivial vacuum phase [i.e., no gap closing, and thus no A p (H (Y, Z p ), Z p ) ⇒ πt−s (Y ) p ,
Ext s,t (E1)
gapless modes and no zero-mode degeneracy (=ground-state
degeneracy)]. where Ext denotes the extension functor, A p is the mod p
(b) it does not break the internal global symmetry given by Steenrod algebra, and Y is any spectrum. The H∗ (Y, Z p ) is
the bulk phase (i.e., G f = Spin(10) for the Spin(10) gauged an A p -module whose internal degree t is given by the ∗.
chiral fermion problem). The πt−s (Y )∧p is the p-completion of the (t − s)-th homo-
This LUT deformation path can be regarded as a path la- topy group of the spectrum Y . We note that for any finitely
beled by g in the quantum phase diagram (at zero temperature, generated Abelian group G, G p∧ = limn→∞ G/pn G is the p-
T = 0) by tuning a parameter (i.e., a coupling constant) g completion of G; if G contains an infinite group Z, then G p∧
of the lattice Hamiltonian Ĥ (g) such that the ground state is the ring of p-adic integers. Here, the G is meant to be
|g.s. (g) is unitarily evolving under this LUT along the defor- substituted by a homotopy group πt−s (Y )∧p in (E2). Here are
mation path. Then we can prove the claim of (a) and (b), either some explanations and inputs:
by “proof by a contradiction” or by directly “constructing such (1) Here the double-arrow “⇒” means “convergent to.”
a LUT deformation path.” The E2 page contains groups Exts,t with double indices
“A proof by a contradiction.” Suppose, given by engineer- (s, t ), and we reindex the bidegree by (t − s, s). There are
ing arbitrary symmetry-preserving (e.g., G f ) local interactions differentials d2 in the E2 page, which are arrows from (t −
for the lattice Hamiltonian, such a path in the phase diagram s, s) to (t − s − 1, s + 2). That is, Ext s,t → Ext s+2,t+1 . Take
is still impossible between two phases (the bulk phase and Kerd2 /Imd2 at each (t − s, s); then we get the E3 page. Repeat
the trivial gapped vacuum phase). Then there must be a phase this procedure, and we get the E4 page, E5 page, and so
transition between two phases, and the two phases should be on. Finally, the Er page equals the Er+1 page (there are no
different quantum phases—in fact, they should be different differentials) for r  N; we call this EN page as the E∞ page,
SPT phases within the G f symmetry. But as we emphasize and we can read the result πD at D = t − s. See further details
that both phases have symmetric gapped trivial tensor product discussed in Ref. [30], Sec. 2.3.

023356-17
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

(2) In Adams spectral sequence, we consider Ext s,t R (L, Z p ). s

Here we have the ring or the algebra R = A p or A2 (1) for
p = 2, and the L is a R-module. The A2 (1) is the subalgebra 5
of A2 generated by the Steenrod square Sq1 and Sq2 . The
index s refers to the degree of resolution, and the index t is
4
the internal degree of the R-module L. Ext groups are defined
by first taking a projective R-resolution P• of L, then, second,
computing the (co)homology group of the (co)chain complex 3
Hom(P• , Z p ). A P• is a resolution, which is an exact sequence
of modules. Here a projective R-resolution P• is an exact 2
sequence of R-modules · · · → Ps → Ps−1 → · · · → P0 → L,
where Ps is projective for s  0. 1

2. Thom-Madsen-Tillmann spectrum and 0

Pontryagin-Thom isomorphism
For Y = MT G, where MT G is the Thom-Madsen- 0 1 2 3 4 5 6 D =t−s
Tillmann spectrum MT G of a group G, the Adams spectral
sequence shows the following: FIG. 2. Adams chart for
Spin×SU(5) .
D
∗
Ext s,t
A p (H (MT G, Z p ), Z p )

∧ The ∧ is the smash product.
⇒ πt−s (MT G)∧p =
G D=t−s p . (E2) Below we will use (E2) and (E3) to compute the D-th
The last equality is that by the generalized Pontryagin-Thom bordism group of G given by
G D . Then we will use the
isomorphism, we have an equality between the D-th bordism (E5) and the techniques around (E6) to compute the D-th
group of G given by
G cobordism group of topological phases of G given by TPD (G).
D and the D-th homotopy group of
MT G given by πD (MT G), namely,
3. Cobordism groups and topological phases for Spin × SU(5):

