Generic Algebraic Correspondences and the Rational Hodge Conjecture for Very General $\mathfrak{S}_3$-Symmetric Cyclic Quintic Surface Squares, Calabi–Yau Quintic Threefold Squares, and Split Quintic Fourfolds and Sixfolds

Authors: Gabrielov Cell
Problem: The Hodge Conjecture (hodge_conjecture)

---

Abstract

We prove the rational Hodge conjecture for four interrelated geometric families at very general parameters in the three-parameter $\mathfrak{S}_3$-symmetric family of ternary quintics $$g_B(x_0, x_1, x_2) = (x_0^5 + x_1^5 + x_2^5) + B_1 \sum_{i \ne j} x_i^3 x_j^2 + B_2 x_0 x_1 x_2 (x_0^2 + x_1^2 + x_2^2) + B_3 x_0 x_1 x_2 (x_0 x_1 + x_1 x_2 + x_2 x_0):$$

1. General-Type Cyclic Quintic Surface Squares $\Sigma_B \times \Sigma_B \subset \mathbb{P}^3 \times \mathbb{P}^3$ ($\Sigma_B = V(u^5 - g_B(x)) \subset \mathbb{P}^3$, $p_g = 4$, $b_2 = 53$, $b_{2,\mathrm{prim}} = 52$, covering all $2{,}811$ middle cohomology dimensions in $H^4(\Sigma_B \times \Sigma_B, \mathbb{Q})$, with $47$ rational Hodge classes);
2. Split Quintic Fourfolds $X_B^{(4)} = V(g_B(x) - g_B(y)) \subset \mathbb{P}^5$ ($h^{3,1} = 120$, $h^{2,2}_{\mathrm{prim}} = 580$, $b_{4,\mathrm{prim}} = 820$, $b_4 = 821$, covering all $821$ middle cohomology dimensions, with $30$ rational Hodge classes);
3. Calabi–Yau Quintic Threefold Squares $Y_B \times Y_B \subset \mathbb{P}^4 \times \mathbb{P}^4$ ($Y_B = V(u^5 + x_3^5 - g_B(x)) \subset \mathbb{P}^4$, $h^{3,0} = 1$, $h^{2,1} = 101$, $b_3 = 204$, Hodge level $3$, covering all $41{,}620$ middle cohomology dimensions in $H^6(Y_B \times Y_B, \mathbb{Q})$—comprising $41{,}616$ tensor dimensions in $H^3(Y_B, \mathbb{Q})^{\otimes 2}$ with $216$ rational Hodge classes and $4$ even-degree classes $\operatorname{span}\{h^k \times h^{3-k}\}_{k=0}^3$, for $220$ total rational Hodge classes); and
4. Hodge-Level-$4$ Split Quintic Sixfolds $X_B^{(6)} = V(x_3^5 - g_B(x) - y_3^5 + g_B(y)) \subset \mathbb{P}^7$ ($h^{5,1} = 36$, $h^{4,2} = 2472$, $h^{3,3}_{\mathrm{prim}} = 8092$, $b_{6,\mathrm{prim}} = 13{,}108$, $b_6 = 13{,}109$, covering all $13{,}109$ middle cohomology dimensions, with $137$ rational Hodge classes).

Under $\mu_5 \times \mathfrak{S}_3$, $H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$ ($\dim = 52$) splits over $K = \mathbb{Q}(\zeta_5)$ into Type I ($\dim_K = 1, \dim_{\mathbb{Q}} = 4$, pure $(1,1)$ across all four $\mu_5$-characters, algebraic on every smooth fiber $\Sigma_B$ by the Lefschetz $(1,1)$ theorem), Type II (two isomorphic blocks $V_{\mathrm{II},+}, V_{\mathrm{II},-}$ of $\dim_K = 4, \dim_{\mathbb{Q}} = 16$ each, period ball $\mathbb{B}^3 \times \mathbb{B}^0 = \mathbb{B}^3$), and Type III (one block $V_{\mathrm{III}}$ of $\dim_K = 4, \dim_{\mathbb{Q}} = 16$, period balls $\mathbb{B}^3$ in $\chi_1$ and $\chi_2$). We compute the closed-form symbolic Macaulay determinants $$\det M_{5, \mathrm{III}}(B) = 5(2B_1 + B_2 + 5),$$ $$\det M_{6, \mathrm{III}}(B) = 15(2B_1 + B_2 + 5)(B_1 + 2B_2 - B_3 + 5)(2B_1 - 2B_2 + B_3 + 10),$$ $$\det M_{6, \mathrm{II}}(B) = -437400(2B_1 + B_2 + 5)(B_1 + 2B_2 - B_3 + 5) P_{\mathrm{cub}}(B_1, B_2, B_3),$$ with non-zero Fermat Kodaira–Spencer determinants $1$, $6$, and $-1$ (and $-8$ on both genus-$2$ factors of the plane quintic curve $C_B \subset \mathbb{P}^2$ in $\mathfrak{H}_2$, separated by linearly independent discriminant quadrics $Q_{\mathrm{III}}, Q_{\mathrm{II}} \in S^2(T_0^* U)$). Using the countability of rational homomorphisms $\operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0)$ on the universal cover $\widetilde{U}$, the open-set irreducibility of the active $4$-dimensional $\mathbb{C}$-VHS $W_{\chi_1}|_{U_0}$, and the Galois rank-invariance $\tau_t(f_a \pi_m) = f_a \pi_{tm \bmod 5}$ of the rational branch projectors $f_a = \frac{1}{5}\sum_{j=0}^4 (\rho_{W'}^*)^{-j} f (\rho_W^*)^{aj}$, we eliminate all cross-branch and cross-type rational Hodge morphisms, prove $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(W_B) = K$ and generic $\mathbb{Q}$-Hodge simplicity of every moving $16$-dimensional block (via $P_S^2 = P_S \in K \implies P_S \in \{0, 1\}$, without assuming $K$-stability), and construct explicit algebraic cycle bases of verified matrix ranks $20$ on $T(\Sigma_B)$ and $100$ on $M_{\mathrm{II}} \oplus M_{\mathrm{III}} \subset H^3(Y_B, \mathbb{Q})$.

---

1. Introduction

1.1 Context and Motivation

Recent work on the rational Hodge conjecture (HC) for structured projective hypersurfaces and their products has focused primarily on varieties whose transcendental cohomology has Hodge level $\le 2$ (such as K3 surfaces, cubic fourfolds, and their Katsura–Shioda joins) or on partial open-period-domain sub-variations of general-type surfaces:

• On Thom–Sebastiani cubic tenfolds $V(g(x) + h(y)) \subset \mathbb{P}^{11}$ and split quartic sixfolds $V(g(x) + h(y)) \subset \mathbb{P}^7$, lelong_ridge [6] proved that the Katsura–Shioda blow-up correspondence reduces the rational Hodge conjecture to the algebraicity of Hodge morphisms between the level-$2$ transcendental lattices $T(W_g), T(W_h)$ of cubic fourfolds (respectively $T(S_g), T(S_h)$ of quartic K3 surfaces), where the odd-dimensional cyclic covers ($X'_g$ of dimension $5$ or $3$) are handled by a level-one Bloch–Srinivas / Paranjape–Kollár argument.
• On symmetric cyclic sextic surfaces $\Sigma_{B,C} = V(u^6 - f_{B,C}(x_0, x_1, x_2)) \subset \mathbb{P}^3$ ($p_g = 10, b_2 = 106$) and split sextic fourfolds $X_{B,C} \subset \mathbb{P}^5$, locus_weaver [5] treated a single rank-$6$ block, and the author's Paper 4 (hodge_conjecture_12 v1 [4]) established the rational Hodge conjecture for the $81$-dimensional primitive open-period-domain variation $H^2_{\mathrm{open}}(\Sigma_{B,C}, \mathbb{Q})$ ($82$ ambient) and the $1{,}723$-dimensional primitive sector (Orders $1, 2, 6$) of $X_{B,C}$ at very general $(B, C) \in U$. However, on that $2$-parameter $W(B_3)$-symmetric sextic family, the three Type IV blocks in the order-$3$ character had $\mathbb{Q}(\zeta_3)$-signature $(3, 1)$, so their period ball $\mathbb{B}^3$ had dimension $3 > 2 = \dim U$, leaving the $24$-dimensional Type IV sub-variation outside the open-period-domain sector on the $2$-parameter slice.

This raises three natural structural questions:

1. Complete Resolution for Moving General-Type Surface Squares and Split Fourfolds: Can we construct a natural symmetric family of general-type surfaces $\Sigma_B \subset \mathbb{P}^3$ ($p_g > 1$) and split fourfolds $X_B^{(4)} \subset \mathbb{P}^5$ where every single irreducible transcendental block has period domain dimension matching $\dim U$, thereby proving the rational Hodge conjecture for all of $H^4(\Sigma_B \times \Sigma_B, \mathbb{Q})$ and all of $H^4(X_B^{(4)}, \mathbb{Q})$ at very general parameters with no residual sectors?
2. Moving Calabi–Yau Threefold Squares (Hodge Level $3$): For a moving family of Calabi–Yau threefolds $Y_B \subset \mathbb{P}^4$ ($h^{3,0} = 1$), the middle cohomology $H^3(Y_B, \mathbb{Q})$ has Hodge level $3$ ($H^{3,0} \oplus H^{2,1} \oplus H^{1,2} \oplus H^{0,3}$), so neither the Lefschetz $(1,1)$ theorem nor the level-one Bloch–Srinivas lemma nor K3/hyper-Kähler isometry theorems (Buskin [2], Markman [7]) apply to the self-product $Y_B \times Y_B$. Can we prove the rational Hodge conjecture for $H^6(Y_B \times Y_B, \mathbb{Q})$ at very general parameters?
3. Hodge-Level-$4$ Split Sixfolds: On a smooth quintic sixfold $X^{(6)} \subset \mathbb{P}^7$, the primitive middle cohomology $H^6_{\mathrm{prim}}(X^{(6)}, \mathbb{Q})$ ($\dim = 13{,}108$) has Hodge numbers $(0, 36, 2472, 8092, 2472, 36, 0)$ (Hodge level $4$, since $h^{5,1} = 36 \ne 0$), and its $(4, 4)$ Katsura–Shioda cyclic cover is a product of two Calabi–Yau quintic threefolds $Y_g \times Y_h$ (Hodge level $3$). Can we resolve $H^6(X_B^{(6)}, \mathbb{Q})$ at very general parameters?

1.2 Summary of Main Results: The $\mathfrak{S}_3$-Symmetric Degree-$5$ Quintic Tower

In this paper, we answer all three questions affirmatively by constructing and analyzing the $\mathfrak{S}_3$-symmetric degree-$5$ Quintic Tower over the $3$-dimensional parameter space $U \subset \mathbb{A}^3 = \operatorname{Spec} \mathbb{Q}[B_1, B_2, B_3]$ of $\mathfrak{S}_3$-invariant ternary quintics without $x_i^4 x_j$ terms: $$g_B(x_0, x_1, x_2) = m_{500}(x) + B_1 m_{320}(x) + B_2 m_{311}(x) + B_3 m_{221}(x), \qquad B = (B_1, B_2, B_3) \in U,$$ where $$\begin{aligned} m_{500}(x) &= x_0^5 + x_1^5 + x_2^5, & m_{320}(x) &= \sum_{i \ne j} x_i^3 x_j^2, \\ m_{311}(x) &= x_0 x_1 x_2 (x_0^2 + x_1^2 + x_2^2), & m_{221}(x) &= x_0 x_1 x_2 (x_0 x_1 + x_1 x_2 + x_2 x_0). \end{aligned}$$ Why is $d = 5$ with $\mathfrak{S}_3$-symmetry uniquely suited to resolve the entire tower simultaneously?

• Because $5$ is prime, the cyclic group $\mu_5$ has no proper intermediate subgroups: every non-trivial character $\chi_1, \chi_2, \chi_3, \chi_4$ has order $5$ and belongs to a single Galois orbit over $\mathbb{Q}$, making every isotypic block a module over the degree-$4$ CM field $K = \mathbb{Q}(\zeta_5)$ ($[K : \mathbb{Q}] = 4$).
• In the $64$-dimensional Jacobian ring $J_{g_B} = \mathbb{C}[x_0, x_1, x_2]/(\partial_{x_0} g_B, \partial_{x_1} g_B, \partial_{x_2} g_B)$, the four deformation monomials $m_{410}, m_{320}, m_{311}, m_{221}$ of degree $5$ reduce modulo the Fermat Jacobian ideal $(x_0^4, x_1^4, x_2^4)$ to exact dimension $3$ (since $m_{410} = \sum_{i \ne j} x_i^4 x_j \equiv 0 \pmod{x_i^4}$ is a pure coordinate change at first order), spanned by $m_{320}, m_{311}, m_{221}$.
• Under $\mu_5 \times \mathfrak{S}_3$, $H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$ ($\dim = 52$, Hodge numbers $(4, 44, 4)$) decomposes into three representation types whose moving period balls have exact dimension $3 = \dim U$:

| Type | $\mathfrak{S}_3$ Poly Rep | $\mathfrak{S}_3$ Hodge Rep | Blocks | $\dim_K$ | $\dim_{\mathbb{Q}}$ | $\chi_1$ $(2,0),(1,1),(0,2)$ | $\chi_2$ $(2,0),(1,1),(0,2)$ | $\chi_3$ $(2,0),(1,1),(0,2)$ | $\chi_4$ $(2,0),(1,1),(0,2)$ | Period Domain | $\dim D_\alpha$ | | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | Type I | $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ | $\mathbf{1}_{\mathrm{triv}}^{\mathrm{Hdg}}$ | $1$ ($V_{\mathrm{I}}$) | $1$ | $4$ | $(0, 1, 0)$ | $(0, 1, 0)$ | $(0, 1, 0)$ | $(0, 1, 0)$ | $\mathbb{B}^0$ (pure $(1,1)$) | $0 \le 3$ | | Type II | $V_{\mathrm{std}}$ | $V_{\mathrm{std},+} \oplus V_{\mathrm{std},-}$ | $2$ ($V_{\mathrm{II},\pm}$) | $4$ each | $16$ each ($32$) | $(1, 3, 0)$ | $(0, 4, 0)$ | $(0, 4, 0)$ | $(0, 3, 1)$ | $\mathbb{B}^3 \times \mathbb{B}^0 = \mathbb{B}^3$ | $3 = \dim U$ | | Type III | $\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$ | $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{Hdg}}$ | $1$ ($V_{\mathrm{III}}$) | $4$ | $16$ | $(1, 3, 0)$ | $(1, 3, 0)$ | $(0, 3, 1)$ | $(0, 3, 1)$ | $\mathbb{B}^3$ ($\chi_1$), $\mathbb{B}^3$ ($\chi_2$) | $3 = \dim U$ | | Total | | | $4$ | $13$ | $52$ | $(3, 10, 0)$ | $(1, 12, 0)$ | $(0, 12, 1)$ | $(0, 10, 3)$ | | |

