Selecting a Reaction Solvent

Some statements in the chemical literature suggest that process chemists should be able to optimize most reactions to greater than 90% yield.  Whether this is true or false depends of course upon the meaning of the terms ‘most reactions’, ‘yield’ and ‘optimize’. 

Most reactions

For the statement to be true, I think its authors must be thinking of name reactions, reactions discussed for example in the series, Organic Reactions, or reactions taught in basic organic textbooks. That is, standard reactions that are taught with the expectations that their analogs can be depended upon to work in prophetic chemical schemes.  To go to the other extreme, the kind of chemical reactions that we encounter as mechanistic puzzles on grad school cumulative exams or reactions proposed to explain an unusual by-product; these, by their very designation as puzzles make the point that they operate because of special circumstances and are unlikely to be successful when applied more generally. 

Yield:

Going still further in restricting the claim, I think the term ‘yield’ in this aphorism also needs some refining.  The Kilomentor Blog, with its emphasis on organic synthesis chemical process development takes pains to distinguish between assay yield (assay value for product as a % of theoretical in the reaction mixture just before isolation begins), isolation yield (the weight of product meeting the specification as a percentage of the theoretical weight one could have obtained with a perfect isolation); and overall percentage yield, which is the weight of product versus the theoretic weight of product as a percent.  Perhaps assay yield is what the rule of thumb refers to.  Isolations from reaction mixtures tend to average about 80% and for the overall reaction yield to be 90% the product of assay fraction multiplied by the isolation fraction must be greater than 0.9. That turns out to require something like an assay yield of 94.8% and an isolation yield of about 94.8%.

Optimization:

Although the word optimize means to find the best combination of all conditions, when seeking an optimum is contemplated, every simple protocol requires that all the parameters being optimized be continuous ones (like temperature, pressure, time, molar ratio etc.) and not discontinuous ones ( like reagent or solvent choice). Thus statistical optimization is really not optimization that properly explores all the potential modes. In any case, optimizing is a theoretical action that assumes an infinite amount of time to perform experiments. As we know or can imagine, chemists and chemical engineers working in the real world have deadlines and the diminishing returns of our optimization efforts soon fall below the opportunity cost for our work on the next problem.

Experienced commentators say the most important discontinuous, discrete variable the manipulation of which can improve a  reaction [ besides the exact substrate for the reaction (which can only be peripherally modified for example by protecting groups without making the substrate irrelevant to the process being contemplated) and the reagent for the reaction (which is constrained to a known few)] is the solvent, for which there are many choices.  

 Important Parameters for the Solvent

Substrate Solubility:

Even though complete solubility of all the reaction ingredients is not necessary to obtain complete reaction and increased concentration, even to the point of using a heterogeneous mixture, is good for throughput; homogeneous reactions have fewer problems on scale up.  Mass transfer is more dependable with a single solution phase and homogeneous reactions are less dependent on thorough mechanical mixing. Because heterogeneous reactions require more mixing power and because mixing power does not scale up proportionally with volume, accepting a heterogeneous reaction makes it more likely that the scale up will be more complex.  

Nevertheless, the actual solubility of reactants may not be apparent to chemists upon mixing solids and the proposed volume of solvent.  A considerable time may be required to reach the equilibrium solubility because dissolution may be kinetically controlled.  Heating the solid component to reflux in the solvent and cooling may result in a solution or at least a finely divided suspension that quickly dissolves when the reaction is started.  Once a small amount of reaction occurs, the first traces of product may rapidly dissolve the remaining starting material even without what would seem to be the full requirement for solvent.  Also, the solubility of a solid reactant in the solvent alone may not be representative of the solubility in the reaction mixture.  

A mixture of solvents can provide a much higher solubility than a single solvent.  Even if the use of a mixed solvent means that the solvent cannot be recovered and recycled, the increased throughput from using a solvent mixture that increases the reactor capacity can be beneficial for the overall cost.  There is too much emphasis upon using a single solvent in organic process chemistry.  Solvent is actually rarely recycled.  It is too costly for a fine chemical plant and there are too many problems associated with meeting the COA for the recycled solvent.

Heat of Vaporization:

The heat of vaporization quantifies how adequately a solvent ‘buffers’ reaction exothermicity? A useful question to ask is, “For the exothermicity of the planned reaction how many times would the heat produced by a hypothetical instantaneous reaction vaporize all the solvent of the reaction?”  This compares the enthalpy of the reaction with the heat of vaporization of the solvent and the dilution together.  It suggests how efficiently the condenser would have to work, at the solvent's boiling point, if the reaction occurred, in the worst possible case, instantaneously.

