MathGroup Archive 2006

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Relatively simple, but problematic, non-linear ODE

  • To: mathgroup at smc.vnet.net
  • Subject: [mg72002] Re: Relatively simple, but problematic, non-linear ODE
  • From: dh <dh at metrohm.ch>
  • Date: Thu, 7 Dec 2006 06:26:22 -0500 (EST)
  • Organization: hispeed.ch
  • References: <el685m$2p9$1@smc.vnet.net>


Hi Alan,

your diff. equ. has a singular solution y==2, that is a solution that is 

tangential to a family of solutions. At a point with y==2 we  have 2 

solutions runing through this point and every numerical procedure will 

have problems.

You can convince yourself by calculating and analytical solution:

rho = 0; z = 3.2;

res = y[x] /. NDSolve[{y'[x]^2 - rho y'[x] y[x] + (1/4) y[x]^2 == 1,y[0] 

== 0}, y[x], {x, 0, z}];

Plot[Evaluate[res], {x, -3, z}]



Daniel



Alan wrote:

> I am trying to find positive, non-decreasing solutions y(z)

> on the interval (0,z), z > 0 to the ODE:

> 

> (y')^2 - rho y y' + (1/4) y^2 = 1,   with y[0] = 0.

> 

> where y' = dy/dz,  and the constant parameter |rho| < 1.

> There are generally two solutions. The one I want has the

> same sign as z.

> 

> Here is my code:

> 

> FSolver[z_, rho_] :=

>   Module[{solnpair, vals, ans},

>       solnpair =  y[x] /. NDSolve[{y'[x]^2 - rho y'[x] y[x] + (1/4) y[x]^2 

> == 1,

>           y[0] == 0}, y[x], {x, 0, z}];

>       vals = Re[solnpair /. x -> z];

>       ans = Select[vals, (Sign[#] == Sign[z]) &][[1]];

>    Return[ans];

> 

> You can see that y[x] = 2 is also a solution to the ODE.

> The solution I want apparently switches to y = 2

> at some critical value of z. The problem is that this causes Mathematica to 

> choke.

> For example, for rho = 0, the exact solution is y(z) = 2 Sin(z/2)

> for |z| < Pi and y(z) = 2 for larger z. I can fix things for rho = 0, since 

> I

> am very confident about the answer. But I am less confident about

> the solution for rho not equal to zero.

> 

> For example, run my fragment with z = 3,  rho = 1/2 to see the problems.

> Any suggestions on how to get some reliable output

> from NDSolve for this problem will be much appreciated.

> 

> Thanks!

> alan 

> 

> 



  • Prev by Date: Re: radical equation
  • Next by Date: Re: Points sampled by N[Derivative[]]
  • Previous by thread: Relatively simple, but problematic, non-linear ODE
  • Next by thread: Re: Relatively simple, but problematic, non-linear ODE