I'm trying to solve, analitically, the transition probability of a spin-1 particle in a magnetic field $$ \vec{B}=B_0\hat{k}+b(\cos{\omega t}\hat{i}+sin{\omega t}\hat{j}). $$

In particular I want to find $$ \mathcal{P}(S_z=\hbar\rightarrow S_z=0)\;\;\;\;\text{and}\;\;\;\;\mathcal{P}(S_z=\hbar\rightarrow S_z=-\hbar). $$

I know that the hamiltonian is of the form $$ H(t)=-\hbar\omega_0\left( \begin{array}{ccc} 1 & 0 & 0 \\ 0 & 0 & 0 \\ 0 & 0 & -1 \end{array} \right) -\frac{\hbar\omega_\perp}{\sqrt2}\left( \begin{array}{ccc} 0 & e^{-i\omega t} & 0 \\ e^{i\omega t} & 0 & e^{-i\omega t} \\ 0 & e^{i\omega t} & 0 \end{array} \right), $$

since the definition of $S_x$, $S_y$ and $S_z$.

I followed the spin-1/2 case. I find the system of differential equations:

$$ i\hbar\dot a(t)=-\hbar\omega_0 a(t)-\frac{\hbar\omega_\perp}{\sqrt{2}}e^{-i\omega t}b(t) $$ $$ i\hbar\dot b(t)=-\frac{\hbar\omega_\perp}{\sqrt 2}[e^{i\omega t}a(t)+e^{-i\omega t}c(t)] $$ $$ i\hbar\dot c(t)=\hbar\omega_0 c(t)-\frac{\hbar\omega_\perp}{\sqrt{2}}e^{i\omega t}b(t) $$

Since it's easier with time independent coefficients I can define $$ A(t)=a(t)e^{i\beta t},\;\;\;\;B(t)=b(t)\;\;\;\;\text{and}\;\;\;\;C(t)=c(t)e^{-i\beta t} $$ and (setting $\beta=\omega$) rewrite the previous system: $$ i\dot A(t)=-\delta A(t)-\frac{\omega_\perp}{\sqrt2}B(t) $$ $$ i\dot B(t)=-\frac{\omega_\perp}{\sqrt2}(A(t)+C(t)) $$ $$ i\dot C(t)=\delta C(t)-\frac{\omega_\perp}{\sqrt2}B(t), $$

where $\delta=\omega-\omega_0$; or in matrix form: $$ i\left( \begin{array}{c} \dot A(t)\\ \dot B(t)\\ \dot C(t) \end{array} \right) =- \left( \begin{array}{ccc} \delta & \frac{\omega_\perp}{\sqrt2} & 0 \\ \frac{\omega_\perp}{\sqrt2} & 0 & \frac{\omega_\perp}{\sqrt2} \\ 0 & \frac{\omega_\perp}{\sqrt2} & -\delta \end{array} \right) \left( \begin{array}{c} A(t)\\ B(t)\\ C(t)\end{array} \right). $$

Now, the solution of an equation like this, where the coefficient matrix is $M$, sohould be $$ \left( \begin{array}{c} A(t)\\ B(t)\\ C(t)\end{array} \right)=e^{iMt} \left( \begin{array}{c} A(0)\\ B(0)\\ C(0)\end{array} \right). $$

I should find now an explicit solution in order to evaluate the probability (I need $|B(t)|^2$ and $|C(t)|^2$). The problem is that in the spin-1/2 case one can use the trick of writing the exponential of the matrix as a sum of a sine and a cosine (since all odd powers $\propto$ $M$ and all the even $\propto$ $\mathbb{I}$), but now it's not possible.

How should I proceed?

I can diagonalize $M$ and use the property $e^{-iMt}=Pe^{-iDt}P^{-1}$, then with the exponential of the diagonal matrix there shouldn't be any problem, but it's long and tedious. Isn't there any other method that I can use? Or did I make any mistake before?

  • $\begingroup$ This problem should be solvable by transforming to a rotating frame. See pp. 391--394 of Shankar's book on quantum mechanics to understand how this method works for spin 1/2. Then, I guess you could post your own answer. $\endgroup$
    – higgsss
    Commented Oct 1, 2015 at 0:22
  • $\begingroup$ I'm sorry but I don't understand how a rotation can solve my problem. For what I see Shankar isn't interested in finding a solution for the state in the perturbed field case. $\endgroup$
    – Charlie
    Commented Oct 2, 2015 at 10:13

2 Answers 2


Finally I solved it using the Cayley–Hamilton theorem.

Since the characteristic polynomial of $M$ is $$ \lambda^3=\Omega^2\lambda, $$

where $\Omega=\sqrt{\delta^2+\omega_{\perp}^2}$, we have that $$M^3=\Omega^2M.$$

Then ($M^4=\Omega^2M^2$, $M^5=\Omega^4M$, $\dots$)

$$ e^{iMt}=\mathbb{I}+iMt+\frac{(iMt)^2}{2!}+...=\mathbb{I}-\frac{M^2}{\Omega^2}+\frac{M^2}{\Omega^2}\cos{(\Omega t)}+i\frac{M}{2}\sin{(\Omega t)}. $$

One can finally compute the probabilities:

$$ \mathcal{P}(1\rightarrow 0,t)=|B(t)|^2=\frac{\delta^2\omega_\perp^2}{\Omega^4}2\sin^4{\left(\frac{\Omega t}{2}\right)}+\frac{\omega^2_\perp}{2\Omega^2}\sin^2{(\Omega t)}, $$ $$ \mathcal{P}(1\rightarrow -1,t)=|C(t)|^2=\frac{\omega_\perp^4}{\Omega^4}\sin^4{\left(\frac{\Omega t}{2}\right)}. $$


You can transform to the rotating frame as follows: \begin{equation} \psi_{\mathrm{rot}} (t) = \hat{U}(t)\psi(t), \end{equation} where the time-dependent unitary transformation $U(t)$ is defined by \begin{equation} U(t) \equiv \exp\left(i\omega t S_{z}/\hbar\right) = \begin{pmatrix} e^{i\omega t}&0&0\\ 0&1&0\\0&0&e^{-i\omega t}\end{pmatrix}. \end{equation} One can verify that $\psi_{\mathrm{rot}} (t)$ satisfies \begin{equation} i\hbar\dot{\psi}_{\!\mathrm{rot}}(t) = H_{\mathrm{rot}}\psi_{\mathrm{rot}}(t), \end{equation} where \begin{equation} H_{\mathrm{rot}} = (\omega_{0}-\omega)S_{z} + \sqrt{2}\omega_{\perp} S_{x}. \end{equation} In the rotating frame, the Hamiltonian is time-independent.

  • $\begingroup$ Ok, but now if I try to find the explicit solution for $\psi_{rot}$ I have the same problem as before. $\endgroup$
    – Charlie
    Commented Oct 2, 2015 at 11:28
  • $\begingroup$ @Charlie Right. But this is a Hamiltonian of the form $B\, \textbf{S}\cdot\hat{\textbf{n}}$, and there exists a unitary operator $V$ such that $V S_{z} V^{-1} = \textbf{S}\cdot \hat{\textbf{n}}$. For this particular problem, $V$ is a rotation about the $y$-axis. $\endgroup$
    – higgsss
    Commented Oct 2, 2015 at 20:15

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.