Chapter 1
Dynamics of Photochemical Reactions of Organic Carbonyls and their Clusters

Dorit Shemesh1 and R. Benny Gerber1,2,3

1Institute of Chemistry and the Fritz Haber Research Center, The Hebrew University of Jerusalem, Jerusalem, Israel

2Department of Chemistry, University of California, Irvine, CA, USA

3Laboratory of Physical Chemistry, University of Helsinki, Helsinki, Finland

1.1 Introduction

Photochemical reactions are central to organic chemistry and are playing a key role in atmospheric aerosols [1]. A variety of reactions occur, and the majority of them involve more than one molecule, mostly surrounded by a cluster (e.g., water). A common approach in modeling is to simplify the system and to treat only unimolecular reactions [2, 3]. However, understanding the processes involved in cluster is of great interest itself and approaches for modeling those processes need to be developed.

Theoretical simulation of photochemistry of organic molecules in clusters is very complicated and challenging due to the following reasons. The size of the system usually is too large for the usage of quantum chemistry-based methods. Condensed phase systems are commonly treated using classical mechanics [4]. On the contrary, force-field potentials utilized for large systems are not applicable here, since those are unable to treat reactions at all. Additional complications arise from the lack of potential energy surfaces for open shell systems, as well as the correct treatment of nonadiabatic transitions between different surfaces. The objective of this chapter is to provide theoretical tools for describing photochemical reactions of organic molecules in clusters and in condensed phase. The chapter focuses on the modeling of photochemical organic reactions using on-the-fly molecular dynamics on a semiempirical potential energy surface. It will be shown that semiempirical methods have the great advantage of being computationally fast enough and simultaneously provide qualitatively an accurate enough description of the systems.

An additional objective of this chapter is to provide mechanistic details and timescales for the important class of reactions in organic carbonyls. Organic carbonyls are of interest due to their importance in atmospheric chemistry and related areas [1, 5-7]. Carbonyls are important in addition to atmospheric chemistry in combustion, petroleum chemistry, biochemistry, and food chemistry. In several of these, the issue of photochemistry/photostability arises. Mechanistic details are known currently for relatively small carbonyls, as seen in the work by [8-30]. Large molecules are challenging for the simulation.

Pentanal, an aliphatic aldehyde, has been chosen for representing the unimolecular reactions of carbonyls [31, 32]. cis-Pinonic acid has been chosen as a representative ketone [33]. The abundance of carbonyl compounds in the atmosphere is very high. Certain carbonyls are directly emitted by various sources, but the vast majority of them are produced in the atmosphere by oxidation of hydrocarbons [1]. Photolysis serves as an important removal pathway for atmospheric carbonyls. In the lower atmosphere, where the availability of radiation is limited to wavelength above ~290 nm, the photolysis of carbonyls is driven by their weak absorption band in the wavelength range of 240-360 nm as a result of a dipole-forbidden n p* transition [1, 34]. Photolysis of aldehydes, such as pentanal, is known to occur through the following pathways:

Process A is the molecular fragmentation channel. Process B represents the fragmentation into two free radicals (Norrish type I splitting). Process C is called a Norrish type II splitting, and it results in acetaldehyde and an alkene as the products. Norrish type II splitting is only possible for aldehydes larger than butanal, and the reason pentanal was selected as the model for this work is to make sure this important channel is included in the calculation. Process D is an H abstraction process and has been found to be minor in small aldehydes [35]. In the microscopic picture, the photoexcitation promotes the system to the first excited singlet state (S1) of np* character. The S1 state can either switch to the ground S0 state via internal conversion (IC) or reach the lowest triplet state T1 via intersystem crossing (ISC). There is evidence that process B can occur either on the ground state or on the triplet state [36]. Reactions for ketones differ in part, but the two types of mechanisms are related, and both involve Norrish processes.

The above processes describe the possible unimolecular reactions occurring after photoexcitation of aldehydes. For assessing the effect of solvation on the photochemistry, the following model systems have been used. Pinonic acid has been studied with one and five water molecules, and a pentanal cluster using five identical pentanal molecules has been built.

The structure of this chapter is as follows: Section 1.2 provides an overview of the methodological approach of this study, in particular with emphasis on providing supporting validation for the semiempirical method. Section 1.3 gives results and discussion and finally Section 1.4 summarizes conclusions.

1.2 Methodology and Systems

The simulation of the photoexcitation process has been done here using on-the-fly molecular dynamics on a semiempirical potential energy surface. In brief, the first step involves molecular dynamics on the singlet ground state at 300 K. From this run, excitation from selected geometries to the singlet excited state occurs. After a short run on the singlet excited state, surface structures with a small S1-T1 gap are chosen. Those are used as starting geometries for the triplet state molecular dynamics simulation. In some of the simulation, the molecular dynamics simulation on the singlet state surface is omitted. For a detailed description and justification of this approach, the reader is referred to Section 1.2.3 and Refs [31, 33, 37].

Here we first discuss, in Section 1.2.1, the potential energy surface employed in this simulation. In Section 1.2.2, we discuss on-the-fly molecular dynamics approach. The different steps of the excitation process itself are explained in Section 1.2.3. The systems used are described in Section 1.2.4.

1.2.1 Potential Energy Surfaces

As already discussed in the introduction, the two edges regarding the accuracy of potential energy surfaces are (i) computationally very demanding high-level quantum methods for relatively small systems yielding very accurate data and (ii) comparable cheap force-field-based potential energy surfaces for very large systems (e.g., proteins) yielding qualitative information of nonreactive processes. In some cases of photochemical processes in condensed phases, ad hoc empirical potentials were constructed that seem to offer a reasonable description of the process [4]. However, in general, this does not seem a practical approach. Since the simulation of the processes required on the one hand potentials capable of describing reactions and on the other hand computational cheap ones, the choice has fallen on semiempirical methods. Semiempirical methods have been derived from the Hartree-Fock method and are therefore close to quantum methods [38]. Integrals are simplified by using parameters adapted to experimental data (e.g., heat of formation) and are therefore calculated very fast. Thus, semiempirical methods are capable of providing mechanistic information on photochemical processes in a relatively short time and are able to treat medium-size systems in a reasonable time.

The semiempirical method used in this study is the OM2/MRCI method developed by Thiel and coworkers [39, 40]. In the OM2 method, proper orthogonality of the orbitals has been introduced, which leads to a better description of energy barriers between different conformers. This method has been already applied to a great success in several studies by our group [31-33, 41, 42]. Other previously applied semiempirical method used in similar studies of our group is PM3 on modeling the vibrational spectra and the dynamics of relevant organic systems [37, 43-49]. All those studies have established semiempirical methods as a reliable tool.

All of the systems were initially optimized using a much higher level of theory, in order to ensure that the OM2 method provides a realistic description of the structure. The method employed was the second-order Møller-Plesset perturbation theory (MP2) [50] using the cc-pVDZ basis set [51]. The resolution-of-identity (RI) approximation for the evaluation of the electron-repulsion integrals implemented in Turbomole was utilized [52].

The advantage of the OM2 semiempirical method compared to other forms of approximations of related semiempirical methods lies in the ability of treating excited states and open shell systems. Excited states can be calculated using the OM2/MRCI variant of this method [40]. In this variant, an active space of orbitals is chosen, similar to a CASSCF calculation. In addition, different reference wave functions are employed and excitations from these reference functions are allowed. With this method, multireference systems can be treated by using several configurations for a particular state. For additional validation of the semiempirical method, the excitation energies were calculated...