After years of speculation on the origins of symmetry-making and -breaking during crystallization, time-resolved in situ scanning probe microscopy and all-atom molecular dynamics simulations have shown that the formation of olanzapine crystals largely occurs by the incorporation of centrosymmetric dimers into growth sites.
Classical models of crystallization would have it that crystals nucleate and grow monomer by monomer; however, evidence is mounting for roles that other, more complex mechanisms may play1. Indeed, non-classical models, based on multi-step mechanisms and involving different intermediate phases (pre-nucleation clusters, liquid-like precursors, amorphous phases, oligomers), have been proposed to explain experimental observations surrounding crystal nucleation and growth. Understanding how crystals form is useful for a host of materials science and technology applications.
In the absence of direct atomic-scale observations, crystal growth mechanisms have historically been speculated and inferred from crystal structures and solute assemblies in solution. An intriguing aspect is that the symmetry elements of a crystal are not correlated to those of its constituent molecules. Now, after years of debate over the origins of inversion symmetry in olanzapine (OZPN) crystal structures, a team led by Peter Vekilov, Alastair Florence and Jeremy Palmer describe in Nature Chemistry how OZPN crystals can grow by incorporating centrosymmetric dimers, pre-formed in solution, that accumulate at the surface of the growing crystal2.
OZPN, an atypical antipsychotic agent originally marketed as Zyprexa for the treatment of bipolar disorder and schizophrenia3, is known to crystallize into more than 60 forms — all but one featuring a dimer that is mainly stabilized by dispersion forces (Fig. 1). The dimer features so prominently in the observed crystal structures that it has been designated the ‘crystal building block’ of OZPN. However, a recent crystal structure prediction (CSP) study found that most of the lattice energy minima of OZPN, a few of which are competitive in energy with the observed forms, are not based on dimers at all4,5. In fact only recently, after many years of polymorph screening and several failed attempts to break the dimer in crystallization, was the sole non-dimer-based polymorph found by crystallizing OZPN from a polymer dispersion designed to disrupt the dimers6. Clearly, without intervening measures, the crystallization of OZPN appears to be strongly biased to dimer-based crystal forms, prompting much speculation about the role of the dimer during crystal nucleation and growth.
The researchers carefully studied the formation of OZPN ethanol–water mixed solvate (OZPN·½H2O·EtOH; CSD refcode: WEXQEW) crystals at the nanoscale, monitoring the step growth along screw dislocations on the dominant (002) face of single crystals (Fig. 1). Unexpectedly, rather than the linear correlation assumed by classical crystal growth theory for monomer incorporation, they found a superlinear dependence of step velocity on solute concentration, ν(C). By combining time-resolved in situ atomic force microscopy with molecular dynamics (MD) simulations, they were able to provide molecular-level understanding of the self-assembly processes leading to the non-classical growth of OZPN crystals. Most notably, they submit that the second-order kinetics of crystal growth — characterized by a quadratic relationship between the step velocity and solute concentration — is a general criterion for identifying growth by incorporation of dimeric units.
Among the possible sources of crystal growth acceleration at high supersaturations are mesoscopic clusters7 — those account for 10–7 to 10–5 of the dissolved OZPN in EtOH–water solutions. Solute-rich clusters have previously been shown to facilitate hydrate formation on the surface of crystals of a neat (non-solvated) polymorph of OZPN (form I) by delivering high OZPN concentrations to the steps to accelerate growth8. Nonetheless, while the prevalence of the centrosymmetric dimer in OZPN crystal structures may suggest cluster formation is favourable, the researchers showed that the contribution of such clusters to the crystal growth of OZPN·½H2O·EtOH was minimal. In fact, when clusters were deposited on the surfaces of growing crystals, they persisted there for hours, retaining their structure and size, independent of the OZPN concentration. Additionally, their removal from the surface (by filtration) has no impact on the step velocities — these two findings are consistent with the clusters not contributing to crystal growth.
Other factors can also be envisaged to account for growth acceleration at high supersaturation. Among these, an increasing density of kinks, which are the only sites where solute molecules incorporate in crystals, and step pinning caused by adsorbed impurities interfering with step propagation, were ruled out. Instead, the researchers propose that crystal growth occurs predominantly by incorporation of dimers, formed in solution in a pool of monomers. Using Raman spectroscopy to measure the relative populations of monomers and dimers in EtOH–water solutions, a non-linear shift was observed in the dimerization equilibrium, from exclusively monomers toward dimers as the concentration increased — showing that the population of dimers increases exponentially. Further support for the formation of dimers in solution came from all-atom classical MD simulations, which showed that two OZPN monomers in close proximity in solution fluctuate around the structure of the dispersion-bound dimer. Once formed in solution, the dimers strongly adsorb to the terraces, concentrating on the surface, where they are able to diffuse to the high-energy kink sites of the growing crystal. Thus, even though the dimers are the minor components in a solution that predominantly consists of monomers, an overwhelming accumulation of adsorbed dimers, as well as the dimerization of adsorbed monomers at the surface, would account for the observed superlinear dependence of ν(C).
This study, in establishing a dimer-based growth mechanism for OZPN, demonstrates how time-resolved atomic-level monitoring of crystal growth, along with emerging all-atom MD simulation techniques, can provide unprecedented insights into non-classical crystallization mechanisms. More generally, the non-linear relationship between growth velocity and solute concentration provides a new way to establish the growth units in crystallization. Such fundamental understanding of crystal growth is essential for crystallization process models that will ultimately enable exquisite process control for materials synthesis.
De Yoreo, J. J. et al. Science 349, aaa6760 (2015).
Warzecha, M. et al. Nat. Chem. https://doi.org/10.1038/s41557-020-0542-0 (2020).
Fulton, B. & Goa, K. Drugs 53, 281–298 (1997).
Bhardwaj, R. M. et al. Cryst. Growth Des. 13, 1602–1617 (2013).
LeBlanc, L. M. & Johnson, E. R. CrystEngComm 21, 5995–6009 (2019).
Askin, S. et al. Cryst. Growth Des. 19, 2751–2757 (2019).
Warzecha, M., Safari, M. S., Florence, A. J. & Vekilov, P. G. Cryst. Growth Des. 17, 6668–6676 (2017).
Warzecha, M. et al. Cryst. Growth Des. 17, 6382–6393 (2017).
The author declares no competing interests.
About this article
Cite this article
Reutzel-Edens, S.M. Symmetry in the making. Nat. Chem. 12, 887–888 (2020). https://doi.org/10.1038/s41557-020-0547-8