gsk approach to enhancing process understanding using dynochem: reaction kinetics examples. james...

14
1 GSK Approach to Enhancing Process Understanding using Dynochem: Reaction Kinetics Example James Wertman Process Engineering GlaxoSmithKline James Wertman Dynochem User Group Meeting 15-16 May 2007 Page 1 Overview Introduction Example for Kinetic Modeling Experimental Design Dynochem Model Data Reduction Reaction Scheme Preliminary Parameter Fitting Model Refinement and Selection Complete Parameter Fitting Optimization Conclusions and Future Work

Upload: scale-up-systems

Post on 05-Jul-2015

470 views

Category:

Technology


3 download

TRANSCRIPT

1

GSK Approach to Enhancing Process Understanding using Dynochem: Reaction Kinetics Example

James WertmanProcess EngineeringGlaxoSmithKline

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 1

Overview

• Introduction

• Example for Kinetic Modeling

• Experimental Design

• Dynochem Model

– Data Reduction

– Reaction Scheme

– Preliminary Parameter Fitting

– Model Refinement and Selection

– Complete Parameter Fitting

• Optimization

• Conclusions and Future Work

2

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 2

Introduction

• Introduction and implementation of Quality by Design

– Strong focus on process understanding to support process

development and project decisions

• Maximize process understanding over operating ranges

from discrete experimental points

– Dynochem and other tools facilitate this implementing mechanistic

and empirical models

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 3

Reaction Example

O

OH

H

O

O

OHO

OO

OH

H

MeO

O

H

MeO

Pleuromutilin Epi-Pleuromutilin "Alkene"

• Homogeneous reaction

• Considerable experimental experience

– General effects of concentration, reagents, and temperature understood

– Scaled reproducibly from < 1g to 40kg scales

– Wide range of conditions already proven acceptable

3

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 4

Reaction Example

O

OH

H

O

O

OHO

OO

OH

H

MeO

O

H

MeO

Pleuromutilin Epi-Pleuromutilin "Alkene"

• Starting material is a fermentation product

– Highly variable impurity profile

– Large number of impurities (30+)

• Limited analytical information early in project development

– Few identified impurities

• Numerous feasible reaction pathways, mechanistically complex

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 5

Experimental Design

• Overall process focus on crystallization and isolation,

constraints imposed by the work-up procedure

– Reduce solvent/reagent amounts to safe minimum, ~3 volumes

– Maintain acid below 1.7 molar equivalents

– Maintain temperature below 40°C

• Solvent/reagent ratio fixed at 2:1

– Previous experimentation suggests negligible effect of ratio on

reaction performance

• Investigation focus on temperature and acid effect on

reaction profile

4

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 6

Experimental Design

• Experimental design selected as a two level factorial design with center points to allow for statistical analysis of data if desired

– Methanol and TMOF fixed at 2 and 1 volume respectively

– Temperature: 25 to 40°C

– Sulfuric Acid: 0.85 to 1.7equivalents

– Chromatography purified Pleuromutilin utilized to minimize impurity complication of analysis

• Time zero is acid charge, exothermic temperature spike minimized by equipment configuration but still present

• Sampling at 0.5, 1, 2, 4, 8...min

– Sampling extended up to 48 hours to capture “Alkene” formation

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 7

min0 2.5 5 7.5 10 12.5 15 17.5 20

Norm.

0

250

500

750

1000

1250

1500

1750

DAD1 D, Sig=210,4 Ref=360,100 (JW092\035-3501.D)

7.842

10.504

12.705

12.900

12.971

13.183

13.844

14.195

14.532

14.837

15.093

15.277

15.918

16.705

17.115

17.634

17.886

18.076 18.241

18.766

Raw Data

min0 2.5 5 7.5 10 12.5 15 17.5 20

Norm.

