Associative Yang-Baxter equation and Fukaya categories of square-tiled surfaces (Q1711914)

The article under review is very explicit. It starts out with the precise formulation of the associative Yang-Baxter equation (AYBE) \[ r^{12}(-u^\prime,v)r^{13}(u+u^\prime,v+v^\prime)-r^{23}(u+u^\prime,v^\prime)r^{12}(u,v)+r^{13}(u,v+v^\prime)r^{23}(u^\prime,v^\prime)=0, \] where \(r:\mathbb C^2\rightarrow\operatorname{Mat}_n(\mathbb C)\otimes\operatorname{Mat}_n(\mathbb C)\) is a meromorphic function of two complex variables \((u,v)\) near \((0,0)\) with values in \(\operatorname{Mat}_n(\mathbb C)\otimes\operatorname{Mat}_n(\mathbb C).\) The notation is \(r^{12}=r\otimes 1\in\operatorname{Mat}_n(\mathbb C)\otimes\operatorname{Mat}_n(\mathbb C)\otimes\operatorname{Mat}_n(\mathbb C)\) etc. Usually, the AYBE is coupled with the skew-symmetry (unitary) condition \(r^{12}(-u,-v)=-r(u,v),\) \(r^{21}\) obtained from \(r\) by the transposition of tensor factors \(a_2\otimes a_1\rightarrow a_1\otimes a_2.\) The AYBE is an associative analog of the classical Yang-Baxter equation (CYBE) \[ [r^{12}(v),r^{13}(v+v^\prime)]+[r^{12}(v),r^{23}(v^\prime)]+[r^{13}(v+v^\prime),r^{23}(v^\prime)]=0, \] with \(r(v)\) a meromorphic function near \(0\) with values in a Lie algebra \(\mathfrak g\otimes\mathfrak g.\) This CYBE results in Poisson-Lie groups and classical integrable systems, and the AYBE comes from the CYBE with \(\mathfrak g=\mathfrak{sl}_n(\mathbb C).\) \textit{A. Polishchuk} [Adv. Math. 168, No. 1, 56--95 (2002; Zbl 0999.22023)] found that solutions of the AYBE often comes from 1-Calabi-Yau \(A_\infty\)-categories. The introduction of the present article contains an explicit construction from a minimal \(A_\infty\)-category with two selected sets of isomorphism classes of objects \(\mathcal X\) and \(\mathcal Y\) with special properties that leads to the general AYBE \[ (r^{x_2x_1}_{y_2y_3})^{12}(r^{x_1x_3}_{y_1y_3})^{13}-(r^{x_1x_3}_{y_1y_2})^{23}(r^{x_2x_3}_{y_2y_3})^{12}+(r^{x_2x_3}_{y_1y_3})^{13}(r^{x_1x_2}_{y_1y_2})^{23}=0. \] In the case when \(\mathcal C\) is the derived category of coherent sheaves on a degeneration of an elliptic curve, there exist natural choices of \(\mathcal X\) and \(\mathcal Y\) with abelian group structures such that the general AYBE gets the previously described form. The authors prove that there is a bijection between formal solutions of the general AYBE and a class of \(A_\infty\)-structures given by deformations of the formal \(A_\infty\)-category \(\mathcal A=\mathcal A_n\) with two objects \(X, Y\) and: \(\operatorname{Hom}(X,Y)=\operatorname{Hom}^0(X,Y)=\mathbb Z\theta_1\oplus\dots\oplus\mathbb Z\theta_n,\) \(\operatorname{Hom}(Y,X)=\operatorname{Hom}^1(Y,X)=\mathbb Z\eta_1\oplus\dots\oplus\mathbb Z\eta_n,\) \(\operatorname{Hom}^0(X,X)=\mathbb Z\operatorname{id}_X,\) \(\operatorname{Hom}^0(Y,Y)=\mathbb Z\operatorname{id}_Y,\) \(\operatorname{Hom}^1(X,X)=\mathbb Z\xi_X,\) \(\operatorname{Hom}^1(Y,Y)=\mathbb Z\xi_Y.\) The elements \(\operatorname{id}_X,\operatorname{id}_Y\) acts as strict units, and \(\mathcal A\) can be viewed as a graded category. Also, the symmetric perfect pairing on the \(\operatorname{Hom}\)-spaces of \(\mathcal A\) is explicitly defined. Let \(\mathbf k\) be a field. The article considers \(A_\infty\)-structures on \(\mathcal A\otimes\mathbf k\) with given \(\mathfrak m_2,\) cyclic with respect to the pairing. A strictly cyclic \(A_\infty\)-category of dimension \(1\) is a strictly unital, proper \(A_\infty\)-category with nondegenerate pairings \(\langle a_1,a_2\rangle:\hom^\ast(X,Y)\otimes\hom^{1-\ast}(Y,X)\rightarrow\mathbf k\) such that \[ \langle a_1,a_2\rangle=(-1)^{(|a_1|-1)(|a_2|-1)+1}\langle a_2,a_1\rangle \] together with the cyclic symmetry condition \begin{align*} & \langle a_{k+1},\mathfrak m_k(a_k,a_{k-1},\dots,a_1)\rangle\\ =(-1)^{(|a_{k+1}|-1)(|a_1|+|a_2|+\dots +|a_k|-k)} & \langle a_k,\mathfrak m_k(a_{k-1},a_{k-2},\dots,a_1,a_{k+1})\rangle\end{align*} For a commutative \(\mathbf k\)-algebra \(R\), let \(\mathcal M_\infty(\mathcal A\otimes R)\) be the set of cyclic, strictly unital, minimal \(A_\infty\)-structures on \(\mathcal A\otimes R\) upto a strict cyclic \(A_\infty\)-equivalence. Put \(P=\sum_{i,j}e_{ij}\otimes e_{ji}\in\operatorname{Mat}_n(\mathbf k)\otimes\operatorname{Mat}_n(\mathbf k),\) this says that \(P\) is the transposition operator given by \(P(x\otimes y)=y\otimes x.\) Now, the first main result of the paper is that there exists a natural explicit bijection between \(\mathcal M_\infty(\mathcal A\otimes R)\) and the equivalence classes of formal skew-symmetric solutions \(r^{x_1x_2}_{y_1y_2}\) of the general AYBE of explicitly given type. The key idea of the proof is to use a version of the construction of the AYBE to a pair \(\mathcal X,\mathcal Y\) of formal deformations of objects \(X\) and \(Y\) in \(\mathcal A.