active pharmaceutical ingredient api production involving continuous processes a process system...

7/23/2019 Active Pharmaceutical Ingredient API Production Involving Continuous Processes a Process System Engineering PSE…

http://slidepdf.com/reader/full/active-pharmaceutical-ingredient-api-production-involving-continuous-processes 1/20

Research paper

Active pharmaceutical ingredient (API) production involving continuous

processes – A process systems engineering (PSE)-assisted design framework

Albert E. Cervera-Padrell a, Tommy Skovby b, Søren Kiil a, Rafiqul Gani a, Krist V. Gernaey a,⇑

a Department of Chemical and Biochemical Engineering, Technical University of Denmark (DTU), Building 229, DK-2800 Kgs. Lyngby, Denmarkb Chemical Production Development, H. Lundbeck A/S, Oddenvej 182, DK-4500 Nykoebing Sj., Denmark

a r t i c l e i n f o

Article history:

Received 25 April 2012

Accepted in revised form 3 July 2012

Available online 20 July 2012

Keywords:

Design framework

Methodology

Continuous pharmaceutical manufacturing

Continuous processes

Microfluidic

Active pharmaceutical ingredient

a b s t r a c t

A systematic framework is proposed for the design of continuous pharmaceutical manufacturing pro-

cesses. Specifically, the design framework focuses on organic chemistry based, active pharmaceutical

ingredient (API) synthetic processes, but could potentially be extended to biocatalytic and fermenta-

tion-based products. The method exploits the synergic combination of continuous flow technologies

(e.g., microfluidic techniques) and process systems engineering (PSE) methods and tools for faster process

design and increased process understanding throughout the whole drug product and process develop-

ment cycle. The design framework structures the many different and challenging design problems

(e.g., solvent selection, reactor design, and design of separation and purification operations), driving

the user from the initial drug discovery steps – where process knowledge is very limited – toward the

detailed design and analysis. Examples from the literature of PSE methods and tools applied to pharma-

ceutical process design and novel pharmaceutical production technologies are provided along the text,

assisting in the accumulation and interpretation of process knowledge. Different criteria are suggested

for the selection of batch and continuous processes so that the whole design results in low capital and

operational costs as well as low environmental footprint. The design framework has been applied to

the retrofit of an existing batch-wise process used by H. Lundbeck A/S to produce an API: zuclopenthixol.

Some of its batch operations were successfully converted into continuous mode, obtaining higher yieldsthat allowed a significant simplification of the whole process. The material and environmental footprint

of the process – evaluated through the process mass intensity index, that is, kg of material used per kg of

product – was reduced to half of its initial value, with potential for further reduction. The case-study

includes reaction steps typically used by the pharmaceutical industry featuring different characteristic

reaction times, as well as L–L separation and distillation-based solvent exchange steps, and thus consti-

tutes a good example of how the design framework can be useful to efficiently design novel or already

existing API manufacturing processes taking advantage of continuous processes.

2012 Elsevier B.V. All rights reserved.

1. Introduction

Pharmaceutical companies have traditionally invested their re-search and development efforts in bringing new products to the

market in the shortest time [1]. Limited patent lifetime and strin-

gent regulations practically meant that focus was on drug discov-

ery rather than on optimal process design [2]. The advent of globalization, the growing importance of generic manufacturers,

the ever increasing awareness of environmental impact, and the

encouragement by the regulatory authorities to increase process

understanding and improve quality and efficiency while minimiz-

ing risk have led the pharmaceutical industry to reconsider the

way drug products are manufactured and process development is

approached [3]. This self-reflection has resulted in the promotion

of continuous pharmaceutical manufacturing (CPM) [2] and pro-

cess analytical technology (PAT) [4] in order to significantly

increase the efficiency and sustainability of production processes.

Concomitantly, microfluidic technologies have emerged [5],

demonstrating enhanced mass and heat transfer compared to

0939-6411/$ - see front matter 2012 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.ejpb.2012.07.001

Abbreviations: API, active pharmaceutical ingredient; CFD, Computational Fluid

Dynamics; CPM, continuous pharmaceutical manufacturing; CPME, cyclopenthyl-

methyl-ether; DOE, design of experiments; FDA, Food and Drug Administration, US;

ICH, International Conference on Harmonization; IND, investigational new drug

(application); LCA, life cycle assessment; Me-THF, 2-methyl-tetrahydrofuran; NDA,

new drug application; NIR, near-infrared; OED, optimal experimental design; PAT,

process analytical technology; PMI, process mass intensity; PSE, process systems

engineering; QbD, quality by design.⇑ Corresponding author. Department of Chemical and Biochemical Engineering,

Technical University of Denmark (DTU), Building 229, DK-2800 Kgs. Lyngby,

Denmark. Tel.: +45 45252970; fax: +45 45932906.

E-mail address: [email protected] (K.V. Gernaey).

European Journal of Pharmaceutics and Biopharmaceutics 82 (2012) 437–456

Contents lists available at SciVerse ScienceDirect

European Journal of Pharmaceutics and Biopharmaceutics

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e j p b

http://dx.doi.org/10.1016/j.ejpb.2012.07.001

mailto:[email protected]


http://www.sciencedirect.com/science/journal/09396411

http://www.elsevier.com/locate/ejpb

http://www.elsevier.com/locate/ejpb

http://www.sciencedirect.com/science/journal/09396411


mailto:[email protected]




batch reactors, resulting in higher yield, increased safety, improved

product quality, and shorter development time.

The use of microfluidic techniques and other continuous flow

processes, optimally combined with PAT, can very effectively con-

tribute to improved and faster process understanding [6]. Process

knowledge can optimally be stored as mechanistic models [7]

and used by process systems engineering (PSE) methods and tools

in order to tackle the process design problem. However, few stud-

ies (e.g., [8]) have focused on the effective use of continuous flow

techniques in order to facilitate process design and vice versa,

exploiting the structured and systematic way of solving problems

of PSE in order to assist in the planning and effective design of

experiments. Furthermore, although very significant contributions

have been made for the development of continuous processes [5],

it is still not clear when batch-wise production and/or continuous

manufacturing should be selected. In this contribution, a PSE-as-

sisted design framework for the development of continuous phar-

maceutical manufacturing processes is proposed. The objective of

the design framework is to identify PSE methods and tools that

can assist in the efficient acquisition of process knowledge and

the systematic application to the different process design problems

that arise throughout the drug product and process development

cycle. This manuscript presents the structure of the design frame-

work step by step, using information from the literature about con-

tinuous flow technologies and PSE methods and tools for process

design applied to pharmaceutical process development. Next, the

design framework is applied to a case study: the conversion of a

batch-wise process used for the production of an active pharma-

ceutical ingredient (API) developed by H. Lundbeck A/S (zuc-

lopenthixol). It is shown that the application of the design

framework and the development of continuous flow technologies

resulted in a significant simplification of the manufacturing pro-

cess, implying capital and operational costs savings, and lower pro-

cess environmental footprint.

2. Design framework

2.1. Scope

From medicinal chemists to chemical engineers, a considerable

number of cross-disciplinary professionals are involved in the

development of a pharmaceutical product. Experience has demon-

strated that decisions taken in the early development stages of a

drug (e.g., finding a suitable synthesis route to an API) establish

the core features of the industrial scale process [9]. Hence, it is

hypothesized that the whole pharmaceutical product-process

development cycle (Fig. 1) must be involved – perhaps in different

degrees – in the frame of a strategic evolution toward more sus-

tainable discovery, development, and operation activities [9,10].

The very large attrition rate through the drug development cycle

practically means that generic approaches toward process devel-

opment are required [11]. Due to the very short time frame avail-

able for process development, a cross-disciplinary strategic

approach toward development must be coordinated.

2.2. Aim

The aim of the proposed design framework is to assist in the de-

sign of continuous pharmaceutical manufacturing (CPM) pro-

cesses, that is, processes taking advantage of continuous flow

when relevant. The workflow of design activities should follow

the drug product-process development cycle (Fig. 1). The design

framework should identify already existing process systems engi-

neering (PSE) methods and tools that can assist in each design

problem. While these PSE methods and tools are well established

in other chemical industries, they may require further develop-

ment in the context of pharmaceutical production [11].

2.3. Structure of the design framework

In each step of the proposed design framework (Fig. 2), a model-

centric approach [7] – complemented by relevant PSE methods and

tools – is used to assist experimental studies for the most efficient

acquisition and storage of process knowledge. A cross-disciplinary

strategy is established, where a sustainability mindset [9] and a

commitment to the application and development of novel process

technologies (e.g., process intensification and continuous pro-

cesses) are communicated throughout the workflow. From the de-

sign problem point of view, the proposed design framework is

generic, that is, both ‘‘new process designs’’ and ‘‘retrofit of old pro-

cesses’’ are treated in a similar manner. The main difference is to be

found in the initial step in the methodology – it is assumed that a

retrofit problem (for example, converting an already existingbatch-wise process into a continuous one) is based on a substantial

amount of previously acquired process experience/knowledge.

However, since continuous processing opens the gate to ‘‘novel

process windows’’ [12], acquisition of novel process knowledge

outside conventional process conditions will be a requirement.

The design framework makes use of the knowledge that is initially

available and extends it with novel methods based on continuous

flow experiments when relevant.

2.3.1. Step 1: Drug discovery and development

This step is typically carried out by medicinal chemists and bio-

chemists and should – at least in our opinion – be optimally com-

plemented by a ‘‘chemical engineering perspective’’, introducing

novel technologies and simplified PSE methods and tools. Microflu-idic techniques can be used to perform high throughput experi-

mentation throughout the drug discovery and development

phases, using lower amounts of reagents and achieving faster reac-

tion times [13]. In this way, a flow chemistry mindset can be pro-

moted, where chemistries that are not feasible in batch mode (e.g.,

involving unstable reaction intermediates), or which are advanta-

geous under non-conventional process conditions, may provide

new opportunities for simpler processes [12,14]. Introducing sim-

plified versions of solvent selection methodologies and promoting

sustainability awareness in research laboratories [9] may also

simplify process development activities. Furthermore, the

development of generic methodologies for substrate adoption

[15] – that is, using a generic plant/microfluidic platform to per-

form one type of reaction with different substrates – could contrib-ute to faster and more efficient compound library generation for

Fig. 1. Phases of the drug development cycle [79]. (For interpretation of the

references to color in this figure legend, the reader is referred to the web version of this article.)

438 A.E. Cervera-Padrell et al. / European Journal of Pharmaceutics and Biopharmaceutics 82 (2012) 437–456



Fig. 2. Design framework for the development of API manufacturing processes involving continuous processes – workflow of design activities.

A.E. Cervera-Padrell et al. / European Journal of Pharmaceutics and Biopharmaceutics 82 (2012) 437–456 439



high throughput screening (HTS). The use of real-time, in-line

monitoring techniques [6], leads to information-rich experiments

that could be automatically driven by optimal experimental design

(OED) algorithms [8]. The use of well-structured ontologies should

be encouraged to organize the large amounts of information gener-

ated in process databases (e.g., [16,17]).

2.3.2. Step 2: Preliminary data collection for process designIn this step, all information obtained during the drug discovery

phase and preclinical tests relevant for process design is collected.

This could include API and intermediate product molecular

weights, chemical structures, solubility data, degradation profiles,

boiling and melting points, synthetic route, by-product formation,

reaction enthalpies, potential hazards, etc. If some of these data are

missing, additional data may be obtained by using property predic-

tion tools (e.g., ICAS ProPred [18]) and by performing experiments

as discussed in step 1.

