# Prior-art scout: D4-lattice degree-4 isotropy (non-hypercubic lattice field theory)

Scout session: 2026-09-20. Four parallel fronts: (1) group theory, (2) lattice field theory, (3) lattice gas / lattice Boltzmann (FCHC), (4) tensor-isotropy + other fields.

---

## VERDICT: PARTIALLY-KNOWN

The core mathematical fact — claim (B) — is **known verbatim**, but in the lattice-gas / lattice-Boltzmann literature, not in lattice field theory: the 4D FCHC lattice (= the D4 lattice, 24 nearest neighbours, the vertices of the 24-cell, all permutations of (±1,±1,0,0)) has exactly isotropic lattice tensors E^(n) through rank 4, with anisotropy returning at rank 6 (Wolfram 1986 states this for the lattice he explicitly lists as "D4"; Frisch et al. 1987 prove the rank-4 isotropy; Wolf-Gladrow 2000 and Chen-Goldhirsch-Orszag 2008 restate it). The group-theoretic mechanism, claim (C), is textbook-established: full Aut(D4 lattice) = W(F4), order 1152, invariant degrees 2,6,8,12 (hence no independent degree-4 invariant), vs B4 = 2,4,6,8. What was NOT found anywhere: the dispersion-relation framing for a free scalar field theory — the explicit expansion ω² = k² − (a²/12)(k²)² + O(a⁴k⁶) on the D4 lattice, the exact fractional anisotropy (ak)⁴/720, and the "free extra power of a² of Lorentz-suppression, no Symanzik tuning needed" consequence (D). Closest lattice-field-theory statements: Neuberger 1987 arguing F4 (co)root lattices are "singularly well suited" for scalar-field regularization; Katz & Nogradi 2026 (QCD on the 16-cell honeycomb) with D†D = p² − (1/6)a²p⁴ + O(a⁴) and Lorentz breaking first at O(a⁴); de Soto & Roiesnel 2007 for the hypercubic p_hat² ≈ p² − (a²/12)p^[4] expansion. No source joins the invariant-theory, lattice-gas, and QFT-dispersion pieces into the single claim.

