Defect-Bound Modes of the Naive Dirac Operator on the FCC
Lattice:
Spectral Multiplet Structure at Tetrahedral and Octahedral Voids
Raghu Kulkarni
∗
SSMTheory Group, IDrive Inc., Calabasas, CA 91302, USA
May 2026
Abstract
The face-centered cubic (FCC) lattice has two classes of interstitial void: tetrahedral,
with coordination 4, and octahedral, with coordination 6. A trapped extra node at either
void is a localized point defect coupled to its surrounding FCC sites through nearest-neighbor
bonds. We compute the bound-mode multiplet of the naive Dirac operator at each defect
class and decompose it under the residual point-group symmetry. At a tetrahedral defect
the multiplet has dimension 4 and decomposes under T
d
∼
=
S
4
as A
1
⊕ T
2
(Schoenflies T
d
convention, in which the three-dimensional irrep is the polar vector). At an octahedral defect
the multiplet has dimension 6 and decomposes under O
h
as A
1g
⊕ E
g
⊕ T
1u
. Both three-
dimensional pieces are spatial-vector representations of their point groups, so the spectral
labels alone do not distinguish the two defect classes; the distinction is combinatorial. The
three-dimensional T
2
piece has the dimension of, and is S
3
-module-isomorphic to, the internal
color space built in the matter paper of Kulkarni [Phys. Open 27, 100423 (2026)] on the
three perfect matchings of K
4
, the bonded graph of the tetrahedral defect’s surrounding
sites. The bonded graph of the octahedron’s six surrounding sites is instead the complete
tripartite K
2,2,2
rather than K
6
, because the three antipodal vertex pairs sit at second-
nearest-neighbor distance in the FCC lattice and are not physical bonds; K
2,2,2
has 8 perfect
matchings and 30 skew-edge pairs, neither giving the 3-color structure of the matter-paper
construction. The spectral analysis is thus a dimensional consistency check on a pre-existing
combinatorial color assignment, not a derivation of color from the spectrum: the tetrahedral
bound triplet has the dimension of the K
4
color space, while the two defect classes are
separated not by their spectral content—which is a spatial-vector triplet in both cases—but
by this combinatorial matching count. A companion dark-matter paper [15] reaches the
same K
2,2,2
obstruction by independent combinatorial means and, on that basis, proposes a
dark-sector identification of the octahedral defect; the present work supports the structural
distinction underlying that proposal, not its mass scale, abundance, or phenomenology.
As supporting background we classify the free-field FCC zero-mode spectrum from an exact
factorization of the kinetic vector field. The comparison with HCP, BCC, and bulk-modified
fermion constructions, and the reasons FCC is uniquely selected as an isotropic substrate,
are developed in the body.
1 Introduction
The naive Dirac operator on a hypercubic lattice in D dimensions has 2
D
geometrically equiv-
alent doublers [1, 2]. Wilson fermions [3] give them a mass term that breaks chiral symmetry.
Staggered fermions [4, 5] cut the count to 2
⌈D/2⌉
tastes at the cost of a residual flavor symmetry.
∗
raghu@idrive.com
1
Minimally doubled formulations [6, 7, 8, 9, 10] keep only two species in the continuum limit by
breaking hypercubic symmetry. Supersymmetric lattice constructions [11, 12] use root lattices
such as A
∗
4
to preserve part of the supersymmetry algebra exactly. Each of these acts on the
bulk operator.
We take a different starting point. Consider the standard naive Dirac operator on the face-
centered cubic (FCC) lattice, and look at the spectrum it produces when a localized point defect
is inserted at an interstitial void. FCC has two void classes per conventional cubic cell: eight
tetrahedral voids (coordination 4) and four octahedral voids (coordination 6). A trapped extra
node at either is a natural physical defect. The bound modes at the defect form a multiplet
under the residual point-group symmetry, and the structure of that multiplet decides whether
the defect can carry an internal symmetry algebra of a given type.
The setting is the matter paper [14] of the SSMTheory program, in which physical matter
is identified with localized lattice defects rather than with a bulk fermion field. There, color
quantum numbers are assigned by combinatorial counting: the complete graph K
4
on the
four valence bonds of a tetrahedral defect has exactly three perfect matchings, giving a three-
dimensional internal color Hilbert space on which an su(3) algebra acts. That construction
is group-theoretic and uses the local bond connectivity only, with no input from the Dirac
spectrum. The question this paper takes up is whether the spectral content of the Dirac operator
at the same defect is consistent with the assigned color structure — that is, whether the bound-
mode multiplet contains a representation of the right dimension and the right transformation
properties — and what the corresponding analysis gives for the octahedral defect.
Results. At a tetrahedral defect the bound-mode multiplet has dimension 4 and decomposes
under the tetrahedral point group T
d
∼
=
S
4
as
4
tet
= A
1
⊕ T
2
, dim 1 + 3. (1)
The three-dimensional T
2
piece (the polar vector of T
d
; see the convention note in Section 4)
has the dimension of the matter paper’s K
4
color space and is S
3
-module-isomorphic to it
(Section 4). At an octahedral defect the multiplet has dimension 6 and decomposes under the
octahedral point group O
h
as
6
oct
= A
1g
⊕ E
g
⊕ T
1u
, dim 1 + 2 + 3. (2)
The 3-dimensional T
1u
piece is the vector representation of O
h
. The tetrahedral T
2
piece is
likewise a spatial-vector representation (of T
d
), so neither triplet is singled out as “internal”
by its spectral label, and the two defect classes are separated instead by their bonded-graph
combinatorics. The bonded graph of the six surrounding sites is the complete tripartite graph
K
2,2,2
, not K
6
: the three antipodal pairs of the octahedron sit at second-nearest-neighbor
distance in the FCC lattice, not nearest-neighbor, so they are not physical bonds. K
2,2,2
has
eight perfect matchings, not the three on which the matter-paper algebra is built, so no K
4
-
type color structure is available at an octahedral defect (Section 5). The companion dark-
matter paper [15] reaches this K
2,2,2
obstruction by independent combinatorial means and, on
that basis, proposes a dark-sector identification of the octahedral defect. The present spectral
analysis supports the structural distinction underlying that proposal (that octahedral defects
do not carry the K
4
-type color algebra), but says nothing about its mass scale, abundance, or
phenomenology, which are the subject of [15]. Tetrahedral defects support the matter-paper
color algebra; octahedral defects do not.
