chain complexes

References

(co)Homology

Exact Sequences

From our discussions on homology, we have that exact sequences are like spaces that are very calm

But they can still have twists (like a Möbius band)

Split Exact Sequences

Correspond to untwisted, calm spaces.

Summary

Consider a polynomial algebra k[x]. We can perform a resolution on the algebra k[x]/x

Set C0 = k[x] and C1 = k[x, ε], where ε is a formal element in degree 1, and defining d(ε) = x.

If we further require the algebra structure to be graded commutative, we have ε2 = 0. So we can interpret this as something like an infinitesimal

Now the elements in degree 1 have the form p(x) ε , where p(x) ∈ C0 is a polynomial with boundary d(p) ε + p d(ε) = 0 + p(x) x.

It follows that if there is a cycle in degree 1, we'd have d(p ε) = p(x) x = 0.

If there are no zero-divisors, this implies that there are no cycles. So H1  ≅ 0.

Similarly, if we take the higher degrees, we're trivial, because they would need to be built out of elements in degree 1 (to increase the degree). But since ε2 = 0, this produces no elements above degree 1. Thus Hn  ≥ 1  ≅ 0, and so we get a (very easy) resolution of k[x]: 0 → k[x, ε] → k[x]  → 0 which is quasi-isomorphic to k[x] (as a trivial chain complex: 0 → k[x]  → 0).

Functional Analysis

Fredholm Operators

By definition, these are operators with

The Fredholm index is dim(ker) - dim(coker) It counts solutions to equations

Atiyah-Singer Index Theorem

If T is an elliptic differential operator on a manifold, the analytical index is equal to the topological index

Harmonic Analysis

A locally compact group is amenable if and only if its first Hochschild cohomology group of its convolution algebra with coefficients in dual modules is zero: H1(L1(G), X*)  ≅ 0

Complex Analysis

Hartog's Extension Theorem

Holomorphic functions are those that are locally complex differentiable

This is enough to conclude that the function is analytic

Sheaf Cohomology

Analytic Continuation

This is actually related to relative cohomology

Mikio Sato viewed real functions (which you'd like to analytically continue) as being boundaries between two holomorphic functions

You build relative cohomology out of this idea: take … → O(C) → O(C ∖ R) → H1R(C, O)  → 0 where O is the sheaf of holomorphic functions. This is a long exact sequence.

An element of O(C ∖ R) is really a pair of holomorphic functions: one defined on the upper half plane, one defined on the lower.

The codifferential into H1 is a computation of the difference between a pair of holomorphic functions, as you approach the real axis.

Functions that admit analytic continuations are precisely coboundaries in the first degree

Wick Rotation

When you take a path integral, say over time, you're integrating ∫-∞∞ ω where ω is some complex differential form.

When we do wick rotation, we want to switch into imaginary time. How does this work?

Consider the upper right and lower left quadrants of the complex plane.

Well say ω is closed (a cocycle). Then we can think of the real and imaginary axes as two half-coverings of a larger bulk made up of the upper right and lower left quadrants

To be a bit more precise, typically what you do is you have to consider finite parts of the quadrants and then take the limits to infinity

I guess another way of thinking about it, is you compactify C (say, into the Riemann sphere) and then play the game

If the pole at ∞ is regular or logarithmic, everything is fine

Cup Product