barrier-based micro/milli channels reactor

Barrier-based micro/milli channels reactor

Citation for published version (APA):Al-Rawashdeh, M. I. M. (2013). Barrier-based micro/milli channels reactor. Technische Universiteit Eindhoven.https://doi.org/10.6100/IR751990

DOI:10.6100/IR751990

Document status and date:Published: 01/01/2013

Document Version:Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can beimportant differences between the submitted version and the official published version of record. Peopleinterested in the research are advised to contact the author for the final version of the publication, or visit theDOI to the publisher's website.• The final author version and the galley proof are versions of the publication after peer review.• The final published version features the final layout of the paper including the volume, issue and pagenumbers.Link to publication

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, pleasefollow below link for the End User Agreement:www.tue.nl/taverne

Take down policyIf you believe that this document breaches copyright please contact us at:[email protected] details and we will investigate your claim.

Download date: 08. Jan. 2022

https://doi.org/10.6100/IR751990

https://doi.org/10.6100/IR751990

https://research.tue.nl/en/publications/843d47ea-1fbd-45aa-bedc-421bb130f401

Barrier-based Micro/Milli ChannelsReactor

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan deTechnische Universiteit Eindhoven, op gezag van de

rector magnificus, prof.dr.ir. C. J. van Duijn, voor eencommissie aangewezen door het College voor

Promoties in het openbaar te verdedigenop woensdag 15 mei 2013 om 16.00 uur

door

Ma’moun Ibrahim Mohammad Al-Rawashdeh

geboren te Irbid, Jordanie

Dit proefschrift is goedgekeurd door de promotor:

prof.dr.ir. J.C. Schouten

Copromotor:dr.ir. T.A. Nijhuis

Eindhoven University of Technology, 2013A catalogue record is available from the Eindhoven University of Technology LibraryMa’moun Ibrahim Mohammad Al-RawashdehBarrier-based Micro/Milli Channels ReactorISBN: 978-90-386-3363-3

To QumaimTo my family

To my father and mother

Table of Contents

Summary ix

1 Introduction 11.1 Micro/milli channel reactors . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Scaling-up via numbering-up . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Design of Micro/Milli-Reactors for Large Scale Processing - DeMiR project . 5

1.4 Aim and lay-out of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Design criteria for a barrier-based gas-liquid flow distributor for parallel mi-crochannels 132.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.2 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.3 Distributor design parameters . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.3.1 Fabrication tolerance in the diameter of barrier channels . . . . . . . 17

2.3.2 Hydraulic resistance in the barrier channel . . . . . . . . . . . . . . . 19

2.3.3 Pressure difference between the manifolds . . . . . . . . . . . . . . . 20

2.3.4 Gas and liquid flow ranges . . . . . . . . . . . . . . . . . . . . . . . 20

2.3.5 Fabrication tolerance in the inner diameter of mixers and microchannels 21

2.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.4.1 Taylor flow in four microchannels operating in parallel . . . . . . . . 21

2.4.2 Fabrication tolerance in the diameter of barrier channels . . . . . . . 22

2.4.3 The hydraulic resistance in the barrier channel . . . . . . . . . . . . 23

2.4.4 Pressure difference between the manifolds . . . . . . . . . . . . . . . 25

2.4.5 Gas and liquid flow ranges . . . . . . . . . . . . . . . . . . . . . . . 25

2.4.6 Fabrication tolerance in the inner diameter of mixers and microchannels 27

2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

vi Table of Contents

2.6 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3 Design methodology for barrier-based two phase flow distributor 33

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.1.1 2-phase resistive network model . . . . . . . . . . . . . . . . . . . . 37

3.1.2 Design methodology . . . . . . . . . . . . . . . . . . . . . . . . . . 38

3.2 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3.2.1 Influence of the manifold flow non-uniformity factor, σ(qM) . . . . . 43

3.2.2 Influence of the barrier channels flow non-uniformity factor, σ(qB) . 44

3.2.3 Influence of the mixers and reaction channels flow non-uniformityfactor, σ(qC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.2.4 The influence of the three flow non-uniformity factors combined . . . 46

3.2.5 Correlation of σ(∆PC) for Taylor flow regime . . . . . . . . . . . . . 48


3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

3.4 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.4.1 A. 2-phase resistive network model . . . . . . . . . . . . . . . . . . 52

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

4 Numbered-up gas-liquid micro/milli channels reactor with modular flow distrib-utor 59

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

4.1.1 Design and fabrication . . . . . . . . . . . . . . . . . . . . . . . . . 63

4.2 Experiments and operating conditions . . . . . . . . . . . . . . . . . . . . . 65

4.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

4.3.1 Liquid versus gas-liquid flow distribution . . . . . . . . . . . . . . . 69

4.3.2 Stainless steel plate reactor - Effect of physical properties . . . . . . 71

4.3.3 Effect of reaction channel types and dimensions . . . . . . . . . . . . 73

4.3.4 Comparison to single channel - Bubble generation frequency and slugand bubble lengths . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

4.3.5 The BMMR versus other barrier-based flow distributors operated inthe Taylor flow regime . . . . . . . . . . . . . . . . . . . . . . . . . 76

4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

Table of Contents vii

5 Designing flow and temperature uniformities in parallel microchannels reactor 815.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

5.2 Modeling and methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

5.2.1 Temperature non-uniformity . . . . . . . . . . . . . . . . . . . . . . 84

5.2.2 Resistive network model . . . . . . . . . . . . . . . . . . . . . . . . 86

5.2.3 Energy balance model . . . . . . . . . . . . . . . . . . . . . . . . . 86


5.3 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

5.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

5.4.1 Effect of temperature on the pressure drop ratio, ∆PB . . . . . . . . . 92

5.4.2 Effect of temperature on flow distribution . . . . . . . . . . . . . . . 92

5.4.3 Numerical validation of the design methodology- Effect of all com-bined contributions on flow distribution . . . . . . . . . . . . . . . . 95

5.4.4 Effect of flow on temperature deviation . . . . . . . . . . . . . . . . 97

5.4.5 Experimental studies - effect of temperature and flow distributions . . 99

5.4.6 Conclusions and outlook . . . . . . . . . . . . . . . . . . . . . . . . 101

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

6 Phenylacetylene hydrogenation over [Rh(NBD)(PPh3)2]BF4 catalyst in numbered-up microchannels reactor 1056.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

6.2 Experimental section . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108

6.2.1 Catalyst preparation . . . . . . . . . . . . . . . . . . . . . . . . . . 108

6.2.2 Experimental procedure - batch setup . . . . . . . . . . . . . . . . . 108

6.2.3 Experimental procedure - BMMR setup . . . . . . . . . . . . . . . . 109

6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.3.1 Batch setup experiments . . . . . . . . . . . . . . . . . . . . . . . . 111

6.3.2 Initial rate studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.3.3 Kinetic models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

6.3.4 Catalyst deactivation . . . . . . . . . . . . . . . . . . . . . . . . . . 115

6.3.5 Determination of kinetic parameters . . . . . . . . . . . . . . . . . . 116

6.3.6 Parametric study- Kinetic results . . . . . . . . . . . . . . . . . . . . 117

6.3.7 BMMR setup experiments . . . . . . . . . . . . . . . . . . . . . . . 118

6.3.8 Reactor modeling - BMMR . . . . . . . . . . . . . . . . . . . . . . 121

6.3.9 Proof of concept of numbering-up in the BMMR . . . . . . . . . . . 122

6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

6.5 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

viii Table of Contents

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

7 Oxidation of ethylbenzene - Case study to reach bulk scale production usingnumbering-up 1297.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

7.1.1 Maximum external numbering-up of modular units . . . . . . . . . . 1317.1.2 Flow capacity and size of reactor modular unit . . . . . . . . . . . . 1317.1.3 Maximum number of parallel channels in one Barrier-based Micro/milli

Channels Reactor (BMMR) . . . . . . . . . . . . . . . . . . . . . . 1337.1.4 Scale of reaction channels in the BMMR. . . . . . . . . . . . . . . . 134

7.2 Case study - Oxidation of ethylbenzene . . . . . . . . . . . . . . . . . . . . 1367.2.1 Objectives of the case study . . . . . . . . . . . . . . . . . . . . . . 1377.2.2 Configuration and operating conditions for Horizontal Bubble Col-

umn Reactor (HBCR) and BMMR . . . . . . . . . . . . . . . . . . . 1387.2.3 HBCR size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1407.2.4 BMMR size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1417.2.5 Dimensions of BMMR reaction channels . . . . . . . . . . . . . . . 1437.2.6 Comparative result of the HBCR versus BMMR . . . . . . . . . . . 145

7.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

8 Conclusions and outlook 1498.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1498.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

Nomenclature 155

Acknowledgments 159

List of Publications 163

About the Author 166

Summary

Gas-liquid processing in microreactors remains mostly restricted to the laboratory scale witha few exceptional successes at the pilot scale. This is due to the complexity and expenditureneeded for an adequate numbering-up with a uniform flow distribution. It is aimed here toovercome this technical challenge using the barrier-based distributor which is a multiphaseflow distributor that assures good flow uniformity and prevents channeling between the twophases.

Experimentally, design criteria for the barrier-based distributor are obtained in a setupmade of capillaries and tube fittings. The design criteria are illustrated for the case of anitrogen-water Taylor flow in four parallel microchannels. Gas-liquid channeling is preventedat equal pressures in the gas and liquid manifolds. An optimal operational window is realizedwhen the gas to liquid flow ratio is kept constant and the ratio of the maximum over the mini-mum flow rates remain less than 15. The effect of variations in the inner diameters due to thefabrication tolerance of the barrier channels and the microchannels on the flow distribution isdemonstrated.

Numerically, the flow distribution is studied using a method based on hydraulic resistivenetworks (RN). The single phase hydraulic RN model is extended to account for two phasesgas-liquid Taylor flow. For ReGL < 30, the accuracy of the model was above 90% for the cap-illaries and tubing’s setup. The model was used to study the effect of fabrication tolerance onflow distribution and barrier channel dimensions. A design methodology has been proposedto determine the required hydraulic resistance in the barrier channels and cut-off values ofthe maximum allowed fabrication tolerance.

Based on this design methodology, the barrier-based micro/milli reactor (BMMR) is de-signed and fabricated to deliver flow nonuniformity of less than 10%. The BMMR consistsof eight parallel channels operated in the Taylor flow regime with a liquid flow rate up to 150

x Summary

mL/min which is suitable for a production capacity in the order of kg/h. The quality of theflow distribution is considered by studying two aspects. The first aspect is the influence ofdifferent viscosities, surface tensions and flow rates. The second aspect is the influence ofmodularity by testing three different reaction channels types: (1) square channels fabricatedin a stainless steel plate, (2) square channels fabricated in a glass plate, and (3) circular chan-nels (capillaries) made of stainless steel. Additionally, the BMMR is compared to a singlechannel and shows the same performance regarding the slug and bubble lengths and bubblegeneration frequency.

Under a range of flow rates without any reactions, the influence of temperature is studiedin the BMMR. The methodology provides a cut-off value of the maximum allowed temper-ature deviation in each part of the BMMR. Temperature deviation effect on flow distributionis quantified using a hydraulic resistive network model. Flow rate effect on the temperaturedeviation is demonstrated using a one dimensional energy balance model. Experiments in theBMMR were conducted to validate these models. Temperature deviation in the barrier chan-nels affects flow non-uniformity by ten times more than in the mixer and reaction channels.Above a certain critical liquid residence time, the flow rate had no significant effect on thetemperature deviation. The critical liquid residence time depends on the liquid used, BMMRmaterial of construction, and its geometrical dimensions. The design methodology providesengineering steps to give a first estimation on the effect of temperature on flow distribution.

Under reaction conditions, the BMMR is tested using the hydrogenation reaction forphenylacetylene to styrene and ethylbenzene using a homogenous cationic rhodium cata-lysts [Rh(NBD)(PPh3)2]BF4. In a semi-continuous batch reactor, a parametric study wasperformed by changing the hydrogen pressure, catalyst concentration, initial concentrationof phenylacetylene and styrene. The kinetic parameters were estimated by fitting the ki-netic model to the experiments. Catalyst deactivation was observed and incorporated in thekinetic model. The kinetic model predicts the experimental result within an accuracy of20%. Preliminary results for performing the hydrogenation reaction in the BMMR reactorare demonstrated. The reaction was conducted in the BMMR and the reactant and productconcentrations of a single channel were compared to that of eight parallel channels in theBMMR. For 95% of the obtained results, the difference in concentrations between the singleand the eight channels remains within ± 10% and depended on the gas and liquid flow rates.As a proof of concept, the numbering-up concept of gas-liquid Taylor flow in the BMMR hasbeen successfully realized.

As a case study, the oxidation of ethylbenzene was used as a relevant industrial applica-

Summary xi

tion for process intensification using the micro/milli channel reactor. The BMMR was usedto scale-up ethylbenzene oxidation via numbering-up to total flow capacity of 10 m3/h whichequals to 80000 tons/y. The size and operational aspects of the BMMR were compared tothe currently used industrial reactor, the horizontal bubble column reactor HBCR. The sizeof BMMR is five to ten times smaller than the HBCR if the residence time is less than 2minutes. In principle, the BMMR shows its potential capability to scale-up multiphase flowapplications to reach industrial bulk scale capacities in the range of m3/h and m3/min forliquids and gases, respectively. This opens a wide variety of opportunities to further invest inthis field and conduct research to fully exploit this technology.

Introduction 111.1 Micro/milli channel reactors

Micro/milli channel reactor technology emerged in the 1990s as an outcome from the ad-vancements in microfabrication technology driven by the computer revolution. By allow-ing chemical reactions to take place in a confinement with a typical lateral dimensions be-low a few millimeters, the fluid mixing (Nguyen and Wu (2005)), heating (Rebrov et al.(2011)), and reaction (McMullen and Jensen (2010)) rapidly increased, which re-determinedthe boundary of what can/cannot be processed by the chemical industry (Hessel et al. (2009,2005)). For example, micro/milli channel reactors allow ”process intensification” to fulfillits aims by substantially reducing the size of the equipment making it more energy efficient,safer, and/or reducing the waste produced (Stankiewicz and Moulijn (2000)). This can befor single phase processes as well as for multiphase ones like for gas-liquid processes (Den-cic et al. (2012)) which is the focus of this work. Moreover, reaction rates can be safelyaccelerated in the micro/milli channel reactor via the use of ”novel process windows” (hightemperature, high pressure, or alternative heating source) resulting in a process intensifica-tion (Hessel et al. (2011)). Process intensification and the miniaturization of devices (Bayeret al. (2005)) make it possible to construct an entire process in a mobile container which canbe moved according to the required demand (Lang et al. (2011)), thus offering the chemicalindustry with a wider range of flexibility and reduced risks.

The simplest form of a micro/milli channel reactor is shown in Figure 1.1. Two fluids Aand B are pumped through a mixer upstream of a micro/milli channel where heating is sup-plied or removed to deliver the required product C. Generally, micro/milli channel reactorsare operated under laminar flow conditions. It is generally known that laminar flow provides

2 Introduction

Figure 1.1: Concept and fundamentals of the micro/milli channel reactor.

poor mixing quality. However, because of the reduced transport length due to the micro/millicross sectional channel dimensions, the mass-transfer performance of the micro/milli chan-nel reactor is relatively higher, compared to conventional reactors. For example, when thechannel internal diameter is reduced by a factor of four (from 1 mm to 0.25 mm), the diffu-sion time (estimated by the square of the channel diameter divided by the molecular diffusioncoefficient- a value of 1x10−9 m2/s is used) reduced by 16 times (Haber et al. (2012)). Ad-ditionally, heat transfer is also accelerated. The large surface to volume ratio (can reachmore than 10,000 m2/m3) minimizes the required heat exchange area and assures a saferreactor operation with a rapid and precise temperature control. Moreover, the small chan-nel cross sectional dimensions eliminates radial temperature profiles which in the case ofhighly exothermic reactions, avoids any hot spot formation (Haber et al. (2012), Rebrov et al.(2011)). In addition, such advantages allow operating micro/milli channel reactors in theexplosive regime which is virtually impossible to do safely using conventional reactor tech-nology (Chen et al. (2011), Leclerc et al. (2008)). The recombination of propagating radicalsat the channel walls prevents ignition of the mixture, which makes this type of reactor safer.As well, the enhanced heat transfer capability of micro/milli channel reactors minimizes thechance for thermal explosion. Micro/milli channel reactors also reduce axial dispersion, spe-cially when operated in the Taylor flow regime which prevents back mixing thus allowing thereactor to operate as an ideal plug flow reactor (Gunther et al. (2005), Salman et al. (2005)).

Micro/milli channel reactors operate continuously which can have several advantages

Scaling-up via numbering-up 3

over batch production. For example, when a chemical route contains toxic or radical interme-diates, multi-step synthesis can be performed and the intermediate chemicals will only existfor a short time until the reaction proceeds to the consecutive step (Allian et al. (2011), Wilesand Watts (2012)). Moreover, an addition of a new chemical can be precisely added at thelocation in the reactor where it is required. Thus, the newly added chemical will never mixwith any of the chemicals upstream of the addition point. Additionally, heat managementcan be better utilized in a continuous production. In the case of a process which containsendothermic and exothermic reactions, heat integration can be coupled which results in auto-thermal operation (Delsman et al. (2004)).

Micro/milli channel reactors are a vital tool at the laboratory scale especially when usedin process automation modes(McMullen and Jensen (2011)). Kinetic studies can be madewithout any worry about the influence of external mass and heat transfers. It reduces theamount of chemical waste and can optimize the reaction much faster than batch processes. Ifat the laboratory scale a reaction was optimized in a single micro/milli channel reactor, largerindustrial scale production can be reached without any need to re-optimize the system. This isachieved via numbering-up or scale-out, i.e. assembling several micro/milli channel reactorsin parallel. Scaling-up via numbering-up maintains the improved mass and heat transfer char-acteristics of the laboratory scale (Hartman and Jensen (2009), Hessel et al. (2005), Masonet al. (2007), Mills et al. (2007)). Thus, numbering-up allows industrial production directlyfrom the laboratory scale. This will shorten the time to reach to the market which is crucialfor applications in the pharmaceutical industry.

1.2 Scaling-up via numbering-up

A major consequence from using a micro/milli channel reactor is the small production capac-ity. A typical flow rate of a micro/milli scale channel is in the range of mL/min. To reach alarger production capacity, multiple channels can be arranged in parallel via numbering-up,depending on the target production capacity. In Figure 1.2 the numbering-up is demonstratedfor three different target production capacities. For a bulk scale production which corre-sponds to a flow rate of 1000 L/h, three ranges can be identified based on the flow rate perchannel: (1) Flow rates less than a few mL/min, which will require excessive numbering-upof thousands even million of channels; (2) Flow rates equals to tenths of a mL/min, whichwill require a moderate numbering-up of hundreds of channels; and (3) flow rates aroundL/min, which will need little numbering-up of few channels. Such a classification helps toanticipate the corresponding challenges associated with the numbering-up.

4 Introduction

Figure 1.2: Number of parallelization versus the flow rate for different production capacities.

A route is suggested to reach the bulk scale production as shown in Figure 1.3. This routewill serve as a backbone to explain the objectives and layout of this thesis. The scaling routeconsists of three major steps. First, the scale-up of a single channel by smartly increasing thecross sectional dimension of the channel while keeping the mixing and heating characteristicsof the microchannel scale. Second is numbering-up, placing multiple channels in parallel, ina modular unit. The modular unit is defined as a device which contains different functionalelements such as: distributor, mixer, reaction channels, heat exchanger and separator, andis fed by a single feed for each phase as shown in Figure 1.3. The third step is externalnumbering-up by arranging multiple modular units in parallel.

Numbering-up is an essential step if micro/milli reactor technology will be used for bulkscale production and depends strongly on the phases used. For heterogeneously catalyzed gasphase reactions, scale-up and numbering-up to thousands of parallel channels has been suc-cessfully demonstrated (Hessel et al. (2008), Lerou et al. (2010), Tonkovich et al. (2004)). Forliquid phase reactions, scale-up has been successfully demonstrated, but only with a limitednumber of parallel channels due to the complexity of flow distribution (Roberge et al. (2009),Saber et al. (2009), Schenk et al. (2003)). For multiphase flow, the progress in numbering-upis still restricted to smaller flow capacities and depends on the contacting of the phases. For

Design of Micro/Milli-Reactors for Large Scale Processing - DeMiR project 5

Figure 1.3: Scheme for the route of scale-up via numbering-up for micro/milli channel reac-tors. (i) scale-up of a single channel, (ii) modular unit, (iii) multi-modular units.

a continuous contact between the phases such as in the falling film microreactor, a largerproduction capacity of hundreds of mL/min has already been demonstrated (check stackedfalling film microreactor from the Institut fur Mikrotechnik Mainz). For dispersed contact-ing, numbering-up is still mostly restricted at the laboratory scale of mL/min (Chambers et al.(2005), Kashid et al. (2010), Mendorf et al. (2010), Natividad et al. (2007), Yue et al. (2010)).This is due to the complexity and expenditure needed to assure equal flow conditions.

Currently, most progress toward larger scale production in liquid phase applications fo-cuses in the direction of minimizing numbering-up. Instead the focus is on ”smart scale-up”of the single channel. This is achieved by using smart micro structures in a single channelwhich allows maximizing the internal channel cross sectional dimensions while keeping thegood mixing and heating of the micro channel scale, thus increasing the flow capacity ofthe single channel. Such a practical approach is hardly transferable to multiphase flow - es-pecially to gas-liquid systems, as this would lead to a change in the flow regime, the mostcrucial engineering point here. Moreover, the approach to accelerate the reaction rate viaincreasing the temperature and pressure (using the concept of novel process window) whichis used successfully for liquid phase is still largely un-explored for gas-liquid phases.

1.3 Design of Micro/Milli-Reactors for Large Scale Process-ing - DeMiR project

The research described in this thesis is part of a project funded by the Dutch TechnologyFoundation (STW) and the Industrial Advisory Board (IROP) of the Netherlands ResearchSchool in Process Technology (OSPT) under the name: Design of Micro/Milli-Reactors forLarge Scale Processing - DeMiR. The objective of the DeMiR project is to develop a generic

6 Introduction

methodology or strategy for selecting and designing the best scale of reactor operation, eitherat the ”micro-scale” or at the ”milli-scale”, in case of G/L and L/L catalytic reactions andmultiphase food processing systems suitable for bulk scale production, viz. in the range ofm3/h and m3/min for liquids and gases, respectively.

Within the DeMiR project, three universities (Eindhoven University of Technology, DelftUniversity of Technology, and Wageningen University) and industrial partners are cooperat-ing. Two other PhD projects were granted at these universities. Delft University is studyingthe computational fluid dynamics and Wageningen University is exploring the prevention offouling in microchannels. The focus at Eindhoven University was on designing, building,and testing a micro/milli-reactor prototype which can be suited for G/L or L/L catalyzedreactions. At the end of the project, the gained knowledge will be used to evaluate if multi-phase flow systems can reach bulk scale productions using micro-process technology.

1.4 Aim and lay-out of the thesis

The objective of this thesis and its layout will be demonstrated based on the suggested scal-ing route in Figure 1.3. The scaling route in Figure 1.3 contains different challenges. In thisthesis the focus will be on two of them, the numbering-up of channels in the modular unit andthe design of the modular unit. The numbering-up challenge aims at developing a multiphaseflow distributor, which can be used to maintain flow non-uniformity below an acceptablelimit. The design challenge of the modular unit focuses on how to successfully integrate allof these functional devices (distributor, mixer, heat exchanger, and reaction channels) intoone unit. This modular unit should successfully demonstrate the micro/milli-reactor proto-type which can be suited for G/L or L/L catalyzed reactions for bulk scale production.

When mixing multiphase flow in a microchannel, many flow patterns can be generateddepending on the flow rates and the channel dimensions as shown in Figure 1.4 (Shao et al.(2009)). In this thesis, the focus will be on one flow regime, Taylor flow (Angeli and Gavri-ilidis (2008), Gunther et al. (2005), Kreutzer et al. (2005)). Taylor flow is attractive due toits well defined gas-liquid interface, reduced axial dispersion due to the separation of liquidslugs by gas bubbles approaching ideal plug flow (Kreutzer et al. (2008), Pedersen and Hor-vath (1981), Salman et al. (2005)). Furthermore, the improved radial mixing in the slugs andbubbles further improves the mass and heat transfer characteristics (He et al. (2010), Iran-doust and Andersson (1989), Kececi et al. (2009), Narayanan and Lakehal (2008)). Taylorflow regime will be used as a case study to demonstrate the objective of this thesis. The final

Aim and lay-out of the thesis 7

Figure 1.4: Flow patterns for gas-liquid flow depending on the gas and liquid flow rates. Themodified image is taken from Shao et al. (2009).

aim is to develop a generic design methodology that can be safely transferred to other flowregimes. This argument will be evaluated at the end of this thesis.

The barrier-based distributor is a multiphase flow distributor for a multichannel microre-actor which assures flow uniformity and prevents channeling between the two phases. Inchapter two, the concept of the barrier-based flow distributor and its design criteria are pre-sented. In an experimental setup consisting of capillaries and tube fittings, design rules areextracted over the required barrier channels dimensions to remain below a target flow non-uniformity. The design criteria are illustrated for the case of a nitrogen-water Taylor flow (1< ReGL < 30 and 3x10−5 < CaGL < 4x10−4) in four parallel microchannels of 0.9 mm innerdiameter.

In chapter three, a generic design methodology for the numbering-up is presented. Thebarrier-based flow distributor is studied numerically using a method based on the hydraulicresistive network model. The single phase hydraulic resistive network model from Com-menge et al. (2002) is extended for two phases (gas-liquid) Taylor flow. For ReGL < 30, theaccuracy of the model in comparison to the experimental result from the capillary setup wasabove 90%. The developed model was used to study the effects of fabrication tolerance andbarrier channel dimensions. A design methodology has been proposed as an algorithm todetermine the required hydraulic resistance in the barrier channels and the maximum allowedfabrication tolerance in each part of the reactor. This methodology is demonstrated using a

8 Introduction

numerical example.

The flow distribution challenge is solved in chapter two and chapter three. In chapterfour, the challenge of building a modular unit is undertaken. Based on the barrier-based dis-tributor, the barrier-based micro/milli reactor (BMMR) is designed and fabricated to deliverflow non-uniformity of less than 10%. The BMMR consists of eight parallel channels alloperated in the Taylor flow regime and with a total liquid flow rate up to 150 mL/min. Themodularity of the BMMR is demonstrated by testing three different reaction channels type:(1) square channels fabricated in a stainless steel plate, (2) square channels fabricated in aglass plate, and (3) circular channels (capillaries) made of stainless steel. The quality of theflow distribution was quantified under the influence of different viscosities, surface tensionsand flow rates. Finally, the BMMR was compared to a single channel regarding slug andbubble lengths and bubble generation frequency.

In all of the previous studies, experimental investigations were conducted at ambient con-ditions. In chapter five, the effects of temperature on flow distribution are presented and visaversa. Using a hydraulic resistive network model, the effect of temperature on the flow dis-tribution is quantified. The effect of flow on the temperature deviation is demonstrated usinga one dimensional energy balance. Experiments in the BMMR were conducted to validatethese models. This chapter concludes by proposing a design methodology to determine themaximum allowed temperature deviation in each part of the reactor, to maintain flow non-uniformity below an acceptable limit.

In chapter six, the BMMR is demonstrated in a real case scenario by conducting hydro-genation of phenylacetylene to styrene and ethylbenzene using homogenous cationic rhodiumcatalyst [Rh(NBD)(PPh3)2]BF4. First, a parametric study in a semi continuous batch reac-tor is performed by changing the hydrogen pressure, catalyst concentrations, initial concen-trations of phenylacetylene and styrene. A mechanism for this reaction system has beenextended from the work of Esteruelas et al. (1998). Kinetic parameters were estimated byfitting the kinetic model to the batch experiments. Catalyst deactivation was observed andincorporated in the kinetic model. The kinetic model predicts the experimental result withinan accuracy of 20%. One of the results were selected to demonstrate the performance of thereaction using the BMMR. The reactant and products concentrations from a single channelwere compared to the outlet from eight parallel channels combined. A proof of concept overthe capability of the BMMR to number-up multiphase flow under reactive flow conditions isdemonstrated.

Bibliography 9

In chapter seven, the BMMR capability to fulfill the outlined objective of the DeMiRproject is evaluated. The oxidation of ethylbenzene as an industrial application is used for thisevaluation. Chapter eight summarizes the most important research results and conclusionspresented in this thesis.

Bibliography

Allian, A. D., Richter, S. M., Kallemeyn, J. M., Robbins, T. A., Kishore, V., 2011. Thedevelopment of continuous process for alkene ozonolysis based on combined in situ FTIR,calorimetry, and computational chemistry. Organic Process Research & Development 15,91–97.

Angeli, P., Gavriilidis, A., 2008. Hydrodynamics of Taylor flow in small channels: a review.Proceedings of the Institution of Mechanical Engineers, Part C: Journal of MechanicalEngineering Science 222, 737–751.

Bayer, T., Jenck, J., Matlosz, M., 2005. IMPULSE - a new approach to process design.Chemical Engineering & Technology 28, 431–438.

Chambers, R. D., Fox, M. A., Holling, D., Nakano, T., Okazoe, T., Sandford, G., 2005.Elemental fluorinepart 16. versatile thin-film gas-liquid multi-channel microreactors foreffective scale-out. Lab on a Chip 5, 191–198.

Chen, J., Halin, S. J. A., Schouten, J. C., Nijhuis, T. A., 2011. Kinetic study of propyleneepoxidation with H2 and O2 over Au/Ti-SiO2 in the explosive regime. Faraday Discuss.152, 321–336.

Commenge, J. M., Falk, L., Corriou, J. P., Matlosz, M., 2002. Optimal design for flow uni-formity in microchannel reactors. AIChE Journal 48, 345–358.

Delsman, E. R., de Croon, M. H. J. M., Pierik, A., Kramer, G. J., Cobden, P. D., Hofmann,C., Cominos, V., Schouten, J. C., 2004. Design and operation of a preferential oxidationmicrodevice for a portable fuel processor. Chemical Engineering Science 59, 4795 – 4802.

Dencic, I., Hessel, V., de Croon, M. H. J. M., Meuldijk, J., van der Doelen, C. W. J., Koch, K.,2012. Recent changes in patenting behavior in microprocess technology and its possibleuse for gas-liquid reactions and the oxidation of glucose. ChemSusChem 5, 232–245.

Esteruelas, M. A., Gonzalez, I., Herrero, J., Oro, L. A., 1998. Kinetic studies on the selec-tive hydrogenation of phenylacetylene catalyzed by [Rh(NBD)(PPh3)2]BF4 (NBD=2,5-norbornadiene). Journal of Organometallic Chemistry 551, 49 – 53.

Gunther, A., Jhunjhunwala, M., Thalmann, M., Schmidt, M. A., Jensen, K. F., 2005. Mi-cromixing of miscible liquids in segmented gas-liquid flow. Langmuir 21, 1547–1555.

Haber, J., Kashid, M. N., Renken, A., Kiwi-Minsker, L., 2012. Heat management in singleand multi-injection microstructured reactors: Scaling effects, stability analysis, and role ofmixing. Industrial and Engineering Chemistry Research 51, 1474–1489.

Hartman, R. L., Jensen, K. F., 2009. Microchemical systems for continuous-flow synthesis.Lab on a Chip 9, 2495–2507.

He, Q., Hasegawa, Y., Kasagi, N., 2010. Heat transfer modelling of gas-liquid slug flowwithout phase change in a micro tube. International Journal of Heat and Fluid Flow 31,126 – 136.

10 Introduction

Hessel, V., Cortese, B., de Croon, M. H. J. M., 2011. Novel process windows - concept,proposition and evaluation methodology, and intensified superheated processing. ChemicalEngineering Science 66, 1426 – 1448.

Hessel, V., Knobloch, C., Loewe, H., 2008. Review on patents in microreactor and microprocess engineering. Recent Patents on Chemical Engineering 1, 1–16.

Hessel, V., Loeb, P., Loewe, H., 2009. Industrial Microreactor Process Development up toProduction in V. Hessel, A. Renken, J.C. Schouten, J. Yoshida. Micro Process Engineering: A Comprehensive Handbook. Wiley-VCH.

Hessel, V., Loewe, H., Schoenfeld, F., 2005. Micromixers–a review on passive and activemixing principles. Chemical Engineering Science 60, 2479–2501.

Irandoust, S., Andersson, B., 1989. Simulation of flow and mass transfer in Taylor flowthrough a capillary. Computers and Chemical Engineering 13, 519–526.

Kashid, M., Gupta, A., Renken, A., Kiwi-Minsker, L., 2010. Numbering-up and mass transferstudies of liquid-liquid two-phase microstructured reactors. Chemical Engineering Journal158, 233 – 240.

Kececi, S., Worner, M., Onea, A., Soyhan, H. S., 2009. Recirculation time and liquid slugmass transfer in co-current upward and downward Taylor flow. Catalysis Today 147, S125– S131.

Kreutzer, M. T., Gunther, A., Jensen, K. F., 2008. Sample dispersion for segmented flow inmicrochannels with rectangular cross section. Analytical Chemistry 80, 1558–1567.

Kreutzer, M. T., Kapteijn, F., Moulijn, J. A., Heiszwolf, J. J., 2005. Multiphase monolithreactors: Chemical reaction engineering of segmented flow in microchannels. ChemicalEngineering Science 60, 5895–5916.

Lang, J., Stenger, F., Richert, H., 2011. Small is beautiful. Elements 37, 12–17.Leclerc, A., Alame, M., Schweich, D., Pouteau, P., Delattre, C., de Bellefon, C., 2008. Gas-

liquid selective oxidations with oxygen under explosive conditions in a micro-structuredreactor. Lab on a Chip 8, 814–817.

Lerou, J. J., Tonkovich, A. L., Silva, L., Perry, S., McDaniel, J., 2010. Microchannel reactorarchitecture enables greener processes. Chemical Engineering Science 65, 380 – 385.

Mason, B. P., Price, K. E., Steinbacher, J. L., Bogdan, A. R., McQuade, D. T., 2007. Greenerapproaches to organic synthesis using microreactor technology. Chemical Reviews 107,2300–2318.

McMullen, J. P., Jensen, K. F., 2010. Integrated microreactors for reaction automation: Newapproaches to reaction development. Annual Review of Analytical Chemistry 3, 19–42.

McMullen, J. P., Jensen, K. F., 2011. Rapid determination of reaction kinetics with an auto-mated microfluidic system. Organic Process Research & Development 15, 398–407.

Mendorf, M., Nachtrodt, H., Mescher, A., Ghaini, A., Agar, D. W., 2010. Design and controltechniques for the numbering-up of capillary microreactors with uniform multiphase flowdistribution. Industrial and Engineering Chemistry Research 49, 10908–10916.

Mills, P. L., Quiram, D. J., Ryley, J. F., 2007. Microreactor technology and process minia-turization for catalytic reactions–a perspective on recent developments and emerging tech-nologies. Chemical Engineering Science 62, 6992–7010.

Narayanan, C., Lakehal, D., 2008. Two-phase convective heat transfer in miniature pipesunder normal and microgravity conditions. Journal of Heat Transfer 130, 074502.

Natividad, R., Cruz-Olivares, J., Fishwick, R. P., Wood, J., Winterbottom, J. M., 2007.Scaling-out selective hydrogenation reactions: From single capillary reactor to monolith.

Bibliography 11

Fuel 86, 1304 – 1312.Nguyen, N.-T., Wu, Z., 2005. Micromixers - a review. Journal of Micromechanics and Mi-

croengineering 15, R1–R16.Pedersen, H., Horvath, C., 1981. Axial dispersion in a segmented gas-liquid flow. Industrial

and Engineering Chemistry Fundamentals 20, 181–186.Rebrov, E. V., Schouten, J. C., de Croon, M. H. J. M., 2011. Single-phase fluid flow dis-

tribution and heat transfer in microstructured reactors. Chemical Engineering Science 66,1374–1393.

Roberge, D. M., M., G., Eyholzer, M., Kockmann, N., 2009. Industrial design, scale-up, anduse of microreactors. Chemistry Today 27, 8–11.

Saber, M., Commenge, J. M., Falk, L., 2009. Heat-transfer characteristics in multi-scale flownetworks with parallel channels. Chemical Engineering and Processing: Process Intensifi-cation 49, 732–739.

Salman, W., Angeli, P., Gavriilidis, A., 2005. Sample pulse broadening in Taylor flow mi-crochannels for screening applications. Chemical Engineering and Technology 28, 509–514.