G
D = πD (MT G). (E3) SU(5) grand unification
We also compute the cobordism group of topological We consider G = Spin × SU(5) for the Georgi-Glashow
phases (TP) defined in [29] as SU(5) grand unification [5]; the Thom-Madsen-Tillmann
spectrum MT G of the group G is
TPD (G). (E4)
MT G = MSpin ∧ (BSU(5))+ . (E8)
The TPD (G) classifies deformation classes of reflection posi-
tive invertible d-dimensional extended topological field theo- The T in MT G means the G-structures are on tangent bun-
ries with symmetry group GD . The TPD (G) and the bordism dles instead of normal bundles. For Spin, the Thom-Madsen-
group
G D are related by a short exact sequence,
Tillmann spectrum MT Spin = MSpin is equivalent to the


G  Thom spectrum, which splits MSpin = ko ∨  8 ko ∨ · · · . The
0 → Ext 1
G D , Z → TPD (G) → Hom
D+1 , Z → 0. ko is the (−1)-connected cover of the real K-theory spectrum.
(E5) The ∧ is the smash product and the ∨ is the wedge sum.
We can compute the E2 page of the A2 (1)-module based The (BSU(5))+ is the disjoint union of the classifying space
on Lemma 11 of [30]. More precisely, in order to compute BSU(5) and a point.27
A2 (1) (L2 , Z2 ), we find a short exact sequence of A2 (1)-
Ext s,t For the dimension D = t − s < 8, since there is no odd
modules, torsion,28 by MT G = MSpin ∧ X , then the D-th homotopy
0 → L1 → L2 → L3 → 0, (E6) group πD (MT G) = πD (ko ∧ X ) for D < 8. So, for the dimen-
sion D = t − s < 8, we have
then we apply Lemma 11 of [30] to compute Ext s,t A2 (1) (L2 , Z2 ) ∗

G ∧
A2 (1) (H (X, Z2 ), Z2 ) ⇒
D=t−s 2 .
Exts,t (E9)
by the given data of Ext s,t A2 (1) (L 1 , Z2 ) and Ext A2 (1) (L3 , Z2 ).
s,t

Our strategy is choosing L1 to be the direct sum of suspensions Hence, for MT G = MSpin ∧ (BSU(5))+ , for the dimen-
of Z2 on which Sq1 and Sq2 act trivially; then we take L3 to sion D = t − s < 8, by (E9), we have the Adams spectral
be the quotient of L2 by L1 . We can use this procedure again sequence,
A2 (1) (L3 , Z2 ) is determined.
and again until Ext s,t ∗ Spin×SU(5)
If G = Spin × G , then BG = B(Spin × G ) = BSpin × A2 (1) (H (BSU(5), Z2 ), Z2 ) ⇒
t−s
Ext s,t . (E10)
BG . By definition, the Madsen-Tillmann spectrum MT G =
Thom(BG, −V ) where V is the induced virtual bundle of
27
dimension 0 by the map BG → BO. By the properties of For a topological space X , it is a standard convention to denote
Thom space (see the discussions in Ref. [30], Sec. 1.3), X+ as the disjoint union of X and a point. Note that the reduced
we have cohomology of X+ is exactly the ordinary cohomology of X .
28
By computation using the mod p Adams spectral sequence for an
MT (Spin × G ) = MSpin ∧ (BG )+ . (E7) odd prime p, we find there is no odd torsion.

023356-18
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

TABLE I. Bordism group

Spin×SU(5)
D . σ is the signature of s
manifold. The c j is the jth Chern class of the associated vector
bundle of SU(n). Note that c3 = Sq2 c2 = (w2 + w12 )c2 = 0 mod 2
5
on Spin 6-manifolds. Actually,
Spin×SU(n)
D =
Spin×SU(n+1)
D for n  3
and 0  D  6. See, also, Ref. [55].
4
Bordism group
3
D
Spin×SU(5)
D Generators
σ
4 Z2 16
, c2 2
5 0
c3
6 Z 2 1

