There is an obvious and increasing need to preserve valuable resources and reduce waste and pollution. Maximising the functional lifetime of materials with reuse and recycling practices has long-term benefits (1). These themes are embodied in the circular economy concept, where materials are considered in terms of the service they provide when fabricated into products (2). Extended product lifespans deliver more service from a material, while waste represents lost potential.

The EU and China are the two regions with the most prominent circular economy strategies. Specific policies have been established since 2015 in the EU (2) and even earlier by the Chinese government (3). Although the regulatory measures are broad, encompassing critical materials and product (eco)-design, European law focuses on recycling targets. China has additional policies encouraging industrial symbiosis so responsibility for waste is shared, including heat and material (waste) outputs of one industry being provided as the input for another. Academic interest in the circular economy concept is high, ranging from policy to product design to improved recycling technologies.

Although recycling targets are an obvious, easily monitored and (potentially) enforceable legislative measure to promote a circular economy, there are many end of life options that preserve a much greater degree of product functionality. Waste avoidance is not enough, if it were then current trends towards biodegradable packaging, waste incineration and landfill reduction would be sufficient. Maintaining and extending the maximum value of limited resources is necessary for a sustainable society. In order to reuse, repair, remanufacture and refurbish products, manufacturers need to be involved in the value chain beyond production. This could be in formal partnerships with waste processing agencies or by implementing extended producer responsibility, pledging to return defunct products to use (4, 5). A change of emphasis towards valuing a product’s service not its material worth prompts a reduction in waste and better use of resources. For example, chemical leasing is a business model where payment for a service is based on productivity, not how much material changes hands (6). Under this business model it is possible to buy paint on the basis of what surface area is to be coated or industrial solvents according to how much apparatus needs to be cleaned or degreased. It is now important to the selling party to provide as little product as possible to maximise profit and in doing so minimise waste. Similar principles are being applied to consumer purchases of clothes and electronic devices on a leasing basis, rather than buying an article outright and eventually disposing of it (7).

Inevitably all products will become obsolete and the obvious way to extend the value provided by finite materials at this point is through recycling. Recycling processes for most types of material produce an inferior product that enters lower value applications, known as open-loop recycling or downcycling. Coupled with poor collection rates, this means 95% of the economic value of the plastic market is lost after a single use (8). Mechanical recycling is effective for PET, polyethylene (PE) and polypropylene (PP), whereby the waste is shredded, melted and remoulded (9). Recycling infrastructure for other polymers is more limited internationally and for composites and thermoset plastics the design and chemical composition of the material excludes conventional recycling completely as an end of life option (10). The presence of additives in many plastic products results in a recyclate with unknown impurities, some of which are toxic and they may be unnecessary or undesirable for the secondary uses of the material.

Closed-loop recycling, returning materials back to their original use, is prevented by product designs that irreversibly combine different types of materials, but also by waste collection and separation processes and the recycling processes themselves. These three aspects of waste management can be addressed by proactive product design, policy action regarding waste collection and recycling infrastructure and engineers and scientists motivated to create new recycling technologies.

Solvents can be used to selectively dissolve waste polymers at end of life for the separation of mixed wastes and composites. The advantage of this technology compared to mechanical recycling is that it is capable of returning a plastic with the same quality as virgin materials as judged by tensile strain and other properties. Recyclate specification sheets often include space for this technical information alongside a description of its appearance (such as colour and particle size) (11). Quality control and the communication of recyclate properties is important to ensure the most value is obtained when deciding what materials are used to make products. Chemical recycling is another alternative recycling technique that takes material a step further back in the production chain by depolymerising it back to monomers (12). This is advantageous for polymers that degrade during use, including biodegradable polymers wrongly captured by recycling practices or that are unstable at the elevated temperatures used in recycling processes. Chemical recycling can potentially be solvent free but in many examples a solvent is required to homogenise the polymer with reactants and catalysts.

In this work, three important case studies will be discussed where a solvent-based process is used to recycle a polymer. The emphasis is on commercial applications, exploring their advantages and limitations. For a theoretical examination of polymer solubility and the related phenomena of gelation and swelling, other literature is available that provides the background knowledge for solvent-based recycling methods (13).

For completeness, it must be said there are less desirable end of life options for waste in a circular economy whereby the value of materials is significantly reduced or completely eliminated. This includes increasingly popular energy recovery (incineration), as well as biodegradation and landfill. Incineration offers some value and offsets energy demand that would otherwise likely be obtained from fossil fuels. Despite the additional use of waste material as a fuel, ultimately the material is lost. Carbon emissions and any other form of pollution represents a loss of resource and the material value it could have provided to society. Biodegradable products are designed to avoid litter. There are also some instances where it is impossible to collect a product for reuse or recycling. One example is lubricants. Forestry regulations require chainsaw and other ‘total-loss’ lubricants to be biodegradable (14). To prevent avoidable resource depletion and waste, the only articles suitable for incineration or biodegradation in a circular economy are bio-based products made only of sustainably sourced renewable materials (15).

One of the most ubiquitous forms of plastic waste is the plastic bottle. Typically made of PET, these single use articles can be effectively recycled, although most often this is in an open recycling loop to make polyester fabrics. Despite this, the recycling rate of PET bottles in Europe is below capacity at only 57% (16), indicating flaws in collection and sorting. Product design also limits recycling. Once (recycled) PET is combined with other materials to make textile products, the inability of conventional recycling processes to separate the PET means there is no option to further recycle the material. For textiles consisting of a mix of cotton and PET, a solvent-based process can perform the separation and recovery of both components.

There are a large number of patented procedures for recycling textile waste containing mixed polyester and cotton items, typically clothes. A solvent can be applied to selectively dissolve either cellulose or PET. The remaining, undissolved polymer can also be recycled after filtration and drying or alternatively converted into a derivative compound. To selectively dissolve cellulose, the solvents used to make rayon fibres are applicable, such as N-methylmorpholine N-oxide (NMMO) which is used in the Lyocell process. The high flammability and oxidising potential of NMMO does not make it an ideal solvent from a safety point of view but it is typically recycled within processes with high efficiency. It has been reported that processes dissolving the PET component of composite textiles, for example in sulfolane (17), reduce the quality of the cellulose fibres (18). Nevertheless, the difficulty in dissolving cellulose has meant research efforts have focused on the solvent-based recovery of PET from textiles rather than the cotton.

Worn Again is a UK-based company that has developed technology for the closed-loop recycling of PET from textiles. A demonstrator pilot plant is due to be operational in 2021 (19). The principal technology describes a solvent added to blended polyester-cotton textiles at an elevated temperature (for example 100°C) (20). Suitable PET solvents include aromatic esters and aldehydes, as well as dipropylene glycol methyl ether acetate. Hot filtration removes undissolved cellulose from the solution of PET. The polyester is obtained with the use of isopropanol acting as an antisolvent. Characterisation of the separated polymers is not available, aside from a statement in the patent that the recovered PET has an identical infrared (IR) spectrum to the virgin material (20). Other works indicate that dissolution-precipitation cycles do not impact the polymer molecular weight, but the crystallinity of the recyclate is significantly lower than virgin PET (21). Here N-methyl-2-pyrrolidone (NMP) was used as the solvent and an alkane for the antisolvent. The use of reprotoxic NMP is not sustainable in the presence of tightening regulations (22) and the forced precipitation by antisolvents is probably responsible for the crystallinity of the isolated polymer. Greater attention is needed at the precipitation phase of the process to produce higher quality polymers.

