Cette thèse, ayant pour thème les algorithmes de la chimie quantique, s'inscrit dans le cade du changement de paradigme observé depuis une douzaines d'années, dans lequel les méthodes de calcul séquentielles se doivent d'être progressivement remplacées par des méthodes parallèles. En effet, l'augmentation de la fréquences des processeurs se heurtant à des barrières physiques difficilement franchissables, l'augmentation de la puissance de calcul se fait par l'augmentation du nombre d'unités de calcul. Toutefois, là où une augmentation de la fréquence conduisait mécaniquement à une exécution plus rapide d'un code, l'augmentation du nombre de cœurs peut se heurter à des barrières algorithmiques, qui peuvent nécessiter une adaptation ou un changement d'algorithme. Parmi les méthodes développées afin de contourner ce problème, on trouve en particulier celles de type Monte-Carlo (stochastiques), qui sont intrinsèquement "embarrassingly parallel", c'est à dire qu'elles sont par construction constituées d'une multitudes de tâches indépendantes, et de ce fait particulièrement adaptées aux architectures massivement parallèles. Elles ont également l'avantage, dans de nombreux cas, d'être capables de produire un résultat approché pour une fraction du coût calculatoire de l'équivalent déterministe exacte. Lors de cette thèse, des implémentations massivement parallèles de certains algorithmes déterministes de chimie quantique ont été réalisées. Il s'agit des algorithmes suivants : CIPSI, diagonalisation de Davidson, calcul de la perturbation au second ordre, shifted-Bk, et Coupled Cluster Multi Références. Pour certains, une composante stochastique a été introduite en vue d'améliorer leur efficacité. Toutes ces méthodes ont été implémentées sur un modèle de tâches distribuées en TCP, où un processus central distribue des tâches par le réseau et collecte les résultats. En d'autres termes, des nœuds esclaves peuvent être ajoutés au cours du calcul depuis n'importe quelle machine accessible depuis internet. L'efficacité parallèle des algorithmes implémentés dans cette thèse a été étudiée, et le programme a pu donner lieu à de nombreuses applications, notamment pour permettre d'obtenir des énergies de références pour des systèmes moléculaires difficiles. / This thesis, whose topic is quantum chemistry algorithms, is made in the context of the change in paradigm that has been going on for the last decade, in which the usual sequential algorithms are progressively replaced by parallel equivalents. Indeed, the increase in processors' frequency is challenged by physical barriers, so increase in computational power is achieved through increasing the number of cores. However, where an increase of frequency mechanically leads to a faster execution of a code, an increase in number of cores may be challenged by algorithmic barriers, which may require adapting of even changing the algorithm. Among methods developed to circumvent this issue, we find in particular Monte-Carlo methods (stochastic methods), which are intrinsically "embarrassingly parallel", meaning they are by design composed of a large number of independent tasks, and thus, particularly well-adapted to massively parallel architectures. In addition, they often are able to yield an approximate result for just a fraction of the cost of the equivalent deterministic, exact computation. During this thesis, massively parallel implementations of some deterministic quantum chemistry algorithms were realized. Those methods are: CIPSI, Davidson diagonalization, computation of second-order perturbation, shifted-Bk, Multi-Reference Coupled-Cluster. For some of these, a stochastic aspect was introduced in order to improve their efficiency. All of them were implemented on a distributed task model, with a central process distributing tasks and collecting results. In other words, slave nodes can be added during the computation from any location reachable through Internet. The efficiency for the implemented algorithms has been studied, and the code could give way to numerous applications, in particular to obtain reference energies for difficult molecular systems.