2.3.3. Step 3: Generate/retrieve base-case design: preliminary process

flowsheet

A preliminary process flowsheet must be found before the clin-

ical trials begin (Fig. 2), since significant amounts of the drug prod-

uct (with different formulations) are to be administered to humans

involved in the studies. The drug product will then gradually be

fine-tuned to accomplish the desired bioavailability and pharma-

cokinetic and pharmacodynamic behavior until a final formulation

is found [13]. The preliminary process flowsheet may be quickly

generated by the process development team using simplified ver-

sions of the design methods and tools explained throughout step

4 or simply retrieved from the chemical development team who

found the synthetic route to the API. The preliminary process flow-

sheet (base-case design) may include somewhat redundant,

unnecessary or inefficient washing, isolation, and purification

steps, since at this point, the priority is to ensure that the drug

product complies with the required quality specifications, and pro-

cess knowledge may still be rather limited.

2.3.4. Step 4: Retrofit analysis

In this step, either a preliminary process flowsheet from step 3

or a traditional batch-wise process from an already existing phar-

maceutical product is retrofitted to remove unnecessary separation

or purification steps and optimize solvent use, reaction yields, and

selectivities.

2.3.4.1. Step 4.1 Solvent selection and solvent exchange. Solvents may

account for about 80–90% of the mass utilization in a typical phar-

maceutical process [19]. Solvent selectionmust be considered in an

early phase of the process design, since solvents might strongly

influence reactivity [20,21], and work-up operations obviously de-

pend on the solvent(s) choice [22]. Regardless of the selection

strategy, for example, [18,20,23–25], two main objectives shouldbe sought. Firstly, the solvent or combination of solvents should

perform its/their intended function at the lowest economic and

environmental costs while avoiding health and safety issues. Sec-

ondly, solvent exchange [26,27] between different operations

should be enabled and if possible minimized, pursuing similar

aims in terms of costs and safety. Alternatively, solvent mixtures

leading to improved performance [25], and potential reaction yield

improvements obtained through telescoping reactions can be

investigated [14,28]. A holistic selection approach thus becomes

necessary. Fig. 3 shows the workflow followed for solvent selection

in this design framework, together with the required methods and

tools and the data-flow (input/output).

2.3.4.2. Step 4.2 Reaction analysis and reactor design. Guidelines areneeded to decide whether a certain reaction should be operated in

batch or in continuous mode, or if the reaction has the potential to

be intensified to a point where it can be performed in continuous

mode. Subsequently, reactions can be optimized according to dif-

ferent criteria. Process knowledge must be progressively added in

order to solve these design questions (Fig. 4).

Steps 4.2.1–4.2.3 in the design framework are used to charac-

terize the different reactions. Heterogeneity – and in particular, so-

lid handling – is a major challenge that hinders widespread use of

microreactor technology [29,30]. However, other continuous flow

applications have been developed to deal with solids and need to

be identified accordingly [31]. Calorimetric studies, preliminary ki-

netic analysis, and safety assessment should be performed [32] –

preferably using in situ monitoring techniques [6,33] – in order

to get familiar with the reaction and identify potential challenges.

An exploratory design of experiments (DOE) may be particularly

useful to find out (at least qualitatively) the influence of process

variables on the reaction yield, product selectivity and kinetics

[32], while optimal experimental design (OED) [34] can be used

to minimize the number of experiments to be performed.

In Step 4.2.4, the reaction is classified in terms of kinetics and

mass and heat transfer demands, for example, according to types

A–C as proposed by Roberge et al. [29]. In this way, a suitable reac-

tor design (e.g., microstructured vs. mesoflow) may be suggested

[29,35]. Alternatively, the classification criteria proposed by Hart-

man et al. [36] in the context of laboratory-scale process develop-

ment could be extended to large-scale processes. In this case, non-

dimensional numbers are used to identify the relative importance

of kinetics to mixing, heat exchange, and axial dispersion.

Step 4.2.5 is used to find a ‘‘business case’’ based on economic

and/or environmental cost factors (e.g., sustainability indicators)

and thereby compare batch and continuous flow strategies. Note

that the cost drivers (e.g., quality, speed to market, throughput,

operational and capital costs) may change along the drug develop-

ment cycle [37] and may thus be difficult to identify. Regarding

sustainability, the process mass intensity – defined as total mass

of materials used to produce a specified mass of product – can be

used as the key metric for fast evaluation and benchmarking of whole processes or single units [38]. For a detailed analysis, other

methodologies and software tools could be used, e.g., SustainPro

[39].

Finally, in step 4.2.6, the information obtained throughout steps

4.2.1 to 4.2.4, combined with the objectives arising from step 4.2.5,

is used to select a batch or a continuous flow configuration. This

subject is too broad to be covered in detail in this text. However,

this step consumes a large amount of the process development

time, and in many scenarios, highly customized equipment is re-

quired. Hence, considering the multiproduct/multipurpose nature

of pharmaceutical plants, generic approaches toward substrate

adoption are highly desired [15,35,40]. Finding the optimal reactor

network has been the subject of intense study [41–43], while system-

atically identifying opportunities for process intensification – forexample, combining reaction and separation phenomena – could

be considered at this point [44,45]. To find a solution to the reactor

configuration and/or network problem, solving an optimization

problem may be needed (step 4.2.7), either rigorously (mathemat-

ically) or approximately (using heuristics and expertise). The prac-

tical application of the methodology needs to take into account

time and equipment or space constraints, which will depend on

the infrastructures and resources available. Ultimately, an eco-

nomic analysis will evaluate the feasibility of the project (step 7).

An approximate economic evaluation of the reactor proposed could

however be advanced at this point (see Fig. 4).

2.3.4.3. Step 4.3 Separation and purification operations. While al-

ready in step 4.1 solvents and solvent exchange operations wereproposed, separation and purification operations (Fig. 5) should




be designed after assessment of the reaction products obtained. In

steps 4.3.1–4.3.5, the intermediate products obtained after each

reaction step are analyzed, and the washing, isolation, and solvent

exchange operations proposed in steps 3 and 4.1 are tested exper-

imentally in order to understand the distribution of the com-

pounds of interest in the resulting streams. Subsequently,

degradation profiles and cross-interactions between API interme-

diates and impurities are studied, in order to find incompatible

compounds/reactions. Very unstable compounds will require

immediate downstream processing upon formation, which can

either be accomplished by cascade reactions [12,14,28] or by con-

tinuous downstream processing [5]. If a full (quantitative) chemi-

cal and physical characterization of by- and side-products is

available, thermodynamic models could be employed to predict

the distribution of API intermediates and impurities in different

streams. An attempt is made to streamline the process – eliminat-

ing unnecessary separation and purification steps – through en-

hanced reaction selectivity obtained in flow mode [37,46].

Designing the separation and purification operations (step

4.3.6) completes the basic features of the simplified process flow-

sheet. As with reactor design, this step involves much of the design

efforts. The optimal flowsheet generation for the separation of

multi-component mixtures has been a subject of wide interest

for the PSE community [47–51]. Computer-aided methods and

tools and simulation software based on thermodynamic models

(with predictive ability) and experimental data are essential for

efficient design and sequencing of separation operations (e.g.,

[25]).

PSE methods and tools for flowsheet synthesis and design are

therefore well established. However, in order to sequence the sep-

aration and purification operations and establish the basic operat-

ing conditions, the first decision to make is whether the

separation/purification operation(s) should be performed batch-

wise or continuously. This is a more difficult question to answer

compared to reactor design, since batch stirred vessels have the

advantage of being multipurpose (reaction + separation), and re-

search on small-scale (micro/meso-scale) continuous separation

units [5,27,52–54] lags to some degree behind the development

of micro- and microstructured reactors. Hence, in our opinion,

deciding whether a separation operation should be carried out in

batch mode or in continuous mode is largely dependent on the

operating mode of the reactions immediately before and immedi-

ately after the separation operation, which requires a holistic pro-

cess design approach. Flow technology specific advantages are

subject of current research and development.

2.3.4.4. Step 4.4 Process flowsheet simulation and/or experimentalvalidation. A process flowsheet should at this point be available,

whose actual feasibility must be verified. Using a model-centric

design approach [7] throughout process development will allow

Fig. 3. Workflow, methods and tools, and data-flow corresponding to step 4.1 of the design framework: solvent selection and solvent exchange.




Fig. 4. Workflow, methods and tools, and data-flow corresponding to step 4.2 of the design framework: reactor design.




Fig. 5. Workflow, methods and tools, and data-flow corresponding to step 4.3 of the design framework: design of separation and purification operations.




deeper process understanding and will probably yield a larger de-

sign space [55]. However, in the last instance, an experimental val-

idation of the process will be required.

2.3.4.5. Step 4.5 Scale-up/out. If the simplified process flowsheet is

successfully validated, the next step is to demonstrate its scalabil-

ity, that is, propose the design of operation units able to meet the

desired industrial-scale throughput while being flexible to respond

to large variations of the demand. The pharmaceutical industry has

traditionally handled this uncertainty by using large multi-purpose

batch units and dividing the yearly production in a number of cam-

paigns as required. In contrast, for continuous operating units, a

modular concept (toolbox concept, [37]) has been proposed, where

parallel (small) units could be replicated as needed (scaling-out or

numbering-up approach) [2,29,56]. However, fluid distribution and

parallelization may be complex [57] and prohibitively costly.

According to Kockmann et al. [40], the numbering-up approach

should actually be kept as a last option. Furthermore, solids han-

dling is still a major limitation for microreaction technology

[29,30]. Hence, a scale-up approach may still be the most econom-

ical method to process large flow rates in challenging scenarios

while maintaining some of the advantages of operating at small

length scales (e.g., using meso-units). In order to accomplish these

objectives, Barthe et al. [58] and Kockmann et al. [40] have shown

that scaling-up should be driven by a thorough description of mass

and heat transfer characteristics. Computational Fluid Dynamics

(CFD) [59] could also be useful to investigate hownon-ideal mixing

influences reaction performance at larger scales [11].

2.3.5. Step 5: Process evaluation and intensification, integration, and

optimization

The process can be evaluated in order to find potential further

improvements, which should be communicated to the upper levels

of the design framework when relevant (Fig. 2). Software-assisted

methodologies have been proposed in order to find process

streams where significant improvements could be achieved in

terms of mass or energy use, for example, SustainPro [39]. One op-tion could be finding opportunities for further process intensifica-

tion [44,45] – for example, combining reaction and separation.

Another opportunity arising from the more extended use of contin-

uous operating units is to introduce solvent integration [46]. En-

ergy consumption has also been generally disregarded by the

pharmaceutical industry as a potential gate to cost and environ-

mental impact reduction [60]. Once all structural decisions form-

ing the process flowsheet have been taken, all process variables

that have not been fixed may be subject to a reduced optimization

problem. The objective could be single or multiple (cost, environ-

mental impact, energy use, etc.).

2.3.6. Step 6: Process monitoring and control tools

With the introduction of PAT and QbD in the pharmaceuticalindustry, the design and validation of processes in a range of pro-

cessing conditions – known as the design space – have been encour-

aged [4,55]. Implementing QbD practically means that the process

can be regulated to respond to disturbances or to meet a certain

optimization criterion, without the need to re-validate the whole

process. However, in order to do so, one must demonstrate suffi-

cient process knowledge, which can for example be stored using

a model-centric approach [7]. The use of data-driven models

(e.g., partial least squares regression) has also been proposed to

respond to, for example, variability in the raw materials via feed-

forward control [61]. Singh et al. [62] created a systematic

computer-aided framework to develop a software (ICAS-PAT) for

design, validation, and analysis of PAT systems. Operability and

flexibility considerations may also be integrated into the processsynthesis procedures themselves [63,64].

