3

Determine the general term of the sequence $(a_n)_{n\ge1}$, strictly decreasing, of strictly positive numbers, which satisfies the properties:

a) $na_n \in \mathbb{N} \setminus \{0\}$ for every $n \in \mathbb{N} \setminus \{0\}$.

b) $\sum_{k=1}^{n} (-1)^{k-1} a_k \binom{n}{k} = \sum_{k=1}^{n} a_k$.

I tried taking $n=1,2,3,4$ to use induction but I couldn't find a pattern. Any ideas on how to approach it?

My approach: I'm pretty sure I made a lot of mistakes but this may help finding the correct proof.
Let's denote the general term of the sequence $(a_n)_{n\geq1}$ as $a_n = \frac{c}{n}$, where $c$ is some positive constant.

First, let's consider property a): $$ na_n = n \cdot \frac{c}{n} = c. $$ So, $na_n$ is always a positive integer, as required.
Now, let's consider property b). We need to find $c$ such that: $$ \sum_{k=1}^{n} (-1)^{k-1} a_k \binom{n}{k} = \sum_{k=1}^{n} a_k. $$ We substitute $a_k = \frac{c}{k}$ into the given equation: $$ \sum_{k=1}^{n} (-1)^{k-1} \left(\frac{c}{k}\right) \binom{n}{k} = \sum_{k=1}^{n} \frac{c}{k}. $$ For convenience, let's rewrite the binomial coefficient as: $$ \binom{n}{k} = \frac{n!}{k!(n-k)!}$$ Substituting this into our equation: $$ \sum_{k=1}^{n} (-1)^{k-1} \left(\frac{c}{k}\right) \frac{n!}{k!(n-k)!} = \sum_{k=1}^{n} \frac{c}{k}. $$ Rearranging the terms, we get: $$ \begin{split} c&\sum_{k=1}^{n} (-1)^{k-1} \frac{n!}{k!(n-k)!k} = c\sum_{k=1}^{n} \frac{1}{k} \\ c&\sum_{k=1}^{n} (-1)^{k-1} \frac{n!}{k!(n-k)!k} - c\sum_{k=1}^{n} \frac{1}{k} = 0 \\ c&\sum_{k=1}^{n} \frac{(-1)^{k-1}n!}{k!(n-k)!k} - c\sum_{k=1}^{n} \frac{1}{k} = 0 \end{split} $$ Now, let's factor out $c$: $$ c\left(\sum_{k=1}^{n} \frac{(-1)^{k-1}n!}{k!(n-k)!k} - \sum_{k=1}^{n} \frac{1}{k^2}\right) = 0 \iff \sum_{k=1}^{n} \frac{(-1)^{k-1}n!}{k!(n-k)!k} = \sum_{k=1}^{n} \frac{1}{k} $$ Given that this equation must hold for all $n$, we can conclude that the sum of the coefficients in both series must be equal. $$ \begin{split} \frac{(-1)^{k-1}n!}{k!(n-k)!k} &= \frac{1}{k} \\ (-1)^{k-1}n! &= \frac{(n-k)!k^2}{(n-k)!} \\ (-1)^{k-1}n! &= k^2 \end{split} $$ Since this equation should hold for all $k$, the only way this can happen is if $n! = k^2$. This is not possible for all $n$ and $k$, so we need to revise our assumption about $a_n$.

The correct approach is to consider a sequence such that $a_n = \frac{c}{n^k}$ for some positive integer $k$ and a positive constant $c$. We will explore what value of $k$ would make the sequence satisfy both properties.

For property a): $ na_n = n \cdot \frac{c}{n^k} = c \cdot n^{1-k} $ For $c \cdot n^{1-k}$ to be a positive integer, $1 - k$ must be zero, meaning $k = 1$.

So, $a_n = \frac{c}{n}$ seems to be the correct form for satisfying property (a).

