Holomorphic functions on a domain are the solutions to a set of homogeneous partial differential
equations called the Cauchy-Riemann equation, and CR functions are the solutions to the analogous
equations on the boundary. Many problems in complex analysis can be reduced to finding appropriate
solutions to the inhomogeneous versions of these equations. These solutions have been successfully
constructed when the geometry of the domain is sufficiently simple. I hope to show how these constructions
follow a pattern based on the singular integral operators of Calder\'on and Zygmund. I then plan to
discuss some more recent examples where this well-understood paradigm breaks down.

I will consider the random walk on Z^d driven by a field of random i.i.d.
conductances.
The law of the conductances is bounded from above; no restriction is posed on the
lower tail (at zero) except that the bonds with positive conductances percolate.
The presence of very weak bonds allows the random walk in a finite box mix pretty
much arbitrarily slowly. However, when we focus attention on the return probability
to the starting point -- i.e., the heat-kernel -- it turns out that in dimensions
d=2,3 the
decay is as for the simple random walk. On the other hand, in d>4 the heat- kernel
at time 2n may decay as slowly as o(1/n^2) and in d=4 as slowly as O(n^{-2}log n).
These upper bounds can be matched arbitrarily closely by lower bounds in particular
examples. Despite this, the random walk scales to Brownian motion under the usual
diffusive scaling of space and time. Based on joint works with N. Berger, C. Hoffman,
G. Kozma and T. Prescott.

B. Howard, B. Mazur and K. Rubin proved that the existence of Kolyvagin systems relies on a cohomological invariant, what they call the core Selmer rank. When the core Selmer rank is one, they determine the structure of the Selmer group completely in terms of a Kolyvagin system. However, when the Selmer core rank is greater than one such a precision could not be achieved. In fact, one does not expect a similiar result for the structure of the Selmer group in general, as a reflection of the fact that Bloch-Kato conjectures do not in general predict the existence of special elements, but a regulator, to compute the relevant L-values.

An example of a core rank greater than one situation arises if one attempts to utilize the Euler system that would come from the Stark elements (whose existence were predicted by K. Rubin) over a totally real number field. This is what I will discuss in this talk. I will explain how to construct, using Stark elements, Kolyvagin systems for certain modified Selmer structures (that are adjusted to have core rank one) and relate them to appropriate ideal class groups, following the machinery of Kolyvagin systems and prove a Gras-type conjecture.