The Increased Scale-Up Risk with Catalyzed Reactions

The probability of failure is increased for catalyzed reactions of which, for example, enantioselective reactions are a prominent contemporary class. The special additional risk is that catalytic system may be more easily shut down by small, even trace, impurities that are difficult to measure much less control. Put another way, a catalyzed reaction is susceptible to poisoning and this can lead to slowing or complete interruption in conversion with no easily identifiable c ause. Catalyzed reactions are inherently less rugged than the uncatalyzed because the catalytic substance, by definition, is used in lower than stoichiometric quantity and so would be disproportionately affected by a particular quantity of a catalyst poison. Impurities in the inputs to a catalyzed process can also accelerate reaction. When, as after a switch to a different source of an input, they are not added performance may deteriorate or fail completely. Neil G. Anderson wrote in Practical Process Research & Development, First Edition, pg. 194: “the importance of trace beneficial impurities may become evident only by failure of the reaction when using different lots of starting materials, reagents, or solvents.” Thus the recommendation to perform laboratory experiments with the same materials to be used in the plant goes double for catalyzed reactions and this includes chemicals used to wash and prep the reactor.

A catalyzed reaction can more easily be shut down without leaving forensic evidence. If it is a catastrophic failure it can poison our minds as much as our reactions. We may start to harbor conspiracy delusions. “Have we been harmed by some disgruntled or mentally disoriented employee? Have some operators made an error and covered it up? Are we now engaged in a long, expensive, and ultimately fruitless failure investigation?” It may seem far fetched, but I was embroiled in such a situation once. Human minds, in the absence of a clear causal connection for a phenomena, are programmed to find signs suggesting hypotheses even in random data.

When a procedure that has been running successfully at large scale  suddenly fails and if laboratory experiments with the identical raw materials run immediately afterwards succeed, these ideas come to mind and make the resulting further inquiry even more difficult to bear.

A suggestion that may be just too inconvenient and divisive to implement should at least be contemplated. When a clear most probable cause cannot be detected after a failure the next run performed at that scale, to be fair, should use a completely different group of operators or  should be run with special laboratory oversight. If the team is al completely different a second failure will at least rule out malevolent intervention by a team member. What must be avoided is  the situation where a second failure would throw what is likely unwarranted suspicion upon employees who would have participated in both failing runs.

The Relationship between the Risk of Catastrophic Failure and the Size of the Scale Up Steps in Chemical Process Development

What do I Mean by Catastrophic Failure?

In the context used herein, I am defining a catastrophic failure of a process step trial as a very large loss of product quality and/or isolated yield from which there is no recovery. That is, by definition, there is no patch known and reprocessing is not viable. Characteristically, the failure, when it occurs, comes as a complete surprise. Catastrophic failures at scale usually create serious financial losses and make project schedule extension necessary. It is the risk we face when we ‘put too many eggs in one basket’.

How is the Size of the Scale Up linked to the Risk of Catastrophic  Failure?

What is risked when a process step is increased in scale? It is fair l widely accepted that at first and quite normally, for any reaction step the yield is likely to fall somewhat. More serious, but still not unexpected, is that the type and quantity of impurities in the isolated product may change in unanticipated ways. Worse still and getting to the catastrophic, the reaction may create a mixture that cannot be purified enough to give an isolable physical form. Still worse, the reactor contents may become unprocessible (can’t cut, can’t stir, can’t cool, can’t filter can’t distil). When these latter things, for which there has been no preparation occur, unacceptable time and money is lost. More material must be ordered. The project milestone are missed. These possibilities limit the size of the scale up steps in development. Consequently, as the cost of the inputs at risk and/or the probability of catastrophic failure fall, the size of the steps in scale up can increase.

The approximately optimal conditions determined with laboratory equipment can still be quite different with respect to a  number of variables from what must be done in a pilot plant. Just for starters, some parameters such as heating, cooling, stirring and the times for reagent additions most often cannot be physically matched after increasing scale because of  equipment limitations. Surprises can occur as one increases the  size of operations and these lead to product with unacceptable properties.

How does One Rank Risks?

