Correlation function of the Schur process with a fixed final partition

We consider a generalization of the Schur process in which a partition evolves from the empty partition into an arbitrary fixed final partition. We obtain a double integral representation of the correlation kernel. For a special final partition with only one row, the edge scaling limit is also discussed by the use of the saddle point analysis. If we appropriately scale the length of the row, the limiting correlation kernel changes from the extended Airy kernel.

💡 Research Summary

The paper studies a natural generalization of the Schur process in which the system starts from the empty partition at time 0 and ends at a prescribed, possibly non‑trivial, final partition µ(4N) at time 4N. The classical Schur process corresponds to the special case µ = ∅, and its correlation functions are known to be determinantal with a kernel that can be written as a double contour integral (Okounkov–Reshetikhin, Johansson). When the final partition is arbitrary, the weight (1.2) is still a product of skew Schur functions, but the associated matrix A (2.9) is no longer Toeplitz, so the standard Wiener–Hopf factorisation does not apply.

The authors first embed the weight into the general class of “product‑of‑determinants” measures (2.1) and invoke the Lindström–Gessel–Viennot theorem to obtain a determinantal expression for the multi‑point correlation function (2.5). The kernel K is expressed as ˜K−φ, where ˜K involves the inverse of A. The main technical contribution is a new method for estimating A⁻¹ when A is not Toeplitz. By assuming the parameters a(i) lie in (0,1) and using the Jacobi–Trudi identity, they rewrite the entries of A in terms of complete symmetric polynomials h_m(a). This enables them to derive an explicit double‑contour integral representation for the kernel (Theorem 2.1, equations (2.18)–(2.19)). The contours C_{r₁}, C_{r₂} are circles with radii r₁>r₂, and the integrand contains products of factors (1−a_{k} z) and (1−a_{k}/z) together with the complete symmetric functions evaluated at the set of all a‑parameters.

Having obtained a usable integral formula, the paper proceeds to analyze the edge scaling limit in a concrete setting: the final partition is taken to be (m,−1,−2,…), while all a(i) are equal to (α,0,0,…). For large N the macroscopic shape of the particle system is a semicircle A(t) (2.26). The authors scale particle positions and time according to the KPZ 1/3 and 2/3 exponents (2.27)–(2.28) and also scale the length m of the top row as m = A(2N) N + B N^{2/3} ω (2.30). Under this scaling, they prove (Theorem 2.2) that the kernel converges to a sum of the extended Airy kernel K₂(τ₁,ξ₁;τ₂,ξ₂) and an additional term involving a single Airy integral with exponential weight e^{λ(τ₁+ω)}. When ω→−∞ (i.e., m=0, the original Schur process) the extra term disappears and one recovers the pure extended Airy kernel, confirming consistency with known results.

The derivation of the scaling limit uses a saddle‑point analysis of the double integral (2.18). The authors identify the dominant critical point, expand the exponent to cubic order, and recognize the resulting integrals as Airy functions. The extra term originates from the contribution of the final row of the partition; its sign and presence depend on whether τ₁+ω is negative or positive, reflecting a phase transition analogous to the one observed in PNG models with an external source.

In summary, the paper makes three significant contributions: (1) it extends the Schur process to arbitrary fixed final partitions and provides a rigorous double‑integral formula for the correlation kernel; (2) it develops a new technique to invert the non‑Toeplitz matrix arising from the general final condition; (3) it shows how the edge scaling limit interpolates between the pure extended Airy kernel and a kernel with an extra Airy integral, thereby elucidating the effect of a non‑trivial final partition on KPZ‑type fluctuations. These results deepen the connection between integrable probability, random matrix theory, and growth models with boundary conditions, and open the way for further investigations of more general initial/final data and multi‑row final partitions.