Designing Microreactors in Chemical Synthesis – Residence-time Distribution

Designing Microreactors in Chemical

Synthesis –Residence-time Distribution

of Microchannel Devices

KLAUS GOLBIG

ANSGAR KURSAWE

MICHAEL HOHMANN

SHAHRIYAR TAGHAVI-MOGHADAM

THOMAS SCHWALBE

CPC-Cellular Process Chemistry Systems GmbH,

Mainz,Germany

This contribution deals with the application of numerical methods in CPC-Systems’

microreaction development process and demonstrates the feasibility and significant

benefits for the suggested design method.It is focused on the design of capillary resi-

dence tubes,which is necessary if the residence time provided by the original micro-

reactor is not sufficient to complete the reaction.The residence time distribution is

calculated from a straightforward numerical model based on the common assump-

tion that axial gradients can be neglected.The results can be adapted easily to other

capillary diameters or reaction conditions.As an example,the method is applied to

the case of sequential synthesis.

Keywords :Microreactor;Residence time distribution;Sequential synthesis;

Dispersion;Fluid dynamics

Introduction

Synthesis via Microreactors

Microreaction devices are beneficial innovative tools in the improvement and opti-mization of reactions as well as in fast supply of sufficient quantities of target com-pounds.This exciting new technology improves control of reaction parameters such as temperature and concentration equipartition and thus allows in many cases higher yields,fewer by-products,and higher selectivities.Therefore it has become increas-ingly interesting for small-scale production and mobile reaction systems,e.g.,for automotive applications (Wegeng and Drost,1998).

In order to realize these benefits a proper design of the microreaction system is crucial:heat transfer area,mixing channel dimensions,and flow capillaries must be sized very carefully to achieve,for example,reasonable mixing,pressure drops,and

Received 2March 2001;in final form 19June 2003.

Address correspondence to Klaus Golbig,CPC-Systems GmbH,Hanauer Landstrasse 526,G58III,Frankfurt,Mainz 60343,Germany.E-mail:wille@https://www.wendangku.net/doc/b5515871.html,

620

https://www.wendangku.net/doc/b5515871.html,m.,192:620–629,2005

Copyright #Taylor &Francis Inc.

ISSN:0098-6445print =1563-5201online

DOI:10.1080=

00986440590495197

Designing Microreactors in Chemical Synthesis621 flow equipartition(Ehrfeld et al.,1997).To open the field of general application of microreactors in chemical synthesis,a comprehensive analysis of the requirements for such a microreactor has to be performed.Such an analysis identifies the design parameters for multipurpose microreactors,based on a feed rate corresponding to a typical bench-scale synthesis.The resulting solution should be suitable for a variety of applications with reactions or transformations of miscible liquids.

Microreactor Approach to a Better Chemistry

CPC-Systems,Cellular Process Chemistry Systems GmbH,located in Mainz (Germany),is engaged in the development and optimization of standardized modu-lar microreaction systems as well as in their application to organic synthesis.This approach is based on the idea of shortening development periods of new active substances by consequent numbering-up.This means that the same microreactor units are used in laboratory and in pilot plant,thus avoiding the scale-up risk.

Numerous microreactor applications are stated in the current literature that can be considered as proof of the principle of microreaction technology in general. Microreactors have been used in commodity synthesis,i.e.,ethylene oxide(Richter et al.,1998),the preparation of hydrocyanic acid(Hessel et al.,1999)and for poly-merizations.CPC-Systems’modular microreaction system,based on the CYTOS1 microreactor,can be applied to promising fields in drug discovery and development processes.For instance,chemical syntheses of different targets like quinoline acid derivates(Ciprofloxacin1)and the Paal-Knorrpyrrole synthesis have been investi-gated(Taghavi-Moghadam et al.,2001).Further examples from our day-to-day microreactor lab experience could be found in Schwalbe et al.(2002)and Autze et al.(2000);they cover the complete range from common nitrations over rearrange-ments to Suzuki couplings and Wittig-Horner reactions.Furthermore,the CYTOS1 modular microreaction system can be used in a setup for sequential synthesis to generate compound libraries.