0

200

400

600

800

DAD1 D, Sig=210,4 Ref=360,100 (JW092\012-1201.D)

7.837

10.494

11.307

12.167

12.708

13.187

13.848

14.558

14.838

15.277

15.497

16.705

min0 2.5 5 7.5 10 12.5 15 17.5 20

Norm.

0

200

400

600

800

1000

1200

1400

DAD1 D, Sig=210,4 Ref=360,100 (JW092\021-2101.D)

7.838

10.503

12.706

12.901

13.184

13.845

14.548

14.837

15.275

16.701

18.255

min0 2.5 5 7.5 10 12.5 15 17.5 20

Norm.

0

200

400

600

800

1000

1200

DAD1 D, Sig=210,4 Ref=360,100 (JW092\008-0801.D)

6.848

10.487

Pleuromutilin

Epi-Pleuromutilin

“Alkene”

Intermediate Species?t = 0min

t = 4min

t = 64min

t = 640min

5

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 8

Data Reduction

• HPLC areas extracted to Excel

• Area percents calculated adjusting for relative response

factors

– Pleuromutilin 5

– Epi-Pleuromutilin 1

– “Alkene” ½

– Σ(unknowns) 1 assumed

• Adjusted area percent utilized assumed to correspond to

weight percent

– Modeling performed with and without inclusion of unknown peaks in

area percent calculation

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 9

Reaction Mechanism

O

OH

H

O

O

OHO

OO

OH

H

MeO

O

H

MeO

MeOH

CH(OCH3)3

H+

H+

Pleuromutilin Epi-Pleuromutilin "Alkene"

• Propose relevant mechanism while incorporating historical experience

– Without methanol or trimethyl orthoformate (TMOF) no product or major

byproduct is detected

– Strongest dependence to temperature and acid content

– Pleuromutilin to Epi-Pleuromutilin transformation observed to be reversible

– Multiple intermediates observed by HPLC, but no analytical information

6

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 10

Reaction Mechanism

O

OO

OH

H

MeO

O

OH

H

O

O

OHO

OH

OH

O

OH O

OHO

OH

H

O

O

OHO

OH

H

O+

O

OHO

OH

H

OHO

H+

H+

H+

H2O

H+

MeOH

H+

O

H

MeO

+

+

+

A B

CD

Pleuromutilin

Epi-Pleuromutilin"Alkene"

• Start with reference

information

• Fits experimental

observations except

for the incorporation

of TMOF

1 H. Berner, G. Schulz, and H. Schneider,

Tetrahedron 36, 1807 (1979).

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 11

Reaction Scheme for Modeling

O

OO

OH

H

MeO

H2SO

4H+ SO

4

H+

O+

O

O+

O O

OH

H

O

O

OHO

OH

H

O

OH

OO

O

O

OH

H

O

OH

OMe

O OH

O

O

H

MeO

OH

O

OHH+

CH(OCH3)3 MeOH

MeOH

O

OHO

OH

H

MeO

Epi-Pleuromutilin

2

+ +

+

+

Pleuromutilin

+

+

++

"Alkene"

2-

+

+

• Mechanistically relevant

reaction scheme

• Actual structures of

intermediates and byproducts

selected only to fulfill mass

balances, no analytical

information available

• Concurs with experimental

observations

7

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 12

Dynochem Setup

O

OO

OH

H

MeO

H2SO

4H+ SO

4

H+

O+

O

O+

O O

OH

H

O

O

OHO

OH

H

O

OH

OO

O

O

OH

H

O

OH

OMe

O OH

O

O

H

MeO

OH

O

OHH+

CH(OCH3)3 MeOH

MeOH

O

OHO

OH

H

MeO

Epi-Pleuromutilin

2

+ +

+

+

Pleuromutilin

+

+

++

"Alkene"

