Machine-Learning-Based Prediction of the Glass Transition Temperature of Organic Compounds Using Experimental Data.

Knowledge of the glass transition temperature of molecular compounds that occur in atmospheric aerosol particles is important for estimating their viscosity, as it directly influences the kinetics of chemical reactions and particle phase state. While there is a great diversity of organic compounds present in aerosol particles, for only a minor fraction of them experimental glass transition temperatures are known. Therefore, we have developed a machine learning model designed to predict the glass transition temperature of organic molecular compounds based on molecule-derived input variables. The extremely randomized trees (extra trees) procedure was chosen for this purpose. Two approaches using different sets of input variables were followed. The first one uses the number of selected functional groups present in the compound, while the second one generates descriptors from a SMILES (Simplified Molecular Input Line Entry System) string. Organic compounds containing carbon, hydrogen, oxygen, nitrogen, and halogen atoms are included. For improved results, both approaches can be combined with the melting temperature of the compound as an additional input variable. The results show that the predictions of both approaches show a similar mean absolute error of about 12-13 K, with the SMILES-based predictions performing slightly better. In general, the model shows good predictive power considering the diversity of the experimental input data. Furthermore, we also show that its performance exceeds that of previous parameterizations developed for this purpose and also performs better than existing machine learning models. In order to provide user-friendly versions of the model for applications, we have developed a web site where the model can be run by interested scientists via a web-based interface without prior technical knowledge. We also provide Python code of the model. Additionally, all experimental input data are provided in form of the Bielefeld Molecular Organic Glasses (BIMOG) database. We believe that this model is a powerful tool for many applications in atmospheric aerosol science and material science.

Ice nucleation in aqueous solutions of short- and long-chain poly(vinyl alcohol) studied with a droplet microfluidics setup.

Poly(vinyl alcohol) (PVA) has ice binding and ice nucleating properties. Here, we explore the dependence of the molecular size of PVA on its ice nucleation activity. For this purpose, we studied ice nucleation in aqueous solutions of PVA samples with molar masses ranging from 370 to 145 000 g mol-1, with a particular focus on oligomer samples with low molar mass. The experiments employed a novel microfluidic setup that is a follow-up on the previous WeIzmann Supercooled Droplets Observation on a Microarray (WISDOM) design by Reicher et al. The modified setup introduced and characterized here, termed nanoliter Bielefeld Ice Nucleation ARraY (nanoBINARY), uses droplet microfluidics with droplets (96 ± 4) µm in diameter and a fluorinated continuous oil phase and surfactant. A comparison of homogeneous and heterogeneous ice nucleation data obtained with nanoBINARY to those obtained with WISDOM shows very good agreement, underpinning its ability to study low-temperature ice nucleators as well as homogeneous ice nucleation due to the low background of impurities. The experiments on aqueous PVA solutions revealed that the ice nucleation activity of shorter PVA chains strongly decreases with a decrease in molar mass. While the cumulative number of ice nucleating sites per mass nm of polymers with different molar masses is the same, it becomes smaller for oligomers and completely vanishes for dimer and monomer representatives such as 1,3-butanediol, propan-2-ol, and ethanol, most likely because these molecules become too small to effectively stabilize the critical ice embryo. Overall, our results are consistent with PVA polymers and oligomers acting as heterogeneous ice nucleators.

Water-organizing motif continuity is critical for potent ice nucleation protein activity.

Bacterial ice nucleation proteins (INPs) can cause frost damage to plants by nucleating ice formation at high sub-zero temperatures. Modeling of Pseudomonas borealis INP by AlphaFold suggests that the central domain of 65 tandem sixteen-residue repeats forms a beta-solenoid with arrays of outward-pointing threonines and tyrosines, which may organize water molecules into an ice-like pattern. Here we report that mutating some of these residues in a central segment of P. borealis INP, expressed in Escherichia coli, decreases ice nucleation activity more than the section's deletion. Insertion of a bulky domain has the same effect, indicating that the continuity of the water-organizing repeats is critical for optimal activity. The ~10 C-terminal coils differ from the other 55 coils in being more basic and lacking water-organizing motifs; deletion of this region eliminates INP activity. We show through sequence modifications how arrays of conserved motifs form the large ice-nucleating surface required for potency.

