
A special type of GagliardoNirenbergSobolev (GNS) inequalities in
$\mathbb{R}^d$ has played a key role in several proofs of LiebThirring
inequalities. Recently, a need for GNS inequalities in convex domains of
$\mathbb{R}^d$, in particular for cubes, has arised. The purpose of this
manuscript is twofold. First we prove a GNS inequality for convex domains,
with explicit constants which depend on the geometry of the domain. Later,
using the discrete version of Rumin's method, we prove GNS inequalities on
cubes with improved constants.

We consider Dirac operators defined on planar domains. For a large class of
boundary conditions, we give a direct proof of their selfadjointness in the
Sobolev space $H^1$.

The twodimensional Dirac operator describes lowenergy excitations in
graphene. Different choices for the boundary conditions give rise to
qualitative differences in the spectrum of the resulting operator. For a family
of boundary conditions, we find a lower bound to the spectral gap around zero,
proportional to $\Omega^{1/2}$, where $\Omega \subset \mathbb{R}^2$ is the
bounded region where the Dirac operator acts. This family contains the
socalled infinite mass and armchair cases used in the physics literature for
the description of graphene quantum dots.

We prove that in ThomasFermiDiracvon Weizs\"acker theory, a nucleus of
charge $Z>0$ can bind at most $Z+C$ electrons, where $C$ is a universal
constant. This result is obtained through a comparison with ThomasFermi theory
which, as a byproduct, gives bounds on the screened nuclear potential and the
radius of the minimizer. A key ingredient of the proof is a novel technique to
control the particles in the exterior region, which also applies to the liquid
drop model with a nuclear background potential.

We study the ionization problem in the ThomasFermiDiracvon Weizs\"acker
theory for atoms and molecules. We prove the nonexistence of minimizers for the
energy functional when the number of electrons is large and the total nuclear
charge is small. This nonexistence result also applies to external potentials
decaying faster than the Coulomb potential. In the case of arbitrary nuclear
charges, we obtain the nonexistence of stable minimizers and radial minimizers.

We consider an atom described by M\"uller theory, which is similar to
HartreeFock theory, but with a modified exchange term. We prove that a nucleus
of charge Z can bind at most Z+C electrons, where C is a universal constant.
Our proof proceeds by comparison with ThomasFermi theory and a key ingredient
is a novel bound on the number of electrons far from the nucleus.

We consider the Pauli operator in $\mathbb R^3$ for magnetic fields in
$L^{3/2}$ that decay at infinity as $x^{2\beta}$ with $\beta > 0$. In this
case we are able to prove that the existence of a zero mode for this operator
is equivalent to a quantity $\delta(\mathbf B)$, defined below, being equal to
zero. Complementing a result from [Balinsky, Evans, Lewis (2001)], this implies
that for the class of magnetic fields considered, Sobolev, Hardy and CLR
inequalities hold whenever the magnetic field has no zero mode.

Probabilistic Cellular Automata (PCA) are simple models used to study
dynamical phase transitions. There exist mean field approximations to PCA that
can be shown to exhibit a phase transition. We introduce a model interpolating
between a class of PCA, called majority voters, and their corresponding mean
field models. Using graphical methods, we prove that this model undergoes a
phase transition.