Using slisemap to interpret physical data.

Manifold visualisation techniques are commonly used to visualise high-dimensional datasets in physical sciences. In this paper, we apply a recently introduced manifold visualisation method, slisemap, on datasets from physics and chemistry. slisemap combines manifold visualisation with explainable artificial intelligence. Explainable artificial intelligence investigates the decision processes of black box machine learning models and complex simulators. With slisemap, we find an embedding such that data items with similar local explanations are grouped together. Hence, slisemap gives us an overview of the different behaviours of a black box model, where the patterns in the embedding reflect a target property. In this paper, we show how slisemap can be used and evaluated on physical data and that it is helpful in finding meaningful information on classification and regression models trained on these datasets.

Computational Tools for Handling Molecular Clusters: Configurational Sampling, Storage, Analysis, and Machine Learning.

Computational modeling of atmospheric molecular clusters requires a comprehensive understanding of their complex configurational spaces, interaction patterns, stabilities against fragmentation, and even dynamic behaviors. To address these needs, we introduce the Jammy Key framework, a collection of automated scripts that facilitate and streamline molecular cluster modeling workflows. Jammy Key handles file manipulations between varieties of integrated third-party programs. The framework is divided into three main functionalities: (1) Jammy Key for configurational sampling (JKCS) to perform systematic configurational sampling of molecular clusters, (2) Jammy Key for quantum chemistry (JKQC) to analyze commonly used quantum chemistry output files and facilitate database construction, handling, and analysis, and (3) Jammy Key for machine learning (JKML) to manage machine learning methods in optimizing molecular cluster modeling. This automation and machine learning utilization significantly reduces manual labor, greatly speeds up the search for molecular cluster configurations, and thus increases the number of systems that can be studied. Following the example of the Atmospheric Cluster Database (ACDB) of Elm (ACS Omega, 4, 10965-10984, 2019), the molecular clusters modeled in our group using the Jammy Key framework have been stored in an improved online GitHub repository named ACDB 2.0. In this work, we present the Jammy Key package alongside its assorted applications, which underline its versatility. Using several illustrative examples, we discuss how to choose appropriate combinations of methodologies for treating particular cluster types, including reactive, multicomponent, charged, or radical clusters, as well as clusters containing flexible or multiconformer monomers or heavy atoms. Finally, we present a detailed example of using the tools for atmospheric acid-base clusters.

Atomic structures, conformers and thermodynamic properties of 32k atmospheric molecules.

Low-volatile organic compounds (LVOCs) drive key atmospheric processes, such as new particle formation (NPF) and growth. Machine learning tools can accelerate studies of these phenomena, but extensive and versatile LVOC datasets relevant for the atmospheric research community are lacking. We present the GeckoQ dataset with atomic structures of 31,637 atmospherically relevant molecules resulting from the oxidation of α-pinene, toluene and decane. For each molecule, we performed comprehensive conformer sampling with the COSMOconf program and calculated thermodynamic properties with density functional theory (DFT) using the Conductor-like Screening Model (COSMO). Our dataset contains the geometries of the 7 Mio. conformers we found and their corresponding structural and thermodynamic properties, including saturation vapor pressures (pSat), chemical potentials and free energies. The pSat were compared to values calculated with the group contribution method SIMPOL. To validate the dataset, we explored the relationship between structural and thermodynamic properties, and then demonstrated a first machine-learning application with Gaussian process regression.

Assessment of physicochemical properties of sorbent materials in passive and active sampling systems towards gaseous nitrogen-containing compounds.

The adsorption and desorption behavior of volatile nitrogen-containing compounds in vapor phase by solid-phase microextraction Arrow (SPME-Arrow) and in-tube extraction (ITEX) sampling systems, were investigated experimentally using gas chromatography-mass spectrometry. Three different SPME-Arrow coating materials, DVB/PDMS, MCM-41, and MCM-41-TP and two ITEX adsorbents, TENAX-GR and MCM-41-TP were compared to clarify the selectivity of the sorbents towards nitrogen-containing compounds. In addition, saturated vapor pressures for these compounds were estimated, both experimentally and theoretically. In this study, the adsorption of nitrogen-containing compounds on various adsorbents followed the Elovich model well, while a pseudo-first-order kinetics model best described the desorption kinetics. Pore volume and pore sizes of the coating sorbents were essential parameters for the determination of the adsorption performance for the SPME-Arrow sampling system. MCM-41-TP coating with the smallest pore size gave the slowest adsorption rate compared to that of DVB/PDMS and MCM-41 in the SPME-Arrow sampling system. Both adsorbent and adsorbate properties, such as hydrophobicity and basicity, affected the adsorption and desorption kinetics in SPME-Arrow system. The adsorption and desorption rates of studied C6H15N isomers in the MCM-41 and MCM-41-TP sorbent materials of SPME-Arrow system were higher for dipropylamine and triethylamine (branched amines) than for hexylamine (linear chain amines). DVB/PDMS-SPME-Arrow gave fast adsorption rates for the aromatic-ringed pyridine and o-toluidine. All studied nitrogen-containing compounds demonstrated high desorption rates with DVB/PDMS-SPME-Arrow. Chemisorption and physisorption were the sorption mechanisms in MCM-41- and MCM-41-TP- SPME-Arrow, but additional experiments are needed to confirm this. An active sampling technique ITEX gave comparable adsorption and desorption rates on the selective MCM-41-TP and universal TENAX-GR sorbent materials for all the compounds studied. Vapor pressures of nitrogen-containing compounds were experimentally estimated by using retention index approach and these values were compared with the theoretical ones, calculated using the COnductor-like Screening MOdel for Real Solvent (COSMO-RS) model. Both values agreed well with those found in the literature proving that these methods can be successfully used in predicting VOC's vapor pressures, e.g. for the formation of secondary organic aerosols.

