CMYK

Scale-Up

How To Minimise

Scale UP Difficulties Dr Trevor Laird

Scale-up of chemical processes, particularly those involving batch or semi-batch manufacture is well-known to be a problematic area of chemistry and chemical engineering, and can be costly when it goes wrong. By correctly choosing and designing the synthetic route to a fine chemical or drug substance, as well as controlling the reaction and work up/product isolation parameters, many of the difficulties in scale up can be avoided. The more complex a process is in terms of chemistry and unit operations, the more there is to go wrong. This article discusses what chemists and engineers can do in advance, both in the laboratory and kilo laboratory, to prevent or at least minimise scaleup issues.

F

or robust and efficient scale-up it is important to choose good syntheses and this is often the decision of the chemist rather than the chemical engineer. One of the failings in university education of chemists is the teaching of organic synthesis from a discovery rather than a manufacturing perspective. In the latter approach, convergent synthetic routes using low cost raw materials, with a minimal waste output, suitable for scale up are designed. For the generic pharmaceutical industry – a major moneyearner in India – it is essential to have the best synthesis (hopefully leading to the best process) to maintain price competitiveness; for the fine chemical, agrochemical, colour chemicals and flavour fragrance industries, this has always been the case, so the best examples of synthetic efficiency are often from these industries. A simple example from the pharmaceutical industry is shown in Scheme 1 for the manufacture of the key tetralone intermediate, used in most routes to make the Pfizer drug Sertraline, which is now off-patent. The original synthesis used for early manufacture compares unfavourably with a later 1-step process

Dr Trevor Laird, Managing Director and Founder of Scientific Update LLP, is an expert in organic process R&D and scale-up of chemical processes and has been an editor-in-chief of the American Chemical Society journal Organic Process Research and Development since its launch in 1996. With over 4 decades of involvement with the chemical and pharmceutical industry, Trevor has consulted for companies in USA, Canada, Australia, South Africa, Singapore, Israel and India, as well as in Western Europe. He is an expert witness on patent cases, and is visiting professor at the University of Sussex, UK. He has authored numerous articles and book chapters. He was educated at London and Sheffield Universities and worked for Imperial Chemical Industries in, and Smith Kline and French, where he was in charge of the Chemical Development Department and of the Pilot Plant. Chemical Industry Digest. July 2010

CMYK

51

Scale-Up from the much cheaper raw material α-naphthol.(6)

the meaning of stoichiometry in semi-batch processes, preferring to think only about the stoichiometry based on the reaction equation – for example the equation might Scheme 1 say that 2 moles of A react with 1 mole of B to give the product C, and O byproducts D. However, if reagent A is O being added to a solution of reaction B 1. succinic anhydride, AlCl3 Cl Cl 1. succinic anhydride, AlCl3 in a stirred tank reactor, the stoichiometry in the reactor varies with Cl Cl 2.2.sodium borohydride sodium borohydride benzene, AlCl3 AlCl benzene, 3 time. The amount of product C formed then cyclise then cyclise along with by products D will depend on the ratio of A to B in the vessel at any one time, and this of course depends on O the rate of addition of A, the rate of reaction of A with B [and this depends OH on the temperature, the solvent, the dichlorobenzene, AlCl3 dichlorobenzene, AlCl3 solvent purity (eg water content), concentration, pH, presence of catalysts 1-step process 1-step process or inhibitors etc.] and especially the Cl mixing in the vessel, which may be scale Cl dependent. Understanding the kinetics of the process, as chemical engineers usually want to do, will assist in the design of the process, allowing the Choosing appropriate conditions correct choice of temperature along To avoid scale up problems, it with the optimal dosing rate for that is important for chemists – with To avoid scale up problems, it particular scale. For an exothermic expert advice from chemical is important for chemists – with reaction, of course, the dosing rate engineers – to choose the correct expert advice from chemical may be limited by the cooling conditions to scale. Many engineers – to choose the correct capacity of the vessel. So it is chemists, however, believe that conditions to scale. important to understand exactly reaction selectivity can only be when the heat is generated in the controlled by operating at low process. temperature, often at minus 78 oC. Apart from the expense of cooling large batches to this temperature, there Hazards of Scale Up is a danger that the increased viscosity of the solvent, When scaling up chemical processes in batch reactors, coupled with the low solubility of reactants/ products/ byproducts at this temperature, leads to reaction media there is always the potential for loss of control if the which are difficult to mix on large scale and resulting in reaction is exothermic, since the change in heat transfer inhomogeneity in the batch with concomitant impurity area per unit volume varies with scale. Whereas a generation. Encrustation of raw materials, intermediates laboratory 0.5L reactor has a heat transfer area of about and products on the vessel walls can also be a problem 0.02m2, a production 3800L vessel has only 10.7m2; thus the heat transfer area per unit volume is 0.04m2/L in the at these low temperatures. lab and only 0.0028 m2/L in the plant, a factor of 7 Whilst some batches occasionally do need such low difference. The consequences of this are increased cycle temperatures, many processes can be operated at higher times and particularly increased addition times for temperatures, provided the reagent is carefully dosed in reagents. The question is usually to the process at the same rate as the whether these changes affect the reaction is proceeding. This Understanding the kinetics of yield and quality of the product – the ensures that the stoichiometry is the process, will assist in the design answer often is yes! constant throughout the addition. of the process, allowing the correct Even more problematic is that if choice of temperature along with reagent addition is too fast Correct dosing regime the optimal dosing rate for that compared to heat removal, particular scale. Chemists are often vague about accumulation can occur and can 52