Schenk, R., Hessel, V., Hofmann, C., Loewe, H., Schoenfeld, F., 2003. Novel liquid-flowsplitting unit specifically made for numbering-up of liquid/liquid chemical microprocess-ing. Chemical Engineering & Technology 26, 1271–1280.

Shao, N., Gavriilidis, A., Angeli, P., 2009. Flow regimes for adiabatic gas-liquid flow inmicrochannels. Chemical Engineering Science 64, 2749 – 2761.

Stankiewicz, A. I., Moulijn, J. A., 2000. Process intensification: transforming chemical engi-neering. Chemical Engineering Progress 96(1), 22–34.

Tonkovich, A., Perry, S., Wang, Y., Qiu, D., LaPlante, T., Rogers, W. A., 2004. Microchannelprocess technology for compact methane steam reforming. Chemical Engineering Science59, 4819 – 4824.

Wiles, C., Watts, P., 2012. Continuous flow reactors: a perspective. Green Chemistry 14,38–54.

Yue, J., Boichot, R., Luo, L., Gonthier, Y., Chen, G., Yuan, Q., 2010. Flow distribution andmass transfer in a parallel microchannel contactor integrated with constructal distributors.AIChE Journal 56, 298–317.

Design criteria for abarrier-based gas-liquid flowdistributor for parallelmicrochannels 22

This chapter has been published as:Al-Rawashdeh, M., Fluitsma, L.J.M., Nijhuis, T.A., Rebrov, E.V., Hessel,V. & Schouten, J.C. 2012. Design criteria for a barrier-based gas-liquidflow distributor for parallel microchannels. Chemical Engineering Journal,181-182, 549-556.

AbstractThis chapter presents criteria for the design of a flow distributor for even distribution of gasand liquid flows over parallel microchannels. The design criteria are illustrated for the caseof a nitrogen-water Taylor flow (1 < ReGL < 30 and 3x10−5 < CaGL < 4x10−4) in fourparallel microchannels of 0.9 mm inner diameter. The distributor consists of a gas manifold,a liquid manifold, four barrier channels for the gas and four for the liquid, and four T-mixersfor mixing of the gas and liquid flows. The four barrier channels have equal inner diametersand length; four different diameters have been studied: 0.05, 0.1, 0.15 and 0.2 mm. Uniformdistribution of the gas and liquid flows over the microchannels is achieved when the pressuredrop over the barrier channels is in the range of around 4 to 25 times the pressure drop overthe corresponding T-mixers and microchannels. Gas-liquid channeling is prevented at equalpressures in the gas and liquid manifolds. An optimal operational window is realized whenthe gas to liquid flow ratio kept constant and the ratio of the maximum over the minimumflow rates remain less than 20. The effect of variations in the inner diameters (result of thefabrication process) of the barrier channels and the microchannels on the flow distribution isdemonstrated. It is suggested that these design criteria can also be applied at larger numbersof parallel microchannels.

14 Design criteria for a barrier-based gas-liquid flow distributor

2.1 Introduction

Microchannel reactors are considered prime-enabling technologies for flow chemistry pro-cessing to achieve process intensification. For some applications, operations in microchan-nel reactors demonstrate an order-of-magnitude increase in reaction speed (Anxionnaz et al.(2008), Hessel (2009), Kockmann and Roberge (2009), Razzaq et al. (2009), Verboom (2009)).Multi-phase (gas-liquid) flow in a microchannel results in different flow regimes (Chen et al.(2006), Shao et al. (2009), Triplett et al. (1999)). Among them, Taylor flow (Angeli andGavriilidis (2008), Gunther et al. (2005), Kreutzer et al. (2005b), Shao et al. (2009)) is attrac-tive due to its well defined gas-liquid interface, reduced axial dispersion approaching idealplug flow, (Kreutzer et al. (2008), Pedersen and Horvath (1981), Salman et al. (2005)) andits excellent mass transfer (Irandoust and Andersson (1989), Kececi et al. (2009)) and heattransfer characteristics (He et al. (2010), Narayanan and Lakehal (2008)). Scaling conven-tional multiphase reactors from lab to industrial scale requires a considerable effort due tothe challenges in keeping similar hydrodynamics and mass and heat transfer characteristicsat different reactor scales. As opposed to traditional reactors, microchannel reactors keepthe same reactor scale and achieve high throughput by numbering-up or scale-out, i.e. as-sembling several microchannel reactors in parallel. Microchannel reactors thus provide analternative scaling route in which they maintain their improved mass and heat transfer char-acteristics (Hartman and Jensen (2009), Hessel et al. (2005), Mason et al. (2007), Mills et al.(2007)).

For industrial scale applications, numbered-up microchannel reactors have so far onlybeen demonstrated for single-phase reactions (Tonkovich et al. (2005)). Exception of thefalling film microreactor (Hessel et al. (2009)), numbering-up for multi-phase operation re-main mostly restricted to the laboratory scale. For example, De Mas et al. (2005) has showna successful numbering-up of gas-liquid in a silicon-based microreactor that consists of 60reaction channels with a flow non-uniformity of less than 10%. However the reactor is onechip composed of 3 stacked plates and has dimensions of less than 6 cm. It is operated ata liquid throughput of 80 mL/h. The distribution of phases, e.g. gas and liquid, to parallelchannels remains a real challenge, since a poor distributor design can result in some channelsbeing filled only with liquid and others with gas, a phenomenon referred to as ”gas-liquidchanneling” (De Mas et al. (2005)). Multiphase flow distribution can be realized via eitherbranching (Link et al. (2004)), internal distribution (Kreutzer et al. (2005a)) or external distri-bution (Kashid et al. (2010)) as shown in Figure 2.1. Numbered-up microchannel reactors areoften based on external distribution (Chambers et al. (2005), De Mas et al. (2005), Haverkampet al. (2006), Losey et al. (2002, 2001), Wada et al. (2006), Yue et al. (2010)).

Introduction 15

Figure 2.1: Schematic representation of gas-liquid flow distribution over parallel microchan-nels operated in the Taylor flow regime. Distributor types are (A) branching, (B) internaldistribution and (C) external distribution. The single phase flow distributor of (C) is shownin Figure 2.2.

External distributor, which is the focus of this work, is based on mixing equally dis-tributed gas and liquid flows at the entrance of each of the parallel channels, as shown inFigure 2.1. The distribution of each phase over the parallel microchannels is realized in asingle-phase flow distributor (Rebrov et al. (2011)) such as a consecutive manifold struc-ture (Commenge et al. (2002)), tree branching (Bejan and Lorente (2006), Tondeur and Luo(2004)), large manifold volume (Schenk et al. (2003)), perforated meshes and grids (thin-wallscreen) (Idelchik (1991)), or thick wall screens (Rebrov et al. (2007a,b)). The design princi-ple for single phase-flow distributors relies on controlling the hydraulic resistance, expressedby the pressure drop, in each parallel channel. The hydraulic resistance for multiphase flowdoes not only depend on the physical properties of the fluids, flow rates and on the hydraulicdiameter of the microchannel as in single phase flow, but also on the gas to liquid flow ra-tio, specific gas-liquid interfacial area and the flow regime (Chen et al. (2006), Shao et al.(2009), Triplett et al. (1999)). Most single phase flow distributors fail to prevent gas-liquidchanneling when implemented for two-phase flow distribution (Chambers et al. (2005), Yueet al. (2010)). Channeling can be prevented when operating in the annular flow regime, whichrequires gas flows of 10 to 100 times larger than that of the liquid flow. Taylor flow is oftenpreferred over annular flow, due to the much lower gas velocities.

When the hydraulic resistances, the ”barrier channels”, are placed between the singlephase flow distributors and the parallel microchannels as shown in Figure 2.2, (1) gas-liquidchanneling is prevented, (2) all flow regimes, viz. Taylor, churn and annular can be success-fully realized, and (3) the flow uniformity is significantly improved to more than 80%(De Mas


et al. (2005), Haverkamp et al. (2006), Wada et al. (2006)).

Figure 2.2: Schematic representation of barrier-based gas-liquid flow distributor for fourparallel microchannels. Symbols used are: (G) gas, (L) liquid, (M) manifold, (B) barrierchannel, and (T) T-mixer.

The aim of this work is to provide a criteria for the design of a gas-liquid flow distributorfor parallel microchannels based on the concept of barrier channels. The optimal distribu-tor design is the one that prevents gas-liquid channeling, minimizes non-uniformity of thegas and liquid flows and has a minimum hydraulic resistance. In this chapter the optimaldistributor design is obtained via an experimental parametric approach in which five designparameters are evaluated: (1) fabrication tolerance in the diameter of barrier channels (varia-tion in the inner channel diameter due to the fabrication process); (2) the hydraulic resistance(i.e. diameter) of the barrier channel; (3) the hydraulic resistance of the barrier channels onthe gas side compared to the one on the liquid side; (4) the maximum gas and liquid flowranges for the distributor; and (5) fabrication tolerance in the inner diameter of the mixersand microchannels.

The following sections describe the experimental setup, followed by the experimental ap-proach to investigate the effect of the five design parameters. Afterwards, results are providedfor variation of the gas and liquid flow distribution, followed by the results for each of thefive design parameters. Finally, criteria for the design of an optimal distributor design areprovided.

2.2 Experimental setup

The experimental studies were performed using nitrogen-water flow in a setup consisting of4 parallel microchannels. The setup, shown in Figure 2.3, consists of four main parts: (M)

Distributor design parameters 17

gas and liquid manifolds, (B) barrier channels, (T) T-mixers, and (C) microchannels; all builtfrom capillaries and tube fittings. The gas and liquid manifolds are made of Swagelok stain-less steel tube fittings with an inner diameter of 7.1 mm. The barrier channels are PEEKcapillaries (IDEX) of 5 cm length. Four different barrier channels are evaluated with innerdiameters of 0.05, 0.10, 0.15 and 0.20 mm. The T-mixers (Valco Instruments) are made ofstainless steel with an inner channel diameter of 1 mm. The microchannels (see Figure 2.3,C) have a nominal diameter of 0.90 ± 0.05 mm and are 40 cm in length. Water is fed by anHPLC pump (see Figure 2.3, I) type 305 (Gilson) and nitrogen is fed via a mass flow con-troller (see Figure 2.3, II) type F-200CV (Bronkhorst High-Tech B.V.).

The pressure is measured using pressure sensors type 26PC (Sensortechnics GmbH) withthree ranges of 0-50 mbar, 0-1 bar and 0-16 bar relative to atmospheric pressure with a relativeaccuracy of 1, 0.5 and 0.5%, respectively. The liquid flow rate per channel is measured byweighing the collected liquid over time (see Figure 2.3, III). The gas flow rate is calculatedusing the Hagen-Poiseuille equation based on the measured pressure drop over the barrierchannels at the gas side. The slug and bubble sizes are determined from image analysis.Images are taken using a digital SLR camera (Canon 400D) with 10.1 megapixels. Usingthe image analysis toolbox in Matlab, the slug and bubble sizes are calculated by countingthe number of pixels. For each image the pixel size (one pixel corresponds to 10 µm) iscalibrated relative to the outer diameter of the capillary. The start-up procedure is as follows:the gas flow is switched on until a steady-state manifold pressure is reached. Subsequently,the liquid flow is started. After 30 min, constant pressure in both manifolds is obtained andmeasurement commenced.

2.3 Distributor design parameters

2.3.1 Fabrication tolerance in the diameter of barrier channels

The first design parameter is the maximum allowed variation in the inner diameters of thebarrier channels. The inner diameter of the barrier channels were varied by choosing dif-ferent four barrier channels from a larger set of barrier channels. The inner diameters of allbarrier channels were determined using single phase flow experiments. The liquid manifoldand the liquid barrier channels were connected, while the mixer, microchannels and the gasmanifold were removed. The pressure sensors were re-positioned upstream of the liquid bar-rier channels, similar to their position at the gas side in Figure 2.3. The liquid flow rate andthe pressure drop were measured per channel. The nominal hydraulic diameter per barrierchannel was calculated using Hagen-Poiseuille equation, assuming a constant diameter over


Figure 2.3: Top: Image of the main components of the experimental setup. Bottom: Flowdiagram of the experimental setup, (I) liquid pump, (II) mass flow controller, (III) liquidcollector, (PS) pressure sensors shown at channel 1 as an example on their locations. Thesymbols (M), (B), and (T) are presented in Figure 2.2; (C) is the microchannel. (i), (ii) and(iii) refer to the three possible blockage locations.

its length. Experiments were performed at arbitrary conditions as mentioned in 2.1. Theconditions where a compromise between the pressure sensor range, pump capacity and timeneeded to collect the sample. For the 0.15 capillary diameter, the effect of the flow rate onthe inner diameter calculations where checked, and it was insignificant, less than 0.05%.

Table 2.1: The inner diameter of the barrier channel (dB) as provided by the supplier, pressuredrop over the barrier channel (∆PB), Reynolds number (ReB) in the barrier channels and liquidflow rate per channel (qL) . Experimental conditions are chosen arbitrarily and conducted atatmospheric conditions.

dB,(mm) ∆PB,(bar) ReB,(−) qL,(ml/min)

0.05 0.85 7 0.020.10 0.40 26 0.120.15 0.30-0.89 67-210 0.49-1.510.20 0.10 50 0.48

The flow non-uniformity (as well the variation in the inner diameter) is quantified via therelative standard deviation according to Equation 2.1, where (qi) refers to the flow rate per

Distributor design parameters 19

channel and (q) is the average flow rate calculated according to Equation 2.2.

σ(q) =1q

√Σi(qi− q)2

N−1100% (2.1)

q =

i=N

∑i=1

qi

N(2.2)

2.3.2 Hydraulic resistance in the barrier channel

The second distributor design parameter is the hydraulic resistance of the barrier channels.Usually the hydraulic resistance is defined independent of the flow rate (Amador et al. (2004),Pan et al. (2009)). The hydraulic resistance for two phase flow is a function of the flow rate.Thus the hydraulic resistance of the barrier channels is defined here according to Equation 2.3as the relative pressure drop (∆PB). It is the average pressure drop over the barrier channels(∆PB) normalized by the average pressure drop over the T-mixers and microchannels (∆PC).Averaging of the pressure drops are defined in relation to the total number of channels similarto Equation 2.2. Using different inner diameters of the barrier channels and flow rates, theflow non-uniformity is determined as a function of the relative hydraulic resistance in thebarrier channels (∆PB).

∆PB =∆PB

∆PC(2.3)

Experiments at eight different operating conditions were performed as shown in Table(2.2). Each experiment is named according to an index (∆PB,G - ∆PB,L- qG

qL), where (∆PB,G)

and (∆PB,L) are the relative hydraulic resistances of the barrier channels at the gas and liquidsides, respectively; ( qG

qL) is the gas to liquid flow ratio, (dB,G) and (dB,L) are the diameters of

the barrier channels as provided by the supplier for the gas and liquid, respectively; (qL) isaverage liquid flow rate over the microchannels.

The relation between flow non-uniformity and ∆PB was best demonstrated by operatingeach experiment under an extreme and systematic non-uniformity of the flows. The extremescenario was generated via blocking the inlet line to the mixer at location (i) or (ii) as shownin Figure 2.3. If the liquid inlet feed line was blocked, then the liquid redistributed to theremaining un-blocked channels, while in the blocked channel gas flowed freely. Thereforegas-liquid Taylor flow only existed in the three unblocked microchannels. The channel block-age experiments were done by realizing a steady state operation in the four channels. A plugwas positioned at the desired location to block the flow. The non-uniformity in the flows was


Table 2.2: Designed operating conditions used in the experimental setup. The experimentsare named as ∆PB,G - ∆PB,L - qG

qL.

Experiment dB,G ∆PB,G dB,L ∆PB,LqGqL

qL

(µm) (-) (µm) (-) (-) (ml/min)

0.5−0.5−1 1000 0.5 1000 0.5 1.0 0.3825−0.5−1 50 25 1000 0.5 1.0 0.3825−75−1 50 25 100 75 1.0 0.3825−10−1 50 25 200 10 1.0 0.384−4−3.5 100 4 200 4 3.5 0.1410−10−0.2 50 10 200 10 0.2 0.5825−25−1 50 25 150 25 1.0 0.3850−50−4 50 50 100 50 4.0 0.12

determined for the normalized flow rate (q), which is defined as the flow rate after blockagenormalized by the flow rate before blockage at each channel.

2.3.3 Pressure difference between the manifolds

The third design parameter addresses if equal or different barrier channels are required on thegas and liquid side. Non-equal hydraulic resistances between the gas barrier channels andthe liquid barrier channels result in non-equal pressure in the manifolds. A systematic crosstalk between the gas manifold and the liquid manifold was created by first obtaining steady-state operation in the four parallel microchannels. One of the microchannels was blockedat location (iii) as is shown Figure 2.3. Once the blockage occurred, in a similar manner toblocking the feed line, the flow in the microchannel was stopped. The cross talk betweenthe gas manifold and the liquid manifold was solely determined by the pressure differencebetween the manifolds.

2.3.4 Gas and liquid flow ranges

The fourth design parameter is the maximum gas and liquid flow ranges for the distributor.Flow distribution based on the barrier channels has one drawback which is that the barrierchannel diameters and lengths are not adjusted to changes in flow rates. Thus when the flowrate changes, the operating pressure and the hydraulic resistance in the barrier channels (∆PB)change, which influences the uniformity of the flows. Therefore, the operating pressure and∆PB were measured at different flow rates using fixed barrier channels diameters of 0.05 and

Results and discussion 21

0.15 mm for gas and liquid, respectively. The flow rates of gas and liquid were changed bythe same factor (F) to a fixed reference flow rates of 0.34 and 0.47 ml/min for the gas andliquid flows, respectively.

2.3.5 Fabrication tolerance in the inner diameter of mixers and mi-crochannels

The fifth design parameter is the maximum allowed variation in the inner diameters of themixers and microchannels downstream of the barrier channels. It is difficult to determine theinfluence of these variations on the flow non-uniformity without the influence of manifoldsand barrier channels. Therefore, a qualitative conclusion is drawn by carrying out singlechannel experiments for 4 T-mixers and 4 microchannels at fixed gas and liquid flow ratesof 0.34 ml/min. By measuring the average slug and bubble sizes and computing the non-uniformity in the slug and bubble sizes, an indication on the influence of variation in theinner diameters was obtained.

2.4 Results and discussion

2.4.1 Taylor flow in four microchannels operating in parallel

Even flow distribution along with Taylor flow was achieved over the 4 parallel microchannels.Table 2.3 and Figure 2.4 show a typical example of the obtained flow non-uniformity in theparallel channels. The flow non-uniformities in the gas and liquid flows were less than 4%and 7%, respectively. The non-uniformity in the residence time was less than 5%. Amongall the non-uniformities, the non-uniformities of slug and bubble lengths over the parallelchannels were the largest and reached 15%. The gas hold-up and the gas-liquid interfacialarea were largely unaffected and the non-uniformity varied by less than 5%.

Figure 2.4: Four parallel microchannels operating in the Taylor flow regime with ReGL of 22.Right to the figure are shown: slug length (Zs), bubble length (Zb), gas superficial velocity(UG), and liquid superficial velocity (UL) per channel.


Table 2.3: Non-uniformity in the 4 parallel microchannels. Residence time (τ), pressure drop(∆PC), specific interfacial area (as), and gas hold up (εg).

Channel 1 2 3 4 σ ,(%)

τ,(s) 16.7 15.3 16.3 15.3 5∆PC,(mbar) 17.3 16.5 14.0 19.2 13as,( m2

m3microchannnel

) 2300 2170 2110 2330 5

εg,(−) 0.45 0.44 0.46 0.45 2

2.4.2 Fabrication tolerance in the diameter of barrier channels

The flow non-uniformity as a function of the variation in the inner diameter of the barrierchannels σ(dB) is shown in Figure 2.5. This experiment is made using single phase flowaccording to the conditions mentioned in Table 2.1. The flow non-uniformity is proportionalto four times the variation in the inner diameter of barrier channels as expressed by Equation2.4.

0.0 0.5 1.0 1.5 2.00

2

4

6

8 Experiments y = 4x

(q),

(%)

(dB), (%)

Figure 2.5: Flow non-uniformity versus variation in the inner diameters of the barrier chan-nels. This experiment is made using single phase flow according to the conditions mentionedin Table 2.1.

σ(q) = 4σ(dB) (2.4)

The hydraulic resistance in the barrier is described by the Hagen-Poiseuille equation (Dels-man et al. (2005)). The relationship between pressure drop and channel diameter is to the


power four which explains the slope four in Equation 2.4. A variation in the inner diameterof the barrier channels of less than or equal to 1%, is beyond the manufabrication tolerances ofthe technology used in their production. Therefore, the minimal practical flow non-uniformityis larger than or equal to 5%.

2.4.3 The hydraulic resistance in the barrier channel

Of the entire range of operating conditions in Table 2.2, 0.5−0.5−1 and 25−0.5−1 resultedin gas-liquid channeling. At these operating conditions the relative hydraulic resistance in thebarrier channel (∆PB) was less or equal to one for the gas or liquid sides. The irregular slugand bubble size formation resulted in chaotic behavior which leads to gas-liquid channeling.With a ∆PB of 4 or larger (as in 4−4−3.5), channeling was avoided and Taylor flow existedin the four channels. Therefore, relative hydraulic resistance in the barrier channel larger orequal to 4 is necessary to prevent gas-liquid channeling.

In Figure 2.6, both experiments 4−4−3.5 and 50−50−4 operate at similar flow ratesbut using different ∆PB. At 50− 50− 4, the flow non-uniformity is five times lower than at4−4−3.5. Using the same barrier channels and at similar flow rate, such as in 4−4−3.5,a slight change in ∆PB has a around 25% effect on flow non-uniformity. The difference of∆PB at similar flow rate is due to the variation of the slug and bubble lengths before andafter the reaction chancel blockage has been made. The flow non-uniformity seems to bemainly controlled by the value of ∆PB and not much by the flow rates. This is due to thecross talks between the flow distribution in the manifold and the variation in the down streamparts (mixer and reaction channels). Depending on the value of ∆PB the degree of cross talksbetween the manifold and the downstream parts is determined. The flow non-uniformity asa function of ∆PB follows a certain trend. To quantify such a trend or to quantify the effectof flow non-uniformity as a function of flow rates, a detailed analytical analysis are needed.Here only qualitative analysis will be made.

The flow non-uniformity depends on the overall hydraulic resistance which is the sumof that over the barrier channels and the downstream parts. Figure 2.6 shows how much therelative hydraulic resistance in the barrier channels (∆PB) influences the flow non-uniformity.When a large ∆PB of 25 and larger are used, the flow non-uniformity for both phases isalmost similar. Meaning that, the flow non-uniformity is dominated mainly by the hydraulicresistances in the barrier channels, and not much by the downstream parts. When ∆PB is lessthan 25, there is a difference for the gas and liquid flow non-uniformity for the same ∆PB.Meaning that, the flow non-uniformity is not completely dominated by ∆PB but also by the


hydraulic resistances in the downstream parts.

0 15 30 45 600

5

10

15

20

25

50 - 50 - 4

10 - 10 - 0.2

25 - 25 - 1

L G

( q

), (%

)

PB , (-)

4 - 4 - 3.5

Figure 2.6: Flow non-uniformity in the parallel microchannels at different average hydraulicresistance in the barrier channels ∆PB. L (N) and G (•) are for gas and liquid flows, respec-tively. Each result shown in the graph represents one of the last four experiments mentionedin Table 2.2.

As ∆PB increases, the flow non-uniformity decreases. The flow non-uniformity reducesfrom around 22% and 10% to less than 3% as ∆PB increases from 4 to 25 for both gas and liq-uid, respectively. When ∆PB increases from 25 to 50 (which doubles the energy dissipation)the flow non-uniformity reduces by only 1%. Since the minimal practical flow non-uniformityis 5%, the upper limit for the hydraulic resistance in the barrier channels can thus be definedas ∆PB in between 20 and 25. Since a ∆PB of 4 or more is required to prevent gas-liquid chan-neling, the optimal range of the hydraulic resistance in the barrier channels can be definedwith lower and upper limits as: (∆PB)MIN ≥ 4 and 20≤ (∆PB)MAX ≤ 25.

If we compare our finding to that of De Mas et al. (2005), the barrier-channel dimensionsin their design were fixed. At the designed flow rates, their distributor operated with a ∆PB

of 50−25−3.5. When they vary the flow rates, the achieved flow non-uniformity remainedsimilar to this work and it was less than 10%. Since De Mas et al. (2005) did their experimentwith ∆PB range larger than (∆PB)MAX , thus the flow non-uniformity is mainly dominated bythe fabrication tolerance in the inner diameter of the barrier-channels and not much by thethe relative hydraulic resistances in the downstream parts. If the barrier-channel dimensionsin their distributor are designed to deliver a ∆PB near to the minimum optimal range of thehydraulic resistance in the barrier channels, (∆PB)MIN , it is worth noticing if the fabricationaccuracy of the reaction microchannels and mixers will maintain a flow non-uniformity ofless than less than 10%.


2.4.4 Pressure difference between the manifolds

Pressure difference between manifolds, which occur as result of non-equal hydraulic resis-tance of the gas barrier channels and liquid barrier channels, can lead to gas-liquid channeling.Channeling occurred when cross talk between the gas manifold and the liquid manifold wasinduced by blocking a microchannel. Channeling occurred chaotically and in two stages. Inthe first stage, the fluid at higher pressure pushed the fluid at lower pressure towards its man-ifold. In the second stage, the higher pressure fluid accumulated in the low pressure manifoldand then escaped out through the adjacent unblocked channels as shown in the insert of Fig-ure 2.7. Once the higher pressure fluid escaped from the low pressure manifold, channelingstopped and the same channeling cycle repeated.

The time needed for one channeling cycle to occur is referred to ”channeling time” (tchan

in Figure 2.7). The channeling time depends on the manifolds pressure ratio and on the mani-folds volume. Usually in microreactor low liquid hold-up is preferred, that is why minimizingthe manifold volume is often desired. Since here the manifold volume are kept the same, thechanneling time is plotted as a function of the manifolds pressure ratio as shown in Figure2.7. If another manifold volume is used, the trend shown in Figure 2.7 will remain the samebut its magnitude will shift. That is why the channeling time shown in Figure 2.7 is only validfor the current setup.

As the difference between the manifolds pressures increases, a reduction in the channelingtime is observed. When equal manifold pressures (last four experiments in Table 2.2) wereinvestigated, channeling occurred but only after a couple of hours. Once a microchannel wasblocked, the flows redistributed equally over the unblocked channels with a change in flownon-uniformity of less than 2% compared to before blocking. Even though the flow non-uniformity was largely unaffected, the flow in the remaining unblocked channels did increaseby 4/3. To minimize channeling, an equal pressure in the manifolds is needed.

2.4.5 Gas and liquid flow ranges

Figure 2.8 shows the manifolds pressure (PM) and relative hydraulic resistance in the barrierchannels (∆PB) when the gas and liquid flow rates were changed proportionally. The gasphase PM and ∆PB are equal to that of the liquid phase because the flow rate of both phaseschanges proportionally. The higher the flow rate is, the larger is the pressure in the manifoldsand that of ∆PB. The non-linear nature of ∆PB occurs since the pressure drop for Taylor flowdepends on the film thickness, which scales non-linearly with the flow rates (Warnier et al.(2010)).

The maximum distributor flow range can be determined from Figure 2.8 based on two


0 2 4 6 81

10

100

1000

10000

25-75-1

25-10-1

t = tchan

t = 0

t chan

, (s)

PM,L

/PM,G

, (-)

Figure 2.7: The channeling time at different manifolds pressure ratio PM,LPM,G

. 25− 75− 1and 25− 10− 1 are the results of two experiments. The thick line is an estimation of thechanneling time. Right top: Gas-liquid channeling in unblocked channel number 1 over timewhen channel number 2 was blocked. tchan is the time needed for one channeling cycle tooccur.

0.0 0.3 0.6 0.9 1.20

7

14

21

28

PB

P M, (

bar)

P B, (

-)

F, (-)

0.0

0.1

0.2

0.3

0.4 PM

Figure 2.8: The manifolds pressure PM and ∆PB when the gas and liquid are changed in thesame proportion by the factor F . The average gas and liquid flow rates per channel whenF = 1 is 0.34 and 0.47 ml/min, respectively.


factors: (1) the operating pressure should remain less than the device maximum operatingpressure; (2) the hydraulic resistance in the barrier channels should remain within the optimalrange, as discussed in section 4.3. In our case for Taylor flow regime, the optimal range of∆PB is maintained when the flow rates changes from F = 0.05 to F = 1. Therefore, theoptimal operational window is realized when the gas and liquid flow rates change in the sameproportion and the ratio of the maximum over the minimum flow rate remains less than 20.

2.4.6 Fabrication tolerance in the inner diameter of mixers and mi-crochannels

The non-uniformity in slug and bubble sizes due to the variations in the inner diameter of themixers and microchannels downstream of the barrier channels is shown in Figure 2.9. Eachcolumn represents a microchannel which was tested in one of the 4 T-mixers.

1 2 3 40

2

4

Zs, (

mm

)

1 2 3 40

2

4

Zb, (

mm

)

Mixer, (−)

Figure 2.9: Slug and bubble lengths in a single channel experiment, using four T-mixers andfour microchannels each represented by one column.

The variation in the inner diameters of the mixers has a larger influence on the slug andbubble sizes compared to that of the microchannels. A non-uniformity of around 15% wasobtained for different mixers. Using the same mixer with different microchannels, a lowernon-uniformity has observed (between 2% to 7%). According to the supplier, the variation inthe inner diameters of the mixers is around 10%. Applying this percentage in the analyticalmodel of van Steijn et al. (2010) (which predicts the bubble and slug sizes in T-mixers) theexpected non-uniformity in slug and bubble sizes is 15%.


Comparing the current result to that in Figure 2.4, the slug and bubble non-uniformity arealso less than 15%. Is this 15% non-uniformity due to the variation in the inner diameters ofthe mixers or to the flow non-uniformities? By applying a 10% flow non-uniformity usingone mixer dimension and using the analytical model of van Steijn et al. (2010), the estimatedslug and bubble non-uniformities are less than 10%. Thus it is more likely that the observedslug and bubble non-uniformities are due to the fabrication precision of the mixers and subse-quently to the flow non-uniformities and to the variation in the inner diameter of the reactionchannels. By reducing the variation in the inner diameter downstream of the barrier chan-nels (improving the fabrication precision), the flow non-uniformity and the needed hydraulicresistance in the barrier channels will reduces.

2.5 Conclusions

Even distribution of gas and liquid flows over parallel microchannels was successfully achievedusing the passively controlled barrier channels concept with a flow uniformity of more than90%. Criteria were established for the design of a gas-liquid flow distributor that prevents gas-liquid channeling, has a minimum hydraulic resistance and minimizes flow non-uniformity.The design criteria are:

• Gas-liquid channeling is prevented when the gas and liquid manifolds have equal pres-sures.

• The optimal range of hydraulic resistances in the barrier channels has lower and upperlimits of: (∆PB)MIN ≥ 4 and 20 ≤ (∆PB)MAX ≤ 25. The ∆PB is the average pressuredrop over the barrier channels divided by the average pressure drop over the T-mixersand microchannels.

• The optimal operational window is realized when the the gas to liquid flow ratio is keptconstant and the ratio of the maximum over minimum flow rates remains less than 20.

• The manufacturing tolerance of the barrier channels has the highest impact on the flownon-uniformity. The flow non-uniformity is proportional to four times the variation inthe inner diameter of the barrier channels.

• A variation in the inner diameters of the T-mixers and microchannels of 10%, results inflow non-uniformity of less than 10%. The required hydraulic resistance in the barrierchannels reduces by lowering the variation in the inner diameters of the T-mixers andmicrochannels.

Outlook 29

2.6 Outlook

The development of an effective gas-liquid distribution system for multiple parallel channelsis more than an academic play. It is the key if such systems in future will be used for indus-trial production. So far, reports about industrial usage of gas-liquid microreactors are verylimited compared to those for gas and liquid single-phase reactions. For liquid chemistry,often ”smart scale-out” is used, i.e. a smart increase in internal dimensions. Such prac-tical approach is hardly transferable to gas-liquid systems, as this would lead to a changein the flow pattern, the most crucial engineering point here. The work reported here addsone brick to such development. A solid understanding is needed, as the flow distributionwith two media with very different properties such as density and viscosity demands fora well-balanced adjustments of pressures over the system. The current design criteria aremerely developed based on an empirical approach for four parallel channels, performed forthe squeezing regime of Taylor flow (Garstecki et al. (2005)), and at cold (without reaction)flow condition. Despite this limited range, we aimed at extracting a detailed and quantifiedexperimental result for the influence of: hydraulic resistance of barrier channels, the fabrica-tion tolerances, variation in the two phase flow. Our next step is to utilize this experimentalresult for developing an analytical model and a design methodology which will allow to trans-late the current design criteria for broader operating conditions (other flow regimes and twophases flow).

Bibliography

Amador, C., Gavriilidis, A., Angeli, P., 2004. Flow distribution in different microreactorscale-out geometries and the effect of manufacturing tolerances and channel blockage.Chemical Engineering Journal 101, 379–390.


Anxionnaz, Z., Cabassud, M., Gourdon, C., Tochon, P., 2008. Heat exchanger/reactors (HEXreactors): Concepts, technologies: State-of-the-art. Chemical Engineering and Processing:Process Intensification 47, 2029–2050.

Bejan, A., Lorente, S., 2006. Constructal theory of generation of configuration in nature andengineering. Journal of Applied Physics 100, 041301.


Chen, L., Tian, Y. S., Karayiannis, T. G., 2006. The effect of tube diameter on vertical two-


phase flow regimes in small tubes. International Journal of Heat and Mass Transfer 49,4220 – 4230.


De Mas, N., Gunther, A., Kraus, T., Schmidt, M. A., Jensen, K. F., 2005. Scaled-out multi-layer gas-liquid microreactor with integrated velocimetry sensors. Industrial and Engineer-ing Chemistry Research 44, 8997–9013.

Delsman, E. R., de Croon, M. H. J. M., Elzinga, G. D., Cobden, P. D., Kramer, G. J.,Schouten, J. C., 2005. The influence of differences between microchannels on micro reac-tor performance. Chemical Engineering & Technology 28, 367–375.

Garstecki, P., Fischbach, M. A., Whitesides, G. M., 2005. Design for mixing using bubblesin branched microfluidic channels. Applied Physics Letters 86, 244108.

Gunther, A., Jhunjhunwala, M., Thalmann, M., Schmidt, M. A., Jensen, K. F., 2005. Mi-cromixing of miscible liquids in segmented gas-liquid flow. Langmuir 21, 1547–1555.


Haverkamp, V., Hessel, V., Loewe, H., Menges, G., Warnier, M. J. F., Rebrov, E. V., de Croon,M. H. J. M., Schouten, J. C., Liauw, M. A., 2006. Hydrodynamics and mixer-inducedbubble formation in micro bubble columns with single and multiple-channels. ChemicalEngineering & Technology 29, 1015–1026.


Hessel, V., 2009. Novel process windows - gate to maximizing process intensification viaflow chemistry. Chemical Engineering & Technology 32, 1655–1681.

Hessel, V., Loeb, P., Loewe, H., 2009. Industrial Microreactor Process Development up toProduction in V. Hessel, A. Renken, J.C. Schouten, J. Yoshida. Micro Process Engineering: A Comprehensive Handbook. Wiley-VCH.

Hessel, V., Loewe, H., Schoenfeld, F., 2005. Micromixers–a review on passive and activemixing principles. Chemical Engineering Science 60, 2479–2501.

Idelchik, I. E., 1991. Fluid dynamics of industrial equipment: Flow distribution design meth-ods. Taylor & Francis.

Irandoust, S., Andersson, B., 1989. Simulation of flow and mass transfer in Taylor flowthrough a capillary. Computers and Chemical Engineering 13, 519–526.



Kockmann, N., Roberge, D. M., 2009. Harsh reaction conditions in continuous-flow mi-croreactors for pharmaceutical production. Chemical Engineering & Technology 32, 1682–1694.

Kreutzer, M. T., Bakker, J. J. W., Kapteijn, F., Moulijn, J. A., Verheijen, P. J. T., 2005a.Scaling-up multiphase monolith reactors: Linking residence time distribution and feedmaldistribution. Industrial and Engineering Chemistry Research 44, 4898–4913.

BIBLIOGRAPHY 31


Kreutzer, M. T., Kapteijn, F., Moulijn, J. A., Heiszwolf, J. J., 2005b. Multiphase monolithreactors: Chemical reaction engineering of segmented flow in microchannels. ChemicalEngineering Science 60, 5895–5916.

Link, D. R., Anna, S. L., Weitz, D. A., Stone, H. A., 2004. Geometrically mediated breakupof drops in microfluidic devices. Physical Review Letters 92, 054503–.

