Inspired by this question I was wondering whether "recursive" integrals have been studied or if they appear anywhere in applications. What I mean is the following: Let $I(x, y) = \int_x^y g(t) dt$ and define

$$\begin{aligned} I_0 &= \int_x^y g(t) dt &&= I(x, y)\\ I_1 &= \int_{\int_y^xg(t)dt}^{\int_x^yg(t)dt}g(t) dt &&= I(-I(x,y), +I(x, y)) \\ I_2 &= \int_{\int^{\int_y^xg(t)dt}_{\int_x^yg(t)dt}g(t) dt}^{\int_{\int_y^xg(t)dt}^{\int_x^yg(t)dt}g(t) dt} g(t) dt &&=I(-I(-I, +I), +I(-I, +I)) \\ &&\vdots \\ I_{n+1} &= I(-I_n, +I_n) \end{aligned}$$

What can be said about convergence criteria for this process? It should be straight-forward to show that for non-negative functions a blow up happens if $\int^x g(t) dt$ grows asymptotically faster than $x$. Otherwise it should converge I think. But who knows what could happen when fast oscillating functions taking both positive and negative values are involved.

  • $\begingroup$ Let $F(u) = \int_{-u}^u g(t)\,dt$. Then you ask about the iterates of $F$; a subject with vast literature. So, how does the assertion that $F(u) = \int_{-u}^u g(t)\,dt$ for some $g$ restrict the choice of the function $F$? $\endgroup$
    – GEdgar
    Feb 6 at 15:46

1 Answer 1


Let $$z_0 = \int\limits_x^y g(t)\,\text dt,\quad F(z)=\int\limits_{-z}^z g(t)\,\text dt.$$

Then we have a simple iteration $$z_{i+1}=F(z_i),$$ with the sufficient condition of convergency in the form of $$|F(z)|<\gamma|z|,\qquad(|z|\le|z_0|)$$ where $\;\gamma<1.\;$


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.