Total curvature of graphs in space (Q2470770)

It is well-known that the total curvature of a closed curve in \(\mathbb R^n\) is at least \(2\pi\) (Fenchel's theorem), moreover if it is at most \(4\pi\) then the curve is unknotted (Fary-Milnor theorem). The paper is aimed to extend these results to finite graphs in \(\mathbb R^n\). The main problem is to understand what does ``total curvature'' mean for graphs. The author discusses three rather elementary notions (variational total curvature, maximal total curvature, cone total curvature), which possess intuitively expected geometric properties (consistency with the classical notion, invariance under homotheties, lower semi-continuity, etc). Unfortunately, it is not clear how these notions are related to isotopical properties of graphs in \(\mathbb R^3\). The author presents another notion, called the net total curvature, whose definition is inspired by Milnor's approach to the curvature of curves. Namely, let \(\Gamma\) be a finite graph in \(\mathbb R^3\), whose edges are \(C^2\)-smooth. For each unit tangent vector \(T_i\) at a vertex \(q\in \Gamma\), let \(\chi_i:S^2\to\{-1,1\}\) be equal to \(-1\) on the hemisphere with center at \(T_i\) and \(+1\) on the opposite hemisphere (values along the equator are arbitrary). The author defines then \[ \text{nc}(q)=\frac{1}{4}\int\limits_{S^2}\left[\sum\limits_i\chi_i(e)\right]^{+}dA_{S^2}(e). \] The net total curvature of \(\Gamma\) is \[ N(\Gamma)=\sum \text{nc}(q_j)+\int_{\Gamma_{reg}}|\vec k| ds, \] where the sum is taken over the vertices of \(\Gamma\), the integral is taken over the separate \(C^2\)-edges of \(\Gamma\), \(\vec k\) and \(s\) are the curvature vector and the arc length of edges, respectively. The net total curvature possesses the geometric properties mentioned above. Moreover, the author demonstrates that if \(\Gamma\) is homeomorphic to the \(\Theta\)-graph, then \(N(\Gamma)\geq 3\); and if \(N(\Gamma)\leq 4\pi\), then \(\Gamma\) is isotopic in \(\mathbb R^3\) to a planar \(\Theta\)-graph. It is expected that similar results hold for another classes of finite graphs.