An equivalent inequality for integers follows:

$$(3n^2-5)\left\lceil n/\sqrt{3} \right\rceil^2 > n^4.$$

This has been checked for n = 2 to 60000. Perhaps there is some connection to the convergents to $\sqrt{3}$.

$\lceil \frac{n}{\sqrt{3}} \rceil > \frac{n^2}{\sqrt{3n^2-5}}$

  • 1
    $\begingroup$ Should follow from $|a/b -\sqrt{3}| > \alpha/b^2$, which is true for all relatively prime $a,b$ for a suitable $\alpha$. $\endgroup$ – Felipe Voloch Nov 6 '14 at 22:59
  • $\begingroup$ If you like my answer, please accept it officially (so that it turns green). Thanks in advance! $\endgroup$ – GH from MO Aug 19 '18 at 23:22

The inequality is true for $n\geq 2$. Let $m:=\lceil n/\sqrt{3}\rceil$, then $$ m^2-\frac{n^2}{3} = \frac{3m^2-n^2}{3} \geq \frac{2}{3}, $$ because $3m^2-n^2>0$ and $3m^2-n^2\not\equiv 1\pmod{3}$. Hence the inequality follows from $$ \frac{2}{3}>\frac{n^4}{3n^2-5}-\frac{n^2}{3}=\frac{5n^2}{3(3n^2-5)}. $$ The latter is equivalent to $n^2>10$, hence we have a proof for $n\geq 4$. The remaining cases $n=2$ and $n=3$ can be checked by hand.

  • $\begingroup$ Why is $\frac{3m^2-n^2}{3}\geq \frac{2}{3}$? $\endgroup$ – Alan Nov 7 '14 at 7:00
  • 2
    $\begingroup$ @Alan, the left side is something-nonnegative-over-3, and GH tells you why that something isn't zero or one, so it's at least 2. $\endgroup$ – Gerry Myerson Nov 7 '14 at 8:58

Equivalently, we want to know if

$\mathrm{sqceiling}(n^2/3) > n^4/(3n^2-5) = n^2/3 + 5/9 + 25/(27n^2) + ...$

where $\mathrm{sqceiling}()$ is the function taking a real to the next exact square.

This is equivalent to

$\mathrm{sqceiling}(n^2/3) - n^2/3 > 5/9 + 25/(27n^2) + ...$

but since $n^2/3$ is never a square for $n>0$, and $n^2 \equiv 0\ \textrm{or}\ 1 (\textrm{mod}\ 3)$, we have

$\mathrm{sqceiling}(n^2/3) - n^2/3 \ge 2/3$.

Clearly for large enough $n$

2/3 > 5/9 + 25/(27n^2) + ...

  • $\begingroup$ We wrote much the same, but you beat me by 4 seconds! $\endgroup$ – GH from MO Nov 6 '14 at 23:15
  • $\begingroup$ not quite! I was still editing my latex. $\endgroup$ – Yaakov Baruch Nov 6 '14 at 23:21

A tight result is $$\lceil \frac{n}{\sqrt{3}}\rceil \ge \frac{n^2}{\sqrt{3n^2-6+\frac{12}{n^2+2}}}$$

This means that the desired inequality holds for $n \ge 4.$ The cases $n=2,3$ can be checked separately.

As noted elsewhere, for $m=\lceil \frac{n}{\sqrt{3}}\rceil$ we have $n^2 \le 3m^2-2$ with equality exactly for the cases $n=1,5,19,71,265,989,\cdots$ A001834 with $\frac{n}{m} \lt \sqrt{3}$ a convergent to that square root. A bit of manipulation then gives the result above and the information that equality holds for exactly those values of $n.$

Details: Change the desired inequality to $$m^2 \ge \frac{n^4}{3n^2-k}$$ with $k$ to be determined. Since $n^2=3m^2-j$ with $j \ge 2,$ we have $m^2=\frac{n^2+j}{3}.$ This yields $$k \le \frac{3jn^2}{n^2+j} =3j-\frac{3j^2}{n^2+j}.$$ Since we do sometime have $j=2,$ The bound $k \le 6-\frac{12}{n^2+2}$ is sometime an equality. It is not hard to show that $\frac{3jn^2}{n^2+j}$ increases when $j \ge 2$ does.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.