A Proposal for Separating a Mixture of Water and a High Boiling Dipolar Aprotic Solvent from a Drown-Out Precipitation.

Frequently in organic synthesis, a product synthesized in dimethylformamide (DMF), dimethylsulfoxide (DMSO) or another high boiling dipolar aprotic solvent is precipitated by adding a substantial excess of water. At present, this drowned-out aqueous solvent mixture cannot be economically recovered but needs to be sent for destruction.  This is wasteful and environmentally questionable. It is also more expensive than burning a completely combustible, purely organic waste. A drown out also usually raises the point of maximum volume thereby reducing throughput.

These difficulties separating away dipolar parotid liquids are not only because there is a strong affinity between these higher boiling organics and the water but also, because after a drown-out there is just such a high proportion of water. 

 Pyridine Treatment

Let us do a thought experiment. Imagine what would happen if pyridine, which forms a binary, minimum-boiling azeotrope with the water in the mixture, is added. Pyridine makes a stronger hydrogen bonding possibility available to the water.  The result may be that pyridine will form an azeotrope with water, which can be removed as the low bp. azeotrope at 92.6 C.  If this were to be the result the still-pot residue would be DMF. A further advantage of this method of purification would be that the potential pyridine impurity left in the residual DMF  is aprotic and is therefore unlikely to interfere in reactions requiring anhydrous conditions. That is, the recovered DMF is aprotic and can probably be reused.

But wait a minute, we now have a large amount of a homogeneous solution of pyridine and water; the pyridine/water azeotrope, which is 57% pyridine and 43% water. Are we better off or have we just changed one problem for another while spending more time and more money?

Pyridine Recycling

The difference is that this mixture can be separated into two liquid phases by the addition of sodium hydroxide.  The separated pyridine layer can be further dried with solid sodium hydroxide or can be used in a wet condition to purify more DMF. The alkaline water can be used in the plant to neutralize an aqueous acid fraction from any other process in preparation for sending it to the sewer.

Imagine a mixture containing 90% water and 10% DMF. Add to this a convenient portion of pyridine and distill the water/pyridine azeotrope away from the DMF containing mixture. In the distillate vessel, the azeotrope comes into contact with a reservoir of solid sodium hydroxide or perhaps just liquid caustic. The distillate separates into a strongly basic aqueous lower layer and an upper pyridine layer. The pyridine layer is led back into the still pot to remove more water/pyridine as an azeotrope. When all the water is transferred out the head temperature rises and the pyridine is distilled leaving purer DMF separated from all the water! If there is any pyridine residue in the DMF it does not contain any available hydrogens and may be suitable for reuse as is. The pyridine fraction is drummed off for reuse and the aqueous alkali kept to neutralize another plant waste.

The Use of Silver Nitrate Complexing to Separate Olefin Containing Compounds

In Organic Synthesis Coll. Vol. III a mixture of cis and trans cyclooctene is separated by mixing somewhat more than two molar equivalents of an aqueous silver nitrate solution with a pentane solution of the cis and trans compounds. The cis compound does not form any adduct and so remains in the pentane solvent. With vigorous stirring the trans compound forms a complex and dissolves in the aqueous phase. Although silver is an expensive reagent, , at least in principle, it is recoverable, so it can be considered for use at scale. Winstein and Lucas studied the complexes formed between silver nitrate and unsaturated hydrocarbons and found that in some cases these are solids useful for isolation and purification.[J. Am Chem. Soc. 60, 836 (1938)]. Complexes that are solids can be recrystallized, often from hot alcohol. It is not clear whether functional groups besides double bonds interfere with separation in this way, although only hydrocarbons have been described in the literature. It is not apparent why a number other functional groups would be incompatible with the method. It may just be that, when other functionalities are present, there are better known options for separations.

The Diels-Alders adduct between norbornadiene and cyclopentadiene contains two double bonds and forms a 2:3 hydrocarbon/silver nitrate adduct [Am. Soc. 81, 4273 (1959)]. The Diels-Alder adduct between norbornene and cyclopentadiene contains only one olefin group and was purified using its 1:1 adduct [Am. Soc. 86, 2188 (1964).].From work with the mixtures of 1,3; 1,4; and 1,5- cyclooctadiene, it has been shown that silver nitrate forms complexes with each of these, but they have different stabilities, and can be separated by using different temperatures. The weak complexes are only isolable at low temperatures [J. Chem. Soc. 312 (1954)]. The natural triene humulene was purified as a silver nitrate complex containing 2 molar equivalents of silver nitrate [Australian J. Chem. 14,272 (1961)][Tet. Let. 1977 (1965)].