This thesis is concerned with a range of chemical problems which are amenable to theoretical investigation via the application of current methods of computational quantum chemistry. These problems include the calculation of accurate thermochemical data, the prediction of reaction kinetics, the study of molecular potential energy surfaces, and the investigation of molecular structures and binding. The heats of formation (from both atomisation energies and isodesmic schemes) of a set of approximately 120 C1 and C2 fluorocarbons and oxidised fluorocarbons (along with C3F6 and CF3CHFCF2) were calculated with the Gaussian-3 (G3) method (along with several approximations thereto). These molecules are of importance in the flame chemistry of 2-H-heptafluoropropane, which has been proposed as a potential fire retardant with which to replace chloro- and bromofluorocarbons (CFC�s and BFC�s). The calculation of the data reported here was carried out in parallel with the modelling studies of Hynes et al.1-3 of shock tube experiments on CF3CHFCF3 and on C3F6 with either hydrogen or oxygen atoms. G3 calculations were also employed in conjunction with the experimental work of Owens et al.4 to describe the pyrolysis of CFClBr2 (giving CFCl) at a radiation wavelength of 265 nm. The theoretical prediction of the dissociation energy of the two C-Br bonds was found to be consistent with the energy at which carbene production was first observed, thus supporting the hypothesis that the pyrolysis releases two bromine radicals (rather than a Br2 molecule). On the basis of this interpretation of the experiments, the heat of formation of CFClBr2 is predicted to be 184 � 5 kJ mol-1, in good agreement with the G3 value of 188 � 5 kJ mol-1. Accurate thermochemical data was computed for 18 small phosphorus containing molecules (P2, P4, PH, PH2, PH3, P2H2, P2H4, PO, PO2, PO3, P2O, P2O2, HPO, HPOH, H2POH, H3PO, HOPO and HOPO2), most of which are important in the reaction model introduced by Twarowski5 for the combustion of H2 and O2 in the presence of phosphine. Twarowski reported that the H + OH recombination reaction is catalysed by the combustion products of PH3 and proposed two catalytic cycles, involving PO2, HOPO and HOPO2, to explain this observation. Using our thermochemical data we computed the rate coefficients of the most important reactions in these cycles (using transition state and RRKM theories) and confirmed that at 2000K both cycles have comparable rates and are significantly faster than the uncatalysed H + OH recombination. The heats of formation used in this work on phosphorus compounds were calculated using the G2, G3, G3X and G3X2 methods along with the far more extensive CCSD(T)/CBS type scheme. The latter is based on the evaluation of coupled cluster energies using the correlation consistent triple-, quadruple- and pentuple-zeta basis sets and extrapolation to the complete basis set (CBS) limit along with core-valence correlation corrections (with counterpoise corrections for phosphorus atoms), scalar relativistic corrections and spin-orbit coupling effects. The CCSD(T)/CBS results are consistent with the available experimental data and therefore constitute a convenient set of benchmark values with which to compare the more approximate Gaussian-n results. The G2 and G3 methods were found to be of comparable accuracy, however both schemes consistently underestimated the benchmark atomisation energies. The performance of G3X is significantly better, having a mean absolute deviation (MAD) from the CBS results of 1.8 kcal mol-1, although the predicted atomisation energies are still consistently too low. G3X2 (including counterpoise corrections to the core-valence correlation energy for phosphorus) was found to give a slight improvement over G3X, resulting in a MAD of 1.5 kcal mol-1. Several molecules were also identified for which the approximations underlying the Gaussian-n methodologies appear to be unreliable; these include molecules with multiple or strained P-P bonds. The potential energy surface of the NNH + O system was investigated using density functional theory (B3LYP/6-31G(2df,p)) with the aim of determining the importance of this route in the production of NO in combustion reactions. In addition to the standard reaction channels, namely decomposition into NO + NH, N2 + OH and H + N2O via the ONNH intermediate, several new reaction pathways were also investigated. These include the direct abstraction of H by O and three product channels via the intermediate ONHN, giving N2 + OH, H + N2O and HNO + N. For each of the species corresponding to stationary points on the B3LYP surface, valence correlated CCSD(T) calculations were performed with the aug-cc-pVxZ (x = Q, 5) basis sets and the results extrapolated to the complete basis set limit. Core-valence correlation corrections, scalar relativistic corrections and spin orbit effects were also included in the resulting energetics and the subsequent calculation of thermochemical data. Heats of formation were also calculated using the G3X method. Variational transition state theory was used to determine the critical points