2022.04.09 — two level system driven with noise

last update: 2023.01.20

in this post I want to collect some results / formulas concerning population transfer between two quantum states.

dynamical equations

we assume that the hamiltonian $h$ governing the system dynamics is composed of a time independent "unperturbed" part $h_o$ and a time dependent "perturbation" $h'$, i.e. \begin{equation} h = h_o + h' \end{equation} let $\ket{-}$ and $\ket{+}$ be respectively the low and high energy eigenstates of $h_o$, i.e. \begin{equation} h_o = \Delta \epsilon \sigma_z \end{equation} where \begin{equation} \sigma_z = \op{+}{+} - \op{-}{-} \end{equation} is the $z$ pauli matrix, so that the energy separation between $\ket{+}$ and $\ket{-}$ is \begin{equation} \bra{+} h_o \ket{+} - \bra{-} h_o \ket{-} = 2 \Delta \epsilon \end{equation} the perturbation $h'$ has the form \begin{equation} h' = \epsilon(t) \sigma_x \end{equation} where \begin{equation} \sigma_x = \op{+}{-} + \op{-}{+} \end{equation} is the $x$ pauli matrix, and \begin{equation} \epsilon(t) = \epsilon_o \cos{\phi(t)} \end{equation} where $\epsilon_o$ is some constant and \begin{equation} \phi(t) = \phi_o + \int_{t_o}^\infty dt' \omega(t') \end{equation} where $\phi_o$ is the phase of the perturbation at $t=t_o$ and \begin{equation} \omega(t) = \omega_o + \Delta \omega + \delta \omega(t) \end{equation} where \begin{equation} \hbar \omega_o = 2 \Delta \epsilon \label{omegao} \end{equation} and $\Delta \omega$ is some fixed "detuning", while $\delta \omega(t)$ is a randomly fluctuating phase with the following statistical properties \begin{align} \mathbb{E} \left[ \delta \omega(t) \right] & = 0 \\ \mathbb{E} \left[ \delta \omega(t_1) \delta \omega(t_2) \right] & = 2 \gamma \delta(t_1 - t_2) \end{align} where $\mathbb{E}\left[ \cdot \right]$ denotes a (classical) expectation value, $\gamma$ is a constant and $\delta(t)$ is the dirac delta function. by the way, if we want to, we can write the total hamiltonian $h$ in the following, somewhat more "geometric" form \begin{equation} h = \bm{h} \cdot \bm{\sigma} \end{equation} where \begin{equation} \bm{h} = \left( h_x,h_y,h_z \right) = \left( \epsilon(t),0,\Delta \epsilon \right) \end{equation} and \begin{equation} \bm{\sigma} = \left( \sigma_x,\sigma_y,\sigma_z \right) \end{equation}

