Sharp inequalities for the conditional square function of a martingale (Q1180578)

Let \((f_ n,{\mathcal F}_ n)_{n\geq 1}\) be a martingale with values from a real or complex Hilbert space \((H,|\cdot|)\), and \((d_ n)_{n\geq 1}\) the sequence of its differences. Let us introduce \[ s^ 2(f)=\sum_{n=1}^ \infty E(| d_ n|^ 2\mid{\mathcal F}_{n- 1}),\quad S^ 2(f)=\sum_{n=1}^ \infty| d_ n|^ 2\hbox{ and } f^*=\sup_{n\geq 1}| f_ n|, \] the conditional square function, the quadratic variation and the maximal function of \((f_ n)_{n\geq 1}\), resp. The following inequalities are proved to be sharp: \[ \begin{aligned} \| f\|_ p \leq\sqrt{2/p}\| s(f)\|_ p, \qquad &0<p \leq 2,\\ \sqrt{2/p}\| s(f)\|_ p \leq\| f\|_ p, \qquad &p \geq 2,\\ \sqrt{2/p}\| s(f)\|_ p \leq\| f^*\|_ p,\qquad &p \geq 2,\\ \| S(f)\|_ p \leq\sqrt{2/p}\| s(f)\|_ p, \qquad &0<p \leq 2,\\ \sqrt{2/p}\| s(f)\|_ p \leq\| S(f)\|_ p, \qquad& p \geq 2.\end{aligned} \] These inequalities are still the best possible in the special case of real-valued conditionally symmetric martingales.