# What is the value of the integral$\int_{0}^{+\infty} \frac{1-\cos t}{t} \, e^{-t} \, \mathrm{d}t$?

I have a question about evaluating

$$\int_{0}^{\infty} \frac{1-\cos t}{t} \, e^{-t} \, \mathrm{d} t$$

Since $\lim_{t \to 0} \frac{1-\cos(t)}{t} =0$, we know that the integrand is integrable near $t=0$.

But when evaluating the integral, how do you deal with the $t$ in the denominator of the integrand?

• I hope the method is the simplest:) – Paul Jun 5 '14 at 13:21
• The fact that this question is so heavily upvoted is baffling... – tomasz Jun 8 '14 at 11:10

Consider $$\mathcal{I}(a)=\int_0^{\infty}\frac{1-\cos at}{t}e^{-t}dt.$$ Differentiating w.r.t. $a$, we find $$\mathcal{I}'(a)=\int_0^{\infty}\sin at\;e^{-t}dt=\frac{a}{1+a^2}.$$ Now integrating back and using that $\mathcal{I}(0)=0$ yields $$\mathcal{I}(a)=\int_0^a\frac{a\,da}{1+a^2}=\frac12\ln\left(1+a^2\right).$$ It remains to set $a=1$.

• You are too fast! :P I solved it by considering $I(a)=\int_0^{\infty} \frac{1-\cos t}{t}e^{-at}\,dt$ but I think both the approaches are similar so I didn't post it. :D – Pranav Arora Jun 5 '14 at 13:37
• @Paul You are welcome! – Start wearing purple Jun 5 '14 at 13:47
• Does this method have a name? I haven't seen an integral evaluated this way before. – Achal Jun 5 '14 at 20:37
• @Achal Differentiation under the integral sign. Sometimes people also refer to Feynman. – Start wearing purple Jun 5 '14 at 21:09
• What is the justification for moving the differentiation into the integral? – Brofessor Aug 26 '18 at 7:37

Assume that $a>1$.

Using the Maclaurin series of the cosine function, we get

\begin{align} \int_{0}^{\infty} \frac{1-\cos t}{t} \, e^{-at} \, dt &= -\int_{0}^{\infty} \sum_{n=1}^{\infty} \frac{(-1)^{n}t^{2n}}{(2n)!} \frac{e^{-at}}{t} \, dt \\ &= \sum_{n=1}^{\infty}\frac{(-1)^{n+1}}{(2n)!} \int_{0}^{\infty}t^{2n-1}e^{-at} \, dt \\ &=\sum_{n=1}^{\infty}\frac{(-1)^{n+1}}{(2n)!} \frac{(2n-1)!}{a^{2n}} \\ &= \frac{1}{2}\sum_{n=1}^{\infty} \frac{(-1)^{n+1}}{n} \left( \frac{1}{a^{2}} \right)^{n} \\ &=\frac{1}{2} \,\ln \left(1+ \frac{1}{a^{2}} \right) \, , \end{align} where interchanging the order of summation and integration is justified by Fubini's theorem.

Letting $a \to 1^{+}$, we get $$\int_{0}^{\infty} \frac{1-\cos t}{t} \, e^{-t} \, dt = \frac{\ln 2}{2}.$$

(See this question about justification for moving the limit inside the integral.)

• Brilliant! +1. Never crossed my mind to use this method. I tried to use Frullani's integral but I couldn't. – Tunk-Fey Jun 5 '14 at 15:13
• @Tunk-Fey: I have posted a solution using Frullani's integral. :) – Pranav Arora Jun 5 '14 at 17:36
• And as always, great solution Random Variable! :) – Pranav Arora Jun 5 '14 at 17:37

Consider the integral $$\int_0^\infty \frac{1-\cos t}{t} e^{-st} dt.$$ If we define $f(t)=1-\cos t$, this integral is the Laplace transform of $f(t)/t$. Moreover, we have the following identity for Laplace transforms:

$$\mathcal{L} \left\{ \frac{f(t)}{t} \right\}(s)=\int_s^\infty F(\sigma)d \sigma$$ where $F(\sigma)$ is the Laplace transform of $f(t)$, which in our case is known to be $$F(\sigma)=\frac{1}{\sigma}-\frac{\sigma}{\sigma^2+1}.$$ Thus we have $$\int_0^\infty \frac{1-\cos t}{t} e^{-st}=\int_s^\infty \left( \frac{1}{\sigma}-\frac{\sigma}{\sigma^2+1} \right) d \sigma= \log \sigma-\frac{1}{2} \log (\sigma^2+1) \big]^{\sigma=\infty}_{\sigma=s} .$$ Plugging in $s=1$ gives $$\int_0^\infty \frac{1-\cos t}{t} e^{-t} dt=\frac{1}{2} \log 2- \log 1=\frac{1}{2} \log 2 .$$

