Minimum Stirrable Volume: A Critical Reactor Parameter for Working at Scale

In a chemical  pilot plant or plant one cannot choose a reactor as easily as one chooses a 250, 500,1000 or 5000 mL flask. One has to work with what is available and that may be constrained not just by the equipment within those four walls by what other processes are planned to be run at the same time as yours.

A plant reactor can be characterized by two crucially important volumes: the Minimum Stirrable Volume and the Maximum Stirrable Volume. The former will be discussed here and the latter in another blog article.

The minimum stirrable volume is exactly what the name teaches. It is the minimum volume of liquid that needs to be in the reactor so that the turning stirrer paddles effectively stir the reactor contents. To some extent, this minimum stirrable volume depends upon what is being stirred and what will happen during the initial stage of the reaction of interest. For example, a homogeneous solution in which an exothermic refluxing occurs at the beginning of the reaction may only need the moving blades to touch the liquid surface because gas evolution and thermal convection are going to move the homogeneous liquid phase around. At the other extreme, if zinc powder, tin granules, or magnesium turning need to be swept up off the bottom of the reactor, the immersion of the stirrer blades probably needs to be complete and the stirring rapid.

The reason chemists, who are more accustomed to working in the laboratory, tend to forget the minimum stirring volume constraint is that two of the most common laboratory stirrer types are the magnetic stirrer and the crescent bladed overhead mechanical type. The magnetic stir bar sits on the bottom of the flask so that the minimum stirrable volume is very small while the crescent bladed stirrer is very often set up with its curved edge nestled up against the bottom of the round-bottomed flask in which it is installed so that again the minimum stirrable volume is close to zero.

In order to make use of this parameter, the chemist must examine the procedure which he intends to use and decide at what points in the process stirring will be required. The chemist estimates what the reactor volume will be throughout the process. The process cannot be run with less material than will allow stirring at the point of lowest volume where stirring is essential. This is particularly important in the very earliest pilot plant run when one does not want to risk more material than is required to see whether the scale-up proceeds without problems.

An example may clarify this. Suppose you want to run a process step for the first time in a 2000-liter reactor that is planned for production. The starting materials are expensive and you do not want to risk any more materials in this first run than are necessary. The minimum stirrable volume in this 2000L reactor is 200L. Your prototype process will produce 45 gm of the desired intermediate from 40 g of starting material in the laboratory. Twenty grams of insoluble anhydrous potassium carbonate must be stirred in the solvent during the reaction.  The point in the process where stirring is essential and the total volume is lowest is right at the beginning when you start adding a reagent drop-by-drop. At this point, you have 40 gm of starting material, 20 gm of potassium carbonate, and 300 ml of solvent. Very roughly the total volume at this point of minimum volume is 360 ml.

The minimum amount of starting material you can use without any change in your lab procedure is 200/360 X40 =22.22kg. If you try using less after loading the reactor with the starting material, the potassium carbonate and the organic solvent the slurry will not touch the stirrer blades.

Suppose management is not happy with this. They don’t want to risk more than 10 kg of starting material in the first run.  The best answer is to perform the first run in the smallest reactor that can be found that has the same configuration as the reactor planned for the full production. A smaller reactor will typically have a smaller minimum stirrable volume. This is one reason that a process is often scaled up in steps.

If you must use the 2000 liter reactor you will need to perform your reaction with a larger proportion of solvent in that first run. Run your reaction at this higher dilution at laboratory scale to be sure that the change makes essentially no difference. Then you are ready for that first plant run.    

Crystal Engineering-Optimization of a Generic Drug Synthesis

Improved Crystallizability of Intermediates

A generic version of a drug is often prepared following different chemistry from that used by the originator. This may arise because some of the intermediates in the originator’s process may be still protected by patents while the drug substance itself can have become free to use. so the generic route must proceed by a new route that avoids the patented intermediates. Such a process would have been worked out for the API provider by its chemists. Although the route may be ‘optimized’ according to what can be done in the laboratory, it is not optimized for the scaled up version. One of the areas where improvements can be made without changing the overall chemistry including the trace solvent analysis, is to make improvements in the efficiency of crystallization of isolated intermediates.

By this I do not mean improvements in the crystallization yield because changes in the crystallization solvent or in the degree of precipitation of the product can have unacceptable effects, in the former case, on the trace solvent analysis and in the latter upon the level of impurities.

I am thinking here of reducing the time for crystallization and filtration. These improvements can be particularly significant for crystallization of intermediates from the early steps of the process where the number of batches of each step can be large. Yet, starting out, synthetic chemists attach little importance to the time spent crystallizing, filtering or drying before they have matured into process chemists.
The size and shape of crystals of a process intermediate are the main factors determining the rate of filtration, the efficiency of washing away impurities, and the rate of drying of the solid.

