I am reading the paper "The quantum-to-classical transition and decoherence" by Maximilian Schlosshauer (https://arxiv.org/abs/1404.2635). I doubt the most basic claims in this paper.

The story is as follows: an isolated microscopic system $S$ (Hilbert space $H_S$) is put into a superposition $\phi=\phi_1+\phi_2$ of states $\phi_1,\phi_2\in H_S$ and then brought into contact with an environment $E$ (Hilbert space $H_E$) which initially is in state $e\in H_E$. Before interaction the composite system $(S,E)$ is in state $$ \Phi=\phi\otimes e = \phi_1\otimes e+\phi_2\otimes e\in H_S\otimes H_E. $$ Now an interaction between $S$ and $E$ takes place and it is claimed that this puts the combined system $(S,E)$ into a state of the form $$ \Psi = \phi_1\otimes e_1+\phi_2\otimes e_2\in H_S\otimes H_E $$ (cited paper, equation (1), p2). This is surprising, since we do not know anything about the interaction (no assumptions made).

I believe that such a principle cannot be maintained: we can write $\phi$ in other ways as a superposition $\phi=\psi_1+\psi_2$ and it must then follow that $\Psi$ is also of the form $$ \Psi = \psi_1\otimes f_1 + \psi_2\otimes f_2\in H_S\otimes H_E, $$ for some vectors $f_1,f_2\in H_E$.

In other words the vectors $e_1,e_2\in H_E$ have the following interesting property: $$ \psi_1+\psi_2=\phi_1+\phi_2\ \Rightarrow\ \exists f_1,f_2\in H_E:\thinspace \phi_1\otimes e_1+\phi_2\otimes e_2 = \psi_1\otimes f_1 + \psi_2\otimes f_2. $$ We will now exhibit an example where such vectors $e_1,e_2$ do not exist. Let $H_S=H_E=L^2([0,1])$. Then the tensor product $H_S\otimes H_E$ is isomorphic to the Hilbert space $L^2([0,1]^2)$ with a canonical isomorphism mapping $f(x)\otimes g(y)$ to the function $f(x)g(y)\in L^2([0,1]^2)$.

Now let $\phi_1(x)=1$, $\phi_2(x)=x$, $\psi_1(x)=1-\sqrt x$ and $\psi_2(x)=x+\sqrt x$. Then $\phi_1+\phi_2=\psi_1+\psi_2$ but there are no functions $e_1(y),e_2(y)\in H$ such that the function $\phi_1(x)e_1(y)+\phi_2(x)e_2(y)$ can be written in the form $$ \phi_1(x)e_1(y)+\phi_2(x)e_2(y) = \psi_1(x)f_1(y)+\psi_2(x)f_2(y), $$ for some functions $f_1,f_2\in H$ since this amounts to the equality $$ e_1(y)+xe_2(y) = (1+\sqrt x)f_1(y)+(x-\sqrt x)f_2(y), $$ for almost all $x,y\in[0,1]$. Simply evaluate the equality at enough different values of $x$ to obtain an overdetermined system of equations in $e_i(y),f_j(y)$ with no solution. Clearly this same argument works if we replace $H=L^2([0,1])$ with a finite dimensional Euclidean space of dimension at least 5.

Why is this not a problem?

  • $\begingroup$ From the paper, $S$ assumed to be two-dimensional. Thus, any state will be of the form $\phi_1\otimes e_1 + \phi_2\otimes e_2$, if $\phi_1$ and $\phi_2$ for a basis for $S$, which means the initial state must evolve into a state of that form. So whatever calculations you've done can't be a problem. $\endgroup$
    – march
    Commented Dec 30, 2023 at 18:23
  • 1
    $\begingroup$ @march OK, thanks, I overlooked this, since it is not explicitly stated in the paper. If you make it an official answer it will be accepted. $\endgroup$
    – gcc
    Commented Dec 30, 2023 at 20:14

1 Answer 1


In the paper, the states that OP labels $\phi_1$ and $\phi_2$ correspond to the states $\lvert s_1\rangle$ and $\lvert s_2 \rangle$, which correspond to passage through one of two slits in a double-slit experiment. For that reason, the Hilbert space $S$ is two-dimensional, and so any state in the tensor product space of $S$ and $E$ can be written in the form $$ \phi_1\otimes e_1 + \phi_2\otimes e_2\,, $$ where $s_j$ are some arbitray states in $E$. In particular, the state after some time-evolution must be write-able in this way, so there can be no contradiction.

The issue with the OP's "counterexample" is that they are implicitly assuming that $S$ is of dimension greater than 2, because they are using the four states $\phi_1(x)=1$, $\phi_2(x)=x$, $\psi_1(x)=1-\sqrt x$, and $\psi_2(x)=x+\sqrt x$, which spans a space of dimension 3.

  • $\begingroup$ It should be noted that the paper is very confusing as a position variable x is postulated on the 2 dimensional space for S, see equation (3) and first line of p3 (the inner product (x,s_i) is defined. This would make the Hilbert space H_S infinite dimensional. $\endgroup$
    – gcc
    Commented Dec 31, 2023 at 19:23

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.