A note on what this result is and is not. The Dirac analysis does not derive SU(3) color
from the spectrum; the matter-paper construction is logically prior and uses different input (K
4
perfect matchings, not Dirac eigenmodes). What we add is a dimensional consistency check:
the tetrahedral bound triplet is three-dimensional, matching the dimension of the K
4
color
2
space, so the matter-paper assignment can be carried on it. Because both bound triplets are
spatial-vector representations of their point groups, the spectrum does not by itself distinguish
the two defect classes; the substantive distinction is the combinatorial K
2,2,2
obstruction (eight
perfect matchings rather than three), which removes the three-element matching space at the
octahedral void.
Bulk background. Sections 2 and 3 give the free-field bulk spectrum on FCC. The ki-
netic vector field factors exactly (Theorem 1), so the full zero-mode structure can be classi-
fied by cases. Three topological classes appear: a singlet at Γ, a quartet at the four L-points
(±π/2, ±π/2, ±π/2), and three boundary nodal loops along the X–W segments of the truncated-
octahedron BZ. The kinetic-map index sum is +1−4+3 = 0, the Nielsen–Ninomiya consistency
relation. The three classes carry distinct Wilson masses (0, 12/a, 16/a) and are not equivalent
to the eight uniformly-massed doublers of the naive simple-cubic operator. This bulk analysis
is supporting context for the defect-bound-mode analysis in Sections 4 and 5, which is the main
new content.
Relation to other reduced-fermion constructions. The defect mechanism we use is un-
like minimally doubled fermions [6, 7, 8], supersymmetric lattices [11, 12], twisted-mass, and
domain-wall constructions: those modify the bulk operator. Here the bulk operator is un-
changed; the physical content sits in localized topological excitations whose internal symmetries
are set by the defect’s local geometry. Section 6 discusses the comparison in detail and works
out where the cuboctahedral first-shell structure of FCC fits among the alternatives. The clos-
est 3D candidate is HCP, with the same coordination number K = 12, but HCP’s first shell
is the anticuboctahedron and gives a preferred-axis anisotropy at the rank-4 bond-tensor level;
FCC is the only 3D close-packed lattice with a fully cubic-symmetric rank-4 bond tensor and
the corresponding full T
d
/ O
h
symmetry at its interstitial voids.
2 The FCC Lattice and the Naive Dirac Operator
The FCC lattice has twelve nearest-neighbor vectors (in units of a/2):
N = {(±1, ±1, 0), (±1, 0, ±1), (0, ±1, ±1)}. (3)
The convex hull of these twelve points is the cuboctahedron [17], with symmetry group O
h
of
order 48. The reciprocal lattice is body-centered cubic; the first Brillouin zone is a truncated
octahedron [18] with high-symmetry points
Γ = (0, 0, 0), L = (±
π
2
, ±
π
2
, ±
π
2
), X = (π, 0, 0), W = (π,
π
2
, 0) (4)
(and cubic permutations of X, W ). Throughout, nearest-neighbor vectors are given in units of
a/2, and Brillouin-zone coordinates and all figure axes are in units of π/a.
The naive Dirac operator in three spatial dimensions uses two-component spinors with no
γ
5
. The local degrees we track below are kinetic-map indices of f : T
3
→ R
3
, not eigenvalues
of γ
5
[16]. The kinetic vector field is
f
µ
(k) =
1
a
X
n∈N
n
µ
sin(k · n), µ = 1, 2, 3, (5)
and the Dirac operator is D(k) = iσ ·
f(k). Zero modes occur wherever |
f(k)| = 0.
Theorem 1 (Factorization). The kinetic vector field of Eq. (5) factors as
f
µ
(k) =
4
a
sin(k
µ
)
cos(k
ν
) + cos(k
ρ
)
, (6)
where (µ, ν, ρ) is any permutation of (1, 2, 3).
3
Proof. The twelve FCC neighbors partition into three sheets S
µν
containing the four vectors
with zero ρ-component. Only sheets S
µν
and S
µρ
contribute to f
µ
(sheet S
νρ
has n
µ
= 0). From
S
µν
, the four vectors (n
µ
, n
ν
) ∈ {(±1, ±1)} contribute
sin(k
µ
+ k
ν
) + sin(k
µ
− k
ν
) −sin(−k
µ
+ k
ν
) −sin(−k
µ
− k
ν
) = 4 sin(k
µ
) cos(k
ν
),
using sin A + sin B = 2 sin
A+B
2
cos
A−B
2
. Sheet S
µρ
contributes 4 sin(k
µ
) cos(k
ρ
) identically.
Summing gives Eq. (6).
3 Bulk Zero Modes and Their Topological Indices
Theorem 2 (Complete classification). The zero set f
−1
(0) on the FCC Brillouin zone consists
of exactly:
1. Γ = (0, 0, 0): one isolated zero.
2. L = (±π/2, ±π/2, ±π/2): four isolated zeros after antipodal identification.
3. Nodal lines on the BZ boundary connecting X- to W -points.
No other zeros exist.
Proof. By Theorem 1, f
µ
= 0 requires sin(k
µ
) = 0 (call this S
µ
) or cos(k
ν
) + cos(k
ρ
) = 0 (C
µ
).
Exhausting all 2
3
= 8 sign combinations:
Case SSS (S
1
∧S
2
∧S
3
): each k
µ
∈ {0, π}. BZ constraints admit Γ and the three X-points.
Case CCC: subtracting pairs 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 four L-points (inside the BZ).
Cases SSC, SCS, CSS: two momenta in {0, π} with their cosines summing to zero; the third
is free. These produce nodal lines: (0, π, t), (π, 0, t), and permutations — the X–W boundary
segments.
Cases SCC, CSC, CCS: 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 case SSS).
The eight cases are exhaustive.
3.1 Topological indices
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
P
n
n
µ
n
ν
cos(k · n).
Proposition 1 (Γ-point). At Γ, 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
= 8δ
µν
: each diagonal entry receives
a contribution of +1 from the eight FCC vectors with |n
µ
| = 1, and off-diagonal entries cancel
by inversion symmetry.
