# Math 22a Fall 2018

## 22a Linear Algebra and Vector Analysis

# Discrete Calculus

In the last part of the Proof seminar, we have done a bit of discrete calculus. These were Unit 33 and Unit 37. Calculus in the discrete started with Gustav Kirchhoff. The fact that the sum of the divergences values are zero ∑_{x}d

^{*}F(x) = 0 is called the Kirchhoff's current law. It is a conservation law which we also know in the continuum. If we have a n-dimensional closed surface which is also bounded, then the integral of the divergence is zero. The reason is the divergence theorem which tells that this integral is the flux through the boundary. But there is no boundary for a closed manifold. In the discrete, this applies more and less with traffic of goods or money and be seen in the context of input-output analysis. The matrix L= d

^{*}d = div grad is the analogue of the the classical Laplacian L f (x,y) = f

_{xx}+ f

_{yy}. If written as a matrix it is called the Kirchhoff matrix. You can get this matrix in Mathematica with

*Normal[KirchhoffMatrix[WheelGraph[5]]]*Kirchhoff can be seen as

**the**pioneer and early champion of discrete calculus. It must be said that discrete calculus has been reemerged again and again, in particular by computer scientists. It is useful in computer graphics for example, where surfaces are just graphs (discrete meshes of triangles or tetrahedra). The discrete setup is implicitly used by any algebraic topologist since the beginnings where algebra has started to be used for describing geometric objects. In order to compute things, we need finite matrices and the matrices can be obtained by discretizing space. You find in this document of 2012 the discrete Stokes theorem, the discrete Poincaré-Hopf theorem (a generalization of the Island theorem we have seen in this course) as well as Gauss-Bonnet all stated and proven in 2 pages. In classical mathematics, this takes maybe 100 times more, if everything is done in arbitrary dimensions (Gauss-Bonnet-Chern in higher dimensions). Poincare-Hopf and Gauss-Bonnet are always done in differential geometry. Gauss-Bonnet-Chern are rarely proven even in graduate level differential geometry courses. On this talk of 2013 you see all of single variable calculus (with Fundamental theorem of calculus, Taylor series, differential equations up to Stokes theorem) done and proven in 20 x 20 =400 seconds. The format of presenting in 20x20 seconds (20 slides which run for exactly 20 seconds) is called Pecha Kucha. See here for the annotated slides. But here are the slides (was recorded with the Math 22 Tatoo, you might want to see in 4K resolution to see that).