Double box motive (Q2035735)

The paper concerns the concept of periods of motives constructed from graphs. It employs a conventional combinatorial way to define homogeneous hypersurfaces by assigning indeterminates to edges of graphs. Then one is interested the mixed Hodge structure (MHS) of the complement of the hypersurface in the projective space. These motives and their periods also have origins in physics by the amplitudes of the statistical generating series related to Feymann diagrams. The periods of the MHS also appear as interesting (quasi)-modular forms. The family of motives defined by graphs also show very attractive singular families with dynamical locus of singularities. This perhaps makes the analysis of the MHS of cohomology of the fibres more complex, however most of the methods of variation of mixed Hodge structure (VMHS) is applicable to them. The MHS of the complement (or singular) variety is explained by the double complex associated to a stratification of the projective completion (or in the blowup) of variety, worked out by P. Deligne and J. Steenbrink and others, see [\textit{P. Deligne}, Equations différentielles à points singuliers réguliers. Cham: Springer (1970; Zbl 0244.14004); \textit{C. A. M. Peters} and \textit{J. H. M. Steenbrink}, Mixed Hodge structures. Berlin: Springer (2008; Zbl 1138.14002)]. Many examples of these motives appear in the author former works [J. Number Theory 148, 328--364 (2015; Zbl 1319.81044)], see also [\textit{D. Huybrechts}, ``The geometry of cubic hypersurfaces'', Preprint (2020)]. The author specifically considers the motive associated to the complement of the following graph polynomial \[ X_{3,1,3}:0=\Psi_{3,1,3}=Q(x_5,x_6,x_7)(x_1+x_2+x_3+x_4)+Q'(x_1,x_2,x_3)(x_4+x_5+x_6+x_7)+x_4 A\tag{1} \] where \(A\) has degree 2 and is linear separately in each variable \(x_1, \dots, x_7\). In the text the polynomial \(\Psi_{m,1,n}\) has been explained in the appendix as the second Symanzik polynomial, [Proposition A1], see also [\textit{D. Kreimer}, ``The master two-loop two-point function. The general case'', Phys. Lett. B 273, 277--281 (1991)]. The variety \(X_{3,1,3}\) is a cubic hypersurface in \(\mathbb{P}^6\), i.e it has dimension \(5\). The interesting cohomology of \(X_{3,1,3}\) is the middle cohomology \(H^5(X_{3,1,3}, \mathbb{Q})\). The MHS has all weights \(\leq 5\). The main result of the paper is \[ \mathrm{Gr}_5^WH^5(X_{3,1,3}, \mathbb{Q}) \cong H^1(E, \mathbb{Q}(-2))\tag{2} \] where \(E\) is an elliptic curve. The proof has 3 steps \begin{itemize} \item[1.] to construct resolution of singularities \(Z \to X_{3,1,3}\) and show \(F^3H^5(Z, \mathbb{C})=\mathbb{C}\), [Theorem 5.1] \item[2.] to show \(F^4H^5(Z, \mathbb{C})=0\) and deduce \(H^5(Z)\) is HS of type \((0,0,1,1,0,0)\), [Proposition 6.1]. \item[3.] to show \(H^3(Z, \mathbb{Q})=\mathrm{Gr}_5^WH^5(X_{3,1,3}, \mathbb{Q})\), [Proposition 6.2]. \end{itemize} In Proposition 2.1 the singular locus of \(X_{3,1,3}\) is described as \(C \sqcup C' \sqcup S\) where \[ \begin{aligned} C &:Q(x_5, x_6 , x_7)=0=x_1=x_2=x_3=x_4,\\ C' &:Q'(x_1, x_2, x_3)=0=x_4=x_5=x_6=x_7 \end{aligned}\tag{3} \] and \(S\) is a finite number of ordinary double points on \(X_{3,1,3}\) disjoint from \(C\) and \(C'\). In Section 3 the author builds up the blow up \(Z\) to calculate the MHS through the spectral sequence. The blow up diagram is \[ \begin{tikzcd} Z \arrow[r] \arrow[d] & W \arrow[d, "\sigma=\operatorname{blow}S"] \\ Y \arrow[r] \arrow[d] & V \arrow[d, "\pi=\operatorname{blow}C\sqcup C'"] \\ X_{3,1,3} \arrow[r] &\mathbb{P}^6 \end{tikzcd}\tag{4} \] The calculation of \(F^3H^5(Z)\) uses the Leray spectral sequence of \(\pi:W \to \mathbb{P}^6\). The Hodge filtration on \(H^6(W-Z)\) is given by pole order filtration. Because \(W\) is obtained by blowing up points and smooth rational curves all cohomology classes on \(W\) are Hodge. By the Gysin sequence \[ F^3H^5(Z) \cong F^4H^6(W-Z) \cong \mathbb{H}^2[W, (\Omega_W^4(Z) \stackrel{d}{\rightarrow} \Omega_W^5(Z) \stackrel{d}{\rightarrow} \Omega_W^6(Z))]\tag{5} \] Proposition 4.1 states that \[ R^i\pi_*\Omega_W^j(kZ)=0, \qquad i,k \geq 1, j \geq 0\tag{6} \] The proof uses the fact that \(R^i\pi_*\Omega_W^j(kZ)\) is killed by the ideal of functions on \(C \sqcup C' \sqcup S\), [Lemma 4.2]. It follows that in order to calculate \(\pi_*\Omega_W^{5}(mZ)\), one needs to finally calculate \(\pi_*\Omega_W^{3+m}(mZ)\) for \(m=1,2,3\). The explicit analysis and computation of \(F^3H^5(Z, \mathbb{C})\) is given in Section 5. In Section 6 the author proves the vanishing of \(F^4H^3(Z, \mathbb{C})\). The proof is algebraic geometric and uses explicit analysis of the exceptional divisors. The proof of the last step namely \(H^3(Z, \mathbb{Q})=\mathrm{Gr}_5^WH^5(X_{3,1,3}, \mathbb{Q})\) is based on the Leray spectral sequence of \(\pi_Z=\pi|_Z:Z \to X_{3,1,3}\) as \[ E_2^{p,q}=H^p(X_{3,1,3}, R^q\pi_{Z,*}\mathbb{Q}_Z) \Rightarrow H^{p+q}(Z, \mathbb{Q}).\tag{7} \] The author indicate that \(E^{3,0}=H^5(X_{3,1,3}, \mathbb{Q})\) and calculates the MHS by suitable stratification of \(X_{3,1,3}\).