Crystal nucleation: Rare made common and captured by Raman

In PNAS, Urquidi et al. (1) combine techniques to induce nucleation in an optical trap (2), with Raman microspectroscopy (3) to monitor the steps before, during, and after crystal nucleation. Nucleation is the first step in the formation of a new material from less stable starting materials. For crystallization, the starting material is a supersaturated solute, that is, a solution concentrated beyond its solubility limit. Depending on the solute concentration, the thermodynamically allowed products may include multiple crystal structures (“polymorphs”) (4) each comprising the same solute molecule or compound. Starting with Ostwald (5), generations of scientists have attempted to understand and predict which material will emerge from nucleation and growth. The answer is complicated (6, 7). The process is somewhat like survival of the fittest in the materials context. A given phase can win in three ways: 1) by getting a head start at the nucleation stage; 2) by outracing the competition during growth; or 3) by being the most thermodynamically stable, if it survives to the ripening and/or coarsening endgame.

Among these three processes, nucleation is notoriously difficult to study because it is a rare event (8); that is, it occurs at random locations, occurs at random times, and yields a nanoscopic cluster that usually cannot be seen until it grows far beyond its size at inception. By enhancing the solute concentration at a focused laser spot, and monitoring that spot with Raman spectroscopy, the technique by Urquidi et al. (1) stands to provide unprecedented insight on nucleation mechanisms.

Until recently, ideas about crystal nucleation mechanisms were built around the classical nucleation theory (CNT) (9). According to the classical theory, nuclei are nanoscopic, equilibrium-shaped clusters with macroscopic crystal structure (10). The theory further assumes that nuclei form and redissolve via a series of monomer attachment and detachment events (11, 12), until a rare fluctuation generates a nucleus large enough for the chemical potential driving force to overcome surface forces (more generally, forces arising from the excess free energy). A few alternative nucleation pathways were recognized, but these involved external agents, for example, collision-induced attrition to generate postcritical fragments (13), ordering and enrichment at charged surfaces (14), or clustering around impurities like charged polymers in solution (15).

In 1997, with rare event simulation methods leading the way (16), it became clear that crystals could nucleate via local density fluctuations, long-lived droplets, and amorphous intermediates (17–20). State-of-the-art methods like atomic force microscopy (AFM) and in situ liquid phase transmission electron microscopy (LP-TEM) methods provide stunning footage of nucleation (19), including many examples of two-step nucleation (21) and other nonclassical nucleation pathways (22). Initial rumors about the death of CNT were sometimes exaggerated, but a more nuanced view has emerged in which the classical pathway is just one of many potential mechanisms, as shown in Fig. 1.

Fig. 1.

Menagerie of nucleation mechanisms. Acronyms and symbols in the diagrams are as follows: CNT, classical nucleation; Hom, homogeneous (in solution) (9); Het, heterogeneous (at an interface) (9); 2°, secondary (from one existing particle to many) (13); 2SN, two-step nucleation (17); PNC, prenucleation cluster (20); and MSC, magic-sized cluster (long-lived metastable intermediates) (23). Nonclassical typically refers to all but the secondary (2°) mechanisms and the mechanisms labeled CNT.

The challenge ahead is to understand which factors can help anticipate the dominant mechanism, and to develop predictive rate theories for each of the various nonclassical mechanisms. For mechanistic studies, AFM is limited to interfacial processes, and the 100-nm-thick LP-TEM cell is also quite different from the bulk solution. For example, the high surface to volume ratio, cluster adhesion effects, and reactions driven by electron irradiation (24) can potentially alter the dominant pathway.

In view of these limitations, the study of Urquidi et al. (1) is a particularly important advance. Their single-crystal nucleation spectroscopy approach enriches the solute concentration in a focused laser spot. The locally elevated concentration accelerates nucleation while the rate remains negligible in the surrounding solution. There are no interfaces and no obvious photochemical side reactions with the 532-nm (green) laser. And, most importantly, they can monitor the Raman signal emanating from the focus spot with a fast 46-ms time resolution.

Urquidi et al. (1) illustrate the technique by studying glycine nucleation from an aqueous solution. Glycine is the “fruit fly” of nucleation research, and, at their conditions, α-glycine is the expected product (25). Instead, they observe a series of transitions, from disordered glycine oligomers, to β-glycine, and, finally (consistent with expectations), to α-glycine. All of this happens in a span of about 1 s—so fast that the intermediates would likely be missed if Urquidi et al. had been waiting for a large crystallite while collecting data from the entire solution.

The results present an exciting opportunity for theorists. There are some questions about the mechanism of local enrichment, because the induced force on a molecule should be extremely small (26). However, if the mechanism of local enrichment can be understood, then the spatial intensity profile near the laser spot could be used to obtain a spatial concentration profile. Given nucleation rates and mechanisms from rare events simulations at each concentration, one could then build spatiotemporal survival probability models and make direct comparisons with experiment. Similar analyses have been done for other systems (27, 28), but without the advantage of a time-resolved sequence of events along the nucleation trajectory.

Urquidi et al. combine techniques to induce nucleation in an optical trap, with Raman microspectroscopy to monitor the steps before, during, and after crystal nucleation

Interestingly, the most stable glycine polymorph, γ-glycine, is a polar crystal with charged faces. The γ-glycine results when glycine is crystallized from an aqueous solution with at least 0.5 molal salt (NaCl) (29, 30). Simulations and Debye–Huckel theory show that the salt stabilizes γ-glycine nuclei, but not the nonpolar α-glycine structure (31). Thus far, these results stand as a satisfying confirmation that salt promotes the nucleation of polar crystals—a prediction derived from the classical picture, but perhaps also consistent with two-step mechanisms. Popper (32) wrote that the most convincing test of a theory comes from special cases where the theory should fail if it is not true. Salt should destabilize dimers and oligomers, making conditions less favorable for two-step nucleation and more favorable for a classical pathway in the experiments of Urquidi et al. (1). Thus, it would be fascinating to see whether a two-step nucleation pathway to γ-glycine still dominates with salt. If the nonclassical pathway still prevails, then the classical mechanism will have been refuted under exactly those conditions where it seems most likely to prevail. Then those early rumors about the classical picture’s demise would, indeed, seem prescient.