given a state $\ket{\psi(t)}$, we can define amplitudes $c_\pm(t)$ by the rule \begin{equation} c_\pm (t) = \ip{\pm}{\psi(t)} e^{\pm i \frac{\phi(t)}{2}} \end{equation} if $\ket{\psi (t)}$ satisfies the schrodinger equation, i.e. $\frac{d}{dt} \ket{\psi(t)} = \left(i \hbar\right)^{-1} h \ket{\psi(t)}$, then the $c_\pm(t)$ satisfy the equation: \begin{equation} \begin{split} \frac{d}{dt} c_\pm(t) & = \ip{\pm}{\dot{\psi}} e^{\pm i \frac{\phi(t)}{2}} + \ip{\pm}{\psi(t)} \frac{d}{dt} e^{\pm i \frac{\phi(t)}{2}} \\ & = \bra{\pm} \left[ \left(i \hbar\right)^{-1} h \pm \frac{i}{2} \frac{d}{dt} \phi(t) \right] \ket{\psi} e^{\pm i \frac{\phi(t)}{2}} \\ & = \bra{\pm} \left[ \left(i \hbar\right)^{-1} h + \frac{i}{2} \frac{d}{dt} \phi(t) \sigma_z \right] \ket{\psi} e^{\pm i \frac{\phi(t)}{2}} \end{split} \label{cpm} \end{equation} inspecting the bracketed expression, we find \begin{equation} \begin{split} & \left(i \hbar\right)^{-1} h + \frac{i}{2} \frac{d}{dt} \phi(t) \sigma_z \\ = & \left(i \hbar\right)^{-1} \left( h_o + h' \right) + \frac{i}{2} \frac{d}{dt} \phi(t) \sigma_z \\ = & \left(i \hbar\right)^{-1} \left( \Delta \epsilon \sigma_z + \epsilon(t) \sigma_x \right) + \frac{i}{2} \frac{d}{dt} \phi(t) \sigma_z \\ = & \left(i \hbar\right)^{-1} \left( \Delta \epsilon \sigma_z + \epsilon(t) \sigma_x \right) + \frac{i}{2} \left( \omega_o + \Delta \omega + \delta \omega(t) \right) \sigma_z \\ = & i \left( \left( \frac{\omega_o}{2} - \frac{\Delta \epsilon}{\hbar} + \frac{1}{2} \left(\Delta \omega + \delta \omega(t)\right) \right) \sigma_z - \frac{\epsilon(t)}{\hbar} \sigma_x \right) \\ (!) = & i \left( \frac{1}{2} \left(\Delta \omega + \delta \omega(t)\right) \sigma_z - \frac{\epsilon(t)}{\hbar} \sigma_x \right) \end{split} \end{equation} where in the step marked (!) we apply equation \eqref{omegao}. inserting this expression back into equation \eqref{cpm}, we get \begin{equation} \begin{split} \frac{d}{dt} c_\pm(t) & = \bra{\pm} \left[ i \left( \frac{1}{2} \left( \Delta \omega + \delta \omega(t) \right) \sigma_z - \frac{\epsilon(t)}{\hbar} \sigma_x \right) \right] \ket{\psi} e^{\pm i \frac{\phi(t)}{2}} \\ & = \left( i \hbar \right)^{-1} \left( \mp \frac{1}{2} \hbar \left( \Delta \omega + \delta \omega(t) \right) c_\pm(t) + \left[ \bra{\pm} \epsilon(t) \sigma_x \ket{\psi(t)} e^{\pm i \frac{\phi(t)}{2}} \right] \right) \end{split} \label{hp} \end{equation} zooming in now on the bracketed expression: \begin{equation} \begin{split} & \bra{\pm} \epsilon(t) \sigma_x \ket{\psi(t)} e^{\pm i \frac{\phi(t)}{2}} \\ = & \bra{\mp} \epsilon_o \cos \phi(t) \ket{\psi(t)} e^{\pm i \frac{\phi(t)}{2}} \\ = & \bra{\mp} \epsilon_o \cos \phi(t) \ket{\psi(t)} e^{\pm i \frac{\phi(t)}{2}} \\ = & \bra{\mp} \epsilon_o \frac{1}{2} \left( e^{i\phi(t)} + e^{-i\phi(t)} \right) \ket{\psi(t)} e^{\pm i \frac{\phi(t)}{2}} \\ (!) = & \frac{1}{2} \bra{\mp} \epsilon_o \ket{\psi(t)} e^{\mp i \frac{\phi(t)}{2}} \\ = & \frac{1}{2} \epsilon_o c_\mp(t) \end{split} \label{cmp} \end{equation} where in the step marked (!) we apply the rotating wave approximation, valid for $\omega_o \gg \Delta \omega, \delta \omega, \gamma, \frac{\epsilon_o}{\hbar}$. substituting \eqref{cmp} into \eqref{hp}, we arrive at \begin{equation} \dot{c}_\pm = (i \hbar)^{-1} \left( \mp \frac{1}{2} \hbar \left( \Delta \omega + \delta \omega \right)c_\pm + \frac{1}{2} \epsilon c_\mp \right) \end{equation} this equation can be recast in the following form of a "modified" schrodinger equation: \begin{equation} \ket{\dot{\tilde{\psi}}} = \left( i \hbar \right)^{-1} \tilde h \ket{\tilde{\psi}} \end{equation} where \begin{equation} \ket{\tilde{\psi}} = e^{i \frac{\phi(t)}{2} \sigma_z} \ket{\psi} \label{hpp} \end{equation} and \begin{equation} \tilde{h} = \bm{\tilde{h}} \cdot \bm{\sigma} \end{equation} where \begin{equation} \bm{\tilde{h}} = \left( \tilde{h}_x, \tilde{h}_y, \tilde{h}_z \right) = \left(\frac{1}{2} \epsilon_o, 0, -\frac{1}{2} \hbar \left( \Delta \omega + \delta \omega(t) \right) \right) \end{equation} the time evolution of the expectation values $\Sigma_i = \bra{\tilde{\psi}} \sigma_i \ket{\tilde{\psi}}$ of the spin operators $\sigma_i$ for a state $\ket{\tilde{\psi}}$ satisfying \eqref{hpp} is given by \begin{equation} \begin{split} \frac{d}{dt} {\Sigma}_i(t) & = \frac{d}{dt} \bra{\tilde{\psi}} \sigma_i \ket{\tilde{\psi}} \\ & = \bra{\tilde{\psi}} \left[ (i \hbar)^{-1} \comm{\tilde{h}}{\sigma_i} \right] \ket{\tilde{\psi}} \\ & = \bra{\tilde{\psi}} \left[ (i \hbar)^{-1} \sum_{j=1}^3 \tilde{h}_j \comm{\sigma_j}{\sigma_i} \right] \ket{\tilde{\psi}} \\ (!) & = (i \hbar)^{-1} \sum_{j=1}^3 \tilde{h}_j \left( \sum_{k=1}^3 2 i \epsilon_{jik} \bra{\tilde{\psi}} \sigma_k \ket{\tilde{\psi}} \right) \\ & = (i \hbar)^{-1} \sum_{j=1}^3 \tilde{h}_j \left( \sum_{k=1}^3 2 i \epsilon_{jik} \Sigma_k \right) \\ & = \sum_{j,k=1}^3 \epsilon_{ikj} \Sigma_k \frac{2\tilde{h}_j}{\hbar} \\ & = \left( \bm{\Sigma} \times \bm{\Omega} \right)_i \end{split} \label{rabi} \end{equation} i.e. \begin{equation} \dot{\bm{\Sigma}} = \bm{\Sigma} \times \bm{\Omega} \label{rabivec} \end{equation} where \begin{equation} \bm{\Sigma} = \left( \Sigma_x, \Sigma_y, \Sigma_z \right) \end{equation} and \begin{equation} \begin{split} \bm{\Omega} & = \left( \Omega_x, \Omega_y, \Omega_z \right) \\ & = \frac{2 \bm{\tilde{h}}}{\hbar} \\ & = \left( \tilde{\omega}, 0, -\Delta \omega - \delta \omega(t) \right) \end{split} \end{equation} where \begin{equation} \tilde{\omega} = \frac{\epsilon_o}{\hbar} \end{equation} note that back in equation \eqref{rabi} we apply in the step marked (!) the commutation relations for the paul spin operators.

