The integration of NeSSI with Continuous Flow Reactors and PAT for Process Optimization Michael F. Roberto, Thomas I. Dearing and Brian J. Marquardt U.S. Food and Drug Administration 1
FDA Sponsored CFR Project Goal: to improve reaction development and optimization through the use of continuous glass flow reactors, NeSSI and process analytics Funded by the FDA to demonstrate the benefits of improved reactor design, effective sampling and online analytics to increase process understanding and control Demonstration of Quality by Design for continuous processing - QbD Partners: FDA, CPAC, Parker, Corning, Kaiser Optical Systems Initial feasibility phase of the project began Nov. 2008 First full phase implementation began Jan. 2010 Currently finalizing Phase III work by Dec. 2012 2
Continuous Flow Reactors Flow cells optimized for continuous production Efficient mixing schemes Rapid heat transfer Alternative to large-scale batch chemistry. More energy efficient Require less solvent Easy to optimize and control Assure quality product thru control 3
Limitations of CFR Production Quantitative determination of material from continuous flow reactors currently uses batch techniques Requires continuous sampling to monitor process Resource intensive materials, equipment, people, time Validation using chromatography lengthens analysis cycle Delay between formation and quality measurement Alternative has to be fast, accurate, integrated Our goal: demonstrate the ability to assess and control the quality of product in-line, using PAT, design of experiment, and statistical models 4
Analytical Control System Design Optimization Feedback & Control Process Modeling Chemistry Analytics and Data Handling Hardware Control and Monitoring (Software) Hardware NeSSI, Raman, Other PAT, Reactor 5
System Control and Analytics Reagent 1 HPLC Pump (Flow Rate) Back Pressure Regulator Flow Meter Pressure Gauge Thermocouple Raman Other PAT Control Analytics HPLC Pump (Flow Rate) Back Pressure Regulator Flow Meter Pressure Gauge Thermocouple Raman Other PAT Reagent 2 Temperature Control Flow Meter Pressure Gauge Thermocouple Raman Other PAT Needle Valve Product 6
What is NeSSI? Industry-driven effort to define and promote a new standardized alternative to sample conditioning systems for analyzers and sensors Standard fluidic interface for modular surface-mount components ISA SP76 Standard wiring and communications interfaces Standard platform for micro analytics 7
What does NeSSI Provide Simple Lego-like assembly Easy to re-configure No special tools or skills required Standardized flow components Mix-and-match compatibility between vendors Growing list of components Standardized electrical and communication Plug-and-play integration of multiple devices Simplified interface for programmatic I/O and control Advanced analytics Micro-analyzers Integrated analysis or smart systems Provides platform for coupling analytics to flowing systems 8
NeSSI Raman Ballprobe Ballprobe Specs. Hastelloy C-276 Ti, SS, Monel Sapphire optic Std. temp range: -40 350 C Pressure: 0-350 Bar Matrix Solutions: www.ballprobe.com 9
Kaiser Multi-Channel Raman 10
Corning Advanced-Flow TM LF Low-Flow Capability (1-10 ml/min) Corning has introduced a reduced flow-rate reactor that retains the outstanding mixing and heat exchange performance of its Advance-Flow TM glass reactors while providing: Low internal volume (2 ml flow) High flexibility Metal-free reaction path Scalability Compatibility with analytics T, Flow and Pumping control 11
Integrated Reactor System With Control Units 2.5 ft 12
Software Development Software developed in-house to monitor critical process parameters Flow, Temp, Pressure Also controls reactor Flow rates Heating control Developed in LabVIEW 13
Analytical Control System Design Optimization Feedback & Control Process Modeling Chemistry Analytics and Data Handling Hardware Control and Monitoring (Software) Hardware NeSSI, Raman, Other PAT, Reactor 14
Experimental Objectives Demonstrate ability to monitor reaction using online process analytics Determine optimal technology for the analysis of this chemistry Find limitations of the systems with respect to the chemistry being performed Establish boundaries for Design of Experiments Determine largest independent contributors towards reaction progress Create a Calibration curve to correlate off-line HPLC results with Raman spectroscopy 15
