Exhaustive Digestion as a Simple Prelude to High Yield Crystallization

 

This article is about the technique I call ‘exhaustive digestion’. This is a phrase that I used in my blog entitled, Getting Better Recovery from Crystallization. In this blog, I make the point that chemists scaling up a laboratory process often fail to do adequate preliminary purification of a solid and as a consequence get a lower recovery from their crystallization than is possible if a preliminary treatment were done. I proposed digestion or trituration as pretreatments before crystallization. I don’t know whether there are any widely accepted differences in meaning between the words digestion and trituration but, when used in my blogs, ‘trituration’ can be understood to apply equally to oils or solids while ‘digestion’ will only be applied to solids. Second, ‘trituration’ can be done with either hot or cold solvent while ‘digestion’ and particularly ‘exhaustive digestion’ as described hereunder is to be done with hot, boiling liquid, usually a rather poor solvent for the desired product. 

Exhaustive digestion is related to swish trituration which I have written about earlier. The difference is that the swish trituration employs its anti-solvent system at ambient temperature and does not measure the boiling point depression of the liquid to assess the extent of purification. 

In the old literature, the progress of purification by digestion on the laboratory scale was followed by observing the change in the boiling point of the digesting solvent or solvent mixture using a Beckmann thermometer, which is a thermometer that can measure differences as small as one-tenth of one Centigrade degree. I don’t know how long it has been since I have seen anyone use a Beckmann thermometer but, at least in my undergraduate Phys. Chem., (over 50 years ago) it was done quite often. The scale on a Beckmann thermometer is only 10-20 degrees but its range can be adjusted by removing or adding mercury to the column of mercury used for measurement by moving mercury back and forth from an attached mercury reservoir. Today a digital thermometer is more likely to easily achieve the same or better precision of measurement without managing liquid mercury.

I will illustrate exhaustive digestion by quoting from an experiment taken from Laboratory Technique in Organic Chemistry by Avery Adrian Morton, First Edition, McGraw-Hill, 1938,  pg. 128-229.

“The apparatus consists of a tube or flask of such size that a large part of the bulb is covered by the solid material being extracted, a thermometer graduated in tenths of a degree, a reflux condenser, and a water or oil bath. Place the sample in the container, insert the thermometer so the bulb is immersed in or covered by the sample, and add solvent to cover the solid and the thermometer bulb. The solvent is usually one that will dissolve impurities without appreciable quantities of the desired material. Petroleum ether or ligroin is often suitable. A mixture of solvents may be employed without affecting their utility in this experiment, as long as the composition of the mixture is not varied. Reflux the solution [I think slurry is intended] using a water or oil bath as a source of heat. After about 20 minutes the temperature has reached a constant level. Usually, no trouble will be experienced from superheating as long as a solid phase is present, although a little agitation with the thermometer is sometimes needed. Filter either by pouring into a Buchner funnel or by using a filter stick. Add more of the same solvent, reflux once more, and observe the temperature after equilibrium conditions have been reached. Continue the operation until the temperature of two or more successive operations is identical. The more soluble impurities have now been removed, and the solution contains only the pure compound or the compound with [less soluble type] impurities that have not been entirely removed…. The operation may be continued if desired until all the product has been dissolved. A change of solvent [may be implemented at some point for further purification]….. Constancy in the boiling points of successive portions constitutes further evidence of the purity of the material, whereas a drop in the boiling point is evidence of the exhaustion of still another component. Usually, the final portions of exhaustive digestion are pure materials.”

Continuously measuring and recording the changes, in real-time, of the boiling point without any actual sampling makes this an early application of Process Analytical  Technology (PAT). The improvements in the size and sensitivity of temperature measuring devices for following small changes in boiling point under conditions where superheating does not occur deserve more frequent consideration.

Exhaustive digestion if used at scale would require a more rugged method for measuring small changes in the boiling point of the slurry than the old Beckmann thermometer. Fortunately, modern digital thermometers meet that need.

Exhaustive digestion, similar to swish trituration, may profit from using either water-rich or hydrocarbon-rich binary azeotropes,  as the anti-solvents, as the desired product’s solubility dictates. Some are suggested in the following list:

97.0% water 3.0% acetic acid azeotrope bp 76.6ºC

91.0% water 9.0% benzyl alcohol azeotrope bp 99.9ºC

87.1% heptane 12.9% water azeotrope bp 79.2ºC

94.4% hexane 5.6% water azeotrope bp 61.6ºC

95.5% hexane 4.5% allyl alcohol azeotrope bp 65.5ºC

97.0% hexane 3.0% 1-butanol azeotrope bp 67.0ºC

91.5% cyclohexane 8.5% water azeotrope bp 69.8ºC

83.7% acetonitrile 16.3% water azeotrope bp 76.5 C

72.9% allyl alcohol 27.1% water azeotrope bp 88.2ºC

66.0% allyl cyanide 34.0% water azeotrope bp 89.4ºC

77.5% formic acid 22.5% water azeotrope bp 107.1ºC

Genotoxic Impurity Considerations

Genotoxicity is the capacity of a substance to damage genetic material such as DNA, and thus cause mutations or possibly cancer. Genotoxic impurities need to be avoided in a pharmaceutical product because they can possibly cause serious health consequences even at levels less than standard analytical methods can detect.