Any risk to workers’ physical safety must be made inconsequential. It would be immoral to knowingly add to risks to health and safety. Even from a completely selfish perspective, a lost time industrial accident can put a chemist manager’s professional career at risk. Safety issues are paramount and signs of a hazard dictate slow scaling.

A loss of starting material is both a loss of time and of money. The budget can perhaps be repaired but the time required for the delivery and qualification of  fresh starting materials is lost forever. If the inputs are inexpensive as a proportion of total costs and are quickly available from multiple sources, one risk of more aggressive scaling is reduced. It is usually the early steps in a process where inputs can be replaced cheaply and quickly and other things being approximately equal, early steps can be scaled up in larger increments for that reason.

Can One Estimate the Likelihood of a Particular Type of Scale-Up Failure?

Perhaps we ought to ask instead: How well am I able to  scale down the pilot plant environment and reproduce it in my laboratory equipment? Scaling down is the exercise of selecting the bench-scale equipment, operating conditions, and mathematical models to successfully simulate pilot or production scale operations in the lab.

Risk can be reduced by testing with such equipment. If the experimentation has been conducted using exactly the same quality for solvents, reagents, processing aids and catalysts, the biggest sources of deviation in scale-up are removed. If the processing times including times of addition, times for transfers, and time for filtrations approximate those necessitated in the pilot plant, risk is reduced. If the corrosiveness and abrasiveness of the reactants have been tested on the reactor’s materials of construction, it reduces risk. If the procedure is insensitive to rate over a wide range of agitation speeds then another sensitivity has been allowed for. If the sensitivity to traces of air and moisture is known and taken into consideration, life is simplified. If none of the reactants reagents or coproducts in the process step are more completely swept out of the reactor at one scale compared to the other, another frequent source of deviation is accounted for.

There are auguries of danger that can be divined while still in the laboratory and addressed before moving to higher scale:

·        Addition or removal of a gas

·        High viscosity of the reaction medium

·        High exothermicity

·        Need for a low reaction temperature

·        Drown out quenching

·        Rapid addition rates

·        Fast reaction relative to the rate of addition of a reacting component

·        Decomposition on the reactor walls

·        Presence of byproduct polymer

·        Use of polymer reagents that may disintegrate with stirring

·        High speed stirring

·        Asymmetric synthesis catalysis

When one scales up, it is advantageous if the first step is of sufficient size that all the changes in the main discontinuous variables (reactor materials, reactor shape, minimal stirrable volume, type of agitation, heat transfer etc.) are introduced together. Making these changes together often can be better accommodated by also including initially an increase in the amount of solvent in the reactor, to give an overall dilution. Often the biggest risk impediment to moving into the pilot plant is the cost of materials to operate at the minimum stirrable volume in the larger reactor. Making an initial dilution that can later we reversed, may set up a more acceptable combination of risks at a more acceptable price.

Said another way, it may be better to delay the optimization of the throughput, which is very often the result of the consequence of increasing the concentration of the reactants and reducing the amount of diluents (ie solvent) until after  the transition to the pilot plant or manufacturing equipment. This will result in a less expensive transition from laboratory to pilot plant. It will require less of the expensive chemical to reach the minimum stirrable volume at the start of the reaction.

Catalyst poisoning will be treated in a separate article.

Balancing Chemical Equations and Calculating Heats of Reaction: Two Often Overlooked Helps for Chemical Process Developers.

Many years ago, when I was a graduate student with Professor R.B. Woodward, at a group party at Harvard University, my wife asked Woodward, who was arguably the finest synthetic organic chemist in the world, what kept him modest. Although the question was expressed with a fair measure of pique, Dr. Woodward did not seem flustered.  He replied something to the effect that Nature itself did this, since even the most accomplished chemist more often proposes what turns out to be unsuccessful than what is successful.