Specific Ion-Protein Interactions Influence Bacterial Ice Nucleation.

Ice nucleation-active bacteria are the most efficient ice nucleators known, enabling the crystallization of water at temperatures close to 0 °C, thereby overcoming the kinetically hindered phase transition process at these conditions. Using highly specialized ice-nucleating proteins (INPs), they can cause frost damage to plants and influence the formation of clouds and precipitation in the atmosphere. In nature, the bacteria are usually found in aqueous environments containing ions. The impact of ions on bacterial ice nucleation efficiency, however, has remained elusive. Here, we demonstrate that ions can profoundly influence the efficiency of bacterial ice nucleators in a manner that follows the Hofmeister series. Weakly hydrated ions inhibit bacterial ice nucleation whereas strongly hydrated ions apparently facilitate ice nucleation. Surface-specific sum-frequency generation spectroscopy and molecular dynamics simulations reveal that the different effects are due to specific interactions of the ions with the INPs on the surface of the bacteria. Our results demonstrate that heterogeneous ice nucleation facilitated by bacteria strongly depends upon the nature of the ions, and specific ion-protein interactions are essential for the complete description of heterogeneous ice nucleation by bacteria.

Ice Nucleation Properties of Ice-binding Proteins from Snow Fleas.

Ice-binding proteins (IBPs) are found in many organisms, such as fish and hexapods, plants, and bacteria that need to cope with low temperatures. Ice nucleation and thermal hysteresis are two attributes of IBPs. While ice nucleation is promoted by large proteins, known as ice nucleating proteins, the smaller IBPs, referred to as antifreeze proteins (AFPs), inhibit the growth of ice crystals by up to several degrees below the melting point, resulting in a thermal hysteresis (TH) gap between melting and ice growth. Recently, we showed that the nucleation capacity of two types of IBPs corresponds to their size, in agreement with classical nucleation theory. Here, we expand this finding to additional IBPs that we isolated from snow fleas (the arthropod Collembola), collected in northern Israel. Chemical analyses using circular dichroism and Fourier-transform infrared spectroscopy data suggest that these IBPs have a similar structure to a previously reported snow flea antifreeze protein. Further experiments reveal that the ice-shell purified proteins have hyperactive antifreeze properties, as determined by nanoliter osmometry, and also exhibit low ice-nucleation activity in accordance with their size.

Contrasting Behavior of Antifreeze Proteins: Ice Growth Inhibitors and Ice Nucleation Promoters.

Several types of natural molecules interact specifically with ice crystals. Small antifreeze proteins (AFPs) adsorb to particular facets of ice crystals, thus inhibiting their growth, whereas larger ice-nucleating proteins (INPs) can trigger the formation of new ice crystals at temperatures much higher than the homogeneous ice nucleation temperature of pure water. It has been proposed that both types of proteins interact similarly with ice and that, in principle, they may be able to exhibit both functions. Here we investigated two naturally occurring antifreeze proteins, one from fish, type-III AFP, and one from beetles, TmAFP. We show that in addition to ice growth inhibition, both can also trigger ice nucleation above the homogeneous freezing temperature, providing unambiguous experimental proof for their contrasting behavior. Our analysis suggests that the predominant difference between AFPs and INPs is their molecular size, which is a very good predictor of their ice nucleation temperature.

Second inflection point of water surface tension in the deeply supercooled regime revealed by entropy anomaly and surface structure using molecular dynamics simulations.

