5
$\begingroup$

Let $a_1, \ldots, a_N$ be boson operators, $[a_i, a_j^\dagger ] = \delta_{ij}$. One often considers Bogoliubov transformation $a_i \to \sum_j (A_{ij} a_j + B_{ij} a_j^\dagger)$, where the matrices $A=(A_{ij}), B=(B_{ij})$ satisfy $$AA^\dagger - BB^\dagger = I, \quad AB^T - BA^T=0,$$ to preserve the commutation relation.

Question: Can we find a unitary operator $U$ (acting on the usual bosonic Hilbert space) that satisfies $U a_i U^\dagger = \sum_j(A_{ij} a_j + B_{ij} a_j^\dagger)$?

I think finding $U$ is an entirely different issue than checking the invariance of the commutation relation.

$\endgroup$
1

2 Answers 2

5
$\begingroup$

There is a general argument for the existence of the unitary operator. From the Stone-von Neumann theorem, if you know where $a$ is mapped, then there is only one unitary operator that can represent the transformation (up to a global phase).

You can construct this operator more explicitly. One way is to realize that your transformations form a path connected Lie group. Formally, you can view it as the subgroup of matrices of size $2N\times2N$ of the form: $$ \begin{pmatrix} A & B\\ B^* & A^* \end{pmatrix} $$ with your additional constraints: $$ A A^\dagger -BB^\dagger =1\\ AB^T-BA^T =0 $$ You can consider a path from the identity to your transformation, namely $A_t,B_t$ with $t\in[0,1]$ and: $$ \begin{align} A_0&=1 & B_0&=0 \\ A_1&=A& B_1&=B \end{align} $$ This gives a varying $a_t$ satisfying the ODE: $$ \dot a_t= \alpha_ta_t+ \beta_ta_t^\dagger $$ with the matrices $\alpha_t,\beta_t$ defined by: $$ \frac{d}{dt} \begin{pmatrix} A_t & B _t\\ B _t ^* & A _t ^* \end{pmatrix} = \begin{pmatrix} \alpha_t& \beta_t\\ \beta_t^* & \alpha_t^* \end{pmatrix} \begin{pmatrix} A_t & B _t\\ B _t ^* & A _t ^* \end{pmatrix} $$ You want to recognize the ODE as a Heisenberg equation from a suitably chosen quadratic Hamiltonian: $$ \dot a_t =-i[a,H] $$ Using the CCR’s you can check that a suitable candidate (unique up to additional multiple of identity) is: $$ H_t= i(\alpha_t)_{ij}a_i^\dagger a_j+\frac{i}{2}(\beta_t)_{ij}a_i^\dagger a_j^\dagger-\frac{i}{2}(\beta_t^*)_{ij}a_i a_j $$ $H_t$ is hermitian and gives the correct equations of motion using the relations (taking the derivatives of the constraints): $$ \alpha_t+\alpha_t^\dagger =0 \\ \beta_t-\beta_t^T=0 $$ Your unitary operator is given by the time ordered exponential: $$ U=\mathcal T \exp\left(\int_0^1dt H_t\right) $$ Actually, it turns out that the exponential is surjective on the group, so you can choose Choose a path with constant $\alpha_t,\beta_t$. This gives a constant $H_t$ and the time ordered exponential is a simple exponential.

Physically, you are constructing the unitary operator from harmonic oscillators and squeeze operators. The process is explicit in the case $N=1$.

Hope this helps.

$\endgroup$
2
  • $\begingroup$ Thanks for your answer! I found the abstract argument based on Stone-von Neumann theorem interesting. Could you elaborate more why knowing where $a_i$'s are mapped implies that there exists a unique unitary operator that realizes that transformation? Maybe one transforms $x_i \sim a_i + a_i^\dagger$ and $p_i \sim a_i - a_i^\dagger$ to obtain the "canonical commutation relation"? $\endgroup$
    – Laplacian
    Commented Mar 11, 2023 at 11:18
  • 2
    $\begingroup$ Yes exactly, and since your Hilbert space is constructed as an irreducible representation, this guarantees existence and uniqueness. Note that the exact mathematical statement is not about the CCR $[x,p] = i$, but rather its exponentiated form (see Weyl form of CCR for more). $\endgroup$
    – LPZ
    Commented Mar 11, 2023 at 13:01
2
$\begingroup$

The general case (more than one mode) is complicated. A formula is given in eqs 2.17, 2.18 in https://arxiv.org/pdf/1608.03289.pdf, unfortunately the author's notation is rather obscure and ius defined only in their cited book.

There is a discussion in my online notes http://people.physics.illinois.edu/stone/bose_bogoliubov.pdf. It is too long to copy here.

$\endgroup$
1
  • $\begingroup$ Thanks for your answer! I will take time to read your notes. $\endgroup$
    – Laplacian
    Commented Mar 11, 2023 at 11:14

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.