Time for some induction

k = 1 n n k ϕ ( k ) \large \sum_{k=1}^n \left \lfloor \frac nk \right \rfloor \phi(k)

Let ϕ \phi denote the Euler's Totient Function . If the summation above is equal to n ( n + A ) B \dfrac{n(n+A)}B , where A A and B B are integers, find the value of A + B A+B .

1 2 3 4 7 5 6

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.

3 solutions

Ivan Koswara
Nov 16, 2015

Combinatorial proof:

Consider the set S S of points ( x , y ) (x,y) on the plane where 1 y x n 1 \le y \le x \le n and x , y x,y are positive integers.

Consider a line y = m k x y = \frac{m}{k} x where 1 m k n 1 \le m \le k \le n and m , k m,k are coprime. This line passes exactly t = n k t = \left\lfloor \frac{n}{k} \right\rfloor points of S S , namely ( k , m ) , ( 2 k , 2 m ) , ( 3 k , 3 m ) , , ( t k , t m ) (k,m), (2k,2m), (3k,3m), \ldots, (tk,tm) . ( x x has to be a positive integer in order for the point to be an element of S S , and moreover a multiple of k k so that y y is an integer, but any such positive integer not greater than n n works and gives a unique point for S S .) Let S k S_k be the set of all points of S S lying on a line in the form y = m k x y = \frac{m}{k} x where 1 m k 1 \le m \le k , m , k m,k are coprime. There are ϕ ( k ) \phi(k) choices of m m by definition of ϕ \phi , and each of these gives n k \left\lfloor \frac{n}{k} \right\rfloor points, so S k S_k contains n k ϕ ( k ) \left\lfloor \frac{n}{k} \right\rfloor \phi(k) points.

Now, consider S 1 , S 2 , , S n S_1, S_2, \ldots, S_n . We will show that every point of S S is in at least one of the sets. Consider a point ( x 0 , y 0 ) S (x_0, y_0) \in S . This point lies on the line y = y 0 x 0 x y = \frac{y_0}{x_0} x where 1 y 0 x 0 n 1 \le y_0 \le x_0 \le n . Suppose x 0 = c x 1 , y 0 = c y 1 x_0 = cx_1, y_0 = cy_1 , where x 1 , y 1 x_1, y_1 are coprime; then simplifying the fraction gives that ( x 0 , y 0 ) (x_0, y_0) lies on the line y = y 1 x 1 x y = \frac{y_1}{x_1} x . By definition of S i S_i , this means ( x 0 , y 0 ) (x_0, y_0) is in S x 1 S_{x_1} . Thus every point of S S is in at least one of the sets.

We will also show that, on the other hand, every point of S S is in at most one of the sets. Consider a point ( x 0 , y 0 ) S (x_0, y_0) \in S . Suppose for the sake of contradiction that this point is in two sets S i , S j S_i, S_j ; without loss of generality, suppose i < j i < j . Then there exists m i , m j m_i, m_j such that y 0 = m i i x 0 y_0 = \frac{m_i}{i} x_0 and y 0 = m j j x 0 y_0 = \frac{m_j}{j} x_0 , where m i , i m_i, i are coprime, and m j , j m_j, j are coprime. This implies y 0 x 0 = m i i = m j j \frac{y_0}{x_0} = \frac{m_i}{i} = \frac{m_j}{j} . But a fraction only has one reduced form (with a positive denominator), while if m i , i m_i, i are coprime and m j , j m_j, j are coprime, both of m i i , m j j \frac{m_i}{i}, \frac{m_j}{j} are reduced fractions, contradiction. Thus no point of S S is in more than one of the sets.

In total, this means each point of S S is in exactly one of the sets S 1 , S 2 , , S n S_1, S_2, \ldots, S_n . Thus S i = S = 1 + 2 + + n = n ( n + 1 ) 2 \sum |S_i| = |S| = 1+2+\ldots+n = \frac{n(n+1)}{2} . But S i = n k ϕ ( k ) |S_i| = \left\lfloor \frac{n}{k} \right\rfloor \phi(k) , so S i \sum |S_i| is the given expression. Thus A = 1 , B = 2 A = 1, B = 2 , and A + B = 3 A+B = \boxed{3} .

Moderator note:

Nice combinatorial bijection that you set up.

We shall prove that the statement is true for n=1: k = 1 1 1 k ϕ ( k ) = 1 × ϕ ( 1 ) = 1 = 1 ( 1 + 1 ) 2 \sum_{k=1}^1\lfloor{\frac{1}{k}}\rfloor \phi(k)=1\times \phi(1)=1=\frac{1(1+1)}{2} , which is indeed true.

