The alternating Hamiltonian may be written as:

$$H = t \sum_{n} (-1)^{n} \left[c^{\dagger}_{n+1}c_{n} + c^{\dagger}_{n}c_{n+1} \right] \; \; .$$

I wanted to know the energy dispersion for this system, so I wrote in mommentum space; After some calculations, I got:

$$H = t\sum_{k} 2\cos\left( k \right)\, c^{\dagger}_{k} c_{k+\pi} \; \; .$$

However, I can't diagonalize this. What should I do? 

  • 2
    $\begingroup$ For each $k$, you now need to diagonalize a $2\times 2$ matrix for $k$ and $k+\pi$. $\endgroup$
    – Meng Cheng
    Jun 14 at 18:48
  • $\begingroup$ Would you mind writing down the matrix I need to diagonalize? I'm having trouble getting to it. $\endgroup$
    – RFeynman
    Jun 14 at 19:59

1 Answer 1


Due to the factor of $(-1)^n$, we can see that this lattice is actually periodic with a period equal to two sites. That is, we've got a lattice with a two-site basis. As such, when we define our momentum-space operators, we have to take this periodicity into account. To do this, we define two sets of operators via \begin{align} \hat{a}_k&=\frac{1}{\sqrt{N}}\sum_{j=0}^{N-1}e^{-ika(2j)/2}\hat{c}_{2j}\,,\\ \hat{b}_k&=\frac{1}{\sqrt{N}}\sum_{j=0}^{N-1}e^{-ika(2j+1)/2}\hat{c}_{2j+1}\,, \end{align} where $a$ is the lattice spacing (i.e., the distance between every two sites), $N$ is the number of primitive unit cells (and so half the total number of sites), and we apply periodic boundary conditions, i.e., $\hat{c}_{N}=\hat{c}_0$, etc. The allowed values of $k$ are determined by assuming periodic boundary conditions, yielding $$ k\to k_m=m\frac{2\pi}{Na}\,,~~~~~~~m=0,1,\dots,N-1\,. $$ Accordingly, we have the orthogonality condition, $$ \frac{1}{N}\sum_ne^{i(k-k')an} = \delta_{kk'}\,, $$ which allows us to invert both of the previous definitions as \begin{align} \hat{c}_{2j}&=\frac{1}{\sqrt{N}}\sum_ke^{ika(2j)/2}\hat{a}_{k}\,,\\ \hat{c}_{2j+1}&=\frac{1}{\sqrt{N}}\sum_ke^{ika(2j+1)/2}\hat{b}_{k}\,. \end{align} Using these definitions, let's rewrite terms in the Hamiltonian as follows. First $$ \sum_n(-1)^n\hat{c}_{n+1}^{\dagger}\hat{c}_n = \sum_n\hat{c}_{2n+1}^{\dagger}\hat{c}_{2n} - \sum_n\hat{c}_{2n+2}^{\dagger}\hat{c}_{2n+1}\,. $$ Plugging in for the operators yields \begin{align} \sum_n\hat{c}_{2n+1}^{\dagger}\hat{c}_{2n} &= \sum_n\frac{1}{\sqrt{N}}\sum_ke^{-ika(2n+1)/2}\hat{b}_{k}^{\dagger} \frac{1}{\sqrt{N}}\sum_{k'}e^{ik'a(2n)/2}\hat{a}_{k}\\ &= \sum_{kk'} \hat{b}_{k}^{\dagger}\hat{a}_{k'} e^{-iak/2}\frac{1}{{N}}\sum_ne^{ia(k'-k)n}\\ &= \sum_{k} e^{-iak/2}\hat{b}_{k}^{\dagger}\hat{a}_{k}\,, \end{align} and \begin{align} \sum_n\hat{c}_{2n+2}^{\dagger}\hat{c}_{2n+1} &= \sum_n\frac{1}{\sqrt{N}}\sum_ke^{-ika(2n+2)/2}\hat{a}_{k}^{\dagger} \frac{1}{\sqrt{N}}\sum_{k'}e^{ik'a(2n+1)/2}\hat{b}_{k}\\ &= \sum_{kk'} \hat{a}_{k}^{\dagger}\hat{b}_{k'} e^{-ika}e^{ik'a/2}\frac{1}{{N}}\sum_ne^{ia(k'-k)n}\\ &= \sum_{k} e^{-iak/2}\hat{a}_{k}^{\dagger}\hat{b}_{k}\,, \end{align} where we have used the orthogonality relation to compute the sums over $n$ and then collapsed the double sum over $k$ and $k'$ using the resulting $\delta_{kk'}$.