The A2 (1)-module structure of H∗ (BSU(5), Z2 ) below

0 1 2 3 4 5 6 D =t−s
degree 6 is shown in Ref. [55], Sec. 6, Fig. 29, and the E2 page
is shown in Fig. 2. Here we have used the correspondence Spin×Spin(10) Spin×Spin(18)
between the A2 (1)-module structure and the E2 page shown ZF ZF
FIG. 3. Adams chart for
D 2
; also for
D 2
.
in Appendix A of Ref. [55].
In the Adams chart, the horizontal axis labels the inte-
ger degree D = t − s and the vertical axis labels the in- 4. Cobordism groups and topological phases for Spin×Spin(10)
f and
teger degree s. The differential drs,t : Ers,t → Ers+r,t+r−1 is Spin×Spin(18)
Z2

an arrow starting at the bidegree (t − s, s) with direction f : SO(10) and SO(18) grand unification
Z2
Kerdrs,t
s,t
(−1, r). Er+1 := Imd s−r,t−r+1 for r  2. There exists N such We consider G = Spin×Spin(10)
for the Fritzsch-Minkowski
r
that EN+k = EN stabilized for k > 0; we denote the stabilized ZF2
SO(10) grand unification [6]; the Thom-Madsen-Tillmann
page E∞ := EN .
spectrum MT G of the group G is
To read the result from the Adams chart in Fig. 2, we look
at the stabilized E∞ page, where one dot indicates a finite MT G = MSpin ∧  −10 MSO(10). (E11)
group Z p , and a vertical finite line segment connecting n dots
indicates a finite group Z pn . But when n = ∞, the infinite The T in MT G means the G-structures are on tangent bundles
line connecting infinite dots indicates an infinite group, an instead of normal bundles. In this case, we have w2 (T M ) =
integer Z. Here, p is given by the mod p Steenrod algebra w2 (VSO(10) ).
A p in (E2). Here, in Fig. 2, p = 2, and we can read from the For the dimension D = t − s < 8, since there is no odd
Adams chart
Spin×SU(5) = Z (an infinite line),
Spin×SU(5) = torsion (see footnote 28), by MT G = MSpin ∧ X , then
0 1
Spin×SU(5) πD (MT G) = πD (ko ∧ X ) for D < 8; so, for the dimension
Z2 (a dot),
2 = Z2 (a dot),
Spin×SU(5) = 0 (noth-
3 D = t − s < 8, from (E2), we have
ing),
Spin×SU(5)
4 = Z 2
(two infinite lines),
Spin×SU(5)
5 =0
G ∧
∗
A2 (1) (H (X, Z2 ), Z2 ) ⇒
D=t−s 2 .
Exts,t
Spin×SU(5)
(nothing), and
6 = 0 (an infinite line). (E12)
Hence, we have the Adams spectral sequence,
Spin×Spin(10)
Classification of all invertible anomalies ZF
∗+10
of Spin × SU(5) fermion theories Exts,t
A2 (1) (H (MSO(10), Z2 ), Z2 ) ⇒
D=t−s 2
. (E13)
By (E2) and (E3), we obtain the bordism group
Spin×SU(5)
D Actually, we find [55]
shown in Table I, focusing on D = 4, 5, 6. Spin×Spin(10) Spin×Spin(18) Spin×Spin(n) Spin×Spin(n+1)
By (E5) and (E6), we obtain the cobordism group ZF ZF ZF ZF
TPD (Spin × SU(5)) shown in Table II, focusing on D = 4, 5.
D 2
=
D 2
=
D 2
=
D 2
,
for n  7 and 0  D  6.

TABLE II. Topological phase classification (≡ TP) as a cobor- TABLE III. Bordism group. The same result holds for
dism group TPD (Spin × SU(5)), following Table I. See, also, Spin×Spin(18) Spin×Spin(n)
ZF ZF
Ref. [55].
D 2
and
D 2
with n  7 and 0  D  6. See Ref. [55].

Cobordism group Bordism group

d TPD (Spin × SU(5)) Generators Spin×Spin(10)
f
Z
D
D 2
Generators
4 0
5 Z 1
2
CSSU(5)
5 5 Z2 w2 (T M )w3 (T M ) = w2 (VSO(10) )w3 (VSO(10) )

023356-19
JUVEN WANG AND XIAO-GANG WEN PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

TABLE IV. Topological phase classification (≡ TP) as a cobor- line connecting infinite dots indicates an infinite group, an
dism group, following Table III. Same result for TPD ( Spin×Spin(18)
F
) integer Z. Here, p is given by the mod p Steenrod algebra
Z2
and TPD ( Spin×Spin(n) ) with n  7 and 0  D  5. See, also, Ref. [55]. A p in (E2). Here, in Fig. 3, p = 2, and we can read from the
F Spin×Spin(10) Spin×Spin(10)
Z2 f f
Adams chart
0 = Z (an infinite line),
1 =
Z Z
2 2

