The self-interlacing conjecture for the geometric-sum polynomial
Let
Let be the finite difference operator from the paper, and let a polynomial be self-interlacing when it has real simple zeros and the zeros of strictly interlace those of .
Self-interlacing conjecture. The polynomial is self-interlacing, or is a self-interlacing polynomial multiplied by .
The claim is motivated by calculations concerning the action of the operator on subclasses of complex polynomials. The source gives no proof or resolution.
References
- Olga Katkova, Mikhail Tyaglov and Anna Vishnyakova, “Hermite-Poulain theorems for linear finite difference operators”, arXiv:1901.06398 (2019).
Progress summary
A reader-written calculation claims the conjecture is false already in three dimensions, but the counterexample has not been independently verified.
Katkova, Tyaglov, and Vishnyakova state this assertion as Conjecture 2 in their 2019 paper: applying to the geometric-sum polynomial should produce a self-interlacing polynomial, possibly after multiplication by . The paper presents calculations motivating the conjecture but gives no proof or resolution.
Posted attempt
A reader-written calculation claims a counterexample at , , and . It obtains , whose discriminant is negative for , so the output has nonreal zeros and violates the proposed dichotomy. This is a claimed complete disproof, but it has not been independently verified.
Current status (as of August 2026): the original conjecture has no verified proof, while a reader-written calculation claims a counterexample for ; the conjecture therefore remains unsettled pending verification of that calculation.
Sources
Solutions 1
This solution needs a summarySee full solution
Take , and consider the geometric-sum polynomial
Choose the admissible parameters
By the definition of the finite-difference operator,
Consequently,
The discriminant of its numerator is
Therefore this polynomial has two nonreal zeros. In particular, it is neither self-interlacing nor times a self-interlacing polynomial.
More generally, every
gives a counterexample, because the discriminant of is
Thus the proposed dichotomy fails already for and .