Losey, M. W., Jackman, R. J., Firebaugh, S. L., Schmidt, M. A., Jensen, K. F., 2002. De-sign and fabrication of microfluidic devices for multiphase mixing and reaction. Journal ofMicroelectromechanical Systems 11, 709–717.

Losey, M. W., Schmidt, M. A., Jensen, K. F., 2001. Microfabricated multiphase packed-bed reactors: Characterization of mass transfer and reactions. Industrial and EngineeringChemistry Research 40, 2555–2562.



Narayanan, C., Lakehal, D., 2008. Two-phase convective heat transfer in miniature pipesunder normal and microgravity conditions. Journal of Heat Transfer 130, 074502.

Pan, M., Tang, Y., Yu, H., Chen, H., 2009. Modeling of velocity distribution among mi-crochannels with triangle manifolds. AIChE Journal 55, 1969–1982.

Pedersen, H., Horvath, C., 1981. Axial dispersion in a segmented gas-liquid flow. Industrialand Engineering Chemistry Fundamentals 20, 181–186.

Razzaq, T., Glasnov, T. N., Kappe, C. O., 2009. Accessing novel process windows in a high-temperature/pressure capillary flow reactor. Chemical Engineering Technology 32, 1702–1716.

Rebrov, E. V., Ekatpure, R. P., de Croon, M. H. J. M., Schouten, J. C., 2007a. Design of athick-walled screen for flow equalization in microstructured reactors. Journal of Microme-chanics and Microengineering 17, 633–641.

Rebrov, E. V., Ismagilov, I. Z., Ekatpure, R. P., de Croon, M. H. J. M., Schouten, J. C., 2007b.Header design for flow equalization in microstructured reactors. AIChE Journal 53, 28–38.

Rebrov, E. V., Schouten, J. C., de Croon, M. H. J. M., 2011. Single-phase fluid flow dis-tribution and heat transfer in microstructured reactors. Chemical Engineering Science 66,1374–1393.

Salman, W., Angeli, P., Gavriilidis, A., 2005. Sample pulse broadening in Taylor flow mi-crochannels for screening applications. Chemical Engineering and Technology 28, 509–514.



Tondeur, D., Luo, L., 2004. Design and scaling laws of ramified fluid distributors by theconstructal approach. Chemical Engineering Science 59, 1799–1813.


Tonkovich, A., Kuhlmann, D., Rogers, A., McDaniel, J., Fitzgerald, S., Arora, R., Yuschak,T., 2005. Microchannel technology scale-up to commercial capacity. Chemical Engineer-ing Research and Design 83, 634–639.

Triplett, K. A., Ghiaasiaan, S. M., Abdel-Khalik, S. I., Sadowski, D. L., 1999. Gas-liquidtwo-phase flow in microchannels part i: Two-phase flow patterns. International Journal ofMultiphase Flow 25, 377–394.

van Steijn, V., Kleijn, C. R., Kreutzer, M. T., 2010. Predictive model for the size of bubblesand droplets created in microfluidic T-junctions. Lab on a Chip 10, 2513–2518.

Verboom, W., 2009. Selected examples of high-pressure reactions in glass microreactors.Chemical Engineering & Technology 32, 1695–1701.

Wada, Y., Schmidt, M. A., Jensen, K. F., 2006. Flow distribution and ozonolysis in gas-liquidmultichannel microreactors. Industrial and Engineering Chemistry Research 45, 8036–8042.

Warnier, M. J. F., de Croon, M. H. J. M., Rebrov, E. V., Schouten, J. C., 2010. Pressuredrop of gas-liquid Taylor flow in round micro-capillaries for low to intermediate Reynoldsnumbers. Microfluidics and Nanofluidics 8, 33–45.


Design methodology forbarrier-based two phase flowdistributor 33

This chapter has been published as:

Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V, Hessel, V. & Schouten,J.C. 2012. Design methodology for a barrier-based two phase flow distrib-utor. AIChE Journal, 58-11, 3482-3493.

AbstractThe barrier-based distributor for a multichannel microreactor is a multiphase flow distribu-tor which assures flow uniformity and prevents channeling between the two phases. For Nnumber of reaction channels, the barrier-based distributor consists of a gas manifold, a liquidmanifold, N barrier channels for the gas, N barrier channels for the liquid, and N mixers formixing the phases prior to the reaction channels. The flow distribution is studied numericallyusing a method based on the hydraulic resistive networks. The single phase hydraulic resis-tive network model (Commenge et al. (2002)) is extended for two phases gas-liquid Taylorflow. For ReGL < 30, the accuracy for the model was above 90%. The developed model wasused to study the effects of fabrication tolerance and barrier channel dimensions. A designmethodology has been proposed as an algorithm to determine the required hydraulic resis-tance in the barrier channels and their dimensions. This methodology is demonstrated usinga numerical example.

34 Design methodology for barrier-based two phase flow distributor

3.1 Introduction

Reactions in micro and millimeter scale channels can benefit from enhanced mass and heattransfer characteristics. This is why micro and millimeter channels are excellent for ”ProcessIntensification” and for operation in the concept of ” Novel Process Windows” (Hartman andJensen (2009), Hessel (2009), Hessel et al. (2011), Mason et al. (2007), Mills et al. (2007)).The throughput of a single micro or millimeter scale reaction channel is often not more thanfew milliliter per minute, which can be increased via two possible routes. The first scalingroute is by ”smart scale-up”, increasing the internal channel cross section, while keepingthe enhanced mass and heat transfer characteristics. Smart scale-up has been successfullydemonstrated for single phase processes. Lonza Company has presented a smart increaseof channel diameter, with even changing microreactor types, from laboratory over pilot toproduction-scale (Roberge et al. (2009)). Additionally, operation in the Novel Process Win-dows allows a rate acceleration of liquid-phase reactions so that even with keeping a channeldiameter small, a large productivity in one channel can be realized.

Figure 3.1: Scheme for the route of scale-up via numbering-up for micro/milli channel reac-tors. (i) scale-up of a single channel, (ii) modular unit, (iii) multi-modular units.

For multi-phase flows, the situation is different. Increasing the channel diameter slightlycan change the flow patterns (Shao et al. (2009)), which provide the characteristic of the pro-cess and the key to its process intensification. For gas-liquid flows difficulties might arisethrough the different impacts of temperature and pressure on gas and liquid; e.g. a gas willbe compressed under pressure which will change the flow pattern notably.

The second scaling route is ”numbering-up”, i.e. placing channels in parallel. The keyfor successful numbering-up is equalizing the flow distribution over the channels. The scal-ing route for numbering-up is shown in 3.1. Distribution of multiphase flow takes place onlyin the parallel channels (ii) where mixing of the two phases occurs. A single phase flowfeeding system is required at the subsequent levels of parallel plates and parallel modules

Introduction 35

(iii). This simplifies the handling of the connectors in a confined system considerably. In fuelprocessing, catalytic gas-phase microreactors with thousands of parallel microchannels arestate of the art and flow distributors are available, give sufficient equipartition, and are partof patent literature (Hessel et al. (2008), Rebrov et al. (2007a,b)). In liquid-phase process-ing, numbering-up has been demonstrated as well, although often at a much lower numberof channels operated in parallel or even as external numbering-up with a very few devices(Schenk et al. (2003)).

Flow distribution for multi-phase flows in microchannels, and in particular for gas-liquidflow, is not as straightforward as for single phase flow. Improper flow distribution does notonly change the flow uniformity, but it also can result in a deformation of the flow patternor in gas-liquid channeling, some channels filled only with liquid while others are filled withgas.

Figure 3.2: Schematic representation of barrier-based gas-liquid flow distributor for fourparallel microchannels. Symbols used are: (G) gas, (L) liquid, (M) manifold, (B) barrierchannel, (T) T-mixer, and (C) reaction channel.

The barrier-based distributor, which is shown in 3.2, is a multiphase flow distributor whichprevents gas-liquid channeling, operates in the whole range of gas and liquid flow patterns,and significantly reduces flow non-uniformity (De Mas et al. (2005), Losey et al. (2001),Wada et al. (2006)). It consists of hydraulic flow resistances between the single phase flowdistributors and the parallel channels to minimize the interaction between these parts. Thehydraulic flow resistance of the barrier channels can be quantified using Equation 3.1 as ∆PB.It is the average pressure drop over the barrier channels ∆PB divided by the average pressuredrop over the corresponding mixers and microchannels ∆PC. Since ∆PB is a ratio of pressuredrops, it is dimensionless.


∆PB =∆PB

∆PC(3.1)

Mixing two phases in a microchannel can result in different flow patterns; Taylor flow(Angeli and Gavriilidis (2008), Shao et al. (2009)) is industrially most attractive due to itswell defined shape, reduced axial dispersion almost approaching ideal plug flow (Kreutzeret al. (2008)) and improved mass and heat transfer (Fries and von Rohr (2009), Gupta et al.(2009), He et al. (2010), Kececi et al. (2009)). In a setup of four channels operated in the Tay-lor flow regime, we experimentally studied the influence of ∆PB on the flow non-uniformityin chapter two. The most important findings were that a flow non-uniformity less than 10% isreached when ∆PB is in the range of 4 to 25, and that the flow non-uniformity is proportionalto four times the variation in the inner diameter of the barrier channels.

By comparing our finding to that of De Mas et al. (2005) who were one of the first tointroduce the barrier-channel concept in microchannels, the dimensions of the gas and liquidbarrier channels were chosen to create a ∆PB of 50 and 25, respectively. This value for ∆PB

is already beyond the optimal which should be between 4 and 25. Larger than this optimallimit, the flow distribution is mainly controlled by the barrier channels. This is why the flownon-uniformity linearly relates to the variation in the channel diameter of the barrier channelsdue to the fabrication tolerance, which has been confirmed by De Mas et al. (2005). Theymentioned that the pressure drop scales with the diameter of the channels to the power four(d4).

The design approach of De Mas et al. (2005) can be regarded as the safe approach, be-cause they were not concerned about the energy dissipation - the pressure drop of the dis-tributor is beyond the optimal range. Their design approach can be summarized as: first thepressure drop over the reaction channels is measured; second the dimensions for the barrierchannels are chosen to provide a ∆PB larger than 25; last, the flow non-uniformity is linearlyscaled with the variation in the channel diameter of the barrier channels with a factor of 4.

In case an optimal distributor design is needed (optimal refers to when ∆PB is between4-25), the approach by De Mas et al. (2005) is not valid any more. The cross talk betweenthe distributor and the downstream parts increases. An improved insight in the relation of thefabrication tolerance, not only in the barrier-channels, but also in all parts of the device arecrucial to reach a desired target flow uniformity.

Our previous work in chapter two quantified the flow non-uniformity as a function of

Introduction 37

∆PB and the variation in the inner diameter for barrier, mixer and reaction channels. How-ever for a target flow non-uniformity, it did not provide steps on how to determine the exactbarrier channels dimensions, fabrication tolerance and how to extend the work to other flowregimes. Here, we aim at determining these targets by developing a design methodology forthe barrier-based distributor using a two phases resistive network model. In the next section,this model is introduced followed by the conceptual idea of the design methodology. After themethods, results and discussion, we conclude by providing the step-by-step design method-ology. A numerical example is provided to illustrate design of a barrier-based gas-liquid flowdistributor for parallel microchannels operated in the Taylor flow regime.

3.1.1 2-phase resistive network model

The design principle for flow distributors relies on controlling the hydraulic flow resistance,described by the pressure drop in each channel. In micro-fluidic devices, the resistive network(RN) model is often used to design and understand the influence of these resistances (Amadoret al. (2004), Commenge et al. (2002)). Most RN models are developed for single phase flowand are only valid at relatively low Reynolds number (Pan et al. (2009)) in the laminar regime.Here the RN model is extended to account for multi-phase flow, the 2-phase resistive network(2-PRN) model. A simplified example for gas-liquid resistive flow network is shown in 3.3.Gas and liquid are distributed to parallel channels using a consecutive ladder manifold. Thegas flow passes through its own barrier channels and the liquid flow passes through its ownbarrier channels. The gas and liquid flows are mixed in a T-mixer and the gas-liquid flowpasses through the reaction channels. The pressure drop in each segment line can be writtenaccording to Equation 3.2, as the multiplication of the hydraulic resistance (R) by the flow rate(q). An explanation of the hydraulic resistances and their types in such a resistive network isgiven in detail in the appendix.

∆P = R q (3.2)

Regardless of the flow path, the pressure drop from in to out is equal to Pin−Pout . Thepressure of the gas and liquid manifold are equal, because gas-liquid channeling is preventedwhen Pin,G = Pin,L as shown in chapter two. The pressure drop for each flow path can bewritten as a summation of the pressure drop in series and equals to Pin−Pout . The pressuredrop for the flow path of channel number i can be written as:

Pin−Pout = ∆PM,L,i +∆PB,L,i +∆PC,i = ∆PM,G,i +∆PB,G,i +∆PC,i (3.3)

∆PM,L,i is the pressure drop in the liquid manifold for the region between channels i-1


Figure 3.3: Schematic view for simplified two-phases resistive network. R is the hydraulicflow resistance defined in Equation 3.2. The index used for each parameter is M manifold,B barrier, C reaction channel, q flow rate, G gas, L liquid, i is a variable for the number ofchannels N.

and i. ∆PB,L,i is the pressure drop for the liquid flow in barrier channel number i. ∆PC,i isthe pressure drop of the gas-liquid flow in mixer and reaction channel number i. Equation3.3 can be rewritten in the hydraulic resistance form to Equation 3.4 and 3.5 for the liquidand gas phase, respectively. By simultaneously solving the pressure drop equations and themass balances for both gas and liquid and for all parallel channels, the flow distribution canbe computed. A detailed analysis of how to arrange and solve these equations is provided inthe appendix.

Pin−Pout = RM,L,i (qM,L,i)+RB,L,i (qB,L,i)+RC,i (qC,L,i +qC,G,i) (3.4)

Pin−Pout = RM,G,i (qM,G,i)+RB,L,i (qB,G,i)+RC,i (qC,L,i +qC,G,i) (3.5)

3.1.2 Design methodology

The 2-PRN model provides the basis of the design methodology. The conceptual idea of thedesign methodology is to analytically decouple the target flow non-uniformity σ(q) into cu-mulative contributions of three flow non-uniformity factors: manifold σ(qM), barrier σ(qB),and mixers and reaction channels σ(qC) as shown in Equation 3.6. Each flow non-uniformityfactor represents part of the hydraulic resistances that compose the total two phase resis-tive network. Before proceeding to explain these flow non-uniformity factors, the followingdefinitions are explained.

σ(q) =√

σ 2(qM)+ σ 2(qB)+σ 2(qC) (3.6)

Introduction 39

σ(q) is the target flow non-uniformity quantified as the relative standard deviations σaccording to Equation 3.7; qi is the flow rate per channel normalized by q as defined inEquation 3.8; q is the average flow rate per channel (Equation 3.9); q is the flow rate perchannel; N is the number of channels.

σ(q) =1q

√Σi(qi− q)2

N−1100% (3.7)

qi =qi

q(3.8)

q =

i=N

∑i=1

qi

N(3.9)

Figure 3.4: Schematic view of how the simplified two-phases resistive network shown inFigure 3 is decoupled into three steps. Step (a) shows the manifold and reaction channels;step (b) extends step (a) to include the barrier channels; step (c) extends step (b) by includingmixers and second phase.

The three flow non-uniformity factors are:

3.1.2.1 The manifold flow non-uniformity factor, σ(qM)

This factor quantifies the flow non-uniformity due to the influence of manifold type anddimensions. It is computed for single phase flow only. In other words, it quantifies theinfluence of the the manifold on the flow distribution without the influence of the (1) secondphase, (2) variation in the hydraulic resistance due to the fabrication tolerance, and (3) withoutthe use of the barrier channels. This is shown in a simplified manner in Figure 3.4 (a). Theresulted flow per channel is qM . By normalizing qM according to Equation 3.10, the relativestandard deviation of qM can be computed which is the manifold flow non-uniformity factorσ(qM).

qM,i =qM,i

q(3.10)


The following conditions have been used to evaluate σ(qM):- Fixed physical properties.- Uniform manifold cross section area, RM,i = constant.- The hydraulic resistances of the barrier channels are negligible, RB,i = 0.- No variation in the channel diameters due to the fabrication tolerance: RB,i = constant,RC,i = constant.

3.1.2.2 The barrier flow non-uniformity factor, σ(qB)

This factor quantifies the flow non-uniformity due to the variation in the inner diameter of thebarrier channels. It is computed for a single phase flow. By extending the scheme shown inFigure 3.4(a) to include the barrier channels with variations in their inner diameters, σ(qB)are computed. So σ(qB) quantifies how much the flow distribution changes as a result ofintroducing the barrier channels. It is shown in a simplified manner in Figure 3.4 (b). Theresulting flow per channel for this step is qB. By normalizing qB according to Equation 3.11,the relative standard deviation of qB can be computed which is the barrier factor σ(qB).

qB,i =qB,i

qM,i(3.11)

The following conditions have been used to evaluate σ(qB):- Fixed physical properties.- Uniform manifold cross section area, RM,i = constant.- The hydraulic resistances for the barrier channels are taken into account, RB,i > RC,i.- The variation in the barrier channel diameters due to the fabrication tolerance are considered,RB,i 6= constant.- The variations in the mixer and reaction channel diameters due to the fabrication toleranceare not taken into account, RC,i = constant.

3.1.2.3 The mixers and reaction channels flow non-uniformity factor, σ(qC)

This factor quantifies the influence of variation in the mixers and reaction channels diame-ters and the influence of the second phase on the flow non-uniformity. It is computed fortwo-phases flow and is shown in a simplified manner in Figure 3.4 (c). By extending thescheme shown in Figure 3.4(b) to take into account: (1) the second phase, (2) mixers, and(3) variations in the mixers and reaction channels diameters due to the fabrication tolerance,σ(qC) can be calculated. So σ(qC) quantifies how much the flow distribution is affected dueto these additions. The flow per channel for this step is qC. By normalizing qC according toEquation 3.12, the relative standard deviation of qC can be computed which is the mixer and

Introduction 41

reaction channels flow non-uniformity factor, σ(qC).

qC,i =qi

qB,i(3.12)

The following conditions have been used to evaluate σ(qC):- Fixed physical properties.- Uniform manifold cross section area, RM,i = constant.- The hydraulic resistances for the barrier channels are taken into account, RB,i > RC,i.- The variation in the barrier channels diameters due to the fabrication tolerance are consid-ered, RB,i 6= constant.- The variations in the mixer and reaction channel diameters due to the fabrication toleranceare considered, RC,i 6= constant.

The product of qM , qB, and qC yields the actual flow per channel q (Equation 3.13).

qi = qM,i qB,i qC,i (3.13)

Correlations for the three flow non-uniformity factors σ(qM), σ(qB), and σ(qC) are ob-tained using the 2-phase resistive network model. These correlations are derived for a rangeof conditions as shown in Table 3.1, Tabel 3.2, and Table 3.3. When the channel geometry isrectangular, H, W , and L are the depth, width and length of the channel. When the channelgeometry is circular, d and L are the diameter and length of the channel, respectively. Hmixer

and Wmixer,G,in are the depth of the channels in a T-mixers and width of inlet gas channel in aT-mixer, respectively.

Table 3.1: Dimensions of channels in (mm) used in the simulation study of Figure 3.6. The ∗refers to channels where fabrication tolerance are added according to Equation 3.14. For eachset of channel dimensions, simulation of the fabrication tolerance is repeated 1000 times. Nis kept fixed to 40, ReGL to 85 and qG

qLto 1.

H W L d

Manifold 4 40 4 -Reaction 1 1 2000Slit liquid barrier 0.05-1.35∗ 0.6 150 -Slit gas barrier 0.01-0.3∗ 0.45 150 -Circular liquid barrier - - 150 0.05-0.5∗

Circular gas barrier - - 150 0.05-0.5∗

The simulation study is performed for gas-liquid flow operating in the Taylor flow regime.


The two phase Taylor flow pressure drop was calculated by the model proposed by Warnieret al. (2010) The slug and bubble lengths in a T-mixer were estimated by the model proposedby van Steijn et al. (2010) For investigating the effect of fabrication tolerance, a randomvariation (δ f ) was added to the inner channel diameters using Equation 3.14 following anormalized random distribution. In the barrier channels, the influence of fabrication tolerancewas studied for two types of channel cross sectional geometry: circular and slit - rectangularchannel with a width over depth larger than 2 as shown in Table 3.1.

di = d ± δ f (3.14)

Table 3.2: Dimensions of channels in (mm) used in the simulation study of Figure 3.7. The ∗refers to channels where fabrication tolerance are added according to Equation 3.14. For eachset of channel dimensions, simulation of the fabrication tolerance is repeated 1000 times. Nis changed from 10-200, and ReGL from 0-250 and qG

qLis kept fix to 1.

H W L Hmixer Wmixer,G,in

Manifold 4 40 4 - -Liquid barrier 0.045-0.225∗ 0.6 150 - -Gas barrier 0.01-0.05∗ 0.45 150 - -Mixer - - - 1∗ 1∗

Reaction 0.1-3∗ 0.1-3∗ 2000 - -

Table 3.3: Dimensions of channels in (mm) used in the simulation study of Figure 3.8. The ∗refers to channels where fabrication tolerance are added according to Equation 3.14. For eachset of channel dimensions, simulation of the fabrication tolerance is repeated 1000 times. Nis kept fixed to 40, ReGL to 85 and qG

qLto 1.

H W L Hmixer Wmixer,G,in

Manifold 4 40 4 - -Liquid barrier 0.05-1.35∗ 0.06-1.8 150 - -Gas barrier 0.01-0.3∗ 0.05-1.35 150 - -Mixer - - - 0.1-3∗ 0.1-3∗

Reaction 1∗ 1∗ 2000 - -


3.2 Results and discussion

3.2.1 Influence of the manifold flow non-uniformity factor, σ(qM)

At different hydraulic flow resistance ratios RC/RM and numbers of parallel channels N, thecalculated manifold flow non-uniformity factor σ(qM) is shown in 3.5. As RC/RM increasesor as N decreases, the manifold flow non-uniformity factor σ(qM) decreases. This behavior issimilar to that shown by Amador et al. (2004). To provide a better physical interpretation forFigure 3.5, a dimensionless analysis for two channels in parallel (arranged consecutively) isderived for the case shown in Figure 3.4 (a). Using Equation 3.2 and assuming that the overallpressure drop is constant, the flow rate q depends just on the overall equivalent hydraulicresistance Req, shown as q = ∆P

Req. Using the equivalent resistance concept in an electrical

network, the equivalent resistance Req for two parallel channels is given by Equation 3.15.This equation explains the flow non-uniformity trend observed in 3.5. For the case when RC

is much larger than RM , the equivalent resistance equals to 1Req

= 2RC

. The 2 here refers tothe number of parallel channels, which explains the dependency on the number of parallelchannels.

1Req

=(RC)1 +(RC +RM)2

(RC)1 (RC +RM)2(3.15)

1000 10000 100000 10000000

2

4

6

8

10 N 20 50 100 150 200

(qM

) , (%

)

RC / RM , (-)

Figure 3.5: The influence of the hydraulic flow resistance ratio RC/RM on the manifold flownon-uniformity factor, σ(qM), as a function of parallel channels N. This is obtained usingthe 2-PRN model for a manifold with a uniform and constant cross sectional area shown inFigure 3.4(a).

σ(qM) = 14.71 N1.90(RC

RM)−0.96 (3.16)


The aim in this section is to extract an empirical correlation for σ(qM) to demonstratethe design methodology. For this purpose the result in Figure 3.5 were fitted, which resultedin Equation 3.16. The fitted values are for a manifold with a uniform and constant manifoldcross sectional area. For other types of manifolds (Rebrov et al. (2011)) the fitted valuesmight not be appropriate. The 95% confidence intervals of the fitting parameters in Equation3.16 are 14.47-15.06, 1.89-1.95, and -0.98 - -0.95, respectively.

3.2.2 Influence of the barrier channels flow non-uniformity factor, σ(qB)

For different hydraulic resistance of the barrier channels ∆PB, the barrier flow non uniformityfactor σ(qB) is studied as a function of the variation in the inner diameter of the barrierchannels σ(dB). By dividing the flow non-uniformity over σ(dB), the influence of ∆PB isshown in Figure 3.6. The result for the circular and slit are compared at the same value ofhydraulic resistance. When a large hydraulic resistance in the barrier channels is used, therelation between σ(qB), σ(dB) and ∆PB starts approaching a constant values of 4 and 3 for thecircular and slit channel geometry, respectively. This trend can be explained by performingdimensionless analysis similar to the one made in the previous section but for the case shownin Figure 3.4 (b). The equivalent resistance for two channels in parallel are derived, as shownin Equation 3.17. The flow rate q depends linearly on 1

Req.

1Req

=(RB +RC)1 +((RB +RC)2 +(RM)2)(RB +RC)1((RB +RC)2 +(RM)2)

(3.17)

When RB is much larger than RM and RC (large hydraulic resistance in the barrier chan-nels), then the equivalent resistance equals to 1

Req= 2

RB. This implies that the flow non-

uniformity is controlled by one part only which is by the barrier channels. The hydraulicresistance for circular and slit barrier channels are shown in Equations 3.18 and 3.19, respec-tively. The relation between RB and the hydraulic diameters are to the power 4 and 3 forcircular and slit geometry, respectively. Which matches well the result shown in Figure 3.6.Note that Equation 3.19 is for infinite parallel-plate channels and only accurate for the slitgeometry when W >> H.

RB,circular =128µL

d4 (3.18)

RB,slit =12µLH3 W

(3.19)

When using a low hydraulic resistance in the barrier channels, the hydraulic resistancesof the manifold and reaction channels start to contribute to the flow non-uniformity. That iswhy σ(qB)

σ(dB) reduces as ∆PB reduces. For the same hydraulic resistance in the barrier channels


and when the fabrication tolerance in the slit and circular barrier channels geometries are thesame, the flow non-uniformity for the slit channel geometry is less by more than 20%. Thus,barrier channels with increasing width over depth are preferable.

1 10 100 10000

1

2

3

4

5 Circle Slit

(qB)

/ (d

B) ,

(-)

PB, (-)

Figure 3.6: The barrier channels flow non-uniformity σ(qB) divided over the variation in theinner diameter of the barrier channels, σ(dB) at different value of ∆PB. This study is donefor circular and slit barrier channels geometries for the system shown in Figure 3.4(b). Theresult for both geometry are computed at the same value of hydraulic resistance.

σ(qB)circular = 4∆PB

1+∆PBσ(dB) (3.20)

σ(qB)slit = 2.65∆PB

1.56+∆PBσ(dB) (3.21)

To extract an empirical correlation for σ(qB) so that it can be used in the design method-ology, the result in Figure 3.6 were fitted numerically, which are shown in Equation 3.20 and3.21 for circular and slit barrier channels, respectively.

3.2.3 Influence of the mixers and reaction channels flow non-uniformityfactor, σ(qC)

The variation in the inner diameters of the mixers, reaction channels can be grouped intoone parameter which is σ(∆PC). This is the variation in pressure drop over the mixers andreaction channels. By varying the inner diameter of the mixers and reaction channels a widerrange of σ(∆PC) is generated numerically. For each σ(∆PC), σ(qC) can be calculated at


different ∆PB as shown in 3.7.

0 5 10 15 200

10

20

30

40

( PC), (%)

(qC),

(%)

PB, (-)

Figure 3.7: The influence of σ(∆PC) on σ(qC) at different values of ∆PB. The study isperformed for the system shown in Figure 3.4(c).

The mixer and reaction channels flow non-uniformity factor σ(qC) linearly increases asσ(∆PC) increases, while it inversely decreases as ∆PB increases. This trend can be fitted withEquation 3.22 which provides the required correlation for the mixers and reaction channelsflow non-uniformity factor, σ(qC).

σ(qC) =σ(∆PC)

˜∆PB(3.22)

By considering the experimental results obtained in chapter two for the influence of ∆PB

on the flow non-uniformity, it was possible to experimentally validate Equation 3.22 as shownin Figure 3.8. Equation 3.22 predicts the liquid flow non-uniformity with an accuracy of morethan 90%. For the gas phase, Equation 3.22 predicts well the flow non-uniformity but only athigher value of ∆PB. At lower ∆PB, Equation 3.22 under predicts the gas flow non-uniformity.It is possible that this deviation is due to some inaccuracy in the flow and pressure dropmeasurements, which were for experiments performed at atmospheric conditions and lower∆PB of less than 50 mbar.

3.2.4 The influence of the three flow non-uniformity factors combined

The obtained correlations for σ(qM), σ(qB) and σ(qC) can be used for evaluating how mucheach of these factors contribute to the overall flow non-uniformity. This is done by plotting


0

5

10

15

20

25

0 5 10 15 20 25

y = x Liquid Gas

( PC)/ PB , (%)

(qC),

(%)

Figure 3.8: Experimental flow non-uniformity σ(qC) at different ∆PB and σ(∆PC) taken from? compared to Equation 3.22.

∆PB versus the variation in the inner diameter of the barrier channels, σ(dB), at differentconditions.

Figure 3.9: The influence of the three flow non-uniformity factors demonstrated by plotting∆PB versus σ(dB). In (a), σ(∆PC) is varied at constant σ(q) of 10% and σ(qM) of 1%. (b)shows σ(q) when it changes at constant σ(∆PC) of 40% and σ(qM) of 1%. In (c), σ(qM) isvaried at constant σ(q) of 10% and σ(∆PC) of 40%.

By varying σ(∆PC) from 5% to 100%, the influence of σ(qC) is analyzed as shown inFigure 3.9(a). As σ(∆PC) increases, ˜∆PB needed to keep σ(q) less than 10% increases lin-early with a ratio of almost one. Independent of σ(∆PC), there is a cut-off value for σ(dB) toreach a given target flow non-uniformity, which is here 10%. σ(dB) should not exceed thiscut-off value. When σ(dB) is larger than the cut-off value, whatever value of ˜∆PB is used, the


target flow non-uniformity will not be obtained.

By varying σ(q), the influence of the barrier channels flow non-uniformity factor, σ(qB)is analyzed as shown in Figure 3.9(b). When allowing a larger non-uniformity of the targetflow, a lower ˜∆PB is required and the maximum allowed variation in the barrier diametersσ(dB) increases. As σ(q) increases from 5% to 30%, the maximum allowed σ(dB) increasesfrom 1% to 7%. This demonstrates the trade off between flow non-uniformity and the fabri-cation tolerance.

The influence of flow non-uniformity factor due to the manifold is shown in Figure 3.9(c).As σ(qM) increases to 5% (half that of σ(q)) its influence on ∆PB and σ(dB) remains almostthe same. When it increases to 8%, which is 80% of σ(q), the maximum allowed σ(dB)reduces and ∆PB increases. It can be conclude that when σ(qM) is less than half of the targetflow non-uniformity, then σ(qM) has little influence on σ(dB) and ∆PB.

3.2.5 Correlation of σ(∆PC) for Taylor flow regime

An important finding of this work is Equation 3.22, in specific σ(∆PC). This parameter con-tains, the variation in the diameter of the mixer and reaction channels, the flow regime, phys-ical properties, and the multi-phase flow non-uniformity. It represents the scaling-parameterfor the process of numbering-up. From a design point of view, this parameter provides themaximum allowed variation in the channel diameters of the mixer and reaction channels.Therefore, a direct correlation to link σ(∆PC) to the variation in the inner diameters is re-quired. Obtaining such a correlation for gas-liquid Taylor flow is demonstrated.

Figure 3.10: Influence of the variation in the inner diameter of the mixer and reaction chan-nels on σ(∆PC)

∆PBat different value of ∆PB. (a), (b) and (c) shows the result of the variation in

σ(dC), σ(WMixer,G,in) and σ(HMixer,in) all normalized by ∆PB, respectively.


Table 3.4: Values of the fitting parameter and their 95% confidence intervals for differentequations.

a 95% confidence intervals

fdC = a σ(dC)∆PB

1.41 1.40-1.42

fWmixer,G,in = a σ(Wmixer,G,in)∆PB

0.52 0.52-0.53

fHmixer = aσ(δHmixer )

∆PB0.28 0.27-0.29

Using the 2-PRN model, the variation in the inner diameter of the mixer and reactionchannels is shown in Figure 3.10. A 10% maximum variation in the channels diameter isgenerated numerically following Equation 3.10 as presented in the design methodology. Thevariation in the diameter of the mixer channels is implemented for the mixer channels depthHmixer and for the width of the inlet gas channel Wmixer,G,in. At each variation, σ(∆PC) iscalculated at different values of ∆PB. The result are then normalized by ∆PB as shown in3.10.

fd = aσ(d)∆PB

(3.23)

σ(∆PC)∆PB

= σ(qC) =√

( f )2dC + ( f )2

Wmixer,G,in+( f )2

Hmixer(3.24)

A linear relations are observed between σ(∆PC) and the variation in the channels diame-ter. By fitting the result using Equation 3.23, correlations for dC, Wmixer,G,in and Hmixer can beobtained. The values of the constant (a) for the three correlations and their 95% confidenceintervals are shown in Table 3.4. By combing the correlations for dC, Wmixer,G,in and Hmixer,one can predict σ(∆PC)

∆PBas shown in Equation 3.24.

Assessing the role of fabrication tolerance on σ(∆PC), is a major contribution of thisdesign methodology. Depending on the application, it is possible that fouling or depositionof material occur over time. In such a scenario what should be done is anticipating thelevel of fouling in advanced. The influence of fouling on the variation in the pressure drop,σ(∆PC) should be determined or estimated. In this way σ(∆PC) takes into account not onlythe fabrication tolerance but also the influence of fouling or deposits. This will result in anew requirement for the barrier channels. Thus σ(∆PC) is a generic parameters which takesinto account all the effects in the mixer and reaction channels.



The design methodology is summarized in Figure 3.11. In the first step the given inputis specified: physical properties, mixer and reaction channels dimensions, target flow non-uniformity. Based on the fabrication process technology, in the second step the maximumacceptable variation in the inner diameter for the barrier, mixer and reaction channels is spec-ified. Next, the hydraulic resistances which can be considered zero are specified. The fourthstep is to find the manifold dimensions. By specifying the percentage of the manifold flownon-uniformity of the target flow non-uniformity, Equation 3.16 or the 2-PRN model can beused to estimate the manifold hydraulic resistance RM . This provides an estimate for themanifold dimensions. The fifth step is to calculate the remaining flow non-uniformity fac-tors. Depending on the barrier channel geometries (slit or circular), Equations 3.20 is usedto calculate the barrier channels flow non-uniformity factor. Using Equation 3.6, the mixerand reaction channel flow non-uniformity factor are calculated. The sixth and last step is tocalculate the barrier channel dimensions. The 2-PRN model or Equation 3.22 are used to cal-culate ∆PB, which can be directly used to calculate the needed ∆PB and the barrier channelsdimensions. Since there is a trade off between the fabrication tolerance and the barrier chan-nels dimensions, the design methodology may require iteration to reach a realistic design. Inthe appendix, the design methodology is demonstrated using a numerical example.

3.3 Conclusions

This paper presents a design methodology for a multi-phase barrier-based flow distributor. Itis based on the 2-phase resistive network model which is applied to different microchannelsdimensions and geometries. The model quantitatively demonstrates the effects of fabricationtolerances on the flow non-uniformity. It shows that:

• For the same fabrication tolerance, barrier channels with slit geometry have a 20%better flow uniformity than those with a circular geometry.

• For a given target flow non-uniformity, there is a cut-off value for the allowed variationin the diameter of barrier channels σ(dB). σ(dB) should not exceed this cut-off value.For example, for a flow non-uniformity of less than 10%, the cut-off value of σ(dB) is2.5% and 3.5% for circular and slit barrier channels geometries, respectively.

• When the flow non-uniformity due to the manifold σ(qM) is less than 50% of the targetflow non-uniformity, reducing σ(qM) does not affect the maximum allowed σ(dB) orthe needed ∆PB.

Conclusions 51

Figure 3.11: Step-by-step design methodology for barrier-based multi-phase flow distributor.

A key outcome of the design methodology is the correlation for σ(∆PC). The variation inthe pressure drop over the mixers and reaction channels. This parameter contains, the varia-tion in the diameter of the mixer and reaction channels, the flow regime, physical properties,and the multi-phase flow non-uniformity. Thus it can be regarded as a scaling parameter fornumbering-up. If a pressure drop correlation is available for a new flow regime or multiphaseflows, the current design methodology can easily be modified by adjusting the correlation ofσ(∆PC).

