Fermion Chirality from Non-Bipartite Topology:
Geometric Doubler Lifting on the FCC Lattice
via Holographic U(1)/Z
2
Phase Projection
Raghu Kulkarni
SSMTheory Group, IDrive Inc., Calabasas, CA 91302, USA
raghu@idrive.com
April 2026
Abstract
We construct and analyse the bond-direction Dirac operator on the Face-Centred Cubic
(FCC) lattice using all 12 nearest-neighbour unit bond directions:
D
SSM
(k) =
P
12
j=1
(γ ·
ˆ
n
j
) e
ik·
ˆ
n
j
. The operator satises
{γ
5
, D
SSM
} = 0
exactly at nite lattice spacing (exact
chiral symmetry, proved algebraically).
Our main analytical result is the
Irrational Doubler Theorem
: every non-
Γ
zero of
D
SSM
has at least one irrational FCC fractional coordinate
f
i
= ±1/(2
√
2)
, placing
it permanently outside any nite rational momentum grid. Two types of doubler zero
exist (Section 3): isolated Z
2
zeros (Type-1, all bond phases
∈ {+1, −1}
) and at-band
U(1) zeros (Type-2, generically complex bond phases); both types share the irrational
fractional coordinate property. This is proved analytically via the V-form structure and
conrmed by a dense
32
3
FCC BZ scan: among
32,768
sampled momenta, no non-
Γ
mode has
E = 0
.
This Z
2
/U(1) distinction provides a geometric basis for the
holographic projection
of the SSM framework [9]: physical electrons carry U(1) electric charge and require
genuinely complex (propagating) bond phases; staggered Z
2
congurations are static,
carry no U(1) charge, and are identied as non-propagating codespace states rather than
physical particles. After this projection, all zone-boundary high-symmetry modes (
X
,
W
,
L
,
K
) carry U(1) phases and are lifted to UV energies
E ≈ 1
5/a
, while a single
massless Dirac mode with exact chiral symmetry persists at
Γ
.
The Nielsen-Ninomiya theorem is satised globally on the continuous torus: all non-
Γ
zeros reside at irrational FCC fractional coordinates and are inaccessible to any nite
integer-
L
simulation grid. In the SSM physical sector, the unique zero is at
Γ
.
1 Introduction
The Nielsen-Ninomiya (NN) theorem [1] is a fundamental topological obstruction to single-
species lattice fermions: any local, Hermitian, translationally invariant Dirac operator on
the continuous torus
T
3
must carry equal numbers of left- and right-handed zeros. On the
standard hypercubic lattice these doublers appear at the rational zone-boundary momenta
k
µ
= π/a
and are directly accessible to nite simulations. Standard remedies all impose a cost:
Wilson fermions [2] lift doublers at the price of chiral symmetry; staggered fermions [3] reduce
but do not eliminate doubling; domain-wall [4] and overlap fermions [6] restore modied chiral
1
symmetry (Ginsparg-Wilson [5]) at high computational overhead.
In this paper we study the bond-direction operator on the non-bipartite FCC lattice. The
main analytical contribution is the Irrational Doubler Theorem (Section 4): every non-
Γ
zero
of
D
SSM
has at least one irrational FCC fractional coordinate
f
i
= ±1/(2
√
2)
, making it
unreachable by any nite FCC lattice of integer size
L
. This is a rigorous characterisation of
the doubler structure that is new to the lattice fermion literature.
The physical interpretation of this theorem comes from the SSM (Sparse-Simplex Matrix)
holographic framework: in the SSM vacuum (a quantum error-correcting code on the FCC
lattice), physical electrons are U(1)-charged propagating defects whose kinematics require
complex bond phases. Staggered Z
2
congurations are static codespace states without U(1)
charge. The doublers are therefore physically inert: they are present in the full BZ (NN
theorem is respected) but are not accessible to U(1)-charged particles.
Scope.
