Spectral Structure of the Naive Dirac Operator
on the Face-Centered Cubic Lattice
Raghu Kulkarni
SSMTheory Group, IDrive Inc., Calabasas, CA 91302, USA
raghu@idrive.com
Abstract
We construct the naive Dirac operator on the Face-Centered Cubic (FCC) lattice
(
K = 12
) in three spatial dimensions. The kinetic vector eld factors exactly as
f
µ
(k) = 4 sin(k
µ
)[cos(k
ν
) + cos(k
ρ
)]
, where
(µ, ν, ρ)
is any permutation of
(1, 2, 3)
.
This factorization yields a complete, analytic classication of all zero modes: a
singlet at
Γ
, an isolated quartet at the four L-points, and nodal lines on the Brillouin
zone boundary. The Jacobian determinant assigns topological indices
χ
Γ
= +1
and
χ
L
= −1
(per point); a winding-number computation gives
χ = +1
per nodal loop,
with 3 independent loops. The Nielsen-Ninomiya sum
+1 − 4 + 3 = 0
is satised.
The L-point quartet transforms under the chiral octahedral group
O
∼
=
S
4
and
decomposes as
4 → 1 ⊕ 3
; an anisotropic Wilson term splits singlet from triplet with
∆M = 4/a
. The stratied spectrum three topological classes with distinct Wilson
masses (
0
,
12/a
,
16/a
) and distinct chiralities has no parallel on hypercubic
lattices, where all
2
3
= 8
doublers are geometrically equivalent.
1 Introduction
On a hypercubic lattice in
D
spatial dimensions, the naive Dirac operator produces
2
D
zero modes (fermion doublers), all geometrically equivalent [1, 2]. Wilson fermions [3] and
staggered fermions [4, 5] suppress or reduce these doublers at the cost of chiral symmetry
or fermion content.
The Face-Centered Cubic (FCC) lattice oers a dierent starting point. It is the densest
sphere packing in 3D [6], has coordination number
K = 12
, and possesses the full oc-
tahedral point group
O
h
. To our knowledge, the zero-mode structure of the naive Dirac
operator on the FCC lattice has not been systematically characterized.
We work in 3 spatial dimensions throughout. The primary application of lattice fermions
QCD requires 4D, but the 3D case is of independent interest for condensed mat-
ter applications (Weyl semimetals, topological insulators on non-cubic lattices) and as a
testing ground for non-hypercubic constructions.
In this paper we prove a complete classication of all zero modes (Section 3), compute
chiralities at isolated zeros (Section 4), verify the Nielsen-Ninomiya theorem via winding-
1
number computation (Section 5), and analyze the representation theory of the L-point
quartet (Section 6).
2 The FCC Lattice
2.1 Real space
The FCC lattice is generated by the 12 nearest-neighbor vectors (in units of
a/2
, where
a
is the cubic lattice constant):
N = {(±1, ±1, 0), (±1, 0, ±1), (0, ±1, ±1)}
(1)
The convex hull of these 12 points is the cuboctahedron [7]. The symmetry group is
O
h
(order 48).
2.2 Reciprocal space
The reciprocal lattice is BCC. The rst Brillouin zone (FBZ) is a truncated octahedron [8],
with high-symmetry points:
Γ = (0, 0, 0)
L = (±π/2, ±π/2, ±π/2)
(4 independent after antipodal ID)
X = (π, 0, 0), (0, π, 0), (0, 0, π)
W = (π, π/2, 0)
and permutations (2)
3 The Naive Dirac Operator
3.1 Construction
In 3D Euclidean space, the minimal spinor representation is
2 × 2
(Pauli matrices
σ
µ
). There is no
γ
5
in 3D; the chirality computed below is the topological index
χ
i
= sgn(det J)
the local degree of the map
f : T
3
→ R
3
not a
γ
5
eigenvalue.
The relevant no-go theorem [1, 2, 9] requires
P
i
χ
i
= 0
.
The kinetic vector eld is:
f
µ
(k) =
1
a
X
n∈N
n
µ
sin(k · n), µ = 1, 2, 3
(3)
The naive Dirac operator is
D(k) = iγ · f(k)
. Zero modes occur wherever
|f(k)| = 0
.
2
3.2 Exact factorization
Theorem 1
(Factorization)
.
The kinetic vector eld factors exactly as:
f
µ
(k) =
4
a
sin(k
µ
)
cos(k
ν
) + cos(k
ρ
)
(4)
where
(µ, ν, ρ)
is any permutation of
(1, 2, 3)
.
