I came across the following exercise.

The quantum Heisenberg ferromagnet, which is specified by the Hamiltonian:

$$ \hat{H} = - J\sum_{\langle m n \rangle} \hat{\vec{S}}_m \cdot \hat{\vec{S}}_n, $$

where $J>0, \hat{\vec{S}}_m$ represents the quantum mechanical spin operator at lattice site m, $\langle mn \rangle$ denotes the summation over neighboring sites and $\vec{S}_m ^2 = S(S+1)$. Now we introduce the following transformation:

$$ \hat{S}_m^- = a^{\dagger}_m (2S - a^{\dagger}_m a_m )^{1/2}, \ \ \ \hat{S}_m^+ = (2S - a^{\dagger}_m a_m )^{1/2} a_m, \ \ \ \hat{S}_m^z = S- a_m^{\dagger}a_m. $$

Now we look at the one dimensional problem and put the lattice constant to unity. At low temperatures, for $S<<1/2$ we expect the deviations of the magnetization from its value to be very small, i.e. $S - \langle S_m^z \rangle = \langle a_m^{\dagger}a_m \rangle <<S$. In this case we may expand $(2S - a^{\dagger}_m a_m )^{1/2}$ in powers of $ a^{\dagger}_m a_m $.

Show that to first order in $a_m^{\dagger}a_m /S$ the Heisenberg Hamiltonian takes the form

$$ \hat{H} = - J N S^2 + J S \sum_m \left( a_{m+1}^{\dagger} - a_{m}^{\dagger} \right) \left( a_{m+1} - a_{m} \right) + \text{higher order terms.} $$

Now also show that if we Fourier transform the operators, the Hamiltonian takes the form

$$ \hat{H}= - J N S^2 + \sum_k \hbar \omega_k \alpha_k^{\dagger} \alpha_k + \text{higher order terms} $$

with $\hbar \omega_k = 4 J S \sin^2(k/2)$.

Now I don't understand for the first Hamiltonian what I need to do. My first guess is to do a mean field approximation, since we have the expectation value. But this approach doesn't give me the desired cross terms. Furthermore I have a hard time on what to do with the dot product between $S_m$ and $S_n$. Would we need to write out the inner product, invert our operators to get $\hat{S}_x$ and $\hat{S}_y$? Since this seems really over complicated.

Then for the Fourier transform I get $2 \sin^2(k/a)$ and I wonder if the exercise is incorrect or if I made a small error.


2 Answers 2


$ \newcommand{\Sz}[1]{\hat {S^{z}_\text{$#1$}} }$ $ \newcommand{\Sx}[1]{\hat {S^{x}_\text{$#1$}} }$ $ \newcommand{\Sy}[1]{\hat {S^{y}_\text{$#1$}} }$ $ \newcommand{\Splus}[1]{\hat {S}^{+}_\text{$#1$} }$ $ \newcommand{\Smin}[1]{\hat {S}^{-}_\text{$#1$} }$ $ \newcommand{\ad}[1]{ a^{\dagger}_\text{$#1$} }$ $ \newcommand{\a}[1]{ a_\text{$#1$} }$

Allright I have finished your problem.

Firstly write the quantum mechanical spin operator in terms of the ladder operators. $\hat{\textbf{S}}_m = ( \Sx{m},\Sy{m},\Sz{m} )$. Now we use the known relations of the ladder operators and these spatial operators to rewrite in terms of ladder operators. $\Sx{m} = \frac{1}{2}(\Splus{m} + \Smin{m}), \Sy{m} = \frac{1}{2i}(\Splus{m} - \Smin{m})$ and it follows from the commutation relations that $\Sz{m} = S - a^{\dagger}_m a_m$ as is given in your exercise.

Now let's calculate this inner product that you have in the Hamiltonian for some $l,m$:

