0
$\begingroup$

I think I didn't not really understand a comment found on "Quantum Mechanics" by Claude Cohen-Tannoudji.

Talking about the fifth postulate of quantum mechanics in the case in which $a_n$ is a degenerate eigenvalue is said

"We stress the fact, however, that is not an arbitrary ket of the eigensubspace $E_n$, but the part of $ | \psi \rangle $ which belongs to $E_n$".

I did not understand very well probably. In that case I would say that immediately after the measure of $a_n$ the state of the system is a linear combination of the bases of $E_n$, eigensubspace related to $a_n$.

Fifth postulate:

Immediately after the measurement of an observable $A$ has yielded a value $a_n$, the state of the system is the normalized eigenstate $ | u_n \rangle $.

$\endgroup$
  • 3
    $\begingroup$ There is no standard way of numbering the postulates of QM. Could you include the "5th postulate" in your question? $\endgroup$ – By Symmetry May 22 '18 at 12:34
  • $\begingroup$ I'm sorry, I added it. $\endgroup$ – Stefano Barone May 22 '18 at 12:40
  • $\begingroup$ You need to define the notation. Nobody can tell what this is about based on these two very brief snippets. $\endgroup$ – Ben Crowell May 22 '18 at 13:07
  • $\begingroup$ For further reference @Stefano Barone, please use "von Neumann projection postulate" or the "collapse postulate" in Quantum Mechanical slang. $\endgroup$ – DanielC May 22 '18 at 14:16
  • $\begingroup$ What is your question? $\endgroup$ – sammy gerbil May 26 '18 at 17:53
2
$\begingroup$

Let $\mathscr{H}$ be the relevant Hilbert space and $\Psi\in \mathscr{H}$. Let furthermore, $P : \mathscr{H}\to \mathscr{H}$ be a projection operator onto some subspace. It is a fact that $P$ defines the complementary projection $Q : \mathscr{H}\to \mathscr{H}$ given by $Q = \mathbf{1} - P$ where $\mathbf{1}$ is the Hilbert space identity transformation.

Notice that by definition a projection satisfies $P^2 = P$. It turns out that $Q$ is indeed a projection operator. We have the proof:

$$Q^2=(1-P)^2=1-2P+P^2=1-2P+P=1-P=Q,$$

where we've used that $P$ is a projection. Another fact is that $PQ\Psi =0$. This is true because we have

$$P(Q\Psi)=P(1-P)\Psi=P\Psi-P^2\Psi=P\Psi-P\Psi=0.$$

Similarly $QP\Psi = 0$ since $[P,Q]=0$.

But if $P$ and $Q$ are projections, they are projections onto where? Well, we can define

$$\mathscr{B}_p=P(\mathscr{H})$$

$$\mathscr{B}_Q=P(\mathscr{H}).$$

These are the subspaces of $\mathscr{H}$ onto which the projectors project. They are not independent however, since these are complementary projections. We have the following result:

$$\mathscr{H}=\mathscr{B}_P\oplus \mathscr{B}_Q.$$

Indeed, by definition of $Q$ we have that $\mathbf{1} = P+Q$ and hence any $\Psi$ is

$$\Psi=P\Psi+Q\Psi.$$

The decomposition is unique, however. Indeed if $\Psi=\Psi_P+\Psi_Q$ with $\Psi_P\in \mathscr{B}_P$ and $\Psi_Q\in \mathscr{B}_Q$ then since $PQ\Psi=QP\Psi=0$ by applying $P$ and $Q$ to the decomposition it follows $\Psi_P = P\Psi$ and $\Psi_Q = Q\Psi$.

This indeed shows that the decomposition is a direct sum decomposition.

Why bother with all this? Well, if $A$ is an observable, corresponding to each eigenvalue there is an eigenspace. The eigenspace corresponding to $\lambda$ has a projector $P_\lambda$ associated to it. Then $Q_\lambda = 1-P_\lambda$ projects out of it.

Thus, all of this allow us to say that a state $\Psi$ may have a "part of it" contained in the eigenspace of $A$ with eigenvalue $\lambda$ and this part is $P_\lambda \Psi$.

The properties of a projection are what allow us to say that $P_\lambda \Psi$ is its part inside that eigenspace.

What the postulate says is that immediately after the measurement the state is $$\frac{P_\lambda \Psi}{\sqrt{(P_\lambda \Psi, P_\lambda \Psi)}}$$

since this projection won't usually be normalized. In summary what is says is: the system doesn't collapse to an arbitrary element of $P_\lambda \mathscr{H}$ but rather to specificaly the part of $\Psi$ lying there properly normalized, which is given by the above formula.

$\endgroup$

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.