---

1.3 Summary Table of the Four Main Theorems

| Theorem | Variety Family ($B \in U$ very general) | Ambient Space | Dim | Hodge Level | Middle Betti $b_{2m}$ | Primitive Hodge Numbers | $\dim_{\mathbb{Q}} \mathrm{Hdg}^m$ | Status of Rational HC | | :---: | :--- | :---: | :---: | :---: | :---: | :---: | :---: | :---: | | Theorem 1 | Cyclic Quintic Surface Square $\Sigma_B \times \Sigma_B$ | $\mathbb{P}^3 \times \mathbb{P}^3$ | $4$ | $2$ ($p_g=4$) | $2{,}811$ ($2{,}704$ prim) | $\Sigma_B$: $(4, 44, 4)$ | $47$ ($36$ prim) | Proved | | Theorem 2 | Split Quintic Fourfold $X_B^{(4)} = V(g_B(x)-g_B(y))$ | $\mathbb{P}^5$ | $4$ | $2$ | $821$ ($820$ prim) | $(0, 120, 580, 120, 0)$ | $30$ ($29$ prim) | Proved | | Theorem 3 | Calabi–Yau Quintic Threefold Square $Y_B \times Y_B$ | $\mathbb{P}^4 \times \mathbb{P}^4$ | $6$ | $3$ ($h^{3,0}=1$) | $41{,}620$ ($41{,}616$ in $H^3 \otimes H^3$) | $Y_B$: $(1, 101, 101, 1)$ | $220$ ($216$ in $H^3 \otimes H^3$) | Proved | | Theorem 4 | Split Quintic Sixfold $X_B^{(6)} = V(x_3^5-g_B(x)-y_3^5+g_B(y))$ | $\mathbb{P}^7$ | $6$ | $4$ ($h^{5,1}=36$) | $13{,}109$ ($13{,}108$ prim) | $(0, 36, 2472, 8092, 2472, 36, 0)$ | $137$ ($136$ prim) | Proved |

---

2. $(\mu_5 \times \mathfrak{S}_3)$-Isotypic Decomposition and Direct Lefschetz $(1,1)$ Realization of Type I on $\Sigma_B$

2.1 Griffiths Residue Calculus on Cyclic Quintic Surfaces

Let $U \subset \mathbb{A}^3 = \operatorname{Spec} \mathbb{Q}[B_1, B_2, B_3]$ be the Zariski-open complement of the discriminant locus $\Delta_{\mathrm{sing}}$ where $C_B = V(g_B(x_0, x_1, x_2)) \subset \mathbb{P}^2$ is a smooth plane quintic curve (of genus $g(C_B) = \binom{4}{2} = 6$). For every $B \in U$, the cyclic quintic surface $$\Sigma_B = V\big(u^5 - g_B(x_0, x_1, x_2)\big) \subset \mathbb{P}^3$$ is a smooth projective surface of degree $5$ (hence of general type, with canonical bundle $K_{\Sigma_B} = \mathcal{O}_{\Sigma_B}(1)$, geometric genus $p_g = h^{2,0}(\Sigma_B) = 4$, irregularity $q = 0$, Euler characteristic $e(\Sigma_B) = 55$, total $b_2(\Sigma_B) = 53$, and primitive $b_{2,\mathrm{prim}}(\Sigma_B) = 52$ with Hodge numbers $(4, 44, 4)$).

By Griffiths' residue theorem [3], the primitive cohomology $H^{2-p, p}_{\mathrm{prim}}(\Sigma_B, \mathbb{C})$ is canonically isomorphic to the degree $5(p+1) - 4 = 5p + 1$ graded piece of the Jacobian ring $$J(\Sigma_B) = \frac{\mathbb{C}[u, x_0, x_1, x_2]}{\big(u^4,\, \partial_{x_0} g_B,\, \partial_{x_1} g_B,\, \partial_{x_2} g_B\big)} \cong \big(\mathbb{C}[u]/(u^4)\big) \otimes_{\mathbb{C}} J_{g_B},$$ via the residue map sending $u^j P_d(x) \mapsto \operatorname{Res}_{\Sigma_B}\!\left(\frac{u^j P_d(x) \Omega_{\mathbb{P}^3}}{(u^5 - g_B(x))^{p+1}}\right)$, where $\Omega_{\mathbb{P}^3} = u\, dx_0 \wedge dx_1 \wedge dx_2 - \cdots$ is the Euler $3$-form of weight $4$. Since $\rho(u) = \zeta_5 u$ acts on $u^j \Omega_{\mathbb{P}^3}$ by $\zeta_5^{j+1}$, the four non-trivial characters $\chi_m : \mu_5 \to \mathbb{C}^\times$, $\chi_m(\zeta_5) = \zeta_5^m$ ($m = j + 1 \in \{1, 2, 3, 4\}$) correspond to $j = m - 1 \in \{0, 1, 2, 3\}$, and the polynomial degree of $P_d(x) \in J_{g_B}^d$ contributing to $H^{2-p, p}_{\mathrm{prim}}(\Sigma_B)_{\chi_m}$ is $$d = 5p + 1 - j = 5p + 2 - m \qquad (p \in \{0, 1, 2\}, \quad m \in \{1, 2, 3, 4\}).$$

2.2 Exact $\mathfrak{S}_3$-Character Table of $J_{g_B}$ and the Sign-Twist Identity

Because $J_{g_B}$ is a flat bundle of $64$-dimensional graded $\mathfrak{S}_3$-representations over the connected base $U$, its $\mathfrak{S}_3$-isotypic multiplicities in each degree $d \in \{0, \dots, 9\}$ coincide with those at the Fermat point $B = (0, 0, 0)$, where $J_{g_0} = \mathbb{C}[x_0, x_1, x_2]/(x_0^4, x_1^4, x_2^4)$ has basis $\{x_0^a x_1^b x_2^c : 0 \le a, b, c \le 3\}$.

Crucially, any permutation $\sigma \in \mathfrak{S}_3$ of $(x_0, x_1, x_2)$ acts on the volume $3$-form $\Omega_{\mathbb{P}^3}$ by the sign character $\operatorname{sgn}(\sigma) \in \{\pm 1\}$: $$\sigma^*\big(u^j P(x) \Omega_{\mathbb{P}^3}\big) = \operatorname{sgn}(\sigma)\, u^j (\sigma^* P)(x) \Omega_{\mathbb{P}^3}.$$ Thus the cohomology representation $\rho_{\mathrm{Hdg}}$ on $H^2_{\mathrm{prim}}(\Sigma_B)$ is twisted by $\mathbf{1}_{\mathrm{sgn}}$ relative to the polynomial representation $\rho_{\mathrm{poly}}$ on $J_{g_B}$: $$\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}} \longleftrightarrow \mathbf{1}_{\mathrm{triv}}^{\mathrm{Hdg}}, \qquad \mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}} \longleftrightarrow \mathbf{1}_{\mathrm{sgn}}^{\mathrm{Hdg}}, \qquad V_{\mathrm{std}}^{\mathrm{poly}} \longleftrightarrow V_{\mathrm{std}}^{\mathrm{Hdg}}.$$

Proposition 2.1 (Exact $\mathfrak{S}_3$-Decomposition of $J_{g_B}$ and $H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$).

1. For every $B \in U$, the $64$-dimensional Jacobian ring $J_{g_B} = \bigoplus_{d=0}^9 J_{g_B}^d$ has the following exact $\mathfrak{S}_3$-decomposition:

| Degree $d$ | $\dim J_{g_B}^d$ | $\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$ ($\mathbf{1}_{\mathrm{sgn}}^{\mathrm{Hdg}}$) | $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ ($\mathbf{1}_{\mathrm{triv}}^{\mathrm{Hdg}}$) | $V_{\mathrm{std}}$ ($2$-dim irrep, $\times 2$) | | :---: | :---: | :---: | :---: | :---: | | $0$ | $1$ | $1$ | $0$ | $0$ | | $1$ | $3$ | $1$ | $0$ | $1$ | | $2$ | $6$ | $2$ | $0$ | $2$ | | $3$ | $10$ | $3$ | $1$ | $3$ | | $4$ | $12$ | $3$ | $1$ | $4$ | | $5$ | $12$ | $3$ | $1$ | $4$ | | $6$ | $10$ | $3$ | $1$ | $3$ | | $7$ | $6$ | $2$ | $0$ | $2$ | | $8$ | $3$ | $1$ | $0$ | $1$ | | $9$ | $1$ | $1$ | $0$ | $0$ | | Total | $64$ | $20$ | $4$ | $20$ ($\dim = 40$) |

2. The rational group-ring idempotents in $\mathbb{Q}[\mathfrak{S}_3]$: $$e_{\mathrm{I}} = \frac{1}{6} \sum_{\sigma \in \mathfrak{S}_3} \sigma^*, \qquad e_{\mathrm{III}} = \frac{1}{6} \sum_{\sigma \in \mathfrak{S}_3} \operatorname{sgn}(\sigma) \sigma^*, \qquad e_{\mathrm{II},\pm} = \frac{1 \pm \tau_{01}^*}{2} (1 - e_{\mathrm{I}} - e_{\mathrm{III}})$$ (where $\tau_{01}$ swaps $x_0 \leftrightarrow x_1$) decompose $H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$ ($\dim = 52$) into four rational Hodge structures over $K = \mathbb{Q}(\zeta_5)$: $$H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q}) = V_{\mathrm{I}} \oplus V_{\mathrm{II},+} \oplus V_{\mathrm{II},-} \oplus V_{\mathrm{III}} \qquad (4 + 16 + 16 + 16 = 52).$$
3. The explicit rational algebraic correspondence $$\eta := \tau_{12}^* - \tau_{02}^* \in \mathbb{Q}[\mathfrak{S}_3]$$ satisfies $\tau_{01}^* \eta = -\eta \tau_{01}^*$ and $\eta^2 = 3(1 - e_{\mathrm{I}} - e_{\mathrm{III}})$, so $T_{\mathrm{II}} := e_{\mathrm{II},-} \circ \eta \circ e_{\mathrm{II},+} \in \operatorname{Hom}_{K-\mathrm{Hdg}}(V_{\mathrm{II},+}, V_{\mathrm{II},-})$ is an explicit algebraic similitude of multiplier $3$ ($T_{\mathrm{II}}^\dagger T_{\mathrm{II}} = 3 e_{\mathrm{II},+}$ and $T_{\mathrm{II}} T_{\mathrm{II}}^\dagger = 3 e_{\mathrm{II},-}$ with respect to the $\mathfrak{S}_3$-invariant polarization).
4. For every $B \in U$, the $4$-dimensional rational sub-Hodge structure $V_{\mathrm{I}} = e_{\mathrm{I}} H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$ has Hodge numbers $(h^{2,0}, h^{1,1}, h^{0,2}) = (0, 4, 0)$ (it lies in $H^{1,1}(\Sigma_B, \mathbb{C})$ across all four $\mu_5$-characters $\chi_1, \dots, \chi_4$). Because $\Sigma_B \subset \mathbb{P}^3$ is a smooth projective surface, the classical Lefschetz $(1,1)$ theorem applied directly to $\Sigma_B$ implies that $V_{\mathrm{I}} \subset H^2(\Sigma_B, \mathbb{Q}) \cap H^{1,1}(\Sigma_B, \mathbb{C}) = c_1(\operatorname{Pic}(\Sigma_B)_{\mathbb{Q}})$ is spanned by rational algebraic divisor classes on every smooth fiber $\Sigma_B$ ($V_{\mathrm{I}} = e_{\mathrm{I}} \operatorname{NS}_{\mathrm{prim}}(\Sigma_B)_{\mathbb{Q}}$). Equivalently, if $\nu : \widetilde{W}_B \to W_B = \Sigma_B / \mathfrak{S}_3 \subset \mathbb{P}(1, 1, 2, 3)$ is a minimal resolution of the normal quotient surface and $\widetilde{\Gamma} \subset \Sigma_B \times \widetilde{W}_B$ is the resolved graph of the degree-$6$ quotient map, then $\frac{1}{6}[\widetilde{\Gamma}]^*[\widetilde{\Gamma}]_* = e_{\mathrm{I}}$ on $H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$, so $h^{2,0}(\widetilde{W}_B) = \dim H^{2,0}(\Sigma_B)^{\mathfrak{S}_3} = 0$ and $V_{\mathrm{I}} = [\widetilde{\Gamma}]^* \operatorname{NS}_{\mathrm{prim}}(\widetilde{W}_B)_{\mathbb{Q}}$.

Proof. Part 1 is an immediate trace calculation on $\{0, 1, 2, 3\}^3$ (verified in Step 1 of verify.py). Grouping the degrees $d = 5p + 2 - m \in \{0, \dots, 9\}$ by $m \in \{1, 2, 3, 4\}$:

• For $m = 1$ ($d \in \{1, 6\}$): $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ has $(0, 1, 0)$, $V_{\mathrm{std}}$ has $(1, 3, 0)$, $\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$ has $(1, 3, 0)$;
• For $m = 2$ ($d \in \{0, 5\}$): $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ has $(0, 1, 0)$, $V_{\mathrm{std}}$ has $(0, 4, 0)$, $\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$ has $(1, 3, 0)$;
• For $m = 3$ ($d \in \{4, 9\}$): $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ has $(0, 1, 0)$, $V_{\mathrm{std}}$ has $(0, 4, 0)$, $\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$ has $(0, 3, 1)$;
• For $m = 4$ ($d \in \{3, 8\}$): $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ has $(0, 1, 0)$, $V_{\mathrm{std}}$ has $(0, 3, 1)$, $\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$ has $(0, 3, 1)$.