2.3.7. Step 7: Process assessment

The final step in the design framework is the economical and/or

environmental assessment of the final process. The environmental

impact related to the process can be evaluated using different

guidelines and software tools as reviewed by Linninger and Mal-

colm [65]. The use of a full LCA should also be encouraged to eval-

uate the origin and nature of the impacts and understand how they

can be reduced [38,60]. One example of an economical analysis of a

pharmaceutical production process including continuous opera-

tions has been given by Schaber et al. [66].

2.3.8. Step 8: Implementation

A verified simplified process including continuous operating

units can be implemented in this step. This step must be reached

at the same time as the regulatory agency approves manufacturing

of the pharmaceutical product. The implementation of continuous

operating units will require training of operators, where a transi-

tion from manual operation to supervision of automated process

regulation through PAT will be a key innovation factor.

3. Results and discussion

One possible application of the proposed design framework is

illustrated through a case study, where a batch-wise process used

by H. Lundbeck A/S to produce the API zuclopenthixol was retrofit-

ted by exploring the advantages of continuous processing. The pro-

duction of zuclopenthixol is a multistep process including

reactions in different solvents and with different reaction rates

and thus constitutes a good example of a typical organic-chemistry

based API production process. Since the original process was al-

ready known and previous experience was available, the design

procedure started from step 4 of the design framework (retrofit

analysis), taking as a base-case design (step 3) the known batch-

wise process. The text below describes the step-by-step applica-

tion of the design framework and summarizes the achievements

obtained throughout the process.

3.1. Case study – synthetic route – steps 1 and 2 of the design

framework

Zuclopenthixol (traded by H. Lundbeck A/S as clopixol) can be

isolated from a mixture of cis- and trans-clopenthixol (4-[3-(2-

Chlorothioxanthen-9-ylidene)propyl]-1-piperazineethanol), where

the trans-isomer is recycled and isomerized to a cis–trans mixture.

This case study focuses on the synthesis of cis- and trans-clope-

nthixol (compound 7) through 4 reaction steps (Scheme 1), result-

ing in an almost equimolar mixture of the two isomers. The first

reaction step is a Grignard alkylation, where allylmagnesiumchlo-

ride (AllylMgCl, compound 2) reacts with chlorothioxanthone

(CTX, compound 1) to produce an alkoxide product (compound3). In the second step, the alkoxide reacts with acidic water to pro-

duce 9-Allyl-2-Chlorothioxanthen-9-Ol (short name ‘‘allylcarbi-

nol’’, compound 4) and magnesium salts. The next step is a

dehydration of the ‘‘allylcarbinol’’ to 9H-Thioxanthene, 2-chloro-

9-(2-propenylidene)-(9CI) (short name ‘‘butadiene’’, compound

5). The last step is a hydroamination of the ‘‘butadiene’’ with 1-

(2-Hydroxyethyl)piperazine (short name ‘‘HEP’’, compound 6) to

produce clopenthixol (7).

3.2. Description of the base-case design (batch-wise process) – step 3

of the design framework

The original batch-wise process flowsheet is divided into two

stages (Fig. 6). The first stage involves the synthesis and isolationof the ‘‘allylcarbinol’’ intermediate. After alkylation of CTX and




hydrolysis of the alkoxide product, the organic phase (containing a

mixture of THF, water, and ‘‘allylcarbinol’’) is separated from the

aqueous phase (containing a mixture of water, THF, Mg salts, and

polar impurities). Since a small amount of unknown impurities is

found after the batch alkylation, ‘‘allylcarbinol’’ is isolated by crys-

tallization before continuing the synthesis. Therefore, a solvent ex-

change step from THF to the antisolvents ethanol and water is

performed. The ‘‘allylcarbinol’’ crystals are separated by filtrationand subsequently dried and collected for storage.

In the second stage (Fig. 6), ‘‘allylcarbinol’’ is dissolved in tolu-

ene and dehydrated to the ‘‘butadiene’’ intermediate. Acetic acid

anhydride is used as dehydrating agent, while acetyl chloride is

used as acid catalyst. The final reaction step (hydroamination) is

a slow reaction (24 h batch). An excess of HEP is desired in order

to increase the reaction rate. Because of its solubility and reactivity

properties, HEP can act both as solvent and as reactant. In a solvent

exchange step, toluene is distilled off and HEP is gradually added.

When the solvent exchange is complete, the reaction proceeds un-

til all ‘‘butadiene’’ is converted. This step ends the process as stud-

ied in this case study. However, a final extraction of clopenthixol

with toluene and separation of HEP with water is performed after

the hydroamination reaction. It is so far assumed that these steps

will be carried out as in the original batch-wise process (cf. Tables

6 and 9).

3.3. Retrofit analysis – step 4 of the design framework

In this step, the original batch-wise process was streamlined as

much as possible, converting its unit operations to continuous

mode when favorable. The following text describes how the pro-

posed design framework was used to systematically resolve the

different design problems.

3.3.1. Solvent selection and exchange – step 4.1

The solvent selection procedure has been summarized in Ta-

ble 1, where for both the original batch-wise process and the ret-

rofit process, and for each synthetic or isolation step (the

crystallization of ‘‘allylcarbinol’’), plausible solvent candidateshave been marked with a cross. The different process options have

been indicated with arrows, where a dashed line represents a nec-

essary solvent exchange operation. In the base-case design, four

solvents are employed: THF is used in the alkylation reaction, eth-

anol and water are used for the controlled crystallization of ‘‘allyl-

carbinol’’, and toluene is used for the dehydration of ‘‘allylcarbinol’’

to ‘‘butadiene’’. The hydroamination is ‘‘solvent-free’’ after evapo-

ration of toluene, where the ‘‘butadiene’’ is mixed (and in fact, dis-

solved) in an excess of HEP.

For the retrofit design, different options were found. The selec-

tion procedure (a simplified version of the methodology by Gani

et al. [18]) began by identifying solvent candidates for each step.

The alkylation reaction requires the use of an ethereal solvent

[67]. Alternatives to THF have been claimed to be greener, forexample, 2-methyl-THF (Me-THF) [68] and cyclopenthyl-methyl-

ether (CPME) [69]. Some of their advantages compared to THF in-

clude low miscibility with water, enabling two-phase separation

in aqueous quench/extraction and thus facilitating wastewater

treatment [9], and a higher boiling point. The hydrolysis reaction

is a work-up step after the alkylation and only requires the use

of acidic water to solubilize Mg salts. The dehydration of ‘‘allyl-

carbinol’’ allows a large spectrum of solvents to be used, with basic

Fig. 6. Flowsheet representing the original batch-wise process used by H. Lundbeck A/S to produce clopenthixol.

Scheme 1. Synthetic route followed to produce clopenthixol from CTX.




requirements including a high boiling point in order to perform the

reaction at high temperature (increasing the reaction rate as dis-

cussed below) and a low miscibility with water in case work-up

washing steps are needed. Traditionally, toluene has been the sol-

vent of choice.

It was hypothesized that by changing the alkylation reaction to

continuous mode, the yield could be improved and a lower amount

of impurities would be formed [31,70]. This would make the crys-tallization step (used to isolate ‘‘allylcarbinol’’ crystals) unneces-

sary, as well as the solvent exchange step to ethanol/water.

Consequently, stages 1 and 2 in the original process could be inte-

grated in one stage. This could be achieved in two ways (see

Table 1):

Option 1: Perform the dehydration reaction in an ethereal sol-

vent (THF, Me-THF, or CPME), thereby streamlining the process.

Only a solvent evaporation prior to the hydroamination reaction

would in this case be needed. However, it would be necessary to

demonstrate the dehydration reaction in the ethereal solvent of

choice. Me-THF and CPME have the advantage of higher boiling

points (potentially enabling faster high temperature dehydra-

tions) and low miscibility with water (facilitating L–L separa-tion and washing steps).

Option 2: Perform the dehydration reaction in toluene as in the

traditional batch-wise process. A solvent exchange step from

the ethereal solvent chosen for the alkylation reaction would

in this case be needed, as well as toluene evaporation prior to

the hydroamination reaction.

Option 1 is clearly the simplest solution, and thus, it was chosen

as the priority. THF was kept as the preferred ethereal solvent due

to its availability and price compared to Me-THF and CPME, under

the condition that the dehydration reaction could be performed in

THF with high conversions in short time. A simplified process flow-

sheet was thereby proposed (Fig. 7). Interestingly, the same acid

used for Mg salt solubilization in the hydrolysis (HCl) could be

used for ‘‘allylcarbinol’’ dehydration after L–L separation, avoiding

the use of acetic acid anhydride and acetyl chloride.

3.3.2. Reaction analysis and reactor design – step 4.2

In this step, the feasibility of the simplified process flowsheet

shown in Fig. 7 was experimentally verified. The reactor analysis

and design procedure was repeated for each reaction step, that

is, alkylation, hydrolysis, dehydration, and hydroamination. For

illustration purposes, Tables 2–5 contain a summary of the infor-

mation collected and the design decisions taken throughout the

application of the design framework to the four reactions. A brief description of the reactor designs proposed is provided in the fol-

lowing subsections.

3.3.2.1. Alkylation (Grignard reaction). The alkylation of CTX is a fast

and exothermic Grignard reaction (reaction type A), most suitable

for microstructured flow reactors as long as homogeneous condi-

tions can be guaranteed [29,35,58,71]. In this particular case, CTX

has a low solubility in THF, while AllylMgCl and the alkoxide prod-

uct have high solubility. The suitability of a continuous filter reac-

tor followed by a tubular reactor with multiple injections has been

demonstrated elsewhere [31,70]. In short, the filter reactor pro-

vides solvent savings, while the multiple injection tubular reactor

provides accurate titration of CTX excess with low impurity forma-

tion. The reactor system can be monitored and potentially con-trolled by real-time in-line NIR spectroscopy measurements [70].

3.3.2.2. Hydrolysis of alkoxide product. According to the process

flowsheet in Fig. 7, HCl can be used to convert MgCl(OH) into more

soluble MgCl2. The reaction is very fast and exothermic (type A).

Four phases may be present in the reaction, a solid phase (poorly

soluble Mg salts and potentially CTX precipitated), two liquid

phases (organic and aqueous phase), and a gas phase, in case the

alkoxide product contains excess Grignard reagent (forming pro-

pene gas when reacting with water [67]). While the alkylation

reactor is operated and monitored such that a small amount of

CTX is always present in the product (the monitoring strategy used

to ensure this condition is described by Cervera-Padrell et al. [70]),

the hydrolysis reactor must be designed to handle the eventualityof propene formation (safe venting must be ensured) as well as the

high exothermicity of the Grignard reagent quench reaction. Right

Table 1

Solvent selection and solvent exchange strategies followed in the traditional batch-wise process and the simplified process containing continuous operation units. Plausible

solvent candidates have been marked with a cross. The different process options have been indicated with arrows, where a dashed line represents a necessary solvent exchange

operation. The followed selection procedure is a simplified version of the methodology developed by Gani et al. [18].

Alkylation Hydrolysis Crystallization Dehydration Hydroamination

Batch

THF x x

Ethanol/Water x

Toluene x

HEP x

Retrofit

Ethereal solv. x x ?

Ethanol/Water

Toluene x

HEP x

Fig. 7. Flowsheet representing the simplified process proposed for the production of clopenthixol using continuous flow operations.




after the hydrolysis reaction, liquid–liquid separation should be

performed, since only the organic phase is used further on in the

process. Therefore, the reactor design may be integrated with the

subsequent phase separation. One option could be using a hydro-

cyclone [72] to perform mixing, reaction and separation in one de-

vice, properly adjusted to handle the solid and gas phases. A

different solution has been experimentally tested, using a small-

scale PTFE tubular reactor with segmented flow followed by a PTFE

membrane separator [73]. It is expected that the formation of li-

quid slugs with internal mixing [74] helps to avoid solid precipita-

tion on the reactor walls.