The other two terms in the Hamiltonian are just the Hermitian conjugates of these ones, and so all told we have \begin{align} \hat{H} &= t\sum_{k} e^{-iak/2}\hat{b}_{k}^{\dagger}\hat{a}_{k} + t\sum_{k} e^{iak/2}\hat{a}_{k}^{\dagger}\hat{b}_{k} - t\sum_{k} e^{-iak/2}\hat{a}_{k}^{\dagger}\hat{b}_{k} - t\sum_{k} e^{iak/2}\hat{b}_{k}^{\dagger}\hat{a}_{k}\\ &= 2it\sum_{k}\sin\left(\frac{ka}{2}\right)\left( \hat{a}_{k}^{\dagger}\hat{b}_{k} -\hat{b}_{k}^{\dagger}\hat{a}_{k} \right)\,. \end{align} We can see that we have succeeded in diagonalizing the Hamiltonian in the quasi-momentum $k$. Now we just need to diagonalize the remaining piece.

To do this, wee define new operators as $$ c_{k,\pm} = \frac{\pm i\hat{a}+\hat{b}}{\sqrt{2}}\,, $$ in which case $$ \hat{a}_k = \frac{i(\hat{c}_{k,-} - \hat{c}_{k,+})}{\sqrt{2}}\,,~~~~~~~~\hat{b}_k = \frac{\hat{c}_{k,-} + \hat{c}_{k,+}}{\sqrt{2}}\,. $$ Then, plugging this into the terms above, we get $$ \hat{a}_k^{\dagger}\hat{b}_k = \frac{-i(\hat{c}_{k,-}^{\dagger} - \hat{c}_{k,+}^{\dagger})}{\sqrt{2}}\frac{\hat{c}_{k,-} + \hat{c}_{k,+}}{\sqrt{2}} =\frac{-i}{2}( \hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} +\hat{c}_{k,-}^{\dagger}\hat{c}_{k,+} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,-} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+} )\,, $$ which implies that $$ \hat{b}_k^{\dagger}\hat{a}_k=(\hat{a}_k^{\dagger}\hat{b}_k)^{\dagger} =\frac{i}{2}( \hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} +\hat{c}_{k,+}^{\dagger}\hat{c}_{k,-} -\hat{c}_{k,-}^{\dagger}\hat{c}_{k,+} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+} )\,. $$ Then, \begin{align} \hat{a}_k^{\dagger}\hat{b}_k - \hat{b}_k^{\dagger}\hat{a}_k &= \frac{-i}{2}( \hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} +\hat{c}_{k,-}^{\dagger}\hat{c}_{k,+} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,-} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+} ) - \frac{i}{2}( \hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} +\hat{c}_{k,+}^{\dagger}\hat{c}_{k,-} -\hat{c}_{k,-}^{\dagger}\hat{c}_{k,+} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+} ) \\ &= -\frac{i}{2}( 2\hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} -2\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+}) \end{align} Finally, plugging this back into the Hamiltonian, we get \begin{align} \hat{H} &= 2it\sum_{k}\sin\left(\frac{ka}{2}\right)\left( -\frac{i}{2}( 2\hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} -2\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+}) \right) = \sum_{k}2t\sin\left(\frac{ka}{2}\right)\left( \hat{c}_{k,-}^{\dagger}\hat{c}_{k,-} -\hat{c}_{k,+}^{\dagger}\hat{c}_{k,+} \right) \end{align}

  • $\begingroup$ Thank you so much for your answer! I have one more question: the final Hamiltonian you write produces an energy dispersion $\left[ E = \pm \sum_{k} 2tsin \left(\frac{ka}{2} \right) \right]$ with a gap, correct? $\endgroup$
    – RFeynman
    Jun 15 at 9:44
  • 1
    $\begingroup$ @RFeynman Yes. Note that it's worth verifying that the final operators still satisfy the correct commutation relations (whether bosonic or fermionic). $\endgroup$
    – march
    Jun 15 at 15:09

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.