Worn Again has also patented a procedure for recycling PET packaging, including drinks bottles (23). The key innovation that distinguishes this from mechanical recycling is the removal of dyes that otherwise dictate the quality of recyclate (Figure 1). Synthetic textiles are also appropriate feedstocks for this process. Coloured plastics and dyed textiles are far less valuable as a secondary feedstock for products compared to uncoloured transparent materials. The Worn Again technology is based on a solvent or temperature switch to firstly dissolve any dyes (but not PET) and then the polymer is dissolved at a higher temperature or in a different solvent. It is important that the first solvent swells but does not dissolve PET under the operating conditions. For instance, dyes are dissolved in ethyl benzoate at 120°C and liberated from the swollen plastic. After removing the dye solution, a second batch of ethyl benzoate is added at 180°C to dissolve the polymer. It is necessary to implement this second step to remove any insoluble impurities. For this to be economically viable the solvent will need to be recycled and in this regard the process is simplified by using the same solvent throughout. A PET recovery of 96% is satisfactory.

Fig. 1.

Solvents described as able to dissolve PET are provided in Table I (20, 23). Due to solvent residue potentially trapped in the recyclate, it is important to consider toxicity as part of solvent selection. The CHEM21 solvent selection guide categorises hazards into safety (S), health (H) and environmental (E) impact using a 1–10 scale where high scores reflect severe hazards (24). Benzyl acetate and ethyl benzoate are listed as having the best health and safety profile. High boiling solvents such as these are penalised in the environmental category because recovery by distillation is energy intensive. Depending on the proposed applications of the recycled PET, residual solvent limits for food contact applications or other regulations must also be considered.

Ultimately any disruptive PET recycling technology needs to provide significant advantages over efficient and widely practiced conventional mechanical recycling processes. The ability to separate combinations of materials is a crucial aspect of solvent-based recycling. With still much to be done to improve recovery rates of easier to recycle products, new technologies will only become commonplace if there is a political will to approach very high, near complete recycling rates, including composites.

Recycling techniques primarily aim to preserve the chemical structure of materials, but polyesters, with their susceptibility to hydrolysis and alcoholysis, are also possible to depolymerise into monomers. Alcoholysis of PET is a favourable chemical recycling approach because its reaction with ethylene glycol produces bis(2-hydroxyethyl) terephthalate as an appropriate monomer to remake PET. Alternatively methanol will produce dimethyl terephthalate and ethylene glycol which can be combined as they are to produce virgin PET (25). Hydrolysis creates an aqueous solution of terephthalic acid and ethylene glycol which thermodynamically discourages esterification.

Chemical recycling is appropriate when other recycling methods produce a poor quality recyclate. This could be due to contamination that is possible to remove during chemical recycling or because the polymer is prone to decomposition. PLA is thought to be responsible for both these issues by the recycling industry. As a biodegradable polymer, mechanical recycling causes degradation into shorter fibres and as a polyester it is also likely to contaminate PET recyclate. However, PLA is suited to chemical recycling. Hydrolysis or alcoholysis produces a single monomer and it is more rapidly decomposed than PET. This means PET waste destined for recycling can be pretreated to remove any PLA by chemical recycling. It has also been shown that mixtures of PLA and PET can be sequentially chemically recycled into their respective monomers in a two-step process so that the polymers no longer contaminate one another (Figure 2) (26). This concept proves useful where conventional sorting techniques (such as near-IR) cannot distinguish between polyesters (27), although new analytical systems are being developed to address this (28).

Fig. 2.

Although recycling PET mechanically without the need for depolymerisation or solvents is the prevailing technology, interest in alternatives is increasing, for example by Carbios, France and DEMETO, EU Framework Programme for Research and Innovation Horizon 2020. The understanding of PLA chemical recycling is arguably more advanced, but commercialisation is constrained by the small market share of PLA and the types of product it is used in. Many PLA containing products are designed for composting at end of life (for instance, plastic lined disposable coffee cups, transparent films for food packaging and other applications). As a bio-based polymer, PLA films are suitable for composting in a circular economy (if other end of life options that preserve more value are not accessible) as there is no net loss of material or emissions from a material perspective. Having said that, there are also many other components and articles made of PLA that will not biodegrade in the conditions provided by industrial composting units (PLA is not suitable for home composting). Thicker PLA materials, such as those that result from three-dimensional (3D) printing with PLA filaments are unlikely to be adequately decomposed by biodegradation on a viable timespan. The possibility that PLA is collected together with PET waste is increasing with the advent of reusable PLA drinks bottles, creating a reason to consider chemically recycling PLA.

Zeus Industrial Products, USA, has patented a process for depolymerising PLA using conditions where PET is unreactive (29). An inert solvent (chloroform) is added to the polymer, along with reactant (methanol) and catalyst (tin dioctanoate) to complete depolymerisation at 57°C. Disadvantages of this process include the use of chloroform, which is toxic if inhaled and suspected of causing cancer and reprotoxicity (24). Full depolymerisation also requires several hours (30).

The Futerro LOOPLA^® process (a joint enterprise formed by Galactic, Belgium and Total, France) is another method for chemically recycling PLA. The company has expertise in PLA production as well as its hydrolysis and alcoholysis at end of life. Either chemical recycling method is potentially able to remake a feedstock suitable for PLA production (31). Hydrolysis can occur in a solution of PLA in ethyl lactate at 130–140°C (32). Ethyl lactate is a significantly less hazardous solvent than the chlorinated solvents that are often used to dissolve PLA and other polyhydroxyalkanoates (24, 33). Without the addition of a catalyst, 97% recovery of lactic acid (isolated by crystallisation) is achieved with minimal hydrolysis of the solvent. Potential contamination by PE, PP or PET is resolved because ethyl lactate does not dissolve these polymers, which can be used advantageously to separate PLA from other plastic wastes by hot filtration. If ethanol is added to the recycling process instead of water, alcoholysis occurs (34). The product is identical to the solvent, ethyl lactate, and so separation is simplified. Distillation removes excess ethanol and residues (such as pigments and contamination). An acid catalyst is required and triazabicyclodecene is preferred.

An issue with the described recycling procedures is the product (lactic acid or its esters) is subject to racemisation which produces inferior polymers with lower crystallinity (12). This must be controlled in order to perform closed-loop recycling. Furthermore, the electricity demand is too high for chemical recycling to compete with mechanical recycling (35, 36). While this is a valid concern for PET, mechanical recycling is not appropriate for PLA anyway due to its degradation (37). The first major barrier preventing chemical recycling of PLA being operated at any appreciable scale is the lack of feedstock and therefore an absence of designated PLA waste collection (38). However, the market growth of PLA products indicates future measures to capture PLA waste will need to be implemented.

Many solvent-based recycling research projects and pilot trials have been successful, but few are viable commercial processes because of the competition from mechanical recycling and in the case of PLA the limited feedstock. The most prominent example of a successful recycling process conducted in a solvent was the VinyLoop^® process, yet after 16 years of operation the plant was closed in 2018. It is important to understand the reasons why to ensure more recycling operations do not close and waste materials are not considered a burden and unnecessarily incinerated or landfilled when more value could be obtained from them.

The VinyLoop^® process took PVC waste streams, often contaminated with textiles and other materials, and selectively dissolved the PVC in an organic solvent. The PVC was then precipitated by steam-driven evaporation of the solvent which itself was recycled. The PVC was said to be of the same quality as the original material. VinyLoop^® was a Solvay, Belgium, technology commercialised as a joint venture in 2002 and ran until 2018 (39). The plant in Ferrara, Italy was established to recycle up to 10,000 tonnes of waste a year, primarily cable insulation (40). In 2008 the plant was updated to treat textile composites as well.

Methyl ethyl ketone (MEK) is a good PVC solvent and in the VinyLoop^® process was used with the cosolvent n-hexane (Figure 3) (41). In a typical example of the process, 9.3 kg of 82% MEK, 5% water, 13% hexane was added for every kilogram of PVC. After mixing at 100°C (2.8 bar) for 10 min, a dispersant was added (0.2% relative to PVC of METHOCEL^TM K100, a cellulose ether). The dispersion agent was needed to make fine particles of PVC. Then the temperature and pressure were reduced and steam injected (3.6 kg per kilogram of PVC). The addition of water allowed evaporation of a MEK-water azeotrope. Precipitation of PVC occurred at 64–65°C, below the boiling point of the azeotrope. Over 99% of the recovered PVC was able to pass through a 1 mm sieve. The water-MEK-n-hexane mixture was also collected. The presence of n-hexane improved the separation of the organic phase from water for reuse. An earlier patent describes the addition to salts to achieve the same effect (42).

