RADical cyclic function, dude!

Algebra Level 3

If a periodic non-zero function f ( x ) f(x) exists such that 3 f ( x ) = f ( x 1 ) + f ( x + 1 ) , \sqrt{3}f(x) = f(x - 1) + f(x + 1), which of these options could be its fundamental period?

Note : A function g ( x ) g(x) is periodic if there exists some T T such that g ( x ) = g ( x + T ) g(x) = g(x + T) for all x . x. The fundamental period is the minimum positive T T such that this property holds.

1 3 6 12 A function with this property does not exist

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.

5 solutions

Timothy Samson
Jan 24, 2018

3 f ( x + 1 ) = f ( x ) + f ( x + 2 ) \sqrt{3} f(x + 1) = f(x) + f(x+2)

3 f ( x 1 ) = f ( x 2 ) + f ( x ) \sqrt{3} f(x - 1) = f(x - 2) + f(x)

--- From this:

3 ( f ( x 1 ) + f ( x + 1 ) ) = 2 f ( x ) + f ( x 2 ) + f ( x + 2 ) \sqrt{3} ( f(x - 1) + f(x + 1)) = 2f(x) + f(x - 2) + f(x+2)

3 ( 3 f ( x ) ) = 2 f ( x ) + f ( x 2 ) + f ( x + 2 ) \sqrt{3} (\sqrt{3} f(x)) = 2f(x) + f(x - 2) + f(x+2)

f ( x ) = f ( x 2 ) + f ( x + 2 ) f(x) = f(x-2) + f(x + 2)

--- Do it again

f ( x + 1 ) = f ( x 1 ) + f ( x + 3 ) f(x + 1) = f(x-1) + f(x + 3)

f ( x 1 ) = f ( x 3 ) + f ( x + 1 ) f(x - 1) = f(x-3) + f(x + 1)

We get:

f ( x 3 ) = f ( x + 3 ) -f(x - 3) = f(x + 3)

f ( x ) = f ( x + 6 ) f(x) = -f(x+6)

therefore f ( x ) = f ( x + 12 ) f(x) = f(x + 12)

Technically you have to prove such a function exists, and an example is f ( x ) = sin π 6 x f(x) = \sin{\frac{\pi}{6}x} .

Zain Majumder - 3 years, 4 months ago

Log in to reply

You can define f(x) however you like on [0,2).

Peter Byers - 3 years, 4 months ago

Log in to reply

Did you mean [0,12)?

Pau Cantos - 3 years, 4 months ago

Not really. Even though you have "such a function does not exist" as a possibile option, the question is: IF it exists, what is its fundamental period? Well, should this function exist, its fundamental period would be 12.

Steve Gualtieri - 3 years, 4 months ago

Assume root 3 is the cubic root of -1 in the complex plane

Ryan Matthew - 3 years, 4 months ago

This shows that 12 is a period of f, but not necessarily that it is the fundamental period. (It is easy to show that 1, 3, and 6 are not periods. But consider g(x) = f(11x), g satisfies all the requirements for f, but has a period of 12/11, and thus a fundamental period that is some integer fraction of that.

Robert Beggs - 3 years, 4 months ago

Log in to reply

Robert, you are right on. In fact, the periods such functions can take on are 12/n, where n is any positive integer not divisible by 2 or 3 (n = 5, 7, 11, 13, 17, ...).

Bottom line here is the problem does not have a correct choice, but analyzing why was an interesting excursion into trigonometry and elementary number theory. I posted an outline of the proof as a comment among these Solutions.

Will Heierman - 3 years, 3 months ago
Ivo Zerkov
Feb 5, 2018

I found it easiest to just solve the recurrence relation over the whole numbers; we can then keep shifting the interval(*) over which we solved the equation to make it hold for all numbers:

( 3 ) f ( x ) = f ( x 1 ) + f ( x + 1 ) \sqrt(3)f(x)=f(x-1)+f(x+1)

Characteristic equation:

r 2 3 r + 1 = 0 \displaystyle r^{2}-\sqrt{3}r+1=0

r = 3 ± i 2 = e ± i θ \displaystyle r=\frac{\sqrt{3}\pm i}{2} = e^{\pm i\theta} .

