3

To prove:

$1.$ $$\sum_{k=1}^n \sin(2k-1)\theta = \frac{\sin^2 n\theta}{\sin \theta}.$$

$2.$ $$\sum_{k=1}^n \sin^2(2k-1)\theta = \frac{n}{2} - \frac{\sin 4n\theta}{4\sin 2\theta}.$$

jimjim
  • 9,675
Bob
  • 31

2 Answers2

7

They become easy (sums of geometric progressions) if you use

$$\sin(x)=\frac{e^{ix}-e^{-ix}}{2i}$$

and

$$\cos(x)=\frac{e^{ix}+e^{-ix}}{2}.$$

When $\sin$ and $\cos$ are replaced in this way by rational functions in $e^{ix}$, you can then replace $e^{ix}$ by $z$ and then many trigonometric identities become either polynomial identities or identities between power series.

If the equality is between rational functions in $\sin(x)$ and $\cos(x)$ then the above substitutions will give you an equality between rational functions. Clearing denominators it becomes an identity between polynomials. This can be checked just by writing the polynomials in the standard form (sums of powers of $z$). This turns the solution into an algorithm.

For example:

$$\cos(2x)=\cos^2(x)-\sin^2(x)$$

It turns into

$$\frac{z^2+z^{-2}}{2}=\left(\frac{z+z^{-1}}{2}\right)^2+\left(\frac{z-z^{-1}}{2i}\right)^2$$

Opening parenthesis, bringing all to one side, and writing as sum of powers of $z$ we get $$0z^4+0z^3+0z^2+0z+0=0.$$

In this problem you get sums of progressions with hypergeometric terms. The sum of which can also be determined algorithmically, whenever there is a nice sum.

OR.
  • 5,941
  • 1
    This really is the proper way to answer these types of questions. The product formula used in other answers is just a hidden application of the exponential definition, and the exponential definition solves a much wider collection of problems. It really is the method competition takers should get very familiar with. – ex0du5 Jul 12 '13 at 22:43
  • @ex0du5 I disagree. One can define the trigonometric functions independently of the exponential function. It seems you're implying the identities in my answer are somehow obscure, but the trigonometric functions can be defined axiomatically with very few properties, a fundamental one being $$\cos(x+y)=\cos x\cos y-\sin x\sin y$$ – Pedro Jul 12 '13 at 22:45
  • (And I disagree even more with the initial phrase "...is the proper way". You're simply stating your opinion, but it comes off as if you're claiming some to posses some absolute truth.) – Pedro Jul 12 '13 at 22:49
  • 1
    @ex0du5 "The ultimate goal in mathematics is to eliminate any need for intelligent thought" Whitehead. I usually favor the way that allows to solve more problems without thinking. – OR. Jul 12 '13 at 22:49
  • @PeterTamaroff I don't think s/he is implying your identities are obscure. And while you could define trigonometric functions without the exponential function, why would you want to? It's clearer and more elegant to do it with $e^{ix}$ if you're going to be doing it rigorously. – Potato Jul 12 '13 at 22:49
  • @Potato Because the trigonometric functions have their own flavour. I don't think they should be seen as appendage of the exponential function. They can be defined over the real numbers very nicely, while the connection to the exponential function pops up with complex numbers. (And still, how would you define $e^{ix}$?) – Pedro Jul 12 '13 at 22:55
  • @PeterTamaroff One of their fundamental properties is being an "appendage of" the exponential functional. There's really no reason to hide that. Perhaps for kids who are just being introduced to them via triangles, but for people who want a rigorous treatment or people doing contest math? That seems absurd to me. – Potato Jul 12 '13 at 22:58
  • @Potato Since I see we will get nowhere, I will not keep discussing this. – Pedro Jul 12 '13 at 22:59
  • @PeterTamaroff Oh, I'm not trying to rile you up! This is just a view I haven't encountered before. – Potato Jul 12 '13 at 23:00
  • @PeterTamaroff (And to answer the question, you define $e^{ix}$ through power series, of course!) – Potato Jul 12 '13 at 23:01
  • @Potato Oh, well, but using power series is already "non-elementary", to my taste. Sometimes I find axiomatic approaches more appealing than constructive ones, the case of the trigonometric functions being one. – Pedro Jul 12 '13 at 23:06
  • 1
    @PeterTamaroff: I didn't mean to imply your solution was more difficult. However, I do consider the exponential formulae much deeper. They directly follow from the bisection formula of Simpson for series, immediately giving the parity of the functions. The sum formulae follow directly from the orthogonality of the exponential, and the product are simple (matrix) inversion. Getting good at the exponential orthogonality gets you tons of nontrigonometric series, integrals, etc. and is useful in physics and other applied fields. – ex0du5 Jul 12 '13 at 23:09
  • 1
    @PeterTamaroff: And I don't just say this theoretically. When I first started playing with exponentials like this, my skills across the spectrum just took off. Very quickly, I was deriving my own q-series and hypergeometric series identities using the exact same skills. They are that fundamental and useful. Now, many wouldn't get as excited manipulating exponentials and series as I did and may not have the same results, but learning the skills is far more important than the specialized sum identities of trigonometrics in the field of competitions considered here too. – ex0du5 Jul 12 '13 at 23:12
  • @ex0du5 I think both approaches have their upsides and downsides. But, of course, using $e^{ix}$ has tons of advantages in the practice. I agree. I just disagree with the idea that because one provides practical ease it should overrun other theoretical approaches which are also very enlightening. I think conformism and ease are not currency in mathematics. – Pedro Jul 12 '13 at 23:15
  • I've been teaching some geometry to some high schools students and of course trig poped all over the place. Now I just found out that I was doing the wrong thing, I needed to wait until the learn complex Analysis to do that, since basic geometry is not really the proper way to introduce trigs. – N. S. Jul 12 '13 at 23:26
  • @N.S. No one said that! – Potato Jul 12 '13 at 23:28
  • 2
    @N.S. The role of doing plane geometry, cracking trigonometry the hard way, etc. is not only teaching them those topics. It is also to develop deductive skills. But yes, most of the problems they will be solving belong to the problems a computer can solve. – OR. Jul 12 '13 at 23:29
  • @N.S. Also, if you really want, the identities with complex numbers are purely geometrical. You just need to introduce, together with addition, which is easily seen as translation, rotation and inversion as a form of multiplication. The thing is that traditionally, books and school programs are written following a 'historicist' order. Roughly speaking, people find taboo teaching younger ages what was invented later in history, even if it makes more sense. – OR. Jul 12 '13 at 23:41
  • @RGB I really want to see a computer solving IMO type problems...And unless I teach them complex analysis, the exp/trig identities make absolutely no sense, they are actually very artificial. And I don't think we should teach young kids which are really interested in math "those formulas hold because I said so".... – N. S. Jul 13 '13 at 00:36
  • 1
    @N.S. Most IMO geometry problems can be solved by a computer. This is a well known application of Groebner bases. Also many of those sums of binomial coefficients and other hypergeometric terms too, although those are being less used in IMO than geometry. Complex analysis (i.e. derivatives and limits) is not necessary. You just have to be clever. Most IMO contestants know inversion. Inversion is just complex multiplication (actually division) in disguise. Complex numbers arithmetic is as elementary as inversion in geometry. The fact that it might be seen as more advanced or something is just a – OR. Jul 13 '13 at 02:30
  • 1
    ... historical handicap that comes from the mystic shadow they have been carrying since their introduction to solve the cubic equations in radicals. – OR. Jul 13 '13 at 02:31
4