Proposition 2 (L-points). At L = (π/2, π/2, π/2), J(L) = −(4/a)(E − I), where E is the
all-ones matrix and I the identity. Eigenvalues are {+4, +4, −8}/a, det J = −128/a
3
< 0,
χ
L
= −1.
Proof. At L, k ·n ∈ {0, ±π}. Six neighbors have cos = +1; six have cos = −1. Diagonal entries
P
n
n
2
µ
cos(k ·n) = 0 (equal numbers of +1 and −1 along each axis). For the off-diagonal entries
the four neighbors with n
µ
n
ν
= 0 all give cos(k · n) = −1 and n
µ
n
ν
= ±1, contributing −4/a.
So J(L) = −(4/a)(E − I), eigenvalues {+4/a, +4/a, −8/a}.
With four independent L-points, χ
total
L
= 4 ×(−1) = −4.
4
3.2 Boundary winding number
Along a nodal line in k
2
(for example k = (π, t, 0)), the transverse field (f
1
, f
3
) winds around
the origin. We computed the winding number on a small circle of radius ϵ = 0.01 in the (k
1
, k
3
)
plane, sampled with 2000 angular points, obtaining w = +1.000±0.003 independent of position
t along the line. Topological stability follows from the factorization in Eq. (6): at k
1
= π,
sin(k
1
) = 0 forces f
1
≡ 0, so any zero of |f | on this face must satisfy f
2
= f
3
= 0, confining
the locus to the nodal line. The three independent boundary loops (one per spatial axis) each
contribute χ = +1, so χ
boundary
= +3.
The Nielsen–Ninomiya balance must be stated with care, because the three classes are
objects of different codimension: the Γ- and L-point zeros are isolated (codimension three)
and carry an integer index χ
i
= sgn det J(k
i
), while the boundary loops are codimension-
two nodal lines carrying a transverse winding number. We therefore state the result as a
consistency relation among these counts under fixed Brillouin-zone identifications, rather than
as a single index theorem over heterogeneous objects. The identifications are: (i) the four
L-points are counted after antipodal identification of the truncated-octahedron BZ (the eight
points (±π/2, ±π/2, ±π/2) pair into four inequivalent zeros); and (ii) the six square boundary
faces pair under inversion k → −k into three independent X–W nodal loops, each of winding
+1. With these identifications,
χ
total
= +1
|{z}
Γ
+ −4
|{z}
L
+ +3
|{z}
boundary
= 0, (7)
the Nielsen–Ninomiya consistency relation for the FCC naive operator. Figure 1 shows the
three classes on the truncated-octahedron Brillouin zone.
0
k
1
[ /
a
]
0
k
2
[ /
a
]
0
k
3
[ /
a
]
( = +1)
L
( = 1)
X
--
W
nodal loop ( = +1)
Figure 1: FCC first Brillouin zone (truncated octahedron) with the three zero-mode classes of
the naive Dirac operator. Green circle: Γ-point singlet, χ = +1. Red squares: four L-points at
(±π/2, ±π/2, ±π/2), χ = −1 each, giving χ
total
L
= −4. Blue outlines: the six square boundary
faces, paired by inversion into three independent X–W nodal loops with χ = +1 per loop. The
Nielsen–Ninomiya sum +1 − 4 + 3 = 0 is realized by these three classes.
5
3.3 Wilson masses
A standard isotropic Wilson term gives the bulk zero-mode classes mass M
W
= 0 at Γ, 12/a
at each L-point, and 16/a on the boundary loops. The three classes are therefore stratified by
Wilson mass at distinct scales [3], in contrast with the uniform 6/a produced by the standard
Wilson term on the simple cubic lattice. An anisotropic Wilson term with enhanced coupling
along (1, 1, 1) splits the L-quartet under S
4
⊃ S
3
as 1 ⊕ 3, with ∆M = 4/a.
4 Tetrahedral Defects
This section gives the main new content of the paper. The bulk classification of Theorem 2
sets the stage; we now turn to the localized bound-mode multiplet at a tetrahedral interstitial
defect (Figure 2(a)).
4.1 Defect geometry
A tetrahedral void of the FCC lattice is bounded by four FCC sites that form a regular tetra-
hedron. An interstitial node placed at the void centroid sits at equal distance from each of
these four sites. Label the surrounding sites {1, 2, 3, 4} and the bonds from the defect to them
{b
1
, b
2
, b
3
, b
4
}. The defect site and the four bonds form a star graph whose point group is
T
d
∼
=
S
4
, the full tetrahedral group of order 24.
The naive Dirac operator restricted to this star has matrix elements between the defect
spinor and each of the four surrounding-site spinors. The Hilbert space localized on the star
factors as C
2
⊗ (C ⊕ C
4
): the spinor index times (defect site ⊕ four surrounding sites). What
sets the bound-mode multiplet structure is the four-dimensional permutation representation of
T
d
on the surrounding sites, with the spinor index decorating each multiplet member.
4.2 The defect Hamiltonian
We work with two-component spinors on a finite FCC cluster C (the parity-even sites x + y + z
even with |r|
2
≤ 9.5; |C| = 55), open boundary conditions, together with one interstitial site at
the void centroid. The Hamiltonian is
H = H
FCC
+ H
defect
+ H
coupling
. (8)
(i) Bulk term. H
FCC
is the naive Dirac hopping operator of Section 2: for each nearest-neighbor
vector n ∈ N and each bond (r, r + n) in C,
H
FCC
r+n, r
=
i
2a
σ · n,
H
FCC
r, r+n
= h.c., (9)
whose Bloch symbol on the infinite lattice is the Hermitian kinetic operator σ ·
f(k) built from
Eq. (5) (up to the overall sign fixed by hermiticity; the bulk Dirac operator D = iσ ·
f of
Section 2 is its anti-Hermitian counterpart, with the same zero set |
f| = 0). (ii) Defect term.
The trapped node carries an on-site energy H
defect
= ε
d
c
†
0
c
0
(two-component). (iii) Coupling
term. The defect site bonds to its M bounding sites {a = 1, . . . , M } at the unit bond length
through a hopping t:
H
coupling
= t
M
X
a=1
c
†
0
c
a
+ h.c.
