On the log-convexity of combinatorial sequences (Q2467343): Difference between revisions

The paper studies log-convexity of combinatorial sequences. It is shown that log-convexity is preserved under componentwise sum, under binomial convolution, and by linear transformations given by matrices of binomial coefficients and Stirling numbers of both kinds. Techniques are developed to prove the log-convexity of sequences satisfying a three-term recurrence. These techniques show the log-convexity of the central binomial coefficients, the Catalan numbers, the Motzkin numbers, the Fine numbers, the central Delannoy numbers, the little and large Schröder numbers, and derangements. If \(F_n\), \(L_n\) and \(P_n\) denote Fibonacci, Lucas and Pell numbers respectively, then the bisections \(\{F_{2n+1}\}\), \(\{L_{2n}\}\), and \(\{P_{2n+1}\}\) are log-concave, while the bisections \(\{F_{2n}\}\), \(\{L_{2n+1}\}\), and \(\{P_{2n}\}\) are log-convex. Stanley suggested the following definition for \(q\)-log-convexity. For two real polynomials \(f(q)\) and \(g(q)\), write \(f(q) \leq_q g(q)\) if \(f(q) - g(q)\) has non-negative coefficients as a polynomial. A sequence of real polynomials \(P_n(q)\) is called \(q\)-log-convex, if \(P_n^2(q) \leq_q P_{n-1}(q)P_{n+1}(q)\). Bell and Eulerian polynomials, \(q\)-Schröder and \(q\)-Delannoy numbers are shown to be log-convex. The \(q\)-log-convexity of Narayana polynomials is posed as a conjecture, together with conjectures about certain transformations keeping \(q\)-log-convexity.