A trigonometric inequality

Algebra Level 5

Find the smallest integer N N such that for all positive integers n n sin 1 100 + sin 2 100 + . . . + sin n 100 < N \lvert \sin \frac{1}{100}+\sin \frac{2}{100}+...+\sin \frac{n}{100} \rvert <N

Details and assumptions

You may use the fact that π \pi is an irrational number.

Clarification: The measurements are done in radians, not degrees.


The answer is 200.

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

We shall extend the problem to the complex plane and use the useful identities:- sin x = e i x e i x 2 \sin x = \frac{ e^{ix} - e^{-ix}}{2} and e i x = c o s ( x ) + i s i n ( x ) e^{ix} = cos(x) + i \ sin(x) . The sum provided is S = k = 1 n e i k / 100 e i k / 100 2 S= \sum \limits_{k=1}^{n} \frac{e^{ik/100} - e^{-ik/100}}{2} . Note that k = 1 n e i k / 100 = e i / 100 ( e i n / 100 1 ) ( e i / 100 1 ) \sum \limits_{k=1}^{n} e^{ik/100} = \frac{e^{i/100}(e^{in/100} - 1)}{(e^{i/100} - 1)} (this forms a geometric progression with initial term e i / 100 e^{i/100} and common ratio e i / 100 e^{i/100} ). Also note that k = 1 n e i k / 100 = e i n / 100 1 1 e i / 100 \sum \limits_{k=1}^{n} e^{-ik/100} = \frac{e^{-in/100} - 1}{1-e^{i/100}} . Thus, 2 S = e i / 100 ( e i n / 100 1 ) ( e i / 100 1 ) e i n / 100 1 1 e i / 100 2S= \frac{e^{i/100}(e^{in/100} - 1)}{(e^{i/100} - 1)} - \frac{e^{-in/100} - 1}{1-e^{i/100}} . Using the second identity mentioned, 2 S = 2 i c o s e c 1 200 sin n 200 sin n + 1 200 2S= 2 \ i \ cosec \frac{1}{200} \sin \frac{n}{200} \sin \frac{n+1}{200} . Hence, S = c o s e c 1 200 sin n 200 sin n + 1 200 |S|= cosec \frac{1}{200} \sin \frac{n}{200} \sin \frac{n+1}{200} . To maximize this, we have to maximize sin n 200 sin n + 1 200 \sin \frac{n}{200} \sin \frac{n+1}{200} . Let f ( n ) = sin n 200 sin n + 1 200 f(n)= \sin \frac{n}{200} \sin \frac{n+1}{200} . Then note that f ( n ) = sin ( 2 n + 1 ) / 200 200 f'(n)= \frac{\sin (2n+1)/200 }{200} . As n n varies over the positive integers, this value will be the closest to zero when n = 200 n= 200 . Also, f ( n ) = cos ( 2 x + 1 ) / 200 100 f''(n)= \frac{\cos (2x+1)/200}{100} . At n = 200 n= 200 , this value is c o s 401 / 200 100 \frac{ cos 401/200}{100} , which is negative. So, the smallest positive integer value for N N is 200 200 .

Moderator note:

The proof of the formula for the sum is very nice: a classical use of complex numbers to avoid trigonometry. The estimate after that, though, is not rigorous. The standard tools of Calculus do not really apply here. One can get an upper bound using Calculus, allowing n n to be real, and calculating the exact values at the critical points. But then one has to show that this upper bound cannot be lowered, and this is a bit more complicated.

Daniel C.'s comment " Now, if we choose a k k , ( 2 k + 1 ) π (2k+1)\pi can be arbitrarily close to a multiple of 1 100 \dfrac{1}{100} (this is probably a problem)" really points at the heart of the matter. All known complete proofs are variations on a proof of the Dirichlet approximation theorem, which was one of the origins of the Dirichlet's Box Principle (a.k.a. Pigeon-Hole Principle).

