I learned that a unitary matrix generated by time-dependent Hamiltonians is written down as

\begin{equation} U(t) = \mathcal{T}\exp\Big(-i\int_0^t H(t') dt' \Big),\tag{1} \end{equation} where $\mathcal{T}\exp$ is the time-ordered exponential operator.

My question is: is the $k$th power of the unitary operator $$U^k(t) = \mathcal{T}\exp\Big(-i\int_0^t kH(t') dt' \Big)~?\tag{2}$$

  • 2
    $\begingroup$ Have you tried a simple example with three time points? $\endgroup$ Aug 3, 2023 at 0:27
  • $\begingroup$ Also, don't forget that the notation of a "time-ordered exponential" is just a notational convenience. It must be taken to mean the power series of time-ordered operators. The "time-ordering operator" is not a linear operator and we are not "applying it" linearly to each term in an expansion of an exponential. $\endgroup$
    – hft
    Aug 5, 2023 at 15:53

2 Answers 2


TL;DR: No, OP's eq. (2) does not necessarily hold.

Counterexample. Take e.g. $$H(t)~=~A~ \delta(t-\frac{1}{3})~+~B~ \delta(t-\frac{2}{3}), $$ where the operators $A$ and $B$ do not commute: $[A,B]\neq 0$. Then $$ T\exp\left[2\int_0^1\! dt~H(t)\right]~=~e^{2B}e^{2A},$$ while $$ T\left[\exp\int_0^1\! dt~H(t)\right]~ T\left[\exp\int_0^1\! dt~H(t)\right]~=~e^Be^Ae^Be^A.$$

  • $\begingroup$ Thanks for the great answer. Actually I confronted a small confusion while rethinking about this problem after looking at your answer. Since the superoperator $\text{ad}_{H(t)}$ commutes with itself for any choice of $H(t)$, isn't the following equality hold: $T \exp(-i \int_0^T \text{ad}_{2H(t)} dt) = T\exp(-i \int_0^T \text{ad}_{H(t)} dt) T\exp(-i \int_0^T \text{ad}_{H(t)} dt) = \mathcal{U} \circ \mathcal{U}$, where $\mathcal{U}$ is the superoperator, i.e. $\mathcal{U}(\rho) = U \rho U^\dagger$?. This contradicts to your answer. What am I missing here? $\endgroup$
    – Hailey Han
    Aug 6, 2023 at 21:47

For time dependent Hamiltonian $U$ does not necessarily have the time translation symmetry. You should write $U(t,t')$ and then I don't really know what $U^k$ represents.

For time independent Hamiltonian $U^k(t) = U(kt)$ is just a time propagation by $kt$ due to the group structure of $U$ operators. On the other hand $ U(kt) = \mathcal T \exp\left(-i \int_0^{kt} H dt'\right) = \mathcal T \exp\left(-i \int_0^{t} k H dt''\right)$

which is what you asked for.

  • 3
    $\begingroup$ If the Hamiltonian is not time dependent the time ordering is the identity and everything is trivial…. $\endgroup$ Aug 3, 2023 at 7:16
  • $\begingroup$ Yeah, you're right $\endgroup$ Aug 3, 2023 at 7:22

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.