The Ginzburg-Landau free energy is defined as $$F = \int dV \left \{\alpha |\psi|^2 + \frac{\beta}{2}|\psi|^4 + \frac{1}{2m^*} \left| \left(\frac{\hbar}{i}\nabla - \frac{e^*}{c}\vec{A}\right)\psi \right|^2 + \frac{h^2}{8\pi}\right \},$$ where $\vec{h}=\nabla\times \vec{A}$. The current equation is derived by taking the variation of the vector field $\vec{A}$. My concern is mainly about the variation of the last term in the free energy: $$\delta \int dV\,\left(\nabla\times \vec{A}\right)^2 = 2\oint_S \left(\delta \vec{A} \times \vec{h}\right)\cdot d\vec{S} + 2\int dV \,\delta\vec{A}\cdot \left(\nabla\times \vec{h}\right).$$ However, in arriving at the final current equation, usually the surface integral is ignored. Why is that so?

In the original paper of Ginzburg (above equation 9), he said:

... by varying this expression with respect to $\psi^*$ and $A$, we obtain the equations defining $\psi$ and $A$ (it must be assumed that ${\rm div} A = 0$)....

Is he implying that the surface term is ignored because $\nabla\cdot \vec{A}=0$? If so, why?

I can also think of other two possibilities:

  1. We manually require $\delta\vec{A}$ to vanish at surface boundary. If so, why is a similar requirement not applied to $\delta \psi^*$?
  2. We can also ignore the surface integral by an appropriate choice of gauge, which is complicated but exist.

What is your opinion?

  • $\begingroup$ I am not able to answer your question but I find very nice your nickname and the associated image "un'offerta che non puoi rifiutare". Very funny! $\endgroup$ Dec 7 '20 at 10:45
  • $\begingroup$ @ValterMoretti :) $\endgroup$ Dec 7 '20 at 10:54
  • 3
    $\begingroup$ When computing variations of the action (e.g. deriving the Euler-Lagrange equations) it is typically assumed that boundary terms vanish. Is there a particular reason why it should be different here? $\endgroup$
    – fqq
    Dec 10 '20 at 2:13

Short answer: the key is that $\delta \vec{A}$ is arbitrary, so you can set it to whatever you need.

In the equation $$\delta \int dV\,\left(\nabla\times \vec{A}\right)^2 = 2\oint_S \left(\delta \vec{A} \times \vec{h}\right)\cdot d\vec{S} + 2\int dV \,\delta\vec{A}\cdot \left(\nabla\times \vec{h}\right),$$ you get rid of the surface term in the same breath as you get rid of $\int dV$ to pass from an integral expression to an expression valid point by point:

$\delta \vec{A}$ is arbitrary, so you set it equal to zero everywhere apart from an infinitesimal volume $dV$ around an arbitrary point $x$.


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.