for: Flow Grignard and Lithiation: Screening Tools and Development of Continuous Processes for a Benzyl Alcohol Starting Material Michael E. Kopach, *, Kevin P. Cole, Patrick M. Pollock, Martin D. Johnson, Timothy M. Braden, Luke P. Webster, Jennifer McClary Groh, Adam D. McFarland, John P. Schafer, # Jonathan J. Adler # and Morgan Rosemeyer # Small Molecule Design and Development, Eli Lilly and Company, Indianapolis, Indiana 46285, USA # D&M Continuous Solutions, LLC, Greenwood, IN 46143, USA Contents Continuous reactors for cryogenic lithiation... S2 Lithiation Reactor Systems 2 and 3. PFRs in series.... S2 Lithiation Reactor Systems 1 and 4. Fill-Empty Stirred tanks in series.... S5 Temperature and pressure trends for continuous lithiation run... S12 Formylation and quench temperature trends... S15 Grignard formylation and reduction impurity trends...s17 S1
Continuous Reactors for cryogenic lithiation: Lithiation Reactor Systems 2 and 3. PFRs in series. These two reactor designs use plug flow reactors (PFRs) in series with static mixers, which are common for cryogenic lithiation, coupling, quench reactions. The innovative features of these compared to standard published equipment sets are the feed coil for BuLi pumping into the reactor, and the expansion chambers in series back pressure regulators for handling some slurry flow out in the pressure step down section without fouling or plugging. The cryogenic lithiation and coupling reactions are accomplished in typical static mix tubes and coiled residence time tubes, and the continuous quench is accomplished in a continuous stirred tank reactor (CSTR) or semi-batch quench vessel. These two systems are designed to perform liquid phase reactions in plug flow, tubular reactors. They include Isco syringe pumps for accurate, precise, pulse-free pumping of liquid reagents under pressure, a stainless steel feed coil for BuLi so that it does not contact the mechanical pump seals, stainless steel static mixers and tube reactors, a cryogenic bath, pre-cooling coils for all feeds inside the cryogenic baths, and a glass CSTR for continous quench. In order to prevent the evolution of gasses and the accompanying distortions in residence time, the 2 PFRs in series operate under about 50 to 100 psig pressure. Reactor system 2 uses a commercially available back pressure regulator, while reactor system 3 uses expansion chambers in series for back pressure regulation. Due to the potential precipitation of high viscosity gels at cryogenic temperatures, the minimum tubing and fitting size is 1.78 mm inside diameter (i.d.). In order to overcome mixing limitations, static mixing tubes are used in the reactor after the mixing tees. Care should be taken to ensure proper mixing of feed streams, as mixing rates are affected by process flow rates. Minimum volumetric flow rates through the 3.18 mm i.d. static mixers should generally be about 10 ml/min. The pre-cooling coils are each 3 m long coils of 3.18 mm o.d. 1.75 mm i.d. stainless steel tubing. Due to the inexpensive nature of the tube reactors, reactors are often changed between chemistries. As mentioned above, back pressure for reactor system 3 is provided by expansion chambers in series with a nitrogen pressure supply and 2 automated block valves. This enables the S2
system to handle a small amount of solid products, at the cost of intermittent product flow. We try to keep the process homogeneous and avoid solids. However, there may be solids during startup transition or shutdown transition when the stoichiometry ratio may be too high in the mixing tee, or for brief periods during transitions or flow rate upsets during a flow chemistry run. If there are solids at steady state, then the reactor system can run for a time before clogging, probably about 1 or 2 hours, depending on how large and concentrated the solids are. Solution or light slurry flows out of the reactor through the expansion chambers in series through sequenced automated open/close block valves. When the valve is open, the flow path is wide, and the flow is high velocity because of high pressure difference. In contrast, the restricting orifice in the back pressure regulator for reactor system 2 is much more prone to plugging with solids. The feed coil is an important feature of this equipment set. It is designed for low axial dispersion so that 90% of the charged BuLi can be used without dilution by the push solvent. It is a 305 m long coil of 3.18 mm o.d., 1.75 mm i.d. tubing. The main advantage is that the mechanical pump does not contact the BuLi. BuLi is known to ruin mechanical seals because of abrasive salts. The safety and quality of the BuLi experimental work is greatly improved by this technique. It is not intended for manufacturing plant production, but it is suitable for research and development runs with 50 ml