Chemical Industry Digest. July 2010

Scale-Up

Cl +

150-153oC

Et H2N

Et NH

n

Bu

n

Bu

a re-assessment of whether it is safe to carry out the process at the increased scale.

Companies who themselves have no hazard testing equipAutomate Yellow 96 Dye ment, such as a reaction Scheme 2 calorimeter or differential scanning calorimeter, can sometimes be lead to a runaway reluctant to reaction, especially outsource the if loss of cooling hazard testing to capacity occurs, competent comsimultaneously. panies because of The consequences expense. They may depend on then risk scale up whether decombased on inadeposition of the quate knowledge. reaction mixture From an accountthen occurs ancy viewpoint, (probably with gas destruction of a evolution) and on plant facility with the boiling point of possible loss of the solvent used. (7) life affects the has Stoessel bottom line so defined 5 classes of much more than criticality with Morton International Inc after the explosion and fire, caused due to a capital expendiincreasing hazard runaway reaction that released flammable materials ture on caloripotential in his excellent article in OPRD, and more details of this topic metric equipment, or a small contract for hazard evalucan be found in this recent book (8) . Despite the ation. The spin-off from these studies is a better widespread knowledge about the hazards of scale up, understanding of kinetics and by-product formation ; particularly of compounds with potentially hazardous calorimetric evaluation can usually pay for itself since it groupings, such as nitro compounds, it is disturbing to usually leads to increased yield, quality or productivity. see runaways still occurring due to poor hazard testing, and poor knowledge of the thermochemistry in the Mass transfer issues process. Chemists are often surprised when changes in An example from 1998 at Morton International in selectivity occur on scale up, owing to differences in mass USA, where an explosion and fire occured during the transfer across phases. An example from Merck (USA) production of Automate Yellow 96 dye, is shown in is the solid dosing of N-bromosuccinimide (NBS) reagent Scheme 2, with a picture taken from the Chemical Safety to a heterocyclic amino derivative, where the selectivity Board report into this incident(9). differences on scale up are dramatic Hopefully lessons have been (Scheme 3). Solution addition of the When scaling up chemical learned from this incident that NBS (NB acetonitrile is often the best processes in batch reactors, there is batch (all in and heat) processes solvent for this reagent) gives a always the potential for loss of should be replaced with semi-batch better result but still the laboratory control if the reaction is exothermic, operations (controlling the rate of selectivity was not able to be since the change in heat transfer addition of one component) matched(10). area per unit volume varies with whenever possible and that a When dosing solutions of scale. change in batch size always needs

NO2

NO2

Chemical Industry Digest. July 2010

53

Scale-Up Scheme 3 Me

Me

Me

Me

Br

Br

Br

Br

NBS

+

acetone Me

NH2

N

Me

N

NH2

+ Me

5-bromo

X

X N

ArOH; NaOMe

N

Scheme 4

3-bromo

X

OAr N

NH2

N

N

desired product

N

NH2

dibromo

X

OMe

+ N

Me

N

byproduct

mimicking the typical plant vessel in the laboratory at “bench” scale. The shape (height to diameter ratio) of laboratory vessels should also mimic those in the plant.

Even if a critical reagent is added via a dip-pipe close to the tip of the agitator, lack of selectivity may occur in fast reactions if the dip pipe diameter is too large, resulting in slow flow rate from the dip pipe into the bulk solution. In the example shown in Scheme 4 the byproduct was formed owing to back-diffusion in the dip-pipe (12).