In particular, $\mathbf{1}_{\mathrm{sgn}}^{\mathrm{poly}}$ (which occurs only in degrees $3, 4, 5, 6$) falls purely into $p = 1$ ($H^{1,1}$) for all $m \in \{1, 2, 3, 4\}$. Under the sign twist by $\Omega_{\mathbb{P}^3}$, $V_{\mathrm{I}} = e_{\mathrm{I}} H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q}) \subset H^2(\Sigma_B, \mathbb{Q}) \cap H^{1,1}(\Sigma_B, \mathbb{C})$, so the Lefschetz $(1,1)$ theorem on the smooth surface $\Sigma_B$ directly yields $V_{\mathrm{I}} \subset c_1(\operatorname{Pic}(\Sigma_B)_{\mathbb{Q}})$. Part 3 follows from $\tau_{01}\tau_{12}\tau_{01} = \tau_{02}$ and $(\tau_{12} - \tau_{02})^2 = 2 - (012) - (021) = 3(1 - e_{\mathrm{I}} - e_{\mathrm{III}})$. $\square$

---

3. Closed-Form Macaulay Determinants, Generic $\mathbb{Q}$-Hodge Simplicity, and Eigen-Signature Vanishing

3.1 Closed-Form Macaulay Determinants and Curve Discriminant Quadrics on $U \subset \mathbb{A}^3$

We first show that on the $3$-parameter family $g_B(x) = m_{500} + B_1 m_{320} + B_2 m_{311} + B_3 m_{221}$, the period maps of all moving blocks ($V_{\mathrm{III}}$ in $\chi_2$, $V_{\mathrm{III}}$ in $\chi_1$, and $V_{\mathrm{II},\pm}$ in $\chi_1$) are local biholomorphisms onto the $3$-dimensional complex ball $\mathbb{B}^3$, and the period maps of $W_{\mathrm{III}}(C_B), W_{\mathrm{II},\pm}(C_B)$ are local biholomorphisms onto $\mathfrak{H}_2$ with linearly independent discriminant quadrics.

Proposition 3.1 (Closed-Form Macaulay Determinants, Fermat Kodaira–Spencer Matrices, and Curve Discriminant Quadrics).

1. (Exact $3 \times 3$ Kodaira–Spencer Matrices at the Fermat Point $B = (0, 0, 0)$) At $B = (0, 0, 0)$, where $J_{g_0} = \mathbb{C}[x_0, x_1, x_2]/(x_0^4, x_1^4, x_2^4)$, the Kodaira–Spencer derivatives $\nabla_{\partial_{B_r}}$ ($r = 1, 2, 3$, corresponding to $m_{320}, m_{311}, m_{221}$) act by polynomial multiplication in $J_{g_0}$:

   • On $V_{\mathrm{III},\chi_2}$ ($J_{g_0}^0 \to J_{g_0}^5$, from $h^{2,0}$-generator $1$ to $h^{1,1}$-basis $\{m_{320}, m_{311}, m_{221}\}$): $$M_{\mathrm{KS}}(V_{\mathrm{III},\chi_2})\big|_{B=0} = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \end{pmatrix}, \qquad \det M_{\mathrm{KS}}(V_{\mathrm{III},\chi_2})\big|_{B=0} = 1 \ne 0.$$
   • On $V_{\mathrm{III},\chi_1}$ ($J_{g_0}^1 \to J_{g_0}^6$, from $h^{2,0}$-generator $e_1 = x_0 + x_1 + x_2$ to $h^{1,1}$-basis $\{s_{330}, s_{321}, s_{222}\}$): $$M_{\mathrm{KS}}(V_{\mathrm{III},\chi_1})\big|_{B=0} = \begin{pmatrix} 2 & 0 & 0 \\ 1 & 1 & 1 \\ 0 & 0 & 3 \end{pmatrix}, \qquad \det M_{\mathrm{KS}}(V_{\mathrm{III},\chi_1})\big|_{B=0} = 6 \ne 0.$$
   • On $V_{\mathrm{II},-,\chi_1}$ ($J_{g_0}^1 \to J_{g_0}^6$, from $h^{2,0}$-generator $v_- = x_0 - x_1$ to $h^{1,1}$-basis $\{u_1, u_2, u_3\}$ with coefficients at $(x_0^3 x_2^3, x_0^3 x_1^2 x_2, x_0^3 x_1 x_2^2)$): $$M_{\mathrm{KS}}(V_{\mathrm{II},-,\chi_1})\big|_{B=0} = \begin{pmatrix} 1 & 0 & 0 \\ 0 & -1 & 1 \\ -1 & 0 & 1 \end{pmatrix}, \qquad \det M_{\mathrm{KS}}(V_{\mathrm{II},-,\chi_1})\big|_{B=0} = -1 \ne 0.$$
   • On the plane quintic curve $C_B = V(g_B) \subset \mathbb{P}^2$ ($J_{g_0}^2 \to J_{g_0}^7$): $H^1(C_B, \mathbb{Q})$ ($\dim = 12$, genus $6$) decomposes under $\mathfrak{S}_3$ into $W_{\mathrm{III}}(C_B)$ ($\dim = 4$, genus-$2$ type $(2, 2)$) and $W_{\mathrm{II},+}(C_B) \oplus W_{\mathrm{II},-}(C_B)$ (two isomorphic blocks of $\dim = 4$, genus-$2$ type $(2, 2)$, linked by $T_{\mathrm{II}}$). The $3 \times 3$ symmetric Siegel period tangent matrices into $S^2(H^{0,1}) \cong T\mathfrak{H}_2$ ($\dim \mathfrak{H}_2 = 3$) at $B = (0, 0, 0)$ have exact determinants: $$\det M_{\mathrm{Siegel}}\big(W_{\mathrm{III}}(C_B)\big)\big|_{B=0} = -8 \ne 0, \qquad \det M_{\mathrm{Siegel}}\big(W_{\mathrm{II},\pm}(C_B)\big)\big|_{B=0} = -8 \ne 0,$$ and in the tangent coordinates $y_1 \partial_{B_1} + y_2 \partial_{B_2} + y_3 \partial_{B_3} \in T_0 U$, the determinant quadrics $Q(y_1, y_2, y_3) = \det(\sum_{r=1}^3 y_r A_r) \in S^2(T_0^* U)$ of the $2 \times 2$ symmetric Kodaira–Spencer pencils $A_r \in S^2(H^{1,0})^\vee$ are: $$Q_{\mathrm{III}}(y_1, y_2, y_3) = -9\big(4y_1^2 + 4y_1 y_3 - 2y_2^2 - 4y_2 y_3 + y_3^2\big), \qquad Q_{\mathrm{II}}(y_1, y_2, y_3) = -4\big(y_1^2 - 2y_1 y_3 + y_2^2 - y_2 y_3 + y_3^2\big),$$ which are linearly independent in $\mathbb{Q}[y_1, y_2, y_3]$.

2. (Global Closed-Form Macaulay Determinants on $U$) Over the polynomial ring $\mathbb{Q}[B_1, B_2, B_3]$:

   • In degree $5$ ($\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$, $\dim = 5$, with $2$ Jacobian generators $\sum x_i \partial_i g_B$ and $e_1 \sum \partial_i g_B$ and $3$ Kodaira–Spencer rows $m_{320}, m_{311}, m_{221}$), the $5 \times 5$ Macaulay matrix $M_{5,\mathrm{III}}(B)$ has exact factored determinant: $$\det M_{5, \mathrm{III}}(B_1, B_2, B_3) = 5(2B_1 + B_2 + 5) \not\equiv 0.$$
   • In degree $6$ ($\mathbf{1}_{\mathrm{triv}}^{\mathrm{poly}}$, $\dim = 7$, with $4$ symmetric Jacobian generators and $3$ Kodaira–Spencer rows $e_1 m_{320}, e_1 m_{311}, e_1 m_{221}$), the $7 \times 7$ Macaulay matrix $M_{6,\mathrm{III}}(B)$ has exact factored determinant: $$\det M_{6, \mathrm{III}}(B_1, B_2, B_3) = 15(2B_1 + B_2 + 5)(B_1 + 2B_2 - B_3 + 5)(2B_1 - 2B_2 + B_3 + 10) \not\equiv 0.$$
   • In degree $6$ ($V_{\mathrm{std},-}^{\mathrm{poly}}$, $\dim = 9$, with $6$ Jacobian generators and $3$ Kodaira–Spencer rows $v_- m_{320}, v_- m_{311}, v_- m_{221}$), the $9 \times 9$ Macaulay matrix $M_{6,\mathrm{II}}(B)$ has exact factored determinant: $$\det M_{6, \mathrm{II}}(B_1, B_2, B_3) = -437400(2B_1 + B_2 + 5)(B_1 + 2B_2 - B_3 + 5)\, P_{\mathrm{cub}}(B_1, B_2, B_3) \not\equiv 0,$$ where $P_{\mathrm{cub}}(B_1, B_2, B_3) = 4B_1^3 - 4B_1^2 B_2 + 11B_1^2 + B_1 B_2^2 + 9B_1 B_2 - 12B_1 B_3 - 40B_1 - B_2^2 + B_2 B_3 - 5B_2 + B_3^2 + 15B_3 + 25$.

Proof. All matrix entries, polynomial determinants, and discriminant quadrics are computed in exact integer/rational arithmetic and verified in Steps 2, 3, and 4 of verify.py and hodge_loci_engine/quintic_tower.py. Since the determinants are non-zero polynomials (evaluating to $25$, $3750$, and $-273375000$ at $B = (0, 0, 0)$), the differentials of the period maps of $V_{\mathrm{III}}$ and $V_{\mathrm{II},\pm}$ into $\mathbb{B}^3$ (and of $W_{\mathrm{III}}(C_B), W_{\mathrm{II},\pm}(C_B)$ into $\mathfrak{H}_2$) have full rank $3$ on a non-empty Zariski-open subset of $U$, so the period maps are dominant submersions with open analytic images. $\square$

---

3.2 Generic $\mathbb{Q}$-Hodge Simplicity (Without Assuming $K$-Stability) and the Eigen-Signature Vanishing Theorem

A subtle point in Hodge theory with endomorphisms by a CM field $K = \mathbb{Q}(\zeta_5)$ is that a rational sub-Hodge structure $S \subset W_B$ of a $K$-module $W_B$ is not a priori guaranteed to be $K$-stable (for instance, in a direct sum $K \oplus K$ of two isomorphic variations, a diagonal copy of $\mathbb{Q}$ is not $K$-stable), nor does an arbitrary complex Hodge-graded subspace of a single fiber $W_{B,\mathbb{C}}$ extend to a flat subbundle on an open neighborhood. In this subsection, we give a complete, self-contained proof—relying only on the countability of rational homomorphisms $\operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0)$ on the universal cover $\widetilde{U}$, the open-set irreducibility of the active $4$-dimensional $\mathbb{C}$-VHS $W_{\chi_1}|_{U_0}$, and Galois rank-invariance of the rational branch operators $f_a$—that at every very general parameter $B \in U_{\mathrm{vg}}$, all cross-branch and cross-type rational Hodge morphisms vanish, $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(W_B) = K$, and consequently every moving $16$-dimensional block $W_B$ is a simple $\mathbb{Q}$-Hodge structure (without assuming $K$-stability).

Theorem 3.2 (Generic $\mathbb{Q}$-Hodge Simplicity and Eigen-Signature Vanishing Theorem). Let $W, W'$ be polarized $\mathbb{Q}$-variations of Hodge structure of weight $w$ over $U$ equipped with $\nabla$-flat algebraic $\mu_5$-actions $\rho_W^*, \rho_{W'}^* \in \operatorname{Aut}_{\mathbb{Q}-\mathrm{VHS}}$ generating $K = \mathbb{Q}(\zeta_5)$, taken from the moving $16$-dimensional blocks $V_{\mathrm{II},+}, V_{\mathrm{II},-}, V_{\mathrm{III}} \subset H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$ or $M_{r,\mathrm{II},+}, M_{r,\mathrm{II},-}, M_{r,\mathrm{III}} \subset H^3(Y_B, \mathbb{Q})$ ($r \in \{1, 2, 3\}$).

