The Kaluza–Klein Spectrum on the Poincaré Homology Sphere

In 1904 Henri Poincaré constructed a closed three-manifold with the same homology as the three-sphere but a nontrivial fundamental group, refuting the conjecture that homology alone determines $S^3$. That manifold—$S^3/\mathrm{2I}$, the quotient of the three-sphere by the binary icosahedral group—has 120-fold symmetry inherited from the 600-cell, and its ring of invariant polynomials is generated by three forms $H$, $T$, $f$ of degrees $20$, $30$, and $12$ satisfying the single relation $T^2 = H^3 - 1728 f^5$: the $E_8$ surface equation. The projection from a 6D periodic lattice to 3D produces a quasicrystal in physical space and this same orbifold in the internal space. The constraint geometry and the spectral content of this orbifold are two outputs of that single projection. The constraint geometry provides the $\beta$-function. The spectral content provides the effective dimension $\mathcal{D}$ that enters the $\beta$-function’s dimensional correction. The full chain connecting these is developed in From Lattice Projection to Cosmic Expansion. This post computes the scalar Kaluza–Klein spectrum on $S^3/\mathrm{2I}$. The extension to $p$-form spectra (vectors, spinors, symmetric tensors) requires analogous branching rules and has not been carried out.

The Internal Space

The binary icosahedral group $\mathrm{2I} \cong \mathrm{SL}(2,5)$ is the double cover of the icosahedral rotation group $\mathrm{I} \cong A_5$. It has 120 elements—the largest finite subgroup of $\mathrm{SU}(2)$—and acts freely on $S^3 \simeq \mathrm{SU}(2)$ by left multiplication. The quotient $\Sigma = S^3/\mathrm{2I}$ is a smooth, compact, oriented three-manifold with $\pi_1(\Sigma) = \mathrm{2I}$ and $H_1(\Sigma;\mathbb{Z}) = 0$. Poincaré constructed it in 1904 as a counterexample to the conjecture that homology determines the three-sphere¹. It is the unique spherical space form with icosahedral holonomy.

In a Kaluza–Klein compactification on $\Sigma$, the scalar field equation reduces to the eigenvalue problem for the Laplace–Beltrami operator $\Delta_\Sigma$. The eigenvalues of $\Delta_{S^3}$ on the unit three-sphere are

with degeneracy $(k+1)^2$². Each eigenspace carries the representation $D_{k/2} \otimes D_{k/2}$ of $\mathrm{SU}(2)_L \times \mathrm{SU}(2)_R$ under the Peter–Weyl decomposition of $L^2(\mathrm{SU}(2))$, where $D_j$ denotes the spin-$j$ irreducible representation of dimension $2j+1$³⁴.

On $\Sigma = S^3/\mathrm{2I}$, scalar wavefunctions must be $\mathrm{2I}$-invariant under the left action. Since $-I \in \mathrm{2I}$ acts as $(-1)^k$ on $D_{k/2}$, only even values $k = 2l$ contribute (integer spin $j = l$). For each surviving spin $l$, the eigenvalue is

and the full orbifold multiplicity is

where the factor $(2l+1)$ comes from the unbroken $\mathrm{SU}(2)_R$ action on the opposite Peter–Weyl factor, and $n_l$ counts the $\mathrm{2I}$-invariant vectors in the spin-$l$ representation. The KK mass tower is therefore determined by the branching multiplicities $n_l$.

Branching Rules

The group $\mathrm{2I}$ has nine conjugacy classes. Each element $g \in \mathrm{2I} \subset \mathrm{SU}(2)$ has eigenvalues $e^{\pm i\alpha}$ in the fundamental representation, where $\alpha \in [0, \pi]$ is the half-angle parameter.

The total is $1 + 1 + 12 + 12 + 12 + 12 + 20 + 20 + 30 = 120$. The character of the spin-$l$ representation of $\mathrm{SU}(2)$ at half-angle $\alpha$ is

with $\chi_l(0) = 2l+1$ and $\chi_l(\pi) = (-1)^{2l}(2l+1)$. The multiplicity of the trivial representation of $\mathrm{2I}$ in the restriction of spin-$l$ is

where the sum runs over the nine conjugacy classes $C_i$ with half-angle parameters $\alpha_i$. This is an exact finite sum over nine terms.

