2

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

edit(12 JAN 2025)- when I posted my proof, I was particularly doubtful about my idea on which my proof was based, My primary aim is just get my idea verified by people at MSE. My proof: The given equation gave me the intuition of a third order recurrence relation solution (after trying for 55 minutes by induction :/ ), so from the given equation of $n$th term I created the characteristic equation of the recurrence relation which comes out to be $$ x^3 - 20x^2 + 108x - 80 = 0 $$

Now in recurrence relation form it can be rewritten as $$ a_{n+3} = 20a_{n+2} - 108a_{n+1} + 80a_n $$ Now to apply induction, where our inductive hypothesis is that every $a_n$ is divisible by $2^{n+1}$. We check the base cases, that is, $a_2=16$, $a_3=240$, and $a_4=3072$, which looks fine.

Then we assume its true for some $a_k$, $a_{k+1}$, and $a_{k+2}$, and then we get that $a_{k+3}$ is divisible by $2^{k+4}$ for some $k$ hence completing our induction. I have cut short the last part because from the equation only its quite visible.

Now is my proof correct? I don't have much experience in proof writing. Moreover, firstly I was trying to prove it using induction only but failed repetitively. As soon as this idea struck me, I published it here, so I apologise for all my errors.

If you have any helpful advice or suggestion please do drop down in the comments. Thanks!

K41
  • 1,662
  • 2
    Your proof works. $\quad$ Note that you could have started the induction at $f(1)$, so you didn't need to calculate $f(4)$. – Calvin Lin Jan 10 '25 at 17:12
  • @CalvinLin yah but in the question its given for all n>1 that is why i thought it would be safe to start from 2 – K41 Jan 10 '25 at 17:14
  • Agreed with your thinking. $\quad$ Sometimes I'm lazy and want to avoid calculation. EG I even considered starting at $f(0)$, but unfortunately that doesn't work out. – Calvin Lin Jan 10 '25 at 17:15
  • @J.W.Tanner till the third last step i was on same path but dont know how i messed up and didn't reach at the end.. :/ but anyway to be honest it gave me an insight that recurrence can be used here, I never saw something like this as a solution to reccurence. – K41 Jan 10 '25 at 17:24
  • Consider the numbers $f(n)/2^n$. For them, the characteristic equation is $(2x)^3 - 20(2x)^2 + 108(2x) - 80 = 0$, i.e. $x^3-10x^2+27x-10=0$. So, you only need to show that the sequence given by the recursion $a_{n+3} = 10a_{n+2} - 27a_{n+1} + 10a_n$ and the initial conditions $a_1=0/2^1=0$, $a_2=16/2^2=4$, $a_3=240/2^3=30$ contains only even entries. But that's easy by induction, since all the coefficients are integers and all 3 initial terms are even. – Alexander Burstein Jan 10 '25 at 22:21
  • @AlexanderBurstein right Mr.Alexander but yah the conclusion immediately follows as soon as we see the equation eithr ways. – K41 Jan 10 '25 at 23:44
  • By cancelling $2^n$ it's equivalent to $,2\mid 5^n-w^n-\bar w^n,\ w = (5!+!\sqrt{17})/2,,$ so it suffices to show $,w^n+\bar w^n,$ is an odd integer, which follows by methods in the dupe, e.g. both methods here work (i.e. using the power sum recurrence or quadratic integer parity). $\ \ $ – Bill Dubuque Jan 11 '25 at 21:06
  • 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. $\ \ $ – Bill Dubuque Jan 11 '25 at 21:08

1 Answers1

2

For what it's worth, after I finished this answer, I noticed the comment by J.W.Tanner, which links to a method similar to mine. Since my answer is somewhat different, I am leaving it undeleted.


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

Induction also can be made to work, although it is a little tricky.

Throughout this response, I will use $~M~$ to denote some (unspecified) odd positive integer, and $~E~$ to denote some (unspecified) even positive integer.