All results are for the 3D spatial FCC lattice. We do not claim a spectral gap
in the strict thermodynamic limit
L → ∞
: the U(1) mode energies can be made small by
approaching the Type-2 at-band surfaces, though never exactly zero on the FCC lattice
with nite integer
L
(since all non-
Γ
zeros reside at irrational fractional BZ coordinates,
Section 3). The SSM holographic projection provides the physical mechanism by which Z
2
modes are excluded from the physical spectrum. The 4D extension to the
D
4
root lattice is
left as future work.
2 The FCC Bond-Direction Dirac Operator
2.1 Construction
The FCC lattice with cubic cell parameter
a
has 12 nearest-neighbour unit bond-direction
vectors:
ˆ
n
j
∈
1
√
2
{(±1, ±1, 0), (±1, 0, ±1), (0, ±1, ±1)}.
(1)
The bond-direction Dirac operator in momentum space is:
D
SSM
(k) =
12
X
j=1
(γ ·
ˆ
n
j
) e
ik·
ˆ
n
j
,
(2)
with
γ
µ
the anti-Hermitian spatial Dirac matrices (
γ
†
µ
= −γ
µ
). Setting
a = 1
throughout. The
bond vectors satisfy
S
µν
=
P
j
ˆn
µ
j
ˆn
ν
j
= 4δ
µν
(exact spatial isotropy, Fermi velocity
c
F
= 4
).
The FCC lattice achieves the kissing number
K = 12
in three dimensions the maximum
number of non-overlapping unit spheres that can simultaneously touch a central sphere
making
D
SSM
the unique maximally symmetric bond-direction operator in 3D.
Continuum limit.
Expanding
e
ik·
ˆ
n
j
≈ 1 + ik ·
ˆ
n
j
for small
|k|
:
D
SSM
→ iγ
µ
k
ν
P
j
ˆn
µ
j
ˆn
ν
j
=
4iγ · k
, recovering the standard massless Dirac equation with Fermi velocity
c
F
= 4
. The
FCC non-bipartite geometry does not break Lorentz symmetry at long wavelengths.
Relation to the standard FCC tight-binding operator.
The
standard
FCC oper-
ator uses physical bond position vectors
(a/2)(±1, ±1, 0)
(length
a/
√
2
) in the exponent:
D
std
(k) =
P
j
(γ ·
ˆ
n
j
)e
ik·(a/2)(±1,±1,0)
.
D
std
has zeros at the L and X high-symmetry points of
the FCC BZ (the bond phase
e
i(π/a)·(a/2)·2
= e
iπ
= −1
at L causes the vector sum to cancel),
2
consistent with the Nielsen-Ninomiya theorem.
D
SSM
uses unit-length directions in the ex-
ponent and is related to
D
std
by a momentum rescaling
k → k
√
2
: this shifts the doublers
from the rational L, X positions of
D
std
to the irrational positions characterised in Section 3.
2.2 Hermiticity and Exact Chiral Symmetry
Proposition 1
(Hermiticity)
.
D
†
SSM
(k) = D
SSM
(k)
for all
k
.
Proof.
The 12 bond vectors form 6 antipodal pairs
ˆ
n
j
′
= −
ˆ
n
j
. Taking the adjoint:
D
†
=
P
j
(−γ ·
ˆ
n
j
)e
−ik·
ˆ
n
j
. Relabelling
j → j
′
cancels the sign:
D
†
= D
. Veried numerically:
∥D − D
†
∥ < 10
−14
.
□
Proposition 2
(Exact chiral symmetry)
.
{γ
5
, D
SSM
(k)} = 0
for all
k
and all
a
.
Proof.
Since
{γ
5
, γ
µ
} = 0
for all spatial
µ
:
{γ
5
, γ ·
ˆ
n
j
} = ˆn
µ
j
{γ
5
, γ
µ
} = 0
for all
j
, hence
{γ
5
, D
SSM
} =
P
j
0 · e
ik·
ˆ
n
j
= 0
.