How true. How true. But even so in our false sophistication we often fail to take simple precautions, which would easily avoid silly errors. I would like to speak of two of these:

·        failing to balance the chemical equation, and

·        failing  to calculate the enthalpy change in a reaction using simple bond energies

Balancing the Chemical Equation

Inorganic chemists rarely show this delinquency but organic chemists, because they are so accustomed to writing a starting material, a reagent and a product as part of a proposed series of reactions, almost never balance equations.  The result is that more often than you would expect even experienced lab workers do not get the stoichiometry correct and add either an excess or a deficiency of a reagent.  This is particularly true of oxidation reactions.  The second difficulty that results is that they cannot see the importance of the coproduct, which is formed along with their desired product, because the coproduct only ecomes import when one tries to balance and so they cannot see the possibilities and the complications that may arise from its presence. 

Calculating the Enthalpy of the Reaction

We organic chemists often seem to have gotten it into our heads that so long as we can draw a self-consistent series of arrows, showing the movement of electron pairs, then a reaction has a reasonable possibility to proceed. Usually we are protected from error by the fact that the transformation we are contemplating is completely analogous to a known reaction.  Nevertheless, it is a simple matter using bond energy tables to calculate the net enthalpy change of the reaction we are hoping will occur.  The result is that it will become more apparent to us whether a desired reaction is just weakly favored (so that steric hindrance, inadequate solvation etc. can inhibit it), disfavored or so strongly favored that we need to be concerned about the exothermicitry of the process and take appropriate precautions.

To be sure, it is negative free energy not a negative enthalpy, which is necessary to have a favorable equilibrium. but it is the less common situation when the entropy of the reaction makes the difference in driving the reaction and when it does, this is almost always when gases are involved or when the reaction is a fragmentation.

To make simpler the calculation of Enthalpy of Reaction, I have gathered together typical bond energies for the covalency between different atoms listed them below.  These should be treated as median or average values.  You may be able by inspection of the substrate or the reagent you intend to employ to recognize bonds, which can be expected to be stronger or weaker than these representative values.

Bond	Value	Bond	Value	Bond	Value
F-C	108	O-H	110-111	N≡C	212.6
F-F	37	O-C	85-91	N≡N	225.8
F-H	135	O=C	173-181	N-Si	76.5
F-O	45	O-O	35	P-P	41
F-C	108-116	O-N	53	P-C	65
F-Si	193	O=N	145	P-H	76
Cl-Cl	58	O-Cl	52	Si-Si	81
Cl-Br		O-Br	48	Si-C	77
Cl-I		O-P	91	Si-H	94
Cl-S		O=P	119-130	C-H	98.7
Cl-N	46	O-Si	111	C-H (vinyl)	108
Cl-P		O-I	56	C-H(acetylene)	128
Cl-C	81	O=S	132	C-C	82.6
Cl-Si	113	S-H	88	C=C	145.8
Cl-H	103.2	S-S	54	C≡C	199.6
Br-Br	46.1	S-C	60
Br-I		S-Si	70
Br-P		S=C	128
Br-S		N-N	39
Br-N		N=N	100
Br-Si	97	N-H	93.4
Br-C	66	N-C	72.8
I-I	36.1	N=C	147

What Might be the Best Cleaning Solvent for Cleaning the Reactor Walls of a Plant Reactor

acetamide model

The walls of a large-scale reactor can sometimes be difficult to clean.  The problem is compounded because they are not easily accessible and cannot be inspected closely. Methods that can be applied in the laboratory for many reasons are off-limits. Scrubbing is impractical, dangerous, and potentially damaging to the equipment. What is needed is a powerful but innocuous solvent that can work by vapor condensation not just below the surface of the refluxing liquid cleaner but above the surface and on the reactor walls where the reaction mixture may have splashed, caked, and baked.

