5

Does $[n \sqrt{2}] = [m (2+\sqrt{2})]$ for $m,n$ natural have no solution where $[x]$ is the floor function of $x$? I tried calculating some examples ($1,000,000$ examples on Python) and it seems as though the left and right hand side never equal the same number- no matter the choice of $m$ and $n$. This is from a Calculus $1$ worksheet.

Alessio K
  • 10,599
  • Desmos should give you a good picture. – sai-kartik Sep 05 '20 at 03:44
  • I upvoted; interesting question. The best way to attack this is to first give as much detail as you can of the background of the question. You indicated that "This is from a Calculus 1 worksheet." Please try to determine what theorems or previously (solved) problems that your book or class focused on that you think might be pertinent here? Please edit your query to include as much of this background as you can. ...see next comment. – user2661923 Sep 05 '20 at 03:56
  • Next, (temporarily) forgo using computer software and try to attack the problem analytically. That is, consider the problem's background (re previous comment) as tools, and show as much work as you can, trying to use the tools to attack the problem. Regardless of how far you get, also edit your query to show this work. – user2661923 Sep 05 '20 at 03:56
  • 2
    @user2661923 I've seen this problem before. This was administered to a calculus class that had only met once prior since the semester literally just started. – Matthew Graham Sep 05 '20 at 03:59
  • @user2661923 This is a problem on the first week (started by reviewing pre-calc and learning limits ) – Sean Nemetz Sep 05 '20 at 04:04
  • @MatthewGraham It's time to meta-cheat. The problem would probably not have been given unless both (1) its solution is known and (2) the problem can be solved in a reasonably straight forward manner via Calculus or pre-calculus. Attempting to solve it "in a vacuum" is generally bad. I continue to recommend that the OP identify the tools that might be pertinent and then use those tools. – user2661923 Sep 05 '20 at 04:04
  • 2
    https://en.m.wikipedia.org/wiki/Beatty_sequence – Qiaochu Yuan Sep 05 '20 at 04:14
  • @RossMillikan I was in the process of commenting on your answer when your answer was deleted. My guess is that your (deleted) answer is "on to something." However, your answer involves continued fractions which the OP might not be expected to be conversant with. Can you repair your answer so that it is completed, without using convergents? – user2661923 Sep 05 '20 at 04:15
  • @QiaochuYuan I skimmed the link that you cited. It is unclear to me that the Rayleigh theorem (A.K.A. Beatty's theorem) is immediately conclusive. Perhaps it is, and I simply didn't spend enough time studying the link. This begs the question: is it intended that the OP be familiar enough with this material to use it to answer the question? If not, is there a (simpler) version of the analysis that is usable that the OP might reasonably be expected to originate? – user2661923 Sep 05 '20 at 04:23
  • Look at the proofs of Beatty theorem. One of them may apply to your situation. – markvs Sep 05 '20 at 04:47
  • @JCAA After following the trailer in ilovebulbasaur's answer, which explicitly links the Beatty theorem to this query, I examined the first proof (only) in https://en.m.wikipedia.org/wiki/Beatty_sequence. I found it very creative and I (surprisingly) was able to follow it. Personally, no way could I have ever originated something like that. – user2661923 Sep 05 '20 at 05:36

2 Answers2

6

Here's a partial answer (let me know if partial answers are unwelcome on math.SE). The equation has no solutions in natural numbers $\mathbb{N}$. Suppose for contradiction that there exists $m,n\in\mathbb{N}$ with $[n\sqrt{2}]=[m(2+\sqrt{2})]$. Then there are three cases:

Case 1: $m=n$. Impossible because then $$[n\sqrt{2}]=[m(2+\sqrt{2})]=[n(2+\sqrt{2})]=[n\sqrt{2}]+2n$$ which implies that $2n=0$ and $n=0\notin\mathbb{N}$.

Case 2: $m>n$. Then if we write $m=n+l$ for some $l\in\mathbb{N}$, $$\begin{align*}[n\sqrt{2}]&=[(n+l)(2+\sqrt{2})] \\ & = [(n+l)\sqrt{2}]+2(n+l) \\ & >[n\sqrt{2}] \end{align*}$$ which is a contradiction.

Case 3: $m<n$. Then if $m=n-l$ for some $l\in\mathbb{N}$, we have $$\begin{align*}[n\sqrt{2}]&=[(n-l)(2+\sqrt{2})] \\ & = [(n-l)\sqrt{2}]+2(n-l). \end{align*}$$ But this won't get us anywhere (unless if someone can continue from here and prove me wrong).

And of course the result follows easily from the link Qiaochu provided: https://en.m.wikipedia.org/wiki/Beatty_sequence. In layman's terms, it states that $$[nr]\neq [m\frac{r}{r-1}]$$ for all positive integers $n,m$ and positive irrationals $r$. In particular, if $r=\sqrt{2}$, then $\frac{r}{r-1}=\frac{\sqrt{2}}{\sqrt{2}-1}=2+\sqrt{2}$, so $$[n\sqrt{2}]\neq [m(2+\sqrt{2})]$$ for all $n,m$ positive integers.

  • @Sean Nemetz it seems like you have a really difficult instructor for calculus! – ilovebulbasaur Sep 05 '20 at 04:44
  • I (also) upvoted, nice analysis. FWIW, for me personally, even partial answers are welcome. Your answer's "closing" still begs two questions: (1) Is it reasonable to expect the OP to be very familiar with the material in the Beatty Sequence link? (2) If not, is there a simpler version of the analysis that the OP might reasonable be expected to know that can be used to complete the problem? – user2661923 Sep 05 '20 at 04:46
  • It depends on what kind of calculus class this is, and what results the OP has seen in class. I have been a teaching assistant for calculus in a US college, and this isn't something I would expect anyone in the class to be able to solve. – ilovebulbasaur Sep 05 '20 at 04:48
1

The less detailed treatment, but more intuitve, is that the two real numbers are extremely close only when their ratio is a continued fraction convergent for $1+ \sqrt 2.$ The numerators and denominators are consecutive elements in the sequence $$ 1, 2, 5, 12, 29, 70, 169, 408, $$

https://oeis.org/A000129

https://en.wikipedia.org/wiki/Pell_number#Pell_numbers

while the relevant pair of real numbers, rounded to the nearest integer, are in the sequence $$ 1, 3, 7, 17, 41, 99, 239, 577, 1393, $$

https://oeis.org/A001333

These are one half the Pell-Lucas numbers,

https://en.wikipedia.org/wiki/Pell_number#Pell%E2%80%93Lucas_numbers

Both integer sequences obey the linear recursion $$ w_{n+2} = 2 w_{n+1} + w_n $$

The characteristic roots are $1+ \sqrt 2$ and $1- \sqrt 2.$ Both sequences can be written in the form $$ A (1 + \sqrt 2)^n + B (1- \sqrt 2)^n $$ The exact coefficients should be given in the OEIS entries for each sequence.

Will Jagy
  • 139,541
  • Very interesting answer. The only reason that I did not upvote this answer is because first I would have to verify it. This verification would take a significant amount of time and effort on my part. – user2661923 Sep 05 '20 at 17:21