Problem with number of intervals chosen by quadgk

I am running the following code being called in nested loops that vary n and k:

 g = @(x) alpha(x, n, fun);
 I = @(x) x.*g(R.*x).*besselj(n,j(n+1,k).*x);
 [int, err] = quadgk(I,0,1);

with:

function res = alpha(r, n, fun)

      func = @(theta) feval(fun,r,theta).*cos(n.*theta);
      res = (1/pi) .* quadgk(func,-pi,pi);

end

and the function passed as fun is:

function y = f(r,th)
    y = (4-r.^2).*sin(th).^2;
end

Now the problem I have is that the first call of quadgk chooses a differen number of points than the call in alpha. This causes me to get a matrix dimension error if f since r and th are different sizes. Is there any way to work around this problem?

p.s. I know I can give them a max number of points, but that doesn't help since they could still choose different numbers under that max.

Answer 1

[Titus Edelhofer]

Hi Nicolas,

since the "r" in the function alpha is the variable, it looks to me, as if you have to rewrite alpha to loop on r, something like

function res = alpha(r, n, fun)
  res = zeros(size(r));
  for i=1:length(r)
    func = @(theta) fun(r(i), theta).*cos(n.*theta);
    res(i) = 1/pi * quadgk(func, -pi, pi);
  end
end

Titus