2

Problem:

$f(n) = 10^n-(5+\sqrt{17})^n-(5-\sqrt{17})^n$ is a function which is valid for all integers $n\geq1$. Prove that $f(n)$ is always perfectly divisible by $2^{n+1}$.

My approach:

I have used the theory of recurrence relations and induction.

The general form of third order homogeneous recurrence relation is $$c_1a_n+c_2a_{n-1}+c_3a_{n-2}+c_4a_{n-3}=0$$

If we change $a_n$ to $x^n$, we get $$c_1x^n+c_2x^{n-1}+c_3x^{n-2}+c_4x^{n-3}=0$$

If we want to get the characteristic equation of it, the equation will be $c_1x^3+c_2x_2+c_3x+c_4=0$.

If we solve the equation, we will get three solutions $k_1$, $k_2$, $k_3$.

Then $a_n=Ak_1^n+Bk_2^n+Ck_3^n$ where $A$, $B$ and $C$ are constants according to the equation.

$$\begin{equation}\tag{i}a_n=Ak_1^n+Bk_2^n+Ck_3^n\end{equation}$$

Let $$\begin{equation}\tag{ii}a_n=10^n-(5+\sqrt{17})^n-(5-\sqrt{17})^n\end{equation}$$

If we compare $(\mathrm{i})$ and $(\mathrm{ii})$, we get $A=1$, $B=-1$, $C=-1$. If we calculate $k_1=10$, $k_2=5+\sqrt{17}$, $k_3=5-\sqrt{17}$

So, the polynomial will be (in terms of $b$) $$(b-10)[b-(5+\sqrt{17})][b-(5-\sqrt{17})]=0$$ or $b^3-20b^2+108b-80=0$.

If we multiply the equation by $b^{n-3}$ and replace $b^n$ by $a_n$, it will give the result of $$a_n=20a_{n-1}-108a_{n-2}+80a_{n-3}$$ or $$a_{n+1}=20a_{n}-108a_{n-1}+80a_{n-2}$$

We will use strong induction now.

Base case:

$2^{2+1}$ divides $f(2)$.

Induction step:

Suppose it is true for all the way up to $k+1$. Then, $2^{k+1} \mid a_n$, $2^{k} \mid a_{n-1}$ and $2^{k-1} \mid a_{n-2}$. Therefore $a_{n}=2^{k+1}m$, $a_{n-1}=2^{k}n$ and $a_{n-2}=2^{k-1}p$ where $m$, $n$ and $p$ are integers.

If we input the values of $a_{n}$, $a_{n-1}$, $a_{n-2}$ in $a_{n+1}=20a_{n}-108a_{n-1}+80a_{n-2}$, we will get $a_{n+1}=2^{k+2}(10m-27n+10p)$, and so $2^{k+2} \mid a_{n+1}$. Therefore $2^{n+1} \mid f(n)$ for all positive integers. $\square$

Conclusion: I don't like this proof because it includes recurrence relation.I don't want to solve it using recurrence relation. Feel free to help me.

Source: Problem 8, BDMO (Bangladesh Mathematical Olympiad) 2024, junior, national

  • Umm, what are you trying to do? Induction? –  Aug 17 '24 at 08:06
  • What is the question you wish to ask in your post? For a solution-verification question to be on topic you must specify precisely which step in the proof you question, and why so. This site is not meant to be used as a proof checking machine. It is not clear what you mean in your prior comment. – Bill Dubuque Aug 17 '24 at 08:07
  • @Angelo Because the same methods apply here (and are already explained in many prior answers). I am waiting for the OP to clarify what they seek before closing it as a dupe. – Bill Dubuque Aug 17 '24 at 08:19
  • @BillDubuque, he wants to prove it without using the recurrence relation $a_n=20a_{n-1}-108a_{n-2}+80a_{n-3}$ – Angelo Aug 17 '24 at 08:24
  • I didn't get it similar https://math.stackexchange.com/questions/48508/how-are-the-integral-parts-of-9-4-sqrt5n-and-9-%e2%88%92-4-sqrt5n-relate –  Aug 17 '24 at 09:47
  • In the induction step, $k=n$? – Bob Dobbs Aug 17 '24 at 11:08
  • Do we need to exclude $n=1$? The function is zero at that argument, and every nonzero integer is surely divisible into (as distinct from by) zero, no? – Oscar Lanzi Aug 17 '24 at 11:44
  • 1
    @Math12, why have you approved Bob Dobbs’ answer if he uses a recurrence relation which is $p_n=2(5p_{n-1}-4p_{n-2})$ ? You said that you do not want to solve your problem by using recurrence relations. – Angelo Aug 17 '24 at 12:33
  • Where did I approved?I didn't approved anyone's answer. –  Aug 27 '24 at 12:36

