Higher order Turán inequalities for unimodal sequences

Koustav Banerjee, Kathrin Bringmann, Ben Kane · 2025

Read Higher order Turán inequalities for unimodal sequences by Koustav Banerjee, Kathrin Bringmann, Ben Kane (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-15236.txt.

Introduction

Koustav Banerjee, Kathrin Bringmann, and Ben Kane study a counting sequence from combinatorics in a 2025 paper. The sequence u(n) counts unimodal sequences of total size n, arrangements that rise then fall. The authors derive a precise growth formula for it, complete with a guaranteed bound on the error, for sequences admitting exact formulas.

With that formula in hand, they prove the sequence obeys a family of stricter pattern rules, the higher order Turán inequalities, for all n from thirty-three upward. Certain shifted second differences of the sequence also come out positive. The generic method is the point: the same approach should settle such inequalities for other sequences with exact counting formulas.

Catalog record

CREATORKoustav Banerjee, Kathrin Bringmann, Ben Kane
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE117 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2505-15236

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Higher order Turán inequalities for unimodal sequences Koustav Banerjee, Kathrin Bringmann, Ben Kane Abstract . We derive an asymptotic expansion with effective error bound for u ⁡ ( n ) u(n) , counting the number of unimodal sequences of size n n . We prove that u ⁡ ( n ) u(n) satisfies the higher order Turán inequalities for n ≥ 33 n\geq 33 and that certain second j j -shifted difference of u ⁡ ( n ) u(n) are positive. 2020 Mathematics Subject Classification 05A16, 11P82 1. Introduction and statement of results A polynomial with real coefficients is hyperbolic if it has real roots. Newton [ 31 , Theorem 2] proved that if a polynomial \sum_{j=0}^{d}(\begin{smallmatrix}d\\ j\end{smallmatrix})a(j)x^{j} is hyperbolic, then a^{2}(j)\geq a(j-1)a(j+1) for 0<j\leq d-1 . A sequence (a(n))_{n\geq 0} of positive real numbers is log-concave if for all n\in\mathbb{N} , (1.1) a^{2}(n)\geq a(n-1)a(n+1). The Jensen polynomial of degree d associated to a sequence with shift n , a(n) is defined by J^{d,n}_{a}(x):=\sum_{0\leq j\leq d}(\begin{smallmatrix}d\\ j\end{smallmatrix})a(n+j)x^{j} . By [ 12 , Section 6] , ( 1.1 ) is equivalent to J^{2,n}_{a} being hyperbolic. First, Nicolas [ 24 ] and later, DeSalvo and Pak [ 13 ] proved that the partition function p(n) is log-concave for n\geq 26 . By [ 12 , Section 6] , J^{3,n}_{a} is hyperbolic iff the discriminant of J^{3,n}_{a} is non-negative. Following [ 11 , equation (1.2)] , a sequence (a(n))_{n\geq 0} of positive real numbers satisfies the higher order Turán inequalities for n\in\mathbb{N} if \displaystyle 4\left(a^{2}(n)\!-\!a(n-1)a(n+1)\right) \displaystyle\!\left(a^{2}(n+1)\!\geq\!a(n)a(n+2)\right) (1.2) \displaystyle\!-\!\left(a(n)a(n+1)\!-\!a(n-1)a(n+2)\right)^{2}\!