---

## 1. GROUP THEORY — claims (B), (C) — ALL VERIFIED (Established)

| Claim | Status | Source |
|---|---|---|
| F4 Weyl group: order 1152, invariant degrees 2, 6, 8, 12 | **verified** | Encyclopedia of Mathematics "Coxeter group": F4 exponents 1,5,7,11, "exponents are one less than the degrees of the generating invariants"; Wikipedia Coxeter table: order 1152 |
| B4/C4 (hyperoctahedral, point group of Z^4): order 384, degrees 2,4,6,8 | **verified** | EOM B_n exponents 1,3,…,2n−1 → degrees 2,4,6,8; order 2^4·4! = 384 (Wikipedia hyperoctahedral group) |
| W(D4) Weyl group: order 192, degrees 2,4,4,6 (NOT the point group of the D4 lattice) | **verified** | EOM D_n exponents 1,3,…,2n−3,n−1 → 1,3,3,5 → degrees 2,4,4,6; Wikipedia Coxeter table: D4 order 192. So the claimed subtlety is right: the D4 *lattice* point group is not W(D4). |
| Aut(D4 lattice) = W(F4), order 1152 | **verified** | Nebe–Sloane lattice catalogue, D4 entry: GROUP_ORDER 1152, GROUP_NAME W(F4) (math.rwth-aachen.de/~Gabriele.Nebe/LATTICES/D4.html); Hirao–Nozaki–Tasaka, arXiv:2303.09000: "W(F4) is a discrete subgroup of O(R4) of order 1152 and coincides with the automorphism group Aut(D4) := {σ ∈ O(R4) | σ(D4) = D4}" |
| Z^4 point group B4 order 384 vs D4 lattice order 1152, ratio exactly 3 | **verified** | Nebe–Sloane Z4 entry GROUP_ORDER 384; 1152/384 = 3. (Same "3 times as large" as in Celmaster's abstract.) |
| D4 nearest neighbours = 24 permutations of (±1,±1,0,0) = 24-cell vertices | **verified** | Hirao–Nozaki–Tasaka: "The D4 lattice is a root lattice in R4 generated by all permutations of (±1,±1,0,0) over Z"; Wikipedia 24-cell |
| D4 densest lattice packing in 4D (Korkine–Zolotareff) | **citation verified** (original proof not re-read) | A. Korkine & G. Zolotareff, "Sur les formes quadratiques positives quaternaires", Mathematische Annalen 5(4) (1872) 581–583, DOI 10.1007/BF01442912 (Crossref-verified); follow-up "Sur les formes quadratiques", Math. Ann. 6 (1873) 366–389, DOI 10.1007/BF01442795. Densest-lattice attribution: inherited-unchecked from secondary sources. |
| SPLAG locations | **TOC-verified** | Conway & Sloane, *Sphere Packings, Lattices and Groups* (3rd ed.): Ch.1 §1.5 p.12; Ch.3 §4.1 "The Automorphism Group of a Lattice" p.90; Ch.4 §7.1 "The Lattice Dn" p.117, §7.2 "The Four-Dimensional Lattice D4" p.118, §7.3 p.119, §7.4 p.120 |
| Humphreys location | **TOC-verified, table entries NOT VERIFIED AT SOURCE** (OCR-unreadable scan) | James E. Humphreys, *Reflection Groups and Coxeter Groups*, Ch.3, §3.20 "Exponents and degrees of Weyl groups", p.82. The degree/orders themselves are independently verified via EOM/Wikipedia/Nebe–Sloane/arXiv above. |
| Coxeter 1951 (Duke Math. J. 18, 765) | not fetched; the invariant degrees above follow from standard tables (Bourbaki plates) — treat Coxeter 1951 as inherited-unchecked | — |

**Conclusion (C) is sound:** F4 has no degree-4 invariant besides (k²)², so degree-4 isotropy is symmetry-protected, not tuned — for a lattice whose point group is exactly W(F4). W(D4) alone (order 192, degrees 2,4,4,6) WOULD allow a degree-4 invariant, which is why the "point group of the D4 lattice" vs "D4 Weyl group" distinction in the claimed result is load-bearing and correct.

**Caveat (from tensor scout):** F4 is not the largest *finite* 4D group with no quartic anisotropy — H4 (600-cell/120-cell, non-crystallographic) is isotropic to rank 8 (Wolfram 1986, p.492–493). F4 is the largest among *crystallographic/Weyl-lattice* groups, which is the relevant class for a lattice regulator.

---

## 2. THE SAME RESULT IN LATTICE GAS / LATTICE BOLTZMANN — THE MOST DIRECT PRIOR ART (Established, verified-at-source)

**This is where claim (B) lives, in tensor language.** The 4D FCHC lattice is exactly the D4 lattice; its 24-neighbour shell is exactly the 24-cell vertices; its rank-4 isotropy and rank-6 return are stated and proved.

