Hydrotropes as Solvents for Extraction and Separation

Hydrotropes are water solutions, most frequently but not always of ionized substances, with concentrations of at least 1 mole/liter. They increase the solubility of organic solutes which would have very poor solubility in water alone. For practical reasons, useful hydrotrope-forming materials need to be inexpensive. Typical hydrotrope-forming substances are:

• aromatic sulfonate salts
• aromatic sulfonic acids
• salts of benzoic acid and substituted benzoic acid
• glycols
• urea
• 4-isopropyl benzenesulfonic acid calcium salt
• 2,4-dimethyl benzenesulfonic acid sodium salt 40%
• p-toluenesulfonate sodium
• ethylene glycol monobutyl ether O-sulfonate potassium
• potassium salicylate

One can have a partitioning of substrate compounds between the hydrotrope and an apolar solvent such as heptane, cyclohexane, and methylcyclohexane. The use of these very non-polar solvents increases the proportion of the substrate extracted into the aqueous phase particularly when a substrate has limited solubility in the alternative hydrocarbon phase.

An example from recent literature is provided:

Hydrotropic Separation of Mixtures of o/p-hydroxy-acetophenone

Koparkar Y.P. and Gaikar V.G. Separation Science and Technology  (Sep. Sci. Technol.) 2004, vol. 39, n^o16, pp. 3879-3895 

A new extractive separation technique has been developed for the separation of o-/p-hydroxy acetophenone (HAPs) using a hydrotropy. Hydrotropes are freely water-soluble organic salts, which enhance the solubility of otherwise water-insoluble or sparingly soluble organic compounds in aqueous solutions. The ability of hydrotropes to differentiate even isomeric organic compounds is explored in this extractive separation. o/p-HAPs were extracted from their solutions in organic solvents of different polarities using aqueous solutions of hydrotropes. The solvent nature has a significant effect on the selective extraction of both phenols. The combination of heptane and aq. Na-p-toluenesulfonate (at least 194 g/1000 g. water) solution gave almost pure p-HAP in the aqueous phase, whereas with chloroform as the solvent, it was possible to retain with complete selectivity p-HAP in the chloroform layer.

Any organic solvent can be used as the 2^nd phase in separation using a hydrotrope so long as that solvent is not strongly soluble in the hydrotrope. Such a combination can be used particularly when one only wants to remove a particular element of a mixture ie. taking a small amount of impurity into the hydrotrope phase. This could be particularly used to purify a material with less than a percent of a particular impurity.  When the impurity seems to be partially soluble in water but not sufficiently soluble to remove it using a practical amount of washing, resorting to a hydrotrope could solve the problem. 

The use of hydrotropes for extraction is an example of ‘salting in’ when it uses a concentrated solution of an inexpensive hydrophobic salt.

Another possibility is that the solid crude substance could be partially dissolved in the hydrotrope solution.

The solubilization capacity of a hydrotrope is usually an exponential function of the hydrotrope concentration, and mere dilution with water can be enough to recover the dissolved materials.

Hydrotropes can be expected to be useful to replace water when it is used in an aqueous trituration of an impure solid mixture as in ‘swish’chromatography. This can be expected to be advantageous when very little of either the predominant component or the minor impurities dissolve in water alone.

The Manufacturing Process Route Selection: The Inclusion of Phase Switching and its Relationship to Validation

Neal G. Anderson in his monograph, Practical Process Research & Development, says nothing about validation but he does speak about the need “to freeze the final process step early; that is, to identify early the most desirable final chemical transformation hat gives the drug substance.” He credits this concept to P. Shutts. [“Freeze the Commercial Process-Issues and Challenges” a talk contributed at the Third International Conference on Process Development Chemistry, Amelia Island, FA, March 26 1997].

Earlier in his book, Anderson says specifically,  “In order to establish routine impurity profiles and levels, ideally the final isolation of drug substance should be optimized, and this process should be used for the preparation of material destined for toxicological studies and later Phase I studies. The types of impurities found in drug substance will be largely determined by the starting materials and reagents used in the final step to prepare the drug substance. Thus the ideal penultimate compound should be identified in investigations, and parallel development work should converge on this penultimate compound.”

In this particular passage the last step Anderson is talking about is not the formation of a pharmaceutical salt. Almost certainly he is talking about a covalent bond forming step that completes the final structure.

Also Anderson is speaking about the ideal situation but what this assumes is not just that the last step, from penultimate compound to actual API, does not change, but also that this final intermediate is so efficiently purified that the steps preceding do not contribute to its impurity profile other than as inconsequential trace substances below 0.1%. It would only be then that different routes to this penultimate compound would not leave behind any of their own impurities in the drug substance. If this is translated into the vocabulary of process validation, this would declare that there should be no critical steps before the purification of the ultimate intermediate! This is indeed a truly ideal situation and would constitute an impractical goal to routinely seek for real-life penultimate compound purifications.

Even though Anderson does not actually use the word validation in his book, even if we accept that validation is the most significant economic outcome of a developed pharmaceutical process, Anderson does convey the idea that certain characteristics of the final process step simplify validation and so are very, very important.

To get there requires a very demanding choice for this penultimate compound. It must be easily and efficiently isolated and purified. As Kilomentor has argued this suggests that the step in which it is prepared and worked up needs to comprise several phase switches, because it is phase switches that provide robust purification and it is compounds containing acid and/or basic functionalities that provide the most common opportunities for these simple phase switches. 

A Practical Scheme for Working Up a Reaction Mixture based upon real Liquid-Liquid Extraction Possibilities and Logical Solubility Testing

General schemes have been devised for examining unknown mixtures such as those one learns as an undergraduate chemist for a laboratory examination. Such schemes as those in the classic text by Shriner, Fuson, and Curtin. [The Systematic Identification of Organic Compounds: A Laboratory Manual. Fifth Edition John Wiley & Son 1964] The more complex of these schemes, no matter what their wisdom, have never to my knowledge been adopted by real chemists to work up real reaction mixtures. The reason is simple. These schemes are designed to handle complete unknowns. The bench chemist always knows something about his mixture, even if it is no more than the fervent hope that a particular product will be present; therefore, the working chemist or chemical engineer is in a more knowledgeable position and so the protocol for a complete unknown is going to be inefficient.