The surface tension of supercooled water is of fundamental importance in physical chemistry and materials and atmospheric sciences. Controversy, however, exists over its temperature dependence in the supercooled regime, especially on the existence of the "second inflection point (SIP)". Here, we use molecular dynamics simulations of the SPC/E water model to study the surface tension of water (σw) as a function of temperature down to 198.15 K, and find a minimum point of surface excess entropy per unit area around ∼240-250 K. Additional simulations with the TIP4P/2005 water model also show consistent results. Hence, we predict an SIP of σw roughly in this region, at the boundary where the "no man's land" happens. The increase of surface entropy with decreasing temperature in the region below the inflection point is clearly an anomalous behavior, unknown for simple liquids. Furthermore, we find that σw has a near-linear correlation with the interfacial width, which can be well explained by the capillary wave theory. Deep in the supercooled regime, a compact water layer at the interface is detected in our simulations, which may be a key component that contributes to the deviation of surface tension from the International Association for the Properties of Water and Steam relationship. Our findings may advance the understanding of the origin of the anomalous properties of liquid water in the supercooled regime.

Global distribution of particle phase state in atmospheric secondary organic aerosols.

Secondary organic aerosols (SOA) are a large source of uncertainty in our current understanding of climate change and air pollution. The phase state of SOA is important for quantifying their effects on climate and air quality, but its global distribution is poorly characterized. We developed a method to estimate glass transition temperatures based on the molar mass and molecular O:C ratio of SOA components, and we used the global chemistry climate model EMAC with the organic aerosol module ORACLE to predict the phase state of atmospheric SOA. For the planetary boundary layer, global simulations indicate that SOA are mostly liquid in tropical and polar air with high relative humidity, semi-solid in the mid-latitudes and solid over dry lands. We find that in the middle and upper troposphere SOA should be mostly in a glassy solid phase state. Thus, slow diffusion of water, oxidants and organic molecules could kinetically limit gas-particle interactions of SOA in the free and upper troposphere, promote ice nucleation and facilitate long-range transport of reactive and toxic organic pollutants embedded in SOA.

Boreal pollen contain ice-nucleating as well as ice-binding 'antifreeze' polysaccharides.

Ice nucleation and growth is an important and widespread environmental process. Accordingly, nature has developed means to either promote or inhibit ice crystal formation, for example ice-nucleating proteins in bacteria or ice-binding antifreeze proteins in polar fish. Recently, it was found that birch pollen release ice-nucleating macromolecules when suspended in water. Here we show that birch pollen washing water exhibits also ice-binding properties such as ice shaping and ice recrystallization inhibition, similar to antifreeze proteins. We present spectroscopic evidence that both the ice-nucleating as well as the ice-binding molecules are polysaccharides bearing carboxylate groups. The spectra suggest that both polysaccharides consist of very similar chemical moieties, but centrifugal filtration indicates differences in molecular size: ice nucleation occurs only in the supernatant of a 100 kDa filter, while ice shaping is strongly enhanced in the filtrate. This finding may suggest that the larger ice-nucleating polysaccharides consist of clusters of the smaller ice-binding polysaccharides, or that the latter are fragments of the ice-nucleating polysaccharides. Finally, similar polysaccharides released from pine and alder pollen also display both ice-nucleating as well as ice-binding ability, suggesting a common mechanism of interaction with ice among several boreal pollen with implications for atmospheric processes and antifreeze protection.

A physically constrained classical description of the homogeneous nucleation of ice in water.