Fig. 3.

The PVC waste being processed had been plasticised into flexible products. The VinyLoop^® process maintained the additive composition of the PVC, which in theory may be advantageous for closed-loop recycling, but in practice the ability to introduce new additives to create new products for contemporary markets and meet changing regulatory requirements would have been preferable. It was the latter that caused the closure of the VinyLoop^® plant. Phthalate esters are used extensively to plasticise PVC. The toxicity of phthalate esters has prompted action by the European Chemicals Agency (ECHA), resulting in a ban on many phthalates, including bis(2-ethylhexyl) phthalate, since 2015 (43). The European Registration, Evaluation, Authorisation and Restriction of Chemicals (REACH) regulation that dictates the nature of bans or restrictions on chemical use requires any company producing, importing, using or isolating bis(2-ethylhexyl) phthalate (present above 0.1%) to have obtained authorisation to do so (44). Recycling of materials containing substances subject to authorisation is also within scope of the REACH regulation. How much this is appreciated, adhered to and policed in Europe is a subject of interesting debate with significant consequences. The operators of VinyLoop^® did have authorisation (45), but these permits are time limited, in this case less than three years. As the expiry date of the authorisation drew near, the recycling plant was closed. The expectation is that most companies will stop handling the banned substances and find alternatives where possible because authorisation is very expensive to obtain. For a recycler, they are subject to the nature of waste produced by others, including legacy materials and plastics produced by manufacturers with authorisation to include otherwise banned plasticisers. A further complication is that medical products are exempt from the plasticiser ban and so there is the possibility of materials containing bis(2-ethylhexyl) phthalate still entering recycling streams.

This case study raises some important questions. How much PVC currently in use contains banned plasticisers? Many articles such as the cable insulation that was recycled by VinyLoop^® has a long lifetime and was made before the EU phthalate bans were implemented. Can solvents remove additives in a compliant way? This question can be addressed by studying the solubility of PVC and phthalate esters. bis(2-Ethylhexyl) phthalate is a liquid expected to be miscible with a large number of organic solvents. Techniques for phthalate determination use solvent extraction methods, albeit on a small analytical scale. Therefore it is logical to add a pre-step to future recycling methods where phthalates (if present) are extracted by swelling but not dissolving the PVC or by dissolving both polymer and additives but later selectively precipitating the PVC. Distillation, as practiced by VinyLoop^®, leaves non-volatile components unseparated (i.e. PVC and bis(2-ethylhexyl) phthalate). A condition of handling substances subject to authorisation in Europe is not to isolate or store refined batches of the chemical(s) in question without a permit. For recycling this is an issue as what can be considered an impurity cannot be removed without destroying it in situ. Incineration or chemical transformation may be suitable and legal approaches.

If a process were to be developed that could remove additives in a compliant way, the cosolvents MEK and n-hexane may no longer be the ideal combination for PVC recycling. This creates scope to reduce the hazards posed by n-hexane in particular. In solvent selection it is important to know what solvents are restricted or subject to authorisation by REACH of course. Recently some ether and chlorinated solvents have been subjected to authorisation and a large number of restrictions on how many others can be used are also in place (22, 33).

The polyolefins PE (high and low density grades) and PP are produced in greater quantities than any other synthetic plastics. As for PET, mechanical recycling is viable because of the availability of the waste and the quality of the recyclate is appropriate for large markets. However, the high calorific content of these hydrocarbons means they are favoured as a feedstock for energy recovery plants (46). Plastic pyrolysis to make oils suitable for refining into fuels and base chemicals is being investigated as a more flexible alternative to incineration (47). The technology is proven on a multi-tonne scale (48, 49). BASF has now used pyrolysis oils made from waste plastic to feed the steam cracker at its primary chemical production plant (50). This indicates there is tangible interest in diversifying the uses of waste polyolefins.

It is also feasible to recover polyolefins from solution. Pappa et al. found xylene at 85°C dissolves PE but not PP (51). The undissolved PP could be removed by filtration and then the PE precipitated with an antisolvent (propanol). Recovery on a 3 kg scale was greater than 99% (Figure 4). The authors report no loss in performance attributes of the recovered polymers and actually an increase in crystallinity. This is unusual compared to the previous case studies (12, 21). Other research also reports that the elastic modulus of PE and PP increases while other properties are the same or slightly improved after solvent-based recycling (52). One explanation is that while recovery is high, the small losses probably represent the more soluble lower molecular weight polymers with less desirable properties.

Fig. 4.

Solvent-based recycling can offer a major advantage when it is used for separation of wastes. Extraction of polymers from mixed waste streams with selective solubility has been known for decades (53), but it is not cost competitive with flotation and near-IR sorting. However, multilayer materials cannot be separated effectively with current technology. This must be considered as a design flaw in a circular economy, which if impossible to resolve by product designers must be addressed by recyclers. Multilayer packaging typically contains a film of aluminium and a number of plastic layers, including PE sealing layers. The use of switchable-polarity solvents can delaminate these materials by dissolving the PE (54, 55). The principle of a switchable-polarity solvent is based on a hydrophobic amine that is converted into an ammonium bicarbonate solution with the addition of water and carbon dioxide (Figure 5) (56). The resultant hydrophilic antisolvent precipitates the PE. Releasing the carbon dioxide pressure then reforms the original amine ready for reuse.

Fig. 5.

Current policies and investment for waste collection, separation and recycling limit the circularity of materials. Product design, consumer choices and conventional business models also share the blame. Despite academic interest in novel polymers designed to self-heal, rapidly biodegrade or depolymerise on command, they are met with resistance by established petrochemical plastic markets. The major reason is that new, synthetically complex products will be more expensive. The introduction of new plastic materials also increases the complexity of the plastic waste market and that is generally unhelpful for recycling practices. Recycling rejection rates are overall already increasing in the UK, now standing at over 4% of post-consumer material collected from households (57). At end of life, small volume plastics are contamination in PET, PE and PP recycling streams, which increases the likelihood that waste is not returned to use because of the low quality of the recyclate. We see this in the recycling of PET, where the presence of PVC at 100 ppm can cause discolouration and degradation of the recyclate (58). Solvent extraction makes it possible to remove PVC from PET (59), in the same way that it might become necessary to remove PLA from PET waste in the future (27).

The potential of polystyrene recycling is also high (60), but recycling rates of consumer waste are low due to the very few districts willing to collect it. Significant barriers to polystyrene recycling include its smaller market size compared to the other major plastics and its low density. Expanded polystyrene is uneconomical to collect, transport and sort for this reason. A number of solvent-based approaches have been proposed to dissolve and densify polystyrene, which in turn could make recycling more economical. Limonene is an effective solvent (61, 62) and Ran et al. recently reported the use of binary solvent systems to dissolve polystyrene (63). The use of switchable-polarity solvents is also known for this purpose (64), but no commercial plants are operational at this time.

The potential for solvent-based recycling to make a significant contribution to a circular economy depends on willingness to invest in end of life processes that recycle difficult waste streams. Start-up and maintenance costs are certainly higher than a conventional recycling plant. There is a social benefit to recycling composites and layered materials that relates to the avoidance of litter, including topical concerns about ocean pollution and microplastics. Waste management of electrical and electronic equipment is infamous for exports to Africa exploiting vulnerable people and exposing them to toxic substances (65). The Basel Convention now makes this practice illegal. With responsibility now placed on treating this waste domestically, research has shown solvents assist the separation and recovery of the complex and valuable components found in these articles (66–69). Removing or at the very least monitoring additives will become hugely important to the recycling industry. Addressing brominated flame retardants is a key step in the reprocessing of electrical and electronic equipment (70, 71). Solvent-based recycling processes have been shown to successfully remove brominated flame retardants from plastics by firstly dissolving the waste and then adding a second solvent to selectively precipitate the polymers (72, 73). Ultimately the possibility of future feedstock shortages and subsequent price increases, coupled with countries’ refusal to accept foreign waste (74), will demand a change to recycling practices beyond simply increasing the capacity of conventional processes. Whether this will occur in the short term or many decades from now depends on the prioritisation of a circular economy in the ambitions of world leaders.