It is interesting to speculate whether an aqueous solution of silver nitrate could be used to treat a mixture of olefin and the dihalocarbene adduct to remove the olefin. The dihalocarbene adduct would be expected to be reactive with silver nitrate if they were both in a homogenous solution but if the dihalocarbene adduct was in a saturated hydrocarbon solvent and the silver nitrate was in water, they might not come into sufficient contact to react. It might also be a problem for complex formation if the olefin that one sought to complex was itself not sufficiently soluble in water to allow reaction.

Use in Column Chromatography

Olefin containing compounds are separated on reverse phase columns when silver nitrate is dissolved in the mobile phase. It would be interesting to see whether the reverse phase HPLC mobility of unsaturated compounds in an aqueous silver nitrate eluate might give an indication of their complexing ability.
 

Liquid-liquid Extraction

Silver nitrate in methanol improves the separation of saturated fatty acids from unsaturated acids that can be held in solution better when silver nitrate is added. This suggests the possibility of liquid-liquid extraction between pentane or hexane and aqueous silver nitrate.

Direct Isolation as a Method in Process Development

In  2004, Neil G. Anderson wrote an article,[ Organic Process Research & Development, 2004, 8, 260-265] where he assessed the benefits of what he called direct isolation processes. Anderson is the author of the finest book about organic process development, Practical Process Research & Development.  In the ‘direct isolation’ article he described three choices for work-up and isolation: direct isolation from the reaction mixture, extraction followed by isolation, and telescoping. 

Kilomentor opines that such categorization is too simple, while providing no useful organizing insight to balance the inexactness introduced by the simplification. Rather, what is needed in the 21^st century to advance the process development art is a more widely held view that quench, purification, and isolation each have many options and these combinations of choices need to be tailored to the particular instance as carefully as the chemical reaction conditions to which the overall work-up applies. 

Anderson quotes with approval P. Zurer [Chem. Eng. News  [2000, 78(1), 26] “In a typical chemical operation, 60-80% of both capital expenditures and operating costs go to separations.” Dr. Anderson and I would agree that in process development, where optimization is cost driven, more attention needs to be directed to fishing the desired reaction product cleanly out of the chemical soup; that is, from the reactor contents. For too long chemists have been disadvantaged by training that taught them to “work up the reaction in the usual way.”

Anderson’s simplification makes it more difficult to consider separations conducted after a number of treatments of the reaction mixture, such as, to identify just a few examples, by separation by solvent exchange, by-product precipitation, product derivatization, clathrate complex formation, scavenger resins, or capture and release resins.

Unisolable Reactive Intermediate Compounds

When we organic chemists talk about reactive intermediates, often we mean transition states like carbonium ions, radicals, or carbanions that are undetectable by normal analytical methods and that exist at such low levels that they can be treated kinetically as being at some constant but very low concentration throughout a chemical transformation. In addition to these; however, there are authentic substances existing at analytically detectable concentrations that are simply too unstable to be isolated in good yield under convenient conditions. They are also called reactive intermediates. They are intermediates in a reaction sequence rather than a single reaction. The greater numbers of these situations arise when compounds are not stable at convenient isolation temperatures or when compounds are too reactive to be concentrated down to a solid or neat liquid state.  An answer for the problem of handling most of such substances is to ‘telescope’ the first reaction into the second. No attempt is made either to store that first unstable compound or to remove the diluting reaction milieu that surrounds it.

Flow Systems

Some such reactive intermediates are too unstable even for these methods to work. Sometimes the conditions required to obtain a practical rate of formation of the intermediate are still too vigorous to allow it to accumulate without further degradation. For such materials some sort of flow system is required to control how long reagents and starting materials are in contact before the accumulating intermediate is carried into an environment where the subsequent transformation can take place. Such flow reactors are a popular expedient today; however, such methods are not new and do not have to be expensive or high tech.

Flow reactors can be advantageous when the reaction mixture passes through a very viscous intermediate stage where much stronger stirring is required {Om P Goel, Continuous reactor model for the use of butyl lithium in the pilot plant, 1974 ???] ; this can be frequently caused by the low temperature required for the stability of the intermediate.

Sometimes the intermediate is too unstable to remain in the reactor during the time required for the mixing of the reagents on a large scale [J.A. Foulkes and J. Hutton, Synthetic Communications, 9(7), 625-630 (1979).]