Thus, since f ( x ) = a cos x θ + b sin x θ f(x)=a\cos{x\theta}+b\sin{x\theta} , we have the period 2 π θ \displaystyle \frac{2\pi}{\theta} = 2 π π / 6 \displaystyle \frac{2\pi}{\pi/6} = 12 12 .

Edit: as it turns out, other (smaller) fundamental periods could be generated for differing f ( x ) f(x) -s. For example, if one uses θ = 11 π 6 \theta=\frac{11\pi}{6} instead of π 6 \frac{\pi}{6} , the general solution yields a function with period 12 11 \frac{12}{11} . Similarly, any period 12 12 k ± 1 \frac{12}{12k\pm 1} can be achieved. Furthermore, one can combine two different θ \theta -s satysfying the equation (and yielding by themselves a solution with a period of the forementioned form 12 12 k ± 1 \frac{12}{12k\pm 1} by using them for different parts of the interval(*) from the beginning to yield a function with any period 12 k \frac{12}{k} , as long as ( 12 ; k ) = 1 (12;k)=1 . Credit to @Will Heierman.

How did u get that characteristic equation??

Pulkit Agrawal - 3 years, 4 months ago

Log in to reply

because , it is like recursive progression

Azzam Labib - 3 years, 4 months ago

I am not sure why functions cannot exist with shorter periods, for example, how about one of period 12/5, where f(x+6/5)=-f(x). This would satisfy f(x+6)=-f(x) and f(x+12)=f(x).

It would almost certainly be discontinuous, piecewise defined on sets whose elements differ by muntipkes of 12/5. In other words, I did not see any indication that all possibilities were considered.

Will Heierman - 3 years, 4 months ago

Log in to reply

I think the function f(x)=sin(11[pi]x/6) satisfies the equation, but its period is 12/11. None of the choices allow for this.

Will Heierman - 3 years, 4 months ago

Log in to reply

Well, it doesn't. Try x = 1 x=1 .

Edit: yes it does, I can't substitute.

Ivo Zerkov - 3 years, 3 months ago

Log in to reply

@Ivo Zerkov I think when f(x) = sin (11 pi x/6) we have f(1) = sin(11 pi /6) = -1/2 and f(0)+f(2) = 0 + sin(11 pi/3) = 0 +[-SQRT(3)]/2 so the equation holds! No counterexample here!

Will Heierman - 3 years, 3 months ago

Log in to reply

@Will Heierman My bad, it does indeed work! Notice this is equivalent to using a θ \theta different from ± π 6 \frac{\pm\pi}{6} that satisfies e ± i θ = 3 + i 2 e^{\pm i\theta}=\frac{\sqrt{3}+i}{2} , namely 2 π + π 6 = 11 π 6 2\pi+\frac{-\pi}{6}=\frac{11\pi}{6} , in my equation for the general solution. I'm not so sure what the answer is anymore, I'll update if I figure it out.

Ivo Zerkov - 3 years, 3 months ago

Log in to reply

@Ivo Zerkov I have the whole answer. Periods that can occur are 12/n where n is an odd number not divisible by 3. The proof is elementary but a little tricky. I posted an outline in a reply to your query to Arjen.

Will Heierman - 3 years, 3 months ago

Thank you for pointing out that there are multiple possible periods. I have edited the wording to account for this.

Andrew Hayes Staff - 3 years, 3 months ago

Just because for your f ( x ) f(x) , f ( x + 12 ) = f ( x ) f(x+12)=f(x) , it doesn't necessarily follow you can make your f ( x ) f(x) match up with what we need in the problem. This becomes obvious when you consider that my f ( x ) f(x) , namely f ( x ) = a cos ( π 6 x f(x)=a\cdot\cos{(\frac{\pi}{6}\cdot x} ) ) + b sin ( π 6 x +b\cdot\sin{(\frac{\pi}{6}\cdot x} ) ) isn't periodic over 12 5 \frac{12}{5} , whereas yours is.