We can do it by trigonometry We know for a series of type Sin (a), sin (a+b), sin (a+2b), sin (a+3b),..............., sin (a+(n-1) b)

The sum of terms is given by

Sin (nb/2) ×sin (a+(n-1)b/2) /sin (b/2)

In this let the sum till sin nth terme be Sn Here value of a=1/100, b=1/100

Sn=sin (n/200)×sin (1/100+(n-1)/200) /sin (1/200)

=) Sn=sin (n/200)×sin ((n+1)/200) /sin (1/200)

We know for small value of x , sin (x) approx equal to x So we can write

sin(1/200) as 1/200

=) Sn=200 sin (n/200)×sin ((n+1)/200)

=) Sn=100 (2sin (n/200)×sin ((n+1)/200)

Using 2sin (a)×sin (b)=cos ((a-b)/2)-cos ((a+b)/2)

We write Sn=100 (cos (1/2)-cos ((2n+1)/400)

Maximum cos (1/2)-cos (2n+1)/400

where cos (1/2) will be approx 1 will be when cos (2n+1/400)=-1

So it is at max 2Sn=100×(approx 2)

And the result will always be less than 2 So N=200

Tushar gautam - 7 years, 10 months ago

Log in to reply

The way you have presented your solution, it seems that the sum of sine series is a standard formula. However I googled it and found no satisfactory result, so (probably) you have to include a proof of the formula you have used (maybe by the same method I have used). Also, I can't understand your last paragraph. You might want to compile your equations in latex, which can be done easily by putting a \ ( and \ ) [spaces removed] around your equations. Also it would be better if you could explain your steps a bit thoroughly as I get lost reading your post.

Sreejato Bhattacharya - 7 years, 10 months ago

Log in to reply

Let S = sin(A) + sin(A+B) + sin(A+2B) + …. + sin(A+(n-1)B)

The key is to multiply by sin(B/2).So S.sin(B/2) = sin(A)sin(B/2) + sin(A+B)sin(B/2) …. + sin(A+(n-1)B)sin(B/2)

Using the identity sin(X)sin(Y) = ½{cos(X-Y)−cos(X+Y)} …

2S.sin(B/2) = cos(A-B/2)−cos(A+B/2) + cos(A+B/2)−cos(A+3B/2) …... +…................ cos(A+(2n-3)B/2)−cos(A+(2n-1)B/2)

Terms cancel in pairs →

2S*sin(B/2) = cos(A-B/2) − cos(A+(2n-1)B/2)

2S.sin(B/2) = 2sin(A+(n-1)B/2).sin(nB/2)

∴ S = sin(A+(n-1)B/2).sin(nB/2)/sin(B/2)

Tushar gautam - 7 years, 10 months ago

Log in to reply

@Tushar Gautam The multiplication by sin B 2 \sin \frac{B}{2} which yields a telescoping series is a standard approach.

Another approach would be to treat sin ( A + k B ) \sum \sin (A+kB) as the imaginary part of e i ( A + k B ) \sum e^{i(A+kB)} and then evaluate the (now) geometric progression.

Calvin Lin Staff - 7 years, 10 months ago

@Tushar Gautam Ok, I got it. That's a nice approach. You should have submitted a solution to this question. :)

Sreejato Bhattacharya - 7 years, 10 months ago

Log in to reply

@Sreejato Bhattacharya I already submitted one solution in this week's set so I can't subbit this one

Tushar gautam - 7 years, 10 months ago

And about representation of answer the language of writting answers on brilliant is difficult

Tushar gautam - 7 years, 10 months ago

It's tricky because s i n n 100 sin \frac{n}{100} has a period of \infty .

Michael Tong - 7 years, 10 months ago

Log in to reply

Indeed. This question would be very different if sin n 100 \sin \frac{n}{100} was replaced with sin n π 100 \sin \frac{n\pi}{100} . The irrationality of p i 100 \frac{pi}{100} is extremely important.

