On way to describe "pure" decoherence (that is, decoherence with respect to a basis that doesn't involve transitions between basis states) between a system $\mathcal{S}$ and an environment $\mathcal{E}$ is with a unitary operator that acts conditional on the state of $\mathcal{S}$: $$ U=\sum_i \vert S_i \rangle \langle S_i \vert \otimes U^\mathcal{E}_i. $$ Here, $U^\mathcal{E}_i$ describes the evolution of $\mathcal{E}$ conditional on $\mathcal{S}$ being in state $\vert S_i \rangle$. A CP map on $\mathcal{S}$ constructed by applying $U$ and tracing out $\mathcal{E}$ will leave fixed the diagonal matrix elements of the initial state $\rho_0^\mathcal{S}$ in the basis $\{\vert S_i \rangle\}$, but will in general add phases and decoherence (i.e. suppression of the norm) to off-diagonal elements. Importantly, $U$ is a valid unitary for any choice of unitary $U^\mathcal{E}_i$.

Now suppose I want to construct a unitary that acts like this on an overcomplete basis. I'll specialize to the coherent states $\vert \alpha \rangle$ where $\alpha = x + i p$ is a point in phase space, but the solution for a general basis would be interesting. If I write down $$ U= \int \mathrm{d}\alpha \vert \alpha \rangle \langle \alpha \vert \otimes U^\mathcal{E}_\alpha. $$ one can check that this is not a valid unitary for arbitrary $U^\mathcal{E}_\alpha$. (This can be easily seen when $\mathcal{S}$ has only two dimensions using an overcomplete basis of 3 or 4 vectors, and choosing random conditional unitaries.)

Is there a compact way to write down the requirements on the $U^\mathcal{E}_\alpha$ for $U$ to be unitary? Obviously one can expand $U^\dagger U = I$ using the above definition, but I'm unable to transform this to something with a clear interpretation. My intuition is that $U^\mathcal{E}_\alpha$ and $U^\mathcal{E}_\beta$ should be "close" for $\vert \alpha - \beta\vert^2 \ll 1$ since they ought to be unrestricted for $\vert \alpha - \beta\vert^2 \gg 1$, but I can't formalize this in a useful way.

(Incidentally, this is closely related to my previous questions which did not generate much interest. I'm giving this one last shot.)

Edit: In the case where the unitary is taking the environment from a certain initial state $\vert E_0 \rangle$ to a conditional state $\vert E_\alpha \rangle$, one can see that the associated CP map is well behaved (i.e. that $\sum_i K_i^\dagger K_i = I$, where the $K_i$ are the Kraus operators) iff the operator $$ \int \mathrm{d}\alpha\mathrm{d}\beta \vert \alpha \rangle \langle \alpha \vert \beta \rangle \langle \beta \vert \cdot f(\alpha,\beta) $$ equal the identity, where $f(\alpha,\beta) = \langle E_\alpha \vert E_\beta \rangle$ is the Gram matrix of inner products of the (normalized) conditional states. A sufficient condition for this is that $f(\alpha,\beta)$ depends only on $\alpha - \beta$ but not $\alpha + \beta$. I believe this is a necessary condition, but I am unable to prove it (or find a counter-example). This condition, along with the fact that $f(\alpha,\beta)$ is a Gram matrix, makes $f(\alpha-\beta)$ a positive-definite function, which I presume is important but don't know how to exploit it.



Your Answer

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