to 5 L BuLi. If we are using 1.5 to 4 liters of BuLi, then we use feed coils made from 6.35 mm o.d., 4.57 mm i.d. coiled tubing. Before it is filled with BuLi, the feed coil starts out filled with solvent, for example hexane. BuLi is manually pushed into the feed coil in reverse direction from an inert, pressurized container, before the start of the continuous reaction experiment. For example, in Figure S1, valves 1, 2, 3, and 4 start out closed. The pressure bottle could be a 1 L vessel (not shown in the figure). The pre-filled BuLi pressure bottle is connected to valve V3 via flexible tubing. An inerted waste container is connected to valve 1. Valves 1 and 3 are opened. The desired volume BuLi (for example 500 ml) is pushed into the feed coil. This may take about 10 minutes depending on the pressure on the BuLi vessel. Valves 1 and 3 are closed, and valves 2 and 4 are opened. Now the feed coil contains BuLi and is ready for the start of the continuous reaction experiment. S3
Figure S1. Lithiation Reactor System 2. PFRs in series with dome-loaded diaphragm style back pressure regulator. Lithiation Reactor System 3. A simplified drawing of lithiation reactor system is shown in Figure S2. Downstream from the PFR reactors, Vessel V1 is pressurized to constant pressure with regulated nitrogen (for example 50 psig). Valves A and B are automated open/close block valves, and their timing and sequence is controlled by the distributed control system (DCS) or PLC. Valve A opens for 5 seconds then closes. This allows the product solution or slurry to quickly flow into vessel V2, pushed by about 50 psig pressure difference. It also pressurizes the gas pocket in the top of vessel V2 to 50 psig. Then valve B opens for 5 seconds and then closes. This allows the product solution or slurry to quickly flow out of vessel V2 and into the quench CSTR, pushed by the 50 psig gas pocket in the headspace of V2. This automated valve sequence repeats about once every 30 seconds. Therefore, flow into the aqueous quench is intermittent slug flow with high linear velocity, about once every 30 seconds. The S4
solids are normally dissolved in the aqueous quench reactor R3. The key to minimizing solids fouling is the high velocity intermittent flow through wide open block valves, and no restricting orifices. With all that being said, if you know that you will constantly generate solids in flow or you are unsure, it is better to use lithiation reactor system 1 or 4, which have 3 intermittent flow stirred tanks in series rather than PFRs. Figure S2. Lithiation Reactor System 3. PFRs in series with expansion chambers in series back pressure regulation. Lithiation Reactor systems 1 and 4. Fill-Empty Stirred tanks in series. The other two reactor designs consist of three intermittent flow stirred tanks in series. They operate in fully automated fill-empty mode. The smallest internal diameter in tubing and fittings between vessels is 3.18 mm, and intermittent slurry flow out of each stirred tank has high linear velocity. These systems were originally designed to be used for testing of cryogenic lithiation, coupling, and quench reactions in series at small scale. However, the systems are capable of running many types of chemistries the temperatures between -78 C and >100 C. These reactor S5
systems can be used for 1 gram experiments, and they can also be used for 1 kg per day production. Automation is used for all of the reagent feeds, and for the timing for each reaction in sequence. The timing and volume of each reagent addition is completely adjustable by the user. For example, the user can choose reaction time 5 minutes in the first reactor, 30 seconds in the second reactor, in one minute in the third reactor in series. This is an advantage compared to running research scale flow experiments in PFRs-in-series, because you cannot change reaction time independently for each reaction in series in a PFRs without taking it apart and changing sizes of each tube. Reactor system 1 uses Isco syringe pumps because it operates at elevated pressure, but reactor system 4 uses Masterflex peristaltic pumps because it operates under atmospheric pressure. The BuLi feed coil is designed and operated the same as was described for lithiation reactor system 2. However, flow is intermittent rather than truly continuous, therefore the BuLi feed zone is used to prefill with an accurate and precise volume of BuLi, which is subsequently pushed into the first stirred tank in series. Another benefit of the fill-empty stirred tanks in series is that the data from these reactors can be used to quantify both kinetics and thermodynamics of the cryogenic reactions. The temperature versus time trends are logged by the DCS. Overall heat transfer coefficients for the reactors are calculated by modeling the cooling rate data from the blank solvent runs. The temperature profiles from the reactions, the known overall heat transfer coefficients, and the known molarities are modeled numerically to quantify reaction rates and heats of reaction. Lithiation Reactor system 1: In this reactor system, the pumps are not automated. The Isco syringe pumps are set to a constant pumping rate and pump against constant pressure at all times. However, the flow of reagents into the stirred tanks is intermittent. The Isco pumps constantly flow into transfer zones, labeled TZ1, TZ2, and TZ3 on the drawing in Figure S3. S6
