Never seen all of these together

Calculus Level 5

0 ( 1 1 + 4 x 2 cos ( 2 x ) ) d x x \large \int_0^{\infty} \left( \frac{1}{1 + 4x^2} - \cos(2x) \right) \frac{dx}{x}

If the value of the above integral is I I . Evaluate 100 I \displaystyle \lfloor 100 I \rfloor .


The answer is 57.

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

2 solutions

Ishan Singh
Jun 18, 2016

Proposition : 0 ( cos x 1 1 + x k ) d x x = γ ; k > 0 \int_{0}^{\infty}\left( \cos x - \dfrac{1}{1+x^k} \right) \dfrac{\mathrm{d}x}{x} = -\gamma \quad ; \quad k>0

Proof : J = 0 ( cos x 1 1 + x k ) d x x = 0 0 cos x e a x e a x 1 + x k d x d a \text{J} = \displaystyle \int_{0}^{\infty}\left( \cos x - \dfrac{1}{1+x^k} \right) \dfrac{\mathrm{d}x}{x} = \int_{0}^{\infty} \int_{0}^{\infty} \cos x \cdot e^{-ax} - \dfrac{e^{-ax}}{1+x^k} \mathrm{d}x \ \mathrm{d}a

= 0 a a 2 + 1 d a 0 0 e a x 1 + x k d x d a \displaystyle = \int_{0}^{\infty} \dfrac{a}{a^2+1} \mathrm{d}a -\int_{0}^{\infty}\int_{0}^{\infty} \dfrac{e^{-ax}}{1+x^k}\mathrm{d}x \ \mathrm{d}a

Substitute a x x ax \mapsto x

J = 0 a a 2 + 1 d a 0 0 a k 1 e x x k + a k d x d a \displaystyle \implies \text{J} = \int_{0}^{\infty} \dfrac{a}{a^2+1} \mathrm{d}a -\int_{0}^{\infty}\int_{0}^{\infty} \dfrac{a^{k-1}e^{-x}}{x^k+a^k}\mathrm{d}x \ \mathrm{d}a

Since 0 e x d x = 1 \displaystyle \int_{0}^{\infty} e^{-x}\mathrm{d}x = 1

J = 0 0 a e x a 2 + 1 a k 1 e x x k + a k d x d a \displaystyle \implies \text{J} = \int_{0}^{\infty} \int_{0}^{\infty} \dfrac{ae^{-x}}{a^2+1} - \dfrac{a^{k-1}e^{-x}}{x^k+a^k}\mathrm{d}x \ \mathrm{d}a

= 0 e x 0 a a 2 + 1 a k 1 x k + a k d a d x \displaystyle = \int_{0}^{\infty} e^{-x} \int_{0}^{\infty} \dfrac{a}{a^2+1} - \dfrac{a^{k-1}}{x^k+a^k}\mathrm{d}a \ \mathrm{d}x

= 0 e x [ ln ( a 2 + 1 ( a k + x k ) 1 k ) ] a = 0 a d x \displaystyle = \int_{0}^{\infty} e^{-x} \left[\ln\left(\dfrac{\sqrt{a^2+1}}{{(a^k+x^k)}^{\frac{1}{k}}} \right)\right]_{a=0}^{a \to \infty} \mathrm{d}x

= 0 e x ln x d x \displaystyle =\int_{0}^{\infty} e^{-x} \ln x \ \mathrm{d}x

= γ \displaystyle = -\gamma \quad \square

Now,

I = 0 ( 1 1 + 4 x 2 cos ( 2 x ) ) d x x \displaystyle \text{I} = \int_{0}^{\infty} \left( \dfrac{1}{1+4x^2} - \cos(2x) \right)\dfrac{\mathrm{d}x}{x}

Substitute 2 x x 2x \mapsto x

I = 0 ( 1 1 + x 2 cos ( x ) ) d x x = γ \displaystyle \implies \text{I} = \int_{0}^{\infty} \left( \dfrac{1}{1+x^2} - \cos(x) \right) \dfrac{\mathrm{d}x}{x} = \boxed{\gamma}

Mark Hennings
Dec 9, 2015

This is (almost) a standard integral. See Gradshteyn & Ryzhik, 3.783, 2. The integral is equal to γ \gamma , the Euler constant.

Could you explain how the integral can be said to converge?