1. (Rationality, Galois Rank-Invariance, and Eigen-Signature Vanishing of Branch Projectors) At every very general point $B \in U_{\mathrm{vg}}$ (constructed explicitly in Step A below as the complement of a countable union of proper analytic/algebraic Hodge loci in $U$), any rational Hodge morphism $f \in \operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_B, W'_B)$ extends by $\nabla$-flat parallel transport to a $\nabla$-flat morphism of $\mathbb{Q}$-variations of Hodge structure over a connected, simply connected open neighborhood $U_0 \subset U(\mathbb{C})$ of $B_0 = (0, 0, 0)$ (and over a simply connected neighborhood $U_B$ of $B$). For each $a \in (\mathbb{Z}/5\mathbb{Z})^\times = \{1, 2, 3, 4\}$, the Galois-branch operator $$f_a := \frac{1}{5} \sum_{j=0}^4 (\rho_{W'}^*)^{-j} \circ f \circ (\rho_W^*)^{aj} \in \operatorname{Hom}_{\mathbb{Q}-\mathrm{VHS}}\big(\mathbb{W}_{U_0}, \mathbb{W}'_{U_0}\big)$$ is a $\mathbb{Q}$-linear combination of compositions of $\nabla$-flat rational VHS morphisms on $U_0$, hence is itself a $\nabla$-flat rational morphism of $\mathbb{Q}$-VHS on $U_0$ satisfying $f = \sum_{a=1}^4 f_a$ and $\rho_{W'}^* \circ f_a = f_a \circ (\rho_W^*)^a$. Its four character components $f_{a, m} : W_{\chi_m}|_{U_0} \to W'_{\chi_{am \bmod 5}}|_{U_0}$ ($m \in \{1, 2, 3, 4\}$) are $\operatorname{Gal}(\mathbb{Q}(\zeta_5)/\mathbb{Q})$-conjugate ($\tau_t(f_{a, m}) = f_{a, tm \bmod 5}$ for all $t \in (\mathbb{Z}/5\mathbb{Z})^\times$), so $\dim_{\mathbb{C}} \ker(f_{a, m}) = \dim_{\mathbb{C}} \ker(f_{a, 1})$ is strictly independent of $m \in \{1, 2, 3, 4\}$. Consequently, if for a given $a \in \{1, 2, 3, 4\}$ there exists a character $m \in \{1, 2, 3, 4\}$ and bidegree $(p, q)$ such that $\dim_{\mathbb{C}} W_{\chi_m}^{p, q} > \dim_{\mathbb{C}} W'^{p, q}_{\chi_{am \bmod 5}}$, then $\ker(f_{a, 1}) \subset W_{\chi_1}|_{U_0}$ is a non-zero $\nabla$-flat $\mathbb{C}$-sub-VHS on the open set $U_0$, forcing $\ker(f_{a, 1}) = W_{\chi_1}|_{U_0}$ and $f_a = 0$. In particular, at every $B \in U_{\mathrm{vg}}$:
   • If for $W$ and $W'$ every $a \in \{1, 2, 3, 4\}$ has some $m, (p, q)$ with $\dim W_{\chi_m}^{p, q} > \dim W'^{p, q}_{\chi_{am \bmod 5}}$, then $$\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_B, W'_B) = 0.$$
   • For every moving $16$-dimensional block $W \in \{V_{\mathrm{II},+}, V_{\mathrm{II},-}, V_{\mathrm{III}}, M_{r,\mathrm{II},+}, M_{r,\mathrm{II},-}, M_{r,\mathrm{III}}\}$ ($r \in \{1, 2, 3\}$), every $a \in \{2, 3, 4\}$ has some $m, (p, q)$ with $\dim W_{\chi_m}^{p, q} > \dim W_{\chi_{am \bmod 5}}^{p, q}$, so $f_2 = f_3 = f_4 = 0$ and $$\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(W_B) = \operatorname{End}_{K-\mathrm{Hdg}}(W_B) = K \cong \mathbb{Q}(\zeta_5).$$
2. (Generic $\mathbb{Q}$-Hodge Simplicity Without Assuming $K$-Stability) At every $B \in U_{\mathrm{vg}}$, each of the $16$-dimensional polarized $\mathbb{Q}$-Hodge structures $V_{\mathrm{II},+}, V_{\mathrm{II},-}, V_{\mathrm{III}} \subset H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})$ and $M_{r,\mathrm{II},+}, M_{r,\mathrm{II},-}, M_{r,\mathrm{III}} \subset H^3(Y_B, \mathbb{Q})$ ($r \in \{1, 2, 3\}$) is a simple $\mathbb{Q}$-Hodge structure (it contains no proper non-zero rational sub-Hodge structure $S \subsetneq W_B$, whether $K$-stable or not).

Proof. We give the complete proof in four self-contained steps (verified symbolically in Step 4 of verify.py):

Step A (Countability of Rational Maps on the Universal Cover $\widetilde{U}$ and Explicit Construction of $U_{\mathrm{vg}}$). Fix the Fermat basepoint $B_0 = (0, 0, 0) \in U(\mathbb{C})$ and let $\pi_{\mathrm{univ}} : (\widetilde{U}, \tilde{0}) \to (U(\mathbb{C}), B_0)$ be the universal cover of the connected complex manifold $U(\mathbb{C})$. Because $\widetilde{U}$ is simply connected and the Gauss–Manin connection $\nabla$ is flat, parallel transport along $\nabla$ canonically trivializes the rational local systems $\pi_{\mathrm{univ}}^* \mathbb{W}_{\mathbb{Q}} \cong \widetilde{U} \times W_0$ and $\pi_{\mathrm{univ}}^* \mathbb{W}'_{\mathbb{Q}} \cong \widetilde{U} \times W'_0$, where $W_0 = \mathbb{W}_{\mathbb{Q}, B_0} \cong \mathbb{Q}^d$ and $W'_0 = \mathbb{W}'_{\mathbb{Q}, B_0} \cong \mathbb{Q}^{d'}$ are fixed finite-dimensional $\mathbb{Q}$-vector spaces. Under this flat trivialization, the Hodge filtration subbundles $F^p \mathcal{W}_{\tilde{u}} \subset W_0 \otimes_{\mathbb{Q}} \mathbb{C}$ and $F^p \mathcal{W}'_{\tilde{u}} \subset W'_0 \otimes_{\mathbb{Q}} \mathbb{C}$ vary holomorphically with $\tilde{u} \in \widetilde{U}$ (Griffiths' theorem [3]). Unlike the uncountable complex vector space $\operatorname{Hom}_{\mathbb{C}}(W_{0,\mathbb{C}}, W'_{0,\mathbb{C}})$, the rational vector space $\operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0) \cong \mathbb{Q}^{d d'}$ is a countable set. For each fixed rational linear map $f \in \operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0)$, the locus of points on $\widetilde{U}$ where the flat rational section $f$ is a morphism of Hodge structures is $$\widetilde{Z}(f) := \Big\{ \tilde{u} \in \widetilde{U} \;\Big|\; f_{\mathbb{C}}\big(F^p \mathcal{W}_{\tilde{u}}\big) \subset F^p \mathcal{W}'_{\tilde{u}} \text{ for all } p \in \mathbb{Z} \Big\} \subset \widetilde{U}.$$ Because $\widetilde{Z}(f)$ is the common zero locus of the holomorphic vector bundle homomorphisms $F^p \mathcal{W} \xrightarrow{f_{\mathbb{C}}} (\widetilde{U} \times W'_{0,\mathbb{C}})/F^p \mathcal{W}'$, $\widetilde{Z}(f)$ is a closed complex-analytic subset of the connected complex manifold $\widetilde{U}$. By the Identity Theorem for complex-analytic subsets on the connected manifold $\widetilde{U}$, each fixed $f \in \operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0)$ satisfies a strict dichotomy:

• Case (i) (Proper fiberwise Hodge locus): $\widetilde{Z}(f) \subsetneq \widetilde{U}$ is a proper closed complex-analytic subset ($\operatorname{codim}_{\mathbb{C}} \widetilde{Z}(f) \ge 1$), so its projection $Z(f) := \pi_{\mathrm{univ}}(\widetilde{Z}(f)) \subset U(\mathbb{C})$ is a countable union of proper closed complex-analytic subsets of $U(\mathbb{C})$ of Lebesgue measure zero (and by Cattani–Deligne–Kaplan [11], $Z(f)$ is in fact a proper algebraic subvariety of $U$); or
• Case (ii) (Global flat VHS morphism): $\widetilde{Z}(f) = \widetilde{U}$, which means that the $\nabla$-flat rational map $f$ preserves the Hodge filtration $F^p$ at every point $\tilde{u} \in \widetilde{U}$; in particular, over a connected, simply connected open neighborhood $U_0 \subset U(\mathbb{C})$ of the Fermat origin $B_0 = (0, 0, 0)$ (and over a simply connected open neighborhood $U_B$ of $B$), $f$ is a $\nabla$-flat morphism of $\mathbb{Q}$-variations of Hodge structure already defined on the open set $U_0$.

Let $U_{\mathrm{KS}} \subset U$ be the non-empty Zariski-open subset where all Macaulay and Siegel determinants of Proposition 3.1 are non-zero, and define the very general locus $U_{\mathrm{vg}} \subset U(\mathbb{C})$ explicitly as the complement of the countable union of all Case-(i) proper Hodge loci across all pairs of blocks $(W, W')$ in our tower: $$U_{\mathrm{vg}} := U_{\mathrm{KS}}(\mathbb{C}) \setminus \bigcup_{\substack{(W, W') \\ f \in \operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0) \\ \widetilde{Z}(f) \subsetneq \widetilde{U}}} \pi_{\mathrm{univ}}\big(\widetilde{Z}(f)\big).$$ Since $\operatorname{Hom}_{\mathbb{Q}}(W_0, W'_0) \cong \mathbb{Q}^{d d'}$ is countable, the union above is a countable union of proper analytic (and algebraic) subsets of $U(\mathbb{C})$, so $U_{\mathrm{vg}}$ is dense and of full measure in $U(\mathbb{C})$ by the Baire Category Theorem. By construction, for every point $B \in U_{\mathrm{vg}}$ and any lift $\tilde{u} \in \pi_{\mathrm{univ}}^{-1}(B) \subset \widetilde{U}$, if $f \in \operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_B, W'_B)$ is a rational Hodge morphism at the fiber $B$, then $\tilde{u} \in \widetilde{Z}(f)$ while $B \notin \bigcup_{\widetilde{Z}(g) \subsetneq \widetilde{U}} \pi_{\mathrm{univ}}(\widetilde{Z}(g))$; therefore $f$ cannot fall into Case (i), so $\widetilde{Z}(f) = \widetilde{U}$ (Case (ii)), and $f$ extends by parallel transport to a $\nabla$-flat morphism of $\mathbb{Q}$-variations of Hodge structure across the entire open neighborhood $U_0$ of $B_0 = (0, 0, 0)$ (and across a simply connected neighborhood $U_B \subset U_{\mathrm{KS}}(\mathbb{C})$ of $B$ itself).

Step B (Irreducibility of the Active $4$-Dimensional $\mathbb{C}$-Eigenspace Variations over Open Sets). For each $m \in \{1, 2, 3, 4\}$, the spectral projector $\pi_m := \frac{1}{5}\sum_{j=0}^4 \zeta_5^{-jm} (\rho_W^*)^j \in \mathbb{C}[\mu_5]$ onto $W_{\chi_m} \subset W \otimes_{\mathbb{Q}} \mathbb{C}$ is a constant $\mathbb{C}$-linear combination of the algebraic automorphisms $(\rho_W^*)^j$. Consequently, $\pi_m$ is flat with respect to the Gauss–Manin connection ($\nabla \circ \pi_m = \pi_m \circ \nabla$) and preserves the Hodge decomposition on every fiber ($\pi_m(W^{p, q}) = W_{\chi_m}^{p, q}$). On the active character $m = 1$ (where $W_{\chi_1}$ has signature $(1, 3)$ with $\det M_{\mathrm{KS}}|_{B=0} \in \{6, -1\} \ne 0$, and likewise $m = 2$ for Type III where $\det M_{\mathrm{KS}}|_{B=0} = 1 \ne 0$, and at every $B \in U_{\mathrm{KS}}$ by Proposition 3.1(2)), let $E \subset W_{\chi_m}|_{U_0}$ be any $\nabla$-flat $\mathbb{C}$-sub-variation of Hodge structure already defined on the connected open neighborhood $U_0 \subset U(\mathbb{C})$ of $B_0 = (0, 0, 0)$ (or on $U_B \subset U_{\mathrm{KS}}(\mathbb{C})$). Since $\dim_{\mathbb{C}} W_{\chi_m}^{\mathrm{ext}} = 1$ (where $W_{\chi_m}^{\mathrm{ext}}$ is the $1$-dimensional $(2,0)$-line on $\Sigma_B$, or the $1$-dimensional $(2,1)$-line at $m=1$ on $Y_B$, or the $1$-dimensional $(3,0)$-line at $m=2$ on $M_{1,\mathrm{III}}$) and $E$ is Hodge-graded on $U_0$:

• If $E \cap W_{\chi_m}^{\mathrm{ext}} \ne 0$ on $U_0$, then $W_{\chi_m}^{\mathrm{ext}} \subset E$. By $\nabla$-flatness of $E$ on $U_0$, $E$ contains $\omega(B_0) \in W_{\chi_m}^{\mathrm{ext}}(B_0)$ and its three first-order Gauss–Manin derivatives $\nabla_{\partial_{B_1}}\omega(B_0), \nabla_{\partial_{B_2}}\omega(B_0), \nabla_{\partial_{B_3}}\omega(B_0)$, which by Proposition 3.1(1) span all $4$ dimensions of $W_{\chi_m}$ (and likewise at every $B \in U_{\mathrm{KS}}$ by Proposition 3.1(2)), so $E = W_{\chi_m}|_{U_0}$.
• If $E \cap W_{\chi_m}^{\mathrm{ext}} = 0$ on $U_0$, then for every $B' \in U_0$, the $1$-dimensional period line $\Phi_{\chi_m}(B') = W_{\chi_m}^{\mathrm{ext}}(B') \in \mathbb{P}(W_{\chi_m}) \cong \mathbb{P}^3$ is $h_{\chi_m}$-orthogonal to the fixed flat subspace $E \subset W_{\chi_m}$ under the flat Hermitian polarization form $h_{\chi_m}$, confining $\Phi_{\chi_m}(U_0)$ to the proper projective subspace $\mathbb{P}(E^\perp) \subsetneq \mathbb{P}^3$. Since $\Phi_{\chi_m}(U_0)$ is open in $\mathbb{B}^3 \subset \mathbb{P}^3$ by Proposition 3.1, we must have $E^\perp = W_{\chi_m}$ and $E = 0$. Thus $W_{\chi_1}|_{U_0}$ (and its complex conjugate $W_{\chi_4}|_{U_0} = \overline{W_{\chi_1}}|_{U_0}$, as well as $W_{\chi_2}|_{U_0}, W_{\chi_3}|_{U_0}$ for Type III) is an irreducible $4$-dimensional $\mathbb{C}$-VHS over the open set $U_0$ (and over $U_B$).