Ivo Zerkov - 3 years, 3 months ago

These solutions are hidden in Ivo's solution: the solutions of r = 1 2 ( 3 ± i ) r = \tfrac12(\sqrt 3 \pm i) are θ = 2 π n ± π 6 n Z , \theta = 2\pi n \pm \frac \pi 6\ \ \ \ n\in\mathbb Z, so that the solutions for f f are of the form f ( x ) = n = 0 A n , ± cos ( 2 π ( n ± 1 12 ) x ) . f(x) = \sum_{n = 0}^\infty A_{n,\pm}\cos(2\pi(n \pm \tfrac1{12})x). The fundamental periods of the individual terms are T n , ± = 1 n ± 1 12 = 12 12 n ± 1 , T_{n,\pm} = \frac{1}{n \pm \tfrac1{12}} = \frac{12}{12n \pm 1}, and the fundamental period of the sum is T = gcd { 12 12 n ± 1 } . T = \text{gcd} \{\frac{12}{12n \pm 1}\}.

Arjen Vreugdenhil - 3 years, 3 months ago

Thank you for pointing out that there are multiple possible periods. I have edited the wording to account for this.

Andrew Hayes Staff - 3 years, 3 months ago

Fourier analysis: f ( x ) f(x) restricted to domain Z \mathbb Z is the sum of harmonics based on a period with integer value T T ; it is the real part of: f ( x ) = n = 0 T 1 A n exp ( 2 n π i x T ) . f(x) = \sum_{n = 0}^{T - 1} A_n \exp\left( \frac{2n\pi i x}T\right). Then f ( x ± 1 ) = n = 0 T 1 A n exp ( 2 n π i x T ) exp ( ± 2 n π i T ) , f(x \pm 1) = \sum_{n = 0}^{T - 1} A_n \exp\left( \frac{2n\pi i x}T\right)\exp\left(\pm \frac{2n\pi i}T \right), so that (with exp ( a i ) + exp ( a i ) = 2 cos a \exp(ai) + \exp(-ai) = 2\cos a ) f ( x + 1 ) + f ( x 1 ) = 2 n = 0 T 1 A n exp ( 2 n π i x T ) cos ( 2 n π T ) , f(x+1) + f(x-1) = 2\sum_{n = 0}^{T - 1} A_n \exp\left( \frac{2n\pi i x}T\right)\cos\left(\frac{2n\pi}T \right), which for all integers x x should be equal to 3 f ( x ) \sqrt 3 f(x) : 2 n = 0 T 1 A n exp ( 2 n π i x T ) cos ( 2 n π T ) = 3 n = 1 T 1 A n exp ( 2 n π i x T ) . 2\sum_{n = 0}^{T - 1} A_n \exp\left( \frac{2n\pi i x}T\right)\cos\left(\frac{2n\pi }T \right) = \sqrt 3 \sum_{n = 1}^{T-1} A_n \exp\left( \frac{2n\pi i x}T\right). This requires that for every n n , either A n = 0 A_n = 0 or 2 cos ( 2 n π / T ) = 3 . 2\cos(2n\pi/T) = \sqrt 3. This means that cos 2 n π T = 1 2 3 2 n π T = ± π 6 n = T 12 . \cos\frac{2n\pi}T = \tfrac12\sqrt 3\ \ \ \ \therefore\ \ \ \ \frac{2n\pi}T = \pm \frac\pi 6\ \ \ \therefore\ \ \ n = \frac T{12}. Thus only the term with n / T = 1 / 12 n/T = 1/12 will be non-zero. Therefore the function has period 12 \boxed{12} and must be of the form f ( x ) = R e A exp ( 2 π i x 12 ) = A 1 cos ( 2 π x 12 ) + A 2 sin ( 2 π x 12 ) , f(x) = \mathrm{Re}\ A \exp\left( \frac{2\pi i x}{12}\right) = A_1\cos\left( \frac{2\pi x}{12}\right) + A_2\sin\left( \frac{2\pi x}{12}\right), where A = A 1 + A 2 i A = A_1 + A_2i .