Taking the inductive step, we now assume that the statement is true up to a number n, and then we will prove it true for the number n+1. We observe that: k = 1 n n k ϕ ( k ) = k = 1 n k ϕ ( k ) \sum_{k=1}^n\lfloor{\frac{n}{k}}\rfloor \phi(k)=\sum_{k=1}^\infty\lfloor{\frac{n}{k}}\rfloor \phi(k) because n k = 0 \lfloor{\frac{n}{k}}\rfloor=0 for every k n + 1 k\geq{n+1}

Then using the fact that every number n for any given k can be written in the form: n = k q ( n , k ) + r ( n , k ) , 0 r ( n , k ) < k n=kq(n,k)+r(n,k),\ 0\leq r(n,k)<k , we can set out to prove that: n k = k q + r k = q + r k = q ( n , k ) \lfloor\frac{n}{k}\rfloor=\lfloor\frac{kq+r}{k}\rfloor=\lfloor q+\frac{r}{k}\rfloor=q(n,k) , since r k < 1 \frac{r}{k}<1 and q is an integer.

In the light of the above, the statement for the number n+1 reads : k = 1 n + 1 k ϕ ( k ) = k = 1 q ( n + 1 , k ) ϕ ( k ) = k = 1 ( q ( n , k ) + r ( n , k ) + 1 k ) ϕ ( k ) \sum_{k=1}^\infty\lfloor{\frac{n+1}{k}}\rfloor \phi(k)=\sum_{k=1}^\infty q(n+1,k)\phi(k)=\sum_{k=1}^\infty (q(n,k)+\lfloor \frac{r(n,k)+1}{k}\rfloor)\phi(k)

However, we note that : r ( n , k ) + 1 k = { 1 if n + 1 0 m o d k 0 all other cases \lfloor \frac{r(n,k)+1}{k}\rfloor=\begin{cases}1 & \quad \text{if } n+1\equiv 0\ mod \ {k}\\ 0 & \quad \text{all other cases} \\ \end{cases}

We conclude that the second part of sum contributes only whenever k|n+1 and thus : k = 1 n + 1 k ϕ ( k ) = k = 1 n k ϕ ( k ) + k n + 1 ϕ ( k ) = n ( n + 1 ) 2 + n + 1 = ( n + 1 ) ( n + 2 ) 2 \sum_{k=1}^\infty\lfloor{\frac{n+1}{k}}\rfloor \phi(k)=\sum_{k=1}^\infty\lfloor{\frac{n}{k}}\rfloor \phi(k)+\sum_{k|n+1}\phi(k)=\frac{n(n+1)}{2}+n+1=\frac{(n+1)(n+2)}{2} , which concludes the proof.

In order to prove that k n ϕ ( k ) = n \sum_{k|n}\phi(k)=n we first calculate for p a prime number: p a n ϕ ( k ) = k = 0 a 1 ϕ ( p k ) = ( 1 1 p ) k = 1 a 1 p k + 1 = ( 1 1 p ) p p 1 ( p a 1 ) + 1 = p a \sum_{p^{a}|n}\phi(k)=\sum_{k=0}^{a-1}\phi(p^k) =(1-\frac{1}{p}) \sum_{k=1}^{a-1}p^k+1=(1-\frac{1}{p})\frac{p}{p-1}(p^{a}-1) +1=p^a

and then we assert that: k p 1 a 1 . . . p l a l ϕ ( k ) = k p 1 a 1 o r . . . o r k p l a l ϕ ( k ) = i = 1 l p i a i n ϕ ( k ) = i = 1 l p i a i \sum_{k|p_{1}^{a_{1}}...p_{l}^{a_{l}}}\phi(k)=\sum_{k|p_{1}^{a_{1}}\hspace{0.1cm} or ... or \hspace{0.1 cm} k| p_{l}^{a_{l}}}\phi(k)=\prod_{i=1}^{l}\sum_{p_{i}^{a_{i}}|n}\phi(k)=\prod_{i=1}^{l}p_{i}^{a_{i}} which means that for generic n = i = 1 l p i a i n=\prod_{i=1}^{l}p_{i}^{a_{i}} , the lemma holds.

Abe Vallerian
Nov 16, 2015

Sorry, this is not a proper solution, but I think this is quite interesting guess.

Notice that n ( n + A ) B \frac{n(n+A)}{B} must have an integer solution. Since n n is an arbitrary integer, then B B has to be either 1 1 or 2 2 ; otherwise, n ( n + A ) B \frac{n(n+A)}{B} is not integer. Next, we will show that B B can't be 1 1 plug-in n = 1 n=1 and n = 2 n=2 for example. Thus, B = 1 B=1 . A A can be easily obtained afterwards.

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...