It is important to note that the hydraulic resistive network used in this design method-ology is only valid for relatively low Re number in the laminar regime. However for largerRe numbers, the hydraulic flow resistances due to flow turning, contraction, expansion andmixing (singularity losses) can not be neglected and strongly influence the manifold perfor-mance. This is an issue of flow distribution for single phase flow which can be approachedseparately. For the influence of the variation in the inner diameter of the barrier channels, nochanges are anticipated at larger Re numbers. For the influence of mixer and reaction chan-nels factor, which is the essence of this work, a case study using a detailed fluid mechanics


analysis (numerical and/or experimental) can only bring clear answers to this question.

3.4 Appendix

3.4.1 A. 2-phase resistive network model

The 2-phase resistive network (2-PRN) model is used to compute the the flow distributionin parallel channels by solving the pressure drop equations and the mass balances for bothgas and liquid simultaneously. An extended 2-phase resistive network for gas-liquid flow isshown in Figure 3.12.

Figure 3.12: Schematic view for gas-liquid flow in multiple parallel microchannels. Variablesmentioned are: (M) manifold, (B) barrier channels, (C) reaction channels, (S) gas-liquidseparator channels, (E) Exit or collector channels, (N) number of channels, (G) gas, (L)liquid.

3.4.1.1 Pressure drops balance

For any consecutive loops, such as the one shown in Figure 3.13, there are two possible flowpaths between one junction in the manifold to one junction in the exit. The pressure drop ofthe two flow baths are equal. For each flow path, the pressure drop is a summation of thepressure drops of the flow segments as shown in Equations 3.25 and 3.26, respectively. Foreach phase, an (N−1) equations are obtained.

Appendix 53

Figure 3.13: Schematic view for the pressure drop balance for one loop (loop rule) for 2reaction channels. The pressure drop balance for liquid and gas is made between the completeand dotted circle, respectively.

∆PB,L,i +∆PC,i +∆PS,L,i +∆PE,L,(N−i+1)− (∆PM,L,i+1 +∆PB,L,i+1 +∆PC,i+1 +∆PS,L,i+1) = 0,

(3.25)

∆PB,G,i +∆PC,i +∆PS,G,i +∆PE,G,(N−i+1)− (∆PM,G,i+1 +∆PB,G,i+1 +∆PC,i+1 +∆PS,G,i+1) = 0,

(3.26)i = 1,2, ......,(N−1)

The pressure drop in Equations 3.25 and 3.26 can be written in hydraulic resistance formatas ∆P = R q. For each ∆Pi, the hydraulic resistance R is a summation of different types ofhydraulic resistances as shown in Table 3.5.

Table 3.5: Types of hydraulic resistances in the flow network shown in 3.12 and their sum-mation to compute ∆Pi.

Flow rate Resistance R f riction Rsingularity Rmixing Rmultiphase

qM,i+1 RM,i+1 = (Ri+1 + Rinlet,i+1)qi RB,i = (Ri)qi RC,i = (Ri)qG,i + qL,i RC,i = (Ri + Ri)qi RS,i = (Ri)qE,(N−i+1) RE,(N−i+1) = (R,(N−i+1) + Routlet,(N−i+1))


R f riction is the hydraulic resistance due to the frictional losses caused by the wall shearin the channel. For laminar flow in channels, the Hagen-Poiseuille equation can be used toestimate the pressure drop for a given flow rate q (Commenge et al. (2002)). Thus R f riction

can be written as shown in Equation 3.27, where µ is the viscosity, L is the channel length,λNC is a non-circularity factor that depends on channel geometry (Amador et al. (2004)), A isthe channel cross section area, and d is the hydraulic diameter.

R f riction =32µLλNC

d2A(3.27)

Rsingularity is the hydraulic resistance due to the losses caused by splitting the flow intotwo branches, sudden contraction in the cross section area and stream turning losses. For theinlet singularity losses, Rsingularity,inlet , it is computed according to Equation 3.28. where asfor the exit singularity Rsingularity,outlet , it is computed according to Equation 3.29.

Rsingularity,inlet = (ζs +ζc +ζt)ρ2A

U (3.28)

Rsingularity,outlet = (ζm +ζe +ζt)ρ2A

U (3.29)

ζs, ζc, ζt , ζm, and ζe are the singularity losses factors due to splitting of two branches,sudden contraction, turning, combining two branches, and sudden increase in cross sectionarea, respectively (Pan et al. (2009)).Rmixing is the hydraulic resistance caused by mixing the gas and liquid, computed accordingto Equation 3.30.

Rmixing = (ζm)ρ2A

(UG +UL) (3.30)

Rmultiphase is the hydraulic resistance for the multiphase flow in the reaction channel. Inour case, Rmultiphase corresponds to the gas-liquid Taylor flow resistance which is caused bythe wall shear in the channel, the capillary force and the slug sizes as shown in Equations3.31, 3.32 and 3.33 (Warnier et al. (2010)). ReGL and CaGL are the Reynolds and capillarynumbers for the gas-liquid flow, Ab is the gas bubble cross section area, and δs is the lengthof the cap in the liquid slug.

Rmultiphase =2µLλNC

d2A(ReL fs) (3.31)

fs =16

ReGL

(1+a

dc

Ls +δs(

ReGL

CaGL)0.33

)(3.32)

Appendix 55

a =7.163

23

32

(AAb

)(ReGL

CaGL

) 13 (

Ca2b +3.34Cab

)−1(3.33)

Note that Rsingularity, Rmixing, and Rmultiphase are flow rate dependent. Since they are usedto calculate the flow rate, a numerical iteration will be required in the solving procedure.

3.4.1.2 Mass balance

By applying mass balance at each junction (junction rule), the flow rates in the manifolddistributor, qM,(i+1), and exit collector, qE,(N−i+1), are expressed as a function of the flowrates in the parallel channels. They are presented in Equations 3.34 and 3.35, respectively.

qM,(i+1) =k=N

∑k=1

qk−k=i

∑k=1

qk (3.34)

qE,(N−i+1) =k=i

∑k=1

qk (3.35)

The total gas and liquid inlet flow rates can be written as in Equations 3.36 and 3.37, respec-tively.

qG,inlet =k=N

∑k=1

qG,k (3.36)

qL,inlet =k=N

∑k=1

qL,k (3.37)

3.4.1.3 Solver

By rearranging the mass balances and the pressure drop balances, a set of algebraic equationsis generated as shown below. The Mresistances is a systematic matrix which includes all thehydraulic resistances. Mresistances is quite large to show here and can be found in the publishedarticle (Al-Rawashdeh et al. (2012)).

The procedure to generate the Mresistances and solve the algebraic equations is presented inFigure 3.14. Matlab 7.5.0 is used to perform the calculations. Once the model generates theresults, the manifold, barrier, mixer and reaction channels dimensions can be tuned to reachthe desired design.


Figure 3.14: Sequence procedure for solving the 2-PRN model.

Mresitances×

qL,1

qG,1...

qL,i

qG,i...

qL,N−1

qG,N−1

qL,N

qG,N

=

RM,L,2×qL,in

RM,G,2×qG,in...

RM,L,i×qL,in

RM,G,i×qG,in...

RM,L,N ×qL,in

RM,G,N ×qG,in

qL,in

qG,in

Bibliography

Al-Rawashdeh, M., Nijhuis, T. A., Rebrov, E. V., Hessel, V., Schouten, J. C., 2012. Designmethodology for barrier-based two phase flow distributor. AIChE Journal 58-11, 3482–3493.

Amador, C., Gavriilidis, A., Angeli, P., 2004. Flow distribution in different microreactor

BIBLIOGRAPHY 57

scale-out geometries and the effect of manufacturing tolerances and channel blockage.Chemical Engineering Journal 101, 379–390.




Fries, D. M., von Rohr, P. R., 2009. Liquid mixing in gas-liquid two-phase flow by meander-ing microchannels. Chemical Engineering Science 64, 1326 – 1335.

Gupta, R., Fletcher, D. F., Haynes, B. S., 2009. Cfd modelling of flow and heat transfer in theTaylor flow regime. Chemical Engineering Science 65, 2094–2107.








Losey, M. W., Schmidt, M. A., Jensen, K. F., 2001. Microfabricated multiphase packed-bed reactors: Characterization of mass transfer and reactions. Industrial and EngineeringChemistry Research 40, 2555–2562.




Rebrov, E. V., Ekatpure, R. P., de Croon, M. H. J. M., Schouten, J. C., 2007a. Design of athick-walled screen for flow equalization in microstructured reactors. Journal of Microme-chanics and Microengineering 17, 633–641.

Rebrov, E. V., Ismagilov, I. Z., Ekatpure, R. P., de Croon, M. H. J. M., Schouten, J. C., 2007b.


Header design for flow equalization in microstructured reactors. AIChE Journal 53, 28–38.Rebrov, E. V., Schouten, J. C., de Croon, M. H. J. M., 2011. Single-phase fluid flow dis-

tribution and heat transfer in microstructured reactors. Chemical Engineering Science 66,1374–1393.







Numbered-up gas-liquidmicro/milli channels reactor withmodular flow distributor 44

This chapter has been published as:Al-Rawashdeh, M., Yu, F., Nijhuis, T.A., Rebrov, E.V, Hessel, V. &Schouten, J.C. 2012. Numbered-up gas-liquid micro/milli channels reac-tor with modular flow distributor. Chemical Engineering Journal, 207-208,645-655.

AbstractGas-liquid processing in microreactors remains mostly restricted to the laboratory scale dueto the complexity and expenditure needed for an adequate numbering-up with a uniform flowdistribution. Here, the numbering-up is presented for multiphase (gas-liquid) flow in microre-actor suitable for a production capacity of kg/h. Based on the barrier channels concept, thebarrier-based micro/milli reactor (BMMR) is designed and fabricated to deliver flow non-uniformity of less than 10%. The BMMR consists of eight parallel channels all operated inthe Taylor flow regime and with a liquid flow rate up to 150 mL/min. The quality of theflow distribution is reported by studying two aspects. The first aspect is the influence ofdifferent viscosities, surface tensions and flow rates. The second aspect is the influence ofmodularity by testing three different reaction channels type: (1) square channels fabricated ina stainless steel plate, (2) square channels fabricated in a glass plate, and (3) circular channels(capillaries) made of stainless steel. Additionally, the BMMR is compared to that of a sin-gle channel regard the slug and bubble lengths and bubble generation frequency. The resultspave the ground for bringing multiphase flow in microreactor one step closer for large scaleproduction via numbering-up.

60 Numbered-up gas-liquid micro/milli channels reactor

4.1 Introduction

The high rates of mass and heat transfer, minimum axial dispersion and the high interfacialarea allow micro/milli channel reactors to run highly exothermic, toxic or even explosive re-actions safely, permitting greener routes for processing (Hartman and Jensen (2009), Hessel(2009), Kockmann and Roberge (2009), Razzaq et al. (2009), Wiles and Watts (2012)). Mi-croreactors are very attractive devices for many different applications (Gunther and Jensen(2006), Mason et al. (2007), Pennemann et al. (2004), Roberge et al. (2005)). Different fromthe traditional scale-up, micro/milli channel reactors reach bulk chemicals productions via socalled numbering up, placing multiple channels in parallel (Gavriilidis et al. (2002), Hasebe(2004), Natividad et al. (2007), Tonkovich et al. (2005)). Because the dimensions of themicrochannel where mixing, heating and reaction remains the same as those of the labora-tory scale, industrial production starts directly from the lab (Bayer et al. (2005), Charpentier(2007), Kralisch and Kreisel (2007)).

The simplest scheme for scale-up via numbering-up is shown in Figure 4.1. In the labora-tory, scale-up of a single channel is investigated while ”smartly” keeping the excellent prop-erties of the micro/milli channels reactor (Kockmann et al. (2011), Tonkovich et al. (2005)).The second scale-up step is to number-up the single channel in one single device - the mod-ular unit. The last step is to arrange all these modular units together in what Hasebe (2004)named the plant lay-out.

Figure 4.1: Scheme for the route of scale-up via numbering-up for micro/milli channel reac-tors. (i) scale-up of a single channel, (ii) modular unit, (iii) Multi-modular units.

The main block for the numbering-up is the modular unit. The modular unit can be de-fined as a device which contains different functional elements such as: distributor, mixer,reaction channels, heat exchanger and separator, and being fed by one single feeding unit foreach phase. The modular unit should maintain equal flow conditions in the parallel channels,all of the functional elements should be integrated in one device, and the fabrication methodshould be suitable for bulk production of the reactor.

Introduction 61

For single phase flow, many modular units are already available in the market for in-dustrial production (Anxionnaz et al. (2008), Kockmann et al. (2011), Rebrov et al. (2011),Tonkovich et al. (2005)). For multi-phase flow, development of modular units is still in apreliminary stage (De Mas et al. (2005), Hessel et al. (2008), Marre et al. (2010)). This ismainly due to the difficulty in managing the flow distribution for multi-phase flow (De Maset al. (2005), Kashid et al. (2010), Mendorf et al. (2010), Wada et al. (2006)). Improper flowdistribution, specially for gas-liquid flow, can result in a deformation of the flow pattern or ingas-liquid channeling (Haverkamp et al. (2006), Yue et al. (2010)), some channels filled onlywith liquid while others are filled with gas.

The flow distribution depends on the hydraulic resistance in each of the parallel channels(Amador et al. (2004), Commenge et al. (2002), Pan et al. (2009), Rebrov et al. (2011)). Insingle phase flow, the hydraulic resistance depends on the physical properties of the fluids andon the hydraulic diameter of the channel. For multi-phase flow, the flow distribution dependson the properties of the single phase (Warnier et al. (2010)) and in addition on the flow rates,the specific gas-liquid interfacial area, the flow regime (Shao et al. (2009)), and on the waythe phases are in contact. The contact between the phases can be continuous like in the fallingfilm microreactor (Chambers et al. (2005)) or dispersed like in segmented Taylor flow (Hesselet al. (2005)). Here we only focus on gas-liquid flow in channels operated under the Taylorflow regime (Angeli and Gavriilidis (2008), Song et al. (2006)). Taylor flow is attractivedue to its well-defined gas-liquid interface, reduced axial dispersion almost approaching plugflow, and high mass and heat transfer (Angeli and Gavriilidis (2008), Gupta et al. (2010)).

Distributing gas and liquid flows to achieve Taylor flow regime in parallel channels canbe achieved via branching, internal distribution (like in the monolith using a douche type), orby using separate gas and liquid feeding for each parallel channel as shown in chapter two.When hydraulic resistances, so called barrier channels, are placed between the single phaseflow distributor and the separate gas and liquid feeding for the parallel micro channels asshown in Figure 4.2, (1) gas-liquid channeling is prevented, (2) all flow regimes, viz. Taylor,churn and annular can be successfully realized, and (3) the flow uniformity is substantiallyimproved (De Mas et al. (2005), Wada et al. (2006)).

The barrier-based distributor is an excellent gas-liquid distributor for parallel channelsoperated in the Taylor flow regime. A major characteristic for this distributor is the hydraulicresistance needed to achieve equal flow distribution. This parameter can be quantified in ageneric way as ∆PB as given in Equation 4.1. It is the average pressure drop over the barrier


channels ∆PB divided by the average pressure drop over the corresponding mixers and microchannels ∆PC. Since ∆PB is a ratio of pressure drops, it is dimensionless.

∆PB =∆PB

∆PC(4.1)

De Mas et al. (2005) were among the first to demonstrate this type of distributor in microchannel reactors. Their design was successfully run but with barrier channels designed in therange of ∆PB larger than 25 and 50 for liquid and gas, respectively. In chapter three it wasdemonstrated that ∆PB can be designed in the range of 4 to 25 by following a specific de-sign methodology. The design methodology determines the maximum acceptable fabricationtolerance in the barrier channels, mixers and reaction channels.

Figure 4.2: Left, schematic of barrier-based gas-liquid flow distributor for four parallel mi-crochannels. Center, drawing of the BMMR showing its components. Right, the barrier-mixerchip and the meandering of the reaction channels. Symbols used are: (G) gas inlet, (L) liquidinlet, (M) manifold, (B) barrier channels, (T) T-mixer, (C) reaction channels,(BT) barrier-mixer chip, (I) inspection window, (O) collector block.

In this work, the barrier-based micro/milli reactor (BMMR) shown in Figure 4.2 was de-signed and fabricated according to the specific design methodology. The BMMR consists ofeight parallel reaction channels all operated in the Taylor flow regime. It is designed to holdpressure up to 20 bar and temperature up to 200 oC, however these two parameters are notexamined in this chapter. The BMMR is a modular type of reactor with a maximum liquidthroughput of 150 mL/min and gas to liquid flow ratio up to 10.

The BMMR demonstrates the numbering-up concept for gas-liquid Taylor flow possiblefor a production capacity reaching kg/h. In this chapter the quality of the flow distribution inthe BMMR is reported by studying two aspects. The first aspect is to experimentally examinesix different fluids with different viscosities, surface tensions and flow rates. The secondaspect is to study the reactor modularity by testing three different reaction channels type: (1)square channels fabricated in a stainless steel plate, (2) square channels fabricated in a glassplate, and (3) circular channels (capillaries) made of stainless steel. Finally, the BMMR is

Introduction 63

compared to that of a single channel regard the slug and bubble lengths and bubble generationfrequency. This chapter present the quality of flow distribution in the BMMR which is anelementary step before operating a reaction which is the next aim. In the next section thedesign methodology and fabrication are presented. This is followed by a description of theexperimental parts and operating conditions; then the results, and finally the discussion andconclusions.

4.1.1 Design and fabrication

The barrier-based micro/milli reactor was designed according to the design methodology aspresented in chapter three. The design is made to deliver flow non-uniformity in the parallelchannels of less than 10%. The main functional elements of the reactor are shown in Figure4.2:

The manifold (M): It is a triangular consecutive manifold made from stainless steel. Boththe gas and liquid manifold dimensions are equal as given in Table 4.1. The volume ofeach manifold is half that of the reaction channels. The flow passes from the inlet, to themanifold volume and then split and delivered to the barrier-channels through a transportchannels which were drilled in the manifold with an inner diameter of 2 mm.

The barrier-mixer chip (BT): This chip is made from glass and it contains the barrier-channels and T-mixer as shown in Figure 4.2. The gas and liquid from the manifolds aredelivered to the inlet of the barrier-channels. Taylor flow is generated in the T-mixer whichthen goes to the reaction channels through a transport channel. The glass chip is connected tothe manifold and the reactor using O-rings. Dimensions of the mixer and barrier channels aregiven in Table 4.1. Fully developed laminar flow is maintained before the fluid reaches themixers. The mixer and barrier channels were fabricated using powder blasting and chemicalwet etching (Micronit), respectively. The fabrication tolerance of the barrier channels weremeasured using nano optical profiler (Bruker) giving an accuracy in the depth as shown inFigure 4.3.

The reaction channels (C): The generated Taylor flow passes to each of the eight reactionchannels separately. Three types of reaction channels are fabricated as shown in Figure 4.4.The design of the reaction channels are arbitrarily made to cover different varieties of reactionchannels design. However, the channel diameters and lengths were adjusted, as given in Table4.1, to deliver similar pressure drops in all of them. Pressure drop in the reaction channels isthe key parameter to design the flow distributor.

The first channel type is square channels milled in a stainless steel plate and then closedby a metal sheet using brazing. Channels were fabricated in a meandering way as shown in


Figure 4.3: (a) Ultrasonic inspection for the brazing of the steel plate. (b) Histogram of themeasured depth of the barrier channels at several positions in the BT chip shown in Figure4.2 and for different chips.

Figure 4.4: Photographs of the BMMR. Top right is the fixed manifolds (M) and barrier-mixer chips (BT) which used to connect the three reaction channels type. (i) Stainless steelplate with a drawing for the meandering reaction channels; (ii) Glass plate with a drawing forthe meandering reaction channels; and (iii) Stainless steel capillaries.

Experiments and operating conditions 65

Table 4.1: Dimensions and Reynolds number for the barrier-based micro/milli reactor at anaverage operating condition of qL = 74 ml/min and qG/qL = 2. Superficial velocity of gasand liquid in the reaction channels are 0.2 and 0.1 m/s, respectively. Symbols refer to thoseexplained in Figure 4.2; Subscript G and L refer to the gas and liquid, T is for the inletchannels of the mixer. ∗ The width is decreasing by an 8 degree angle.

d , (mm) W, (mm) H, (mm) L , (mm) Re, (-) ∆P, bar

MG 6.4 41∗ 5 155 3 0.001ML 6.4 41∗ 5 155 20 0.001BG - 0.4 0.1 ± 0.001 340 65 1BL - 1.0 0.1 ± 0.001 37 183 1TG - 1.3 1.3 13 13 -TL - 1.3 1.3 10 78 -CPlate - 1.23 1.23 2000 245 0.15CGlass - 1.1 0.87 1500 307 -CCapillary 0.75 - - 667 403 -

Figure 4.2 and in Figure 4.4. The quality of the brazing was tested using ultrasonic inspection(a technique used to test welding) as shown in Figure 4.3. Excellent brazing is obtained. Tovisualize the slug and bubbles in the steel plate, an inspection window is made by directingthe flow to the top of the plate for a distance of 40 mm and then re-directing it back into thereaction channels in the steel plate. To measure the pressure drop over the reaction channelsindividually, an extra opening is made at the inlet and outlet of the reaction channels asshown in Figure 4.5. The second reaction channels type are square channels fabricated, usingpowder blasting (Lionix), in a glass plate with the dimensions given in Table 4.1 and shownin Figure 4.4(ii). The third reaction channels tested are the circular stainless steel capillaries.The steel capillaries were commercially available (Valco).

4.2 Experiments and operating conditions

The experiments were performed over a range of flow rates, surface tensions and viscositiesas given in Table 4.2. All chemicals were ordered from VWR International. The viscositywas measured using a falling piston viscometer. The surface tension was measured using atensiometer.

A process flow diagram of the experimental setup is shown in Figure 4.5. Liquid is beingpumped using a gear pump (NHK Mikrosysteme GmbH, MZR-7205) with a liquid mass flowcontroller (Bronkhorst). Nitrogen is fed from a gas bottle and controlled using a mass flow


Table 4.2: Density ρ , viscosity µ and surface tension γ of all six liquids used in the experi-ments in weight percentage. Liquid flow rate changes from 5 mL/min to 150 mL/min. Gasto liquid flow ratio changes from 0.5 to 5.

Fluid No ρ , (kg/m3) µ , (Pa.s) γ , (N.m−1)

100% Water 1 998 1.0 72.095% Water+5% Ethanol 2 989 1.5 52.480% Water+20% Ethanol 3 969 2.5 38.5100% Ethanol 4 789 1.6 22.370% Water+30% Glycerol 5 1072 2.5 70.350% Water+50% Glycerol 6 1126 6.0 69.1

Figure 4.5: Process flow diagram of the experimental setup and the locations of the pressuresensors. Symbol used: (i) liquid tank, (ii) gas bottle, (iii) gear pump, (iv) mass flow controller,(v) BT glass chip, (vi) pressure sensors at the inlet, (vii) manifold, (viii) reaction channelplate, (ix) inspection window, (x) differantial pressure sensor, (xi) connection block, and (xii)collector block. The dotted circles are enlarged view of the connection and channel of areaction channel in the steel plate.

controller (Bronkhorst). The pressure is measured at the manifold using a pressure sensor(range 0-25 bar, Endress+Hauser,PMP131). The pressure drop over the reaction channelsis measured using a differential pressure sensor in the range of 0-250 mbar (Sensortech-nics GmbH, 24PC). The bubble frequency in each barrier-mixer chip was measured using aportable stroboscope (Check.Line, DS-2000LED) which has a frequency range between 30- 300,000 FPM. By synchronizing the bubble generation frequency with that of the strobo-scope, it was possible to generate a static image of the slug and bubble of several unit cells ofTaylor flow consisting of liquid slugs separated with gas bubbles. A handheld digital micro-

Results 67

scope (Dino-Lite, AD413TL) was used to record the image. By calibrating the image pixelwith the width of the mixer channel, the slug and bubble lengths were measured in everychannel with an accuracy of ± 50 µm. Slug length was measured as the length between twoconsecutive bubble caps as shown in the Figure 4.6.

The measured bubble generation frequency and slug and bubble lengths allowed to calcu-late the bubble velocity per channel according to Equation 6.1. By quantifying the differencebetween the bubble velocities over the eight parallel channels, the flow non-uniformity wascalculated using the relative standard deviation according to Equations 4.3 and 4.4.

uB = f (LS +LB) (4.2)

σ(uB) =1

uB

√Σi(uB,i− uB)2

N−1100% (4.3)

uB =

i=N

∑i=1

uB,i

N(4.4)

The bubble velocity does not take into account the liquid film thickness as given in Equa-tion 4.5 (Warnier et al. (2008)). AB is cross section area of the bubble, A is the channel crosssection area, UL is the liquid superficial velocity, and UG is the gas superficial velocity. Equa-tion 4.5 shows that part of the liquid flow in the channel (the one in the liquid film) is nottaken into account when calculating the flow rate per channel. The amount of the liquid filmin the channel depends on the capillary numbers and on the mode of operation for Taylor flow(Abiev (2009), Taylor (1961)). In the flow range investigated here, Taylor flow is operatedin the recirculation mode and with a capillary number less than 0.04. Therefore the liquidfilm occupy less than 0.17 of the channel cross section area (Warnier et al. (2008)). Becausebubble velocity is the most convenient way to measure flow rate per channel and because theliquid film will exist in all channels, the amount of the liquid film is not accounted for in theflow non-uniformity calculations.

uB =A

AB(UG +UL) (4.5)

4.3 Results

The BMMR is a modular reactor which integrates all of the functional elements (distribu-tor, mixer, reaction channels, and heat exchanger) in a compatible and smooth way. The


modularity of the BMMR using three reaction channels type is shown by the photographsin Figure 4.4. Exchanging the reaction channels while keeping the same manifolds and thebarrier-mixer chips is relatively simple. The only fixed parameters in the BMMR are the out-side dimensions of the manifolds and barrier-mixer chips and the location for the openings ofthe barrier-mixer chips. The inside dimensions and material of constructions of the reactionchannels and the heat exchangers can be chosen freely. This is valid as long as the value ofthe pressure drop of the reaction channels matches the limits set by the design methodologygiven in chapter three. If this is not the case, fabrication of a new set of barrier-mixer glasschips is needed which can be made according to the mentioned design methodology.

Figure 4.6: Typical result of the slug (LS) and bubble (LB) lengths distribution in the BMMR.Slug length is the length between two consecutive bubble caps as shown in the figure. Resultshown is for 100% ethanol with nitrogen in the steel plate reaction channels with flow ratesequal to 5 mL/min and 10 mLn/min for liquid and gas, respectively.

For demonstration, a typical experimental result obtained using the steel plate is shown inFigure 4.6 for an experiment using 100% ethanol with nitrogen. At relatively low flow rates,the slugs and bubbles were captured in a single image at the inspection window. In all of theeight channels, slugs and bubbles were uniform and a stable Taylor flow was observed in thechannels. By varying the gas and liquid flow rates as mentioned in Table 4.2, the residencetime and specific interfacial area varied in the range of 1-120 (s) and 1000-5000 (m2/m3),respectively.

The flow non-uniformity was quantified using the relative standard deviation given in

Results 69

0 1 2 3 4 5 6 7 80.0

0.4

0.8

1.2

qL - qG 15 - 29 63 - 36 64 - 86 104 - 53 103 - 90 150 - 73 150 - 106 65 - 28 63 - 112

Rel

ativ

e (u

B) ,

(-)

Channel, (-)

Figure 4.7: Relative bubble velocity (divided by average velocity) per channel for the caseof ethanol-nitrogen over a wide rate of flow rates for gas (qG, mLn/min) and liquid (qL,mL/min).

Equation 4.3. To use that equation, the flow rate of each channel must be constant overtime. Fluctuation of flow rate over time was observed in some cases when bubble or slugcoalescence occur and when pump fluctuated specially at large flow rates. The degree of thatfluctuation was quantified by measuring the range of bubble generating frequency where fluc-tuations observed using the stroboscope. Fluctuations in the frequency were in all cases lessthan 3%. Because of that, fluctuation of flow rate of a channel over time was neglected andaverage bubble generating frequency was used instead. For a wide range of flow rates, theflow rate per channel is shown in Figure 4.7 to demonstrate the profile of the flow distribu-tion. The relative bubble velocity per channel for the case of nitrogen-ethanol flow is plottedover the eight parallel channels. In center channels, flow rate is the largest and decreaseselsewhere. Over the entire flow rate tested, profile of the flow distribution remains the same.However the broadness of that profile depended on the flow rate. Quantifying that broadnesswhich is the flow non-uniformity at varied conditions will be discussed in further details inthe next sections.

4.3.1 Liquid versus gas-liquid flow distribution

The first experiment to examine the flow non-uniformity is made by studying the influenceof flow rate for each phase separately. This was done in two separate experiments both us-ing ethanol-nitrogen flow in the steel plate. In the first, only liquid phase was measured bycollecting the outlet of each reaction channel into a separate vessel, then measuring the col-


lected weight over time. In the second experiment, experiment 1 was repeated but insteadthe gas and bubble velocity per channel was measured. The liquid flow non-uniformity isshown in Figure 4.8 (i), the non-uniformity in the bubble velocity shown in Figure 4.8 (ii)and the relative pressure difference shown in Figure 4.8 (iii). For the liquid phase only, theflow non-uniformity remains less than 3%. For the bubble velocity, the non-uniformity istwice larger; it is between 5 % and 10%. This demonstrates that the gas flow non-uniformityis twice larger than that of the liquid.

Figure 4.8: (i) Liquid flow non-uniformity σ(qL) and (ii) bubble velocity non-uniformityσ(uB) at varied gas and liquid (100% ethanol) flow rates. (iii) Experimental result of ∆PB.

The only difference between the gas and liquid manifolds and barrier sections of thereactor is the width of the barrier channel. In the fabrication process, wet chemical etching isused simultaneously to fabricate the barrier channels of the gas and liquid channels. Thereforethe absolute fabrication tolerance in the width for both channels is the same. But since thewidth of the gas barrier channel is 2.5 times less than that of the liquid barrier channel, therelative tolerance is larger. However one should keep in mind that for both experiments, thenon-uniformity remains within the acceptable margin of 10%. Moreover, in both experiments∆PB remains within the optimal range of 4 to 25 for the entire flow rate tested as shown inFigure 4.8 (iii).

Results 71

4.3.2 Stainless steel plate reactor - Effect of physical properties

The influence of viscosities, surface tensions and flow rates on the flow non-uniformity in thestainless steel plate are shown in Figure 4.9. All of these parameters were included in onedimensionless number, the capillary number CaB = µ.uB

γ . CaB is used because it contains theviscosity, surface tension, bubble generation frequency, and slug and bubble lengths. For allof the six fluids and for the entire range of flow rates, the flow non-uniformity remains withinthe acceptable flow non-uniformity of 10%, with two exceptions. The first is at large flowrate, when CaB approaches 0.04. The non-uniformity for the 50% glycerol and 30% glycerolapproaches the maximum limit of 10%. This can be explained by the influence of manifoldon the flow distribution. The flow non-uniformity in a consecutive type of manifold increasesas the flow rate or the viscosity increases (Griffini and Gavriilidis (2007), Pan et al. (2009)).

0.00 0.01 0.02 0.03 0.040

5

10

15

20 Water100% Ethanol 5% Ethanol 20% Ethanol 100% Glycerol 30% Glycerol 50%

(uB)

, (%

)

CaB, (-)

Figure 4.9: Bubble velocity non-uniformity for six fluids given in Table 4.2 versus capillarynumber CaB = µ.uB

γ . Ca is calculated as an average over the eight parallel channels.

The second exception (where the flow non-uniformity exceeds the 10%) is at low flowrate when CaB is less than 2.5x10−3. This exception can be explained by the wettabilityand the liquid film thickness. Before explaining that, it is important to notice that at lowflow rates (low CaB), ∆PB is the lowest. As ∆PB decreases, the influence of variations in thereaction channels (flow rates, slug and bubble lengths, and fabrication tolerance) on the flowdistribution increases. This relation was mathematically obtained in chapter three and givenin Equation 4.6.


σ(qC) =σ(∆PC)

∆PB(4.6)

σ(∆PC) is the variation in pressure drops over the reaction channels, and σ(qC) is theflow non-uniformity due to the flow rates and all variations in the mixers and reaction chan-nels. Keeping Equation 4.6 in mind, as the liquid flow rate or the viscosity decreases, theliquid film thickness decreases (Fries et al. (2008)). As the liquid film thickness decreasesand because there are sharp bends in the transport channels and reaction channels (see Figure4.5), it is possible that partially dry walls could form. The partially dry walls can induce bub-ble coalescence especially at lower slug lengths (when the length is similar or lower than thechannel diameter (Oztaskin et al. (2009))). Bubble coalescence generate pressure fluctuatingover the reaction channels (σ(qC) increase) and produces larger flow non-uniformity.

Figure 4.10: Steady state pressure drop of the eight reaction channels over time for 100%ethanol (i and iii) and 100% water (ii and iv). The operating conditions are qL = 14 mL/minand qG = 30 mLn/min for i and ii, and qL = 50 mL/min and qG = 130 mLn/min for iii and iv.Figure is printed in color.

Figure 4.10 shows the steady state pressure drop of the eight reaction channels over timefor 100% water and 100% ethanol. At low flow rate, the pressure drop for the 100% ethanol isvery smooth. Thus, uniform and stable Taylor flow is formed. Using 100% water, large fluctu-ations in pressure drops are observed which indicates that Taylor flow is not stable and bubble

Results 73

coalescence occurs which was also visually observed. As the flow rates increases, a smoothsteady state pressure drop is observed for both fluids. To maintain flow non-uniformity as lowas possible, assuring a good wetting in the channel where Taylor flow passes is mandatory.This was obtained when CaB is between 2.5x10−3 and 3.8x10−2.

4.3.3 Effect of reaction channel types and dimensions

The influence of modularity and reaction channels type on the flow non-uniformity are shownin Figure 4.11 and Figure 4.12 using 100% ethanol and 100% water, respectively. Using100% ethanol, the flow non-uniformity remains within the acceptable range of less than 10%.Using 100% water and at lower flow rate (CaB less than 0.002), the flow non-uniformity ofthe channels made of steel exceeds 10%. However for the glass plate, the non-uniformityremains less than 10%. The glass plate has smaller channel diameter and shows a betterwettability compared to the steel plate. That could explains why the flow non-uniformityremains less than 10% for the glass plate. The circular channels did not perform better thanthe square channels. Most probably this is due to the transport channels shown in Figure 4.5.Transport channels are the ones which transport Taylor flow from the barrier-mixer chip tothe reaction channels through a connector block made of stainless steel. Transport channelsare connected to the eight capillaries via capillary fittings. The connection contains bendsand sharp edges. It is possible that the wetting in the transport channels is not good, whichcould result in bubble coalescence. If bubble coalescence occurs, the pressure drop over thereaction channels starts to fluctuate significantly as shown in Figure 4.10. At low flow rate,the value of ∆PB is the smallest. Therefore, the interaction between the pressure fluctuationsand the flow distribution is the largest as shown in chapter two.

Stable and uniform Taylor flow was observed in the three reaction channels type for al-most the entire range examined here. This proofs that the choice to keep same pressure dropin the channels is the key for modularity to use same distributor for different reaction chan-nels and dimensions. In addition result shows that reaction channel geometry and dimensionhas no significant influence on flow distribution if pressure drop is maintained similar to eachother.

4.3.4 Comparison to single channel - Bubble generation frequency andslug and bubble lengths

The BMMR is compared to that of a single channel regard the bubble generating frequencyas a function of flow rates. In Figure 4.12, the bubble generation frequency f is plotted ver-sus the flow rate. The flow rate is represented by Reynolds number ReB to allow comparison


0.00 0.01 0.02 0.03 0.040

5

10

15

20 Steel plate Glass plate Capillaries

(uB)

, (%

)

CaB, (-)

Figure 4.11: Bubble velocity non-uniformity using 100% ethanol for the 3 reaction channelsgiven in Table 4.1 and shown in Figure 4.4 versus capillary number CaB = µ.uB

γ .

0.000 0.002 0.004 0.006 0.008 0.0100

5

10

15

20 Steel plate Glass plate Capillaries

(uB)

, (%

)

CaB, (-)

Figure 4.12: Bubble velocity non-uniformity σ(uB) using 100 % water for the 3 reactionchannels given in Table 4.1 and shown in Figure 4.4 versus capillary number CaB = µ.uB

γ .

Results 75

to that from the single channel results (Laborie et al. (1999)). Average values of f and ReB

are calculated over the eight parallel channels. The bubble generating frequency is a linearfunction of the Reynolds number. As the viscosity increases the slope increases in the samemanner as that of Laborie et al. (1999). The BMMR result matches with that of the singlechannel of Laborie et al. (1999). Therefore, even with the non-uniformity in the flow ratesand slug and bubble lengths, the BMMR reactor still shows similar performance to that of asingle channel.

