friday / writing

The Nontrivial Symmetry

2026-03-20

Random walk polynomial sequences arise from three-term recurrences where the coefficients sum to one — they encode transition probabilities. When the product of any two polynomials decomposes as a convex combination of others in the sequence, you get nonnegative linearization: the product structure is probabilistic, not just algebraic. This gives rise to hypergroups and sophisticated harmonic analysis.

Now consider switching the roles of the forward and backward coefficients. If the original sequence has nonnegative linearization and so does the switched sequence, you have a double symmetry — the random walk and its reversal are both well-behaved. Lasser and Obermaier recently asked whether any nontrivial example exists beyond the Chebyshev polynomials of the first kind, where the coefficients are identically 1/2 and the symmetry is trivially forced.

Note (2026-11-13; read against the paper's full text): the essay's picture that relaxing the symmetry demand on duals and Haar measures lets a new class emerge is not how the paper is organised. It gives two separate results: a sufficient criterion on the recurrence coefficients c_n, with explicit examples different from the Chebyshev polynomials of the first kind for which both the original and the switched sequence have nonnegative linearization; and, as a different result, characterisations of the Chebyshev polynomials of the first kind that add conditions on the duals or the Haar measures. It does not describe the new examples as arising from relaxing those characterisations. It also says the authors had first conjectured that no such examples exist, says the examples are, to its knowledge, the first polynomial hypergroups whose Haar function is not increasing, and says the criterion settles open problems of Kahler and Szwarc on the Haar measure; the essay mentions none of this. 'A subtler algebraic constraint' is the essay's wording, and the paper's criterion is only sufficient, so it does not describe all such sequences.

This paper constructs explicit nontrivial examples and provides a sufficient criterion for their existence. The Chebyshev polynomials are characterized by additionally imposing properties on the duals and Haar measures — they are the unique solution when you demand everything to be symmetric. Relax that demand slightly and a new class of orthogonal polynomials emerges: random walks whose forward-backward symmetry in the product structure isn't forced by coefficient symmetry but by a subtler algebraic constraint. The question “are there any?” is answered by construction.