, M = 4 (tetrahedral), 6 (octahedral). (10)
6
Effective on-site model. Integrating out the defect site 0 produces a shell self-energy
Σ
s
(E) = t
2
/(E − ε
d
) that acts only in the symmetric (A
1
) combination of the bounding sites
and is energy-dependent; in the deep-level limit |E − ε
d
| ≫ |t| it becomes a static shift. The
verification script uses the simpler, manifestly point-group-symmetric model in which the defect
shifts the on-site energy of each bounding site uniformly,
H
defect-model
= V
M
X
a=1
c
†
a
c
a
, V < 0, (11)
an attractive well on the bounding shell. The multiplet decomposition (Propositions 3 and 5)
is forced by the site point group (T
d
or O
h
) of the perturbation alone and is independent of the
radial profile of the potential; the two models differ only in which members are pulled to bound
energies (the rank-one self-energy populates A
1
; the full-rank well of Eq. (11) populates every
irrep). We refer to Eq. (11) as the defect model below, and avoid the term “Schur complement”
as imprecise.
What is proved. We establish two statements, and only these: (a) the M -dimensional
bounding-shell amplitude space carries the point-group decomposition of Propositions 3 and 5;
and (b) in the strong-coupling regime there exist eigenstates localized on the defect-plus-shell
whose weights reproduce that decomposition, confirmed by exact diagonalization (Section 4.5).
We do not claim an infinite-lattice bound-state theorem, an asymptotic decay rate, or a deriva-
tion of internal symmetry from the spectrum; the spectral content is a consistency check on the
geometric color assignment of the matter paper [14], not a derivation of it.
4.3 Multiplet structure under T
d
Convention. We use the Schoenflies T
d
convention throughout: the polar vector (x, y, z)
transforms as T
2
and axial vectors as T
1
. The four-site permutation representation of T
d
∼
=
S
4
is then 4
tet
= A
1
⊕T
2
, the same reducible representation as the four σ-bonds (equivalently the
four H positions) in CH
4
. In the abstract-S
4
convention the same three-dimensional irrep is
often written T
1
; the relabeling does not affect any amplitude or dimension, only the symbol.
The permutation representation of S
4
on four sites has character χ = (4, 2, 0, 1, 0) on the
conjugacy classes (e, (12), (12)(34), (123), (1234)). Projecting onto T
d
irreps gives the following
decomposition.
Proposition 3. The permutation representation of T
d
∼
=
S
4
on the four surrounding sites of a
tetrahedral defect decomposes as
4
tet
= A
1
⊕ T
2
, (12)
with dimensions 1 + 3 = 4.
Proof. Direct projection against the S
4
character table. For the trivial representation A
1
with
character (1, 1, 1, 1, 1) the multiplicity is
⟨A
1
, χ⟩ =
1
24
1 ·1 · 4 + 6 · 1 ·2 + 3 ·1 · 0 + 8 · 1 · 1 + 6 · 1 · 0
= 1.
For the polar-vector representation T
2
with character (3, 1, −1, 0, −1) the multiplicity is
⟨T
2
, χ⟩ =
1
24
1 ·3 · 4 + 6 · 1 ·2 + 3 ·(−1) · 0 + 8 · 0 · 1 + 6 · (−1) · 0
= 1.
The other irreps A
2
, E, T
1
have multiplicity 0.
The A
1
singlet is the symmetric combination |s⟩ = (|1⟩ + |2⟩ + |3⟩ + |4⟩)/2. The T
2
triplet
is spanned by three orthogonal combinations transforming as the polar vector. With spinor
decoration each of the four multiplet members carries a two-component spinor index, giving
eight bound-mode states in total.
7
4.4 Compatibility with the K
4
color construction
The matter paper [14] builds a three-color quark Hilbert space at a tetrahedral defect from the
perfect-matching combinatorics of K
4
, the complete graph on the four valence bonds. K
4
has
six edges; they partition into three perfect matchings (pairs of disjoint edges covering all four
vertices):
M
1
= {b
12
, b
34
}, M
2
= {b
13
, b
24
}, M
3
= {b
14
, b
23
}, (13)
where b
ij
joins surrounding sites i and j. The Weyl group of the resulting su(3) algebra is
S
3
= S
4
/V
4
, where V
4
= {e, (12)(34), (13)(24), (14)(23)} is the Klein four-group; V
4
acts trivially
on the three matchings, so the quotient acts faithfully.
The matter-paper color Hilbert space H
C
= span{|M
1
⟩, |M
2
⟩, |M
3
⟩} is logically independent
of the Dirac bound-mode space, since it lives on matchings rather than on sites. What needs to
be checked is whether the bound-mode space of Proposition 3 can carry an SU(3) action that
respects the assigned color structure.
Proposition 4. The three-dimensional T
2
sub-multiplet of the tetrahedral bound-mode space has
the dimension of the K
4
matching space H
C
= span{|M
1
⟩, |M
2
⟩, |M
3
⟩} of the matter paper [14]
and is isomorphic to it as an S
3
-module, where S
3
= S
4
/V
4
is the residual Weyl group acting
on the three perfect matchings.
There is a linear identification
Φ : T
2
−→ span{|M
1
⟩, |M
2
⟩, |M
3
⟩}. (14)
Φ is (i) dimension-preserving and (ii) S
3
-equivariant, where S
3
acts on T
2
through the restriction
S
4
→ S
4
/V
4
and on the matching states by permutation. It is basis-dependent: Φ is fixed only up
to the choice of an ordered basis of T
2
, and it is neither canonical nor S
4
-equivariant (the Klein
group V
4
acts nontrivially on the site amplitudes spanning T
2
but trivially on the matchings, so
no S
4
-equivariant identification exists). Concretely, the standard three-dimensional irrep of S
4
restricted to S
3
⊂ S
4
decomposes as A ⊕E (dimensions 1 + 2), and the matching representation
A
1
⊕ E of S
3
on H
C
has the same decomposition; the two 3-dimensional spaces are therefore
isomorphic as S
3
-modules. The matter-paper su(3) algebra, defined on H
C
, acts on T
2
by
transport through Φ. This is a statement of representational compatibility, not a derivation of
color from the Dirac spectrum.