Process intensification (PI) is a design framework which aims to create smaller, safer and more efficient processes. There have been many reviews on process intensification and attempts to define it since its inception over 20 years ago. Process intensification approaches often involve the reduction in the size of the process units to increase heat and mass transfer rates and, in multiphase processes, to manipulate and control the flow patterns and increase the interfacial areas.

The benefits of operating in small scale units stem from the thin fluidic films and the decreased diffusion distances, which increase the heat and mass transfer rates resulting in homogeneous concentration and temperature fields. Residence times can be shortened, thus avoiding side reactions, increasing selectivity and reducing waste. The decrease in length scales also increases the importance of surface or interfacial forces over inertial, viscous and gravitational ones; as a result, flow patterns in two-phase systems tend to be regular. Channel walls with different wettabilities can be fabricated to separate two-phase mixtures or impose certain patterns in the channels. The large interfacial area-to-volume ratios benefit mass transfer, while the large channel surface-to-volume ratios improve heat transfer. Flows are laminar in many cases and, combined with the regular flow patterns, can be modelled more easily. The small volumes reduce the risks of handling hazardous materials, while accidents are better contained. Intensification is often linked to continuous flow processing. The homogeneous conditions in channels facilitate monitoring and allow modularity where the process steps are separated by controlling, for example, the temperature or the addition of reactants along the channel.

Processes involving two immiscible liquids are widespread industrially and intensification has already been shown to benefit emulsifications (1–3) and reactions including hydrogen peroxide oxidations and (trans-)esterifications (4, 5). Among two-phase liquid processes, extractions are commonly used for the separation of materials in, among others, the pharmaceutical (proteins, antibiotics in aqueous two-phase systems), energy (uranium in nuclear spent fuel reprocessing; carbon dioxide or hydrogen sulfide removal) and mining (copper and precious metals) sectors (6, 7). Industrially, extractions are carried out in mixer settler units or pulsed columns which suffer from inhomogeneous and not well-characterised flow fields and large inventories. Intensified approaches have already been applied in the extraction of bio-based chemical precursors (8, 9), transition metals (10) including platinum group metals (11), lanthanides (12, 13) and actinides (14, 15), acetone in toluene-water systems (16–18). Because of the improved efficiency and reduced volumes of the small units, the amount of solvent required is reduced; this paves the way for the use of novel, efficient but sometimes expensive solvents (such as ionic liquids). In addition, external fields such as centrifugal, magnetic and ultrasonic can easily be applied to improve mixing, separation or reaction rates. The majority of the intensified demonstrations are in single channels. For industrial applications it is necessary to increase throughput by increasing the number of channels, which is presently a major challenge.

In what follows, liquid-liquid extractions in intensified small-scale contactors are reviewed. These include extractions in single channels and in confined impinging jets cells as well as approaches to increase throughput via scale out, where many parallel channels are combined with appropriate manifolds. Developments on the use of centrifugal forces to enhance separations in small channels in CCC systems are discussed. The combination of intensified technologies with novel ionic liquid solvents is also considered.

When two immiscible liquids flow together in small channels, many flow patterns can form, ranging from segmented to annular and dispersed flows, as can be seen in the example in Figure 1. Parallel flows, where the two liquids flow in continuous layers next to each other, can also occur, usually by modifying either the wetting properties or the geometry of the channel walls.

Fig. 1.

The segmented, plug or slug flow pattern has been extensively studied because it appears for a wide range of phase flow rates and has been linked to high mass transfer rates. In this pattern, the dispersed phase moves as drops with size larger than the channel diameter (plugs) separated by slugs of the continuous phase (see Figure 1). Usually, there is a thin film of the continuous phase between the plugs and the channel wall. As the film is usually very thin, axial dispersion is limited. In addition, within each phase, circulation patterns are established which improve radial mixing (see Figure 2). As a result, a plug-flow reactor configuration establishes with improved radial and decreased axial mixing that ensures uniform residence times for the reactants.

Fig. 2.

In the small channel contactors, the flow characteristics and the mass transfer performance are closely related. The plug and slug lengths determine not only the interfacial area but also the mixing characteristics within the two phases. Apart from the flow rate ratio and the properties of the two phases, the plug and slug lengths depend significantly on the geometry of the inlet. Plug lengths have been reported by many investigators (19, 20) but there is no single model to predict them a priori.

Interfacial areas can be calculated from the measured plug and slug lengths and shapes of the front and back ends of the plugs. It has been found that the specific interfacial area (interfacial area per unit volume of the contactor) depends on the channel diameter, the flow rate ratio, the total velocity and the inlet geometry. Interfacial areas ranging from 2760 m² m^–3 to 4800 m² m^–3 in 0.5 mm, 0.75 mm and 1 mm internal diameter (ID) channels have been reported by Kashid et al. (21), whilst Li and Angeli (13) measured specific interfacial areas up to 8500 m² m^–3 for smaller channels of 0.2 mm ID, using high-speed imaging. In larger channels, with a diameter of 4 mm, the specific interfacial area decreased to values up to 880 m² m^–3. The circulation times within the plugs or slugs can be calculated when the velocity fields in the phases are known. Velocity field measurements have been carried out with particle image velocimetry (PIV) (22) or predicted from computational fluid dynamics (CFD) simulations (23). The results have shown that mixing is improved as the velocity increases and the plug and slug lengths decrease.

Segmented flow contactors have been used for many liquid-liquid mass transfer and reaction operations. In the case of fast reactions, the overall rate of the process is primarily controlled by the rate of mass transfer and microreactors have been shown to intensify processes. A typical example is the transesterification reaction of vegetable oils to produce biodiesel, where yields over 90% can be achieved under 30 s residence time in 240 μm hydraulic diameter channels (24). Regarding liquid-liquid metal extractions, combined with ionic liquids as the solvent phase, microchannels have been applied to the separation of uranium (25) and europium (13) for spent nuclear fuel reprocessing and for analysis in nuclear waste management. Channel sizes between 0.2 mm and 2 mm were used and extraction efficiencies >80% were achieved in <30 s. Pedersen et al. (26) achieved separation >90% of titanium-45 for use in positron imaging in less than 15 s in 0.75 mm perfluoroalkoxy (PFA) tubing.

The overall volumetric mass transfer coefficient for a solute being transferred from the aqueous to the organic phase is given by Equation (i) (15):

(i)

where ɛ_aq is the volume fraction of the aqueous phase, C_aq,eq is the concentration of the solute in the aqueous phase at equilibrium, C_aq,init is the concentration of the solute in the aqueous phase at residence time T₁ and C_aq,fin is the concentration of the solute in the aqueous phase at residence time T₂.

High-value precious metals such as platinum and palladium have also been extracted using intensified contactors. Yin et al. (11) and Kriel et al. (27, 28) extracted high-value metals (Pt, Pd) using parallel flow contactors. These metals are often found in mixtures at low concentrations (for example, from spent automotive catalysts) and their extraction may not be economic using conventional devices. In particular, Kriel et al. (27) demonstrated the extraction, scrub and stripping processes in flow channels with overall recovery rates over 95%. By modelling the flowsheet of spent nuclear fuel reprocessing using intensified extractors for the first time, Bascone et al. (29) showed important reductions in solvent use and in radiolytic degradation.

To increase throughput, large channel sizes should be considered that still preserve the benefits of small-scale operations, such as thin fluid films and enhanced heat and mass transfer rates. The extraction efficiencies and the volumetric mass transfer coefficients for channels between 0.5 mm and 4 mm ID were measured by Tsaoulidis and Angeli (25) and k_Lα as high as 0.06 s^–1 were found, even at the largest channels. The k_Lα increased as the channel diameter decreased. In smaller channels, however, the pressure drop increased and the throughput decreased. There is, therefore, a trade-off between mass transfer performance, throughput and energy requirements which needs to be carefully considered when designing plug flow separators.