3.3.2.3. Dehydration of ‘‘allylcarbinol’’. The organic phase obtained

after L–L phase separation (Fig. 7) contains THF, water, and HCl

acid. Since this acid works as a catalyst for the dehydration of

‘‘allylcarbinol’’, the solution is ready for the reaction step, provided

that it is verified that the reaction is irreversible and thus water

does not affect the reaction equilibrium. A kinetic model was

developed based on batch experiments at normal pressure,

describing the effect of temperature and concluding that the reac-

tion is indeed irreversible. Due to the low boiling point of THF, the

maximum temperature achieved was 67.5 C, requiring more than

40 min for full conversion [75]. It was therefore decided to increase

the pressure so that the temperature could be increased above the

normal boiling point of the solution. At 120 C and 5 bar, it was

predicted that 99% conversion could be obtained in about 2 min,thereby simplifying the continuous flow reactor design. This was

confirmed experimentally using a small-scale tubular reactor,

where water was added to saturation as would occur after the

hydrolysis and L–L separation. Further discussion on the model

building procedure and experimental results can be found else-

where [75].

3.3.2.4. Hydroamination of ‘‘butadiene’’ with HEP. The hydroamina-

tion of ‘‘butadiene’’ with HEP is carried out in batch mode at

90 C for 24 h, obtaining a yield of around 70%. Unfortunately,

the acceleration of this reaction is not as obvious as in the dehydra-

tion, since at high temperatures, the conversion may be thermody-

namically limited [76]. The ‘‘butadiene’’ conversion rate could beincreased with temperature to a point where the time for almost

full conversion could be reduced to a few hours, while decreasing

the selectivity. Since the most important cost driver in this reaction

is yield and not throughput, reaction rate improvements implying

loss in yield cannot be justified. Hence, a different approach toward

reaction rate and selectivity improvements was explored using

catalysis.

Anti-markovnikov hydroamination of non-activated olefins has

been listed as one of the so-called ‘‘ten challenges for catalysis’’

[76]. Studies have shown that palladium-based catalysts could be

used for the hydroamination of cyclohexadiene with a broad

spectrum of amines [77,78]. Unfortunately, in a fast screening

experiment, it was found that tetrakis-triphenylphosphine-

palladium(0) (Pd(PPh3)4) (promising according to Löber et al.[78]) did not provide any significant improvements of the rate of

Table 2

Information collected and design decisions taken throughout the application of the design framework to the alkylation reaction (step 4.2 of the framework).

Alkylation (Grignard reaction) – THF solvent

Step 4.2 – reactor design Information collected Output/decision taken

Step 4.2.1 Phases – reactant solubility CTX sparingly soluble in THF

Qualitative analysis Phases – reactant solubility Grignard reagent soluble in THF

Phases – product solubility Alkoxide soluble in THF

Reaction generic type Grignard reaction, alkylation on ketone

Reactivity issues Grignard reagent reactive toward air/humidity

Heterogeneity Heterogeneous/homogeneous (solvent use?)

Step 4.2.2 Kinetics Very fast reaction

Calorimetric studies and short kinetic analysis Reaction enthalpy Exothermic (167 kJ/mole)

Step 4.2.3 Stoichiometry Excess of Grignard reagent leads to side-products

Exploratory design of experiments Concentration High Grignard reagent concentration leads to impurity formation

Temperature Temperature below 35–40 C does not have significant effect

on impurity formation

Kinetics/mixing Very fast kinetics, mixing controlled

Solubility CTX solubility curve vs. temperature obtained

Dissolution rate CTX dissolution rate found

Step 4.2.4 Reaction class Reaction class A (very fast, exothermic)

Reaction classification and identification of limitations Limitations Mass and heat transfer limited

Step 4.2.5 Cost driver 1 Yield (low side-product formation)

Determination of cost drivers and sustainability indicators Cost driver 2 Solvent consumption

Cost driver 3 Man-power and safetySustainability driver Solvent consumption

Batch or continuous? Decision Small-scale continuous reactor with improved mixing and heat

transfer, improved safety, more automation

Step 4.2.6 Cost driver 1 – yield Multiple injection microreactor (side-entries)

Reactor design Cost driver 2 – solvent use Filter reactor

Cost driver 3 – operational Automation, in-line monitoring

Step 4.2.7 Problem formulation Minimize solvent use and impurity formation

Solve optimization problem Solution method Heuristics and experimentation

Constraints Operational constraints dominate (precipitation)

Economic evaluation of the project Capital and operational costs Economic advantage of removing large reactor and

substitute by small-scale reactor

Improved yield expected

Lower solvent consumption

Work-up greatly facilitated

Increased automation and safety

Less manual operations (e.g., solid feeding)




hydroamination of ‘‘butadiene’’ with HEP, with or without addition

of a co-catalyst (trifluoroacetic acid). Future work should thus

investigate whether it is possible to increase the reaction rate

and/or yield of the hydroamination reaction, where the latter is

the priority both from an economical and environmental point of

view. With the current limited knowledge about this reaction, it

is most advisable to perform this reaction in batch mode.

3.3.3. Separation and purification design – step 4.3

Once the four reaction steps were characterized, it was investi-

gated whether it was indeed possible to simplify the process flow-

sheet as in Fig. 7. If the simplified process was infeasible, other

process options could be considered as discussed in Section 3.3.1.

3.3.3.1. Analysis of intermediate products and by-/side-product

formation – step 4.3.1. The following information was obtained

from the study of the four reaction steps:

The continuous alkylation produced very low amount of side-

products under the conditions described by Müller Christensen

et al. [31] and Cervera-Padrell et al. [70].

The continuous hydrolysis did not produce any side-products.

However, if an excess of Grignard reagent is present after the

alkylation, propene gas will be produced. CTX is unaffected by

the hydrolysis.

The continuous dehydration in THF mainly produces ‘‘butadi-

ene’’, but it may also produce an unknown impurity in very

low concentration. The product, however, was considered to

be of very high quality [75].

The hydroamination has not been studied with enough detail in

this work. However, preliminary studies show that the reaction

yield is typically between 50% and 70% with almost total con-

version of ‘‘butadiene’’. It is speculated that a fraction of the

‘‘butadiene’’ would polymerize.

3.3.3.2. Evaluation of base-case design and solvent exchange opera-

tions – step 4.3.2. The original batch-wise process used to produce

clopenthixol (base-case design) includes many solvent exchange

operations, washing steps, and separation steps. Table 6 lists all

the tasks and subtasks involved in the traditional process. For

every subtask, the table indicates the amount of feed and waste

materials, that is, materials entering and leaving the process. The

materials that are kept in the process (going from one subtask to

the next one) are either transformed in a reaction or they remain

in solution. These are not indicated in the table, since the purpose

here is to evaluate the process footprint in the following steps of

the design framework. The function of each washing/separation/

solvent-exchange task in the original process has been analyzed

and its relevance with respect to a simplified process flowsheet

(e.g., Fig. 7) questioned. This information has been summarized

in Table 7.

3.3.3.3. Assessment of intermediate product degradation profiles and

cross-interactions – step 4.3.3. Potential degradation and cross-

interaction issues have been summarized in Table 8 taking as a ref-erence the base-case design (batch-wise process). Some of these

Table 3

Information collected and design decisions taken throughout the application of the design framework to the hydrolysis of the alkoxide intermediate (step 4.2 of the framework).

Hydrolysis of alkoxide product – THF solvent


Step 4.2.1 Phases – reactant solubility Alkoxide soluble in THF

Qualitative analysis Phases – product solubility ‘ ‘Allylcarbinol’’ soluble in THF (organic phase)

Phases – product solubility Mg salts soluble in acidic water (aqueous phase)

Reaction generic type Hydrolysis of alkoxide from Grignard reaction

Safety issues Excess Grignard reagent produces propene gas

Heterogeneity Two liquid phases. Partial solid heterogeneity can be expected: potential

CTX excess and insoluble Mg(OH)Cl. Potential propene formation

Step 4.2.2 Kinetics Very fast reaction

Calorimetric studies and short kinetic analysis Reaction enthalpy Exothermic

Step 4.2.3 Stoichiometry Use enough water to solubilize MgCl2 salts

Exploratory design of experiments Use enough acid to convert Mg(OH)Cl to MgCl2

Concentration No concentration issues

Temperature Temperature only affects solubility

Kinetics/mixing Very fast kinetics, mixing controlled

Solubility High solubility of ‘ ‘allylcarbinol’’ and MgCl2 salts

Dissolution rate Not measured – could be important to measure time

for Mg(OH)Cl solubilization in form of MgCl2; however, expected fast

Step 4.2.4 Reaction class Reaction class A (very fast, exothermic)

Reaction classification and identification of limitations Limitations Mass and heat transferred limited

Step 4.2.5 Cost driver 1 Capital cost (small-scale reactor)Determination of cost drivers and sustainability

indicators

Cost driver 2 Man-power

Batch or continuous? Decision Small-scale continuous reactor with improved mixing and heat

transfer, improved safety, more automation. Should handle solids

Step 4.2.6 Cost driver 1 – capital cost Simple tubular reactor

Reactor design Cost driver 2 – operational Automation, in-line monitoring

Heterogeneity handling Segmented-flow tubular reactor or hydrocyclone

(4 phases, i.e., S–L–L–G) Integrate gas venting device

Step 4.2.7 Problem formulation Minimize capital cost, water use

Solve optimization problem Solution method Heuristics and experimentation

Constraints Operational constraints dominate (precipitation)

Economic evaluation of the project Capital and operational

costs

Economic advantage of removing large reactor and substitute by small-scale

reactor

Hydrocyclone integrates reaction + separation





issues may not be relevant in a simplified process flowsheet (e.g.,

Fig. 7).

3.3.3.4. Setting constraints for composition of intermediate streams –

step 4.3.4. Based on the information collected in the previous step,

the general recommendation is to keep solutions under nitrogen

cover and avoid extreme temperatures. No important issues were

identified, except the potential degradation of ‘‘butadiene’’. If pos-

sible, the ‘‘butadiene’’ solution in THF should be stored for a short

period (the degradation profile still needs to be quantified) before

the hydroamination, at low temperature, and with nitrogen cover.

There could be a potential interaction between the acid used in the

dehydration step and the hydroamination reaction. However, it isnot known whether this interaction would be positive or negative,

since acid co-catalysts are typically used in combination with cat-

alysts of the hydroamination reaction [76–78].

3.3.3.5. Elimination of unnecessary separation/purification steps – step

4.3.5. Since the quality of the ‘‘allylcarbinol’’ product obtained in

the continuous alkylation reactor is higher than in batch mode, it

is expected that the isolation of ‘‘allylcarbinol’’ by crystallization

is not needed. This means that subtask 2.3 and tasks 4, 5, 6, and

7 in Table 6 are not needed. Furthermore, since the dehydration

of ‘‘allylcarbinol’’ with HCl/water in THF proved successful, the

use of toluene, acetic acid anhydride, and acetyl chloride in the

dehydration step can be avoided (Fig. 7). Finally, if it is experimen-

tally demonstrated that the acid catalyst used in the dehydrationdoes not have a negative effect on the hydroamination reaction,

subtasks 8.1–8.17 (Table 6) would not be needed. These changes

greatly simplify the original process and result in a considerably

reduced list of tasks as shown in Table 9.

3.3.3.6. Design of separation and purification operations – step

4.3.6. According to Table 9, two types of separation operations

are present in this case-study: L–L separation of an organic and

an aqueous phase, and solvent exchange by distillation. Due to

the relatively simple structure of the process and the lack of rigor-

ous knowledge of the thermodynamics of these separation steps

(this process is only handled by H. Lundbeck A/S), a knowledge-

based design approach based on basic physical insights was fol-

lowed, while much of the development time was invested indesigning, constructing, and experimentally validating a surface-

tension based continuous L–L membrane separator [73].