4.5 Numerical confirmation
The 1 + 3 site-permutation decomposition is confirmed by direct diagonalization of the naive
Dirac operator on the finite cluster C of Section 4.2. The verification script (dirac_verify_defects.py,
included with the submission) builds the 55-site FCC cluster and models the trapped-node de-
fect at the tetrahedral void (1/2, 1/2, 1/2) as the uniform attractive on-site well of Eq. (11) on
the four surrounding FCC sites. This T
d
-symmetric perturbation creates well-localized bound
modes without modifying the bulk operator. The resulting Hermitian 110 × 110 Hamiltonian
is diagonalized exactly. The eight most-localized eigenstates (= 1 + 3 spatial ×2 spinor) carry
the A
1
and T
2
site representations with total weights converging to 2 and 6 respectively as the
well depth increases, with ratio T
2
/A
1
→ 3, the value expected for the A
1
⊕ T
2
decomposition.
Table 1 reports the convergence; all hopping and energy values are in units of the bulk coefficient
1/2a (i.e. a = 1).
5 Octahedral Defects
Unlike the tetrahedral case of Section 4, we do not perform a separate finite-cluster diagonaliza-
tion at the octahedral void. The result of this section is (i) the symmetry decomposition of the
8
|V | top-8 band A
1
wt. T
2
wt. T
2
/A
1
min. loc. wt.
8 [−10.12, −6.92] 1.814 5.624 3.100 0.900
16 [−17.95, −14.55] 1.953 5.906 3.024 0.980
32 [−33.85, −30.40] 1.988 5.977 3.006 0.995
64 [−65.79, −62.33] 1.997 5.994 3.001 0.999
128 [−129.76, −126.30] 1.999 5.999 3.000 1.000
Table 1: Finite-cluster Dirac diagonalization at a tetrahedral defect (55-site cluster, 110-
dimensional Hermitian Hamiltonian). As the well depth |V | increases, the eight most-localized
eigenstates have total A
1
weight → 2 and T
2
weight → 6 (ratio → 3), and their localization
weight on the four bounding sites → 1, confirming the A
1
⊕T
2
decomposition. Reproduced by
dirac_verify_defects.py.
six-dimensional bounding-shell amplitude space under O
h
(Proposition 5), and (ii) the combi-
natorial obstruction on the bonded graph K
2,2,2
(Proposition 6), whose enumeration is carried
out in full in the companion paper [15]. No claim is made here about the detailed bound-state
spectrum at an octahedral defect beyond the representation content of the localized shell space.
5.1 Defect geometry
An octahedral void of the FCC lattice is bounded by six FCC sites at the vertices of a regular
octahedron. The surrounding sites and the defect site form a local star graph with point group
O
h
of order 48. The defect couples to each of the six surrounding sites through one bond.
5.2 Multiplet structure under O
h
The permutation representation of O
h
on the six octahedron vertices decomposes into O
h
irreps
as follows.
Proposition 5. The permutation representation of O
h
on the six surrounding sites of an octa-
hedral defect decomposes as
6
oct
= A
1g
⊕ E
g
⊕ T
1u
, (15)
with dimensions 1 + 2 + 3 = 6.
This is the standard O
h
decomposition on six octahedron vertices, derivable from the O
h
character table [18, 17]. The A
1g
singlet is the symmetric sum. The E
g
doublet captures two
diagonal differences between antipodal-pair amplitudes. The T
1u
triplet is the vector represen-
tation, with basis vectors transforming as (x, y, z) under O
h
.
5.3 What distinguishes the two triplets
Both three-dimensional sub-multiplets are spatial-vector representations of their point groups:
the tetrahedral T
2
is the polar vector of T
d
, and the octahedral T
1u
is the polar vector of O
h
.
Because both are spatial, the Dirac decomposition alone cannot single out one as “internal”
and the other as “spatial.” What distinguishes the two defect classes is not the spectral label
but the bonded-graph combinatorics on which the matter paper’s color algebra is built: K
4
has
three perfect matchings, supporting the three-element color space H
C
, whereas K
2,2,2
has eight
perfect matchings and 30 skew-edge pairs (enumerated in [15]), neither factoring as a three-color
representation. The spectral analysis contributes only a dimensional consistency check at each
defect class: dim T
2
= 3 = dim H
C
at the tetrahedral void, with no analogous three-dimensional
matching space at the octahedral shell.
9
defect
1
2
3
4
(a) Tetrahedral defect:
K
4
, 3 perfect matchings
K
4
perfect matchings
M
1
= {12, 34}
M
2
= {13, 24}
M
3
= {14, 23}
defect
1
1'
2
2'
3
3'
(b) Octahedral defect:
K
2, 2, 2
, 8 perfect matchings
K
2, 2, 2
edges (12)
antipodal pairs (not bonds)
one perfect matching
(8 total)
Figure 2: Bond graphs for the two interstitial defect classes of the FCC lattice. (a) Tetra-
hedral: the four surrounding FCC sites form a regular tetrahedron, all six pairs at nearest-
neighbor distance, giving the complete graph K
4
; its three perfect matchings M
1
, M
2
, M
3
(color-
coded) carry the matter paper’s three colors [14]. (b) Octahedral: the three antipodal vertex
pairs sit at second-nearest-neighbor distance
√
2 L and are therefore not physical bonds (red dot-
ted). Only the twelve non-antipodal nearest-neighbor pairs are bonds (gray), so the bonded
graph is the complete tripartite K
2,2,2
, not K
6
, and has eight perfect matchings rather than
three (one shown in green). The matter-paper color algebra does not extend to this graph
(Proposition 6; independently [15]).
5.4 Combinatorial obstruction from K
2,2,2
The matter paper’s su(3) algebra uses perfect-matching combinatorics on the complete graph
of the four mutually nearest-neighbor valence bonds of a tetrahedral defect. For the octahedral
defect, the analogous graph is built from the six surrounding sites and the bonds joining those
pairs at nearest-neighbor distance in the FCC lattice. The six octahedron vertices fall into
three antipodal pairs at second-nearest-neighbor distance
√
2 L (with L the nearest-neighbor
distance), and these antipodal pairs are not bonded. Only the twelve non-antipodal pairs are
physical bonds. The bonded graph is therefore the complete tripartite graph
K
2,2,2
: 6 vertices in three antipodal pairs, 12 edges between non-antipodal pairs, (16)
not the complete graph K
6
on six vertices (compare Figure 2(a) and (b)). Direct enumeration
on K
2,2,2
(verified by the accompanying script) gives:
• Number of edges: 12.