Alternatively, from $\cos(a+b)=\cos a\cos b-\sin a\sin b$ we get $$\sin a\sin b=\frac{\cos(b-a)-\cos(b+a)}2$$

Giving appropriate values to $a,b$, obtain a telescopical sum on the left.

Let $a=x,b=(2k-1)x$. Then $$\eqalign{ & \sin x\sin \left( {2k - 1} \right)x = \frac{{\cos (2k - 2)x - \cos 2kx}}{2} \cr & \sin x\sum\limits_{k = 1}^n {\sin \left( {2k - 1} \right)x} = \frac{{1 - \cos 2nx}}{2} \cr} $$

so $$\sum\limits_{k = 1}^n {\sin \left( {2k - 1} \right)x} = \frac{{{{\sin }^2}nx}}{{\sin x}}$$

For the second one, note that $${\sin ^2}(2k - 1)\theta = \frac{{1 - \cos \left( {2k - 1} \right)2\theta }}{2}$$ so it suffices to find $$\sum\limits_{k = 1}^n {\cos \left( {2k - 1} \right)x} $$

And again, we can use $$\frac{{\sin \left( {a + b} \right) - \sin \left( {a - b} \right)}}{2} = \sin b\cos a$$ to get $$\sin x\cos \left( {2k - 1} \right)x = \frac{{\sin 2kx - \sin \left( {2k - 2} \right)x}}{2}$$ $$\sum\limits_{k = 1}^n {\cos \left( {2k - 1} \right)x} = \frac{{\sin 2nx}}{{2\sin x}}$$

so $$\sum\limits_{k = 1}^n {{{\sin }^2}(2k - 1)x}  = \frac{n}{2} - \frac{{\sin 4nx}}{{4\sin 2x}}$$

ADD Note that above gives $$\sum\limits_{k = 0}^{n - 1} {\frac{{\sin \left( {k + \frac{1}{2}} \right)t}}{{\sin \frac{t}{2}}}} = {\left( {\frac{{\sin \frac{{nt}}{2}}}{{\sin \frac{t}{2}}}} \right)^2}$$ which is fundamental when computing the Fejér Kernel as the Cesaro mean of the Dirichlet Kernel.

Pedro
  • 122,002