A new upper bound for the complex Grothendieck constant (Q1103816): Difference between revisions

In the present work a new estimate for the Grothendieck constant \(K^{{\mathbb{C}}}_ G\) (i.e. the best constant in the Grothendieck inequality) in the complex case is obtained. It is proved that \(K^{{\mathbb{C}}}_ G\leq 8/\pi (k_ 0+1)\) where \(k_ 0\) is the unique solution of the equation \[ \pi (k+1)/8=k\int^{\pi /2}_{0}\frac{\cos^ 2 t}{(1-k^ 2 \sin^ 2 t)^{1/2}}dt=:\phi (k) \] in [0,1], numerically \(K^{{\mathbb{C}}}_ G\leq 1.40491... \). The author's method consists in a subtile extension of Krivine's proof that \(K^{{\mathbb{R}}}_ G\leq \pi /(2 \log (1+\sqrt{2}))\) [\textit{J. L. Krivine}, Adv. Math. 31, 16-30 (1979; Zbl 0413.46054)]. A crucial fact is the formula \({\mathbb{E}}(sign X\cdot sign \bar Y)=\phi (| {\mathbb{E}}(X\bar Y)3/)sign {\mathbb{E}}(X\bar Y)\) for complex valued standard Gaussian variables, which corresponds to \({\mathbb{E}}(sign X sign Y)=(2/\pi)arc \sin {\mathbb{E}}(XY)\) in the real valued case. Major part of the proof is devoted to an estimation of the Taylor coefficients of \(\phi^{-1}\).