Proof.
The 12 FCC neighbors partition into three sheets:
S
µν
contains the 4 vectors
with zero
ρ
-component. Only the sheets
S
µν
and
S
µρ
contribute to
f
µ
(the sheet
S
νρ
has
n
µ
= 0
).
From
S
µν
, the 4 vectors
(n
µ
, n
ν
) ∈ {(±1, ±1)}
contribute:
(+1) sin(k
µ
+ k
ν
) + (+1) sin(k
µ
− k
ν
)
+ (−1) sin(−k
µ
+ k
ν
) + (−1) sin(−k
µ
− k
ν
)
= 2[sin(k
µ
+ k
ν
) + sin(k
µ
− k
ν
)] = 4 sin(k
µ
) cos(k
ν
)
(5)
using
sin A + sin B = 2 sin
A+B
2
cos
A−B
2
. The sheet
S
µρ
contributes
4 sin(k
µ
) cos(k
ρ
)
by the
identical argument. Their sum gives Eq. (4).
3.3 Complete zero-mode classication
Theorem 2
(Completeness)
.
The zero set
f
−1
(0)
on the FCC Brillouin zone consists of
exactly:
1.
Γ = (0, 0, 0)
: a single isolated zero.
2.
L = (±π/2, ±π/2, ±π/2)
: four isolated zeros (after antipodal identication).
3. Nodal lines on the BZ boundary connecting X-points to W-points.
No other zeros exist.
Proof.
By Theorem 1,
f
µ
= 0
requires either
sin(k
µ
) = 0
or
cos(k
ν
) + cos(k
ρ
) = 0
. Denote
these conditions
S
µ
and
C
µ
. The system
f
1
= f
2
= f
3
= 0
splits into
2
3
= 8
cases.
Case SSS
(
sin k
µ
= 0
for all
µ
): Each
k
µ
∈ {0, π}
, giving 8 candidate points. The BZ
constraint
|k
µ
| + |k
ν
| ≤ π
admits
Γ
and the 3 X-points.
Case CCC
(
cos k
ν
+ cos k
ρ
= 0
for all pairs): Subtracting equations gives
cos k
1
=
cos k
2
= cos k
3
; substituting back gives
2 cos k
1
= 0
, so
k
µ
= ±π/2
. These are the 8
L-points (4 after antipodal identication), all inside the BZ.
Cases with two
S
and one
C
(SSC, SCS, CSS): Two momenta lie in
{0, π}
with their
cosines summing to zero (one is
0
, the other
π
); the third is free. These are the nodal
lines:
(π, 0, t)
,
(0, π, t)
, and permutations.
3
Cases with two
C
and one
S
(SCC, CSC, CCS): These require
sin k
µ
= 0
and
cos k
µ
=
− cos k
ν
= − cos k
ρ
. If
k
µ
= 0
:
k
ν
= k
ρ
= π
, but
|k
ν
| + |k
ρ
| = 2π > π
(outside BZ). If
k
µ
= π
:
k
ν
= k
ρ
= 0
(X-point, already in SSS).
The 8 cases exhaust all possibilities.
Figure 1 shows the zero-mode locations in the Brillouin zone. Figure 2 displays
|f(k)|
2
on three slices conrming the classication.
3
2
1
0
1
2
3
k
1
3
2
1
0
1
2
3
k
2
3
2
1
0
1
2
3
k
3
L
X
FCC Brillouin Zone: Zero-Mode Structure
(singlet, = +1)
L (quartet, = 1 each)
X-W nodal lines ( = +1 each)
Figure 1: FCC Brillouin zone (truncated octahedron) with zero-mode locations:
Γ
(green),
4 L-points (red squares), and XW nodal lines on the square boundary faces (blue dashed).
4 Topological Index at Isolated Zeros
The topological index at an isolated zero
k
i
is
χ
i
= sgn(det J)
, where:
J
µν
(k) =
∂f
µ
∂k
ν
=
1
a
X
n∈N
n
µ
n
ν
cos(k · n)
(6)
Proposition 1
(
Γ
-point index)
.
At
Γ = (0, 0, 0)
:
J
µν
= (8/a) δ
µν
,
det J = 512/a
3
> 0
,
χ
Γ
= +1
.
Proof.
