2
$\begingroup$

This arises in the context of trying to rigorously understand quantum dynamics but it's a functional analysis issue. For simplicity suppose we are in dimension $1$.

Let $\hat{Q}$ be the operator that's multiplication by $x$ and $\hat{P} = \frac{h}{i}\frac{d}{dx}$. Suppose $A$ is a denseley defined self-adjoint operator on $L^2(\mathbb{R})$ and that $$e^{\frac{-itA}{\hbar}}(\hat{Q}+ \hat{P})e^{\frac{itA}{\hbar}} = a\hat{Q} + b\hat{P}$$ where $a,b \in\mathbb{R}$

The author then says it follows from "exponentiation" that $$e^{\frac{-itA}{\hbar}}e^{\frac{i}{\hbar}(\hat{Q}+ \hat{P})}e^{\frac{itA}{\hbar}} = e^{\frac{i}{\hbar}(a\hat{Q} + b\hat{P})}$$ I can't think of a rigorous explanation for what exponentiation means and why the first identity implies the second. In particular I'm uncomfortable with the fact that conjugating by a unitary operator goes through the exponentiation. I'm assuming this is a functional calculus thing. I was wondering if anyone knew of a rigorous explanation for this.

$\endgroup$
  • $\begingroup$ In most case this should be enough : let $D(A)\subset L^2$ be a dense subspace and $A$ be self-adjoint $D(A) \to L^2$ (that is for $f,g \in D(A)$, $\langle Af,g \rangle = \langle f, Ag \rangle$) and $(A-cI)^{-1}$ be compact for some $c$, then the SVD gives $(A-cI)^{-1} f= \sum_n \lambda_n \langle f,\phi_n \rangle \phi_n$, $A f= \sum_n (\lambda_n^{-1}+c) \langle f,\phi_n \rangle \phi_n$ where $\lambda_n^{-1}+c$ are real and $e^{i t A} f = \sum_n e^{it(\lambda_n^{-1}+c)} \langle f,\phi_n \rangle \phi_n$ is well-defined and unitary $L^2 \to L^2$ $\endgroup$ – reuns Apr 16 at 5:15
  • $\begingroup$ What do you mean by SVD? $\endgroup$ – Questions Apr 16 at 5:27
2
$\begingroup$

I'm not sure this is the level of rigor you need. But in physics this identity is normally motivated as follows (let me set $h=1$ for readability): $$ U = e^{it A}\;\Longrightarrow\; U U^{\dagger} = \mathbb{1}\,. $$ Then $$ \begin{aligned} U^\dagger\, e^{i(\hat{Q}+\hat{P})} U &= \sum_{n=0}^\infty U^\dagger\, \frac{i^n}{n!}(\hat{Q}+\hat{P})^n \,U \\&= \sum_{n=0}^\infty\frac{i^n}{n!}\, U^\dagger\, (\hat{Q}+\hat{P})\,UU^\dagger (\hat{Q}+\hat{P}) \cdots U U^\dagger (\hat{Q}+\hat{P}) \,U \\& =\sum_{n=0}^\infty\frac{i^n}{n!}\, \big(U^\dagger\, (\hat{Q}+\hat{P})\,U\big)^n = \\&= e^{i \,U^\dagger\, (\hat{Q}+\hat{P})\,U}\,. \end{aligned} $$ From which the desired identity follows. In the second line I simply inserted $\mathbb{1}$ in between each factor in the form of $UU^\dagger$.

I guess one should also motivate that the exponential of the operator $\hat{Q} + \hat{P}$ is well defined as a power series. For the operator $\hat{Q}$ alone this is just multiplication by $e^x$ and it clearly matches the sum $\sum_n x^n/n!$. For $\hat{P}$ the convergence of the exponential is the same as the convergence for the Taylor expansion $$ e^{ia\hat{P}} f(x) = \sum_{n=0}^\infty \frac{a^n}{n!}f^{(n)}(x) = f(x+a)\,. $$ Assuming $f \in C^\infty(\mathbb{R}) \cap L^2(\mathbb{R})$. In order to exponentiate $\hat{Q} + \hat{P}$ you may use Baker–Campbell–Hausdorff formula. $$ e^{\hat{Q}}e^{\hat{P}} = e^{\hat{Q}+\hat{P} + \frac12 i}\,, $$ where I used $[\hat{Q},\hat{P}] = i \mathbb{1}$ and the fact that all other terms are commutators with $\mathbb{1}$ which vanish.

$\endgroup$
  • $\begingroup$ Right, this is how I typically think about these things, but I'm not sure how to make sense of this formal series mathematically $\endgroup$ – Questions Apr 16 at 5:01
  • $\begingroup$ en.wikipedia.org/wiki/… $\endgroup$ – MannyC Apr 16 at 5:03

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.