3
$\begingroup$

The block-encoding framework shows the following statement in general, as discussed in the paper https://arxiv.org/abs/1804.01973.

A block-encoding of a matrix $A \in \mathbb{C}^{N \times N}$ is a unitary $U$ such that the top left block of $U$ is equal to $A / \alpha$ for some normalizing constant $\alpha \geq\|A\|$: $$ U=\left(\begin{array}{cc} A / \alpha & \\ \cdot & \cdot \end{array}\right) $$ In other words, for some $a$, for any state $|\psi\rangle$ of appropriate dimension, $$\alpha\left(\left\langle\left. 0\right|^{\otimes a} \otimes I\right) U\left(|0\rangle^{\otimes a} \otimes\right.\right |\psi\rangle)=A|\psi\rangle.$$

My question is to find out the dimension of the elements from $$\alpha\left(\left\langle\left. 0\right|^{\otimes a} \otimes I\right) U\left(|0\rangle^{\otimes a} \otimes\right.\right.|\psi\rangle)=A|\psi\rangle.$$ I am not sure where $a$ came from in $|0\rangle^{\otimes a}$, and not sure what the dimension of $I$ is in this expression.

Any help could be highly appreciatedted!

$\endgroup$

1 Answer 1

3
$\begingroup$

$U$ can be written as: $$U=\begin{pmatrix}\frac{1}{\alpha}A&B\\C&D\end{pmatrix}$$ while $|0\rangle^{\otimes a}\otimes|\psi\rangle$ can be written as: $$|0\rangle^{\otimes a}\otimes|\psi\rangle=\begin{pmatrix}|\psi\rangle\\0\end{pmatrix}$$ We thus have: $$U\left(|0\rangle^{\otimes a}\otimes|\psi\rangle\right)=\begin{pmatrix}\frac{1}{\alpha}A&B\\C&D\end{pmatrix}\begin{pmatrix}|\psi\rangle\\0\end{pmatrix}=\begin{pmatrix}\frac{1}{\alpha}A|\psi\rangle\\C|\psi\rangle\end{pmatrix}$$ and then: $$\alpha\left(\langle0|^{\otimes a}\otimes I\right)U\left(|0\rangle^{\otimes a}\otimes|\psi\rangle\right)=\alpha\begin{pmatrix}I & 0\end{pmatrix}\begin{pmatrix}\frac{1}{\alpha}A|\psi\rangle\\C|\psi\rangle\end{pmatrix}=A|\psi\rangle$$

Is is now easy to see that $I$ must have the same size as $A$, which is the $N\times N$ identity matrix, while $a$ is equal to $\frac{n}{N}$, where $n$ is the dimension of $U$. You can also see in the last equation that the $\alpha$ at the beginning of the product cancels out with the $\frac{1}{\alpha}$ factor applied to $A$ in $U$.

$\endgroup$
3
  • 1
    $\begingroup$ Thank you for the detailed explanation and this helps a lot! But I thought the dimension for the tensor product of two matrices is multiplicative, not additive?. $\endgroup$ Sep 15, 2021 at 18:00
  • $\begingroup$ @JohnParker You are totally right! I've updated the answer accordingly $\endgroup$ Sep 16, 2021 at 12:03
  • $\begingroup$ Well, to me this seems like an abuse of notation, not really clear. $\endgroup$
    – dnnagy
    Nov 21, 2022 at 20:25

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.

Not the answer you're looking for? Browse other questions tagged or ask your own question.