To integrate the small channel extractors with the rest of the process, the separation of the two phases at the end of the channel should be considered. In small channels, wettability and interfacial effects are important and have been successfully implemented for the separation of the organic and aqueous phases. Separators with side channels or membranes that are preferentially wetted by the organic or aqueous phases have been tested. Currently, however, there are few commercially available options (for example, Zaiput Flow Technologies, USA); for parallel processing with many channels (see Section 4 below) the separator capital costs would scale linearly with the number of channels. Alternatively, gravity separators can still be used for systems with high throughput and sufficient density difference (>0.1 g cm^–3, (30)). In gravity settlers, however, the mass transfer between the phases can continue and diminishes the benefits of well-controlled conditions of the small channel.

An alternative option for increasing throughput in small channels is to increase the velocities of the two fluidic streams in an impinging jets inlet configuration. In recent years, there has been a renewed interest in using confined impinging jets reactors (CIJR) for many applications, such as crystallisation (31), nanoparticle synthesis using liquid precipitation (32), micromixing (33), extraction (15) and bioreactions (34). The high energy dissipation rates due to collision and redirection of the fluidic jets in the impingement zone make the contactor particularly suited for applications where rapid mixing of the fluids is necessary. When the two jets are immiscible liquids, then the large energy dissipation rates result in the formation of dispersions. A typical configuration of a cylindrical CIJR at 180° nozzle angle is shown in Figure 3(a).

Fig. 3.

The mixing and dispersed phase size are affected by a number of geometric characteristics, such as main channel (D) and nozzle (d_j ) size, main channel to nozzle size ratio, inter-nozzle distance (Id), nozzle height and impingement angle. The main challenge in developing confined impinging jets contactors for a particular application is the quantification of the effects of the parameters on the resulting drop sizes. There are many studies on impinging jets with miscible liquids, which demonstrate that for improved mixing the two opposing jets should have similar momentum so that they collide in the middle plane (35). Studies of impinging jets in confined spaces with immiscible liquids are very limited. The drop size has been related to the energy dissipation rate, while the uniformity of the dispersions depends on the geometric design of the contactor, the phase ratio and the intensity of mixing in the impingement zone (36). The energy dissipation rate (ɛ) can be described as the ratio of the power available due to kinetic energy change (K) at collision over the mixing volume of the impingement zone (V_iz ), according to Equation (ii):

(ii)

where

(iii)

and and are the mass flow rates (kg s^–1) of Phase 1 and Phase 2 respectively, u₁ and u₂ are the average velocities of Phase 1 and Phase 2, respectively, and ρ is the density of the mixture.

The Sauter mean diameter, given by Equation (iv):

(iv)

(where n is the number of drops and d_i is the diameter of the drop i in the distribution) has been related to the specific energy dissipation rate as follows (37), Equation (v):

(v)

The dependence of the average drop size on the energy dissipation in the impingement zone is presented in Figure 4 for two different aqueous/organic phase systems (15). As can be seen, at low energy dissipation values, the drop size depends on the geometry of the system and the fluid properties, while at larger ɛ, above 600 W kg^–1, the drop sizes converge. Similar results were also found by Siddiqui (2) for aqueous/organic systems in impinging jets contactors with the addition of emulsifiers, while in less viscous systems a stronger dependence of drop size on energy dissipation rate was observed.

Fig. 4.

It has been found that the dispersions formed in impinging jets have low polydispersity. Tsaoulidis and Angeli (36) reported polydispersity indices (PdI) as low as 0.05, for an oil/water system, for a wide range of jet velocities from 0.17 m s^–1 to 6.2 m s^–1. Interfacial area-to-volume ratios were significantly affected by the velocities of the jets and the values varied between 2000 m² m^–3 and 12,000 m² m^–3. Siddiqui (2, 3) also reported very narrow drop size distributions in a sunflower oil/water emulsification process with surfactants, for dispersed phase fractions up to 10% and drop sizes less than 10 μm.

The few mass transfer studies available have also revealed high mass transfer coefficients compared to other contactors as can be seen in Table I. k_Lα can be one to two orders of magnitude higher than in conventional contactors and two to three times higher than in microchannels. Values are similar to those of centrifugal contactors, however, the specific power input for centrifugal contactors can be two to three orders of magnitude higher when compared to a confined impinging jets cell (16).

The studies revealed that high mass transfer coefficients were obtained at short residence times (<4 s), with values up to 1 s^–1. Several parameters were found to affect mass transfer, including geometric characteristics and flow rate ratio. In Figure 5(a), mass transfer rates were calculated for two different main channel sizes i.e. 2 mm and 3.2 mm. It is shown that k_Lα depend on channel size at short residence times but are independent of the channel sizes at long times. The 3.2 mm channel should then be preferred because it has higher throughput and reduced pressure drop compared to the 2 mm channel. The mass transfer coefficient also increases with increasing collision velocities of the two jets (shown as the sum of the two velocities, u_tot in Figure 5(b)). The velocity of the liquid jets will define the position of the point of impingement (Figure 6) and will affect the uniformity of the flow pattern in the main channel as well as the drop size distribution.

Fig. 5.

Fig. 6.

Impinging jets systems have been used for more than four decades now but it is still not possible to determine accurately the effects of the dominant variables including aspect ratio of jet inlets, dead volume at the mixing section, fluid properties on the drop size and the mass transfer performance of the contactors.

Harmsen (41) identified four hurdles that any PI innovation must address before it can be implemented industrially. These are: (a) risk of failure by combining novel aspects; (b) scale-up knowledge uncertainty; (c) equipment unreliability; and (d) improved safety, health and environmental risks. The main hurdle faced by single-channel contactors is the scale-up uncertainty. Currently, there are no applications of small-channel extractors at large commercial or pre-commercial scales reported in the literature.

While single small channel flow contactors and reactors are highly efficient, the scale-up of the throughput without losing the small-scale advantages remains a main challenge. Scale-up can be achieved by increasing the number of channels operating in parallel (‘scale out’ or ‘number-up’). It is not trivial, however, to reproduce accurately the flow conditions of a single channel in many parallel ones. The challenge is to design a flow distributor within the process-specific maldistribution tolerance of the total flow rate and of the flow rate ratio of the two phases for each channel.

Scale out for single-phase processes requires a flow distributor that can achieve almost the same flow rate and thus residence time, in all channels. Single-phase flow distributors are commonly encountered in multi-tubular reactors, catalytic converters and other honeycomb catalysts. The distributors usually take one of two forms, bifurcation or consecutive manifolds, as shown in Figure 7. The consecutive manifold has a small footprint compared to the bifurcation one. Approaches based on resistance networks have been used to design manifolds that reduce flow non-uniformities among the channels (42).

Fig. 7.

In the case of multiphase systems, both the residence time and the flow rate ratio of the two phases are critical for the performance of the process and should be constant among the many channels of the manifold. There are two types of two-phase flow distributors: split-combine and combine-split (Figure 8). Combine-split designs first bring the two fluids in contact and then distribute the two-phase mixture into as many channels as necessary. Combine-split distributors do not have dead volumes and have been successfully designed by Hoang et al. (43) at a chip-scale; the design has been tested for up to eight channels but it may be unfeasible to increase further the number of channels.

Fig. 8.

Split-combine distributors (or double manifolds) distribute each of the single-phase fluids independently into as many channels as necessary and then bring them in contact. This type of distributor can, in turn, be either a bifurcation or a consecutive manifold.