Crystal Habit Modification for Pharmaceuticals

CA2509796 (WO2004064806A2; US20060122265A1)

PROCESS FOR MODIFYING DRUG CRYSTAL FORMATION

Published 2004-08-05

The patent application teaches a method increasing the bulk density of acicular (needle-like) crystals by cyclical temperature programming of a solid only partially dissolved in a recrystallization solvent.

Generic API manufacturers may be missing out both on a competitive advantage and a cost reducing methodology by not working to control the crystal habits of the general products they manufacture.

CA2509796 (WO2004064806A2; US20060122265A1) is from a family of patent applications filed by Novartis. The alleged invention proposes to use a cycle of temperature fluctuations applied to the recrystallized slurry of a pharmaceutical product to convert needles into more compact solid forms.  The general claims of these applications are unlikely ever to issue. The authors have adopted the strategy of ignoring the prior art rather than trying to distinguish themselves from it.  Novartis is attempting to monopolize something closely akin to the ancient technique called aging, digestion, or Oswald ripening which is known to produce more easily filterable precipitates. This method has been routine in gravimetric analysis for a very long time. I learned it in my undergraduate days reading Inorganic Vogel.

Some additional time ‘ripening’ a product before filtration can more than pay for itself in shorter filtration and drying times. Saving time in a drying step can save plenty of energy. Additionally ripened crystals frequently contain lower levels of impurities. 

Flavianic Acid as a Possible Candidate as the Electron Acceptor Component in Charge-Transfer Complexes

Flavianic acid

Picric acid is a common constituent of charge-transfer complexes with aromatic hydrocarbons at least as large as naphthalene. It also forms salts with amines. Unfortunately, its explosiveness is a problem working at scale.

Flavianic acid (8-hydroxy-5,7-dinitro napthalene-2-sulfonic acid)  contains the same phenolic aryl nucleus with three strong electron-withdrawing substituents as in picric acid. For picric acid, there are three nitro groups. For flavianic acid, there are two nitro groups and one vinylogous sulfonic acid as substituents.

Like picric acid, flavianic acid is known to form insoluble salts with amines, but unlike picric acid flavianic acid forms these by donating a proton from the sulfonic acid moiety rather than the acidic phenol. As a consequence, the electron-poor aromatic ring remains substantially electron-deficient even after the transfer.

My hypothesis is that flavianic acid may form a preferred precipitant for organic amine compounds that also comprise an electron-rich aromatic ring because there are those two binding motifs. Simplistically thinking this could produce larger intermolecular attractions and higher melting points.  

The 1,2-Diol Functionality as a Possible Phase Separating Tag

In CA2677670, a monoglyceride ester is separated from other impurities by absorbing the mixture on silica gel and washing with hexanes/ethyl acetate 90:10 v/v. This was not a column chromatography as can be determined from the experimental details. The 90:10 mixture of hexanes/ethyl acetate (10 ml) was used to dissolve the approx. 16 g of ester and to this solution 40 g of silica gel was added.  The slurry was put on a fritted funnel and eluted with 150 ml of the mixed solvents to remove the impurities. A second elution with 300 ml of ethyl acetate  removed the monoglyceride which was concentrated in vacuo. This seems to show that diols seem to bind tenaciously to polar solid adsorbants.

It is well known that mono alcohols often form insoluble complexes with CaCl₂, LiCl, LiBr, CaBr₂ and MnCl₂ for example. So it not surprising that diols would form strong complexes with such inorganic salts.  As evidence of this there is a patent, US 3,846,450 titled Purification of Oxygenated Compounds that describes the removal of diols by passing a liquid comprising some of these through solid alkali earth halides. This would trivialize their separation from compounds without this substructure. 

It has been reported that complex steroidal and prostaglandin structures can be purified by precipitating as LiBr complexes [GB2094795]. The prostaglandin structures typically contain more than one alcohol functionality. This should increase the likelihood that metal halide complexes with 1,2-diols are more likely to produce solid precipitates.  Kilomentor has already published a note about using such metal complexes to separate alcohols from non-alcohols and some alcohol mixtures from each other.

I have not found work showing that substances containing two or more non-adjacent alcohol groups dependably form lithium bromide or calcium bromide precipitates even though the work with lithium bromide and prostaglandin intermediates is promising in this respect. What is clear is that neutral 1,2-diols can be separated from other functionalities ruggedly and dependably.The 1,2-diol functionality most probably can be covalently attached to a very wide variety of intermediates as a ‘phase-separating tag’.