In comparison to conventional parallel synthesizers,sequential synthesis offers higher flexibility because the reaction conditions can be controlled independently and individually for each reaction https://www.wendangku.net/doc/b5515871.html,pared to the limited reaction vol-ume in a reaction block of a parallel synthesizer,access to variable amounts of target compounds is ensured by a system running continuously,even if the reaction sequence is demanding or divergent.The automation of the system,which can be established by using an auto-feed-sampler and a fraction collector,leads to perform-ing a maximum of chemistry with minimum effort.

The automated microreaction system also opens up an easy way to automatic reaction optimization.For this purpose a mathematical software generates a statisti-cal design of experiments and evaluates the results automatically via inline analytical sensors.Therefore CPC-Systems’microreaction systems are suited for the pro-duction of specimens as well as for the process development.

Design Model for Capillary Flows

The Problem

CPC-Systems’CYTOS1microreaction system,shown in Figure1,is composed of a pumping module,an exchangeable microreaction unit,and an optional

residence-time providing module,all connected by a convenient bayonet coupling.The residence-time providing unit,for example,a stainless steel capillary,is required for realization of larger reaction times.

For sequential operations the residence-time distributions (RTD)of the micro-reactor and the residence module have to be considered carefully because they may suffer from the laminar flow regime in the capillary tubing,especially if dead volumes are present.On the other hand,diffusive mixing can be very fast on a small scale,so that laminar flow is not necessarily considered a drawback.If more than one capillary is operated in parallel,flow equipartition also becomes an important issue.However,laminar flow regime is known to be modeled very precisely with minor effort.

In our microreaction system corrosion-resistant ceramic piston pumps are used,in contrast to some micro-analytic devices,e.g.,electrophoresis,which are based on electroosmotically driven flows.The velocity profiles differ in both cases.In a pressure-driven capillary,the laminar velocity profile in Figure 2shows a pro-nounced parabolic shape,where the liquid at the wall is nearly stagnant.Influenced by such a velocity profile a concentration peak injected at the inlet will soon

be

Figure 1.CYTOS 1

microreaction system,the first commercial turn-key microreaction system.622K.Golbig et al.

dispersed,yielding a broad residence-time distribution at the outlet.Nevertheless,the molecular diffusion in radial direction limits this peak broadening.In this article design rules consider both effects.

Modeling

There are only a few sources available on this particular problem.In the engineering literature usually only an axial dispersion process is considered (Levenspiel,1958,1999),which was shown first by Taylor (1953)to be sufficient.He developed an analytical solution of the interaction between radial diffusion and axial convection in tubes.On the assumption that axial concentration gradients can be neglected in comparison to radial gradients and the introduction of a relative coordinate system moving with the mean velocity of the fluid,he was successful in reformulating the original problem into a simpler unidimensional axial dispersion process.The role of this process in the case of heterogeneous gas-phase reactions was investigated by Matlosz and coworkers (Commenge et al.,2001).

In order to uncover the details of this peak-broadening phenomenon,a numerical approach was developed that is more flexible regarding geometry and boundary con-ditions.The model uses finite volume balances as well as the following assumptions:

.

Fully developed parabolic velocity profile .

Restriction to radial diffusion .

Conservation of mass (no reaction).

Discretization in axial and radial direction .

Limitation of maximum concentration change to 10%by appropriate choice of D t .Length of one axial discretization element D l

D l ?D t áu max

The program computes the diffusive

c i ;j ;k t1?c i ;j ;k t

D n i à1;k àD n i ;j ;k p D l er i àr i à1Twith D n i ;k ?2p D D l r i c i ;j ;k àc i t1;j ;k r i t1àr i

and the convectional mass transport from cell i,j à1to cell i,j

c i ;j ;k t1?e1àg i Tc i ;j ;k tg i c i ;j à1;k

in an alternating mode.In the equations n denotes amount of moles,c the cell concen-tration,D the diffusion coefficient,r i the outer radial position of the cell i ,and g i

the

Figure 2.Velocity profile and peak broadening inside a capillary tube due to combination of axial convection and radial diffusion.