□
This distinguishes
D
SSM
from Wilson fermions, which add an identity-valued
O(a)
term that
commutes with
γ
5
, breaking chiral symmetry.
3 Zero Structure on the Continuous Torus
3.1 The V-Form
Using antipodal symmetry the operator rewrites as
D
SSM
(k) = iγ
µ
V
µ
(k)
, where
V
µ
(k) =
P
j
ˆn
µ
j
sin(k ·
ˆ
n
j
)
. Factoring:
V
x
= 2
√
2 sin(k
x
/
√
2)
cos(k
y
/
√
2) + cos(k
z
/
√
2)
,
(3)
and cyclically.
D
SSM
= 0
if and only if
V
µ
= 0
for all
µ
.
3.2 The
Γ
-Point Physical Zero
At
k = 0
, all phases equal
+1
and
D
SSM
(0) = γ ·
P
j
ˆ
n
j
= 0
by antipodal cancellation. The
long-wavelength expansion gives
D
SSM
≈ 4iγ · k
(see Section 2 for the derivation).
3.3 Doubler Zeros and Their FCC Fractional Coordinates
On the continuous torus
T
3
, additional zeros required by the NN theorem include:
Type-1 (isolated):
k
d
=
π
√
2
(±1, ±1, ±1)
, where
k
d
·
ˆ
n
j
∈ {0, ±π}
for all bonds, so
e
ik
d
·
ˆ
n
j
∈
{+1, −1}
. In FCC fractional coordinates
k = f
1
b
1
+f
2
b
2
+f
3
b
3
(with
b
i
the FCC reciprocal
basis):
f
i
= ±1/(2
√
2) ≈ ±0.3536
irrational.
Type-2 (at surfaces):
Entire planes such as
k
y
= −π
√
2
,
k
z
= 0
(for all
k
x
), where (3)
forces
V
x
= V
y
= V
z
= 0
for all
k
x
(since
sin(k
y
/
√
2) = sin(−π) = 0
and
cos(k
y
/
√
2) +
cos(k
z
/
√
2) = cos(−π)+cos(0) = 0
). The relevant component has FCC fractional coordinate
|f
i
| = 1/(2
√
2)
(irrational).
Both types lie at irrational FCC fractional coordinates and are never exactly sampled by any
nite periodic FCC lattice of integer size
L
(veried for
L = 1
to
500
). The NN theorem is
globally satised:
Γ
carries chiral index
+1
; the doublers carry compensating indices such
that all topological charges sum to zero on
T
3
.
3
4 The Irrational Doubler Theorem
Denition 3
(Phase character)
.
k
is a
Z
2
mode
if
e
ik·
ˆ
n
j
∈ {+1, −1}
for all 12 bonds. It is
a
U(1) mode
if at least one bond phase
e
ik·
ˆ
n
j
/∈ R
. Z
2
modes have
sin(k ·
ˆ
n
j
) = 0
for all
j
;
U(1) modes have
sin(k ·
ˆ
n
j
) = 0
for some
j
.
Theorem 4
(Irrational Doubler Theorem)
.
Every non-
Γ
zero of
D
SSM
has at least one ir-
rational FCC fractional coordinate
f
i
= ±1/(2
√
2)
. Consequently, no non-
Γ
zero of
D
SSM
is
sampled by any nite FCC lattice of integer size
L
.
Proof. V-form structure.
D
SSM
= 0
i
V
µ
= 0
for
µ = x, y, z
simultaneously. From (3),
V
x
= 0
requires either
sin(k
x
/
√
2) = 0
or
cos(k
y
/
√
2) + cos(k
z
/
√
2) = 0
(and cyclically). In
either case, one component satises
k
i
/
√
2 = n
i
π
(
n
i
a nonzero integer for
k = 0
), giving
k
i
= n
i
π
√
2
.
Irrational fractional coordinates.