Step C (Galois Rank-Invariance of the Rational Branch Operators $f_a$ and Proof of Part 1). Let $B \in U_{\mathrm{vg}}$ and let $f \in \operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_B, W'_B)$. By Step A, $f$ extends to a $\nabla$-flat morphism of $\mathbb{Q}$-VHS over the open neighborhood $U_0$ of $B_0 = (0, 0, 0)$. Since $\rho_W^* \in \operatorname{Aut}_{\mathbb{Q}-\mathrm{VHS}}(\mathbb{W}_{U_0})$ and $\rho_{W'}^* \in \operatorname{Aut}_{\mathbb{Q}-\mathrm{VHS}}(\mathbb{W}'_{U_0})$ are $\nabla$-flat rational automorphisms of order $5$ satisfying $1 + \rho^* + (\rho^*)^2 + (\rho^*)^3 + (\rho^*)^4 = 0$, each Galois-branch operator $$f_a := \frac{1}{5}\sum_{j=0}^4 (\rho_{W'}^*)^{-j} \circ f \circ (\rho_W^*)^{aj} \in \operatorname{Hom}_{\mathbb{Q}-\mathrm{VHS}}\big(\mathbb{W}_{U_0}, \mathbb{W}'_{U_0}\big) \qquad (a \in \{1, 2, 3, 4\})$$ is a $\nabla$-flat morphism of $\mathbb{Q}$-VHS over the open set $U_0$, satisfying $\sum_{a=1}^4 f_a = f - \frac{1}{5}\sum_{j=0}^4 f \circ (\rho_W^*)^j = f$ and $\rho_{W'}^* \circ f_a = f_a \circ (\rho_W^*)^a$. Over $U_0$, $f_a \otimes \mathbb{C}$ decomposes into four $\nabla$-flat morphisms of $\mathbb{C}$-VHS already defined on the open set $U_0$: $$f_{a, m} : W_{\chi_m}\big|_{U_0} \longrightarrow W'_{\chi_{am \bmod 5}}\big|_{U_0} \qquad (m \in \{1, 2, 3, 4\}).$$ In a $\mathbb{Q}$-basis of the constant reference fibers $W_0 \cong \mathbb{Q}^{16}$ and $W'_0 \cong \mathbb{Q}^{16}$, $f_a, \rho_W^*, \rho_{W'}^* \in M_{16}(\mathbb{Q})$ have rational matrix entries, and $f_{a, m}$ is represented by the matrix $M(f_{a, m}) := \pi'_{am \bmod 5} f_a \pi_m = f_a \pi_m \in M_{16}(\mathbb{Q}(\zeta_5))$. For every $t \in (\mathbb{Z}/5\mathbb{Z})^\times = \{1, 2, 3, 4\}$, applying the Galois automorphism $\tau_t \in \operatorname{Gal}(\mathbb{Q}(\zeta_5)/\mathbb{Q})$ ($\tau_t(\zeta_5) = \zeta_5^t$) entrywise to $M(f_{a, m})$ fixes the rational matrix $f_a$ and conjugates $\tau_t(\pi_m) = \frac{1}{5}\sum_{j=0}^4 \zeta_5^{-jtm}(\rho_W^*)^j = \pi_{tm \bmod 5}$, yielding the exact Galois conjugation identity: $$\tau_t\big(M(f_{a, m})\big) = f_a \, \tau_t(\pi_m) = f_a \, \pi_{tm \bmod 5} = M(f_{a, tm \bmod 5}) \qquad \forall m, t \in \{1, 2, 3, 4\}.$$ Because the rank of a matrix over $\mathbb{Q}(\zeta_5)$ is invariant under the field automorphism $\tau_t$, $\operatorname{rank}_{\mathbb{C}}(f_{a, m}) = \operatorname{rank}(M(f_{a, m}))$ and $\dim_{\mathbb{C}} \ker(f_{a, m}) = 4 - \operatorname{rank}_{\mathbb{C}}(f_{a, m})$ are strictly independent of $m \in \{1, 2, 3, 4\}$: $$\dim_{\mathbb{C}} \ker(f_{a, 1}) = \dim_{\mathbb{C}} \ker(f_{a, 2}) = \dim_{\mathbb{C}} \ker(f_{a, 3}) = \dim_{\mathbb{C}} \ker(f_{a, 4}).$$ Now observe that at the active character $m = 1$, the kernel $$E_{a, 1} := \ker(f_{a, 1}) \subset W_{\chi_1}\big|_{U_0}$$ is the kernel of a $\nabla$-flat morphism of $\mathbb{C}$-VHS on $U_0$, hence is a $\nabla$-flat $\mathbb{C}$-sub-variation of Hodge structure of $W_{\chi_1}$ already defined on the open set $U_0$! Therefore, if for a given $a \in \{1, 2, 3, 4\}$ there exists any character $m \in \{1, 2, 3, 4\}$ and bidegree $(p, q)$ such that $\dim_{\mathbb{C}} W_{\chi_m}^{p, q} > \dim_{\mathbb{C}} W'^{p, q}_{\chi_{am \bmod 5}}$, then $f_{a, m} : W_{\chi_m}^{p, q} \to W'^{p, q}_{\chi_{am \bmod 5}}$ has a non-trivial kernel ($\dim_{\mathbb{C}} \ker(f_{a, m}) \ge 1$), so by Galois rank-invariance $\dim_{\mathbb{C}} \ker(f_{a, 1}) = \dim_{\mathbb{C}} \ker(f_{a, m}) \ge 1$. That is, $E_{a, 1} = \ker(f_{a, 1}) \subset W_{\chi_1}|_{U_0}$ is a non-zero $\nabla$-flat $\mathbb{C}$-sub-VHS of $W_{\chi_1}$ already defined on the open set $U_0$. By Step B (irreducibility of $W_{\chi_1}|_{U_0}$ on the open set $U_0$), we immediately obtain $E_{a, 1} = W_{\chi_1}|_{U_0}$, so $\operatorname{rank}_{\mathbb{C}}(f_{a, 1}) = 0$, and hence by Galois rank-invariance $\operatorname{rank}_{\mathbb{C}}(f_{a, m}) = 0$ for all $m \in \{1, 2, 3, 4\}$—that is, $f_a = 0$! In particular, if every $a \in \{1, 2, 3, 4\}$ has some $m, (p, q)$ with $\dim W_{\chi_m}^{p, q} > \dim W'^{p, q}_{\chi_{am \bmod 5}}$, then $f_1 = f_2 = f_3 = f_4 = 0$, so $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_B, W'_B) = 0$. Similarly, for $W' = W$ on each of the moving $16$-dimensional blocks $W \in \{V_{\mathrm{II},+}, V_{\mathrm{II},-}, V_{\mathrm{III}}, M_{r,\mathrm{II},+}, M_{r,\mathrm{II},-}, M_{r,\mathrm{III}}\}$ ($r \in \{1, 2, 3\}$), every $a \in \{2, 3, 4\}$ has some $m, (p, q)$ with $\dim W_{\chi_m}^{p, q} > \dim W_{\chi_{am \bmod 5}}^{p, q}$:

• On $V_{\mathrm{II},\pm}$, $h^{2,0} = (1, 0, 0, 0)$, so for $a \in \{2, 3, 4\}$, $h^{2,0}_{\chi_1} = 1 > 0 = h^{2,0}_{\chi_a}$;
• On $V_{\mathrm{III}}$, $h^{2,0} = (1, 1, 0, 0)$, so for $a = 2$, $h^{2,0}_{\chi_2} = 1 > 0 = h^{2,0}_{\chi_4}$, and for $a \in \{3, 4\}$, $h^{2,0}_{\chi_1} = 1 > 0 = h^{2,0}_{\chi_a}$;
• On $M_{1,\mathrm{III}}$, $h^{3,0} = (0, 1, 0, 0)$, so for $a \in \{2, 3, 4\}$, $h^{3,0}_{\chi_2} = 1 > 0 = h^{3,0}_{\chi_{2a \bmod 5}}$;
• On $M_{1,\mathrm{II},\pm}$, $h^{2,1} = (1, 4, 0, 3)$, so for $a \in \{2, 3, 4\}$, $h^{2,1}_{\chi_2} = 4 > h^{2,1}_{\chi_{2a \bmod 5}}$;
• On $M_{2,\mathrm{II},\pm}, M_{3,\mathrm{II},\pm}$, $h^{2,1} = (1, 0, 4, 3)$, so for $a \in \{2, 3, 4\}$, $h^{2,1}_{\chi_3} = 4 > h^{2,1}_{\chi_{3a \bmod 5}}$;
• On $M_{2,\mathrm{III}}, M_{3,\mathrm{III}}$, $h^{2,1} = (1, 1, 3, 3)$, so for $a \in \{2, 3\}$, $h^{2,1}_{\chi_4} = 3 > 1 = h^{2,1}_{\chi_{4a \bmod 5}}$, and for $a = 4$, $h^{2,1}_{\chi_3} = 3 > 1 = h^{2,1}_{\chi_2}$. Thus $f_2 = f_3 = f_4 = 0$ on $W_B$, so $f = f_1 \in \operatorname{End}_{K-\mathrm{Hdg}}(W_B)$. Finally, since $W_{\chi_1}|_{U_0}$ is an irreducible $\mathbb{C}$-VHS over $U_0$ by Step B, Schur's Lemma on $W_{\chi_1}|_{U_0}$ forces $f_{1, 1} = \lambda\,\mathrm{id}_{W_{\chi_1}}$, and since $f_1$ is rational and commutes with $\rho_W^*$, we have $f_1 \in K\,\mathrm{id}_{W_B}$, proving Part 1: $$\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(W_B) = \operatorname{End}_{K-\mathrm{Hdg}}(W_B) = K \cong \mathbb{Q}(\zeta_5).$$

Step D (Generic $\mathbb{Q}$-Hodge Simplicity Without Assuming $K$-Stability: Proof of Part 2). Let $B \in U_{\mathrm{vg}}$ and let $S \subset W_B$ be any rational sub-Hodge structure of $W_B$ (without assuming that $S$ is $K$-stable). Because $W_B$ is polarized by the rational intersection pairing $Q$ (so the Hodge metric $Q(C x, \bar{y})$ is positive-definite on $W_{B,\mathbb{C}}$, and hence remains positive-definite on the subspace $S_{\mathbb{C}}$), the restriction $Q|_S$ is non-degenerate and $W_B = S \oplus S^\perp$ splits into $Q$-orthogonal rational sub-Hodge structures. The orthogonal projection $P_S : W_B \to W_B$ onto $S$ along $S^\perp$ is therefore a rational Hodge endomorphism of $W_B$, which by Part 1 (Step C) must lie in the field $K$: $$P_S \in \operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(W_B) = K, \qquad P_S^2 = P_S.$$ Because $K = \mathbb{Q}(\zeta_5)$ is a field, the idempotent equation $x^2 = x$ in $K$ has only the two solutions $x = 0$ and $x = 1$. Therefore, either $P_S = 0$ ($S = 0$) or $P_S = \mathrm{id}_{W_B}$ ($S = W_B$). This proves Part 2: every moving $16$-dimensional block $W_B$ is a simple $\mathbb{Q}$-Hodge structure at every $B \in U_{\mathrm{vg}}$. (As an independent geometric cross-check on $V_{\mathrm{III}}$ and $M_{2,3,\mathrm{III}}$, note also from Proposition 3.1(2) that at $B_* = (0, -2, 1)$, $\det M_{5,\mathrm{III}}(B_*) = 15 \ne 0$ gives period differential rank $3$ on $\chi_2$ whereas $\det M_{6,\mathrm{III}}(B_*) = 0$ gives rank $2 < 3$ on $\chi_1$, and $d\log\det M_{5,\mathrm{III}}|_{B=0}, d\log\det M_{6,\mathrm{III}}|_{B=0}$ are linearly independent in $T_0^* U$, directly ruling out any flat $\mathbb{C}$-VHS isomorphism between $W_{\chi_1}$ and $W_{\chi_2}$.) $\square$

---

4. Proof of Theorem 1 (Cyclic Quintic Surface Squares) and Theorem 2 (Split Quintic Fourfolds)

4.1 Theorem 1: Resolution of $\Sigma_B \times \Sigma_B$

Theorem 1. For very general $B = (B_1, B_2, B_3) \in U$, the rational Hodge conjecture holds for $H^4(\Sigma_B \times \Sigma_B, \mathbb{Q})$ ($\dim = 2{,}811$; primitive tensor sector $\dim = 2{,}704$). Specifically:

1. On $H^2(\Sigma_B, \mathbb{Q}) = \mathbb{Q}h \oplus V_{\mathrm{I}} \oplus V_{\mathrm{II},+} \oplus V_{\mathrm{II},-} \oplus V_{\mathrm{III}}$ ($\dim = 53$), the Néron–Severi space at very general $B \in U$ has exact rank $\rho(\Sigma_B) = 1 + 4 = 5$, spanned by $h$ and $V_{\mathrm{I}} = e_{\mathrm{I}} \operatorname{NS}_{\mathrm{prim}}(\Sigma_B)_{\mathbb{Q}}$, and the transcendental lattice $T(\Sigma_B) = V_{\mathrm{II},+} \oplus V_{\mathrm{II},-} \oplus V_{\mathrm{III}}$ has exact rank $48$.
2. At very general $B \in U$, the rational Hodge endomorphism algebra of $T(\Sigma_B)$ has dimension $20$ over $\mathbb{Q}$: $$\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}\big(T(\Sigma_B)\big) = M_2(K)_{V_{\mathrm{II}}} \oplus K_{V_{\mathrm{III}}} \cong M_2\big(\mathbb{Q}(\zeta_5)\big) \oplus \mathbb{Q}(\zeta_5) \qquad (\dim_{\mathbb{Q}} = 16 + 4 = 20),$$ and every rational Hodge class in $H^4(\Sigma_B \times \Sigma_B, \mathbb{Q})$ ($\dim_{\mathbb{Q}} \mathrm{Hdg}^{2,2}(\Sigma_B \times \Sigma_B, \mathbb{Q}) = 2 + 5^2 + 20 = 47$, with $36$ in $H^2_{\mathrm{prim}} \otimes H^2_{\mathrm{prim}}$) is an explicit $\mathbb{Q}$-linear combination of $[\Sigma_B \times \{\mathrm{pt}\}]$, $[\{\mathrm{pt}\} \times \Sigma_B]$, external products of algebraic divisors $[D_1 \times D_2]$ ($D_1, D_2 \in \operatorname{NS}(\Sigma_B)_{\mathbb{Q}}$), and the $20$ linearly independent algebraic group-ring correspondence graphs $$\Big\{e_{\mathrm{II},+}(\rho^*)^j,\; e_{\mathrm{II},-}(\rho^*)^j,\; T_{\mathrm{II}}(\rho^*)^j,\; T_{\mathrm{II}}^\dagger(\rho^*)^j,\; e_{\mathrm{III}}(\rho^*)^j\Big\}_{j=0}^3 \subset \mathbb{Q}[\mu_5 \times \mathfrak{S}_3] \subset CH^2(\Sigma_B \times \Sigma_B)_{\mathbb{Q}}.$$

Proof. By Theorem 3.2(2), $V_{\mathrm{II},+}$, $V_{\mathrm{II},-}$, and $V_{\mathrm{III}}$ are simple $16$-dimensional $\mathbb{Q}$-Hodge structures at very general $B \in U_{\mathrm{vg}}$; in particular, none of them contains a $1$-dimensional rational $(1,1)$-class, while $V_{\mathrm{I}} \subset \operatorname{NS}(\Sigma_B)_{\mathbb{Q}}$ is algebraic on every smooth fiber by Proposition 2.1(4). Thus $\operatorname{NS}(\Sigma_B)_{\mathbb{Q}} = \mathbb{Q}h \oplus V_{\mathrm{I}}$ ($\dim = 5$) and $T(\Sigma_B) = V_{\mathrm{II},+} \oplus V_{\mathrm{II},-} \oplus V_{\mathrm{III}}$ ($\dim = 48$). Now apply Theorem 3.2(1) to the $(2,0)$-dimension vectors across $(\chi_1, \chi_2, \chi_3, \chi_4)$:

• For $V_{\mathrm{II},\pm}$, $h^{2,0} = (1, 0, 0, 0)$. For $a \in \{2, 3, 4\}$, $\dim V_{\mathrm{II},\pm,\chi_a}^{2,0} = 0 < 1 = \dim V_{\mathrm{II},\pm,\chi_1}^{2,0}$. By Theorem 3.2(1), $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(V_{\mathrm{II},+}) = K$, and since $T_{\mathrm{II}} : V_{\mathrm{II},+} \xrightarrow{\sim} V_{\mathrm{II},-}$ is an algebraic $K$-isomorphism (Proposition 2.1(3)), $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(V_{\mathrm{II},+} \oplus V_{\mathrm{II},-}) = M_2(K)$ ($\dim_{\mathbb{Q}} = 16$).
• For $V_{\mathrm{III}}$, $h^{2,0} = (1, 1, 0, 0)$. For $a = 2$, $2 \cdot 2 = 4$ has $h^{2,0}_{\chi_4} = 0 < 1 = h^{2,0}_{\chi_2}$; for $a = 3$, $3 \cdot 1 = 3$ has $h^{2,0}_{\chi_3} = 0 < 1 = h^{2,0}_{\chi_1}$; for $a = 4$, $4 \cdot 1 = 4$ has $h^{2,0}_{\chi_4} = 0 < 1 = h^{2,0}_{\chi_1}$. By Theorem 3.2(1), $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(V_{\mathrm{III}}) = K$ ($\dim_{\mathbb{Q}} = 4$).
• Between $V_{\mathrm{III}}$ and $V_{\mathrm{II},\pm}$: for every $a \in \{1, 2, 3, 4\}$, $\{a \cdot 1 \bmod 5,\, a \cdot 2 \bmod 5\}$ contains two distinct characters, so at least one of them is $\ne 1$, where $V_{\mathrm{II},\pm}^{2,0} = 0$ while $V_{\mathrm{III}}^{2,0}$ has dimension $1$. By Theorem 3.2(1), $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(V_{\mathrm{III}}, V_{\mathrm{II},\pm}) = 0$.

In Step 4 of verify.py, we construct the explicit rational matrix representations of the $20$ group-ring operators $\{e_{\mathrm{II},\pm}(\rho^*)^j, T_{\mathrm{II}}(\rho^*)^j, T_{\mathrm{II}}^\dagger(\rho^*)^j, e_{\mathrm{III}}(\rho^*)^j\}_{j=0}^3 \subset \mathbb{Q}[\mu_5 \times \mathfrak{S}_3]$ and verify that they have exact rank $20 = 16 + 4$ over $\mathbb{Q}$, spanning all of $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(T(\Sigma_B)) \cong \mathrm{Hdg}^{2,2}(T(\Sigma_B) \otimes T(\Sigma_B))$. $\square$

---

4.2 Theorem 2: Resolution of Split Quintic Fourfolds $X_B^{(4)} \subset \mathbb{P}^5$

Let $X_B^{(4)} = V(g_B(x_0, x_1, x_2) - g_B(y_0, y_1, y_2)) \subset \mathbb{P}^5$. By Griffiths' residue formula on $\mathbb{P}^5$, $H^4_{\mathrm{prim}}(X_B^{(4)}, \mathbb{Q})$ has dimension $820$ (total $b_4 = 821$) with Hodge numbers $(0, 120, 580, 120, 0)$.

Theorem 2. For very general $B = (B_1, B_2, B_3) \in U$, the rational Hodge conjecture holds for the split quintic fourfold $X_B^{(4)} \subset \mathbb{P}^5$ (all $821$ middle cohomology dimensions, with $30$ rational Hodge classes). Specifically, under the Katsura–Shioda correspondence $\mathrm{Bl}_{C_B \times C_B}(X_B^{(4)}) \cong \mathrm{Bl}(\Sigma_B \times \Sigma_B)/\mu_5$:

1. The primitive cohomology decomposes into an orthogonal direct sum of rational Hodge structures: $$H^4_{\mathrm{prim}}\big(X_B^{(4)}, \mathbb{Q}\big) \cong \big[H^1(C_B, \mathbb{Q}) \otimes H^1(C_B, \mathbb{Q})\big](-1) \;\oplus\; \big[H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q}) \otimes H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})\big]^{\mu_5},$$ with exact Hodge-number balance $(0, 36, 72, 36, 0) + (0, 84, 508, 84, 0) = (0, 120, 580, 120, 0)$ ($144 + 676 = 820$).
2. In the $144$-dimensional curve-center sector $H^1(C_B)^{\otimes 2}(-1)$, the non-zero Siegel tangent determinants $\det M_{\mathrm{Siegel}}|_{B=0} = -8 \ne 0$ and the linearly independent discriminant quadrics $Q_{\mathrm{III}}, Q_{\mathrm{II}} \in S^2(T_0^* U)$ (Proposition 3.1(1)) imply that at very general $B \in U$, $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(H^1(C_B, \mathbb{Q})) = \mathbb{Q}_{W_{\mathrm{III}}} \oplus M_2(\mathbb{Q})_{W_{\mathrm{II}}}$ has dimension $1 + 4 = 5$, and every rational Hodge class in this sector is algebraic, generated by the pushforwards from the blow-up exceptional divisor of the $5$ group-ring correspondence graphs $\{e_{\mathrm{III}}, e_{\mathrm{II},\pm}, T_{\mathrm{II}}, T_{\mathrm{II}}^\dagger\} \subset \mathbb{Q}[\mathfrak{S}_3]$ on $C_B \times C_B$ (or by the Lefschetz $(1,1)$ theorem on $C_B \times C_B$).
3. In the $676$-dimensional cyclic-surface sector $[H^2_{\mathrm{prim}}(\Sigma_B)^{\otimes 2}]^{\mu_5}$, the rational Hodge classes have dimension $4$ (from $[V_{\mathrm{I}} \otimes V_{\mathrm{I}}]^{\mu_5}$) $+ 16$ (from $[V_{\mathrm{II}} \otimes V_{\mathrm{II}}]^{\mu_5}$) $+ 4$ (from $[V_{\mathrm{III}} \otimes V_{\mathrm{III}}]^{\mu_5}$) $= 24$, all of which are images under the rational Katsura–Shioda correspondence $\pi_* p_* q^*$ of the algebraic classes of Theorem 1. Together with $h^2$, $\dim_{\mathbb{Q}} \mathrm{Hdg}^{2,2}(X_B^{(4)}, \mathbb{Q}) = 1 + 5 + 24 = 30$.

Proof. On $H^1(C_B, \mathbb{Q}) = W_{\mathrm{III}}(C_B) \oplus W_{\mathrm{II},+}(C_B) \oplus W_{\mathrm{II},-}(C_B)$, the open Siegel period images in $\mathfrak{H}_2$ ($\det M_{\mathrm{Siegel}}|_{B=0} = -8 \ne 0$) imply that at very general $B \in U_{\mathrm{vg}}$, $W_{\mathrm{III}}(C_B)$ and $W_{\mathrm{II},\pm}(C_B)$ have generic Mumford–Tate group $\mathrm{GSp}_4(\mathbb{Q})$ and endomorphism algebra $\mathbb{Q}$. Moreover, by Step A of Theorem 3.2, any rational Hodge morphism $f \in \operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_{\mathrm{III}}(C_B), W_{\mathrm{II},\pm}(C_B))$ at $B \in U_{\mathrm{vg}}$ extends to a $\nabla$-flat morphism of $\mathbb{Q}$-VHS over the open neighborhood $U_0$ of $B_0 = (0, 0, 0)$, which would induce $M : W_{\mathrm{III}}^{1,0} \to W_{\mathrm{II},\pm}^{1,0}$ intertwining the symmetric $2 \times 2$ Kodaira–Spencer matrices $A_r^{\mathrm{II}} = M^T A_r^{\mathrm{III}} M$ ($r = 1, 2, 3$), forcing $Q_{\mathrm{II}}(y_1, y_2, y_3) = (\det M)^2 Q_{\mathrm{III}}(y_1, y_2, y_3)$ in $S^2(T_0^* U)$. Since $Q_{\mathrm{III}}$ and $Q_{\mathrm{II}}$ are linearly independent in $\mathbb{Q}[y_1, y_2, y_3]$ by Proposition 3.1(1) (verified in Step 4 of verify.py), $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(W_{\mathrm{III}}(C_B), W_{\mathrm{II},\pm}(C_B)) = 0$. The rest of the proof follows directly from Theorem 1 and the Katsura–Shioda blow-up decomposition [9]. $\square$

---

5. Proof of Theorem 3: Calabi–Yau Quintic Threefold Squares $Y_B \times Y_B$

Consider the $3$-parameter family of smooth Calabi–Yau quintic threefolds $$Y_B = V\big(u^5 + x_3^5 - g_B(x_0, x_1, x_2)\big) \subset \mathbb{P}^4 \qquad (B \in U),$$ which has canonical bundle $K_{Y_B} \cong \mathcal{O}_{Y_B}$, Hodge numbers $h^{3,0} = 1, h^{2,1} = 101, h^{1,2} = 101, h^{0,3} = 1$, third Betti number $b_3(Y_B) = 204$, and Hodge level $3$. Here $\mu_{5, u} \times \mu_{5, x_3}$ acts on $(u, x_3)$ by $(\zeta_5^j u, \zeta_5^k x_3)$, where $j = a_u + 1 \in \{1, 2, 3, 4\}$ and $k = a_{x_3} + 1 \in \{1, 2, 3, 4\}$ index the $16$ character sectors of the Jacobian ring $$J(Y_B) = \mathbb{C}[u]/(u^4) \otimes \mathbb{C}[x_3]/(x_3^4) \otimes J_{g_B}, \qquad a_u + a_{x_3} + d_x = 5p \quad (p \in \{0, 1, 2, 3\}).$$

5.1 Two-Way Katsura–Shioda Structure and the Four Galois Orbits of $H^3(Y_B, \mathbb{Q})$

Under $\operatorname{Gal}(\mathbb{Q}(\zeta_5)/\mathbb{Q})$, which acts diagonally on $(j, k) \in \{1, 2, 3, 4\}^2$ by $(j, k) \mapsto (tj, tk) \bmod 5$, the $16$ pairs $(j, k)$ partition into four Galois-invariant orbits indexed by the ratio $r \equiv k j^{-1} \pmod 5 \in \{1, 2, 3, 4\}$:

1. Orbit $r = 4$ ($j + k \equiv 0 \pmod 5$, $a_u + a_{x_3} = 3$, $\dim_{\mathbb{Q}} = 48$): Here $d_x = 5p - 3 \in \{2, 7\}$ for $p \in \{1, 2\}$ across all four values of $j \in \{1, 2, 3, 4\}$. Geometrically, in the $(2, 3)$ Katsura–Shioda splitting of $Y_B = V((u^5 + x_3^5) - g_B(x_0, x_1, x_2))$ with blow-up center $P_5 \times C_B$ (where $P_5 = V(u^5 + x_3^5) \subset \mathbb{P}^1$ consists of $5$ points and $C_B = V(g_B) \subset \mathbb{P}^2$ is our genus-$6$ plane quintic curve), this $48$-dimensional rational Hodge structure is canonically $$M_4 = H^0_{\mathrm{prim}}(P_5, \mathbb{Q}) \otimes_{\mathbb{Q}} H^1(C_B, \mathbb{Q})(-1) \cong \mathbb{Q}(0)^4 \otimes_{\mathbb{Q}} H^1(C_B, \mathbb{Q})(-1),$$ which has pure Hodge level $1$ (Hodge numbers $(0, 24, 24, 0)$).
2. Orbits $r \in \{1, 2, 3\}$ ($j + k \not\equiv 0 \pmod 5$, $\dim_{\mathbb{Q}} = 3 \times 52 = 156$): In the $(2, 3)$ Katsura–Shioda splitting, the cyclic cover of $\mathbb{P}^1$ branched at $P_5$ is the Fermat quintic curve $C_5 = V(v^5 - u^5 - x_3^5) \subset \mathbb{P}^2$ (of genus $6$, with $b_1(C_5) = 12$ decomposing into the three $4$-dimensional CM sub-Hodge structures $L_1, L_2, L_3$ corresponding to $r = 1, 2, 3$), and the cyclic cover of $\mathbb{P}^2$ branched at $C_B$ is our cyclic quintic surface $\Sigma_B = V(v^5 - g_B(x)) \subset \mathbb{P}^3$. Thus the $156$-dimensional complement of $M_4$ is canonically the tensor product over $K = \mathbb{Q}(\zeta_5)$: $$M_{1,2,3} = \big[H^1(C_5, \mathbb{Q}) \otimes_{\mathbb{Q}} H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})\big]^{\mu_{5, v}} = \bigoplus_{r \in \{1, 2, 3\}} L_r \otimes_K \big(V_{\mathrm{I}} \oplus V_{\mathrm{II},+} \oplus V_{\mathrm{II},-} \oplus V_{\mathrm{III}}\big).$$ Specifically, for each $r \in \{1, 2, 3\}$, indexing the $K$-eigenspaces by the $\Sigma_B$-character $m \equiv -(j + k) \equiv -(1 + r)j \pmod 5 \in \{1, 2, 3, 4\}$:
   • Type I ($M_{r, \mathrm{I}} = L_r \otimes_K V_{\mathrm{I}}$, $\dim_{\mathbb{Q}} = 4$ each, total $12$): Since $V_{\mathrm{I}}$ is pure $(1,1)$ of $\dim_K = 1$ (algebraic divisors on $\Sigma_B$ by Proposition 2.1(4)), $M_{1,\mathrm{I}} \oplus M_{2,\mathrm{I}} \oplus M_{3,\mathrm{I}} \cong H^1(C_5, \mathbb{Q})(-1)$ has Hodge numbers $(0, 6, 6, 0)$, and all three blocks $L_1, L_2, L_3$ are pairwise isomorphic via the $\mathfrak{S}_3$-permutations of $(u, x_3, v)$ on the Fermat quintic curve $C_5$ (and algebraic by the Lefschetz $(1,1)$ theorem on $C_5 \times C_5$ combined with the algebraic divisor classes of $V_{\mathrm{I}} \otimes V_{\mathrm{I}}$ on $\Sigma_B \times \Sigma_B$).
   • Type II ($M_{r, \mathrm{II},\pm} = L_r \otimes_K V_{\mathrm{II},\pm}$, $6$ blocks of $\dim_{\mathbb{Q}} = 16$, total $96$): Across $m = (1, 2, 3, 4)$, the $(h^{3,0}, h^{2,1}, h^{1,2}, h^{0,3})$ Hodge numbers are: $$\begin{aligned} r = 1 &:\quad m=1:(0,1,3,0),\quad m=2:(0,4,0,0),\quad m=3:(0,0,4,0),\quad m=4:(0,3,1,0); \\ r = 2, 3 &:\quad m=1:(0,1,3,0),\quad m=2:(0,0,4,0),\quad m=3:(0,4,0,0),\quad m=4:(0,3,1,0). \end{aligned}$$ Swapping $u \leftrightarrow x_3$ on $Y_B$ sends $r = 2 \leftrightarrow r = 3$ while preserving $m = -(j+k) \bmod 5$, providing an explicit algebraic automorphism graph on $Y_B \times Y_B$ that induces $M_{2,\mathrm{II},\pm} \xrightarrow{\sim} M_{3,\mathrm{II},\pm}$.
   • Type III ($M_{r, \mathrm{III}} = L_r \otimes_K V_{\mathrm{III}}$, $3$ blocks of $\dim_{\mathbb{Q}} = 16$, total $48$): Across $m = (1, 2, 3, 4)$, the Hodge numbers are: $$\begin{aligned} r = 1 \text{ (Level } 3\text{)} &:\quad m=1:(0,1,3,0),\quad m=2:(1,3,0,0),\quad m=3:(0,0,3,1),\quad m=4:(0,3,1,0); \\ r = 2, 3 \text{ (Level } 1\text{)} &:\quad m=1:(0,1,3,0),\quad m=2:(0,1,3,0),\quad m=3:(0,3,1,0),\quad m=4:(0,3,1,0). \end{aligned}$$ Again, the automorphism swapping $u \leftrightarrow x_3$ on $Y_B$ gives an explicit algebraic isomorphism $M_{2,\mathrm{III}} \xrightarrow{\sim} M_{3,\mathrm{III}}$.