Spin×Spin(10) Spin×Spin(10)
Cobordism group f f

Spin×Spin(10)  0 (nothing),
2
Z
2
= 0 (nothing),
3
Z
2
= 0 (noth-
D TPD f Generators Spin×Spin(10) Spin×Spin(10)
Z2 f f
ing),
4 = Z2 (two infinite lines),
5 = Z2 (a
Z Z
2 2
5 Z2 w2 (T M )w3 (T M ) = w2 (VSO(10) )w3 (VSO(10) ) Spin×Spin(10)
f
dot), and
6 = Z2 (a dot).
Z
2

The A2 (1)-module structure of H∗+10 (MSO(10), Z2 ) be-

Spin×Spin(10)
low degree 6 is shown in Ref. [55], Sec. 6, Fig. 27, and Classification of all invertible anomalies of f and
Z2
the E2 page is shown in Fig. 3. Here we have used the Spin×Spin(18)
fermion theories
correspondence between the A2 (1)-module structure and the
f
Z2
Spin×Spin(10)
E2 page shown in Appendix A of Ref. [55]. f
By (E2) and (E3), we obtain the bordism group
D 2
Z
To read the result from the Adams chart in Fig. 3, we look
at the stabilized E∞ page, where one dot indicates a finite shown in Table III, focusing on D = 5.
group Z p , and a vertical finite line segment connecting n dots By (E5) and (E6), we obtain the cobordism group
indicates a finite group Z pn . But when n = ∞, the infinite TPD ( Spin×Spin(10)
f ) shown in Table IV, focusing on D = 5.
Z2

[1] S. L. Glashow, Nucl. Phys. 22, 579 (1961). [30] Z. Wan and J. Wang, Ann. Math. Sci. Appl. 4, 107 (2019).
[2] S. Weinberg, Phys. Rev. Lett. 19, 1264 (1967). [31] Z. Wan, J. Wang, and Y. Zheng, arXiv:1912.13504.
[3] A. Salam and J. Ward, Phys. Lett. 13, 168 (1964). [32] Z. Wan and J. Wang, Nucl. Phys. B (2019),
[4] D. Tong, J. High Energy Phys. 07 (2017) 104. doi:10.1016/j.nuclphysb.2020.115016.
[5] H. Georgi and S. L. Glashow, Phys. Rev. Lett. 32, 438 (1974). [33] D. Gaiotto, A. Kapustin, N. Seiberg, and B. Willett, J. High
[6] H. Fritzsch and P. Minkowski, Ann. Phys. 93, 193 (1975). Energy Phys. 02 (2015) 172.
[7] T. D. Lee and C. N. Yang, Phys. Rev. 104, 254 (1956). [34] A. Kitaev, Ann. Phys. 303, 2 (2003).
[8] H. B. Nielsen and M. Ninomiya, Phys. Lett. B 105, 219 (1981). [35] X.-G. Wen, Phys. Rev. Lett. 90, 016803 (2003).
[9] J. B. Kogut, Rev. Mod. Phys. 51, 659 (1979). [36] M. Levin and X.-G. Wen, Phys. Rev. B 67, 245316
[10] P. H. Ginsparg and K. G. Wilson, Phys. Rev. D 25, 2649 (2003).
(1982). [37] B. Yoshida, Ann. Phys. 326, 15 (2011).
