Mathematically Absurd

by SomeonecoolLovesMaths, Jan 3, 2025, 6:52 PM

Prove that for any positive integers $a_1, a_2, \dots, a_n$,
\[ \prod_{1 \leq i < j \leq n} \frac{a_i-a_j}{i-j} \in \mathbb{Z} .\]
Proof:
We shall show that for any prime $p$
\[ v_p \left( \prod_{1 \leq i < j \leq n} (a_i-a_j) \right) \geq v_p \left( \prod_{1 \leq i < j \leq n} (i-j) \right). \]The divisibility immediately follows. Indeed, the problem statement is equivalent to showing that the quantity
\[ v_p \left( \prod_{1 \leq i < j \leq n} (a_i-a_j) \right) \]is minimised when $a_i=i$ for all $i$. To prove this we find some reasonable way of computing the above quantity.
Denote $A_{p^k} =  \{ (i,j) \mid p^k \text{ divides } a_i-a_j\} $. Then
\[ v_p \left( \prod_{1 \leq i < j \leq n} (a_i-a_j) \right) = \sum_{1 \leq i < j \leq n} v_p \left( a_i-a_j \right) = \sum_{k \geq 1} |A_{p^k}| .\]We proceed by computing the size of each individual set $A_{p^k}$. I claim the cardinality of this set is minimised when $a_i=i$ for all $i$. Indeed, let
\[ A_{(p^k, x)} = \left\{  a_i \mid a_i \equiv x \pmod{p^k} \right\} \]for $0 \leq x \leq p^k-1$. We remark that the union of $A_{(p^k, x)}$ partitions the set $\{a_1, a_2, \dots a_n\}$ because modular congruence is an equivalence relation. Moreover, $(i, j) \in A_{p^k}$ if and only if $\{a_i, a_j \} \subseteq A_{(p^k,x)}$ for some $x$. Thus
\[ |A_{p^k}| = \sum_{x=0}^{p^k-1} \binom{|A_{(p^k,x)}|}{2} .\]I claim that the above sum is minimised if there exists an integer $k$ such that $|A_{(p^k,x)}| \in \{k, k+1 \}$ for all $x$. (Note that this is very close to Jensen's Inequality.) Suppose for the sake of contradiction that the function
\[ f(a_0, a_1, \dots, a_{p^k-1} )=\sum_{x=0}^{p^k-1} \binom{a_x}{2} \]has minimal value at $(a_0, a_1, \dots , a_{p^k-1})$ that does not satisfy the above claim. Then there exists $i, j$ such that $a_i-a_j \geq 2$.
Rearrange to $a_i -a_j-2 \geq 0$ and rewrite as
\begin{align*} 0 < a_i -a_j-1 &= \frac{2(a_i-1)-2(a_j)}{2}\\
&=\frac{(a_i-1)(a_i-(a_i-2))-a_j(a_j+1-(a_j-1))}{2} \\
&= \frac{a(a_i-1)}{2}-\frac{(a_i-1)(a_i-2)}{2} -\frac{(a_j+1)a_j}{2} + \frac{a_j(a_j-1)}{2} \\
&= \binom{a_i}{2} + \binom{a_j}{2}-\binom{a_i-1}{2}-\binom{a_j+1}{2}.
\end{align*}Then
\[ \binom{a_i-1}{2} + \binom{a_j+1}{2} < \binom{a_i}{2}+\binom{a_j}{2} .\]Finally, if we assume without loss of generality that $j > i$ we conclude
\[ f(a_0, a_1, \dots ,a_{p^k-1} ) > f(a_0, a_1, \dots , a_i-1, \dots, a_j+1, \dots, a_{p^k-1} ) \]contradicting the minimality of $(a_0, a_1, \dots a_{p^k-1})$.
Because $\{ a_1 ,a_2 , \dots ,a_n \}$ is partitioned by the $A_{(p^k,x)}$, we have that
\[ |A_{(p^k,0)}|+|A_{(p^k,1)}|+\dots+|A_{(p^k,p^k-1)}|=n . \]We may say that $n$ has remainder $r$ upon Euclidean division by $p^k$. Then from our above work it follows that,
\[ |A_{p^k}|=f \left( | A_{(p^k,0)}|, |A_{(p^k,1)}|, \dots, |A_{(p^k,p^k-1)}|\right) \geq r\binom{\left\lfloor \frac{n}{p^k}\right\rfloor +1 }{2} + (p^k-r)\binom{\left\lfloor \frac{n}{p^k}\right\rfloor}{2}. \]It remains to show that this value is taken when $a_i=i$. For that case, we define $A_{p^k}^{\star}=\{ (i,j) \mid p^k \text{ divides } i-j\}$ and $A^{\star}_{(p^k,x)}= \left\{  i \mid i \equiv x \pmod{p^k} \right\} $. It's fairly easy to see that
\[ |A_{(p^k,x)}|= \space
\begin{cases}\left\lfloor \frac{n}{p^k}\right\rfloor +1 &\text{ if } 0 \leq x \leq r \\
\left\lfloor \frac{n}{p^k}\right\rfloor & \text{ if } r < x \leq p^k-1. 
\end{cases} \]Then
\[ |A_{p^k}^{\star}|=f \left( | A_{(p^k,0)}^{\star}|, |A_{(p^k,1)}^{\star}|, \dots, |A_{(p^k,p^k-1)}^{\star}|\right) =  \sum_{x=0}^{p^k-1} \binom{|A_{(p^k,x)}^{\star}|}{2} = r\binom{\left\lfloor \frac{n}{p^k}\right\rfloor +1 }{2} + (p^k-r)\binom{\left\lfloor \frac{n}{p^k}\right\rfloor}{2} \]when $a_i=i$. Thus $A_{p^k} \geq A_{p^k}^star$.
\[ v_p \left( \prod_{1 \leq i < j \leq n} (i-j) \right) = \sum_{1 \leq i < j \leq n} v_p \left( i-j \right) = \sum_{k \geq 1} |A^{\star}_{p^k}|. \leq \sum_{k \geq 1} |A_{p^k}|= v_p \left( \prod_{1 \leq i < j \leq n} (a_i-a_j) \right) ,\]completing the proof.