$$ \begin{align} \hat{\textbf{S}}_l \cdot \hat{\textbf{S}}_m &= \Sx{l} \Sx{m} + \Sy{l} \Sy{m} + \Sz{l} \Sz{m} \\ &= \frac{1}{2}(\Splus{l} + \Smin{l})\frac{1}{2}(\Splus{m} + \Smin{m}) + \frac{1}{2i}(\Splus{l} - \Smin{l}) \frac{1}{2i}(\Splus{m} - \Smin{m}) + S^2 - Sa^{\dagger}_l a_l - Sa^{\dagger}_m a_m + a^{\dagger}_l a_la^{\dagger}_m a_m \\ &= \frac{1}{4} ( \Splus{l} \Splus{m} + \Smin{l} \Splus{m} + \Splus{l} \Smin{m} + \Smin{l}\Smin{m}) - \frac{1}{4} ( \Splus{l} \Splus{m} - \Smin{l} \Splus{m} - \Splus{l} \Smin{m} + \Smin{l}\Smin{m}) + S^2 - Sa^{\dagger}_l a_l - Sa^{\dagger}_m a_m + a^{\dagger}_l a_la^{\dagger}_m a_m \\ &= \frac{1}{2}(\Smin{l} \Splus{m} + \Smin{l}\Smin{m}) + S^2 - Sa^{\dagger}_l a_l - Sa^{\dagger}_m a_m \end{align} $$

Where we have omitted the higher order term $a^{\dagger}_l a_la^{\dagger}_m a_m$. We now rewrite the ladder operators in terms of creation operators as is suggested in your question, but first let's rewrite these expressions a bit using $S - \langle S_m^z \rangle = \langle a_m^{\dagger}a_m \rangle <<S$ so $(1 - \frac{a_m^{\dagger}a_m}{2S})^{1/2} = 1$.

$$ \begin{align} \hat{S}_m^- &= a^{\dagger}_m (2S - a^{\dagger}_m a_m )^{1/2} = a^{\dagger}_m \sqrt{2S} (1 - \frac{a^{\dagger}_m a_m}{2S} )^{1/2} = \sqrt{2S} \ a^{\dagger}_m \\ \hat{S}_m^+ &= (2S - a^{\dagger}_m a_m )^{1/2} a_m = \sqrt{2S} \ a_m\\ \end{align}$$

Your question simplifies the solution by only checking the nearest neighbour. That reduces the Hamiltonian to just:

$$\hat{H} = - J\sum_{m=1}^N \hat{\textbf{S}}_m \cdot \hat{\textbf{S}}_{m+1}$$

Plugging all our derivations and rewriting a bit gives:

$$ \begin{align} \hat{H} &= - J\sum_{m=1}^N \frac{1}{2}(\Smin{m} \Splus{m+1} + \Smin{m+1}\Smin{m}) + S^2 - Sa^{\dagger}_m a_m - Sa^{\dagger}_{m+1} a_{m+1} \\ &= - J\sum_{m=1}^N S(\ad{m}\a{m+1} + \ad{m+1}\ad{m} ) + S^2 - S\ad{m}\a{m} - S\ad{m+1}\a{m+1} \\ &= - J\sum_{m=1}^N S^2 + S(\ad{m}(\a{m+1} - \a{m}) + \ad{m+1}(\a{m} - \a{m+1})) \\ &= -JNS^2 - JS \sum_{m=1}^N(\ad{m} - \ad{m+1})(\a{m+1} - \a{m}) \\ &= -JNS^2 + JS \sum_{m=1}^N(\ad{m+1} - \ad{m})(\a{m+1} - \a{m}) \end{align} $$


Since this is the one - dimensional problem, one can rewrite the Hamiltonian as follows: $$H=-J\sum_{m=1}^{N}\vec{S}_m\vec{S}_{m+1}$$ with $S_{N+1} = S_1$ (periodic boundary condition,for example)

$S^+$ and $S^-$ are ladder operators of the spin angular momentum $\vec{S}$ and have the following form: $$S^{\pm} = S^x\pm iS^y$$ So: $$S^x = \frac{S^+ + S^-}{2}, S^y = \frac{S^+ - S^-}{2i}$$ Now, we calculate $\vec{S}_m\vec{S}_{m+1}$.

$$\vec{S}_m\vec{S}_{m+1} = S^x_mS^x_{m+1} + S^y_mS^y_{m+1} + S^z_mS^z_{m+1} = \frac{1}{2}\left(S^+_mS^-_{m+1}+S^-_mS^+_{m+1}\right)+S^z_mS^z_{m+1}$$ We also have (to the first order): $$(2S - a^\dagger a)^{1/2}\approx (2S)^{1/2}\left(1 - \frac{a^\dagger a}{4S}\right)$$ Subsituting this into the expressions of $S^+$ and $S^-$, then expanding the previous equation (neglecting terms with second order and higher), you will get the desired answer (note that $[a_m,a^\dagger_n]=\delta_{mn}$)


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.