In the present research, the structure and the stability of native defects in diamond will be investigated under different conditions of temperature. These defects are formed by the simulation of a ``cold'' implantation, as usually done by the CIRA scheme. Since the computation time available enables only the simulation of physical phenomena that last few tens of picoseconds, the diffusion of defects in diamond (which is very slow) cannot be investigated for appreciable time. Thus, the fast crystal reconstruction or relaxation is the major topic of the thesis, as well as rapid vacancy-interstitial recombinations that take place in the early stages of the annealing.

In the MSc research of D. Saada [20], an ion-impact was
simulated by giving a sudden momentum to a bulk atom in a random
direction, at a temperature of 0K (section 6.4 in ref. [20]
and ref. [57]). The energy of the atom was eight time (416 eV)
larger than the displacement threshold energy to create a
vacancy-interstitial pair, calculated to be 52 eV. Consequently, a
significant damage volume was created along the irradiation track,
with an average size of 1 nm, including *sp*^{3}-like as well as
*sp*^{2}-like bonds. In that manner, a ``cold'' ion implantation was
simulated, leading to a configuration whose annealing will be studied
in the present research. One can change the density of defects created
by modulating the number of carbon atoms knocked from their lattice
site, their initial kinetic energy and the size of the sample. Thus,
from samples generated with different densities of defects, the
effects of the annealing process in the crystal structure can be
investigated.

One has to keep in mind that ion implantation induces volume expansion
[22], due to the addition of atoms, to the formation of
vacancies and interstitials and to the graphitization of the
irradiated volume, the density of diamond being greater than that of
graphite (3.51 *gr*/*cm*^{3} and 2.26 *gr*/*cm*^{3} respectively). Thus
particular care must be taken of the pressure applied to the sample.
In the present research, the volume of the sample and the number of
atoms are kept fixed. The defects created during the bombardment are
therefore embedded in a diamond matrix that should apply a high
pressure to the damaged region. This situation is similar to that
occurring at the end of the ion track in experiment, where the kinetic
energy of the atoms is down to about few hundreds of eV, and the
damage is mainly due to nuclear stopping. This could also occur at Mev
implantation energies, where the ion injected enters very deeply the
diamond lattice, and creates deeply buried layers. The present
simulation relates therefore to the ``end of track'' damage and its
annealing.
The annealing process is simulated by a molecular dynamic technique,
and the interatomic interactions are described by the potential of
Tersoff [21], which has been shown to properly describe
both diamond and graphite bonding. To study the structure of a
particular sample, the mean displacement of each atom is calculated at
several intervals of time. To optimize the mean calculation and reduce
the error, few samples are generated with the same parameters
(temperature, pressure and defects density), and computation of the
average of the mean displacement of the atoms are done for the same
time intervals, leading to a configuration of mean displacement of the
atoms. The potential energy of this configuration is then minimized to
obtain the final configuration of the sample that is
investigated. This computation procedure is applied to obtain results
independent of the thermal fluctuations.

It is also of great importance to understand the mechanism of dopant compensation that was observed in experiment. To this purpose, electronic energy levels induced by specific defects (namely vacancy and <100> split interstitial) has been calculated in the present research, by using a tight binding model described below. With the same model, the behavior of hydrogen in diamond has been investigated. A detailed comprehension, at an atomistic level, of hydrogen in diamond is of basic importance for the reason explained above, and requires understanding the relative energetics of different possible sites for hydrogen. The electronic energy levels induced by hydrogen are also studied, and the motion of hydrogen is inquired as a function of the temperature.

Few models are available to describe the interatomic interactions and
the dynamical motion of atoms. The most accurate is the *ab
initio* model based on density functional theory [86]. In
this model, the local density approximation (LDA) [87] is
very often used for the exchange-correlation interaction, and can also
be modified with a gradient correction (GGA) [88]. In a
large majority of calculations found in literature, plane waves are
used as a basis for the electronic wave functions, and
pseudopotentials [89] describe the interaction between the
valence electrons and the ionic core. The atoms are considered as
classical, and the Newton's equations are solved to obtain the atomic
positions and velocities as a function of the time.

The second model, which is widely used, is the tight binding model
[90], explained below in more details. Here, only a reduced
basis of four orbitals is used, and an empirical functional describes
the repulsive part of the interatomic interaction. This model is less
accurate than the precedent but much less computationally
expensive. With these two models the energy levels induced by defects
or impurities in diamond can be calculated. In the present research on
hydrogen in diamond, relatively long computation times are needed to
study the motion of hydrogen (see below), precluding, thus, the use of
the *ab initio* model. The tight binding model is accurate enough
and has been therefore chosen here over *ab initio* model for the
calculations of hydrogen in diamond.

Finally the classical potential (see next section) is applied in the
first part of the research concerning the simulation of damage in
diamond. In these calculations, a sufficiently large defect volume
must be created, embedded in a diamond matrix. This requires therefore
very large samples containing 5120 carbon atom to minimize the size
effects and to obtain reliable statistic. The *ab initio* and
tight binding models cannot be implemented to this kind of problems
due to their expensive resources required.