12
$\begingroup$

I would like to solve a simple 2nd-order ODE with one of the boundary conditions defined at $ -\infty $. The ODE I am looking to solve is:

$$ w''(z)-2i\pi^2w(z)=0 $$

with the corresponding boundary conditions:

$$ w(z=-\infty)=0, \; w'(z=0)=0+i\dfrac{\tau_{0}}{\mu}. $$

My attempt at a solution using DSolve is as follows:

DSolve[{-2 I \[Pi]^2 w[z] + (w^\[Prime]\[Prime])[z] == 0, 
w[-Infinity] == 0, w'[0] == 0 + I Subscript[\[Tau], 0]/\[Mu]}, w[z],z]

but I only get an empty set of curly brackets as an output. I checked the rest of my snipet of code without the w[-Infinity]==0 boundary condition, and that works as expected; therefore, I know that this is a problem with the boundary condition at $z=-\infty$. I am looking for methods with which I can solve simple ODE's with boundary conditions at infinity, and any help would be greatly appreciated.

$\endgroup$
2
  • $\begingroup$ Is the boundary condition at z=0 a condition on w or on w'? You have it both ways. $\endgroup$
    – LouisB
    Nov 10, 2016 at 5:31
  • $\begingroup$ The boundary condition at z=0 is for w'. I am sorry about the typo. $\endgroup$
    – Kaszt
    Nov 10, 2016 at 6:32

2 Answers 2

10
$\begingroup$

This is the solution of your equation without the boundary conditions:

sol = DSolve[-2 I \[Pi]^2 w[z] + w''[z] == 0, w[z], z] // ExpToTrig //
   ComplexExpand

(*  {{w[z] -> 
   C[1] Cos[\[Pi] z] Cosh[\[Pi] z] + C[2] Cos[\[Pi] z] Cosh[\[Pi] z] +
     C[1] Cos[\[Pi] z] Sinh[\[Pi] z] - C[2] Cos[\[Pi] z] Sinh[\[Pi] z] + 
    I (C[1] Cosh[\[Pi] z] Sin[\[Pi] z] - C[2] Cosh[\[Pi] z] Sin[\[Pi] z] + 
       C[1] Sin[\[Pi] z] Sinh[\[Pi] z] + C[2] Sin[\[Pi] z] Sinh[\[Pi] z])}}  *)

Now let us take its limit at z->-Infinity:

Limit[w[z] /. sol, z -> -\[Infinity]]

(* {ComplexInfinity}  *)

Let us now try this limit at C[1]=0 and C[2]=0:

Limit[w[z] /. sol /. {C[1] -> 0}, z -> -\[Infinity]]

(* ComplexInfinity *)

Limit[w[z] /. sol /. {C[2] -> 0}, z -> -\[Infinity]]

(*  0  *)

The latter gives us what we need, therefore, C[2]=0.

Let us now implement the second boundary condition:

    Solve[(D[(w[z] /. sol /. C[2] -> 0), z] /. z -> 0) == I*t/m, C[1]]

(*  {{C[1] -> ((1/2 + I/2) t)/(m \[Pi])}}  *)

Done. Have fun!

$\endgroup$
5
  • $\begingroup$ Nice post. I have not see you in a while, I hope you are all good. Regards. $\endgroup$
    – user21
    Nov 10, 2016 at 9:19
  • $\begingroup$ @user21 Thank you, I am fine. $\endgroup$ Nov 10, 2016 at 12:40
  • $\begingroup$ What if some linear combination of the coefficients is zero, but both C[1]and C[2] are non-zero? Then you can't find the solution by testing simple cases. In this particular problem we are lucky: the correct boundary condition at infinity is C[2]==0 instead of Pi C[1]-7C[2]==0 or something. $\endgroup$ Nov 10, 2016 at 16:21
  • $\begingroup$ @JoonasIlmavirta With sol = w[z] /. DSolve[-2 I \[Pi]^2 w[z] + w''[z] == 0, w[z], z][[1]]' check how does List @@ sol look like and the perform Limit[#, z -> -Infinity] & /@ List @@ sol - this explains why C[2]=0 and gives a hint how to proceed with less trivial cases. $\endgroup$
    – corey979
    Nov 10, 2016 at 21:52
  • $\begingroup$ @Joonas Ilmavirta In the case you mentioned one needs to invent an approach. It is exactly the point, where science turns into art. $\endgroup$ Nov 11, 2016 at 8:36
2
$\begingroup$

This method is somewhat similar to Alexei's. We first solve the equation with the boundary condition (b.c.) at 0 i.e. the b.c. DSolve can handle:

generalsol = 
 DSolve[{-2 I π^2 w[z] + w''[z] == 0, w'[0] == 0 + (I Subscript[τ, 0])/μ}, 
   w[z], z][[1, 1, -1]]

(* ((1/2 - I/2) E^((-1 - 
I) π z) ((1 + I) π μ C[1] + (1 + I) E^((2 + 2 I) π z) π μ C[
 1] - I Subscript[τ, 0]))/(π μ) *)

This solution involves E^((-1 - I) π z), which shouldn't exist in a solution that goes to 0 at -Infinity, so coefficient of E^((-1 - I) π z) should be 0. To make the coefficient clearer, let's Collect:

Collect[generalsol, Exp[_], Simplify]
(* E^((1 + I) π z) C[1] + 
 E^((-1 - I) π z) (C[1] - ((1/2 + I/2) Subscript[τ, 0])/(π μ)) *)

Apparently, C[1] - ((1/2 + I/2) Subscript[τ, 0])/(π μ) should be equal to 0, so the particular solution satisfies the b.c. at -Infinity is:

sol = Function[z, #] &[
  E^((1 + I) π z) C[1] /. C[1] -> ((1/2 + I/2) Subscript[τ, 0])/(π μ)]

(* Check: *)
{-2 I π^2 w[z] + w''[z] == 0, w[-∞] == 0, 
   w'[0] == 0 + (I Subscript[τ, 0])/μ} /. w -> sol // Simplify
(* {True, True, True} *)
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.