2

Using this logic, there should be total $p^{p^2}$ functions $c: C_p \times C_p \to \Bbb Z/p\Bbb Z$ where $p$ is a prime.

Out of these many functions, is there any simple way to see how many (and which) functions satisfy the 2-cocycle identity $$c(b_2,b_3) - c(b_1 b_2,b_3) + c(b_1,b_2 b_3) - c(b_1,b_2) = 0$$ where $b_i \in C_p$, i.e., $b_i \in \{0, 1, 2, \ldots ,p-1\}$?

It's certainly impractical to write down all the $p^{p^2}$ possible functions explicitly for $p>2$.

Note: We have not been taught cohomology. Please don't use results from there.

Captain Lama
  • 25,743

1 Answers1

3

You can use basic exact sequences of cohomology to compute that number. In what follows, to keep notations light, I will omit the cyclic groups, so for instance $H^2=H^2(C_p,\mathbb{Z}/p\mathbb{Z})$. What you want to compute is the order of $Z^2$.

We have $0\to B^2\to Z^2\to H^2\to 0$ and $0\to Z^1\to C^1\to B^2\to 0$, so we are done if we can compute the order of $H^2$, $Z^1$ and $C^1$. Explicitly, $|Z^2|=\frac{|H^2|\cdot |C^1|}{|Z^1|}$.

Now $C^1$ is just the set of functions $C_p\to \mathbb{Z}/p\mathbb{Z}$, so $|C^1|=p^p$. Also $Z^1$ is the set of group morphisms $C_p\to \mathbb{Z}/p\mathbb{Z}$, so $|Z^1|=p$.

Finally, the most complicated is $H^2$, but the cohomology of cyclic groups is very well-known, and $H^2=\mathbb{Z}/p\mathbb{Z}$, so $|H^2|=p$.

In the end, $|Z^2|=p^p$.


EDIT: What is written above is the "correct" way to compute the number you are looking for. But since you want a computation that does not explicitly involve cohomology, here is a completely elementary proof.

We can write an explicit bijection between $Z^2$ and the set $X$ of functions $C_p\to \mathbb{Z}/p\mathbb{Z}$: to each $c\in Z^2$ we associate $$\varphi_c: x\mapsto c(x,1),$$ and to each $\varphi\in X$ we associate $$c_\varphi: (x,y)\mapsto \varphi(0) + \sum_{i=0}^{y-1}(\varphi(x+i)-\varphi(i))$$ where we assume that $1\leqslant y\leqslant p$.

The trickiest part is to show that $c_\varphi$ is actually a $2$-cocycle. We will do that at the end. For now let's assume this is true, and show that the above associations are mutually inverse bijections. It is immediate that $c_\varphi(x,1)=\varphi(x)$, so in other words $\varphi_{c_\varphi}=\varphi$. It is a little less obvious that $$ \varphi_c(0) + \sum_{i=0}^{y-1}(\varphi_c(x+i)-\varphi_c(i)) = c(x,y)$$ (in other words $c_{\varphi_c}=c$), but not very hard to check either. We can proceed inductively: if $y=1$, then the equality just means that $\varphi_c(0)+\varphi_c(x)-\varphi_c(0)=c(x,1)$, which is the definition of $\varphi_c$. Now if the equality is true for some $1\leqslant y<p$, then using the cocycle condition: $$\begin{align} c(x,y+1) & = c(x,y)+c(x+y,1)-c(y,1) \\ & = c(x,y)+\varphi_c(x+y)-\varphi_c(y) \\ & = \varphi_c(0) + \sum_{i=0}^{y-1}(\varphi_c(x+i)-\varphi_c(i))+\varphi_c(x+y)-\varphi_c(y) \\ &= \varphi_c(0) + \sum_{i=0}^{y}(\varphi_c(x+i)-\varphi_c(i)) \end{align}$$ as we wanted. In the end $c\mapsto \varphi_c$ and $\varphi\mapsto c_\varphi$ are inverse of each other, which shows that $|Z^2|=|X|=p^p$.

It remains to show that $c:=c_\varphi$ is a $2$-cocycle. We want to establish that if $1\leqslant x,y,z\leqslant p$ $$c(y,z)+c(x,y+z) = c(x+y,z)+c(x,y).$$

First, suppose $y+z\leqslant p$. Then on the left-hand side, we have $$2\varphi(0)+\sum_{i=0}^{z-1}\varphi(y+i)-\varphi(i) + \sum_{i=0}^{y+z-1}\varphi(x+i)-\varphi(i)$$ and on the right-hand side: $$2\varphi(0)+\sum_{i=0}^{z-1}\varphi(x+y+i)-\varphi(i) + \sum_{i=0}^{y-1}\varphi(x+i)-\varphi(i).$$ So clearing the terms that appear on both sides, the equality amounts to $$\sum_{i=0}^{z-1}\varphi(y+i) + \sum_{i=y}^{y+z-1}\varphi(x+i)-\varphi(i) = \sum_{i=0}^{z-1}\varphi(x+y+i),$$ which is easy to check since $\sum_{i=y}^{y+z-1}\varphi(x+i)-\varphi(i) = \sum_{i=0}^{z-1}\varphi(x+y+i)-\varphi(y+i)$.

Finally, suppose $y+z>p$; let us write $y+z=p+r$, with $r\geqslant 1$. Note that $r\leqslant y$. Then the left-hand side is $$2\varphi(0)+\sum_{i=0}^{z-1}\varphi(y+i)-\varphi(i) + \sum_{i=0}^{r-1}\varphi(x+i)-\varphi(i)$$ and the right-hand side is the same: $$2\varphi(0)+\sum_{i=0}^{z-1}\varphi(x+y+i)-\varphi(i) + \sum_{i=0}^{y-1}\varphi(x+i)-\varphi(i).$$ If we simply the terms appearing on both sides, we get the equality $$\sum_{i=0}^{z-1}\varphi(y+i) = \sum_{i=0}^{z-1}\varphi(x+y+i) + \sum_{i=r}^{y-1}\varphi(x+i)-\varphi(i).$$ The left-hand side can be rewritten as $$\sum_{i=y}^{p-1}\varphi(i)+\sum_{i=0}^{r-1}\varphi(i)$$ and the right-hand side as $$\sum_{i=y}^{p-1}\varphi(x+i) + \sum_{i=0}^{r-1}\varphi(x+i) + \sum_{i=r}^{y-1}\varphi(x+i) - \sum_{i=r}^{y-1}\varphi(i) = \sum_{i=0}^{p-1}\varphi(i)-\sum_{i=r}^{y-1}\varphi(i)$$ and now both sides are easily seen to be equal.

Captain Lama
  • 25,743
  • Why is the number of group morphisms $C_p \to \Bbb Z/p\Bbb Z$ equal to $p$? Is this a theorem? Also group morphism means the same thing as group homomorphism, right? –  Apr 20 '20 at 13:13
  • Also, well, we're not supposed to use any cohomology result as such... –  Apr 20 '20 at 13:18
  • 1
    I edited to give a completely elementary proof, although of course it's a lot longer and a lot less natural. – Captain Lama Apr 21 '20 at 17:56
  • Thanks for the edit; this looks helpful. However, I do not understand why $c \mapsto \varphi_c$ and $\varphi \mapsto c_{\varphi}$ being inverses of each other implies $|Z^2| = |X| = p^p$. I know that the number of possible functions $C_p \to \Bbb Z/p$ is $p^p$. –  Apr 22 '20 at 14:50
  • 1
    When you have a bijection between two sets, they have the same number of elements. When two functions are inverse of each other, they are bijections. – Captain Lama Apr 22 '20 at 15:23