Note that we have only considered the restriction of f f to the whole numbers Z \mathbb Z . We can give the same reasoning for the restriction to any subdomain of the form x 0 + Z x_0 + \mathbb Z , with 0 x 0 < 1 0 \leq x_0 < 1 . In general, then, f ( x ) = A 1 , { x } cos ( 2 π x 12 ) + A 2 , { x } sin ( 2 π x 12 ) , f(x) = A_{1,\{x\}}\cos\left( \frac{2\pi x}{12}\right) + A_{2,\{x\}}\sin\left( \frac{2\pi x}{12}\right), where { x } : = x x \{x\} := x - \lfloor x\rfloor and ( A 1 , x , A 2 , x ) x [ 0 , 1 (A_{1,x},A_{2,x})_{x \in [0,1\rangle} is an uncountable collection of parameters, not all zero.

Hey Arjen,

What happens if we use, say, 11 π 6 \frac{11\pi}{6} instead of π 6 \frac{\pi}{6} ? Could this yield a function with a period of 12 11 \frac{12}{11} , or similarly any period 12 12 k ± 1 \frac{12}{12k\pm 1} with k k whole? Check out what @Will Heierman commented on my solution, if you'd like an explicit example of such a function.

Ivo Zerkov - 3 years, 3 months ago

Log in to reply

Hi Ivo and Argen,

It turns out that the periods which functions satisfying the additive identity can have are 12/n, where n is any positive integer which is not divisible by 2 or 3 (i. e. is relatively prime to 12).

I have a proof which I have generated on a computer, but do not know how to insert it here properly formatted. If you email me at wheierman@corunduminium(dot)com, I will try to Messenger a photo to you.

The idea is this. Some of the details are left to you guys.

1: The trig polynomials f(x)=sin(n pi x/6) satisfy the equation if n differs from a multiple of 12 by 1. They have period 12/n.

2: If n is divisible by 2 or 3, the equation can be used to imply violation of periodicity by showing f must be identically zero.

3: If n is any other positive integer, we can find two positive integers, both differing from a multiple of 12 by 1, whose greatest common divisor is n. Call them j and k.

4: Define f as follows. Let f(x)= sin(j pi x / 6) if x is an integer and f(x)=sin(k pi x / 6) if x is not.
Its period is 12/n.

Ba-da-BING!

Will Heierman - 3 years, 3 months ago

Log in to reply

If we just took the sum f(x)=sin(j pi x /6)+sin(k pi x / 6) on the real line in the last argument part 4, we still get period 12/n. This is a simpler construction than the one used there.

Will Heierman - 3 years, 3 months ago

You are right... here is how I would approach it:

Let n N n \in \mathbb N and f ( x ) = sin n π x / 6 f(x) = \sin n\pi x/6 . Choose an arbitrary x 0 R x_0 \in \mathbb R . Then

f ( x 0 1 ) + f ( x 0 + 1 ) = sin ( n π x 0 6 n π 6 ) + sin ( n π x 0 6 + n π 6 ) = 2 sin n π x 0 6 cos n π 6 = 2 cos n π 6 f ( x 0 ) . f(x_0 - 1) + f(x_0 + 1) = \sin \left(\frac{n\pi x_0 }6 - \frac{n\pi}6\right) + \sin \left(\frac{n\pi x_0 }6 + \frac{n\pi}6\right) \\ = 2\sin \frac{n\pi x_0 }6 \cos \frac{n\pi}6 = 2\cos \frac{n\pi}6 f(x_0). This satisfies the condition on f f iff cos n π / 6 = 1 2 3 \cos n\pi/6 = \tfrac12\sqrt 3 , i.e. n = ± 1 n = \pm 1 modulo 12.