it is worth asking of what relevance is equation \eqref{rabivec}, since it gives the time evolution of a state evolving under the influence of $\tilde{h}$, while our system evolves under $h$. if we are only concerned in populations $P_\pm(t) = \left| \ip{\pm}{\psi} \right|^2$ (as opposed to coherences), however, we find that we are ok: \begin{equation} \begin{split} P_\pm(t) & = \left| \ip{\pm}{\psi} \right|^2 \\ & = \left| \bra{\pm}\left(e^{-i \frac{\phi(t)}{2} \sigma_z} e^{+i \frac{\phi(t)}{2} \sigma_z}\right) \ket{\psi} \right|^2 \\ & = \left| \left( e^{+i \frac{\phi(t)}{2} \sigma_z} \ket{\pm} \right)^\dagger \left( e^{+i \frac{\phi(t)}{2} \sigma_z} \ket{\psi} \right) \right|^2 \\ & = \left| \left( e^{ \pm i \frac{ \phi(t) }{ 2 } } \ket{\pm} \right)^\dagger \ket{ \tilde{\psi} } \right|^2 \\ & = \left| e^{\mp i \frac{\phi(t)}{2}} \ip{\pm}{\tilde{\psi}} \right|^2 \\ & = \left| \ip{\pm}{\tilde{\psi}} \right|^2 \\ & = \ip{\tilde{\psi}}{\pm}\ip{\pm}{\tilde{\psi}} \\ & = \frac{1}{2} \left( \ip{\tilde{\psi}}{\pm}\ip{\pm}{\tilde{\psi}} + \ip{\tilde{\psi}}{\mp}\ip{\mp}{\tilde{\psi}} \right) + \frac{1}{2} \left( \ip{\tilde{\psi}}{\pm}\ip{\pm}{\tilde{\psi}} - \ip{\tilde{\psi}}{\mp}\ip{\mp}{\tilde{\psi}} \right) \\ & = \frac{1}{2} \left( P_\pm + P_\mp \right) + \frac{1}{2} \left( \pm \Sigma_z \right) \\ & = \frac{1}{2} \left( 1 \pm \Sigma_z \right) \end{split} \end{equation} an expression for the dynamics of the classical expectation value $\bm{\tilde{\Sigma}} = \mathbb{E} \left[ \bm{\Sigma} \right]$ of the spins $\bm{\Sigma}$ over the distribution of phase noises $\delta \omega(t)$ can be determined using results summarized in appendix b of agarwal's "quantum statistical theory of optical-resonance phenomena in fluctuating laser fields" (link). for a dyamical system of the form \begin{equation} \dot{\bm{x}} = M \bm{x} + F(t) \bm{x} \label{ag} \end{equation} where $M$ is a time-independent operator and $F(t)$ is a stochastic operator satisfying \begin{align} \mathbb{E} \left[ F(t) \right] & = 0\,, \label{age1} \\ \mathbb{E} \left[ F_{ij}(t_1) F_{kl}(t_2) \right] & = 2 Q_{ijkl} \delta(t_1 - t_2) \label{age2} \end{align} for some $Q_{ijkl}$, the dynamics of the expectation value $\mathbb{E} \left[\bm{x}\right] = \bm{y}$ satisfies the equation \begin{equation} \dot{\bm{y}} = M \bm{y} + Q \bm{y} \end{equation} where \begin{equation} Q_{ij} = \sum_{k} Q_{ikkj} \end{equation} defining \begin{align} \bm{\Omega_o} & = \left(\tilde{\omega},0,-\Delta \omega\right) \\ \bm{\delta \Omega}(t) & = \left(0,0,-\delta \omega(t)\right) \\ \end{align} so that \begin{equation} \bm{\Omega} = \bm{\Omega_o} + \bm{\delta \Omega}(t) \end{equation} and making the identifications \begin{align} \bm{x} & \to \bm{\Sigma} \, , \\ M & \to \cdot \times \bm{\Omega_o} \, , \\ F(t) & \to \cdot \times \bm{\delta \Omega}(t) \, , \\ \end{align} we find that conditions \eqref{ag}, \eqref{age1}, and \eqref{age2} are satisfied and that \begin{equation} Q \to - \gamma \left( 1 - \bm{\hat{z}} \bm{\hat{z}}^\intercal \right) \end{equation} with \begin{equation} \bm{\hat{z}} = \left(0,0,1\right) \end{equation} so we arrive finally at the following dynamical equation for $\bm{\tilde{\Sigma}}$ \begin{equation} \dot{\bm{\tilde{\Sigma}}} = \bm{\tilde{\Sigma}} \times \bm{\Omega_o} - \Gamma \bm{\tilde{\Sigma}} \label{dyn} \end{equation} where \begin{equation} \Gamma = \gamma \left( 1 - \bm{\hat{z}} \bm{\hat{z}}^\intercal \right) \end{equation}