The Spectrum

Evaluating $n_l$ for $l = 0$ through $l = 120$ partitions the spectrum into three regimes. Below the Coxeter number ($l < 30$), the tower is sparse. Of the 30 spin values $l = 0, \ldots, 29$, exactly 15 survive with $n_l = 1$, and 15 are forbidden with $n_l = 0$. The surviving spin values are

and the forbidden spin values are

In plain terms: below the Coxeter number $h = 30$, the spectrum is sparse and every surviving mode is non-degenerate—each admissible spin carries exactly one invariant harmonic, and the admissible spins are a fixed algebraic set that does not depend on where one chooses to truncate. At the threshold $h$, two things change simultaneously: every spin value becomes admissible, and the first doubly-degenerate invariant appears. The Coxeter number is the sharp boundary between gap-controlled and complete regimes.

At and above the Coxeter number ($30 \leq l \leq 59$), every spin value is present. The invariant count begins to increase, with $n_{30} = 2$ the first level carrying two independent $\mathrm{2I}$-invariants. Below $l = 30$, every surviving mode has $n_l = 1$; the transition at $l = 30$ marks both the onset of completeness and the first multiplicity enhancement.

In the asymptotic regime ($l \gg 30$), the invariant count satisfies $n_l \to (2l+1)/120$ on average, and the full multiplicity $m_l = (2l+1) \cdot n_l \to (2l+1)^2/120$, recovering the expected $1/|\mathrm{2I}|$ factor relative to $S^3$.

The first massive mode occurs at $l = 6$, with eigenvalue $\lambda_6 = 4 \cdot 6 \cdot 7 = 168$. The five consecutive forbidden levels $l = 1, \ldots, 5$ create the largest gap in the tower. The eigenvalue gap between the zero mode ($\lambda = 0$) and the first excitation ($\lambda = 168$) is a factor of 56 above the generic $S^3$ first excitation at $k = 1$ ($\lambda = 3$). This protection factor—56×—is the largest among all finite subgroup quotients of $S^3$.

The Numerical Semigroup

The forbidden spin values have algebraic structure. The generators $\{6, 10, 15\}$ are the degrees of the Klein invariant polynomials for $\mathrm{2I}$ acting on $\mathbb{C}^2$, divided by two. Klein showed that the ring of polynomial invariants $\mathbb{C}[x,y]^{\mathrm{2I}}$ is generated by three homogeneous polynomials⁵,

satisfying the single relation $T_{30}^2 = H_{20}^3 - 1728 \, f_{12}^5$. This relation defines the $E_8$ simple singularity in the ADE classification of du Val singularities⁶. The factor of two between the Klein degrees $\{12, 20, 30\}$ and the semigroup generators $\{6, 10, 15\}$ arises because the Molien series counts invariants at polynomial degree $k$, while the spin quantum number is $l = k/2$.

Define the numerical semigroup

Its gap set is

The semigroup has Frobenius number $F = 29$ (the largest gap) and genus $g = 15$ (the number of gaps). The identity $g = h/2$—the genus equals half the Coxeter number—is specific to the $E_8$ case.

Theorem (Semigroup selection rule). Let $\Gamma \subset \mathrm{SU}(2)$ be a finite subgroup and let $d_1, \ldots, d_r$ be the degrees of the homogeneous generators of the invariant ring $\mathbb{C}[x,y]^{\Gamma}$. Define the numerical semigroup $S = \langle d_1/2, \ldots, d_r/2 \rangle$. Then the spin-$l$ representation of $\mathrm{SU}(2)$ contains a $\Gamma$-invariant vector ($n_l \geq 1$) if and only if $l \in S$. Equivalently, the forbidden KK levels on $S^3/\Gamma$ are the gaps of $S$.

Proof. The Molien series for $\Gamma$ acting on $\mathrm{Sym}^{\bullet}(\mathbb{C}^2)$ is

where $P(t)$ encodes the relations among the generators⁵⁷. Under the identification $\mathrm{Sym}^{2l}(\mathbb{C}^2) \cong D_l$, the coefficient of $t^{2l}$ in $M(t)$ equals $n_l$. Writing $M(t)$ in the variable $u = t^2$ gives $\widetilde{M}(u) = \sum_{l \geq 0} n_l \, u^l$, whose denominator is $\prod_i (1 - u^{d_i/2})$. The set of exponents $l$ with $n_l \geq 1$ is the numerical semigroup generated by the denominator exponents $d_i/2$⁸, and the set with $n_l = 0$ is its gap set. For $\Gamma = \mathrm{2I}$, the Molien series specializes to

and $S = \langle 6, 10, 15 \rangle$ with gap set identical to the forbidden KK levels. $\square$

The theorem reduces the determination of forbidden KK levels to a purely algebraic problem: given $\Gamma$, look up the Klein invariant degrees and compute the gap set of the resulting semigroup. The character sum is required only for multiplicities $n_l > 1$ above the Coxeter threshold.