Heat Capacity:

The heat capacity of a solvent controls how rapidly the temperature of the solution could rise when a particular reaction takes place and releases the energy of reaction. 

Ease of a Solvent Change:

How easy will it be to make a solvent change in order to use a different solvent for the isolation or to prepare for a subsequent reaction when there is no product isolation? Lower boiling solvents are easier to switch away from because they can be displaced by higher boiling solvents.  Kilomentor has suggested in other articles other options for switching solvents.

Drying Simplicity: 

Drying solvents on scale with inorganic salts followed by filtration of the inorganic salt hydrates uses labour, equipment, and time inefficiently.  It is greatly disfavoured for work at scale. The preferred method for solvent drying selects a solvent that forms an azeotrope with water and distils a portion of solvent as the azeotrope.

Freezing Point:

At what temperature does the solvent that is being considered solidify or become highly viscous? The freezing point can limit the range of temperatures that can be used in the optimization.  Lowering the temperature is often the best option for increasing the selectivity of the desired reaction over other competing reactions that produce by-products. If low temperatures create viscous reaction mixtures, these can result in hot-spots during reagent additions, inadequate mixing leading to incorrect stoichiometry, creating in turn by-products, and poor crystallization control.

A Change of Solvent in a Process Step will more often than not dramatically change the Optimal Values of all the Other Variables

There is actually very little information about reactions that have been really thoroughly explored.  An exception is the Willgerodt reaction. The optimal reaction conditions for acetophenone have been investigated using statistical methods in thirteen different solvents  varying 4 parameters. [ R. Carlson, Acta Chem. Scand. 40, 694 (1986)].  The optimal conditions are dramatically different depending upon the solvent used in the reaction.

in the case of the Willgerodt reaction, the optimal moles of sulfur per mole of acetophenone varied between 2 and 17. The optimal moles of morpholine reagent per mole of acetophenone varied between 6 and 13.7. The optimal reaction temperature varied between 70 and 145 C and the optimal reaction time varied between 2 and 22 hours. In two of the solvents major side products were observed. In one solvent no product was obtained and in another a different substance was the major product.

In contrast, when the substrates were differently substituted acetophenones, PROVIDED THAT THE SOLVENT WAS KEPT CONSTANT, there was little variation from the optimum of the major parameters:

• the yields were in the range 88-95%
• the sulfur/acetophenone molar ratio at optimum varied from 7-10
• the morpholine/acetophenone ratio varied from 8.4-10.6 and 
• the optimal temperature varied between 116-133 C

Although the breadth of evidence is narrow since few reactions have been sufficiently broadly investigated, this points to a hypothesis that solvent is the most significant parameter in reaction optimization.  Because it is a discrete and not a continuous parameter, it is necessary to decide before optimization is started what solvent or pair of solvents (one could optimize with a binary mixture of varying proportions) to systematically test using your optimization algorithm.  

When a first candidate solvent has been identified, either by finding a literature example in that solvent or by trial experiments, then other candidate solvents can be found by comparing principle properties and choosing other solvent candidates by their proximity in principle property space to the solvent that is the original candidate. 

Actually it is impossible to make a general statement based on data about reactions optimized in different solvents because such data is so rare.

In order to optimize a reaction the most important thing to know is whether the reaction has ever been performed in a solvent that will dissolve the substrate you need to use as starting material. For example, if you have a polypeptide as starting material and the reaction you are planning has only been conducted up to now in heptane, this is not promising. Your peptide is most likely insoluble in heptane and when you use an alternative solvent all the optimization parameters will be different. Put another way, preferentially you want to start optimizing a reaction by running the reaction on your substrate in the solvent that was used in a successful literature example.  In order to start an optimization, one must be able to detect and quantify the desired product in a first reaction mixture.

A surprising corollary is that the most important information to have in advance of an optimization study is the range of solvents that have been used in the prior art for the proposed reaction.

Investigational vs Empirical Approach to Optimizing Reactions

One might have a thorough and correct theoretical basis for understanding what favours or disfavours a particular chemical reaction or, conversely, one might know nothing at all about the reaction.  Normally the state of knowledge is somewhere in the middle. Sometimes applying the generalizations offered by the literature quickly results in step improvements; sometimes this common knowledge fails entirely and leads away from the better conditions. 

Screening of Suitable Solvents in Organic Synthesis. Strategies for Solvent Selection

Rolf Carlson, Torbjorn Lundstedt and Christer Albano. Acta Chim. Scandanavica B 39(1985) 79-91.