Accurate prediction of ice nucleation from room temperature water.

Crystal nucleation is one of the most fundamental processes in the physical sciences and almost always occurs heterogeneously with the aid of a nucleating substrate. No example of nucleation is more ubiquitous and impactful than the formation of ice, vital to fields as diverse as geology, biology, aeronautics, and climate science. However, despite considerable effort, we still cannot predict a priori the efficacy of a nucleating agent. Here we utilize deep learning methods to accurately predict nucleation ability from images of room temperature liquid water-generated from molecular dynamics simulations-on a broad range of substrates. The resulting model, named IcePic, can rapidly and accurately infer nucleation ability, eliminating the requirement for either notoriously expensive simulations or direct experimental measurement. In an online poll, IcePic was found to significantly outperform humans in predicting the ice nucleating efficacy of materials. By analyzing the typical errors made by humans, as well as the application of reverse interpretation methods, physical insights into the role the water contact layer plays in ice nucleation have been obtained. Moving forward, we suggest that IcePic can be used as an easy, cheap, and rapid way to discern the nucleation ability of substrates, also with potential for learning other properties related to interfacial water.

Routes to cubic ice through heterogeneous nucleation.

The freezing of water into ice is one of the most important processes in the physical sciences. However, it is still not understood at the molecular level. In particular, the crystallization of cubic ice ([Formula: see text])-rather than the traditional hexagonal polytype ([Formula: see text])-has become an increasingly debated topic. Although evidence for [Formula: see text] is thought to date back almost 400 y, it is only in the last year that pure [Formula: see text] has been made in the laboratory, and these processes involved high-pressure ice phases. Since this demonstrates that pure [Formula: see text] can form, the question naturally arises if [Formula: see text] can be made from liquid water. With this in mind, we have performed a high-throughput computational screening study involving molecular dynamics simulations of nucleation on over 1,100 model substrates. From these simulations, we find that 1) many different substrates can promote the formation of pristine [Formula: see text]; 2) [Formula: see text] can be selectively nucleated for even the mildest supercooling; 3) the water contact layer's resemblance to a face of ice is the key factor determining the polytype selectivity and nucleation temperature, independent of which polytype is promoted; and 4) substrate lattice match to ice is not indicative of the polytype obtained. Through this study, we have deepened understanding of the interplay of heterogeneous nucleation and ice I polytypism and suggest routes to [Formula: see text] More broadly, the substrate design methodology presented here combined with the insight gained can be used to understand and control polymorphism and stacking disorder in materials in general.

Ice is born in low-mobility regions of supercooled liquid water.

When an ice crystal is born from liquid water, two key changes occur: (i) The molecules order and (ii) the mobility of the molecules drops as they adopt their lattice positions. Most research on ice nucleation (and crystallization in general) has focused on understanding the former with less attention paid to the latter. However, supercooled water exhibits fascinating and complex dynamical behavior, most notably dynamical heterogeneity (DH), a phenomenon where spatially separated domains of relatively mobile and immobile particles coexist. Strikingly, the microscopic connection between the DH of water and the nucleation of ice has yet to be unraveled directly at the molecular level. Here we tackle this issue via computer simulations which reveal that (i) ice nucleation occurs in low-mobility regions of the liquid, (ii) there is a dynamical incubation period in which the mobility of the molecules drops before any ice-like ordering, and (iii) ice-like clusters cause arrested dynamics in surrounding water molecules. With this we establish a clear connection between dynamics and nucleation. We anticipate that our findings will pave the way for the examination of the role of dynamical heterogeneities in heterogeneous and solution-based nucleation.

Crystal Nucleation in Liquids: Open Questions and Future Challenges in Molecular Dynamics Simulations.

The nucleation of crystals in liquids is one of nature's most ubiquitous phenomena, playing an important role in areas such as climate change and the production of drugs. As the early stages of nucleation involve exceedingly small time and length scales, atomistic computer simulations can provide unique insights into the microscopic aspects of crystallization. In this review, we take stock of the numerous molecular dynamics simulations that, in the past few decades, have unraveled crucial aspects of crystal nucleation in liquids. We put into context the theoretical framework of classical nucleation theory and the state-of-the-art computational methods by reviewing simulations of such processes as ice nucleation and the crystallization of molecules in solutions. We shall see that molecular dynamics simulations have provided key insights into diverse nucleation scenarios, ranging from colloidal particles to natural gas hydrates, and that, as a result, the general applicability of classical nucleation theory has been repeatedly called into question. We have attempted to identify the most pressing open questions in the field. We believe that, by improving (i) existing interatomic potentials and (ii) currently available enhanced sampling methods, the community can move toward accurate investigations of realistic systems of practical interest, thus bringing simulations a step closer to experiments.