The generators themselves have a direct group-theoretic origin. The icosahedral rotation group $\mathrm{I} \cong A_5$ has rotation axes of orders $\{2, 3, 5\}$—edge midpoints, face centers, and vertices of the icosahedron. The semigroup generators $\{6, 10, 15\} = \{\operatorname{lcm}(2,3),\; \operatorname{lcm}(2,5),\; \operatorname{lcm}(3,5)\}$ are the pairwise least common multiples of these axis orders. In the character sum for $n_l$, each conjugacy class contributes with a periodicity in $l$ determined by its half-angle: the $C_4$ class ($\alpha = \pi/2$) has character period 2, the $C_3$/$C_6$ classes ($\alpha = 2\pi/3$, $\pi/3$) have period 3, and the four icosahedral classes ($\alpha = k\pi/5$, $k = 1,2,3,4$) have period 5. These periods match the rotation axis orders $\{2, 3, 5\}$ of the icosahedron. The first spin value at which all three periodicities simultaneously produce constructive interference is $\operatorname{lcm}(2,3) = 6$—the smallest generator. The remaining generators 10 and 15 mark the first pairwise constructive interferences of the $\{2, 5\}$ and $\{3, 5\}$ axis-order pairs.

The E8 Exponents

The $E_8$ root system has eight exponents $m_i$ defined by the eigenvalues of the Coxeter element: $e^{2\pi i m_i/h}$ where $h = 30$. The exponents are

All eight $E_8$ exponents appear among the 15 forbidden KK levels. The remaining seven forbidden levels are $\{2, 3, 4, 5, 8, 9, 14\}$, which complete the gap set of the semigroup. The $E_8$ exponents are the integers $m \in [1, h-1]$ coprime to $h = 30$⁸: $\gcd(m, 30) = 1$. Since $30 = 2 \times 3 \times 5$, the exponents avoid multiples of 2, 3, and 5—they are maximally incompatible with the generator structure $\{6, 10, 15\} = \{2 \times 3, \; 2 \times 5, \; 3 \times 5\}$ and the hardest gaps to fill by semigroup combinations.

The $E_8$ exponents have a dual role: they are both the eigenvalue phases of the Coxeter element in the root system and a distinguished subset of the forbidden KK harmonics on the Poincaré homology sphere. The McKay correspondence⁷—the bijection between finite subgroups of $\mathrm{SU}(2)$ and simply-laced Dynkin diagrams—provides the bridge. Under this correspondence, the representation graph of $\mathrm{2I}$ has the connectivity of the extended $\hat{E}_8$ Dynkin diagram. The KK spectrum inherits $E_8$ structure because the branching problem and the McKay graph encode the same algebraic data.

The Coxeter Threshold

The Coxeter number $h = 30$ marks a sharp transition in spectral structure. Below $l = 30$, the density of surviving modes increases stepwise:

The transition at $l = 30$ is also the first appearance of $n_l = 2$: the second independent $\mathrm{2I}$-invariant harmonic appears at the same threshold where the tower achieves completeness. Below the Coxeter number, every surviving mode has a single invariant ($n_l = 1$).

Spectral Dimension and Weyl Scaling

The Weyl counting function $N(\lambda) = \#\{\text{eigenvalues} \leq \lambda\}$, counted with multiplicity, satisfies

asymptotically on any Riemannian three-manifold $\Sigma$⁹. For $\Sigma = S^3/\mathrm{2I}$ with unit radius, $\mathrm{Vol}(\Sigma) = 2\pi^2/120$, giving a Weyl coefficient of $1/360$. At spin $l = 500$ ($\lambda = 1{,}002{,}000$), the Weyl prediction is $N_{\mathrm{Weyl}} \approx 2{,}786{,}115$, while the exact count gives $N_{\mathrm{exact}} = 2{,}794{,}378$—a ratio of 1.003, confirming the multiplicity formula and three-dimensional Weyl asymptotics to within 0.3%.

The cumulative eigenvalue count used for fitting excludes the $l = 0$ zero mode, which is the constant function and does not propagate. Restricting the fit to the sparse regime through the Coxeter number yields a different exponent:

The deficit $\Delta d_{\mathrm{eff}} \approx 0.38$ between Region I and the asymptotic regime arises from the semigroup sparsity: 50% of spin values are forbidden in $[0, 30)$, each surviving level contributes multiplicity $(2l+1) \times 1$, and the resulting growth of $N(\lambda)$ is sub-Weyl.

