Microreactors in Discovery and Development

Klavs F. JensenDepartment of Chemical Engineering

Massachusetts Institute of TechnologyCambridge, MA 02139

kfjensen@mit.edu web.mit.edu/jensenlab

PoaC Symposium, Eindhoven May 22, 2013

Advantages of continuous synthesis

• Safety - Small scale, no headspace, no accumulation of reactive/toxic intermediates

• Expansion of reaction space/toolbox/feasibility - Many “Forbidden Reactions” become feasible

• Robustness, stability, QbD - Steady-state, continuous processes

• Scalability – scale-up faster and reliable

• Versatility and flexibility - customizable and adjustable equipment

• Leverage and efficiency - increase in throughput, with a dramatically reduced equipment footprint

Elements needed to realize continuous synthesis

• Chemically compatible, easily integrated continuous flow reactors with operation conditions that can be scaled from research to production

• Knowledge based design, in-line monitoring and optimization approaches of synthesis, workup, and purification steps

• Ability to scale predictably over several orders of magnitude of production for homogeneous as well as multiphase processes

• Immobilization and/or recycling of catalysts

• Development of workup techniques for integration with reaction sequences and when economically advantageous, recycle of regents

• Methods for handling (addition, formation, separation,…) of solids in continuous flow reactors

Components of a microreactor system

4

Reagents

Product

Quench

Reactor: microstructured metal, Si, glass, ceramic, or stainless steel, polymer tubing

Pump: syringe, HPLC, gear, pressure driven …

Mixer: active (stirring) , passive (diffusion)

Mixer

Temperature control

In-line Monitoring:UV-visFTIRRamanHPLC

Thermal quench

MixerPump Reactor

Microstructured Systems Tube Flow Systems

http://www.vapourtec.co.uk/

www.thalesnano.com

www.imm-mainz.de

www.imvt.kit.edu

www.chemtrix.com

5

Production Systems

DSM/KIT

Lonza/Ehrfeld/BTSD.M. Roberge et al. Org. Process Res. Develop. 2008, 12, 905–910www.ehrfeld.com

CorningS. Braune et al., Chemistry Today

2008 26 (5) Sep-Oct www.corning.com

ESKwww.esk.com

6

When is a premixing necessary?

1 20

4

ntkC d

DaD

=rate of transport

rate of reaction

• Diffusion time alone is not sufficient

• The Damköhler number is the appropriate scaling of reaction and transport effects

2

m

L

D

23

4td

DaD

K. D. Nagy et al., Org. Process Res. Dev. 2012, 16 (5) 976–9810 100 200 300 400 500 600

0

0.2

0.4

0.6

0.8

1

Residence time - s

Tu

be

dia

me

ter

- m

m

No mixing needed

• 1st order reaction 95% conversion after k = 3

• Similar numbers can be derived for other reaction expressions or experimental data

Dispersion effects

D = 10-9 m2/s

Bodenstein number (Bo) enables estimate of deviation from plug flow

0 100 200 300 400 500 6000

0.2

0.4

0.6

0.8

1

Residence time - s

Tu

be

dia

me

ter

- m

m tube

Convection model

Large deviations from plug flow

Small deviations from plug flow

Plug flow

plug flow

laminar flow

2 2

4D tu d

DD

uLBo

D

= 48 tube= 30 square channel

K. D. Nagy et al., Org. Process Res. Dev. 2012, 16 (5) 976–981

Application to a glycosylation reaction

• Improvement of mixing: 3x throughput

• Elimination of dispersion: identical behavior across tubes, no byproduct

Conversion

Dis

per

sio

n e

ffec

ts

dtube

(m) SS TPEEK

T

mixer

500 83 89 91

750 76 89 92

1000 48 73 84

= 10 s

K. D. Nagy et al., Org. Process Res. Dev. 2012, 16 (5) 976–981

+GL

94

93

89

Rapid reaction optimization using inline analysis and feedback

• Automated system with feedback capabilities for optimization• Plug-and-play

• Different in-line analysis methods, e.g., HPLC, FTIR

• Self calibrating

• Customizable, experimental optimization algorithms

• ‘Black-box’ approaches

• Model based methodsParameters: temperature, residence time, catalyst/ligand loading, solvent composition, ionic strength, pH

J.P. McMullen and K.F. Jensen, Annu. Rev. Anal. Chem. 3, 19–42 (2010).Org. Proc. Res. Dev. 14, 1169–1176 (2010)

Yield = 80%T = 88°C RT= 50s

• Nelder-Mead Simplex algorithm for online optimization

• 46 automated experiments to move from benzaldehyde yield of 21% to 80%

• Achieved 1 experiment per 10 minute throughput rate

11

Multi-variable reaction optimization

J.P. McMullen and K.F. Jensen, Org. Proc. Res. Dev. 2010 , 14, 1169–1176.