- **Frisch, d'Humières, Hasslacher, Lallemand, Pomeau & Rivet, "Lattice Gas Hydrodynamics in Two and Three Dimensions", Complex Systems 1 (1987) 649–707.** Verified at source (Complex Systems page/PDF).
  - p.656 §2.3: "The residing lattice is face-centered-hypercubic (FCHC), defined as the set of signed integers (x1,x2,x3,x4) such that x1+x2+x3+x4 is even. Each node is connected via links of length c = sqrt(2) to 24 nearest neighbors, having two coordinates differing by ±1."
  - p.673: "In order to eventually obtain the Navier-Stokes equations, the tensor T… given by equation (6.1) must be isotropic, that is, invariant under the full orthogonal group."
  - p.673: "Crucial observations … are the isotropy of pairwise symmetrical tensors for the triangular FHP lattice in two dimensions and the face centered-hypercubic (FCHC) lattice in four dimensions."
  - p.674: proof that on FCHC the anisotropic term vanishes: "invariance requires φ = 0, which proves isotropy."
  - p.692 Appendix A lists the 24 FCHC velocities: (±1,±1,0,0) and permutations.
- **S. Wolfram, "Cellular Automaton Fluids 1: Basic Theory", J. Stat. Phys. 45 (1986) 471–526.** Verified at source. The closest thing to a stated theorem:
  - p.492: "The {3,4,3} polytope has 24 vertices with coordinates corresponding to permutations of (±1,±1,0,0). It yields E(n) that are isotropic up to n=4."
  - Table V (p.496), root-vector lattices: "D4 SO(8) 24 4", where the last column nmax = "maximum even n at which the E(n) are found to be isotropic" — i.e. anisotropy returns at rank 6. **Wolfram explicitly names the lattice D4.** (He does not literally write "FCHC = D4" in one sentence.)
  - p.491 2D theorem: "E(n) is isotropic if and only if M does not divide any of integers n, n−2, n−4, …" (M-gon velocity sets).
  - p.493–494 group-theory mechanism (irreducible representation of E^(n) ⇒ one component ⇒ rotationally invariant): rank-4 under the octahedral group has two components ⇒ anisotropic.
  - p.495 3D: "None yield isotropic E^(4)" for the most symmetrical 3D lattices; "A system with icosahedral point symmetry would be guaranteed to yield an isotropic E^(4), but … it is not possible to tessellate three-dimensional space with regular icosahedra."
- **Wolf-Gladrow, *Lattice-Gas Cellular Automata and Lattice Boltzmann Models* (Springer LNM 1725, 2000).** Verified at source. p.8: "Wolfram (1986) showed that lattice tensors over the face-centered hypercube (FCHC) are isotropic up to rank 4." p.108: "The {3,4,3}-polytop is referred to as face-centered hypercube (FCHC). It has 24 corners with coordinates which are permutations of (±1,±1,0,0). The corresponding lattice tensors are isotropic up to 4th rank inclusively."
- **Chen, Goldhirsch & Orszag, "Discrete Rotational Symmetry, Moment Isotropy, and High Order Lattice Boltzmann Models", J. Sci. Comput. 34 (2008) 87–112 (arXiv:0709.1464).** Verified at source (abstract). p.11: "All velocities … have the same magnitude sqrt(2)c. It is well known that this FCHC lattice is isotropic up to 4th order …, and its moments at 6th order and higher are not isotropic."
- **Frisch, Hasslacher & Pomeau, "Lattice-Gas Automata for the Navier-Stokes Equation", PRL 56, 1505 (1986).** Verified. p.1506: HPP square lattice "is invariant under π/2 rotations. Such a lattice symmetry is insufficient to insure the isotropy of the fourth degree tensor relating momentum flux to quadratic terms in the velocity." 3D: "the face centered cubic, with twelve equal-speed velocity directions … the relevant tensors … depend now on three constants."

**Why FCHC was chosen (explicit in the sources):** no 3D regular/crystallographic nearest-neighbour lattice gives isotropic rank-4 tensors; the 4D FCHC/D4 shell does; project/pseudo-periodize the 4th dimension to simulate 3D. **Anisotropy order:** rank 6 is where isotropy fails (Wolfram nmax=4; Chen et al. "6th order and higher are not isotropic") — the same fact the claimed result expresses as "anisotropy first appears at O(a⁴k⁶)".

Note: the historical wording is "isotropy of fourth-rank lattice tensors", not "k⁴-cancellation in the dispersion relation" — mathematically the same content (the fourth moment Σ_v v_i v_j v_k v_l ∝ δ-pairs ⇒ the degree-4 term of the small-k expansion is exactly isotropic).