Calvin Lin Staff - 7 years, 10 months ago

I have another solution:

We can see that the above-given sum of sines will simply cancel out to approximately 0 after 2n\pi for all integer n. Hence, we only need to look at sin x \sin x for 0 < x < π 0 < x < \pi The sum will simply become 100*(area under graph of sin x \sin x for 0 < x < π 0 < x < \pi ), which can easily be evaluated to 200 ny integration, giving the required value of N. :)

Dhruv Baid - 7 years, 9 months ago

There is a typo in the second last line:- f ( n ) = cos ( 2 n + 1 ) / 200 100 f''(n)= \frac{ \cos (2n+1)/200 }{100} . The n n was replaced by the x x . Also sorry for poor latex formatting. :(

Sreejato Bhattacharya - 7 years, 10 months ago

Here is a potential solution (it uses some of the info here ). It isn't quite rigorous, but I think it might be close:

The given expression looks like a Riemann Sum. Let S S be our desired sum. It is clear that, where k k is a whole number n = ( 2 k + 1 ) π 100 n=\left\lfloor\dfrac{(2k+1)\pi}{100}\right\rfloor This is simply because we want n n to be at the last point where sin ( n ) \sin(n) is still positive for that period. The integral from 0 to ( 2 k + 1 ) π (2k+1)\pi of sin ( x ) \sin(x) is 2 2 . The Riemann sum from 0 to ( 2 k + 1 ) π (2k+1)\pi of sin ( x ) \sin(x) is 1 100 ( sin 1 100 + sin 2 100 + + sin n 100 ) = 1 100 S \dfrac{1}{100}\left(\sin\dfrac{1}{100}+\sin\dfrac{2}{100}+\cdots+\sin\dfrac{n}{100}\right)=\dfrac{1}{100}S Now, using the bounds in the linked article (we have the left Riemann sum), 2 S 100 2 1 100 π 2 = π 100 2-\dfrac{S}{100}\le 2\cdot\dfrac{1}{100}\cdot \dfrac{\pi}{2}=\dfrac{\pi}{100} I split the area into two parts, the area from 0 to π 2 \dfrac{\pi}{2} , and the area from π 2 \dfrac{\pi}{2} to π \pi .

Therefore, 2 π 100 S 2 2-\dfrac{\pi}{100}\le S\le 2 200 π S 200 200-\pi\le S\le 200 Now, if we choose a k k , ( 2 k + 1 ) π (2k+1)\pi can be arbitrarily close to a multiple of 1 100 \dfrac{1}{100} (this is probably a problem). Therefore, we can get S S very close to 200 \boxed{200} (also probably a problem).

The ending is not quite rigorous; can anyone fix it?

Daniel Chiu - 7 years, 10 months ago

Log in to reply

Hint: We want to show that multiples of 1 100 \frac{1}{100} can get arbitrailty close to ( 2 k + 1 ) π (2k+1) \pi . What does arbitrarily close mean? Well, let's first try and get within 0.001 of ( 2 k + 1 ) π (2k+1) \pi .

For each j j , let A j = j 100 ( 2 k j + 1 ) π A_j = \frac{j}{100} - (2k_j+1) \pi , where k j k_j is an integer such that 0 A j < 2 π 0 \leq A_j < 2 \pi . This defines how close we are to ( 2 k + 1 ) π ( 2k+1) \pi , which is what we want to work with.

Consider the values of A j A_j for j = 1 j = 1 to 10. Can you show that one of these values must be less than 0.001 0.001 ? How about if we use j = 1 j=1 to 10000?

How would you generalize this approach, to get within ϵ \epsilon of ( 2 k + 1 ) π (2k+1) \pi ? Is the fraction 1 100 \frac{1}{100} important? Can we replace it by 1 50 \frac{ 1}{ 50} ? p i 100 \frac{pi}{100} ? 1 π \frac{1}{\pi} ?

Notice how I flipped your condition around. If we fix a k k , then multiples of 1 100 \frac{1}{100} cannot get arbitrarily close to ( 2 k + 1 ) π (2k+1) \pi . E.g. If k = 0 k=0 , then the closest we can get is 3.14 3.14 .