Online optimization with ReactIRflow cell

Conjugate GradientSteepest Descent

• Black box optimization is feasible with a variety of optimization techniques.• The choice of methods significantly influences the number of experiments

J.S. Moore et al, Org. Procss. Des. Devel. 2012, 16 ,1409–1415 12

Microreactors can produce chemical kinetics in relatively few experiments

• Benefits of using integrated microreactors for kinetic studies:

• With the appropriate approach, microreactors yield:– Rate law (mechanism formulation)

– Rate parameters

Better control of residence time and temperature

Study reactions at unconventionalconditions

Small amounts of expensive reagent material required

Intelligent design of sequential experiments

Easier to scale up Investigate fast reactions

No temperature gradient No head space

No mixing gradient Accurate sampling

13

No informationPlenty ofinformation

Rat

e

Time

J.P. McMullen and K.F. Jensen, Org. Process Res. Dev., 15 398–407 (2011)13

Reaction example

Use automated microreactor system to investigate reaction

– Select rate model from small list of potential rate forms

– After rate model is chosen, estimate rate parameters

– Optimize reaction conditions

by a small number of experiments varying time and inlet concentrations

1 1 2

22 1 2

I

II

r k C C

r k C C

23 1 2

4 1 2 4 3

III

IV f r

r k C C

r k C C k C

Pro

bab

ility

Dis

trib

uti

on

Isoprene Concentration

RateModel IRateModel IIRateModel IIIRateModel IV

Exp 5: = 60s, CIso = CMA = 0.5 M

0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20

Pro

bab

ility

Dis

trib

uti

on

Isoprene concentration

Rate ModelIRate ModelIIRate ModelIII

Exp 6: =60s, CIso = 2.0, CMA = 1.5 M

(1)

InlineHPLC

(2)

DMF

DMF

14J.P. McMullen and K.F. Jensen, Org. Process Res. Dev., 15 398–407 (2011)

Determine rate parameters

Optimize conditions

Scale-up

After 8 experiments

Parameter Estimate CI

EA (kJ/mol) 50.7 ± 0.9 (1.8%)

A x 105 (M-1 s-1) 1.71 ± 0.6 (35.5%)

Scale-up based on kinetics and reactor models

• Use chemistry information with reactor flow and heat transfer models to predict performance of scaled up system

– Example: incorporating kinetics from microreactor to scale up by a factor of 500 using a Corning Advanced Flow Reactor System

– Predicted performance agrees with experimental data

15

120 L 60 mL

Entry Time (min) Temp. (oC) Experimental Predicted

1 1.5 110 78.1 ± 0.4 79.3

2 2.0 100 82.6 ± 0.1 83.7

3 2.5 110 85.2 ± 1.0 86.5

4 1.0 126 83.5 ± 3.1 83.9

J.P. McMullen and K.F. Jensen, Org. Process Res. Dev., 15 398–407 (2011)

Using high pressure and temperature to accelerate reactions - Aminolysis of epoxides

Flow reactor yields similar to microwave batch yields without headspace issues, shorter reaction times, reduced bis-alkylation, and scalable performance – performance scales to larger reactors

16M.W. Bedore, et al. Org. Proc. Res. Dev. 2010, 14, 432-440

A

BC

A Metoprolol B C

CB

Kinetics and scaling of Aminolysis of epoxides

• Single preparation of two reagent solutions, 0.5 to 2 g of each reagent to scan up to 35 sets of conditions with samples in triplicate, generating over 100 data points within one 8-h period.