Now, let's check if this satisfies property (b). $$ \sum_{k=1}^{n} (-1)^{k-1} a_k \binom{n}{k} = \sum_{k=1}^{n} a_k. $$ Substituting $a_k = \frac{c}{k}$: $$ \sum_{k=1}^{n} (-1)^{k-1} \left(\frac{c}{k}\right) \binom{n}{k} = \sum_{k=1}^{n} \frac{c}{k}. $$ This is the same equation we started with. Since we already proved that $a_n = \frac{c}{n}$ satisfies property a), we conclude that it also satisfies property b).

Therefore, the general term of the sequence $(a_n)_{n\ge1}$ is $a_n = \frac{c}{n}$, where $c$ is a positive constant.

Source: Romanian National Olympiad 2022 shortlist problem

  • Condition a is useful as it gives $a_n$ is rational, and particularly with a denominator that divides $n$. – D S Mar 27 '24 at 06:46
  • To the OP: I haven't tried this out, but I feel with condition b, you can solve for $a_i$ in terms of $a_1$. – D S Mar 27 '24 at 06:50
  • @DS I only got a1=2a2 but after that I couldn’t find the rest – math.enthusiast9 Mar 27 '24 at 07:57
  • @math.enthusiast9 for $n=4,5$ you get two equations in terms of $a_1,a_2,a_3,a_4$, but you already know $a_2 = 0.5a_1$. Unless they're equivalent (again, I haven't tried) you can solve for them. – D S Mar 27 '24 at 08:39
  • Also, can you make a guess from condition a? Your question will be closed if you don't provide more context. – D S Mar 27 '24 at 08:41
  • 2
    Experiments with Mathematica shows that this system of equation is highly undetermined. Indeed, if we are willing to find a solution of the form $a_n=F_n(a_1,\ldots,a_{n-1})$, then the experiments suggests that the values of $a_2, a_4, a_6, \ldots$ are uniquely determined for any given choice of the values of $a_1, a_3, a_5, \ldots$. This means that the other conditions on $(a_n)$ must play a crucial role. (Note: If we only require that $(a_n)$ is non-increasing instead of strictly increasing, then we have a solution of the form $a_n=\frac{1}{2}a_1$ for $n\geq 2$.) – Sangchul Lee Mar 27 '24 at 09:01
  • 1
    Possible Duplicate : You can easily show this $a_n$ works, but hard to show there are no other solutions apart from integer multiples of this. – D S Mar 27 '24 at 09:03
  • @SangchulLee but we also have the decreasing condition and rational condition. We can easily show $a_n = \frac{k}{n}$ for any $k \in \Bbb N$, but difficult to show other solutions are (im)possible. – D S Mar 27 '24 at 09:10
  • 2
    @DS, That will be very hard, because the following solution $$a_n=\alpha(1+\mathbf{1}{{k=1}}) + \frac{\beta}{k}$$ solves the problem for any $\alpha\in\mathbb{N}{\geq 0}$ and $\beta\in\mathbb{N}{>0}$. (Indeed, this $(a_n)$ is strictly decreasing, positive, and satisfies $na_n\in\mathbb{N}{>0}$ together with the given system of equations.) I am not sure if there are other types of solutions, though. – Sangchul Lee Mar 27 '24 at 09:11
  • @DS "for =4,5 you get two equations in terms of 1,2,3,4, but you already know 2=0.51. Unless they're equivalent (again, I haven't tried) you can solve for them." I tried doing this but they are equivalent – math.enthusiast9 Mar 27 '24 at 09:19
  • Not sure if it will ever help, but writing $f(x)=\sum_{n\geq 1}a_n x^n$ for the (formal) generating function of $(a_n)$, the system of equations is equivalent to $$f(x)+f(\mu(x))=0, \qquad \mu(x)=-\frac{x}{1-x}.$$ Since $\mu(\mu(x))=x$, substituting $x\mapsto\mu(x)$ in the equation above reveals no further information. – Sangchul Lee Mar 27 '24 at 09:35
  • 2
    To OP: I understand that this is indeed a challenging problem. However, you could still provide additional context, including sources, related areas, etc., to increase the likelihood of receiving an answer. – Sangchul Lee Mar 27 '24 at 09:39

0 Answers0