Let $~g(n)~$ denote $~\displaystyle (5 + \sqrt{17})^n - (5 - \sqrt{17})^n.$

$\underline{\text{Claim:}}$
$g(n) = 2^n \times M.$

Assuming that the claim is true, then you have that

$$10^n - g(n) = 2^n \times (5^n - M),$$

which is divisible by $2^{n+1}.$


$\underline{\text{Proof of Claim:}}$

By inspection, $~f(2) = 2^2 \times 21, ~f(3) = 2^3 \times 95, ~f(4) = 2^4 \times 433, ~f(5) = 2^5 \times 1975.$

Assume that the claim is true for $~n = N,~$ and $~n = N-2.$

Then,

$$g(N) \times g(2) = M \times 2^{N+2}$$

$$= \left[ ~(5 + \sqrt{17})^N + (5 - \sqrt{17})^N ~\right] \times \left[ ~(5 + \sqrt{17})^2 + (5 - \sqrt{17})^2 ~\right]$$

$$=g(N+2)$$

$$ + \left\{ ~\left[ ~(5 + \sqrt{17})^N \times (5 - \sqrt{17})^2 ~\right] + \left[ ~(5 - \sqrt{17})^N \times (5 + \sqrt{17})^2 ~\right]~\right\}$$

$$= g(N+2) + \left\{ ~g(N-2) \times \left[ (5 + \sqrt{17})^2 \times (5 - \sqrt{17})^2 ~\right] ~\right\}$$

$$= g(N+2) + \left[ ~2^{N-2} \times M \times 2^6 ~\right]$$

$$= g(N+2) + \left[ ~M \times 2^{N+4} ~\right]$$

$$= g(N+2) + \left[ ~E \times 2^{N+2} ~\right].$$

Therefore,

$$g(N+2) = M \times 2^{N+2} - E \times 2^{N+2} = M \times 2^{N+2}.$$

user2661923
  • 42,303
  • 3
  • 21
  • 46
  • 1
    +1, your proof is neat and more elegant than that which JW Tanner has given link to in comment. Thanks ! – K41 Jan 11 '25 at 01:28
  • Indeed $2^n|g(n)$:

    We may write:

    $17\equiv 1\bmod 2^4\Rightarrow \sqrt {17}\equiv 1\bmod 2^2$

    $A=(5+\sqrt{17})^n\equiv (1\bmod 2^2+1\bmod 2^2)^n\equiv 0\bmod 2^{2n}$

    $B=(5+\sqrt{17})^n(1\bmod 2^2-1\bmod 2^2)^2\equiv 0\bmod 2^{2n}$

    $\Rightarrow A+B\equiv 0\bmod 2^{2n}$

    That is:

    $2^n|g(n)=(5+\sqrt{17})+(5-\sqrt{17})$

    – sirous Jan 11 '25 at 14:00
  • @sirous interesting about modular arithmetic around $~\sqrt{17}.~$ Assumed typo in your comment re $~B = (5 ~\color{red}{-}~\sqrt{17})\cdots .~$ Also, my approach requires also that $~2^{n+1}~$ does not divide $~g(n),~$ which I could not find any easier way of showing. After the fact, I did notice that the induction would have worked just as well by focusing on $~g(N-1),g(N),g(N+1),~$ rather than $~g(N-2),g(N),g(N+2).$ – user2661923 Jan 11 '25 at 14:47
  • @user2661923, You claimed $2^n|G(n)$ and I showed it is true. Except the typo I do not see anything wrong to get down vote. I do not mean you down voted me. – sirous Jan 11 '25 at 16:56
  • @sirous All that I meant was that if you were going to attack the problem in an alternative manner, the comment that you gave provides a creative way of showing that $~2^n ~| ~g(n).~$ Then, if you happened to be looking for some alternative way of completely conquering the problem, you would then have to find some alternative way of showing that $~2^{n+1}~$ does not divide $~g(n).$ – user2661923 Jan 11 '25 at 17:07
  • Please strive not to post more (dupe) answers to PSQs & dupes of FAQs. This is enforced site policy, see here. – Bill Dubuque Jan 11 '25 at 21:08
  • @BillDubuque i think i have made the necessary changes, this post should be reopened to accept answers. – K41 Jan 13 '25 at 11:59