14

Hi I am trying to find out for what values of the real parameter does the integral $$ I=\int_0^\infty \frac{\sin x}{x^s}dx $$ (a) convergent and (b) absolutely convergent.

I know that the integral is convergent if $s=1$ since $$ \int_0^\infty \frac{\sin x}{x}dx=\frac{\pi}{2}. $$ For $s=0$ it is easy to see divergent integral since $\int_0^\infty \sin x\, dx$ is divergent. However I am stuck on figuring out when it is convergent AND or absolutely convergent.

I know to check for absolute convergence I can determine for an arbitrary series $\sum_{n=0}^\infty a_n$ by considering $$ \sum_{n=0}^\infty |a_n|. $$ If it helps also $$\sin x=\sum_{n=0}^\infty \frac{(-1)^{2n+1}}{(2n+1)!} {x^{2n+1}}$$. Thank you all

5 Answers5

14

This is a good problem to analyze. We can solve it by just series methods and careful thought.

Given the following integral \begin{equation} \int_{0}^\infty \frac{\sin x}{x^s}dx, \tag1 \end{equation} for what values of the real parameter s is the integral convergent and absolutely convergent.

(a) In order to solve this problem we break (1) into two pieces \begin{equation} \int_{0}^\infty \frac{\sin x}{x^s}dx=\int_{0}^1 \frac{\sin x}{x^s}dx + \int_{1}^\infty \frac{\sin x}{x^s}dx \tag2 \end{equation} We can analyze each term separately. It is easy to see that the term $$ \int_{1}^\infty \frac{\sin x}{x^s}dx $$ is divergent for $ s \leq 0$ since integral is proportional to $x^s$ which diverges as $x \to \infty$. For $ s > 0$, the series is convergent since $x^{-s} \downarrow 0 \ \text{as}\ x \to \infty$. We now consider the other term in (2) and write it explicitly in terms of a sum $$ \int_{0}^1 \frac{\sin x}{x^s}dx=\int_{0}^1 \sum_{n=0}^{\infty}\frac{(-1)^n x^{2n+1}}{(2n+1)!}{x^{-s}}dx= \int_{0}^{1}\sum_{n=0}^{\infty}\frac{(-1)^n x^{2n+1-s}}{(2n+1)!}dx. $$ We can evaluate if this integral is convergent by analyzing the series inside which is \begin{equation} \sum_{n=0}^{\infty}\frac{(-1)^n x^{2n+1-s}}{(2n+1)!}\equiv \xi \end{equation} Using the ratio test on $\xi$, we have $$ \lim_{n\to \infty}\bigg| \frac{(-1)^{n+1} x^{2n+3-s} \cdot (2n+1)}{(2n+3)! \cdot (-1)^n x^{2n+1-s}} \bigg|=\lim_{n\to \infty} \frac{x^2}{4n^2+10n+6}=0. $$ By the definition of the ratio test, this series is absolutely convergent since $$ \lim_{n\to \infty} \bigg|\frac{\xi_{n+1}}{\xi_n}\bigg| =0 <1. $$ We now check for uniform convergence by swapping the order of summation and integration, that is doing the integral first which yields $$ \sum_{n=0}^{\infty} \frac{(-1)^n} {(2n+1)!}\int_{0}^{1} x^{2n+1-s} dx=\sum_{n=0}^{\infty} \frac{(-1)^n} {(2n+1)! \cdot (2n+2-s)}. $$ Note, the $(2n+2-s) >0$ to be defined. Computing the sum for $n=0$ we have the condition $2 -s > 0$, or $ 2>s$. Evaluating the integral at $n=0, s=2$ we have $$ \int_{0}^{1} x^{2n+1-s} dx=\int_{0}^{1} {x^{-1}} dx $$ which diverges as the logarithm.

We can conclude that (1) is convergent for $s \in (0,2)$.

(b):For absolute convergence we check the convergence of $$ \int_{0}^{\infty} \bigg|\frac{\sin x}{x^s}\bigg| dx. $$ Once again, we break the integral into two parts $$ \int_{0}^{\infty} \bigg|\frac{\sin x}{x^s}\bigg| dx=\int_{0}^{1} \bigg|\frac{\sin x}{x^s}\bigg| dx + \int_{1}^{\infty} \bigg|\frac{\sin x}{x^s}\bigg| dx. $$ The second term on the right converges for $s > 1$ and is seen easily since $$ \int_{1}^{\infty} \bigg|\frac{\sin x}{x^s}\bigg| dx < \int_{1}^{\infty} \bigg|\frac{1}{x^s}\bigg| dx $$ which is convergent for $s > 1$. We check the other term for convergence by noting that $$ \bigg|\frac{\sin x}{x^s}\bigg|=\frac{\sin x}{x^s} $$ for $ x \in [0,1]$. Thus we conclude that $$ \int_0^1 \frac{\sin x}{x^s} $$ is absolutely convergent for $s \in (0,2)$.

