Consider the Hamiltonian $f(t)H,$ where $H$ is time-independent and $f(t)$ is classical white noise. Then I would write a Schrodinger equation

$$\mathrm{d}\psi=-iH\psi\ \mathrm{d}W_t,$$

where $W_t$ is a Wiener process. But using Ito calculus, the solution is un-normalized$^{[a]}$: $$\psi(t,W_t)=e^{+H^2 t/2}e^{-iHW_t}\psi(0) \, .$$

What happened here? I could modify the Schrodinger equation to get rid of the $e^{H^2 t/2}$ term, but I don't know how to justify that.

$[a]$: You can get this solution as a special case of the example on Wikipedia: $$\mathrm{d}S_t=S_t(\mu\mathrm{d}t+\sigma\mathrm{d}W_t)\Rightarrow S(t,W_t)=S(0)e^{(\mu-\sigma^2/2)t + \sigma W_t}.$$

  • $\begingroup$ Not sure if this really helps or if it is at all what you're asking, but $\operatorname{E}(S)=\operatorname{E}(S_0e^{(\mu-\sigma^2/2)t+\sigma W_t})=e^{\mu t}$, "thus" the drift term. Or I guess had you gone with Stratonovich as one usually does in physics, $\mathrm{d}S_t=S_t(\mu\mathrm{d}t+\sigma\circ\mathrm{d}W_t) \Rightarrow S_t=S_0e^{\mu t+\sigma W_t}$. $\endgroup$
    – alarge
    Jan 26, 2019 at 13:17

1 Answer 1


The Schrödinger equation you wrote conserves the averaged probability density if it is seen in the Stratonovich form. Now, its solution, which is normalized if $||\psi(0)||^2 = 1,$ would be $$\psi(t) = U(t)\psi(0) = \exp(-iX_tH)\psi(0) \quad (*), $$ where $X_t = \int_0^t \xi(s)d s$ is the Brownian motion in $\mathbb{R}$ and the integral is in the Stratonovich form, i.e., whenever evaluating, say, expectation values and delta functions ($\Theta(t)$) pop-up, you must take $\Theta(0) = 1/2$ --Conversely, in the Ito form, $\Theta(0) = 0$.

However, if you want to write your equation in the Ito form, the correct equation would be $$ d \psi(t) = -iH\psi(t)d W_t - \frac{1}{2}H^2\psi(t)d t. \quad (**) $$ One can demonstrate this equation either by just transforming Eq. (*) to the Ito form (see page 11 in https://arxiv.org/abs/quant-ph/9702030), or by more "physical" arguments (see, e.g., page 45 of van Kampen's Stochastic processes in physics and chemistry). Namely, starting with a stochastic Schrödinger equation with two generators: a unitary, fluctuating one $-iHd X_t$; and a dissipative one of the form $-V d t$, i.e., $$ d \psi(t) = -iH \psi d X_t - V\psi(t)d t. $$

After imposing the conservation of averaged probability density in infinitesimal times, i.e., $\mathbb{E}\left[||\psi(t + d t)||^2\right] = \mathbb{E}\left[||\psi(t)||^2\right]$, one finds that $U = \frac{1}{2}H^2$.

To conclude, Eq. (*) is normalized and, when taken in the Ito form, solves Eq.(**): $$ \begin{aligned} d \psi(t) &= \psi(t + d t ) -\psi(t) \\ &=\left[U(t+d t,t)- I\right]U(t,0) \psi(0)\\ &= \left[\exp(-iH d X_t) - I \right]\psi(t) \\ &= -iH\psi(t)d X_t - \frac{1}{2}H^2\psi(t) d t. \end{aligned} $$


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.