reagents in the laboratory, chemists usually add to the surface of the agitated liquid and this works fine on small scale. If this is done on the plant, however, poor mixing may occur in large reactors. The key is to add the reagents into a region of high turbulence, such as close to the tip of the agitator, using a dip pipe, or to add Reactions involving two liquid reagents via a recirculation loop. phases, such as phase-transfer If reagent addition is too fast Ensuring that the temperature and catalysed reactions, can also be very viscosity of the added reagent is compared to heat removal, accusensitive to the position of the similar to that in the vessel will also mulation can occur and can lead to agitator in the vessel, as well as the a runaway reaction, especially if aid in good mixing. agitator type/diameter/shape, and loss of cooling capacity occurs, Scale up of agitation can be based differences in yield and quality of simultaneously. on a number of factors but Paul (11) products may be seen depending on recommends using constant power the choice of vessel. Guidelines have been advocated for per unit volume or mass. This mixing energy dissipation scale up of such processes, such as “maintain the ratio is given by: of interfacial area to total volume a constant”. Interfacial area is also significant during product isolations Ei = NpN3d5/V involving phase separations (13). Where d Np N V

= = = =

impeller diameter (m) power number rotational speed (sec -1) volume (m3)

Solvent extraction problems on Scale Up

Note here the dependence on the fifth power of the impeller diameter and the third power of the rotational speed, and it is easy to see why scale up can be problematic. One answer is to focus on scale down, with the mixing energy dissipa-tion being perhaps the most useful parameter for “scale down” experiments, i.e.

54

During extractive work-ups in the laboratory, the chemist often does not record whether the aqueous phase was added to the organic, or vice-versa, since in the laboratory it usually makes no difference to the outcome of the experiment. In the plant, if the reaction mixture is organic and the vessel is not full, it may be convenient to add water to it, whereas if the vessel is full and a larger vessel needs to be used for the extractions, the water could be added first or last, depending on preferred

Chemical Industry Digest. July 2010

Scale-Up operations. The order of addition stainless steel or Hastelloy Despite the widespread knowledge can make a big difference as to equipment in the plant and whether emulsions form, and to about the hazards of scale up, particularly corrosion testing will be necessary the time of disengagement of the of compounds with potentially hazardous to study compatibility. Many layers, as well as their separation groupings, such as nitro compounds, it is reaction mixtures, however, are efficiency. Clearly this can affect disturbing to see runaways still occurring incompatible with certain metals, yield, but also may impact on due to poor hazard testing, and poor even though the individual knowledge of the thermochemistry in the product quality, since a saturated components may all be compatible solution of water in an organic process. in use tests. It is essential, solvent is an excellent hydrolysis therefore to use test the reaction medium for esters and other hydrolysable groups if traces mixture as well as the individual components for of acid base are present. At extended separation times compatibility with the materials of construction of the hydrolysis may then occur leading to greater amounts of vessel; one company I consulted for failed to do this and by-products. ended up with purple, rather than white product, when For this reason the solvent ethyl acetate is a poor the reaction intermediate picked up iron from the reactor choice of extraction solvent for scale up, particularly in and formed a coloured complex! acid/base work-ups, since the extended times the ethyl acetate is in contact with water when trace acids/bases are present will initiate hydrolysis of the solvent, leading to more acid (acetic acid) which further catalyses hydrolysis. The high solubility of water in ethyl acetate and vice-versa means that aqueous layers, unless heavily salted, are rich in organics (and thus more difficult to dispose of) whereas the ethyl acetate layers have high water contents and may need drying before further processing. Isopropyl acetate and butyl acetate, though more expensive initially, may actually be more costeffective overall in scale up, particularly since solvent recovery is easier because the low water content in the solvent leads to higher recoveries.

In another case, one company failed to realise that a CF3 group on an organic molecule may yield traces of HF in solution, formed in small amounts in byproduct processes. Since the reaction was being carried out in a glass-lined reactor, this equipment then became etched, and had to be removed from the plant and expensively relined before it was available for manufacturing use again and was a few months out of commission .

Crystallisation and Polymorphism Of all the unit operations which cause difficulty on scale up, crystallisation and drying are the most prevalent, particularly when the intermediate or final product is polymorphic or can form solvates (14). Of course in India, the generic pharmaceutical industry is highly active in investigating alternative crystalline forms of drugs in order to circumvent existing patents, or to provide new IP opportunities. However, consistent manufacture of the desired form on large scale can be a problem. Key parameters controlling which form is produced, and the particle size distribution, (PSD, which determines filterability and drying times) include the number and level of trace impurities in solution (even as low as 0.01%), which may vary from batch to batch.

In the laboratory, liquid-liquid separations are carried out at ambient temperature, which varies from lab to lab! In the plant, depending on the its location and whether it is summer or winter, extractions may be carried at anywhere from 2-3oC to 40oC unless a temperature is specified, and different results from the laboratory will be encountered. Since extractions are more efficient at higher temperature and separations are usually better (no emulsions usually), it is often preferable on plant to extract in the 50-100oC range, if this is not detrimental to the product quality; it is certainly more efficient in both raw materials When dosing solutions of reagents in usage and time. the laboratory, chemists usually add to

Compatibility with glass, stainless steel and hastelloy To avoid metal contamination, chemists mostly use glass equipment in the laboratory, but, for many reasons, engineers may prefer

the surface of the agitated liquid and this works fine on small scale. If this is done on the plant, however, poor mixing may occur in large reactors. The key is to add the reagents into a region of high turbulence, such as close to the tip of the agitator, using a dip pipe, or to add reagents via a recirculation loop.