• Number of perfect matchings (triples of disjoint edges covering all six vertices): 8.
• Number of skew-edge pairs (pairs of disjoint edges): 30.
The same counts appear in the companion dark-matter paper [15], where the skew-edge pair
count c
(O)
skew
= 30 enters as combinatorial input to the mass derivation.
Proposition 6. The matter-paper su(3) algebra constructed from the three perfect matchings
of K
4
does not extend to a structurally analogous algebra on the bond graph K
2,2,2
. K
2,2,2
has
eight perfect matchings, not three, and a skew-pair count of 30; neither factors as a three-color
representation [15]. No internal SU(3) algebra of the matter-paper form acts on the bound-mode
multiplet of an octahedral defect.
10
The bound-mode multiplet at an octahedral defect therefore carries the A
1g
⊕ E
g
⊕ T
1u
decomposition derived above, but no embedded three-color matching space of the matter-paper
type. The same obstruction is reached combinatorially in [15]; the spectral analysis here is
independent representation-theoretic support for that structural conclusion.
5.5 Implication: a structurally distinct second sector
Tetrahedral defects support the matter-paper color construction. Octahedral defects do not.
The two defect classes are not equivalent matter sectors of the FCC substrate. Only the tetrahe-
dral class, with coordination 4, is consistent with the K
4
-based color algebra of the SSMTheory
program. On the basis of this same structural distinction, the companion dark-matter paper [15]
proposes a dark-sector identification of the octahedral class; the O
h
multiplet decomposition of
Proposition 5 and the bond-graph identification of Eq. (16) support that distinction, while the
mass, abundance, and indirect-detection phenomenology of the octahedral sector are developed
in [15] and are neither used nor evaluated here.
6 Comparison with Other Lattices and Reduced-Fermion Con-
structions
The defect mechanism of Sections 4–5 is unlike the established reduced-fermion constructions
in lattice gauge theory. We retain the standard naive Dirac operator on FCC; physical fermion
content is carried by localized excitations at substrate defects, not by changes to the bulk
operator. The point group of the defect, together with the bonded graph it generates, fixes the
multiplet structure. Table 2 summarizes the comparison across the lattices considered below.
HCP: same K = 12, broken isotropy along the c-axis. Hexagonal close-packed (HCP)
and FCC are the two close-packed lattices in 3D, both with coordination K = 12 and the
same nearest-neighbor distance. The first shells differ structurally. FCC’s first shell is the
cuboctahedron, with point group O
h
(order 48) including inversion symmetry. HCP’s first shell
is the anticuboctahedron (the Johnson solid J
27
, also called the triangular orthobicupola), with
point group D
3h
(order 12) and a distinguished c-axis. The two solids are related by rotating
one triangular face of the cuboctahedron by 60
◦
relative to its opposite face. That single twist
breaks O
h
to D
3h
and singles out the stacking direction.
The consequence at the bond-tensor level is sharp. Both lattices give T
(2)
µν
= K
2
δ
µν
, so
ω ∝ |
k| at long wavelengths in both cases. The rank-3 tensor T
(3)
µνρ
vanishes for both: by full
inversion symmetry for FCC, and by the horizontal mirror σ
h
acting on the anticuboctahedron
for HCP. At rank 4 the two diverge. For FCC, direct computation gives
T
(4), FCC
xxxx
= T
(4), FCC
yyyy
= T
(4), FCC
zzzz
,
T
(4), FCC
xxxx
T
(4), FCC
xxyy
= 2, (17)
which is anisotropic at rank 4 relative to the fully isotropic value 3, but with all three coordinate
axes equivalent. The unbroken symmetry of the rank-4 corrections is the full cubic group, so
the anisotropy is the standard discrete cubic anisotropy of any 3D lattice. For HCP at ideal
c/a =
p
8/3, direct computation gives
T
(4), HCP
xxxx
= T
(4), HCP
yyyy
=
5
2
= T
(4), HCP
zzzz
=
8
3
,
T
(4), HCP
xxxx
T
(4), HCP
xxyy
= 3,
T
(4), HCP
zzzz
T
(4), HCP
xxzz
= 4. (18)
The in-plane ratio 3 is what fully-isotropic rank-4 tensors give, but the c-axis breaks this: T
zzzz
differs from T
xxxx
, and the c-axis ratio differs from the in-plane ratio. HCP’s rank-4 anisotropy
contains a preferred-axis term forbidden by full SO(3), not merely a discrete cubic anisotropy.
11
A substrate intended to support an emergent rotationally invariant continuum theory cannot
single out a spatial direction at the O(a
2
) correction level of its bulk dispersion. HCP fails this
requirement, FCC passes it (subject to the standard discrete cubic anisotropy expected of any
3D lattice). Of the two close-packed 3D lattices, only FCC is therefore a candidate for a
substrate of isotropic physical spacetime. In four dimensions the analogous question singles out
the D
4
root lattice, whose 24-cell first shell gives an exactly isotropic rank-4 bond tensor.
At the local defect-bound-mode level relevant to the rest of this paper, HCP’s reduced
symmetry has a different consequence. The combinatorial bond graphs K
4
at a tetrahedral void
and K
2,2,2
at an octahedral void are preserved (they are determined by which pairs of bounding
vertices sit at nearest-neighbor distance, which depends only on the local coordination), so the
matter-paper K
4
perfect-matching construction still defines a three-element color basis at an
HCP tetrahedral defect. The local point group at the voids, however, is reduced: from T
d
to a
C
3v
subgroup at a tetrahedral void, and from O
h
to D
3d
at an octahedral void. The bound-mode
multiplet decompositions split more finely under the reduced symmetry. The A
1
⊕T
2
structure
of Section 4 refines under T
d
→ C
3v
as A
1
⊕A
1
⊕E (dimensions 1 + 1 + 2), with the T
2
triplet
of FCC splitting into a C
3v
singlet plus doublet (T
2
→ A
1
⊕ E). The T
d
-degeneracy among
the three colors is lifted by the broken symmetry. Whether this HCP refinement is physically
interesting in its own right is a question we do not pursue here; HCP is not a candidate substrate
for the SSMTheory framework on the rank-4 isotropy grounds described above.
