1
$\begingroup$

In this question, the OP asked to prove the Cosine Integral identity $-\int_x^\infty \frac{\cos(t)}{t}\,dt=\gamma+\log(x)+\int_0^x \frac{\cos(t)-1}{t}\,dt$, where $\gamma $ is the Euler-Mascheroni constant. However, the two posted answers are incomplete and seem unsatisfactory.


Using Complex Anaysis

One can show for $x>0$ that $-\int_x^\infty \frac{\cos(t)}{t}\,dt=\gamma+\log(x)+\int_0^x \frac{\cos(t)-1}{t}\,dt$ using contour integration. To wit, Cauchy's Integral Theorem guarantees that

$$\begin{align} 0&=\oint_C \frac{e^{iz}}{z}\,dz\\\\ &=\int_\epsilon^R \frac{e^{ix}}{x}\,dx +\int_0^{\pi/2}\frac{e^{iRe^{i\phi}}}{Re^{i\phi}}\,iRe^{i\phi}\,d\phi+\int_R^\epsilon \frac{e^{-x}}{ix}\,i\,dx+\int_{\pi/2}^0 \frac{e^{i\epsilon e^{i\phi}}}{\epsilon e^{i\phi}}\,i\epsilon e^{i\phi}\,d\phi\\\\ &=\int_\epsilon^R \frac{e^{ix}}{x}\,dx -\int_\epsilon^R \frac{e^{-x}}{x}\,dx-i\frac\pi2+O(\epsilon)+O\left(\frac1R\right)\tag1 \end{align}$$

whence after taking the real part of both sides of $(1$ and integrating by parts the integral $\int_\epsilon^R \frac{e^{-x}}{x}\,dx$ with $u=e^{-x}$ and $v=\log(x)$, we find

$$\begin{align}-\int_x^R \frac{\cos(x')}{x'}\,dx'&=\log(x)-\int_\epsilon^R e^{-x}\log(x)\,dx+\int_\epsilon^x\frac{\cos(x')-1}{x'}\,dx'\\\\ &-\log(\epsilon) -e^{-R}\log(R)+e^{-\epsilon}\log(\epsilon)+O\left(\frac1R\right)+O(\epsilon)\tag2 \end{align}$$

Letting $R\to\infty$ and $\epsilon\to 0$ in $(2)$ yields the sought relationship.


Using Real Analysis Alternatively, we can use either the Laplace Transform or "Feynman's Trick" to show that

$$\int_0^\infty \frac{\cos(x)-e^{-x}}{x}\,dx=\int_0^\infty \left(\frac{x}{x^2+1}-\frac1{x+1}\right)=0\tag3$$

Starting with $(3)$, is tantamount to starting with the real part of $(1)$ and we are done.


So, what are other ways to prove the coveted relationship using real analysis tools only?

$\endgroup$
0
$\begingroup$

Note that for each $x>0$, the improper integral $\int_{x}^{\infty}\frac{\cos(t)}{t}dt$ converges conditionally (in the sense that $\lim_{A\rightarrow\infty}\int_{x}^{A}\frac{\cos(t)}{t}dt$ exists but $\lim_{A\rightarrow\infty}\int_{x}^{A}\left|\frac{\cos(t)}{t}\right|dt=\infty$). Define $F:(0,\infty)\rightarrow\mathbb{R}$ by $F(x)=\int_{x}^{\infty}\frac{\cos(t)}{t}dt$. Note that we still have $F'(x)=-\frac{\cos(x)}{x}$. For, \begin{eqnarray*} F'(x) & = & \lim_{h\rightarrow0}\frac{\int_{x+h}^{\infty}\frac{\cos(t)}{t}dt-\int_{x}^{\infty}\frac{\cos(t)}{t}dt}{h}\\ & = & -\lim_{h\rightarrow0}\frac{\int_{x}^{x+h}\frac{\cos(t)}{t}dt}{h}\\ & = & -\frac{\cos(x)}{x}. \end{eqnarray*}

On the other hand, $0$ in the integral $\int_{0}^{x}\frac{\cos(t)-1}{t}dt$ is a removable singularity. For $t\neq0$, we have \begin{eqnarray*} \frac{\cos(t)-1}{t} & = & \frac{(1-\frac{1}{2!}t^{2}+\frac{1}{4!}t^{4}-\cdots)-1}{t}\\ & = & -\frac{1}{2!}t+\frac{1}{4!}t^{3}-\frac{1}{6!}t^{5}+\cdots. \end{eqnarray*} Define $\phi:\mathbb{R}\rightarrow\mathbb{R}$ by $\phi(t)=-\frac{1}{2!}t+\frac{1}{4!}t^{3}-\frac{1}{6!}t^{5}+\cdots$. It can be proved that the power series converges everywhere (using root-test), and hence $\phi$ is an analytic function. Therefore $\int_{0}^{x}\frac{\cos(t)-1}{t}dt=\int_{0}^{x}\phi(t)dt$.

Now $\frac{d}{dx}\int_{0}^{x}\frac{\cos(t)-1}{t}dt=\frac{\cos(x)-1}{x}$.

Finally, define $G:(0,\infty)\rightarrow\mathbb{R}$ by $$ G(x)=\int_{x}^{\infty}\frac{\cos(t)}{t}dt-\left[\ln(x)+\int_{0}^{x}\frac{\cos(t)-1}{t}dt\right]. $$ Then $G$ is differentiable and $G'(x)=-\frac{\cos(x)}{x}-\frac{1}{x}-\frac{\cos(x)-1}{x}=0$. This shows that $G$ is a constant function.

$\endgroup$
  • $\begingroup$ You have not evaluated the constant, which renders this development incomplete. But I appreciate your attempt. $\endgroup$ – Mark Viola Mar 16 at 16:49
  • $\begingroup$ I thought that you just asked to prove that there exists a constant $\gamma$ such that ... You did not ask explicitly "compute the value of $\gamma$". $\endgroup$ – Danny Pak-Keung Chan Mar 16 at 16:59
  • $\begingroup$ Read the question of the OP. "Real Analysis Methodologies to Prove $-\int_x^\infty \frac{\cos(t)}{t}\,dt=\gamma+\log(x)+\int_0^x \frac{\cos(t)-1}{t}\,dt$ for $x>0$." The constant $\gamma$ is explicitly written. How did you infer that I was interested in something considerably less than that. Furthermore, in the body of the question, I actually show two derivations that include $\gamma$. $\endgroup$ – Mark Viola Mar 16 at 17:08
  • $\begingroup$ I interpret that line as "Prove that there exists $\gamma$ such that ... for all $x>0$. $\endgroup$ – Danny Pak-Keung Chan Mar 16 at 17:12
  • 1
    $\begingroup$ You should state explicitly: Compute the value $\gamma$ and express it in close-form. Otherwise, I can write $\gamma = G(1) = \int_1^\infty \frac{\cos t}t dt - \int_0^1 \frac{\cos t -1}{t} dt$. $\endgroup$ – Danny Pak-Keung Chan Mar 16 at 17:17

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.