Hamming cube and martingales (Q1677879)

In the present paper the authors investigate analysis on the Hamming cube \(\{-1,\;1 \}^n\). For any \(f: \{-1,\;1 \}^n \rightarrow \mathbb{R}\) the discrete gradient is defined by \[ |\nabla f|^2 (x)=\sum_{y \sim x}(\frac{f(x)-f(y)}{2})^2 \] where the summation runs over all neighbor vertices of \(x\) in the Hamming cube. The authors prove the following results Theorem A. Let \(1 < p \leq 2, \;n \geq 1\) and \(f: \{-1,\;1 \}^n \rightarrow \mathbb{R}\). We have \[ s_{q}(\mathbb{E}_{n}|f|^p-|\mathbb{E}_{n}f|^p)^{1/p} \leq(\mathbb{E}_{n}|\nabla f|^p)^{1/p}. \] Here \(q\) is the conjugate exponent of \(p\), \(\mathbb{E}_{n}\) is the expectation on the Hamming cube and \(s_{q}\) is denoted the smallest positive zero of the confluent hypergeometric function \(_{1}F_{1}(-\frac{q}{2}, \frac{1}{2}, \frac{x^2}{2})\). Theorem B. For any \(n \geq 1\) and any \(f: \{-1,\;1 \}^n \rightarrow \mathbb{R}\), we have \[ \mathbb{E}_{n}M(f , |\nabla f|) \leq M(\mathbb{E}_{n}f, 0) \] where \(M(x, y)=\mathrm{Re}(x+iy)^{\frac{3}{2}}=\frac{1}{\sqrt 2}(2x-\sqrt{x^2+y^2})\sqrt{\sqrt{x^2+y^2}+x}\).