It has been argued that most genotoxic reagents/intermediates present in syntheses that have four or more stages prior to API isolation/purification are likely to be deactivated by reaction with other reagents or purged by dissolution in solvents, removed by vacuum distillation or other phase switching operations; however, regulatory reviewers who have a vested interest in emphasizing the seriousness of genotoxins show little sympathy for such arguments. Government reviewers tend predominantly to request quantitative analytical information on the API and/or intermediates possibly combined with the results of spiking experiments (impurity fate analysis) and could well require these in order to demonstrate the absence of carryover of potential genotoxic impurities (PGIs), at a suitably low (TTC) level. Such investigations can be highly resource-intensive and challenging, particularly in respect of developing validated analytical methods for low levels of multiple PGIs. Moreover, limits for certain PGIs may become part of the API specification and so become an ongoing quality-control commitment.

However, concerns usually relate to late-stage operations that involve potential genotoxic impurities PGIs. One EU review raised as a major objection, potential residues of mesityl oxide (4-methyl-3-penten-2-one) in a drug substance crystallized from acetone.  Another cited possible traces of alkyl mesylates (methyl, ethyl and isopropyl mesylates) in a mesylate salt drug substance.

Unless the threatened concern about the PGI is identified at the early planning stage, modifying a route that is otherwise the best proposal should be the last resort.  Nonetheless, modification of the API synthesis in a way that minimizes PGI levels in the drug substance has been undertaken in some cases; recent published examples pertain to formaldehyde, chloroalkanes and acetamide.

Given the difficulties in obtaining reliable predictions, particularly for aromatic amines, it may be prudent to undertake an Ames assay on any PGI containing an aromatic amine structural alert, if data are not already available in the public domain. Aromatic amines that otherwise make admirable process intermediates because of the multiple options for their purification, should where it is feasible be transformed to more acceptable functionalities early in a process route.

The potential action of small alkyl alkylating agents presents a particularly frequent and acute problem for pharmaceutical products because so many of them in their API salt form are hydrochlorides, methanesulfonates, tosylates and besylates as well as to a much smaller extent, hydrobromides.

These pharmaceutical salts are most often prepared in the final chemical transformation of a sequence and so any PGI has the highest probability of lingering behind in the API.

The best evidence that these concerns about PGIs arising from making the aforementioned pharmaceutical salts are without any foundation is the epidemiological evidence of the health of medicinal and process chemists who have been making and using these salts for years and years without any relative health damage.

The apparent lack of health damage is readily explained since the reactivity of halo compounds in biological systems can be predicted to a significant extent on the basis of their relevant chemical properties such as alkylating potential and susceptibility to hydrolysis. Thus, bromo compounds are expected to be more reactive than chloro compounds. SN1 and SN2 characteristics will determine the nature and the extent of reactivity towards nucleophiles, and steric factors may also play a part in some cases. In the 4-(p-nitrobenzyl)pyridine alkylation assay, alkyl halides generally show negligible activity, methyl methanesulfonate (MMS) being at least 40 times more active than ethyl, propyl or butyl bromide. Allyl bromide appears to be more active (around one-eighth of the activity of MMS although allyl chloride shows minimal activity. Benzyl chloride, while rather active for a chloro compound, is around 20-fold less active than allyl bromide. Owing to their volatility and/or hydrophobicity many alkyl halides show negative results in conventional Ames Salmonella assays, and it is often necessary to employ vapour phase exposure in a closed system (using a desiccator for example) in order to obtain positive results. Most alkyl halides, especially bromides, are Ames positive (using a closed test system if necessary), although 1-chloropropane, 1-chlorobutane and neopentyl bromide are all Ames-negative. As expected, based on their lack of alkylating activity, both chloro- and bromobenzene are Ames-negative. Some unsaturated halo compounds have the potential to be metabolized to form quite active mutagenic molecular species. For example, evidence suggests that oxidative biotransformation of vinyl chloride produces chloroethylene oxide and 2-chloroacetaldehyde as active metabolites. Binding of bromobenzene-3,4-oxide to liver proteins is thought to account for the hepatotoxicity of bromobenzene. The predominant metabolic pathway for simple alkyl halides is halide displacement by glutathione, although some C-hydroxylation reactions may occur.

Rodent bioassay data on alkyl halides strongly suggest that these compounds are either non-carcinogens (1- chlorobutane, bromomethane) or low-potency carcinogens (chloroethane, bromoethane). Both chloroethane and bromoethane produced an increased incidence of a rare type of endometrial tumour in female mice and it seems highly plausible that the carcinogenic effect is caused by a species-/ gender-specific stress-related adrenal overstimulation and excessive corticosteroid production. The rodent carcinogenicity profile for chloroethane (increased incidence of a rare tumour type at an extremely high concentration of 15,000 ppm in one species/gender) is thus much closer to that for a non-genotoxic carcinogen than for a genotoxic carcinogen. Thus, the (feeble) alkylating activity of chloroethane seems largely incidental to its carcinogenic activity, a scenario likely to apply to many other similar alkyl halides. This prediction is strongly supported by the fact that benzyl, ethyl, isopropyl, and trityl bromides were inactive as carcinogens at doses up to 0.83, 12.5, 8.3, and 0.25 mmol/kg, respectively, when administered by single subcutaneous injection to female rats. A number of independent expert assessments are available on halo compounds.

Acceptable/tolerable exposures are expressed in various ways, for example as minimal risk levels (MRLs) by the Agency for Toxic Substances and Disease Registry, or as reference concentrations/doses (RfCs/RfDs) by the U.S. Environmental Protection Agency. There is a clear consensus that chloroethane is less hazardous than the more reactive chloromethane, although recommended safe exposures for the former range from the highly conservative OEHHA value of 150 μg/day to 200 mg/day (10 mg/m³ at an average air intake of 20 m³/day⁴³) based on the EPA IRIS assessment.