• Scale from 0.12 mL to 12 mL with accurate prediction of yield

• Dean flow off-sets increased dispersion

O

PhNH2

Ph

OH

NH

Ph

OH

NPh

OH

3

1

Ph NH

OH

OPh

Ph

OH

N

Ph

OH

Ph

HON

Ph

OH

2

6

4

5

7

27.2 kJ/mol

16.2 kJ/mol

8.4 kJ/mol

12.2 kJ/mol

Sudarsan, Ugaz, PNAS. 2006, 103, 7228-7233.

N. Zaborenko, et al. Org. Proc. Res. Dev. 2011, 15, 131-139. 17

Continuous flow systems for manufacturing

• Production scale systems can be achieved by scale-out or scale-up

• Scale-out – parallelization of microreactors– Retain heat and mass transfer benefits– Large challenges in fluidic connections

and control• Scale-up – increasing size of reactor

– Realize higher production rates – Retain heat transfer advantages– Reduce micro-scale concerns (i.e.

clogging)– Use knowledge of chemistry and

process to scale • Example:

– Corning/DSM - 5 weeks of operation - 40 metric tons processed (Chemistry Today, Vol 26 , Sep -Oct 2008)

0.1 mL

10 mL

18

Examples of reactor scaling

10 c

m

Lonza/Ehrfeld/BTSwww.ehrfeld.com

Corningwww.corning.com

19

spiral design Corning LFR Corning AFR

Volume: 220L 450L 8.7mL

Length:400m 1 mm700m

Scaling multiphase reactions

• Gas-liquid flow: spiral design (silicon nitride) and Corning AFR

• Liquid-liquid flow: spiral design, Corning LFR & AFR

*

L

A

A A A

dCr a Ck C

dt

Mass transfer in gas-liquid flow: pH-measurements by laser induced fluorescence (LIF) and velocities by particle image velocimetry (PIV)

pH

Velocity

Regions of high mass transfer correspond to regions with increased flow

21S. Kuhn and Klavs F. Jensen, Ind. Eng. Chem. Res., 2012, 51, 8999–9006.

Mass transfer in gas-liquid flow

• Microreactors: – Mass transfer is connected to internal recirculation

– Interfacial slip governs the strength of this recirculation

– Mass transfer is dominated by the bubble cap

GL

GG VV

V

Holdup:

G=0.82

G=0.54

S. Kuhn, M.J. Nieves Remacha 22

Gas-liquid flow in Corning AFR

10 mm1 mm

M.J. Nieves Remacha , A.A. Kulkarni

Bubble creation, recombination, and split enhance mass transfer

• Periodic merging and break-up of the bubbles

• Constant renewal of the gas-liquid interface

• Interfacial area is more efficiently used (entire gas bubble vs. bubble caps)

8.7 mL

0.45 mL

Butanol-water: mass transfer in Corning AFR and LFR modules

• Splitting and combining droplets provide effective mass transfer over different length scales

M. J. Nieves Remacha A. A. Kulkarni

S. Kuhn,A. Woitalka 24

Hexane/Water Hexane always dispersed phase due to hydrophilicity of glass walls

Drop size decreaseswith water flow rate

Number of drops increaseswith hexane flow rate

Flow patterns of liquid/liquid flow:

25

Qw = 10 mL/min, Qh = 10 mL/min Qw = 40 mL/min, Qh = 40 mL/minQw = 40 mL/min, Qh = 10 mL/min

Qh, Qw = 10 – 80 mL/min

dp decreases from inlet towardsoutlet

dp = 0.3 – 1.4 mm a = 1000 – 10 000 m-1

kLa = 0.9 - 41 s-1

∆P = 2.5 - 120 kPaRow

1 23456

M. J. Nieves Remacha A. A. Kulkarni

Experimental interFoamCompression

Qw = 40 ml/min, Qh = 10 ml/min

CFD will ultimately allow prediction of mass transfer performance

Qw = 20 ml/min, Qh = 20 ml/min

interFoamCompressionExperimental

26

0

0.2

0.4

0.6

0.8

1

1.2

1 2 3 4 5 6

Co

nve

rsio

n

Residence time (minute)

standard (spiral)

1M NaCl (spiral)

pH=9 (spiral)

Toluene (spiral)

pH=5 (spiral)

pH=9 (corning)0

0.2

0.4

0.6

0.8

1

1.2

0 1 2 3 4 5

Co

nve

rsio

n

Flow rate ratio (QA/QO)

