Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

That is:


I don't know how to differentiate a integral if functions of $x$ are at its limits.

Can you guys show me how to do this?

share|cite|improve this question
Dear Voldemort, you might take a look at this answer. If it doesn't help, I can make additional suggestions. --- The general formula is $$\frac{d}{dt}\ g(t,\dots,t)=\frac{\partial g}{\partial x_1}(t,\dots,t)+\cdots+\frac{\partial g}{\partial x_n}(t,\dots,t).$$ – Pierre-Yves Gaillard Aug 20 '11 at 3:13
Dear "You know who", look at all the answers that were provided at… not so long ago. – Álvaro Lozano-Robledo Aug 20 '11 at 3:18
The variable $x$ occurs 3 times in your function. The recipe is: replace the first occurrence by $x_1$, the second by $x_2$, the third by $x_3$, call $g$ the new function, and apply the above formula. [I hope you will be given enough time to find the solution by yourself. --- I see that a full solution has already be given... Note there are 3 terms in the solution because of these 3 occurrences.] – Pierre-Yves Gaillard Aug 20 '11 at 3:22
Still not quite get it. In Pierre-Yves Gaillard's formula what does "g(t,...,t)" mean? Is it g(t1,...,t2)? – Voldemort Aug 20 '11 at 4:14
Put $$h(x):=\int_{a(x)}^{b(x)}\!f(x,t)\,dt.$$ You want to compute $h'(x)$. Put $$g(x_1,x_2,x_3):=\int_{a(x_1)}^{b(x_2)}\!f(x_3,t)\,dt.$$ (In particular $h(x)=g(x,x,x)$.) Then compute $$\frac{\partial g}{\partial x_1}(x_1,x_2,x_3)$$ (just differentiate with repect to $x_1$ - keeping $x_2$ and $x_3$ fixed). Then compute $$\frac{\partial g}{\partial x_1}(x,x,x)$$ (just replace $x_1$ by $x$, $x_2$ by $x$, $x_3$ by $x$). Etc. Finally: $$h'(x)=\frac{\partial g}{\partial x_1}(x,x,x)+\frac{\partial g}{\partial x_2}(x,x,x)+\frac{\partial g}{\partial x_3}(x,x,x).$$ – Pierre-Yves Gaillard Aug 20 '11 at 5:27
up vote 1 down vote accepted

You can apply the following rule (Leibniz rule): [edited in response to Didier Piau's comment]:

If $$I(x)=J(u(x),v(x),x),\quad\text{with}\ J(\alpha,\beta,z)=\int_\alpha^\beta f(t,z)dt,\tag{1}$$

then, under suitable conditions, we have

$$I^{\prime }(x)=\displaystyle\int_{u(x)}^{v(x)}\dfrac{\partial f(t,x)}{\partial x}dt+f(v(x),x)v^{\prime }(x)-f(u(x),x)u^{\prime }(x).\tag{2}$$

For further detais see this answer of mine.

Added 3. Pierre-Yves Gaillard's comment proves $(2)$. I derive $(2)$ as follows, by using this old blog post of mine (hereafter $t$ is the independant variable):

  • (i) Let $f\left( x,t\right)$ be a real function defined in a rectangle $R=[ a,b]\times[c,d]\in\mathbb{R}^{2}$. If $f(x)$ is integrable in $x$ for each real value of $t$ and $\dfrac{\partial f\left( x,t\right) }{\partial t}$ is continuous in $x$ and $t$ in $R$, then the derivative of the following parametric integral $$I(t)=\displaystyle\int_{a}^{b}f\left( x,t\right) dx\tag{3}$$ is given by $$I^{\prime }(t)=\displaystyle\int_{a}^{b}\dfrac{\partial f\left( x,t\right) }{\partial t}dx.\tag{4}$$
  • (ii) If the limits of integration are functions of $t$ $$I(t)=J(u(t),v(t),t)=\displaystyle\int_{u(t)}^{v(t)}f\left( x,t\right) dx,\tag{5}$$ then $$\dfrac{d}{dx}\displaystyle\int_{a}^{x}g\left( t\right) dt=g\left( x\right)\tag{6}$$ and $$\dfrac{d}{dx}\displaystyle\int_{x}^{b}g\left( t\right) dt=-\dfrac{d}{dx}\displaystyle\int_{b}^{x}g\left( t\right) dt=-g\left( x\right).\tag{7}$$ By the chain rule, we have $$I^{\prime }(t)=\dfrac{dI}{dt}=\dfrac{\partial J}{\partial t}\dfrac{dt}{dt}+\dfrac{\partial J}{\partial v}\dfrac{dv}{dt}+\dfrac{\partial J}{\partial u}\dfrac{du}{dt}\tag{8}$$ or