 Acceptable exposures in the context of genotoxic impurities can also be calculated on the basis of the TD50 values as described above, resulting in PDEs of 1810 and 149 μg/day for chloroethane and bromoethane respectively. A PDE for non-carcinogenic 1-chlorobutane could be determined using ICH Q3C (R3) methodology on the basis of the most appropriate NOAEL in lifetime studies.

Quotes from Org. Process Res. Dev. 2011, 15, 1243–1246

[Compounds can be grouped together that have] strikingly different reactivities, isopropyl chloride and isopropyl mesylate for example. Isopropyl chloride is Ames-negative in assays using standard conditions; testing has to be carried out in a desiccator to obtain a (feebly) positive result; on the other hand, isopropyl mesylate (Swain-Scott s = 0.29) is a potent mutagen in the standard Ames test and gives positive results in several in vivo assays. Hydrolysis half-lives at pH 7.0 and 25 C are 38 days (the same as for chloroethane; TD50 1810 mg/kg/day) and 4.5 h, respectively. Unfortunately, no rodent bioassay data appear to be available for either compound, although 1,2- dichloropropane is Ames-positive in standard assays and has a mouse TD50 of 276 mg/kg/day (negative in the rat). A Risk Specific Dose (RSD) for isopropyl chloride of approximately 37 μg/day (for a 50 kg patient) has been determined by Bercu et al. and Contrera based on QSAR (quantitative structure- activity relationship) techniques using regression analysis of “training sets” of TD50 data. [The choice of compounds in the isopropyl chloride training set could be questioned in that it contained several non-genotoxic polychloro compounds but not 1,2-dichloropropane or chloroethane.] Use of QSAR models to predict genotoxic/carcinogenic potency relies on rule-based techniques (such as DEREK) or statistical techniques (using regression analysis of training data sets), and particularly in relation to carcinogenicity, the latter approach is considered to provide more reliable results. Since halo compounds in particular and other compounds containing “Ashby alerts” are often identified as PGIs, it is interesting to note that in the determination of the TTC limit Kroes et al. classified only two such compounds (5% of the total in the data set) in the lowest potency category (equivalent to 1.5 μg/day), but since the data set is non-transparent it is not possible to identify these two key compounds. [In a prior publication using essentially the same data set, the TD50 for methyl methanesulfonate is listed as 0.178 mg/kg/day, 179 times lower than the true value, and it is unclear whether this miss-transcription was carried forward to the data set used by Kroes et al. in the more “definitive” publication.] In the Kroes publication, no carbamates were classified in the lowest potency category, suggesting that the standard TTC of 1.5 μg/day may be unnecessarily conservative for this structural class.  

How does one Produce the Most Likely Impurities in the Product from a Process Step?

In order to develop a good purity analysis for an organic substance, one needs to have some method to assess different methods. A better method separates and quantifies more impurities from the product. A better method increases the degree of separation between the closest impurity and the product without losing separation for any other impurity. A better method separates more cleanly an impurity designed to have a very minor structural difference from the product.

 

But is there a way to prepare a  resolution standard with larger amounts of the most likely potential impurities of the product? This becomes an important practical matter.

There are two types of impurities. Impurities that are derived from product degradation; that is, they come via the desired product and arise from reactions of the desired product occurring after it has been isolated and purified. A separate type of impurity is one that is formed at the time of the synthesis of the desired product and which is not completely removed by the isolation, separation, and purification processes performed before packaging the final product. Such impurities are characteristic of the process. 

It is this second type that is considered here. These impurities are produced in greater or lesser amounts by variations from the proper continuous variables controlling the process.

 

New previously unobserved impurities are usually created by changing the discontinuous variables of the process step, such as reagents, reagent purity, reagent/substrate ratio, solvent, solvent purity, substrate purity, processing chemicals…

The most significant continuous variables are time and temperature.

The one that can produce the most profound or substantial changes in chemical reactivity is temperature. The effect of time is usually more limited because if we are seeking a complete reaction it is likely that the transformation is self-terminating and will essentially stop when the correct time has expired with relatively less occurring after this required reaction time. Increasing the temperature by 10 C according to a rule of thumb should double the rate of reaction. This will also allow competing reactions that are limited under the most preferred conditions to compete and produce by-products.

Thus, an increase in the temperatures of each of the different stages of the reaction by 10C and a decrease in the time by half in each stage should produce more impurities in the final product and these impurities should reflect realistic possible impurities.   If possible the extent of disappearance of starting material should be kept about the same.

Product separated from a process step stressed in this way should show increased amounts of the most likely process impurities.

If it does not, this is important information for your determination of critical parameters that you will need to work on at some point. Tests that provide a resolution standard go hand-in-hand with examining critical parameters!

Use of free radical inhibitors or antioxidants to increase the overall yield of organic synthesis steps

The use of radical inhibitors or antioxidants to improve yields does not appear to have many precedents in organic synthesis. A keyword search in 2011 provided only two references- both related to the stabilization of m-chloroperbenzoic acid towards thermal degradation during the epoxidation of resistant olefins.

Y. Kishi, M. Aratani, H. Tanino, T. Fukuyama and T. Goto, J.C.S. Chem. Comm. 1972  64 and 

D.M. Tal, Steroids (1989),  54(1), 113-22.

