Fermion Mass Formula from Spectral Geometry on $S^3/2I$

The Standard Model contains 12 fundamental fermions spanning 12 orders of magnitude in mass. The Higgs mechanism explains how particles acquire mass. It does not explain why they have the masses they do. This paper constructs a mass formula from four ingredients, each traced to a single topological postulate: $S^1 = \partial(\text{Mobius}) \hookrightarrow S^3, \quad \partial S^3 = \emptyset$.

The formula is applied to the 8 nontrivial irreducible representations of the binary icosahedral group across 3 isolated flat connections, producing 24 mass predictions. Of these, 10 correspond to Standard Model fermions: 9 are reproduced within a factor of 3 and all 10 within a factor of 4. Two SM fermions have no assignment: $\nu_2$ falls between adjacent entries, and the charm quark is displaced by the $T_3$ resolution at rank 17. Eight predictions have no SM assignments.

I. The Formula

$\Large \boxed{m(\rho, \sigma) = \mu_\Lambda \cdot C_{\text{geom}}(\rho) \cdot \left(\sqrt{\Omega_\Lambda}\right)^{\text{dist}(\rho)/30} \cdot T^2(\rho \otimes \sigma)}$

Four factors. Four sources. Each traces independently to the topological postulate.

II. The Factors

1. Neutrino Floor $\mu_\Lambda$

The vacuum energy density of the cosmological constant defines the overall mass scale:

$\mu_\Lambda \equiv \rho_\Lambda^{1/4} \approx 2.25 \text{ meV}$

This is the fourth root of Λ, set by the ground mode ($m_h = 0$) of the Mobius surface. All particle masses trace back to this vacuum energy floor, scaled by the hierarchical factors that place each irrep at its position on the spectrum.

The neutrino mass sector provides direct access to this scale:

The multipliers (4, 22) emerge from parity violation due to the Mobius twist. KATRIN and cosmological bounds provide the falsification window.

2. Kostant Sunflower $C_{\text{geom}}(\rho)$

Each irrep sits at a specific position on the finite domain, encoded by its Kostant exponents. The geometric mean of the phase factor $C(e/D) = 2\sin^2(\pi e/D)$ over these exponents gives the irrep's amplitude on the spectrum.

$C_\text{geom}(\rho) = \bigl(\prod_e 2\sin^2(\pi e/D)\bigr)^{1/(2\,\dim\rho)}$

The domain size depends on spin: $D = 60$ for integer-spin, $D = 120$ for half-integer. This encodes the fundamental distinction between bosons and fermions in the geometry. The domain size traces to the edge stabilizer $Z_4 \subset 2I$: integer-spin irreps carry only real $Z_4$ content ($D = 60 = |I|$), half-integer carry only complex pairs ($D = 120 = |2I|$).

3. McKay Elevator $(\sqrt{\Omega_\Lambda})^{\text{dist}/30}$

The McKay graph encodes the distance of each irrep from the trivial representation $R_0$. Distance determines orders of magnitude separation from the vacuum floor via the hierarchy exponent $(\sqrt{\Omega_\Lambda})^{\,\text{dist}/30}$, where the denominator $h(E_8) = 30$ is the Coxeter number.

R0(1) -- R1(2) -- R3(3) -- R6(4) -- R7(5) -- R8(6) -- R5(4) -- R2(2)
                                                 |
                                                R4(3)  dist 6
  0        1        2        3        4         5        6        7

Half-integer spin: R₁, R₂, R₆, R₈. Integer spin: R₀, R₃, R₄, R₅, R₇.

The $j_\text{first}$ rule follows from the McKay multiplicity structure: the spin-$j$ representation of $SU(2)$, restricted to $2I$, first contains irrep $\rho$ at exactly $j = \text{dist}(\rho)/2$. This holds for both spin parities and is a consequence of the McKay correspondence between the $2I$ representation graph and the extended $E_8$ Dynkin diagram.

