I'm trying to solve the Fokker-Planck equation

$$\frac{\partial p}{\partial L}(L, \eta)= \frac{\partial}{\partial \eta}\left[\left(\eta^{2}-1\right) \frac{\partial p}{\partial \eta}(L, \eta)\right] = (\eta^2-1)\frac{\partial^2p}{\partial\eta^2}(L,\eta)+2\eta\frac{\partial p}{\partial\eta}(L,\eta), \quad \eta>1$$

with initial condition

$$p(L=0, \eta)=\delta(\eta-1).$$

Ignoring the initial condition my code is

FPE = D[P[L, eta], 
   L] == (1/Lloc)*((eta^2 - 1)*D[P[L, eta], {eta, 2}] + 
     2*eta*D[P[L, eta], eta])

s = NDSolve[{FPE, P[0, eta] == DiracDelta[eta-1]}, P, {L, 0, 30}, {eta, 1, 30}]

to which I get the error "Encountered non-numerical value for a derivative at L == 0." Is there something I have done wrong? I know this FPE has an analytical solution in terms of Legendre polynomials. Any help would be much appreciated. Thanks.

  • 2
    $\begingroup$ The meaning of the error message seems pretty straightforward to me: you are trying to solve for three unknowns (i.e. the function and its derivatives), but only provided one equation. You need more. Notice also that NDSolve is a numerical solver: you also need information on the range for L and eta, and initial conditions. Everything should have a numerical value. If you want solutions depending on parameters, look at ParametricNDSolve. Perhaps start from the docs and try to reformulate your problem. $\endgroup$ – MarcoB Dec 6 '19 at 17:21
  • $\begingroup$ Thanks for your comment. I have updated my question. $\endgroup$ – rami_salazar Dec 6 '19 at 17:32
  • $\begingroup$ @rodger_kicks Don't you need two boundary conditions? You only have an initial condition. Without the boundary conditions, you cannot solve. $\endgroup$ – dearN Dec 8 '19 at 16:37


n = 20;
FPE = D[P[L, eta], L] - D[(eta^2 - 1) D[P[L, eta], eta], eta]
s = Quiet @ NDSolve[{FPE == 0, P[0, eta] == D[Tanh[n (eta - 1)], eta]}, P, {L, 0, 30}, {eta, 1, 30}][[1]]
Plot3D[Evaluate[P[L, eta] /. s], {L, 0, 30}, {eta, 1, 30}]

Here the problem is with DiracDelta[]. Use instead an approximation as

DiracDelta[eta] ~ D[Tanh[n eta], eta]

for $n \gg 1$ and $\eta \ge 0$

  • $\begingroup$ I can't thank you enough. That approximation is amazing. I didn't think of that. $\endgroup$ – rami_salazar Dec 7 '19 at 19:04

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.