I have a fermionic system with states 1,2. They are coupled by a harmonic oscillator. The Hamiltonian of the system should then be

$$ H=\left[\gamma(a^\dagger+a)-\delta\right]\left( c^\dagger_1 c_2+c^\dagger_2 c_1 \right) + \xi \left ( c^\dagger_1 c_1-c^\dagger_2 c_2 \right) + \hbar\omega\left(a^\dagger a + \frac{1}{2} \right) $$

where $\gamma$ is the coupling constant, $\delta$ some initial overlap integral and $\xi$ the energy difference between the uncoupled states.

In order to get the eigenvalues of this system I would like to diagonalize $H$. However I fail to see how to approach this because of the terms with three operators in them such as $a^\dagger c^\dagger_1 c_2$, which make me wonder how to even display $H$ as a matrix.

  • $\begingroup$ I assume that should that be $c_1^\dagger c_2 + c_2^\dagger c_1$, right? $\endgroup$ – J. Murray May 2 '20 at 21:09
  • $\begingroup$ @J.Murray right, thanks. Changed it $\endgroup$ – Liberty May 2 '20 at 21:17

Actually, after reading my reply I understood that all of this is by no means necessary. The solution is extremely simple. \begin{align} \left( \hat{H}_0 + \hat{H}_{int} \right) \vert \Psi_k \rangle = E_k \vert \Psi_k \rangle \end{align} The most general state $\vert \Psi_k \rangle$ is in the tensor space therefore it is of the form \begin{align} \vert \Psi_k \rangle = \sum_{n_1,f_1,f_2} c_{1,2,3} \;\vert n_1 \rangle \times \vert f_1\rangle \times \vert f_2 \rangle \end{align} The fermion product $\vert f_1 \rangle \times \vert f_2 \rangle $ is one of the 4 possible candidates 1) $\vert0,0\rangle$ 2) $\vert0,1\rangle$ 3) $\vert1,0\rangle$ 4) $\vert1,1\rangle$

The free hamiltonian has eigenvectors $\vert n_b, f_1, f_2 \rangle$ with eigenvalues $E(n_b,f_1,f_2) = \hbar\left(n_b + \frac{1}{2}\right) + \xi(f_1-f_2)$. Here the term proportional to $\xi$ contributes only if $\vert f_1 \neq f_2 \vert$. However, the states 1-4 are not eigenstates of $H_{int}$. A linear combination of these eigenstates is still an eigenstate of $H_0$. Therefore I seek for eigenstates of $H_{int}$ in the form of such linear combination.

Let us analyse the interaction term $\hat{H}_{int}$. It is the sum of two term, the first in which there are three operators ($a c^\dagger c$ etc.) and the last with two operator. \begin{align} \hat{H}_{int} &= \hat{H}_{int,1} + \hat{H}_{int,2}\\ &=\gamma \Big( \hat{a}^\dagger + \hat{a} \Big)\left(\hat{c}_1^\dagger \hat{c}_2 + \hat{c}_2^\dagger \hat{c}_1\right) - \delta \left(\hat{c}_1^\dagger \hat{c}_2 + \hat{c}_2^\dagger \hat{c}_1\right) \end{align} Let us start from the latter - which is easier. The eigenvectors of $H_{int2}$ are the rays $c_1\Big( \vert \alpha ,1,0\rangle + \vert \alpha ,0,1 \rangle\Big)$ for any $\vert\alpha \rangle \equiv \sum_{n_b=0}^{\infty} c_{n_b} \vert n_b\rangle$. Also, we have the linear combinations $c_1 \vert \alpha ,0,0\rangle + c_2 \vert \alpha ,1,1\rangle $.

Finally, let us turn to $H_{int1}$. The fermionic part is analogue, however this time we have the bosonic creation/annihilation operators. Therefore any linear combination $c_1 \vert \alpha ,0,0\rangle + c_2 \vert \alpha ,1,1\rangle $ is eigenstate (with zero eigenvalue). The other case is where I use the state $c_1\Big( \vert \alpha ,1,0\rangle + \vert \alpha ,0,1 \rangle\Big)$. Here $\vert \alpha\rangle $ is a generic linear combination of number-eigenstates of the bosonic operator $n_b$. The result is \begin{align} \hat{H}_{int1} \vert \alpha\rangle \times \vert(...)\rangle &=\gamma \Big(\hat{a}^\dagger + \hat{a} \Big) \sum_{n_b=0}^{\infty} c_{n_b} \vert n_b\rangle \times \vert(...)\rangle \\ &=\gamma \Big(\sum_{n_b=1}^{\infty} \sqrt{n_b}\,c_{n_b-1} \vert n_b\rangle + \sum_{n_b=0}^{\infty} \sqrt{n_b+1}\,c_{n_b+1} \vert n_b\rangle \Big)\times \vert(...) \\ &=\gamma \Big(\sum_{n_b=1}^{\infty} (\sqrt{n_b}\,c_{n_b-1}+ \sqrt{n_b+1}\,c_{n_b+1}) \vert n_b\rangle\Big)\times \vert(...) + c_1 \vert0\rangle\times \vert(...) \end{align}

