## ganeshie8 one year ago A regular $$n-gon$$ is inscribed in an unit circle. Let $$c_{ij}$$ be the length of chord joining vertices $$i$$ and $$j$$. Show that $\sum\limits_{i\ne j} {c_{ij}}^2 = n^2$ and $\sum\limits_{i\ne j}\dfrac{1}{{c_{ij}}^2} = \dfrac{n^2-1}{12}$

1. freckles

|dw:1441591152116:dw| how to show: $c_{ij}=c_{jk}=c_{ki}=\sqrt{3}$

2. thomas5267

$\operatorname{crd}(x)=2\sin\left(\frac{x}{2}\right)\\$

3. thomas5267

What would be the sum of n=3? \begin{align*} \sum\limits_{i\ne j} {c_{ij}}^2 &= c_{12}+c_{13}+c_{23}?\\ &=c_{12}+c_{13}+c_{23}+c_{21}+c_{31}+c_{32}? \end{align*}

4. ganeshie8

sry i think order doesn't matter freckles has correct interpretation it seems.. there are only $$\binom{n}{2}$$ chords, not $$n^2$$

5. ganeshie8

What would be the sum of n=3? \begin{align*} \sum\limits_{i\ne j} {c_{ij}}^2 &= c_{12}+c_{13}+c_{23}?\\ \end{align*} i meant this, i see the ambiguity with my notation but i just don't know how to fix it...

6. freckles

|dw:1441592259514:dw| well this my interpretation for a 4-gon...$c^2_{12}+c^2_{23}+c^2_{34}+c^2_{41}=4^2 \\ \text{ but this a a reg n-gon } \\ \text{ so } c_{12}=c_{23}=c_{34}=c_{41} \\ \text{ so this really becomes } 4 c_{12}^2=4^2 \\ \text{ and so we really just want to show } c_{12}^2=4 \text{ or that } c_{12}=2$

7. thomas5267

Oops! Not fully awake. $\sum_{i=1}^{n-1}\sum_{j=i+1}^nc_{ij}?$

8. ganeshie8

that works!

9. thomas5267

10. ganeshie8

|dw:1441592568527:dw|

11. freckles

oh

12. ganeshie8

basically the first identity says : sum of squares of all the possible chords is $$n^2$$

13. ganeshie8

second identity says : sum of squares of reciprocals of all the possible chords is $$(n^2-1)/2$$

14. anonymous

mainly we need to show how many cycles the chords finish of the unit circle :D like if we have only n=3 then its one cycle and sum c_ij=1

15. thomas5267

$\sum_{k=1}^{n-1}\sum_{j=k+1}^n(c_{ij})^2= \sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1}(c_{ij})^2$ \operatorname{crd}(x)=2\sin\left(\frac{x}{2}\right)\\ \begin{align*} &\phantom{=}\sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1}\left(2\sin\left(\frac{2\pi (k-j)}{n}\right)\right)^2\\ &=2(n-1)(n)\sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1}\sin^2\left(\frac{2\pi (k-j)}{n}\right)\\ &=2(n-1)(n)\sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1}\left(1-\cos\left(\frac{4\pi (k-j)}{n}\right)\right)\\ &=n^2(n-1)^2-\sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1}\cos\left(\frac{4\pi (k-j)}{n}\right)\\ &=n^2(n-1)^2-\operatorname{Re}\left(\sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1} e^{\frac{4i\pi (k-j)}{n}}\right)\\ &=n^2(n-1)^2-\operatorname{Re}\left(\sum_{k=0}^{n-2}\sum_{j=k+1}^{n-1} e^{\frac{4i\pi k}{n}}e^{\frac{-4i\pi j}{n}}\right)\\ &=n^2(n-1)^2-\operatorname{Re}\left(\sum_{k=0}^{n-2}e^{\frac{4ki\pi }{n}}\left(\frac{1-e^{\frac{-4ni\pi }{n}}}{1-e^{\frac{-4i\pi }{n}}}-\frac{1-e^{\frac{-4(k+2)i\pi }{n}}}{1-e^{\frac{-4i\pi }{n}}}\right)\right)\\ \end{align*} I give up!