Designing Microreactors in Chemical Synthesis 623

velocity weighting factor of cell i.The velocity weighting factor g i represents the ratio of fluid leaving cell i,j à1into the next cell i,j downstream during the time period D t .

g i ?1àr 2i tr 2i à12R 2

where R denotes to the inner radius of the capillary.The subscripts contain infor-mation about the radial and axial position of the cell,and the last subscript indicates the numeric operation step.

Simulation Results

As presented in Figure 3,the resulting residence-time distribution of a short capillary for each individual radial position clearly shows that material is short cutting through the center whereas the RTD curves near the wall are delayed.

Generalized RTD Diagram.To avoid time-consuming calculations,the fact is used that,simply put,the calculation procedure is just computing the same figures of dimensionless concentrations in each run.Only the scaling of the time steps differs depending on the diffusion time constants d 2=D .Thus,the resulting dimensionless RTD curve is dependent on only one parameter,the dimensionless diffusion time or Fourier number Fo d :

Fo d ?t mean D

d wher

e t mean is mean residence time,D is diffusivity,and d 2is tube diameter is

1.Figure 3.Residence-time distribution for capillary with 5min mean residence time and

1.75mm inner diameter (D ?6.9á10à10m 2=s).

624K.Golbig et al.

Thus,previously generated output data can be reused for other conditions.Examples for dimensionless RTD curves are illustrated in Figure 4.Each curve represents a set of many different cases that are analogous from a physical point of view.For the required case,only the Fourier number Fo d has to be determined,and the corresponding calculation run can be taken from the output database,which should be arranged in terms of Fo d number.

Dimensionless Dispersion Curve.This straightforward numerical model was vali-dated by experiments with capillaries and toluene as tracer,which was analyzed by gas chromatography (GC).The double logarithmic plot in Figure 5reflects the width of the RTD curve peaks at half of their height showing a minor deviation between the experimental results (plotted as dots)and the numerical calculated dis-persion curve (dashed line),but we found excellent agreement between the theory of Taylor (Levenspiel,1999;Taylor,1953)(gray line)and our calculated dispersion curve.

During the evaluation it became apparent that there is a need for faster solvers that give results for high Fourier numbers Fo d greater than three.One possibility to reach this aim is to exploit the fact that the dispersion process in the laminar flow regime can be regarded as a linear signal transmission process.Thus,the trans-fer functions of two capillaries,which have to be determined first,can be convo-luted.Then the convolution is done from each inlet radial position of the first capillary to each outlet position of the second one by averaging over all possible combinations.

The advantage of such a procedure is that it proceeds exponentially in time or capillary length.Although convolutions are very time-consuming calculations,they offer a speed advantage for high Fo d

numbers.

Figure 4.Generalized RTD curves.

Designing Microreactors in Chemical Synthesis 625

Application to Sequential Synthesis

One major application of microreaction systems is the generation of a variety of substances by automatically exchanging the starting materials (and if necessary also modifying the reaction conditions).A photo of such a sequential synthesis setup using the SEQUOS 2Lab System is given in Figure 6.

An auto-sampler and a fraction collector complete the setup of the CYTOS 1microreaction system.The individual substances are separated by short purge-flushes of solvents.The intensity of these purge-flushes can be modified in order to optimize the process.While too long flushes waste operating time,too short flushes may result in cross-contamination of the different substances.

Figure 7illustrates an example with short pumping and purging periods.Reac-tant A is pumped over a time period of t ?4min 30s,the spacer over 2min 20s.Because of the RTD,the resulting response at the system’s exit after 9min 40s is slightly broadened.The product is collected directly after the elapse of the mean residence time for the adjusted pumping time.While collecting the first reactant A,the next reactant B is already fed into the reactor separated by the spacer.It can be clearly seen that both reactant peaks are nevertheless separated.An inte-gration of the RTD signal yields that 85%of the fed reactants are collected,slightly diluted by spacer solvent.