In the very old literature, a common, inexpensive, and innocuous compound was claimed to be the best solvent known and one that would dissolve both organic and inorganic materials; salts as well as uncharged covalent molecules: molten boiling acetamide.  Acetamide can be synthesized in situ in the reactor by heating ammonium carbonate and acetic acid and distilling out water. This in fact is the first preparation in the First Collective Volume of Organic Synthesis. 
Acetamide has bp (760 mm) 222.0 C ; bp (100 mm) 158 C; bp (40 mm) 136 C;  bp (20 mm) 120 C; bp (10 mm) 105 C; or bp (5 mm) 92 C. 
As a white solid, it has mp 82.3 C. 
The solubility is 2 grams per ml of water. 
Acetamide has been advocated as a “green” solvent [http://acs.confex.com/acs/green07/techprogram/S3384.HTM]
The ninth edition of the Merck Index describes it as:
 “Solvent; molten acetamide is an excellent solvent for many organic and inorganic compounds. Solubilizer; renders sparingly soluble substances more soluble in water by mere addition or by fusion.”  Way back in 1933, Professor O.F. Stafford of the University of Oregon wrote that acetamide dissolved more different chemicals than any other known solvent. [J. Am. Chem. Soc., 1933, 55 (10), pp 3987–3988].

Process chemists sometimes forget that their responsibility is for the minimized overall cost of the process and this is much more than the chemicals only cost. The throughput per unit of time is a major factor in the overall cost. That time includes the equipment cleaning time required between batch runs and between the end of one campaign and a new one for another product that is to be run with the same equipment. It makes little sense to invest extensive research efforts in reducing processing time when the same throughput efficiencies can be more easily achieved by reducing cleaning time between runs.

Many plants use a standard cleaning protocol implemented as an SOP.  Special cleaning procedures are resorted to only when it fails to remove all the contamination. In some cases, the standard cleaning or rinsing will even exaggerate a problem. For example, in the synthesis of adamantane described in Organic  Synthesis Coll. Vol. III pg. 16-19, specific instructions are provided to avoid treating the vessel with water until acetone is used first to completely remove the tar.

Development chemists are the first to get an indication that special cleaning problems could arise after certain processing. Giving the plant scale-up people a heads-up and some suggestions will improve both teamwork and overall efficiency.

Since acetamide is Ames negative it is a moot point as to whether it should be considered as a “genotoxic impurity”; it is, however, a modest-potency carcinogen with a TD50 value of 180 mg/kg/day.

Another Way to Separate Phenolics by Crystallizing of Co-crystals?

When Kilomentor comes upon some very specific information that might have general utility for separations of a function group class, he saves it in his personal files, until an appropriate process chemistry situation arises.  The trick is (i) to have saved the information and (ii) to read these notes over sufficiently so that when the possible application comes up, some internal mental alert will sound to remind himself that he has some information that might be useful. Then, it is easy enough to retrieve it, examine it in more depth and see whether it really could be part of a rugged, time-saving, and even perhaps an elegant solution.

In this blog, I would like to examine the content of the patent US5081263 which on its face teaches an improved means to purify meta or para-substituted hydroxylphenyl or hydroxylnaphthyl carboxylic acids.

The inventive trick is that the authors have discovered that aryl carboxylic acids bearing a phenolic group, not in an ortho position to the carboxyl group, can be advantageously crystallized from p-dioxane because co-crystals are formed using this particular solvent.

The inventors explain that “the particular feature of the said adducts is that hydrogen bridge bonds exist between the hydroxyl groups of the aromatic compounds and the oxygen atoms of the dioxane, so that the adducts are 2:1 adducts…..and the carboxyl groups of two hydroxycarboxylic acid molecules are, in turn, dimerized, so that relatively long chain-like arrangements can form.”

In other words, (and this is my interpretation), the carboxylic acid functionality has a strong preference in this medium to exist as acid dimers leaving the phenol hydroxyls un-associated, and in p-dioxane they strongly prefer making two hydrogen bonds between the two phenols and the two ether oxygens of a single dioxane molecule.  This leads to high molecular weight co-crystals.

The patent provides information to suggest that the molecules that might do this can have other non-interfering functional groups and they propose fluorine, chlorine, bromine or a nitro group as potentially not interfering. Interestingly, this nitro can be ortho to the phenol and the dioxane co-crystal will still form. A specific example is a crystallization of 4-hydroxy-3-nitrobenzoic acid.  Other teachings in the patent indicate that the crystallization of the cocrystals can be from mixtures of dioxane and water or dioxane and ethanol, so it would seem that hydroxyalkyl is also a  non-interfering. 