Since at least the functional groups and molecular weight of the desired product are known and so some educated guesses can be made about the likely physical properties and a theoretical proposal can be made for a rational separation.  For example, if the hoped-for product is a neutral lipophilic aldehyde, a mild aq. acid extraction and a mild aq. base extraction can be applied to an ether solution. Then some aldehyde-specific reagents such as bisulfite or Girard’s P or T can be contemplated.

Another reason that these classical schemes are set aside is that they apply solubility tests in unappealing solvents such as carbon tetrachloride and benzene. Today these are unacceptably toxic solvents.

What would be more interesting and have more likelihood of application would be a scheme, that could be adapted to take account of predicted estimated properties of a particular reaction mixture and would have an inherent awareness of the most helpful procedures for separations at scale.

To explore this, let us assume first that at the end of the reaction period, the reactor content is homogeneous and the TLC or other in-process check is encouraging.  Let us also assume that the trial reaction has been conducted on a scale of at least several grams. I make this assumption first because improving the throughput in a series of reactions in a process scheme should focus on the early steps and these are intended to be the ones using the less expensive materials.  

If the content is not homogeneous, the phases should be separated and treated separately.  This practice is based on the wise rule that the chemist should never immediately refuse a phase separation offered by nature while exploring further manipulations. If there is a solid in a liquid, filter the solid and retain the filtrate. If there are two liquids, cut the phases and examine each. Such natural separations are not likely to be quantitative recoveries of any component, but the constitution of the phases may provide a guide to a modification that can deliver a more complete separation.

Supposing in the alternative now that the phase is a single one or that we are examining multiple phases separately, the chemist will be fairly confident that the main constituent of this phase is solvent and the solvent is the reaction solvent.  The first step will be to remove that solvent completely without exposing the other constituents to conditions so severe that they may resume reacting with each other.  The chemist is the best judge of how to achieve this and how to make sure that no further chemical reaction happens.

Let us therefore suppose now that the solvent has been removed under mild conditions and the residue, either as an oil, a mixture of solids and liquids, or a glass is available to be examined.

It is not possible to separate effectively a viscous oil from a solid. In this situation, oil and solid probably must be handled together although again this is a situation where the chemist’s powers of observation and judgment are more useful than any strict rule.

The mixture should be examined for its solubility properties much as in the classic approach in Shriner, Fuson, and Curtin but the solvents should be different and be based on a different principle.  

The solvents, which might be used are hexane, acetonitrile, methanol, and ethyl ether. The first three solvents are chosen because methanol and hexane in the presence of a few percent of water give two immiscible phases and the same is true of acetonitrile and hexane. Different solubility information among these three can not only direct us to a trituration step but differential partitioning among these pairs can identify liquid-liquid extraction possibilities. 

Although diethyl ether is not a solvent acceptable in a general-purpose chemical plant, its remarkable ability to avoid forming emulsions makes it irreplaceable for acid and basic aqueous extraction tests.

So in practice, one gram of the mixture is placed in a small r.b. flask treated with 7 ml of methanol and swirled.  If any solid remains undissolved the slurry can be cooled in ice to maximize the quantity of solid and filtered cold and washed with a little cold methanol. Such solid is examined.

If there is no solid in the solution we could add 7 ml/gm of hexane(s) and mix the phases together.  Again we look for any solid, which might separate and treat it appropriately. If two immiscible liquid layers are not present a drop or two of water is added to make the methanol layer separate.  The two phases are separated and each is evaporated to dryness, pumped under vacuum, and weighed.  Each phase should be examined, if it is convenient, by the same analysis that was used for the reaction’s in-process check. The combination of the weights obtained and the analyses of the separate layers are useful properties of the mixture.  They may provide the first hints of the most efficient methods of isolation. If one phase or the other contains essentially all the contents of the mixture all one can say is that the mixture is substantially polar or apolar, depending on to which solvent the mixture components have migrated.

If the mixture is substantially apolar take a new sample of the mixture in a small r.b. flask and tread it with 7 ml of acetonitrile and swirl, repeating the procedure that was used with methanol.  In the case of acetonitrile, water will very rarely be needed to get two liquid phases upon adding hexane. In fact, try to avoid using water here. The two phases are separated and each is evaporated to dryness, pumped under vacuum, and weighed.  Each phase should be examined, if it is convenient, by the analysis that was used for the reaction’s in-process check. The combination of the weights obtained and the analyses of the separate layers again are useful properties of the mixture.  They may provide the first hints at the most efficient methods of isolation. If one phase or the other contains essentially all the contents of the mixture all one can say is that the mixture is substantially polar or apolar.

Water is not used in these tests. Nevertheless, there is a good likelihood that if an inorganic salt is present it will be insoluble in one of methanol or acetonitrile and will have been filtered off.

A frequent result of this work will be that the substantial majority of the reaction mixture will remain in the hexane layer. This is to be expected since the vast majority of organic compounds are substantially lipophilic and noncrystalline when present as mixtures; nevertheless, when a useful separation is made at this stage it may be particularly useful for its simplicity. 

Take a new portion of the mixture and try to dissolve it in 7 ml/g of diethyl ether. Again if there is a solid separate it.  Now, classically extract the diethyl ether with an equal volume of 1N aq. HCl and separate the phases. Adjust the pH of the aqueous phase back to neutrality observing any cloudiness or solid separation and then back extract the murky neutralized water with ether, dry, evaporate to dryness, and weight.

In the same classical fashion extract the ether, which has been acid extracted with aq. base of pH about 9.0 and recovery the acid fraction.

Recover the neutral constituents from the residual ether. That has now been treated both with aq. acid and aq. base.

Each phase should be examined if it is convenient by the analysis that was used for the reaction’s in-process check. The combination of the weights obtained and the analyses of the separated layers are useful properties of the mixture.

Quite often very little more will have been accomplished than would have been achieved following the tried and true rules of thumb, but a useful number of times something really exciting and simplifying will have been drawn to your attention.

If the material, that you are seeking is either in the acidic or the basic fractions, even if it is still a serious mixture, your problems are well on their way to resolution because the means for rugged separations of such mixtures at scale are plentiful, and these methods I explore elsewhere. See for example Kilomentor’s blog on extractive crystallization.