L–L phase separation – task S3 (Table 9). The separation of THF

containing ‘‘allylcarbinol’’ and water containing magnesium

salts is challenging due to the partial miscibility of THF and

water even with dissolved API and salts (salting-out effect).

The densities of the two phases differ by only about 25 kg/m3,

meaning that separation by decantation is a very slow process,

where coalescence of droplets is a great challenge. Since the

alkylation and hydrolysis reactions can be performed in small

continuous reactors, it is more convenient to perform the L–L

phase separation continuously as well. A surface-tension based

separation method using a hydrophobic membrane has beenproposed [73].

Table 4

Information collected and design decisions taken throughout the application of the design framework to the dehydration of ‘‘allylcarbinol’’ (step 4.2 of the framework).

Dehydration of ‘‘allylcarbinol’’ – THF solvent


Step 4.2.1 Phases – reactant solubility ‘‘Allylcarbinol’’ soluble in THF

Qualitative analysis Phases – product solubility ‘‘Butadiene’’ soluble in THF

Phases – product solubility Water partially soluble in THF with API interm.

Reaction generic type Dehydration of a tertiary alcohol

Safety issues Potential ‘‘butadiene’’ polymerization or oxidation (many double bonds)?

Heterogeneity Two liquid phases when water is formed

Step 4.2.2 Kinetics Slow reaction at normal pressure (>45 min)

Calorimetric studies and short kinetic analysis Reaction enthalpy Slightly endothermic

Step 4.2.3 Stoichiometry Not relevant (only one reactant)

Exploratory design of experiments Concentration ‘‘Allylcarbinol’’ and acid concentration affect reaction rate (kinetic model found)

Temperature Temperature increases reaction rate (kinetic model found)

Kinetics/mixing Slow kinetics

Solubility No issues found

Dissolution rate Not relevant

Step 4.2.4 Reaction class Reaction class B (slow reaction, potentially faster by increasing pressure and

temperature)Reaction classification and

identification of limitations Limitations Kinetic-limited

Step 4.2.5 Cost driver 1 Capital cost (small-scale reactor)

Determination of cost drivers and sustainabilityindicators

Cost driver 2 Man-powerCost driver 3 and

environmental impact

No need to use acetyl chloride and acetic acid

anhydride

Batch or continuous? Decision Small-scale pressurized high-temperature

continuous reactor, improved safety, more automation

Step 4.2.6 Cost driver 1 – capital cost Small-scale pressurized tubular reactor

Reactor design Cost driver 2 – operational Automation, in-line monitoring

Cost driver 3 – catalyst Use of HCl catalyst from hydrolysis and phase separation

Step 4.2.7 Problem formulation Minimize reaction time (and reactor volume)

Solve optimization problem Solution method Kinetic modeling, experimentation, heuristics

Constraints Potential impurity formation at high temperature

Economic evaluation of the project Capital and operational costs Economic advantage of removing large reactor and substitute by small-scale

reactor

Avoid use of acetyl chloride and acetic acid anhydride, use inexpensive HCl





Distillation-based solvent exchange to HEP – task S5 (Table 9).

This operation is coupled with the hydroamination reaction.

So far, limited knowledge has been accumulated about the

hydroamination reaction and the solvent exchange to HEP.

Experimental results show that the reaction is very slow and

is probably best combined to the solvent exchange step by

batch distillation. It has also been considered to evaporate a

fraction of the solvent in the ‘‘butadiene’’ solution after releas-ing the pressure of the dehydration reaction.

L–L phase separation – tasks S6.3, S6.5 and S6.7 (Table 9). This

part of the process has not been experimentally tested yet. Sol-

vent selection has not been considered either. Therefore, the

process has been assumed as in the base-case design. If needed,

the operations could be performed in continuous mode as with

the previously described THF-water separation.

3.3.4. Process simulation and/or experimental validation – step 4.4

The simplified process flowsheet (Fig. 7) has been experimen-

tally validated. Alkoxide products of different concentrations were

obtained using a continuous filter reactor in series with a side-en-

try reactor, demonstrating that it was possible to maintain impu-

rity formation at a similar level to the product isolated bycrystallization using the traditional production method [31,70],

thereby confirming that this purification step was not required.

The alkoxide solutions were subsequently hydrolyzed in continu-

ous mode and the organic and aqueous phases were continuously

separated using a hydrophobic membrane device [73].

The dehydration of ‘‘allylcarbinol’’ in THF using HCl as catalyst

was successfully performed in continuous mode with low impurity

formation [75]. The ‘‘butadiene’’ obtained in THF was hydroami-

nated with HEP in batch mode. First, THF and water were removedby batch distillation. Then, ‘‘butadiene’’ was consumed to produce

clopenthixol, a slowprocess that produces some impurities. In con-

clusion, all the operations have been validated, but the hydroamin-

ation reaction requires further development.

3.3.5. Scale-up/scale-out – step 4.5

Due to the limited annual production of the API studied (ca. 4

tonnes/year), the scale-up factor of the industrial alkylation reactor

(filter reactor + side-entry reactor) with respect to the laboratory

equipment was only about one order of magnitude (industrial-

scale flow rate is in the order of 300 mL/min). No units in parallel

were required to scale-up the process, and operations involving

solids (e.g., continuous solid charging) were actually simpler inindustrial scale than in laboratory scale, due to the lack of small-

Table 5

Information collected and design decisions taken throughout the application of the design framework to the hydroamination reaction (step 4.2 of the framework).

Hydroaminaton of ‘‘butadiene’’ with HEP – solvent-free (HEP excess)

Step 4.2 – reactor design Information

collected

Output/decision taken

Step 4.2.1 Phases – reactant

solubility

‘‘Butadiene’’ soluble in HEP

Qualitative analysis Phases – product

solubility

Clopenthixol soluble in HEP

Reaction generic

type

Anti-Markovnikov hydroamination of diene

Safety issues Potential ‘‘butadiene’’ polymerization or oxidation (many double bonds)?

Heterogeneity Not to be expected unless polymerization occurs

Step 4.2.2 Kinetics Very slow reaction (24 h at 90 C)

Calorimetric studies and short kinetic

analysis

Reaction enthalpy Slightly exothermic [76]

Step 4.2.3 Stoichiometry HEP excess favorable

Exploratory design of experiments Concentration No issues found

Temperature Temperature increases reaction rate, but may decrease product yield (selectivity). Reaction

equilibrium may be limited at high temperature.

Kinetics/mixing Slow kinetics

Solubility No issues found

Dissolution rate No issues found

Catalyst Pd(PPh3)4 – no effect on reaction rate

Step 4.2.4 Reaction class Reaction class C (slow reaction, potentially faster by increasing pressure and temperature)

Reaction classification and identification

of limitations

Limitations Kinetic-limited

Step 4.2.5 Cost driver 1 Yield (product selectivity)

Determination of cost drivers and

sustainability indicators

Cost driver 2 Capital cost (large reactor)

Cost driver 3 Low throughput

Batch or continuous? Decision Batch most obvious and simple solution.

If reactionrate couldbe increased, a continuous tubular reactor or series CSTR shouldbe considered.

Continuous reactor could lead improved temperature control and yield.

Step 4.2.6 Cost driver 1 – yield T emperature control, avoid hot-spots

Reactor design Cost driver 2 –

capital cost

Batch reactor probably easiest solution

Cost driver 3 –

throughput

No solutions found unless reaction rate can be increased

Step 4.2.7 Problem

formulation

Maximize product yield

Solve optimization problem Solution method Kinetic modeling, experimentation, heuristics

Constraints Potential impurity formation at high temperature

Economic evaluation of the project Capital and

operational costs

Yield is to be maximized




Table 6

List of tasks and subtasks needed to obtain clopenthixol using the traditional batch-wise process, indicating material inputs (positive values) and waste streams (negative values)

in L/kg reference. The PMI is calculated summing the material inputs, assuming that the densities are approximately 1 kg/L. Values with symbol ‘‘–?’’ indicate that a certain

amount is released but the exact value is not known. Data obtained from H. Lundbeck A/S internal documents. Panel a corresponds to the production and isolation of

‘‘allylcarbinol’’. Values are referred to 1 kg of CTX as reference, while the PMI is calculated for 1 kg of ‘‘allylcarbinol’’. Panel b corresponds to the production of clopenthixol from

‘‘allylcarbinol’’. Values are referred to 1 kg of ‘‘allylcarbinol’’, while the PMI is calculated for 1 kg of clopenthixol. Comments: (1) Mg salts are solubilized using acetic acid; (2) the

pH is increased to avoid formation of impurities downstream; (3) the miscibility of THF and water is unknown; (4) remaining water is unknown; (5) dry conditions are avoided to

prevent ‘‘butadiene’’ formation; (6) distillation stopped at 85 C to avoid drying; (7) cristallization is initiated with ‘‘allylcarbinol’’ crystals; (8) yield expected 80–115%; (9) PMI

referred to kg of ‘‘allylcarbinol’’ product; (10) used to evaporate traces of water; (11) dehydration agent; (12) catalyst; (13) removes polar impurities (e.g., acetic acid); (14)

removes polar impurities; (15) increase to pH 9–11; (16) 6.5 M excess of HEP to ‘‘butadiene’’; solvent exchange lasts ca. 4 h; hydroamination is prolonged for 20 h; (17) aqueousstream sent for regeneration of HEP; (18) remove polar impurities; (19) assuming 70% yield; (20) value will be ca. 26 kg/kg if recycled HEP is subtracted.

Panel a

Task/subtask THF CTX (s) AllylMgCl

(1.4M in THF)

Water Acetic acid

(80% aq)

Ammonia

(25% aq)

Ethanol Allylcarbinol Comments Unit

1. Alkylation 3.5 Ref (1 kg) 3.25 R.

B1

2. Hydrolysis R.

B2

2.1 Water addition 5.1 R.

B2

2.2 Mg salts solub. 0.45 1 R.

B2

2.3 pH correction 0.06 2 R.

B2

3. L–L separation –? 5 –? –? 3 R.

B24. Solvent exchange R.

B2

4.1 Batch distillation 2.1 –? 4 R.

B2

4.2 Water addition 0.4 5 R.

B2

4.3 Batch distillation –? –? 6 R.

B2

4.4 Ethanol addition 5.8 R.

B2

4.5 Water addition 2 R.

B2

5. Crystallization Seed 7 R.

B2

5.1 Water addition 3.8 R.

B2

6. Filtration N. 1

6.1 Filter crystals –? 6 5.8 Approx. N. 1

6.2 First wash 1.2 N. 1

6.3 Filter crystals 1.2 N. 1

6.4 Second wash 0.6 0.6 N. 1

6.5 Filter crystals 0.6 0.6 N. 1

7. Drying –? –? N. 1

Total inputs (L) 3.5 1 kg 3.25 13.1 0.45 0.06 6.4

Product 0.9–1.35 kg 8

PMI (kg/kg) 20–

30

9

Panel b

Task/subtask Water Ammonia

(25% aq)

Allylcarbinol Toluene Acetic acid

anhydride

Acetyl

chloride

HEP E/Z

clopenthixol

Comments Unit

8. Dehydration R.

B3

8.1 Charging/loading Ref (1 kg) 3 R.

B38.2 Evaporate toluene 2 10 R.

B3

8.3 Add acetic acid anhydride 0.36 11 R.

B3

8.4 Add acetyl chloride 0.005 12 R.

B3

8.5 Add toluene 2.4 R.

B3

8.6 First wash 1.15 R.

B3

8.7 L–L separation 1.15 13 R.

B3

8.8 Second wash 1.15 R.

B3


B3

(continued on next page)




scale instruments able to perform certain operations in a robust

manner at laboratory scale.

