I cannot calculate the following type of integrals numerically :

$\int_0^1 dy \int_0^y f(x) dx $

$f(x)$ can be a complicated function. The problem is due to the fact that the upper limit of one of the integrals is not a number. Can anyone suggest how to compute it. Many thanks!

Edit : Actually my aim is to calculate a residue. I didn't want to bother you with these long expressions. Here is what I wish to compute (numerically at least).

$$\text{Res}\left(\frac{\int_0^x \text{P}(w) S(w) \, dw}{\text{P}(x)},\left\{x,\frac{1}{\pi }\right\}\right)$$ where, $$\text{P}(\text{z}) \text{ := }\frac{1-\pi ^4 z^4}{z^3}$$ $$S(\text{z})\text{ := } \pi ~ A(z) F(z)-\pi ^2 B(z)+\frac{\pi F'(z)}{2 (1-\pi z)} $$ $$F(\text{z})\text{ :=}\int_0^z \frac{\int_0^y \pi A(u) \text{P}(u) \, du}{\text{P}(y)} \, dy-\frac{3 \left(\log \left(\pi ^2 z^2+1\right)+\log (\pi z+1)+\log (1-\pi z)\right)}{64 \pi ^4}$$ $$A(\text{z})\text{ :=}\frac{\pi z (\pi z+2)+3}{4 z (\pi z-1) (\pi z+1) \left(\pi ^2 z^2+1\right)}$$ $$B(\text{z})\text{ :=}\frac{(\pi z (\pi z+2)+3) (\pi z (\pi z (\pi z+1)+1)+5)}{16 (\pi z-1) (\pi z+1)^2 \left(\pi ^2 z^2+1\right)^2}$$

  • $\begingroup$ Can you give us an example of $y$, and possibly a reduced version of $f(x)$ as well? $\endgroup$
    – MarcoB
    May 13, 2015 at 8:45
  • $\begingroup$ Possible duplicate of mathematica.stackexchange.com/questions/17766/… $\endgroup$
    – mattiav27
    May 13, 2015 at 9:38
  • $\begingroup$ Also this: mathematica.stackexchange.com/questions/83335/… $\endgroup$
    – mattiav27
    May 13, 2015 at 9:40
  • 1
    $\begingroup$ It's a repeated integral; what you have is equivalent to $$\int_0^1 (1-x)f(x)\mathrm dx$$ That being said, and do forgive me for my bluntness, you've been quite vague as to the nature of your $f(x)$, and this makes it difficult to assist you. Methods that are efficient for one class fail spectacularly for another, so more elaboration is necessary. $\endgroup$ May 13, 2015 at 9:54
  • $\begingroup$ You see, you should have included those expressions at the outset. It's gnarly now, but it should be amenable to some simplification. $\endgroup$ May 15, 2015 at 0:41

1 Answer 1


Something like this should work (assuming that y is not a complicated function, but it is easily generalized to that case):

f[x_?NumberQ]:=(*your function*)



As an example:

f[x_?NumberQ] := x;
I1[y_?NumberQ] := NIntegrate[f[x], {x, 0, y}];
NIntegrate[I1[y], {y, 0, 1}]

with the result:


You can check with Integrate that theresult is correct.

The _?NumberQ in the definition of the functions mean that the function will be evaluated only if the input variable has a numerical value. A link to its use.

And here a link to a problem similar to yours.

  • $\begingroup$ Thanks @mattiav27 ! Can you suggest how to modify it for complicated functions. Actually my functions are really complicated. Your code works fine for simple functions, I have checked. I don't understand why it isn't working for complicated functions. $\endgroup$ May 13, 2015 at 10:15
  • $\begingroup$ @pinu I need your function: edit you first post $\endgroup$
    – mattiav27
    May 13, 2015 at 10:34
  • $\begingroup$ @pinu: meanwhile you can take a look at this old question of mine: there are two nested numerical integrations with complicated functions mathematica.stackexchange.com/a/56212/8822 $\endgroup$
    – mattiav27
    May 13, 2015 at 12:20

Your Answer

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

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