If the substance you are seeking still seems only to be found in the hexane or neutral diethyl ether phases more sophisticated means need to be applied.  If 30% or more of a target substance has ended up in the methanol or acetonitrile phases there is reason to hope that more intensive extractions may give you what you need.

If TLC of the methanol, acetonitrile, or hexane solutions showed a substantial amount of material remaining at the origin, the presence of high molecular weight or even polymeric materials is likely.  If the mixture is strongly colored and the product sought is not expected to show color, polymer and tars are likely and the mixture should at the outset be cleaned up before looking for the desired species.  Filtering through a plug of adsorbent, which retains the origin material is usually successful.  Charcoaling a portion in an alcohol solvent often works.  Sometimes steam distillation, regular distillation, or codistillation with a high boiling hydrocarbon can be useful. In codistillation, with kerosene be mindful that you will need to get the mixture back from the high boiling solvent!

Because cyclohexane combined with nitromethane or nitroethane, or any mixture of the two, also forms two immiscible phases; the same methods illustrated above can be applied in this system.  The same goes apparently with cyclohexane and mixtures of dimethylformamide/dimethylacetamide. With these combinations, the temperature needs to be kept not too far above ambient to preserve the two-phase behavior.

Hydrotropes can also be used to dissolve components of an unknown mixture into water.

30% Acetic acid- 70% Water and Methylene Chloride: Two Liquid Phases for Partition Extractive Work-ups

Because Kilomentor puts a huge emphasis on simple, rugged, powerful separation methodologies, I am very interested in pairs of liquids that are not miscible with each other but where both have good solvent properties for a substantial range of  organic compounds. The reason is that liquid-liquid extraction is a simple and scalable purification method.

I have in my old laboratory notes mention that a mixture of two phases: one- methylene chloride, and the other- 30% acetic acid-70% water, was used to remove the non-polar contaminants from a peptide mixture made up from among neutral amino acids. It seems to me that the addition of the very good solvent acetic acid will have increased the number of polar or semi-polar substrates so substantially in this predominantly aqueous phase that it in competition with methylene chloride could separate some important mixtures. Furthermore, the dissolving power of methylene chloride probably can be further modulated by adding into it toluene or hexanes

.  

Using Reduced Pressure Distillation of Azeotropes to Perform Solvent Switches in Chemical Process Development at Scale

Many organic chemists, including Kilomentor, have accumulated tables of lower boiling azeotropes that can be useful when designing work up procedures or solvent switching protocols. These azeotropes are mostly ones that exist at atmospheric pressure. When one thinks more carefully about the actual isolation in a particular scale up, because a solvent exchanging distillation is going to take quite some time, it would be better if the distilling were done at reduced pressure where boiling points are lower and where we can expect less degradation of the solute product we want. The difficulty is to know whether an azeotrope exists at the reduced system pressure we can achieve dependably in the scaled reactor. Fortunately, for a group of common organic solvents the Vapor-Liquid phase diagrams are available for free at http://vle-calc.com/phase_diagram.html

All you need to do is select the solvents you are working with, choose isobaric conditions, choose an x, y, T type diagram, select the pressure you can expect to easily achieve ( pressure units can be selected by clicking on the pressure unit shown) and press the calculate button.

There is a limitation on the solvents that the software can deal with.

Understanding Distillation is still important for Chemical Process Development and Organic Synthesis At Scale.

One of the non-obvious outcomes of structural identification using spectroscopy (particularly NMR and MS) is reduced experience with distillation among organic synthetic chemists. This is because even an inexperienced student researcher can now routinely identify a substance using milligrams of a pure compound; first using flash chromatography. Then high-performance preparative liquid chromatography or preparative gas chromatography can replace old-fashioned distillation for making samples big enough for identification of the products from most steps in laboratory synthesis. 

Corroborating this trend is the virtual disappearance of boiling points as part of physical characterizations in the chemical literature.

Finally, as the catalogs of suppliers of chemical intermediates become thicker, more of the products of early steps in syntheses can simply be purchased. It is these lower molecular weight entities that formerly were prepared and distilled in the lab.

Standard distillation has an inherent problem that became a further reason for the substantial abandonment of distillation from the laboratory. Unless a small-scale distillation column receives an input of heat supplied by vigorously boiling the liquid mixture in the still pot, it cannot achieve liquid-vapor equilibrium. Thus, on the lab scale, there is a hold-up of distillate that is inevitably lost and this can be up to 30%. Compounding this inherent difficulty is the annoyance that all glass laboratory distillation equipment is expensive and does not easily accommodate the particular amount of crude that you may have. That is, the amount of crude distillate must be selected to fit the size of the physical assembly that you have and not the other way around. Fractional distillation assemblies are not available in your lab drawer in 100 ml, 200 ml, 500 ml, 1L, 5L, and 15L sizes like round bottom flasks are!

The days when distillation units were patched together with hardened cork or rubber stoppers between pieces of blown glass are long past. Now all glass assemblies are a single piece or pieces joined with ground glass joints.  

Because of this, now more than ever, distillation assemblies for vacuum distillation often use the same equipment as for simple distillation and lab workers don’t appreciate the special requirements imposed by the low-pressure condition. The boiling point of the fluid mixture in the still pot of a distilling assembly depends upon the pressure at the surface of the liquid, not the pressure recorded on a pressure gauge, which may be, and usually is, closer to the vacuum pump. For pressures from 760 mm down to 15 mm of mercury, a regular distillation flask is satisfactory. For pressures below this level, and particularly pressures 2 mm or less, the diameter and location of the vapor port linking the distillation portion of the apparatus to the condensing portion become very important. This is not usually understood.
The increment in vapor pressure at the surface of the boiling liquid, over and above the vacuum pressure reading taken at the receiver is proportional to the length and inversely proportional to the fourth power of the diameter in centimeters of that sidearm plus any other narrow portion of the path between still pot and condenser.

 

Advantages Filtering Solids at Scale

Crystallization in the laboratory is rarely performed completely under an inert atmosphere. Most commonly crystal filtration is done in the open air on a Buchner filter followed by washing with ice-cold wash liquid and then partial drying by the passage of air drawn through the filter cake by vacuum derived from a water aspirator.

