Microreactors with Inline IR Analysis for Reaction Screening

Transcription

1 Microreactors with Inline IR Analysis for Reaction Screening Jason Moore* Jerry Keybl Prof. Klavs Jensen Massachusetts Institute of Technology Department of Chemical Engineering 17 th International Process Development Conference May 17, 2010

2 Outline Background Homogeneous Gas-Liquid Catalysis Rhodium catalyzed hydroformylation Measure liquid phase concentrations of 2-phase flow using ReactIR flow cell to determine reaction kinetics Automated Optimization Optimization of a Paal Knorr reaction Use flow cell to monitor conversion and approach to steady state at each reaction condition 2

3 Microreactors in Brief Rapid mixing by diffusion across channels due to small width High heat transfer through silicon allows for precise temperature control Low volumes reduce materials consumption and allow for safer handling of hazardous reactions * Broader range of reaction conditions available than in batch Inlets Outlet 5 mm * Sahoo, H. PhD thesis, MIT, (2008). Kralj, J., Sahoo, H., and Jensen, K. Lab on a Chip 7, (2007). Mixer Channels: 100 mm Extraction Channels: 300 mm 3

4 ReactIR Real-time FT-IR measurements of mid-range IR spectrum using attentuated total reflectance (ATR) probe Able to extract concentrations of individual species from IR spectrum to determine profiles of transient species Wetted materials: Hastelloy body, silicon or diamond window, gold or teflon seal 2 μm penetration depth with several reflections Flow cell removes probe fiber alignment issues and has higher IR throughput 4

5 Published ReactIR Work Previous use of the ReactIR flow cell has been qualitative: Consumption of reactants and monitoring of reaction intermediates Monitoring chromatographic effects and dispersion in flow systems Reactor failure (PAT) 5 Carter, C.; Lange, H.; Ley, S.; Baxendale, I.; Wittkamp, B.; Goode, J.; Gaunt, N. Org. Process Res. Dev. 2010, 14,

6 Homogeneous Gas-Liquid Catalysis Jerry Keybl

7 Motivation Hydroformylation is an important commercial process for the manufacture of aldehydes and alcohols Olefins studied: 1-Hexene, 1-Heptene, 1-Octene, 1-Decene Some kinetics are available in literature few data have been published on the kinetics of hydroformylation and the data known are often contradictory. * 7 *Van Leeuwen et al, Rhodium Catalyzed Hydroformylation 2000

8 Partial Hydroformylation Mechanism 8

9 Catalysis System Design Sample Valves Gas Manifold ReactIR Microreactor 5 cm 9 Isco Pumps Backpressure Regulator (up to 30 bar)

10 Conversion (Product Basis) Initial Sampling Issue Collect multiple samples at same conditions Sample holdover apparent Carbon-basis mass balances have been poor by GC analysis (<50%) Sampling technique leads to large losses, particularly the volatile alkene to sample headspace Use of less volatile alkenes would minimize this effect Use internal standard of similar volatility Inline monitoring with ReactIR solves these problems First Sample Second Sample Temperature (K) 4 6

11 Hydroformylation of 1-Decene Online analysis: data looks good qualitatively Undecanal production increases with temperature 1-decene concentration decreases with temperature Lose resolution of 1-decene at higher conversions T ( C)= P = 25 bar Residence Time = 2.7 min Undecanal 1-Decene 11

12 Poor Mass Balance: Side Reaction Quantitatively, data makes little sense mass balance does not close Compare reaction solution over time Fresh solution contains much more primary alkene than does week old solution, as measured by alkene stretch New peak observed around 970 cm -1 -> associated with C-H stretch of transalkene Reactant isomerizes to 2-alkene at room temperature - Isomerization reaction with Rh catalyst is known in literature, but not always reported in hydroformylation experiments* 1-Alkene Rh Cat. 2-Alkene Rh Cat. 1-Alkene Rh Cat. 12 *Van Leeuwen et al, Organometallics 2001

13 Concentration (M) 1-Octene Hydroformylation Mass balance closes T = C P = 25 bar Residence Time = 2.7 min Octene Aldehyde 2-Octene Mass Balance Time (min)

14 Conversion (Aldehyde Basis) ln(k) Hydroformylation Kinetics Pressure: 25 bar Residence Time: 2.7 min Gas Ratio (CO:H 2 :N 2 ): 0.45:0.45:0.1 Flowrates (Gas:Octene:Catalyst μl/min): 20:12: E a = 28.1 kcal/mol y = x R² = Temperature ( o C) E E E E E-03 1/T 14

15 Automated Optimization of a Paal Knorr Reaction Jason Moore

16 Goals and Motivation Expand upon flow reaction monitoring techniques Quantitative IR analysis using ReactIR flow cell - Track the concentration of multiple species - Does not require dilution like with HPLC and UPLC UV/Vis analysis - Increases knowledge of reaction system Automated optimization and process modeling for improved efficiency Reduces labor required and increases reaction efficiency Minimal usage of reactants and solvents 16