Calvin Lin Staff - 7 years, 10 months ago

Reply to Challenge Master note
I found out later that S |S| could be maximized in a much simpler method. Note that f ( n ) = 1 2 ( cos 1 200 cos 2 n + 1 200 ) f(n)= \frac{1}{2} (\cos \frac{1}{200} - \cos \frac{2n+1}{200}) . To maximize this, we have to minimize cos 2 n + 1 200 \cos \frac{2n+1}{200} , whose minimum is 1 -1 , and attained when n = 100 π ( 2 c ± 1 ) 1 2 n= 100 \pi (2c \pm 1) - \frac{1}{2} ( c ϵ Z c \epsilon Z ).

Sreejato Bhattacharya - 7 years, 10 months ago

Log in to reply

n has to be an integer!

Sambit Senapati - 7 years, 10 months ago

Log in to reply

Yes that is true, and that's why the required minimum cannot be reached. However, we can determine a suitable n n so that its value is close to 1 -1 . Tushar G has proved this (I agree that my differentiation approach was a bit hazy and also somewhat incorrect).

Sreejato Bhattacharya - 7 years, 10 months ago

the summation of the series will be maximum when

n/100 =3.14

now applying summation of sine series

difference= 0.01 no.of terms = 314

Now,

{sin(100 3.14 0.5* .01) * sin([3.14+.01]/2)} / sin(0.02)

taking pi & 3.14 as interchangeable

we get,

{1*1} / 0.02

= 200(approx)

Purvam Modi - 7 years, 9 months ago
Hitesh Yadav
Jul 26, 2020

Well I would love for someone to correct me if I'm wrong on this one. By observing denominators I found that it would be sufficient to approximate value of pi to 2 digit after decimals . Then we can say that if we terminate n to 314 314 we will get max value of the sum. Therefore I terminated the sum there. Cause If I went a step further I would have included negatives. Having done that using the formula as described by Tushar Gautam I evaluated the sum by taking s i n ( x ) 1 / 200 sin(x)\approx1/200 . Now I know I can't justify or tell you how tight bounds are. And since by now I haven't entered the complex realm therefore this was pretty much the only option .Also there are couple of other approximations I made which you'll encounter if you try to do this way.

There are several issues

  1. While adding the specific term of n = 315 n = 315 will become negative, how do you know that there is no further sequence of terms we could add that (eventually) becomes positive? In particular, n = 315 942 sin n 100 0.000578 \sum_{n=315}^{942} \sin \frac{n}{100 } \approx 0.000578 , which means that the sum till 942 is larger than the sum till 314.
  2. The sum till n = 314 n = 314 gives 199.99900 199.99900 \ldots , so if we had an addition value of 0.001 0.001 , then the answer would be larger than 200. This seems to be a very small margin of error, esp since the next series of terms adds > 50% of the error.
  3. Note that we can get close to 200 (but not exactly). In fact, the supremum is 100 ( 1 + cos 1 200 100 ( 1 + \cos \frac{1}{200 } .

Calvin Lin Staff - 10 months, 2 weeks ago

Log in to reply

Ah! I see the problem now .I tried it on desmos and you are correct. I tried to add setting upper bounds like 942(3 pi) , then I also tried for 7853(25 pi), 14765(47 pi) and the answer kept closing to 200 . So I think I would have to go till infinity to ever reach 200. Can I still fix this answer somehow?

Hitesh Yadav - 10 months, 2 weeks ago

Log in to reply

Not easily. Your approach of only focusing on a finite truncated sum doesn't help address the infinitely large values of n n .

As a slightly easier way to see why that's the case, consider finding the supremum of cos n \sum \cos n . The cycling of numbers is much more aggressive, and you can quickly convince yourself that the max isn't cos 1 \cos 1 .

Calvin Lin Staff - 10 months, 2 weeks ago

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...