The continuous alkoxide hydrolysis and phase separation is also

relatively simple to scale up, since high flow rates may be achieved

with relatively small membrane areas [73]. Alternatively, a hydro-

cyclone (or centrifugal contactor separator) integrating mixing,

reaction, and phase separation could be used [72]. The main chal-

lenge of the hydrolysis reaction is the possible co-existence of 4

phases (propene gas in case of AllylMgCl excess, two liquid phases

and possible Mg salts precipitation), which mayentail safety issues.

The rate of the dehydration reaction was increased to a point

where almost total conversion could be obtained in less than 2 or

3 min (120 C, 5 bar), which is manageable using a tubular reactorof reasonable length, even at high throughputs. However, it is ex-

pected that the reaction rate could be increased even further with

low impurity formation if the residence time is optimized. The

hydroamination reaction has not beenstudiedwith sufficient detail

to predict how scale-up should be approached. It is speculated that

very long batch reaction times would entail substrate or product

loss, and thus, future research will focus on the optimization of the

reaction rate. Solvent exchange from THF/water to HEP should be

considered in combination with thehydroaminationreactordesign.

3.4. Monitoring and control – step 5 of the design framework

The selection of monitoring techniques has been done by in-house expertise, and in-line and at-line applications have been

integrated with the process development activities whenever pos-

sible. The large potential of NIR spectroscopy for in-line and at-line

analysis has been demonstrated for a variety of processes

[70,73,75], which can lead to a reduction of process development

time.

Thus far, the simplest approach toward adopting continuous

pharmaceutical production is to convert individual synthetic or

separation steps from batch to continuous mode, and use buffer

tanks to store intermediate compounds. This is due to potential

drastic changes in process conditions, differences in characteristic

times, and limited experience on continuous production. There-

fore, one of the main challenges for the control of a continuous

pharmaceutical manufacturing plant resides in obtaining a contin-uous stream from raw materials to product while being able to re-

spond to process disturbances, which is left for further research.

3.5. Intensification, integration, optimization – step 6 of the design

framework

No further intensification or optimization efforts were done

other than the ones discussed in the previous sections. Mass and

heat integration have not been considered yet. However, an impor-

tant step has been the quantification of the amount of THF solvent

released in the aqueous stream after phase separation, which was

surprisingly high (150–300 g/L [73]). Therefore, it should be con-sidered to recover this solvent from the aqueous stream rather

Table 6 (continued)

Panel b

Task/subtask Water Ammonia

(25% aq)

Allylcarbinol Toluene Acetic acid

anhydride

Acetyl

chloride

HEP E/Z

clopenthixol

Comments Unit

8.10 Extraction with aq ammonia 1.15 0.125 15 R.

B3

8.11 L–L separation 1.15 0.125 14 R.

B3

8.12 Third wash 1.15 R.B3


B3

8.14 Fourth wash 1.15 R.

B3


B3

8.16 Fifth wash 1.15 R.

B3


B3

9. Solvent exchange and

hydroamination

3.4 2.95 16 R.

B4

10. Solvent exchange to toluene &

aqueous extraction of HEP

R.

B4

10.1 Add toluene 5 R.

B410.2 Add water 2.6 R.

B4

10.3 L-L separation 2.6 -2.5 17 R.

B4

10.4 First wash 2 R.

B4

10.5 L-L separation 2 18 R.

B4

10.6 Second wash 2 R.

B4

10.7 L-L separation 2 18 R.

B4

TOTALS in (kg) 13.5 0.125 1 10.4 0.36 0.005 2.95

PRODUCT 1 19

PMI (kg/kg) 28 20

R. Reactor; N. Nutsche.




than treating it as waste. Alternatively, the use of a different ethe-

real solvent should be considered, as discussed in Section 3.3.1.

3.6. Process assessment – step 7 of the design framework

The PMI (material footprint) has been evaluated for the tradi-

tional batch process (see PMI result in Table 6) and the simplified

process (see PMI result in Table 9), assuming that the final work-

up step (extraction of clopenthixol with toluene and extraction of HEP with water) is the same for both (which could potentially be

simplified). Although the analysis does not consider the recovery

of excess HEP for simplicity, it is noteworthy that the material

footprint for the simplified process is roughly half as much as

the footprint of the traditional process. This figure could be de-

creased by further optimization of the simplified process (note

that a worst-case scenario has been assumed for the simplified

process, for example, regarding the water consumption in the

phase separation step), as well as by implementing solvent recov-

ery strategies. The PMI is assumed to be correlated to processoperating costs. Nevertheless, previous economic analyses show

Table 7

Analysis of the function of every task in the base-case design and comparison with the proposal of a simplified process.

Task

#

Operation Tasks/subtasks performed Effect on impurity separation Relevant in simplified process flowsheet?

1 Alkylation Reaction No work-up included Yes

2 Hydrolysis Reaction+ work-up: This is a reaction step but some

of the subtasks are used as preparation of following

tasks. Acetic acid is used to improve the solubility of

the Mg salts formed. The acid is added until the pH is

5–6. However, if the acid was left in the solution, it

would catalyze the formation of an impurity in task 4

that would impair the crystallization (task 5). For this

reason, aqueous ammonia is added to remove the

acetic acid from the organic phase. Besides, acetic

acid is used instead of a stronger acid (e.g., HCl) to

avoid formation of ‘‘butadiene’’, which also impairs

the crystallization

Mg salts solubilized in the

aqueous phase

Alkoxide quench and magnesium salt

solubilization are needed. All other subtasks

are only needed when ‘‘allylcarbinol’’ is

crystallized (see task 5)

3 L–L separation The aqueous phase is discarded Mg salts and the aqueous phase

are separated. Polar impurities are

removed.

Yes

4 Solvent exchange

to ethanol/water

THF and water are removed by batch distillation.

Ethanol and water are added. This solvent

combination is optimal to perform the controlled

crystallization of ‘‘allylcarbinol’’ in the subsequent

step

None No (see task 5)

5 Crystallization of

‘‘allylcarbinol’’

‘‘Allylcarbinol’ ’ is crystallized Impurities are separated if the

crystallization is well controlled

If the alkoxide is obtained in a continuous

Grignard reactor with higher quality, this

step could be avoided

6 Filtration The mother liquor is separated from the

‘‘allylcarbinol’’ crystals. Next, the crystals are washed

with water to remove remaining impurities. A final

wash with water and ethanol is done to facilitate

drying

Remaining impurities are

separated

No (see task 5)

7 Drying Ethanol and water are evaporated until dry

‘‘allylcarbinol’’ powder is obtained

Ethanol and water are evaporated No (see task 5)

8 Dehydration of


Toluene is distilled off to ensure that water (with a

lower boiling point) is not present before the catalyst

(acetyl chloride) is added to start the dehydration. A

molar excess (ca. 110%) of acetic acid anhydride is

used to eliminate water by forming acetic acid. When

the reaction is complete, up to 2 washing steps are

done to remove most of the acetic acid and polar

impurities formed during the dehydration. Then, an

extraction with aqueous ammonia is carried out to

remove any remains of acetic acid from the organic

phase. Up to 3 washing steps, finalize the reaction

work-up by removing remains of acetic acid or

aqueous ammonia

Catalyst and dehydration agent

(both converted to acetic acid) are

removed. Polar impurities are also

washed out

The dehydration is needed. The work-up

washing steps are only required if the acid

catalyst impairs the hydroamination (task

9)

9 Solvent exchange

and

hydroamination

The toluene-‘‘butadiene’’ solution is added to a warm

solution of HEP, from which toluene is slowly

removed by batch distillation. The hydroamination

reaction starts with the ‘‘butadiene’’ addition to HEP

Solvent used in the dehydration

(in this case toluene) is removed

Yes

10 Solvent exchange

to toluene and

aqueous

extraction of HEP

Toluene is added to the product from the

hydroamination reaction to dissolve the clopenthixol

product. At high pH, the amine groups in the

clopenthixol molecule are not protonized and the

product is more soluble in toluene than in HEP. Then,water is added to extract the HEP. The aqueous phase

is sent for regeneration of HEP (by distillation).

Finally, the organic phase is washed twice in order to

remove polar impurities

HEP and polar impurities are

removed

Yes




that raw material costs are typically 30–80% of the operating

costs, meaning that yield and quality are typically the main cost

drivers [29,37,66].

Considering the last column of Tables 6 and 9, it is interesting

that while there is an obvious reduction in the number of tasks

performed when switching to a simplified process containing con-

tinuous units, the number of physical units needed to perform

these tasks is kept constant or even increases. This occurs because

batch reactor B2 in the original process can perform the hydrolysis,

L–L phase separation, solvent exchange, and crystallization opera-

tions. In contrast, a continuous process requires two continuous

alkylation reactors and one continuous hydrolysis reactor, option-

ally integrated with a continuous L–L phase separator. Despite the

fact that there may be a large decrease in unit sizes (smaller phys-

ical footprint), the capital cost may be similar for a batch reactor

and a continuous unit [29,37]. Furthermore, the potential savings

Table 8

List of inputs and outputs to relevant tasks in the base-case design, analysis of possible cross-interactions, and potential degradation issues.

Task # Reactants, reagents,

catalysts, and solvents

Products, catalysts, and

solvents

Degradation issues Cross-interactions

1 – Alkylation CTX, AllylMgCl, THF Alkoxide, CTX?,

AllylMgCl?, THF

Keep N2 atmosphere to avoid

hydrolysis. Avoid high temperature

None

2 – Hydrolysis Alkoxide, CTX?, AllylMgCl?,

THF, Water, Acid

‘‘Allylcarbinol’’, CTX?,

Propene?, THF, Water,

Acid?, MgCl2

None Acid may dehydrate


to ‘‘butadiene’’

4 – Solvent exch. ‘‘Allylcarbinol’’, CTX?, THF,

Water, Ethanol, Acid?

‘‘Allylcarbinol’’, CTX?,

Ethanol, Water, Acid?

Avoid high temperature using

vacuum

distillation

Acid may catalyze the formation

of

an impurity from ethanol and


5 – Crystallization As product #4 As product #4 None ‘‘Butadiene’’ or impurities may

impair the crystallization

6 – Filtration As product #4 ‘‘Allylcarbinol’’, Ethanol

traces, Water traces

None None

7 – Drying As product #6 ‘‘Allylcarbinol’’ None None

8 – Dehydration ‘‘Allylcarbinol’’, Toluene,

Acetic acid anhydride,

Acetyl chloride

‘‘Butadiene’’, Toluene,

Acetic acid?

‘‘Butadiene’’ may degrade by

polymerization. Avoid high

temperature. Use N2 atmosphere

None

9 – Hydroaminaton ‘‘Butadiene’’, HEP, Toluene, Acetic

acid?

Clopenthixol, HEP,

Aceticacid?