Therefore, the integral in (1) is absolutely convergent for $s \in (1,2)$.

Hakim
  • 10,213
Jeff Faraci
  • 9,878
  • Hi is there a reason for the down vote? Thank you. – Jeff Faraci May 13 '14 at 19:45
  • Thank you!!!! I would have missed the evaluating the integral at n=0,s=2. I did not down vote though. Check as answer and +1 – user143444 May 13 '14 at 19:52
  • Yes that is a subtle point. Glad it helped. – Jeff Faraci May 13 '14 at 19:57
  • 2
    2 downvoters? It is quite interesting you down vote the correct solution. Thank you – Jeff Faraci May 13 '14 at 20:14
  • Nice answer - literally (+1) – Ron Gordon May 14 '14 at 18:12
  • @RonGordon Thank you sir. – Jeff Faraci May 15 '14 at 04:00
  • @Integrals Am I misunderstanding how you're saying the series $x^{-s}$ is convergent since $x^{-s} \to 0$ as $x \to \infty$ to mean $\int_{1}^\infty x^{-s} ds$ converges? If so This simply isn't true for every $s>0$, for example $s=1$. Oop just reread, if you're saying that $\int_1^\infty \frac{sin(x)}{x^{s}}$ converges for $s > 0$ then I think I can see what you're saying (I would need to actually work it out to verify this however). – DanZimm Jul 04 '14 at 02:54
  • Also a few nit picks, you say "if $s\le0$ then the integral is proportional to $x^s$" when $x^s = \frac{1}{x^r}$ where $r=-s$ so that $r > 0$ (and a similar issue for the following statement). – DanZimm Jul 04 '14 at 02:56
6

$\newcommand{\+}{^{\dagger}} \newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\down}{\downarrow} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\isdiv}{\,\left.\right\vert\,} \newcommand{\ket}[1]{\left\vert #1\right\rangle} \newcommand{\ol}[1]{\overline{#1}} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large A}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert} \newcommand{\wt}[1]{\widetilde{#1}}$ $\ds{I \equiv \int_{0}^{\infty}{\sin\pars{x} \over x^{s}}\,\dd x:\ {\large ?}}$

\begin{align} I&=\lim_{\epsilon \to 0^{+}}\int_{\epsilon}^{\infty} {1 \over x^{s - 1}}{\sin\pars{x} \over x}\,\dd x =\lim_{\epsilon \to 0^{+}}\int_{\epsilon}^{\infty}{1 \over x^{s - 1}}\bracks{% \half\Re\int_{-1}^{1}\expo{\ic\verts{k}x}\,\dd k}\,\dd x \\[3mm]&=\half\Re\int_{-1}^{1}\bracks{\color{blue}{% \lim_{\epsilon \to 0^{+}}\int_{\epsilon}^{\infty} {\expo{\ic\verts{k}x} \over x^{s - 1}}\,\dd x}}\,\dd k\tag{1} \end{align}

\begin{align} &\overbrace{\color{blue}{% \lim_{\epsilon \to 0^{+}}\int_{\epsilon}^{\infty}{\expo{\ic\verts{k}x} \over x^{s - 1}} \,\dd x}}^{\ds{\ic\verts{k}x = -t\ \imp\ x = {\ic \over \verts{k}}\,t}} =\lim_{\epsilon \to 0^{+}}\int_{-\epsilon\ic}^{-\infty\ic}\pars{\expo{\ic\pi/2}t \over \verts{k}}^{1 - s} \expo{-t}\,{\ic \over \verts{k}}\,\dd t \\[3mm]&=-\,{\expo{-\pi s\ic/2} \over \verts{k}^{2 - s}} \lim_{\epsilon \to 0^{+}}\int_{-\epsilon\ic}^{-\infty\ic}t^{1 - s}\expo{-t}\,\dd t \\[3mm]&=-\,{\expo{-\pi s\ic/2} \over \verts{k}^{2 - s}}\times \\[3mm]&\lim_{\epsilon \to 0^{+}}\bracks{% -\int^{\epsilon}_{\infty}t^{1 - s}\expo{-t}\,\dd t -\lim_{R \to \infty}\int_{-\pi/2}^{0} R^{1 - s}\expo{\ic\pars{1 - s}\theta}\exp\pars{-R\expo{\ic\theta}}R \expo{\ic\theta}\ic\,\dd\theta}\qquad\pars{2} \end{align}