The best inhibitor found by Kishi for stabilizing m-chloroperbenzoic acid was 4,4’-thiobis-(6-t-butyl-3-methyl-phenol) that allowed 100% of an m-chloroperenzoic acid charge to be retained after 3 hours heating at 90 C in ethylene dichloride. Octene-1, dodecene-1 and methyl methacrylate were quantitatively epoxidized using such stabilized oxidant.

Synthetic chemists apparently assume that free radical reactions do not occur unless free radical initiators are present in the reaction mixture or unless the reaction mixture is irradiated. It might seem they think it can’t happen unless they are intending it to happen. Obviously, this is not true! Free radical reactions can take place not just during the contemplated reaction phase but during the work-up of the reaction when we might think that all the reacting is stopped. Actually, the opportunity is greater in the work-up phase; this phase usually takes more time, particularly when the process is being scaled up.

Are free-radical reactions inhibited by particular pH ranges of the solvent medium? No, they are not. The most frequent type of free radical reaction is oxidation and only the relative amounts of different species that can be oxidized are affected by pH not particularly the oxidation rates.

Oxidation often produces coloured products when it can introduce new unsaturation into molecules. The presence of unexpected colour in a reaction is suggestive of unanticipated oxidation. I recall that in the preparation of some aniline compounds the procedure teaches the addition of hydrogen sulfide to the aqueous phase during isolation to prevent colour development from exposure to air during workup and crystallization. The usual response to a colored product is to use charcoal in the recrystallization rather than trying to prevent colored by-products in the first place.

If you are performing a distillation and the contents of the still pot are darkening why wouldn't you add an antioxidant? Answer- I've never thought of it.

Strategies for Characterizing Trace Impurities Important for Regulatory Compliance in Pharmaceutical Synthesis

In setting the specifications for a pharmaceutical substance, its unknown impurities must typically be less than 0.1% area/area with respect to the main signal peak using the standard detector for that particular method. The usual instrumentation is reverse-phase liquid chromatography.  Even so, regulatory agencies prefer that impurities be identified.  Impurities that might structurally resemble genotoxic substances should be absent. If an unequivocal structure has been assigned to a minor component, it is possible a higher concentration of that impurity level can be accepted by regulators providing an extra incentive to discover that structure.

HPLC-Mass Spectroscopy & HPLC-MS-MS

With the correct instrumentation and method development, a skilled analyst can greatly reduce the number of possible structures for an HPLC impurity peak.  Nearly always this requires that the HPLC mobile phase consist of either a volatile salt buffer, such as ammonium acetate, or no buffer at all. When developing new analytical test methods the first choice for buffers should be volatile buffers.  General analytical methods are well described and easily searchable so no more need be said here.

 It is worth discussing what can be done when the above standard approaches fail or are inapplicable for some reason.  For one thing, it is sometimes possible to narrow down a list of suspect structures and devise means to isolate or synthesize a sample of the hypothesized material.

Clues Suggestive of Structure

1. Is the impurity acidic, basic or neutral?  This can be determined by aqueous acidic and basic extractions.

2. Does the quantity of impurity remain fairly constant with respect to the main component even when different methods of purification are tried? If yes, this is suggestive,  either of a homologous structure in which there is some slight difference in a side-chain between the impurity and the active drug, or of a positional isomer relationship between the impurity and the main constituent.  Such impurities usually come from impurities in one of the starting material building blocks.

3. HPLC is the most common present-day method of pharmaceutical analysis and diode array variable UV wavelength detectors are routinely available for such analyses. So how does the UV spectrum of the impurity measured by the diode array detector compare to the UV spectrum of the main product?  Does this narrow down structural possibilities?

4. Is the impurity’s increase most probably a function of the degree of scale-up? 

5. Have you performed a laboratory run in which the duration of the addition times are the same as in the plant?  Even if obtaining very slow rates of addition on the laboratory scale is too technically difficult or requires unavailable equipment the same effect might be obtained by adding a quarter of the dropping funnel charge, then waiting for ¼ of the plant addition time; adding the next quarter of the charge and waiting another quarter of the plant addition time and so on.  Impurities often arise from the wider variation in the stoichiometric ratios that are present during the lengthened addition period on scale.

6. Is the impurity occurring only in the most recent runs and not occurring under somewhat different earlier conditions?  This is a key question that arises out of Kepner-Tregoe problem analysis.  The answer may trigger an insightful guess at the structure of the impurity connected with the possible change that caused it.

7. Do you already know some means to obtain a sample free of this impurity?  Even if this is expensive and impractical it provides information to fashion a separation/identification method.

8. Is the main component (active API) reactive with some easily removed and quantitatively reacting material (such as hydrogen)?

9. If one prepares the sample for analysis differently, does the impurity increase, decrease, or remain the same? That is to say- is it actually an artefact of the analytical method?

Swish Chromatography

Does “swish chromatography” increase the relative concentration of the impurity versus the main peak? 

Could trituration (swish TLC) with an atypical liquid in which the main component is poorly soluble give an enriched composition? Among atypical solvent I would include methylamine, ethylamine, sulphur dioxide, dinitrogen-pentoxide, carbon disulfide, nitromethane, acetonitrile, perfluoromethylcyclohexane, tetrachloroethylene, trifluoroacetic acid,  carbon tetrachloride, dicyclopentadiene and perylene sulfone. These solvents are quite volatile and can be readily removed. The perylene sulfone and dicyclopentadiene both  decompose to volatiles upon heating under vacuum and methylamine, ethylamine, carbon disulfide, sulphur dioxide and dinitrogen pentoxide are gases under normal conditions or are very volatile.