---

## 3. NON-HYPERCUBIC LATTICE FIELD / GAUGE THEORY (Established; small but real literature)

- **W. Celmaster, "Gauge Theories on the Body-Centered Hypercubic Lattice", Phys. Rev. D 26, 2955 (1982).** Verified at source (APS + INSPIRE). Abstract verbatim: "The four-dimensional body-centered hypercubic lattice has a point symmetry group which is three times as large as that of the simple hypercubic lattice. This enlarged symmetry is implemented by introducing an action consisting of a sum over triangular plaquettes." (The BCH vertex set is the same as D4*/F4 — per Katz & Nogradi 2026.) **NOT verified:** any dispersion-relation / leading-anisotropy-order computation — the APS PDF is paywalled and the abstract does not state the k⁴ result. INSPIRE keywords (continuum limit, propagator, symmetry: rotation) suggest propagator/continuum-limit analysis, but the degree-4 cancellation is not stated in any accessible source.
- **Celmaster follow-ups (verified at source):** Celmaster & Krausz, "Fermion mutilation on a body-centered tesseract", Phys. Rev. D 28, 1527 (1983) — chiral fermions on BCH, "the resulting continuum field theory is not Lorentz invariant"; Celmaster, "Average plaquette of SU(2) gauge theory on a body-centered hypercubic lattice", Phys. Rev. D 28, 2076 (1983); Celmaster, "Evidence for Improved Scaling of SU(2) Gauge Theory on a Body-Centered Hypercubic Lattice", Phys. Rev. Lett. 52, 403 (1984). **No** Celmaster–Krausz "face-centered cubic" gauge paper found (searched; not found).
- **H. Neuberger, "Spinless fields on F4 lattices", Phys. Lett. B 199, 536–540 (1987).** Verified (INSPIRE/Elsevier abstract): "It is argued that four-dimensional lattices based on the (co)roots of the exceptional Lie algebra F4 are singularly well suited for the regularization of scalar fields." **This is the closest lattice-QFT statement to claim (D)** — but the abstract does not spell out the k⁴ cancellation.
- **Bhanot, Bitar, Heller & Neuberger, "phi4 on F4: Analytical results", Nucl. Phys. B343, 467 (1990) and "phi4 on F4: Numerical results", Nucl. Phys. B353, 551 (1991)** (Crossref-verified; erratum B375, 503).
- **S. D. Katz & D. Nogradi, "QCD on the 16-cell honeycomb", Phys. Rev. D 114, 054504 (2026), arXiv:2512.10604.** Verified at source. Explicit for the Wilson-Dirac operator on the 16-cell honeycomb (same 24-neighbour shell family): "D†D = p² − (1/6)a²p⁴ + O(a⁴)", the O(a²) correction Lorentz invariant, "Lorentz breaking first appears at O(a⁴)", E = |p| + O(a⁴); sites have 24 nearest neighbours; 16-cell/24-cell vertex-set symmetry groups have 1152 elements vs 384 (cubic). **This is an explicit degree-4-isotropy statement in QFT — for the fermion D†D, not the free-scalar dispersion.**
- **F. de Soto & C. Roiesnel, "On the reduction of hypercubic lattice artifacts", JHEP 09 (2007) 007 (arXiv:0705.3523).** Verified at source. States the standard hypercubic expansion: "p_hat² ≈ p² − (a²/12) p^[4] + (a⁴/360) p^[6] − …" with p^[n] = Σ_mu p_mu^n — **this verifies claim (A)'s hypercubic formula**, and says the improved rotational restoration on the BCH lattice "can be analyzed in terms of the primitive invariant p[4]", citing Neuberger.
- **Symanzik tuning (the alternative to changing the lattice):** K. Symanzik, "Continuum limit and improved action in lattice theories", Nucl. Phys. B226, 187 (1983) and II, B226, 205 (1983) (Crossref-verified).
- **Random/simplicial lattices (separate, larger line, statistical restoration):** Christ, Friedberg & Lee, "Random Lattice Field Theory: General Formulation", Nucl. Phys. B202, 89 (1982); "Gauge theory on a random lattice", Nucl. Phys. B210, 310 (1982); "Weights of Links and Plaquettes in a Random Lattice", Nucl. Phys. B210, 337 (1982) (INSPIRE-verified; note the task brief's page assignments for the two B210 papers were swapped); Drouffe & Moriarty, "Gauge theories on a simplicial lattice", Nucl. Phys. B220, 253 (1983).