\) In fact, the authors have to apply a kind of triple Massey products instead of just \(\mathfrak m_3.\) In a seminal paper, \textit{A. A. Belavin} and \textit{V. G. Drinfel'd} [Funct. Anal. Appl. 16, 159--180 (1983; Zbl 0504.22016); translation from Funkts. Anal. Prilozh. 16, No. 3, 1--29 (1982)] classified nondegenerate solutions of the classical Yang-Baxter equation for simple complex Lie algebras. They proved that they can be either elliptic, trigonometric or rational. They classified the trigonometric solutions in terms of combinatorial data, involving Belavin-Drinfeld triples. Similarly, the authors pose the problem of classifying nondegenerate solutions \(r(u,v)\) of the AYBE. If \(r\) is strongly nondegenerate, the Laurent expansion of the solution at \(u=0\) has the form \(r(u,v)=\frac{1\otimes 1}{u}+ r_0(v)+\dots,\) and it was shown that the projection \(\overline r_0(v)\) of \(r_0(v)\) to \(\mathfrak{sl}_n(\mathbb C)\otimes\mathfrak{sl}_n(\mathbb C)\) is a nondegenerate solution of the CYBE, and that if \(r_0(v)\) is either elliptic or trigonometric, then \(r(u,v)\) is determined by \(\overline r_0(v).\) Polischuk [loc. cit.] proved that all elliptic solutions of the CYBE extend to those of the AYBE. Also, \textit{T. Schedler} [Math. Res. Lett. 10, No. 2--3, 301--321 (2003; Zbl 1053.17013)] observed that this is not the case for trigonometric solutions. The authors ask which solutions of the AYBE that can be realized by families \((\mathcal X,\mathcal Y)\) of objects in some geometric 1-Calabi-Yau categories. They have a source in the derived categories of coherent sheaves on elliptic curves and their degenerations where one can take \(\mathcal X\) as a universal deformation of a simple vector bundle and \(\mathcal Y\) as the family of the structure sheaves of the points, and they find that all the solutions of the AYBE for which \(\overline r_0(v)\) is elliptic arise from simple vector bundles on elliptic curves and that they can be explicitly computed in terms of elliptic functions, all the solutions coming from the nodal degenerations of elliptic curves were shown to be trigonometric. What was not expected was that not all trigonometric solutions of the AYBE appear in this way. In fact trigonometric solutions corresponding to the data \((S,C_1,C_2,A)\) comes from a simple vector bundle on a cycle of projective lines if and only if the corresponding transitive permutations \(C_1\) and \(C_2\) commute. The authors solve the problem of finding other 1-Calabi-Yau \(A_\infty\)-categories and objects in them which give the missing solutions by looking at appropriate Fukaya categories. Starting from an associative Belavin-Drinfeld structure, they construct a square-tiled surface \(\Sigma\) with a local symplectomorphism \(\pi:\Sigma\rightarrow\mathbb T\) to the square torus. When \(A=\emptyset,\) \(\Sigma\) is the \(n\)-fold covering space of the punctured torus \(\mathbb T_0\) associated to the permutations \(C_1,\;C_2.\) Lifts of standard Lagrangian curves in \(\mathbb T\) to \(\Sigma\) give a pair of exact Lagrangians \(L_1,L_2\) in \(\Sigma\) such that \(\bigoplus_{i,j=1}^2\operatorname{HF}^\ast(L_i,L_j)\simeq\mathcal A\otimes\mathbb C.\) Complex push-offs of the Lagrangians \(L_1\) and \(L_2\) give \(1\)-parameter families \(L_1^x\) and \(L_2^y\) which taken as families \(\mathcal X\) and \(\mathcal Y\) in the general construction of solutions of the AYBE, results in a solutions recording triple products between \((L_1^{x_1},L_2^{y_1},L_1^{x_2},L_2^{y_2}).\) This gives exactly the trigonometric solution of the AYBE associated with \((S,C_1,C_2,A),\) and this is a main result in the article. For the final main result, the surface \(\Sigma\) has genus \(1\) if and only if \(C_1\) and \(C_2\) commute. This explains why only such solutions comes from simple vector bundles on nodal degenerations of elliptic curves, and as an application one gets: Let \(C\) be a cycle of projective lines over \(\mathbb C.\) For any simple vector bundle \(V\) on \(C\) there exists a composition \(\Phi\) of \(1\)-spherical twists and their inverses such that \(\Phi(\mathcal O_C)\simeq V.\) The article is explicit and with good examples, and the authors prove that in this special situation, the solution of the formal general AYBE associated with the pair \((\mathcal O,V),\) respectively \((L_1,L_2)\) split generates the perfect derived category of \(C\) (resp. the Fukaya category of \(\Sigma\)). Thus the main technique is to reduce the problem to a similar question about Lagrangians in the Fukaya category where they can use the action of the mapping class group.