2 Answers2

4

A Freshman's $\rm\color{#0af}{Dream}$ Binomial Theorem yields a simple proof: cancelling $\:\!2^n$ reduces it to

$\qquad\quad\ \ \exists\:\!j\in\Bbb Z\!:\ a^n\! + b^n\ \ \, =\,\ \ \color{#0a0}5^n\ +\ \color{#e80}2\,j\in\Bbb Z,\,$ for $\,a,b = \frac{5\pm\sqrt{17}}2,\,$ $\small \begin{align}\color{#0a0}{a\!+\!b = 5}\\ \color{#e80}{ab = 2}\end{align}\:\!;\ $ it's true by ${\rm\color{#0af}{Dream}}\ \ \exists\:\!j\in\Bbb Z^{\phantom{|^{|}}}\!\!\!\!:\ a^n\! + b^n\! = (\color{#0a0}{a\!+\!b})^n \!+\!\color{#e80}{ab}j\in \Bbb Z,\,$ if $\ \color{#0a0}{a\!+\!b},\color{#e80}{ab} \in \Bbb Z,\,$ by $\rm\color{#c00}{induction},$ with step: $\qquad\quad\ \ a^{n+1}\!+b^{n+1} = (a\!+\!b)\:\!\overbrace{(a^n\!+b^n)}^{\large \!\!\!\color{#c00}{(a+b)^n\:\!\:\!+\:\color{#e80}{ab}^{\phantom{|^|}}\!\!j}\!\!\!} $ ${-}\:\color{#e80}{ab}\:\!(\overbrace{a^{n-1}\!+b^{n-1}}^{\color{#c00}{\large k\ \in\ \Bbb Z}}) \,=\, \overbrace{(\color{#0a0}{a\!+\!b})^{n+1}\!+\color{#e80}{ab}\:\!\color{#90f}J}^{ (a+b)\ \color{#c00}{j\,\ -\,\ k}\,\ =\:\!:\,\ \color{#90f}{J}\ \in\ \Bbb Z\!\!\!\!\!\!\!\!\!}\ \ \ \ \bf\small QED$

Bill Dubuque
  • 282,220
  • This shows how closely related are proofs using the binomial theorem and power sum recurrence. We could prove the Lemma w/o explicitly using the recurrence, but this way is more enlightening. See here for a conceptual proof of the recurrence. – Bill Dubuque Aug 18 '24 at 03:03
  • This is really a genious solution. +1 – Bob Dobbs Aug 18 '24 at 03:05
  • What is Freshman's dream? letting $$(a+b)^n=a^n+b^n$$. But that is wrong and i know about binomial theorem but binomial theorem is not that short. All in all, you worked with freshman's dream binomial theorem. Is it the combination of two theory or something else? Could please explain it to me? –  Aug 18 '24 at 12:56
  • @Math12 This Dream states that if both the sum and product of $,a,b,$ are integers then for all $,n\ge 1,$ we have $,a^n+b^n,$ is an integer of form $,(a+b)^n + k_n,\color{#0af}{ab},,$ for some integer $,k_n.\ \ $ – Bill Dubuque Aug 18 '24 at 17:49
  • Could you please explain it a little bit? –  Aug 20 '24 at 17:18
  • @Math12 Please be more precise about what is not clear to you. – Bill Dubuque Aug 22 '24 at 18:41
  • How did you write $a^n+b^n \equiv 5^n\mod{2}$? –  Aug 25 '24 at 07:07
  • @Math12 As I said "Cancel $2^n,\ldots,,$ i.e. $$\begin{align} 2^{n+1}&\mid 10^n-(2a)^n-(2b)^n\ \overset{\large \div\ 2^n_{\phantom{,}}}\iff\ \ \ \ \ \ \ \ , 2,&\mid\ \ 5^n\ - ,\ a^n,\ -,\ b^n\[.5em] \iff \bmod 2!:!&\ \ \ \ \ , 5^n,\equiv, a^n\ ,+,\ b^n \end{align}\qquad$$ – Bill Dubuque Aug 25 '24 at 07:59
  • How did you write $a^n+b^n \equiv (a+b)^n \pmod{ab}$? –  Aug 25 '24 at 13:24
  • @Math12 We prove by strong induction that for all $,n\ge 1!:,$ $ a^n+b^n = (a+b)^n + j,ab,\ {\rm some}\ j\in\Bbb Z.,$ Base case $,n=1,$ is clear. The induction step follows by the final displayed equation, iwhich shows that $,\color{#c00}{P(n),\ P(n!-!1)\Rightarrow} P(n!+!1).\ \ $ – Bill Dubuque Aug 25 '24 at 23:30
  • How did you write $a^{n+1}+b^{n+1}=(a+b)\cdot (a^n+b^n)-ab\cdot (a^{n-1}+b^{n-1})$? –  Aug 26 '24 at 07:23
  • @Math12 It is trivial algebra to verify. I gave a link in my first comment for a conceptual derivation. – Bill Dubuque Aug 26 '24 at 08:12
  • Or, simpler than the above 2nd-order power sum recurrence is this 1st-order (power) recurrence $$\begin{align}(a+b)^{n+1} &= (a+b)(a+b)^n\ &= (a+b)(a^n+b^n+j,ab),\ j\in\Bbb Z,,\ \rm by\ induction\ &= a^{n+1}+b^{n+1} + ab,(\underbrace{a^{n-1}+b^{n-1}+j,(a+b)}_{=\ J\ \in\ \Bbb Z,\ \rm by\ induction})\end{align}\qquad\qquad$$