$$\!\!\!\!\!\!\!\!\!\!\!\!\!\!\! \mbox{When}\quad\epsilon \to 0^{+},\ \mbox{the first integral converges when}\ \Re\pars{1 - s} > -1\ \imp\ \Re\pars{s} < 2\tag{3} $$

Let's study the second integral in the limit $\ds{R \to \infty}$: \begin{align} &\verts{\int_{-\pi/2}^{0} R^{1 - s}\expo{\ic\pars{1 - s}\theta}\exp\pars{-R\expo{\ic\theta}}R \expo{\ic\theta}\ic\,\dd\theta} \leq R^{2 - s}\int_{-\pi/2}^{0}\exp\pars{-R\cos\pars{\theta}}\,\dd\theta \\[3mm]&=R^{2 - s}\int_{0}^{\pi/2}\exp\pars{-R\sin\pars{\theta}}\,\dd\theta <R^{2 - s}\int_{0}^{\pi/2}\exp\pars{-R\,{2\theta \over \pi}}\,\dd\theta \\[3mm]&={\pi \over 2}\pars{R^{1 - s} - R^{1 - s}\expo{-R}} \to 0\ \mbox{when}\ \Re\pars{1 - s} < 0\ \imp\ \Re\pars{s} > 1\tag{4} \end{align}

$\pars{3}$ and $\pars{4}$ show that both terms in $\pars{2}$ converge whenever $\ds{1 < \Re\pars{s} < 2}$: $$ \color{blue}{\lim_{\epsilon \to 0^{+}} \int_{\epsilon}^{\infty}{\expo{\ic\verts{k}x} \over x^{s - 1}}\,\dd x} =-\,{\expo{-\pi s\ic/2} \over \verts{k}^{2 - s}}\,\Gamma\pars{2 - s}\,,\qquad\qquad 1 < \Re\pars{s} < 2 $$ where $\ds{\Gamma\pars{z}}$ is the Gamma Function. This result is replaced in $\pars{1}$ to find: \begin{align} I&=-\,\half\,\cos\pars{\pi s \over 2}\Gamma\pars{2 - s} \int_{-1}^{1}\verts{k}^{s - 2}\,\dd k =-\,\half\,\cos\pars{\pi s \over 2}\Gamma\pars{2 - s}\,{2 \over s - 1} \end{align}

$$\color{#00f}{\large% I = \lim_{\epsilon \to 0^{+}}\int_{\epsilon}^{\infty} {\sin\pars{x} \over x^{s}}\,\dd x = \cos\pars{\pi s \over 2}\Gamma\pars{1 - s}}\,,\qquad 1 < \Re\pars{s} < 2 $$ where we used the Gamma Recurrence Formula ${\bf\mbox{6.1.15}}$.

Felix Marin
  • 89,464
  • This one is quite similar. Above I included the convergence along an arc ( Jordan Lemma ) which yields $\large\Re\left(s\right) > 1$ as one of the conditions. – Felix Marin May 14 '14 at 18:43
  • @WillieWong Thanks a lot. You're very kind. – Felix Marin May 15 '14 at 08:08
  • @FelixMarin: since you are around, I hope you don't mind that I edited your first comment above to remove the reference to the erasure. I find it very worthwhile to keep your comment pointing out the other answer. – Willie Wong May 15 '14 at 08:13
0

$$\varphi_1(\alpha) =\int_0^\infty \frac{\sin t}{t^\alpha}\,dt\tag{I}$$

case $\alpha\gt 0$

Near $t=0$, $\sin t\approx t.$ Which yields, $\frac{\sin t}{t^{\alpha}}\approx \frac{1}{t^{\alpha -1}}$ and the convergence of the integral in (I) holds nearby $t=0$ if and only if $\alpha<2 $.