Heterogeneous seeded molecular dynamics as a tool to probe the ice nucleating ability of crystalline surfaces.

Ice nucleation plays a significant role in a large number of natural and technological processes, but it is challenging to investigate experimentally because of the small time scales (ns) and short length scales (nm) involved. On the other hand, conventional molecular simulations struggle to cope with the relatively long time scale required for critical ice nuclei to form. One way to tackle this issue is to take advantage of free energy or path sampling techniques. Unfortunately, these are computationally costly. Seeded molecular dynamics is a much less demanding alternative that has been successfully applied already to study the homogeneous freezing of water. However, in the case of heterogeneous ice nucleation, nature's favourite route to form ice, an array of suitable interfaces between the ice seeds and the substrate of interest has to be built, and this is no trivial task. In this paper, we present a Heterogeneous SEEDing (HSEED) approach which harnesses a random structure search framework to tackle the ice-substrate challenge, thus enabling seeded molecular dynamics simulations of heterogeneous ice nucleation on crystalline surfaces. We validate the HSEED framework by investigating the nucleation of ice on (i) model crystalline surfaces, using the coarse-grained mW model, and (ii) cholesterol crystals, employing the fully atomistic TIP4P/ice water model. We show that the HSEED technique yields results in excellent agreement with both metadynamics and forward flux sampling simulations. Because of its computational efficiency, the HSEED method allows one to rapidly assess the ice nucleation ability of whole libraries of crystalline substrates-a long-awaited computational development in, e.g., atmospheric science.

Communication: Truncated non-bonded potentials can yield unphysical behavior in molecular dynamics simulations of interfaces.

Non-bonded potentials are included in most force fields and therefore widely used in classical molecular dynamics simulations of materials and interfacial phenomena. It is commonplace to truncate these potentials for computational efficiency based on the assumption that errors are negligible for reasonable cutoffs or compensated for by adjusting other interaction parameters. Arising from a metadynamics study of the wetting transition of water on a solid substrate, we find that the influence of the cutoff is unexpectedly strong and can change the character of the wetting transition from continuous to first order by creating artificial metastable wetting states. Common cutoff corrections such as the use of a force switching function, a shifted potential, or a shifted force do not avoid this. Such a qualitative difference urges caution and suggests that using truncated non-bonded potentials can induce unphysical behavior that cannot be fully accounted for by adjusting other interaction parameters.

The Many Faces of Heterogeneous Ice Nucleation: Interplay Between Surface Morphology and Hydrophobicity.

What makes a material a good ice nucleating agent? Despite the importance of heterogeneous ice nucleation to a variety of fields, from cloud science to microbiology, major gaps in our understanding of this ubiquitous process still prevent us from answering this question. In this work, we have examined the ability of generic crystalline substrates to promote ice nucleation as a function of the hydrophobicity and the morphology of the surface. Nucleation rates have been obtained by brute-force molecular dynamics simulations of coarse-grained water on top of different surfaces of a model fcc crystal, varying the water-surface interaction and the surface lattice parameter. It turns out that the lattice mismatch of the surface with respect to ice, customarily regarded as the most important requirement for a good ice nucleating agent, is at most desirable but not a requirement. On the other hand, the balance between the morphology of the surface and its hydrophobicity can significantly alter the ice nucleation rate and can also lead to the formation of up to three different faces of ice on the same substrate. We have pinpointed three circumstances where heterogeneous ice nucleation can be promoted by the crystalline surface: (i) the formation of a water overlayer that acts as an in-plane template; (ii) the emergence of a contact layer buckled in an ice-like manner; and (iii) nucleation on compact surfaces with very high interaction strength. We hope that this extensive systematic study will foster future experimental work aimed at testing the physiochemical understanding presented herein.

Machine Learning C-N Couplings: Obstacles for a General-Purpose Reaction Yield Prediction.

Pd-catalyzed C-N couplings are commonplace in academia and industry. Despite their significance, finding suitable reaction conditions leading to a high yield, for instance, remains a challenging and time-consuming task which usually requires screening over many sets of conditions. To help select promising reaction conditions in the vast space of reagent combinations, machine learning is an emerging technique with a lot of promise. In this work, we assess whether the reaction yield of C-N couplings can be predicted from databases of chemical reactions. We test the generalizability of models both on challenging data splits and on a dedicated experimental test set. We find that, provided the chemical space represented by the training set is not left, the models perform well. However, the applicability domain is quickly left even for simple reactions of the same type, as, for instance, present in our plate test set. The results show that yield prediction for new reactions is possible from the algorithmic side but in practice is hindered by the available data. Most importantly, more data that cover the diversity in reagents are needed for a general-purpose prediction of reaction yields. Our findings also expose a challenge to this field in that it appears to be extremely deceiving to judge models based on literature data with test sets which are split off the same literature data, even when challenging splits are considered.