Is extractive crystallization possible to selectively phase switch the impurity?

Switching from HPLC to TLC for Isolation

Is there some means to find the equivalent TLC,  Rf for the impurity, which you are identifying in HPLC by a RRT?

On this question, I do not know what the literature provides but here is a possible method: Suppose one runs a preparative TLC using a solvent system that gives an Rf of 0.05-0.1 for the main component but one performs multiple elutions on the plate and then removes the adsorbent in horizontal strips from that plate and runs these according to one’s HPLC procedure.  Multiple elutions with a poor solvent system provides the best chance for separation of impurities. If the impurity is not co-eluting on TLC, this HPLC analysis of bands from the prep plate will locate the TLC Rf range where the HPLC impurity of interest is located. Even when this Rf region has been located, the impurity is not necessarily one that can be readily visualized on the TLC plate. The HPLC impurity may be undetectable by conventional TLC at the concentration being spotted.  Nevertheless, if it works, you have found a preparative scaleable method for separating the impurity even if you cannot detect the impurity by normal visualization.  If the location of the HPLC impurity is well resolved from the main compound, you will be able to simplify your TLC method by increasing the polarity of the elution solvent trying to get a method that only requires a single development of the plate, but if the separation is difficult, the simplification may not be possible and you will have to live with a multiple elution method.  

Run a preparative scale column, collecting multiple fractions in the Rf region where your qualitative study has shown the impurity to come and analyze these fractions by HPLC to pick out those with the highest concentration of the unknown impurity.

Hypothesizing the Structure of Potential Impurities

Perhaps you can hypothesize a possible identity for the troublesome impurity from the answers to series of questions. 
What modifications of conditions increase the impurity? What modifications decrease it? Does the Rf provide a clue to the polarity and so make some structures more likely and others less likely? Is the Rf consistent with the proposed functional groups? Does the degree that it partitions between different liquid phases provide a clue?

If you have a potential mechanism for formation of the hypothesized impurity, is it going to be easier to simply synthesize this potential impurity and test it instead of trying to isolate and purify the impurity from the product mix?

Is a composition that results from intramolecular self-reaction possible? Think of possible side reactions that could yield such products.  These materials, because they have a molecular weight about double the API itself and similar functional groups can be difficult to crystallize out.  Size exclusion chromatography can be very powerful for distinguishing materials differing by 700 atomic units.  Advantageously the high molecular weight impurity elutes first!

Is reaction with a solvent, solvent impurity, or reagent impurity possible? 

Could the impurity be present in a very small amount but have a very large extinction coefficient compared to the main substance, so that the HPLC detector signal was exaggerated in a molar comparison?  The diode array spectrum may be useful to assess the likelihood of such a situation. Absorptions with very high extinction coefficients usually have extensive conjugation and longer wavelength absorption.

Can color-forming TLC reagents be useful to identify the functionality in the impurity?  If the impurity is separable by TLC one can often perform derivatizations on the TLC plate, which signal some functionality by a coloration of the spot or band. In contrast, this is not readily applicable to HPLC separation. 

Could you obtain a KBr/IR by concentrating a preparative plate sample on a triangle piece of KBr. Could you obtain a mass spectrum or ms/ms from a TLC or HPLC sample?

Does steam distillation help to decrease the impurity level?  If steam distillation reduces an impurity the most frequent conclusion is that the impurity was some solvent-like material.

Combining Yield Optimization with Impurity Identification

Performing simplex or other process step optimization can provide some conditions, which result in a dramatic increase in the impurity level. These conditions are unsuitable for optimizing but such a sample may be an easier mixture from which to purify the impurity.

Using Coloured Derivatizing Agents

Can colored derivatives be useful? Note you will not know whether the impurity of interest formed a derivative unless you know that it contained the prerequisite functional group.  The main component should not form the derivative.

Chromatography of colored derivatives is simpler because the experiment can be followed visually on the chromatographic plate or on a column whichever is used.

4-phenyl-azo-benzenesulfonic acid chloride has been used as a derivatizing agent for primary and secondary amines [R.D. Desai and C.V. Mehta, Indian J. Pharm. 13, 211 (1951).] Hydrolysis in conc. HCl-dioxane E.O. Woolfork, W.E. Reynods and J.L. Mason J. Org. Chem. 24, 1445 (1959).

4-phenyl-azo-benzoyl chloride is a derivatization for alcohols. [E.O. Woolfork, F-E. Beach and P. McPherson, J. Org. Chem. 20, 391 (1955).]

4-(4-nitrophenyl-azo)-benzoylchloride can be used to form derivatives from primary and secondary alcohols, amines and thiols. E. Hecker, Ber. 88, 1666(1955).

E.S. Amin and E. Hecker Ber. 89, 695 (1956).

A. Butenandt, T. Beckmann and E. Hecker, Z. Physiol. Chem.324, 71 (1961).

A. Butenandt, D. Stamm and E. Hecker, Ber. 94, 1931 (1961).

Because an impurity that moves closely with the main product probably has the same polar functionalities as the main compound, if one can make a colored derivative of the main composition, the unidentified impurity will also most likely form the same derivative. Now, however, a TLC chromatographic separation will much more easily show up the impurity and allow sensitive variation of the elution system to separate this small minor band. Swish TLC can also probably be usefully applied to the now colored, highly crystalline derivative of probably decreased solubility, and the impurity itself recovered by breaking apart the derivative.

