(328f) Estimating the Three Characteristic Lengths of Plate-like Particles in Suspension

The size and shape of crystals are relevant properties of a powder, as they determine its rheology. In particular, ensuring a good filterability, flowability, and compressibility is crucial in the pharmaceutical industry, aiming for fast processing from crystallization to the final tablet manufacturing.

State-of-the-art characterization techniques, such as laser diffraction, focused beam reflectance measurement, or Coulter counter, typically rely on the assumption of sphericity, and therefore characterize the particles with a single size descriptor. Due to their lower symmetry, needle and plate-like crystals require two and three generic length descriptors, respectively, for an accurate size and shape characterization.

For the case of needle-like crystals, both mono and dual imaging systems have been successfully used to obtain the particle size and shape distribution[1,2]. Thanks to the larger amount of information gathered, dual imaging offers both improved accuracy and reliability. In addition, it is less sensitive to issues arising due to particle orientation with respect to the cameras. With the Dual Imaging System for Crystallization Observation (DISCO)[2], a flow-through device developed at ETH Zurich, we have been able to measure the particle size and shape distribution to a degree of accuracy such that the estimation of the suspended solid mass is made possible[3]. Further, the device has been successfully employed to estimate kinetics of fundamental phenomena and to effectively manipulate the size and shape of needle-like crystals[4,5].

For the case of plate-like crystals, to the best of our knowledge, no commercial device is able to accurately estimate their three characteristic lengths. This unavailability severely restricts designing processes to manipulate the size and shape of particulate populations with a plate-like morphology, even though they are frequently encountered in pharmaceutical processes. Recently, a neural network-based machine learning model built on the stereoscopic information provided by the DISCO has been proposed[6], thus laying the foundations for successful handling of particulate systems that require a 3D characterization.

In this contribution, we validate the aforementioned machine learning model and demonstrate that the DISCO can characterize plate-like particles with remarkable accuracy in terms of their three characteristic lengths, i.e., their length, width, and thickness.

Since no analytical standards are available for plate-like particles of different sizes and shapes, we produced populations of custom-designed “LithoPlatelets”. The LithoPlatelets were fabricated by SU-8 photolithography. The SU-8 photoresist was spin-coated onto a silicon wafer to achieve the desired layer thickness, which determines the thickness of the platelets (in the order of tenths of micrometers). The structures were exposed through a photomask, and gently scraped off the wafer surface once fully polymerized[7]. Finally, to be measured with the DISCO, these reference platelet populations were suspended in mixtures of water and propan-2-ol. The measurement process is outlined in Figure 1.

Thanks to the flexibility of the fabrication method, several monodisperse population with a range of sizes and aspect ratios have been successfully characterized both individually and combined, thus proving the potential of the characterization technique when applied to real, polydisperse crystalline systems. We believe that this work therefore addresses two important challenges in the particle technology community. First, to exploit a reliable methodology to produce monodisperse populations of particles with nonstandard dimensions, which can be used as analytical references. Second, to provide a characterization platform featuring a robust, experimentally validated machine learning model that is able to accurately estimate the three characteristic lengths of plate-like particles.

[1] Cardona, J.; Ferreira, C.; McGinty, J.; Hamilton, A.; Agimelen, O. S.; Cleary, A.; Atkinson, R.; Michie, C.; Marshall, S.; Chen, Y.-C.; Sefcik, J.; Andonovic, I.; Tachtatzis, C. Image Analysis Framework with Focus Evaluation for in Situ Characterisation of Particle Size and Shape Attributes. Chem. Eng. Sci. 2018, 191, 208–231.

[2] Rajagopalan, A. K.; Schneeberger, J.; Salvatori, F.; Bötschi, S.; Ochsenbein, D. R.; Oswald, M. R.; Pollefeys, M.; Mazzotti, M. A Comprehensive Shape Analysis Pipeline for Stereoscopic Measurements of Particulate Populations in Suspension. Powder Technol. 2017, 321, 479–493.

[3] Bötschi, S.; Rajagopalan, A. K.; Morari, M.; Mazzotti, M. An Alternative Approach to Estimate Solute Concentration: Exploiting the Information Embedded in the Solid Phase. J. Phys. Chem. Lett. 2018, 9 (15), 4210–4214.

[4] Salvatori, F.; Binel, P.; Mazzotti, M. Efficient Assessment of Combined Crystallization, Milling, and Dissolution Cycles for Crystal Size and Shape Manipulation. Chem. Eng. Sci. X 2019, 1, 100004.

[5] Bötschi, S.; Rajagopalan, A. K.; Rombaut, I.; Morari, M.; Mazzotti, M. From Needle-like toward Equant Particles: A Controlled Crystal Shape Engineering Pathway. Comput. Chem. Eng. 2019, 131, 106581.

[6] Jaeggi, A.; Rajagopalan, A. K.; Morari, M.; Mazzotti, M. Characterizing Ensembles of Platelike Particles via Machine Learning. Ind. Eng. Chem. Res. 2021, 60 (1), 473–483.

[7] Wyatt Shields Iv, C.; Zhu, S.; Yang, Y.; Bharti, B.; Liu, J.; Yellen, B. B.; Velev, O. D.; López, G. P. Field-Directed Assembly of Patchy Anisotropic Microparticles with Defined Shape. Soft Matter 2013, 9 (38), 9219–9229.