(368ad) Vibrational Spectroscopies As Process Analytical Technologies (PATs) for Continuous Manufacturing Monitoring and Chemical Kinetic Determination

In-situ/operando spectroscopy of liquids (Raman, IR, NMR) for qualitative and quantitative analysis, chemometric model development, reaction engineering and kinetic determination, pharmaceutical manufacturing, consumer product research

Spectroscopy is a powerful tool for understanding the world around us. It can be used for qualitative means, interpreting the intermolecular interactions between solute and solvent or exploring the impact of a process. Spectroscopy can also be used for quantitative means, such as determining kinetic rate parameters and for determining concentrations for process control.

At Rutgers, I have worked on four major projects using spectroscopy in chemical engineering applications. The first is exploring intermolecular interactions between biomass relevant chemicals. The second is model development of Critical Quality Attributes (CQAs) for bioreactors. The third is the rate determination of kinetic information for diphenhydramine synthesis. And finally, my most recent project is implementing PAT into a continuous pharmaceutical project.

In biomass-solvent interactions, we used ATR-FTIR spectroscopy to analyze the interaction between hydroxymethylfurfural (HMF) in dimethylsulfoxide (DMSO)-water mixtures. As the individual changes between peaks are difficult to see, we utilized two-dimensional correlation spectroscopy (2D-COS). Through this work, we were able to visualize the interactions between the aldehyde of HMF and the sulfoxide of DMSO and see the preference for interaction between HMF against the DMSO dimers and monomers. This, in turn, is believed to help stabilize HMF in DMSO, resulting in higher yields compared to solvent such as water which do not appear to have any form of selective interactions. This information may be used in selecting or designing solvent systems for higher HMF yields without the processing penalties of DMSO.

In the next project, Raman spectra were collected in bioreactors with the intention of developing models to track CQAs such as glucose content, ammonia, lactate, viable cell count, and others. While models can be developed, the importance of preprocessing has been found vital for model performance. To this end, I developed Porchlight (Konkol and Tsilomelekis, 2023). This is a GUI Python application intended for the educational audience, but is also used in my own research, designed to help teach the impact of preprocessing parameters. It allows students to import spectroscopic data and try a variety of standard and less common preprocessing techniques. A graph adjacent to the preprocessing selector plots the data with appropriate labels to help get an intuition for what the preprocessing steps did to the spectra. From there, it can be exported for further analysis. Of course, for researchers and professionals, Porchlight is accessible for scripts for simple slotting into scripts and notebooks.

At Colgate-Palmolive, I participated in the Experiential Learning Opportunity Program fellowship with Colgate-Palmolive, I worked on using Raman spectroscopy to explore treatment of skin, acid etching of dentin, and used spectroscopy to help explain the performance of delivery systems. One of these projects were published in a paper I coauthored (Doss et al., 2023).

In my most recently published work, we determined the kinetics of diphenhydramine synthesis. The work is challenging because of the various extremes present here. In the first step, concentrated hydrochloric acid is used to chlorinate benzhydrol in a biphasic reaction that readily proceeds at room temperature in a large continuous stirred tank batch reactor. All you need is a mix. On the other hand, the second step is performed at high temperatures in a microfluidic reactor with very small volumes. For this work, we used Raman spectroscopy in the batch tank reactor to get qualitative profiles for the generation/consumption of the reactants and products. In both cases, a low-field flow NMR provided a quantitative concentration profiles. With the NMR as the reference, the Raman profiles can be scaled accordingly, giving a far greater density of data compared to the periodic NMR sample.

My current project is on the implementation of spectroscopy and other forms of sensors into a continuous pharmaceutical manufacturing process for injectable formulations. In this work, I have implemented a UV-Vis spectrophotometer alongside flow cells and a multiplexer to control the light path. A single UV-vis spectrometer is used to collect data at two different locations at the line. A bespoke application written in Python monitors when a spectrum is collected, and then communicates with a microcontroller which switches the light path to the next position. With the flow meters, pressure sensors, pH, and conductivity sensors I implemented, all aspects of the line can be monitored and collected for process control. Using the sensor data from the UV-Vis, pH and conductivity probes, a machine learning model has been built to predict the concentration of acid, salt, and drug content. The microcontroller work has also been spun out into a low-cost laboratory project for measuring residence time distributions in chemical engineering laboratory courses.

My work as a PhD student has prepared me by having a comprehensive understanding on how to use sensors and especially spectroscopy to solve challenges in chemical engineering. I have implemented spectroscopic tools in hardware and software, developed applications for the processing and analysis of data, and used spectra to analyze real solutions for monitoring processes and calculating kinetic information. I would like to continue leveraging these many skills in a professional capacity in an R&D role in a field such as in pharmaceuticals or consumer products.