D G - 5 years, 6 months ago

Log in to reply

The function ( 1 + 4 x 2 ) 1 cos 2 x (1+4x^2)^{-1} - \cos 2x vanishes at x = 0 x=0 and has a well -defined Maclaurin series, so the whole integrand is continuous. There is therefore no problem integrating all the way down to 0 0 .

The function x 1 ( 1 + 4 x 2 ) 1 x^{-1}(1+4x^2)^{-1} can be integrated up to \infty . The only problem is integrating cos 2 x x \frac{\cos 2x}{x} out to \infty . But this is a standard improper Riemann integral.

On the other hand, it does not exist as a Lebesgue integral.

Mark Hennings - 5 years, 6 months ago

@Sudeep Salgia , do you have an elementary solution?

Pi Han Goh - 5 years, 6 months ago

Log in to reply

It all depends on how elementary you want to go! I will assume the following formula for the Euler-Mascheroni constant: γ = 0 e x ln x d x . \gamma \; = \; -\int_0^\infty e^{-x}\, \ln x\,dx \;. Define functions a , b , c a,b,c on ( 0 , ) (0,\infty) by a ( x ) = ln x b ( x ) = 0 x cos t 1 t d t c ( x ) = x cos t t d t . a(x) \; = \; \ln x \qquad b(x) \; = \; \int_0^x \frac{\cos t - 1}{t}\,dt \qquad c(x) \; = \; \int_x^\infty \frac{\cos t}{t}\,dt \;. Taking Laplace transforms of each of these functions ( p > 0 p > 0 throughout): ( L a ) ( p ) = 0 ln x e p x d x = 0 ln x p e x 1 p d x = 1 p 0 ( ln x ln p ) e x d x = γ p 1 1 p ln p ( L b ) ( p ) = 0 ( 0 x cos t 1 t d t ) e p x d x = 0 d t t cos t 1 t e p x d x = 1 p 0 cos t 1 t e p t d t = 1 p 0 ( 0 1 sin s t d s ) e p t d t = 1 p 0 1 d s 0 e p t sin s t d t = 1 p 0 1 s p 2 + s 2 d s = [ 1 2 p ln ( p 2 + s 2 ) ] 0 1 = 1 2 p ln ( p 2 + 1 ) + 1 p ln p . \begin{array}{rcl} (La)(p) & = & \displaystyle \int_0^\infty \ln x e^{-px}\,dx \; = \; \int_0^\infty \ln\tfrac{x}{p} e^{-x} \,\tfrac{1}{p}dx \\ & = & \tfrac{1}{p} \displaystyle\int_0^\infty (\ln x - \ln p) e^{-x}\,dx \; = \; -\gamma p^{-1} - \tfrac{1}{p}\ln p \\ (Lb)(p) & = & \displaystyle \int_0^\infty\left(\int_0^x \frac{\cos t - 1}{t}\,dt\right) e^{-px}\,dx\; = \; \int_0^\infty dt \int_t^\infty \frac{\cos t - 1}{t} e^{-px}\,dx \\ & = & \displaystyle \tfrac{1}{p} \int_0^\infty \frac{\cos t - 1}{t}e^{-pt}\,dt \; = \; -\tfrac{1}{p}\int_0^\infty \left(\int_0^1 \sin st\,ds\right) e^{-pt}\,dt \\ & = & \displaystyle -\tfrac{1}{p} \int_0^1 ds \int_0^\infty e^{-pt}\sin st\,dt \; = \; -\tfrac{1}{p}\int_0^1 \frac{s}{p^2 + s^2}\,ds \\ & = & \displaystyle \Big[-\tfrac{1}{2p}\ln(p^2 + s^2)\Big]_0^1 \; = \; -\tfrac{1}{2p}\ln(p^2 + 1) + \tfrac{1}{p}\ln p \;. \end{array} Since d d p 0 cos t t ( 1 e p t ) d t = 0 e p t cos t d t = p p 2 + 1 \tfrac{d}{dp}\int_0^\infty \frac{\cos t}{t} (1 - e^{-pt})\,dt \; = \; \int_0^\infty e^{-pt}\cos t\,dt \; = \; \tfrac{p}{p^2+1} we see that ( L c ) ( p ) = 0 ( x cos t t d t ) e p x d x = 0 d t 0 t cos t t e p x d x = 1 p 0 cos t t ( 1 e p t ) d t = 1 2 p ln ( p 2 + 1 ) . \begin{array}{rcl} (Lc)(p) & = & \displaystyle\int_0^\infty \left( \int_x^\infty \frac{\cos t}{t}\,dt \right) e^{-px}\,dx \; = \; \int_0^\infty dt \int_0^t \frac{\cos t}{t} e^{-px}\,dx \\ & = & \displaystyle \tfrac{1}{p}\int_0^\infty \frac{\cos t}{t}(1 - e^{-pt})\,dt \; = \; \tfrac{1}{2p}\ln(p^2+1) \;. \end{array} Putting these results together yields ( L a ) ( p ) + ( L b ) ( p ) + ( L c ) ( p ) + γ p 1 0 (La)(p) + (Lb)(p) + (Lc)(p) + \gamma p^{-1} \; \equiv \; 0 and hence that a ( x ) + b ( x ) + c ( x ) + γ 0 a(x) + b(x) + c(x) + \gamma \; \equiv \; 0 This improves an important result about the Cosine Integral function C i ( x ) = x cos t t d t = c ( x ) = a ( x ) + b ( x ) + γ = γ + ln x + 0 x cos t 1 t d t Ci(x) \; = \; -\int_x^\infty \frac{\cos t}{t}\,dt \; = \; -c(x) \; =\; a(x) + b(x) + \gamma \; = \; \gamma + \ln x + \int_0^x \frac{\cos t - 1}{t}\,dt This is the equation we need. This formula is frequently stated without proof; the above gives a sketch of why it is true (there is a lot of integration theory that is required to justify all the changes of integral order, and the differentiation within the integrals that I have done!).