4. Reidemeister Torsion $T^2(\rho \otimes \sigma)$

Three flat SU(2) connections on $S^3/2I$ provide the generation mechanism. Each has $H^1 = 0$: no moduli, no mixing between vacua. The Reidemeister torsion encodes the fine structure within each mass shell.

Integer-spin irreps have exact closed forms. The Galois pair ratio $T^2(R_3)/T^2(R_4) = \varphi^{-4}$ is exact to 70+ digits.

Half-integer irreps ($R_1$, $R_2$, $R_6$, $R_8$) are computed from the same spectral data by the same method: the Laplacian eigenvalues and their multiplicities are exact (from $SU(2)$ representation theory and the $2I$ character table), and the spectral zeta function has a unique meromorphic continuation. The Hurwitz decomposition expresses each torsion as a finite sum of $L$-function derivatives at $s = 0$, each computable to arbitrary precision.

The difference is selectivity: integer-spin irreps retain only 4 of 16 Dirichlet characters (conductors 2, 3, 5, 5), producing algebraic closed forms. Half-integer irreps retain 28 to 32 characters. The resulting sums are equally exact; they remain as finite combinations of $L'(0, \chi)$ values rather than reducing to algebraic expressions. The decimal values displayed throughout this paper are truncations of well-defined mathematical constants, computable to any desired precision by the same methods that verified the integer-spin values to 79 digits.

The 24 vacuum torsion values follow from $\log T^2(\rho \otimes \sigma) = \sum_\tau N_{\rho\sigma\tau} \log T^2(\tau)$:

III. The 24 Predictions

24 predictions from the mass formula. The table below shows assignments to Standard Model (SM) fermions and unassigned predictions.

The assignment is constrained, not free. The formula produces 24 ranked entries; four structural filters narrow which entries can correspond to which fermions. The Coxeter-Galois gate (§IV.4) assigns weak isospin $T_3 = \pm 1/2$ to each entry. The $Z_3$ face decomposition (§IV.1) restricts color-charged assignments to irreps carrying colored pairs. The $Z_4$ edge decomposition (§IV.2) fixes the domain for each spin parity. The eta sign gate (§IV.3) restricts positive electric charge to negative-eta entries. Together these filters determine the quantum numbers $(T_3, Y, Q)$ at each $(\rho, \sigma)$ pair before any mass comparison is made.

Notes on the table.

Neutrino masses. The $\nu_1$ observed value is inferred from $\Delta m^2_{21}$ assuming normal hierarchy with a near-zero lightest mass; the absolute mass scale is experimentally unknown. KATRIN, JUNO, and Project 8 will constrain this. The $\nu_3$ value uses $\sqrt{\Delta m^2_{31}}$ as a proxy. The ratio 1.98 for $\nu_1$ depends on this assumption.

The down quark (rank 14). The ratio of 3.22 is the weakest assigned hit after $\nu_2$, and the only charged fermion outside the ×3 window. The down quark mass itself carries large uncertainty (4.67 $\pm$ 0.5 MeV from lattice QCD), but even at the upper end of the allowed range the tension remains. Whether this reflects a systematic residual at high McKay distance or a needed correction in the $R_8$ Galois vacuum sector is open.

Rank 15 resolution. Rank 15 sits between the muon (105.7 MeV) and strange quark (93.4 MeV). The Coxeter-Galois gate (§IV.4) assigns $T_3 = -1/2$ to this entry. The $Z_3$ face decomposition then splits the mass shell: the singlet component carries the muon ($Q = -1$), the colored component carries the strange quark ($Q = -1/3$). Both particles occupy the same $(\rho, \sigma)$ address in different color sectors.