Because it is conducted in this fashion the final crystallization temperature and the temperature of the wash liquid is rarely taken below zero degrees Centigrade because this would cause moisture from the air to contaminate the solvents used and/or to condense on the porcelain or glass filter funnel.  But conversely, if the Buchner filter is not sufficiently cold, it becomes more difficult to draw off the mother liquors and the wash solvent without partially redissolving the filtrand. Thus laboratory filtration in the open air often has contradictory preferences.

These are not problems in the kilolab, pilot plant, or plant. For safety and to avoid contamination all operations are done in a closed system that is easily kept dry and inert. As a result, cooling to a lower temperature, such as -20° C  is simple during all the operations of crystal formation, collection, and washing.

Sometimes efforts to find a suitable solvent system for recrystallization of a compound, which is crystalline within the typical ambient temperature range, can be replaced with a low-temperature recrystallization from hexane, pentane, or other hydrocarbon liquid. The large temperature range between the liquid’s boiling point and -20°C diminishes the need for a dramatic difference in solubility between the refluxing hot solvent and that same solvent at 0°C.

The special need that must be met to explore such an approach is met by providing means for checking out low-temperature recrystallization at a laboratory scale. Roger Giese described such an apparatus and its mode of use in the Journal of Chemical Education, 45, 610 (1968). Step-by-step instructions are provided. The apparatus is sufficiently simple that it can be put together by modifying a chromatography column that has a fritted glass disc as the plug. Because it operates with its own jury-rigged cold bath made from a plastic bottle, it does not need to fit in a Dewar for cooling, unlike the apparatus described by C. Frank Shaw, III and A. L. Alfred in the Journal of Chemical Education, 47, 165 (1970).

The Trapping of Excess Benzyl Halides

Benzyl halides are lachrymators. Their presence in a reaction mixture even as a residue makes work-up troublesome. In some cases, human exposure can cause serious skin problems and so a method for removing them from a reaction mixture before work-up and a simple method to separate the resultant materials in the isolation would be valuable. This is an unproven suggestion. Thiourea is a relatively inexpensive and innocuous material that reacts readily with benzyl halides upon heating together. The products are called phenylmethyl isothiuronium salts and they form insoluble salts with carboxylic and sulfonic acids which can be easily filtered off. 

A Piperidino Phase Switching Functional Group Tag

KiloMentor blog articles have emphasized of having intermediate substances that are easy to purify in a synthetic scheme.  An intermediate that acts as a base and can be extracted into an aqueous phase as a salt and subsequently taken back into an organic solvent in its free base form will usually be easier to separate and purify. This is consistent with the teaching of Dennis P. Curran who first drew attention to the importance of phase switching in purification and isolation.

R. A. Olofson and Duain E. Abbott wrote a paper, Tests of a Piperidino Mask for the Protection of Functionalized Carbon Sites in Multistep Synthesis, J. Org. Chem., 49, 2795-2799 (1984) which I believe needs more consideration. It teaches that a piperidine functionality attached to carbon at a primary or secondary carbon within a complex structure can be converted, in their respective cases, into a primary chloride or secondary chloride/ alkene in high yield and good selectivity.  To quote from the paper, “…. advantages can be envisioned in schemes using this mask for multistep protection. First, acid-base extraction methods are readily adapted for the isolation of process intermediates. Second, the formation of readily crystallized amine salts can further facilitate the purification of these intermediates.”

The reagent that affects this loss of a piperidine group from a complex molecule and its replacement by chlorine is α-chloroethyl chloroformate {Cl-CO-O—CHCl-CH₃; called ACE-Cl}. The conversion can be pictured as:

The reason the cleavage is so clean with the piperidine ring moiety unopened is that competition between S_N1 (E1) and S_N2 cleavages of the intermediate quaternary salt is occurring with relative rates such that ArCH2, allyl, t-alkyl >> sec-alkyl >methyl > primary 

alkyl >> piperidine ring opening. That is, the piperidino ring 

functionality is cleaved much more slowly than anything else!

Treatment with methanol hydrogen chloride removes the ACE protection liberating the piperidine.

Use of Molecular Sieves in Process Chemistry

Using inorganic salts for drying organic solutions on scale requires adding a solid into a reactor containing organic solvent. Considerable time is needed for filtering and washing steps and this contaminates several vessels, the pumps, and connecting lines.  The use of azeotropes is sometimes an alternative but there are many cases where these do not exist. Drying with a column of molecular sieves is an alternative. It is not without its own drawbacks but in a situation where there are few options it needs to be considered.

If the reaction being optimized involves an equilibrium involving a small molecule, molecular sieves are an effective tool for shifting the equilibrium using the principle of Le Chatelier.  Commercial molecular sieves are synthetic aluminosilicate solids called zeolites. These inorganic polymers have structures that contain cavities interconnected by channels that normally contain water. Many solid materials contain interstitial water. What sets zeolites apart is that even after the water has been driven off the inorganic architecture remains the same and the unaltered cavities remain. Different zeolite structures have channels with different pore diameters and so can hold and retain molecules of different sizes removing them from a solution.  Sieves designated 3A can only host water or ammonia and nothing larger. Type 4A can also accept methanol, ethanol, hydrogen sulfide, carbon dioxide, sulfur dioxide, ethylene, ethane, or propylene. Type 5A absorbs small straight chain alkanes but excludes branched molecules and rings larger than four-membered.
The operational problem with molecular sieves is that either the solution in contact with the sieves must be very vigorously agitated or the sieves must be first ground to a powder to expose a high surface area to the solution. The high stirring grinds the sieves even more and makes recycling of a consistently sized material impossible; therefore, at-scale the most common way to operate is to rapidly pump the solution that is to be treated in a cycle through a bed of sieves.
This can achieve a high rate of absorption of any molecule,which fits the interstices of the channels. They are taken out of the solution and into the zeolite.  When a zeolite's capacity is exhausted it can be reactivated with vacuum and heat. 

Managing Competing Reactions

The yield of a sought-after transformation in an organic synthesis sequence can often be reduced by the loss of a portion of the starting material in a competing reaction.

