$L_x$ and $L_y$ don't commute since $[L_x,L_y] = i \hbar L_z$. From the uncertainty principle, we know that $\Delta L_x \cdot \Delta L_y \geq \frac{1}{2} \cdot |\langle i \hbar L_z \rangle|$. There would be no simultaneous eigenstates of $L_x$ and $L_y$ if $\langle L_z \rangle$ was zero for all eigenstates of $L_x$ and $L_y$. Is it possible to show that ?

Otherwise I would be happy with an other explanation of why or why not there would be simultaneous eigenstates of $L_x$ and $L_y$.


3 Answers 3


According to the closure relationship we have for any two components of the angular momentum: $$[\hat{J}_i,\hat{J}_j]=i\hbar\epsilon_{ijk}L_k\tag{1}$$ where $\epsilon_{ijk}$ is Levi-Civita symbol and the sum is understood over the repeated indexes. A non-zero commutator implies that there is no common basis of eigenvectors.

That said, common eigenstates can still exist: in this case the state vector such that all components of angular momentum are zero. Please notice this happens only because the commutator is an operator: if it were a number (I mean a multiple of the identity operator), not a single common eigenstate would be possible. This is the case of position and momentum. Consider the 1D case: $$[\hat{x},\hat{p}]=i\hbar\mathbb{1}\tag{2}$$ Suppose there exists a common eigenstate $|\alpha\rangle$ such that: $$\hat{x}|\alpha\rangle=x|\alpha\rangle \\ \hat{p}|\alpha\rangle=p|\alpha\rangle$$ Then $$[\hat{x},\hat{p}]|\alpha\rangle=xp|\alpha\rangle-px|\alpha\rangle=0$$ that is contradiction with $(2)$.

On the other hand if the commutator were an operator as in $(1)$ it could possibly be zero for that particular state, as it happens for angular momentum.

  • 1
    $\begingroup$ The first part of your answer is good but I don't think it's wise in the second part to use $\hat x$ and $\hat p$ as an example given it is plagued by subtleties with normalization and unboundedness. $\endgroup$ Jan 3, 2022 at 17:45
  • $\begingroup$ Do you mean because their eigeinbases are non-normalizable states? $\endgroup$ Jan 3, 2022 at 17:48
  • 2
    $\begingroup$ In part yes. Also because in finite dimensional spaces as there is the contradiction of $\hbox{Tr}[\hat x,\hat p]=\hbox{Tr}(\hat x \hat p)-\hbox{Tr}(\hat p\hat x)=i\hbar \hbox{Tr}(\mathbb{1})$, whereas in infinite dimensional spaces of course this identity must be handled carefully because of the unbounded nature of the operators. $\endgroup$ Jan 3, 2022 at 18:38

It is true that $\langle \alpha\vert L_z\vert\alpha\rangle=0$ if $\vert\alpha\rangle$ is an eigenstate of $L_x$ OR an eigenstate of $L_y$. Indeed, if it’s an eigenstate of $L_x$ then $\Delta L_x=0$ so the uncertainty relation yields $$ 0\times \Delta L_y=0\ge \frac{1}{2}\vert\langle \alpha\vert L_z\vert\alpha\rangle\vert \tag{1} $$ which implies the right hand side (which is necessarily real and non-negative) must be $0$. The same argument can be trivially modified to show the same result if $\vert\alpha\rangle$ is an eigenstate of $L_x$ instead.

I don’t know how this proves there are no simultaneous eigenstates of $L_x$ and $L_y$ and nothing in (1) prevents $\Delta L_y=0$.

The converse is not true: there are states with $\langle L_z\rangle=0$ which are not eigenstates of $L_x$ or $L_y$.

There is one state which is a simultaneous eigenstate of $L_x$ AND $L_y$: it is the $L=0$ state.

Now if you have a basis where $L_x$ is diagonal, then you can construct ladder operators $\tilde L_\pm$ in terms of $L_y$ and $L_z$, and in particular express $L_y$ as a combo of $\tilde L_\pm$, which do not act diagonally on the eigenstates of $L_y$. The exception is (again) the $L=0$ state, which is killed by $\tilde L_\pm$.


One tool for finding your way towards a proof is to construct some representation matrices for a few different $L_{x,y,z}$ and see how you feel about them. Here’s a representation for spin two, which gives you a sense of how it goes. The size of the matrix is $2\ell+1$. The $z$-axis matrix has the eigenvalues on the diagonal already,

$$ L_z = \left(\begin{array}{ccccc} \ell \\ &\ell-1 \\ && \ddots \\ &&& -\ell+1 \\ &&&& - \ell \end{array}\right), $$

and the raising and lowering matrices are one-off-the-diagonal,

$$ L_+ = \left(\begin{array}{ccccc} 0& \sqrt{2\ell} & \\ &0& \sqrt{\cdots} & \\ &&\ddots \\ && \cdots & 0& \sqrt{2\ell} \\ &&&&0 \end{array}\right) \qquad L_- = (L_+)^T \renewcommand{\ket}[1]{\left|{#1}\right>} $$

so that $L_\pm\ket{\ell,m} = \sqrt{\ell(\ell+1) - m(m\pm1)} \ket{\ell, m\pm1}$. From these raising and lowering matrices $L_\pm$ you construct

\begin{align} 2 L_x &= L_+ + L_- \\ 2i L_y &= L_+ - L_- \end{align}

and then just find the eigenvectors. For spin-half (which isn’t allowed for orbital angular momentum $L$, but the algebra is similar), the $x$-axis eigenvectors are ${1}\choose{\pm1}$, with eigenvalues $\pm1$, and the $y$-axis eigenvectors are $1\choose{\pm i}$, with eigenvalues $\pm i$. Those basis vectors are orthogonal to each other. However, it is the case that the eigenstates of $\sigma_x$ have expectation value $\left<\sigma_y\right> = \left<\sigma_z\right> = 0$.


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.