$$I^{\prime }(t)=\left( \dfrac{\partial }{\partial t}\displaystyle\int_{u}^{v}f\left( x,t\right) dx\right) \dfrac{dt}{dt}+\left( \dfrac{\partial }{\partial v}\displaystyle\int_{u}^{v}f\left( x,t\right) dx\right) \dfrac{dv\left( t\right) }{dt}+\left( \dfrac{\partial }{\partial u}\displaystyle\int_{u}^{v}f\left( x,t\right) dx\right) \dfrac{du\left( t\right) }{dt}.$$ $$\tag{9}$$ Hence $$I^{\prime }(t)=\displaystyle\int_{u\left( t\right) }^{v\left( t\right) }\dfrac{\partial f\left( x,t\right) }{\partial t}dx+f\left( v\left( t\right) ,t\right) v^{\prime }\left( t\right) -f\left( u\left( t\right) ,t\right) u^{\prime}\left( t\right)\tag{10},$$

which proves $(2)$ (with $x,t$ interchanged).

Added. Remark: for my convenience I changed the name of your functions $a(x),b(x)$. The functions $a(x),b(x)$ are the above functions $u(x),v(x)$.

Added 2. Example (here $t$ is the independant variable): If $$I(t)=\displaystyle\int_{2t}^{t^{2}}e^{tx}dx,$$

then $f\left( x,t\right) =e^{tx},u\left( t\right) =2t$ and $v\left( t\right) =t^{2}$. Hence $v^{\prime }\left( t\right) =2t, u^{\prime }\left( t\right) =2$, and $$\dfrac{\partial f\left( x,t\right) }{\partial t}=\dfrac{\partial }{\partial t}e^{tx}=xe^{tx}.$$

Thus, we get

$$I^{\prime }(t)=\displaystyle\int_{u\left( t\right) }^{v\left( t\right) }\dfrac{\partial f\left( x,t\right) }{\partial t}dx+f\left( v\left( t\right) ,t\right) v^{\prime }\left( t\right) -f\left( u\left( t\right) ,t\right) u^{\prime }\left( t\right),$$

$$I^{\prime}(t)=\displaystyle\int_{2t}^{t^{2}}xe^{tx}dx+2te^{t^{3}}-2e^{2t^{2}}=\dfrac{e^{t^{3}}\left( 3t^{3}-1\right) -e^{2t^{2}}\left( 4t^{2}-1\right)}{t^{2}}.$$

share|cite|improve this answer
The first displayed formula/definition is misleading precisely on the cause of the OP's trouble... The expression J(u,v,x) should be defined on its own or, a minima, replaced by J(u(x),v(x),x). This applies to the linked answer as well. – Did Aug 20 '11 at 10:50
Yes, well I would find this even more convincing if you used $J$ in your solution (possibly to write things like $\partial J/\partial\alpha$, $\partial J/\partial\beta$ and $\partial J/\partial z$). But I guess this amounts to a complete reformulation of your text to satisfy my whims... So you can just forget the whole thing. :-) – Did Aug 20 '11 at 12:58
@Américo: careful, German orthography is not so straightforward: there are Steinitz and Lipschitz, for – t.b. Aug 20 '11 at 13:14
Yes. But VfB Fortuna Chemnitz takes a t... Gottverdammt! – Did Aug 20 '11 at 13:16
@Américo: Come, come... If you or I wanted to never look stupid (sic) and blind (sic), unless we are one of those very few almost supernatural people, we should stop doing mathematics. Now. So... – Did Aug 20 '11 at 15:25

$$\frac{d}{dx}\int_{a(x)}^{b(x)} f(x,t)dt=f(x,b(x))\frac{d}{dx}b(x)-f(x,a(x))\frac{d}{dx}a(x)+\int_{a(x)}^{b(x)}\frac{\partial}{\partial{x}}f(x,t)dt,$$

where I have used Leibniz's Rule.

Note that if $a(x)$ and $b(x)$ are constants, then we have a special case of Leibniz's Rule.

share|cite|improve this answer
As @Theo said elsewhere, this is Leibniz and not Leibnitz, see – Did Aug 20 '11 at 13:01

There is an intermediate function of three variables involved, namely $$F(u,v,w):=\int_u^v f(w,t)\ dt\ .$$ One has $$F_u(u,v,w)=-f(w,u)\ ,\quad F_v(u,v,w)=f(w,v)\ ,\quad F_w(u,v,w)=\int_u^v f_w(w,t)\ dt\ ,$$ where the last formula is Leibniz' Rule "without extras".

When the variables $u$, $v$, $w$ become functions of $x$: $$u(x):=a(x)\ ,\quad v(x):=b(x)\ ,\quad w(x):=x\ ,$$ then the composition with $F$ defines a function $\phi(x):=F\bigl(a(x),b(x),x\bigr)$. In order to compute the derivative $\phi'$ we have to use the chain rule and obtain the formulas given in Nana's answer.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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