Rank 17 resolution. Rank 17 ($R_4$, std) has $j_\text{first} = 2$ (integer). Stage 1 of the Coxeter-Galois gate assigns $T_3 = -1/2$, resolving the former $c/\tau$ ambiguity in favor of the tau lepton ($Q = -1$, from the singlet $Z_3$ sector). The charm quark ($T_3 = +1/2$) does not occupy rank 17; its assignment within the 24-entry table is open (see §VI).

The b-quark cluster. Five entries (ranks 18-22) from four distinct $(\rho, \sigma)$ pairs cluster near the b mass at 4.18 GeV, with ratios from 1.28 to 2.80. This is a dense region of the spectrum where multiple irrep-vacuum combinations land in the same mass shell. The closest entry is rank 18 ($R_2$, gal, ratio 1.28). Whether the remaining entries represent structural redundancy, mixing contributions, or unassigned states is open.

IV. Particle Identity

The mass formula assigns each Standard Model fermion to a pair $(\rho, \sigma)$. The mass comes from the torsion. The identity comes from the stabilizer structure of the icosahedron.

The binary icosahedral group $2I$ inherits three stabilizer subgroups from the icosahedron: face ($Z_3$, order 3), edge ($Z_2$, lifting to $Z_4$ in the double cover), and vertex ($Z_5$, order 5). Each irrep of $2I$ restricts to these subgroups, producing three independent decompositions. Two encode known Standard Model structure. The third connects to the electroweak sector through the Möbius twist.

1. Color from Faces

The face stabilizer $Z_3 \subset 2I$ is generated by order-3 elements. Restricting each irrep to $Z_3$ decomposes it into color singlets (lepton-type) and color triplet/anti-triplet pairs (quark-type):

Every assigned fermion's color charge is accommodated by the $Z_3$ decomposition of its irrep. The electron (color singlet) sits on $R_7$, which carries 1 singlet. The quarks sit on irreps with colored pairs. The neutrinos sit on $R_1$, which has no singlet content at the irrep level; the tensor product $R_1 \otimes \sigma$ with the vacuum connection generates the singlet channel through which the neutrino propagates as a color-neutral state.

The face stabilizer $Z_3$ corresponds to the $SU(3)$ color factor under the McKay correspondence: removing $R_3$ (the node killed by $Z_3$ alongside $R_4$ and $R_8$) from the extended $E_8$ Dynkin diagram produces the maximal subalgebra $SU(3) \times E_6$.

Color is generation-independent. The equivariant eta at the order-6 face class is perfectly vacuum-invariant: $\eta = 2$ in all three vacua. The face geometry looks the same from every vacuum. This matches the Standard Model: color charge is the same across generations.

2. Domain from Edges

The domain size $D = 60$ vs $120$ introduced in §II.2 traces to a deeper structure. The edge stabilizer $Z_2$ lifts to $Z_4 \subset 2I$, generated by order-4 elements. The $Z_4$ decomposition enforces an exact binary:

• Half-integer irreps $\{R_1, R_6, R_8, R_2\}$: all $Z_4$ content is complex pairs. Domain $D = 120$.
• Integer-spin irreps $\{R_0, R_3, R_7, R_5, R_4\}$: all $Z_4$ content is real. Domain $D = 60$.

This is the spin-statistics connection, built into the group. The mass formula encodes it through $C_\text{geom}$, which evaluates the Kostant exponents on the $D = 60$ or $D = 120$ grid.

3. The Eta Sign Gate

The Dirac eta invariant $\eta(\rho, \sigma)$ varies with the vacuum. Across all SM-assigned entries in the mass formula, a strict constraint holds:

$\eta(\rho, \sigma) > 0 \implies Q \leq 0$

Equivalently: positive electric charge requires negative eta. All entries with $Q = +2/3$ ($u$, $c$, $t$) have $\eta < 0$. All entries with $\eta > 0$ (ranks 1, 14, 18, 22) carry fermions with $Q = 0$ or $Q = -1/3$.

The eta invariant measures spectral asymmetry: the parity content of the mode. This gate connects parity to electric charge through the spectral geometry.

