0
$\begingroup$

I want to prove that the particle number operator $$N = \sum_i a^\dagger_i a_i $$ commutes with a general product of creation / annihilation operators, e.g., $$(a^\dagger_1)^2 (a^\dagger_2) (a_1)^7 (a^\dagger_3)^2 ...$$ as long as the number of creation operators is equal to the number of annihilation operators.

So I plan to compute the below commutator $$ [N, \prod_j\, (a^\dagger_j)^{n_j} (a_j)^{k_j}] $$

using the commutator basic identities and $$[a_i,a^\dagger_j]=\delta_{ij}$$

expecting to be zero only when the number of $a^\dagger$ is the same of $a$.

Is this logically correct and sufficient to prove my question or am I missing something?

$\endgroup$
2
  • 1
    $\begingroup$ Why are there factorials on $n$ and $k$? Going off the line above, I would expect you'd want an operator $\prod_{j} (a_{j}^{\dagger})^{n_{j}} (a_{j})^{k_j}$ in your commutator (for some collection of numbers $\{n_{j}, k_{j}\}_{j}$, where your line above would have $n_{1}=2$, $k_{1} = 7$, then $n_2=1$, $k_2 =0$ and so on) $\endgroup$ Oct 26, 2020 at 1:11
  • 1
    $\begingroup$ It does makes more sense, on the contrary to my assumption. I'm going to edit it $\endgroup$
    – EinRock
    Oct 26, 2020 at 1:39

2 Answers 2

3
$\begingroup$

The result you seek is $$ \left[ N , \prod_{j} (a_j^{\dagger})^{n_j} (a_{j})^{k_j} \right] \ = \ \left( \sum_{i} ( n_{i} - k_{i} ) \right) \prod_{j} (a^{\dagger})^{n_{j}} (a_{j})^{k_j} \ . $$ As you guessed, this commutes with a RHS of $0$ when $n_j = k_j$ for all $j$. However, interestingly there are other combinations of operators which make this commute (for example, if you have two oscillators, then picking $n_1=2$, $n_2 = 0$ and $k_1=1$, $k_2=1$ makes the above commute ie. $N = a_1^\dagger a_1 + a_2^\dagger a_2$ would commute in this example with the operator $(a_1^\dagger)^2 a_1 a_2$).

To see how this result follows, note that this commutator has the form $$ [ \sum_{j} A_j , \prod_{j} X_{j} ] \ = \ \sum_{j} [ A_{j}, X_j ] \prod_{\ell \neq j} X_\ell \ , $$ which follows when you are careful about the tensor product structure of the operators (see Appendix A below). This means that our commutator becomes $$ \left[ N , \prod_{j} (a_j^{\dagger})^{n_j} (a_{j})^{k_j} \right] \ = \ \sum_{j} \left[ a_j^\dagger a_j, (a_j^{\dagger})^{n_j} (a_{j})^{k_j} \right] \; \prod_{\ell \neq j} (a_\ell^{\dagger})^{n_\ell} (a_{\ell})^{k_\ell} $$ It follows that $\left[ a_j^\dagger a_j, (a_j^{\dagger})^{n_j} (a_{j})^{k_j} \right] = ( n_j - k_j ) (a_j^{\dagger})^{n_j} (a_{j})^{k_j}$ (see Appendix B below). With this we get the quoted result.

Appendix A: To see why this is true, imagine there are only two oscillators, then we have $$ [A_1 + A_2, X_1 X_2] \to [ A_{1} \otimes \mathbb{I} + \mathbb{I} \otimes A_2 , X_1 \otimes X_2 ] = [ A_{1}, X_1 ] \otimes X_2 + X_1 \otimes [ A_2 , X_2 ] , $$ and the generalization of this is quoted above.