Spiral (τ=2.73min)

Corning (τ=2.75min)

Example of mass transfer effects

• With increasing volumetric aqueous-to-organic phase ratio, the rate of catalyst transfer across the aqueous slug-cap interface increases

• Corning LFR behaves better due to higher mass transfer rate

Y. Zhan, S Born

Multi-step synthesis using solvent switch

• Preparation of aryl triflatein first reaction step

• Extract base and residual reagent by liq-liqextraction

• Solvent switch by single stage microfluidic distillation

• Carry out subsequent reaction step via Heck reaction

28R.L. Hartman, et al., Angew. Chem. Int. Ed., 2010, 49, 899-903

Liquid-LiquidExtraction

enables multistepsynthesis

DPPA on-demand

• Biphasic conditions solve NaN3 solubility.

• Outlet 0.2 M excess NaN3 pH = 9.5 (pKa HN3 = 4.75).

• Membrane separation removes water and salts.

• In-line FTIR allows product stream to be immediately quenched/used

S. Born

90

91

92

93

94

95

96

97

98

99

100

0 10 20 30 40 50 60

Yie

ld %

Time at Steady State (min)

• Corning Low Flow Reactor• 9-12 mL/min tres = 1-1.3 min• vf = 12 mL/min, FDPPA = 1 mol/hr, 6 kg/day• Use in downstream chemistry (Curtius,…)

Scaling-up: Continuous run

S. Born

spiralsampling valve

membranepacked-bed

MFC

O2

N2

Gas-liquid separation

Multistep synthesis of amides directly from alcohols and amines

X. Liu, X.; Jensen, K.F. Green Chemistry 2013 ASAP

Oxidation with O2:

• Lower cost

• Simplify workup

• Catalyst of choice: Ru/Al2O3-5wt%

Packed-bed microreactor

Spiral microreactor

Ru/Al2O3, 80 oC

90-130oC, 20 min

Gas-liquid

separator

Continuous microreactor synthesis

X. Liu, X.; Jensen, K.F. Green Chemistry 2013 ASAP

Catalyst immobilization and recycle• Packed beds – high surface area

supports

– Potential for high catalyst loading

– Physical (e.g., fluorous) and chemical (covalent) attachment to catalyst surface

• Biphasicsystems

– Homogeneous catalysis - catalyst in one phase, substrates in another

– Integration of separators and pumps for continuous operations and recycling

• Nano filtrationLow molecular weight cut-off solvent stable membranes

4 mm 80-240 L

200 m

33

Challenges

– Retain activity of immobilized catalyst

– Leaching of catalyst, especially metal centers , requiring the use of downstream metal capture beds

– Solvent selection

– Sufficiently low and sharp molecular cut-off and solvent stability of membranes

K.D. Nagy and K.F. Jensen, ChimicaOggi/Chemistry Today, 29 (4) (2011).

Recycling of homogeneous catalysts

• Continuous recycling of unmodified homogeneous palladium catalysts via liquid-liquid phase separation

• Continuous catalyst recycling in palladium catalyzedhydroxylation of aryl halides. 5 runs with nearly constant yield of ~80%, but then slow degradation of the catalyst.

Li, P. et al. ChemCatChem 2013

Br OMe HO OMe[Pd] L3

toluene, 2M KOH,

TBAB 5 mol%, 100 oC

P(t-Bu)2

i-Pr i-Pr

i-Pr

OMe

MeO

L3

Continuous nano-filtration and recycle of a metathesis catalyst in a micro flow system

• Evonik Puramem® 280 membrane retains catalyst in the system

• Low co-fee of catalyst to make up for deactivated catalyst

• Continuous ethylene removal to reduce catalyst deactivation

E. O’Neal

Continuous nanofiltration and recycle of a metathesis catalyst in a micro flow system

• TOF ~ 750 h-1 at the start of the experiment

• 10 minute residence time with > 94% product at the reactor outlet

• Total TON of 935 was obtained using this system.