Liquid water can persist in a supercooled state to below 238 K in the Earth's atmosphere, a temperature range where homogeneous nucleation becomes increasingly probable. However, the rate of homogeneous ice nucleation in supercooled water is poorly constrained, in part, because supercooled water eludes experimental scrutiny in the region of the homogeneous nucleation regime where it can exist only fleetingly. Here we present a new parameterization of the rate of homogeneous ice nucleation based on classical nucleation theory. In our approach, we constrain the key terms in classical theory, i.e., the diffusion activation energy and the ice-liquid interfacial energy, with physically consistent parameterizations of the pertinent quantities. The diffusion activation energy is related to the translational self-diffusion coefficient of water for which we assess a range of descriptions and conclude that the most physically consistent fit is provided by a power law. The other key term is the interfacial energy between the ice embryo and supercooled water whose temperature dependence we constrain using the Turnbull correlation, which relates the interfacial energy to the difference in enthalpy between the solid and liquid phases. The only adjustable parameter in our model is the absolute value of the interfacial energy at one reference temperature. That value is determined by fitting this classical model to a selection of laboratory homogeneous ice nucleation data sets between 233.6 K and 238.5 K. On extrapolation to temperatures below 233 K, into a range not accessible to standard techniques, we predict that the homogeneous nucleation rate peaks between about 227 and 231 K at a maximum nucleation rate many orders of magnitude lower than previous parameterizations suggest. This extrapolation to temperatures below 233 K is consistent with the most recent measurement of the ice nucleation rate in micrometer-sized droplets at temperatures of 227-232 K on very short time scales using an X-ray laser technique. In summary, we present a new physically constrained parameterization for homogeneous ice nucleation which is consistent with the latest literature nucleation data and our physical understanding of the properties of supercooled water.

Sensitivity of liquid clouds to homogenous freezing parameterizations.

Water droplets in some clouds can supercool to temperatures where homogeneous ice nucleation becomes the dominant freezing mechanism. In many cloud resolving and mesoscale models, it is assumed that homogeneous ice nucleation in water droplets only occurs below some threshold temperature typically set at -40°C. However, laboratory measurements show that there is a finite rate of nucleation at warmer temperatures. In this study we use a parcel model with detailed microphysics to show that cloud properties can be sensitive to homogeneous ice nucleation as warm as -30°C. Thus, homogeneous ice nucleation may be more important for cloud development, precipitation rates, and key cloud radiative parameters than is often assumed. Furthermore, we show that cloud development is particularly sensitive to the temperature dependence of the nucleation rate. In order to better constrain the parameterization of homogeneous ice nucleation laboratory measurements are needed at both high (>-35°C) and low (<-38°C) temperatures. KEY POINTS: Homogeneous freezing may be significant as warm as -30°CHomogeneous freezing should not be represented by a threshold approximationThere is a need for an improved parameterization of homogeneous ice nucleation.

Size dependence of phase transitions in aerosol nanoparticles.

Phase transitions of nanoparticles are of fundamental importance in atmospheric sciences, but current understanding is insufficient to explain observations at the nano-scale. In particular, discrepancies exist between observations and model predictions of deliquescence and efflorescence transitions and the hygroscopic growth of salt nanoparticles. Here we show that these discrepancies can be resolved by consideration of particle size effects with consistent thermodynamic data. We present a new method for the determination of water and solute activities and interfacial energies in highly supersaturated aqueous solution droplets (Differential Köhler Analysis). Our analysis reveals that particle size can strongly alter the characteristic concentration of phase separation in mixed systems, resembling the influence of temperature. Owing to similar effects, atmospheric secondary organic aerosol particles at room temperature are expected to be always liquid at diameters below ~20 nm. We thus propose and demonstrate that particle size should be included as an additional dimension in the equilibrium phase diagram of aerosol nanoparticles.

Glass formation processes in mixed inorganic/organic aerosol particles.

Recent experiments suggest that organic aerosol particles may transform into a glassy state at room temperature under dry conditions. Information on glass forming processes in mixed inorganic/organic aerosol particles is sparse, however, because inorganic crystal nucleation is usually very likely in such mixtures. Here we investigate the glass transition temperatures Tg of various organics (trehalose, sucrose, citric acid, sorbitol, and glycerol as well as 3-MBTCA) in binary mixtures with either NaNO3 or NH4HSO4 at different mass fractions. The glassy samples were prepared with the MARBLES technique by atomizing dilute aqueous solutions into aerosol particles and subsequent diffusion drying. The resulting aerosol particles were collected and their phase behavior was investigated using differential scanning calorimetry. At small and intermediate inorganic mass fractions salt crystallization did not occur. Instead, the single-phase mixtures remained in an amorphous state upon drying such that determination of their Tg was possible. From these measurements the Tg value of pure NaNO3 and pure NH4HSO4 could be inferred through extrapolation, resulting in values of Tg(NaNO3) ≈ 290 K and Tg(NH4HSO4) ≈ 220 K. Upon drying of NH4HSO4/3-MBTCA mixtures, phase-separated samples formed in which the inorganic-rich and organic-rich phases each show an independent glass transition. Our measurements provide a route toward establishing Tg values of inorganic salts that usually crystallize readily, and they may explain the reported contradicting observations of NaNO3 aerosol particles to either crystallize or remain amorphous upon drying at room temperature.

