Given the problem;

$\displaystyle\cases{ y'(t)=y^2(t) & \cr y(0)=1 }$ for $t\in[0,1]$

Using the implicit euler method, find an explicit formula to get $y_{n+1}$

HINT: The method has to be consistent.

The problem is a special case of Bernoulli equation with the solution $y(t)=\frac{1}{1-t}$, now do I have to find an approximate formula without recurrence relation, is that correct ?

Consistency is defined as: $\epsilon_n=y(t_{n+1})-y(t_n)-h_nf(t_{n+1},y(t_{n+1}))$ for the implicit euler method, but since I must get here an explicit formula, what does it mean ?

  • $\begingroup$ My guess: You write up the formula for implicit Euler and try to solve it for $y_{n+1}$. You get a quadratic equation, so you have to pick one of its roots. That is where the consistency comes in. $\endgroup$ – Harald Hanche-Olsen Oct 22 '14 at 15:33
  • $\begingroup$ @Harald Hanche-Olsen I don't know if I'm on the right track but, I got $y_{n+1}=y(t_0)+\sum_{i=0}^{n}h_iy^2(t_{i+1})$ $\endgroup$ – derivative Oct 22 '14 at 15:44
  • $\begingroup$ I think whoever posed the question only wants you to write an explicit formula for $y_{n+1}$ in terms of $y_n$, not an explicit formula in terms of $n$ and the initial data, which is much harder. But if I'm right, this should have been stated more clearly. $\endgroup$ – Harald Hanche-Olsen Oct 22 '14 at 16:50
  • $\begingroup$ @Harald Hanche-Olsen the assistant has only mentioned that, it has to do something with taylor formula $\endgroup$ – derivative Oct 22 '14 at 18:51

With steplength $h$, the implicit Euler method for this equation would say $$y_{n+1}-y_n=hy_{n+1}^2.$$ Seeing this as a quadratic equation to solve for $y_{n+1}$, we get $$y_{n+1}=\frac{1\pm\sqrt{1-4hy_n}}{2h},$$ but you need to pick the correct sign. For concistency, you absolutely need to have $y_{n+1}$ converging to $y_n$ if you let $h\to0$, and that immediately eliminates the plus sign (which would have the right hand side going off to $\pm\infty$), leaving $$y_{n+1}=\frac{1-\sqrt{1-4hy_n}}{2h}$$ as the only sensible choice.

You may, and probably should, go further and verify that this is in fact consistent. This might involve taking a Taylor expansion of the square root, or perhaps my favourite trick for fractions of this kind: Multiplying numerator and denominator by $1+\sqrt{1-4hy_n}$. I'll stop here, as you do need to do some if this work on your own.

| cite | improve this answer | |

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.