On small homotopies of loops (Q929992)

Anomalous behavior in homotopy theory arises when an essential map is the uniform limit of inessential maps. Such behavior manifests itself in such oddities as pointed unions of contractible spaces being non-contractible, and (infinite) concatenations of nulhomotopic loops being essential [cf. \textit{J. W. Cannon} and \textit{G. R. Conner}, Topology Appl. 106, 225-271 (2000; Zbl 0955.57002)]. Often topologists attempt to control such behavior by requiring ``small'' maps to be nulhomotopic. This is the flavor of the k-ULC property from geometric topology. In the present article there are two natural notions of ``small'' curves which are studied by the authors -- curves which can be homotoped into arbitrarily small neighborhoods of a point, and curves which can be uniformly approximated by nulhomotopic curves. This article describes how various embodiments of these notions are related. Informally, the article is meant to clarify the following question. Question 1.1. If \(X\) is a space in which small nulhomotopic loops bound small homotopies, then is a loop which is the uniform limit of a family of nulhomotopic loops necessarily nulhomotopic? The answer is in the negative. Formally, the paper studies two relatively new and subtly different separation axioms: homotopically Hausdorff and \(\pi_1\)-shape injective. These notions were introduced in a number of papers by \textit{J. W. Cannon} and \textit{G. R. Conner} [op. cit., and Topology Appl. 153, 2648--2672 (2006; Zbl 1105.55008)], \textit{G. R. Conner} and \textit{H. Fisher} [Topology Appl. 129, 73--78 (2003; Zbl 1025.57005)], \textit{G. R. Conner} and \textit{J. Lamoreaux} [Fundam. Math. 187, 95--110 (2005; Zbl 1092.57001)], \textit{A. Zastrow} [Generalized \(\pi_1\)-determined covering spaces, preprint (1995)] and were put to good use by \textit{H. Fisher} and \textit{A. Zastrow} [Algebr. Geom. Topol. 5, 1655--1676 (2005; Zbl 1086.55009)], and \textit{H. Fisher} [Fundam. Math. 197, 167--196 (2007; Zbl 1137.55006)]. The intuition behind a space being homotopically Hausdorff is that curves which can be homotoped into arbitrarily small neighborhoods of a point are nulhomotopic, whereas in a \(\pi_1\)-shape injective space one has the intuition that curves which can be homotoped arbitrarily close to a nulhomotopic curve are themselves nulhomotopic. The authors show that the property of small nulhomotopic loops bounding small nulhomotopies implies homotopically Hausdorff. This motivates the authors to formulate the following more formal question. Question 1.2. Does the homotopically Hausdorff property imply \(\pi_1\)-shape injectivity? As solution the authors construct two examples \(A\) and \(B\), neither of which is \(\pi_1\)-shape injective, by rotating a topologist's sine curve in \(\mathbb{R}^3\) to create a ``surface'' and adding a null sequence of arcs to make the space locally path connected. Both spaces are homotopically Hausdorff and \(B\) is strongly homotopically Hausdorff. For the sake of completeness, the authors show that shape injective implies strongly homotopically Hausdorff, and strongly homotopically Hausdorff implies homotopically Hausdorff. The proof that \(A\) and \(B\) have the desired properties requires a lemma which answers, in the negative, the following natural question. Question 1.3. Can adding arcs to a space turn an essential loop into a nulhomotopic loop?