How do interfaces alter the dynamics of supercooled water?

The structure of liquid water in the proximity of an interface can deviate significantly from that of bulk water, with surface-induced structural perturbations typically converging to bulk values at about ∼1 nm from the interface. While these structural changes are well established it is, in contrast, less clear how an interface perturbs the dynamics of water molecules within the liquid. Here, through an extensive set of molecular dynamics simulations of supercooled bulk and interfacial water films and nano-droplets, we observe the formation of persistent, spatially extended dynamical domains in which the average mobility varies as a function of the distance from the interface. This is in stark contrast with the dynamical heterogeneity observed in bulk water, where these domains average out spatially over time. We also find that the dynamical response of water to an interface depends critically on the nature of the interface and on the choice of interface definition. Overall these results reveal a richness in the dynamics of interfacial water that opens up the prospect of tuning the dynamical response of water through specific modifications of the interface structure or confining material.

Predicting heterogeneous ice nucleation with a data-driven approach.

Water in nature predominantly freezes with the help of foreign materials through a process known as heterogeneous ice nucleation. Although this effect was exploited more than seven decades ago in Vonnegut's pioneering cloud seeding experiments, it remains unclear what makes a material a good ice former. Here, we show through a machine learning analysis of nucleation simulations on a database of diverse model substrates that a set of physical descriptors for heterogeneous ice nucleation can be identified. Our results reveal that, beyond Vonnegut's connection with the lattice match to ice, three new microscopic factors help to predict the ice nucleating ability. These are: local ordering induced in liquid water, density reduction of liquid water near the surface and corrugation of the adsorption energy landscape felt by water. With this we take a step towards quantitative understanding of heterogeneous ice nucleation and the in silico design of materials to control ice formation.

What can reaction databases teach us about Buchwald-Hartwig cross-couplings?

Despite the widespread and increasing usage of Pd-catalyzed C-N cross couplings, finding good conditions for these reactions can be challenging. Practitioners mostly rely on few methodology studies or anecdotal experience. This is surprising, since the advent of data-driven experimentation and the large amount of knowledge in databases allow for data-driven insight. In this work, we address this by analyzing more than 62 000 Buchwald-Hartwig couplings gathered from CAS, Reaxys and the USPTO. Our meta-analysis of the reaction performance generates data-driven cheatsheets for reaction condition recommendation. It also provides an interactive tool to find rarer ligands with optimal performance regarding user-selected substrate properties. With this we give practitioners promising starting points. Furthermore, we study bias and diversity in the literature and summarize the current state of the reaction data, including its pitfalls. Hence, this work will also be useful for future data-driven developments such as the optimization of reaction conditions via machine learning.

Physisorption of Water on Graphene: Subchemical Accuracy from Many-Body Electronic Structure Methods.

Wet carbon interfaces are ubiquitous in the natural world and exhibit anomalous properties, which could be exploited by emerging technologies. However, progress is limited by lack of understanding at the molecular level. Remarkably, even for the most fundamental system (a single water molecule interacting with graphene), there is no consensus on the nature of the interaction. We tackle this by performing an extensive set of complementary state-of-the-art computer simulations on some of the world's largest supercomputers. From this effort a consensus on the water-graphene interaction strength has been obtained. Our results have significant impact for the physical understanding, as they indicate that the interaction is weaker than predicted previously. They also pave the way for more accurate and reliable studies of liquid water at carbon interfaces.

Pre-critical fluctuations and what they disclose about heterogeneous crystal nucleation.

Heterogeneous crystal nucleation is ubiquitous in nature and at the heart of many industrial applications. At the molecular scale, however, major gaps in understanding this phenomenon persist. Here we investigate through molecular dynamics simulations how the formation of precritical crystalline clusters is connected to the kinetics of nucleation. Considering heterogeneous water freezing as a prototypical scenario of practical relevance, we find that precritical fluctuations connote which crystalline polymorph will form. The emergence of metastable phases can thus be promoted by templating crystal faces characteristic of specific polymorphs. As a consequence, heterogeneous classical nucleation theory cannot describe our simulation results, because the different substrates lead to the formation of different ice polytypes. We discuss how the issue of polymorphism needs to be incorporated into analysis and comparison of heterogeneous and homogeneous nucleation. Our results will help to interpret and analyze the growing number of experiments and simulations dealing with crystal polymorph selection.