\geq\!0. Chen–Jia–Wang [ 11 ] proved that p(n) satisfies the higher order Turán inequalities for n\geq 95 . Their proof was based on Rademacher’s exact formula for p(n) [ 29 ] which was a substantial improvement of the Hardy–Ramanujan divergent series for p(n) (see [ 18 ] ) and Lehmer’s error estimates [ 20 , 22 ] . Due to the seminal work of Griffin–Ono–Rolen–Zagier [ 16 ] , for each degree d , J^{d,n}_{p} is hyperbolic for n sufficiently large. In [ 6 ] , studying the asymptotic expansion (up to any order) of p(n) with an effective error bound, a unified framework was given to prove various inequalities for p(n) . For another notable instance Ono–Pujahari–Rolen [ 27 ] proved that the plane partition function 1 1 1 For more background on \operatorname{pp}(n) , we refer the reader to [ 3 , Chapter 11] . \operatorname{pp}(n) is log-concave for n\geq 12 by significantly improving the Wright’s asymptotic formula [ 32 ] for \operatorname{pp}(n) . Pandey [ 28 ] studied the hyperbolicity of J^{d,n}_{\text{pp}} and in particular, showed that \operatorname{pp}(n) satisfies the higher order Turán inequalities for n\geq 26 . Let u(n) count the number of unimodal sequences of n , 1\leq a_{1}\leq\dots\leq a_{r}\leq c\geq b_{s}\geq\dots\geq b_{1}\geq 1\ \ \ \sum_{j=1}^{r}a_{j}+c+\sum_{j=1}^{s}b_{s}=n\ \ r,s\in\mathbb{N}_{0}. The generating function for unimodal sequences is given by (see [ 4 , equation (3.2) and (3.3)] ) (1.3) \sum_{n\geq 0}u(n)q^{n}=\sum_{n\geq 0}\frac{q^{n}}{(q;q)^{2}_{n}}=\frac{1}{\left(q;q\right)_{\infty}^{2}}\sum_{n\geq 0}(-1)^{n}q^{\frac{n(n+1)}{2}}, where for n\in\mathbb{N}_{0}\cup\{\infty\} , (a;q)_{n}:=\prod_{j=0}^{n-1}(1-aq^{j}) . Unlike p(n) which occur as Fourier coefficients of the modular form P(q):=(q;q)_{\infty}^{-1} , we see from ( 1.3 ) that u(n) is associated to the product of a modular form P^{2}(q) and the partial theta function \sum_{n\geq 0}(-1)^{n}q^{\frac{n(n+1)}{2}} . The second author and Nazaroglu [ 10 ] gave a Rademacher-type exact formula for u(n) . Using this exact formula, Bridges and the second author [ 8 ] proved that u(n) is log-concave for n\in\mathbb{N}\setminus\{1,5,7\} . In this paper, we determine the asymptotic expansion (up to any order) for u(n) and then for its shifted version u(n+s) with s\in\mathbb{N} . We then use this to study the higher order Turán inequalities and second-order shifted differences for u(n) . To state the main results, we use the notation f(n)=O_{\leq c}(g(n)) if |f(n)|\leq c\left|g(n)\right| for n\geq n_{0} and n_{0}\in\mathbb{N} . Theorem 1.1 . Let N\in\mathbb{N}_{\geq 3} . For n\geq n_{N} , we have u(n)=\frac{e^{2\pi\sqrt{\frac{n}{3}}}}{8\cdot 3^{\frac{3}{4}}\sqrt{\pi}n^{\frac{5}{4}}}\left(\sum_{m=0}^{N+1}\dfrac{A(m)}{n^{\frac{m}{2}}}+O_{\leq C_{N}}\left(n^{-\frac{N+2}{2}}\right)\right), where n_{N} , A(m) , and C_{N} are defined in ( 3.61 ), ( 3.120 ), and ( 5.1 ), respectively. Remark 1.2 . (1) For N=3 , the constants A(m) (0\leq m\leq 4) are given in [ 8 , equation (4.16)] , and n_{N}=100000 , C_{3}=478 are as in [ 8 , Theorem 1.1] . (2) The convolution of non-negative sequences (a(n))_{n\geq 0} , (b(n))_{n\geq 0} is (c(n))_{n\geq 0} with c(n):=\sum_{k=0}^{n}a(k)b(n-k) . From [ 23 ] , log-concavity is preserved under convolution, i.e., if (a(n))_{n\geq 0} , (b(n))_{n\geq 0} are log-concave