Converting to FCC fractional coordinates
k =
P
i
f
i
b
i
via
the inverse of (1), one nds
f
j
= ±n
j
/(2
√
2)
for the component(s) with
k
j
= n
j
π
√
2
. Since
√
2
is irrational,
f
j
= n
j
/(2
√
2) = p/L
for any integers
p, L
with
n
j
= 0
. Hence every non-
Γ
zero lies at an irrational FCC fractional coordinate.
Classication of zero types.
Type-1 (isolated, Section 3): all sines zero,
f
1
= f
2
= f
3
=
±1/(2
√
2)
, all bond phases
∈ {+1, −1}
(Z
2
). Type-2 (at surfaces): one sine and one cosine-
sum cancel simultaneously; one fractional coordinate is
±1/(2
√
2)
(the others vary). On
these surfaces the bond phases are generically complex (U(1)), but the irrational coordinate
f
i
= ±1/(2
√
2)
persists for all points on the surface, guaranteeing they are never sampled by
any integer-
L
grid.
Numerical conrmation.
A dense
32
3
scan of rational FCC BZ momenta (Section 6) nds no
non-
Γ
zero, consistent with the irrational coordinate result.
□
Within the Z
2
class, two subclasses are physically distinct:
Remark 5
(Zero classication)
.
Uniform Z
2
(
k = 0
, all phases
= +1
): the physical
Dirac cone.
Staggered Z
2
(Type-1 isolated zeros, all phases
∈ {+1, −1}
): at
f
i
= ±1/(2
√
2)
, ex-
cluded by holographic projection and irrational coordinates.
Type-2 at-band zeros
(partially U(1), generically complex phases): also at
f
i
=
±1/(2
√
2)
in at least one component, never sampled by nite integer-
L
grids.
Table 1 shows the phase character at standard high-symmetry points. All zone-boundary
high-symmetry points are U(1) (complex phases, gapped); the K-point minimum
E
K
=
0.01372/a
is the smallest energy on the standard high-symmetry path.
5 Holographic Projection and the Physical Spectrum
5.1 Physical Motivation in the SSM Framework
In the SSM vacuum [9], the FCC vacuum is modelled as a quantum error-correcting (QEC)
code. Physical electrons are propagating topological defects carrying U(1) electric charge.
For a charge-carrying defect to propagate through the bond network, each hopping amplitude
must be a genuinely complex U(1) phase, enabling Berry phase accumulation and net charge
transport.
A mode with all bond phases
∈ {+1, −1}
is a
staggered Z
2
conguration
: the hopping
4
Table 1: Phase character and energy at FCC BZ high-symmetry points. All zone-boundary
points are at rational FCC coordinates and are strictly gapped. The unique zero at rational
coordinates is the physical
Γ
mode.
∗
Type-2 at-band zeros have
E = 0
on the continuous
torus but lie at irrational coordinates (
f
i
= ±1/(2
√
2)
) and are never sampled by any nite
integer-
L
lattice; max
|Im| > 0
for generic points on the surface.
Point Fractional coords
max |Im(e
ik·
ˆ
n
)|
Phase type
E
Γ (0, 0, 0) 0
Z
2
(uniform)
0
X (0,
1
2
,
1
2
) 0.964
U(1)
5.45/a
W (
1
4
,
1
2
,
3
4
) 0.964
U(1)
1.97/a
L (
1
2
,
1
2
,
1
2
) 0.964
U(1)
4.72/a
K (
3
8
,
3
8
,
3
4
) 0.372
U(1)
0.01372/a
Type-1 doubler
f
i
=
1
2
√
2
0
Z
2
(staggered)
0
Type-2 at band
|f
relevant
| =
1
2
√
2
> 0
U(1) (irrational coords)
0
∗
amplitudes are real-valued, no net current can ow, and no U(1) charge is transported. Such
modes are frozen codespace states of the QEC vacuum they correspond to standing-
wave patterns whose interference destructively cancels all charge currents. They are present
in the quantum vacuum of the SSM, but as inert background congurations, not as physical
propagating particles.
Denition 6
(Holographic projection)
.
