I'd like to know if one can define a pertinent Hilbert space where the operator $$A_p v := -\frac{1}{2} v" + (vF + v\int_\mathbb{R} Sp + p\int_\mathbb{R} Sv )'$$ is symmetric. Here, $p$ satisfies the following differential equation with $p \in L^2_1(\mathbb{S})\cap C^\infty(\mathbb{S}) $ $$\frac{p'}{2p} = F + \int Sp $$ and $u,v$ and $F,S$ can be taken to satisfie at least the following weak regularity conditions:

  • $u,v \in L^2_0(\mathbb{R})\cap C^2(\mathbb{R})$
  • $F,S \in C^\infty(\mathbb{R})$ and odd

but the regularity will probably be restricted much more by choosing the nice Hilbert space that we wish. We note that $\int_\mathbb{R} Sp= \int_\mathbb{R} S(x) p(x) dx$.

My motivation for a possible positive answer for this question comes from the following similar problem. Given a fixed $p \in L^2_1(\mathbb{S})\cap C^\infty(\mathbb{S}) $ ($\mathbb{S} = \mathbb{R}/(2\pi\mathbb{Z})$ and $p \neq 0$), solution of $\frac{p'}{2p}= S*p$ ($S(\theta) = \sin(\theta)$) for, the operator

$$B_p v := -\frac{1}{2} v" + ( v\int_\mathbb{S} S*p + p\int_\mathbb{S} S*v )'$$

a such Hilbert space is $H_ {-1,1/p}$ provideded with $\left\langle f,g\right\rangle_{H_ {-1,1/p}} := \int \frac{\mathcal{F}\mathcal{G}}{p} $ where $\mathcal{F}$ and $\mathcal{G}$ are respectively the primitives in $ L^2(\mathbb{R})$ of $f$ and $g$ such that $\int \frac{\mathcal{F}} {p} = \int \frac{\mathcal{G}} {p} =0$ . Here $H_ {-1,1/p}:= H_{1,p}'$ (notation:$V'$ is the dual space of $V$), $H_{1,p}:= ${$ \overline{ f \in C^1(\mathbb{S}): \int f =0 } ^H$} (wich is also a Hilbert space with scalar product $\left\langle f,g\right\rangle_{H_ {1,p}} := \int (f'g'p)$ ) and $H =L^2 _0 (\mathbb{S})$.

We can prove for this case that $$\left\langle f,B_pg\right\rangle_{H_ {-1,1/p}} = \left\langle B_pf,g\right\rangle_{H_ {-1,1/p}}=-\frac{1}{2}\int\frac{fg}{p}+\int fS'*g$$ then $B_p$ is clearly symetric in $H_ {-1,1/p}$

I expecte had well motivated my question. I've tried several integral and derivative restrictions under spaces similar to $H_ {-1,1/p}$ but I've not been sucsecefull yet. I'd be glad for some advice. If you visit the topic please leave a message with your opinion.

  • 3
    $\begingroup$ Is this a homework problem? I see that you included a "hint". $\endgroup$
    – Nik Weaver
    Commented Jun 18, 2012 at 16:44
  • $\begingroup$ No, I've just tried to make a clear question giving as a hint the technic that I've saw in a article in which a similar problem is solved. So it inspires me to triyng solve this subproblem of my real and bigger problem of reaserch. $\endgroup$
    – Paul
    Commented Jun 18, 2012 at 23:45
  • $\begingroup$ If this question seems to be a homework question for you, you must be able to help me. If you want I can show you, Nik Weaver, the solution of the problem that motivates me (which I give as Hint). Do you think you have an idea? thank you $\endgroup$
    – Paul
    Commented Jun 18, 2012 at 23:49
  • 4
    $\begingroup$ Paul, I suggest you don't use the word "hint" (and there is no need to feel insulted by Nik Weaver's suggestion, it was just a reaction to your unfortunate choice of words). Instead, call it a "Observation which may be helpful", and perhaps even include a reference to the paper where you saw this. $\endgroup$
    – Yemon Choi
    Commented Jun 19, 2012 at 0:14
  • $\begingroup$ Thanks Yemon Choi. I really don't want to be misinterpreted. Just to make clear I did not feel insulted by Nik Weaver's comment, but at the same time I didn't understended that it was a sugestion, differently of yours comment explicitly sugesting me to change it. Sorry, I'm new here. And I still need help guys ... $\endgroup$
    – Paul
    Commented Jun 19, 2012 at 0:49

1 Answer 1


I don't completely understand the question, but I would answer "no". If you take $f = 1$ and $u = S = 0$ then you get the operator $Av = -{1\over 2}v'' + v'$, which is not even formally symmetric.

I'm especially confused by the motivating example, where the inner product seems to depend on the vector $v$.

  • $\begingroup$ Thank's for you answer and for make me constact some errors of typing. $\endgroup$
    – Paul
    Commented Jun 19, 2012 at 0:56
  • $\begingroup$ Actually, it's $u$ where before was $v$ in the definition of the spaces and scalar product. Please, see if now it makes more sense, because before it had any. $\endgroup$
    – Paul
    Commented Jun 19, 2012 at 1:02
  • $\begingroup$ Okay, it makes more sense now. I still don't see how to help because your operator isn't formally symmetric --- the $v'$ part is skew symmetric --- but if the inner product can be anything I'm not sure how to answer you. $\endgroup$
    – Nik Weaver
    Commented Jun 19, 2012 at 1:34
  • $\begingroup$ You have reason. The operator is formally not symetric but it can maybe be symetrized by chosing a good scalar product $\endgroup$
    – Paul
    Commented Jun 19, 2012 at 11:19
  • 1
    $\begingroup$ yes. it is the same trick to represent the ornstein-uhlenbeck operator $u\mapsto \Delta u+\nabla B\cdot \nabla u$ as the gradient of the energy functional $u\mapsto \frac{1}{2} \int_\Omega |\nabla u|^2 e^{-B}dx$ -- this is done in the space $L^2(\Omega)$, taking the inner product $(u|v):=\int_\Omega u v e^{-B}dx$. $\endgroup$ Commented Feb 8, 2013 at 8:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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