Free-radical Bromination Scale-up

Let us, as a simple example, suggest useful modifications for the scale-up of the bromination of a heterocycle which has been previously performed in the laboratory with N-bromosuccinimide in carbon tetrachloride.  Since this is, superficially at least, a straightforward problem it can highlight the principles without overpowering us with the details.  Kilomentor will make the assumption from the reagent/solvent combination that this is a benzylic bromination of an alkyl substituent that is contemplated.

N-bromosuccinimide is not the most economical source of bromine atoms. 1,3-dibromo-5,5-dimethylhydantoin has a higher weight percentage of transferable bromine than other reagents(56%). This compares with 45% for NBS and 50% for liquid bromine wherein the latter only one half the bromine atoms get incorporated into the product. The cost of the hydantoin derivative per kilogram is also lower than for NBS . The Aldrich per kilogram prices are $109.60 and $135.00 respectively.

Unless there is a strong reason to the contrary, an optimization should use the least expensive reagent available at the expected scale.

To make the most realistic interpretations from the experimentation and to be able to make the soundest hypotheses to explain whatever we find, we should know whatever we can about the probable mechanism of the transformation. The benzylic bromination with NBS in carbon tetrachloride catalyzed by dibenzoylperoxide is a free radical catalyzed reaction. The reactions in the process are not all free radical however. The brominating agent is bromine radicals from molecular bromine not from the NBS itself.  NBS is just a reagent for converting each molecule of the co-product hydrogen bromide into one new molecular bromine and one succinimide. What is really happening is the free radical reaction of a low concentration of bromine with the aralkyl substrate.  

The development reactions should be done with shielding from laboratory lighting. When one scales up into a closed tank, there will no light entering the reaction vessel and so we do not want any photochemical reactions to complicate our study.

Dibenzoyl peroxide is most often the free radical initiator used in the literature; however it is not necessarily the best. Dibenzoyl peroxide decomposes into two types of radicals benzoic acid radicals and phenyl radicals produced from by decarboxylation from the former.

Azo initiators can also be used and have some advantages. According to William A. Pryor, Free Radicals. McGraw-Hill Inc. 1966:

“The most common azo initiator is azobis isobutyronitrile (AIBN). This useful initiator was first prepared in 1949 and has been widely used since. It has a half-life of 17 hours at 60 C and 1.3 hours at 80 C. This azo compound decomposes at the same temperature at almost the same rate in benzene,toluene, xylene, acetic acid, aniline, nitrobenzene, dodecylmercaptan, and isobutyl alcohol. This contrasts very markedly with benzoyl peroxide, for example, which has a rate of decomposition that is very solvent dependent. Furthermore, the rate of decomposition of AIBN in solvents such as toluene is essentially unaffected by inhibitors such as chloranil, iodine, or diphenylpicrylhydrazyl (DPPH). This implies that radicals do not attack the azo compound to produce an induced decomposition; if they did, the rate of disappeaance of the azo compound would be lower in the presence of substrates that would inhibit the radical chain decomposition. Induced decomposition is negligible for most azo compounds in solution; this sometimes makes them the preferred initiator for studies of radical reactions.” 

Furthermore, AIBN has worked where benzoyl peroxide has performed poorly. For example, see Fieser and Fieser, Vol. 1 pg. 45; H. Stockmann, J. Org., 29, 245 (1964).

We need a radical which will initiate a chain by removing a benzylic hydrogen. A benzylic radical is a nucleophilic radical so the faster reaction will be with an electrophilic initiator radical.  This is taught by Brian P. Roberts in his papers. A good review of this theory is in Polarity-reversal catalysis of hydrogen-atom abstraction reactions: Concepts and applications in organic chemistry, Chem. Soc. Rev.. 1999, 28, 25-35. The isobutyronitrile radical can be regarded as electrophilic since the alpha cyano carbanion would be a stabilized anion. This theory would predict that AIBN would be a better initiator than benzoyl peroxide. Advantageously, AIBN will be less sensitive to whatever changes in solvent we wish to implement.

Phenyl radicals from dibenzoyl peroxide could be regarded as electrophilic radicals since an sp² orbital can sustain a negative charge better than an ordinary sp³ orbital. Nevertheless, its electrophilicity cannot match an alpha cyano carbanion. The latter would be expected to react more quickly to generate the nucleophilic benzylic radical we need to achieve overall bromination.  If we still want to use dibenzoylperoxide as reagent a catalytic amount of trimethylamine thexylborane or triethylamine borane could act as catalyst by polarity reversal.