Figure S3. Lithiation Reaction System 1. Intermittent flow stirred tanks is series with a cascade of reaction pressure. The transfer zones gradually fill, and then empty into the reactors quickly when the automated valves open. The BuLi feed coil is filled with BuLi and operated as described above for the other reactor systems. The magnetic stir bars in the stainless steel reactors are oriented vertically, and the magnetic drives are mounted to the side of the constant temperature cryogenic baths. The reactors are submerged down into the cryogenic cooling baths. The reactors are made from 19.1 mm o.d., 15.7 mm i.d. stainless steel tubing, 30.5 cm long, mounted vertically so that the bottom half of the reactor is submerged in a dry ice/acetone batch. The magnetic stir bars are about 12.7 cm long, and they vibrate back and forth and rotate. Constant pressure of 15 psig is applied to the headspace of the first stirred tank R1 at all times, 7 psig is applied to the headspace of the second stirred tank R2 in series at all times, and the 3 rd stirred tank in series R3 is always vented to the atmosphere at 0 psig. This way, pressure is used to push lithiation reaction product into the coupling reactor, and pressure is used to push from the coupling reactor to the quench reactor. These transfers occur when automated block S7
valves between the stirred tanks open. Cryogenic slurry transfer from reactor 1 to 2 takes about 1 second, and from reactor 2 to 3 takes about 4 seconds. The automated, repeating sequence is as follows: 1. Open valves B, J for 5 seconds, then close valves B, J. This transfers the aryl bromide from transfer zone TZ1 into reactor R1 and the DMF from transfer zone TZ3 into reactor R2. The transfer zones are maintained at pressures 20 psig, and the reactors are maintained at lower pressures, therefore the feed solutions from the transfer zones flow into the reactors quickly by pressure driving force. 2. Wait cooling time, 1 minute (user input). 3. Close valve O. 4. Open valve F for 5 seconds, then close valve F. This transfers the BuLi into reactor 1, where it mixes all-at-once with the aryl bromide and starts the lithiation reaction. The transfer zone is maintained at pressure 20 psig, and the reactor is maintained at pressure 15 psig, therefore the BuLi from the transfer zone TZ2 flows into the reactor R1 quickly by pressure driving force. It takes about 1 second for the BuLi transfer into the reactor. 5. Open valve O. This supplies a constant nitrogen blowback to keep the contents in the reactor instead of entering the bottom outlet tubing. 6. Wait lithiation time, example 10 seconds (user input). 7. Close valve O. 8. Open valve M for 5 seconds, then close valve M. This transfers the contents from reactor R1 into reactor R2, where the lithiated intermediate mixes all-at-once with the precooled DMF. The transfer takes about 1 second because of the pressure difference and the large diameter tubing (4.57 mm i.d.) 9. Open valve O. 10. Wait coupling time, example 10 seconds (user input). 11. Close valve N. 12. Open valve P for 10 seconds, then close valve P. This transfers the contents from reactor R2 into the quench reactor R3, where the coupled product mixes all-at-once with the precooled aqueous quench. The transfer takes about 5 seconds. 13. Open valve N. 14. Wait quench time, example 30 seconds (user input). 15. Repeat the sequence. Lithiation Reactor system 4: The glass reactors had no pressure rating, therefore the reactors R1, R2, R3 operated at 0 psig (Figure S4). The glass reactors were 250 ml volume and cone bottom. Minimum stir volume with overhead stirring was about 10 ml. These reactors operated at about 10 20 ml fill volume. The bottom half of each conical glass reactor was submerged into a cryogenic cooling bath. In this system, each stirred tank in series can be operated closer to end of reaction kinetics like a true CSTR, closer to all-at-once addition kinetics like a PFR, or closer to S8
