The document discusses using ReactIR technology to provide insights into chemical reactions and processes. It presents three case studies where ReactIR was used: (1) monitoring an unstable acid chloride intermediate in a Vilsmeier reaction, (2) studying mixed anhydride formation with unstable intermediates, and (3) gaining understanding of a chiral resolution process. ReactIR allowed observing reaction components in real-time, identifying side reactions, and gaining mechanistic insights in all three cases.

1. David W. Place Scientific Update Basel, Switzerland ReactIR as a Diagnostic Tool for Developing Robust, Scalable Synthetic Processes 29-OCT-2007 InPACT

2. InPACT ( In tegrated P arallel A utomated C hemistry T echnologies) 8 dedicated personnel on 2 sites High throughput screening Rapidly screen multiple combinations of reaction parameters utilizing state-of-the-art automated reactor technologies. Reaction optimization screening Quickly generate data on improved reaction conditions to assist SRD’s timely development and optimization of processing conditions. Technology assessment (ReactIR) Identify and evaluate new process screening and development technologies and demonstrate their application to SRD’s process research & development.

3. Outline Brief Introduction Some Definitions and Arguments Systems, Software and Techniques ReactIR for Process Monitoring Solvent Exchange Unit Operations Atmospheric Azeotropic Distillation ReactIR as Reaction Diagnostic Case Study #1: Vilsmeier Reaction Case Study #2: Unstable in-situ intermediates Case Study #3: A Peek into Chiral Resolution Summary

4. “ Robust” Wikipedia – “ Healthy, strong, durable, often adaptable, innovative, flexible” “ Statistics: Robust statistical test performs even if its assumptions are violated…” “ Engineering: …the design is capable of functioning correctly (or at the very minimum, not failing catastrophically) under a great many conditions” How to achieve robustness in a process Myth busting Definition/Challenge of assumptions Rigour through Repetition Definition of Tolerances (Fail points vs. Fail safe) ReactIR DoE

5. Creation of Process Myths It is often difficult to define reaction response when inadequate analytical techniques are available to study that response The # of myths created for a given reaction or process increases as the # of simple and rapid analytical methods to study that reaction decreases ReactIR helps fill the gap

6. InPACT - ReactIR ( In tegrated P arrallel A utomated C hemistry T echnologies) ReactIR Technology Attentuated Total Reflectance (ATR) In-situ monitoring Non-perturbational Mobile Fiber optics (iC10 below) ReactIR Capabilities Max IR Range 4000-600 cm -1 pH Range 0-14 Temperature Range –80 to 200 o C SiComp Detector K6 Conduit IR Source And Data Collector iCIR software And ConcIRT ReactIR™ iC10

7. ConcIRT and ConcIRT Live! Deconvolution Software Any Spectrum at any given point in time is a composite of all IR active components What do you do when a single isolated band is not available?  ConcIRT  A1 A2 A3 A4 A5 A6 A7  A8

8. ConcIRT A users observations Things to be aware of: Equimolar amounts of chemicals in a reaction mixture that change at the same rate will not be deconvoluted into separate components by ConcIRT A mixture of chemicals charged into the reactor prior to initiating IR collection will not be deconvoluted into separate components by ConcIRT Take reference spectra of all known materials if possible Wet Sensor with solvent to be used Initiate ReactIR data collection Add solid(s) to reactor allowing time for dissolution/equilibration -OR- Add liquids/solutions via syringe pump at a rate 5x collection interval For each change/addition made to the reaction allow 5x the collection interval when feasible. Continue with the rest of the experiment. A + B C ConcIRT is useful for monitoring change To ensure proper deconvolution by ConcIRT

9. David Place ReactIR and Unit operations 10-APR-2007

10. Solvent Exchange A Practical Approach to Process Monitoring Problem : Solvent A is replaced with Solvent B before final crystallization of API. Residual Solvent A inhibits final crystallization of API Question : Is there a way to monitor the solvent exchange and determine when Solvent A concentration < specification without GC? Benefit : Able to pinpoint when a confirmation sample should be taken.