for the barrierless reactions and the resulting B3LYP energetics were scaled to be compatible with the G3X and CCSD(T)/CBS values. As the results of modelling studies are critically dependent on the heat of formation of NNH, more extensive CCSD(T)/CBS calculations were performed for this molecule, predicting the heat of formation to be 60.6 � 0.5 kcal mol-1. Rate coefficients for the overall reaction processes were obtained by the application of multi-well RRKM theory. The thermochemical and kinetic results thus obtained were subsequently used in conjunction with the GRIMech 3.0 reaction data set in modelling studies of combustion systems, including methane / air and CO / H2 / air mixtures in completely stirred reactors. This study revealed that, contrary to common belief, the NNH + O channel is a relatively minor route for the production of NO. The structure of the inhibitor Nd-(N'-Sulfodiaminophosphinyl)-L-ornithine, PSOrn, and the nature of its binding to the OTCase enzyme was investigated using density functional (B3LYP) theory. The B3LYP/6-31G(d) calculations on the model compound, PSO, revealed that, while this molecule could be expected to exist in an amino form in the gas phase, on complexation in the active site of the enzyme it would be expected to lose two protons to form a dinegative imino tautomer. This species is shown to bind strongly to two H3CNHC(NH2)2+ moieties (model compounds for arginine residues), indicating that the strong binding observed between inhibitor and enzyme is partially due to electrostatic interactions as well as extensive hydrogen bonding (both model Arg+ residues form hydrogen bonds to two different sites on PSO). Population analysis and hydrogen bonding studies have revealed that the intramolecular bonding in this species consists of either single or semipolar bonds (that is, S and P are not hypervalent) and that terminal oxygens (which, being involved in semipolar bonds, carry negative charges) can be expected to form up to 4 hydrogen bonds with residues in the active site. In the course of this work several new G3 type methods were proposed, including G3MP4(SDQ) and G3[MP2(Full)], which are less expensive approximations to G3, and G3X2, which is an extension of G3X designed to incorporate additional electron correlation. As noted earlier, G3X2 shows a small improvement on G3X; G3MP4(SDQ) and G3[MP2(Full)], in turn, show good agreement with G3 results, with MAD�s of ~ 0.4 and ~ 0.5 kcal mol-1 respectively. 1. R. G. Hynes, J. C. Mackie and A. R. Masri, J. Phys. Chem. A, 1999, 103, 5967. 2. R. G. Hynes, J. C. Mackie and A. R. Masri, J. Phys. Chem. A, 1999, 103, 54. 3. R. G. Hynes, J. C. Mackie and A. R. Masri, Proc. Combust. Inst., 2000, 28, 1557. 4. N. L. Owens, Honours Thesis, School of Chemistry, University of Sydney, 2001. 5. A. Twarowski, Combustion and Flame, 1995, 102, 41.

<p>This thesis presents briefly the application of quantum mechanics on systems ofchemical interest, i.e., the field of quantum chemistry and computational chemistry.The molecules of the ten-electron series, hydrogen fluoride, water, ammonia,methane and neon, are taken as computational examples. Some applications ofquantum chemistry are then shown on these systems, with emphasis on the natureof the molecular bonds. Conceptual methods of chemistry and theoreticalchemistry for these systems are shown to be valid with some restrictions, as theseinterpretations does not represent physically measurable entities.The orbitals and orbital energies of neon is studied, the binding van der Waalsinteractionresulting in a Ne2 molecule is studied with a theoretical bond lengthof 3.23 °A and dissociation energy of 81.75 μEh. The equilibrium geometries ofFH, H2O, NH3 and CH4 are studied and the strength and character of the bondsinvolved evaluated using bond order, dipole moment, Mulliken population analysisand L¨owdin population analysis. The concept of electronegativity is studied in thecontext of electron transfer. Lastly, the barrier of inversion for NH3 is studied, withan obtained barrier height of 8.46 mEh and relatively constant electron transfer.</p>

<p>In dye-sensitized solar cells, the functions of light absorption and charge transport are separated. A photosensitive ruthenium-polypyridine dye in the cell absorbs light, injects an electron to a semiconductor and is then regenerated by a redox couple, typically iodide/triiodide. Quantum chemical calculations of the electronic structure of triiodide have been carried out with the restricted active space SCF method, including spin-orbit coupling, and with density functional theory. It was shown that the difference in charge density between the terminal and central atoms results in a splitting of the core levels. The calculations gave a value of the splitting of 0.8 - 1.0 eV for the <em>3d</em> and <em>4d</em> levels. Experimentally, the electronic structure has been investigated with photoelectronspectroscopy. The measured terminal/center splitting is 1.1 eV.The spin-orbit interaction of the <em>4d </em>levels of triiodide has also been calculated. The splitting was determined to be 1.6 eV. The experimental value is 1.7 eV. An assignment of the peaks in the computed spectrum of triiodide was made and the features of the experimental spectrum have beenidentied.The theoretical valence spectrum of triiodide has been computed and assigned. The results can be used in the analysis of photoelectron spectra of the molecule. Information about the electronic structure of the redox couple can help in the understanding of the electron transfer processes and forfurther development of the solar cells.  