I am trying to linearize the following GP eq: \begin{equation} i\partial_{t}\psi(r,t)=\left[-\frac{\nabla^{2}}{2m}+g\left|\psi(r,t)\right|^{2}+V_{d}(r)\right]\psi(r,t) \end{equation}

The ansatz for the mean-field wavefunction is: \begin{equation} \psi_{0}(r,t)=\psi_{0}\, e^{i(k_{0}r-\omega_{0}t)} \end{equation}

One then has to add the fluctuations on top: \begin{equation} \psi(r,t)=\big[\psi_{0}(r)+\delta\psi(r,t)\big]\, e^{-i\omega_{0}t} \end{equation}

Pluggin this into the original equation we get to 0 order that \begin{equation} \omega_{0}-\frac{k_{0}^{2}}{2m}=g|\psi_0|^2. \end{equation}

Expanding to first order (linear response) we get \begin{equation} i\partial_{t}\delta\vec{\psi}=\mathcal{L}\cdot\delta\vec{\psi}+\vec{F}_{d} \end{equation}

with \begin{equation} \delta\vec{\psi}(r,t)=\left(\begin{array}{c} \delta\psi(r,t)\\ \delta\psi^{\star}(r,t) \end{array}\right) \end{equation}

\begin{equation} \vec{F}_{d}(r)=V_{d}(r)\,\left(\begin{array}{c} \psi_{0}(r)\\ -\psi_{0}^{*}(r) \end{array}\right) \end{equation}

\begin{equation} \mathcal{L}=\left(\begin{array}{cc} -\frac{k_{0}^{2}}{2m}-\frac{\nabla^{2}}{2m}+g|\psi_{0}|^2 & g\psi_{0}^{2}\, e^{2ik_{0}r}\\ -g\psi_{0}^{2\star}\, e^{-2ik_{0}r} & -\left(-\frac{k_{0}^{2}}{2m}-\frac{\nabla^{2}}{2m}+g|\psi_{0}|^2\right) \end{array}\right) \end{equation}

The goal here is to determine $\delta\vec{\psi}(r,t)$, by diagonalizing $\mathcal{L}$ and expanding on the corresponding eigenmodes. I am trying to follow these notes http://arxiv.org/abs/cond-mat/0105058v1, which give the general formalism starting on page 66.

However, the author says $\mathcal{L}$ is not diagonalizable in general therefore one has to do the trick of splitting $\delta\psi$ into a part along $\psi_0$ and a part orthogonal to it (see eq. 229). This leads to a new operator $\mathcal{L}$, given by (235) which is diagonalizable.

How can I apply this formalism to my problem? How do I construct the new operator $\mathcal{L}$? How do I deal with the projection operators (233) and (234), etc

  • $\begingroup$ What is $V_d(r)$ in your Lagrangian? Is it just the harmonic trap? What is your end goal - is it to obtain the eigenmodes of the system, or is something else? $\endgroup$
    – jarm
    Commented Apr 18, 2015 at 12:36
  • $\begingroup$ $V_d(r)$ is a defect potential, which in the simplest case is a delta function $\delta(r-r_0)$. The end goal is to calculate the effect of this perturbation caused by the defect, i.e. calculate $\delta \psi$. $\endgroup$
    – Andrei
    Commented Apr 18, 2015 at 14:53
  • $\begingroup$ But then I do not understand the question, as it seems to me that Y. Castin does just what you want in his review. His perturbation $\delta U$ can depend on the coordinate and be time independent, such that $\delta U = V_d(\boldsymbol r)$. It seems to me Eq. 255 is exactly what you need. Could you please clarify if I am wrong? $\endgroup$
    – jarm
    Commented Apr 22, 2015 at 8:58
  • $\begingroup$ I agree that his formalism should be applicable, that is why I mention that reference :) I just don't know how to apply it in practice to my problem. $\endgroup$
    – Andrei
    Commented Apr 24, 2015 at 10:01
  • $\begingroup$ I think that your problem was actually explicitly treated in the reference! Otherwise, it could be that I do not understand your problem. Could you explain to me what is the difference between your problem and the one treated in the reference? :) $\endgroup$
    – jarm
    Commented Apr 24, 2015 at 15:44

1 Answer 1


Apologies for a very late response. I recently joined this community and am posting this answer for the benefit of the OP of this unanswered question.

The decay of a density of particles governed by stochastic differential equation (SDE) dynamics to the equilibrium density is a well known topic in statistical mechanics, for which you may see the book by Risken. For the stochastic process $x_s \sim p(s,x)$, for all $0 \leq s$, with the SDE dynamics $d x_s = b(x_s) ds + \sigma dw_s$ and the corresponding linear Fokker-Planck (FP) partial differential equation (PDE) dynamics $\partial_t p = \mathcal{L}^\dagger p$, the question is whether any initial density will decay to the invariant density $p^\infty(x)$ which solves the stationary FP equation and at what rate? This problem has been answered for certain systems, namely, Ornstein-Uhlenbeck and Langevin systems. The key idea is to consider the initial density as a perturbation from the equilibrium given as $p(0,x) = p^\infty(x) + p^\infty(x) \tilde{p}(0,x)$ and prove that $\tilde{p}(s,x)$ vanishes under the limit $s \rightarrow +\infty$, by using the eigen properties of the generator $\mathcal{L}$ of the SDE which is indeed the adjoint of the forward FP operator $\mathcal{L}$. Thus, properties of the SDE dynamics determine whether or not we can show that the equilibrium density is indeed attained by particles with any arbitrary initial density.

Similarly, in the case of the linear Schrödinger equation, eigen properties and corresponding stability results have been explored, in particular please consider the book by Berezen on this topic, with comprehensive exploration for various classes of Schrödinger potentials. More recently, this analysis has found application in the theory of mean field games (MFGs) starting with the seminal work in the integrator dynamics regime (similar to the Ornstein-Uhlenbeck case in the physics literature) and more recently in the case of Langevin dynamics. The former work considers the case of simple integrator SDE dynamics for a particular class of MFGs, which shows strong analogies with the non-linear Schrödinger and explicitly treating the Gross-Pitaevskii equation. The latter work is more general in terms of the nonlinear Langevin SDE dynamics for a general class of large-population systems, but shows stability results using the linear Schrödinger equation.

More specifically, the cited references indeed treat the diagonalization of PDE systems which correspond to that mentioned in the original question, with the key idea being the linearization and subsequent linear analysis using the eigen properties of the linear operators governing the local dynamics close to the equilibrium.


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.