11. Peak Selection

12. Depletion of Ethanol from 1:1 Ethanol/Ethyl Acetate Mixture GC Assay = 0.5 wt% EtOH ATR Assay (reflux)= 4.4 wt% EtOH ATR Assay (r.t.) = 1.4 wt% EtOH

13. Depletion of Ethyl Acetate from 1:1 Ethanol/ Ethyl Acetate mixture with 0.16M “API” “ API” 1513 cm -1 GC Endpoint Assay = 0.23 wt% EtOAc ATR Endpoint Assay (raw) = 2.1 wt% EtOAc ATR Endpoint Assay (ConcIRT) = 0.3 wt% EtOAc 1740 cm -1 EtOAc depletion 883 cm -1 EtOH enrichment ATR Spectra 1740 cm -1 (EtOAc)

14. A Real Example TBME/EtOH solvent Swap ConcIRT Analysis (a) (b) (c) 610 mL EtOH Charge Distillation starts Best TBME/EtOH ratio

15. Summary ReactIR 4000 can be used to monitor solvent exchanges ReactIR Data agrees with GC data in determination of endpoint concentration for Ethyl Acetate depletion The detection limit will depend on Peak absorbativity, spectral overlap, and “thermal baseline” Study of exchange process will be required before implementation into large scale operations ATR Calibration to solvent system Temperature sensitivity of solvents used Small Scale simulation ConcIRT will be necessary to monitor more complex mixtures

16. Azeotropic Distillation A Practical Approach to Process Monitoring Problem : Reagent in next step will be destroyed by residual water > 0.03wt% Question : Can ReactIR be used to follow the removal of H 2 O by azeotropic distillation with Toluene to ensure consistent removal?

17. Azoetrpoic Distillation Unit Operation ConcIRT Deconvolution Component Profiles Charge 30g FMoc Glutamate mono hydrate and 150 mL Toluene; Start heating to Tj = 140C, dissolution of FMoc Glutamate monohydrate during heat ramp between 24C and 71C; Distillation commences Tj = 140C, Tr = 111C; Distillation stopped collected 132mL distillate; (e) Tj = -15C, Tr = -11.5C (a) (b) (c) (d) (e)

18. Comparison to KF data Simple azeotropic distillation to remove water Fmoc Glutamate Deconvoluted ReactIR Data agrees with KF at the distillation endpoint Component trace for H 2 O disagrees with KF data during distillation

19. Solvent Assisted Dehydration? Stable Hot Unstable Cold? C=O 1733 cm -1 C=O 1725 cm -1 C=O 1729, 1702 cm -1

20. ConcIRT Components Signal due to Toluene Signal due to Toluene Signal due to Toluene C=O str FMOC Glutamate

21. Summary ReactIR can be used to monitor azeotropic distillation Detection of free Water down to the ppm level appears to be possible Endpoint detection using ConcIRT ReactIR techniques give you physio-mechanistic information even for “simple” operations Define interrelationships and interdependencies of reaction components ReactIR as PAT empowers you to: Pinpoint physical changes in the system relative to process observations Cross-correlate process steps with real-time changes Generate a “snapshot” of your entire process for Process Validation

22. David Place Case Study #1: Vilsmeier Reaction 10-APR-2007

23. Understanding Stability An Acid Chloride intermediate Situation– Late Stage Phase 2  3 process Acid Chloride formation via Vilsmeier strategy Inconsistent results on Plant scale led to multiple failed starts Random variation in impurity profile in API traced to this step Limited information from HPLC analysis (EtOH quench)

24. Vilsmeier Catalytic Cycle Me 2 NH 2 + Cl - + CO + HCl H 2 O H 2 O CO + CO 2 + 2HCl H 2 O H 2 O Thermal Barrier? Degradents Thermal Barrier? Solubility Solubility