E. O’Neal

0

20

40

60

80

100

120

140

0 15 30 45 60 75

Reactor Volumes, t/

Re

lati

ve P

res

su

re D

rop

Constriction

Bridging

Exploring the origins of clogging

• Handling of solid reagents and products raise challenges for microreactors

R. Hartman et al., Org. Proc. Res. Dev. 2010 14 (6), 1347–1357

Salt by-product

37

Teflon reactor with integrated ultrasound

• Continuous operation without plugging

• Piezo @ 50kHz, 30 Watt

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0 1 10 100 1000

Particle size [m]

Par

ticl

e si

ze f

ract

ion

no sonication

40kHz

50kHz

60kHz

Kuhn, S. et al., Lab Chip 11 2488-92 (2011) 38

Precursors

solventligands∆

• Optical properties and average size depend on factors difficult to control: injection, local temperature and concentration fluctuations, mixing and cooling rates

• Controlled growth of heterostructures difficult to realize, specifically core/shells and graded alloys

PO

Si, Ge, InAs,

InGaN, …

CdSe, CdS, ZnS,

PbSe

Injection

In(MA)3(TMS)3P

Batch synthesis has numerous challenges and limits realization of novel nanostructures

Two stage system : Separate investigations of mixing and aging processes

ConstantAging Temp.320°C

ConstantMixing Temp.170°C

Aging temperature is significant

J. Baek, et al., Angew. Chem. Int. Ed., 2011, 50, 627 –630.

• Residence time : 4 min (fixed)• Solvent, octane at 65 Bar

Reactor design for gradual addition

Side Stream 1

Main Stream

Side Stream 2

Product

• Consideration Distribution of each side flows Viscosity and flowrate changes Residence time after each injection Reactor size restriction Prevent back flows

R1 R2 R3 R4 R5 R6

R8

R7 R9

R10

R11

R12

R13

R14

R15

R16

R17

F2

F1

F3

α∙(F1+F2+F3)

Sequential addition of molecular precursors increase InP QD size

J. Baek, et al., Angew. Chem. Int. Ed., 2011, 50, 627 – 630.

Multistep synthesis core-shell structures

43

InP Growth Investigation

InP / ZnS Core-Shell QD Synthesis

• In(MA)3 : Indium myristate• (TMS)3P : tris(Trimethylsilyl) phosphine

• MA : Myristic acid• Zn(OA)2 : Zinc oleate

• (TMS)2SBistrimethylsilyl sulfur

1. Mixing 2. Aging3. Seq-Growth

4. Shell Formation

44

Synthesis of InP/ZnS core shell structures

ZnS

InP

1. Mixing 2. Aging3. Seq-Growth4. Shell Formation

45

Hierarchical assembly nanostructures for catalysis: nanoparticle synthesis

S.K. Lee, et al., Lab Chip, 2012, 12, 4080 - 4084

Hierarchical assembly nanostructures for catalysis: self assembly

S.K. Lee, et al., Lab Chip, 2012, 12, 4080 - 4084

Hierarchical assembly nanostructures for catalysis: Microreactor application

Oxidation of 4-isopropyl benzaldehyde

S.K. Lee, et al., Lab Chip, 2012, 12, 4080 - 4084

Summary

• Microreactors are coming of age

• Understanding of reaction and mass transfer enables implementation and scaling of flow processes

• Automated tools for extraction of chemical kinetics by using statistical techniques

• Automated optimization based on black box approaches, or better, models with chemical kinetics

• Scaling of homogeneous systems can be accomplished with classical reaction engineering concepts

• Quantitative understanding of heat and mass transfer needed for scaling of multiple phase systems

• Emerging techniques for handling of solids, including the synthesis of nanoparticles

• Ability to synthesize complex micro-nano structures

AcknowledgementsMIT Colleagues:

Moungi G. Bawendi, Peter M. Allen,S. L. Buchwald, Pengfei Li T.F. Jamison, Bo Shen , David Snead

Students and postdocs: Andrea Adamo, Jinyoung Baek, Stephen Born, Patrick Heider, Simon Kuhn, Xiaoying Liu, Amol Kulkarni, Seung-Kon Lee, Jonathan McMullen, Jean Christophe Monbaliu, Jason Moore, Kevin Nagy, Maria-Jose Nieves Remacha, Everett O’Neal, Victor Sebastian, N. Weeranoppanant, Yanjie Zhang

Funding: Corning, DARPA, ISN, NSF, Novartis-MIT Center for Continuous Manufacture,

Thank you!

Questions?

51