Liquid-Liquid Extraction using Hydrotropes as an Alternative to Fractional Crystallization for Purification at Scale

How does one purify a mixture of structurally similar neutral compounds that is about 80% one isomer and 20% the other? If you adopt fractional crystallization the most likely outcome is that you do purify the major compound but the recovery is about 60%. The lost material is in the mother liquors in an approximate 50:50 w/w ratio with the minor constituent.

You could try liquid-liquid partition, even trying several of these in series resembling a rough counter-current extraction. The problem is that there aren’t that many liquid phases that are mutually immiscible and more frequently at least one component of any pair that is immiscible will exhibit poor solubility for most of the multifunctional organic compound mixtures that you want to separate. Yes- water and hydrocarbons are immiscible but neither one dissolves most organics well. Yes- acetonitrile and hydrocarbons are immiscible, but most organic mixtures do not partition competitively between them. Yes, hydrocarbons and perfluorocarbons are immiscible but again distribution between them is usually overwhelmingly into one or the other. Then there are less well-known ones such as MIBK/sulfolane which might be promising, but these are few.

What is needed is a way to modify water so that it has an increased capacity to dissolve organic compounds of interest while still remaining substantially immiscible with those common organic solvents which also have a good ability to dissolve a target mixture. This is what hydrotropes can do.

Two important strengths of the methodology: (i) the solubilization capacity of the hydrotrope is a strong function, usually exponential, of the hydrotrope concentration and 
(ii) mere dilution of the hydrotrope with water is enough to recover dissolved materials.

Producing the most likely impurities of a Given Product for Use in Developing a Powerful Analytical Method

In order to develop a good purity analysis for an organic substance one needs to have some method to assess different methods. The better the method separates and quantifies more impurities from the product. A better method the more impurities it separates and the greater the degree of separation between the closest impurity and the product without losing some degree of separation for any single impurity. A better method separates distinctly even an impurity designed to have a very minor difference from the product. Is there a way to prepare a product sample with larger amounts of the most likely potential impurities?

There are two classes of impurities. Impurities that are product degradation derived; that is, they come from the desired product and arise from reactions upon the desired product after it has been isolated and purified. A different class of impurities are formed at the time of the synthesis of the desired product and which were not completely removed by the isolation and purification process performed before packaging of final product. These impurities are process characteristic. It is this second type that are considered here. These impurities are produced in greater or lesser amounts by variation in the continuous variables controlling the process.

A third group of impurities are created by changing the discontinuous variables of the process step, such as reagents, reagent purity, solvent, solvent purity, substrate purity, processing chemicals When the discontinuous variables are not altered this set of impurities do not appear and do not need to be further mentioned herein.

Temperature is the most significant continuous variable and it can produce the most substantial changes in chemical reactivity. If we are seeking a complete reaction it is likely that the transformation is self terminating and will essentially stop when the correct time has expired with very little occurring after this required reaction time. Increasing the temperature by 10 C according to a rule of thumb will double the rate of reaction. This will also allow competing reactions which are limited under the more preferred conditions to compete and produce by-products.

Thus an increase in the temperatures of each of the different stages of the reaction by 10 C and a decrease in the time by half in each stage should produce more impurities in the final product and these impurities should reflect realistic possible impurities.   If possible the extent of disappearance of starting material should be kept about the same.

Removal of Hydroxyl Impurities from a Solid Product at Scale by Chemical Reaction

KiloMentor has several times proposed the purification of a substance by selective reaction of its impurities to produce new impurities that can be separated by simple aqueous base extraction. One of the proposed methods for removing an alcohol impurity from a predominantly non-alcohol product is the reaction with succinic anhydride or phthalic anhydride and then water extraction of the carboxylic acid product impurity with dilute aqueous base. 

This is precisely the method patented for the purification of some samples of the drug substances citalopram and escitalopram in CA558198 ( WO2005/084643).

In these particular patented instances, the reason for needing to get these hydroxyl impurities reduced was that the size and the crystal polymorph being formed was dependent on their concentrations. The hydroxyl-containing impurity in citalopram or escitalopram was Z-4-(4-dimethylamino-1-(4-fluorophenyl)-but-1-enyl)-3-hydroxymethyl- benzonitrile. Reduction of this impurity by a factor of 10 was easily achieved heating with succinic anhydride or phthalic anhydride and then extracting. 

Example 1 

Scavenging of hydroxyl-containing impurity by succinic anhydride

 A mixture of R- and S-Citalopram (55.5 g) containing 0.6% of Z-4-(4-dimethylamino1-( 4-fluorophenyl)-but-l-enyl)-3-hydroxymethylbenzonitriIe is dissolved in dry toluene (145.0 g). Succinic anhydride (0.5 g) and aqueous ammonia (25% by weight) (3 ml) is added (pH = 10.5-11.0). The phases are separated and the toluene phase is washed with water (3x 120 ml). The toluene phase is evaporated and the yield is 53.0 g (95%). The product contains 0.06% of Z-4-(4-dimethylamino1-( 4-fluorophenyl)-but-1-enyl)-3-hydroxymethyl-benzonitrile.