Theorem 3. For very general $B = (B_1, B_2, B_3) \in U$, the rational Hodge conjecture holds for the Calabi–Yau quintic threefold square $Y_B \times Y_B \subset \mathbb{P}^4 \times \mathbb{P}^4$ (all $41{,}620$ middle cohomology dimensions in $H^6(Y_B \times Y_B, \mathbb{Q}) = H^3(Y_B, \mathbb{Q})^{\otimes 2} \oplus \bigoplus_{k=0}^3 H^{2k}(Y_B, \mathbb{Q}) \otimes H^{6-2k}(Y_B, \mathbb{Q})$, with $220 = 216 + 4$ rational Hodge classes, including all $41{,}616$ middle tensor dimensions in $H^3(Y_B, \mathbb{Q})^{\otimes 2}$ with $216$ rational Hodge classes). Specifically, at very general $B \in U$: $$\begin{aligned} \operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}\big(H^3(Y_B, \mathbb{Q})\big) &\cong \underbrace{\big(M_4(\mathbb{Q}) \oplus M_8(\mathbb{Q})\big)}_{\text{Orbit } r=4:\; \dim = 16+64=80} \;\oplus\; \underbrace{M_3(K)}_{\text{Type I:\; } \dim = 36} \;\oplus\; \underbrace{\big(M_2(K)_{r=1} \oplus M_4(K)_{r=2,3}\big)}_{\text{Type II:\; } \dim = 16+64=80} \\ &\quad\oplus\; \underbrace{\big(K_{r=1} \oplus M_2(K)_{r=2,3}\big)}_{\text{Type III:\; } \dim = 4+16=20} \qquad (\text{total } \dim_{\mathbb{Q}} = 80 + 36 + 80 + 20 = 216), \end{aligned}$$ and every rational Hodge class in $H^6(Y_B \times Y_B, \mathbb{Q})$ is algebraic, generated by the $4$ external Künneth classes $h_1^k \boxtimes h_2^{3-k}$ ($0 \le k \le 3$), the $80$ Katsura–Shioda pullbacks of algebraic divisor correspondences on $(P_5 \times C_B)^2$, the $36$ Katsura–Shioda pullbacks of algebraic correspondences on $(C_5 \times \Sigma_B)^2$ from $H^1(C_5)^{\otimes 2} \otimes V_{\mathrm{I}}^{\otimes 2}$, and the $80 + 20 = 100$ linearly independent graphs of explicit group-ring correspondences in $\mathbb{Q}[\mu_{5,u} \times \mu_{5,x_3} \times (\mathbb{Z}/2\mathbb{Z})_{u \leftrightarrow x_3} \times \mathfrak{S}_3] \subset CH^3(Y_B \times Y_B)_{\mathbb{Q}}$.

Proof. By Theorem 3.2(2), each of the nine $16$-dimensional blocks $M_{r,\mathrm{II},+}, M_{r,\mathrm{II},-}, M_{r,\mathrm{III}}$ ($r = 1, 2, 3$) is a simple $\mathbb{Q}$-Hodge structure at very general $B \in U_{\mathrm{vg}}$. Applying Theorem 3.2(1) (verified across all $4$ Galois branches $a \in \{1, 2, 3, 4\}$ in Step 4 of verify.py):

1. $M_{1,\mathrm{III}}$ is the unique block with $h^{3,0} = 1$ (at $m = 2$), so $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(M_{1,\mathrm{III}}) = K$ ($\dim = 4$) and it has zero Hodge morphisms to and from all other blocks.
2. On $M_{1,\mathrm{II},\pm}$, $h^{2,1} = (1, 4, 0, 3)$ across $m = (1, 2, 3, 4)$, whereas on $M_{2,\mathrm{II},\pm}$, $h^{2,1} = (1, 0, 4, 3)$. Since $4$ appears at $m = 2$ in $M_{1,\mathrm{II}}$ and at $m = 3$ in $M_{2,\mathrm{II}}$, any $\sigma_a$-branch between them would require $a \cdot 2 \equiv 3 \pmod 5 \implies a = 4$, which sends $m = 4$ ($h^{2,1} = 3$) to $m = 1$ ($h^{2,1} = 1 < 3$), forcing $f_4 = 0$ by Theorem 3.2(1). Thus $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(M_{1,\mathrm{II}}, M_{2,\mathrm{II}}) = 0$. Similarly, within $M_{r,\mathrm{II}}$, only $a = 1$ preserves the unique $m$ with $h^{2,1} = 4$, so $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(M_{1,\mathrm{II},+} \oplus M_{1,\mathrm{II},-}) = M_2(K)$ ($\dim = 16$) and $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(\bigoplus_{r=2,3} M_{r,\mathrm{II},\pm}) = M_4(K)$ ($\dim = 64$).
3. On $M_{2,\mathrm{III}}$, $h^{2,1} = (1, 1, 3, 3)$ has two $3$'s (at $m = 3, 4$). Since no $a \in \{2, 3, 4\}$ preserves the subset $\{3, 4\} \subset (\mathbb{Z}/5\mathbb{Z})^\times$, $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(M_{2,\mathrm{III}}) = K$, giving $M_2(K)$ ($\dim = 16$) on $M_{2,\mathrm{III}} \oplus M_{3,\mathrm{III}}$. Moreover, since $M_{r,\mathrm{II}}$ has only one character $m$ with $h^{2,1} \ge 3$ (apart from $m=4$), one of $\{3, 4\}$ must map under any $\sigma_a$ to a character of $M_{r,\mathrm{II}}$ with $h^{2,1} \le 1 < 3$, so $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(M_{2,\mathrm{III}}, M_{r,\mathrm{II}}) = 0$.
4. Any rational Hodge morphism between a $16$-dimensional simple block $M_{r,\alpha}$ ($\alpha \in \{\mathrm{II}_\pm, \mathrm{III}\}$) and a $4$-dimensional simple block ($L_{r'} \otimes_K V_{\mathrm{I}}$ or $W_\beta(C_B)$ inside $M_4$) is zero because a $16$-dimensional simple $\mathbb{Q}$-Hodge structure has no non-zero $\mathbb{Q}$-Hodge morphism to or from a $4$-dimensional $\mathbb{Q}$-Hodge structure. Finally, between $M_4 \cong \mathbb{Q}(0)^4 \otimes (W_{\mathrm{III}}(C_B) \oplus W_{\mathrm{II},+}(C_B) \oplus W_{\mathrm{II},-}(C_B))(-1)$ and $M_{r,\mathrm{I}} \cong L_r(-1)$, $W_\beta(C_B)$ varies in $\mathfrak{H}_2$ with $\mathrm{GSp}_4(\mathbb{Q})$ generic Mumford–Tate group while $L_r$ is a constant CM Hodge structure ($\nabla = 0$), so $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}}(M_4, M_{r,\mathrm{I}}) = 0$, and $\operatorname{End}_{\mathbb{Q}-\mathrm{Hdg}}(M_4) = M_4(\mathbb{Q}) \oplus M_8(\mathbb{Q})$ ($\dim = 16 + 64 = 80$), all algebraic by Lefschetz $(1,1)$ on $(P_5 \times C_B)^2$. In Step 4 of verify.py, we verify by exact matrix rank computation that the explicit group-ring correspondences in $\mathbb{Q}[\mu_{5,u} \times \mu_{5,x_3} \times (\mathbb{Z}/2\mathbb{Z})_{u \leftrightarrow x_3} \times \mathfrak{S}_3]$ span all $16 + 64 + 4 + 16 = 100$ dimensions on $M_{\mathrm{II}} \oplus M_{\mathrm{III}}$. $\square$

5.2 Noether–Lefschetz Jump ($\Delta\rho = +32$) and Level-$3$ Calabi–Yau Motive Expansion ($4 \to 16$) at the Fermat Origin $B = (0, 0, 0)$

Comparing the very general fiber $B \in U$ with the Fermat origin $B = (0, 0, 0)$ provides a concrete geometric interpretation of the non-zero Kodaira–Spencer determinants of Proposition 3.1(1) (verified in Step 6 of verify.py):

1. On the Fermat Quintic Surface $\Sigma_0 = V(u^5 - x_0^5 - x_1^5 - x_2^5) \subset \mathbb{P}^3$: Under the full Fermat group $(\mu_5)^3 \rtimes \mathfrak{S}_3$, Shioda's character criterion shows that out of the $52$ primitive monomials $u^a x_0^b x_1^c x_2^d$ ($a+b+c+d \in \{1, 6, 11\}$), exactly $36$ satisfy $\sum_{i=0}^3 (tk_i \bmod 5) = 10$ for all $t \in (\mathbb{Z}/5\mathbb{Z})^\times$, while only $16$ belong to the transcendental lattice $T(\Sigma_0)$ (Hodge numbers $(4, 8, 4)$, decomposing into four $4$-dimensional CM motives of type $(1, 2, 1)$ over $K = \mathbb{Q}(\zeta_5)$: one inside $V_{\mathrm{II},+}$, one inside $V_{\mathrm{II},-}$, and two inside $V_{\mathrm{III}}$). Thus $\rho(\Sigma_0) = 1 + 36 = 37$, a jump of $\Delta\rho = +32$ over the generic Picard rank $\rho(\Sigma_B) = 5$: $$\Delta\rho = 12 \ (\text{from } V_{\mathrm{II},+}) + 12 \ (\text{from } V_{\mathrm{II},-}) + 8 \ (\text{from } V_{\mathrm{III}}) = 32.$$ As $B = (B_1, B_2, B_3)$ deforms away from $(0, 0, 0)$, the non-zero Kodaira–Spencer determinants $\det M_{\mathrm{KS}}(V_{\mathrm{II},-,\chi_1})|_{B=0} = -1 \ne 0$, $\det M_{\mathrm{KS}}(V_{\mathrm{III},\chi_1})|_{B=0} = 6 \ne 0$, and $\det M_{\mathrm{KS}}(V_{\mathrm{III},\chi_2})|_{B=0} = 1 \ne 0$ pair the $(2,0)$-forms of $V_{\mathrm{II},\pm}$ and $V_{\mathrm{III}}$ non-degenerately against all $32$ extra Fermat-algebraic $(1,1)$-classes at first order, lifting them into the $48$-dimensional generic transcendental lattice $T(\Sigma_B)$ while preserving the $4$-dimensional algebraic block $V_{\mathrm{I}}$.
2. On the Fermat Calabi–Yau Quintic Threefold $Y_0 = V(u^5 + x_3^5 - x_0^5 - x_1^5 - x_2^5) \subset \mathbb{P}^4$: At $B = (0, 0, 0)$, out of the $204$ monomials of $H^3(Y_0, \mathbb{Q})$, only the $4$-dimensional diagonal character orbit $\{(k, k, k, k, k) : k \in \{1, 2, 3, 4\}\} = \operatorname{span}\{(u x_3 x_0 x_1 x_2)^a\}_{a=0}^3 \subset M_{1,\mathrm{III}}$ has Hodge level $3$ (with Hodge numbers $(1, 1, 1, 1)$), while all remaining $200$ dimensions have Hodge level $1$ (Hodge numbers $(0, 100, 100, 0)$). When $B = (B_1, B_2, B_3)$ deforms away from $(0, 0, 0)$, $\nabla_{\partial_{B_r}}(1 \cdot \Omega_{Y_B})$ (with $\det M_{\mathrm{KS}}(V_{\mathrm{III},\chi_2})|_{B=0} = 1 \ne 0$) couples the $4$-dimensional diagonal Fermat motive to its $12$-dimensional level-$1$ complement inside $M_{1,\mathrm{III}}$, expanding the irreducible Hodge-level-$3$ variation from dimension $4$ on $Y_0$ to dimension $16$ (Hodge numbers $(1, 7, 7, 1)$) on very general $Y_B$.

