Limits and Integrals (hum, yummy)

Calculus Level 5

lim x 1 x 1 x 1 t tan ( π t 2 t + 1 ) d t \large \lim_{x\to \infty} \frac{1}{x}\int_{1}^x \frac{1}{t}\tan\left(\frac{\pi t}{2t + 1}\right) dt

The limit above has a closed form. Find the limit to 3 decimal places.


The answer is 1.273.

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

Take g : I = [ 1 , ) R , such that g ( t ) = π t 2 t + 1 g : I =[1, \infty) \to \mathbb{R},\text{ such that } g(t) = \frac{\pi t}{2t + 1} . This function is a strictly function because g ( t ) = π ( 2 t + 1 ) 2 > 0 , t ( 1 , ) g'(t) = \frac{\pi}{(2t +1)^2} > 0,\space \forall t \in (1,\infty) , and g ( 1 ) = π 3 g(1) = \frac{\pi}{3} and lim t g ( t ) = π 2 \displaystyle \lim_{t\to \infty} g(t) = \frac{\pi}{2} .

Since tan ( x ) > x , x ( π 4 , π 2 ) \tan(x) > x , \space \forall x \in (\frac{\pi}{4},\frac{\pi}{2}) we have: lim x 1 x 1 t tan ( π t 2 t + 1 ) d t lim x 1 x 1 t π t 2 t + 1 d t lim x 1 x π 2 t + 1 d t = \displaystyle\lim_{x\to \infty} \int_{1}^x \frac{1}{t}\tan\left(\frac{\pi t}{2t +1}\right) dt \ge \displaystyle\lim_{x\to \infty} \int_{1}^x \frac{1}{t} \cdot \frac{\pi t}{2t + 1} dt \ge \displaystyle\lim_{x\to \infty} \int_{1}^x \frac{\pi}{2t + 1} dt = = π 2 [ lim x ln ( 2 x + 1 ) ln ( 3 ) ] = = \frac{\pi}{2} \cdot \left[\displaystyle\lim_{x\to \infty} \ln(2x + 1) - \ln(3)\right] = \infty Thus we have lim x 1 x 1 x 1 t tan ( π t 2 t + 1 ) d t = \displaystyle\lim_{x\to \infty} \frac{1}{x} \int_{1}^x \frac{1}{t}\tan\left(\frac{\pi t}{2t +1}\right) dt = \frac{\infty}{\infty} so that we can apply L'Hopital's rule and the fundamental theorem of calculus (version 1), obtaining lim x 1 x 1 x 1 t tan ( π t 2 t + 1 ) d t = lim x 1 x tan ( π x 2 x + 1 ) = \displaystyle\lim_{x\to \infty} \frac{1}{x} \int_{1}^x \frac{1}{t}\tan\left(\frac{\pi t}{2t +1}\right) dt = \lim_{x \to \infty} \frac{1}{x} \tan\left(\frac{\pi x}{2x + 1}\right) = = lim x 1 x cos ( π x 2 x + 1 ) × lim x sin ( π x 2 x + 1 ) = 0 0 . = \displaystyle\lim_{x\to \infty} \frac{\frac{1}{x}}{\cos\left(\frac{\pi x}{2x + 1}\right)} \times \lim_{x\to \infty} \sin\left(\frac{\pi x}{2x + 1}\right) = \frac{0}{0}. Now, it's sufficient that there exists the first limit above because the second limit above is 1 1 , and applying L'Hopital's rule again for the first limit we get lim x 1 x cos ( π x 2 x + 1 ) = lim x 1 x 2 sin ( π x 2 x + 1 ) π ( 2 x + 1 ) 2 = lim x ( 2 x + 1 ) 2 π x 2 = 4 π . \displaystyle\lim_{x\to \infty} \frac{\frac{1}{x}}{\cos\left(\frac{\pi x}{2x + 1}\right)} = \lim_{x\to \infty} \frac{\frac{-1}{x^2}}{- \sin\left(\frac{\pi x}{2x + 1}\right) \cdot \frac{\pi}{(2x +1)^2}} = \lim_{x\to \infty} \frac{(2x +1)^2}{\pi x^2} = \frac{4}{\pi}. Therefore, lim x 1 x 1 x 1 t tan ( π t 2 t + 1 ) d t = 4 π \displaystyle\lim_{x\to \infty} \frac{1}{x} \int_{1}^x \frac{1}{t}\tan\left(\frac{\pi t}{2t +1}\right) dt = \frac{4}{\pi}

great work

Ujjwal Mani Tripathi - 4 years, 9 months ago

Log in to reply

thank you buddy

Guillermo Templado - 4 years, 9 months ago

Superb u are awesome mind blowing how can u get these ideas

Sudhamsh Suraj - 4 years, 9 months ago

Log in to reply

Thank you, nevertheless, this problem and solution are not mine. If you want a problem from myself,for instance check this last one

