Abstract:Let $K$ be the fraction field of a two-dimensional henselian, excellent, equi-characteristic local domain. We prove a local-global principle for Galois cohomology with finite coefficients over $K$. We use classical machinery from étale cohomology theory, drawing upon an idea in Saito's work on two-dimensional local class field theory. This approach works equally well over the function field of a curve over an equi-characteristic henselian discrete valuation field, thereby giving a different proof of (a slightly generalized version of) a recent result of Harbater, Hartmann and Krashen. We also present two applications. One is the Hasse principle for torsors under quasi-split semisimple simply connected groups without $E_8$ factor. The other gives an explicit upper bound for the Pythagoras number of a Laurent series field in three variables. This bound is sharper than earlier estimates.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is about the cohomological Hasse principle on two - dimensional local rings. Specifically, the author tries to prove that under certain conditions, Galois cohomology groups satisfy the local - global principle. The following is the specific description of the problem:
### Research Background and Problem Description
#### 1. Cohomological Hasse Principle
For a two - dimensional Henselian excellent equicharacteristic local field \(K\), the author hopes to prove that for some Galois cohomology groups with finite coefficients, there exists a local - global principle. That is, for given integers \(r\geq2\) and \(n\geq1\), consider whether the following natural map is injective:
\[H^{r}(K,\mathbb{Z}/n(r - 1))\to\prod_{v\in\Omega_{K}}H^{r}(K_{v},\mathbb{Z}/n(r - 1))\]
where \(\Omega_{K}\) represents the set of all normalized discrete valuations on \(K\), and \(K_{v}\) is the completion of \(K\) at \(v\).
#### 2. Hasse Principle for Torsor Spaces
As one of the applications of the above - mentioned cohomological Hasse principle, the author also studies the local - global principle for torsors of a semisimple, simply - connected algebraic group \(G\). Specifically, consider whether the following natural map has a trivial kernel:
\[H^{1}(K,G)\to\prod_{v\in\Omega_{K}}H^{1}(K_{v},G)\]
### Main Results
The author has successfully proved that in the equicharacteristic case, the answers to the above two problems are positive. The specific results are as follows:
#### Theorem 1.2
Suppose \(A\) is an equicharacteristic Henselian excellent normal local domain, and denote \(K\) as its fraction field. Then for any \(r\geq2\), the natural map
\[H^{r}(K,\mathbb{Z}/n(r - 1))\to\prod_{v\in X^{(1)}}H^{r}(K_{v},\mathbb{Z}/n(r - 1))\]
is injective, where \(X^{(1)}\) represents the set of discrete valuations corresponding to all points of codimension 1 in \(X\).
#### Theorem 1.3
Suppose \(G\) is a quasi - split semisimple simply - connected group without \(E_{8}\) factors, and the order of the Rost invariant of each simple factor is not divisible by \(\text{char}(K)\), then the natural map
\[H^{1}(K,G)\to\prod_{v\in\Omega_{K}}H^{1}(K_{v},G)\]
has a trivial kernel.
### Applications
#### 1. Hasse Principle for Torsor Spaces
Theorem 1.3 provides the Hasse principle for many types of semisimple algebraic groups, especially for quasi - split semisimple simply - connected groups.
#### 2. Upper Bound Estimation of Pythagoras Numbers
As another application, the author gives a more precise upper bound estimation of the Pythagoras number of the Laurent series field in three variables. The specific results are as follows:
#### Theorem 1.4
Let \(k\) be a field with characteristic different from 2, and \(r\geq2\) be an integer. If \(p(k(x,y))\leq2^{r}\), then
\[p(k((x,y,z)))\leq p(k((x,y))(t))\leq p(k((x,y))((z,t)))\leq2^{r}\]
If \(p(k(x,y)) = 2^{r}\), then
\[p(k((x,y,z)))=p(k((x,y))(t))=p(k((x,y))((z,t)))=2^{r}\]
These results not only improve the previous estimations but also are applicable to more types of fields, such as those finitely generated over \(\mathbb{R}\) or \(\mathbb{C}\)