Appendix B: We just need to evaluate a commutator of the form $[ a^\dagger a, (a^\dagger)^n a^k]$ here. One way to do this is consider the cases of $n>k$, $n=k$ and $n<k$ separately. Clearly for $n=k$ we have find that $[ a^\dagger a, (a^\dagger)^n a^k] = [ N, N^n ] = 0$. For $n>k$, we can set $n = k+m$ and the commutator has the form $$ [ a^\dagger a, (a^\dagger)^n a^k] = [ N, (a^\dagger)^m N^k] $$ which is easy to evaluate case-by-case for $m=1,2,3,\ldots$ once you know that $[N,a^{\dagger}] = a^{\dagger}$. It then follows that the above is equal to $$ \cdots = m (a^{\dagger})^m N^k = (n-k) (a^{\dagger})^n a^k \ . $$ The remaining case for $n<k$ follows similarly and gives the same result.

$\endgroup$
3
  • $\begingroup$ Remarkable. However I didn't quite understood the explanation related to appendix A and the appendix A itself. Care to explain or indicate me some resources so I can profound my knowledge on it? $\endgroup$
    – EinRock
    Oct 26, 2020 at 3:25
  • 1
    $\begingroup$ For two oscillators as in Appendix A: There's a separate Hilbert space for each oscillator. Note that you label states for two oscillators with two labels $| n_1, n_2 \rangle$, where the operators with "$1$" subscripts only talk to the $n_1$ in the above, and similarly for the "$2$" subscripts. For example, you have operations like $a_1^{\dagger} a_{1} | n_1, n_2 \rangle = n_1 | n_1, n_2 \rangle$ or also $a_2^{\dagger} a_{2} | n_1, n_2 \rangle = n_2 | n_1, n_2 \rangle$. This should convince you that the overall Hilbert space is a tensor product of each of the two oscillators Hilbert spaces. $\endgroup$ Oct 26, 2020 at 20:37
  • 1
    $\begingroup$ It is annoying to include the $\otimes$ symbols, so for brevity people write $| n_1 \rangle \otimes | n_2 \rangle \to | n_1,n_2 \rangle$, or $N = a_1^{\dagger} a_1 \otimes \mathbb{I} + \mathbb{I} \otimes a_2^{\dagger} a_2 \to a_1^\dagger a_1 + a_2^\dagger a_2$. So you've secretly been using a tensor product structure (there's lots of good literature out there about tensor products). Appendix A is just being explicit calculating $[ a_1^{\dagger} a_1 \otimes \mathbb{I} + \mathbb{I} \otimes a_2^{\dagger} a_2, (a_1^{\dagger})^{n_1} a_1^{k_1} \otimes (a_1^{\dagger})^{n_2} a_2^{k_2} ]$ $\endgroup$ Oct 26, 2020 at 20:44
2
$\begingroup$

Actually it does not commute. What you have is $[\hat N,a_i^\dagger]=a_i^\dagger$ so that \begin{align} [\hat N,(a_i^\dagger)^2]=[\hat N,a_i^\dagger]a_i^\dagger + a_i^\dagger [\hat N,a_i^\dagger]= 2a_i^\dagger \end{align} by the derivative rule. The induction hypothesis is that $$ [\hat N,(a_i^\dagger)^p]=p (a_i^\dagger)^p $$ so that \begin{align} [\hat N,(a_i^\dagger)^p]&= [\hat N,a_i^\dagger (a_i^\dagger)^{p-1}]\\ & =[\hat N,a_i^\dagger](a_i^\dagger)^{p-1}+ a_i^\dagger[\hat N,(a_i^{\dagger})^{p-1}]\\ &= a_i^\dagger(a_i^\dagger)^{p-1} + a_i^\dagger (p-1)(a_i^\dagger)^{p-1}\\ &= p (a_i^{\dagger})^p \end{align} As similar argument can be made for $[\hat N, (a_i)^p]$.

$\endgroup$
1
  • $\begingroup$ You are absolutely right! However I am studying the case of the N operator commuting with a general product of $a^\dagger$s and $a$s as in my example above. Said that, actually this question helped me getting the result $\endgroup$
    – EinRock
    Oct 26, 2020 at 3:20

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.