1
$\begingroup$

Let $\text{lcm}(x)$ be the least common multiple of $\{1,2,3,\dots,x\}$

Let $x\#$ be the the primorial for $x$.

It occurs to me that for $x \ge 10$:

$$\frac{\text{lcm}(x^2+x)}{\text{lcm}(x^2)} < 2^x\frac{(x^2+x)\#}{(x^2)\#}$$

Am I right?

Here is my thinking:

(1) $\dfrac{\text{lcm}(x+1)}{\text{lcm}(x)} > 1$ if and only if $x+1$ is a power of a prime.

(2) If a prime $x^2 \ge p > x$, then $p^2 \ge (x+1)^2 > x^2 + x$ so it will divide out in the ratio.

(3) If a prime $x \ge p \ge (\sqrt{x}+1)$, then $p^2 \le x^2$ and $p^4 \ge (\sqrt{x}+1)^4 \ge {4\choose0}x^2 + {4\choose1}x^{3/2} + {4\choose2}x + {4\choose3}x^{1/2} + 1 > x^2+x$

(4) But there can only be one such $x \ge p \ge \sqrt{x} + 1$ such that $p^3 | \frac{(x^2+x)!}{(x^2)!}$ since $(p+2)^3 - p^3 \ge {3\choose1}2p^2 + {3\choose2}4p + 1 > 2x$

(5) Using Hanson's result that $\text{lcm}(x) < 3^x$, we have:

$$\frac{\text{lcm}(x^2 + x)}{\text{lcm}(x^2)} < 3^{\sqrt{x}+1}x\frac{(x^2+x)\#}{(x^2)\#}$$

(6) $2^x > 3^{\sqrt{x}+1}x$ since $2^{10} = 1024 > 969>3^{\sqrt{10}+1}\times10$ and $2^{\sqrt{x}} > 3(3x)^{1/\sqrt{x}}$


Edit: Forgot the case where a prime might be between $\sqrt{x}$ and $\sqrt{x}+1$

So, I've updated my argument.

$\endgroup$
1
+50
$\begingroup$

HINT $1$

Note that $$r(x)={\large \dfrac {\mathrm{lcm}(x^2+x)}{\mathrm{lcm}(x^2)}\cdot{\dfrac {(x^2)\#}{(x^2+x)\#}} = \prod_{p_k<x+1} p_k^{[\log_{p_k}(x^2+x)]-[\log_{p_k}(x^2)]}}.$$ Calculations using Wolfram Alpha of function $$\bar r(x) = \left(\dfrac45\right)^x\prod_{p_k<x+1} p_k^{[\log_{p_k}(x^2+x)]-[\log_{p_k}(x^2)]}$$ gives the plot of

r(x)

which shows that $\bar r(x) \leq 1,$ and that leads to the empiric estimation $$r(x)\leq \left(\dfrac54\right)^x,\quad x\in\mathbb N.$$ The reason is that all primes greater than $x,$ contains both in the ratio for lcm and the ratio for primorial. Therefore, correction of $r(m)$ happens at least for the squares of primes.

Let $x=11,$ then the interval $[x^2+1,x^2+x] = [122,132]$ contains new degrees $2^7=128$ and $5^3=125.$ So $r(11)=2\cdot5=10.$

This means that complex consideration leads to the better estimation.

HINT $2$

For $x>>1$ $$\dfrac{(x^2+x)\#}{x^2\#}\approx \left(x^2+\frac12x\right)^d,$$ where $$d=\pi(x^2+x)-\pi(x^2)\approx \int\limits_{x^2}^{x^2+x}\dfrac{\mathrm dt}{\log t} = x^2\int\limits_{x^2}^{x^2+x}\dfrac1{\log t}\mathrm d\dfrac t{x^2} = x^2\int\limits_{1}^{1+1/x}\dfrac{\mathrm du}{\log(x^2u)}$$ $$ = \dfrac{x^2}{2\log x}\int\limits_{1}^{1+1/x}\dfrac{\mathrm du}{1+\dfrac{\log u}{2\log x}} \approx \dfrac{x^2}{2\log x}\int\limits_{1}^{1+1/x}\left(1-\dfrac{\log u}{2\log x}\right)\,\mathrm du$$ $$ = \dfrac{x^2}{2\log x}\left(u - \dfrac{u\log u -1}{2\log x}\right)\bigg|_1^{1+\frac1x} = \dfrac{x^2}{2\log x}\left(\dfrac1x-\dfrac{(1+\frac1x)\log (1+\frac1x)}{2\log x}\right),$$ $$d\approx\dfrac x{2\log x} - \dfrac{x+1}{4\log^2x}.$$

$\endgroup$
  • $\begingroup$ Thanks for your analysis! Using my equation, I noticed that for $x \ge 63, (5/4)^x > 3^{\sqrt{x}+1}x$. It is very interesting that the number can go so significantly lower than $2^x!$. $\endgroup$ – Larry Freeman Dec 8 '18 at 18:47
  • $\begingroup$ @LarryFreeman This anaysis has significant modeling part... $\endgroup$ – Yuri Negometyanov Dec 8 '18 at 18:50
  • $\begingroup$ @LarryFreeman You are welcome! I like real analysis. $\endgroup$ – Yuri Negometyanov Dec 8 '18 at 22:09

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.