Maldistribution of gas-liquid flows in double manifolds and the effects of manufacturing tolerances were studied experimentally by Al-Rawashdeh et al. (44) for square main channels with 1 mm side. The authors achieved deviations in the residence time below 20% by controlling the pressure drop in the distribution sections. Garciadiego Ortega et al. (45) developed a method to analyse the two-phase flow maldistribution and used a resistance network model to simulate the double manifolds. The effect of the number of channels on the maldistribution was also studied and scaling-laws for the design of these distributors were proposed as well as a procedure for an effective and economic double manifold design. The first step is to define the single-channel size and flow rates for a particular process and define the sensitivity of its performance to flow maldistribution. This defines the maldistribution tolerances, which determine the dimensions of the flow distributor. Finally, the pumping requirements are calculated. There is a trade-off between pumping requirements and maldistribution, with low maldistribution tolerances resulting in high pumping costs.

A promising intensified separation technology is high-performance CCC. It is a form of liquid-liquid extraction that achieves separation by repeated partitioning of solutes between two immiscible liquid phases, as they interact in a continuous length of coiled tubing under centrifugal and Archimedean forces. The tubing is wrapped around a cylindrical drum (called a bobbin) to form typically a three-dimensional (3D) helical configuration with one or several layers. Within the CCC column, one of the liquid phases is held stationary by a combination of hydrodynamic and hydrostatic forces generated as a result of rotating the column in planetary motion, while the other mobile phase is continuously pumped through the coil and serves to transport the solutes through the system. The J-type CCC is the most commonly used, where the bobbin is mounted on a planetary axis, driven by a central axis so that the column rotates about its own axis while it revolves around the central axis at the same velocity in the same direction (Figure 9(a)). The double rotation of the column during its planetary motion produces a variable centrifugal force field. This force field creates a unique mixing pattern in which a series of sequential mixing and settling zones are generated simultaneously along the length of the column. These alternating mixing and settling steps are essential to the chromatographic process as they promote solute transfer between the phases and therefore, separation of species with different partition coefficients.

Fig. 9.

In the last decade, the use of ionic liquids either as solvents or additives in liquid-liquid extractions has expanded considerably because of their unique properties. Ionic liquids are organic salts that are liquid at room temperature. The stability, phase behaviour and greater solvating power of ionic liquids, together with the ability to design their structure, can increase both the flexibility and performance of separations and allow separations that were not previously considered possible. The combination of the two technologies, ionic liquids and CCC, therefore represents an exciting approach to intensified liquid-liquid separations. However, the use of ionic liquids in CCC is not a trivial task due to their relatively high viscosities, which can introduce significant problems for the majority of traditional CCC machines that are mostly low pressure. To overcome the pressure limitations previously encountered using the CCC technique, AECS-QuikPrep^TM Ltd, UK in collaboration with the QUILL Research Centre have reported on the design and construction of a modified high backpressure CCC instrument (Figure 9(b)). The high solvating power of ionic liquids allows separations to be run at very high sample loadings which gives rise to high space-time yields for ionic liquid and CCC separations. The ability to custom design ionic liquids allows a greater range of mobile phases to be employed and enables separations with pH neutral water as the mobile phase, where previously toxic organic solvents (such as acetonitrile), concentrated salt solutions (such as aqueous dipotassium phosphate), polymers (such as polyethylene glycol) or acids (such as nitric acid) were used. Scale up can be achieved with the increased capacity CPC instrument (46). Ionic liquids have been applied successfully as a major solvent system component for a wide range of separations (47–49) including:

The combination of ionic liquids with CCC has been successfully used in the separation of the anticancer drug lentinan at a scale 10–100 times the scales of earlier separations (50). Lentinan is found in shiitake mushrooms (Lentinus edodes) and is used as an adjunct to therapy in combination with chemotherapeutic drugs such as fluorouracil to modulate the body’s immune system activity. Lentinan naturally exists in water and salt solutions but is easily denatured by solvents. This means that for the isolation and purification process of lentinan, water based solvent systems are required. The conventional purification of lentinan normally involves up to 10 steps. An ionic liquid-based aqueous biphasic solvent system (ABSS) was developed using 1-n- butyl-3-methylimidazolium salts [C₄mim]Cl / 2.5 M K₂[HPO₄]_(aq) (1:1) mixture (51), which allowed lentinan separations on the 1–3 g scale, without denaturing the lentinan. This CCC process used aqueous [C₄mim]Cl as the mobile phase and the lentinan was separated from the [C₄mim]Cl solution by the addition of ethanol to the [C₄mim]Cl phase. The [C₄mim]Cl can be recovered and reused after the lentinan has been precipitated. The ethanol can also be recovered by evaporation allowing it to be reused. This leads to a separation process that does not consume solvents or reagents.

An improved lentinan process has also been developed with a novel ABSS based on microemulsions. Surface active ionic liquids such as 1-dodecyl-3-methylimidazolium di(iso-octyl)phosphinate ([C₁₂mim][DiIOP]), when mixed with water and hexane produce a water immiscible microemulsion phase, which contains 75 mol% water (Figure 10). The aqueous phase is composed of >99% water, which allows water to be used as the mobile phase in CCC separations, with the microemulsion as the stationary phase. This greatly simplifies product isolation since the product does not end up mixed with large quantities of involatile chromatography solvent constituents. Also, this approach does not produce any solvent waste (other than water) making this a very green and inexpensive separation to run. The full lentinan process takes the freeze dried hot water extract of shiitake mushrooms (Figure 11(a)) and precipitates lentinan from this crude extract dissolved in [C₄mim]Cl, using ethanol. The precipitated 80% lentinan (Figure 11(b)) is then purified with the water-microemulsion solvent system shown in Figure 10 to give the off-white 95% lentinan shown in Figure 11(c) on the 25 g per run scale. Recent industrial uses of CCC and CPC instruments are in the refining of galantamine from daffodils (for example, BioExtractions (Wales) Ltd, UK) or the red spider lily (52), the purification of cannabinoids and metal ion separations associated with the nuclear industry.

Fig. 10.

Fig. 11.

Liquid-liquid extractions are widely used for the separation and purification of many compounds. Small channels (up to 4 mm in diameter) and a combination with external fields, such as centrifugal forces, can significantly intensify the process by reducing residence times, improving extraction and extraction efficiencies and reducing the amount of solvent required. Mass transfer coefficients up to 1 s^–1 have been measured in the impinging jet contactors. These characteristics have made possible the implementation of novel and often expensive solvents such as ionic liquids with significant improvements to the separation. Droplet-based flows (dispersed or plug flow patterns) in particular have been shown to enhance mass transfer and increase interfacial areas. However, the throughputs are small and scale out would be required before they can be applied to industry. On the other hand, the fast mass transfer rates and well-characterised flow patterns render small channels suitable for analysis and for research on new extractants. Impinging jets have increased throughputs and can produce dispersions with narrow size distribution and large interfacial areas. CCC devices with alternating mixing and settling steps allow separation of species with different partition coefficients and have been used to optimise solvent systems and conditions for separations. High throughputs can be achieved with the increased capacity CPC which has simpler rotor design and fewer moving parts compared to CCC (46, 53).

At small scales, the contactor geometry significantly affects the flow and mass transfer characteristics. Possibilities are open for novel contactor designs that exploit interfacial and wettability effects to establish desirable flow patterns and enhance mass transfer. For the commercial application of the technology in production, robust scale-out designs for two-phase systems need to be further developed and the sensitivity of their performance against flow maldistribution needs to be tested.

“Process Systems Engineering for Pharmaceutical Manufacturing” is an ambitious reference comprising 24 chapters covering process systems engineering (PSE) methods and case studies of interest to engineers working in pharmaceutical process development, model development, process simulation, process optimisation and supply-chain or enterprise optimisation. Business model optimisation, including optimisation of clinical trials and supply chain, are topics covered in Chapters 1 and 21–24. Continuous manufacturing of drug product (downstream) is a key theme covered in Chapters 6 and 16–20, while process control, flowsheet modelling and key unit operation modelling are covered in Chapters 5, 7, 8–11 and 13–15. Of particular interest is the topic of small molecule upstream development and workup solvent selection and optimisation discussed in Chapters 3–4, with case studies involving separation solvent selection presented for ibuprofen, artemisinin and diphenhydramine in Chapter 4.