Comparing the lhs with rhs I've $c_1=c_0$ and $c_{n_b} = (\sqrt{n_b}\,c_{n_b-1}+ \sqrt{n_b+1}\,c_{n_b+1})$ therefore $c_{n_b + 1} = \frac{c_{n_b} - \sqrt{n_b}\,c_{n_b-1}}{\sqrt{n_b+1}}$

Modulo few errors I've made in ladder operators factors etc., you can find the state $\vert \alpha \rangle$ and the eigenvectors of the full $\hat{H}$ taking the intersection of the three sets.


The hamiltonian operator we have is that of three harmonic oscillator (two fermionic and one bosonic) in the presence of a interaction. If we assume of treating this term as a perturbation over the free theory. We can find the spectrum of the full hamiltonian by using time-independent perturbation theory. If ${\rm O}\left(H_{int}\right) $ where $\lambda$ is the perturbative parameter, the eigenvalue problem reads \begin{gather} \left( \hat{H}_0 + \hat{H}_{int} \right)\left\vert n\right\rangle = E_n \left\vert n \right\rangle \\ {\rm where}\\ H_0 = \xi\left( \hat{n}_{f1} - \hat{n}_{f2} \right) + \hbar\omega\left( \hat{n}_b + \frac{1}{2}\right) \\ H_{int} = \left[ \gamma\left(\hat{a}^\dagger + \hat{a} \right) - \delta \right] \Big( \hat{c}_1^\dagger \hat{c}_2 + \hat{c}_2^\dagger \hat{c}_1 \Big) \end{gather} here $n_i$ is a shorthand for the oscillator's occupation number, while $n$ is the label of the eigenstate.

If we schematically indicate the eigenstates of the free theory with $\vert n^{(0)} \rangle=\left\vert n_b,n_{f1},n_{f2} \right\rangle$, that is \begin{align} \hat{H}_0 \vert n^{(0)} \rangle = E^{(0)}_n \vert n^{(0)} \rangle \end{align} the full theory eigenstates and eigenvalues, in time-independent perturbation theory, are given in terms of the exact free-theory solutions as eq.(1.44) of Sakurai,Napolitano - Modern quantum mechanics \begin{align} \vert n \rangle &= \vert n^{(0)} \rangle + \lambda \sum_{k\neq n} \vert k^{(0)} \rangle \frac{\langle k^{(0)} \vert H^{(int)} \vert n^{(0)} \rangle}{E^{(0)}_n - E^{(0)}_k} + \lambda^2... \end{align}

Before you apply to the problem you should notice that the fermionic creation operator gives zero if the corresponding state is occupied. There are few possible states

  1. $ \vert n_b,0,0 \rangle$
  2. $ \vert n_b,0,1 \rangle$
  3. $ \vert n_b,1,0 \rangle$
  4. $ \vert n_b,1,1 \rangle$

And $H_{int}$ conserves the sum of occupation numbers of fermionic oscillators, that is $n_{f1} + n_{f2}$ of the ket must be the same of the bra here $\langle k^{(0)} \vert H^{(int)} \vert n^{(0)} \rangle = 0$ otherwise it gives zero. The energy spectrum you're looking for is eq. (1.53b) of Sakurai, Napolitano.

Therefore, I think that the problem may further reduced to finding only the state $\vert\alpha\rangle$ since the hamiltonian eigenstates can be written as the tensor product $ \vert \alpha \rangle \times \Big( \vert 0,1 \rangle + \vert 1,0 \rangle \Big)$

  • $\begingroup$ So some of the matrix elements $H_{ij}$ can also contain ladder operators? How does that make sense? Also what about the term $\frac{1}{2}\hbar\omega$ that isn't linked with any operators? $\endgroup$ – Liberty May 2 '20 at 21:54

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.