16. beginnersmind

I think I have an idea for the first identity. Let's start by looking at $\sum\limits_{j=2}^{n} {c_{1j}}^2$ That is all the chords starting from the vertex labelled with 1. We'll label vertices counterclockwise, like this: |dw:1441594979932:dw| We note that all other vertices have the same kinds of chords, so $\frac{n}{2}\sum\limits_{j=2}^{n} {c_{1j}}^2 = \sum\limits_{i\ne j} {c_{ij}}^2 = n^2$ or $\sum\limits_{j=2}^{n-1} {c_{1j}}^2 = 2n$ Let's define $$\LARGE\alpha=\frac{2\pi}{n}\$$ and use the law of cosines. $c_{12} = 1^2 + 1^2 - 2cos\alpha$ $c_{13} = 1^2 + 1^2 - 2cos2\alpha$ $c_{1n} = 1^2 + 1^2 - 2cos[(n-1)\alpha]$ I think for even n you can prove that the cosine terms pair up to add up to zero, except the middle one, which gives +2 (corresponding to the diagonal).

17. beginnersmind

Here's what I mean by 'pairing up' |dw:1441595804013:dw|

18. beginnersmind

@ganeshie8 @freckles can you check if my reasoning is correct?

19. thomas5267

Barbarian's approach: $\sum_{k=1}^{n-1}\sum_{j=k+1}^n(c_{ij})^2= \frac{1}{2}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}(c_{ij})^2 \quad c_{ij}=c_{ji},\,c_{ii}=0\\$ \operatorname{crd}(x)=2\sin\left(\frac{x}{2}\right)\\ \begin{align*} &\phantom{=}\frac{1}{2}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\left(2\sin\left(\frac{2\pi (k-j)}{n}\right)\right)^2\\ &=2\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\sin^2\left(\frac{2\pi (k-j)}{n}\right)\\ &=\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\left(1-\cos\left(\frac{4\pi (k-j)}{n}\right)\right)\\ &=n^2+\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\cos\left(\frac{4\pi (k-j)}{n}\right)\\ &=n^2+\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\operatorname{Re}\left( e^{\frac{4i\pi (k-j)}{n}}\right)\\ &=n^2+\operatorname{Re}\left(\sum_{k=0}^{n-1}\sum_{j=0}^{n-1} e^{\frac{4i\pi k}{n}}e^{\frac{-4i\pi j}{n}}\right)\\ &=n^2+\operatorname{Re}\left(\sum_{k=0}^{n-1} e^{\frac{4i\pi k}{n}}\frac{1-e^{\frac{-4i\pi n}{n}}}{1-e^{\frac{-4i\pi }{n}}}\right)\\ &=n^2+\operatorname{Re}\left(\frac{1-e^{-4i\pi}}{1-e^{\frac{-4i\pi }{n}}}\sum_{k=0}^{n-1} e^{\frac{4i\pi k}{n}}\right)\\ &=n^2+\operatorname{Re}\left(\frac{1-e^{-4i\pi}}{1-e^{\frac{4i\pi }{n}}}\frac{1-e^{4i\pi}}{1-e^{\frac{-4i\pi }{n}}}\right)\\ &=n^2+\operatorname{Re}\left(\frac{1-e^{-4i\pi}-e^{4i\pi }+1}{1-e^{\frac{4i\pi }{n}}-e^{\frac{-4i\pi }{n}}+1}\right)\\ &=n^2+\operatorname{Re}\left(\frac{2-\left(e^{-4i\pi}+e^{4i\pi }\right)}{1-e^{\frac{4i\pi }{n}}-e^{\frac{-4i\pi }{n}}+1}\right)\\ &=n^2+\operatorname{Re}\left(\frac{2-2}{1-e^{\frac{4i\pi }{n}}-e^{\frac{-4i\pi }{n}}+1}\right)\\ &=n^2 \end{align*}

20. thomas5267

I feel really old using this method. I have use this method at least 30 times.