Now let take into play the case where $t $ is large.

case $\alpha\leq 0$

Employing integration by part, \begin{eqnarray*} \Big| \int_{\frac{\pi}{2}}^\infty \frac{\sin t}{t^\alpha}\,dt\Big| &= & \Big| -\alpha \int_{\frac{\pi}{2}}^\infty \frac{\cos t}{t^{\alpha+1}}\,dt\Big|\\ % &\leq & \alpha \int_{\frac{\pi}{2}}^\infty \frac{ 1 }{t^{\alpha+1}}\,dt< \infty \qquad\text{since} \qquad \alpha +1>1~~\text{with} ~~\alpha >0. \end{eqnarray*} Thus for $\alpha>0 $

$\varphi_1(\alpha)$ exists if and only if $0<\alpha<2$.

We will later these are the only values of $\alpha$ which guarantee the existence of $\varphi_1$. For now let have a look on the integrability of functions under (I). In other to see that, one can quickly check the following

$$ \mathbb{R}_+ = \bigcup_{n\in\mathbb{N}} [n\pi, (n+1)\pi).$$

Then, $$\int_0^\infty \frac{|\sin t|}{t^\alpha}\,dt = \int_{0}^{\pi} \frac{\sin t}{{t}^\alpha} \,dt+ \sum_{n=1}^{\infty} \int_{n\pi}^{(n+1)\pi} \frac{|\sin t|}{t^\alpha}\,dt \\:= \int_{0}^{\pi} \frac{\sin t}{{t}^\alpha} \,dt+\sum_{n=1}^{\infty} a_n$$

With suitable change of variable ($u = t-n\pi$) we get

\begin{eqnarray*} a_n &=& \int_{0}^{\pi} \frac{\sin t}{{(t+n\pi)}^\alpha} \,dt\qquad\text{since } \sin(t+n\pi)= (-1)^n\sin t \end{eqnarray*} On the oder hand, it is also easy to check

\begin{eqnarray} \frac{2}{(n+1\pi)^\alpha} \leq a_n \leq \frac{2}{(n\pi)^\alpha}. % \end{eqnarray} These inequality together with the Riemann sums show that the series of general terms $(a_n)_n$ and $(b_n)_n$ converge if and only if $\alpha>1.$ Moreover we have seen from the foregoing that

$$\int_{0}^{\pi} \frac{\sin t}{{t}^\alpha} \,dt$$ converges only for $\alpha <2$

Taking profite of the tricks above, we get the result for the case $\alpha \leq 0$ as follows

$$\int_0^\infty \frac{\sin t}{t^\alpha}\,dt = \int_{0}^{\pi} \frac{\sin t}{{t}^\alpha} \,dt+ \sum_{n=1}^{\infty} \int_{n\pi}^{(n+1)\pi} \frac{\sin t}{t^\alpha}\,dt \\:= \int_{0}^{\pi} \frac{\sin t}{{t}^\alpha} \,dt+\sum_{n=1}^{\infty} a'_n $$

With

\begin{eqnarray*} |a'_n| &=&\left|\int_{n\pi}^{(n+1)\pi} \frac{\sin t}{{(t+n\pi)}^\alpha} \,dt\right|= \left|\int_{0}^{\pi} \frac{\sin t}{{(t+n\pi)}^\alpha} \,dt\right| \geq \frac{2}{(\pi+n\pi)^\alpha} \qquad\qquad\text{since } \sin(t+n\pi) = (-1)^n\sin t . \end{eqnarray*}
and the equalities hold in both cases when $\alpha = 0.$ Therefore, $$\lim |a'_n|= \begin{cases} 2 &~~if ~~\alpha = 0 \nonumber\\ \infty & ~~if ~~\alpha <0. \nonumber \end{cases}$$ What prove that the divergence of the series $\sum\limits_{n=0}^{\infty} a'_n$ since $a_n'\not\to 0$. Consequently the left hand side of the previous relations always diverge since $\int_{0}^{\pi} \frac{\sin t}{{t}^\alpha} \,dt $ converges for $\alpha\leq 0.$

Conclusion$ \frac{\sin t}{t^\alpha} $ converges for $0<\alpha<2$ and converges absolutely for $1<\alpha <2$.

Guy Fsone
  • 23,903
0

Note that when x close to $0$ the integrand behaves as

$$ \frac{x}{x^s} .$$

On the other hand at infinity behaves as

$$ \frac{1}{x^s} .$$

Now check the integrability of the above funtions and see the conditions on $s$.