solution

together $\left[ \cdot \times \bm{\Omega_o} - \Gamma \right]$ form a linear operator, which we will call $A$, so that \begin{equation} \dot{\bm{\tilde{\Sigma}}} = A \bm{\tilde{\Sigma}} \end{equation} as a linear operator, $A$ possesses 3 eigenvectors $\bm{u}_i,\,i=1,2,3$ with associated eigenvalues $\lambda_i.$ if the system is initiatilized in a spin state $\bm{\tilde{\Sigma}}_o$, then after a time interval $T$ the spin state $\bm{\tilde{\Sigma}}'$ will be \begin{equation} \bm{\tilde{\Sigma}}' = \sum_{j=1}^3 \bm{u}_j \left( \bm{\hat{x}}_j \cdot S^{-1} \bm{\tilde{\Sigma}}_o \right) e^{\lambda_j T} \end{equation} where \begin{equation} S_{ij} = \bm{\hat{x}}_i \cdot \bm{u}_j \end{equation} where \begin{equation} \bm{\hat{x}}_1 = \bm{\hat{x}},\,\, \bm{\hat{x}}_2 = \bm{\hat{y}},\,\, \bm{\hat{x}}_3 = \bm{\hat{z}} \end{equation} we are interested in the $z$ component $\bm{\hat{z}} \cdot \bm{\tilde{\Sigma}}'$ of the final spin state starting from an initial state of $\bm{\tilde{\Sigma}}_o = -\bm{\hat{z}}$. this is given by \begin{equation} \bm{\hat{z}} \cdot \bm{\tilde{\Sigma}}' = - \sum_{j=1}^3 \left( \bm{\hat{z}} \cdot \bm{u}_j \right) \left( \bm{\hat{x}}_j \cdot S^{-1} \bm{\hat{z}} \right) e^{\lambda_j T} \end{equation}

