Henselianity in the language of rings (Q2636534)

Consider the following four properties of characteristic zero fields $K$ in the language $\mathcal{L}_{\text{ring}} = \{0,1,+,-,\cdot\}$: \begin{itemize} \item[(\textbf{h})] $K$ is Henselian (i.e., $K$ admits a nontrivial Henselian valuation), \item[(\textbf{eh})] every $L$ which is $\mathcal{L}_{\text{ring}}$-elementarily equivalent to $K$ is Henselian, \item[($\emptyset$-\textbf{def})] $K$ admits a $\emptyset$-definable nontrivial Henselian valuation, and \item[(\textbf{def})] $K$ admits a definable nontrivial Henselian valuation. \end{itemize} This paper considers the following question: What are the general relationships and implications between these four properties? If $K$ is algebraically closed then the answer is clear: $K$ has properties (\textbf{h}) and (\textbf{eh}) and does not have properties (\textbf{$\emptyset$-def}) and (\textbf{def}), by strong minimality. Accordingly, the authors focus their attention on the class of fields $\mathcal{K}_0$ of all non-algebraically closed fields of characteristic zero. Using the canonical Henselian valuation $v_K$ of $K$ (i.e., the coarsest Henselian valuation on $K$ with separably closed residue field if such a valuation exists, otherwise the finest Henselian valuation on $K$), they further refine $\mathcal{K}_0$ into subclasses: \[ \mathcal{K}_{0,0} \ := \ \{K\in \mathcal{K}_0: \text{char}(Kv_K)=0\} \] and for any prime $p$: \[ \mathcal{K}_{0,p} \ := \ \{K\in\mathcal{K}_0: \text{char}(Kv_K)=p\}. \] The main result of the paper is the following: {Theorem.} \begin{itemize} \item[(A)] In the class $\mathcal{K}_{0,0}$ the complete picture is \[ \begin{tikzcd} (\emptyset\text{-}\mathbf{def}) \arrow[d,Rightarrow] \arrow[r, Leftrightarrow]& (\mathbf{eh}) \arrow[d,Rightarrow]\\ (\mathbf{def}) \arrow[r,Rightarrow]& (\mathbf{h}) \\ \end{tikzcd} \] \item[(B)] For each prime $p$, in the class $\mathcal{K}_{0,p}$ the complete picture is \[ \begin{tikzcd} (\emptyset\text{-}\mathbf{def}) \arrow[d,Leftrightarrow] \arrow[r,Rightarrow] & (\mathbf{eh}) \arrow[d,Leftrightarrow] \\ (\mathbf{def}) \arrow[r,Rightarrow] & (\mathbf{h}) \\ \end{tikzcd} \] \end{itemize} Here, ``complete picture'' refers to the fact that if an implication is given in one direction $\Longrightarrow$, then the theorem says the converse $\Longleftarrow$ does not hold in that class. Several components of the theorem were known previously. The main technical results of this paper are as follows: \begin{itemize} \item[(1)] They show for $K\in\mathcal{K}_{0,0}$ that (\textbf{eh}) implies (\textbf{$\emptyset$-def}). This uses a result on defining Henselian valuations from [the second author and \textit{J. Koenigsmann}, Proc. Edinb. Math. Soc., II. Ser. 60, No. 3, 665--687 (2017; Zbl 1436.03199)]. \item[(2)] They construct a non-Henselian $t$-Henselian field of characteristic $p$ (prime or zero) of divisible-tame type which is not separably closed. This is used to show that (\textbf{h}) does not imply (\textbf{$\emptyset$-def}) in $\mathcal{K}_{0,p}$. \item[(3)] They show that if $(K,v_K)$ has mixed-characteristic, then (\textbf{def}) implies (\textbf{$\emptyset$-def}). This relies heavily on the machinery of $q$-Henselian valuations as developed in [the second author and \textit{J. Koenigsmann}, Ann. Pure Appl. Logic 166, No. 7--8, 741--754 (2015; Zbl 1372.03077)]. \end{itemize}