Useful as all this might be for separating hydroxylaryl carboxylic acids, it would seem that the usefulness might be broader and more significant. Carboxylic acids are not typically difficult to purify. In many other articles, KiloMentor has argued that in fact carboxylic acids are preferred intermediates in synthetic process design precisely because if a mixture is produced during synthesis, they can be separated by simple acid-base extraction from all non-acids and a mixture of acids can be separated by pH-controlled extraction, or extractive crystallization or by reversible formation of a myriad of salt derivatives.

The gift the patent may be providing is the possibility that phenolic, diphenolic, or even polyphenolic compounds may form co-crystals with p-dioxane, and simple phenols may form simple 2:1 adducts with dioxane. Now the separation of diphenols, phenols, and non-phenols is a more challenging goal than the separation of a group of carboxylic acids. Yes, phenols are weakly acidic and some of the strategies for separating acidic compounds, in general, do work but it is not as rugged a methodology and interfering reactivity from the more alkaline conditions (such as oxidation) can raise several ugly problems. 

High molecular weight phenols, called pseudophenols because their alkali salts are not extracted into water, may be extracted from hydrocarbon solvents with Claisen's alkali. 

Phenols also can form O-sulfate water-soluble salts that are easily extracted and crystallized.

At the same time it is quite true that this idea may not work out in any particular situation, but the key pedagogical point is that if you have collected the concept and have sufficient familiarity to recall it in the appropriate situation, you get one more simple isolation possibility to evaluate. Choosing from more potential and distinctly different approaches increase your chances for simple, rugged, elegant solutions.

Salts of Orotic Acid: Neutralizing excess Base without increasing the Inorganic Salt Concentration

Orotic acid is a pharmaceutically acceptable salt former that is safe to use as a counter ion in drug salts.

Intuitively it seems correct to conjecture that to obtain a high melting salt it might be sensible to derivatize with the highest melting salt-former available. Although such a prediction would have no ineluctable basis, it is the way to bet. Such salt would share some substructure with the acid-derivatizing agent from which it was formed. If these features contribute to the high melting property of the acid, those features, if preserved in the salt’s structure, might be anticipated to contribute to some retained high melting characteristic.

Among pharmaceutically acceptable acids for making salts, the highest melting is orotic acid at around 325 C. Its structure contains multiple hydrogen bond donors and acceptors and some of these hydrogen-bonding sites may be retained in the pharmaceutical salt formed.

 Despite this possible desirable property orotic acid salts are rarely prepared. This may be because orotic acid itself is poorly soluble even in water. Its alkali metal salts of orotic acid are also poorly soluble.  For example, sodium orotate can be precipitated by adding a solution of  N, N-dimethyl-2-hydroxyethylammonium orotate in 80% aqueous ethanol to sodium hydroxide in water. 

 N,N-dimethyl-2-hydroxyethyl ammonium orotate combined with a sodium halide would give a precipitate of sodium N, N-dimethyl-2-hydroxyethylorotate and N,N-dimethyl-2-hydroxyethyl ammonium halide. This might be a method of removing an alkali cation from an aqueous solution of an organic compound. The sodium cation would be precipitated as an alkaline salt by orotic acid.  N,N-dimethyl-2-hydroxyethyl ammonium halide would be left in the aqueous solution but this could be removed by extraction with a neutral organic, water-immiscible solvent.

Orotic acid or a salt of orotic acid might be expected to form a complex with triphenylphosphine oxide which is a good hydrogen bond acceptor. Orotic acid has an imide NH which is a good hydrogen bond donor.

Orotic acid may be a good substance to neutralize aqueous base since both the alkaline salt and the free acid are essentially insoluble and will precipitate.

The unusual solubility properties of orotic acid and its metal salts make it worth bearing in mind when trying to isolate organic bases, particularly since, as a pharmaceutically acceptable salt, trace residues are not as critical.

The salts of orotic acid with volatile amines such as ammonia, methylamine, ethylamine, trimethylamine, and triethylamine are all expected to be substantially more soluble than those of hard metallic elements so that these hard cations are likely to be precipitated. Heating the ammonium salts of large nonvolatile acids will likely drive the volatile amine off leaving the free acid.