0 250 500 750 10000

40

80

120

160 Water 100% Ethanol 5% Ethanol 20% Ethanol 100% Glycerol 30% Glycerol 50%

Bub

ble

freq

uanc

y , (

s-1)

ReB, (-)

Figure 4.13: Bubble generating frequency as a function of Reynolds number (Re = ρLduBµL

) forthe six fluids given in Table 4.2 .

In Figure 4.14, The BMMR is compared to that of a single channel regard the slug andbubble lengths as a function of the gas flow rates at a fixed liquid flow rate of 50 ml/min.As the gas flow rate increases, the slug length decreases while the bubble length increaseslinearly. The non-uniformity in the bubble velocity is plotted as a function of gas flow rate.As the flow rate increases, the non-uniformity decreases reaching a kind of minimum. Athigh flow rate, the slug lengths are lower than that of the reaction channels. Oztaskin et al.(2009) demonstrated that as the slug lengths is equal to or lower than the channel diameter,Taylor flow is not stable because the velocity profile in the liquid slug is not fully developed.The non-stable Taylor flow result in bubble coalescence. Thus fluctuation occurs in pressuredrops in the reaction channels, which result in larger flow non-uniformities.

As the viscosity and surface tension changed, there was no significant influence on theslug and bubble lengths. This is different than what was reported in literature (Fries et al.


(2008)) for studies made in a single channel. Most probably this is because of slug andbubble lengths non-uniformity over the eight parallel channels is comparable to those fromchanging viscosity and surface tension.

Figure 4.14: Upper, slug and bubble length as a function of gas flow rate at fixed liquid flowrate of 50 mL/min. Lower, Bubble velocity non-uniformity as a function of gas flow rate atfixed liquid flow rate.

4.3.5 The BMMR versus other barrier-based flow distributors operatedin the Taylor flow regime

A comparison is made between the BMMR versus two other earlier developed barrier-basedflow distributors as shown in Table 4.3. The other two distributors are the scaled-out chip ofDe Mas et al. (2005) and the micro bubble column MBC of Haverkamp et al. (2006). TheBMMR clearly outperformed the other two distributors. The maximum liquid flow rate andresidence time are 100 times higher in the BMMR while the pressure drop ratio ∆PB is at leasttwo times lower. Additionally, the BMMR provides a much wider flexibility in the choiceof fabrication technology, material of construction and modularity as demonstrated earlier inthis chapter. Therefore, the BMMR shows a great potential to fulfill the requirements neededto bring micro/milli channel reactors to industrially relevant production capacity.

Conclusion 77

Table 4.3: Comparison of three barrier-based distributors suitable for the Taylor flow regime:Scaled-out chip (De Mas et al. (2005)), micro bubble column MBC, (Haverkamp et al.(2006)), and this work, the BMMR. Dimensions of the channels are given as width x depth xlength.

Scaled-out chip MBC This work

Parallel channels 60 64 8Max qL,(mL/min) Less than 2 Less than 2 150Flow non-uniformity,(%) Less than 10 NA Less than 10Pressure drop ratio ∆PB,(-) Larger than 25 Larger than 25 4-25Max residence time Less than 1 sec Less than 1 sec 2 min

Dimensions, (mm)Barrier channels - Gas 0.10 x 0.07 x 6 0.005 x 0.02 x 0.6 0.4 x 0.1 x 340Barrier channels - Liquid 0.07 x 0.05 x 10 0.02 x 0.02 x 0.6 1.0 x 0.1 x 37Reaction channels 0.40 x 0.28 x 20 1.10 x 0.17 x 20 1.2 x 1.2 x 2000

4.4 Conclusion

The barrier-based micro/milli reactor has been successfully designed according to the method-ology purposed in chapter three to provide a flow non-uniformities of less than 10%. The flownon-uniformity is experimentally examined by studying two aspects. The first aspect is bychanging the viscosities, surface tensions and the flow rates for six different fluids. The sec-ond aspect is by studying the reactor modularity using three reaction channels type: (1) squarechannels fabricated in stainless steel plate, (2) square channels fabricated in glass plate, and(3) circular channels (capillaries) made of stainless steel. Finally the BMMR is comparedto that of a single channel regards the slug and bubble lengths and the bubble generatingfrequency. Conclusions obtained are:

1. The flow non-uniformity for the BMMR remains less than the acceptable margin of10% when: liquid flow rate changed from 10-150 mL/min, gas to liquid ratio 0.5 - 5,viscosity 1.25 - 6.71 (Pa.s), and surface tension 0.028 - 0.083 (N.m−1).

2. To maintain flow non-uniformity as low as possible and less than the 10%, assuring agood wetting in the channel where Taylor flow passes is mandatory. This was obtainedwhen CaB is between 2.5x10−3 and 3.8x10−2.

3. To prevent pressure fluctuations and reduce the flow non-uniformities, slug and bubblelengths should be larger than 2 times the channel diameters.


4. Reaction channel geometry and dimension have no significant influence on flow distri-bution as far as the pressure drop maintain the same. The key parameter for modularityover varied channel dimensions and geometries is the pressure drop. To exchange var-ious reaction channels using the same distributor (same barrier-mixer chips), similarvalue for the pressure drop is required.

In summary, this chapter presented the BMMR which demonstrate the numbering-up ofgas-liquid Taylor flow in microreactor suitable for a production capacity of kg/h. A uniformflow distribution is achieved at varied conditions even at larger viscosity which can be at-tractive for certain applications like sulfonation, biodiesel or polymerization reactions. Thestudy of the flow distribution is an elementary step made before performing a reaction in theBMMR which will be the next target.

BibliographyAbiev, R., 2009. Circulation and bypass modes of the slug flow of a gasliquid mixture in

capillaries. Theoretical Foundations of Chemical Engineering 43, 298–306.Amador, C., Gavriilidis, A., Angeli, P., 2004. Flow distribution in different microreactor

scale-out geometries and the effect of manufacturing tolerances and channel blockage.Chemical Engineering Journal 101, 379–390.



Bayer, T., Jenck, J., Matlosz, M., 2005. IMPULSE - a new approach to process design.Chemical Engineering & Technology 28, 431–438.


Charpentier, J. C., 2007. In the frame of globalization and sustainability, process intensifi-cation, a path to the future of chemical and process engineering (molecules into money).Chemical Engineering Journal 134, 84–92.



Fries, D. M., Trachsel, F., von Rohr, P. R., 2008. Segmented gas-liquid flow characterizationin rectangular microchannels. International Journal of Multiphase Flow 34, 1108 – 1118.

Bibliography 79

Gavriilidis, A., Angeli, P., Cao, E., Yeong, K. K., Wan, Y. S. S., 2002. Technology andapplications of microengineered reactors. Chemical Engineering Research and Design 80,3–30.

Griffini, G., Gavriilidis, A., 2007. Effect of microchannel plate design on fluid flow unifor-mity at low flow rates. Chemical Engineering & Technology 30, 395–406.

Gunther, A., Jensen, K. F., 2006. Multiphase microfluidics: from flow characteristics tochemical and materials synthesis. Lab on a Chip 6, 1487–1503.

Gupta, R., Fletcher, D. F., Haynes, B. S., 2010. Taylor flow in microchannels: A review ofexperimental and computational work. The Journal of Computational Multiphase Flows 2,1–32.


Hasebe, S., 2004. Design and operation of micro-chemical plants–bridging the gap betweennano, micro and macro technologies. Computers & Chemical Engineering 29, 57–64.

Haverkamp, V., Hessel, V., Loewe, H., Menges, G., Warnier, M. J. F., Rebrov, E. V., de Croon,M. H. J. M., Schouten, J. C., Liauw, M. A., 2006. Hydrodynamics and mixer-inducedbubble formation in micro bubble columns with single and multiple-channels. ChemicalEngineering & Technology 29, 1015–1026.


Hessel, V., Angeli, P., Gavriilidis, A., Loewe, H., 2005. Gas-liquid and gas-liquid-solid mi-crostructured reactors: Contacting principles and applications. Industrial and EngineeringChemistry Research 44, 9750–9769.



Kockmann, N., Gottsponer, M., Roberge, D. M., 2011. Scale-up concept of single-channelmicroreactors from process development to industrial production. Chemical EngineeringJournal 167, 718 – 726.


Kralisch, D., Kreisel, G., 2007. Assessment of the ecological potential of microreaction tech-nology. Chemical Engineering Science 62, 1094–1100.

Laborie, S., Cabassud, C., Durand-Bourlier, L., Laine, J. M., 1999. Characterisation of gas-liquid two-phase flow inside capillaries. Chemical Engineering Science 54, 5723–5735.

Marre, S., Adamo, A., Basak, S., Aymonier, C., Jensen, K. F., 2010. Design and packag-ing of microreactors for high pressure and high temperature applications. Industrial andEngineering Chemistry Research 49, 11310–11320.


Mendorf, M., Nachtrodt, H., Mescher, A., Ghaini, A., Agar, D. W., 2010. Design and controltechniques for the numbering-up of capillary microreactors with uniform multiphase flow


distribution. Industrial and Engineering Chemistry Research 49, 10908–10916.Natividad, R., Cruz-Olivares, J., Fishwick, R. P., Wood, J., Winterbottom, J. M., 2007.

Scaling-out selective hydrogenation reactions: From single capillary reactor to monolith.Fuel 86, 1304 – 1312.

Oztaskin, M. C., Worner, M., Soyhan, H. S., 2009. Numerical investigation of the stability ofbubble train flow in a square minichannel. Physics of Fluids 21, 042108.


Pennemann, H., Watts, P., Haswell, S., Hessel, V., Loewe, H., 2004. Benchmarking of mi-croreactor applications. Organic Process Research & Development 8, 422–439.

Razzaq, T., Glasnov, T. N., Kappe, C. O., 2009. Accessing novel process windows in a high-temperature/pressure capillary flow reactor. Chemical Engineering Technology 32, 1702–1716.


Roberge, D. M., Ducry, L., Bieler, N., Cretton, P., Zimmermann, B., 2005. Microreactortechnology: A revolution for the fine chemical and pharmaceutical industries? ChemicalEngineering & Technology 28, 318–323.


Song, H., Chen, D. L., Ismagilov, R. F., 2006. Reactions in droplets in microfluidic channels.Angewandte Chemie 45, 7336–7356.

Taylor, G. I., 1961. Deposition of a viscous fluid on the wall of a tube. Journal of FluidMechanics 10, 161–165.




Warnier, M. J. F., Rebrov, E. V., de Croon, M. H. J. M., Hessel, V., Schouten, J. C., 2008. Gashold-up and liquid film thickness in Taylor flow in rectangular microchannels. ChemicalEngineering Journal 135, S153 – S158.



Designing flow and temperatureuniformities in parallelmicrochannels reactor 55

This chapter will be submitted as:Al-Rawashdeh, M., Nijhuis, T.A, Rebrov, E.V., Hessel, V. & Schouten,J.C. 2013. Designing flow and temperature uniformities in parallel mi-crochannels reactor. AIChE Journal.

AbstractA design methodology is proposed to maintain gas and liquid flow non-uniformities below anacceptable limit in a parallel micro/milli channels reactor. The design methodology providesa cut-off value of the maximum allowed temperature deviation in each part of the reactor. Theeffect of temperature deviation on flow distribution was quantified using a hydraulic resistivenetwork model. The effect of flow rate on the temperature deviation was demonstrated us-ing a one dimensional energy balance model. Experiments were conducted to validate thesemodels using the barrier-based micro/milli channels reactor (BMMR). Flow distribution inthe BMMR is based on the barrier channels concept by placing extra hydraulic resistancesin the gas and liquid manifolds to passively regulate the flows. Temperature deviation in thebarrier channels affects flow non-uniformity by ten times more than in the mixer and reactionchannels. Above a certain critical liquid residence time, flow rate has no significant effecton the temperature deviation. The critical liquid residence time depends on the liquid used,reactor material of construction, and its geometrical dimensions. The design methodologyprovides engineering steps which gives a first estimation to reach the desired flow and tem-perature uniformities in the BMMR. Such a methodology is applicable if numbering-up isused to scale-up microreactors to large scale capacities specially for microreactor applica-tions which involve multiphase flow.

82 Designing flow and temperature uniformities in parallel microchannels reactor

5.1 Introduction

Microstructured reactors are viable reactors for highly exothermic and mass transfer limitedreactions. This makes them attractive devices to improve safety, reduce waste, and enhanceproduct selectivity, as well as conversion (Anxionnaz et al. (2008), Charpentier (2007), Hes-sel et al. (2011), Kockmann and Roberge (2009), Wiles and Watts (2012)). For a singlemicrochannel reactor, the flow rate is often in the range of mL/min which is suited for a g/hproduction rate. Scale-up is required to reach kg/h and ton/h production rates. Scale-up routein microchannel reactors can be achieved in three consecutive steps as given in chapter one.First, the microchannel cross sectional dimensions are scaled-up while maintaining the massand heat transfer properties of the single microchannel (Hornung et al. (2007), Kockmannet al. (2011)). Second, numbering-up is achieved by placing multiple channels in parallel inone modular unit. The third step is made by placing multiple modular units in parallel.

For heterogeneously catalyzed gas phase reactions in microreactors, scale-up and num-bering up to thousands of parallel channels was successfully demonstrated (Hessel et al.(2008), Lerou et al. (2010)). Also, for liquid phase reactions, scale-up was successfullydemonstrated, but only for a limited number of parallel channels due to the complexity offlow distribution (Roberge et al. (2009), Saber et al. (2009), Schenk et al. (2003)). For gas-liquid processing, the throughput is still mostly restricted at the laboratory scale at g/h due tothe complexity and expenditure needed for an adequate numbering-up with a uniform flowdistribution (Chambers et al. (2005), Kashid et al. (2010), Mendorf et al. (2010), Natividadet al. (2007), Yue et al. (2010)).

It is difficult to achieve exact uniform flow distribution in numbered-up microchannelreactors. Usually an acceptable target for flow non-uniformity is defined according to theprocess and the required quality of the product. The design principle for most flow distribu-tors relies on controlling the hydraulic flow resistances in the device. This is described by thepressure drop in each of the parallel channels. In microfluidic devices, the resistive networkRN model is often used to predict the influence of these resistances (Amador et al. (2004),Commenge et al. (2002)). The RN model is however only valid for very low Reynolds num-ber in the laminar flow range, which is the case for most microfluidic devices. The pressuredrop in a microchannel is the product of the hydraulic resistance (R) and the flow rate (q) asgiven in Equation 5.1.

∆P = R q (5.1)

For single phase flow, the Hagen-Poiseuille equation can be used to estimate the pressuredrop for a given flow rate q (Commenge et al. (2002)). In that case, the hydraulic resistance

Introduction 83

R is given in Equation 5.2:

R =32µLλNC

d2A(5.2)

where µ is the viscosity, L is the channel length, λNC is a non-circularity factor that dependson the channel geometry, A is the channel cross section area, and d is the hydraulic diame-ter (Amador et al. (2004)). The hydraulic resistance is a function of the channel hydraulicdiameter to the power four. Therefore, channel diameter has a significant influence on flowdistribution. The second parameter which influences the hydraulic resistance is the viscositywhich changes with the fluid composition and temperature. A few studies have been madeto investigate the effect of flow and temperature distribution in microreactors (Hornung et al.(2007), Mason et al. (2007), Rebrov et al. (2011)). This is because many phenomena playa role in such a study: flow hydrodynamics, heat exchanger design and reaction kinetics.Quantifying the interplay between these phenomena, requires advanced analytical techniquessuch as computational fluid dynamics (CFD). Additionally, modeling of multiphase flow inparallel microchannels increases the complexity of the CFD modeling dramatically. Hav-ing an alternative and a more simple technique to study the effect of flow and temperaturedistribution is therefore a valuable asset for microreactor designs specially for microreactorapplications which involve multiphase flow.

Figure 5.1: (i) A schematic view of the BMMR with 4 parallel microchannels. Symbols usedare G gas, L liquid, M manifold, B barrier channel, C gas-liquid microchannel. (ii) Picture ofthe BMMR which consists of 8 parallel microchannels. (iii) An enlarged view of the glasschip which has the gas and liquid barrier channels and the T mixer. (iv) A picture of Taylorflow in the eight parallel microchannels.

The barrier-based micro/milli channel reactor (BMMR) is a structured multiphase mi-croreactor which reaches large scale production via numbering-up (Al-Rawashdeh et al.(2012), De Mas et al. (2005), Wada et al. (2006)). Uniform flow distribution is achievedin the BMMR by placing hydraulic resistances, so called barrier channels (B), between theparallel microchannels (C) and the separate gas and liquid feeding manifolds (M) as shown


in Figure 5.1. This approach has the following advantages: (1) gas-liquid channeling is pre-vented, (2) all flow regimes, viz. Taylor, churn and annular can be successfully realized, and(3) the flow uniformity is substantially improved.

The hydraulic resistances of the barrier channels can be quantified in a generic way as∆PB as given in Equation 5.3. It is the average pressure drop over the barrier channels, ∆PB,divided by the average pressure drop over the corresponding mixers and microchannels, ∆PC.Since ∆PB is a ratio of pressure drops, it is dimensionless.

∆PB =∆PB

∆PC(5.3)

Early studies in the field have demonstrated this type of distributor in microchannel re-actors (De Mas et al. (2005), Wada et al. (2006)). Their designs were successfully run withbarrier channels dimensions giving ∆PB larger than 25 and 50 for liquid and gas, respec-tively. In chapter three, we have demonstrated that ∆PB can be designed in the range of 4 to25 by following a specific design methodology. Cut-off values of the maximum acceptablefabrication tolerance in the barrier channels, mixers and reaction channels diameters weredetermined using the proposed design methodology.

The aim of this chapter is to study the effect of temperature and flow distribution forgas-liquid Taylor flow in the BMMR. A simpler approach is proposed compared to that ofa CFD. Firstly, a pre-defined temperature deviation in the parallel microchannels is applied.Then, the RN model is used to estimate the influence of temperature on the flow distribution.Secondly, one dimensional energy balance model is used to study the influence of flow onthe temperature deviation. The models are validated experimentally using the BMMR. Usingthese results, a design methodology is proposed to determine the maximum allowed tempera-ture deviation in the parallel microchannels to keep flow non-uniformity below an acceptablelimit.

5.2 Modeling and methodology

5.2.1 Temperature non-uniformity

In microchannels reactor, there are several possible temperature non-uniformities. For sake ofsimplification, temperature non-uniformities are represented by three temperature deviationfactors as will be explained next. These factors will be used to study the effect of temperatureon flow distribution.

Modeling and methodology 85

Figure 5.2: (i): Schematic cross sectional view of a single channel BMMR showing themanifolds M, barrier channels B, mixer and reaction channel C, gas G and liquid L. (ii)Schematic top cross view of BMMR for four parallel microchannels. Average temperaturefor M, B, and C are shown in three different colors for microchannel number one.

The BMMR is divided into three parts: manifold M, barrier channels B, and mixer andreaction channels C represented by three colors as shown in Figure 5.2(i). One average tem-perature Tavg is used in each part and per parallel microchannel as shown in Figure 5.2(ii).As a result, a number of N different average temperatures is obtained for each of the BMMRparts. Quantifying the difference between these average temperatures gives the three tem-perature deviation factors as σ(TM), σ(TB), and σ(TC) for manifold, barrier, and reactionchannels, respectively. σ(T ) is obtained according to Equation 5.4, where Tavg,i is the aver-age temperature for microchannel number i for any of the BMMR parts: M, B, or C; Tavg isthe average temperature for any of the parts, M, B, or C, over the entire number of parallelmicrochannels N.


σ(T ) =

√Σi(Tavg,i− Tavg)2

N−1(5.4)

5.2.2 Resistive network model

In chapter three, the RN model was extended to account for multi-phase flow which wascalled the 2-phase resistive network (2-PRN) model. The 2-PRN model was developed forgas-liquid Taylor flow and was experimentally verified. For a given temperature deviation, the2-PRN model can be used to evaluate the effect of temperature on flow distribution. However,the 2-PRN model has some limitations. First, it is valid only at low Reynolds number. Second,the 2-PRN model cannot evaluate the effect of flow on temperature deviation. Third, only onevalue for the temperature can be used for each hydraulic resistance independent of the axiallength and cross sectional temperature profiles. Due to these limitations, an energy balancewas developed to study the effect of flow rate on temperature deviation in the BMMR.

5.2.3 Energy balance model

The effect of flow rate on the temperature deviation was accounted for by using the one di-mensional energy balance model as shown in Figure 5.3. The energy balance was appliedfor a single microchannel C through which gas-liquid Taylor flow was passing without anyreaction. A uniform heat supply Qsupply was applied to the single microchannel C while noheat was supplied to the other parts of the BMMR. Heat exchange between the gas-liquidmicrochannel C and the other parts of the BMMR was taken into account via heat conductionas will be explained later. By changing the flow rate, temperature profiles were generatedwhich allowed studying the effect of flow rate on temperature deviation.

Few assumptions were made to construct the energy balance model: Constant physicalproperties; plug flow for both gas and liquid, negligible axial dispersion; no reaction in thegas-liquid Taylor flow; heat supply to the reaction channel only at a fixed reference tempera-ture along the axial direction; no heat transfer takes place in the direction perpendicular to theaxial direction; Fixed heat loss to the surroundings and heat conduction to the other BMMRparts.

The first energy balance was made for the gas-liquid phase in the microchannel and itconsisted of three terms as shown in Figure 5.3 and in Equation 5.5. The temperature profilein the microchannel was determined by the convective inlet and outlet heat transport (Qin andQout ) and heat exchange at the channel wall (Qexch). Usually the heat taken by the gas phase


Figure 5.3: (i) Axial temperature profiles for gas-liquid Taylor flow, microchannel wall, andfor the water heating bath. (ii) A drawing of the terms in the energy balance for the gas-liquid microchannel (G-L) and the microchannel wall which are given in Equation 5.5 and inEquation 5.10, respectively.

was less than one percent compared to that taken by the liquid phase. Therefore, energybalance over the gas phase was neglected. However, the influence of the gas flow rate wasstill included when computing the gas-liquid heat transfer coefficient.

Qout −Qin = Qexch (5.5)

m cpdTdx

= hwc Ac (Tw−T ) (5.6)

The energy balance was made for the liquid phase in the microchannel as shown in Equa-tion 5.6, where m is the mass flow rate, Cp is the heat capacity of the liquid, hwc is the overallgas-liquid-solid heat transfer coefficient (W m−2 K−1), and Ai is the heat exchanger surfacearea (m2). The overall gas-liquid-solid heat transfer coefficient was estimated using the heatresistance in series model as shown in Equation 5.7. It consisted of the heat transfer resistancein the channel walls ( dw

kw) and the heat transfer resistance in the gas-liquid flow ( 1

hGL).

1hwc

=1

hGL+

dw

kw(5.7)

To calculate the heat transfer coefficient hGL the entrance effects were neglected and acorrelation for the Nusselt number for gas-liquid Taylor flow was taken from literature (Leunget al. (2012)) as shown in Equations 5.8 and 5.9.


Nu =Ls

Ls +LB(4.364+

a1

L∗s +a2L∗1/3s

) (5.8)

L∗s =Ls

ReGL Pr d(5.9)

where a1 = 0.29, a2=0.15, Nu = hGLdk is the Nusselt number for gas-liquid Taylor flow, d

is the channel diameter (m), k is the thermal conductivity of liquid (W/m.K), Ls and Lb arethe slug and bubble lengths (m), ReGL is gas-liquid Reynolds number ReGL = dρuGL

µ and Pr

is the Prandtl number Pr = Cpµk .

The second energy balance was made over the microchannel wall. The energy balanceconsisted of four terms as shown in Figure 5.3 and in Equation 5.10: heat transfer from the hotfluid to the channel wall, Qsupply, heat transfer from the channel wall to the gas-liquid Taylorflow, Qexch, heat loss by conduction to the adjacent devices (manifolds and barrier channels),Qconduction, and heat loss by convection to the surroundings, Qloss. Axial heat dispersion wasneglected because uniform heat supply at a fixed temperature was maintained. The energybalance over the microchannel wall was written as given in Equation 5.11. Explanationsabout the used heat transfer coefficients are given next.

Qsupply−Qexch−Qconduction−Qloss = 0 (5.10)

hmcAsupp (Th−Tw)−hwc Ac (Tw−T )−hcond Acond (Tw−Tjoint)−hloss Aloss (Tw−Tsurr) = 0(5.11)

hmc is the heat transfer coefficient for the heat transfer from the heating source to the mi-crochannel wall. hmc depends on the type of the heat exchanger. In the experimental setup,heat was supplied by putting the gas-liquid microchannel in a water heating bath. Thus, hmc

was assumed as heat transfer coefficient on a free surface, a value of 500 W/m2.K was used(Haber et al. (2012)). hcond is the heat transfer coefficient for the heat transfer between thegas-liquid microchannel and the BMMR assembly (inlet tubes, manifolds, barrier channelsand collectors) which depends on the conductivity of the channel walls and characteristictransfer length. hcond was estimated by dividing the conductivity of the microchannel wallmaterial by the length of gas-liquid microchannel. hloss is the heat transfer coefficient forexternal heat losses which depends on the external surface of the device without insulation.A value equals to 5 W/m2.K was assumed.


Equation 5.7 was solved simultaneously with Equation 5.11 to generate axial tempera-ture profiles for the liquid and the microchannel wall depending on the used flow rates. Thechannel dimensions, geometry, material of construction and flow rates are given in the exper-imental section.


In chapter three, a design methodology was developed for estimating the maximum allowedvariation in the channel diameter and required barrier channel dimensions to stay below a tar-get flow non-uniformity. The same design methodology was used here to quantify the effectof temperature on flow distribution. The working principle of the methodology is to decouplethe effect of temperature deviation of each part of the BMMR, manifolds M, barrier channelsB, and mixers and reaction channels C, alone and then combine their contribution to cumula-tive flow non-uniformity. Full details about the methodology is given in chapter three.

A temperature deviation, quantified by Equation 5.4, was applied to one particular partof the BMMR, M, B, or C, while the temperature of the other parts maintained at a constantand fixed temperature. Hydraulic resistances were calculated and the 2-PRN model was usedto compute the flow distribution. Next, the flow rate qi,a f ter of each microchannel was nor-malized according to Equation 5.12 by its respective flow rate qi,be f ore before applying thetemperature deviation. By using Equation 5.13, the flow non-uniformity factor of that par-ticular BMMR part was then calculated; where ¯q is the average flow rate of the normalizedflow rate and N is the number of channels. The cumulative contribution of the three flow non-uniformity factors gave the overall gas or liquid flow non-uniformity as shown in Equation5.14. σ(q) is the target gas or liquid flow non-uniformity quantified according to Equation5.13. The three flow non-uniformity factors are: (1) σ(qT,M) for the effect of temperaturedeviation in the manifold part, σ(TM), (2) σ(qT,B) for the effect of temperature deviation inthe barrier channels part, σ(TB), and (3) σ(qC) for the effect of temperature deviation in themixer and reaction channels part, σ(TC).

qi =qi,a f ter

qi,be f ore(5.12)

σ(q) =1¯q

√Σi(qi− ¯q)2

N−1100% (5.13)

σ(q) =√

σ2(qT,M)+ σ2(qT,B)+σ2(qT,C) (5.14)


5.3 Experimental section

The experiments were conducted using in the BMMR shown in Figure 5.1. The reactionchannels are square fabricated in a stainless steel plate. Key dimensions of the BMMR aregiven in Table 5.1. Design and fabrication details of the BMMR can be found in chapter four.

Table 5.1: Dimensions for the barrier-based micro/milli reactor BMMR. ∗ The width is de-creasing by an 8 degree angle. The thickness of the microchannl wall is 2 mm.

W, (mm) H, (mm) L , (mm)

Inlet gas manifold 41∗ 5 155Inlet liquid manifold 41∗ 5 155Gas barrier channel 0.4 0.1 340Liquid barrier channel 1.0 0.1 37Inlet gas T-mixer 1.3 1.3 13Inlet liquid Mixer 1.3 1.3 10Reaction channel 1.23 1.23 2000

The experiments were performed using water and nitrogen at different flow rates. The liq-uid flow rate was changed in the range of 10-115 mL/min and the gas flow rate from 10 to 230mLn/min. A process flow diagram of the experimental setup is shown in Figure 5.4. Liquidwas pumped using a gear pump (NHK Mikrosysteme GmbH, MZR-7205) via a liquid massflow controller (Bronkhorst). Nitrogen was fed from a gas bottle in a controller manner byusing a mass flow controller (Bronkhorst). The pressure was measured at the manifold usinga pressure sensor (range 0-25 bar, Endress+Hauser,PMP131) and temperature was measuredusing a thermocouple type pt100.

Table 5.2: Values used for heat transfer experiments. Density ρ , viscosity µ , heat capacitycp, water thermal conductivity k, steel thermal conductivity kw.

ρ , kg/m3 µ , Pa.s cp , J/kg.K k, W/m.K kw, W/m.K

1000 0.001 4181.3 0.57 16

Heat was supplied by placing the reaction channels in a water heating bath as shown inFigure 5.4. The other parts of the reactor assembly were outside the water heating bath. Thevapors from the water heating bath obscured the vision. Thus, it was not possible to measureslug and bubble lengths and bubble generation frequency. Instead, the flow distribution in theliquid phase was measured using the weight method. The tubes between the reactor and the

Experimental section 91

Figure 5.4: Process flow diagram of the experimental setup and the locations of the temper-ature and pressure sensors. Symbol used: (i) liquid tank, (ii) gas bottle, (iii) gear pump, (iv)mass flow controller, (v) BT glass chip, (vi) temperature sensors at inlet, (vii) manifold, (viii)reaction channel plate, (ix) inspection window, (x) temperature sensors at outlet, (xi) connec-tion block, and (xii) collector block. The thick line represents the level in which the reactorwas immersed in the water bath. Bottom: A picture of the BMMR on top of the water bath.

collector (xii) were disconnected. The liquid from each channel flowed into a separate vessel.The collected liquid volume in each vessel was weighed after reaching the steady state con-dition. Flow non-uniformity was calculated using Equation 5.13. The flow distribution wasdetermined at two water bath temperatures, 20oC and at 70oC. Measuring flow distribution athigher temperatures was not possible due to the high vapor pressure because water had to becollected in an open vessel. Over the entire tested flow range, temperature non-uniformity atthe outlet of the eight microchannels was measured.


5.4 Results

5.4.1 Effect of temperature on the pressure drop ratio, ∆PB

Any changes to the viscosity will affect the pressure drop which is the key parameter inflow distribution. Theoretically, the effect of temperature on nitrogen and water viscositiesis shown in Figure 5.5(a) (Chung et al. (1988), Goletz and Tassios (1977)). Increasing thetemperature decreases the water viscosity and increases the nitrogen viscosity. Such an effectwill result in larger pressure drop ratio ∆PB at the gas side compared to the liquid side. Ifthe difference between ∆PB of the gas and liquid sides is too large, gas-liquid channelingis expected as demonstrated in chapter two. Therefore, the range of operating temperatureshould be specified for each barrier channels dimensions in the BMMR.

20 40 60 80 1000.0

0.5

1.0

1.5

2.0 G L

Rel

ativ

e vi

scos

ity, (

%)

T, (oC)

Figure 5.5: Theoretical effect of temperature on nitrogen and water viscosities taken fromChung et al. (1988) an Goletz and Tassios (1977), respectively.

5.4.2 Effect of temperature on flow distribution

5.4.2.1 The manifold flow non-uniformity factor, σ(qT,M)

Temperature deviation in the manifold has a negligible effect on flow distribution. The man-ifolds hydraulic resistances, RM,i, are the smallest if compared to other hydraulic resistancesin the BMMR. This is made on purpose to reduce the influence of manifold on the flow distri-bution. The manifolds cross sectional dimensions were increased to an acceptable limit usingthe 2-PRN model.

Results 93

5.4.2.2 The barrier flow non-uniformity factor, σ(qT,B)

The effect of temperature deviation in the barrier channels on flow distribution is shown inFigure 5.6. The study was conducted at different barrier channel dimensions by reducingthe barrier channel depth. Temperature deviation in the barrier channels has a significantinfluence on flow distribution. The influence is linear with a slope larger than one. Forexample, for a temperature deviation σ(TB) of 4 oC, flow non-uniformity reaches 16%. Astemperature deviation increases, the flow distribution slightly depends on the used pressuredrop ratio ∆PB. But since temperature deviation in the barrier channels must remain very low,due to its significant influence on flow distribution, the influence of ∆PB is neglected.

0 4 8 12 160

10

20

30

40

50~PB ,(-)

22 10 7 5

(qT,

B),

(%)

(TB), (oC)

~

Figure 5.6: Effect of temperature deviation in the barrier channels on flow distribution. Studyis made at different hydraulic resistance in barrier channels ∆PB and at an average temperatureof 20 oC.

When the study in Figure 5.6 is repeated but at higher average temperature of 50 oC and85 oC, the temperature deviation in the barrier channels has a milder influence as shownin Figure 5.7. This is due to the non-linear relation of liquid viscosity with temperature asshown in Figure 5.5. The relation between temperature deviation in the barrier channel andflow distribution is represented in Equation 5.15, where f1(Tdesigned ,µL) is a constant thatdepends on the designed temperature and liquid viscosity.

σ(qT,B) = f1(Tdesigned ,µL)σ(TB) (5.15)


0 4 8 12 160

10

20

30

40

50 T, (oC)

20 50 85

(qT,

B),

(%)

(TB), (oC)

_

~

Figure 5.7: Effect of temperature deviation in the barrier channels on flow distribution. Studyis conducted at different average temperature of 20 oC, 50 oC, and 85oC all with a fixed ∆PBof 10.

5.4.2.3 The mixer and reaction channel flow non-uniformity factor, σ(qC)

The effect of temperature deviation in the mixer and reaction channels is shown in Figure 5.8.As temperature deviation increases, flow non-uniformity increases depending on the pressuredrop ratio, ∆PB. The effect of temperature deviation on flow non-uniformity reduces at larger∆PB. For the entire tested range of temperature deviation, the flow non-uniformity remainsless than 10%. This effect of temperature deviation on flow distribution is much lower whenit is in the mixer and reaction channels than in the barrier channels. For example, for the sametemperature deviation, the barrier channels affect flow distribution 10 times more than in themixer and reaction channels.

σ(qC) =σ(∆PC)

∆PB(5.16)

Temperature deviation effect on flow distribution depends slightly on the operating tem-perature as shown in Figure 5.8 when changing from 20 oC or 50 oC. In chapter three, Equa-tion 5.16 was found to describe accurately the variational effects in the mixer and reactionchannels on flow distribution. Here, Equation 5.16 is found also to describe accurately theeffect of temperature deviation in the mixer and reaction channels on flow distribution. Equa-tion 5.16 is a generic correlation because it is given in term of pressure drops which includesall variations in the mixer and reaction channels, such as: diameter, temperature, slug andbubble lengths or other factors.

Results 95

0 4 8 12 160.0

2.5

5.0

7.5

10.0PB, (-)

3 @ 50 oC 3 5 7 10 22

(qT,

C),

(%)

(TC), (oC)

~

~

Figure 5.8: Effect of temperature deviation in the mixer and reaction channels on flow distri-bution. Study is made at different hydraulic resistance in the barrier channels.

5.4.3 Numerical validation of the design methodology- Effect of all com-bined contributions on flow distribution

In this section, the combined contribution from all flow non-uniformity factors will be demon-strated. The gas or liquid flow non-uniformity can be obtained as a cumulative contribution ofall flow non-uniformity factors as given in Equation 5.17. In this study, the effect of temper-ature deviation in the barrier via σ(qT,B) and in the mixer and reaction channels via σ(qT,C)will be accounted for, while the effect of temperature deviation in the manifold σ(qT,M) willbe neglected. In addition to these two factors, three other flow non-uniformity factors areincluded in Equation 5.17. These are the flow non-uniformity factors due to variations inchannel diameters which always exist in the BMMR as demonstrated in chapter four. Thesethree flow non-uniformity factors are: (1) The flow non-uniformity factor due to the manifoldtype and dimensions, σ(qM); (2) The flow non-uniformity factor due to the variation in thediameter of the barrier channels, σ(qB); and (3) the flow non-uniformity factor due to thevariation in the internal diameter of the mixer and the reaction channels, σ(qC).

