Where have I gone wrong in calculating norm residue symbol.

298 Views Asked by At

It is well known that the reciprocity map in local class field theory gives reciprocity map in Global class field theory. Namely, if $(\cdot,L_\beta/K_P):K_P^\times \rightarrow G(L_\beta/K_P)$ is the local reciprocity map for place $P$ in global field $K$, then global Artin map $[\cdot,L/K]$:$I_K\rightarrow G(L/K)$ is given by $\prod_P (\cdot,L_\beta/K_P)$. The kernel of this contains $K^\times$ and I understand this. But I tried to verify something explicitly and getting negative result so I would like to know where I have gone wrong.

So now let $K$ be a rational function field over one variable over finite field $\mathbb{F}_q$ where $q=p^n$ for a prime $p$ and $L$ be its constant extension of degree $m$.

Let $Q$ be a place in $K$ which does not split in $L$ and $f_Q$ be corresponding irreducible polynomial.

Then $[f_Q,L/K]=\prod(f_Q,L_\beta/K_P)$ where $\beta|P$. All factors in the right hand side is 1 except for the case $P=Q$ and $P=\infty$ as all extension $L_\beta/K_P$ is constant and hence unramifed and image of $f_Q$ in those field is a unit. In the infinity case I get $\phi^{-deg(Q)}$ and when $P=Q$ we get $\phi$ where $\phi$ is used to denote the generator of $G(L/K)$. This is because $G(L/K)\cong G(L_{\beta'}/K_Q)\cong G(L_\infty/K_\infty)$ as those places do not split. (Also because $(a,\bar{\mathbb{F}_q}((T))/\mathbb{F}_q((T)))=\phi^{v(a)}$ where in this case I mean $\phi$ as a generator of $G(\bar{\mathbb{F}_q}((T))/\mathbb{F}_q((T)))$

But there is no reason $-\deg(Q)+1$ must be divisible by $n$. Obviously I am wrong somewhere but I can't seem to figure out at the moment.

2

There are 2 best solutions below

1
On BEST ANSWER

You are beginning with a global field $K$ of char. $p$, containing $\mathbb F_q$ as its maximal finite subfield, and forming the (everywhere unramified) extension $L = \mathbb F_{q^m} \otimes_{\mathbb F_q} K.$

If $\mathfrak p$ is a prime of $K$, with residue field $\kappa(\mathfrak p)$, then the primes $\mathfrak q$ of $L$ over $\mathfrak p$ correspond to the various field factors of the tensor product $\mathbb F_{q^{m}} \otimes_{\mathbb F_q} \kappa(\mathfrak p).$

Note that Gal$(L/K)$ is canonically isomorphic to Gal$(\mathbb F_{q^m}/\mathbb F_q)$. If $\mathfrak q$ lies over $\mathfrak p$, then the extension $\kappa(\mathfrak q)/\kappa(\mathfrak p)$ is a direct factor of $\mathbb F_{q^m} \otimes_{\mathbb F_q} \kappa(\mathfrak p).$ This tensor product has an obvious action of Gal$(\mathbb F_{q^m}/\mathbb F_q)$ (via its action on the left factor in the tensor product), and the decomposition group $D_{\mathfrak q}$ is the subgroup which preserves $\kappa(\mathfrak q)$.

In particular, if $\kappa(\mathfrak p) = \mathbb F_q$, then the tensor product is just equal to $\mathbb F_q$, and so there is a single prime $\mathfrak q$ over $\mathfrak p$, of degree $n$. Furthermore, in this case $D_{\mathfrak q} = $ Gal$(L/K)$.


As in your post, let's now take $K = \mathbb F_q(x)$.

If $\mathfrak p$ corresponds to the point $x = 0$, then the residue field is $\mathbb F_q,$ and so there is a single prime lying over infinity, of degree $m$.

If $\mathfrak p$ is the point at infinity, then $\kappa(\mathfrak p) = \mathbb F_q,$ and thus there is also a single prime lying over infinity, of degree $m$.

In each case the decomposition group is equal to all of Gal$(L/K) =$ Gal$(\mathbb F_{q^m}/\mathbb F_q)$. In particular, in each case the Frobenius element at $\mathfrak p$ is equal to the Frobenius element Frob${}_q$ in Gal$(\mathbb F_{q^m}/\mathbb F_q)$.

Let's take $f = x$. Then the norm residue symbol of $x$ at $\mathfrak p$ is trivial except for $\mathfrak p$ corresponding to $0$ or $\infty$.

At $0$, we see that $x$ has a zero of order $1$, while at $\infty$, it has a pole of order $1$. Thus the Artin map sends $x$ to Frob${}_q \cdot $Frob${}_q^{-1}$, i.e. to $1$, which is what you wanted to check.

You should try some other examples yourself. E.g. if $f$ is irreducible of degree two, and $n$ is odd, then there will again be a unique prime $\mathfrak q$ over $\mathfrak p$, and $\kappa(\mathfrak q) = \mathbb F_{q^{2n}}.$ The Frobenius element Frob${}_{\mathfrak p}$ is then equal to Frob${}_{q^2} = $Frob${}_{q}^2$.

The order of the pole of $f$ at $\infty$ is $2$, so in this case the global Artin map takes $f$ to Frob${}_{q^2}\cdot $ Frob${}_q^{-2} = $ Frob${}_q^2 \cdot$ Frob${}_q^{-2} = 1$ (as it must!).

0
On

I am not 100 percent sure but I think I figured it out.

This happens because image of $\phi$ in $G(L_{\beta'}/K_Q)$ isn't necessarily cannonical Frobenius generator. As those two groups are isomorphic it will be a generator but not necessarily the one we want to make use of local theory.