But this is only indicated if one knows first that your product constituting non-nucleophilic material, does have hydroxyl-containing impurities. One potential means to test for free hydroxyls and indeed all nucleophilic species (NH and SH also) is to first, in an analytical amount, derivatize any nucleophilic functional-group-containing compounds to give colored materials that can be seen in a developed thin layer chromatogram as distinct from the unreactive main component. A colored derivatizing agent such as p-phenylazobenzoyl chloride or 4’-nitroazobenzene-4-carboxylic acid chloride

( Fieser & Fieser Reagents for Organic Synthesis Vol. 1), can be expected to produce colored spots on a TLC of the crude organic solution obtained by treating with such a reagent in an inert organic solvent and then washing with a dilute aqueous base to remove excess reagent. 

If such colored spots are present, a treatment with succinic anhydride or phthalic anhydride, or other hydroxyl scavenging agent can be useful for purification.

Thiazolidinedione as a Potential Reagent for eliminating Carbonyl Impurities from a Reaction Mixture: Purification by Reaction and Extraction

The development of the glitazones (pioglitazone, rosiglitazone etc.) as  drug substances has reduced the price of 2,4-thiazolidine-2,4-dione which is used in these syntheses. 

R=H

KiloMentor has always been attracted to the idea that some neutral carbonyl impurities could be quickly separated away from a complex reaction mixture using chemical agents that would react with the crude and give rise to products that could be extracted into another liquid phase.

Several such methods have already been popularized. For example, that aldehydes and ketones could be separated from materials without these functional groups by treatment of a mixture with semicarbazide adsorbed on silica gel was taught by Suk Dev. These derivatives could be subsequently hydrolyzed and recovered from a cyclohexane solution (so long as they were soluble in this solvent). Another KiloMentor blog promoted formation of oximes which offered two possibilities for solid separation: as a crystalline derivative itself or a hydrochloride salt of the oxime. Oximes have been cleanly decomposed to recover the ketone using many different reagents.

Thiazolidine-2,4-dione, it would seem, is a reagent that could be used to separate a carbonyl component, present as an impurity in a reaction mixture, where a non-carbonyl substance predominates and is the constituent of interest. This works because the condensation product between thiazolidinedione and a carbonyl along with any excess thiazolidinedione reactant will dissolve in a strong base because the thiazolidinedione has an acidic salt-forming imide functionality. It must be noted though, that the carbonyl compound itself cannot be regenerated. This is different from what was possible in the cases already alluded to.

 Thiazolidine-2,4-dione has melting point of 123-126 C. It can be purchased for less than $100/kg. The reaction can be done in a melt of this compound with sodium acetate as catalyst. Thiazolidine-2,4-dione is known to react easily with both aldehydes and ketones to give condensation products at its reactive methylene. With ketones, a mixture of two geometric isomers may be formed but both products upon treatment with aqueous alkali can be deprotonated to salts that have the potential to be extracted into an aqueous solution. Thus, these products show promise to be separable from neutral non-carbonyl materials by liquid-liquid extraction. Although only aryl aldehydes and aryl ketones seem to have been used in the literature, there is no good reason why aliphatic ketones would not also react cleanly particularly in the presence of an excess of the thiazolidinedione reagent. Aliphatic aldehydes might be too sensitive and could polymerize.

US 4703052

In a typical such reaction, the aldehyde or ketone starting material (IV) and thiazolidinedione (VI) are combined in approximately equimolar amounts with a molar excess, preferably a 2-4 fold molar excess, of anhydrous sodium acetate, and the mixture is heated at a temperature high enough to affect melting, at which temperature the reaction is substantially complete in from about 5 to 60 minutes. The desired olefin of formula (III) is then isolated, for example, by mixing with water and filtration, to obtain the crude product, which is purified, if desired, e.g. by crystallization or by standard chromatographic methods.

A mixture of carbonyls or a single carbonyl impurity in a non-carbonyl product might be separated by treating the mixture with thiazolidine-2,4-dione and sodium acetate and acetic anhydride with or without a solvent in order to condense the more reactive compound with the thiazolidine-2,4-dione. This condensed product will have an acidic hydrogen on the imide which can be converted to a sodium or potassium salt that can allow extraction into an aqueous solution. The residual, more hindered carbonyl or the non-carbonyl containing compound will remain unreacted and can be recovered by treatment with an organic solvent in which it readily dissolves. Any excess thiazolidine-2,4-dione present will also dissolve in the aqueous alkaline solution.

This methodology would need to be demonstrated with a mixture of ketones with different degrees of steric hindrance. A similar methodology has been used by partially forming enamines with the more reactive of two carbonyls and distilling the unreactive compound away from the enamine substance. The method proposed here does not require that the compounds be sufficiently low molecular weight to be volatile.

CA2423978

The invention also provides a process for preparing the potassium salt or a solvate thereof, characterized in that 5-[4-[2-(N-methylW(2- pyridyl)amino)ethoxy]benzyl]thiazolidine-2,4-dione (Compound (I)) or a salt thereof,  preferably dispersed or dissolved in a suitable solvent, is reacted with a source of potassium ion and thereafter, if required, a solvate of the resulting potassium salt is recovered.  A suitable reaction solvent is an alkanol, for example, propan-2-ol, or a hydrocarbon, such as toluene, a ketone, such as acetone, an ester, such as ethyl acetate, an ether such as tetrahydrofuran, a nitrile such as acetonitrile, or a halogenated hydrocarbon such as dichloromethane, or water; or a mixture thereof.  Conveniently, the source of potassium ion is potassium hydroxide. The potassium hydroxide is preferably added as a solid or in solution, for example in water or a lower alcohol such as methanol, ethanol, or propan-2-ol, or a mixture of solvents. An alternative source of potassium ion is a potassium alkoxide salt for example potassium tertiary butoxide.  

