Problems which interest and/or annoy me (Q1318440)

The positivity of the integral \(\int^ x_ 0 t^{-\beta} J_ \alpha (t)dt\) for all \(x>0\) is a question which has been studied for many decades. It is known to hold when \(\alpha \geq 0\), \(\beta = - {1 \over 2}\), when \(\alpha>-1\), \(\beta = 0\), and if it holds for \((\alpha,\beta)\) then it holds for \((\alpha - \mu,\beta + \mu)\), \(\mu>0\), as long as the integral converges. It is shown to hold at \((\alpha + \mu,\beta + \mu)\) for \(\mu>0\). For \(-1<\alpha<{1 \over 2}\), it holds for \(\beta>\beta (\alpha)\) with \(\beta (\alpha)\) the value of \(\beta\) where the areas of the first two arches are equal. Numerical data suggests that \(\beta (\alpha)\) is decreasing and convex for \(-1<\alpha<{1 \over 2}\), but this has not been demonstrated. For Jacobi polynomials, the heat kernel \[ \sum^ \infty_{n=0} r^{n(n + \alpha + \beta + 1)}P_ n^{(\alpha,\beta)} (x) P_ n^{(\alpha, \beta)} (1)/h_ n \] was shown to be positive for \(0 \leq r<1\), \(-1 \leq x \leq 1\) by Bochner when \(\alpha = \beta \geq - {1 \over 2}\), and for \(\alpha \geq \beta \geq - {1 \over 2}\) by Gasper. When \(\alpha = \beta = - {1 \over 2}\), the zeros of this function are determined by the triple product of the theta function. They are real and lie on \(x<-1\) when \(0 \leq r<1\). This is shown to hold when \(\alpha = \beta = {1 \over 2}\) and when \(\alpha = {1 \over 2}\), \(\beta = - {1 \over 2}\).