At
k = 0
,
cos(k · n) = 1
for all
n
. The sum
P
n
n
µ
n
ν
over the 12 FCC neighbors
evaluates to
8δ
µν
(o-diagonal entries cancel by inversion symmetry; each diagonal entry
sums to 8).
4
1.0 0.5 0.0 0.5 1.0
k
1
/
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
k
2
/
log
10
|
f
|
2
, slice
k
3
= 0
1.0 0.5 0.0 0.5 1.0
k
1
/
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
L
log
10
|
f
|
2
, slice
k
3
= /2
1.0 0.5 0.0 0.5 1.0
k
2
/
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
entire
plane
is zero
log
10
|
f
|
2
, slice
k
1
=
Figure 2:
log
10
|f(k)|
2
on three momentum slices. Left:
k
3
= 0
, showing the isolated
Γ
zero. Center:
k
3
= π/2
, showing L-point zeros at
(±π/2, ±π/2)
. Right:
k
1
= π
, where
sin(k
1
) = 0
forces
f
1
≡ 0
, producing the nodal structure visible on the BZ boundary face.
Proposition 2
(L-point index)
.
At
L = (π/2, π/2, π/2)
:
J(L) = −
4
a
(E − I)
(7)
where
E
is the all-ones matrix and
I
is the identity. Eigenvalues:
{+4, +4, −8}/a
.
det J = −128/a
3
< 0
, giving
χ
L
= −1
.
Proof.
At
L
, the scalar products
L·n
take values in
{0, ±π}
. Six neighbors have
cos = +1
;
six have
cos = −1
. Diagonal entries vanish by cancellation. Each o-diagonal entry equals
−4/a
: the 4 neighbors with
n
µ
n
ν
= 0
all have
cos(L · n) = −1
, and
n
µ
n
ν
= ±1
.
With 4 independent L-points:
χ
total
L
= 4 × (−1) = −4
.
5 Boundary Chirality and Nielsen-Ninomiya
5.1 Winding number
For a nodal line along
k
2
(e.g.,
k = (π, t, 0)
), the transverse kinetic eld
(f
1
, f
3
)
winds
around the origin on a small circle of radius
ϵ
in the
(k
1
, k
3
)
plane. The winding number
w =
1
2π
I
d arg(f
1
+ if
3
)
(8)
evaluated numerically at
ϵ = 0.01
with 2000 angular samples gives
w = +1
(to within
3 × 10
−3
) for each XW segment, with convergence scaling as
1/N
.
5.2 Loop structure
The nodal lines form 3 independent closed loops, one per spatial axis. The loop at
k
µ
= π
consists of 4 XW segments forming a closed circuit around the square face
5
of the truncated octahedron. Opposite faces (
k
µ
= +π
and
k
µ
= −π
) are identied.
The 3 loops lie in disjoint planes and are separate connected components of
f
−1
(0)
. The
topological classication of such nodal-line structures is standard in the condensed matter
literature [11].
The factorization guarantees a zero-free tubular neighborhood around each loop: at
k
1
=
π
,
sin(k
1
) = 0
forces
f
1
≡ 0
, so zeros of
|f|
require only
f
2
= f
3
= 0
. These vanish only
on the nodal line itself, establishing topological invariance of the winding.
Each loop contributes
χ = +1
, giving
χ
boundary
= 3 × (+1) = +3
.
5.3 Nielsen-Ninomiya verication
χ
total
= +1
|{z}
Γ
− 4
|{z}
L
+ +3
|{z}
boundary
= 0 ✓
(9)
6 Representation Theory of the L-Point Quartet
The 4 L-points transform under the chiral octahedral group
O
∼
=
S
4
. The permutation
representation decomposes as
4 → 1⊕3
. By Schur's lemma, any
S
4
-invariant mass matrix
on the quartet splits into singlet and triplet components with generically dierent masses.
An anisotropic Wilson term with enhanced coupling
r
′
= 1.5
along the
(1, 1, 1)
diagonal
and standard
r = 1
elsewhere breaks
S
4
→ S
3
. This produces
M(L
1
) = 18/a
(singlet)
and
M(L
2
) = M(L
3
) = M(L
4
) = 14/a
(triplet), with splitting
∆M = 4/a
.
7 Comparison with Hypercubic Lattices
Lattice
K
Doublers Stratied? Wilson masses
S
n
Simple cubic 6
2
3
= 8
corners No All
6/a 4 → 1 ⊕ 3
(degen.)
