• $(E,\mathcal E)$ be a measurable space
  • $\mu$ be a measure on $(E,\mathcal E)$ and $$\mu f:=\int f\:{\rm d}\mu$$ for Borel measurable $f:\mathbb R\to\mathbb R$ with $f\ge0$ or $\mu|f|<\infty$
  • $\mathcal A_0$ be a subspace of $\left\{f:E\to\mathbb R\mid f\text{ is bounded and }\mathcal E\text{-measurable}\right\}$ closed under multiplication and dense in $L^p(\mu)$ for all $p\ge1$
  • $\Gamma$ be a bilinear symmetric operator on $\mathcal A_0$ and $$\Gamma(f):=\Gamma(f,f)\;\;\;\text{for }f\in\mathcal A_0$$


  1. $\forall f\in A_0:\exists c\ge0:\forall g\in\mathcal A_0:\left|\mu\Gamma(f,g)\right|\le c\left\|g\right\|_{L^2(\mu)}$
  2. $\mu\Gamma(f^2,g)+2\langle\Gamma(f),g\rangle_{L^2(\mu)}=2\mu\Gamma(fg,f)$ for all $f,g\in\mathcal A_0$
  3. $\Gamma(f)\ge0$ for all $f\in\mathcal A_0$

By 1., there is a unique linear symmetric operator $(\mathcal A_0,L)$ on $L^2(\mu)$ with $$\langle Lf,g\rangle_{L^2(\mu)}=-\mu\Gamma(f,g)\;\;\;\text{for all }f,g\in\mathcal A_0.\tag1$$ By 2., $$L(fg)=2\Gamma(f,g)+fLg+gLf\;\;\;\text{for all }f,g\in\mathcal A_0.\tag2$$

Question 1: How can we show that $${\Gamma(f,g)}^2\le\Gamma(f)\Gamma(g)\tag3$$ for all $f,g\in\mathcal A_0$? By 3., $$0\le\Gamma(f+\lambda g)=\Gamma(f)+2\lambda\Gamma(f,g)+\lambda^2\Gamma(g)\;\;\;\text{for all }\lambda\in\mathbb R.\tag4$$ If $\Gamma$ would be positive definite, then we could choose $$\lambda:=-\frac{\Gamma(f,g)}{\Gamma(g)}$$ and conclude. Are we able to prove positive definiteness of $\Gamma$? If not, is there an other way to show $(3)$?

Let $$\mathcal E(f,g):=\mu\Gamma(f,g)\;\;\;\text{for }f,g\in\mathcal A_0.$$ By $(3)$, $${\mathcal E(f,g)}^2\le\mathcal E(f)\mathcal E(g)\;\;\;\text{for all }f,g\in\mathcal A_0.\tag5$$


  1. $\mu(Lf)=0$ for all $f\in\mathcal A_0$

By 4., $$\mathcal E(f,g)=-\langle f,Lg\rangle_{L^2(\mu)}=-\langle Lf,g\rangle_{L^2(\mu)}\;\;\;\text{for all }f,g\in\mathcal A_0.\tag6$$

Question 2: How can we show that $\mathcal E$ is closable? Let $(f_n)_{n\in\mathbb N}$ be $\mathcal E$-Cauchy with $$\left\|f_n\right\|_{L^2(\mu)}\xrightarrow{n\to\infty}0.\tag7$$ By $(6)$ and $(5)$, $$0\le\mathcal E(f_n)\le\left\|Lf_m\right\|_{L^2(\mu)}\left\|f_n\right\|_{L^2(\mu)}+{\mathcal E(f_n-f_m)}^{\frac12}{\mathcal E(f_n)}^{\frac12}\;\;\;\text{for all }m,n\in\mathbb N.\tag8$$ My problem is that $\mathcal E(f_n)$ is occuring on the right-hand side of $(8)$. Why can we nevertheless conclude $\mathcal E(f_n)\xrightarrow{n\to\infty}0$?

  • $\begingroup$ Concerning question 1: If $\Gamma(g) > 0$, you can your choice of $\lambda$. Otherwise, (4) is linear in $\lambda$ and, hence, it implies $\Gamma(f,g) = 0$. $\endgroup$ – gerw Feb 15 at 7:30
  • $\begingroup$ Concerning question 2: I think you can simply use Young's inequality for the second addend on the right hand side and then bring $E(f_n)/2$ to the left-hand side. $\endgroup$ – gerw Feb 15 at 7:32

Gerw basically answered your questions, but I think it's worth summarizing.

  1. The Cauchy-Schwarz inequality $$ a(x,y)^2\leq a(x,x)a(y,y) $$ holds for every symmetric positive semidefinite bilinear form $a\colon V\times V\to \mathbb{R}$. You have already shown this in the case $a(y,y)>0$. If $a(y,y)=0$, then $$ a(x+\lambda y,x+\lambda y)=a(x,x)+2\lambda a(x,y) $$ is a nonnegative affine function of $\lambda$, hence $a(x,y)=0$.
  2. If $T$ is a symmetric linear operator in $H$ such that $\langle Tx,x\rangle\geq 0$ for all $x\in D(T)$, then the bilinear form $$ a(x,y)=\langle Tx,y\rangle $$ is closable. Let $(x_n)$ be an $a$-Cauchy sequence such that $\|x_n\|_H\to 0$. With the same computation as in the question, one gets $$ a(x_n,x_n)\leq\|Tx_m\|_H\|x_n\|_H+a(x_n-x_m,x_n-x_m)^{1/2}a(x_n,x_n)^{1/2}. $$ The first term goes to zero as $n\to\infty$, while $a(x_n-x_m,x_n-x_m)$ is small for large $m,n$ and $a(x_n,x_n)$ is bounded (since $(x_n)$ is $a$-Cauchy). Thus $a(x_n,x_n)\to 0$.
  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$ – Aloizio Macedo Mar 21 at 18:15
  • $\begingroup$ I would still be interested in a response to the open question (which you can find in the chat). $\endgroup$ – 0xbadf00d Apr 1 at 15:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.