21. thomas5267

I messed up the denominator in the last 5 steps but it does not affect the result.

22. thomas5267

Technically my proof isn't entirely correct since the thing in the real part simplifies to 0/0.

23. thomas5267

Actually no my proof is correct since $$1-e^{-4\pi i}=0$$.

24. ganeshie8

Awesome! @beginnersmind and @thomas5267 you both have the same idea it seems

25. ganeshie8

@beginnersmind is fixing the duplicates in the end @thomas5267 is evaluating the sum straight

26. thomas5267

Like how many times I have used the identity $$\operatorname{Re}\left(e^{ix}\right)=\cos(x)$$ on this site lol...

27. beginnersmind

Can you explain the sums of $$\LARGE e^{\frac{i\pi k}{n}}$$ add up the 1?

28. ganeshie8

i think we can just use the result : sum of roots of unity add up to 0

29. thomas5267

Bring back your awesome analytic proof!

30. beginnersmind

I mean $\LARGE e^{\frac{i4\pi k}{n}}$

31. ganeshie8

$$\sum\limits_{k=1}^n e^{i(k2\pi/n)} = 0$$ because each term in above sum is a root of the polynomial $$x^n-1$$, its easy to see that sum of roots of that polynomial is 0 (vieta formulas)

32. beginnersmind

Yeah, I see how those are equivalent. Not sure I see the easy part. :)

33. ganeshie8

i think you do :) just find the sum of roots of polynomial $$x^n+0x^{n-1}+\cdots +0x-1$$

34. beginnersmind

Ok, I think I see now. Write it as (x-x1)*(x-x2)*...*(x-xn) = 0 Then the sun is the negative of the x^(n-1) term, which is 0.

35. ganeshie8

Exactly..

36. ganeshie8

Okay so one down one more to go

37. beginnersmind

Neat. Plenty of good stuff to remember.

38. ganeshie8

i hope we could approach the other identity same way but there is no easy way to use euler idenity it seems..

39. thomas5267

Cannot use the same method without modification. x/0 occurs. \operatorname{crd}(x)=2\sin\left(\frac{x}{2}\right)\\ \begin{align*} &\phantom{=}\sum_{i=1}^{n-1}\sum_{j=i+1}^n\frac{1}{\left(c_{ij}\right)^2}\\ &=\frac{1}{2}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\left(2\sin\left(\frac{2\pi (k-j)}{n}\right)\right)^{-2}\text{ with some care}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\left(\sin^2\left(\frac{2\pi (k-j)}{n}\right)\right)^{-1}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\left(1-\cos\left(\frac{4\pi (k-j)}{n}\right)\right)^{-1}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1}{1-\cos\left(\frac{4\pi (k-j)}{n}\right)}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1}{1-\cos\left(\frac{4\pi (k-j)}{n}\right)}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{1-\cos^2\left(\frac{4\pi (k-j)}{n}\right)}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{\sin^2\left(\frac{4\pi (k-j)}{n}\right)}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{\cos^2\left(\frac{\pi}{2}-\frac{4\pi (k-j)}{n}\right)}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{\operatorname{Re}\left(e^{\frac{i\pi}{2}-\frac{4i\pi (k-j)}{n}}\right)^2}\\ &=\frac{1}{8}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{\operatorname{Re}\left(e^{\frac{i\pi}{2}}e^{-\frac{4i\pi (k-j)}{n}}\right)^2}\\ &=\frac{1}{2}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{\left(e^{\frac{i\pi}{2}}e^{-\frac{4i\pi (k-j)}{n}}+e^{-\frac{i\pi}{2}}e^{\frac{4i\pi (k-j)}{n}}\right)^2}\\ &=\frac{1}{2}\sum_{k=0}^{n-1}\sum_{j=0}^{n-1}\frac{1+\cos\left(\frac{4\pi (k-j)}{n}\right)}{-e^{-\frac{8i\pi (k-j)}{n}}+2-e^{\frac{8i\pi (k-j)}{n}}}\\ \end{align*}

40. thomas5267

How to count the number of residue classes of 2x+b mod n?