Esterification of Benzoic Acid Straight-forward, pharmaceutically relevant Strong dependence on temperature, flow rates 16 Wiles, C., Watts, P. 2009
Raman Signal (A.U.) Raman Signal (A.U.) Reaction Monitoring with Raman Background-corrected spectrum cropped to region of interest Multivariate statistics combined with x 10 4 5 Raman spectra 4.5 4 used to monitor 3.5 3 reaction 2.5 2 progress 1.5 1 0.5 12000 10000 8000 6000 4000 2000 0 Product Reagent 720 740 760 780 800 820 840 860 Raman Shift (cm^-^1) 0 400 600 800 1000 1200 1400 1600 1800 Raman Shift (cm^-^1) 17
Reaction Procedure and Setup Chemistry performed in reactor On-line Raman monitoring on end of product line Each variable independently controlled in DoE to enable understanding of each critical process parameter Computer control of flow and temperature Catalyst concentration controlled manually 18
Temperature (C) Raman PC1 Scores (AU) Temperature vs. Reaction Progress Clear correlation between temperature and product formation Reaction progress not linearly correlated with temperature Expected for this reaction 160 150 140 130 120 110 100 90 80 70 60 Temperature (C) Raman PC1 Scores 60 C 100 C 140 C 0 20 40 60 80 100 120 140 160 Time (min) 19
Temperature vs. Reaction Progress 20
RAMAN PLS MODEL 21
Online Raman Analysis Establish Raman spectroscopy as a quantitative validation alternative to HPLC for the esterification of benzoic acid Build a calibration model between Raman scores and HPLC area ratio information Model constructed in MATLAB Validate model with live synthesis in reactor 22
Construction of Calibration Model Calibration standards prepared offline 0-100%, 10% intervals 20 ml of both catalyst and standard solutions 3 aliquots taken for HPLC Standards injected into NeSSI product line of reactor system Built separate calibration injection area with NeSSI pieces bypassed backpressure with relief valve Reactor plates not part of calibration standards Would increase product yield 23
Raman Signal (A.U.) Raman Spectra: Calibration Model x 10 4 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 400 600 800 1000 1200 1400 1600 1800 Raman Shift (cm^-^1) Strong activity in the fingerprint region Activity in loadings ignores large methanol peak in center 24
Construction of Calibration Model PLS Model was constructed from baselinecorrected Raman spectrum Corrected full spectrum used to ensure ability to detect process upsets or outliers HPLC peak area ratios used for validation Preprocessing: SNV then Mean Centered 2 Latent Variables used in PLS model Captures 99.8% of variance in Raman spectra 25
Yield Predicted by Raman (%) PLS Raman Calibration Model 100 90 80 PLS Model - Raman Prediction vs. HPLC High Concentration Factorial Calibration R^2 = 0.999 2 Latent Variables RMSEC = 0.91141 Bias = 0 70 60 50 40 30 20 10 0 0 20 40 60 80 100 Yield Measured by HPLC (%) 26
Validation Experimental Chemistry performed in reactor Raman measurements performed on-line at room temperature, in product stream Pressure of ~6 bar maintained liquid phase 5 aliquots taken from product stream Off-line validation was performed using HPLC Prepared same manner as calibration standards Data was validated in PLS Calibration model 27
Raman Signal (A.U.) Raman Spectra: Validation Model x 10 4 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 400 600 800 1000 1200 1400 1600 1800 Raman Shift (cm -1 ) Same activity region as in calibration spectra 28
Yield Predicted by Raman (%) PLS Raman Validation Model 100 90 80 70 60 50 PLS Model - Raman Yield Prediction vs. HPLC High Catalyst Concentration R 2 = 0.999 2 Latent Variables RMSEC = 0.91141 RMSEP = 2.1617 Calibration Bias = 0 Prediction Bias = 0 High Temperature (T), Long residence time (R t) High T, Short R t 40 30 Y Predicted 1 20 Calibration Validation Test 10 Low T, Long R t 1:1 0 Low T, Short R t fit 0 20 40 60 80 100 Yield Measured by HPLC (%) 29 J. Pharm. Innov. 7 (2), 69-75.
Conclusions Raman spectroscopy is a viable quantitative, on-line monitoring method for the esterification of benzoic acid Successful off-line calibration Accurate and rapid determination of product conversion Used HPLC to create calibration curve NeSSI allowed for interface of analytics, PAT, and reactors for a robust continuous flow system 30
Acknowledgements FDA, ORISE Center for Process Analysis and Control (CPAC) Corning Parker Kaiser Optical Systems Mettler-Toledo Medicinal Chemistry, UW 31