Chapter 2, ‘The Development of a Pharmaceutical Oral Solid Dosage Forms’ submitted by Rahamatullah Shaikh, Dónal P. O’Brien, Denise M. Croker and Gavin M. Walker (University of Limerick, Ireland), provides a summary of solid oral dosage form development, covering solubility and dissolution kinetics, pKa, excipient types and the standard formulation processes of direct compression as well as wet and dry granulation and capsule filling. This chapter is recommended reading for anyone not familiar with formulation of drug tablets as it provides a well-organised summary helpful in understanding the types of processes modelled in the chapters on continuous manufacturing, flowsheet and unit operation modelling as it relates to drug product.

I have organised this review according to general topics covered rather than by sequential order of the chapters.

Chapter 1, ‘New Product Development and Supply Chains in the Pharmaceutical Industry’, contributed by Catherine Azzaro-Pantel (Université de Toulouse, France), introduces the pharmaceutical supply chain and summarises the product life cycle of a drug starting from discovery through clinical trials, registration and commercialisation. This chapter provides a concise summary of clinal trial phases, pre-launch and launch activities and is recommended reading for those not familiar with the pharmaceutical business model and drug development process (Figure 1).

Fig. 1.

Chapter 21, contributed by Brianna Christian and Selen Cremaschi (Auburn University, USA), covers ‘Planning of Pharmaceutical Clinical Trials Under Outcome Uncertainty’. The authors reference an increase in attrition rates in clinical trials and state “the time from discovery to product launch of a drug is around 10–15 years with an average research and development (R&D) cost of about $2.6 billion per drug” as motivating factors driving the need for better clinical trial optimisation. This chapter provides details of a “perfect information” deterministic mixed-integer linear programming model (MILP) problem including constraints. By using an innovative heuristic modification to the stochastic programming model a five order of magnitude improvement is reported.

Chapter 22, ‘Integrated Production Planning and Inventory Management in a Multinational Pharmaceutical Supply Chain’ contributed by Naresh Susarla and Iftekhar A. Karimi (National University of Singapore) presents a MILP model for a complex supply chain and provides a strategy to optimise inventory, resources and production schedules in the supply chain to maximise profit. The intent of the model is as a tool for decision making for “production planning and scenario analysis in a multinational pharmaceutical enterprise”. To mitigate risk associated with the complex, multinational network of supply, drug inventories of 180 days are not atypical. However, high levels of inventory come at a cost. A change introduced in the supply network may have impact on inventories, lead-times and dependencies as impacted by other portions of the supply network. In this chapter the authors describe their approach to this optimisation problem. While looking at the authors’ formulation of their case study problem, the value of working with fewer strategic suppliers in a vertically integrated supply network is evident in that it will minimise the complexity, delay and cost associated with a complex network. A takeaway from this chapter is that pharmaceutical companies can anticipate improved access to software tools to compare the impact of supply chain alternatives as research is translated into commercial software offerings.

Chapter 12, ‘PAT for Pharmaceutical Manufacturing Process Involving Solid Dosages Forms’ contributed by Andrés D. Román-Ospino and Ravendra Singh (Rutgers, The State University of New Jersey, USA), Vanessa Cárdenas and Carlos Ortega-Zuñiga (University of Puerto Rico, USA), presents near-infrared (NIR) calibration models and chemometrics. For those not skilled in process analytical technology (PAT) and analytical determination, this chapter is very informative and provides comparison of various methods for analytical data fitting to determine blend uniformity for real-time control of continuous pharmaceutical processes. Principal component analysis (PCA), partial least squares (PLS) and multivariate curve resolution alternating least squares (MCR-ALS) are presented as suitable techniques for multiple parameter determination where linear regression or classical least squares methods are not suitable. Layering of talc and lactose as a specific case study in non-homogeneity is discussed in this chapter. Finally, a process example utilising Unscrambler^® X Process Pulse II (Camo Analytics AS, Norway) and NIR (Viavi Solutions Inc, USA) is presented where Unscrambler^® X software is utilised to generate and upload a calibration model generated via methods presented in the chapter. In the example the NIR data processing system is interfaced to a DeltaV^TM distributed control system (Emerson Electric Co, USA) to provide real time process control of a tableting process.

Chapter 19, ‘Monitoring and Control of a Continuous Tumble Mixer’ contributed by Carlos Velázquez, Miguel Florían and Leonel Quiñones, (University of Puerto Rico, USA), presents a case study for the mixing of naproxen sodium with excipient using a continuous mixer designed by Velázquez. The PAT technology implemented for this case study employed the use of NIR in conjunction with Unscrambler^® X in a PAT implementation similar to that described in Chapter 12. The closed-loop control dynamics for the experimental mixer are evaluated. A finding from the study is that a different control scheme is required for very low dosage active pharmaceutical ingredient (API) vs. higher dosages. The authors identified flowrate control of API addition at very low dosage as variable due to poor powder flow properties as well as limitations of the NIR methods employed in low dosage applications.

Chapter 9, ‘Crystallization Process Monitoring and Control Using Process Analytical Technology’ contributed by Levente L. Simon (Syngenta Crop Protection AG, Switzerland), Elena Simone (University of Leeds, UK) and Kaoutar Abbou Oucherif (Eli Lilly and Co, USA), introduces quality by design (QbD) and reviews online analytical techniques available for crystallisation monitoring and control which include attenuated total reflectance Fourier-transform infrared (ATR-FTIR), Raman spectroscopy, acoustic spectroscopy, conductivity measurement, refractive index measurement, turbidity measurement, focused beam reflectance measurement (FBRM) and particle vision and measurement (PVM).

Automated direct nucleation control (ADNC) along with polymorph determination and control via Raman and attenuated total reflectance ultraviolet (ATR-UV) spectroscopy are presented for batch and continuous crystallisation processes. The ADNC method involves heating and cooling cycles to control crystal count as measured by FBRM to a specified target. In the batch implementation, after initial nucleation, the system automatically heats to dissolve fines and heating and cooling cycles proceed until the crystallisation endpoint (low solution concentration). An advantage of this method is that from PAT data collected, the metastable zone width (MSZW) and solubility curves may be constructed. Since solubility curves are not required prior to running ADNC experiments, this method is useful for process development. An interesting adaptation of the ADNC method to a two-stage continuous mixed-suspension mixed-product removal (MSMPR) crystalliser system is an innovation by Yang et al. (1) where heating and cooling is performed on the jacket of a wet mill while the MSMPR crystalliser is maintained at constant temperature. The MSMPR with wet mill achieves both form control and FBRM particle count control under continuous flow operation.

Chapter 5, ‘Flowsheet Modeling of a Continuous Direct Compression Process’ contributed by Seongkyu Yoon, Shaun Galbraith, Bumjoon Cha and Huolong Liu (The University of Massachusetts Lowell, USA), summarises the scope of individual unit operation models for continuous powder blending, powder feeding (and potency control), tablet press, feed frame and tablet compaction. The authors highlight both a population balance model (PBM) as well as a stirred-tanks-in-series modelling approach to blending. The value of the modelling is in being able to accurately predict the response of perturbations on key quality attributes of finished tablets. An accurate system-wide process model allows implementation of both feedback and feedforward (predictive) control methodologies which can be developed and tested offline, provided that the underlying unit operation models are accurate. Modelling will facilitate development of continuous direct compression (CDC) processes and control schemes for CDC, where elimination of granulation results in simpler, less expensive processes.