To overcome this,  the most common strategy is to vary the reaction conditions including, the reaction time. Frequently, it is changing the reaction solvent or the reaction temperature that contributes massively to that success.  That each different reaction is likely to respond differently to such changes and so a good chance exists to find conditions where the desired reaction is even more favored.  A change in the solvent seeks to alter the difference in the free energies of activation between the different reactions. A change in reaction temperature aims to change the ratio of reaction rates as a function of temperature without changing the actual difference in free energies of activation.

Other, less obvious, but potentially effective targets also exist. If catalyzed versions of the desired reaction, as opposed to the undesired competing reaction are known, or can be conceived, addition of a catalyst can accelerated the desired change. Addition of a catalyst is the most dramatic way to change one free energy of activation relative to another. Furthermore, because catalytic mechanisms are so specific, it is very unlikely that both reactions would be affected to the same degree.

A much, much less frequent situation arises where the less favored reaction can be inhibited.

Such a situation has been described in the paper, Effect of Amines on O-Benzyl Group Hydrogenolysis,[ Bronislaw P. Czech and Richard A. Bartsch, J. Org. Chem. 1984, 49, 4076-4078 ]. In an example, a substance with a monosubstituted olefin and an alkyl benzyl ether was treated with hydrogen gas and Pd/C in alcohol with the addition of a catalytic amount of a non-aromatic amine. In the presence of the added amine only olefin reduction occurred. Benzyl n-nonyl ether as investigated as a test substrate. Using the same conditions but in one case in the presence of 5 mol % n-butylamine and in the other in its absence, the first case gave complete debenzylation and the latter case gave none. This may be a fairly common case in the situation of competing reactions on heterogeneous catalysts. If different sites are responsible for the catalysis of the two reactions one group of sites can be selectively poisoned. Thus we could understand the different activities of a catalyst prepared with different protocols.

How would one create larger differences in the product ratios arising from differing fates of a common intermediate?

This is the most troublesome situation because there is a common free energy of activation going to the shared intermediate, so changing that activation energy will not change a product mixture. Also, any subsequent free energies of activation passing from the high energy common intermediate to the transition states can be expected to be small and the difference between these two even smaller. It is the rate of quenching of the activated common intermediate that may be crucial here. Take the example of the 5-hexenyl radical intermediate. After only a brief existence the material trapped will arise from the cyclopentyl methyl radical.

Making a Good Recrystallization Process Step Better

Recrystallization can efficiently purify organic solids.  The weakness of the method for devising optimal synthetic processes is that a good recrystallization cannot be predicted based on molecular structures of starting materials, co-products, by-products, and product to the same extent one can predict, for example, the results of acid-base extractions or methyl alcohol/heptane solvent partitioning.

It is for this reason that in synthesis planning for chemical process development Kilomentor gives preference to intermediates that are acids, bases, or salts.

Nevertheless, many process intermediates will be compounds that offer no practical alternative to purification by recrystallization and so it is useful to consider simple ways to increase the recovery from recrystallization steps. 

Recrystallization separates impurities in two ways during the operation. Typically, the solid is first dissolved in the minimum amount of a hot solvent.  The temperature for dissolution is most often the boiling point of that solvent, although for high boiling solvents a lower temperature, such as steam bath temperature, may be used.  These temperatures are convenient because there is no problem holding a solution at these points.  The hot solution is then filtered to remove insoluble substances. This filtration is the first phase separation from extraneous insoluble solids. Very often this purification opportunity is not properly recognized because a good solvent usually dissolves essentially everything when warmed or, in instances where it does not, some kind of filter aid is added, obscuring the presence of insolubles.  Then, in the second stage, the clear solution (i) may be cooled to a lower temperature (ii) is mixed with an anti-solvent to reduce its solubility or (iii) is treated in both ways together. Thereafter, a crystalline solid phase appears, is separated by filtration, and the impurities are retained in the mother liquors.

In the most frequently used techniques, recrystallization is conducted from a single solvent or a mixture of two solvents by dissolving the solid hot, filtering hot and then cooling to recover a crop of crystals.

When more of the recrystallizing solvent mixture is needed to completely dissolve the crude solid prior to filtering than is needed to effectively hold impurities in solution after cooling, good product is likely being lost using this simple process.

The following simple test can easily show whether a part of the product is being unnecessarily lost in any particular recrystallization situation. Instead of recrystallizing the solid in a single charge, divide it into two equal homogeneous portions. Recrystallize the first portion as usual with the only difference that if the crystals are washed on the filter, keep the wash liquid separate from the regular filtrate. Dry, and weigh this first portion. Now recrystallized the second portion of crude using as solvent only the mother liquors from the first portion. Again dry and weigh the product and analyze both for purity.

If  the recovery from the second portion is greater than from the first at the same time as the purity profiles of the two portions are not significantly different, changing your processing methodology will save you product.

At scale, recrystallization in two portions rather than one will save product but double processing costs.  The same result, however, can usually be obtained by dissolving and filtering the entire crude amount in a single charge and then reducing the volume by half before cooling and recovering the solid.  When the two conditions are met, the two-stage laboratory experiment provides the proof that you only need half the solvent to efficiently dissolve away the impurities. The second half of the solvent was more than anything else just dissolving away your product.

Note that in order to practice this method without problems the hot solution of crude solid must be stable to any extended boiling during the concentration stage.  Of course, if there is a stability problem, the concentrating can be done under reduced pressure to lower the heat requirement.

To proceed further with the same thought, one should first place the crude solid into a poor solvent such that not the entire mass dissolves.  It is important however that this solvent is more volatile than the good solvent in which one is planning to recrystallized as above. Filtering the mixture partially dissolved in this poorer solvent will remove some material as sludge on the filter. Adding the good solvent and heating to boiling will remove this poorer solvent because it is more volatile. After hot filtering, the last traces of this poorer solvent will be removed when the volume is boiled down to one half.

Recycling Mother Liquors in Chemical Process Development to Raise Yields and Reduce Solvent Usage

Maximizing chemical yield by recycling mother liquors from crystallizations is underutilized in chemical and pharmaceutical processing, particularly outside the developing world. One of the few articles on this subject is Alan A.Smith’s, A Model for Mother Liquor Recycle in Batch Processing, Org. Process Research & Development 1997, 1, 165-167. Dr. Smith himself, in the comments, recommends a further reference.