controlled addition kinetics like a batch reactor. The operating mode depends on what is best for the chemistry and how the chemistry is intended to be scaled up. For example, cryogenic lithiation, coupling, and quench reactions in series were operated with all-at-once addition in each reactor and fill-empty, to most closely mimic the conditions of PFRs with static mixers. The timing and flow rates of all the pumps were completely adjustable by the user. Charging of all reagents was sequenced and fully automated. The operator input values, and the automation controlled the start and stop time of each pump and what rate to pump each reagent, the reaction times, and the times to transfer slurry from one reactor to the next. Figure S4. Lithiation Reactor System 4. Atmospheric pressure intermittent flow stirred tanks in series. Sequence: Referring to the Figure S4, the following sequence was automated and repeating. Initially valves B D G J are open. This vents all 3 reactors and the product collection vessel to the atmospheric pressure vent header. Initially valve M is open. This keeps 30 psig N 2 in TZ1 30 ml BuLi transfer zone. S9
The DCS executes the following automated sequence: 1. Close M, Open K. This will allow the system to pump the desired volume of BuLi into the 30 ml TZ1 zone. Start Pump 1 Start Pump 2 Start Pump 3 Start Pump 4 at P1 Start Time at P2 Start Time at P3 Start Time at P4 Start Time Stop Pump 1 Stop Pump 2 Stop Pump 3 Stop Pump 4 at P1 Stop Time at P2 Stop Time at P3 Stop Time at P4 Stop Time (This pumped (This pumped (This pumped (This pumped 7.0 ml aryl 3.0 mll BuLi 13.7 ml DMF 20.6 ml acid bromide into intotz1 zone.) into reactor R2.) quench solution reactor R1.) (user input) (user input) into reactor R3.) (user input) (user input) 2. Wait pre-cooling time 2.5 minutes (user input). 3. Close K 4. Open L. This quickly pushes the BuLi solution (for example 3 ml) from TZ1 into reactor R1. The transfer takes about 1 second. This starts the lithiation reaction in Reactor R1. 5. Wait L Open Time. 6. Close L 7. Open M.. 8. Wait Reaction Time 1 (10 40) seconds (user input). 9. Close D. This blocks vent line from reactor R2. 10. Open A, C. This pulls vacuum on reactor R2 so that slurry from reactor R1 quickly transfers to reactor R2 through the 6.35 mm o.d. 4.57 mm i.d. PFA tube. The transfer takes about 2 seconds. This starts the coupling reaction in Reactor R2. Opening valve A replenishes the headspace gas so that the vent header and reactor R1 stay at 0 psig. A 150 ml pot downstream from valve C in the vacuum supply (not shown in Figure S4) serves to give a quick vacuum pull for the transfer. A metering valve downstream from the 150 ml pot serves to limit the quantity of the vacuum pull. 11. Wait Transfer Time 1, for example 5 seconds (user input). 12. Close A, C. This stops the vacuum pull. S10
13. Open D. This vents reactor R2 again. 14. Open E, wait 2 seconds, then Close E. This causes the liquid or slurry that was pulled up into the tubing between reactor R2 and valve E to drop back into reactor R2 by gravity. 15. Wait Reaction Time 2, for example 25 seconds (user input). 16. Close B, G. This blocks vent lines from reactors R1 and R3. 17. Open A, E, H. This pulls vacuum on reactor R3 so that slurry from reactor R2 quickly transfers to reactor R3 through the 3.18 mm i.d. PFA tube. The transfer takes about 2 seconds. This starts the quench reaction in reactor R3. Opening valve A replenishes the headspace gas so that the vent header and reactor R2 stay at 0 psig. A 150 ml pot downstream from valve H in the vacuum supply serves to give a quick vacuum pull for the transfer. A metering valve downstream from the 150 ml pot serves to limit the quantity of the vacuum pull. It is critical that the process tubing from R3 to the product collection vessel goes to a higher elevation than the process tubing between R2 and R3. The vacuum creates a pressure difference sufficient enough so that slurry can flow up and out of R2 to R3, but it cannot overcome gravity enough to continue to pull up and out of R3 toward the product collection vessel. In other words, if the process tubing between vessels looks like an upside down U, then the upside down U from R3 to the product collection vessel is taller than the upside down U from R2 to R3. 18. Wait Transfer Time 2, for example 5 seconds (user input). 19. Close A, E, H. This stops the vacuum pull. 20. Open B, G. This vents reactors R1 and R3 again. 21. Wait Reaction Time 3, for example 25 seconds (user input). 22. Close J. This blocks the vent line from product collection vessel. 23. Open F, I. This pulls vacuum on product collection vessel so that the liquid-liquid mixture from reactor R3 quickly transfers to product tank through the 3.18 mm i.d. PFA tube. The transfer takes about 5 seconds. Opening valve F replenishes the headspace gas so that the vent header and reactor R3 stay at 0 psig. A 150 ml pot downstream from valve I in the vacuum supply serves to give a quick vacuum pull for the transfer. A metering valve downstream from the 150 ml pot serves to limit the quantity of the vacuum pull. 24. Wait Transfer Time 3, for example 10 seconds (user input). 25. Close F, I. This stops the vacuum pull. 26. Open J. This vents the product collection vessel again. 27. Wait Cycle Wait Time, for example 10 seconds (user input). 28. Repeat sequence S11