noise source

suppose the phase / frequency noise is supplied from a source $\delta \omega(t)$ with rms value $\delta \omega_o$ and a correlation time $\tau$. the correlation function $\mathbb{E} \left[ \delta \omega(t_1) \delta \omega(t_2) \right]$ of such a noise should be \begin{equation} \mathbb{E} \left[ \delta \omega(t_1) \delta \omega(t_2) \right] = \delta \omega_o^2 e^{-\frac{|t_1 - t_2|}{\tau}} \end{equation} if the correlation time $\tau$ is much shorter than all the other relevant time scales (i.e. $\tilde{\omega}^{-1}$, $\Delta \omega^{-1}$, the interaction time $T$), then we can reasonably make the substitution \begin{equation} \frac{e^{-\frac{|t_1 - t_2|}{\tau}}}{2 \tau} \to \delta(t_1 - t_2) \end{equation} so that for such a noise source we can make the identification \begin{equation} \gamma \to \delta \omega_o^2 \tau \end{equation}

rabi frequency

the parameter $\tilde{\omega}$ corresponds to the rabi (radial) frequency, as can be seen by letting $\gamma = \Delta \omega = 0$ and inspecting the dynamics of equation \eqref{dyn} for an initial $\bm{\tilde{\Sigma}}(t_o) = \bm{\hat{z}}$. the spin will trace out a circular motion along a circle of unit radius with a velocity — and thus angular velocity — of $|\dot{\bm{\tilde{\Sigma}}}(t_o)| = \tilde{\omega}$.

according to hilborn's "einstein coefficients, cross sections, f values, dipole moments, and all that" (link), the rabi frequency is related to the electric field amplitude $E$ of the perturbing radiation and the dipole moment $\mu_{+-}$ of the transition by the relation \begin{equation} \tilde{\omega} = \frac{\mu_{+-}E}{\hbar} \end{equation} the cycle-averaged intensity $I$ of a plane electromagnetic wave with electric field amplitude $E$ is given by \begin{equation} I = \frac{1}{2} c \epsilon_o E^2\,, \end{equation} while from chadwick et al.'s "quantum state specific reactant preparation in a molecular beam by rapid adiabatic passage" (link) we find the transition moment $\mu_{+-}$ to be related to the transition's einstein coefficient $A_{+-}$ by the relationship \begin{equation} \mu_{+-} = c\left(j_-, m_- ; 1, 0 | j_+ m_+ \right) \sqrt{\frac{3 \varepsilon_o h c^3}{2 \omega_o^3} A_{+-}} \end{equation} where we can rearrange terms and obtain the following intuitively-appealing expression for the rabi frequency $\bar{f} = \frac{\bar{\omega}}{2 \pi}$: \begin{equation} \bar{f} = \sqrt{A_{+-}A_\gamma} \end{equation} where \begin{equation} A_\gamma = \frac{3}{2} \times \frac{1}{\pi^4} \times \frac{I \times \pi \left(\frac{\lambda}{2}\right)^2}{E_\gamma} \end{equation} where $\lambda = \frac{c}{2 \pi \omega_o}$ is the (vacuum) wavelength of the radiation and $E_\gamma = \hbar \omega_o$ is the energy of the photons making up the radiation. so we can think of $A_\gamma$ roughly speaking as the rate at which photons pass through a circle of diameter equal to the wavelength. the rate $\bar{f}$ at which radiation stimulates rabi cycling is then seen to be the geometric mean of $A_{+-}$, the spontaneous emission rate, and the rate at which photons strike the molecule.