Kilomentor’s discussion here is indebted to this first paper.

When crystallizing or recrystallizing product (i) from a reaction mixture (ii) a partially treated workup solution or (iii) from a crude solid isolate, none of the impurities’ concentrations exceed their solubility product in the solution. That is, more of each impurity could be dissolved in the filtrate without precipitating any of that solid.  The liquid solution can extract more of the impurities from the crude product. Besides having the residual capacity to dissolve impurities, the filtrate is saturated with the desired product, which is going to be lost if the filtrate is sent as waste.

Both of these situations can very often be improved upon if a portion of the crystallization filtrate from a first batch can be used as part of the crystallization solvent in a subsequent batch.

There are other situations, easily identified, where filtrate recycling is not promising. For example, when an anti-solvent has been added to cause crystallization and this anti-solvent is not easily removed.  The reason is easy to understand. Adding this modified filtrate back into a second batch will not reproduce the precipitation conditions of the first batch.  It is unreasonable to expect an equivalent product.

A similar unpromising situation occurs with crystallization from mixed solvents where the product is dissolved in a first solvent and then that homogeneous solution is diluted with a second solvent to induce crystallization.

In general, it is crystallization achieved by cooling alone that is amenable to partial filtrate recycling because the original condition can be recreated simply by reheating to the original dissolution temperature.

If x is the fraction of mother liquors you contemplate recycling in place of an equal volume of solvent and 0 < x < 1, the paper referenced above shows that the impurities in the mother liquors of each subsequent batch will tend towards a limit that at infinite batches becomes

                                    I_infinite = I₁  X 1/(1- x)

Thus with x=0.5 and the level of impurity in the first batch I₁ = 2%   I_infinite = 2% X 1/(1-0.5) = 4%

Or with  x=0.7 and the level of impurity in the first batch I₁ = 3%                 I_infinite = 3% X 1/(1-0.7) = 10%

Or  with x=0.6 and the level of impurity in the first batch I₁ = 0.8%               I_infinite = 0.8% X 1/(1-0.6) = 0.2%

Of course, if you recycle all the mother liquors no matter what the level of I₁ is Infinite = number  X 1/(1-1.0) = infinite; that is to say the impurities come out on the product.

The motivation for recycling some of the mother liquors is, of course, not usually to save solvent but to increase the recovered yield of the desired product.  To illustrate this, for simplicity let us suppose that what Kilomentor defines as the ’reaction yield’ (the assay of the desired product in the isolation solution as a percentage of the theoretical quantity of desired product) is 100%. The ‘recovery yield’ (the weight % recovered product as a percentage of the weight of desired product based on the assay yield) in this situation becomes equal to what we all call the reaction yield (the weight of product isolated over the theoretical weight of product possible as a percentage). If under these circumstances the reaction yield is 70%, there will be 30% of the material left in the saturated mother liquors (so long as no degradation has occurred).  If half of it is recycled, the overall yield will be increased by ½ X 30% = 15% and will become 85%.

Now even if the solubility product limit of all the different impurities in the mother liquor is never exceeded, the desired product which is isolated by crystallization using some mother liquor recycling will be less pure than when the technique is not used. One reason for this is that the mother liquors do contain a higher concentration of impurities and more of these by-products will either co-precipitate or be adsorbed on the pure solid crystalline product.  

Another possibility is that separation of the desired crystalline solid from mother liquor solution is incomplete. Mother liquor solution is trapped on the surface of the solid and evaporates there or is deposited there when the crystals are placed in the drier. Certainly, the wash solution used on the filtered crystal product becomes more critical both to preserve the yield improvement achieved (by not dissolving the product) and by removing this film of mother liquor without precipitating impurities.

The crystallization from a solution in which a portion of mother liquors is being recycled will likely be different from an isolation without recycling.  Optimal crystallization temperature and cooling time will change as the percentage of impurities changes. Typically crystallization proceeds more slowly in the presence of a higher concentration of impurities and greater care needs to be taken to prevent co-precipitation.

It would be unusual to recycle more than 50% of the mother liquors from one run of a campaign to the next. Remember to save all the mother liquors from run A until the product of run B certified to be trouble-free. If run B has a problem and needs to be investigated, you will not want to use run B mother liquors in run C. You still want to have at least 50% of the mother liquors from run A to use while you check to see if there is some deviation in run B.

If you intend to use mother liquor recycling in a validated process, you will need to use mother liquor recycling in the validation batches and have in place the analytical testing protocols required to show that the mother liquors you are transferring from one batch to the next meet preset standards and have been stored for a validated time under validated conditions. 

A Large Rigid Acid for Making Crystalline Salts : 4’’-n-pentoxy-[1,1’:4,1’’]-terphenyl-4-carboxylic acid

 

The title compound  has a structure containing three phenyl groups strung together end to end through their para positions with an n-pentyloxy cap at one terminal and a carboxylic acid function at the other. All three rings are therefore para disubstituted. This title acid, after coupling with the Echinocandin B macrocycle, gave a product with excellent pharmaceutical properties. The substructure compound, as an amide derivative, constitutes the drug Anidulafungin. Since this linkage would be broken during the drug’s metabolism and since the drug has been found to be safe, this would seem to establish that this acid component is fairly nontoxic in the human metabolism.

This compound is a moderately high melting solid and would be predicted to give high melting salts as well as covalent derivatives. For example, the esters and amides are more likely to be solids using this rather than other carboxyl coupling partners. The rigidity of the carbon skeleton might be expected to cause it to have a significant steric effect even remote to the point of attachment with another substructure. The expense of this potential protecting group would be ameliorated because the acid could be easily recovered and recrystallized to a dependable purity for recycling. Its use might be particularly promising where the process intermediates are expected to be liquids, oils, or waxes such as in prostaglandin synthesis. 

The Problem of Oiling Out in Chemical Process Development

It is often called LLPS (liquid-liquid phase separation). It can be helpful if you are performing a liquid-liquid extraction and are fearful of emulsions. When you are trying to perform a crystallization or recrystallization LLPS is bad news because it is what we practitioners call 'oiling out'. 