Guillermo Templado - 4 years, 9 months ago
Mark Hennings
Sep 1, 2016

Suppose that f ( t ) f(t) is a continuous function on [ 1 , ) [1,\infty) such that lim t f ( t ) = L \lim_{t\to\infty}f(t) = L . Then 1 x 1 x f ( t ) d t L = 1 x 1 x [ f ( t ) L ] d t L x \frac{1}{x}\int_1^x f(t)\,dt - L \; = \; \frac{1}{x}\int_1^x \big[f(t) - L\big]\,dt - \frac{L}{x} Given ϵ > 0 \epsilon > 0 find X > 1 X > 1 such that f ( t ) L < ϵ |f(t)-L| < \epsilon for all t > X t > X . Then 1 x 1 x f ( t ) d t L 1 x 1 X [ f ( t ) L ] d t + ϵ x X x + L x ϵ + 1 x ( ( X 1 ) sup 1 t X f ( t ) L + L ) \left|\frac{1}{x}\int_1^x f(t)\,dt - L\right| \; \le \; \frac{1}{x}\left|\int_1^X \big[f(t) - L\big]\,dt\right| + \epsilon\frac{x-X}{x} + \frac{|L|}{x} \; \le \; \epsilon + \frac{1}{x}\left((X-1)\sup_{1 \le t \le X}|f(t)-L| + |L|\right) for all x > X x > X , so we can find Y > X Y > X such that 1 x 1 x f ( t ) d t L < 2 ϵ \left|\frac{1}{x}\int_1^x f(t)\,dt - L\right| \; < \; 2\epsilon for all x > Y x > Y . In other words, we have shown that lim x 1 x 1 x f ( t ) d t = L \lim_{x \to \infty}\frac{1}{x}\int_1^x f(t)\,dt \; = \; L In this case we have f ( t ) = 1 t tan ( π t 2 t + 1 ) = 1 t cot ( π 2 ( 2 t + 1 ) ) = cos ( π 2 ( 2 t + 1 ) ) g ( π 2 ( 2 t + 1 ) ) × 2 ( 2 t + 1 ) π t f(t) \; = \; \frac{1}{t}\tan\left(\frac{\pi t}{2t+1}\right) \; = \; \frac{1}{t}\cot\left(\frac{\pi}{2(2t+1)}\right) \;=\; \frac{\cos\left(\frac{\pi}{2(2t+1)}\right)}{g\left(\frac{\pi}{2(2t+1)}\right)} \times \frac{2(2t+1)}{\pi t} Where g ( x ) = sin x x g(x)=\frac{\sin x}{x} . Since g ( x ) 1 g(x) \to 1 as x 0 x \to 0 , we deduce that f ( t ) 4 π f(t) \to \frac{4}{\pi} as t t \to \infty . Thus the solution to this problem is 4 π = 1.273 \tfrac{4}{\pi} = \boxed{1.273} .

Hello again, good morning...I never thought to do it like this... Right now, I'm a little confused... However, it seems "correct"... The solution is 4 π \frac{4}{\pi} . I'll see slowly this proof later...The first part seems very interesting, nevertheless... in the penultime line I think there is a typo. Shouldn't it be f ( t ) = 1 t cot ( π 2 ( 2 t + 1 ) ) f(t) = \frac{1}{t} \cot\left(\frac{\pi}{2(2t +1)}\right) ? Could you develope the last lines a little bit more, please? ... Thank you, anyway...

Guillermo Templado - 4 years, 9 months ago

Log in to reply

Typo corrected. The last bit of the argument should now be clear.

Mark Hennings - 4 years, 9 months ago

Log in to reply

The last bit is now clear, thank you very much. I'm reviewing again the first part and I'm starting to fit things, and I think this is again unbelievable. You are fantastic... you are able to take a particular problem and generalize the result and create postulates that can be used to solve other problems. I think there are very few people who can do this. Did you know?

Guillermo Templado - 4 years, 9 months ago

Log in to reply

@Guillermo Templado I am glad you liked it. However, I did not do anything that special. Mathematics is about proving general results, after all.

Mark Hennings - 4 years, 9 months ago

Why couldn't you just directly use L'hopital (accompanied with fundamental theorem of calculus part 1), and we're left to prove that lim x 1 x tan ( π x 2 x + 1 ) = 4 π \displaystyle \lim_{x\to\infty} \dfrac1x \tan \left( \dfrac{\pi x}{2x + 1} \right) =\dfrac4\pi ?

Well of course, you need to show that tan ( π x 2 x + 1 ) > x \tan \left( \dfrac{\pi x}{2x+1} \right) > \sqrt x for x > 1 x>1 first, which is rather trivial. So the original integral diverges to infinity.

