Maths Olympiad Prep

Library / Turkish Mathematical Olympiad /43 of 47

Turkish Mathematical Olympiad

Number theory Difficulty 7.6 National Olympiad, round 2 Prove it Turkey

The sequence a1,a2,…a_1, a_2, \dots satisfies the equality ∑i=1na⌊ni⌋=n10\sum_{i=1}^{n} a_{\lfloor \frac{n}{i} \rfloor} = n^{10} for each positive integer nn. Given an arbitrary positive integer cc, prove that can−can−1n\frac{c^{a_n} - c^{a_{n-1}}}{n} is an integer for each n>1n > 1.

Solution

For any n>1n > 1, we can write the given equality for nn and n−1n-1, then subtract them side by side to obtain the following:
∑i=1n−1(a⌊ni⌋−a⌊n−1i⌋)+a1=n10−(n−1)10. \sum_{i=1}^{n-1} \left( a_{\lfloor \frac{n}{i} \rfloor} - a_{\lfloor \frac{n-1}{i} \rfloor} \right) + a_1 = n^{10} - (n-1)^{10}.
Defining a new sequence by b1=a1=1b_1 = a_1 = 1 and bn=an−an−1b_n = a_n - a_{n-1} for each n>1n > 1, we can express the equality we obtained in the following form:
∑i∣nbni=n10−(n−1)10⇒∑i∣nbi=n10−(n−1)10. \sum_{i|n} b_{\frac{n}{i}} = n^{10} - (n-1)^{10} \Rightarrow \sum_{i|n} b_i = n^{10} - (n-1)^{10}.
It is easy to see that this equality holds for n=1n = 1 as well as n>1n > 1. Now, recall the definition of the well-known μ\mu-function:
μ(n)={(−1)kn is square-free and the number of prime divisors of n is k0n is not square-free \mu(n) = \begin{cases} (-1)^k & n \text{ is square-free and the number of prime divisors of } n \text{ is } k \\ 0 & n \text{ is not square-free} \end{cases}
Also recall the following well-known theorem:
Theorem. Let x1,x2,…x_1, x_2, \dots be a sequence and let yn=∑i∣nxiy_n = \sum_{i|n} x_i for each positive integer nn. Then,
xn=∑i∣nμ(i)⋅yni x_n = \sum_{i|n} \mu(i) \cdot y_{\frac{n}{i}}
holds for each positive integer nn. If n>1n > 1, denoting the prime divisors of nn by p1,p2,…,pkp_1, p_2, \dots, p_k, we can express the equality more explicitly as follows:
xn=yn−ynp1−⋯−ynpk+ynp1p2+⋯+ynpk−1pk−… x_n = y_n - y_{\frac{n}{p_1}} - \dots - y_{\frac{n}{p_k}} + y_{\frac{n}{p_1 p_2}} + \dots + y_{\frac{n}{p_{k-1} p_k}} - \dots
Now, we know that
∑i∣nbi=n10−(n−1)10=(10n9−45n8+⋯−45n2+10n−1). \sum_{i|n} b_i = n^{10} - (n-1)^{10} = (10n^9 - 45n^8 + \dots - 45n^2 + 10n - 1).
Therefore, by the theorem above, we find that
bn=∑i∣nμ(i)⋅((ni)10−(ni−1)10) b_n = \sum_{i|n} \mu(i) \cdot \left( \left( \frac{n}{i} \right)^{10} - \left( \frac{n}{i} - 1 \right)^{10} \right)
=10⋅∑i∣nμ(i)⋅(ni)9−45⋅∑i∣nμ(i)⋅(ni)8+⋯+10⋅∑i∣nμ(i)⋅(ni)−∑i∣nμ(i). = 10 \cdot \sum_{i|n} \mu(i) \cdot \left(\frac{n}{i}\right)^9 - 45 \cdot \sum_{i|n} \mu(i) \cdot \left(\frac{n}{i}\right)^8 + \dots + 10 \cdot \sum_{i|n} \mu(i) \cdot \left(\frac{n}{i}\right) - \sum_{i|n} \mu(i).