Furthermore, the solvation structure of the prototype dye, the tris(bipyridine)ruthenium(II) complex, in water and its interaction with iodide and chloride has been studied by means of molecular dynamics simulations. The trajectory analysis showed that the water molecules in the first solvation shell form a chain in between the bipyridine ligands. It was found that the iodide ions are more likely than chloride to enter between the ligands, which can be important for the electron transfer processin the solar cell.</p>

<p>The basic building blocks in molecular electronics often correspond to conjugated molecules. A compound class consisting of rigid rod-like staffane molecules with the heavier Group 14 elements Si, Ge, Sn and Pb at their bridgehead positions has now been investigated. Herein these oligomers are called heavy-core or Si-, Ge-, Sn- or Pb-core staffanes. These compounds benefit from interaction through their bicyclo[1.1.1]pentane monomer units. Quantum chemical calculations were performed to probe their geometries, stabilities and electronic properties associated with conjugation.</p><p>The stabilities of the bicyclo[<i>n.n.n</i>]alkane and [<i>n.n.n</i>]propellanes (1 ≤ <i>n</i> ≤ 3) with C, Si, Ge and Sn at the bridgehead positions were studied by calculation of homodesmotic ring strain energies. The bicyclic compounds with <i>n</i> = 1 and Si, Ge or Sn at bridgehead positions have lower strain than the all-carbon compound.</p><p>A gradually higher polarizability exaltation is found as the bridgehead element is changed from C to Si, Ge, Sn or Pb. The ratio between longitudinal and average polarizability also increases gradually as Group 14 is descended, consistent with enhanced conjugation in the heavier oligomers.</p><p>The localization of polarons in C-, Si- and Sn-core staffane radical cations was calculated along with internal reorganization energies. The polaron is less localized in Si- and Sn-core than in C-core staffane radical cation. The reorganization energies are also lower for the heavier staffanes, facilitating hole mobility when compared to the C-core staffanes.</p><p>The effect of the bicyclic structure on the low valence excitations in the UV-spectra of compounds with two connected disilyl segments was also investigated. MS-CASPT2 calculations of 1,4-disilyl- and 1,4-bis(trimethylsilyl)-1,4-disilabicyclo[2.2.1]heptanes and 1,4-disilyl- and 1,4-bis(trimethylsilyl)-1,4-disilabicyclo[2.1.1]hexanes revealed that although the bicyclic cage separates the two disilyl chromophores, there is a strong red-shift of the lowest valence excitations when compared to an isolated disilane.</p>

In dye-sensitized solar cells, the functions of light absorption and charge transport are separated. A photosensitive ruthenium-polypyridine dye in the cell absorbs light, injects an electron to a semiconductor and is then regenerated by a redox couple, typically iodide/triiodide. Quantum chemical calculations of the electronic structure of triiodide have been carried out with the restricted active space SCF method, including spin-orbit coupling, and with density functional theory. It was shown that the difference in charge density between the terminal and central atoms results in a splitting of the core levels. The calculations gave a value of the splitting of 0.8 - 1.0 eV for the 3d and 4d levels. Experimentally, the electronic structure has been investigated with photoelectronspectroscopy. The measured terminal/center splitting is 1.1 eV.The spin-orbit interaction of the 4d levels of triiodide has also been calculated. The splitting was determined to be 1.6 eV. The experimental value is 1.7 eV. An assignment of the peaks in the computed spectrum of triiodide was made and the features of the experimental spectrum have beenidentied.The theoretical valence spectrum of triiodide has been computed and assigned. The results can be used in the analysis of photoelectron spectra of the molecule. Information about the electronic structure of the redox couple can help in the understanding of the electron transfer processes and forfurther development of the solar cells.  Furthermore, the solvation structure of the prototype dye, the tris(bipyridine)ruthenium(II) complex, in water and its interaction with iodide and chloride has been studied by means of molecular dynamics simulations. The trajectory analysis showed that the water molecules in the first solvation shell form a chain in between the bipyridine ligands. It was found that the iodide ions are more likely than chloride to enter between the ligands, which can be important for the electron transfer processin the solar cell.

