Truth and tractability: Compromising between accuracy and computational cost in quantum computational chemistry methods for noncovalent interactions and metal-salen catalysis

Computational chemists are concerned about two aspects when choosing between the myriad of theoretical methodologies: the accuracy (the "truth") and the computational cost (the tractability). Among the least expensive methods for the approximation of the electronic Schrodinger equation are the Hartree-Fock (HF), density functional theory (DFT), and second-order Moller-Plesset perturbation theory (MP2) methods. While each of these methods yield excellent results in many cases, the inadequate inclusion of certain types of electron correlation (either high-orders or nondynamical) can produce erroneous results. For example, when utilizing the HF or DFT methods for the computation of noncovalent interactions, both methods fail to predict favorable London dispersion interactions. The MP2 method, on the other hand, dramatically overbinds noncovalent pi-pi interactions compared to the chemically accurate CCSD(T) method. Likewise, for the computation of transition metal-salen spin-state energy gaps, DFT methods yield large errors (tens of kcal mol-1) compared to the robust CASPT3 method. Both CCSD(T) and CASPT3 methods, however, are limited to the computation of small to medium sized molecules with "modest-sized" basis sets (currently, around 30 atoms for single point computations).;The compromise for the computation of noncovalent interactions often comes from empirically scaling DFT and/or MP2 methods to fit benchmark data sets. The DFT method with an empirically fit dispersion term (DFT-D) often yields semi-quantitative results. The spin-component scaled MP2 (SCS-MP2) method parameterizes the same- and opposite-spin correlation energies to approximate CCSD(T) results and often yields less than 20% error for prototype noncovalent systems. Each of these methods have the added benefit of being no more computationally expensive than their canonical counterparts. Therefore, the DFT-D and SCS-MP2 methods can be utilized on rather large systems.;There is no simple fix for cases with a large degree of nondynamical correlation (such as transition metal-salen complexes). While testing the new M0-family of meta-GGA DFT functionals on the spin-state energy gaps of transition metal-salen complexes, no DFT method produced reliable results. Therefore each metal-salen complex must be evaluated on a case-by-case basis to determine which methods are the most reliable. For the 'well-behaved' Al(III)-salen complex, all single reference methods produce reliable results because the Al(III)-salen complex contains little to no important nondynamical correlation. The salen-ligand, however, does contain a large degree of pi-aromaticity. Therefore, when studying the reaction energy profile for the addition of cyanide to unsaturated imides catalyzed by two Al(Cl)-salen complexes, the BP86-D method was used for geometry optimizations followed by SCS-MP2 single point computations to accurately capture London dispersion forces. Compromising between these two methods, the entire reaction energy profile was tractable with resources found in a personal desktop computer while providing enough 'truth' that explained many experimental observations.;As an introduction to quantum computational methods, chapter one will provide a brief development of the HF, electron correlation, multireference, and DFT methods. Particular attention will be paid to the various approximations these methods are built upon. Chapter two will test the performance of these methods for the description of noncovalent interactions. Also within this chapter, empirically scaled quantum computational methods will be explored. All sections in chapter two are based on previously published work by this author. Section 2.2 provides a pioneering article that investigates the new SCS-MP2 method for a variety of noncovalent interactions.[1] Section 2.3 develops the SCS-CCSD method and provides preliminary results for noncovalent interactions.[2] Section 2.4 includes part of a feature article (primarily written by C. D. Sherrill) that assesses a myriad of new theoretical methods for the newly obtained complete basis set CCSD(T) noncovalent interaction potential energy curves.[3] Section 2.5 investigates the intricate indole-benzene complex with the SCS-MP2 method.[4] And section 2.6 tests a variety of approximate methods for a set of 22 highly accurate noncovalent interaction energies.[5] Chapter three will focus on various metal-salen complexes. Section 3.2 tests the performance of standard DFT methods in comparison with very robust multireference wavefunction based methods for metal-salen energetics and is based on previously published work.[6] Section 3.3 extends this work on the metal-salen energetics to include the new M0-family of DFT functionals and is based on submitted material.[7] Lastly, section 3.3 investigates the cyanide addition to unsaturated imides reaction as catalyzed by the Al(Cl)-salen complex. This work is currently in preparation for publication.[8] Chapter four provides concluding remarks and important future work.