More fun in 2015, Part 39

d n μ ( n d ) f ( d ) = n \sum_{d|n}\mu\left(\frac{n}{d}\right)f(d)=n

If f ( d ) f(d) is an arithmetic function such that the equation above holds for all positive integers n n , find f ( 2015 ) f(2015) .

Notation: μ \mu denotes the Möbius function .


The answer is 2688.

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.

4 solutions

Alan Yan
Dec 22, 2015

Note that if d g = n dg = n then the expression is equivalent to g n μ ( g ) f ( n / g ) \sum_{g |n}{\mu(g) f(n/g)} which by Mobius Inversion, means that f ( n ) = d n d f(n) = \sum_{d|n}{d} or the sum of the divisors of n n . The answer follows by summing the divisors of 2015.

Yes, this is a concise solution for those of us who know about Mobius inversion! (+1)

Otto Bretscher - 5 years, 5 months ago
Otto Bretscher
Dec 22, 2015

Here is a solution for those who know the dirichlet convolution , ( f g ) ( n ) = d n f ( n / d ) g ( d ) (f*g)(n)=\sum_{d|n}f(n/d)g(d) . This product turns out to be associative, with an identity e e given by e ( 1 ) = 1 , e ( n ) = 0 e(1)=1,e(n)=0 for n > 1 n>1 . It is easy to see that I μ = e I*\mu=e where I ( n ) = 1 I(n)=1 for all n n ; this simply means that d n μ ( d ) = 0 \sum_{d|n}\mu(d)=0 for n > 1 n>1 .

We are given the equation i d = μ f id=\mu*f , the identity function. To solve for f f , let's multiply both sides with I I to find I i d = I μ f = e f = f I*id=I*\mu*f=e*f=f , meaning that f ( n ) = d n d f(n)=\sum_{d|n}d , the sum of all divisors of n n , a multiplicative function.

Since f ( p ) = p + 1 f(p)=p+1 for primes, we have f ( 2015 ) = f ( 5 13 31 ) = 6 14 32 = 2688 f(2015)=f(5*13*31)=6*14*32=\boxed{2688} .

For prime powers we have f ( p m ) = 1 + p + p 2 + . . . + p m = p m + 1 1 p 1 f(p^m)=1+p+p^2+...+p^m=\frac{p^{m+1}-1}{p-1} .

Moderator note:

Good approach. We are using the Mobius inversion to calculate the function f.

now that i have learn about the Dirichlet product, i understand how awesome this solution is!!

Aareyan Manzoor - 5 years, 5 months ago

Log in to reply

Could you add a wiki to help explain the Dirichlet convolution?

Calvin Lin Staff - 5 years, 5 months ago

Log in to reply

sure! let me create one

Aareyan Manzoor - 5 years, 5 months ago
Tran Hieu
Dec 22, 2015

First, it is easy to see that f ( 1 ) = 1 f(1) = 1

Then we have f ( p ) = p + 1 f(p) = p+1 for every prime p because μ ( p ) f ( 1 ) + μ ( 1 ) f ( p ) = p \mu(p)f(1)+\mu(1)f(p) = p , that means f ( p ) f ( 1 ) = p f(p)-f(1)=p

Next we will prove that if n = p 1 p 2 . . . p k n=p_1*p_2*...*p_k with p i p_i are prime, then f ( n ) = f ( p 1 ) f ( p 2 ) . . . f ( p k ) f(n)=f(p_1)*f(p_2)*...*f(p_k)

From the condition we have
μ ( 1 ) f ( n ) + p i n μ ( p i ) f ( n p i ) + p i , p j n μ ( p i p j ) f ( n p i p j ) . . . = n \mu(1)f(n)+\sum_{p_i|n}\mu(p_i)f(\frac{n}{p_i})+\sum_{p_i,p_j|n}\mu(p_i*p_j)f(\frac{n}{p_i*p_j}) ... = n

or

f ( n ) p i n f ( n p i ) + p i , p j n f ( n p i p j ) . . . + . . . = p 1 p 2 . . . p k = [ f ( p 1 ) 1 ] [ f ( p 2 ) 1 ] . . . [ f ( p k ) 1 ] f(n)-\sum_{p_i|n}f(\frac{n}{p_i})+\sum_{p_i,p_j|n}f(\frac{n}{p_i*p_j}) - ...+... = p_1*p_2*...*p_k = [f(p_1)-1][f(p_2)-1]...[f(p_k)-1]

When expand the RHS, all the sum in LHS will be cancel out and that leaves f ( n ) = f ( p 1 ) f ( p 2 ) . . . f ( p k ) f(n)=f(p_1)*f(p_2)*...*f(p_k)

Apply that to 2015 we have f ( 2015 ) = f ( 5 ) f ( 13 ) . . . f ( 31 ) = 6 14 32 = 2688 f(2015)=f(5)*f(13)*...*f(31) = 6*14*32 = \boxed{2688}

Note: In general, if n = p 1 α 1 p 2 α 2 . . . p k α k n=p_1^{\alpha_1}p_2^{\alpha_2}...p_k^{\alpha_k} :

f ( n ) = i = 1 k p i α i + 1 1 p i 1 f(n) = \prod_{i=1}^{k}\frac{p_i^{\alpha_i+1}-1}{p_i-1} (sum of all divisors of n)

Yes, this is a clear, straightforward approach! (+1) Can you show us how you derive the formula for f ( n ) f(n) in the last line? It seems like your earlier work applies to square-free n n only.

Otto Bretscher - 5 years, 5 months ago

Log in to reply

Well, I don't have any proof that f(n) is multiplicative given only your equation without transform f(n) to sum of all divisors of n

Tran Hieu - 5 years, 5 months ago

Log in to reply

In my solution, I show that f f is multiplicative by writing it as the Dirichlet convolution of two multiplicative functions.

Otto Bretscher - 5 years, 5 months ago
Aareyan Manzoor
Dec 22, 2015

i am quite certain this function is multiplicative, but i dont have proof(i was trying but it became too lengthy). lets use easy method:


put n=1 to get f(1)=1. now put n=5) to find f ( 5 ) f ( 1 ) = 5 f(5)-f(1)=5 , from this we find f ( 5 ) = 6 f(5)=6 . similarly we have f ( 13 ) = 14 f(13)=14 , f ( 31 ) = 32 f(31)=32 . now put n = 65 = 13 5 n=65=13*5 , we get: f ( 65 ) f ( 5 ) f ( 13 ) + f ( 1 ) = 65 f(65)-f(5)-f(13)+f(1)=65 . from this we get f ( 65 ) = 84 f(65)=84 , it seems so multiplicative! similarly we have f ( 155 ) = 192 f(155)=192 , f ( 403 ) = 448 f(403)=448 . put n=2015 to get that [f(2015)-f(65)-f(155)-f(403)+f(5)+f(13)+f(31)-f(1)=2015). plug in values to find f ( 2015 ) = 2688 f(2015)=\boxed{2688} . as for wether a function like this exists or not, try f ( n ) = n p n ( 1 + 1 p ) f(n)=n\prod_{p|n} \left(1+\dfrac{1}{p}\right) . this is just a function satisfying, but any other function satisfying the equation will also have f(2015)=2688.

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...