Set-up: We are given $N$ spin $\frac{1}{2}$ particles, with associated Pauli operators $\{X_i, Y_i, Z_i\}_{i=1}^N$. We can quickly verify that the operators $\bar{X}=X_1+X_2+\ldots X_n$, $\bar{Y}=Y_1+Y_2+\ldots Y_n$, $\bar{Z}=Z_1+Z_2+\ldots Z_n$ satisfy the commutation relations

$$[\bar{X},\bar{Y}]=2i \bar{Z}, [\bar{Y},\bar{Z}]=2i \bar{X}, [\bar{Z},\bar{X}]=2i \bar{Y}.$$

Thus, $\frac{1}{2}\bar{X}, \frac{1}{2}\bar{Y}, \frac{1}{2}\bar{Z}$ satisfy the angular momentum commutation relation (with $\hbar$ set to $1$ for convenience).

The decomposition of $\frac{1}{2}\bar{X}, \frac{1}{2}\bar{Y}, \frac{1}{2}\bar{Z}$ in terms of irreducible representations is also well understood. For example, answers in this and this stack-exchange question discuss the number of times a spin $\frac{k}{2}$ irreducible representation appears in the decomposition. Let us formally write the decomposition as:

$$\frac{1}{2}\bar{X}=\bigoplus_{\ell} J_x^{(\ell)}, \frac{1}{2}\bar{Y} = \bigoplus_{\ell} J_y^{(\ell)}, \frac{1}{2}\bar{Z}=\bigoplus_{\ell} J_z^{(\ell)}$$

where $\ell$ labels each irreducible decomposition and $J_x^{(\ell)},J_y^{(\ell)}, J_z^{(\ell)}$ are the matrices in the $\ell-$th decomposition.

Question: Is there an explicit expression for $J_x^{(\ell)},J_y^{(\ell)}, J_z^{(\ell)}$ (for each $\ell$) in terms of Pauli operators $\{X_i, Y_i, Z_i\}_{i=1}^N$? There is clearly some expression, since Pauli operators span the space of matrices on $N$ spin $\frac{1}{2}$ particles. But I am looking for an explicit one, which hopefully would be a nice multi-variate polynomial in the Paulis.

  • 2
    $\begingroup$ You do know the answer for N=2: it is the Clebsch reduction series. You are also expected to appreciate the complexity for N=3. And you dare expect something "nice" for higher Ns? $\endgroup$ Mar 16, 2022 at 20:08
  • $\begingroup$ @CosmasZachos , its fine to not have a ``nice'' expression. But is there an explicit expression? $\endgroup$ Mar 20, 2022 at 18:29
  • $\begingroup$ Of course not. You believe the standard answers provided on this site multiple times for N=3 are "explicit"? You understand the Clebsch series for N=2 explicit? That's precisely what I'm asking you. $\endgroup$ Mar 20, 2022 at 18:31
  • $\begingroup$ You want an "explicit" version of the $2^N\times 2^N$ matrix T illustrated here for N =2, right? There is no such thing available. $\endgroup$ Mar 22, 2022 at 15:17

1 Answer 1


Each matrix $J_i^{(l)}$ is an angular momentum matrix in some irreducible representation. As mentioned, each decomposition $l$ corresponds to some (nonunique) spin-$k/2$. All that is left is writing the angular momentum matrices for a spin-$k/2$ system, which is a common exercise.

Your question is now whether a given spin matrix in a given dimension can easily be written in terms of Pauli matrices. The answer is actually no, because tensor products of Pauli matrices always correspond to $2^n$-dimensional matrices for some positive integer $n$, while spin matrices must have dimension $k+1$. For example, a spin-$1$ matrix may look like $$J_z=\frac{1}{2}\begin{pmatrix}1&0&0\\0&0&0\\0&0&-1\end{pmatrix}.$$ There's no way of writing this in terms of Pauli matrices; in fact, since $3$ is a prime number, one cannot decompose this matrix in terms of sums of tensor products of matrices of smaller sizes (other than trivial things like scalars multiplied by $3\times 3$ matrices). It is thus impossible in general to write the angular momentum matrices in terms of Pauli matrices alone.

The generalizations of Pauli matrices that you are looking for are probably these spin matrices themselves. Alternatively, you can look at Gell-Mann matrices or clock and shift matrices as described in this answer for generalizing Pauli matrices.

The updated question says we are not directly interested in the matrices $J_i^{(l)}$ but actually in representing the matrices acting on the full system, like $0\oplus 0\oplus J_i^{(l)}\oplus 0\cdots$. Of course these can be achieved, but the results shouldn't be expected to be illuminating...

To make the connection, we need to express each single-element matrix $E_{jk}$ as a linear combination of tensor products of Pauli matrices. This can be done by expanding the locations of the nonzero elements in binary with $N$ digits, like $j=0101000\equiv\mathbf{j}$, to establish the connection $E_{jk}=|\mathbf{j}\rangle\langle \mathbf{k}|$. In this fashion, we have $E_{jk}=|j_1\rangle\langle k_1|\otimes\cdot\otimes |j_N\rangle \langle k_N|$. Then, we use Pauli matrices as a basis for each $|j_i\rangle\langle k_i|$ to write $E_{jk}$ as a linear combination of tensor products of Pauli matrices. Finally, we express each $J_i^{(l)}$ as a linear combination of the single-element matrices $E_{jk}$, which is the same as expanding them in the computational basis, so the matrix elements follow directly from the standard definitions of the angular momentum matrices (like $J_z$ in dimension $3$ quoted above). There are lots of indices flying around, from the location of the $l$th block to the location within the $l$th block, as well as the bit strings vs the locations corresponding to those bit strings, but it is all doable if one insists.

  • $\begingroup$ "The answer is actually no, because tensor products of Pauli matrices always correspond to 2n-dimensional matrices for some positive integer n, while spin matrices must have dimension k+1." I mean writing the $J_i^{\ell}$ matrices as a linear combination of N-qubit Pauli matrices. This is always possible since the $4^N$ Pauli matrices span the space of matrices on N qubits. But I am looking for an explicit expression. $\endgroup$ Mar 20, 2022 at 18:27
  • $\begingroup$ @anuraganshu the big matrices that you have defined as $\bar{X}$ etc can be spanned by Pauli matrices, but the $J_i^l$ matrices have different dimensions $\endgroup$ Mar 21, 2022 at 12:58
  • $\begingroup$ @anuraganshu linear combinations of matrices in some dimension won't take you out from that dimension $\endgroup$ Mar 21, 2022 at 15:38
  • $\begingroup$ We can view $J^{(\ell)}$ matrices as N qubit matrices by putting the zero matrix in every other block. When I write $\frac{1}{2}\bar{X}=\oplus_{\ell} J^{(\ell)}_x$, I am viewing $J^{(\ell)}_x$ as some matrix in the $\ell$-th block followed by zeros in all the other block. $\endgroup$ Mar 22, 2022 at 0:22
  • $\begingroup$ @anuraganshu that is not what the notation means, but of course you are free to come up with your own notation. In your notation you would not need the $\oplus$ symbol but rather the $+$ symbol because you would be adding matrices in the same space $\endgroup$ Mar 22, 2022 at 0:38

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.