As KiloMentor has often repeated, when devising a process, chemists are really guessing when they try to assess how well and how easily they will be able to purify those solid intermediates they need to recover/purify by crystallization. One of the mantras of the KiloMentor blog is: Choose process schemes that incorporate rugged scalable phase switches that either improve purity before a final crystallization or enable process steps to be telescoped to avoid entirely some of these crystallizations.

Having the substance you are trying to crystallize oil out is high on the list of those things you don’t want to happen. This is particularly true on large scale, because you are working in a vessel with a stirrer that does not scrape the walls and in which you can’t easily observe what is happening. Because oiling out at scale occurs down inside a poorly illuminated reactor, in the situation where that oil eventually solidifies, you may never learn what happened. All that may be evident is that the purification failed and the impurities are not uniformly distributed in the product.

Even in the most rugged reaction sequences successful crystallization of solid intermediates may be required and reducing the likelihood of oiling out of low-melting solids will be needed to avoid a major dislocation.

Only one article ever accepted by the Journal of Organic Process Research & Development contained ‘oiling out’ in its title [ Jie Lu et al., Org. Process Res. & Dev. 2012, 16, 442-446]. 

Only three pages in Niel Anderson’s,  Practical Process Research & Development, First Edition pertain to oiling out problems in crystallization (Sorry – I can’t afford to pay for both First and Second Editions to check for updates). In the one example at pg. 280 of the first edition, Anderson cites the case of a pharmaceutical product isolation where oiling out is avoided by adjusting processing to make sure that plenty of seeds are available. In this example, the drug captopril was crystallized by first forming a thick seeding suspension comprising some previously isolated captopril solid, acetic acid, and sodium chloride all together in water, and then followed by adding slowly and simultaneously (i) the strongly basic hydrolyzate obtained by first treating  S-acetyl captopril methyl ester with 3.3 equivalents of sodium hydroxide and (ii) aqueous HCl; the latter, in such amounts that the crystallizer contents always remained acidic.  By forming the captopril in situ in the presence, throughout the entire nucleation, of many preformed captopril crystallites, oil was not formed even though there was a high concentration of sodium chloride in the water.

The oiling-out phenomena can be categorized by two parameters. The first is temperature. Oiling out near or above the solute’s melting point should not be surprising at all. Separation of solid should not be expected if the solution saturation is exceeded at a temperature where that substrate is expected to be a liquid. In this case, the solution is too concentrated for work at that temperature. More serious is oiling out that occurs near and above the melting point of the main solute and most worrisome is oiling out that occurs below that melting point.

The second parameter pertains to solvents. There is oiling out from a single solvent or from a solvent combination. It seems to me that oiling out from a single solvent below the anticipated melting point of the substrate most often arises simply because the rate of phase separation is faster than the rate of nucleation. The antidotes should be one or both slower cooling and seeding. 

Oiling out from a solvent mixture appears more frequently and has a more obvious explanation. The separating solute can cause the solvent combination to demix and separate. This situation will be most common when the solvent mixture is composed of solvents of quite different polarities; for example ethanol-hexane.

Another scenario can play out when the main impurities begin to separate in preference to the desired product and they contaminate the emerging product enough to reduce its melting point below the solution temperature. This is likely to arise when trying to purify a main substance with more polar impurities by crystallizing from a strongly apolar solvent or purifying a main substance with predominantly less polar impurities from a strongly polar solvent. 

Idebenone

It would seem to me that this is the situation in the Jie Lu et al. example cited earlier. Idebenone comprises a dialkyl-dimethoxyl-p-quinone with a primary hydroxyl in the side chain. Idebenone is the subject of the Liu paper. The two impurities of concern in it each have one or the other of the two methoxyls demethylated to a phenolic hydroxyl. Thus these impurities are distinctly more polar than idebenone itself, yet this idebenone is being recrystallized from methylene chloride-hexane. I, myself, have extensive experience with idebenone processing. From this experience, I know that it can be recrystallized in high yield from ethanol-water and this would most likely be a preferred method for getting rid of these phenolic impurities without any risk of oiling out. 

Problems using either Solid Desiccants to Dry Organic Solutions or Charcoal to Decolorize when working At-Scale

Differing Time Requirements for Pilot Plant Operations

Neal G. Anderson in his book, Practical Process Research and Development, breaks down the time required to perform the operation of drying an organic solution with a solid inorganic desiccant such as sodium sulfate. He estimates that, including equipment preparation and equipment cleaning, the unit operation would take two operators 8 hours. Only three hours are actually spent performing the physical operations that parallel the laboratory manipulations. An additional one hour is spent assembling and testing the filter and its associated piping. At the end of the actual filtration, four hours are required to rinse, disassemble, clean, validate, and store the equipment ready for reuse.

These same times apply to the operation of treating an organic solution with decolorizing charcoal and filtering it off.

Chemists more accustomed to working at a laboratory scale cannot initially conceive how these preparation and cleaning phases can add up to so much time. In the lab, we are accustomed to using glassware that is cleaned with a quick acetone rinse, transferring liquids simply by gravity, pouring hot fluids through the air, and assessing cleanliness by visual inspection. On-scale, however, whether the solution is hot or cold, filtration needs to be done in a closed inert system. The generally uncomplicated pilot plant filtration compares more closely to working in the laboratory with a toxic liquid or a pyrophoric solid in Schlenk-tube equipment.

With regard to drying organic solvents, once it is grasped that removing water is a difficult and costly process, chemists who are trying to be process minded can try to avoid adding water into a reaction mixture or alternately think about ways to remove it without the complication of additional equipment. Pouring a reaction mixture into water or adding water into an organic reaction mixture is a frequent operation incorporated into a reaction step. We should be trying to rethink this standard approach, which often uses a humongous excess of water. To further complicate things this drown out is usually done rapidly, which again cannot be duplicated in the plant.