Atmospheric Sulfuric Acid-Multi-Base New Particle Formation Revealed through Quantum Chemistry Enhanced by Machine Learning.

The formation of molecular clusters and secondary aerosols in the atmosphere has a significant impact on the climate. Studies typically focus on the new particle formation (NPF) of sulfuric acid (SA) with a single base molecule (e.g., dimethylamine or ammonia). In this work, we examine the combinations and synergy of several bases. Specifically, we used computational quantum chemistry to perform configurational sampling (CS) of (SA)0-4(base)0-4 clusters with five different types of bases: ammonia (AM), methylamine (MA), dimethylamine (DMA), trimethylamine (TMA), and ethylenediamine (EDA). Overall, we studied 316 different clusters. We used a traditional multilevel funnelling sampling approach augmented by a machine-learning (ML) step. The ML made the CS of these clusters possible by significantly enhancing the speed and quality of the search for the lowest free energy configurations. Subsequently, the cluster thermodynamics properties were evaluated at the DLPNO-CCSD(T0)/aug-cc-pVTZ//ωB97X-D/6-31++G(d,p) level of theory. The calculated binding free energies were used to evaluate the cluster stabilities for population dynamics simulations. The resultant SA-driven NPF rates and synergies of the studied bases are presented to show that DMA and EDA act as nucleators (although EDA becomes weak in large clusters), TMA acts as a catalyzer, and AM/MA is often overshadowed by strong bases.

Impact of Quantum Chemistry Parameter Choices and Cluster Distribution Model Settings on Modeled Atmospheric Particle Formation Rates.

We tested the influence of various parameters on the new particle formation rate predicted for the sulfuric acid-ammonia system using quantum chemistry and cluster distribution dynamics simulations, in our case, Atmospheric Cluster Dynamics Code (ACDC). We found that consistent consideration of the rotational symmetry number of monomers (sulfuric acid and ammonia molecules, and bisulfate and ammonium ions) leads to a significant rise in the predicted particle formation rate, whereas inclusion of the rotational symmetry number of the clusters only changes the results slightly, and only in conditions where charged clusters dominate the particle formation rate. This is because most of the clusters stable enough to participate in new particle formation have a rotational symmetry number of 1, and few exceptions to this rule are positively charged clusters. In contrast, the application of the quasi-harmonic correction for low-frequency vibrational modes tends to generally decrease predicted new particle formation rates and also significantly alters the slope of the formation rate curve plotted against the sulfuric acid concentration, which is a typical convention in atmospheric aerosol science. The impact of the maximum size of the clusters explicitly included in the simulations depends on the simulated conditions. The errors arising from a limited set of clusters are higher for higher evaporation rates, and thus tend to increase with temperature. Similarly, the errors tend to be higher for lower vapor concentrations. The boundary conditions for outgrowing clusters (that are counted as formed particles) have only a small influence on the results, provided that the definition is chemically reasonable and that the set of simulated clusters is sufficiently large. A comparison with data from the Cosmics Leaving OUtdoor Droplets (CLOUD) chamber and a cluster distribution dynamics model using older quantum chemistry input data shows improved agreement when using our new input data and the proposed combination of symmetry and quasi-harmonic corrections.

Hydration of Atmospheric Molecular Clusters III: Procedure for Efficient Free Energy Surface Exploration of Large Hydrated Clusters.

Sampling the shallow free energy surface of hydrated atmospheric molecular clusters is a significant challenge. Using computational methods, we present an efficient approach to obtain minimum free energy structures for large hydrated clusters of atmospheric relevance. We study clusters consisting of two to four sulfuric acid (sa) molecules and hydrate them with up to five water (w) molecules. The structures of the "dry" clusters are obtained using the ABCluster program to yield a large pool of low-lying conformer minima with respect to free energy. The conformers (up to ten) lowest in free energy are then hydrated using our recently developed systematic hydrate sampling technique. Using this approach, we identify a total of 1145 unique (sa)2-4(w)1-5 cluster structures. The cluster geometries and thermochemical parameters are calculated at the ωB97X-D/6-31++G(d,p) level of theory, at 298.15 K and 1 atm. The single-point energy of the most stable clusters is calculated using a high-level DLPNO-CCSD(T0)/aug-cc-pVTZ method. Using the thermochemical data, we calculate the equilibrium hydrate distribution of the clusters under atmospheric conditions and find that the larger (sa)3 and (sa)4 clusters are significantly more hydrated than the smaller (sa)2 cluster or the sulfuric acid (sa)1 molecule. These findings indicate that more than five water molecules might be required to fully saturate the sulfuric acid clusters with water under atmospheric conditions. The presented methodology gives modelers a tool to take the effect of water explicitly into account in atmospheric particle formation models based on quantum chemistry.

Configurational Sampling of Noncovalent (Atmospheric) Molecular Clusters: Sulfuric Acid and Guanidine.

We studied the configurational sampling of noncovalently bonded molecular clusters relevant to the atmosphere. In this article, we discuss possible approaches to searching for optimal configurations and present one alternative based on systematic configurational sampling, which seems able to overcome the typical problems associated with searching for global minima on multidimensional potential energy surfaces. Since atmospheric molecular clusters are usually held together by intermolecular bonds, we also present a cost-effective strategy for treating hydrogen bonding and proton transferring by using rigid molecules and ions in different protonation states and illustrate its performance on clusters containing guanidine and sulfuric acid.