BCC: different coordination, no analog. Body-centered cubic has K = 8 with cubic first
shell. Tetrahedral voids in BCC exist but are distorted; the four bounding vertices are not at
equal nearest-neighbor distance from one another (the bonded graph of a tetrahedral void in
BCC is not K
4
), and the T
d
local symmetry is broken. The matter-paper K
4
perfect-matching
construction does not directly apply. Gauge fields on BCC were studied by Celmaster [13];
the related fermion analysis on BCC, including whether a different local symmetry-based color
construction is available, is outside the scope of the present paper. The local structure of
FCC and HCP — not BCC or hypercubic Z
3
— is what supports the defect-bound-mode color
algebra.
Minimally doubled fermions. Karsten [6], Wilczek [7], and Boriçi–Creutz [8, 9] introduced
operators that produce exactly two continuum doublers, breaking hypercubic symmetry to
obstruct the doubling theorem. Bedaque, Buchoff, Tiburzi, and Walker-Loud [10] analyzed the
symmetry structure systematically. The modification is at the bulk-operator level. The defect
approach here is orthogonal: it leaves the bulk operator alone and looks at localized structure.
A future minimally-doubled construction on FCC could be combined with the defect analysis;
we do not undertake that here.
Supersymmetric lattices. Kaplan, Katz, and Ünsal [11] and Catterall [12] construct super-
symmetric lattice gauge theories on root lattices such as A
∗
4
, exploiting algebraic structure of
the root system to preserve part of the continuum supersymmetry. Defect-bound modes are not
in scope. The FCC lattice is the A
3
root lattice; a supersymmetric extension of the construction
here is consistent with these methods but is not pursued.
Wilson, staggered, twisted-mass. The standard hypercubic constructions [3, 4, 5] remove
or relabel doublers through mass terms, phase factors, or chirally rotated masses. All act on
the bulk operator on Z
D
. Staggered fermions on a 3D hypercubic lattice produce 4 tastes at the
eight cubic-BZ corners, all carrying Wilson mass 6/a. The FCC spectrum derived here has zeros
at the BZ interior (Γ, four L-points at (±π/2, ±π/2, ±π/2)) and on boundary nodal lines, with
three topological classes at three Wilson masses (0, 12/a, 16/a). The two Brillouin zones have
different shapes (truncated octahedron versus cube), and no linear change of lattice variables
12
maps one to the other while preserving the O
h
symmetry and nearest-neighbor structure. The
FCC operator is not a reparametrization of staggered fermions.
Lattice K First shell Point group Defect-mechanism status
Simple cubic (naive) 6 cube O
h
no analog (distorted tet void)
Simple cubic (staggered) 6 cube O
h
no analog
BCC (naive) 8 cube O
h
distorted tet void, K
4
broken [13]
HCP 12 anticuboctahedron D
3h
K
4
, K
2,2,2
preserved; reduced void symmetry; c-axis anisotropy
Minimally doubled varies varies varies bulk modification, no defect content
FCC (this work) 12 cuboctahedron O
h
K
4
, K
2,2,2
with full T
d
, O
h
at voids
Table 2: Comparison across common 3D lattices. Among the close-packed lattices, only FCC
has the cuboctahedral first shell that gives full O
h
symmetry without a preferred axis at the
rank-4 bond-tensor level. HCP shares the combinatorial bond graphs at its voids but breaks
isotropy along its c-axis at O(a
2
) in the bulk dispersion and reduces the local point group at
each void to a C
3v
or D
3d
subgroup. BCC has K
4
-broken tetrahedral voids and does not support
the matter-paper color construction directly.
FCC is uniquely selected by three independent criteria. A substrate compatible with
the matter-paper defect-and-color picture, with rotationally invariant continuum physics, and
with maximum-density packing must satisfy three requirements simultaneously: (i) full O
h
first-
shell symmetry, so the rank-4 bond tensor has no preferred axis beyond the standard discrete
cubic anisotropy of any 3D lattice; (ii) the K
4
bond graph at tetrahedral interstitial voids
and the K
2,2,2
bond graph at octahedral voids, so the matter-paper su(3) algebra on K
4
perfect
matchings is defined at the tetrahedral class; (iii) the maximum kissing number K = 12 in three
dimensions and the maximum Bravais sphere-packing density π/(3
√
2) ≈ 0.7405 established by
Kepler’s theorem (proved by Hales [20]), so the substrate is the densest 3D arrangement allowed
by sphere-packing geometry [19]. Each of the alternatives in Table 2 fails at least one of these.
Simple cubic (K = 6) and hypercubic Z
3
have full O
h
but no tetrahedral void with K
4
bond
structure, and K = 6 is not the maximum kissing number. BCC (K = 8) has full O
h
but its
tetrahedral voids are distorted, with bounding vertices at unequal distances and a K
4
-broken
bond graph; K = 8 is also below the maximum. HCP (K = 12) attains the maximum kissing
number and packing density and has the same combinatorial bond graphs at its voids as FCC,
but its anticuboctahedral first shell has only D
3h
symmetry and develops a preferred c-axis
at the rank-4 bond-tensor level, failing requirement (i). FCC is the unique 3D Bravais lattice
satisfying all three criteria. The construction in this paper is therefore not lattice-agnostic. It
picks out FCC specifically on independent geometric grounds (isotropy of the bulk dispersion),
combinatorial grounds (the K
4
structure at the tetrahedral void), and packing-theoretic grounds
(maximum kissing number and density). The same selection extends to four dimensions through
the D
4
root lattice, where the 24-cell first shell gives exact rank-4 isotropy and the FCC structure
is recovered on each constant-time slice.
