2
$\begingroup$

How to find the general solution of a differential equation with a shift, in the following form?

$$\frac{\partial}{\partial t}g(x,t)=g(x-\Delta,t)+\frac{\partial^2}{\partial x^2} g(x,t)$$

where $\Delta > 0$. And what about the following?

$$\frac{\partial}{\partial t}g(x,t)=g(x,t-\Delta)+\frac{\partial^2}{\partial x^2} g(x,t)$$


Edit1: Here are few follow-up details about my question. Is there a ``nice" way to represent the solution in $x$-space, as opposed to e.g., Fourier? Is the solution real + positive + normalizable? Does it have the correct properties of a probability density function?

$\endgroup$
  • 1
    $\begingroup$ as you can see from the general solution I wrote down, the normalization $N(t)=\int_{-\infty}^\infty g(x,t)dt=G(0,t)=e^{t}G(0,0)=e^{t}N(0)$ increases exponentially with time; so this is not a probability density function (why did you expect that?) $\endgroup$ – Carlo Beenakker Jul 22 '20 at 10:45
3
$\begingroup$

Fourier transform $G(k,t)=\int_{-\infty}^\infty e^{ikx} g(x,t)dx$ with respect to $x$, then $$\frac{\partial}{\partial t}G(k,t)=e^{ik\Delta}G(k,t)-k^2 G(k,t),$$ hence $$G(k,t)=\exp\left(te^{ik\Delta}-tk^2\right)G(k,0).$$ For the second differential equation you would similarly Fourier transform with respect to $t$.

$\endgroup$
  • $\begingroup$ Thanks. Can you please continue the solution for few more steps? Is it real, or have an imaginary part? How do you than transform back from k to x-space? $\endgroup$ – user1611107 Jul 22 '20 at 9:58
  • $\begingroup$ to transform back to $x$-space you calculate $g(x,t)=(2\pi)^{-1}\int_{-\infty}^\infty e^{-ikx}G(k,t)dk$. This is the general solution, it will be real if $g(x,0)$ is real; it cannot be worked out further without further information on $g(x,0)$. $\endgroup$ – Carlo Beenakker Jul 22 '20 at 10:41
  • $\begingroup$ Ok, thank you. About the normalization: I was thinking that it is simpply a diffusion equation, with a drift term (even if the drift dependson x-\Delta). But you are right, it's a source. My mistake. $\endgroup$ – user1611107 Jul 22 '20 at 11:34
  • $\begingroup$ The thing is, the idea of Fourier I am familiar with, but the backward transformation of the exp[t exp(i k Delta)] is something that I can't work out. Is there a simple form of g(x,t) that you can demonstrate the inverse transform on? E.g., if g(x,t)=delta(x) or a narrow Gaussian? Thanks again $\endgroup$ – user1611107 Jul 22 '20 at 11:39
  • 1
    $\begingroup$ the inverse Fourier transform does not have a closed form expression for the delta function initial condition; you would need to calculate integrals of the form $\int_0^{\infty } e^{\cos k-k^2} \cos( k x) \cos (\sin k) \, dk$, which can only be done numerically. $\endgroup$ – Carlo Beenakker Jul 22 '20 at 12:07

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.