The running Weyl exponent—computed cumulatively from $l = 6$ up to successive cutoffs—shows the crossover:

The crossover from $\sim$2.6 to $\sim$3.0 occurs in a window around the Coxeter number. The Region I exponent $d_{\mathrm{eff}} = 2.613$ is within 0.2% of $\varphi^2 = (3+\sqrt{5})/2 \approx 2.618$. An exact identity underlies the regression estimate—the refined spectral quantity that equals $\varphi^2$ identically is the Chebyshev norm ratio at the two icosahedral half-angles,

where the even-index Chebyshev polynomials take values in the period-5 cycle $\{1, \varphi, 0, -\varphi, -1\}$ at $\varphi/2$ and $\{1, -1/\varphi, 0, 1/\varphi, -1\}$ at $1/(2\varphi)$. The squared norm of the first cycle is $2 + 2\varphi^2$; of the second, $2 + 2/\varphi^2$. Their ratio reduces to $(1+\varphi^2)/(1+1/\varphi^2) = \varphi^2$ by multiplying numerator and denominator by $\varphi^2$. The integer terms appear identically in both norms and cancel; what survives is the relative spectral power of the dominant ($\varphi$) and subdominant ($1/\varphi$) eigenvalues of the Penrose substitution matrix. The regression estimate $d_{\mathrm{eff}}(30) = 2.613$ is a finite-sample power-law fit to fifteen discrete points; the 0.2% undershoot is fitting residual, while the Chebyshev identity holds exactly over the full period that controls the gap structure. The triadic tension analysis carries the same identity as the factorization $I = 4\pi\varphi^2$, where the $\varphi^2$ factor is this Chebyshev norm ratio and the $4\pi$ is Gauss–Bonnet on $S^2$.

The Poincaré homology sphere is topologically three-dimensional everywhere, but its low-energy spectral content is thinner than a generic three-manifold. The semigroup gap structure creates a spectral bottleneck: modes are absent, and the cumulative eigenvalue count grows as if the manifold had effective dimension $\varphi^2$. Above $h$, all levels are present, degeneracies grow normally, and standard Weyl behavior recovers. The crossover at $l = h$ is a spectral signature of the $E_8$ algebraic structure imprinted on the geometry.

The appearance of $\varphi^2$ at the Coxeter threshold is controlled by three algebraic facts. First, Chebyshev periodicity: since $\varphi/2 = \cos(\pi/5)$, the Chebyshev polynomial $U_l(\varphi/2) = \sin((l{+}1)\pi/5)/\sin(\pi/5)$ takes values in $\{0, \pm 1, \pm\varphi\}$ with exact period 10. The evaluations at $\cos(4\pi/5) = -\varphi/2$ share the same value set, while those at $\cos(2\pi/5) = 1/(2\varphi)$ and $\cos(3\pi/5) = -1/(2\varphi)$ take values in $\{0, \pm 1, \pm 1/\varphi\}$. The individual periods are 10 for the $\pi/5$ and $3\pi/5$ classes and 5 for the $2\pi/5$ and $4\pi/5$ classes; the combined icosahedral contribution has period $\operatorname{lcm}(5, 10) = 10$. The irrational amplitudes $\varphi$ and $1/\varphi$ are exactly the eigenvalues of the Penrose substitution matrix $M = \bigl[\begin{smallmatrix}1&1\\1&0\end{smallmatrix}\bigr]$, and they enter the spectrum through the four icosahedral conjugacy classes alone—all other classes contribute integer character values. Second, the spectral dimension at the Coxeter threshold ($\varphi^2$) and the asymptotic value ($3$) are both spectral invariants of $M$: $\varphi^2 = \varphi + 1$ is the Perron–Frobenius eigenvalue of $M^2$, while $3 = \operatorname{Tr}(M^2) = \varphi^2 + 1/\varphi^2$. The two values are separated by exactly $1/\varphi^2$, and no free parameter interpolates between them. Third, the Mignosi–Restivo–Salemi periodicity theorem¹⁰ establishes $\varphi^2$ as the exact recurrence threshold separating aperiodic from ultimately periodic behaviour in combinatorics on words. The recurrence quotient of a word measures the maximum gap between consecutive occurrences of any factor relative to the factor length; the Fibonacci word—the canonical aperiodic sequence governed by the substitution $a \to ab$, $b \to a$—has recurrence quotient exactly $\varphi^2$, achieved as a supremum, and is not ultimately periodic. Any word whose recurrence quotient strictly exceeds $\varphi^2$ is ultimately periodic. The constant $\varphi^2$ is therefore the sharp boundary: at and below $\varphi^2$, aperiodic structure can persist; above it, global periodicity is forced. The semigroup gap structure below $h$ instantiates this transition—below the threshold, local periodicity of the character sum does not propagate to global completeness, and the spectrum remains sparse.