7 Discussion
What this paper establishes
Two specific structural results. First, the bound-mode multiplet at a tetrahedral defect of the
FCC lattice has dimension 4 and decomposes under T
d
as A
1
⊕ T
2
(Proposition 3), with the
three-dimensional T
2
sub-multiplet dimensionally and S
3
-module-isomorphic to the K
4
match-
ing space of the matter paper (Proposition 4). Second, the bound-mode multiplet at an octa-
hedral defect has dimension 6 and decomposes under O
h
as A
1g
⊕ E
g
⊕ T
1u
(Proposition 5);
both three-dimensional pieces are spatial-vector representations, so the distinguishing content is
13
combinatorial, not spectral: the bonded graph K
2,2,2
admits 8 perfect matchings and 30 skew-
edge pairs, neither factoring as a three-color structure (Proposition 6). The matter-paper color
construction applies to tetrahedral defects and fails for octahedral defects on both spectral-
dimension and combinatorial grounds. This supports the structural distinction underlying the
quark/dark-sector split of the SSMTheory program: tetrahedral defects with the quark-class
matter, octahedral defects with the dark-sector species proposed in the companion paper [15] on
the same K
2,2,2
combinatorics. The free-field bulk classification (Theorem 2) of three topologi-
cal classes satisfying the vanishing Nielsen–Ninomiya consistency relation provides the spectral
background.
Limitations
The Dirac analysis does not derive SU(3) color from the spectrum; the matter paper’s construc-
tion is logically prior, using K
4
combinatorics independently of the Dirac operator. Specifically:
(a) the spectral result is a dimensional consistency check (the tetrahedral triplet has the di-
mension of the K
4
color space); since both bound triplets are spatial-vector representations,
the spectrum does not by itself separate the two defect classes, and the distinction between
them is combinatorial, not spectral, not a derivation of internal symmetry; (b) the multiplet
decomposition is a symmetry statement about the localized shell space, robust to the radial
profile of the trapping potential (Section 4.2), but is not an infinite-lattice bound-state theo-
rem; (c) the octahedral result is a symmetry-plus-combinatorics statement, with no separate
finite-cluster diagonalization performed; and (d) the physical interpretation of the octahedral
sector (its mass scale, cosmological abundance, and indirect-detection signatures) is the subject
of the companion dark-matter paper [15] and is neither used nor assessed here.
Open questions
Two natural extensions. First is the dynamical analysis: turning on the SU(3) gauge field of
the SSMTheory program on the FCC bonds and computing how the bound-mode multiplet
responds to gauge fluctuations would close the dynamical picture. Second is the 4D extension:
an FCC slice × discrete time gives a D
4
root-lattice structure with exact SO(4) isotropy in
the long-wavelength limit; the analogous defect-bound-mode analysis on that 4D lattice would
address the propagating-fermion content of the substrate.
Acknowledgments
I thank a referee for a careful reading of an earlier version of this work; that feedback prompted
the present reorganization around the defect-bound-mode analysis.
Data availability
A Python script reproducing all numerical and algebraic claims of this paper is included with
the submission as dirac_verify_defects.py; it runs 80 explicit assertions including the finite-
cluster Dirac diagonalization of Section 4 (which reproduces Table 1 across five well depths),
and exits with status 0 on success. The same script is also available online at https://github.
com/raghu91302/ssmtheory/blob/main/dirac_verify_defects.py. The full LaTeX source,
figure-generating Python scripts, and verification script constitute the complete reproducibility
package; no separate data files are required.
14
AI contributions statement
The author used Anthropic’s Claude as an assistant for language and presentation only: refining
prose and clarity, checking the L
A
T
E
X build and formatting, cross-checking internal consistency
of citations, dimensions, point-group conventions, and notation, and drafting supplementary
documentation (README and verification-script comments). The scientific content of this
work (the bulk zero-mode classification, the defect-bound-mode and representation-theory re-
sults, the K
4
/K
2,2,2
analysis, and the verification-script logic, including the finite-cluster Dirac
diagonalization) was conceived, derived, and verified by the human author, who takes full re-
sponsibility for all claims in the manuscript.
References
[1] H. B. Nielsen and M. Ninomiya, Nucl. Phys. B 185, 20 (1981); Erratum: Nucl. Phys. B
195, 541 (1982). doi:10.1016/0550-3213(81)90361-8.
[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] L. H. Karsten, Phys. Lett. B 104, 315 (1981). doi:10.1016/0370-2693(81)90133-7.
[7] F. Wilczek, Phys. Rev. Lett. 59, 2397 (1987). doi:10.1103/PhysRevLett.59.2397.
[8] A. Boriçi, Phys. Rev. D 78, 074504 (2008). doi:10.1103/PhysRevD.78.074504.
[9] M. Creutz, JHEP 04, 017 (2008). doi:10.1088/1126-6708/2008/04/017.
[10] P. F. Bedaque, M. I. Buchoff, B. C. Tiburzi, and A. Walker-Loud, Phys. Lett. B 662, 449
(2008). doi:10.1016/j.physletb.2008.03.020.
[11] D. B. Kaplan, E. Katz, and M. Ünsal, JHEP 05, 037 (2003). doi:10.1088/1126-
6708/2003/05/037.
[12] S. Catterall, JHEP 06, 027 (2008). doi:10.1088/1126-6708/2008/06/027.
[13] W. Celmaster, Phys. Rev. D 26, 2955 (1982). doi:10.1103/PhysRevD.26.2955.
[14] R. Kulkarni, Matter as Incomplete Crystallization: Quark Charges, Color Confinement,
and the Proton Mass from a Single Extra Node in the Vacuum Lattice, Phys. Open 27,
100423 (2026). doi:10.1016/j.physo.2026.100423.
[15] R. Kulkarni, Dark Matter as a Trapped K = 6 Remnant in the Octahedral Voids of the
FCC Vacuum Lattice, Zenodo preprint (2026), doi:10.5281/zenodo.20047901 (submitted to
Phys. Open; not peer-reviewed).
[16] L. H. Karsten and J. Smit, Nucl. Phys. B 183, 103 (1981). doi:10.1016/0550-3213(81)90549-
6.
[17] H. S. M. Coxeter, Regular Polytopes, 3rd ed., Dover, New York (1973).
[18] N. W. Ashcroft and N. D. Mermin, Solid State Physics, Saunders College, Philadelphia
(1976).
15