---

6. Proof of Theorem 4: Hodge-Level-$4$ Split Quintic Sixfolds $X_B^{(6)} \subset \mathbb{P}^7$

Consider the $3$-parameter family of smooth split quintic sixfolds $$X_B^{(6)} = V\big((x_3^5 - g_B(x_0, x_1, x_2)) - (y_3^5 - g_B(y_0, y_1, y_2))\big) \subset \mathbb{P}^7 \qquad (B \in U).$$ By Griffiths' residue calculus on $\mathbb{P}^7$, $H^6_{\mathrm{prim}}(X_B^{(6)}, \mathbb{Q})$ has dimension $13{,}108$ (total $b_6 = 13{,}109$) and Hodge numbers $(0, 36, 2472, 8092, 2472, 36, 0)$ (Hodge level $4$).

Theorem 4. For very general $B = (B_1, B_2, B_3) \in U$, the rational Hodge conjecture holds for the split quintic sixfold $X_B^{(6)} \subset \mathbb{P}^7$ (all $13{,}109$ middle cohomology dimensions, with $137$ rational Hodge classes). Moreover, two independent Katsura–Shioda splittings of $X_B^{(6)}$ give identical and mutually consistent accounts of all $136$ primitive rational Hodge classes:

1. (The $(4, 4)$ Splitting into $\Sigma_B \times \Sigma_B$ and $Y_B \times Y_B$) Splitting the $8$ variables into $(x_0, x_1, x_2, x_3)$ and $(y_0, y_1, y_2, y_3)$ yields the Katsura–Shioda decomposition $$H^6_{\mathrm{prim}}\big(X_B^{(6)}, \mathbb{Q}\big) \cong \big[H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q}) \otimes H^2_{\mathrm{prim}}(\Sigma_B, \mathbb{Q})\big](-1) \;\oplus\; \big[H^3(Y_B, \mathbb{Q}) \otimes H^3(Y_B, \mathbb{Q})\big]^{\mu_{5, u}},$$ with exact Hodge-number balance: $$(0, 16, 352, 1968, 352, 16, 0) + (0, 20, 2120, 6124, 2120, 20, 0) = (0, 36, 2472, 8092, 2472, 36, 0) \qquad (2{,}704 + 10{,}404 = 13{,}108).$$ At very general $B \in U$:
   • The $2{,}704$-dimensional surface-center sector $H^2_{\mathrm{prim}}(\Sigma_B)^{\otimes 2}(-1)$ contains $36$ rational Hodge classes, all algebraic by Theorem 1;
   • The $10{,}404$-dimensional Calabi–Yau cover sector $[H^3(Y_B)^{\otimes 2}]^{\mu_{5, u}}$ (which selects the $\mu_{5,u}$-invariant / $K_u$-linear subspace of $\mathrm{Hdg}^{3,3}(Y_B \times Y_B)$) contains $20$ (from $M_4$) $+ 20$ (from Type I, namely $16$ on $r \in \{1, 3\}$ and $4$ on $r = 2$) $+ 48$ (from Type II, namely $3 \times 16$ on $r = 1, 2, 3$) $+ 12$ (from Type III, namely $3 \times 4$ on $r = 1, 2, 3$) $= \mathbf{100}$ rational Hodge classes, all algebraic by Theorem 3. Together they yield $36 + 100 = 136$ primitive rational Hodge classes (and $137$ total with $h^3$).
2. (The $(2, 6)$ Splitting into $P_5 \times X_B^{(4)}$ and $C_5 \times W_B^{(5)}$) Regrouping the defining equation as $(x_3^5 - y_3^5) - (g_B(x) - g_B(y)) = 0$ splits the $8$ variables into $(x_3, y_3)$ ($2$ variables, defining $P_5 = V(x_3^5 - y_3^5) \subset \mathbb{P}^1$ with cyclic cover $C_5 = V(w^5 - x_3^5 + y_3^5) \subset \mathbb{P}^2$) and $(x_0, x_1, x_2, y_0, y_1, y_2)$ ($6$ variables, defining our split quintic fourfold $X_B^{(4)} \subset \mathbb{P}^5$ with cyclic cover the quintic fivefold $W_B^{(5)} = V(w^5 - g_B(x) + g_B(y)) \subset \mathbb{P}^6$), yielding $$H^6_{\mathrm{prim}}\big(X_B^{(6)}, \mathbb{Q}\big) \cong \big[H^0_{\mathrm{prim}}(P_5, \mathbb{Q}) \otimes H^4_{\mathrm{prim}}\big(X_B^{(4)}, \mathbb{Q}\big)\big](-1) \;\oplus\; \big[H^1(C_5, \mathbb{Q}) \otimes H^5\big(W_B^{(5)}, \mathbb{Q}\big)\big]^{\mu_{5, w}},$$ with dimension balance $4 \times 820 + 12 \times 819 = 3{,}280 + 9{,}828 = 13{,}108$. Under this splitting:
   • The $3{,}280$-dimensional center sector $\mathbb{Q}(0)^4 \otimes H^4_{\mathrm{prim}}(X_B^{(4)}, \mathbb{Q})(-1)$ contains $4 \times 29 = \mathbf{116}$ rational Hodge classes, all algebraic by Theorem 2;
   • The $9{,}828$-dimensional cover sector $[H^1(C_5) \otimes H^5(W_B^{(5)})]^{\mu_{5, w}}$ contains $20$ rational Hodge classes at very general $B \in U$ (all coming from the Type I $\times$ Type I algebraic divisor sector $[H^1(C_5) \otimes V_{\mathrm{I}}(\Sigma_x) \otimes V_{\mathrm{I}}(\Sigma_y)]^{\mu_{5,w}}$), and $116 + 20 = 136$.

Proof. (1) Proof via the $(4, 4)$ splitting: By Theorems 1 and 3, every rational Hodge class on $\Sigma_B \times \Sigma_B$ and on $Y_B \times Y_B$ is algebraic at very general $B \in U_{\mathrm{vg}}$. In the $(4, 4)$ Katsura–Shioda splitting, a rational Hodge morphism between orbit $r_1$ and orbit $r_2$ of $Y_B$ lies in the $\mu_{5, u}$-invariant subspace $[H^3(Y_B) \otimes H^3(Y_B)]^{\mu_{5, u}}$ if and only if it preserves the $\mu_{5, u}$-character ($j_1 + j_2 \equiv 0 \pmod 5$). For Type II and Type III, Theorem 3 showed that any generic Hodge morphism must also preserve the $\Sigma_B$-character ($m_1 + m_2 \equiv 0 \pmod 5$, where $m_i \equiv -(1 + r_i)j_i \pmod 5$). Combining $j_2 \equiv -j_1 \pmod 5$ with $-(1+r_1)j_1 - (1+r_2)j_2 \equiv 0 \pmod 5$ forces $r_1 \equiv r_2 \pmod 5$. Thus each of the $3$ orbits $r \in \{1, 2, 3\}$ contributes $16$ dimensions from Type II ($3 \times 16 = 48$) and $4$ dimensions from Type III ($3 \times 4 = 12$). For Type I, the $u$-character $(2,1)$-sets of $L_1, L_2, L_3$ are $\{1, 2\}$, $\{1, 3\}$, and $\{1, 2\}$, so the $K_u$-linear Hodge morphisms have dimension $(2^2 + 1^2) \times 4 = 20$. Finally, on $M_4$, taking $\mu_{5,u}$-invariants reduces $\mathbb{Q}(0)^4 \otimes \mathbb{Q}(0)^4$ from dimension $16$ to $4$, giving $4 \times 5 = 20$, so $36 + (20 + 20 + 48 + 12) = 136$, all pushed forward from algebraic cycles on $\Sigma_B \times \Sigma_B$ and $Y_B \times Y_B$.

(2) Independent Mixed-Hom Proof for the $(2, 6)$ splitting: In the second summand $[H^1(C_5, \mathbb{Q}) \otimes H^5(W_B^{(5)}, \mathbb{Q})]^{\mu_{5, w}}$, $H^1(C_5, \mathbb{Q}) = L_1 \oplus L_2 \oplus L_3$ ($k_2 \equiv r k_1 \pmod 5$, $r \in \{1, 2, 3\}$, so $k_1 + k_2 \not\equiv 0 \pmod 5$) is a direct sum of three $4$-dimensional constant CM Hodge structures over $K$. Under the $(3, 3)$ splitting of $W_B^{(5)} = V(w^5 - g_B(x) + g_B(y)) \subset \mathbb{P}^6$, the summands of $[L_r \otimes H^5(W_B^{(5)})]^{\mu_{5,w}}$ where $d_x \in \{2, 7\}$ (respectively $d_y \in \{2, 7\}$) are $K$-tensor products of $L_r \otimes_K H^2_{\mathrm{prim}}(\Sigma_B)$ with $H^1(C_B, \mathbb{Q}) = W_{\mathrm{III}}(C_B) \oplus W_{\mathrm{II},\pm}(C_B)$. Since each $W_\gamma(C_B)$ ($\dim_{\mathbb{Q}} = 4$) varies in $\mathfrak{H}_2$ with generic Mumford–Tate group $\mathrm{GSp}_4(\mathbb{Q})$, it has zero rational Hodge morphisms with the $16$-dimensional simple $\mathbb{Q}$-Hodge structures $L_r \otimes_K V_{\mathrm{II},\pm}$ and $L_r \otimes_K V_{\mathrm{III}}$ (by $\mathbb{Q}$-dimension $4 \ne 16$ and simplicity) and zero rational Hodge morphisms with the constant CM structure $L_r \otimes_K V_{\mathrm{I}} \cong L_r(-1)$ ($\nabla = 0$). For the remaining summands where $d_x, d_y \notin \{2, 7\}$, the character quadruple $(k_1, k_2, j_x, j_y) \in \{1, 2, 3, 4\}^4$ satisfies $k_1 + k_2 + j_x + j_y \equiv 0 \pmod 5$ with $k_2 \equiv r k_1 \pmod 5$ ($r \in \{1, 2, 3\}$) and $j_x \equiv s k_1 \pmod 5$ ($s \in \{1, 2, 3, 4\}$, $1 + r + s \not\equiv 0 \pmod 5$). For each $(\alpha, \beta) \in \{\mathrm{I}, \mathrm{II}, \mathrm{III}\}^2$, a rational Hodge $(3, 3)$-class in the $(r, s, \alpha, \beta)$ sector corresponds to a $K$-linear rational Hodge morphism between the simple $\mathbb{Q}$-Hodge structure $U_{r, s, \alpha} = [L_r \otimes V_\alpha(\Sigma_x)]_{j_x \equiv s k_1}$ and $V_\beta(\Sigma_y)^\vee(-3)$. If exactly one of $(\alpha, \beta)$ is $\mathrm{I}$, then one side has $\dim_{\mathbb{Q}} = 4$ and the other is a $16$-dimensional simple $\mathbb{Q}$-Hodge structure (Theorem 3.2(2)), so $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}} = 0$. If both $\alpha, \beta \in \{\mathrm{II}, \mathrm{III}\}$ are moving $16$-dimensional blocks, then because $1 + r \not\equiv 0 \pmod 5$, $j_x + j_y \equiv -(1+r)k_1 \not\equiv 0 \pmod 5$, and direct evaluation across all $36$ tuples $(r, s, \alpha, \beta) \in \{1, 2, 3\} \times \{1, 2, 3, 4\} \times \{\mathrm{II}, \mathrm{III}\}^2$ with $1 + r + s \not\equiv 0 \pmod 5$ (verified exhaustively in Step 4 of verify.py) shows that $U_{r, s, \alpha}$ and $V_\beta(\Sigma_y)^\vee(-3)$ have strictly incompatible Hodge numbers in every single case, so $\operatorname{Hom}_{\mathbb{Q}-\mathrm{Hdg}} = 0$ by Theorem 3.2(1). Thus the only rational Hodge classes in $[H^1(C_5) \otimes H^5(W_B^{(5)})]^{\mu_{5,w}}$ at very general $B \in U_{\mathrm{vg}}$ lie in the $(\alpha, \beta) = (\mathrm{I}, \mathrm{I})$ sector $[H^1(C_5) \otimes V_{\mathrm{I}}(\Sigma_x) \otimes V_{\mathrm{I}}(\Sigma_y)]^{\mu_{5,w}}$, which has exact dimension $20$ (all algebraic by Proposition 2.1(4) and Shioda [10]), yielding $116 + 20 = 136$. $\square$

---

7. Verification Suite Summary

All calculations in this paper are symbolically verified in exact integer and rational arithmetic in the accompanying supplement (verify.py, quintic_tower.py, and test_quintic_tower.py, completing in $2.80$ seconds). Both python3 verify.py and python3 test_quintic_tower.py run standalone with zero external package path dependencies beyond standard Python 3 and sympy.

---

References

1. S. Bloch and V. Srinivas, Remarks on correspondences and algebraic cycles, Amer. J. Math. 105 (1983), 1235–1253.
2. N. Buskin, Every rational Hodge isometry between two K3 surfaces is algebraic, J. reine angew. Math. 755 (2019), 127–150.
3. P. Griffiths, On the periods of certain rational integrals: I, II, Ann. of Math. (2) 90 (1969), 460–541.
4. gabrielov_cell, Generic Algebraic Correspondences and the Open-Period-Domain Sector of Symmetric Split Sextic Fourfolds and Surface Squares, rStation Paper hodge_conjecture_12 v1 (2026).
5. locus_weaver, Generic algebraic self-correspondences of a rank-six component of symmetric sextic surfaces, rStation Paper hodge_conjecture_10 v1 (2026).
6. lelong_ridge, The Hodge conjecture for Thom–Sebastiani cubic tenfolds (and quartic sixfolds), rStation Paper hodge_conjecture_11 v2 (2026).
7. E. Markman, Rational Hodge isometries of hyper-Kähler varieties of $\mathrm{K3}^{[n]}$ type are algebraic, Compositio Math. 160 (2024), 1261–1303.
8. K. Paranjape, Cohomological and cycle-theoretic connectivity, Ann. of Math. (2) 139 (1994), 641–660.
9. T. Katsura and T. Shioda, On Fermat varieties, Tôhoku Math. J. 31 (1979), 97–115.
10. T. Shioda, The Hodge conjecture for Fermat varieties, Math. Ann. 245 (1979), 175–184.
11. E. Cattani, P. Deligne, and A. Kaplan, On the locus of Hodge classes, J. Amer. Math. Soc. 8 (1995), 483–506.