As #8 Acid may have a (positive

/negative?) influence on thehydroamination reaction

10 – Solvent exch. Clopenthixol, HEP, Toluene, Water Clopenthixol, Toluene Use N2 atmosphere None

Table 9

List of tasks and subtasks needed to obtain clopenthixol using the simplified process, indicating material inputs (positive values) and waste streams (negative values) in L/kg of

CTX. The PMI is calculated summing the material inputs, assuming that the densities are approximately 1 kg/L. Values marked with are assumed as for the base-case design but

could potentially be reduced or eliminated. Comments: (1) if the concentration of AllylMgCl was increased to 1.5 M, this value would be reduced to 2.7 L/kg, reducing THF

consumption in the Grignard reagent formation; (2) a volume increase of 35% is expected due to API intermediate in solution; a worst-case scenario has been considered where

the volumetric flow ratio between acidic water and alkoxide solution is 2:1; otherwise, this value could be decreased; (3) THF in the aqueous waste stream has been calculated

assuming a concentration of 200 g/L; (4) more HCl is added in case the pH after L–L separation is not low enough for catalysis; (5) assuming 6.5 M excess of HEP with respect to

‘‘butadiene’’ as in the base-case design, but the values are corrected to refer to 1 kg of CTX; (6) the aqueous stream is sent for regeneration of HEP; (7) removes polar impurities;

(8) removes polar impurities; (9) overall yield from CTX assumed as 70%; (10) value will be ca. 26–27 kg/kg if recycled HEP is subtracted.Task THF CTX

(s)

AllylMgCl (1 M

in THF)

Water HCl (37% aq) Toluene HEP E/Z

clopenthixol

Comments Reactor

S1. Alkylation Ref

(1 kg)

4.1 1 Reactors C1

and C2

S2. Hydrolysis 10.9 0.34 2 Reactor C3

S3. L–L separation 2.4 11.23 Consumed to

MgCl2

3 Separator C1

S4. Dehydration 0.03 4 Reactor C4

S5. Solvent exchange and hydroamination 1.4 0.08 3.5⁄ 5 Reactor B5

S6. Solvent exchange to toluene and

aqueous extraction of HEP

Reactor B5

S6.1 Add toluene 5.9⁄ Reactor B5

S6.2 Add water 3⁄ Reactor B5

S6.3 L–L separation 3⁄

2.9⁄ 6 Reactor B5

S6.4 First wash 2.3⁄ Reactor B5

S6.5 L–L separation 2.3⁄ 7 Reactor B5

S6.6 Second wash 2.3⁄ Reactor B5

S6.7 L–L separation 2.3⁄ 8 Reactor B5

Totals in (kg) 4.1 18.7⁄ 0.37 5.9⁄ 3.5⁄

Product 1.1 9

PMI (kg/kg) 29⁄ 10




obtained from less manual operations may be compensated by

higher costs needed to implement in-line sensors and actuators.

Hence, while a new pharmaceutical process of these characteristics

would most likely be designed and implemented according to the

simplified process shown in Fig. 7 (or a further optimized process),

moving the traditional batch process toward the simplified process

including continuous operations can only be justified based on a

rigorous economical and environmental analysis.

The mode of operation of each unit will most likely be consid-

ered on a case by case basis and depending on the short-term

and long-term strategies of the manufacturing company. In this

particular case study, it is clear that changing the alkylation reac-

tion to continuous mode brings an obvious product quality

improvement that enables the elimination of a solvent exchange

step, crystallization, filtering, drying, collecting the product, and fi-

nally the storage. Therefore, it is an obvious advantage in terms of

operating cost as well as capital cost. It is less clear whether mov-

ing the dehydration reaction to continuous mode will result in

immediate savings. However, avoiding the use of toluene, acetyl

chloride, and acetic acid anhydride in this reaction, as well as elim-

inating the numerous subsequent washing steps (subtasks 8.5–

8.17 in Table 6) may in the long term compensate the investment

needed to establish a small-scale continuous tubular reactor run-

ning at high temperature under pressure. Yet, the eventual connec-

tion of a continuous dehydration reactor with a discontinuous

hydroamination reaction may be a challenge.

3.7. Process implementation – step 8 of the design framework

The continuous alkylation reactor has been implemented at H.

Lundbeck A/S, demonstrating a production of high quality alkoxide

with low solvent consumption using smaller scale equipment. The

product is continuously monitored using in-line NIR spectroscopy

measurements, which assist the operators in troubleshooting situ-

ations. The simplified process in Fig. 7 will be gradually imple-

mented at H. Lundbeck A/S on a step-by-step basis as the processexperience increases. This is probably the safest method for retro-

fitting an existing production plant with respect to guaranteeing

that product supply can be maintained.

4. Conclusions

A systematic framework has been proposed to design continu-

ous pharmaceutical manufacturing processes, that is, processes

that exploit the advantages of continuous flow. The framework fol-

lows the drug product and process development cycle and pro-

motes the synergic interaction of PSE and microfluidic techniques

throughout it, emphasizing the importance of solvent selection,

reactor design, and separation process design. The application of

the framework starts already at the drug discovery level, whereefficient interaction with medicinal chemists can result in reduced

development time, selection of environmentally friendly synthetic

routes, and smoother scale-up. Guidelines are suggested to ascer-

tain when to perform a certain operation in batch or in continuous

mode, while a final process evaluation in terms of cost, environ-

mental footprint, quality, and safety must be performed to evalu-

ate the viability of a design project.

The design framework has been applied to retrofit an existing

batch-wise manufacturing plant used by H. Lundbeck A/S to pro-

duce clopenthixol. The process includes a set of reaction steps with

different characteristic times, L–L phase separations, and solvent

exchange steps by distillation. The use of continuous reactors re-

sulted in improved product quality, thus avoiding the isolation of

an intermediate product by crystallization and eliminating prepa-ration (solvent exchange to anti-solvent) and subsequent steps (fil-

tration, drying, and storage). It was shown that the simplification

of the process used to manufacture clopenthixol yields a reduction

of the material footprint of the process (evaluated by the process

mass intensity index) with at least 50%. This reduction is correlated

with the environmental footprint and to operating costs. The cap-

ital costs of the plant could also be reduced by the elimination of

some of its large units, isolation, and storage facilities. The design

framework assisted in structuring the different and challenging

design problems faced and could especially be useful for the devel-

opment of novel continuous pharmaceutical manufacturing pro-

cesses through increased process understanding.

Acknowledgements

We thank the Technical University of Denmark and H. Lundbeck

A/S for technical and financial support.

References

[1] N.A.M. Tamimi, P. Ellis, Drug development: from concept to marketing!,

Nephron Clin Pract. 113 (2009) 125–131.

[2] K. Plumb, Continuous processing in the pharmaceutical industry - changing

the mindset, Chem. Eng. Res. Des. 83 (2005) 730–738.[3] A. Behr, V.A. Brehme, C.L.J. Ewers, H. Grön, T. Kimmel, S. Küppers, I. Symietz,

New developments in chemical engineering for the production of drug

substances, Eng. Life Sci. 4 (2004) 15–24.

[4] U.S. Department of Health and Human Services, Food and DrugAdministration,

PAT – A Framework for Innovative Pharmaceutical Development,

Manufacturing, and Quality Assurance, 2004.

[5] R.L. Hartman, K.F. Jensen, Microchemical systems for continuous-flow

synthesis, Lab Chip 9 (2009) 2495–2507.

[6] J.P. McMullen, K.F. Jensen, Integrated microreactors for reaction automation:

newapproaches to reaction development, Annu. Rev. Anal.Chem. 3 (2010) 19–

42.

[7] K.V. Gernaey, R. Gani, A model-based systems approach to pharmaceutical

product-process design and analysis, Chem. Eng. Sci. 65 (2010) 5757–5769.

[8] J.P. McMullen, K.F. Jensen, Rapid determination of reaction kinetics with an

automated microfluidic system, Org. Process Res. Dev. 15 (2011) 398–407.

[9] R.K. Henderson, C. Jiménez-González, D.J.C. Constable, S.R. Alston, G.G.A. Inglis,

G. Fisher, J. Sherwood, S.P. Binks, A.D. Curzons, Expanding GSK’s solvent

selection guide – embedding sustainability into solvent selection starting at

medicinal chemistry, Green Chem. 13 (2011) 854–862.

[10] C. Jiménez-González, P. Poechlauer, Q.B. Broxterman, B. Yang, D. am Ende, J .

Baird, C. Bertsch, R.E. Hannah, P. Dell’Orco, H. Noorman, S. Yee, R. Reintjens, A.

Wells, V. Massonneau, J. Manley, Key green engineering research areas for

sustainable manufacturing: a perspective from pharmaceutical and fine

chemicals manufacturers, Org. Process Res. Dev. 15 (2011) 900–911.

[11] K.V. Gernaey, A.E. Cervera-Padrell, J.M. Woodley, A perspective on PSE in

pharmaceutical process development and innovation, Comput. Chem. Eng. 42

(2012) 15–29.

[12] V. Hessel, Novel process windows – gate to maximizing process intensification

via flow chemistry, Chem. Eng. Technol. 32 (2009) 1655–1681.

[13] L. Kang, B.G. Chung, R. Langer, A. Khademhosseini, Microfluidics for drug

discovery and development: from target selection to product lifecycle

management, Drug Discov. Today 13 (2008) 1–13.

[14] D. Webb, T.F. Jamison, Continuous flow multi-step organic synthesis, Chem.

Sci. 1 (2010) 675–680.

[15] R. Singh, R. Rozada-Sanchez, T. Wrate, F. Muller, K.V. Gernaey, R. Gani, J.M.

Woodley, A retrofit strategy to achieve ‘‘fast, flexible, future (F3)’’

pharmaceutical production processes, Comput. Aided Chem. Eng. 29 (2011)291–295.

[16] R. Singh, K.V. Gernaey, R. Gani, An ontological knowledge-based system for the

selection of process monitoring and analysis tools, Comput. Chem. Eng. 34

(2010) 1137–1154.

[17] V. Venkatasubramanian, C. Zhao, G. Joglekar, A. Jain, L. Hailemariam, P. Suresh,

P. Akkisetty, K. Morris, G.V. Reklaitis, Ontological informatics infrastructure for

pharmaceutical product development and manufacturing, Comput. Chem. Eng.

30 (2006) 1482–1496.

[18] R. Gani, P. Arenas Gómez, M. Folic, C. Jiménez-González, D.J.C. Constable,

Solvents in organic synthesis: replacement and multi-step reaction systems,

Comput. Chem. Eng. 32 (2008) 2420–2444.

[19] D.J.C. Constable, C. Jiménez-González, R.K. Henderson, Perspective on solvent

use in the pharmaceutical industry, Org. Process Res. Dev. 11 (2007) 133–137.

[20] R. Gani, C. Jiménez-González, D.J.C. Constable, Method for selection of solvents

for promotion of organic reactions, Comput. Chem. Eng. 29 (2005) 1661–1676.

[21] M. Folic, C.S. Adjiman, E.N. Pistikopoulos, Design of solvents for optimal

reaction rate constants, AIChE J. 53 (2007) 1240–1256.

[22] J. Chen, B.L. Trout, Computer-aided solvent selection for improving the

morphology of needle-like crystals: a case study of 2,6-dihydroxybenzoicacid, Cryst. Growth Des. 10 (2010) 4379–4388.




[23] R. Gani, C. Jiménez-González, A.T. Kate, P.A. Crafts, M. Jones, L. Powell, J.H.

Atherton, J.L. Cordiner, A modern approach to solvent selection, Chem. Eng. –

New York (March) (2006) 30–43.

[24] D.C. Weis, D.P. Visco, Computer-aided molecular design using the signature

molecular descriptor: application to solvent selection, Comput. Chem. Eng. 34

(2010) 1018–1029.

[25] D. Hsieh, A.J. Marchut, C. Wei, B. Zheng, S.S.Y. Wang, S. Kiang, Model-based

solvent selection during conceptual process design of a new drug

manufacturing process, Org. Process Res. Dev. 13 (2009) 690–697.

[26] Y. Li, Y. Yang, V. Kalthod, S.M. Tyler, Optimization of solvent chasing in API

manufacturing process: constant volume distillation, Org. Process Res. Dev. 13(2009) 73–77.

[27] J.C. Lin, A.G. Livingston, Nanofiltration membrane cascade for continuous

solvent exchange, Chem. Eng. Sci. 62 (2007) 2728–2736.

