In this paper https://arxiv.org/abs/math-ph/9901018 the author (on page 4) mentions : "...a standard argument for the absolute convergence of the Dyson series for [the evolution operator] $\hat{U}_{\tau}$ ..." .

Could you please give me a hint of what is a standard way of proving the absolute convergence of the Dyson series for the evolution operator? References are welcome.

(I am not a mathematician, but a physicist)

  • $\begingroup$ Please don't let us do all the work: What is the operator $\hat{U}_\tau$, so at least give us the Hamiltonian, then it is possible to say more. $\endgroup$ – Luke May 23 '17 at 12:33
  • $\begingroup$ Thanks for your reply. The paper just says that the Hamiltonian is a self-adjoint, bounded operator. $\endgroup$ – Salsiccio May 23 '17 at 13:49

By `Dyson series' I suppose you mean the following expression (with $t\geq t_0$): $$D (t,t_0) = \sum_{k=0}^\infty \left( - \frac{i}{\hbar} \right)^k \int_{\Delta^k(t,t_0)} d^k t \, \prod_{l=1}^k V (t_k)$$ $V\colon t \mapsto V(t)$ is a map on the reals, yielding a linear (say bounded) operator $V(t)$ on your Hilbert space, and $\Delta^k(t,t_0)$ is the geometric simplex $$\Delta^k(t,t_0) = \left\lbrace (t_1, \dots, t_k) \in \mathbb{R}^k\middle\vert t \geq t_1 \geq t_2 \geq \dots \geq t_k \geq t_0 \right\rbrace . $$

A sufficient and natural condition for convergence of the series is obtained by first defining $$I \left( t, t_0 \right) := \int_{t_0}^t d t' \, \left\Vert V (t')\right\Vert \, ,$$ and then observing that the last expression of \begin{align} D(t,t_0) &\leq \sum_{k=0}^\infty \frac{1}{\hbar^k} \, \left\Vert \int_{\Delta^k(t,t_0)} d^k t \, \prod_{l=1}^k V (t_k) \right\Vert \\ & \leq \sum_{k=0}^\infty \left( \frac{I (t,t_0)}{\hbar} \right)^k \end{align} is a geometric series. Thus, we obtain (absolute) convergence for $$\int_{t_0}^t d t' \, \left\Vert V (t')\right\Vert \leq \hbar \, .$$ Note that absolute convergence is a must, as else you run into serious problems with defining the limit in the first place.


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.