Let $f(t) = \int_0^t \ln{(s^2+t^2)} ds$, how can I find the derivative $\frac{df(t)}{dt}$?

The function $\,\int_0^t \ln{(s^2+t^2)} ds$ is defined to be continuous in $s^2+t^2 > 0$ and $ s^2+t^2 \ne 0$, then by the Fundamental Theorem of Integral Calculus, we have $$ f´(t) = (\int_0^t \ln{(s^2+t^2)} ds)´= f(t) - f(0) = \ln{(s^2+t^2)} + \ln{(s^2)}$$

Is that correct? can I apply the same procedure to this function $g(t) = \int_0^t \frac{e^{st}}{s} ds$ ? Please help. Thanks.


A non-orthodoxical approach.

Let $G(s,t)$ be a primitive of the integrand $g(s,t)$ (antiderivative on $s$): it is such that $\frac{\partial G}{\partial s}$ is the integrand.

The expression to be derived is thus $G(s,t)|_{s=0}^t=G(t,t)-G(0,t)$. When deriving with respect to $t$, we need to use the chain rule: $$\frac{dG(t,t)}{dt}=\frac{\partial G}{\partial s}(t,t)\frac{dt}{dt}+\frac{\partial G}{\partial t}(t,t)\frac{dt}{dt}=g(t,t)+\frac{\partial G}{\partial t}(t,t),$$ $$\frac{G(0,t)}{dt}=\frac{\partial G}{\partial s}(0,t)\frac{d0}{dt}+\frac{\partial G}{\partial t}(0,t)\frac{dt}{dt}=\frac{\partial G}{\partial t}(0,t).$$ We are still missing the expression of $\frac{\partial G}{\partial t}$. $G(s,t)$ being a primitive (on $s$) of $g(s,t)$, $\frac{\partial G}{\partial t}(s,t)$ must be a primitive (on $s$) of $\frac{\partial g}{\partial t}(s,t)$. So $$\frac{\partial G}{\partial t}(s,t)=\int \frac{2t}{s^2+t^2}ds=2\arctan\frac st+C.$$

Putting it all together, the answer is $$g(t,t)+2\arctan1-2\arctan0=\ln(2t^2)+\frac{\pi}2$$

For the second case, $$\frac{\partial G}{\partial t}(s,t)=\int{e^{st}}ds=\frac{e^{st}}t+C.$$

and the answer, $$\frac{e^{t^2}}t+\frac{e^{t^2}}t-\frac10.$$


That is not correct. The integrand is itself a function of $t$. And clearly, the answer should be independent of $s$.

Check here: http://en.wikipedia.org/wiki/Differentiation_under_the_integral_sign

Hence, $$f'(t)=\ln(t^2+t^2)+\int_0^t \frac{2t}{s^2+t^2}\,ds=\ln(2t^2)+\frac{\pi}{2}$$

Similarly, $$g'(t)=\frac{e^{t^2}}{t}+\int_0^t e^{st}\,ds=\frac{e^{t^2}}{t}+\frac{e^{t^2}-1}{t}=\frac{2e^{t^2}-1}{t}$$

  • 2
    $\begingroup$ Your last term should be $\pi/2$ since $s=t$ as you wrote it just before. $\endgroup$ – Claude Leibovici Jun 22 '14 at 7:47
  • $\begingroup$ @ClaudeLeibovici: Oops, thanks! :P $\endgroup$ – Pranav Arora Jun 22 '14 at 7:49
  • $\begingroup$ Sorry for my dumbs questions but you integrate directly f ? or applied the Fundamental Theorem, really thanx for your patient .... $\endgroup$ – Rachel Jun 22 '14 at 7:52
  • $\begingroup$ @Knight: I applied the Fundamental theorem of Calculus. Click the link I posted. And the integral is very common. What is the derivative of arctan(x)? $\endgroup$ – Pranav Arora Jun 22 '14 at 7:55
  • $\begingroup$ ok, really thanx. $\endgroup$ – Rachel Jun 22 '14 at 7:57

Hint: using the Leibniz integral rule,

$$\begin{align} \frac{d}{dt}f(t)&=\ln{(s^2+t^2)}\bigg{|}_{s=t}\frac{d}{dt}(t)-\ln{(s^2+t^2)}\bigg{|}_{s=0}\frac{d}{dt}(0)+\int_{0}^{t}\frac{\partial}{\partial t}\ln{(s^2+t^2)}\,\mathrm{d}s \end{align}$$


The integration can be carried out explicitly:

$$\,\int_0^t \ln{(s^2+t^2)} ds=(1/2)t(-4 + \pi + 2\ln 2 + 4\ln t)=:f(t)$$


$$f'(t)=\pi/2 + \ln(2t^2)$$

And $\int_0^t \frac{e^{st}}{s} ds$ does not converge (at $s= 0^+$).


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.