Simple task, but I know only extremely overkill solution :)

Let $a,b,c,d,e,f$ be positive irrational numbers.

Suppose that for any positive integer number $n$,

$$\lfloor a n\rfloor \cdot \lfloor b n\rfloor \cdot \lfloor c n\rfloor = \lfloor d n\rfloor \cdot \lfloor e n\rfloor \cdot \lfloor f n\rfloor.$$

Prove that the sets $\{a,b,c\}$ and $\{d,e,f\}$ are equal.

  • $\begingroup$ Your question is incomplete. What fo you mean by {a,b,c} and what are you doing with the product of the floors? $\endgroup$ – Scott Burns Nov 19 '16 at 13:01
  • $\begingroup$ {a,b,c} is an unordered set $\endgroup$ – kotomord Nov 19 '16 at 13:06
  • 2
    $\begingroup$ Interesting question +1. It follows quickly that $abc = def$ from $\lim x / \lfloor x \rfloor = 1$. Would you have a simple (counter)example why the irrational condition is necessary. $\endgroup$ – dxiv Nov 21 '16 at 4:38
  • $\begingroup$ @dxiv - I have no simple (counter)example, but my solution uses irrationality of (at least one) number $\endgroup$ – kotomord Nov 21 '16 at 8:06
  • 1
    $\begingroup$ @AlexYoucis, I'll write it today (Sorry, I'm working now) It uses en.wikipedia.org/wiki/Equidistribution_theorem $\endgroup$ – kotomord Nov 23 '16 at 14:58

Let $E=(\mathbb{R}/\mathbb{Z})^6$. Let $x=(\overline{a},\overline{b},\overline{c},\overline{d},\overline{e},\overline{f})\in E$. Let $F=\mathbb{N}x$.

$\overline{F}$ is a subgroup of $E$.

$a,b,c,d,e,f$ are irrationnal, so it exists $(u_n)\in\mathbb{N}^{\mathbb{N}}$ and $(\delta_a,\delta_b,\delta_c,\delta_d,\delta_e,\delta_f)\in \{0,1\}^6$ such that $u_nx \in E \mapsto 0_E$ with $u_nx=(\overline{a_n},\overline{b_n},\overline{c_n},\overline{d_n},\overline{e_n},\overline{f_n})$ and $0<a_n<1,0<b_n<1,0<c_n<1,0<d_n<1 ,0<e_n<1 ,0<f_n<1$.

and $(a_n,b_n,c_n,d_n,e_n,f_n)\in[0,1]^6 \mapsto (\delta_a,\delta_b,\delta_c,\delta_d,\delta_e,\delta_f)$.

So it exists $(v_n)\in\mathbb{N}^{\mathbb{N}}$ such that $v_nx \in E \mapsto 0_E$ with $v_nx=(\overline{a'_n},\overline{b'_n},\overline{c'_n},\overline{d'_n},\overline{e'_n},\overline{f'_n})$ and $0<a'_n<1,0<b'_n<1,0<c'_n<1,0<d'_n<1 ,0<e'_n<1 ,0<f'_n<1$.

and $(a'_n,b'_n,c'_n,d'_n,e'_n,f'_n)\in[0,1]^6 \mapsto (1-\delta_a,1-\delta_b,1-\delta_c,1-\delta_d,1-\delta_e,1-\delta_f)$

So $\lfloor au_n\rfloor \lfloor bu_n\rfloor \lfloor cu_n\rfloor=(au_n-a_n) (bu_n-b_n)(cu_n-c_n)=abc(u_n)^3-abc_n(u_n)^2-acb_n(u_n)^2-bca_n(u_n)^2+\dots$

So, as already shown by dxiv, $$\lim\frac{\lfloor au_n\rfloor \lfloor bu_n\rfloor \lfloor cu_n\rfloor}{u_n^3}=abc$$

So $abc=def$ because $\lfloor au_n\rfloor \lfloor bu_n\rfloor \lfloor cu_n\rfloor=\lfloor du_n\rfloor \lfloor eu_n\rfloor \lfloor fu_n\rfloor$

And $$\lim \frac{\lfloor au_n\rfloor \lfloor bu_n\rfloor \lfloor cu_n\rfloor-abc(u_n)^3}{u_n^2}=-ab\delta_c-ac\delta_b -bc\delta_a$$

We have also $$\lim \frac{\lfloor av_n\rfloor \lfloor bv_n\rfloor \lfloor cv_n\rfloor-abc(v_n)^3}{v_n^2}=-ab(1-\delta_c)-ac(1-\delta_b) -bc(1-\delta_a)$$

So $$ab+ac+bc=-\lim \frac{\lfloor au_n\rfloor \lfloor bu_n\rfloor \lfloor cu_n\rfloor-abc(u_n)^3}{u_n^2}-\lim \frac{\lfloor av_n\rfloor \lfloor bv_n\rfloor \lfloor cv_n\rfloor-abc(v_n)^3}{v_n^2}$$

So $ab+ac+bc=de+df+ef$.