We can now solve the problem! Since (by partial fractions) ϵ N ( 1 1 + 4 x 2 cos ( 2 x ) ) d x x = ϵ N ( 1 cos ( 2 x ) x 4 x 1 + 4 x 2 ) d x \int_\epsilon^N \left(\frac{1}{1+4x^2} - \cos(2x)\right)\frac{dx}{x} \; = \; \int_\epsilon ^N \left(\frac{1 - \cos(2x)}{x} - \frac{4x}{1 + 4x^2}\right)\,dx we see that 0 N ( 1 1 + 4 x 2 cos ( 2 x ) ) d x x = 0 2 N 1 cos x x d x [ 1 2 ln ( 1 + 4 x 2 ) ] 0 N = γ + ln ( 2 N ) C i ( 2 N ) 1 2 ln ( 1 + 4 N 2 ) = γ + ln ( 2 N 1 + 4 N 2 ) C i ( 2 N ) \begin{array}{rcl} \displaystyle\int_0^N \left(\frac{1}{1 + 4x^2} - \cos(2x)\right) \frac{dx}{x} & = & \displaystyle \int_0^{2N} \frac{1 - \cos x}{x}\,dx - \Big[\tfrac12\ln(1 + 4x^2)\Big]_0^N \\ & = & \displaystyle \gamma + \ln(2N) - Ci(2N) - \tfrac12\ln(1 + 4N^2) \\ & = & \gamma + \ln\left(\frac{2N}{\sqrt{1+4N^2}}\right) - Ci(2N) \end{array} It is now clear that 0 ( 1 1 + 4 x 2 cos ( 2 x ) ) d x x = lim N 0 N ( 1 1 + 4 x 2 cos ( 2 x ) ) d x x = γ . \int_0^\infty \left(\frac{1}{1 + 4x^2} - \cos(2x)\right) \frac{dx}{x} \; = \; \lim_{N \to \infty} \int_0^N \left(\frac{1}{1 + 4x^2} - \cos(2x)\right) \frac{dx}{x} \; = \; \gamma \;.

Mark Hennings - 5 years, 6 months ago

Log in to reply

OH MY GOD!!!! How on Earth?! I'm speechless. This solution needs to be printed and framed.

This link is literrally a trasure chest!!

Please be the next featured member on Brilliant ! I want to know everything about you! Where your talents come from? Your idealogy? Your favorite mathematics topic? etc.

And are you active in any other math platform other than Brilliant? StackOverFlow? MathStackExchange? etc.

And why haven't you proved (disproved?) the Riemann hypothesis yet? haha!

  • Biggest fan.