 Substrates containing the tag would, perchance, be precipitated by stirring with an inorganic salt in non-polar solvent. It might turn out that the 1,2-diol at the end of a hydrocarbon chain might be a substructure that could control precipitation in a wide variety of intermediates using a standard set of conditions ( a particular salt, precipitating solvent, ethanol catalyst and reaction conditions). It is already known for example that a primary alcohol is preferred to a secondary or tertiary one.

After the terminal 1,2-diol had served its purpose for intermediate isolation/ purification it could be selectively cleaved to an aldehyde or cleaved and reduced to a primary alcohol with one  fewer carbons than the diol. The functional group would be expected to work as a phase-separating tag best when the other functional groups in the intermediate were not polar ones that could also interact strongly with the inorganic salt.

It seems that whether a solid complex is formed may depend upon both the crystal lattice energy of the complex and the energy of the crystal lattice of the salt itself. As Sharpless notes, [K.B. Sharpless, A.O. Chong, and J.A. Scott, Rapid Separation of Organic Mixtures by Formation of Metal Complexes, J. Org. Chem., 40, 1252 (1975)}, whether they form solid complexes or not the alcohols do cause the dissolution of the calcium chloride into the hexane. Another important observation provided by Sharpless et al. was that mixtures of alcohols often dissolved but did not even partially precipitate under the complex forming conditions even when the pure components of the mixture formed solid calcium chloride complexes when treated individually but separately. 

Why some alcohols form solid complexes and others just dissolve the inorganic salt ,but do not precipitate, has been hanging unsolved for a long time. The Sharpless strategy has never become popular. This is because, according to a personal communication from Sharpless himself, the best conditions for forming and precipitating the complexes were unfortunately not those recommended in his article. Not a 2:1 molar alcohol inorganic salt ratio, but a large excess of inorganic salt works best taking into account more cases. Perhaps the alcohols and inorganic salt form oil-in-water or water in-oil emulsions which only occasionally break down to precipitated solid. 

If the problem is emulsion formation it might be important to remove completely any residual water. Using aprotic solvents that have fewer degrees of freedom themselves might help. Cyclohexane and diisopropyl ether might be tried. Diisopropyl ether seems to be the solvent of choice when it is difficult to get regular crystallization. Patent GB1555968 suggests that methyl isobutyl ketone (MIBK)or methyl n-amylketone are preferred candidates to form insoluble complexes, at least when calcium bromide is used.

Clearly solvents must be used that do not themselves dissolve these divalent inorganic salts because such solvents present in so large an excess would easily out compete substrates.  Hexanes, methylene chloride, MIBK and methyl n-amyl ketones would meet the criterion of not dissolving much salt alone.

Besides the equilibrium effect sometimes giving rise to useful precipitation there is probably also a kinetic effect upon whether the precipitation/crystallization provides purification. The limited data could be interpreted as suggesting that small alcohols exchange more rapidly than large alcohols and small alcohols, present catalytically, promote exchanges. 

Using a Sulfur Tag for Separations both in the Lab and At Scale

In 1964, G.M.Badger, N. Kowanko and W.H. F. Sasse submitted a short communication to J. Chromatog. 13, (1964) 234 titled, Chromatography on a column of Raney cobalt. The small experimental section read as follows:

“The freshly prepared Raney cobalt (ca 7.5 g) was mixed with clean sand and packed into a chromatographic column (1.2 cm X 10 cm.). A mixture of isoeugenol (0.5 g) and 2,5-dimethylthiophene (0.5 g) was applied to the column and eluted with methanol ( a 3-ft head of liquid was required). Evaporation of the first fraction 930 ml) gave sulfur-free isoeugenol (0.477 g). Subsequent fractions contained only trace amounts of isoeugenol and were also sulfur-free. The dimethylthiophene was subsequently recovered by Soxhlet extraction of the cobalt-containing solid with methanol.” (my italics).

The discussion pointed out that active cobalt metal binds sulfur-containing compounds by chemisorption; however, unlike Raney nickel, cobalt has a much reduced tendency to desulfurize. Nevertheless, this binding is powerful, much stronger than simple adsorption, as the rigorous conditions described for removing the dimethylthiophene from the solid support attested.

What this suggested to me was that the method would not need to be conducted as a column chromatography. It would probably work simply by stirring the solid with a solution containing the sulfurous material, passing through filter aid, and washing. Thus, the method could separate sulfur-containing from sulfur-free materials by filtration as easily as an insoluble polymer is separated from a solution.

That  desulfurization under the conditions of separation is unlikely is further suggested by another paper [1960] by the same authors which contains the sentence “Desulphurisation with Raney cobalt was similar to that with W7-J Raney nickel in that, although little reaction occurred in boiling methanol, it was complete in diethyl phthalate at 220.”

