This article is in Chemistry in Australia magazine: Issue December 2024
Author: Dennis Power
Curtin researchers have drastically reduced the time required to predict crystal formation in a vast selection of solvents, potentially removing the need for “trial and error” experimentation. Dr Peter Spackman, of Professor Julian Gale’s computational chemistry team, developed the innovative CrystalClear protocol, which rapidly generates intermolecular interaction energies, giving key insights into how molecular crystals grow. The protocol has proven its worth to the pharmaceutical industry with a successful ibuprofen case study.
Spackman has created software called Open Computational Chemistry, which is freely available on GitHub and is capable of performing a range of quantum chemistry calculations, including predicting crystal growth in the CrystalClear protocol. The computational method is powerful and simple in its design, only requiring two inputs to predict the associated free energies: the molecular crystal structure and type of solvent used. These energies can then be used in further simulations, such as the Monte Carlo simulator CrystalGrower, to simulate nanoscale surface topography and crystal habit.
“One of my main goals is that the methods I write are usable by people who are not experts to make predictions, hopefully somewhat interactively, in the space of an hour or two – free, available and quick on a standard laptop,” said Spackman.
Several case studies have been undertaken to test CrystalClear and highlight its effectiveness. Ibuprofen was a logical choice for a pharmaceutical case study owing to its well-established crystal habit.
The solvent plays an essential role in the way a crystal grows; the rate of growth will be determined from how strongly, and where, the molecule interacts with the solvent. A functional group that has strong affinity for the solvent will grow at a slower rate because the desolvation process requires greater energy, and vice versa.
Acetonitrile is a key solvent in the pharmaceutical industry and has a large dielectric constant, resulting in strong intermolecular attraction. In the case of ibuprofen, this leads to needle-shaped crystals as the carboxylic acid moiety strongly interacts with the solvent, slowing crystal growth in this direction. The crystal morphologies of ibuprofen predicted by CrystalClear show reasonable accuracy to the experimental shapes (Fig. 1).

Figure 1. Predicted crystal forms (top) and experimental crystal forms (bottom) for ibuprofen.
“An individual chemist may not care if they get the wrong crystal shape, but in a pharmaceutical pipeline or an industrial context that’s really critical,” said Spackman.
Crystals that grow with certain shapes, such as needles, can cause issues throughout the manufacturing process so it makes sense to try to avoid this. Normally chemists would use a “trial and error” experimental approach, but this can take a considerable amount of time. As Spackman put it, “We can do better than educated guesswork”.
The computational method makes many coarse approximations in order to make its predictions, trading accuracy for computational speed. When asked why this particular approach was taken, Spackman commented, “You need to have insights faster than the experiment can be done. Until we can simulate faster than reality happens, we’re never going to get there.”
There are other ways to make these energy predictions, but they can take upwards of several years to calculate accurately. Although informative, this is completely impractical for timely decision making.
Crystallisation is an essential process in the purification of chemicals. The reverse process, dissolution, is equally significant, particularly in the pharmaceutical sector. CrystalClear can quickly generate predictions for both. Although calculating solubility was not an aim of Spackman’s research, it was an incidental outcome.
State-of-the-art methods for solubility prediction are within 1.5 orders of magnitude on an absolute scale. By comparison, CrystalClear successfully predicts trends in solubilities within two orders of magnitude.
“Considering the simplicity of the presented approach … these results are surprisingly good,” he said.
Spackman is continuing to improve CrystalClear, turning his attention to new frontiers in molecular crystal growth prediction, and is looking at expanding its applications to more complex systems: minerals, cocrystals and the challenging world of metal–organic framework structures. When asked what’s next, Spackman pointed to CrystalClear’s already global usage and hopes for its ongoing success.
“It is our vision that automated, rapid protocols for the prediction of crystal growth, such as the present method, will soon be readily integrated into laboratory settings for targeted design and prediction of crystal forms and habits,” he said.
Dennis Power
#Free
#ChemistryinAustralia