  • 2
    -1 For not providing a solution, please post as a comment. – user143444 May 13 '14 at 19:50
  • 2
    @Gregory: This is the main idea (in fact the solution ) and what's left is for the OP to do. Gettingdetailed answers is not a good strategy for learning. – Mhenni Benghorbal May 13 '14 at 20:02
  • 2
    This is not the main idea however. You miss a lot of key points. Please post as a comment next time thanks :) Greg – user143444 May 13 '14 at 20:03
  • @Gregory: yes it is! There are no other key points! – Mhenni Benghorbal May 13 '14 at 20:05
  • No quite a lot of detail is lacking from your solution. This is a good comment though. -1 for solution. +1 for comment. Thanks :) Greg – user143444 May 13 '14 at 20:06
  • @Gregory: have a good day? – Mhenni Benghorbal May 13 '14 at 20:09
  • 2
    Even as a hint, this is false. "On the other hand at infinity behaves as 1/x^s" squarely misses the point and yields the (wrong) answer that s should be in (1,2) although the integral converges for every s in (0,2). @upvoters Can you explain? – Did May 14 '14 at 06:46
  • @Did: It is a correct answer! Why just don't you take care of your answers? – Mhenni Benghorbal May 14 '14 at 06:51
  • @moderatos: This answer has been up voted four times and some people came and down voted intentionally for four times? It is a perfect answer and gives the key for finding the right $s$ with simplisity. – Mhenni Benghorbal May 14 '14 at 06:53
  • 7
    For the nth time, Mhenni, this is how the site is meant to work: everybody nosing in everybody's posts (if this unsettles you that your answers are read and critiqued, don't post). About the four upvotes: I agree, these upvotes are amazing seeing the defective mathematical content of this answer--but once again this is how the site works. – Did May 14 '14 at 07:23
  • @moderators: where are things heading? – Mhenni Benghorbal May 14 '14 at 22:55
  • 1
    I think this might make a nice hint, except for the fact that it really does miss quite a few of the technicalities (which are non-trivial!) of this problem. For $0 < s < 1$, the integral does converge, though not absolutely - but the asymptotic analysis in this answer isn't fine enough to get that. (I did not vote on this answer). –  May 15 '14 at 21:34
  • @T.Bongers: As I said before in my previous comments it has the idea and the crucial steps and gives the desired result. If you feel it misses some details you can work them out. The OP should do some work. By the way mathematians know the value of this answer. Have a good day and thanks for the comment. – Mhenni Benghorbal May 15 '14 at 23:21
  • 1
    @MhenniBenghorbal The basic asymptotic analysis that the integrand is $\sim x^{-s}$ is very insufficient for studying convergence (rather than absolute convergence) of the integral, though - I believe this is why your answer has received such a negative response: It really doesn't provide the idea or the crucial steps, except for the study of absolute convergence. –  May 16 '14 at 00:47
  • 1
    @T.Bongers Even absolute convergence would need an additional argument since the "peaks" of the oscillating function could become narrower and narrower and save the day even when $1/x^s$ alone is not integrable (this does not happen with sine, naturally, but one needs to show it does not). – Did May 17 '14 at 08:26
-2

It is absolutely convergent for $1<s<2$. First, write the integral as

$\int_{0}^{\infty} \left|{\frac{\sin(x)}{x^s}}\right| \;dx = \int_{0}^{1} \left|\frac{\sin(x)}{x^s}\right| \;dx + \int_{1}^{\infty} \left|\frac{\sin(x)}{x^s} \right|\;dx$.

Then $ \int_{1}^{\infty} \left|\frac{\sin(x)}{x^s} \right|\;dx$ converges for any $s>1$ since $\int_{1}^{\infty} \left|\frac{1}{x^s} \right|\;dx$ converges for such $s$ (using the obvious fact that $|\sin(x)| \leq 1$). Moreover, this integral diverges for $s\leq 1$.

For the other summand, recall that $\frac{\sin(x)}{x} < 1$ on $(0,1]$, so we bound the integrand:

$\left|\frac{sin(x)}{x^s}\right| \leq \frac{1}{x^{s-1}}$

for $x\in (0,1].$

It is well known that the integral $\int_0^1 \frac{1}{x^p}\;dx$ converges for $0<p<1$, so $\int_0^1 \frac{1}{x^{s-1}}\;dx$ converges for $1<s<2$. It remains to show that this integral diverges for $s\geq 2$, but this could be accomplished by using your taylor series for $\sin(x)$.