Pi Han Goh - 4 years, 9 months ago

Log in to reply

If you like, I am using de l'Hopital's rule. What I showed was simply a continuous variant of the rather better known result that if a sequence a n a_n converges to L L , then the sequence of arithmetic means 1 n j = 1 n a j L n \frac{1}{n}\sum_{j=1}^n a_j \, \to \, L \qquad n \to \infty

Mark Hennings - 4 years, 9 months ago

Log in to reply

What I showed was simply a continuous variant of the rather better known result that if a sequence a n a_n converges to L L ,

Hmmmm, either I forgot what this means or I'm really tired right now and I can't recall this fact. Can you remind me what technique/theorem this is called? Like Mean Value theorem? Or Continuity Correction Factor? Or etc

Pi Han Goh - 4 years, 9 months ago

Log in to reply

@Pi Han Goh Have a go. Split up the sum n 1 j = 1 n a j L = n 1 j = 1 n ( a j L ) n^{-1}\sum_{j=1}^na_j - L = n^{-1}\sum_{j=1}^n(a_j-L) into the sum for 1 j N 1 \le j \le N and N + 1 j n N+1 \le j \le n . For any fixed N N , the first sum will tend to 0 0 as n n \to \infty . How do you choose N N to make the second sum small?

Mark Hennings - 4 years, 9 months ago

Log in to reply

@Mark Hennings I think you're referencing Riemann Sum, but I don't see how is that relevant (yet).

Split up the sum n 1 j = 1 n a j L = n 1 j = 1 n ( a j L ) n^{-1}\sum_{j=1}^na_j - L = n^{-1}\sum_{j=1}^n(a_j-L) into the sum for 1 j N 1 \le j \le N and N + 1 j n N+1 \le j \le n

Doesn't this equation simplifies to L / n = L n = 1 L/n = L \Rightarrow n = 1 ? Are you sure you wrote it correctly?

And for your question, my intuition says that it's infinity, but given that I can't understand what is going on, I'm can only guess what the intended answer is. Sorry for being ignorant, but can you spell it out for me? I've been pondering on this question for quite some time now.

Pi Han Goh - 4 years, 9 months ago

Log in to reply

@Pi Han Goh Given ϵ > 0 \epsilon > 0 , we can find N N such that a n L < ϵ |a_n-L| < \epsilon for all n > N n > N . This means that 1 n j = N + 1 n ( a j L ) < n N n ϵ < ϵ \left|\frac{1}{n}\sum_{j=N+1}^n(a_j-L)\right| \; < \; \frac{n-N}{n}\epsilon \; < \; \epsilon for all n > N n > N .

The number N N is fixed, and hence j = 1 N ( a j L ) \sum_{j=1}^N(a_j-L) is a fixed number. Thus we can find M > N M > N such that 1 n j = 1 N ( a j L ) < ϵ \left|\frac{1}{n}\sum_{j=1}^N(a_j-L)\right| \; < \; \epsilon for all n > M n > M . This means that 1 n j = 1 n a j L < 2 ϵ \left|\frac{1}{n}\sum_{j=1}^n a_j - L\right| < 2\epsilon for all n > M n > M . QED

Mark Hennings - 4 years, 9 months ago

Log in to reply

@Mark Hennings Okay, I'm literally lost for words right now. I (am pretty sure I) understood what you wrote but I can't fathom how you managed to think of all these steps. Is there a prerequisite / known theorem / lemma / algorithm that you're using in your solution and in this comment? I doubt "epsilon-delta" is the only explanation I need.

Pi Han Goh - 4 years, 9 months ago

Log in to reply

@Pi Han Goh I have to say this is pretty standard bookwork. Try any good book on convergence.

Mark Hennings - 4 years, 9 months ago

Log in to reply

@Mark Hennings Damn. I feel so ashamed right now. Okay, you should know what is my follow up question is right now:

Do you have any good book on convergence? I doubt Jame Stewart or Rudin taught these stuff

Pi Han Goh - 4 years, 9 months ago

Log in to reply

@Pi Han Goh W.L.Ferrar wrote "Convergence", a very old OUP book. You can find scans of it on the internet, and it goes in to the properties of sequences in some detail. It was also the first book I bought as an undergraduate, many years ago!

Mark Hennings - 4 years, 9 months ago

Log in to reply

@Mark Hennings You're the best!! Brilliant should send a notification to everyone everything you post a comment/solution. Your comments are always super duper helpful. Thanks again.

Pi Han Goh - 4 years, 9 months ago

Yup, Right,I used L'Hópital twice and fundamental theorem Calculus. Do you want me to write a particular solution. Anyway, Mark's solution is a more general solution,because his first postulate envolves more possibilities. Do you want me to write my solution, Pi ? If you want it, tomorrow it will be written.

Guillermo Templado - 4 years, 9 months ago

Log in to reply

YES PLEASEEEEEEE!

Pi Han Goh - 4 years, 9 months ago

Log in to reply

@Pi Han Goh Tomorrow, it will be written. I promise. Give me some time, please.

Guillermo Templado - 4 years, 9 months ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...