**Direct answer to the key question:** no lattice-field-theory source states in so many words "the D4/24-cell lattice gives an exactly isotropic k⁴ term for the free scalar dispersion relation". The facts exist piecemeal (Neuberger's F4-scalar argument; the 2026 Katz–Nogradi O(a⁴) result; the LBM rank-4 isotropy; the group theory), never joined into claims (B)+(D).

---

## 4. THE TENSOR-ISOTROPY CRITERION (Partially stated; clean sources exist)

- **Wolfram 1986** (above): the closest to an explicit theorem — E^(n) isotropic iff its representation under the lattice's point group has a single (trivial) component; rank-4 under octahedral group has two components ⇒ anisotropic; p.489 gives the unique isotropic form E^(2n) = M·Δ^(2n)/[d(d+2)…(d+2n−2)].
- **Chen–Goldhirsch–Orszag 2008** (above): necessary-and-sufficient conditions for maximal isotropy order of a velocity set; "2D hexagonal and 4D-FCHC lattices are 4th order isotropic, while the 2D square and 3D cubic lattices are not."
- **Neumann's principle (crystallography):** IUCr "An Introduction to Crystal Physics" §4: "According to Neumann's principle the tensor representing any physical property should be invariant with regard to every symmetry operation of the given crystal class." The *invariant-theory phrasing* ("isotropic rank-4 tensor iff no degree-4 invariant beyond (k²)²") appears nowhere verbatim in the sources found — it is a faithful synthesis of Wolfram §3.4 + standard invariant theory (own inference).
- **The "one extra power of a²" design principle is standard in LBM** under the names "moment isotropy", "rotational symmetry of the generating discrete vector set", "isotropy conditions" — not under the invariant-theory name.

---

## 5. OTHER FIELDS — analogues exist (Established, verified at source)

- **Isotropic finite-difference stencils:** Patra & Karttunen, "Stencils with isotropic discretization error for differential operators", Numer. Methods PDE 22(4) (2006) 936–953 (isotropic-error Laplacian bilaplacian stencils); Mattila et al., "High-Accuracy Approximation of High-Rank Derivatives: Isotropic Finite Differences Based on Lattice-Boltzmann Stencils", ScientificWorldJournal 2014 (LB stencils as isotropic FD operators — the same isotropy order concept).
- **Hexagonal grids / wave FDTD:** Hamilton & Bilbao, "Hexagonal vs. Rectilinear Grids for Explicit Finite Difference Schemes for the Two-dimensional Wave Equation", ICA 2013 — hexagonal grid "the more natural choice to emulate the isotropy of the Laplacian".
- **Electromagnetic FDTD:** Shen et al., "A new FDTD stencil for reduced numerical anisotropy…", Int. J. RF and Microwave CAE 17(5) (2007) 447–454.
- **Phase-field / dendritic growth:** Karma & Rappel, "Quantitative phase-field modeling of dendritic growth in two and three dimensions", Phys. Rev. E 57 (1998) 4323 — §VI "THREE-DIMENSIONAL EQUATIONS AND LATTICE ANISOTROPY"; p.4337 expands the discretized Laplacian as the isotropic Laplacian plus cubic-symmetry ∂x⁴+∂y⁴+∂z⁴ term modifying anisotropy — i.e. claim (A)'s structure in 3D; Ji et al., "Isotropic finite-difference approximations for phase-field simulations…", J. Comput. Phys. 457 (2022) 111069 (arXiv:2110.12448) — modern direct attack on "spurious lattice anisotropy".
- **Not found:** any numerical-analysis source calling D4/FCHC the "best 24-point 4D Laplacian stencil" (null result only).