For n>1n > 1, denoting the prime divisors of nn by p1,p2,…,pkp_1, p_2, \dots, p_k, we see that the following holds for each positive integer rr:
∑i∣nμ(i)⋅(ni)r=nr⋅(1−1p1r)⋯(1−1pkr)=ϕ(n)⋅nr−1⋅(1+1p1+⋯+1p1r−1)⋯(1+1pk+⋯+1pkr−1). \begin{aligned} \sum_{i|n} \mu(i) \cdot \left(\frac{n}{i}\right)^r &= n^r \cdot \left(1 - \frac{1}{p_1^r}\right) \cdots \left(1 - \frac{1}{p_k^r}\right) \\ &= \phi(n) \cdot n^{r-1} \cdot \left(1 + \frac{1}{p_1} + \dots + \frac{1}{p_1^{r-1}}\right) \cdots \left(1 + \frac{1}{p_k} + \dots + \frac{1}{p_k^{r-1}}\right). \end{aligned}
Furthermore, ∑i∣nμ(i)=0\sum_{i|n} \mu(i) = 0 since n>1n > 1, therefore we obtain the following:
bnϕ(n)=10n8∏(1+1p+⋯+1p8)−45n7∏(1+1p+⋯+1p7)+⋯−45n∏(1+1p)+10. \begin{aligned} \frac{b_n}{\phi(n)} &= 10n^8 \prod \left(1 + \frac{1}{p} + \dots + \frac{1}{p^8}\right) \\ &\quad -45n^7 \prod \left(1 + \frac{1}{p} + \dots + \frac{1}{p^7}\right) + \dots -45n \prod \left(1 + \frac{1}{p}\right) + 10. \end{aligned}
We draw the following two conclusions from this formula:
(i) For each positive integer nn, one has ϕ(n)∣bn\phi(n) \mid b_n. This is obvious for n=1n = 1 and follows from the formula for n>1n > 1.
(ii) For each positive integer nn, one has bn>0b_n > 0, thus an≥na_n \ge n. Because,
b1=1>0, b2=210−2>0, b3=310−210−1>0, b4=410−310−210+1>0 b_1 = 1 > 0,\ b_2 = 2^{10} - 2 > 0,\ b_3 = 3^{10} - 2^{10} - 1 > 0,\ b_4 = 4^{10} - 3^{10} - 2^{10} + 1 > 0
and for n≥5n \ge 5,
bnϕ(n)>n7⋅(10n−45)∏(1+1p+⋯+1p7)+…+n⋅(120n−45)∏(1+1p)+10>0⇒bn>0. \begin{aligned} \frac{b_n}{\phi(n)} &> n^7 \cdot (10n - 45) \prod \left(1 + \frac{1}{p} + \dots + \frac{1}{p^7}\right) + \dots \\ &\quad + n \cdot (120n - 45) \prod \left(1 + \frac{1}{p}\right) + 10 > 0 \Rightarrow b_n > 0. \end{aligned}

For n>1n > 1, separate the prime divisors of nn into two subsets as those which divide cc and those which do not divide cc, hence write n=m⋅sn = m \cdot s. Here, mm divides sufficiently large powers of cc and ss is coprime to cc. Hence, m∣cn−1∣can−1m \mid c^{n-1} \mid c^{a_{n-1}} since an−1≥n−1a_{n-1} \ge n-1 and s∣cϕ(s)−1∣cϕ(n)−1∣cbn−1s \mid c^{\phi(s)-1} \mid c^{\phi(n)-1} \mid c^{b_n-1} since ϕ(n)∣bn\phi(n) \mid b_n. Therefore, n∣can−1⋅(cbn−1)=can−can−1n \mid c^{a_{n-1}} \cdot (c^{b_n-1}) = c^{a_n} - c^{a_{n-1}}.

Want a route through all this instead of an archive? The track puts 2,604 problems in a working order, from Junior Challenge level to the IMO shortlist.

Source: MathNet, licensed CC-BY-4.0. Statement reproduced verbatim; metadata (topic, difficulty) added by this project.