Integration of vector fields on cell complexes and Morse theory (Q2685245)

The aim of this paper is to define a nice theory of trajectories of vector fields on polyhedral complexes, so that the set of them captures the topology of the polyhedral complex, as in classical Morse theory. The author's construction relies on Forman's discrete Morse theory. An \textit{affine complex} \(X\) is a finite CW complex such that each cell has the structure of a convex polytope compatible with the gluing. Let \(\alpha^j\) denote an \(j\)-dimensional cell, and write \( \alpha < \beta\) if the cell \(\alpha\) is a face of \(\beta\). Recall that a discrete Morse function \(F :\{\text{cells of } X\} \rightarrow \mathbb{R}\) on a complex \(X\) is a function such that for every cell \(\alpha^i\), the inequalities \[ |\{\beta^{i+1}>\alpha^i : F(\beta^{i+1}) \leq F(\alpha^i)|\leq 1 \] and \[ |\{\beta^{i-1}<\alpha^i : F(\beta^{i-1}) \geq F(\alpha^i)|\leq 1 \] hold. \textbf{Definition:} A discrete Morse function \(F\) is called \textit{generic} if for every any pair of cells \(\alpha, \beta\) satisfying \(\alpha < \beta\) we have \(F(\alpha) \neq F(\beta)\). The function \(F\) is called \textit{tame} if \(\alpha ^i < \beta^j\) with \( j \geq i+2\) implies \(F(\beta^j )> F(\alpha^i)\). \textbf{Definition:} A piecewise affine metric on a simplicial complex is called \textit{sharp} if on each simplex \(\sigma\), the following condition is satisfied: Take any vertex \(a\) of \(\sigma\) and a \(2\)-dimensional plane \(L\) containing \(a\). Assume \(L \cap \sigma \) is a triangle denoted by \(\Delta abc\). Then it is acute angled, this is, the interior angle at each vertex is less than \(\frac{\pi}{2}\) with respect to the metric. Let \(X\) be an affine complex and \(F\) be a discrete Morse function on \(X\). Define a piecewise linear function \(f\) from \(F\) as follows: \begin{itemize} \item[1.] Take the barycentric subdivision \(X_1\) of \(X\). By definition, the vertices of \(X_1\) correspond to the cells of \(X\). \item[2.] To a vertex of \(X_1\), give the value of \(F\) at the corresponding cell. \item[3.] Extend it affine linearly to each maximal cell of \(X_1\). It is possible because \(X_1\) is a simplicial complex. \end{itemize} A vertex of \(X_1\) is called \(i\)-dimensional if it correspond to an \(i\)-dimensional cell of \(X\), and similarly we call an edge of \(X_1\) an \((i,j)\)-edge if its endpoints are \(i\) and \(j\)-dimensional vertices. Define \(\mathcal{M}_i^PL\) to be the free abelian group generated by critical \(i\)-dimensional vertices. It is canonically isomorphic to \(\mathcal{M}_i\). The author constructs a complex \[ 0 \rightarrow \mathcal{M}_n^{PL} \rightarrow \dots, \rightarrow \mathcal{M}_0^{PL} \] by counting gradient trajectories of the piecewise linear flow. More specifically, the linear map \(d_{PL}:\mathcal{M}_i^{PL} \rightarrow \mathcal{M}_{i-1}^{PL}\) is defined by linearly extending the map \[ p \mapsto \sum_{q \in \mathcal{M}_{i-1}^{PL}} \sum_{\gamma \in \Gamma^{PL}(p,q)} m(\gamma)q \] where \(m(\gamma)\) denotes the multiplicity of \(\gamma\) and \(\Gamma^{PL}(p,q)\) denotes the the set of gradient trajectories connecting \(p\) and \(q\). The main theorem of the article is the following: \textbf{Theorem:} The sequence \((\mathcal{M}^{PL}_{\bullet}, d_{PL})\) is a complex, and its homology is isomorphic to the singular homology of \(X\). Furthermore the author shows that if the affine complex \(X\) is a polyhedral decomposition of a manifold without boundary, then there exists the dual of the piecewise linear theory on the dual complex \(\overline{X}\) of \(X\) and defines the stable and unstable complexes as follows: \textbf{Proposition:} Assume \(X\) is a manifold without boundary, and let \(p\) be an \(i\)-dimensional critical vertex. Then the subset of \(X_1\) swept by the trajectories starting from \(p\) is a subcomplex of \(X^{(i)}\) called the \textit{unstable complex} of \(p\). In the dual theory, \(p\) is a critical \(n-i\)-dimensional point, and its unstable complex is the \textit{stable complex} of \(p\) of the original theory.