Affine weighted Motzkin paths and the differential kernel method

Abstract

We study Motzkin paths whose up-, level-, and down-step weights are affine functions of the height: alpha_k = A k + alpha_0, gamma_k = B k + gamma_0, beta_k = C k + beta_0. Let w_{n,k} be the total weight of paths of length n ending at height k, and let W_n = w_{n,0} be the return column. For affine weights the row recurrence becomes a first-order differential equation in the variable marking terminal height, and, unless beta_0 = C, that equation involves the unknown return series as boundary data. We show that the return series is nevertheless determined by the step rule alone: a characteristic ending on the floor forces a cancellation and yields an Abel-Volterra equation for the returns, while the same identity read at an interior point reconstructs the entire triangle (w_{n,k}). We call this boundary cancellation the differential kernel method; it takes the place of substituting an admissible kernel root, which is unavailable because the equation is differential rather than algebraic in the catalytic variable. When the boundary index nu = (beta_0 - C)/C is a positive integer m, the construction becomes finite and combinatorial: shifting every height by m turns the divided-difference evolution into an ordinary weighted path model on a half-line carrying m virtual levels below the visible floor, and a first-entry decomposition expresses (w_{n,k}) through two such local models. Exactly one bridge, of weight alpha_0 - A, leads from the virtual strip back to the visible region; it is closed, and the terminal-height columns factorise, precisely when alpha_0 = A. For the Dyck weights alpha_k = k+1, beta_k = k+nu+1, gamma_k = 0 this gives sum_{n>=0} w_{n,k} t^n / n! = sec^{nu+1}(t) tan^k(t).

42 pages, 8 figures