If given a wave function which is an element of a Hilbert space $\mathcal{H}$ which is a tensor product of two other Hilbert spaces $\mathcal{H}_1\otimes\mathcal{H}_2$, how do you formulate measurement of the function in one of the sub-spaces, i.e. $\mathcal{H}_1$?

For example, say you have a system $\left|\psi\right>$ which consists of two qubits $\left|\psi_1\right>\otimes\left|\psi_2\right>$. The system undergoes a unitary operation, giving $\left|\psi'\right>=U\left|\psi\right>$. It should still be possible to measure just one of the two qubits, causing a partial collapse of the system.

If the two qubits are not entangled after $U$, then they can be separated as $\left|\psi'\right>=\left|\psi_1'\right>\otimes\left|\psi_2'\right>$. It seems clear that measuring i.e. $\left|\psi_1'\right>$ should not cause $\left|\psi_2'\right>$ to collapse.

If the qubits are partially entangled, it may still be possible for one of the qubits to be in some sort of superposition. For example, if $\left|\psi'\right>=\frac{\left|00\right>+\left|01\right>+\left|11\right>}{\sqrt{3}}$, then $\left|\psi_2'\right>$'s state depends on what $\left|\psi_1'\right>$ collapses to. If $\left|\psi_1'\right>\rightarrow\left|0\right>$, then could $\left|\psi_2'\right>$ still be in a superposition, namely $\frac{\left|0\right>+\left|1\right>}{\sqrt{2}}$?

To be more specific about the "formulation" part of my question, you can use e.g. $\left|\left<0\mid\psi_1'\right>\right|^2$ to find the probability that $\left|\psi_1'\right>$ collapses to $\left|0\right>$, assuming a normalized wavefunction. What would be the format for e.g. measuring a single qubit of $\left|\psi'\right>$? Would it be something like $\left(\left<0\right|\otimes\left<\psi_2'\right|\right)\left|\psi'\right>$?

Some insight I'm looking into: the expectation value of an operator $O$ is given by $\left<\psi\mid O\mid\psi\right>$. When $O$ is a projection operator, such as $\left|0\right>\left<0\right|$ (which projects to a $\left|0\right>$-basis), the expectation value is the probability that measurement yields that particular outcome. So perhaps partial measurement of a system involves projection to some subspace spanned by multiple basis states?

  • $\begingroup$ Yes your intuition is correct. You can use these 'partial' scalar product. Since the notation can be confusing I suggest you add lowerscript as in e.g., $_{A}\langle \chi | \psi \rangle_{AB}$ if you measure system $A$. $\endgroup$ – lcv Jan 23 '19 at 1:08

In general, such "partial collapse" of a wavefunction cannot be expressed as the usual inner product (i.e. using $\left<0\mid\psi\right>$). Instead, you must use the expectation value of the corresponding projection operator. This works for "full collapse" to a single basis state, as the corresponding projection operator is just the outer product of the basis with itself ($\left|0\right>\left<0\right|$).


Using the example I gave above, let $P$ be the operator that projects the first qubit to $\left|0\right>$. This can be represented by the matrix $\begin{bmatrix}1&0&0&0\\0&1&0&0\\0&0&0&0\\0&0&0&0\end{bmatrix}$. Suppose there exists some basis state $\left|B\right>$ that represents $\left|\psi_1'\right>=\left|0\right>$ while $\left|\psi_2'\right>$ is unaffected. $P$ can then be written as $\left|B\right>\left<B\right|$. If $\left|B\right>$ is given by the matrix $\begin{bmatrix}x\\y\\z\\w\end{bmatrix}$, then $\left|B\right>\left<B\right|=\begin{bmatrix}xx^*&xy^*&xz^*&xw^*\\yx^*&yy^*&yz^*&yw^*\\zx^*&zy^*&zz^*&zw^*\\wx^*&wy^*&wz^*&ww^*\end{bmatrix}$. Equating with the earlier matrix gives the following equations:

  • $z=w=0$
  • $\left|x\right|^2=\left|y\right|^2=1$
  • $xy^*=x^*y=0$

Which has no solution. Therefore, $\left|B\right>$ does not exist.

This means that the probability of measuring part of a system must be expressed using operator expectation, $\left<\psi\mid P\mid\psi\right>$.

It is possible, however, for $P$ to be expressed as a sum of outer products. In this case, $P=\left|00\right>\left<00\right|+\left|01\right>\left<01\right|$. This could be generalized to be a sum of projections to each basis of the subspace.

| cite | improve this answer | |
  • $\begingroup$ Of course, you can expand it as a sum of overlap-squares. $\endgroup$ – Norbert Schuch Nov 8 '16 at 20:40

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.