and a(0)=b(0)=1 , then so is (c(n))_{n\geq 0} . Since the higher order Turán inequalities hold for p(n) for n sufficiently large [ 11 ] and the partition generating function P(q) appears as a factor in ( 1.3 ), it is natural to ask about the behaviour of the Turán inequalities under convolution. Specifically, for some N_{0}\in\mathbb{N}_{0} , suppose that (a(n))_{n\geq 0} and (b(n))_{n\geq 0} are two sequences of real numbers which satisfy the higher oder Turán inequalities for n\geq N_{0} . Under what conditions does there exist N_{1}\!\in\!\mathbb{N}_{0} such that the convolution of these two sequences also satisfy (if at all) the higher order Turán inequalities for n\geq N_{1} ? How can one determine an effective estimate of the cutoff N_{1} ? In order to prove log-concavity and its higher order analogs, we give an error bound for the asymptotic expansion of the shifted unimodal sequence u(n+s) . Theorem 1.3 . Let N\in\mathbb{N}_{\geq 3} and s\in\mathbb{N} . For n\geq n_{N}(s) , we have u(n+s)=\dfrac{e^{2\pi\sqrt{\frac{n}{3}}}}{8\cdot 3^{\frac{3}{4}}\sqrt{\pi}n^{\frac{5}{4}}}\left(\sum_{m=0}^{N+1}\dfrac{A_{s}(m)}{n^{\frac{m}{2}}}+O_{\leq C_{N}(s)}\left(n^{-\frac{N+2}{2}}\right)\right), where n_{N}(s) , A_{s}(m) , and C_{N}(s) are defined in ( 5.12 ), ( 5.9 ), and ( 5.10 ), respectively. Let \Delta be the backward difference operator defined on a sequence (a(n))_{n\geq 0} by \Delta(a)(n):=a(n)-a(n-1), and for r\in\mathbb{N} , the r -fold application of \Delta is given as \Delta^{r}(a)(n):=\Delta(\Delta^{r-1}(a))(n) . Gupta [ 17 ] proved that (p(n))_{n\geq 2} is convex , i.e., \Delta^{2}(p)(n)>0 for n\geq 2 and furthermore, he investigated the positivity of \Delta^{r}(p)(n) for n sufficiently large. For a more detailed study on finite differences of the partition function, we refer the reader to [ 2 , 17 , 21 , 25 ] . Recently, Gomez–Males–Rolen [ 14 ] extended the definition of \Delta for a sequence (a(n))_{n\geq 0} with 1\leq j\leq n , \Delta_{j}(a)(n):=a(n)-a(n-j). Similarly, the second j -shifted difference is defined as (1.4) \Delta^{[2]}_{j}(a)(n):=a(n)-2a(n-j)+a(n-2j). In [ 14 , Theorem 1.2] , using Rademacher’s exact formula for p(n) [ 29 ] , the authors showed that (p(n))_{n\geq n(j)} satisfies shifted convexity , i.e., j\in\mathbb{N} , for n\geq n(j):=\max\{2,16j^{2}+\frac{1}{24}\} , we have \Delta^{[2]}_{j}(p)(n)\geq 0 . In this paper we prove that u(n) also satisfies shifted convexity. Theorem 1.4 . For j\in\mathbb{N} , there exist effectively computable constants n_{\Delta}(j) (see ( 5.17 )) such that for n\geq n_{\Delta}(j) , \Delta^{[2]}_{j}(u)(n)>0. Remark 1.5 . Positivity of \Delta^{[2]}_{j}(u)(n) for n,j\in\mathbb{N} can also be shown by elementary combinatorial argument via constructing injections. However, in this paper, since our approach is analytic, we do not present a combinatorial proof. We finally use Theorem 1.3 to verify the higher order Turán inequality for u(n) . Theorem 1.6 . For n\geq 33 , u(n) satisfies the higher order Turán inequality. In particular, those n\in\mathbb{N} for which ( 1 ) does not hold are \{n\in\mathbb{N}:n\leq 26\}\cup\{28,30,32\} .

Read the complete text