Topological complexity of unordered configuration spaces of surfaces (Q2311442)

Recall that the (normalized) topological complexity $\mathsf{TC}(X)$ of a space $X$ is the minimal integer $k$ such that $X\times X$ admits an open cover $\{U_0,\ldots ,U_k\}$ by $k+1$ sets on which the free path fibration $\pi:X^{[0,1]}\to X\times X$ given by $\pi(\gamma)=(\gamma(0),\gamma(1))$ admits a local section $s_i:U_i\to X^{[0,1]}$. This notion, motivated by robotics applications, was introduced in [\textit{M. Farber}, Discrete Comput. Geom. 29, No. 2, 211--221 (2003; Zbl 1038.68130); Topology Appl. 140, No. 2--3, 245--266 (2004; Zbl 1106.68107)]. Given a surface $S$ and natural number $n$, let $F(S,n)$ denote the ordered configuration space of $n$-tuples of distinct points in $S$. The unordered configuration space $C(S,n)=F(S,n)/\mathfrak{S}_n$ is the quotient of the ordered configuration space by the action of the symmetric group $\mathfrak{S}_n$ which permutes the coordinates. In this nicely written paper, the authors give estimates for $\mathsf{TC}(C(S,n))$, and in many cases compute the exact value. The main results may be summarized as follows (see Theorems 1.1 and 1.3): (a) If $S$ is a punctured closed surface other than the disc, the annulus or the Möbius band, then $\mathsf{TC}(C(S,n))=2n$; (b) If $S$ is a closed surface other than the sphere or the projective plane, then $2n\leq \mathsf{TC}(C(S,n))\leq 2n+2$; (c) If $A$ is the annulus, then $\mathsf{TC}(C(A,n))=2n-1$; (d) If $M$ is the Möbius band, then $2n-1\leq \mathsf{TC}(C(M,n))\leq 2n$; (e) If $D$ is the disc, then $n-1+\mathsf{cd}([P_n,P_n])\leq \mathsf{TC}(C(D,n))\leq 2n-2$. In part (e), $\mathsf{cd}([P_n,P_n])$ denotes the cohomological dimension of the commutator subgroup of the $n$-strand pure braid group $P_n$, which the authors conjecture equals $n-2$. (The first author of the present paper has in the meantime found a proof of this conjecture, posted as [``On the homology of the commutator subgroup of the pure braid group'', Preprint, \url{arXiv:1905.05099}].) This would give $2n-3\leq \mathsf{TC}(C(D,n))\leq 2n-2$ in the disc case. The authors show that $2n-3 = \mathsf{TC}(C(D,n))$ in case $n=3$, and conjecture this to be the case for all $n$. Until this paper appeared, almost all non-trivial computations of the topological complexity of configuration spaces had concerned the ordered versions. See [\textit{M. Farber} and \textit{S. Yuzvinsky}, in: Geometry, topology, and mathematical physics. Selected papers from S. P. Novikov's seminar held in Moscow, Russia, 2002--2003. Providence, RI: American Mathematical Society (AMS). 145--156 (2004; Zbl 1088.68171); \textit{M. Farber} and \textit{M. Grant}, Proc. Am. Math. Soc. 137, No. 5, 1841--1847 (2009; Zbl 1172.55006)] for configuration spaces of Euclidean spaces, and [\textit{D. C. Cohen} and \textit{M. Farber}, Compos. Math. 147, No. 2, 649--660 (2011; Zbl 1218.55004)] for configuration spaces of surfaces. (An exception is [\textit{S. Scheirer}, Algebr. Geom. Topol. 18, No. 2, 839--876 (2018; Zbl 1393.55004)] which deals with unordered configuration spaces of trees.) The reason for this is that the standard cohomological lower bound in terms of zero-divisor cup-length is usually insufficient for unordered configuration spaces. The authors therefore employ the lower bound given by Lupton, Oprea and the reviewer in [\textit{M. Grant} et al., Topology Appl. 189, 78--91 (2015; Zbl 1317.55003)], which states that for an aspherical space $X$ with fundamental group $\pi$, one has $\mathsf{TC}(X)\geq \mathsf{cd}(A\times B)$ whenever $A$ and $B$ are subgroups of $\pi$ all of whose conjugates intersect trivially. Since all of the spaces $C(S,n)$ considered are aspherical, the authors obtain their lower bounds by identifying suitable subgroups $A$ and $B$ of the surface braid groups $\pi_1(C(S,n))$, and showing that their conjugates intersect trivially by mapping to the homology $H_1(S\times \cdots \times S)$ of the product. The upper bounds come from standard techniques, with the exception of the annulus and disc cases. Here the authors describe explicit and rather ingenious motion planning algorithms for $C(A,n)$ and $C(D,3)$.