Glass-forming properties of 3-methylbutane-1,2,3-tricarboxylic acid and its mixtures with water and pinonic acid.

3-Methylbutane-1,2,3-tricarboxylic acid (3-MBTCA) is an atmospheric oxidation product of α-pinene and has been identified as the most relevant tracer compound for atmospheric terpene secondary organic aerosol (SOA) particles. Little is known, however, of its physicochemical properties such as water solubility and phase state (e.g., liquid, crystalline, glassy). To gain knowledge, we synthesized 3-MBCTA from methyl 2-methylpropanoate and dimethyl maleate via a Michael addition and subsequent hydrolysis with 78% overall yield. It was found that 3-MBTCA transforms into anhydrides upon melting at Tm = 426 ± 1 K, thus preventing a determination of the glass transition temperature Tg by differential scanning calorimetry (DSC) through melting and subsequent cooling. Therefore, we designed the novel technique MARBLES (metastable aerosol by low temperature evaporation of solvent) for transferring a substance into a glassy state without heating. In MARBLES an aqueous solution is atomized into wet aerosol particles that are subsequently dried in several diffusion dryers resulting in glass formation of the residual particles for several solutes. The glassy aerosol particles are collected in an impactor until enough mass has accumulated that the sample's Tg can be determined by DSC. Using this method, the glass transition temperature of 3-MBTCA was found to be Tg ≈ 305 ± 2 K. Moreover, we have determined the glass transition Tg' of the maximal freeze-concentrated aqueous solution of 3-MBTCA, and Tg of mixtures of 3-MBTCA with water and pinonic acid. The latter data indicate a dependence of Tg upon the atomic oxygen-to-carbon ratio of the mixture, with implications for parametrizing the glass-forming behavior of α-pinene SOA particles in the atmosphere.

Formation of highly porous aerosol particles by atmospheric freeze-drying in ice clouds.

The cycling of atmospheric aerosols through clouds can change their chemical and physical properties and thus modify how aerosols affect cloud microphysics and, subsequently, precipitation and climate. Current knowledge about aerosol processing by clouds is rather limited to chemical reactions within water droplets in warm low-altitude clouds. However, in cold high-altitude cirrus clouds and anvils of high convective clouds in the tropics and midlatitudes, humidified aerosols freeze to form ice, which upon exposure to subsaturation conditions with respect to ice can sublimate, leaving behind residual modified aerosols. This freeze-drying process can occur in various types of clouds. Here we simulate an atmospheric freeze-drying cycle of aerosols in laboratory experiments using proxies for atmospheric aerosols. We find that aerosols that contain organic material that undergo such a process can form highly porous aerosol particles with a larger diameter and a lower density than the initial homogeneous aerosol. We attribute this morphology change to phase separation upon freezing followed by a glass transition of the organic material that can preserve a porous structure after ice sublimation. A porous structure may explain the previously observed enhancement in ice nucleation efficiency of glassy organic particles. We find that highly porous aerosol particles scatter solar light less efficiently than nonporous aerosol particles. Using a combination of satellite and radiosonde data, we show that highly porous aerosol formation can readily occur in highly convective clouds, which are widespread in the tropics and midlatitudes. These observations may have implications for subsequent cloud formation cycles and aerosol albedo near cloud edges.

The homogeneous ice nucleation rate of water droplets produced in a microfluidic device and the role of temperature uncertainty.

