Crystallization is the most frequently used method for isolating organic solids. Recrystallization is the most frequently used means to purify them. But this phase switching from a solute in solution to an ordered solid is one of the most unpredictable methods both for isolation and purification in the sense that it is impossible to predict melting point, solubility, or lattice energy much less the comparative values of these for the desired product versus t the most troublesome by-products which one is trying to remove from the reaction mixture from which it was synthesized. This unpredictability is exactly what led KiloMentor to focus on other ways to switch phases to achieve purification in a chemical process chain.
This is not to say that once a certain molecular architecture has been achieved one cannot prove that the target compound is a solid. I am saying first, that no reliable prediction can be made of how to get it separated sufficiently pure that it is not an oily mixture, and second, no reliable prediction is possible of what recrystallization conditions will be needed to purify it from its most predominant and intransigent impurities and, third, what yield of pure product can be anticipated.
The following generalizations have some logical basis:
The lower the melting point of an intermediate, the more difficult it will be to crystallize.
There is a positive correlation between melting point and ease of crystallization.
The presence of impurities in a reaction mixture leads to a melting point depression and contamination of isolated samples of the desired product.
The extent of the melting point depression depends upon the proportion of those impurities.
The free energy of crystallization is roughly proportional to the enthalpy of crystallization and this is roughly proportional to the melting point.
The fewer the chemical conformations (the fewer the rotatable bonds) or the more symmetrical the compound, the more crystallizable the compound will be.
I do not know whether this is latter claim is proven but it seems likely and agrees with some simple observations. Compounds with long hydrophobic chains as part of the structure tend to be lower melting than cyclic substances. Compounds with high symmetry seem to be higher melting than unsymmetrical compounds of the same molecular weight and functional group type (ie t-butanol vs 1-butanol).
This makes sense in physical terms. There would be more entropic resistance to the crystallization of a molecule that can adopt multiple conformations than to one that because of its cyclic form or symmetry can adopt fewer. If only one conformation can be accepted into a crystal lattice, the compound with fewer conformations has a statistically better chance of being added to the lattice. Thus crystallization might be expected to be faster.
Crystallization is accelerated if one has seed crystals and it is for this reason that so much effort is expended to get the first crystalline material. If an intermediate has never been prepared before one should anticipate the possibility that a considerable effort may be needed to obtain the first solid. The methods adopted to obtain the first batch of material sufficiently pure to crystallize do not need to be scaleable. Making these seeds is a valid use of chromatography in process development. Column chromatography can quickly and dependably deliver a high-purity material, which should have a significantly enhanced tendency to crystallize. If a chromatographically purified solid material does not readily crystallize one can properly be pessimistic.
Other classical methods which can lead to that initial crystallization are:
- scratching the flask containing the impure material with a glass rod while cooling the oil
- steam distillation to remove traces of solvent from the oil
- ultrasonic treatment in a sonification bath
- cooling to a low temperature to form a glass followed by slow warming
- trituration with a pure hydrocarbon fraction
- overnight cooling in a sealed vessel (to exclude moisture) in a deep freeze
- formation of a solid derivative, crystallization, followed by regeneration of the compound itself
It is apparent why this should happen if the initial crystallization is being inhibited by small amounts of a particular impurity.
The difficulty with the initial crystallization of a new substance is that two different physical processes must occur in each other's presence. First, initial crystal nuclei must be generated and this usually requires a low temperature AND then, second, these nuclei must become bigger. This latter requires crystal growth. The optimum temperature for crystal growth is consistently more elevated than the best temperature for nucleation. It is for this reason it is thought that raising and lowering the temperature or establishing a temperature gradient within the oil or solution can enhance this first crystallization. Seeds formed in one colder region of the oil or solution migrate into the warmer crystal-growth region.
Once some crystals have been created, even when one performs a recrystallization there are always, it is hypothesized, trace amounts of the crystals remaining which as seeds provide a ‘memory’ when the bulk of the material is taken to conditions optimum for crystal growth. This explains a remarkable phenomenon. One form of crystal may reproducibly form from a substance for years but then by chance, a more stable form crystallizes and thereafter it is impossible to obtain any of the first form because there are always seeds around to catalyze the formation of this later-discovered but more stable form.
Coloured compounds which according to their structure should not be coloured are contaminated by small amounts of polyunsaturated impurities. Charcoaling can remove the colour and at the same time often assist that initial crystallization.
Crystallization is an art. There are the virtuosos and then there are the rest of us.
Low-Temperature Crystallization
In the laboratory, crystallizing and then reducing the temperature of the crystal slurry to below zero and filtering below zero has an excellent chance of failing. The reason is that at the laboratory scale working with one's hands, it is difficult to maintain an inert atmosphere over the cold liquid and over the filtered crystals. This causes moisture to condense into the crystallizing mixture and onto the solid on the filter funnel and this can lead to outright failure, oily crystals, or a solid that melts away on the filter or dissolves in the wash liquid.
At scale however these problems are reduced to nothing. It is simple to retain an inert atmosphere which excludes moisture, it is simple to cool to sub-zero temperatures and hold that temperature and it is much easier to hold the temperature of the filtered solid low and keep down the temperature of the wash liquid. Using a temperature differential between the boiling point of the solvent on the upper side and –20 ℃ on the lower side can provide higher recoveries of crystalline product. Also, the greater temperature gap is less demanding on the solvent properties and allows more inexpensive solvents to work adequately.
Using less common solvents on scale should be more a case of selecting a solvent for improved purity rather than choosing a solvent for enhanced recovery.
Checking for the purification ability of a solvent operating on a particular reaction mixture
If the problem of obtaining crystals of the product from the mixture is resolved, it is useful to explore the ability of the solvent to distinguish between the desired intermediate and other impurities in the reaction mixture that one would expect The following experiment might be very revealing and very simple to do but frankly, I have never done it myself and I know of no one who has ever done it. Take a crude solid contaminated with reaction contaminants and divide the solid into two equal portions. Recrystallize the first portion from the solvent you have identified. Filter the solvent but do not allow any wash liquid to mix with the mother liquors but keep the mother liquors from this first crystallization separate. Dry the solid. Now use the mother liquors to recrystallize the second portion of the solid and isolate it in the exact same way. Now compare the recovery and purity of the two portions. The recovery of the second portion would be expected to be higher than the first. The second recrystallization presumably is done from solvent saturated with the desired product. The purity of the first should be greater than for the second. There will be nearly twice the level of impurities in the second crystallization versus the first; however, the extent of these differences will be quite dependent upon the particular substance and its particular impurity levels. Thus, one experiment might provide a very good idea of to what extent recycling solvent could improve recovery for a process step.
A crystallization in two parts followed by a trituration or slurrying of both parts combined in an anti-solvent might be expected to give a better yield of a homogenous product in a single batch than some of the poor recovery recrystallizations that we often employ.