17 Previous Optimization Work: Pd Catalyzed Heck Specified constraint on equilvalents T= 90 o C Approximately 1 experiment every 20 minutes Optimization procedure complete in 2 days Analysis done by HPLC 17 F 3 C Cl + O 3 equiv 1% Pd(OAc) 2 3% t Bu 2 MePhos 1.2 equiv Cy 2 NMe n-buoh 125 C 6 min McMullen, J.P.; Stone, M.S.; Buchwald, S.L.; Jensen, K.F.; (submitted.) F 3 C O

18 Absorbance Diamond Window Example: Paal Knorr Reaction * Ethanolamine 1: Hexanedione Product 2: 3: : Wavenumber (cm -1 ) 18 * Paal, C. Chem. Ber. 1884,17,2756. Knorr, L. Chem. Ber. 1884, 17, Taghavi-Moghadam, S.; Kleemann, A.; Golbig, K. G. Org. Process Res. Dev. 2001, 5, 652.

19 Concentration (M) ReactIR Calibration Calibrations performed on reaction species give linear trends with high R 2 and small slope confidence intervals: Ethanolamine: 9.58 ± 0.67 Hexanedione: 3.89 ± 0.14 Product: ± 0.28 R² = R² = R² = Peak Height Ethanolamine Hexanedione Product 1521 C=C Product waterfall plot Hexanedione Ethanolamine Product 19

20 -ln(1-x) -ln(1-x) Mechanism Ring-closure is the rate-limiting step Hemiaminal is generally favored over enamine as the intermediate Good candidate for optimization because reaction does not conform to simple kinetic model * First Order Fit R² = R² = R² = R² = Residence Time (min) Residence Time (min) 20 * Mothana, B. and Boyd, R. J. of Molecular Structure: THEOCHEM 2007,

21 Automation System Layout Syringe Pumps 4 cm 232 μl Microreactor ReactIR Temperature Controller IR Excel Auto Export 21 Legend Fluid Flow Data Flow Labview Control Matlab Optimization Algorithm

22 Conversion DoE Results Design 1 T ( o C) τ (min) X IR Design 2 T ( o C) τ (min) X IR Six residence times at each reaction setpoint, but varying approach to steady state High reproducibility of results Design 1 Design Sample Number 22

23 AUTOMATED STEPS Optimization Algorithm Enter starting conditions, parameters, termination Thermally equilibrate reaction ( T-T SP < e T ) Next experiment determined by optimization algorithm Flush system with minimum volume Steady state reached by IR analysis? No Termination criteria achieved? Yes Report optimum Yes Compute objective function 23

24 Gradient Optimization Algorithm Steepest descent algorithm Gradient determined from randomized DoE around initial conditions After DoE, move along gradient from centerpoint until objective function stops increasing X/τ 24

25 Gradient Optimization of Conversion Optimize conversion: J Conversion Start with randomized DoE to calculate gradient Move along gradient to next experiment within T and τ bounds Terminate when conversion decreases or design space corner is met Optimization terminates with both constraints active X 25

26 Gradient Optimization of Production X/τ Optimization of conversion divided by residence time, which is proportional to production rate: J / Conversion Residence Time 26

27 Temperature ( o C) Conjugate Gradient Optimization Algorithm Begins with steepest descent algorithm After first trajectory is complete, the direction of later trajectories considers the previous search direction 130 S 1 = J(x 0 ) 120 S k+1 = J(x k+1 ) + b k S k * b k 2 J( xk 1) 2 J( x ) Fletcher-Reeves k Initial Steepest Descent Steepest Descent Conjugate Gradient Residence Time (min) 27 * Beers, K. Numerical Methods of Chemical Engineering 2007.

28 Conjugate Gradient Optimization of Production X/τ Optimization of conversion divided by residence time, which is proportional to production rate: J Conversion / Residence Time 28

29 ln(k) Conversion Conclusions Homogeneous Gas-Liquid Catalysis Initial sampling issues solved ReactIR allowed for identification of side reaction Octene hydroformylation kinetics were determined Automated Optimization Data precision ensured by monitoring approach to steady state Optimizations of conversion and production carried out y = x R² = E E E-03 1/T Time (min)

30 Acknowledgements Prof. Klavs Jensen Jerry Keybl Jonathan McMullen KFJ Group 30

31 31

32 Backup Slides 32

33 Conversion Gradient Optimization of Conversion Time (min) Move to next experiment after the difference in moving averages on conversion is below a threshold and at least 4 residence times have passed 33

34 Kinetics First Order First-order kinetics match experimental data significantly better than second-order kinetics Induction period while initial reaction approaches equilibrium Second Order 34

35 Automation System Layout Syringe Pumps 4 cm Microreactor ReactIR Computer 35 Temperature Controller Legend Fluid Flow Data Flow Syringe Pumps Matlab Optimization Algorithm Labview Control IR Excel Auto Export Temperature Control

36 Automation System Automation System Mobile IR with Flow Cell 36

37 Published ReactIR Work Previous use of the ReactIR flow cell has been qualitative: Monitoring chromatographic effects and dispersion in flow systems Consumption of reactants Monitoring of reaction intermediates Reactor failure 37 Carter, C.; Lange, H.; Ley, S.; Baxendale, I.; Wittkamp, B.; Goode, J.; Gaunt, N. Org. Process Res. Dev. 2010, 14,