25. Observations Reaction is completely suppressed at 0 o C DMF + Oxalyl chloride requires >0 o C to initiate catalytic cycle forming Acid Chloride DMF is required to form Acid Chloride DMCA.HCl + Oxalyl chloride at >0 o C leads to degradation of Oxalyl chloride w/o acid chloride formation Rapid Acid Chloride formation initiates at 18-20 o C Degradation of acid chloride confounded with crystallization/precipitation SM hygroscopic Successive attempts to produce Acid chloride with same lot  less and less acid chloride is formed

26. Product Formation vs. Oxalyl Chloride Decay Possible explanation of Scaleup disconnect At 0C no reaction with or destruction of Oxalyl chloride is observed Heat ramp  set to maximize acid chloride, minimize Oxalyl degradation

27. ConcIRT Deconvolution of Components Acid Chloride initiation and “decay” DMF Rapid Acid Chloride formation initiates at 18-20 o C Degradation of acid chloride confounded with crystallization/precipitation 1743 cm -1 Trace

28. Acid Chloride Quench with Water Not as unstable in presence of water as suspected THF * Water has been subtracted from this data Myth: The acid chloride will react immediately with water to form SM

29. LC/MS Data for Major Impurity = 193 amu Degradation via cyclic intermediate Acid Catalyzed cyclization Trace Impurity Detected in ReactIR Impurity identity Confirmed by NMR

30. Acid Catalyzed isomerization Isomers can be detected with the SiCOMP ConcIRT Component Spectra: IR “Footprint region” Cis- C=C-H 744, 667 cm -1

31. Summary In-situ or unstable intermediates can be monitored empowering process chemists to: Fine tune addition rates of reagents Optimize reaction times based on real-time measurements Diagnose and circumvent undesirable side reactions Eliminate a vast majority of trial-and-error experimentation in process optimization

32. David Place Case Study #2: Unstable In-Situ Intermediates 10-APR-2007

33. Mixed Anhydride Formation Using ReactIR to monitoring unstable intermediates Problem : Reaction of WAY-284855 to WAY-266123 is slow Question : What is causing slow reaction and intermittent fluctuation in reaction yield? Unstable Intermediates ReactIR

34. Mixed anhydride formation at 0 o C Mixed anhydride is not stable at 0 o C At end of N-methylmorpholine (NMM) addition, consumption of SMs slows Process for anhydride formation is not robust! NMM addition rate = 450 mL/h 4.1 mole/h 9 equiv/h

35. Mixed anhydride formation at -8 o C Mixed anhydride IS stable at -8 o C over at least 90 minutes 20% residual SMs after NMM addition means there is not enough base added to catalyze anhydride formation. This is possible cause of slow formation of amide in the next step NMM addition rate = 12 mL/h 0.11 mole/h 6 equiv/h

36. Mixed anhydride formation using 20% more NMM at -8 o C Excess of NMM does not appear to effect stability of the anhydride Complete conversion to mixed anhydride is achieved with 1.2 equiv of base when the NMM addition rate is slow. NMM addition rate = 6 mL/h 0.05 mole/h 3.6 equiv/h

37. Scaling Reaction Degradation vs. Addition time Problem : Reaction exotherms during addition of NMM and RNH 2 Calculated that addition rate on scale  >2h This will enable control over exotherm in large scale equipment Question : Will the mixed anhydride survive during this extended addition time? To test for worst case, addition of NMM and RNH 2 over 3h was studied. Use ReactIR to monitor formation during telescoped process. Simulated in 250 mL Jacketed reactor at –11 o C (limit of system capability)

38. Scaleup: Mixed anhydride Formation Stressed 3h (0.04 mole/h, 0.6 equiv/h) addition to avoid exotherm on scaleup (a) (b) (e) (f) (g) (d) (c) (a) To (b) IBCF Addition over 10 min (c) To (e) NMM addition over 3 h (f) To (g) RNH 2 addition over 3 h HPLC assay @ 220 nm Inflection at ~1 equiv NMM Incomplete reaction?