EXAMPLES

Example 1 

5-[4-[2-(N-Methyl-N-(2-pyridyl)amino)ethoxy]benzyl]thiazolidine-2,4- dione, potassium salt

A solution of potassium hydroxide (0.56 g) in water (5 ml) was added to a stirred solution of 5-[4-[2-(N-methyl-N-(2-pyridyl)amino)ethyl]thiazolidine-2,4-dione (3.0 g) in tetrahydrofuran (30 ml) at 50°C. The solution was cooled with stirring to 21°C over approximately 1 hour, before the solvent was evaporated under reduced pressure to afford 5-[4-[2-(N-methyl-N-(2-pyridyl)amino)ethoxy]benzyl]thiazolidine-2,4-dione, potassium salt (2.90 g) as a crystalline solid.

Example 2

 5-[4-[2-(N-Methyl-N-(2-pyridy1)amino)ethoxyl benzyl] thiazolidine-2,4- dione, potassium salt

A stirred suspension of 5-[4-[2-(N-methyl-N-(2-pyridyl)amino)ethoxy]benzyl]  thiazolidine-2,4-dione (3.0 g) in acetone (30 ml) was heated to reflux before a solution of potassium hydroxide (0.56 g) in water (5 ml) was added. After 5 minutes a clear solution was formed and the temperature of the stirred solution was lowered to 21°C over approximately 1 hour. The solvent was evaporated under reduced pressure to give the 5-  [4-[2-(N-methyl-N-(2-pyridyl)mino)ethoxy]benzyl]thiazolidine-2,4-dione, potassium salt (3.25 g) as a crystalline solid.

Example 3 

5-[4-[2-(N-Methyl-N-(2-pyridyl)amino)ethoxy]benzyI]thiazolidine-2,4- dione, potassium salt

A solution of potassium hydroxide (0.56 g) in water (1 ml) was added to a stirred suspension of 5-[4-[2-(N-methy1-N-(2-pyridy1)amino)ethoxy]benzy1]thiaz01idine-2,4- dione (3.0 g) in propan-2-ol (30 ml) at reflux. Within 5 minutes the solution became clear before a precipitate began to form. The stirred mixture was cooled to 21°C over approximately 90 minutes. The solid precipitate was collected by filtration, washed with propan-2-01 (1 0 ml) and dried under vacuum for 16 hours to afford 5-[4-[2-(N-methyl-N- (2-pyridyl)amino)ethoxy]benzyl]thiazolidine-2,4-dione, potassium salt (3.14 g) as a white crystalline solid.

Found (%): C: 54.44, H: 4.53, N: 10.45; Expect: C: 54.52, H: 4.83, N: 10.60.

The potassium ion level was determined as 9.9% by wt (expect: 9.9%) by ion

chromatography. Water content (Karl-Fisher): 0.2 % by wt.

Example 4

 5-[4-[2-(N-Methyl-N-(2-pyridy1)amino)ethoxyl benzyl] thiazolidine-2,4- dione, potassium salt

Potassium t-butoxide (1.41 g) was added to a stirred suspension of 5-[4-[2-(N-methyl-N-(2-pyridyl)amino)ethoxy]benzyl]thiazolidine-2,4-dione (3.0 g) in ethyl acetate (30 ml) at reflux. The stirred mixture was maintained at reflux for 15 minutes and then cooled to 21ºC over approximately 1 hour. The solid was collected by filtration, washed with ethyl acetate (10 ml) and dried under vacuum at 50°C for 72 hours to yield the 5-[4-[2-(Nniet1iy1- N-(2-pyridy1)amino)ethoxy]benzy1]thiazo1idine-2,4-dione7 potassium salt (3 30 g)  as a white crystalline solid. 

 Example 5 

5-[4-[2-(N-MethyI-N-(2-pyridyl)amino)ethoxy]benzyl]thiazolidine-2,4- dione, potassium salt

A solution of potassium hydroxide (4.71 g) in water (5.0 ml) was added to a stirred suspension of 5-[4-[2-(N-methyl-N-(2-pyridyl)amino)ethoxy]benzy1]thiazolidin-2,4- dione (25.0 g) in propan-2-ol1 (250 ml) at reflux. The stirred mixture was maintained at reflux for 15 minutes and then cooled to 21ºC over approximately 1 hour. The solid was collected by filtration, washed with propan-2-01 (50 ml) and dried under vacuum at 60°C for 16 hours to afford the 5-[4-[2-(N-methyl-N-(2-pyridyl)amino)ethoxy]benzyl]  thiazolidine-2,4-dione, potassium salt (26.6 g) as a white crystalline solid.

Pertinent reference

J. Org. Chem., 1956, 21 (11), pp 1269–1271

A possible common use

Friedel-Craft acylations usually proceed with o,p-orientation to an electron-donating aromatic substituent. It seems possible that the para-substituted product could preferentially react with an insufficient quantity of 2,4-thiazolidinedione  at its less sterically hindered carbonyl or if the reaction is reversible the para-substituted compound could produce the more thermodynamically stable product compared to the ortho-substituted compound and thus produce mixtures of predominantly unreacted ortho compound and thiazolidinedione adducts of the para compound. Upon dilution with water, the thiazolidinedione product may crystallize leaving unreacted ortho ketone or upon aqueous alkali/organic solvent partition the thiazolidinedione adduct would go to the aqueous phase and the ortho-substituted product to the organic layer.