The physical (U(1)) sector of
D
SSM
consists of:
1. all U(1) modes (
e
ik·
ˆ
n
j
/∈ R
for at least one bond), plus
2. the unique uniform Z
2
mode
k = 0
(the Dirac vacuum).
Non-uniform Z
2
(staggered) modes are excluded as non-propagating.
5.2 Consequence: Single Physical Dirac Mode
Theorem 7
(Single zero on nite FCC lattices)
.
On any nite FCC lattice with integer size
L
,
D
SSM
has exactly one zero: the
Γ
-point. After holographic projection (Denition 6), this
is the unique physical massless mode.
Proof.
By Theorem 4, every non-
Γ
zero has
f
i
= ±1/(2
√
2)
for at least one component. Since
1/(2
√
2)
is irrational, no such
f
i
equals
p/L
for any integers
p, L
. Therefore no non-
Γ
zero
falls on the
L
-site grid. The holographic projection additionally excludes staggered Z
2
modes
from the physical spectrum on physical grounds.
□
5.3 Dual Protection: Irrational BZ Coordinates
The doublers additionally satisfy a kinematic protection on nite lattices. In FCC fractional
coordinates, all Z
2
doubler modes reside at
f
i
= ±1/(2
√
2)
(irrational) or on at surfaces at
the same irrational coordinate in one component. Since
√
2
is irrational,
f
i
= n/L
has no
solution for any integer
L
. This holds regardless of lattice size, providing a second independent
reason the doublers never appear in any nite-lattice simulation of the FCC crystal.
Together: doublers are excluded (i) kinematically, by irrational FCC coordinates (never
sampled by any integer-
L
lattice proved in Theorem 4), and (ii) physically, by holographic
projection for Type-1 Z
2
modes (which carry no U(1) charge). Type-2 at-band zeros are
U(1) modes and are not excluded by holographic projection; they are excluded solely by the
5
irrational coordinate argument.
5.4 Comparison with Existing Approaches
Wilson fermions [2]
: add
r
P
µ
(1 −cos k
µ
a) ·1
, breaking chiral symmetry via an identity-
valued term.
D
SSM
has no such term; all operators anticommute with
γ
5
.
Overlap fermions [6]
: realise the Ginsparg-Wilson relation
{γ
5
, D} = Dγ
5
D
at high com-
putational cost.
D
SSM
achieves the stronger
{γ
5
, D} = 0
through geometry alone.
Creutz/hyperdiamond [7, 8]
: exploit non-bipartite geometry in 2D/4D. The present 3D
construction additionally provides the Z
2
/U(1) mechanism.
6 Numerical Results
6.1 Dispersion Along the High-Symmetry Path
Figure 1 shows the dispersion
E(k)
along the standard FCC path
Γ
-
X
-
W
-
L
-
Γ
-
K
. Blue
segments are U(1) modes (complex phases, all gapped); red segments mark Z
2
-phase path
points (bond phases
∈ {+1, −1}
). A single massless mode exists at
Γ
; all zone-boundary
modes are lifted.
Figure 1: Dispersion
E(k)
along the FCC high-symmetry path
Γ
-
X
-
W
-
L
-
Γ
-
K
. Blue: U(1)
modes (complex bond phases, all gapped). Red: Z
2
-phase path points (bond phases
∈
{+1, −1}
), excluded by holographic projection. Single massless mode at
Γ
; all zone-boundary
modes lifted to UV energies.
6.2
32
3
BZ Scan with Phase Classication
We scan
N = 32
in each FCC BZ direction (
32
3
= 32,768
k-points, fractional coordinates
f
i
∈ [−0.5, 0.5)
) and classify each point by phase character:
6
U(1) modes
:
32,767
sampled points, all with
E > 0
. Minimum
E
across the full scan:
0.0013/a
(from grid proximity to the Type-2 at-band surfaces at irrational BZ positions
these are not exact zeros of the operator). Minimum
E
on the high-symmetry path:
E
K
= 0.01372/a
(K-point).