4. Weak Isospin from the Coxeter-Galois Gate

The same irrep carries different fermions in different vacua. $R_8$ produces the up quark (trivial vacuum), down quark (Galois), and muon or strange (standard). The vacuum $\sigma$ selects the electroweak identity. The rule determining weak isospin $T_3$ at each $(\rho, \sigma)$ is a two-stage filter, computable entirely from Tools 2-4 before any mass is evaluated.

Stage 1: Spectral parity. If $j_\text{first}(\rho, \sigma) \in \mathbb{Z}$ (integer), the mode enters the Dirac spectrum through the bosonic channel: $T_3 = -1/2$.

Stage 2: Coxeter-Galois gate. For half-integer $j_\text{first}$, evaluate two conditions. First: does $\rho$ carry the Coxeter conjugate pair $(13, 17)$ under $h(E_8) = 30$, and is it stripped by the vacuum? The pair lives in the Kostant exponents of $R_2$, $R_6$, and $R_8$ (all half-integer). The trivial vacuum preserves it ($\rho \otimes R_0 = \rho$). The nontrivial vacua strip it (half-integer $\otimes$ half-integer $\to$ integer-spin components, which carry none of the eight $E_8$ exponents). Second: does the tensor product $\rho \otimes \sigma$ contain a Galois-nonfixed irrep? The four Galois-nonfixed irreps are $\{R_1, R_2, R_3, R_4\}$, whose characters involve $\sqrt{5}$; the Galois-fixed irreps $\{R_0, R_5, R_6, R_7, R_8\}$ are invariant under $\sqrt{5} \to -\sqrt{5}$.

Both conditions must hold for $T_3 = -1/2$. If either $(13, 17)$ is preserved (trivial vacuum, or $\rho$ lacks the pair) or the product is entirely Galois-fixed, then $T_3 = +1/2$.

In one line:

$T_3 = -\tfrac{1}{2} \iff j_\text{first} \in \mathbb{Z},\ \text{or}\ (13,17)\ \text{stripped and}\ \rho \otimes \sigma\ \text{has Galois-nonfixed content.}$

Verification against SM-assigned entries:

Ten entries, ten correct assignments. Rank 17 is resolved as $\tau$ ($T_3 = -1/2$). Rank 24 is resolved as $t$ ($T_3 = +1/2$): $R_2 \otimes R_1 = R_5$, and $R_5$ is Galois-fixed, so the Galois involution has nothing to act on and the state retains upper isospin by structural inertia.

Hypercharge and electric charge follow from $Z_3$. Once $T_3$ is fixed, the $Z_3$ face decomposition (§IV.1) determines color, and the Gell-Mann-Nishijima formula $Q = T_3 + Y/2$ closes the circuit:

All ingredients are representation-theoretic: $j_\text{first}$ from SU(2) $\to$ 2I branching rules, Kostant exponents from the McKay correspondence, tensor product decompositions from the character table, and Galois-fixed/nonfixed status from Gal($\mathbb{Q}(\sqrt{5})/\mathbb{Q}$). Identity is determined before mass is computed.

5. The Vertex and the Twist

The vertex stabilizer $Z_5 \subset 2I$ produces a well-defined decomposition of each irrep into three components $(n_0, n_1, n_2)$, where $n_1$ counts $(\zeta, \zeta^4)$ pairs and $n_2$ counts $(\zeta^2, \zeta^3)$ pairs under the fifth roots of unity. The Galois conjugation $\sqrt{5} \to -\sqrt{5}$ swaps $n_1 \leftrightarrow n_2$.

$R_1$ (neutrinos) and $R_2$ (top/bottom) are pure and complementary under the Galois action. $R_7$ (electron) is maximally democratic. The two nontrivial vacua are $R_1$ and $R_2$ themselves, Galois conjugates that differ precisely in their $Z_5$ content: $R_1$ carries only $n_1$, $R_2$ carries only $n_2$. The electroweak distinction between vacua IS the vertex decomposition.