Ice nucleation was investigated experimentally in water droplets with diameters between 53 and 96 micrometres. The droplets were produced in a microfluidic device in which a flow of methyl-cyclohexane and water was combined at the T-junction of micro-channels yielding inverse (water-in-oil) emulsions consisting of water droplets with small standard deviations. In cryo-microscopic experiments we confirmed that upon cooling of such emulsion samples ice nucleation in individual droplets occurred independently of each other as required for the investigation of a stochastic process. The emulsion samples were then subjected to cooling at 1 Kelvin per minute in a differential scanning calorimeter with high temperature accuracy. From the latent heat released by freezing water droplets we inferred the volume-dependent homogeneous ice nucleation rate coefficient of water at temperatures between 236.5 and 237.9 Kelvin. A comparison of our newly derived values to existing rate coefficients from other studies suggests that the volume-dependent ice nucleation rate in supercooled water is slightly lower than previously thought. Moreover, a comprehensive error analysis suggests that absolute temperature accuracy is the single most important experimental parameter determining the uncertainty of the derived ice nucleation rates in our experiments, and presumably also in many previous experiments. Our analysis, thus, also provides a route for improving the accuracy of future ice nucleation rate measurements.

Antifreeze glycopeptide diastereomers.

Antifreeze glycopeptides (AFGPs) are a special class of biological antifreeze agents, which possess the property to inhibit ice growth in the body fluids of arctic and antarctic fish and, thus, enable life under these harsh conditions. AFGPs are composed of 4-55 tripeptide units -Ala-Ala-Thr- glycosylated at the threonine side chains. Despite the structural homology among all the fish species, divergence regarding the composition of the amino acids occurs in peptides from natural sources. Although AFGPs were discovered in the early 1960s, the adsorption mechanism of these macromolecules to the surface of the ice crystals has not yet been fully elucidated. Two AFGP diastereomers containing different amino acid configurations were synthesized to study the influence of amino acid stereochemistry on conformation and antifreeze activity. For this purpose, peptides containing monosaccharide-substituted allo-L- and D-threonine building blocks were assembled by solid-phase peptide synthesis (SPPS). The retro-inverso AFGP analogue contained all amino acids in D-configuration, while the allo-L-diastereomer was composed of L-amino acids, like native AFGPs, with replacement of L-threonine by its allo-L-diastereomer. Both glycopeptides were analyzed regarding their conformational properties, by circular dichroism (CD), and their ability to inhibit ice recrystallization in microphysical experiments.

Influence of sequential modifications and carbohydrate variations in synthetic AFGP analogues on conformation and antifreeze activity.

Certain Arctic and Antarctic ectotherm species have developed strategies for survival under low temperature conditions that, among others, consist of antifreeze glycopeptides (AFGP). AFGP form a class of biological antifreeze agents that exhibit the ability to inhibit ice growth in vitro and in vivo and, hence, enable life at temperatures below the freezing point. AFGP usually consist of a varying number of (Ala-Ala-Thr)(n) units (n=4-55) with the disaccharide ß-D-galactosyl-(1→3)-α-N-acetyl-D-galactosamine glycosidically attached to every threonine side chain hydroxyl group. AFGP have been shown to adopt polyproline II helical conformation. Although this pattern is highly conserved among different species, microheterogeneity concerning the amino acid composition usually occurs; for example, alanine is occasionally replaced by proline in smaller AFGP. The influence of minor and major sequence mutations on conformation and antifreeze activity of AFGP analogues was investigated by replacement of alanine by proline and glycosylated threonine by glycosylated hydroxyproline. The target compounds were prepared by using microwave-enhanced solid phase peptide synthesis. Furthermore, artificial analogues were obtained by copper-catalyzed azide-alkyne cycloaddition (CuAAC): propargyl glycosides were treated with polyproline helix II-forming peptides comprising (Pro-Azp-Pro)(n) units (n=2-4) that contained 4-azidoproline (Azp). The conformations of all analogues were examined by circular dichroism (CD). In addition, microphysical analysis was performed to provide information on their inhibitory effect on ice recrystallization.