Physisorption of CO and N2O on surfaces of ceria (CeO2) was investigated by means of high-level quantum-mechanical embedded cluster calculations. Both systems have high relevance in the field of environmental chemistry and heterogeneous catalysis. The CO/CeO2 system, has been investigated in a couple of both experimental and theoretical studies, but for the N2O/CeO2 system, this is the first study in the literature, experimental or theoretical. In physisorption, the interaction relies entirely on classical electrostatic interactions and electron dispersion forces. No covalent bond is formed between the molecule and the surface. A proper description of the dispersion requires some of the most accurate quantum-mechanical methods available, such as MP2 or CCSD(T). Moreover, even the most sophisticated methods cannot heal errors anywhere else in the theoretical treatment. Standard periodic models cannot be used with methods such as CCSD(T), but embedded cluster models can, and have been thoroughly explored in this thesis. In this thesis, embedded cluster models were constructed for the CeO2(110) and (111) surfaces. Using a range of assessment tests, it was verified that the electronic structure of the central region of a large and fully embedded surface cluster agrees well with the corresponding region in a periodic system. CO physisorption was investigated at the CCSD(T) level. Due to the prohibitively large expenses (in computer time) for standard CCSD(T) calculations, the method of increments, previously used in the literature for bulk systems, was extended to adsorption problems. It was found that, electron correlation contributes by 30 - 80% to the molecule-surface interaction and that the contribution depends on the topology of the surface. The calculated CO-ceria interaction energy is 20 kJ/mol for the (111) surface and 27 kJ/mol for the (110) surface. In low temperature TPD experiments for the N2O/CeO2(111) system, one surface species was found with an adsorption energy of about 29 kJ/mol. IR measurements showed stretching frequencies that are typically assigned to N2O adsorption with the O-end directed towards surface cations. However, theoretical calculations up to the MP2 level predicted two equally favorable adsorption species. Improvements in the structural model (larger clusters, consideration of molecule-induced relaxation) or the computational method (larger basis sets) did not affect this result. Only at the CCSD(T) level was one dominating surface species found, namely N2O adsorbed over a Ce ion, with the O-end of the molecule directed towards the surface. The calculated stretching vibrational frequency shifts (with respect to the gas phase) for this adsorbed species agree well with the measured IR spectra.