It would seem that, besides obviously being able to separate the sulfur-containing from sulfur-free compounds, the technology should be adaptable to separate compounds that have been derivatized with a sulfur-containing reagent from compounds without such an appendage.

It might be that the method of recovery of the chemisorbed compound could be improved. Eluting with a solvent containing carbon disulfide or COS might speed the recovery without irreversibly contaminating the eluting solvent.

Also, a chemisorbant simpler to prepare than Raney cobalt might be available by reducing a cobalt salt with sodium borohydride to give a Cobalt boride analogous to the Nickel boride catalysts called P-1 and P-2 developed by H. C.Brown et al. 

Choose Moderate Reduce Pressure for Distillation At-Scale

The KiloMentor Blog has highlighted methods for isolation and purification from transformations at scale. One might think that a good place to look for examples of such procedures would be the famous series, Organic Syntheses. This doesn’t turn out to be the case. Organic Syntheses procedures may be good for making multigram samples of intermediate molecular weight intermediates but the techniques used very frequently have no kilo scale equivalent. The most common isolation/purification method used is simple or fractional distillation usually under some vacuum. From my limited experience, it is used in more than 50% of the procedures.  Increasingly in the more recent submissions, some chromatography is used and this is even more unacceptable for large scale work.

Moreover, the distillations are most often performed at diminished pressures that are not readily accessible using standard process equipment. It is a rule of thumb for the plant to not count on getting reduced pressure below 50 mm of mercury. In a sample of 63 procedures I examined at random, only 10 distillations were done at or above this pressure. Nine were performed between 20-49 mm; 22 from 5.0-19 mm; 11 from 1.0-4.9 mm; six from 0.2-0.95 mm; and 5 less than 0.19 mm.

The Pilot Plant Test

Although it may be true that the first pilot plant reaction in a process step development should be done as close as possible to the preferred laboratory protocol, it is not true that the sampling for this pilot plant run should be no more substantial than that in the lab. Because a pilot plant test is so expensive, provision should be made so that, if it results in unpleasant surprises, it makes the investigation easy by providing bountiful data to point unambiguously to a most probable cause for the deviation from the expected result.

It is fair to say that in the first pilot plant run, one needs a convincing reason for not taking a sample at each point in the process where the reactor contents are homogeneous- not just at the points where a regular in-process analysis is planned in the final batch sheet, after optimization.

These samples are of two types. One type is forensic samples.

 These need only be examined when there is a deviation that needs to be investigated. The second type are samples that are used to simplify the process as for example to determine the most efficient number of washes, the best volumes of washes and to obtain a mass balance by determining in what discarded phases product was thrown away.

When samples are taken, some of these samples will be kinetically unstable-that is the reaction is continuing to proceed in them even after the sample is taken. For these kinetically ’live’ samples the best course is to take them on to final product in the laboratory. To do this the sample should have the reactants completely added and should be representative of the reaction mixture. Reaction mixtures that are heterogeneous should not generally be completed in the lab because a negative result will only raise the question, “Is the problem the reaction or the sampling?”

Most often samples of reaction mixtures will be treated with some reagent to stop further reaction and to preserve the stoichiometries present at the time the samples was removed from the bulk reaction.

Sometimes reaction mixtures that are intentionally not quenched even though they are known to be poorly stable are stored in very cold conditions to stabilize them and only used for forensic purposes if and when it is discovered that there was a deviation. If all runs smoothly, they can be discarded.

If the quenched reaction mixture is quite stable, samples should be taken for split run testing of varying work-up variants. Also running the work-up of a known fraction of the reaction mixture in the lab allows a quantitative and qualitative comparison of the yield and purity between lab and plant methods.
If multiple extractions are performed, samples of the discarded phases can be used to determine the mass balance and to predict the most efficient number and volumes for the extraction liquid phases. Because the point of maximum volume often occurs during one of the extraction stages, efficiencies in the extraction can increase the throughput and improve the costing of the process step.

Washing of crystals or of insoluble catalysts or by-products can be important for realizing a good yield. Samples of the filtrate at different points can provide information about the best wash quantities. The effectiveness of the washing of a filtered solid will almost always be very different with the plant equipment compared to the laboratory setup. If centrifugation is used in the pilot plant this is even more often true.

A pilot plant run should never be undertaken until a lab batch has been made and analyzed using the solvents, reagents, and substrates to be used in the pilot batch. That is: substrates, reagents, solvents and catalysts that are to go into the pilot plant batch should be use-tested in the laboratory. This the most fundamental rule of scale up, yet, it is the frequently violated. The reason is that sampling the pilot plant materials is tedious or faces bureaucratic documentation hurdles. One solution is to ask well in advance so that the process development use-test sample is taken when the analytical laboratory sample to verify the certificate of analysis is taken.