Birth and death in discrete Morse theory (Q739613)

Discrete Morse theory is a beautiful theory analogous to smooth Morse theory popularized by Milnor. There are many uses and applications of discrete Morse theory. One natural question to ask is: suppose a phenomenon happens in the smooth setting. What is the analogue in the discrete setting and how is it similar to and different from the smooth? The authors of the paper under review undertake this question for the following smooth phenomena: Let \(N\) be a smooth manifold and \(f_t : N \to \mathbb{R}\) a family of generically Morse functions with \(F : N \times I\to \mathbb{R}\) the adjoint. As we let \(t\) vary, we obtain a corresponding varying of critical pints of \(f_t\) in \(N\); that is, critical points can ``die'' (e.g. \(a\) is critical for \(t<t_0\) but not for \(t\geq t_0\)) and be ``born''. The discrete analogue studied in this paper is as follows. Let \(M\) be a cell complex, \(0=t_0< \dots,< t_r=1\) a finite sequence of values, and consider corresponding discrete Morse functions \(F_{t_i}: M \to \mathbb{R}\). What would it mean for the critical cells to vary, be born, and die? How could one determine death and birth? With this as the setting, section 2 introduces the basics of discrete Morse theory of CW complexes. Especially emphasized is the gradient vector field viewpoint, as this point of view is essential to the rest of the paper. Section 3 is devoted to defining and explaining two algorithms needed (to be slightly refined in a subsequent section). Let \(\alpha, \beta\) be \(k\)-cells of \(M\) with \(\alpha\) critical for \(V_i\) (discrete vector field of \(f_{t_i}\)) and \(\beta\) critical for \(V_j\). The author define \(\alpha\) to be \textit{connected} to \(\beta\) if there exists a \(k\)-cell \(\gamma\) and a \(V_i\)-path \(\alpha, \ldots, \gamma\) of \(k\) and \((k-1)\)-cells and a \(V_j\)-path \(\gamma, \dots, \beta\) of \(k\) and \((k-1)\)-cells. Furthermore, \(\alpha\) and \(\beta\) are \textit{strongly connected} if \(\alpha\) is connected to \(\beta\) and \(\beta\) is connected to \(\alpha\). These definitions allow a notion of critical cells varying as \(i\) varies, and furthermore, to define ``birth'' and ``death'' of critical cells. The algorithms in this section return pairs of connected critical cells, as well as a birth/death pairs in the form of diagrams illustrating these connections among critical cells. Section 4 considers how the algorithm and theory can make sense of different cell decompositions of \(M\). This section is a technical and necessary precursor to section 5 which gives algorithms connecting critical cells between different decompositions. Section 6 discusses applications to CT head scans. Applying the theory in this case allows one to follow cavities through the brain via a series of slices of the brain. The final section gives several open questions as well as future directions.