Chapter 6, ‘Applications of a Plant-Wide Dynamic Model of an Integrated Continuous Pharmaceutical Plant: Design of the Recycle in the Case of Multiple Impurities’ submitted by Brahim Benyahia (Loughborough University, UK), takes the continuous methodology described in Chapter 5 a step further by integrating the chemical synthesis steps (upstream) with the formulation and tabletting steps (downstream) into a single continuous flowsheet. Of interest is the impact of wash-factor (i.e. wash volumes) and recycle (purge ratio) on the quantity of in-specification product produced. The recycle of wash streams is not often performed in batch API but in continuous processing this recycle provides potential for optimisation and cost savings. The evaluation of wash factors and their limits as potential critical process parameters (CPP) is performed following a model-driven QbD approach. In the case study presented, plant dynamics are compared for both full purge and full recycle purge ratios.

Chapter 11 ‘Process Dynamics and Control of API Manufacturing and Purification Processes’ submitted by Maitraye Sen, Ravendra Singh and Rohit Ramachandran (Rutgers, The State University of New Jersey, USA) introduces a hybrid model predictive control/proportional-integral-derivative (MPC-PID) controller in which a single model based controller coupled with one PID temperature controller replaced four separate PID controllers in a continuous API/pharmaceutical intermediate process comprised of crystallisation, filtration, drying and excipient blending operations. PBM and discrete element method (DEM) methods were utilised to model the process while PCA was used to generate a reduced-order model for use by the model predictive controller. Various control schemes can be tested and optimised entirely in silico allowing investigations of system or controller response to transient conditions and process upsets to be investigated. The authors used MATLAB^® (MathWorks Inc, USA) to fit data resulting from process simulations to transfer functions useful for model predictive control. gPROMS^® (Process Systems Enterprise Ltd, UK) was utilised for PBM calculations and EDEM^® (DEM Solutions Ltd, UK) was used to simulate the mixer where a PCA method was fit to six components from the DEM model.

Chapter 13, ‘Model-Based Control System Design and Evaluation for Continuous Tablet Manufacturing Processes (via Direct Compaction, via Roller Compaction, via Wet Granulation)’ contributed by one of the editors of the volume, Ravendra Singh (Rutgers, The State University of New Jersey, USA), is a review of model-based control for a formulation process which includes blending, granulation, roller compaction, milling and tableting. For the case study in Chapter 13, a PBM is employed in gPROMS^®, but this time for the roller compactor. Unlike the example in Chapter 11, the API crystallisation, isolation and drying steps are not included as API is taken as the input and blended with excipients prior to granulation.

Chapter 7, ‘Advanced Multiphase Hybrid Model Development of Fluidized Bed Wet Granulation Processes’ submitted by Ashutosh Tamrakar, Dheeraj R. Devarampally and Rohit Ramachandran (Rutgers, The State University of New Jersey, USA), implements a hybrid computational fluid dynamics (CFD)/DEM approach to model the coupled behaviour of fluid flow and collisions. The authors transfer data from their CFD-DEM model to a PBM to provide resulting distributions from the granulation process. The DEM-CFD-PBM approach considers residence time in the spray zone, particle collision frequency, aggregation, attrition, particle temperatures and fluid/particle velocities. Residence time in the two zones (spray zone and drying zone) is impacted by fluid flow within the zones and the passing of particles between zones as modelled via CFD-DEM. Results from the CFD-DEM runs are exported to the PBM to investigate sensitivity to inlet gas temperature and gas flow rate. Excellent fit of experimental data from the fluid bed granulator is achieved.

Chapter 15, ‘Advanced Control for the Continuous Dropwise Additive Manufacturing of Pharmaceutical Products’ was contributed by Elçin Içten (Amgen Inc, USA), Gintaras V. Reklaitis and Zoltan K. Nagy (Purdue University, USA). In this chapter the authors describe a system and control methodology for the generation of solid oral dosage forms via a drop on demand (DoD) additive manufacturing technique involving dropwise deposition of API as solvent solution or as solvent/polymer melt (see Figure 2).

Fig. 2.

The DoD system is particularly useful for generation of personalised medicine for highly potent (low dosage) products. The authors present a control scheme based on image analysis of each drop and investigate various cooling profiles for the substrate (tablets). The authors present a polynomial chaos expansion (PCE) surrogate model for prediction of crystallisation, total dosage and product attributes as a function of drop attributes and cooling profile. The PCE model provides a QbD approach for predictive performance of the tablets’ release profile.

Chapters 16–18 present case studies for automation of continuous pharmaceutical process plants where process control is the focus. Chapters 17 and 18 have a bit of redundancy with Chapter 13 as all three chapters are based on a series of published articles by one of the editors of the volume, Ravendra Singh. Chapter 18 is focused on formulation without granulation but with a control scheme to control tablet hardness by controlling tablet press punch depth and real-time measurement of bulk density is used in a feedforward control scheme. Detailed discussion of the control hardware, sensors and control algorithms for the pilot plant is presented in Chapter 17. Process modelling allows complex system dynamics, interactions and control schemes to be investigated and optimised in silico, as enabling technology in the development of robust continuous drug manufacturing processes.

Chapter 3, ‘Innovative Process Development and Production Concepts for Small-Molecule API Manufacturing’, contributed by John M. Woodley (Technical University of Denmark), summarises innovations in process systems engineering used to facilitate process development and optimisation. After a viable process model is developed, ‘virtual experimentation’ may be used to better focus benchtop experiments. Alternative routes and separation schemes can be evaluated if physical property data is available. The CAPEC-PROCESS Industrial Consortium (now the Process and Systems Engineering Centre (PROSYS)) at the Technical University of Denmark has contributed to the generation of physical property estimation methods to address this need.

The author describes use of template processes in which processes under development are fit to a template scheme based on conditions known to work for similar processes. For instance, a reaction step is evaluated against a process template for which simulation and laboratory models already exist (Figure 3). The sufficiency of the template is tested and then the process is optimised using modelling tools already developed for the template process. The author notes that while the template process approach may only be adaptable to 80% of process candidates, for those processes which are adapted, existing knowledge may be leveraged in the development of the new process. Process templating is a powerful tool in the application of PSE models for process integration and intensification and may be useful in evaluating process scheme alternatives when an API synthetic scheme involves multiple transformations.

Fig. 3.

Chapter 4, ‘Plantwide Technoeconomic Analysis and Separation Solvent Selection for Continuous Pharmaceutical Manufacturing: Ibuprofen, Artemisinin, and Diphenhydramine’ contributed by Samir A. Diab, Hikaru G. Jolliffe and Dimitrios I. Gerogiorgis (University of Edinburgh, UK), provides an evaluation of continuous separation steps vs. their batch separation counterparts. The authors noted that for the three continuous API processes evaluated by others, the evaluations had focused on performing the chemistry steps continuously and had not implemented continuous separation steps. As shown in Figure 4, the authors present a continuous liquid-liquid extraction (LLE) separation scheme as a replacement for the batch scheme found in the literature for ibuprofen (IBU). In addition, the authors evaluated additional solvents including n-heptane, cyclohexane, methylcyclohexane and isooctane and found many of the solvent choices to be suitable when a continuous LLE process is used vs. a continuous process. Using process modelling, the efficiencies of separation, the quantities of solvent and an economic comparison of alternative solvents are presented. For a continuous IBU extraction using heptane, the authors project capital savings of 58% and operating savings greater than 50% vs. the batch process utilising diethylether. The case studies presented in this chapter are based on process simulations performed by the authors and not on actual laboratory data. While it does not validate a final solvent choice, the use and conclusions based on simulation data highlight the value of a modelling-based approach to selecting workup or extraction solvents with environmental, flammability and regulatory suitability.

Fig. 4.

“Process Systems Engineering for Pharmaceutical Manufacturing” is a diverse collection of reviews and case studies, most of which were published previously. While this book provides an excellent summary of process modelling and computing with a view to the increased importance of robust simulation tools in pharmaceutical process development and manufacturing, more recent journal publications may provide additional or more in-depth information on the current state of specific technologies or algorithms described in the book. It is also evident that much of the key work in these areas has yet to be done. One topic missing from discussion in the book is the advent of quantum computing and the potential quantum computing presents in solving optimisation problems in process systems engineering. I would look forward to seeing an additional volume added to the series as the technology develops.