Paper 2, Section II, H

Linear Analysis
Part II, 2009

For 1p<1 \leqslant p<\infty and a sequence x=(x1,x2,)x=\left(x_{1}, x_{2}, \ldots\right), where xjCx_{j} \in \mathbb{C} for all j1j \geqslant 1, let xp=(j=1xjp)1/p.\|x\|_{p}=\left(\sum_{j=1}^{\infty}\left|x_{j}\right|^{p}\right)^{1 / p} .

Let p={x=(x1,x2,):xjC\ell^{p}=\left\{x=\left(x_{1}, x_{2}, \ldots\right): x_{j} \in \mathbb{C}\right. for all j1j \geqslant 1 and xp<}\left.\|x\|_{p}<\infty\right\}.

(a) Let p,q>1p, q>1 with 1/p+1/q=1,x=(x1,x2,)p1 / p+1 / q=1, x=\left(x_{1}, x_{2}, \ldots\right) \in \ell^{p} and y=(y1,y2,)qy=\left(y_{1}, y_{2}, \ldots\right) \in \ell^{q}. Prove Hölder's inequality:

j=1xjyjxpyq\sum_{j=1}^{\infty}\left|x_{j}\left\|y_{j} \mid \leqslant\right\| x\left\|_{p}\right\| y \|_{q}\right.

(b) Use Hölder's inequality to prove the triangle inequality (known, in this case, as the Minkowski inequality):

x+ypxp+yp for every x,yp and every 1<p<\|x+y\|_{p} \leqslant\|x\|_{p}+\|y\|_{p} \quad \text { for every } x, y \in \ell^{p} \quad \text { and every } 1<p<\infty

(c) Let 2p<2 \leqslant p<\infty and let KK be a closed, convex subset of p\ell^{p}. Let xpx \in \ell^{p} with xKx \notin K. Prove that there exists yKy \in K such that

xy=infzKxz.\|x-y\|=\inf _{z \in K}\|x-z\| .

[You may use without proof the fact that for every 2p<2 \leqslant p<\infty and for every x,ypx, y \in \ell^{p},

x+ypp+xypp2p1(xpp+ypp).]\left.\|x+y\|_{p}^{p}+\|x-y\|_{p}^{p} \leqslant 2^{p-1}\left(\|x\|_{p}^{p}+\|y\|_{p}^{p}\right) .\right]