Reaction Quenching

Why is this done? One reason, as suggested above, is that it usually stops the reaction. This is why the operation is often called quenching the reaction. Water is an evaporable, cheap, low strength buffer. Since the strongest acid that can exist in a wet solution is H₃O⁺, and the strongest base  ^-OH, many reactions that require either more acidic or more basic conditions than these are stopped by adding water. Another mechanism also operates when a water quench is performed. The addition of sufficient water to produce two immiscible phases in the reactor quenches by phase separation of reaction participants. Some chemicals are much more soluble in the aqueous phase while others are much more soluble in the organic layer. Separation of reactants, reagents, products, co-products and byproducts into one or the other of water or organic usually reduces the rate of reaction among them to essentially zero. A third mode by which aqueous quenching works is the rapid decomposition of one of the reactants by water.

If we remove a water quenching step in a process, it is still usually necessary to stop the reaction at the correct endpoint by other means. Whatever that operation is, it must be rapid. What can be done depends upon the characteristics of the reaction itself. Whatever the quenching additive used it is very preferably inexpensive.

An option is to use some water, but not the large volumes that result in a second phase. In the laboratory, the volume of water used in a quench often doubles the total volume in the reaction vessel. When the quench is done by pouring, with vigorous stirring, the reaction mixture onto an ice-water slush, the final volume is usually still greater. Such large volumes are frequently used ‘just to be sure’ because the laboratory reactions are not investigated enough to figure out the actual lowest amount of water that is needed. At scale, the total reactor volume immediately after quenching, but before the separation of this water, is very often the point of maximum reactor volume and, as such, it controls the batch size. That is, using larger quench volumes than necessary immediately lowers the possible throughput for the process step. Increasing the volume at the point of maximum volume in the reactor increases the cost per kilogram of intermediate product from the step. Investigating the quench that is sufficient to do the job but keeps the lowest maximum reactor volume can pay off by keeping the throughput of the process step high. This is most critical for the early steps of a multistep process.

Increased Waste and Disposal Costs

Creating a large aqueous phase saturated with organic substances also creates a larger waste disposal cost. The water becomes a waste stream. Water contaminated with a saturation level of organic contaminants cannot be discharged to a municipal sewage treatment system. Moreover, it is one that cannot be treated inexpensively by burning. Reducing the amount of the water quench phase will greatly reduce the kilograms of waste produced per kilogram of product. It is an easy way to make a process step both more environmentally friendly and cheaper.

Replacements for copious amounts of water could be ammonium chloride or ammonium acetate both of which are quite soluble in lower alcohols and which only create volatile residues. These would also satisfy the buffering function. In other situations acetic acid or ammonia might work. Grignard reactions and hydride reduction reactions are often treated with ethyl acetate.

Even if water-organic partitioning in a liquid-liquid extraction cannot be avoided in the work-up procedure, using a minimum volume of quench can still be an advantage. Once the reaction has been stopped it may be possible to reduce the volume of the reaction mixture by distillation before adding the larger portion of the aqueous extraction phase. In this way, the volume/kg of product at the point of maximum volume will still be kept lower and the throughput maintained or increased.  For example, an alkylation under strongly basic anhydrous conditions could be quenched with ammonium chloride in methanol and then the volume of the reaction mixture could be reduced by half using vacuum distillation before adding water for an acid-base extraction to purify the product by phase shifting. The volume/kg at the point of maximum volume might thus be reduced by half and the number of runs required to meet a production target cut in half.

Scientists more accustomed to laboratory synthesis would benefit by recognizing that in a plant setting, one usually cannot do the equivalent of simply switching from a 500 ml to a 1-liter reaction flask when they want to make more material per run. The chemical reactors available are few in number and cannot be simply interchanged since, for example, some may be glass-lined and others stainless steel, or one may have a different heating/cooling capability than another.  In an environment with this rigidity, increasing the throughput per batch by increasing the starting materials charged becomes a big deal!

Alternative Methods for Drying an Organic Liquid

Herein, drying an organic liquid  means reducing the water content; most often to negligible or alternately to any other practically acceptable level.

Drying is often not necessary. In the laboratory drying an organic solution is conducted routinely. Most often no effort is made to determine whether it is useful or necessary. Drying may be performed because the cut between an aqueous and an immiscible organic phase may not be perfect and there may be a concern about small water droplets in the organic layer. In the lab, it is easier to be overly safe rather than sorry. In the plant, discovering that drying is not an operational concern can save significant time and money.

Drying a solution may be essential. The most widely known drying method in the plant setting is azeotropic distillation.  That the overall volume is reduced during the operation is an advantage. Also, because almost all multipurpose reactors have reflux and simple distillation capabilities, no additional equipment is contacted by the product. 

That the operation requires heating and also frequently a vacuum are disadvantages. Also, unfortunately, water is very often less volatile than the organic solvent used in reactions and many lower boiling solvents do not form water azeotropes. It is not much of an answer to suggest that the process step reaction solvent be switched to one that does form a useful azeotrope because this reduces solvent choice which is one of the most influential variables that can be used in reaction yield optimization. Choice of the reaction solvent should be kept as open as possible.

Where heating is a concern or where an azeotropic method is not possible, operators at scale have found that they can pass a solution through a fixed bed of molecular sieves to achieve drying. Molecular sieves that have absorbed water can be renewed by the passage of hot dry air once the sieves have been washed to remove external contamination. 

Another procedure that can remove water from a solution is to add a reagent that reacts preferentially with water and produces products that are more easily removed than water. Triethyl orthoformate or trimethyl orthoacetate, for example, can consume water and produce ester/alcohol products. DMF dimethylacetal can react with water to produce dimethylformamide and methanol. Bis-trimethylsilyl-trifluoracetamide can react with water to give Bis trimethylsilylether and trifluoracetamide.

Replacing Charcoal Treatment

Charcoal decolorizing treatment has even more severe time and equipment disadvantages than drying with inorganic salts. Because charcoal is insoluble and very finely divided in most cases, proper cleaning is harder and the evidence of inadequate cleaning (black particles) is embarrassingly obvious. Methods of purification of crude solids that are less burdensome are preferred but there are occasions when charcoaling cannot be replaced. At scale, producers often have dedicated charcoaling equipment so that most of their equipment is never contacted with charcoal. Sometimes pumping a solution through a fixed bed containing charcoal works just as well as adding it into the organic solution.

Reducing the extraneous color from process intermediates may be intellectually satisfying, but the question needs to be asked, "Is reducing the color a critical parameter of the process. Could the color be ignored? Would it be purged by further processing or removed closer to the end of the process?