This example clearly demonstrates that the residence-time distribution of the system –microreactor and residence tubing –has to be taken into account and is a crucial part of the system’s functionality.In the example of Figure 8a longer pumping time was applied in order to obtain a sample of higher quality (i.e.,

less

Figure 5.Dimensionless dispersion curve.

626K.Golbig et al.

dilution with spacing solvent),slightly increasing the spacing time from 140s up to 270s and improving the separation of both sequences.This setup allows the collect-ion of 77%of the whole pulse with only 1.3%dilution by spacing solvent.Theore-tically,it is possible to gather all spent material if the purging time is as long as the total peak width from the Dirac

pulse.

Figure 7.Short cycle times for high throughput (SEQUOS 2microstructured residence time modules having 9min 40s mean residence time,pumping time 270s,spacing time 140

s).Figure 6.Sequential synthesis with SEQUOS 2microreaction system.

Designing Microreactors in Chemical Synthesis 627

Conclusion and Outlook

Thus it can be stated that

.CPC-Systems has successfully implemented a realistic numerical model for residence-time distribution of capillaries

.Residence capillaries have been designed with adjusted trade-off between peak-broadening and pressure drop demands

.Calculated data can be converted easily to actual conditions

.Experiments can be planned accurately in

advance.

Figure 8.Sequential synthesis with longer cycle times for overall higher product quality and better sequence separation (SEQUOS 2microstructured residence time modules having 9min 40s mean residence time,9min 40s pumping time,and 4min 30s spacing

time).Figure 9.Temperature distribution in a flat microchannel setup (70m m wide capillary gap)with neutralization reaction (55kJ =mol,concentration 4800mol =m 3).

628K.Golbig et al.

Designing Microreactors in Chemical Synthesis629 This method will be used as a pre-optimization tool for future microreactor projects at CPC-Systems.Future work aims at the extension to more complex cases and flow structures.An example for this is presented in Figure9,where the tempera-ture distribution in a flat rectangular channel is shown.The channel is fed with strong acid and strong base from the left.Instantaneous release of reaction heat(55kJ=mol)is presumed as well as isothermal wall condition.Due to micro dimensions of the channel,the hot spot is limited to4.2K in this case.

Literature

Autze,V.,Kleemann,A.,Golbig,K.,and Oberbeck,S.(2000).Nachr.Chem.,48,p.683–685. Commenge,J.M.,Falk,L.,Corriou,J.P.,and Matlosz,M.(2001).In5th International Conference on Microreaction Technology:Book of Abstracts,143.

Ehrfeld,W.,Golbig,K.,Hessel,V.,Lo¨we,H.,and Richter,Th.(1997).In VDI-GVC Jahrbuch,102–116,VCH,Weinheim.

Hessel,V.,Ehrfeld,W.,Golbig,K.,and Wo¨rz,O.(1999).GIT Lab.-Fachz.,10,1100–1103. Levenspiel,O.(1958).Ind.Eng.Chem.,50,343.

Levenspiel,O.(1999).Chemical Reaction Engineering,3d ed.,ch.13,296–301,John Wiley, New York.

Richter,Th.,Ehrfeld,W.,Gebauer,K.,Golbig,K.,Hessel,V.,Lo¨we,H.,and Wolf,A.(1998).

In2nd International Conference on Microreaction Technology:Process Miniaturization, Topical Conference Preprints,146–151,American Institute of Chemical Engineers, New York.

Taghavi-Moghadam,S.,Kleemann,A.,and Golbig,K.(2001).Microreaction technology as a novel approach to drug design,process development and reliability,Org.Process Res.&Dev.,5,652–658.

Taylor,G.I.(1953).Proc.Roy.Soc.,219A,186.

Schwalbe,T.,Autze,V.,and Wille,G.(2002).Chimia,56,636–646.

Wegeng,R.S.and Drost,M.K.(1998).In2nd International.Conference on Microreaction Technology:Process Miniaturization,3–11.