39. 3-D Contour Reveals More Info NMM.HCl is evident in solution R 3 N + -H Str. Vibrations Consumption Of IBCF Mixed anhydride Dissolved NMM.HCl was not observed with NMM Addition over 30 minutes

40. Dissolved NMM.HCl Kinetic Profile Addition of RNH 2 Dissolved NMM.HCl detected ~20 minutes after inflection point Upon addition of RNH 2 to mixed anhydride soln IBCF is consumed NMM.HCl component increases Slight exotherm detected

41. Possible Mechanistic Explanation X 1 2 Route 1 dominates until ~1 equiv NMM Route 2 activated with excess NMM k1 < k2 reason for inflection

42. Summary NMM mol equiv/unit time is important in formation of the mixed anhydride Lab scale reactions Base can be added faster with more control over exotherm Equiv NMM per unit time is higher  Less base required Large Scale simulation Base must be added slower to control exotherm Equiv NMM per unit time is lower  More base required A study of this type without ReactIR technique would have been much more difficult.

43. David Place Case Study #3: A Peek into Chiral Resolution 10-APR-2007

44. Chiral Resolution Case Study Using ReactIR’s “limitation” to your advantage Problem : Chiral resolution gives variable recovery and enantiometric purities on 1L + scales. Question : Can ReactIR be used to follow the resolution, define endpoints or give further insight into the resolution vs. time? Use ReactIR to monitor dissolved components and monitor solid/solution equilibrium.

45. Process Assumptions Acid and both enantiomers of the Substrate react quickly to form the salt Salt crystallizes out of solution due to differential solubility of the salt vs. its free base in ethanol The least soluable diastereomeric salt crystallizes first 50/50 mixture of enantiomers will require 0.5 equiv resolving agent to recover desired enantiomer in 50% maximum yield. Major salt product is the mono salt.

46. Establishing Components Comparison of ConcIRT detected component spectra An important first step in the use of ConcIRT data Not detected As a component

47. Chiral Resolution Unit Operations Charges Chiral Resolving agent (75C) Cooled to 60C (solution becomes turbid) Seeded reaction Cooled to 50C Cooled to room temp. (a) (b) (c) (d) (e)

48. Equilibration of Salt Forms Monitoring what is in solution vs. what is not Cool to RT  Complete ejection of Form A RT Hold  Form A/B equilibration Cannot rule out Form B crystallization during equilibration This all happens with little change to overall SM signal 96.7 %ee, 38.4% yield

49. Summary Two forms of the diasteriomeric salts and their proportionation appear to be factors in resolution outcome 2h for proportionation  96.7 %ee, 38% yield No proportionation  90.3 %ee, 41% yield Difficult to assign specific cause from experiment but Identified a competative process between salt forms  a key to defining robustness? If process is accelerated  leads to unpredictable results ReactIR as a tool for process understanding

50. Future Plan: Implementing in Large Scale Lab Scale work Study and understand dynamics of process through references and reactions Establish components of reaction; Relate these to Unit operations Repeat experiment to generate confidence and Identify critical observables Determine how well process endpoints correlate with other assay techniques Demo Scale (1L-5L) Repeat best process at larger scale to identify discrepancies from Lab scale Determine best sensor technology and positioning Compare critical observables to those generated in Lab scale (Same?) Cross reference to other assay techniques Kilo Lab Scale (10L – 250 L) Purchase/Design interchangeable ways to introduce technology Repeat Best Demo Process and compare critical observables

51. Conclusions ReactIR is a powerful tool to obtain qualitative information about chemical processes Implement this tool  Augment process understanding  Maximize knowledge from every experiment Use technique to Challenge or Confirm Process Myths (untested hypotheses) Make better decisions and ask better questions based real-time AND assay data

52. Acknowledgements Unit Operations Case Study Silvio Iera Case Study #1 Kan Eng Alex Gontcharov Warren Chew Qing Yu Karen Sutherland Sylvain Daigneault Case Study #2 Alice Sebastian Peter Wehrenberg Case Study #3 Sreeni Megati Robert Tinder Lisa Routel Mettler Toledo Will Kowalchyck Carlo Tripodi Tom Holmberg Wyeth InPACT group Cyril Benhaim, Haris Durutlic, Misato Konishi, Mike Macewan, Luc Richard, Rob Tinder Martin Guinn Mike O’Brien