Thanks to HKIS200543

Comment

1 Comment

The post below has been deleted. Click to close.
This post has been deleted. Click here to see post.
This is amazing omg

by Eisestein, Jan 10, 2025, 8:17 PM

My First Blog

avatar

SomeonecoolLovesMaths
Shouts
Submit
  • I will bookmarked this blog

    by giangtruong13, Yesterday at 7:37 AM

  • W blog op

    by quasar_lord, Mar 10, 2025, 5:32 PM

  • orz blog
    how so orzzzz

    by Erratum, Jan 31, 2025, 9:47 AM

  • shouts; your post is too short , it must be atleast 8 characters

    by mqoi_KOLA, Dec 5, 2024, 6:37 PM

  • Guys it took my like 10 hours to do this! Idk why did I even do this, but now it looks kinda satisfying ngl.

    by SomeonecoolLovesMaths, Nov 25, 2024, 5:07 PM

  • add this one new thing to the intergrals that is lim n tends to infinity b-a/n summation k=1 to n f(a+k(b-a)/n))= int_{a}^b f(x) dx

    by Levieee, Nov 21, 2024, 8:37 PM

  • woahh hi HSM main orz blog!!!

    by eg4334, Nov 17, 2024, 3:31 PM

  • Me in 4 years of Aops - 555 posts.

    This guy in 10 months of Aops - 2608 posts

    by HoRI_DA_GRe8, Oct 17, 2024, 10:22 AM

  • Remember; the one who falls and gets back up is stronger than the ones who never tried... Good luck for next year's IOQM

    by alexanderhamilton124, Oct 15, 2024, 7:03 AM

  • fourth shout

    by QueenArwen, Sep 2, 2024, 4:05 PM

  • Hii
    !!!!!!!!!!!

    by Yummo, Apr 2, 2024, 2:43 PM

  • i am shouting. it is very loud.

    by fruitmonster97, Mar 21, 2024, 7:49 PM

  • real!!!!!

    by SirAppel, Mar 17, 2024, 6:22 PM

13 shouts
Tags
About Owner
  • Posts: 3139
  • Joined: Dec 22, 2023
Blog Stats
  • Blog created: Mar 17, 2024
  • Total entries: 20
  • Total visits: 1249
  • Total comments: 19
Search Blog
a