In Situ Measurement of Solution Concentration during the Batch Cooling Crystallization of l-Glutamic Acid using ATR-FTIR Spectroscopy Coupled with Chemometrics

Abstract

The in situ measurement of solution supersaturation associated with the batch cooling crystallization of l-glutamic acid (LGA) at 500 mL and 20 L scale sizes is assessed via ATR-FTIR spectroscopy. A partial least squares chemometric calibration model was developed for the online prediction of LGA concentration from measured FTIR absorbance spectra overcoming some significant challenges related to the low sensitivity of LGA in the mid-IR frequency range, its low solubility in water, and its complex speciation chemistry. The solubility data of LGA in water over the temperature range from 40 to 80 °C, using ATR-FTIR, reveals excellent agreement with those obtained both from using a gravimetric method and literature data. The metastable zone width determined using the turbidimetric methods as a function of heating/cooling rates and solute concentration is found to increase with increasing cooling rate while it decreases with increasing solution concentration. Monitoring online crystallization via both spontaneous and seeded in 500 mL and 20 L crystallizers reveals good concentration predictions for seeded crystallization, while fouling of the ATR crystal prevents its routine use for unseeded crystallization studies. Higher supersaturation levels are found for the larger crystallizer scale-size consistent with enhancement of secondary nucleation at the smaller scale-size.

Problems, potentials and future of industrial crystallization

“…Routinely used techniques are focused beam reflectance measurements (FBRM) and optical reflection monitoring (ORM) in order to analyze the evolving crystal size distribution in situ (particle engineering) [30] 1) , e.g., in-line spectroscopic techniques primarily attenuated total reflectance Fourier transformation infrared (ATR-FTIR) and ATR ultraviolet-visible (UV-vis) spectroscopy. The latter can be applied for solution concentration measurements [31] and to monitor polymorph and solvate conversations in real-time [28,32,33]. Furthermore, in situ imaging techniques for the direct observation of the appearance and growth of crystals such as PVM (particle vision monitor) were developed recently [34].…”

Surface Adsorbed Templates for Directing the Crystal Growth of Entacapone as Monitored Using Process Analytical Techniques

“…Presumably at the lower solute concentrations the propensity for solute-solute interactions in the solution decreases, hence lowering the probability of crystallisation. Similarly, the increase in MSZW with increasing cooling rate is typical of the behaviour of materials which do not spontaneously nucleate easily 27,28 . In such cases, the rate of production of available solute from the decrease in solubility due to temperature lowering is not matched by the nucleation rate resulting in a lowering of the crystallisation on-set.…”

Comparative study on ATR-FTIR calibration models for monitoring solution concentration in cooling crystallization

“…In fact, experiments for ATR-FTIR calibration had been mainly carried out in the undersaturated zone (USZ) [9][10][11][12][13][14][15]. A small number of calibration methods were presented in the literature to analyze ATR-FTIR spectra for in-situ measurement of the solution concentration.…”

“…Liotta et al [10] adopted an amount of ATR-FTIR spectra collected in USZ to build a partial least-squares (PLS) calibration model for predicting the concentration of an active pharmaceutical ingredient (API) during crystallization, resulting in the averaged relative prediction error about 2.9%. For monitoring the cooling crystallization process of L-glutamic acid (LGA), a PLS calibration model of the solution -3-concentration was built based on the ATR-FTIR spectra collected in USZ [11], giving the averaged relative prediction error about 6.9%. Chen et al [12] developed an extended loading space standardization method to enhance the model prediction based on the above type spectra for LGA solution, which reduced the relative prediction error to 0.7%~2.5%.…”