Pi Han Goh - 5 years, 6 months ago

Log in to reply

@Pi Han Goh Glad you like it. I am an analyst by training, and this is part of what analysts do. I am quite happy not to be the next featured member, though.

By the way, the reason why the first integral evaluates to γ -\gamma is that it is equal to Γ ( 1 ) = ψ ( 1 ) \Gamma'(1) = \psi(1) .

As you can see, nothing that has to do with the Euler-Mascheroni constant is that elementary; that is probably why we still know so little about it!

Mark Hennings - 5 years, 6 months ago

Log in to reply

@Mark Hennings So analyst, huh. I'll start focusing on that right now :) .

Héctor Andrés Parra Vega - 5 years, 5 months ago

Even I did the same. I directly used the cosine integral for it. Cosine integral could be evaluated using taylor series. Then u could use the sum as it is. That would reduce the number of steps. Actually I don't know laplace transform. That's y I couldn't get your solution. Anyways nice solution :)

Aditya Kumar - 5 years, 6 months ago

Log in to reply

@Aditya Kumar When I solved the problem, I started at "We can now solve the problem!". The identity C i ( x ) = γ + ln x + 0 x cos t 1 t d t Ci(x) \; = \; \gamma + \ln x + \int_0^x \frac{\cos t - 1}{t}\,dt for the Cosine integral C i ( x ) = x cos t t d t Ci(x) \; = \; -\int_x^\infty \frac{\cos t}{t}\,dt is a standard one, and using it gets the answer almost immediately, since it is clear that C i ( x ) 0 Ci(x) \to 0 as x x \to \infty .

I only took the time, using Laplace transforms, to prove that identity, since an elementary proof was asked for.

Mark Hennings - 5 years, 6 months ago

Log in to reply

@Mark Hennings Hmm I see. Nice way to present your solution!

Aditya Kumar - 5 years, 6 months ago

I did not know these were standard integrals. I started of by x = (tant)/2 and got these integrals to solve.

Satyen Dhamankar - 5 years, 6 months ago

Log in to reply

@Satyen Dhamankar I doubt that's possible. Please show your working, thanks.

Pi Han Goh - 5 years, 6 months ago

Log in to reply

@Pi Han Goh Substituting x = (tant)/2 in the integral and simplifying a bit, you get the integrand as cot(t) - {[cos(tant)].(sect)^2 }/tant, with limits from 0 to pi/2. Now let tant=y. Therefore the integrand reduces to (1-cosy)/y, limits from 0 to infinity. (Sorry for not being proficient in Latex)

Satyen Dhamankar - 5 years, 6 months ago

Log in to reply

@Satyen Dhamankar Not quite. Your pair of substitutions is just substituting 2 x = y 2x = y , which replaces the integral with 0 ( 1 1 + y 2 cos y ) d y y \int_0^\infty \left( \dfrac{1}{1+y^2} - \cos y\right)\,\dfrac{dy}{y} and not 0 1 cos y y d y \int_0^\infty \dfrac{1 - \cos y}{y}\,dy as you claim.

Moreover, this last integral does not exist, since 0 x 1 cos y y d y γ + ln x x . \int_0^x \dfrac{1 - \cos y}{y}\,dy \; \sim \; \gamma + \ln x \qquad \qquad x \to \infty \;.

Anyway, you are still going to have to quote the result about the Cosine Integral to handle your last integral. As I have said in other posts, the work with the Laplace transform was only presented to prove the results about the Cosine Integral; the results themselves are standard.

Mark Hennings - 5 years, 6 months ago

I have written another elementary solution.

Ishan Singh - 4 years, 12 months ago

0.57721566490153286060651209008240243104215933593992….

Lu Chee Ket - 5 years, 6 months ago

Log in to reply

I've always seen that u just provide the answer. You never provide the complete solution. Can't u just post the complete solution whenever u r willing to post an answer?

Aditya Kumar - 5 years, 5 months ago

Log in to reply

Depends onto needs for completion. Please pick anything that helps. To me, the more the information, the more the helps. It is all up to you or me to pick whatever we want. As long as we are not going against the principle, just take with gratitude. Reasonable requests shall always be considered upon requests.

Lu Chee Ket - 5 years, 5 months ago

Log in to reply

@Lu Chee Ket OK so from now on please don't post direct answers in solutions. It freaks people out. If u want attention post a solution properly. U won't gain anything doing this.

Aditya Kumar - 5 years, 5 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...