[28] P.A. Santacoloma, G. Sin, K.V. Gernaey, J.M. Woodley, Multienzyme-catalyzed

processes: next-generation biocatalysis, Org. Process Res. Dev. 15 (2011) 203–

212.

[29] D.M. Roberge, L. Ducry, N. Bieler, P. Cretton, B. Zimmermann, Microreactor

technology: a revolution for the fine chemical and pharmaceutical industries?,

Chem Eng. Technol. 28 (2005) 318–323.

[30] R.L. Hartman, J.R. Naber, N. Zaborenko, S.L. Buchwald, K.F. Jensen, Overcoming

the challenges of solid bridging and constriction during Pd-catalyzed C–N

bond formation in microreactors, Org. Process Res. Dev. 14 (2010) 1347–1357.

[31] K. Müller Christensen, M. Jønch Pedersen, K. Dam-Johansen, T. Lønberg Holm,

T. Skovby, S. Kiil, Design and operation of a filter reactor for continuous

production of a selected pharmaceutical intermediate, Chem. Eng. Sci. 71

(2012) 111–117.

[32] D.M. Roberge, An Integrated approach combining reaction engineering and

design of experiments for optimizing reactions, Org. Process Res. Dev. 8 (2004)

1049–1053.

[33] J.P. McMullen, K.F. Jensen, An automated microfluidic system for online

optimization in chemical synthesis, Org. Process Res. Dev. 14 (2010) 1169–

1176.

[34] G. Franceschini, S. Macchietto, Model-based design of experiments for

parameter precision: state of the art, Chem. Eng. Sci. 63 (2008) 4846–4872.

[35] N. Kockmann, D.M. Roberge, Harsh reaction conditions in continuous-flow

microreactors for pharmaceutical production, Chem. Eng. Technol. 32 (2009)

1682–1694.

[36] R.L. Hartman,J.P. McMullen, K.F. Jensen, Deciding whether to go with the flow:

evaluating the merits of flow reactors for synthesis, Angew. Chem. Int. Edit. 50

(2011) 7502–7519.

[37] D.M. Roberge, B. Zimmermann, F. Rainone, M. Gottsponer, M. Eyholzer, N.

Kockmann, Microreactor technology and continuous processes in the fine

chemical and pharmaceutical industry: is the revolution underway?, Org

Process Res. Dev. 12 (2008) 905–910.

[38] C. Jimenez-Gonzalez, C.S. Ponder, Q.B. Broxterman, J.B. Manley, Using the right

green yardstick: why process mass intensity is used in the pharmaceutical

industry to drive more sustainable processes, Org. Process Res. Dev. 15 (2011)912–917.

[39] A. Carvalho, R. Gani, H. Matos, Design of sustainable chemical processes:

systematic retrofit analysis generation and evaluation of alternatives, Process

Saf. Environ. Prot. 86 (2008) 328–346.

[40] N. Kockmann, M. Gottsponer, D.M. Roberge, Scale-up concept of single-

channel microreactors from process development to industrial production,

Chem. Eng. J. 167 (2011) 718–726.

[41] D. Hildebrandt, D. Glasser, The attainable region and optimal reactor

structures, Chem. Eng. Sci. 45 (1990) 2161–2168.

[42] A.C. Kokossis, A.F. Christodoulos, Optimizationof complex reactor networks—I.

Isothermal operation, Chem. Eng. Sci. 45 (1990) 595–614.

[43] V.L. Mehta, A.C. Kokossis, Nonisothermal synthesis of homogeneous and

multiphase reactor networks, AICHE J. 46 (2000) 2256–2273.

[44] P. Lutze, R. Gani, J.M. Woodley, Process intensification: a perspective on

process synthesis, Chem. Eng. Process. 49 (2010) 547–558.

[45] P. Lutze, R. Gani, J.M. Woodley, Phenomena-based process synthesis and

design to achieve process intensification, Comput. Aided Chem. Eng. 29 (2011)

221–225.

[46] D.I. Gerogiorgis, P.I. Barton, Steady-state optimization of a continuouspharmaceutical process, Comput. Aided Chem. Eng. 27 (2009) 927–932.

[47] S.D. Barnicki, J.J. Siirola, Process synthesis prospective, Comput. Chem. Eng. 28

(2004) 441–446.

[48] J.C. Brunet, Y.A. Liu, Studies in chemical process design and synthesis: 10. An

expert system for solvent-based separation process synthesis, Ind. Eng. Chem.

Res. 32 (1993) 315–334.

[49] C.A. Jaksland, R. Gani, K.M. Lien, Separation process design and synthesis based

on thermodynamic insights, Chem. Eng. Sci. 50 (1995) 211–530.

[50] L. d’Anterroches, R. Gani, Group contribution based process flowsheet

synthesis, design and modelling, Fluid Phase Equilib. 228–229 (2005) 141–

146.

[51] E. Bek-Pedersen, R. Gani, Design and synthesis of distillation systems using a

driving-force-based approach, Chem. Eng. Process. 43 (2004) 251–262.

[52] R.L. Hartman, H.R. Sahoo, B.C. Yen, K.F. Jensen, Distillation in microchemical

systems using capillary forces and segmented flow, Lab Chip 9 (2009) 1843–

1849.

[53] J.G. Kralj, H.R. Sahoo, K.F. Jensen, Integrated continuous microfluidic liquid-

liquid extraction, Lab Chip 7 (2007) 256–263.

[54] A. Ziogas, V. Cominos, G. Kolb, H. Kost, B. Werner, V. Hessel, Development of a

microrectificationapparatus for analytical and preparative applications, Chem.

Eng. Technol. 35 (2012) 58–71.

[55] International Conference on Harmonisation (ICH), Quality Guidelines Q8

(Pharmaceutical Development, R2, 2009), Q9 (Quality Risk Management,

2006) and Q10 (Pharmaceutical Quality System, 2009), U.S. Department of

Health and Human Services, Food and Drug Administration, Center for Drug

Evaluation and Research (CDER), Rockville, MD.

[56] K.F. Jensen, Microreaction engineering – is small better?, Chem Eng. Sci. 56(2001) 293–303.

[57] M. Mendorf, H. Nachtrodt, A. Mescher, A. Ghaini, D.W.Agar, Designand control

techniques for the numbering-up of capillary microreactors with uniform

multiphase flow distribution, Ind. Eng. Chem. Res. 49 (2010) 10908–10916.

[58] P. Barthe, C. Guermeur, O. Lobet, M. Moreno, P. Woehl, D.M. Roberge, N. Bieler,

B. Zimmermann, Continuous multi-injection reactor for multipurpose

production – Part I, Chem. Eng. Technol. 31 (2008) 1146–1154.

[59] H.S. Pordal, C.J. Matice, T.J. Fry, The role of computational fluid dynamics in the

pharmaceutical industry, Pharm. Technol. (February) (2002).

[60] G. Wernet, S. Conradt, H. Isenring, C. Jiménez-González, K. Hungerbühler, Life

cycle assessment of fine chemical production: a case study of pharmaceutical

synthesis, Int. J. Life Cycle Ass. 15 (2010) 294–303.

[61] S. García Muñoz, S. Dolph, H.W. Ward II, Handling uncertainty in the

establishment of a design space for the manufacture of a pharmaceutical

product, Comput. Chem. Eng. 34 (2010) 1098–1107.

[62] R. Singh, K.V. Gernaey, R. Gani, ICAS-PAT: a software for design, analysis and

validation of PAT systems, Comput. Chem. Eng. 34 (2010) 1108–1136.

[63] A. Malcolm, J. Polan, L. Zhang, B.A. Ogunnaike, A.A. Linninger, Integrating

systems design and control using dynamic flexibility analysis, AICHE J. 53

(2007) 2048–2061.

[64] M.K.A. Hamid, G. Sin, R. Gani, Integration of process design and controller

design for chemical processes using model-based methodology, Comput.

Chem. Eng. 34 (2010) 683–699.

[65] A.A. Linninger, A. Malcolm, Pollution prevention for batch pharmaceutical and

specialty chemical processes, in: E. Korovessi, A.A. Linninger (Eds.), Batch

Processes, CRC Press, Taylor & Francis Group, Boca Raton, Florida, US, 2006.

[66] S.D. Schaber, D.I. Gerogiorgis, R. Ramachandran, J.M.B. Evans, P.I. Barton, B.L.

Trout, Economic analysis of integrated continuous and batch pharmaceutical

manufacturing: a case study, Ind. Eng. Chem. Res. 50 (2011) 10083–10092.

[67] G.S. Silverman, P.E. Rakita, Handbook of Grignard Reagents, CRC Press, New

York, 1996.

[68] D.F. Aycock, Solvent applications of 2-methyltetrahydrofuran in

organometallic and biphasic reactions, Org. Process Res. Dev. 11 (2007) 156–

159.

[69] K. Watanabe, N. Yamagiwa, Y. Torisawa, Cyclopentyl methyl ether as a new

and alternative process solvent, Org. Process Res. Dev. 11 (2007) 251–258.

[70] A.E. Cervera-Padrell, J.P. Nielsen, M. Jønch Pedersen, K. Müller Christensen, A.R.Mortensen, T. Skovby, K. Dam-Johansen, S. Kiil, K.V. Gernaey, Monitoring and

control of a continuous Grignard reaction for the synthesis of an active

pharmaceutical ingredient intermediate using in-line NIR spectroscopy, Org.

Process Res. Dev. 16 (2012) 901–914.

[71] D.M. Roberge, N. Bieler, M. Mathier, M. Eyholzer, B. Zimmermann, P. Barthe, C.

Guermeur, O. Lobet, M. Moreno, P. Woehl, Development of an industrial multi-

injection microreactor for fast and exothermic reactions – Part II, Chem. Eng.

Technol. 31 (2008) 1155–1161.

[72] G.N. Kraai, F. van Zwol, B. Schuur, H.J. Heeres, J.G. de Vries, Two-phase

(bio)catalytic reactions in a table-top centrifugal contact separator, Angew.

Chem. Int. Edit. 47 (2008) 3905–3908.

[73] A.E. Cervera-Padrell, S.T. Morthensen, D.J. Lewandowski, T. Skovby, S. Kiil, K.V.

Gernaey, Continuous hydrolysis and L-L phase separation of an active

pharmaceutical ingredient intermediate using a mini-scale hydrophobic

membrane separator, Org. Process Res. Dev. 16 (2012) 888–900.

[74] B.K.H. Yen, A. Günther, M.A. Schmidt, K.F. Jensen, M.G. Bawendi, A

microfabricated gas–liquid segmented flow reactor for high-temperature

synthesis: the case of CDSE quantum dots, Angew. Chem. 117 (2005) 5583–

5587.[75] A.E. Cervera-Padrell, Moving from Batch Towards Continuous Organic-

Chemical Pharmaceutical Production, PhD Thesis, Technical University of

Denmark, Kgs. Lyngby, 2012.

[76] T.E. Müller, K.C. Hultzsch, M. Yus, F. Foubelo, M. Tada, Hydroamination: direct

addition of amines to alkenes and alkynes, Chem. Rev. 108 (2008) 3795–3892.

[77] J.F. Hartwig, Development of catalysts for the hydroamination of olefins, Pure

Appl. Chem. 76 (2004) 507–516.

[78] O. Löber, M. Kawatsura, J.F. Hartwig, Palladium-catalyzed hydroamination of

1,3-dienes: a colorimetric assay and enantioselective additions, J. Am. Chem.

Soc. 123 (2001) 4366–4367.

[79] K.G. Tomazi, A.A. Linninger, J.R. Daniel, Batch processing industries, in: E.

Korovessi, A.A. Linninger (Eds.), Batch Processes, CRC Press, Taylor & Francis

Group, Boca Raton, Florida, US, 2006.


active pharmaceutical ingredient api production involving continuous processes a process system...

Documents