In many-body physics, Holstein-Primakoff transformation is defined as follows:

\begin{align} S_i^+ &= \sqrt{2S}(1-a_i^\dagger a_i/2S)^{1/2}a_i, \\ S_i^- &=\sqrt{2S}(1-a_i^\dagger a_i/2S)^{1/2}, \end{align}

where $S_i^{\pm}, S$ are spin operators and $a_i$ is the transformed operator.

I have two questions regarding this.

  1. Is the square root or division of an operator (not a number!) well-defined?
    (Furthermore, $S$ is a vector operator!)
    It seems that the square root is defined in terms of Taylor series, but how can we guarantee that the series converges?

  2. The transformed operator $a_i$ is defined implicitly by the above relation. Does there actually exist such $a_i$ that satisfies the above relation? If so, is it unique?

  • 1
    $\begingroup$ More on Holstein-Primakoff. $\endgroup$
    – Qmechanic
    Apr 20 '19 at 12:03
  • $\begingroup$ $S$ is a number, NOT an operator or a vector. $\endgroup$ Apr 20 '19 at 15:44

Since you reference Wikipedia, I'll use its definition instead of your garbled one, so consider $$ S_+ = \hbar \sqrt {2s- {a^\dagger a} }~~ a ~, \qquad S_- = \hbar a^\dagger\, \sqrt{2s-{a^\dagger a}} ~, \qquad S_z = \hbar(s - a^\dagger a) ~, $$ where capital S s are operators, $[S_+,S_-]=2\hbar S_z$; and $[a,a^\dagger ] =1$ ab initio.

Crucially, unlike what you seem to suspect, s is a parameter, such that, by straightforward plugin, (check!), $$ S^2= S_z^2 + \tfrac{1}{2}(S_+ S_- + S_- S_+)=\hbar^2 s(s+1) 1\!\!1. $$ The last equality holds, of course, formally, by dint of $a^\dagger a a = a a^\dagger a -a$.

Now to you question: The (number) operator $a^\dagger a$ is a diagonal matrix with eigenvalues 0,1,2,3,4,... so the diagonal matrix under the square root sign has eigenvalues 2s,2s-1,2s-3,... terminating with a 0 in its bottom rightmost component. That is, it has only 2s+1 components, and cannot allow $S_-$ to lower a state past $a^{\dagger ~~2s}|0\rangle$, nor $S_+$ to raise it past $|0\rangle$!

The states connected by these S operators then comprise a 2s+1 multiplet: a spin s representation of angular momentum. So, the diagonal matrix under the square root sign is not even infinite dimensional.

How do you take a square root of a finite diagonal matrix D? It is but the diagonal matrix of the square root of each eigenvalue of the original matrix D.

  • $\begingroup$ Actually, it was other person who linked the Wikipedia article. Anyway, thank you! $\endgroup$
    – eigenvalue
    Apr 26 '19 at 12:18

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.