Lemma: We show $\overline{F}=\overline{\mathbb{N}x}$ is a subgroup. For all $k \in \mathbb{N}$, let the compact $A_k=kx+\overline{\mathbb{N}x}\subset \overline{\mathbb{N}x}$. And for all $k$, $0\in A_k$. So $0 \in G=\cap_k A_k$. $x+G\subset G$. So $\mathbb{N}x+G\subset G$.

$G$ is compact so $\overline{\mathbb{N}x}\subset G$.

As $G \subset A_0=\overline{\mathbb{N}x}$, we have $G=\overline{\mathbb{N}x}$.

As $0 \in\overline{\mathbb{N}x}=G \subset A_1=x+\overline{\mathbb{N}x}$, we have $-x\in \overline{\mathbb{N}x}$. So $ \overline{\mathbb{N}x}= \overline{\mathbb{Z}x}$

Remark: for all $m\in\mathbb{N}$, we can choose $(\delta_a,\delta_b,\dots,\delta_f)=(\{ma\},\{mb\},\dots,\{mf\})\in \overline{F}$, and $(u_n)$ such that $u_n x\mapsto \delta$. So we have $ab\{mc\}+ac\{mb\}+bc\{ma\}=de\{mf\}+df\{me\}+ef\{md\}$ for all $m\in\mathbb{N}$.

  • $\begingroup$ But you have no proof what a+b+c = d+e+f? $\endgroup$ – kotomord Nov 23 '16 at 13:40
  • 2
    $\begingroup$ No, but my comment was too long. $\endgroup$ – marco2013 Nov 23 '16 at 13:41
  • $\begingroup$ I use the property: $\overline{F}$ is a subgroup. So if $(\overline{y_1},\overline{y_2},\overline{y_3},\overline{y_4},\overline{y_5},\overline{y_6})\in \overline{F}$, we have $(\overline{1-y_1},\overline{1-y_2},\overline{1-y_3},\overline{1-y_4},\overline{1-y_5},\overline{1-y_6})\in \overline{F}$ too. $\endgroup$ – marco2013 Nov 23 '16 at 13:57
  • $\begingroup$ In your lemma, why do we have $0\in A_k$ ? $\endgroup$ – Ewan Delanoy Nov 23 '16 at 14:58
  • $\begingroup$ Because $E$ is compact. So from the sequence $nx$, we can find a subsequence $w_nx\mapsto \ell\in E$ (with $w_n$ strictly increasing). Let$\epsilon>0$,$\exists N\in\mathbb{N}$ such that $n>N\implies \|w_nx-\ell\|<\epsilon$. So $\forall n,m>N$, $\|(w_n-w_m)x\|<2\epsilon$. Take $w_n-w_m>k$. So $0\in \overline{A_k}$. But $A_k$ is compact. $\endgroup$ – marco2013 Nov 23 '16 at 15:08

So, the overkill proof;

One of the equidistribution theorems: https://en.wikipedia.org/wiki/Equidistribution_theorem

a is irrational, then sequence $a*p_n$ mod 1 is uniformly distributed.

Lemma: If a > 0 is irrational, then sequence { $\lfloor a*n\rfloor$ } contains infinitely many prime numbers.

Proof of lemma:

Case (a>1):

$\lfloor a*n\rfloor = p_m$ <=> $p_m < a*n < p_m + 1$ <=> $\frac{p_m}{a} < n < \frac{p_m+1}{a}$ <=> n = $\lfloor \frac{p_m+1}{a}\rfloor$, {$\frac{p_m}{a}$} $ \in (1-\frac{1}{a}, 1)$ (where are infinitely many such m by thm.)

Case (a<1) is trivial.

Proof of task:

Without loss of generality $a \ge b \ge c, d \ge e \ge f, a\ge d$

Let a > d. Fix n what $n*a > n*d + 2, n*c > 1, \lfloor n*a \rfloor =p_m$

$\lfloor n*d \rfloor \lfloor n*e \rfloor \lfloor n*f \rfloor > 0 $ and not divisible to $p_m$ - it is the absurd.

So, a = d and we need to prove what if $\lfloor n*e \rfloor \lfloor n*f \rfloor = \lfloor n*b \rfloor \lfloor n*c \rfloor$, then sets {b, c} and {e,f} are equals.

We can do it similary or by method of @marco2013

P.S. Maybe, we can generalize lemma to rational not-integer numbers (with the Dirichlet thm.)


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.