σ(q) =√

σ2(qM)+ σ2(qB)+σ2(qC)+ σ2(qT,B)+σ2(qT,C) (5.17)

The target from Equation 5.17 is that the cumulative contribution of all flow non-uniformityfactors should give an overall flow non-uniformity below an acceptable limit. For demonstra-tion, a maximum acceptable overall flow non-uniformity of 10% was used here. To achievethis target, first a weight was given to each flow non-uniformity factor in Equation 5.17 to


assure that the target overall flow non-uniformity remained below 10%. Then, the assignedweight to each flow non-uniformity factor was used as a target from which the maximumallowed variation in diameter and temperature was calculated. For any of the individual flownon-uniformity factors, there was no exact weight which can be assigned. Therefore, therewere different design options which can be generated. For example, Table 5.3 shows threedifferent design options generated to give cut-off values of the maximum allowed variationin diameters and temperatures of the BMMR to maintain flow non-uniformity below 10%.These three design options were given here for demonstration while other design optionswere possible to generate. Such a variety of the design options gives the designer and thefabrication company extra room and flexibility to find the most realistic and optimal designdimensions.

Table 5.3: Cut-off values of the maximum allowed variation in diameters and temperaturesin the BMMR to maintain flow non-uniformity below 10%. *Manifold flow non-uniformityfactor was assumed less than 1%. Average temperature is 20 oC.

Variable* Option 1 Option 2 Option 3

σ(dB) (%) 2 1 2.3σ(dMixer) (%) 10 15 5σ(dC) (%) 15 25 10σ(TB) (oC) 1 0.5 1.5σ(TC) (oC) 5 12 8

Analytical validation for one of the design options is given next using the 2-PRN model.For demonstration, option 1 in Table 5.3 was inserted in the 2-PRN model. The model wasthen used to compute the overall flow non-uniformity which compared to the target value of10%. The variations in diameter and deviations in temperature are shown in Figures 5.9(c)and (d), respectively. The lines represent average values of these variations. The simulationwas run and repeated 10000 times. In each iteration, random values for diameter and tem-perature were chosen as explained in chapter three with a maximum variation less than thecut-off values specified in Table 5.3. For both gas and liquid, more than 95% of the iterationsgave an over all flow non-uniformity less than 10% as shown in Figure 5.9(a). The remaining5% of the iterations slightly deviated from the target but the maximum obtained overall flownon-uniformity did not exceed 15%. The average pressure drop ratio ∆PB of both gas andliquid remained around a value of 6 as is shown in Figure 5.9(b).

Results 97

Figure 5.9: Effect of 10000 random variations in channel diameter and temperature on (a) gasand liquid flow non-uniformity, (b) pressure drop ratio in barrier channels, (c) variations indiameter in different part of the reactor, (d) deviations in temperature in barrier and reactionchannels.

5.4.4 Effect of flow on temperature deviation

The one dimensional energy balance was used to study the effect of flow on temperaturedeviation. A good way to plot the obtained results is shown in Figure 5.10. The ratio oftemperature deviation to flow non-uniformity σ(T )

σ(q) varies as a function of the liquid residencetime. This ratio is described in Equation 5.18 where τL is the liquid residence time and f2 andf3 are constants which depend on the liquid used, the material of construction of the BMMRand its geometrical dimensions.

σ(T )σ(q)

= f2 exp(− f3 τL) (5.18)

Above a certain liquid residence time, σ(T )σ(q) decreases approaching almost zero. Beyond

that critical liquid residence time, there is no significant effect of flow on temperature devia-tion as shown in Figure 5.10. Meaning that all channels will have similar outlet temperatureregardless of the flow distribution. This is because of the large heat exchange area of theBMMR. When the heat conductivity of the wall material changes from 16 W/m.K for steel to0.93 W/m.K for glass, the critical liquid residence time increases by two times. Reducing hmc


0 5 10 15 20 250.00

0.05

0.10

0.15

0.20 km = 16 W/m.K * km = 0.93 W/m.K 0.5 dC

0.5 hmc

T)/

q), (

o C/%

)

Liquid residence time, (s)

Figure 5.10: Effect of liquid residence time on ratio of temperature deviation to flow non-uniformity. Study is made based on the one dimensionaal energy balance. * reference casefor steel plate given in Table 5.2. km equals to 0.93 W/m.K refer to glass; 0.5 dC is whenchannel diameter reduced by half; 0.5hmc is when the heat transfer coefficient from waterbath to microchannels wall reduced by half.

by half increases the critical residence time by 1.5 times. Moreover, when the microchanneldiameter is reduced by half, the critical residence time decreases by more than 3 times.

0 5 10 15 20 250.0

0.5

1.0

1.5

2.0 km = 16 W/m.K * km = 0.93 W/m.K 0.5 dC

0.5 hmc

T), (

o C)

Liquid residence time, (s)

Figure 5.11: Effect of a 10% flow non-uniformity on temperature deviation at varied liquidresidence time for the case shown in 5.10.

For a specific target flow non-uniformity, 10% was used here as an example, Figure 5.10can be re-plotted as shown in Figure 5.11. For the entire tested flow distribution, temperaturedeviation σ(T ) remained less than 2 oC. The flow rate effect on temperature deviation was

Results 99

not that significant as can already be seen in section 5.4.2.3. However, one should keep inmind that this study was made with uniform heat supply and without any reaction involved.Testing the effect of these factors on the temperature and flow distributions is an essentialstudy to be explored in the future.

5.4.5 Experimental studies - effect of temperature and flow distribu-tions

The effect of liquid flow rate on temperature deviation is shown in Figure 5.12. The gas flowrate has no effect on the obtained result. The investigations were made at steady state situa-tion. Liquid temperatures were measured at four different locations as shown in Figure 5.12,inlet to the manifold, Tinlet , inlet to the gas-liquid microchannels, Tin, outlet of the gas-liquidmicrochannels, Tout , and water heating bath, Tbath. It is good to remind here that only thegas-liquid microchannels were in contact with the water heating bath while the other BMMRparts (inlets, manifolds, barrier channels, and collector) were outside the water heating bath.

0 30 60 90 12020

40

60

80

100 bath out in inlet

T, (o C

)

qL, (mL/min)

Figure 5.12: Steady state temperature deviation at different flow rates. Four liquid tempera-ture are measured: water bath Tbath, outlet from the gas-liquid microchannels Tout , inlet to thegas-liquid microchannels Tin, and inlet to the manifold Tinlet .

The temperature of liquid inlet to the microchannels Tin is much larger than Tinlet . Itreaches more than 60 oC for some flow rates. The increase of Tin even before it enters themicrochannels is due to the heat conduction. That is why Tin decreases as the flow rate in-creases. For the entire tested liquid flow rates, the temperature at the outlet of the gas-liquidmicrochannels Tout always reached the water bath temperature. The difference of Tout over


the eight parallel microchannels remained always less than 1 oC. This is due to the very largeheat transfer area of the BMMR. Thus, the liquid residence time in the BMMR was alwayslarger than the critical liquid residence time.

The effect of temperature on flow distribution is shown in Figure 5.13(a). Experimentsmade at 70 oC in Figure 5.12 were repeated at 20 oC. The liquid flow non-uniformity re-mained less than 10% for the entire range of tested flow rates. Changing gas to liquid flowratio at a constant liquid flow rate had no significant effect on flow distribution.

Figure 5.13: (a) Liquid flow non-uniformity at varied gas (from 10 to 230 mLn/min) andliquid (10-115 mL/min) flow rates at two temperatures at 20 oC and 70 oC . (b) and (c) arethe relative flow rate per channel for all flow rates at 20 oC and 70 oC, respectively. Theschematic drawing show a top view of the liquid manifold, the inlet feed location, and thedirection of increasing temperature in the manifold from Low to High.

When the temperature increased from 20 oC to 70 oC, flow non-uniformity increasedby a factor of two. The increase in flow non-uniformity at 70 oC is due to the temperaturedeviation in the barrier channels. The relative flow rate per channel is plotted as shownin Figure 5.13(b) and Figure 5.13(c). When the experiments were conducted at 70 oC, the

Results 101

profile of flow distribution changes. Flow rate was the smallest near to the manifold inletfor microchannel number 1 and largest at the end of manifold for microchannel number 8.Figure 5.13 shows a top view of the liquid manifold with the position of manifold inlet andoutlet to the barrier channels. The manifold was not in contact with the water heating bath,but gets heated due to the heat conduction via the metal connections. The temperature nearthe manifold inlet was the lowest (Low) while that in the last channel, microchannel 8, wasthe highest (High). When temperature increased, liquid viscosity decreased allowing largerflow rate in that specific channel to maintain similar pressure drop as those of the parallelmicrochannels. The profile of flow distribution took the shape of an inclined line; low nearthe manifold inlet and high near the last microchannel. This result demonstrates clearly theeffect of temperature deviation in the barrier channels on flow distribution.

5.4.6 Conclusions and outlook

In this chapter the effect of temperature on flow distribution was demonstrated for the barrier-based micro/milli channel reactor (BMMR). The 2-PRN model was used to compute the flowdistribution for a pre-defined temperature deviation in three parts of the BMMR: manifolds,barrier channels, and mixer and reaction channels. Temperature deviation in the barrier chan-nels has the strongest effect on flow distribution. For a similar value of temperature deviation,the barrier channels affect flow distribution 10 times more than the mixer and reaction chan-nels. Temperature deviation in the manifold has negligible effect on flow distribution.

The 2-PRN model could not be used to study the effect of flow rate on temperature de-viation. Instead, a one dimensional energy balance was developed. Above a certain criticalliquid residence time, flow rate had no significant effect on the temperature deviation. Thecritical liquid residence time depends on the liquid used, the BMMR material of constructionand its geometrical dimensions.

A design methodology was proposed to provide cut-off values of the maximum allowedtemperature deviation in the barrier channels and in the mixer and reaction channels. Bycombining this design methodology with the previously developed one on the effect of mi-crochannel diameters, a complete engineering methodology was achieved for maintainingflow non-uniformity below a defined target. An example of how to keep flow non-uniformitylower than 10% was demonstrated.

This study was performed with a large degree of simplification. Nonetheless, it provideda quick estimation of the critical design parameters needed for the numbering-up like the


diameter variations and temperature deviations. What is a real challenge next, is to test thisdesign methodology in a real case which involves a reaction in the gas-liquid microchannels.Thus, exploring the relation between flow non-uniformities, temperature deviations and theireffect on the reaction conversion and selectivity. The second challenge is to design a heatexchanger which ensures a uniform heat supply or removal from the reaction microchannels.

Bibliography

Al-Rawashdeh, M., Fluitsma, L. J. M., Nijhuis, T. A., Rebrov, E. V., Hessel, V., Schouten,J. C., 2012. Design criteria for a barrier-based gas-liquid flow distributor for parallel mi-crochannels. Chemical Engineering Journal 181-182, 549–556.

Amador, C., Gavriilidis, A., Angeli, P., 2004. Flow distribution in different microreactorscale-out geometries and the effect of manufacturing tolerances and channel blockage.Chemical Engineering Journal 101, 379–390.



Charpentier, J. C., 2007. In the frame of globalization and sustainability, process intensifi-cation, a path to the future of chemical and process engineering (molecules into money).Chemical Engineering Journal 134, 84–92.

Chung, T. H., Ajlan, M., Lee, L. L., Starling, K. E., 1988. Generalized multiparameter corre-lation for nonpolar and polar fluid transport properties. Industrial and Engineering Chem-istry Research 27, 671–679.



Goletz, E., Tassios, D., 1977. An antoine type equation for liquid viscosity dependency totemperature. Industrial & Engineering Chemistry Process Design and Development 16,75–79.

Haber, J., Kashid, M. N., Renken, A., Kiwi-Minsker, L., 2012. Heat management in singleand multi-injection microstructured reactors: Scaling effects, stability analysis, and role ofmixing. Industrial and Engineering Chemistry Research 51, 1474–1489.



Bibliography 103

Hornung, C. H., Mackley, M. R., Baxendale, I. R., Ley, S. V., 2007. A microcapillary flowdisc reactor for organic synthesis. Organic Process Research & Development 11, 399–405.





Leung, S. S., Gupta, R., Fletcher, D. F., Haynes, B. S., 2012. Effect of flow characteristics onTaylor flow heat transfer. Industrial and Engineering Chemistry Research 51, 2010–2020.


Mendorf, M., Nachtrodt, H., Mescher, A., Ghaini, A., Agar, D. W., 2010. Design and controltechniques for the numbering-up of capillary microreactors with uniform multiphase flowdistribution. Industrial and Engineering Chemistry Research 49, 10908–10916.

Natividad, R., Cruz-Olivares, J., Fishwick, R. P., Wood, J., Winterbottom, J. M., 2007.Scaling-out selective hydrogenation reactions: From single capillary reactor to monolith.Fuel 86, 1304 – 1312.



Saber, M., Commenge, J. M., Falk, L., 2009. Heat-transfer characteristics in multi-scale flownetworks with parallel channels. Chemical Engineering and Processing: Process Intensifi-cation 49, 732–739.





Phenylacetylene hydrogenationover [Rh(NBD)(PPh3)2]BF4

catalyst in numbered-upmicrochannels reactor 66

This chapter will be submitted as:Al-Rawashdeh, M., Zalucky, J., Muller, C., Nijhuis, T.A, Hessel,V. & Schouten, J.C. 2013. Phenylacetylene Hydrogenation over[Rh(NBD)(PPh3)2]BF4 Catalyst in Numbered-up Microchannels reactor.Industrial & Engineering Chemistry Research.

AbstractThis chapter provides a proof of concept for the capability of the barrier-based micro/millichannel reactor (BMMR) to number-up gas-liquid Taylor flow under reactive flow condi-tions. The hydrogenation of phenylacetylene to styrene and ethylbenzene using homogenouscationic rhodium catalysts [Rh(NBD)(PPh3)2]BF4 was used as a model reaction. First, aparametric study in a semi continuous batch reactor was made by changing the hydrogenpressure, the catalyst concentrations, and the initial concentrations of phenylacetylene andstyrene. A mechanism for this reaction system has been proposed by Esteruelas et al. (1998).This mechanism was extended in this chapter to develop a kinetic model which predicts theexperimental result within an accuracy of 20%. Catalyst deactivation was observed and in-corporated in the kinetic model. Second, the reaction was conducted in the BMMR. Thereactant and products concentrations of a single channel were compared to those of eightparallel channels combined. For 95% of the obtained results, the difference in concentrationsbetween the single and the eight channels was within ± 10% and depended on the gas andliquid flow rates. As a proof of concept, the numbering-up of gas-liquid Taylor flow in theBMMR under reactive flow conditions has been successfully realized.

106 Phenylacetylene hydrogenation in numbered-up microchannels reactor

6.1 Introduction

Microstructured reactors are already at the production scale for single phase processes (Re-brov et al. (2011)); for example, in fine-chemical applications in the liquid phase (Kock-mann et al. (2011), Kockmann and Roberge (2009)) and in fuel processing in the gas-phase(Deshmukh et al. (2010), Lerou et al. (2010)). However, gas-liquid micro processing re-mains mostly restricted to the laboratory scale. This is due to the complexity and expenditureneeded for an adequate numbering-up with an equal distribution of the gas and liquid flowsover multiple microchannels (De Mas et al. (2005), Kashid et al. (2010), Losey et al. (2002)).

Figure 6.1: Top: A schematic cross sectional view of the BMMR for one single channel.Symbol used are G gas, L liquid, M manifold, B barrier channel, C reaction channel. Bottom:the BMMR with an enlarged view of the glass chip which hosts the barrier channels and theT mixer, and a picture of Taylor flow in the eight parallel channels.

In chapter two, we developed a barrier-based gas-liquid flow distributor which was usedfor a water-nitrogen flow in the Taylor flow regime (Angeli and Gavriilidis (2008), Guptaet al. (2010)). Gas-liquid channeling was prevented and a flow uniformity of more than 90%was achieved. A general design methodology was developed in chapter three which led tothe ’Barrier-based Micro/Milli Reactor’ (BMMR) which was presented in chapter four andshown in Figure 6.1. The BMMR is a modular reactor for conducting a multiphase (gas-liquid) reaction in 8 parallel micro or milli channels at a pressure up to 20 bar and a temper-ature up to 200 oC with a total liquid flow rate up to 150 mL/min.

The BMMR showed a promising potential for numbering-up gas-liquid flow in the Taylor

Introduction 107

flow regime. In chapter four the flow non-uniformity in the BMMR was larger than 90% un-der varied gas and liquid flow rates, liquid viscosities and surface tensions. In chapter five, theeffect between temperature and flow distribution and visa versa was studied and incorporatedin a design methodology which ensures that flow uniformity remains within the acceptablerange. All of the mentioned studies were conducted under cold flow conditions - without anyreaction. In this chapter, the effect of reactive flow conditions on the numbering-up conceptin the BMMR will be evaluated.

The selective reduction of triple bonds and double bonds is an important transformation insynthetic organic chemistry (Kluwer and Elsevier (2007)). Transition-metal compounds playa key role in catalytic conversion in particular for the homogenously catalyzed hydrogenationreactions. The cationic rhodium complex with the general formula [Rh(diene)La]+ (a = 2 or3) is an active catalyst for the hydrogenation of olefins, diens, alkynes, ketones, polynuclearheteroaromatic compounds and carbon dioxide (Schrock and Osborn (1976a,b)).

The cationic rhodium catalyst [Rh(NBD)(PPh3)2]BF4 was studied by Esteruelas et al.(1998) for the selective hydrogenation of phenylacetylene to styrene. The catalyst showed anexcellent performance using dichloromethane as a solvent. A selectivity of 80% was achievedat mild conditions of 25 oC and atmospheric pressure. In their work, the reaction mechanismwas experimentally verified and the reaction was found to be first order in the catalyst, sub-strate, and second order in the hydrogen pressure.

Phenylacetylene hydrogenation is often used as a model system for evaluating reactordesign and for evaluating alkyne hydrogenation catalysts (Wilhite et al. (2002)). Attempts tocorrelate experimental data with reactor design are limited due to the absence of accurate ki-netic models. For newly developed reactors, model systems are required to fully characterizeand exploit their potential. Microreactors are an attractive reactors for conducting hydrogena-tion reactions (Abdallah et al. (2004), Keybl and Jensen (2011), Nijhuis et al. (2003)). Themicro meter scale reaction channel accelerates the fluid mixing and heating, and increasesthe specific interfacial area by more than one order of magnitude compared to a typical slurrybubble column. This acceleration allows the reaction to occur in a very short time scale andallows precise temperature control resulting in clear cut in the conversion and selectivity.

In this chapter, the hydrogenation of phenylacetylene to styrene and ethylbenzene cat-alyzed by the cationic rhodium catalyst [Rh((NBD)(PPh3)2]BF4 is used as a model reaction tocharacterize the barrier-based micro/milli channel reactor - the BMMR. The aim is to providea proof of concept of the capability of the BMMR to number-up gas-liquid Taylor flow under


reactive flow conditions. First, a parametric study is carried out to identify possible operatingwindows for the BMMR and to provide data for the kinetic model. The parametric study isconducted by changing hydrogen pressure, catalyst concentrations, initial concentrations ofphenylacetylene, and styrene in a semi continuous batch reactor using dichloromethane as asolvent. These parametric experiments are fitted by extending the kinetic model of Esteruelaset al. (1998) to take into account the production of styrene and ethylbenzene. Second, the re-action is conducted in the BMMR under different gas and liquid flow rates while keeping theother operating conditions fixed. The numbering-up concept under reactive flow conditionsis analyzed by comparing the reactant and products concentrations from a single reactionchannel to that from eight reaction channels combined.

6.2 Experimental section

6.2.1 Catalyst preparation

A mixture of Bis(norbornadiene)rhodium (I) tetrafluoroborate (0.08 g, 0.21 mmol) and triph-enylphosphine (0.11 g, 0.42 mmol) was stirred in freshly distilled tetrahydrofuran (2 ml)for 10 minutes (All glassware was dried prior to use). The volume of the solution was re-duced to 1 ml, and it was then poured into diethyl ether (Na-dried and distilled, 25 ml).The sample was isolated by filtration (Craig tube) and washed with ether, giving 0.151 g of[Rh((NBD)(PPh3)2]BF4 as yellow solid.

6.2.2 Experimental procedure - batch setup

Table 6.1: Operating conditions applied in the parametric study experiments. Isothermaltemperature maintained at 70 oC. * Styrene initial concentration is 150 mol/m3

Experiment mcat , (g/L) PH2, (bar) CPA, mol/m3

1 0.48 10 1502 0.90 10 1503 1.24 10 1504 0.21 10 1505 0.12 5 1506 0.12 15 1507 0.51 10 758 0.48 10 2009 0.48 10 0*

Experimental section 109

The experiments in the batch setup were carried out in a stainless steel 75 mL batch re-actor at a constant pressure and temperature as given in Table 6.1. The reactor was equippedwith five side inlets. One was for the thermocouple to monitor the temperature; another wasfor the gas supply; one was for the sampling using a stainless steel capillary of 0.5 mm inter-nal diameter connected to three way valve; and the last side inlet was connected to a funnelthat will contain the concentrated substrate and solvent solution. Agitation was provided viaa magnetic stirring bar. The pressure was maintained by keeping the gas side open to a hy-drogen cylinder while the pressure was fixed using a pressure regulator valve.

The experimental procedure was carried out as follows. The catalyst [Rh(NBD)(PPh3)2]-BF4 was weighed and dissolved in dichloromethane. The solution was pressurized with hy-drogen and then heated up to the required temperature while stirring at 1100 rpm. Aftera stabilization time for 30 minutes, a solution of phenylacetylene in dichloromethane wasadded and this was taken as the starting time zero of the reaction. For each experimentalrun, samples were taken at different reaction times. The samples were analyzed using gaschromatography (Varian 3800) with a FID detection system. The liquid phase concentrationswere measured from calibration curves generated for each species using the internal standard(n-dodecane). Phenylacetylene, styrene, ethylbenzene, and n-dodecane were obtained fromSigma-Aldrich.

6.2.3 Experimental procedure - BMMR setup

A process flow diagram of the BMMR experimental setup is shown in Figure 6.2. Key di-mensions of the BMMR are given in Table 6.2. Reaction channels with square cross-sectionare fabricated in a stainless steel plate. Further details about the design and fabrication of theBMMR can be found in chapter four.

Liquid was pumped (v) using a gear pump (NHK Mikrosysteme GmbH, MZR-7205) viaa liquid mass flow controller (vi) (Bronkhorst). Nitrogen(iii), hydrogen(i) and argon (iv) werefed from gas bottles using mass flow controllers (Bronkhorst). The pressure was measuredat the manifold using a pressure sensor (range 0-25 bar, Endress + Hauser,PMP131) andtemperature was measured using thermocouple type Pt100. The temperature in the reactorwas regulated by placing the entire reactor in a water heating bath (Lauda). The gas-liquidseparation was achieved in a 0.5L stainless steel vessel (xiv) via gravity. The liquid levelwas monitored by a differential pressure sensor (PS4) and maintained by a PID control valve(Bronkhorst)(xvii) placed downstream of the gas-liquid separator. At the gas outlet, a backpressure regulator (Bronkhorst) (xvi) was used to regulate the operating pressure of the entire


Figure 6.2: Process flow diagram of the BMMR setup and the locations of the temperatureand pressure sensors. Symbol used: (i) liquid tank, (ii) hydrogen bottle, (iii) nitrogen bottle,(iv) argon bottle, (v) gear pump, (vi) mass flow controller, (vii) Barrier-mixer glass chip,(viii) manifold, (ix) reaction channel plate, (x) inspection window, (xi) connection block,(xii) collector block and sampling points, (xiv) gas liquid separator, (xv) mist collector, (xvi)back pressure regulator, (xvii) control valve, (TS) temperature sensor, (PS) pressure sensor.Box around reactor represents Lauda water bath. Box around the collector blok represents anice bath.

Table 6.2: Dimensions for the barrier-based micro/milli reactor (BMMR). Symbols used arechannel width W, depth H, and length L. ∗ The width is decreasing by an 8 degree angle.

W, (mm) H, (mm) L , (mm)

Inlet gas manifold 41∗ 5 155Inlet liquid manifold 41∗ 5 155Gas barrier channel 0.4 0.1 340Liquid barrier channel 1.0 0.1 37Inlet gas T-mixer 1.3 1.3 13Inlet liquid Mixer 1.3 1.3 10Reaction channel 1.23 1.23 2000

setup.

The experiments in the BMMR were conducted according to the following procedure.First, the substrate (phenylacetylene) and catalyst ([Rh(NBD)(PPh3)2]BF4) were dissolvedin dichloromethane with the required quantities under inert conditions in a glove box. The

Results 111

solution was placed in a liquid storage tank (i) and transported to the experimental setup. Theliquid storage was then connected to an argon line. Next, the tubing and the liquid manifoldwere filled while the hydrogen was being fed at very low flow rate. A flow of nitrogen tothe gas-liquid separator was started to bring the system to the required operating pressure.During all of these processes, the water heating bath was switched on to reach the requiredtemperature. The collector block and sample points were placed in an ice box to quench thereaction after the gas-liquid flows leave the reactor. Once the required temperature and pres-sure in the reactor were reached, the BMMR experimental setup became ready for conductinga reaction experiment.

The experiments were performed at a constant temperature of 70 oC and pressure of 10bar. The liquid flow rate was changed in the range of 8-125 mL/min and the gas flow rate from60-4750 mLn/min. For each specific experiment, the gas and the liquid flows were run for fewminutes until a steady state situation was reached. Then, samples were taken and analyzedusing GC as stated earlier. Samples were taken at two locations. The first location was beforethe collector block for one of the outlet streams from the reactor. The second location wasafter the collector block for the eight reaction channels combined. At each sampling location,two samples were taken at different times to check the reproducibility. Because it was notpossible to measure slug and bubble lengths, the residence time of Taylor flow was calculatedbased on the combined gas and liquid superficial velocities and not on the bubble velocity.

6.3 Results

6.3.1 Batch setup experiments

A typical result for one of the batch experiments is shown in Figure 6.3. Phenylacetyleneconverts to styrene with a maximum selectivity up to 80%. The production of Ethylbenzenelinearly increases till all styrene is consumed. The results obtained from the batch experi-ments are used to estimate the kinetic parameters via fitting.

6.3.2 Initial rate studies

In Figure 6.4, initial rate studies are presented as a function of catalyst concentrations, hydro-gen pressure and initial concentrations of phenylacetylene. In all studies, a similar initial rateof reaction for styrene and phenylacetylene is obtained. For ethylbenzene, the initial rate is10 times smaller which demonstrates the excellent selectivity of [Rh(NBD) (PPh3)2]BF4 tostyrene.


0 5 10 15 20 25 300

25

50

75

100

125

150

175

C, (

mol

/m3 )

Time, (min)

Figure 6.3: Typical result obtained for the hydrogenation reaction of phenylacetylene using[Rh((NBD)(PPh3)]BF4 in dichloromethane. Operating conditions are for experiment 3 inTable 6.1. Symbols used are: phenylacetylene (circle), styrene (triangle), and ethylbenzene(square).

In Figure 6.4(a), the initial reaction rate shows a first-order dependence to the catalystconcentration which matches with those from literature (Esteruelas et al. (1998), Wilhiteet al. (2002)). At catalyst concentration below 0.38 g/L, the initial rate of reaction is almostzero. It is possible that part of the catalyst is poisoned when it is mixed with the solventwhich makes it non-effective. Most likely, the solvent or reactant contained a small amountof poisoning compound which deactivate a fixed amount of catalyst 0.38 g/L. To accountfor this, in subsequent modeling we will use an effective catalyst concentration which is theactual catalyst concentration minus the inactive catalyst amount of 0.38 g/L.

The second parameter which was investigated is the hydrogen pressure as shown in Fig-ure 6.4(b). The hydrogen pressure varies between 5 bar to 15 bar. The initial reaction rateincreases as the hydrogen pressure increases. The initial reaction rate for the 15 bar experi-ment is equal to that at 10 bar. Increasing the hydrogen pressure to 15 bar does not increasethe reaction rate. In Figure 6.4(c), initial reaction rates of phenylacetylene concentrations areshown. Increasing the initial concentration of phenylacetylene slightly increases the initialreaction rate.

Results 113

Figure 6.4: Initial rate studies for catalyst concentration (a), hydrogen pressure (b), and initialconcentration of phenylacetylene (c). Operating conditions are given in Table 6.1.

6.3.3 Kinetic models

To obtain a kinetic model which takes into account the styrene and ethylbenzene production,the reaction mechanism of Esteruelas et al. (1998) is extended as proposed in Equations 6.1to 6.6.

[Rh(NBD)(PPh3)2]BF41

+H2K6 [RhH2(NBD)(PPh3)2]BF4

2(6.1)

[RhH2(NBD)(PPh3)]BF42

K7 [RhH2(NBD)(PPh3)]BF43

+PPh3 (6.2)


+PhC≡ CHK8 [RhH(CH = CHPh)(NBD)(PPh3)2]BF4

4(6.3)

[RhH(CH = CHPh)(NBD)(PPh3)]BF44

+H2k9→ [RhH2(NBD)(PPh3)]BF4

3+CH2 = CHPh

(6.4)


+CH2 = CHPhK10 [RhH(CH2−CH2Ph)(NBD)(PPh3)]BF4

5(6.5)

[RhH(CH2−CH2Ph)(NBD)(PPh3)]BF45

+H2k11→ [RhH2(NBD)(PPh3)]BF4

3+CH3−CH2Ph

(6.6)


Based on the described reaction mechanism, the equilibrium constants are defined as:

K6 =[2]

[H2][1](6.7)

K7 =[PPh3][3]

[2](6.8)

K8 =[4]

[PhC≡ CH][3](6.9)

K10 =[5]

[CH2 = CHPh][3](6.10)

By re-substituting all complex intermediates as a function of 4, the concentration of complexintermediates 1, 2, 3, and 5 are written as

[1] =[4][PPh3]

K6K7K8[PhC≡ CH][H2](6.11)

[2] =[4][PPh3]

K7K8[PhC≡ CH](6.12)

[3] =[4]

K8[PhC≡ CH](6.13)

[5] =K10[4][CH2 = CHPh]

K8[PhC≡ CH](6.14)

The total amount of catalyst in the system is known and is equal to the sum of all the catalystcomplexes.

[Rh]tot = [1]+ [2]+ [3]+ [4]+ [5] (6.15)

The concentration of [4] is then equal to Equation 6.16.

[4] =K6K7K8[Rh]tot[H2][PhC≡ CH]

(1+K6[H2]+K6K7[H2]PPh3

+K6K7K8[H2][PhC≡CH]

PPh3+K6K7K10

[H2][CH2=CHPh]PPh3

)(6.16)

Following the same analogy as that of Esteruelas et al. (1998), the rate determining steps forthe formation of styrene and ethylbenzene are

r9 = k9[H2][4] (6.17)

Results 115

r11 = k11[H2][5] (6.18)

By substituting complex intermediate [4] and [5], the rate expression for the rate determiningsteps are

r9 =K6K7K8k9[Rh]tot[H2]2[PhC≡ CH]

(1+K6[H2]+K6K7[H2]PPh3


PPh3+K6K7K10

[H2][CH2=CHPh]PPh3

)(6.19)

r11 =K6K7K10k11[Rh]tot[H2]2[CH2 = CHPh]

(1+K6[H2]+K6K7[H2]PPh3


PPh3+K6K7K10

[H2][CH2=CHPh]PPh3

)(6.20)

Via mass balance, the concentrations of phenylacetylene, styrene and ethyl benzene are com-puted in Equations 6.21, 6.21 and 6.23, respectively.

d[PhC≡ CH]dt

=−r9 (6.21)

d[CH2 = CHPh]dt

= r9− r11 (6.22)

d[CH3−CH2Ph]dt

= r11 (6.23)

6.3.4 Catalyst deactivation

Catalyst deactivation was observed during reaction. In all of the conducted experiments, re-action rate reduced significantly after some reaction time. Figure 6.5 showed the result forone of these experiment conducted for three hours. The significant reduction in reaction ratecan be clearly noticed after 30 minutes. The same significant reduction was noticed in all ofthe conducted experiments after 30 minutes. For experiments which were completed in lessthan 30 minutes as the one shown in 6.3, reduction of the reaction rate was not observed.

Another observation which suggests that catalyst deactivation occurred was the repro-ducibility of the experimental results. For two experiments conducted at the same operatingconditions with the same initial concentrations, a bad reproducibility was obtained if the re-action started at different times. Time zero for the reaction is defined when the substratesolution was injected into the catalyst solution. For example if the heating step was madelonger by 20 minutes (so reaction time starts after 50 minutes), inconsistent results were ob-


tained with an error larger than 30%. During the 30 minutes heating step, the catalyst wasin contact with hydrogen. Therefore, the catalyst was activated and because of that it wasprone for deactivation. For example, the norbordiene (NBD) ligand could be hydrogenated(Esteruelas et al. (2000)). To maintain a good reproducibility, the reaction started at the sametime for all of the conducted experiments.

For estimating the kinetic parameters, a model to predict the catalyst deactivation wastaken into account. Among different tested deactivation models, the model shown in Equa-tion 6.24 predicted best the catalyst deactivation. Because it was not so clear which of themetal-complexes deactivates, the deactivation was implemented for the total catalyst concen-tration. In Figure 6.6, catalyst deactivation using Equation 6.24 is demonstrated for the sameexperimental result of Figure 6.5. After 30 minutes, more than 90% of the initial catalystweight was deactivated.

0 30 60 90 120 150 1800

30

60

90

120

150

C, (

mol

/m3 )

Time, (min)

Figure 6.5: Experimental result for experiment 2 in Table 6.1. Reduction of the reaction rateafter 30 minutes is due to catalyst deactivation. Symbols used are:phenylacetylene (circle),styrene (triangle), and ethylbenzene (square).

d[Rh]tot

dt=−kd[H2][Rh]tot (6.24)

6.3.5 Determination of kinetic parameters

By providing values to the kinetic parameters K’s, Equations 6.21 to 6.24 can be numericallyintegrated using the Ode15 Matlab routine to calculate the reaction concentrations over time.

Results 117

0 10 20 30 40 50 600.0

0.2

0.4

0.6

0.8

1.0

mca

t/mca

t, t=

0, (-)

Time, (min)

Figure 6.6: Deactivation of catalyst over time using Equation 6.24 for the same experimentalconditions used in Figure 6.5.

The computed concentrations can then be compared with the experimental concentrations tocalculate the error function given in Equation 6.25, where L is the number of experimentsinvolving M species, with concentrations measured at N reaction times. By minimizing theerror function using the least square method (using Fmin in Matlab), kinetic parameters K’scan be estimated.

error =

√√√√√L∑k

M∑j

N∑i

Cexpj,i −Ccalc

j,i

Cexpj,i

LM(N−1)(6.25)

6.3.6 Parametric study- Kinetic results

All of the experimental results from the parametric study are shown in Figure 6.7. The pointsare the experimental results while the solid lines are the model fittings. In all experiments,the maximum styrene selectivity remains between 60% and 85%.

The best fitted values for the kinetic parameters are given in Table 6.3. The extendedkinetic model of Esteruelas et al. (1998) predicts well the experiments within an accuracy of± 20%. The inaccuracy only increases when using high hydrogen pressure, for the 15 barexperiment. Overall, the model shows to be capable of predicting the reaction behavior.

The kinetic model was tested for the hydrogenation of styrene to ethylbenzene as shown


Figure 6.7: Parametric study result for operating conditions given in Table 6.1. Dots are theexperimental result and solid lines are the model. Experiments (i,ii ii) refer to the pressureeffect from 5, 10, to 15 bar, respectively. Experiments (iv,v,vi) refer to the effect of catalystconcentration from 0.48, 0.9 to 1.24 g/L, respectively. Experiments (vi,vii,ix) refer to initialconcentration of phenylacetylene from 75, 150, to 200 mol/m3, respectively.

in Figure 6.8. The kinetic predicted accurately the hydrogenation of styrene to ethylbenzene.This also confirmed the capability of the model to describe the reaction behavior at variedrange of operating conditions.

6.3.7 BMMR setup experiments

Experimental results obtained for the hydrogenation reaction in the BMMR setup are shownin Figure 6.9. Reactant and products concentrations were measured at varied gas and liquidflow rates for the two sample locations shown in Figure 6.2(xii). One sampling location was

Results 119

Table 6.3: Best fitting for kinetic parameters.

