When is a separating map biseparating? (Q2365008)

\(C(X)\) and \(C(Y)\) denote the rings of real- or complex-valued functions on the Tikhonov spaces \(X\) and \(Y\) endowed with their respective compact-open topologies. For a subalgebra \(W\) of \(C(X)\), an additive map \(T:W\to C(Y)\) is separating if \(fg=0\Rightarrow TfTg=0\); it is biseparating if it is bijective and \(T^{-1}\) is separating. A reason for interest in biseparating maps is the generalization of the Gelfand-Kolmogorov-Hewitt theorem, namely that the existence of a biseparating map \(T:C(X)\to C(Y)\) implies that \(\upsilon X\) is homeomorphic to \(\upsilon Y\). In the context of normed vector (Archimedean) lattices \(E\) and \(F\), Abramovich [\textit{C. B. Huijsmans} and \textit{W. Luxemburg} (eds.), Positive operators and semigroups on Banach lattices (1992; Zbl 0783.47052), p. 143] asked if separating linear bijections \(T:E\to F\) are biseparating. \textit{C. B. Huijsmans} and \textit{B. de Pagter} [Proc. Edinb. Math. Soc., II. Ser. 37, 125-132 (1994; Zbl 0807.47024), Cor. 2.2] and \textit{A. V. Koldunov} [Proc. Am. Math. Soc. 123, 1083-1095 (1995; Zbl 0827.47051), Theorem 3.6] affirmed this for linear separating bijections when \(E\) is a uniformly complete vector lattice and \(F\) a normed vector lattice. We move in the same direction here in the context of spaces \(C(X)\), \(C(Y)\) of continuous functions (which need not be normed vector lattices). Using purely topological methods, we show in Theorem 16 that if \(Y\) is connected then a separating linear bijection \(T:C(X)\to C(Y)\) is biseparating; in Theorem 17 we show that an analogous result holds if \(X\) is 0-dimensional or \(Y\) is pseudocompact. In Section 4 overlap with lattices disappears altogether: we show that the analogues of the main results also hold for 0-dimensional \(X\) and \(Y\) when the continuous functions take values in a non-Archimedean valued field.