Arjen Vreugdenhil - 3 years, 3 months ago

Log in to reply

@Arjen Vreugdenhil Pathological functions may not have Fourier transforms. Though for continuous functions they converge almost everywhere, I still don't think they necessarily converge to the function. This problem has some very deep nuances, fascinating!!

Will Heierman - 3 years, 3 months ago

That's really clever, thank you for sharing! I was much too quick to conclude the answer is 12.

Ivo Zerkov - 3 years, 3 months ago

Log in to reply

@Ivo Zerkov I sent a photo of my manuscript to Arjen. If you want one, friend Will Heierman on Facebook and I will copy it to you also.

Will Heierman - 3 years, 3 months ago
Peter Macgregor
Feb 5, 2018

Let us make the bold assumption that f ( x ) = C a x f(x)=C a^{x} where C and a are constants.

Substituting this into the equation and dividing by C gives

3 a x = a x 1 + a x + 1 \sqrt{3} a^{x}=a^{x-1}+a^{x+1}

Dividing throughout by a x 1 a^{x-1} gives the quadratic equation

3 a = 1 + a 2 \sqrt{3}a=1+a^{2}

with the easy solution

a = 3 ± i 2 = e ± i π 6 a=\frac{\sqrt{3} \pm i}{2}=e^{\pm\frac{i \pi}{6}}

And so the general solution to the functional equation is

f ( x ) = C 1 e i π x 6 + C 2 e i π x 6 f(x)=C_{1}e^{\frac{i \pi x}{6}}+C_{2}e^{-\frac{i \pi x}{6}}

For a full solution we would need boundary conditions to determine the arbitrary constants. However all we want is the period, T, of the function. Knowing how the imaginary exponentials relate to the trig functions which have period 2 π 2 \pi allows us to see from the general solution that

π 6 T = 2 π \frac{\pi}{6}T=2 \pi

which implies that T = 12 T=\boxed{12}

That was elegant.

Timothy Samson - 3 years, 4 months ago

What was the thought that let u make that bold assumption??.. Plz tell

Pulkit Agrawal - 3 years, 4 months ago

Log in to reply

My guess: Function is periodic, therefore trigonometry is involved. Trigonometry is involved in C a x Ca^{x} because of Euler's formula e i θ = cos ( θ ) + i sin ( θ ) e^{i\theta} = \cos(\theta) + i\sin(\theta) . Therefore, guess y = C a x y = Ca^{x}

Timothy Samson - 3 years, 4 months ago

I noticed that the functional equation had the same form as a recurrence relation, where the argument of the function corresponds to the subscripts in the recurrence relation. My assumption about the form of the solution was then inspired by the usual approach to recurrence relations. See https://brilliant.org/wiki/linear-recurrence-relations/.

Peter Macgregor - 3 years, 4 months ago

Nice solution. There's a typo after the "with the easy solution" it must be "a=..." rather than "x=...".

Pau Cantos - 3 years, 4 months ago

Log in to reply

Thanks Pau, I've corrected it now.

Peter Macgregor - 3 years, 4 months ago

When one presumes a particular form, other solutions having other periods might be missed. However, the way the problem is stated with choices, it is logical to assume the period is unique. Had that been true, generating one solution would have been a legitimate way to produce the correct response.

The problem is with the problem itself, so not to worry if it misled you. It misled me so I guessed wrong and looked at the solutions.

I informed Brilliant staff about it, but so far I see no indication the situation will be addressed and recrified. I haven't gotten my point back either! Grrrr!

Will Heierman - 3 years, 3 months ago
John Winkelman
Feb 7, 2018

as usual I took a simple minded approach. Sine is periodic, so I let f(x) = sinks, with k to be determined. It turns out to be Pi/6 to satisfy the relation for f(x). Next we want sin(Pi x/6)=sin (Pi/6 (x+T)), which gives T= 12. Probably the rt(3) made me think of a trig function. I have to admire the other solutions which made no assumptions, well done all!

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...