Branch Comparison

The semigroup structure generalizes across the ADE classification. For each binary polyhedral group $\Gamma \subset \mathrm{SU}(2)$, the Klein invariant ring has specific generator degrees, and the KK gap set on $S^3/\Gamma$ is controlled by the corresponding semigroup.

The Penrose branch is extremal in protection—56× versus 8× for the dodecagonal branch—but intermediate in Weyl deficit. The dodecagonal branch has a larger deficit (0.50 vs 0.39) despite having far fewer forbidden levels. The explanation is geometric: $\mathrm{BD}_6$ has only 24 elements, so $S^3/\mathrm{BD}_6$ has 5× the volume of $S^3/\mathrm{2I}$, and the 3 missing levels create a proportionally larger distortion in a shorter Coxeter window ($h = 14$ vs 30).

The cases $E_6$ and $D_8$ have the same group order, genus ($g = 3$), and Frobenius number ($F = 5$), yet their forbidden sets differ: $\{1, 2, 5\}$ for $E_6$ versus $\{1, 3, 5\}$ for $D_8$. These produce distinct protection factors (16× vs 8×) and distinct Region I effective dimensions (2.70 vs 2.50). The placement of the gap set within the spectral tower, not merely its cardinality or Frobenius number, governs the low-energy effective dimension.

For the binary dihedral family (McKay type $D_n$, $n \geq 4$), the semigroup generators reduce to $\langle 2, p \rangle$ with $p$ the unique odd generator. The gap set is $\{1, 3, 5, \ldots, p-2\}$—the first $g = (p-1)/2$ odd integers—so the first excitation occurs uniformly at $l = 2$ ($\lambda = 24$, protection factor 8×) for the entire family. As $n \to \infty$, $F/h \to 1/2$ and $g/h \to 1/4$, so $g = h/2$ is never achieved in the $D$-family.

For $E_8$, the genus equals $h/2$ exactly. The three generators $\{6, 10, 15\}$ are pairwise coprime ($\gcd(6,10) = 2$, $\gcd(6,15) = 3$, $\gcd(10,15) = 5$, while $\gcd(6,10,15) = 1$), producing a maximally sparse semigroup for the given generator sizes. The normalized Frobenius number $F/h = 29/30 = 0.967$—algebraic protection extends through nearly the entire Coxeter range. For $D_8$, $F/h = 5/14 = 0.357$. The Penrose branch is the unique ADE branch whose spectral protection saturates nearly the full window below $h$.

E8 as a Forced Selection

Every spectral invariant derived here descends from a single algebraic object: the relation $T^2 = H^3 - 1728\,f^5$ that the Klein invariants of $\mathrm{2I}$ satisfy on $\mathbb{C}^2$. The generator degrees $\{12, 20, 30\}$ fix the numerical semigroup $\langle 6, 10, 15\rangle$, whose gap set is the forbidden spectrum, whose Frobenius number saturates the Coxeter range at $F/h = 29/30$, whose genus equals $h/2$—a coincidence occurring nowhere else in the ADE classification—and whose protection factor of 56× exceeds every other finite subgroup quotient of $S^3$. The effective dimension $d_{\mathrm{eff}} = \varphi^2$ is not fitted but forced: it is the Chebyshev norm ratio at the two icosahedral half-angles, the Perron–Frobenius eigenvalue of $M^2$ for the Penrose substitution, and the sharp Mignosi–Restivo–Salemi boundary between aperiodic and ultimately periodic recurrence. The 6D embedding dimension and the $E_8$ algebraic structure answer different questions on different channels—6D is the minimum parent lattice for icosahedral projection (Kramer–Neri, 1984¹¹), $E_8$ is the compactification geometry inherited through the McKay correspondence—but both trace back to the same forced chain: negative selection yields $C_{10}$, $C_{10}$ on $S^3$ yields $\mathrm{2I}$, $\mathrm{2I}$ yields the Klein invariants, and the Klein invariants yield $E_8$. Every link is a theorem.