---

## 6. WHAT WAS NOT FOUND (where the claimed result may still be original)

1. The free-scalar dispersion-relation statement on the D4 lattice — ω² isotropic through k⁴, anisotropy first at O(a⁴k⁶), fractional anisotropy exactly (ak)⁴/720 — **not stated anywhere found** (lattice QFT, lattice-Boltzmann, or numerical analysis).
2. Claim (A)'s exact "fractional anisotropy = (ak)²/16" for Z^4 — the formula p_hat² ≈ p² − (a²/12)p^[4] is standard (de Soto–Roiesnel), but this specific fractional-anisotropy number was not found.
3. The joined claim (D) — "D4 over Z⁴ buys one extra power of a² of Lorentz suppression for free, with no tuned Symanzik coefficients; protection survives radiative corrections because an exact-F4-symmetric regulator generates only F4-invariant operators" — the ingredients exist (Neuberger 1987: F4 lattices "singularly well suited"; Katz–Nogradi 2026: O(a⁴) first Lorentz breaking; standard invariant-theory argument) but are nowhere joined. The radiative-stability argument is a direct corollary of F4-invariant counterterms (own inference; standard).
4. Any statement of the F4/B4 invariant-degree contrast used as lattice-design rationale in QFT.

## 7. WHERE I LOOKED (scope)

Web searches + fetches across: APS (journals.aps.org), INSPIRE-HEP metadata, Crossref, arXiv (incl. HTML/ar5iv), Complex Systems 1987 PDF, Springer/Google Books and Wolfram-hosted J. Stat. Phys. PDF, AWI-hosted Wolf-Gladrow book, Scholarpedia LBM, Nebe–Sloane lattice catalogue, Encyclopedia of Mathematics, Wikipedia, RWTH/arXiv group-theory sources, IUCr teaching material, TU Eindhoven / Illinois / PMC abstracts. Paywalled/unverified: Celmaster PRD 26 body text (APS paywall), d'Humières–Lallemand–Frisch Europhys. Lett. 2 (1986) 291 (IOP bot-block; citation verified indirectly via Complex Systems ref [2]), Humphreys table text (OCR), Bourbaki hardcopy, SPLAG full text (TOC only), Korkine–Zolotareff original proof, Coxeter 1951 Duke Math. J. No attempt was made to read paywalled PDFs beyond abstracts; items above are marked accordingly (verified-at-source vs inherited-unchecked).

## Evidence-class summary

- **Established (verified at source):** all group-theory claims (§1); FCHC/D4 rank-4 isotropy and rank-6 return (§2); existence/claims of Celmaster, Neuberger, Bhanot et al., Katz–Nogradi, de Soto–Roiesnel, Symanzik, Christ–Friedberg–Lee, Drouffe–Moriarty (§3); Wolfram tensor criterion, Chen et al. criterion, Neumann principle (§4); FD/FDTD/phase-field analogues (§5); hypercubic p_hat² expansion (claim A's formula) (§3).
- **Established citation, content inherited-unchecked:** Korkine–Zolotareff 1872 densest-packing attribution (proof not re-read); Coxeter 1951; Bourbaki; SPLAG §7.2 content (TOC-verified location only); Humphreys table entries.
- **Own inference (clearly labeled):** the exact-phrase invariant-theory criterion (rank-4 isotropic iff no degree-4 invariant) as a statement; the radiative-stability argument; the (ak)⁴/720 number as a derivation rather than a found citation.
- **Anomaly/null:** no statement of the scalar dispersion-relation result anywhere; no "24-cell best 4D stencil" in numerical analysis.