Rationality problems for relation modules of dihedral groups (Q2417802)

This paper is densely packed with results and information. Suppose that $K$ is a Galois field extension of $k$ with Galois group $G=\mathrm{Gal}(K/k)$. For a $G$-lattice $M=\bigoplus_{1\leq i\leq n} \mathbb{Z}\cdot e_i$, denote $K(M):= K(x_1,\dots,x_n)$ and define a $G$-action on $K(M)$ by $\sigma \cdot x_j=\prod_{1\leq i\leq n} x_i^{a_{ij}}$ if $\sigma\cdot e_j= \sum_{1\leq i\leq n}a_{ij}e_i\in M$. Then, the fixed field of this action is denoted $K(M)^G$. Now let $D_n=\langle \sigma, \tau |\sigma^n=\tau^2=1$, $\tau\sigma\tau^{-1}=\sigma^{-1}\rangle$ be the dihedral group with $2n$ elements, $n\geq 2$. Let $F=\langle s_1,s_2\rangle$ denote the free group on two elements. The authors mainly consider the above construction in the case when $G=\mathrm{Gal}(K/k)=D_n$, and $M=R^{ab}:=R/[R,R]$, where $R$ is the kernel of the map \[ F\xrightarrow{\epsilon}D_n, \] defined by $\epsilon(s_1)=\sigma$ and $\epsilon(s_2)=\tau$. Then $R^{ab}$ is canonically a $\mathbb{Z}[D_n]$-module, since conjugation by elements of $F$ defines an action of $D_n$ on $R^{ab}$. This is an example of a \textit{relation module} introduced in the 1970's by Grünberg and Roggenkamp. The authors investigate three main properties of $K(R^{ab})^{D_n}$, namely $k$-rationality, stable $k$-rationality, and retract $k$-rationality. A field $L$ is defined to be \textit{rational} over $k$ when it is isomorphic to $k(x_1,\dots, x_n)$. It is called \textit{stably $k$-rational} when, for some $Y_1, \dots , Y_m$ algebraically independent over $L$, $L(Y_1,\dots, Y_m)$ is rational over $k$. Finally, $L$ is called \textit{retract $k$-rational} when there exists a $k$-algebra domain $A$ with field of fractions $L$ and $k$-algebra morphisms \[ \begin{gathered} \varphi : A\rightarrow k[x_1,\dots, x_n][\frac{1}{f}],\\ \psi : k[x_1,\dots , x_n][\frac{1}{f}]\rightarrow A \end{gathered} \] such that $\psi \circ \varphi= Id_A$, for $f\in k[x_1,\dots, x_n]\smallsetminus \{0\}$. Rational implies stably rational, which in turn implies retract-rational. Now, there are two main results in this paper; first, a number $n\geq 2$ is an odd integer if and only if $K(R^{ab})^{D_n}$ is stably $k$-rational, in which case $K(R^{ab})^{D_n}$ is also rational over $k$. (The positive part of the implication follows from previous results by Endo and Miyata, but rationality is proved using additional techniques.) Moreover, if $K(R^{ab})^{D_n}$ is not stably rational over $k$ (i.e. $n$ is even) and $k$ is an infinite field, then $K(R^{ab})^{D_n}$ is not retract-rational over $k$. In addition, the authors also investigate $k(R^{ab})^{D_n}$. They prove that, if $n$ is odd, then $k(R^{ab})^{D_n}$ is isomorphic to $k(D_n)(t)$ for any field $k$, where $t$ is a transcendental element over the field $k(D_n)$. Here $k(G)=k(x_g\mid g\in G)^G$ where $G$ acts on $x_g$ by permutation. Thus, the question of rationality of $k(G)$ is the Noether problem. It follows then that in cases where the Noether problem has a positive solution for $D_n$ (for example, when $\zeta_n+\zeta_n^{-1}\in k$), $k(D_n)^{D_n}$ is always rational over $k$ for $n$ odd. A 1978 theorem of Snider stated that $k(R^{ab})^{D_2}$ is always rational (of which the authors also give an alternate proof) and for $n\geq 3$, $k(R^{ab})^{D_n}$ is rational when $\zeta_n\in k$. One important tool the authors use is given by previous general results on the stable rationality and retract-rationality of $K(M)^G$. Voskresenskii, Endo and Miyata proved that $K(M)^G$ is stably $k$-rational if and only if there exists a resolution \[ 0\rightarrow M\rightarrow P\rightarrow E\rightarrow 0\tag{1} \] with $P,E$ a permutation lattices (i.e. free abelian groups on finite $G$-sets). Saltman proved that if $k$ is an infinite field, then $K(M)^G$ is retract $k$-rational if and only if there exists a resolution (1) where $P$ is a permutation lattice and $E$ is a summand of a permutation lattice. It is worth noting that the proofs of Lemma 3.11 and Theorem 4.6 in this paper are due to the referee.