    Or we could show $,xy\mid f(x+y)-f(x)-f(y),$ for a polynomial $,f(t)\in \Bbb Z[t],$ with $,f(0)=0,,$ then specialize to the power case $,f(t)=t^n.\ \ $

    – Bill Dubuque Dec 01 '24 at 10:23
0

Have you seen these formulae?: $$p_n=p_{n-1}e_1-p_{n-2}e_2+p_{n-3}e_3-...\tag1$$ Let $x_1=w=5+\sqrt{17}$ and $x_2=\bar w=5-\sqrt{17}.$ Then we have two elementary symmetric polynomials: $e_1=x_1+x_2=10$, $e_2=x_1x_2=8.$ Power sum symmetric polynomials are $$p_n=x_1^n+x_2^n,\,\,n\geq0.$$ First examples are $p_0=2,$ $p_1=e_1=10,$ $p_2=84.$

By $(1)$, we have the recursive equation $p_n=2(5p_{n-1}-4p_{n-2}).$ İt is not difficult to show that $p_n=2^n\times \text{odd}$ for $n\geq1$ by induction. Hence, $2^{n+1}$ will divide $10^n-p_n$ for positive integer $n$.

Bob Dobbs
  • 15,712
  • 1
    @BobDobbs, OP said that he does not want to solve it by using recurrence relations. – Angelo Aug 17 '24 at 12:36
  • @Angelo I dont think we can avoid recursion. Binomial expansion is recursive. Can you write an answer without. recursion? It must be very complicated. – Bob Dobbs Aug 17 '24 at 13:15
  • 1
    @BobDobbs, I do have a solution without recursion where I use only induction but this is a duplicate question and I cannot post an answer. – Angelo Aug 17 '24 at 15:16
  • @Angelo You can share the link here. Maybe it has an hidden recursion. Are you sure it is recursion free? – Bob Dobbs Aug 17 '24 at 15:22
  • You can search the title. This is not a duplicate question. –  Aug 17 '24 at 15:48
  • 1
    Do you have other solution? I don't need a solution related with recurrence relation. –  Aug 17 '24 at 15:58
  • I can write a solution without the word "recurruence. " But, isnt induction a recurruence? @Math12 – Bob Dobbs Aug 17 '24 at 16:35
  • This is recurrence related topic but not the same. You can use induction but not direct recurrence relation. –  Aug 17 '24 at 17:24
  • I am used to NG identities that's why I posted a solution related to it. Luckly, they didn't downvote. Your recurrence is third order. Mine is second. Simpler. @Math12 – Bob Dobbs Aug 17 '24 at 17:30