[11] D. B. Kaplan, Phys. Lett. B 288, 342 (1992). [38] H. Bombín, Commun. Math. Phys. 327, 387 (2014).
[12] Y. Shamir, Nucl. Phys. B 406, 90 (1993). [39] Z. Nussinov and G. Ortiz, Proc. Natl. Acad. Sci. 106, 16944
[13] R. Narayanan and H. Neuberger, Phys. Lett. B 302, 62 (1993). (2009).
[14] M. Lüscher, Nucl. Phys. B 549, 295 (1999). [40] Z. Nussinov and G. Ortiz, Ann. Phys. 324, 977 (2009).
[15] X. Chen, Z.-X. Liu, and X.-G. Wen, Phys. Rev. B 84, 235141 [41] L. Kong and X.-G. Wen, arXiv:1405.5858.
(2011). [42] D. S. Freed, arXiv:1406.7278.
[16] X. Chen, Z.-C. Gu, Z.-X. Liu, and X.-G. Wen, Phys. Rev. B [43] S. Galatius, I. Madsen, U. Tillmann, and M. Weiss, Acta Math.
87, 155114 (2013). 202, 195 (2009).
[17] J. Wang and X.-G. Wen, arXiv:1307.7480. [44] R. Thom, Comment. Math. Helvetici 28, 17 (1954).
[18] J. Wang, X.-G. Wen, and E. Witten, Phys. Rev. X 8, 031048 [45] X.-G. Wen, Phys. Rev. B 99, 205139 (2019).
(2018). [46] Z.-C. Gu and X.-G. Wen, Phys. Rev. B 80, 155131 (2009).
[19] E. Eichten and J. Preskill, Nucl. Phys. B 268, 179 (1986). [47] T. Senthil, Ann. Rev. Condensed Matter Phys. 6, 299 (2015).
[20] I. Montvay, Nucl. Phys. B Proc. Suppl. 29, 159 (1992). [48] A. Kapustin, arXiv:1403.1467.
[21] P. D. V. Swift, Phys. Lett. B 145, 256 (1984). [49] A. Kapustin, R. Thorngren, A. Turzillo, and Z. Wang, J. High
[22] J. Smit, Acta Phys. Pol. B 17, 531 (1986). Energy Phys. 12 (2015) 52.
[23] X.-G. Wen, Chin. Phys. Lett. 30, 111101 (2013). [50] M. Guo, P. Putrov, and J. Wang, Ann. Phys. 394, 244 (2018).
[24] M. F. Golterman, D. N. Petcher, and E. Rivas, Nucl. Phys. B [51] C. Wang and T. Senthil, Phys. Rev. B 89, 195124 (2014); 91,
395, 596 (1993). 239902(E) (2015).
[25] L. Lin, Phys. Lett. B 331, 449 (1994). [52] Q.-R. Wang and Z.-C. Gu, Phys. Rev. X 8, 011055 (2018).
[26] M. F. L. Golterman and Y. Shamir, Phys. Rev. D 51, 3026 [53] Q.-R. Wang and Z.-C. Gu, arXiv:1811.00536.
(1995). [54] X. Chen, Z.-C. Gu, and X.-G. Wen, Phys. Rev. B 82, 155138
[27] C. Chen, J. Giedt, and E. Poppitz, J. High Energy Phys. 04 (2010).
(2013) 131. [55] Z. Wan and J. Wang, arXiv:1910.14668.
[28] T. Banks and A. Dabholkar, Phys. Rev. D 46, 4016 (1992). [56] A. P. Schnyder, S. Ryu, A. Furusaki, and A. W. W. Ludwig,
[29] D. S. Freed and M. J. Hopkins, arXiv:1604.06527. Phys. Rev. B 78, 195125 (2008).

