Biased \(2 \times 2\) periodic Aztec diamond and an elliptic curve (Q6095837)

A domino tiling of the Aztec diamond can be viewed as a perfect matching, also called dimer configuration, on the Aztec diamond graph. By putting weights on the edges of the Aztec diamond graph, one defines a probability measure on the set of all perfect matchings, and hence all domino tilings, by saying that the probability of having a particular matching is proportional to the product of the weights of the edges in that matching. In recent years, several works have appeared on domino tilings of the Aztec diamond where the weights are doubly periodic (that is, they are periodic in two independent directions). This paper studies a particular example of a \(2\times 2\) doubly periodic weighting that is a generalization of the model studied in [\textit{V. Beffara} et al., Ann. Probab. 46, No. 5, 2973--3013 (2018; Zbl 1428.60063)] and [\textit{M. Duits} and \textit{A. B. J. Kuijlaars}, J. Eur. Math. Soc. (JEMS) 23, No. 4, 1075--1131 (2021; Zbl 1487.60018)]. The difference is that the authors of the present paper introduce an extra parameter that induces a bias towards horizontal dominos, while model considered in [\textit{V. Beffara} et al., Ann. Probab. 46, No. 5, 2973--3013 (2018; Zbl 1428.60063)] and [\textit{M. Duits} and \textit{A. B. J. Kuijlaars}, J. Eur. Math. Soc. (JEMS) 23, No. 4, 1075--1131 (2021; Zbl 1487.60018)] will be referred to as the unbiased \(2\times 2\) periodic Aztec diamond. From general arguments it follows that the correlation functions in the model are determinantal. In order to perform a rigorous asymptotic study, one aims to find an expression for the correlation kernel that is amenable for an asymptotic analysis. For the unbiased \(2\times 2\) periodic Aztec diamond, a double integral representation was first found in [\textit{S. Chhita} and \textit{K. Johansson}, Adv. Math. 294, 37--149 (2016; Zbl 1344.82021)]. The boundary between the smooth and rough disordered region has been studied extensively in [\textit{V. Beffara} et al., Ann. Probab. 46, No. 5, 2973--3013 (2018; Zbl 1428.60063)]. Unfortunately, it is not obvious how the expression in [\textit{S. Chhita} and \textit{K. Johansson}, Adv. Math. 294, 37--149 (2016; Zbl 1344.82021)] extends to the biased generalization in this paper. Instead, the authors follow the approach of [\textit{T. Berggren} and \textit{M. Duits}, Adv. Math. 356, Article ID 106766, 48 p. (2019; Zbl 1465.60040)]. In [loc. cit.] the authors studied probability measures on particle configurations given by products of minors of block Toeplitz matrices. The biased \(2\times 2\) periodic Aztec diamond can be viewed as a special case of such a probability measure. The main result of [4] is an explicit double integral formula for the correlation kernel, provided one can find a Wiener-Hopf factorization for the product of the matrix-valued symbols for the block Toeplitz matrices. That Wiener-Hopf factorization can in principle be found by carrying out an iterative procedure, in which the total number of iterations is of the same order as the size of the Aztec diamond. In certain special cases, such as the unbiased \(2\times 2\) periodic Aztec diamond (see [\textit{T. Berggren} and \textit{M. Duits}, Adv. Math. 356, Article ID 106766, 48 p. (2019; Zbl 1465.60040)]) the procedure is periodic, and after a few iterations one ends up with the same parameters that one started with. This means that the Wiener-Hopf factorization has a rather simple form, and after inserting that expression in the double integral formula one obtains a suitable starting point for a saddle point analysis (see [\textit{M. Duits} and \textit{A. B. J. Kuijlaars}, J. Eur. Math. Soc. (JEMS) 23, No. 4, 1075--1131 (2021; Zbl 1487.60018)] and [4]). However, generically, the iteration in [\textit{T. Berggren} and \textit{M. Duits}, Adv. Math. 356, Article ID 106766, 48 p. (2019; Zbl 1465.60040)] is too complicated to find simple expressions for the Wiener-Hopf factorization, and other ideas are needed. The biased \(2\times 2\) periodic Aztec diamond is the simplest doubly periodic case in which it is difficult to trace the flow in [\textit{T. Berggren} and \textit{M. Duits}, Adv. Math. 356, Article ID 106766, 48 p. (2019; Zbl 1465.60040)]. The first main result of the present paper is that the Wiener--Hopf factorization can alternatively be computed by following a linear flow on an explicit elliptic curve. Consider an elliptic curve (over \(\mathbf R\)) defined by the equation \(y^2 = x^2 + \frac{4x(x - a^2)(x - 1/a^2)}{(a + 1/a)^2(\alpha + 1/\alpha)^2}\), where \(\alpha\) and \(a\) are the parameters from the dimer model. One easily verifiesthat the curve crosses the \(x\)-axis precisely three times, once at the origin and at two further intersection points in \((-\infty, 0)\). The elliptic curve has therefore two connected components, and one of those, lies entirely in the left half plane. Note also that \((0, 0)\), \((a^2, a^2)\) and \((a^{-2}, a^{-2})\) are the intersection points of the curve with the line \(y = x\). It is well known that an elliptic curve carries an abelian group structure, and one can add points on the curve. Note that in [\textit{M. Duits} and \textit{A. B. J. Kuijlaars}, J. Eur. Math. Soc. (JEMS) 23, No. 4, 1075--1131 (2021; Zbl 1487.60018)] it was shown that the double periodicity leads to matrix-valued orthogonal polynomials. For the unbiased \(2\times 2\) periodic Aztec diamond, these matrix-valued orthogonal polynomials have a particularly simple structure. Somewhat surprisingly, they even have explicit integral expressions that lead to an explicit double integral representation for the correlation kernel. The expression in [\textit{M. Duits} and \textit{A. B. J. Kuijlaars}, J. Eur. Math. Soc. (JEMS) 23, No. 4, 1075--1131 (2021; Zbl 1487.60018)] was re--derived in [4]. For the biased model it is interesting to see what the flow on the elliptic curve implies for the matrix-valued orthogonal polynomials, and if explicit expressions can be given in general and for the periodic case. In the remaining part of this paper the authors prove the main results. In Section 4 the authors show that the linear flow on the elliptic curve can be used to find a Wiener-Hopf factorization. Section 5 returns to the biased Aztec diamond and prove Theorem 3.1, which is by then just an identification of the parameters in the discussion of Section 4. The saddle point analysis starts in Section 5.4. After that, the authors perform a saddle point analysis in Section 5.5. In Appendix A the authors work out the example where \((a^{-2}, a^{-2})\) is a torsion point of order six, compute the boundary of the rough disordered region, and provide numerical results supporting the saddle point analysis of Section 5.5. Finally, Appendix B shows how the notion of division polynomials can be used to find algebraic relations between \(a\) and \(\alpha\) so that \((a^{-2}, a^{-2})\) is a torsion point of order \(d\).