6
$\begingroup$

Suppose $K=\mathbb{F}_q$ is a finite field of characteristic 3. Suppose $q > 3$, then is it true that for all $x \in K$ there exists $a,b,c \in K^{\times}$ such that $x = a^2 + b^2 + c^2$ i.e, is every element the sum of 3 non-zero squares?

My [wrong] idea is: fix $x \in K$. Note that since $q > 3$, there exists $\gamma \in K$ such that $\gamma^2 = -1$. Case 1, if $x = k^2$ for some $k \in K$ then $x = k^2 + \gamma^2 + 1^2$. Case 2, if $x$ is not a square then I'd like to say that either $x + 1$ or $x-1$ is a square and so $x = k^2 + \gamma^2 + \gamma^2$ or $x = k^2 + 1^2 + 1^2$ for some $k \in K$.

But I'm not sure if it's true that, given $x \in K$, one of $x, x+1, x-1$ is a square.

Just realised my idea is nonsense since such a $\gamma$ needn't exist. Take $K = \mathbb{F}_{27}$

$\endgroup$
2
  • $\begingroup$ You can add the [additive-combinatorics] tag. Also an idea that you might find useful. In any group $G$, let $S$ be a set of size greater than $|G|/2$. Then $S+S=G$ because for any $x \in G$, the two sets $S$ and $x-S$ must have an element in common. $\endgroup$
    – Aravind
    Apr 5 '17 at 11:49
  • 1
    $\begingroup$ What you have is a set (non-zero squares), which is very close to half the size of the field (less by 1), but you are considering $S+S+S$, so a small modification to the above argument should work. $\endgroup$
    – Aravind
    Apr 5 '17 at 11:51
2
$\begingroup$

The following works for any finite field $K$ of char $\neq 2$ and size more than $5$.

Idea 0: In any finite group $G$, if $A$ is a set of size greater than $|G|/2$, then $A+A=G$.

Let $S$ be the set of non-zero squares in $K$ and $S_0=S \cup \{0\}$. Using Idea 0, we have: $K=S_0+S_0$.

Idea 1: It is sufficient to show that at least one square of $K^*$ belongs to $3S$, at least one non-square of $K^*$ belongs to $3S$, and $0 \in 3S$. This is because $3S$ is closed under multiplication by non-zero squares.

Idea 2: Let $a \in K$ with $a \notin \{-1,1\}$. Then $a$ can be written as $x^2-y^2$ with $x,y \neq 0$. Just take $x=(a+1)/2$ and $y=(a-1)/2$.

Claim 1: $K^* \subseteq S+S$.

From $K=S_0+S_0$, we know that every non-square can be written as $a^2+b^2$ and clearly $a,b \neq 0$. Thus, every non-square is in $S+S$. Now I show that all non-zero squares are in $S+S$. Using Idea 2, write a non-zero square not in $\{-1,1\}$ as $x^2-y^2$ with $x,y \neq 0$. This shows that $x^2 \in S+S$ and hence all non-zero squares are in $S+S$. This proves the claim.

Now consider $3S$ which contains $K^*+S$. Clearly, $|3S| \geq |K^*|$, therefore $3S$ misses at most one element of $K$. Using the fact that $3S$ is closed under multiplication by $S$, we see that it cannot miss any element other than zero.

To show that $0 \in 3S$, consider an arbitrary non-zero $a$ and write $-a^2=b^2+c^2$ with $b,c \neq 0$ using Claim 1.

$\endgroup$
0
2
$\begingroup$

We write $A+B$ for $\{a+b:a\in A, b\in B\}$.

When $\mathbb F$ is a finite field of order $3$ or $5$, not every element of $\mathbb F$ is a sum of $3$ nonzero squares: in $\mathbb F_3$ there is only one nonzero square, and in $\mathbb F_5$ the only nonzero squares are $1$ and $-1$.

For all other fields of odd order we have the following.

Proposition. If $\mathbb F$ is a finite field of odd order, and $|\mathbb F|\neq 3, 5$, then $\mathbb F = S+S+S$, where $S$ is the set of nonzero squares in $\mathbb F$.

This can be proved using only the cardinality of $S$.

Lemma. If $G$ is a finite abelian group of odd order $\neq 3,5$ and $A\subseteq G$ has $|A|=\frac{|G|-1}{2}$, then $A+A+A=G$.

Observing that $|S|=\frac{|\mathbb F|-1}{2}$ when $\mathbb F$ is a finite field of odd order, the Proposition is a special case of the Lemma.

To prove the Lemma we need:

Fact 1. If $A,B\subseteq G$ have $|A|+|B|>|G|$, then $A+B=G$.

Fact 1 follows from the observation that $(g-A)\cap B \neq \emptyset$ for every $g\in G$ whenever $|A|+|B|>|G|$.

We also need a theorem of Kneser (see Nathanson's book):

Theorem 1. If $A,B$ are finite subsets of an abelian group satisfying $|A+B|<|A|+|B|$ and $H$ is the group of elements $h\in G$ satisfying $A+B=A+B+h$, then $$|A+B| = |A+H|+|B+H| - |H|. \tag{1}\label{K}$$ The group $H$ is called the stabilizer of $A+B$.

(Theorem 1 is a special case of Satz 1 of Kneser's original article where $G$ is discrete.)

To prove the Lemma, assume $|G|>3$ (so that $|A|>1$) and consider two cases.

Case 1. $|A+A|\geq 2|A|.$ In this case we have $|A+A|+|A| > |G|$, so Fact 1 implies $A+A+A = G$.

Case 2. $|A+A|<2|A|.$ In this case Theorem 1 applies, so let $H$ be the stabilizer of $A+A$. Equation (\ref{K}) becomes $$|A+A| = 2|A+H| -|H|. \tag{2}\label{SpecialK}$$ Observe that $A+A=A+A+H = (A+H)+(A+H)$.

Claim. If $A+A+A\neq G$, then $|A+H|=|A|$, and $H = \{0\}.$

If $|A+H|>|A|$, then $2|A+H|>|G|$. Then Fact 1 and the identity $A+A = (A+H)+(A+H)$ imply $A+A=G$. So we conclude that $|A+H|=|A|$, meaning $A$ is a union of cosets of $H$. Consequently $|H|$ divides $|A| = \frac{|G|-1}{2}$, so $|H|$ is relatively prime to $|G|$. But $H$ is a subgroup of $G$, so its order divides $|G|$, and we conclude that $|H|=1$.

The Claim allows us to assume $|H|=1$, so Equation (\ref{SpecialK}) becomes $|A+A|=2|A|-1 = |G|-2$. Applying Fact 1 to $(A+A)+A$ we conclude that either $A+A+A=G$ or $|G|-2+|A| \leq |G|$. So if $A+A+A\neq G$, then $|A|\leq 2$, meaning $|G|=3$ or $|G|=5$. $\square$

$\endgroup$

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.