Z
2
modes
:
1
sampled point the
Γ
-point (all phases
= +1
).
E = 0
.
All non-
Γ
doubler zeros
(Type-1 Z
2
and Type-2 U(1)): at irrational BZ coordinates
(
f
i
= ±1/(2
√
2)
in at least one component), never sampled by any integer-
L
grid.
This conrms Theorem 4: no non-
Γ
zero is sampled by the rational grid. Figure 2 shows the
energy histograms for U(1) and Z
2
modes separately.
Figure 2: U(1)/Z
2
phase decomposition of the
32
3
FCC BZ scan.
Left
: U(1) modes (
32,767
points), all with
E > 0
. The dashed line marks the K-point gap (
0.01372/a
), the minimum on
the high-symmetry path. The scan minimum (
0.0013/a
) arises from grid proximity to Type-
2 at-band surfaces at irrational BZ coordinates; these are not exact zeros of the operator.
Right
: Z
2
modes (
1
sampled point:
Γ
). All non-
Γ
zeros (Type-1 and Type-2) are at irrational
coordinates and absent from the scan grid.
6.3 Zone-Boundary Spectrum
Table 2 gives
E(k) = min |eig(D
SSM
)|
at the standard FCC BZ high-symmetry points.
Table 2: Minimum singular value at FCC BZ high-symmetry points. All zone-boundary
modes are U(1) and strictly gapped. The K-point value
0.01372/a
is the minimum on the
standard high-symmetry path.
Point Fractional coords
(f
1
, f
2
, f
3
) E
Type
Γ (0, 0, 0) 0
Z
2
uniform (physical)
X (0,
1
2
,
1
2
) 5.45/a
U(1), gapped
W (
1
4
,
1
2
,
3
4
) 1.97/a
U(1), gapped
L (
1
2
,
1
2
,
1
2
) 4.72/a
U(1), gapped
K (
3
8
,
3
8
,
3
4
) 0.01372/a
U(1), path minimum
The K-point value
0.01372/a
is a genuine local minimum of the dispersion (the gap rises to
≈ 0.114/a
when displaced by
ε = 0.05/a
along each of the six principal reciprocal directions,
7
ruling out a saddle point). Because K has fully rational FCC fractional coordinates
(
3
8
,
3
8
,
3
4
)
,
this gap is
independent of lattice size
L
: the K-point is always exactly sampled at any
L
divisible by 8, with the same gap
0.01372/a
.
This must be distinguished from the
global
minimum grid gap, which is not xed: by Dirich-
let's approximation theorem, rational fractions
p/L
can approximate
1/(2
√
2)
to within
O(1/L)
, so grid points can approach the Type-2 at-band zero surfaces arbitrarily closely
as
L → ∞
, causing the overall minimum grid gap to approach zero. This is the mechanism
underlying the thermodynamic-limit disclaimer in Section 3: no individual non-
Γ
grid point
has
E = 0
for any nite
L
, but the inmum over all nite lattices is zero. Numerically:
E
min
(L = 48) ≈ 3 × 10
−7
/a
,
E
min
(L = 32) ≈ 1.3 × 10
−3
/a
, oscillating irregularly as
L
varies
(governed by the quality of rational approximation to
1/(2
√
2)
at each
L
).
6.4 Gauge Field Robustness
With a U(1) background eld
A
µ
, link phases
e
iA·
ˆ
n
j
are inserted. For all tested strengths
|A| ≤ 2.0/a
, the spectral gap at all zone-boundary points persists; no new zero-crossings ap-
pear. The
Γ
-point shifts in momentum (expected gauge response) but the physical spectrum
structure is preserved. Figure 3 shows the dispersion under varying eld strength.
Figure 3: Dispersion under U(1) background eld
A = (A, 0, 0)
for
A ∈
{0, 0.05, 0.1, 0.2, 0.5, 1.0, 2.0}/a
(viridis colour scale). The spectral gap at zone-boundary
points persists for all tested eld strengths. The
Γ
-point cone shifts in momentum (gauge
response) but no new zeros appear.