The basic building blocks in molecular electronics often correspond to conjugated molecules. A compound class consisting of rigid rod-like staffane molecules with the heavier Group 14 elements Si, Ge, Sn and Pb at their bridgehead positions has now been investigated. Herein these oligomers are called heavy-core or Si-, Ge-, Sn- or Pb-core staffanes. These compounds benefit from interaction through their bicyclo[1.1.1]pentane monomer units. Quantum chemical calculations were performed to probe their geometries, stabilities and electronic properties associated with conjugation. The stabilities of the bicyclo[n.n.n]alkane and [n.n.n]propellanes (1 ≤ n ≤ 3) with C, Si, Ge and Sn at the bridgehead positions were studied by calculation of homodesmotic ring strain energies. The bicyclic compounds with n = 1 and Si, Ge or Sn at bridgehead positions have lower strain than the all-carbon compound. A gradually higher polarizability exaltation is found as the bridgehead element is changed from C to Si, Ge, Sn or Pb. The ratio between longitudinal and average polarizability also increases gradually as Group 14 is descended, consistent with enhanced conjugation in the heavier oligomers. The localization of polarons in C-, Si- and Sn-core staffane radical cations was calculated along with internal reorganization energies. The polaron is less localized in Si- and Sn-core than in C-core staffane radical cation. The reorganization energies are also lower for the heavier staffanes, facilitating hole mobility when compared to the C-core staffanes. The effect of the bicyclic structure on the low valence excitations in the UV-spectra of compounds with two connected disilyl segments was also investigated. MS-CASPT2 calculations of 1,4-disilyl- and 1,4-bis(trimethylsilyl)-1,4-disilabicyclo[2.2.1]heptanes and 1,4-disilyl- and 1,4-bis(trimethylsilyl)-1,4-disilabicyclo[2.1.1]hexanes revealed that although the bicyclic cage separates the two disilyl chromophores, there is a strong red-shift of the lowest valence excitations when compared to an isolated disilane.

This thesis presents briefly the application of quantum mechanics on systems ofchemical interest, i.e., the field of quantum chemistry and computational chemistry.The molecules of the ten-electron series, hydrogen fluoride, water, ammonia,methane and neon, are taken as computational examples. Some applications ofquantum chemistry are then shown on these systems, with emphasis on the natureof the molecular bonds. Conceptual methods of chemistry and theoreticalchemistry for these systems are shown to be valid with some restrictions, as theseinterpretations does not represent physically measurable entities.The orbitals and orbital energies of neon is studied, the binding van der Waalsinteractionresulting in a Ne2 molecule is studied with a theoretical bond lengthof 3.23 °A and dissociation energy of 81.75 μEh. The equilibrium geometries ofFH, H2O, NH3 and CH4 are studied and the strength and character of the bondsinvolved evaluated using bond order, dipole moment, Mulliken population analysisand L¨owdin population analysis. The concept of electronegativity is studied in thecontext of electron transfer. Lastly, the barrier of inversion for NH3 is studied, withan obtained barrier height of 8.46 mEh and relatively constant electron transfer.

Through the development of physical chemistry and chemical physics over the late-nineteenth and early-twentieth centuries, the relationship between physics and chemistry changed to create a broad interdisciplinary framework in which chemists and physicists could make contributions to problems of common value. It is here argued that evolving disciplinary factors such as physical and chemical responses to the atomic hypothesis, the nature of disciplinary formation in Germany and the United States, the reception of quantum mechanics within physics and chemistry, and the application of quantum mechanics to the problem of chemical bonding by physicists and chemists, formed the chemical bond into a physico-chemical theory. In the late nineteenth-century context of early physical chemistry, the chemical bond was known as a physical link between atoms, which could not be studied by chemical means because of the lack of an adequate atomistic framework. Both chemists and physicists broadly accepted the atomistic hypothesis following the discovery of the electron at the turn of the twentieth century, which afforded theoretical study of chemical bonding. Between 1916 and 1919, Gilbert N. Lewis and Irving Langmuir proposed the valence bond to be a pair of electrons shared between two atoms, within the context of a cubic model of the atom. However, the lack of a physical mechanism for the shared electron pair prevented the formation of a fully physico-chemical view of bonding. In 1927, physicists Walter Heitler and Fritz London showed the stability of the valence bond was caused by the wave mechanical phenomenon of resonance. Chemist Linus Pauling extended their treatment of the valence bond to a theory of structural chemistry in The Nature of the Chemical Bond. His synthesis of the physical and chemical views, his value as a physico-chemical researcher during the 1930s, and the research of his contemporaries John Slater and Robert Mulliken show that a true physico-chemical blend was only realized within the amorphous discipline of chemical physics. Finally, it is seen that this interdisciplinarity of chemical bonding and its supporting framework force a reevaluation of the reductionist criteria, and a re-definition of the chemical bond as a physico-chemical work.