Parameter Fitting Unit

K6 6.89x10−5 m3/molK7 0.50 mol/m3

K8 0.06 m3/molk9 5.90x10−3 m3.m3/mol.g.minK10 0.06 m3/molk11 9.19x10−4 m3.m3/mol.g.minkd 1.10x10−3 m3/mol.min

0 5 10 15 20 250

30

60

90

120

150

C, (

mol

/m3 )

Time, (min)

Figure 6.8: Result for hydrogenation of styrene to ethylbenzene. Operating conditions are forexperiment 9 in Table 6.1. Symbols used are: styrene (triangle), ethylbenzene (square), andsolid lines are for the model.

taken before the collector block at the outlet of one reaction channel; and the other sam-pling location was taken after the collector block for the eight reaction channels combined.Concentrations from both samples were compared with each other to provide indication intothe effect of numbering-up in the BMMR. Three sets of experiments were performed at thesame operating conditions of experiment 7 in Table 6.1. The difference between them wasthe amount of prepared reactant solution as 0.5 L, 1 L, and 1.5 L for (a), (b), and (c), re-spectively. Larger amount of reactant solution allowed taking more sampling points at widerrange of gas and liquid flow rates. It is important to mention that the aim from the study ofFigure 6.9 is to compare the single channel result to that from the 8 collected channels to give


indication about the effect of numbering-up. Thus, variation in the results between (a), (b)and (c) will not be addressed here. However, possible reasons for the differences are catalystdeactivation, effect of slug and bubble lengths, and or bubble coalescence.

Figure 6.9: Experimental result obtained for the hydrogenation reaction in the BMMR setup.Symbols used are: phenylacetylene (circle), styrene (triangle), and ethylbenzene (square);filled symbols are the analysis of a sample collected from a single channel and open symbolsare for sample collected from the eight channels together. Operating conditions are for ex-periment 7 in Table 6.1. (a), (b), and (c) refers to volume of the reactant solution used in theexperiment as 0.5 L, 1 L, and 1.5 L, respectively.

A few observations can be made based on the result in Figure 6.9. First, the reaction per-formance in the BMMR is ten times faster compared to that of the batch experiment (6.7(vii))made at the same conditions. In the BMMR, the hydrogen contacts with the catalyst and sub-strate only in the reaction channel. Thus, activation of the catalyst happened in-situ and thereis less possibility for catalyst deactivation. Whereas in batch setup, the hydrogen comes intocontact with the catalyst before the reaction started, during the heating step, which increasesthe chance of catalyst deactivation as already shown in Figure 6.5. Second, the reactant andproducts trends of the 8 collected reaction channels are the same as for the single channelreaction. Third, some of the single channel concentrations are larger than the 8 channels andat other condition lower. The results from the single channel in comparison to the eight chan-nels vary randomly. Fourth, the difference in performance between the single channel andthe 8 channels seems to depend on the gas and liquid flow rates.

To demonstrate the effect of gas and liquid flow rates, the reaction performance at fixedliquid flow rate and varied gas flow rates is shown in Figure 6.10. Parts (a), (b) and (c) referto three experiments done at different total liquid flow rate as 7.5 mL/min, 15 mL/min, and30 mL/min, respectively. As gas flow rate increases, the gas bubble length increases, sluglength decreases, and residence time decreases. Such an influence on the reaction perfor-

Results 121

Figure 6.10: Experimental result obtained for the hydrogenation reaction in the BMMR setup.Symbols used are the same as those of Figure 6.9. Experiments were done at fixed liquid flowrate with varied hydrogen flow rate as mentioned on top of each point in the figures. Liquidflow rate of (a) is 7.5 mL/min, for (b) 15 mL/min, and for (c) is 30 ml/min

mance is correctly noticed in Figure 6.10(a) and (b). However, in Figure 6.10(c) the reactionperformance seems random even for the single channel. The slug and bubble lengths are themain parameter changed here which seems crucial on the reaction performance. Moreover,the absolute slug and bubble lengths can affect flow distribution as shown in chapter fourwhich could influence the reaction performance. Because in this experimental setup, it wasnot possible to measure slug and bubble lengths and flow non-uniformity, no possibility wasthere to link flow distribution to reaction performance. However, it can be clearly stated thatthe reaction performance of the single and 8 reaction channels are affected by the flow non-uniformity and as well by the slug and bubble lengths. Therefore, selective hydrogenation tostyrene using [Rh(NBD)(PPh3)2]BF4 is a suitable reaction model to characterize the BMMRand the concept of numbering-up.

6.3.8 Reactor modeling - BMMR

The kinetic model determined using the batch reactor experiments was used to model thereaction performance in the BMMR. The model is provided in the appendix. The modelingresults are provided in Figure 6.11. The calculations were performed for the virtual perfor-mance of the BMMR at a residence time similar to the batch reactor experiments in (a), at aresidence time more typical for the BMMR in (b), in the absence of catalyst deactivation andas well at different gas-liquid ratios (five and one) as in (c) and (d), respectively.

When comparing the results of the reactor model in Figure 6.11 to the experimental resultsin Figure 6.9, it can be seen that the reactor model underestimates the reaction rate. In otherwords, the reaction rate proceeds much faster in the BMMR than what is predicted by thereactor model. The reaction performance in the BMMR is ten times faster compared to that


Figure 6.11: Results of the BMMR reactor model at operating conditions similar to Figure6.9 for a residence time similar to the batch reactor experiments (a), at a residence time moretypical for the BMMR experiments (b), without catalyst deactivation (c), and when changingthe gas-liquid flow ratio to one (d) whereas all of the previous three results were performedwith gas-liquid flow ratio of five.

of the batch experiment (6.7(vii)) made at the same conditions. As anticipated earlier inthe text, catalyst deactivation might not occur in the BMMR which could explain the resultobtained by the reactor model. However, the modeling results without catalyst deactivationshown in Figure 6.11(c) show that the reaction rate increases in the reactor model but stillslower than the experimental results. When changing the gas flow rate from five to one, therate of reaction decreases which shows the dependency of the reactor model on the gas andliquid flow rates matching the experimental finding in Figure 6.10. To explain the highercatalyst activity in the BMMR compared to the batch reactor, kinetic experiments are neededin a single channel reactor under the Taylor flow regime.

6.3.9 Proof of concept of numbering-up in the BMMR

The effect of numbering-up in the BMMR can be seen by plotting all of the experimentalresults obtained from the BMMR as shown in Figure 6.12. The reaction concentrations fromthe single reaction channel are compared to concentrations from the eight reaction channelscombined. For more than 95% of the results, the difference in concentrations between the

Conclusions 123

0 20 40 60 800

20

40

60

80

-10%

Col

lect

ed c

hann

els -

C, (

mol

/m3 )

Single channel - C, (mol/m3)

+10%

Figure 6.12: Comparison between the experimental result of a single reaction channel com-pared to the eight reaction channels combined in the BMMR at varied flow rates. Symbolsused are: phenylacetylene (circle), styrene (triangle), and ethylbenzene (square).

single and the eight collected channels remains within ± 10%. Predicting what is the corre-sponding flow non-uniformity which lead to this percentage is not straightforward. This isbecause the reaction depends in addition to the residence time on the interfacial area whichcould not be estimated because no information could be obtained about the slug and bubblelengths in this setup. Additionally, concluding if the ± 10% difference margin is acceptablewill depend on the final application. This margin was obtained before any optimization of thereactor. It is expected that this marginal difference can be reduced by assuring that no slugand bubble coalescence occurs and the slug and bubble lengths remain larger than the channeldiameter. For example, this can be done by avoiding the transport channels in the BMMRsetup (see Figure 6.2) by placing the mixer in the reaction channel. Overall, the BMMR suc-cessfully demonstrated its potential for numbering-up gas-liquid Taylor flow under reactiveconditions.

6.4 Conclusions

This chapter aimed at providing a proof of concept over the capability of the barrier-based mi-cro/milli channel reactor (BMMR) to number-up gas-liquid Taylor flow under reactive flowconditions. The hydrogenation of phenylacetylene over a homogenous catalyst [Rh(NBD)(PPh3)2]BF4 was used as a model reaction.


First, kinetic parameters were estimated for the reaction by extending the work of Esteru-elas et al. (1998). The kinetic model was capable of predicting the reaction behavior for arange of catalyst concentrations (0.48 g/L to 1.24 g/L), hydrogen pressures (5 bar to 15 bar),and phenylacetylene concentrations (75 mol/m3 to 200 mol/m3). The kinetic model predictedthe experiments within an accuracy of 20%. Part of the catalyst showed to be non-effectivewhen it was mixed with the solvent. The catalyst deactivated over the reaction time. A modelto predict catalyst deactivation was incorporated into the kinetic model. Even with catalystdeactivation, the developed kinetic model was of use for evaluating the BMMR performance.

Second, the model reaction was used for characterizing the numbering-up concept forgas-liquid Taylor flow in the BMMR. The reaction performance in the BMMR was ten timesfaster compared to that of the batch experiment made at the same conditions. Most likely thisis because activation of the catalyst in the BMMR happened in-situ and there is less possibil-ity for catalyst deactivation. Whereas in batch setup, the hydrogen comes into contact withthe catalyst before the reaction started, during the heating step, which increases the chance ofcatalyst deactivation. The reaction conversion and selectivity were sensitive to the flow non-uniformity and slug and bubble lengths. For more than 95% of the results, the difference inconcentrations between the single and the eight combined reaction channels remained within± 10%. This result was achieved before any re-optimization of the reactor design and oper-ating conditions. As a proof of concept, the numbering-up of gas-liquid Taylor flow in theBMMR under reactive flow conditions was successfully realized.

6.5 Appendix

Equation in the gas phase:Pure hydrogen was used in the the gas phase. The hydrogen consumed by the reaction whichchanges the gas flow rate as

[H2]GA

dqG

dx=−kGLaGL(

RThH2

[H2]G− [H2]L) (6.26)

Equations in the liquid phase:

Equation for hydrogen:

qL

Ad[H2]L

dx= kGLaGL(

RThH2

[H2]G− [H2]L)− r[H2]L εL (6.27)

Appendix 125

Equation for phenylacetylene:

qL

Ad[PhC≡ CH]

dx= r[PhC≡CH] εL (6.28)

Equation for styrene:

qL

Ad[CH2 = CHPh]

dx= r[CH2=CHPh] εL (6.29)

Equation for ethylbenzene:

qL

Ad[CH3−CH2Ph]

dx= r[CH3−CH2Ph] εL (6.30)

The rate equations are:

r[H2]L = r9 + r11 (6.31)

r[PhC≡CH] =−r9 (6.32)

r[CH2=CHPh] = r9− r11 (6.33)

r[CH3−CH2Ph] = r11 (6.34)

Note that r9 and r11 are in minutes which needs to be converted to seconds in Equations6.31 to 6.34. Equations r9 and r11 are given in Equations 6.19 and 6.20, respectively..Mass transfer parameters:

Mass transfer parameters are calculated from Bercic and Pintar (1997) as:

kGLaGL = 0.11(UG +UL)1.19

[(1− εG)(LS +LB)]0.57 (6.35)

UG and UL are the gas and liquid superficial velocities. The slug LS and bubble LB

lengths are calculated according to van Steijn et al. (2010). The gas hold-up is calculatedfrom Warnier et al. (2008) as

εG = AbUG

UG +UL(6.36)

where Ab is the gas bubble cross section area which is calculated from the film thickness as


given by Warnier et al. (2008). The liquid hold-up is calculated as

εL = 1− εL (6.37)

The ODE’s of the reactor model were solved using matlab ode15s function.

Bibliography

Abdallah, R., Meille, V., Shaw, J., Wenn, D., de Bellefon, C., 2004. Gas-liquid and gas-liquid-solid catalysis in a mesh microreactor. Chemical Communications, 372–373.


Bercic, G., Pintar, A., 1997. The role of gas bubbles and liquid slug lengths on mass transportin the taylor flow through capillaries. Chemical Engineering Science 52, 3709–3719.


Deshmukh, S. R., Tonkovich, A. L. Y., Jarosch, K. T., Schrader, L., Fitzgerald, S. P., Ki-lanowski, D. R., Lerou, J. J., Mazanec, T. J., 2010. Scale-up of microchannel reactorsfor fischer-tropsch synthesis. Industrial and Engineering Chemistry Research 49, 10883–10888.


Esteruelas, M. A., Herrero, J., Martin, M., Oro, L. A., Real, V. M., 2000. Mecha-nism of the hydrogenation of 2,5-norbornadiene catalyzed by [Rh(NBD)(PPh3)2]BF4 indichloromethane: a kinetic and spectroscopic investigation. Journal of OrganometallicChemistry 599, 178 – 184.

Gupta, R., Fletcher, D. F., Haynes, B. S., 2010. Taylor flow in microchannels: A review ofexperimental and computational work. The Journal of Computational Multiphase Flows 2,1–32.


Keybl, J., Jensen, K. F., 2011. Microreactor system for high-pressure continuous flow ho-mogeneous catalysis measurements. Industrial and Engineering Chemistry Research 50,11013–11022.

Kluwer, A. M., Elsevier, C. J., 2007. The handbook of homogeneous hydrogenation. Wiely-VCH Verlag GmbH.


Bibliography 127




Nijhuis, T. A., Dautzenberg, F. M., Moulijn, J. A., 2003. Modeling of monolithic and trickle-bed reactors for the hydrogenation of styrene. Chemical Engineering Science 58, 1113–1124.


Schrock, R. R., Osborn, J. A., 1976a. Catalytic hydrogenation using cationic rhodium com-plexes. i. evolution of the catalytic system and the hydrogenation of olefins. Journal of theAmerican Chemical Society 98, 2134–2143.

Schrock, R. R., Osborn, J. A., 1976b. Catalytic hydrogenation using cationic rhodium com-plexes. ii. the selective hydrogenation of alkynes to cis olefins. Journal of the AmericanChemical Society 98, 2143–2147.


Warnier, M. J. F., Rebrov, E. V., de Croon, M. H. J. M., Hessel, V., Schouten, J. C., 2008. Gashold-up and liquid film thickness in Taylor flow in rectangular microchannels. ChemicalEngineering Journal 135, S153 – S158.

Wilhite, B. A., McCready, M. J., Varma, A., 2002. Kinetics of phenylacetylene hydrogenationover Pt/Al2O3 catalyst. Industrial and Engineering Chemistry Research 41, 3345–3350.

Oxidation of ethylbenzene - Casestudy to reach bulk scaleproduction using numbering-up 77

AbstractThe barrier-based micro/milli channels reactor (BMMR) is a flow reactor based on a genericdesign methodology to reach larger scale throughput via numbering-up. In this chapter theBMMR capability is evaluated to fulfill the outlined objective of the DeMiR project. Theoxidation of ethylbenzene as an industrial application is used for this evaluation. The studycompares the horizontal bubble column reactor (HBCR) to the BMMR which operates in theTaylor flow regime. The comparative study will be based on a production capacity of 1 m3/hof ethylbenzene hydroperoxide which corresponds to 8000 tons/yr. Including the diluentwhich is ethylbenzene and a target conversion of 10 wt%, the total flow capacity is 10 m3/hwhich equals to 80000 tons/yr. The study compares two aspects: the size of the reactor andreactor operation.

130 Case study to reach bulk scale production using numbering-up

7.1 Introduction

The objective of the DeMiR project is to develop a generic methodology for selecting anddesigning a flow reactor with a reaction channel, either at the ”micro-scale” or ”milli-scale”dimensions. This flow reactor should be developed for the case of L/L and G/L catalytic reac-tions and multiphase food processing systems suitable for bulk scale production, in the rangeof m3/h and m3/min for liquids and gases, respectively. In this PhD research, the barrier-based micro/milli channels reactor (BMMR) was developed and showed promising resultsfor numbering-up multiphase flow. The BMMR provided a generic design methodology toreach larger scale throughput. In this chapter the BMMR capability to fulfill the outlined ob-jective of the DeMiR project will be evaluated. The oxidation of ethylbenzene as an industrialapplication is used for this evaluation.

Figure 7.1: Schematic route for scale-up via numbering-up for micro/milli channel reactors.(i) scale-up of a single channel, (ii) modular unit - internal numbering-up, (iii) Multi-modularunits - external numbering-up.

The scaling route for the BMMR to reach bulk scale production is shown in Figure 7.1.The scaling route consists of three steps. The first step is the scale-up of a single chan-nel (increasing cross section dimensions) while ”smartly” keeping the properties of massand heat transfer similar to the micro/milli channel reactor (Kockmann et al. (2011)). Thesecond scale-up step is to number-up the single channel in one single device - the modu-lar unit (Tonkovich et al. (2005)). The last step is to number-up the modular units, externalnumbering-up (Hasebe (2004)). In the previous chapters, no quantification was made regardthis scaling route. For example, no information was provided over the maximum produc-tion capacity each scaling step can reach, the maximum number of externally numbered-upBMMR units, or the maximum number of reaction channels a one BMMR unit can contain.The following sections are an attempt to quantify the scaling route shown in Figure 7.1.

Introduction 131

7.1.1 Maximum external numbering-up of modular units

In water treatment applications, reverse osmosis (RO) units are made of modular units whichcan be arranged in parallel or in series. The state of the art in numbering-up of reverse osmosisallows the operation of hundreds of modular units. The advancement in the field of externalnumbering-up in water treatment applications cannot be transferred directly to reactors andespecially not for microreactors. Microreactors are typically applicable for reactions whichare highly exothermic, mass transfer limited, toxic or even explosive. Moreover, reaction ratein this kind of reactors is often accelerated by operation under a ”Novel Process Windows”(using high temperatures and pressures or an alternative heating route) (Hessel et al. (2011)).Due to these demanding requirements, the maximum numbering-up of microreactors is ex-pected to be much less than that of RO units.

The increased system complexity for numbering-up of microreactors requires advancedprocess system engineering which is not fully mature for this kind of reactor arrangements.For example, each reactor modular unit will operate as an independent unit. Each reactormodular unit will require its own feedings, its own control unit, heating and cooling supply.However, all reactor modular units should follow an overall control system to assure theproduct quality is maintained. Because of that, there is need to monitor the product quality foreach modular reactor separately. Additionally, the entire process needs to be fully automated.So, issues like detecting leakages and reactor blockage become crucial. New ways will beneeded on how to take/remove one reactor from service, and how to handle shut-down/start-up procedures. Additionally, a new arrangement and layout of the manufacturing plant willbe needed. For example where and how to position the pumps and mass flow controllers; thelocation of the storage intermediate and product tanks and their pipe network; the locationof utility units and their control units; the separation unit if it is combined or per reactormodule. Due to all of these complex requirements, in this study the maximum number ofreactor modular units will remain within a small and limited number. We assume a valuenot more than 10 to 20. Developments in the field of external numbering-up of reactors willvalidate this assumption in the future.

7.1.2 Flow capacity and size of reactor modular unit

The minimum flow capacity for one reactor modular unit is shown in Figure 7.2. The cal-culations are based on the assumption that 10 to 20 maximum modular reactors can only beused. For three different production capacities and three residence times, flow capacities andreaction channels volumes are calculated as shown in Figure 7.2.


Figure 7.2: Flow capacity and reaction channels volume of one reactor modular unit calcu-lated for different production capacities and different residence times. Scale of cross sec-tional dimensions of reaction channel as micrometer (Micro), millimeter (Milli), or centime-ter (Centi) are presented.

The volume of reaction channels is determined by the residence time needed for the reac-tion. As the residence time decreases, the reaction channel volume decreases. For example,for a residence time of less than a second, a few mL of reaction channel volume is neededto reach a bulk scale capacity of 1000 L/h. However, when a residence time of minutes isneeded, a liters size reaction channels volume is required. Thus, external numbering-up andhundreds even thousands of parallel reaction channels will be needed to reach such a liter sizereaction channels. Additionally, the residence time is crucial in determining the scale of thereaction channel cross sectional dimensions as ”micro-scale” or ”milli-scale” as can be seenin the case of 1000 L/h.

Based on the analysis of Figure 7.2, two conclusions can be made regarding the bulkscale capacity of 1000 L/h. Firstly, the minimum flow capacity for one reactor modular unitis liters per minute. Secondly, reaction channels with micro-scale dimensions are only suitedfor very fast reactions which require a residence time of less than a second. Whereas forreactions which will require seconds or minutes residence times, milli-scale or even centi-scale reaction channels are the optimal scale for the reaction channels.

Introduction 133

7.1.3 Maximum number of parallel channels in one Barrier-based Mi-cro/milli Channels Reactor (BMMR)

For the BMMR, the maximum number of reaction channels in one modular reactor unit de-pends more on the requirements of the flow non-uniformity rather than on the reaction chan-nel volume and flow capacity. According to the design methodology developed in chapterthree and chapter five, a target flow uniformity is reached if the fabrication tolerance andtemperature deviation in each part of the modular reactor remains below certain limits. Thus,the maximum number of parallel reaction channels in one BMMR will depend more on thetechnology used for the fabrication and on the design architecture of how the reactor partswill be combined.

There is a wide variety of options available for the design architecture of micro/milli chan-nel reactors. However, not all of the design options are optimal. For example, the technicalchallenge and costs are very high when the reactor has all of the functional elements fabri-cated in one device. If a blockage or other problem happens inside such a reactor design, thefailed component cannot be replaced and the entire device has to be discarded. Additionally,fulfilling the requirements of fabrication tolerance and temperature deviation is a challengingtask. Therefore, a practical reactor design requires the reactor to be split into different partswhich will be combined together mechanically. Each part will contain one or more functionalelements depending on the reactor layout and implemented fabrication technology. Such asolution provides a wider flexibility in the choice of fabrication technology and material ofconstruction. Thus, fulfilling the requirements to reach a certain flow uniformity with an op-timal design and a reasonable fabrication cost.

The modular reactor is assembled by combining the reactor parts together. Each reactorpart will contain N number of parallel channels. Channel openings from the different partsneed to be connected to each other accurately and sealed. As a result, the way in which thereactor parts are combined becomes crucial. In some cases, like in the BMMR, it will bethe main factor which will decide the maximum number of parallel channels obtained in onereactor modular unit and its maximum operating temperature and pressure.

The BMMR is divided into four parts as shown in chapter four, gas and liquid manifolds,glass chips which contain barrier channels and mixers, reaction channels with integrated heatexchanger, and a collector block. These four parts are connected together using O-ringswhich are pressed together mechanically. O-rings may create limitations on the maximumtemperature and pressure of the reactor. In microfluidics, high temperature and high pressure


can be safely applied due to the small channel volume. However, the maximum operatingtemperature for O-rings is typically less than 400 oC and is dependent on the material theO-rings are made of. If operation at a higher temperature is required, then the possibilityfor using metal O-rings like copper or silver can be evaluated or modification of the reactordesign is needed. The maximum operating pressure depends on the design of the mechanicalholder which sealed the O-rings. A large operating pressure will require larger size holderwhich will increase the size of the reactor. In the BMMR, the current design is made to han-dle an operating pressure up to 20 bar and a temperature of 200 oC. Extending the designto larger temperature and pressure is a task to be evaluated in the second generation of theBMMR.

In addition to the limitations on temperature and pressure, O-rings here decide the maxi-mum number of parallel channels in one BMMR. For example, 4000 O-rings will be neededif the BMMR contains 1000 parallel reaction channels. Such a large number of O-rings is notpractical to work with. Therefore and due to these practical reasons, the maximum number ofchannels in one BMMR will be limited to a around hundreds of reaction channels. The totalnumber of O-rings in one BMMR will be limited to not more than a couple of hundreds.

7.1.4 Scale of reaction channels in the BMMR.

Previously, it has been determined that: (1) the maximum number of modular BMMR needsto be within a limited number, a value not more than 10 to 20 units is assumed here; (2)the minimum flow capacity of one reactor modular unit for bulk scale capacity is liters perminute; (3) cross sectional dimensions of reaction channel as ”micro”, ”milli”, or ”centi”meter scale will depend on the residence time of the reaction; and (4) the maximum numberof parallel reaction channels in one BMMR will be limited to not more than a couple of hun-dreds, a value not more than 100 to 200 is assumed. These four conditions are obtained fromthe analysis made on the limitations of numbering-up. In other words, these conditions are akind of top bottom approach to decide the operating window and reaction channel dimensionswhich suits the route of scale-up via numbering-up. However, such an operating window andreaction channel dimensions do not take into account the requirements of the reaction kinet-ics, mass and heat transfer of the single channel experiments. Matching between the fourlimiting conditions from the numbering-up and those from the single channel investigationscan be reached via two options. The first option is to accelerate the reaction rate via the useof ”Novel Process Windows”, and the second option is by scaling-up the single channel.

Typical reaction residence times which are suitable for ”micro-scale” or ”milli-scale”

Introduction 135

reaction channels are less than a few minutes (Roberge et al. (2005, 2009)). Within this re-action residence times, three sub-domains can also be classified which will have differentrequirements of the micro/milli channel reactor. These types are: type A which has a reactionresidence time less than a second; type B which has a residence time of a few seconds to min-utes, and type C which has a residence time larger than few minutes (Roberge et al. (2005,2009)). For reactions which require a residence time larger than few minutes, the conceptof novel process windows can be used to accelerate the reaction rate to type A or B. Oncethe reaction rate is accelerated, then the reaction residence time is fixed and the only way toincrease the production capacity of one reaction channel is via scale-up.

Figure 7.3: Different examples of scale-up by smart structuring in micro/milli channel reac-tor.

Increasing the reaction channel cross sectional dimensions can be achieved via smartstructuring, see Figure 7.3. Smart structuring ensures maintaining the same mass and heattransfer characteristics of the micro/milli scale reaction channel even when increasing the re-action channel cross sectional dimensions. Figure 7.3 shows some examples of smart struc-turing in micro/milli channel reactors. For example, changing the geometrical cross sectionaldimensions from circular to slit has been used to scale-up heterogeneously catalyzed gasphase reactions (Tonkovich et al. (2005)). The micro-scale channel dimensions are main-tained but only in one axis of the channel dimensions. For liquid phase reactions, mixing el-ements were introduced along the channel length with non-uniform cross section dimensions(Lavric and Woehl (2009), Roberge et al. (2009)). These mixing elements assure maintainingsimilar mixing and heating intensity to that of the micro/milli scale reaction channels as thereaction requires.


Most of the development on smart structuring is made for single phase flow. For multi-phase flow, the challenge for smart structuring is more demanding because the channel wallsplay a dominant role in stabilizing the interfaces between the phases (Lavric and Cerato-Noyerie (2012), Lavric and Woehl (2009)). Additionally, introducing mixing elements ornon-uniform structures along the reaction channel length will result in bubble splitting andcoalescence. This will lead to different flow dynamic behavior of the phases along the chan-nel length which could change the flow regime, the most important engineering part of thereactor. The success of this approach for multiphase flow depends if these flow dynamics canbe quantified and utilized for certain applications. This is a promising field which can still befurther explored.

7.2 Case study - Oxidation of ethylbenzene

The oxidation of ethylbenzene is used as an industrial application to demonstrate the analysismade previously on the numbering-up. Oxidation of ethylbenzene is one of the steps in theSM/PO process which was developed by Shell Chemicals Company (Klusener et al. (2007)).The SM/PO process produces propylene oxide and styrene. Both chemicals are producedon massive bulk scale with a capacity of a few million metric tons per year. The first stepin the SM/PO process is the air-oxidation of ethylbenzene (EB) to ethyl-benzene hydroper-oxide (EBHP) as shown in Figure 7.4. A brief description of this technology is given by(Klusener et al. (2007)). The oxidation step is conducted in cross-flow operation in a largehorizontal bubble column reactors in series with a very low aspect ratio: height 4.6 m, length15.25 m. Typically, a reactor train consists of 4 to 5 of such horizontal columns, which areequipped with baffles and heating/cooling coils. Air is introduced via separate middle andside sparger systems. The gas outlet stream contains, besides unconverted oxygen, a verysignificant amount of EB from the evaporation/stripping stage, and this EB is recovered in acondensing column and recycled to the reactor train.

The peroxidation of EB to EBHP is autocatalytic, implying that a minimum concentrationof the EBHP product is necessary to initiate the reaction and that the reaction goes faster byincreasing EBHP concentration. On the other hand, the higher EHBP concentration leads toincreased EBHP decomposition, by-product formation and lower selectivity. For this reasonbackmixing is desirable in the front-end of the first reactor, whereas plug flow behavior isdesirable in the remaining reactor train. To facilitate this, the reactors are segmented by a fewtransversal baffles. The way in which these baffles are arranged ensures a good mixing at theinlet of the reactor and plug flow behavior in the rest of the reactor (Klusener et al. (2007)).

Case study - Oxidation of ethylbenzene 137

Typical process operating conditions for the oxidation of EB are shown in Figure 7.4.

Figure 7.4: Schematic view of the SM/PO process developed by Shell Chemicals companiesto produce styrene and propylene oxide. Information is taken from Sumbharaju et al. (2011).

The energy research center (ECN) in the Netherlands has chosen the oxidation of ethyl-benzene as a suitable reaction to demonstrate the process intensification using micro/millireactor channel technology (Sumbharaju et al. (2011)). ECN has conducted the experimentsin a single channel setup made of capillary and tube fittings. The reaction was conductedin the Taylor flow regime. According to their first results, the reaction does proceed in themicro/milli reaction channel but with a lower conversion compared to the conventional pro-cess (Sumbharaju et al. (2011)). Further studies are still needed to reach a similar conversionto the SM/PO process. However, ECN experimental work gives indication that process in-tensification for the oxidation of ethylbenzene could be made using the micro/milli reactortechnology. By process intensification we mean here that the reaction residence time couldbe accelerated to the minutes scale. The minutes residence time is within the range accept-able to operate the BMMR. Thus, it is viable to evaluate the oxidation of ethylbenzene in theBMMR to reach bulk scale production.

7.2.1 Objectives of the case study

The oxidation of EB will be used as a case study to compare the horizontal bubble columnreactor (HBCR) to the BMMR which operates in the Taylor flow regime. The comparativestudy will be based on a production capacity of 1 m3/h of EBHP which corresponds to 8000


tons/yr. Including the diluent, which is EB and a target conversion of 10 wt%, the total flowcapacity will be 10 m3/h which equals to 80000 tons/yr. The study will compare two aspects:the size of the reactor and reactor operation.

Before proceeding, it should be made clear that the flow capacity mentioned here is atleast ten times smaller than a realistic industrial capacity for the SM/PO process. The flowcapacity of the SM/PO process is a large bulk-scale capacity which is beyond the capabilityof the BMMR at this development stage. The aim of this comparative study is to demonstratethe advantages of the BMMR for an industrial reaction at a reasonable bulk scale capacity ofm3/h and m3/min for liquids and gases, respectively.

7.2.2 Configuration and operating conditions for Horizontal Bubble Col-umn Reactor (HBCR) and BMMR

The BMMR configuration is made comparable to that of the HBCR to make it easy for com-parison. The design is made similar to the shell and tube heat exchanger with some excep-tions. In the BMMR, the reaction takes place inside the tubes, while in the HBCR the reactiontakes place in the shell side. Another difference is that the shell side in the HBCR containsair bubbles which are supplied by the sparger system from below. The un-reacted oxygenand gas vapors are accumulated on top of the liquid phase. Evaporation from the liquid phase(latent heat) is used to take away the large amount of heat generated by this highly exother-mic reaction to maintain an isothermal operation (Tavadyan et al. (2003)). The gas phasewhich composed of un-reacted air, vapors from reactant and product solution is condensedand recycled back to the reactor.

Process operating conditions of the HBCR are taken from Sumbharaju et al. (2011) as: 2bar, 140 oC , 10% conversion, 90% selectivity, and 1 hour residence time. A cross sectionalschematic view of the HBCR is shown in Figure 7.5. This figure is taken from Kluseneret al. (2007) where an experimental RTD has been conducted to study the gas and liquid flowdistribution. A CFD study was used to predict the liquid velocity field, gas hold-up, oxygenconcentration in the liquid, and oxygen consumption rate.

The main advantage of the HBCR configuration is the easyness of scale-up since the reac-tor can be arranged in series to accommodate the large reactor volume (volume of one reactoris around 250 m3). Disadvantages are: reaction conversion is limited by heat removal sincehigher conversion leads to a thermal run way reaction (Sumbharaju et al. (2011), Tavadyanet al. (2003)); flow hydrodynamics and gas flow distribution result in temperature flow non-


Figure 7.5: Top: Schematic drawing of the commercial reactor with the 14 detector locations(heat exchange coils not shown). Bottom, CFD predictions: (a) liquid velocity field (m/s);(b) gas hold-up (kg/m3); (c): O2 concentration in the liquid (mol/m3); (d) O2 consumptionrate (Kmol/m3). Result are taken from Klusener et al. (2007).

uniformity which affects the reaction performance. For example, in the CFD study shownin Figure 7.5, gas starvation in the center of the reactor was demonstrated which reduces thereactor productivity and yield.

In the BMMR, EB oxidation occurs inside the tubes as shown in Figure 7.6. The use of abarrier-based distributor assures a uniform flow distribution of the gas and liquid flows overall reaction channels and allows operation in the Taylor flow regime. The reaction occurs inthe liquid slug which maintains an intense mixing performance, as well as fast heat removal.Liquid slugs will not mix with each other and the reactor approaches that of an ideal plugflow. Additionally, radial temperature profile in the reaction channel is eliminated due to thesmall cross sectional dimensions. Therefore, heat removal is always maintained minimizingthe risk of hot spot formation resulting in a safer reactor.


Figure 7.6: Schematic drawing showing gas and liquid inlet directions, reaction and coolingparts for HBCR and BMMR.

In the BMMR, cooling/heating of reaction channels bundle is provided from the shellside. The shell side allows a larger flow capacity for the cooling medium which accommo-dates the excess amount of heat released by the reaction. However, this will depend on thebaffles and type of tube bundles. There are many tube pitches and layouts available in indus-try (Shah (2012)). The selection of the tube pitch is a compromise between a close pitch forincreased shell side heat transfer and surface compactness, and an open pitch for a decreasedshell side plugging and easy in shell side cleaning. For simplicity, a 90 o angle tube pitchlayout will be chosen as shown in Figure 7.7(i).

Currently, ECN is investigating the oxidation of EB to reach optimal intensified reactionconditions. No information is available yet about the optimal process operating conditions.Instead, values from their preliminary work and hypothetical ones will be used to perform thecomparative study (Sumbharaju et al. (2011)). The process operating conditions which willbe used in this study for the BMMR are: 5 bar, 180 oC, 10% conversion, and 90% selectivity.Regarding the reaction residence time, different hypothetical values will be evaluated.

7.2.3 HBCR size

An order of magnitude study is used to estimate the HBCR volume. For a target conversionof 10 wt%, a flow rate of 10 m3/h and a liquid residence time of 1 h, the volume of the liquidin the HBCR is 10 m3. The total reactor volume is estimated by dividing the liquid volume byits hold-up in the reactor. Three factors are used to estimate the liquid hold-up in the reactor:


Figure 7.7: Drawing of a cross section view of BMMR which is similar to shell and tubeheat exchangers (i). Enlarged view (ii) and (iii) are for the tube pitch which shows the crosssection view of one unit cell with nomenclature used to calculate some geometrical values.

(1) the gas hold-up in the liquid, (2) the volume of the heat exchanger and baffles, and (3) thevolume of the gas phase in top of the liquid phase in the reactor. An estimation of these threefactors is made by analyzing the cross sectional view in Figure 7.5. The gas hold-up in thesystem is assumed to be 50% of the liquid volume. The volume of the gas vapors in top ofthe liquid phase is assumed to be 35% of the total reactor volume. The hold-up of the heatexchanger (tube bundles and baffles) is assumed to be 25% of the reactor volume. Hold-upsof the HBCR are shown in Figure 7.8. According to these hold-ups, the total reactor volumecan be estimated at 50 m3. For a more accurate estimation of the reactor volume, reactionkinetics and a reactor model are needed.

7.2.4 BMMR size

The BMMR size is estimated by calculating the volume of one unit cell which is illustratedin Figure 7.7(ii) and (iii) and then multiplying that by the required number for parallelization.An example of how the calculations are made is given next. Before that, factors which areused in the calculations will be explained.


Figure 7.8: Hold-ups in the BMMR and HBCR. The symbol HX refers to the heat exchangerpart, and Vapor refers to the vapor space in the top of the liquid in the HBCR.

The volume of one unit cell consists of the reaction channel which contains the gas andliquid volumes, the channel wall and the heat exchanger. The reaction channel volume isestimated by multiplying the liquid flow rate per channel with the reaction residence timeand then dividing the result by the liquid hold-up. Different reaction channel volumes arecalculated for different hypothetical reaction residence times. A liquid hold-up in the reac-tion channels of 0.5 is assumed for all calculations. Due to the small channel dimensions,the volume of the reaction channel walls cannot be ignored and must be taken into account.Figure 7.9 shows the ratio of cross section area for internal channel diameter (Ad) over out-side channel diameter (At) for standard micro/milli capillary and tube sizes. The ratio Ad /At

varies between 0.1 and 0.6 which demonstrates the large volume of the channel wall forsuch channel sizes. A value of 0.4 for Ad /At is assumed when the volume of the unit cellis calculated. The volume of the heat exchanger around the reaction channel is estimated byassuming a distance equal to the internal reaction channel diameter on all four sides as shownin Figure 7.7(iii).