Temperature and Pressure Trends for Lithiation Run -71-72 -73 Temperature (C) -74-75 -76-77 Bath Temperature Lithiation mixing temp Lithiation after 1 SM temp Coupling mixing temp Coupling after 1 SM temp -78-79 13:04 13:12 13:19 13:26 13:33 13:40 13:48 Time Figure S5. Typical temperature trends of the lithiation and coupling reactions in sequential PFRs. Dry ice was added to the cooling bath at 13:17. 43 42.5 Reactor outlet pressure (psig) 42 41.5 41 40.5 40 Back Pressure 39.5 13:04 13:12 13:19 13:26 13:33 13:40 13:48 Time Figure S6. Typical pressure trend of the lithiation and coupling reactions in sequential PFRs The pressure oscillations are caused by operation of the automated valves and the expansion chambers in series. S12
Mg Sequestering for research scale Grignard formation CSTR In the Grignard CSTR experiments, the most difficult aspect is keeping magnesium in the CSTR, and preventing magnesium fine particles from pumping downstream. This section explains how to do it experimentally at the research scale using 250 ml glass flasks. Product solution is continuously pumped out of the reactor by pulling through a dip tube that is inserted down inside a settling pipe. The settling pipe is inserted through a customized side port into the glass flask at an angle of 30 degrees from horizontal. This can be seen in the picture in Figure. The height of the end of the outlet tubing is equal to the height of the liquid inside the reactor. The total pumping rate out of the reactor is about 50% higher than total liquid flow rate. This way gas and liquid segments flow out of the react door and it maintains liquid level at the dip tube height. When the product solution flows up the angled settling pipe toward the outlet tube, magnesium particles settle back down and tumble back into the reactor by gravity. The vapor space up inside of the settling pipe is connected to the vapor space inside the head of the reactor. This equalizes liquid levels in the settling pipe and the reactor. This is shown in the picture in Figure. The liquid segments entering the outlet tube come up from the settling pipe, and the vapor segments entering the outlet tube come down from the reactor headspace. This way gas bubbles do not disturb the magnesium settling inside the angled settling pipe. S13
Figure S7. Settling pipe and outlet tube in the side of a 250 ml Grignard formation CSTR. S14
Figure S8. Settling pipe and vapor space connection to the top of the CSTR. Formylation and quench temperature trends The temperature versus time data for formulation and quench CSTRs in series is shown in Figure. During reaction, the process temperature in CSTR1 was about 42 C, and the process temperature in CSTR3 was about 33.5 C. The quench reaction was more exothermic as seen by the larger delta between process and jacket temperatures. Delta T between jacket and process was about 3 C for the fomylation reaction and 6.5 C for the quench reaction at steady state. S15
Figure S9. Temperature versus time data for formylation and quench CSTRs in series. Further engineering calculations and characterizations are needed for scale up of the Grignard reaction in the CSTR. The overall heat transfer coefficient should be calculated for the research scale reactor and predicted for production scale. A real time energy balance should be performed and used as a safety feature of the process. This can be done by quantifying the heat removed from the reactor by the heat transfer fluid in the jacket, and it requires understanding of the energetics a reaction. A kinetic model should be generated by measuring and modeling reaction rate as a function of temperature. Mixing calculations are needed for solids suspension at research and pilot scales, and the amount of vortexing as a function of agitation speed considering the baffling. Calculations are needed for sizing and positioning of the settling pipe, and momentum balance to estimate efficiency of settling pipe for keeping Mg particles in the CSTR upon scale up. Also, calculations are needed for minimum required mean residence time in the settling pipe and the Mg trap, size of these relative to CSTR, and the minimum particle size that will be sequestered as a function of fluid properties and flow parameters. S16
Grignard formylation and reduction impurity trends Figures S10 and S11 show impurity trends for the Grignard formylation and reduction steps, respectively. Figure S10. Continuous Bouveault formylation HPLC impurity data: area% vs. time (h). Continuous Reduction: CSTR 4 Figure S11. Continuous reduction HPLC impurity data: area% vs. time (h). S17