023356-20
NONPERTURBATIVE DEFINITION OF THE STANDARD … PHYSICAL REVIEW RESEARCH 2, 023356 (2020)

[57] A. Kitaev, in Advances in Theoretical Physics: Landau Memo- [80] E. Dagotto, E. Fradkin, and A. Moreo, Phys. Rev. B 38, 2926
rial Conference, Chernogolovka, Russia, 2008, AIP Conf. (1988).
Proc. No. 1134, edited by V. Lebedev and M. Feigelman (AIP, [81] R. C. Ball, Phys. Rev. Lett. 95, 176407 (2005).
Melville, NY, 2009), p. 22. [82] F. Wilczek and A. Zee, Phys. Rev. D 25, 553 (1982).
[58] A. P. Schnyder, S. Ryu, A. Furusaki, and A. W. W. Ludwig, [83] X.-G. Wen, Phys. Rev. Lett. 88, 011602 (2001).
in Advances in Theoretical Physics: Landau Memorial Con- [84] X.-G. Wen, Phys. Rev. D 68, 065003 (2003).
ference, edited by V. Lebedev and M. Feigelman, AIP Conf. [85] M. Levin and X.-G. Wen, Phys. Rev. B 73, 035122 (2006).
Proc. No. 1134 (AIP, New York, 2009), p. 10. [86] J. A. Wheeler, Information, physics, quantum: The search for
[59] Z.-C. Gu and X.-G. Wen, Phys. Rev. B 90, 115141 (2014). links, in Feynman and Computation: Exploring the Limits of
[60] D. Gaiotto and A. Kapustin, Int. J. Mod. Phys. A 31, 1645044 Computers (Perseus Books, USA, 1989), pp. 309–336.
(2016). [87] M. A. Levin and X.-G. Wen, Phys. Rev. B 71, 045110 (2005).
[61] A. Kapustin and R. Thorngren, J. High Energy Phys. 10 (2017) [88] X.-G. Wen, Condens. Matter Phys. 20, 198710 (2013).
080. [89] G. ’t Hooft, NATO Adv. Study Inst. Ser. B Phys. 59, 135
[62] J. Wang, K. Ohmori, P. Putrov, Y. Zheng, Z. Wan, M. Guo, (1980).
H. Lin, P. Gao, and S.-T. Yau, Prog. Theor. Exp. Phys. 2018, [90] X.-G. Wen, Phys. Rev. D 88, 045013 (2013).
053A01 (2018). [91] C. Callan and J. Harvey, Nucl. Phys. B 250, 427 (1985).
[63] M. Guo, K. Ohmori, P. Putrov, Z. Wan, and J. Wang, Commun. [92] D. S. Freed, J. Differential Geom. 80, 45 (2008).
Math. Phys. 376, 1073 (2020). [93] E. Witten, Commun. Math. Phys. 100, 197 (1985).
[64] T. Lan, C. Zhu, and X.-G. Wen, Phys. Rev. B 100, 235141 [94] X. Dai and D. S. Freed, J. Math. Phys. 35, 5155 (1994).
(2019).. [95] L. Kong, X.-G. Wen, and H. Zheng, arXiv:1502.01690.
[65] Z. Wan, J. Wang, and Y. Zheng, Ann. Phys. 414, 168074 [96] D. Fiorenza and A. Valentino, Commun. Math. Phys. 338,
(2020). 1043 (2015).
[66] I. García-Etxebarria and M. Montero, J. High Energy Phys. 08 [97] S. Monnier, Commun. Math. Phys. 338, 1327 (2015).
(2019) 003. [98] W. Ji and X.-G. Wen, Phys. Rev. Res. 1, 033054 (2019).
[67] J. Wang, X.-G. Wen, and E. Witten, J. Math. Phys. 60, 052301 [99] J. C. Wang, L. H. Santos, and X.-G. Wen, Phys. Rev. B 91,
(2019). 195134 (2015).
[68] J. Milnor and J. Stasheff, Characteristic Classes, by Milnor [100] A. Kapustin and R. Thorngren, arXiv:1404.3230.
and Stasheff, Annals of Mathematics Studies, Vol. 76 (Prince- [101] E. Witten, Phys. Lett. B 117, 324 (1982).
ton University Press, Princeton, NJ, 1974). [102] X. G. Wen, Int. J. Mod. Phys. B 04, 239 (1990).
[69] C. Arf, J. Reine Angew. Math. 183, 148 (1941). [103] A. Kapustin, arXiv:1404.6659.
[70] A. Y. Kitaev, Phys. Usp. 44, 131 (2001). [104] J. C. Wang, Z.-C. Gu, and X.-G. Wen, Phys. Rev. Lett. 114,
[71] M. DeMarco and X.-G. Wen, arXiv:1706.04648. 031601 (2015).
[72] Y. BenTov and A. Zee, Phys. Rev. D 93, 065036 (2016). [105] A. Vishwanath and T. Senthil, Phys. Rev. X 3, 011016 (2013).
[73] Y.-Z. You, Y. BenTov, and C. Xu, arXiv:1402.4151. [106] S. L. Adler, Phys. Rev. 177, 2426 (1969).
[74] Y.-Z. You and C. Xu, Phys. Rev. B 91, 125147 (2015). [107] J. Bell and R. Jackiw, Nuovo Cimento A 60, 47 (1969).
[75] Y. Kikukawa, Prog. Theor. Exp. Phys. 2019, 073B02 (2019). [108] J. C. Wang and X.-G. Wen, Phys. Rev. B 91, 125124 (2015).
[76] Y. Kikukawa, Prog. Theor. Exp. Phys. 2019, 113B03 (2019). [109] A. Kapustin, Phys. Rev. B 89, 125307 (2014).
[77] J. Wang and X.-G. Wen, Phys. Rev. D 99, 111501(R) [110] A. Kitaev and L. Kong, Commun. Math. Phys. 313, 351
(2019). (2012).
[78] G. Baskaran and P. W. Anderson, Phys. Rev. B 37, 580 (1988). [111] T. Lan, J. C. Wang, and X.-G. Wen, Phys. Rev. Lett. 114,
[79] I. Affleck, Z. Zou, T. Hsu, and P. W. Anderson, Phys. Rev. B 076402 (2015).
38, 745 (1988). [112] N. Seiberg (unpublished).

023356-21