7 Conclusion
We have established the following results for the FCC bond-direction Dirac operator:
1.
Exact chiral symmetry
:
{γ
5
, D
SSM
} = 0
at nite
a
, proved algebraically. No Wilson
terms required.
8
2.
Irrational Doubler Theorem
(Theorem 4): Every non-
Γ
zero of
D
SSM
has at least
one irrational FCC fractional coordinate
f
i
= ±1/(2
√
2)
, making it unreachable on any
nite FCC lattice. Proved via V-form analysis and conrmed by
32
3
BZ scan (zero
exceptions).
3.
Holographic projection
: In the SSM framework, physical U(1)-charged electrons
require complex propagating bond phases. Staggered Z
2
congurations carry no U(1)
charge. After projection to the U(1) physical sector,
D
SSM
has a single zero at
Γ
with
exact chiral symmetry. All zone-boundary modes (
X
,
W
,
L
,
K
) are U(1) and gapped
(
E ≈ 1
5/a
).
4.
Irrational kinematic protection
(Theorem 4): All non-
Γ
zeros of
D
SSM
Type-1
(Z
2
) and Type-2 (U(1)) alike have at least one irrational FCC fractional coordinate
f
i
= ±1/(2
√
2)
, unreachable on any nite FCC lattice of integer size
L
(
L = 1
500
veried numerically; proved analytically by the irrationality of
√
2
).
The Z
2
/U(1) mechanism is specic to non-bipartite lattice geometries where the bond di-
rections carry irrational phase structure relative to the reciprocal lattice. It is absent from
bipartite hypercubic lattices in the following sense: the standard hypercubic doublers also
reside at Z
2
phase positions (bond phases
∈ {+1, −1}
at
k
µ
= π/a
), but those positions
lie at
rational
fractional BZ coordinates (
f = 1/2
), making them directly accessible on any
even-sized nite lattice. For
D
SSM
, the Z
2
doubler positions shift to
irrational
coordinates
f
i
= 1/(2
√
2)
, kinematically inaccessible to all nite FCC lattices. The combination of ge-
ometric irrational doubler positioning and holographic U(1)/Z
2
projection is a qualitatively
new pathway to exact chiral symmetry on discrete structures.
Computational overhead.
D
SSM
requires 12 nearest-neighbour evaluations per site versus
6 for the 3D hypercubic operator, a factor-of-2 overhead in hopping operations substan-
tially less than overlap fermions, which require a matrix square root.
Future directions.
Three open questions are identied: (i) the behaviour of
D
SSM
under
full non-Abelian
SU(N)
gauge elds and anomaly matching; (ii) whether the Z
2
/U(1) mech-
anism can be formalised as a topological invariant (e.g., a Z
2
-valued index on the FCC bond
bundle); and (iii) the explicit construction and verication of the 4D extension to the
D
4
root
lattice (
K = 24
bonds), which is non-bipartite and the natural candidate for a Euclidean
lattice gauge theory application.
References
[1] H. B. Nielsen and M. Ninomiya, Nucl. Phys. B
185
, 20 (1981).
[2] K. G. Wilson, Phys. Rev. D
10
, 2445 (1974).
[3] J. Kogut and L. Susskind, Phys. Rev. D
11
, 395 (1975).
[4] D. B. Kaplan, Phys. Lett. B
288
, 342 (1992).
[5] P. H. Ginsparg and K. G. Wilson, Phys. Rev. D
25
, 2649 (1982).
[6] H. Neuberger, Phys. Lett. B
417
, 141 (1998).
[7] M. Creutz, JHEP
04
, 017 (2008).
[8] T. Kimura and T. Misumi, Prog. Theor. Phys.
127
, 63 (2012).
9
[9] R. Kulkarni, Constructive Verication of
K = 12
Lattice Saturation, Zenodo:
10.5281/zenodo.18294925 (2026).