Chemical Industry Digest. July 2010

Fine control of the crystallisation process (supersaturation, seeding, cooling profile, presence of impurities) leaves a lot to be desired, so batch to batch variations are expected. The control of a crystallisation process needs exact control of nucleation (by seeding at a defined supersaturation) and a

55

Scale-Up programmed cooling programme that allows the crystals to take up the supersaturation very slowly. Reactor or filter/centrifuge contamination from previous batches of the same substance may impact on the ability to produce the correct crystal form and PSD of the product. . These issues can also be relevant to the colour chemicals, agrochemicals and fine chemicals industries, where specific physical properties of the product are desired for further processing, such as formulation, or affect the stability of the product ( eg to oxygen or light). The additional effect of stir out time/temperature on crystal form content and PSD, with a subsequent potential change during filtration/centrifugation (depending on choice and size of equipment) and drying (depending on dryer type, temperature, solvent removal etc) means that batch to batch reproducibility is never guaranteed. Only by careful laboratory studies of this interacting factors using statistical methods such as Design of Experiments can true manufacturing robustness be achieved.

Optimisation using statistical methods Many companies in India still carry out process optimisation using one parameter at a time variations, whereas the trend elsewhere is to use the Design of Experiments (DoE) approach, recognising that variables are rarely independent of each other ( eg rate of addition and temperature). In these detailed parameter studies it is important to study variables which affect scale up, such as dosing time and mixing, and parameters in the work-up and product isolation as well as in the reactions. Only by looking at the effect of all these interacting parameters can a truly optimised process, which works well on scale and is efficient and robust, be developed. For pharmaceutical processes, regulatory authorities are keen to see the DoE approach used in new submissions, since it shows that “quality has been designed into the process” and gives assurance that the process is robust, and that the manufacturer knows the design space in which to operate and where the edge of failure lies. Such data is of course important and extremely useful for a plant manager operating any chemical process. Process understanding always leads to better process control and usually to more successful scale-up!

Conclusion The best way to minimise scale up problems is by data gathering and detailed process understanding. Having technical staff – both chemists and engineers – who are 56

well trained with up-to-date knowledge of current thinking can help with design of better processes with fewer scale up issues. Using outside help in the form of consultants with industry experience can be invaluable for companies wishing to design low-cost processes which can be easily scaled up, and in trouble-shooting persistent manufacturing problems.

References 1. T. Laird “Development and Scale-up of Processes for the Manufacture of Pharmaceuticals” Comprehensive Medicinal Chemistry. Vol 1, 1989, Pergamon Press; T. Laird, The Neglected Science of Chemical Development, Chemistry in Britain, Dec. 1989, p.1208 2. N.G Anderson, Practical Process Development, Academic Press 2000 3. K.G. Gadamasetti, Process Chemistry in the Pharmaceutical Industry, Marcel Dekker, 1999 (Vol 1) and CRC Press 2007 (with T Braish) (Vol 2) 4. W. Hoyle, Pilot plants and Scale Up of Chemical Processes, Royal Society of Chemistry, Vols 1 and 2, 1997 and 1999 5. S. Lee and G Robinson, Process Development: Fine Chemicals from Grams to Kilograms, Oxford Science, 1992 6. M. Williams and G. Quallich, Chem& Ind, 1990, 315; G Quallich, Chirality, 2005, 17, S120-S126 7. F. Stoessel, Org Process R&D, 1997, 1, 428 8. F Stoessel, Thermal Safety of Chemical Prrocesses; Risk Assessment and Process Design, Wiley-VCH, 2008 9. www.csb.gov/assets/document/ Morton_Report.pdf 10. E.L. Paul, presentation at 2 nd International Conference on Scale Up of Chemical Processses, Scientific Update, 1996 11. E.L. Paul, Y.A.Atiemo-Obeng and S.M.Kresta, Handbook of Industrial Mixing, Wiley-Interscience, 2004 12. K.J.Carpenter, Chem Eng Sci, 2001, 56, 305-322 13. J.H Atherton and K.J.Carpenter, Process Development, Physico-Chemical Concepts, Oxford Science, 2000. Editor’s Note: Scientific Update LLC regularly conducts conferences and training courses for industrial chemists and chemical engineers in chemical development and scale-up.

Chemical Industry Digest. July 2010