0

Given the set $S_m=\{[\frac{k(k+1)}{2}]_m\}$ for $k\in\mathbb{N}$. Where $[x]_m$ denotes the equivalence class (modm).

I am trying to show that: $S_m=\mathbb{Z}_m$ if and only if $m=2^s$ for some $s \in\mathbb{N}$.

A hint is also given that for each prime $p>2$ there is a $b\in\mathbb{Z}_p$ such that the equation $x^2\equiv b\pmod{p}$ has no solutions.

I've tried using the condition for what m has to equal in order for the identity to hold but I'm quite lost in how to proceed really, any help would be greatly appreciated.

Bill Dubuque
  • 282,220
  • 1
    Let $t=\frac{k(k+1)}{2}$. What expression do you get when you calculate $8t+1$? Can you factorise that? – Jaap Scherphuis Feb 11 '21 at 16:31
  • @JaapScherphuis Yes, so I factorised it as $8t+1=(2k+1)^2$ and which I can see is an odd number but should I then use the hint to show that the both sets are not equal for odd $m$? or how? – eldoco1311 Feb 11 '21 at 17:56
  • If $m$ is an odd prime, then there is some $b$ that is not a square mod $m$. So you have such a $b$. You can then find $t$ such that $8t+1=b \bmod m$. And so.... – Jaap Scherphuis Feb 11 '21 at 18:42
  • I've given an answer but with the part of the question you've been discussing covered over if you now wish to do that part yourself. –  Feb 11 '21 at 18:42
  • Hint elaboration: $ $ by the (modular) quadratic formula (completing the square)

    $$\bmod {\rm odd}\ m!:,\ (x!+!1)x/2\equiv c,\overset{8(\ \ )+1}\Longrightarrow, (2x!+!1)^2\equiv \color{#0a0}{8c!+!1}\qquad\qquad\qquad$$

    so it has no roots if the $\rm\color{#0a0}{discriminant}$ is $\rm\color{#c00}{nonsquare},,$ i.e. $,\color{#0a0}{8c!+!1} = \color{#c00}b,,$ i.e. if $,c\equiv (b!-!1)/8\ \ $

    – Bill Dubuque Apr 25 '24 at 15:49
  • See here for a more general question. – Bill Dubuque Apr 25 '24 at 15:49

2 Answers2

1

Case 1: $\ m=2^s$

For $k$ and $l$ residues modulo $2^{s+1}$ suppose $$\frac{k(k+1)}{2}\equiv \frac{l(l+1)}{2}\pmod {2^s}.$$

Then $k(k+1)\equiv l(l+1)\pmod {2^{s+1}}$

i.e. $(k-l)(k+l+1)\equiv 0\pmod {2^{s+1}}$.

Now one of the two factors on the LHS is odd and so the other factor has to be divisible by $2^{s+1}$. So equality occurs if and only if $k+l+1\equiv 0\pmod {2^{s+1}}$. Hence everyone of the $2^{s+1}$ values for $\frac{k(k+1)}{2}$ is taken precisely twice and so, by the pigeon-hole principle, $\frac{k(k+1)}{2}$ takes every value modulo $2^s$.

Case 2: If $m$ has an odd prime factor $p$ then, following the given hint, let $b$ be an element such that $x^2\equiv b\pmod p$ cannot be solved.

Suppose that $\frac{k(k+1)}{2}\equiv \frac{b-1}{8}\pmod p$ could be solved. Then $(2k+1)^2\equiv b \pmod p$, a contradiction.

Bill Dubuque
  • 282,220
  • Very nice, and those two cases for m cover all cases? Also how did you get to supposing that the LHS should be congruent to precisely (b-1)/8 – eldoco1311 Feb 11 '21 at 18:44
  • Yes - either m is divisible by an odd prime or its a power of 2. Re. (b-1)/8 its doing the invers of what you had looked at. You had multiplied by 8 and added 1 so subtract 1 and divide by 8. –  Feb 11 '21 at 18:47
  • How do you determine from the last congruence that for every of the 2^(s+1) values for k(k+1)/2 is taken exactly twice? – eldoco1311 Feb 11 '21 at 19:21
  • They are paired up with each pair being k and -1-k. –  Feb 11 '21 at 19:24
  • Ohh, I think I finally got it, thank you very much! Do you have any general tips for approaching problems like these? – eldoco1311 Feb 11 '21 at 19:56
  • One obvious thing is that familiarity with as many ideas as possible is useful. (In this question quadratic residues.) Other than that, always be prepared to do some examples (In this case mod 8 say) just to get a feel for what the problem is about. Good luck! –  Feb 11 '21 at 20:02
0

For each positive integer $m$, let $f_m:Z_m\to Z_m$ be given by $$ f(x) = \Bigl( {\small{\frac{x^2+x}{2}}} \Bigr) {\,\text{mod}\,} m $$ Claim:$\;f_m$ is injective if and only if $m$ is a power of $2$.

Proof:

First assume $m$ is not a power of $2$.

We want to show that $f_m$ is not injective.

Suppose instead that $f_m$ is injective.

Let $n > 1$ be an odd divisor of $m$.

Then $2$ is a unit in $Z_n$.

Since $f_m$ is injective, $f_m$ is surjective, hence $f_n$ is surjective.

Since $f_n$ is surjective, $f_n$ is injective, contrary to $$ f_n(n-1) = (2^{-1})(n^2-n) = 0 = f_n(0) $$ hence $f_m$ is not injective, as was to be shown.

Next assume $m$ is a power of $2$.

We want to show that $f_m$ is injective.

The injectivity of $f_1$ is immediate, so assume $m > 1$.

Since $m$ is a power of $2$, so is $2m$.

It follows that all odd integers are invertible mod $2m$.

Now suppose $f_m(x)=f_m(y)$ for some $x,y\in\{0,...,m-1\}$ with $x > y$.

Our goal is to derive a contradiction.

From $f_m(x)=f_m(y)$ we get \begin{align*} & {\small{\frac{x^2+x}{2}}}\equiv{\small{\frac{y^2+y}{2}}}\;(\text{mod}\;m) \\[4pt] \implies\;& x^2+x\equiv y^2+y\;(\text{mod}\;2m) \\[4pt] \implies\;& (x^2+x)-(y^2+y)\equiv 0\;(\text{mod}\;2m) \\[4pt] \implies\;& (x-y)(x+y+1)\equiv 0\;(\text{mod}\;2m) \\[4pt] \end{align*} Consider two cases . . .

Case $(1)$:$\;x-y$ is even.

Then $x+y+1$ is odd, so $x+y+1$ is invertible mod $2m$.

Also we have $1\le x-y\le m-1$, so $x-y$ is not divisible by $2m$.

But then \begin{align*} & (x-y)(x+y+1)\equiv 0\;(\text{mod}\;2m) \\[4pt] \implies\;& x-y\equiv 0\;(\text{mod}\;2m) \\[4pt] \end{align*} contradiction.

Case $(2)$:$\;x-y$ is odd.

Then $x-y$ is invertible mod $2m$.

Also we have $2\le x+y+1\le 2m-2$, so $x+y+1$ is not divisible by $2m$.

But then \begin{align*} & (x-y)(x+y+1)\equiv 0\;(\text{mod}\;2m) \\[4pt] \implies\;& x+y+1\equiv 0\;(\text{mod}\;2m) \\[4pt] \end{align*} contradiction.

Thus both cases yield a contradiction.

It follows that $f_m$ is injective.

This completes the proof.

quasi
  • 61,115