2-

+

+

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 13

Kinetic Parameter Fitting

• Actual temperature profiles imposed

• Initial guesses, focus on forward rate constants

– First two reactions expected to be fast

– Pleuromutilin through Epi-Pleuromutilin rates assumed faster than the

conversion to Alkene

8

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 14

Kinetic Parameter Fitting

• Manually adjust rate constants for rough fit of data, single scenario

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 15

Kinetic Parameter Fitting

• Manually adjust rate constants for rough fit of data for both scenarios

at 25°C

25°C, 0.85eq Acid 25°C, 1.7eq Acid

9

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 16

Kinetic Parameter Fitting

• Fit reaction rate constants to

both reactions at 25°C

– Acid dissociation and TMOF

‘activation’ reactions not

considered

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 17

Reaction Scheme Refinement

• Incremental reaction scheme refinement employed

• Rate constant fitting repeated for each reaction scheme for both

experiments at 25°C

• Fit statistics compared in selecting the reaction scheme for model

O

OO

OH

H

MeO

H2SO

4H+ SO

4

H+

O

OH

H

O

O

OH

O OH

O

O

H

MeO

OH

O

OHH+

CH(OCH3)3 O

OO

OH

H

MeOH+

Epi-Pleuromutilin

2

+

+

+

Pleuromutilin

++

"Alkene"

2-

+

Epi-Pleuromutilin

+

10

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 18

Reaction Scheme Selection

5.594

5.248

5.296

Model Selection

Criteria

101.2sec

141.9sec

6.72min

Fit Run Time

2.05e40.39913

9.07e30.43494

7.99e30.43275

F-statisticSSQReactions

• Best fit attained with the simplest reaction scheme

• Sufficient for characterizing acid and temperature effects

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 19

Final Parameter Fitting

• Manually adjust activation energies for a rough fit of data at 40°C

11

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 20

Final Parameter Fitting

• Fit activation energies with both

experiments at 40°C

• Fit reaction parameters to

experiments at 25°C and 40°C

• Fit reaction parameters to all

experimental data

• Check fit varying initial guesses

• Visually check fit to data

42.26Solution.Rxn2.Keq

6.322Solution.Rxn3.k>

8.138Solution.Rxn2.k>

9.949Solution.Rxn3.Ea>

8.519Solution.Rxn2.Ea>

C.I (%)Parameter

5.322Model selection criteria

2.64E+04F-statistic

1.571SSQ

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 21

Optimization

• Defining an optimum reaction

– Starting material consumption

– Byproduct minimization

– Reaction time minimization

• ‘Optimum’ at maximum acid and temperature

– Single point during reaction, not accounting for decomposition after

optimum reaction time

• Criteria not sufficient for determining best reaction

conditions in this example

12

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 22

Optimization

• Defining new optimization criteria

– Reaction time = time at maximum product

– Quench window = time to consume 1% of maximum product

• Defining an objective function for optimization requires

discrete reaction time and quench window target times

• Actual goals are constraints

– Reaction time < 8 hours

– Quench window > 2 hours

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 23

Simulated Experimental Design

• Utilize statistical analysis software to model reaction time

and quench window responses

• User defined response surface design

– Simulations utilized to model responses

– Maximize resolution, 29 ‘experiments’

• Acid and temperature fit each response with polynomial

model and logarithmic transformations

– log(reaction time) = a*A + b*T + c*A2 + d*A*T...

13

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 24

Reaction Time Response Surface

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 25

Quench Window Response Surface

14

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 26

Operating Range

James Wertman

Dynochem User Group Meeting

15-16 May 2007 Page 27

Conclusions & Future Work

• Simplest reaction scheme selected for model providing the best parameter fitting

– Model sufficient to describe effects of acid and temperature only

– Additional experimentation and/or analytical information would be required to further refine the reaction scheme and model

• Dynochem results used in combination with other tools to provide a more complete understanding of the process

– Identified operating range based on desired reaction time and quench window

– Developed a clear relationship of factors to process operation

– Incorporate in-situ analytical measurements with model and verification as developed to further process understanding

• Operating conditions provided to deliver reasonable reaction time yet allow time to perform analysis