BCC 8 Multiple No Single scale
6 → 1 ⊕ 5
[10]
FCC 12 1 + 4 + lines Yes
0, 12/a, 16/a 4 → 1 ⊕ 3
(split)
The simple cubic lattice has a formal
4 → 1 ⊕ 3
decomposition under
S
4
, but the singlet
and triplet are degenerate (identical Wilson mass
6/a
at all 4 points). The FCC lattice is
the only common 3D lattice where the decomposition produces distinct physical masses
across the three topological classes. Figure 3 compares the FCC and cubic spectra.
8 Discussion
Minimum fermion content.
A Wilson term with
r = 1
gives L-modes mass
12/a
and boundary modes
16/a
, leaving
Γ
massless. In the continuum limit, the heavy modes
decouple, yielding one massless species.
6
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Wilson mass
M
×
a
L (×4)
X-W lines
= +1
= 1
each
= +1
each
FCC: Stratified Spectrum
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Wilson mass
M
×
a
7 corners
(all equivalent)
all identical
mass & chirality
Simple Cubic: Uniform Spectrum
Figure 3: Wilson mass spectra on FCC (left) and simple cubic (right) lattices. The
FCC spectrum is stratied into three classes with distinct masses; the cubic spectrum is
uniform.
Retaining the quartet.
A momentum-dependent Wilson term
w(k) =
P
µ
cos
2
(k
µ
)
vanishes at L (
cos
2
(π/2) = 0
) and equals 3 at X (
cos
2
(π) = 1
). This selectively gaps
boundary modes while preserving the L-quartet. Whether the resulting action satises
reection positivity is open.
Boundary nodal lines.
The continuous nodal lines carrying
χ
boundary
= +3
have no
hypercubic analog. Domain-wall fermions [12] oer a possible approach if the FCC lattice
can be embedded as the boundary of a higher-dimensional lattice with a domain-wall
mass term.
Extension to 4D.
The 4D analog is the
D
4
root lattice (24 nearest neighbors, 24-cell
Brillouin zone). The factorization of Theorem 1 may generalize, but the case analysis is
substantially more involved.
9 Conclusion
The naive Dirac operator on the FCC lattice has a stratied zero-mode spectrum:
Γ
(singlet,
χ = +1
), L (quartet,
χ = −4
), and boundary nodal lines (
χ = +3
), with
+1 − 4 + 3 = 0
. The kinetic eld factors as
f
µ
= 4 sin(k
µ
)[cos(k
ν
) + cos(k
ρ
)]
, which proves
the classication is complete. The L-quartet decomposes as
1 ⊕ 3
under
S
4
, with an
explicit singlet-triplet splitting
∆M = 4/a
. This stratied structure distinct masses,
chiralities, and topological character across three classes is unique to the FCC lattice
among common 3D lattices.
References
[1] H. B. Nielsen and M. Ninomiya, Nucl. Phys. B
185
, 20 (1981). doi:10.1016/0550-
3213(81)90361-8
7
[2] H. B. Nielsen and M. Ninomiya, Phys. Lett. B
105
, 219 (1981). doi:10.1016/0370-
2693(81)91026-1
[3] K. G. Wilson, Phys. Rev. D
10
, 2445 (1974). doi:10.1103/PhysRevD.10.2445
[4] L. Susskind, Phys. Rev. D
16
, 3031 (1977). doi:10.1103/PhysRevD.16.3031
[5] J. Kogut and L. Susskind, Phys. Rev. D
11
, 395 (1975).
doi:10.1103/PhysRevD.11.395
[6] T. C. Hales, Ann. Math.
162
, 1065 (2005). doi:10.4007/annals.2005.162.1065
[7] H. S. M. Coxeter,
Regular Polytopes
, Dover (1973).
[8] N. W. Ashcroft and N. D. Mermin,
Solid State Physics
, Saunders College (1976).
[9] L. H. Karsten and J. Smit, Nucl. Phys. B
183
, 103 (1981). doi:10.1016/0550-
3213(81)90549-6
[10] W. Celmaster, Phys. Rev. D
26
, 2955 (1982). doi:10.1103/PhysRevD.26.2955
[11] C. Fang, H. Weng, X. Dai, and Z. Fang, Chin. Phys. B
25
, 117106 (2016).
doi:10.1088/1674-1056/25/11/117106
[12] D. B. Kaplan, Phys. Lett. B
288
, 342 (1992). doi:10.1016/0370-2693(92)91112-M
8
