7

A Cayley table of an finite group has to have every element exactly once in every row and exactly once in every column.

Proof that every element of a group has to be at most once in every row and at most once in every column:

Let $(G, \circ)$ be a group and $a, b, c, d \in G$ with:

(I) $a \circ b = d$

(II) $a \circ c = d \Leftrightarrow a = d \circ c^{-1}$

Then:

$\begin{align} (a \circ c) \circ (a \circ b) &= d \circ d \\ \Leftrightarrow d \circ (d \circ c^{-1} \circ b) &= d \circ d \\ \Leftrightarrow d \circ c^{-1} \circ b &= d\\ \Leftrightarrow c^{-1} \circ b &= e\\ \Leftrightarrow b &= c \end{align}$

As the group is finite, this also means it is exactly once in every row/column ($\forall a,b \in G: a \circ b = x$ with $x \in G$).

Now my question is:

Does a group with an infinite number of elements exist, that has not every element in every row/column of its Cayley table?

(I know that Cayley tables usually get used only for finite groups. But if set of the group has a countable number of elements, you can imagine a Cayley table. For example, $(\mathbb{Z}, +)$ has obviously every element in every row/column).

VividD
  • 16,196
Martin Thoma
  • 10,157

3 Answers3

13

The term Cayley table is generally restricted to finite groups. However, it’s certainly possible to generalize the idea. For a group $G$ and an element $a\in G$, the $a$ ‘row’ of the table is essentially just the function $$f_a:G\to G:b\mapsto ab\;,$$ and the $a$ ‘column’ is essentially just the function $$f^a:G\to G:b\mapsto ba\;.$$ If $G$ is countably infinite, you can visualize the Cayley table as an infinite matrix.

Let $G$ be any group, and fix $a\in G$. For each $b\in G$ you have $b=a(a^{-1}b)$, so $b$ appears in row $a$ in column $a^{-1}b$. Similarly, $b=(ba^{-1})a$, so $b$ appears in column $a$ in row $ba^{-1}$. It follows that $b$ appears in every row and column. The cardinality of the group doesn’t matter.

Added: You didn’t ask, but it’s also clear that each element of $G$ appears only once in each row and column: if $ax=ay$ or $xa=ya$, then $x=y$. Thus, each of the maps $f_a$ and $f^a$ for $a\in G$ is a bijection from $G$ onto itself, i.e., a permutation of $G$. The set of all permutations of $G$ is denoted by $\operatorname{Sym}(G)$ and is a group under composition of functions; the maps

$$G\to\operatorname{Sym}(G):a\mapsto f_a$$

and

$$G\to\operatorname{Sym}(G):a\mapsto f^a$$

are isomorphisms of $G$ to subgroups of $\operatorname{Sym}(G)$. This is Cayley’s theorem.

Brian M. Scott
  • 631,399
  • 1
    In other words, the generalization is that $f_a$ and $f^a$ are permutations of $G$. – lhf Aug 29 '12 at 18:53
  • @lhf: Exactly so. – Brian M. Scott Aug 29 '12 at 18:57
  • I think that $a\mapsto f^a$ is an anti-homomorphism. – Kan't Sep 15 '24 at 19:20
  • 1
    @Kan't: (Sorry; I must have missed this back in September.) That would be true iff $f^a(xy)=f^a(y)f^a(x)$ for all $x,y\in G$, i.e., iff $xya=yaxa$ for all $x,y\in G$. But that’s true iff $xy=yax$ for all $x,y\in G$, which is certainly not generally true. If $G$ is Abelian, for instance, it’s true only for $a=1_G$. – Brian M. Scott Dec 02 '24 at 01:38
  • I meant the $f^{{_}}$ is an anti-isomorphism: $f^{ab}(x)=$ $x(ab)=$ $(xa)b=$ $f^b(xa)=$ $f^b(f^a(x))=$ $(f^bf^a)(x)$ for all $x$, namely $f^{ab}=f^bf^a$. – Kan't Dec 02 '24 at 06:04
  • But right-multiplying by $a^{-1}$ solves "the issue". – Kan't Dec 02 '24 at 06:16
  • 1
    @Kan't: Ah, I see: I misread your earlier comment. You’re looking at the map $\varphi:G\to{f^a:a\in G}$ and observing that $\varphi(ab)=\varphi(b)\circ\varphi(a)$ for all $a,b\in G$, so that $\varphi$ is an anti-isomorphism. Yes, absolutely. – Brian M. Scott Dec 02 '24 at 06:38
2

In the general setting, having at least one element in every row and column just means that given any $g$ and any $a$, there exists a $b$ and $b'$ such that $ab = g$ and $b'a = g$. This is true because you can let $b = a^{-1}g$ and $b' = ga^{-1}$.

Having exactly one element in each row or column is equivalent to $ac = g$ and $ac' = g$ implying that $c = c'$. This is because $ac = g = ac'$. Multiplying $a^{-1}$ to both side gives $c = c'$. Do the same thing for $ca = g$ and $c'a = g$.

William
  • 20,245
0

Does a group with an infinite number of elements exist, that has not every element in every row/column of its Cayley table?

Some element $h\in G$ does not occur in some $a$-th row if and only if $ag\ne h$ for every $g\in G$. But for $g:=a^{-1}h$ you get precisely $ag=h$. Therefore, every element occurs in every row(/column) of the (possibly infinite) Cayley's table.

Kan't
  • 4,819