• We have a same approach. I didn't notice your answer because I used my tablet to type the answer. Sorry for that. +1. – Tunk-Fey Jun 5 '14 at 14:34

Another approach

Consider Laplace transform $$\mathcal{L}\left[f(t)\right]=F(s)=\int_0^\infty f(t)\ e^{-st}\ dt$$ and property of the unilateral Laplace transform $$\mathcal{L}\left[\frac{f(t)}{t}\right]=\int_s^\infty F(\omega)\ d\omega,$$ where $F(\omega)$ is Laplace transform of $f(t)$. We choose $f(t)=(1-\cos at)$ and it is easy to show that $$\mathcal{L}\left[1-\cos at\right]=F(s)=\int_0^\infty (1-\cos at)\ e^{-st}\ dt=\frac{a^2}{s(s^2+a^2)},$$ then \eqalign { \mathcal{L}\left[\frac{1-\cos at}{t}\right]&=\int_1^\infty \frac{a^2}{s(s^2+a^2)}\ ds\\ &=\int_1^\infty \left[\frac{1}{s}-\frac{s}{s^2+a^2}\right]\ ds\\ &=\left.\frac12\left[\ln s^2-\ln(s^2+a^2)\right]\right|_1^\infty\\ &=\left.\frac12\ln\left(\frac{s^2}{s^2+a^2}\right)\right|_1^\infty\\ &=\frac12\ln\left(1+a^2\right). } Taking $a=1$ yields $$\int_0^\infty \left[\frac{1-\cos t}{t}\right]\ e^{-t}\ dt=\large\color{blue}{\frac{\ln 2}{2}}.$$

Here's a solution using Frullani's integral.

$$\int_0^{\infty} \frac{1-\cos t}{t}e^{-t}\,dt=\Re\left(\int_0^{\infty} \frac{1-e^{it}}{t}e^{-t}\,dt\right)=\Re\left(\int_0^{\infty} \frac{e^{-t}-e^{-(1-i)t}}{t}\,dt\right)$$ $$=\Re\left(\ln\left(\frac{1-i}{1}\right)\right)=\Re\left(\ln\left(\sqrt{2}e^{i(-\pi/4+2k\pi)}\right)\right)=\boxed{\dfrac{\ln 2}{2}}$$

• Wait!? I got the result until $\Re[\ln(1-i)]$ but when I used W|A the answer is $1$. I forgot to use the way that you used. Darn! – Tunk-Fey Jun 5 '14 at 17:50
• @Tunk-Fey: But still, why does W|A gives 1? Any ideas? :/ – Pranav Arora Jun 5 '14 at 17:51
• I don't know!? Have you clicked the link? – Tunk-Fey Jun 5 '14 at 17:52
• @Tunk-Fey: Yes, I clicked it and it shows 1 as the answer. – Pranav Arora Jun 5 '14 at 17:52
• However, if I enter this: wolframalpha.com/input/… it shows the correct result. – Pranav Arora Jun 5 '14 at 17:54


\begin{align} &\color{#44f}{\large\int_{0}^{\infty}{1 - \cos\pars{t} \over t} \expo{-t}\,\dd t} =\Re\int_{0}^{\infty}\bracks{% \expo{-t} - \expo{-\pars{1 + \ic}t}}\int_{0}^{\infty}\expo{-t\xi}\,\dd\xi\,\dd t \\[3mm]&=\Re\int_{0}^{\infty}\int_{0}^{\infty}\bracks{% \expo{-\pars{1 + \xi}t} - \expo{-\pars{\xi + 1 + \ic}t}}\,\dd t\,\dd\xi =\Re\int_{0}^{\infty}\bracks{{1 \over \xi + 1} - {1 \over \xi + 1 + \ic}}\,\dd\xi \\[3mm]&=\Re\ln\pars{1 + \ic} = \color{#44f}{\large\half\,\ln\pars{2}} \end{align}