A Computational Verication Code
#!/usr/bin/env python3
"""FCC Bond-Direction Dirac Operator: Irrational Doubler Theorem Verification
Licence: MIT | Dependencies: numpy>=1.21, matplotlib>=3.4, python>=3.8
Archived: Zenodo 10.5281/zenodo.18927549
"""
import numpy as np
# Anti-Hermitian spatial gamma matrices
sx=np.array([[0,1],[1,0]],dtype=complex)
sy=np.array([[0,-1j],[1j,0]],dtype=complex)
sz=np.array([[1,0],[0,-1]],dtype=complex)
I2=np.eye(2,dtype=complex); Z2=np.zeros((2,2),dtype=complex)
g1=np.block([[Z2,sx],[-sx,Z2]]); g2=np.block([[Z2,sy],[-sy,Z2]])
g3=np.block([[Z2,sz],[-sz,Z2]]); g5=np.block([[I2,Z2],[Z2,-I2]])
gammas=[g1,g2,g3]
# 12 FCC unit bond directions
n_vecs=[]
for i in [-1,1]:
for j in [-1,1]:
n_vecs+=[np.array([i,j,0]),np.array([i,0,j]),np.array([0,i,j])]
n_vecs=np.array(n_vecs)/np.sqrt(2)
def D_SSM(k, lp=None):
D=np.zeros((4,4),dtype=complex)
for idx,n in enumerate(n_vecs):
ph=np.exp(1j*np.dot(k,n))
if lp is not None: ph*=np.exp(1j*lp[idx])
D+=sum(n[mu]*gammas[mu] for mu in range(3))*ph
return D
def gap(k,lp=None):
D=D_SSM(k,lp)
return np.min(np.sqrt(np.maximum(np.linalg.eigvalsh(D@D.conj().T),0)))
def is_Z2(k, tol=1e-10):
return all(abs(np.exp(1j*np.dot(k,n)).imag)<tol for n in n_vecs)
a=1.0
b1=2*np.pi/a*np.array([-1,1,1]); b2=2*np.pi/a*np.array([1,-1,1])
b3=2*np.pi/a*np.array([1,1,-1])
# 1. Verify symmetries
k0=np.array([1.2,-0.4,2.7]); D=D_SSM(k0)
10
print(f"||D-D*|| = {np.linalg.norm(D-D.conj().T):.2e}")
print(f"||{{g5,D}}|| = {np.linalg.norm(g5@D+D@g5):.2e}")
# 2. High-symmetry table
for name,k in [("Gamma",0*b1),("X",0.5*b2+0.5*b3),
("W",0.25*b1+0.5*b2+0.75*b3),
("L",0.5*b1+0.5*b2+0.5*b3),
("K",0.375*b1+0.375*b2+0.75*b3)]:
print(f" {name}: E={gap(k):.5f} Z2={is_Z2(k)}")
# 3. Z2/U1 BZ scan (Theorem 1 verification)
N=32; n_z2_zero=0; n_u1_zero=0; n_u1_gapped=0; n_z2_gapped=0
for i1 in range(N):
for i2 in range(N):
for i3 in range(N):
k=(i1/N-0.5)*b1+(i2/N-0.5)*b2+(i3/N-0.5)*b3
g=gap(k); z2=is_Z2(k)
if z2:
if g<1e-8: n_z2_zero+=1
else: n_z2_gapped+=1 # Z2 gapped (only Gamma is Z2, so this stays 0)
else:
if g<1e-8: n_u1_zero+=1
else: n_u1_gapped+=1
print(f"Z2_zero={n_z2_zero}, U1_zero={n_u1_zero}, U1_gapped={n_u1_gapped}, Z2_gapped={n_z2_gapped}")
# Expected: Z2_zero=1 (Gamma), U1_zero=0, U1_gapped=32767, Z2_gapped=0
# (Only Gamma is Z2 in the scan; doublers are at irrational positions, unsampled.)
11