The BMMR size is calculated at the maximum number of parallelization and the mini-mum flow rate per channels which are the limiting conditions. They are the limiting condi-tions because adding more parallel channels is not possible and reducing the flow rate perchannel will not give the required target flow capacity. The maximum number of paral-lelization is assumed to be 20 BMMR units and each BMMR will have 200 parallel reactionchannels. In total there are 4000 parallel reaction channels. Dividing the flow rate by the 4000parallel channels gives the minimum flow rate per channel which is needed to reach the targetflow capacity via the route of numbering-up. Larger flow rates per channel is an advantagesince it will reduce the needed number of BMMR’s or the number of channels in one BMMR.


Figure 7.9: Ratio of a cross section area for internal channel diameters (Ad) over outsidechannel diameters (At ) for standard sizes of capillary and tubing. Values are taken fromswagelok tubing.

The procedure for calculating the volume of BMMR is demonstrated for a reaction resi-dence time of 2 minutes. The minimum flow capacity of one modular BMMR unit is 0.5 m3/hwhich is equal to 8.3 L/min. The minimum liquid flow rate per reaction channel is 42 mL/minwhen dividing the BMMR flow rate by the 200 parallel channels. For a 2 minutes reactionresidence time, the liquid volume in the reaction channels is equal to 84 mL. Dividing thevolume by the liquid hold-up in the reaction channel, the volume of one reaction channel is168 mL. Dividing this volume over the hold-up of channel walls and heat exchanger, the vol-ume of one unit cell equals to 2.7 L. Multiplying that by 200 gives the volume of one BMMRunit which will be 0.55 m3. Multiplying that by 20 BMMR units, the total reactor volume is11.0 m3. The reactor size in such a case is 5 times smaller than the size of the HBCR. Thesame calculations can be used to find the BMMR volumes for other reaction residence times.

7.2.5 Dimensions of BMMR reaction channels

In the previous section, the minimum flow rate per reaction channel and the volume of onereaction channel were determined. No information was given about the possible reactionchannel dimensions. Here, the suitable reaction channel dimensions will be estimated basedon the limitations from pressure drop and the required flow regime. Pressure drop is the keyparameter to assure an adequate flow distribution, while the flow regime is the most importantengineering factor and determines the BMMR window of operation.

The pressure drop of the reaction channel is calculated for different reaction channel di-


ameters from 0.25 mm to 5 mm. For each channel diameter, the channel length is adjusted tomatch the required reaction residence time. Some assumptions are made to obtain these re-sults. Physical properties of EB and air are used in the calculations. A circular cross sectiongeometry is assumed for the reaction channel. Slug and bubble lengths are assumed to betwice the reactor channel diameter. These specific slug and bubble lengths can be obtainedvia the design of the T-mixer (van Steijn et al. (2010)).

A conservative operating window for Taylor flow regime is assumed here. A uniformTaylor flow is reached when: (1) the maximum channel diameter is not more than 5 mm,and (2) the gas and liquid velocities are lower than 0.5 m/s. Liquid velocity is obtained bydividing the liquid flow rate obtained from the previous section, with the reaction channelcross sectional area. Gas velocity is calculated based on the gas to liquid flow ratio needed toobtain a gas hold-up of 0.5.

Figure 7.10: Reaction channel length and pressure drop as a function of different reactionchannel diameters. Results are calculated for different reaction residence times given in min-utes near to each line.

Figure 7.10 shows the length of reaction channel and the pressure drop as a function ofdifferent reaction channel diameters. Figure 7.10 can be used to identify optimal reactionchannel dimensions suitable for this case study. Two design limits are specified to help lo-cate the suitable reaction channel dimensions. Values larger than the design limits will bediscarded. The first design limit is when the pressure drop over the reaction channel is largerthan 5 bar. This is because pressure drop over the barrier channels will be substantial, makingthe BMMR a non-attractive design option. Also, a large pressure drop in the reaction channelcould affect the stability of the Taylor flow regime. The second design limitation is when therequired reaction channel length is larger than 50 m. Such a long reaction channel might notbe a very practical solution since reaction channels are grouped as a bundel of 200 parallelchannels.


Under all tested conditions, reaction channel diameters smaller than 2 mm exceed thedesign limits. Thus, only reaction channel dimensions between 2 mm and 5 mm will be con-sidered for locating the suitable reaction channel dimensions. For a reaction residence timeof 10 minutes and using a maximum reaction channel diameter of 5 mm, the required channellength is longer than 40 m. The pressure drop is less than 0.2 bar. When the channel diameterdecreases, the channel length and pressure drop increases. For example, when the channeldiameter reduces by half from 5 mm to 2.5 mm, the required channel length becomes longerthan 100 m with a pressure drop larger than 5 bars. Such a design option is discarded becauseit exceeds the design limits. So, using a reaction channel diameter in the range of 4 mm to5 mm is a more suitable design option for the 10 minutes reaction residence time. When alower reaction residence time is used, a smaller reaction channel dimensions becomes moresuitable as a design option. For example, when the reaction residence time is 1 minute, therequired channel length is 4 m for the 5 mm channel diameter and 27 m for the 2 mm channeldiameter. In both cases pressure drop is lower than 0.5 bar. Therefore, both channel dimen-sions are suitable design options in the BMMR configuration.

The above analysis demonstrates the trade-off between reaction channel diameter, lengthand pressure drop at varied reaction residence times. However, only limitations from theBMMR device and Taylor flow were taken into consideration. Other considerations such asmass and heat transfer characteristics, thermal runaway and hot spot formation, reactor sta-bility, back mixing and axial dispersions were not included. These additional considerationsshould be addressed in the scale-up study and experimentally using a single channel exper-imental setup which is beyond the scope of this study. There is the need to match betweenthese limitations and the ones from the numbering-up. An example is given to illustrate that.In the case of a 10 minutes reaction residence time, the optimal reaction channel dimensionsare: channel diameter between 4 mm to 5 mm and channel length between 40 m to 80 m.These are the limitations from the numbering-up. If then the scale-up study suggests that themost suited reaction channel diameter is 1 mm, smart scale-up by making internal structuresis needed which can be made by converting the 5 mm channel diameter tube to 1 mm con-finements. If that doesn’t work, the choice of the flow regime needs to be revaluated.

7.2.6 Comparative result of the HBCR versus BMMR

A comparison between the reactor size of the HBCR and BMMR is shown in Figure 7.11.The volume of the HBCR is calculated at the process conditions of the SM/PO process. The


volume of BMMR is calculated for hypothetical reaction residence times. The volume of theBMMR is larger than the HBCR when the reaction residence time is 10 minutes. This resultis justified by the difference between the liquid hold-up and total reactor volume for bothreactors as shown in Figure 7.8. When the reaction residence time decreases to less than 10minutes, the volume of the BMMR becomes less than the HBCR. For a reaction residencetime of 2 minutes and less, the volume of the BMMR is 5 to 10 times smaller than the HBCR.Such a reduction in the reactor volume is substantial which can be regarded as process inten-sification.

Figure 7.11: Comparing of reactor volumes between the HBCR calculated at the conditionsof the SM/PO process, and BMMR calculated at hypothetical different reaction residencetimes.

The rate limiting steps of the HBCR are the cooling capacity and the distribution of thegas and liquid flows. Large temperature gradients in the HBCR are observed if the spargersystem does not assure uniform air distribution. This reduces the reactor productivity andyield as shown previously in Figure 7.5. Additionally, the large temperature gradient in theshell side limits the cooling capacity in the tubes and increases the amount of EB evaporation.An additional amount of energy is then needed to recover back the evaporated EB. Usingthe BMMR, the reaction will take place in the reaction channels. Due to the small channeldiameter, radial temperature gradients will be minimized which will ensure constant anduniform heat flux to the shell side. Therefore, generation of hot zones and large temperature

Conclusions 147

gradients are reduced. The barrier-based distributor will assure uniform gas and liquid flowdistribution which will result in a more precise operation with reduced evaporation and energyconsumption. Additionally, the reactor layout via external numbering-up will allow a flexibleproduction capacity. The production capacity of the plant can operate with a factor range of20 which corresponds to the 20 BMMR units.

7.3 Conclusions

The oxidation of ethylbenzene was used as a case study for the process intensification usingthe micro/milli channel reactor. The BMMR can be used to scale-up ethylbenzene oxidationvia numbering-up to a production capacity of 1 m3/h for the EBHP. Process intensificationusing the BMMR occurs only when the rate of reaction is accelerated to a reaction residencetime of less than 2 minutes, with a conversion of 10 wt% and a selectivity of 90%. Thisstudy is based on the condition that the rate of reaction will be accelerated in the micro/millichannel reactor. Work progress at ECN and further studies on a single channel experimentalsetup might validate that. Meanwhile, the case study of using the BMMR to reach bulkscale production for the oxidation of EB can be made more comprehensive. For example,a techno-economical study can be a valuable tool in estimating the expected BMMR capitalexpenditure and operating costs. For that a reaction kinetic and a reactor model will beneeded. Additionally, a new conceptual design of the BMMR can be realized to accommodate200 parallel reaction channels similar to that of a shell and tube heat exchanger design. Sucha work will definitely provide a better overview of the potential of the BMMR to scale-up multiphase flow applications to industrial bulk scale capacities in the range of m3/h andm3/min for liquids and gases, respectively.

Bibliography

Hasebe, S., 2004. Design and operation of micro-chemical plants–bridging the gap betweennano, micro and macro technologies. Computers & Chemical Engineering 29, 57–64.


Klusener, P. A. A., Jonkers, G., During, F., Hollander, E. D., Schellekens, C. J., Ploemen,I. H. J., Othman, A., Bos, A. N. R., 2007. Horizontal cross-flow bubble column reactors:CFD and validation by plant scale tracer experiments. Chemical Engineering Science 62,5495 – 5502.

Kockmann, N., Gottsponer, M., Roberge, D. M., 2011. Scale-up concept of single-channel


microreactors from process development to industrial production. Chemical EngineeringJournal 167, 718 – 726.

Lavric, E. D., Cerato-Noyerie, C., 2012. Mass transfer in gas-liquid flow in CorningAdvanced-Flow reactors. Chemical Engineering Transactions 29, 979–984.

Lavric, E. D., Woehl, P., 2009. Advanced-FlowTM glass reactors for seamless scale-up.Chemistry Today 27, 45–48.

Roberge, D. M., Ducry, L., Bieler, N., Cretton, P., Zimmermann, B., 2005. Microreactortechnology: A revolution for the fine chemical and pharmaceutical industries? ChemicalEngineering & Technology 28, 318–323.


Shah, R. K., 2012. Ullmann (2007): Ullmann’s Encyclopedia of Industrial Chemistry- Elec-tronic version. Wiley-VCH, Weinheim.

Sumbharaju, R., Correia, L. A., de Groot, A., van Delft, Y. C., 2011. Oxidation of ethylben-zene in a millimeter G-L Taylor flow reactor. In: 8th International Conference on Catalysisin Multiphase Reactors - CAMURE8.

Tavadyan, L. A., Martoyan, G. A., Minasyan, S. H., 2003. Determination of the kinetic sig-nificance of elementary steps in the reaction of ethylbenzene oxidation inhibited by ionol.Kinetics and Catalysis 44, 490–498.



Conclusions and outlook 888.1 Conclusions

This thesis aimed at developing design rules and an engineering methodology for scaling-upmultiphase flow in micro/milli channel reactors to reach bulk scale production in the range ofof m3/h and m3/min for liquids and gases, respectively. Scale-up is made here via numbering-up which is a necessary step to reach bulk scale-production for micro/milli channel reactors.The key challenge for numbering-up is to assure equal flow conditions over the parallel reac-tion channels. The barrier-based flow distributor has shown promising results for numbering-up multiphase flow. A key characteristic of the barrier-based distributor is to provide thehydraulic flow resistance needed to achieve equal flow distribution. It can be quantified ina generic way as ∆PB as given in Equation 8.1. It is the average pressure drop over the bar-rier channels ∆PB divided by the average pressure drop over the corresponding mixers andreaction channels ∆PC. Since ∆PB is a ratio of pressure drops, it is dimensionless.

∆PB =∆PB

∆PC(8.1)

Early studies in this field have used a pressure drop ratio ∆PB between 25 to 50 to dis-tribute the flows in a micro channels reactor (De Mas et al. (2005), Losey et al. (2002)). How-ever, such a large pressure drop ratio makes the barrier-based flow distributor a non-attractivesolution for numbering-up. The large energy consumption and lowering of the reactor operat-ing window are some of the reasons for that non-attractiveness. Using an experimental setupmade of capillaries and tube fittings, the minimum required pressure drop ratio was identifiedin chapter two. The following design rules were extracted:

• The minimum optimal pressure drop ratio ∆PB lays between 4 and 25.

150 Conclusions and outlook

• Gas-liquid channeling is prevented when the pressures of gas and liquid manifolds areequal.

• Variations in internal diameter of microchannels due to fabrication tolerances in dif-ferent parts of the reactor have a different impact on flow distribution. The largestimpact on flow distribution comes from the barrier channels. For example, in circu-lar cross sectional barrier channels, each percentage of diameter variation affects flowdistribution by a factor of four.

• The barrier-based flow distributor is optimal for a certain range of gas and liquid flowrates. That optimal range is quantified as the ratio when the maximum flow rate overminimum flow rate used is less than a factor of 20.

Increasing the pressure drop ratio ∆PB alone cannot assure reaching the desired flow uni-formity. A design methodology was developed in chapter three to reach the desired targetflow uniformity. The desired flow uniformity was reached when the fabrication tolerancein each part of the reactor remains below certain cut-off values. These cut-off values werequantified using the 2−PRN model. The 2−PRN model is based on the hydraulic resis-tance network model which was developed and extended to account for multiphase flow. The2−PRN model was experimentally validated using the capillaries and tube fittings setup.

The barrier-based micro/milli channels reactor (BMMR) is a prototype reactor which wasused to demonstrate the numbering-up of multiphase flow. The BMMR was designed to de-liver a flow non-uniformity lower than 10%. The BMMR was designed according to theextracted design rules from chapter two and the analytical model and design methodologyfrom chapter three. In chapter four, the flow uniformity of the BMMR was investigated ex-perimentally. A flow non-uniformity lower than 10% was achieved at different viscosities,surface tensions, and gas and liquid flow rates. Additionally, the modularity of the BMMRwas demonstrated without any loss in flow uniformity for microchannels fabricated in (1)stainless steel plate, (2) glass plate, and in (3) stainless steel capillaries. The barrier-basedflow distributor is a versatile device for varied reaction channel dimensions and materials.The key design parameter which allows such a modularity is the pressure drop over the reac-tion channels. The pressure drop over the reaction channels needs to be within the optimalrange specified for the dimensions of the barrier channels.

The influence of temperature on the flow distribution and vice versa was investigated inchapter five. A desired flow uniformity was reached when the temperature deviation in differ-ent parts of the BMMR remained lower than cut-off values similar to that of the fabrication

Conclusions 151

tolerance. The impact of temperature deviation in the barrier channels on flow distributionwas the largest. For example the flow non-uniformity is ten times larger when compared toa similar temperature deviation in the mixer and reaction channels. For a constant heat fluxover the reaction channels and without any reaction, the temperature deviation at the outletof the reaction channels was a function of the liquid residence time. Larger than a criticalresidence time, there was a negligible temperature deviation even for a wide range of flownon-uniformities. Below this critical liquid residence time, the temperature deviation de-pended on the flow uniformity. The value of this critical liquid residence time is a functionof the liquid used, the BMMR material of construction, and its geometrical dimensions.

The work performed in the previous chapters was conducted without any reaction. Inchapter six, the BMMR was tested using the hydrogenation of phenylacetylene to styrene andethylbenzene using [Rh(NBD)(PPh3)2]BF4 as the catalyst. First, a parametric study was con-ducted to quantify the effect of pressure, catalyst concentration, and substrate concentrationin a semi continuous batch reactor. Catalyst deactivation was observed and incorporated intoa kinetic model which was extended from the work of Esteruelas et al. (1998). The kineticmodel predicted the experimental results within an accuracy of 20%. A preliminary studywas made to analyze the reaction performance in the BMMR. Concentrations of reactant andproducts of a single reaction channel were compared to that from eight reaction channelscombined. For 95% of the obtained results, the difference in concentrations between the sin-gle channel and the eight combined channels remained within ± 10%. The numbering-up ofmultiphase flow under reactive flow conditions in the BMMR has been successfully demon-strated.

Chapter seven revises the objectives of the DeMiR project based on the achieved knowl-edge. The capability of the BMMR was analyzed based on the concept of numbering-up toreach bulk scale production for multiphase flow applications. A case study was made usingthe oxidation of ethylbenzene as an industrial application. The study was made for a totalflow rate of 10 m3/h with a conversion of 10 wt%. Conventionally, the oxidation of ethylben-zene is conducted in a horizontal bubble column reactor (HBCR). The size of the HBCR wascompared to the size of a BMMR. The BMMR achieves process intensification by reducingthe equipment size 5 to 10 times, if the reaction residence time is reduced from one hour inthe HBCR to less than a few minutes in the BMMR.

152 Conclusions and outlook

8.2 Outlook

A comprehensive overview over the capabilities of the barrier-based micro/milli channelsreactor to reach bulk scale production is achieved by doing the following:

• Heat transfer studies: The effect of temperature deviation and flow distribution wasinvestigated in chapter five. Constant and uniform heat supply was maintained in thatstudy. When a heat exchanger is integrated with the reaction channels, the type of theheat exchanger and the reaction performance will determine the temperature deviationand thus the flow distribution. The impact on reaction performance therefore needs tobe reassessed.

• Heterogeneous catalysts: Coating the catalyst at the reaction channel walls is the mostsuited form to implement heterogeneous catalysts in the BMMR. Pressure drop overthe reaction channel will be minimum and the reactants and products at the catalystwill be continuously refreshed due to the good external mass transfer. However, theamount of catalyst per reaction channel will significantly influence the reactor perfor-mance. Previous studies on heterogeneously catalyzed gas phase reactions have shownthat catalyst distribution is one of the most important factors which affects the reac-tor performance (Delsman et al. (2005)). Experimental studies on catalyst distributionover multiple reaction channels are therefore a must to evaluate the capability of theBMMR for heterogeneously catalyzed reactions.

• Annular flow regime and liquid-liquid multiphase flow: The developed design method-ology and engineering correlations were made as a function of pressure drops. Thus,it is possible to use them for other flow regimes if a correlation for pressure drop isknown. Because it is possible to estimate pressure drop for annular flow regime andfor organic-aqueous (liquid-liquid) slug flow, it is possible in principle to use the de-velop design methodology for numbering-up these two attractive flow regimes.

• Application demonstrations: A good approach to evaluate the capabilities of the BMMRis to perform selected demonstration applications. One of the ultimate tests is the poly-merization reaction. The viscosity of the reaction fluid in a polymerization reactionchanges along the reaction channel length. Change of viscosity influences the pressuredrop which is the key for regulating the flow distribution. For example, in a channelwhich has a low flow rate, the reaction will proceed further than in a channel with alarger flow rate. The flow will be more viscous in the channel with the low flow rateand the other way around. This significant dependency of the pressure drop on reac-tion performance makes the polymerization reaction the ultimate test reaction for theBMMR.

Bibliography 153

• Conceptual design of the BMMR with large numbers of reaction channels: This isnot so much a scientific challenge as much as it is a technological challenge. Sucha challenge will help in providing a realistic operating window of the BMMR andprovides a better overview of the limitations and capacity which one BMMR unit canreach.

• Designing an online gas-liquid separator: In this research, a gravity based gas-liquidseparator was used. The volume of the liquid and gas in such a separator was large.Such a solution is an unsuitable one if toxic or explosive mixtures are involved wherelow liquid hold-up is desired. On-line separation of gas and liquid in the microchanneltherefore provides an attractive solution. Additionally, online separation can simplifythe process. Once the gas flow is separated from the liquid flow, the reaction will stopwhich will remove the need for a quench line to stop the reaction.

• Process system engineering and techno-economical study: If the micro/milli reactortechnology will be used on an industrial scale, then a new plant layout is needed. Thisis needed because the scaling route of the technology is based on replication of thereactor unit via numbering-up. Handling externally numbered-up reactors is a complextask, as explained in chapter seven, which will require process system engineeringstudies. Additionally, a techno-economical study will be necessary to illustrate clearlythe advantages and disadvantages of this technology and it is economical attractiveness.

BibliographyDe Mas, N., Gunther, A., Kraus, T., Schmidt, M. A., Jensen, K. F., 2005. Scaled-out multi-

layer gas-liquid microreactor with integrated velocimetry sensors. Industrial and Engineer-ing Chemistry Research 44, 8997–9013.

Delsman, E. R., de Croon, M. H. J. M., Elzinga, G. D., Cobden, P. D., Kramer, G. J.,Schouten, J. C., 2005. The influence of differences between microchannels on micro reac-tor performance. Chemical Engineering & Technology 28, 367–375.



Nomenclature

Roman Symbols

as Specific interfacial area m2

m3reactor

A Channel cross-section area m2

Ab Bubble cross-section area m2

C Concentration mol/m3

Cab Capillary number based on the liquid properties and the gas bubblevelocity -

CaGL Capillary number based on the liquid properties and sum of the super-ficial gas and liquid velocities -

d Channel nominal hydraulic diameter m

F Dimensionless factor -

h Heat transfer coefficient W/m2.K

H Channel height m

Hmixer Depth of T-mixer channels m

k Thermal conductivity W/m.K

L Channel length m

mcat Catalyst concentration g/L

N Number of channels -

P Pressure Pa

q Flow rate m3

Q Energy term W

R Hydraulic resistance Pa sm3

Re Reynolds number, ρUdµ -

156 Nomenclature

Reb Reynolds number based on the liquid properties and the bubble veloc-ity -

ReGL Reynolds number based on the sum of the superficial gas and liquidvelocities -

T Temperature oC

ub Bubble velocity m/s

U Superficial velocity m/s

W Channel width m

Wmixer,G,in Width of inlet gas channel in a T-mixer m

Z Slug or bubble length m

Greek Symbolsεg Gas hold up -

τ Residence time, τ = LC(UG+UL) s

δs Correction on slug length to account for the nose and tail of bubble m

δ f Random tolerance added to the inner diameter of a channel to simulatethe fabrication tolerance m

γ Surface tension N.m−1

λNC Non-circularity factor depends on the channel geometry -

µ Viscosity Pa.s

ρ density kg/m3

cp Heat capacity W/m.K

ζc Singularity loss due to sudden contraction -

ζe Singularity loss due to sudden increase in cross section area -

ζm Singularity loss due to combining two branches -

ζs Singularity loss due to splitting of two branches -

ζt Singularity loss due to turning -

Subscriptss Slug -

b Bubble -

B Barrier -

C Microchannel of gas-liquid flow -

E Exit channel -

157

G Gas -

L Liquid -

M Manifold -

S Separator channel -

T T-mixer -

Acknowledgments

At last the PhD project has come to an end. I am deeply grateful to many people who con-tributed to the realization of this PhD thesis and to those who colored my life in the last fiveyears in Eindhoven. Below is the least I can do to express my sincere appreciation and thanksto all of you.

I would like to start with my promoter Prof. Jaap Schouten. Dear Jaap, I am very gladthat I did my PhD degree under your supervision. I enjoyed learning from the many discus-sions and meetings we had together especially on how to be professional. I appreciate theconfidence you gave me to lead and direct the project in the way I wanted while at the sametime being critical and giving valuable comments on the best way of presenting the achievedresults, in scientific papers and presentations, and achieving the desired goals.

I am very thankful to my daily supervisor Dr. Xander Nijhuis and to my co-supervisorsProf. Volker Hessel and Prof. Evgeny Rebrov. Dear Xander, I greatly appreciate your inputand advice during some of the critical stages of the PhD project which resulted in a workbased on a much stronger foundations. Many thanks for all of your effort and support. DearVolker, your input to reach complete papers was always of great help. Besides the PhDwork, you expanded my perspective and experience of the scientific world especially whenwe worked together on the MIFCAB workshop in Jordan and the follow up ideas. Thanksfor all that you have done. Dear Evgeny, although you left to the Queen’s University Belfastduring the beginning of the project, your scientific input remained very critical and valuableto the progress. I was very glad that you visited me in Jordan and joined the MIFCAB work-shop and I hope that next time you will get confident enough to ride a camel.

This project would not have happened without the financial support of STW and IROPwhich is greatly appreciated. Many thanks to all of the industrial partners ECN, DSM, Fries-land Food, Akzo Nobel, Micronit, Flowid and Delft and Wageningen Universities who al-

160 Acknowledgments

ways provided valuable comments during the progress meetings held. Special thanks to HansKroon who provided me with helpful industrial inputs during the visits I made to DSM; andto Arend de Groot, Luci Correia, Raghavendra Sumbharaju from ECN who always were opento help, support, and provide me with data needed for the case study.

I would like to thank Dr. Christian Muller who allowed me to use the organic laboratoryand conduct the hydrogenation batch experiments and to Leen Broeckx who was very kindto guide me in the lab. I also would like to thank all of the master students who worked hardon some of the experimental parts: Luuk Fluitsma, Fangyuan Yu and Johannes Zalucky. Iwould like to thank my classmate and good Indian friend, Narendra, for helping me in theheat exchanger experiments and to all the wonderful time we spent together. Many thanksas well to Maria ”Anto” who always was ready to lend me fiber optics and other laboratorytools, and letting me discover the Venezuelan leaf culture.

Now I direct my appreciation to the long life survivals of the SCR group who keep itfunctioning and in order. Denise who was always open to answer and help at any time and99% of the time knows the answer to a wide variety of topics. Anton who is super critical atfirst (especially with P&ID), but super helpful when you know him and discover his taste ofart. Peter and Marlies who were always very friendly and ready to help in ordering chemicals,instruments and doing analyses. Carlo ”the marathon runner” for helping with constructingthe labview program of the first experimental setup and keeping all the SCR people safe.Many thanks to Paul from the GTD for doing an excellent automation of the Millipede setup.

Here my gratitude goes to the great men in the workshop, the backbone of the SCR group,who allow all PhD’s to realize their ideas and dreams. Chris with his love to porto wine andships. Dolf, who is always very kind and ready to help at any moment. Erik, with his originaltaste for good cars, especially those of the fire brigade; the project would have take muchlonger without your help and support. Theo, the man with the green uniform who I did notrecognize when I saw him for the first time with normal clothes; I greatly appreciate buildingthe Millipede setup. And Madan, the brain and the creative mind behind all of the innovativedevelopment of the BMMR reactor, I can just say that I enjoy all the wonderful time we spenttogether and I hope to see you soon in Jordan and I promise that you will try some good foodwhich is not dry like sand.

I would like to acknowledge the SCR colleagues and friends, those who left and thosewho are still around. Many thanks guys for making life in the SCR group enjoyable with alot of unforgettable memories, nice conference nights and wonderful parties. I would like to

161

give special thanks to all of my office mates: Marco, the smart person who knows how totake good pictures and do true innovation, Jiaqi with his impressive observation skills withjokes and ideas which you wonder how they come to his mind, the well organized Swiss,Christine, with the enjoyable trip of cycling for a better world, Elnaz who introduced me tothe best international meal ”Abgoshet” and the unforgettable memory of wonderful Persianmusic, Fransizka the energetic, happy and lively person, and Wim with his true separatorwhich works in no gravity, in space, and in the Antarctic.

During the PhD, I met wonderful people with whom I have priceless memories andmade my social life in Eindhoven beautiful. Branito, the Serbian strategist whose experi-ence touches all parts of life, but he is afraid of cats and full moon, all the best with Anina.Carlitos, the true friend who always reignites my passion to knowledge, and made me the firstMoorsih visiting the beautiful Eshkerish. My dear good friend Faysal who always surprisedme with his multi-capacity and diverse skills but he only introduced me to Tofique. The Ital-ian Agnese who is a real angel in all senses even when she loses in the Lasagna competition.Maria Eriksson who allowed me to discover the Viking world, with their social toilet, mid-summer celebration and sophisticated music of Nyckelharpa. The lovely and lively SouthAmerican friends Paty, Loly and George (you are part of the family now), guys I will missyou so much. Nopita who allowed me to discover the beauty of Katarini with its wonderfulpeople and the fantastic time in France eating in Bushane. Ana and Roman, you are the bestcooks and travel companions ever, even though you successfully tricked me and Carolitos inDenmark.

Special thanks to the original salsa dancer Roman who impressed me with his specialpersonal collections of Voodoo doodoo, German castes, passa passa, and fake knives. Wel-come soon in the bamboo and tomato business. The capitalist Dries who let me breath thefresh Belgian air in a wonderful bike trip, all the best with Simonita who will see your hairevery morning. Ana Luiza and Fernando the lovely new married couple. Your wedding wasfantastic and I will never forget it. Mara and Simon with their new Lucy which is under fullexperimentation, many thanks for allowing me to try the Sicilian scooter and Rap Futoristico.Lara Truuuutter (Eeehhheeee Qha Qha- so difficult to write) the very nice and kind Africanperson whom Taught me kuza and the right order of knife, fork and spoon. And Vitaly andLiza, the nicest couple I met in Eindhoven whose lifes are of an inspiration. I hope one dayyou will have a farm which has for sure a pony and cat.

Additionally, I had a wonderful group of friends who always supported me during thePhD period. Abu Slooh shaik el shabab the most honest and truthful person; Abu housain

162 Acknowledgments

the best cook and maker of knaifeh; Abu Turkey and Sami, Abu Al Maati, Taleb, Mansoor,Mohamad Fiad, Abdel Hamid, Khaled Azzam and Jafer-Galob the best friend.

Behind the scene and outside the European continent there was my friends and fam-ily from Jordan, and my home family: Majdi and Asama and Anas el mankoush; Omar andKawlah, Houthifa abu galeb, Banan Um haider, Asoul Asal and Raheeq; Amouneh and Raed,Suhaib the great, my lovely Lolo and Yaser the shrewd; Aymen el herash, Amar abu osheba;Emad and Sara abu and um fawzi; and my beloved father and mother, words cannot describemy appreciation and obligation to you. All my love and thanks to all of you.

Ma’moun Al-Rawashdeh, Eindhoven, 20-03-2013.

List of Publications

Journal Publications

• Al-Rawashdeh, M., Zalucky, J., Muller, C., Nijhuis, T.A., Hessel, V., Schouten, J.C.,2013. Phenylacetylene hydrogenation over [Rh(NBD)(PPh3)2]BF4 catalyst in num-bered up microchannels reactor. Industrial & Engineering Chemistry Research, Inpreparation.

• Al-Rawashdeh, M., Yu, F., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C.,2013. Designing flow and temperature uniformities in parallel microchannels reactor.AIChE Journal, In preparation.

• Al-Rawashdeh, M., Yu, F., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C.,2012. Numbered-up gas-liquid micro/milli channels reactor with modular flow distrib-utor. Chemical Engineering Journal, 207-208, 645-655.

• Cantu-Perez, A., Al-Rawashdeh, M., Hessel, V., Gavriilidis, A., 2012. Reaction model-ing of a microstructured falling film reactor incorporating staggered herringbone struc-tures using eddy diffusivity concepts. Chemical Engineering Journal, online.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2012. De-sign methodology for a barrier-based two phase flow distributor. AIChE Journal, 58-11,3482-3493.

• Al-Rawashdeh, M., Fluitsma, L.J.M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten,J.C., 2012. Design criteria for a barrier-based gas-liquid flow distributor for parallelmicrochannels. Chemical Engineering Journal, 181-182, 549-556.

• Al-Rawashdeh, M., Cantu-Perez, A., Ziegenbalg, D., Lob, P., Gavriilidis, A., Hessel,V., Schonfeld, F., 2012. Microstructure-based intensification of a falling film microre-actor through optimal film setting with realistic profiles and in-channel reduced mixing.Chemical Engineering Journal, 179, 318-329.

• Ziegenbalg, D., Lob, P., Al-Rawashdeh, M., Kralisch, D., Hessel, V., Schonfeld, F.,2010. Use of ’smart interfaces’ to improve the liquid-sided mass transport in a falling

164 List of publications

film microreactor. Chemical Engineering Science, 65(11), 3557-3566.

• Al-Rawashdeh, M., Hessel, V., Lob, P., Mevissen, K.A.G.H., Schonfeld, F., 2008.Pseudo 3-D simulation of a falling film microreactor based on realistic channel andfilm profiles. Chemical Engineering Science, 63(21), 5149-5159.

Oral Presentations

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2012.Numbering up gas-liquid flow in micro/milli channel reactors. Proceedings of the 22ndInternational Symposium on Chemical Reaction Engineering (ISCRE22), 2-5 Septem-ber. Maastricht, The Netherlands.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2012.Scale-up of multiphase micro/milli channel reactors. Proceedings of the 12th Interna-tional Conference on Microreaction Technology (IMRET12), 20-22 February. Lyon,France.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2011.Scale-up of multiphase micro/milli channel reactors. Proceedings of the MicroNanoConference 11, 15 - 16 November. Ede, the Nederlands.

• Al-Rawashdeh, M., Nijhuis, T.A., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten,J.C., 2010. Uniform gas-liquid flow distribution in parallel micro and milli channelsin the Taylor Flow Regime The two phases resistance network model. NetherlandsProcess Technology Symposium (NPS-10), (pp. 80-80). Veldhoven, the Nederlands.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2010. Scal-ing rules for multiphase gas-liquid flow distribution in multiple micro/milli channelsrunning in parallel under segmented flow regime. Proceedings of the 19 InternationalCongress of Chemical and Process Engineering (CHISA 2010) and the 7th EuropeanCongress of Chemical Engineering (ECCE-7), 28 August - 1 September, (pp. 356-357). Prague, Czech Republic.

• Al-Rawashdeh, M., Hessel, V., Lob, P., Schonfeld, F., 2009. Impact of channel geom-etry on the performance of a falling film microreactor - without and with in-channelstructuring. 47th European two-phase flow group meeting 2009, 1st joint ETPFG-EPCE multi-phase meeting, 3-6 June Bled, Slovenia, (pp. 1). Bled, Slovenia.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C. (2009).Multiphase flow distributor for multi micro/milli channels running in parallel undersegmented gas liquid flow regimes. Proceedings of the IMM Young Scientists Work-shop, 19-20 November. Mainz, Germany.

165

• Al-Rawashdeh, M., Hessel, V., Lob, P., Schonfeld, F., 2008. Pseudo 3-D simulation ofa falling film microreactor. Proceedings of the European Comsol Conference, Novem-ber 4-6, (pp. 1-7). Hannover, Germany.

Poster presentations

• Cantu-Perez, A., Al-Rawashdeh, M., Hessel, V., Gavriilidis, A., 2012. Studies of mi-crostructured falling film reactor using eddy diffusivity concepts. 12th InternationalConference on Microreaction Technology (IMRET12). Lyon, France.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2011.Numbering up of gas-liquid flow in micro/milli channel reactors. Proceedings of theNetherlands Process Technology Meeting (NPS-11), 24-26 October. Arnhem, TheNetherlands.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2011. Amultifunctional device to speed-up fine chemical industry. Proceedings of the STWJaarcongres, Simon Stevin Leerlingprijs, 6 October. Nieuwegein, the Nederlands.

• Al-Rawashdeh, M., Nijhuis, T.A., Rebrov, E.V., Hessel, V., Schouten, J.C., 2009. De-sign of micro/milli reactor for large scale processing. Netherlands Process TechnologySymposium (NPS-9) - Special GSPT corner, (pp. 1-1). Veldhoven, The Netherlands.

About the Author

Ma’moun Al-Rawashdeh was born on 21st May 1981 in Irbid, Jordan. He finished his BScdegree in Chemical Engineering in 2004 at the Jordan University of Science and Technol-ogy, Jordan. For a year and half, he worked as a chemical engineer at the Jordan PetroleumRefinery, Jordan. In 2006, he started his MSc study in Chemical Engineering with special-ization in Process Engineering at the Eindhoven University of Technology, the Netherlands.He conducted his master thesis in the Institut fur Mikrotechnik Mainz GmbH in Germany onthe topic ’Process intensification in a falling film microreactor based on pseudo 3-D simula-tion’ under the supervision of Prof.dr. V. Hessel. In 2008, he obtained his MSc degree (CumLaude) and continued as a PhD student in the Laboratory of Chemical Reactor Engineeringunder the supervision of Prof.dr.ir J.C. Schouten. During his PhD project, he developed theBarrier-based Micro/Milli Channels Reactor (BMMR) presented in this thesis. The BMMR isa structured multiphase flow reactor which reaches large scale production via numbering-uprather than scaling-up.

barrier-based micro/milli channels reactor

Documents