The dodecahedral angular defect $\pi/5$ at each vertex, halved by the Möbius $\mathbb{Z}_2$ holonomy to $\pi/10$, enters the weak coupling constant as $\cos(\pi/10) = \sqrt{(2+\varphi)}/2$. In the companion gauge coupling analysis, this correction uniquely improves $\alpha_W$ (from 5.6% to 0.4% error) and uniquely degrades $\alpha$ and $\alpha_s$ if misapplied. The vertex geometry reaches the observer through the Möbius twist, and only the parity-violating force carries the projection.

$R_7$ occupies a special position in this structure. $R_7 \otimes R_1 = R_7 \otimes R_2 = R_6 + R_8$: the two nontrivial vacua produce identical torsion at $R_7$. The Dirac eta invariant captures this through the antisymmetric vacuum combination $(5/2)(\eta_\text{std} - \eta_\text{gal})$, which equals an integer for every irrep and is uniquely zero at $R_7$. The irrep that carries the electron sees both vacua identically. It sits at the center of the vertex structure, where the Galois distinction vanishes.

6. Summary

The three stabilizer orders 2, 3, 5 are the primes dividing $|2I| = 120$ and the conductors of the four surviving Dirichlet characters in the torsion L-basis. They generate color, domain, and (through their interfaces) gravity and the weak force. The mass formula computes how heavy. The stabilizer structure says what.

V. Dead Zone, Targets, and Exclusions

Eight entries have no Standard Model assignments. Six fall in the "dead zone" (ranks 4-9), the mass range between $10^{-9}$ and $10^{-6}$ GeV (eV to keV) where no known fundamental fermion exists and experimental sensitivity to new states is limited. One is a target (rank 16) in a normal mass range where a particle could exist but has no current SM assignment. One is excluded (rank 3) by existing data.

The dead zone is actively probed by sterile neutrino and warm dark matter searches. Physical states at these masses require extremely suppressed non-gravitational couplings. The framework is agnostic about whether these entries correspond to propagating particles or are structural residuals of the spectrum with no physical realization. If physical, they are candidates for sterile neutrino or warm dark matter searches in the eV-keV window.

Rank 16 ($R_5$, std) at 349 MeV sits in a normal mass range between the strange quark and charm quark. No SM fermion occupies this mass. If the entry is physical, it predicts an undiscovered state; if not, it joins the dead zone as a structural residual. Either outcome is informative.

Rank 3 Exclusion

Rank 3 ($R_1$, triv) predicts ~0.4 eV. Cosmological bounds constrain the sum of neutrino masses to $\Sigma m_\nu \lesssim 0.1$ eV, and solar neutrino data place a lower bound of ~10 meV on $\nu_2$ via the matter effect. A single eigenstate at 0.4 eV would saturate or exceed the cosmological sum by itself. Rank 3 is excluded as a SM neutrino. Whether it corresponds to a non-SM state with suppressed couplings (similar to the dead zone entries) or signals a structural correction needed in the $R_1$ triv vacuum remains OPEN.

The $\nu_2$ Gap

$\nu_2$ at 8.6 meV sits between rank 1 (0.2 meV) and rank 2 (66.7 meV), a ratio gap of ~7.75× from rank 2. Together with the charm quark (displaced by the $\tau$ resolution at rank 17), these are the two SM fermions with no $(\rho, \sigma)$ assignment. The experimentally allowed window for $\nu_2$ is roughly 10-50 meV, constrained from below by solar neutrino data and from above by the cosmological mass sum. The normal vs. inverted hierarchy remains experimentally undetermined; JUNO and DUNE are expected to resolve this. Whether vacuum mixing between $R_1$ entries or additional structure at the neutrino scale closes the gap remains OPEN.

The topology permits and Ψ settles. The formula composes and the masses land.