41. Empty

|dw:1441654270423:dw| do this for all n points and you will have found the length of the chords twice, so we need to divide by 2. $\frac{n}{2} \sum_{k=0}^{n-1} (e^{i\frac{ 2 \pi }{n}k}-1)(e^{i\frac{ 2 \pi }{n}k}-1)^*$ so you can also see in the sum itself I have the term with its complex conjugate since $$zz^* =c^2$$ Do boring algebra: $\frac{n}{2} \sum_{k=0}^{n-1} 2 - e^{i\frac{ 2 \pi }{n}k}-e^{-i\frac{ 2 \pi }{n}k}$ Separate the sum: $\frac{n}{2} \sum_{k=0}^{n-1} 2 -\frac{n}{2} \sum_{k=0}^{n-1} e^{i\frac{ 2 \pi }{n}k}+e^{-i\frac{ 2 \pi }{n}k}$ That first term is $$n^2$$ so now we really just have to show that this equals zero and we're home: $0=\sum_{k=0}^{n-1} e^{i\frac{ 2 \pi }{n}k}+e^{-i\frac{ 2 \pi }{n}k}$ Which it does because you can either think of it as cosine terms all added up on a complete rotation so the positive part is symmetric to the negative part and cancel. The second one is a bit trickier, I haven't gotten to doing that one yet.

42. thomas5267

There are $$\left\lfloor \frac{n}{2} \right \rfloor$$ edges with distinct length in a regular n-gon. For odd n, there are n edges of a particular length. For even n, there are n chords for all but one length. There are n/2 chords for that length. In total, there are $$\binom{n}{2}$$ edges. \begin{align*} &\phantom{=} \sum_{i=1}^{n-1}\sum_{j=i+1}^n\frac{2}{\left(c_{ij}\right)^2}\\ &=n\sum_{j=2}^{\left\lfloor \frac{n}{2}\right \rfloor+1}\frac{1}{\left(c_{1j}\right)^2} \text{ for odd n}\\ &=n\sum_{j=2}^{\frac{n}{2}}\frac{1}{\left(c_{1j}\right)^2} +\frac{n}{2}\frac{1}{\left(c_{1\frac{n}{2}}\right)^2}=n\sum_{j=2}^{\frac{n}{2}}\frac{1}{\left(c_{1j}\right)^2} +\frac{n}{8}\text{ for even since }c_{1\frac{n}{2}}=\text{diameter}=2\\ \end{align*} \text{For odd }n\text{:}\\ \begin{align*} &\phantom{=}n\sum_{j=2}^{\left\lfloor \frac{n}{2}\right \rfloor+1}\frac{1}{\left(c_{1j}\right)^2}\\ &=n\sum_{j=2}^{\left\lfloor \frac{n}{2}\right \rfloor+1}\frac{1}{\left(\operatorname{crd}\left(\frac{2\pi j}{n}\right)\right)^2}\\ &=n\sum_{j=2}^{\left\lfloor \frac{n}{2}\right \rfloor+1}\frac{1}{\left(1-e^{\frac{2\pi ij}{n}}\right)\left(1-e^{\frac{-2\pi ij}{n}}\right)}\\ &=n\sum_{j=2}^{\left\lfloor \frac{n}{2}\right \rfloor+1}\frac{1}{2-2\operatorname{Re}\left(e^{\frac{2\pi ij}{n}}\right)}\\ &=n\sum_{j=2}^{\left\lfloor \frac{n}{2}\right \rfloor+1}\frac{1}{2-2\cos\left(\frac{2\pi ij}{n}\right)} \end{align*} Can't evaluate the sum.

43. Empty

If you look at it geometrically and rearrange your indices you'll see because $$\csc(\frac{\pi}{2})=1$$ it looks like the sums will work out by symmetry.

44. thomas5267

How does that work? The problem is that it is squared so it won't cancel out.

Find more explanations on OpenStudy