<P>In 1956, after completing my graduate training and postdoctoral work in physical
organic chemistry, I entered the employment of the Hercules Powder Co. (now Hercules,
Inc.) at their research center in Wilmington, Delaware. I joined a group of chemists
working on the development of a commercial process for producing polyethylene and
polypropylene employing the recently discovered Ziegler-Natta catalysts. This work
required me to learn the literature on transition-metal chemistry and to do laboratory
work with air-sensitive and unstable organometallic compounds. While I accomplished
little of scientific value in the two years that I was in this program, the experience
prepared me for my next assignment. The next project I was given by my supervisor,
Dr. David Breslow, was a rather unusual one. He said, ”Do something with transition
metals." I had the opportunity to choose a project that I thought would benefit the
company.</P><P>At this time, 1958, organotransition-metal chemistry was just beginning.
Only a few cyclopentadienyl, alkyl, and carbonyl complexes of the metals were known
and very little chemistry had been done with these compounds. There were several important
commercial processes being used at that time, however, which employed transition metals
as catalysts. All of these had been discovered by empirical methods and nothing was
known about the chemistry involved. I decided to begin my new project by trying to
determine the mechanism of one of these important reactions in the hope that this
knowledge would suggest new chemistry to study.</P><P>The commercial process I chose
to study was the hydroformylation reaction, which was also known as the Oxo reaction
in Germany where it had been discovered by Otto Roelen in 1938.
[
1 ]
The reaction involved a cobalt-catalyzed combination of carbon monoxide and hydrogen
with an alkene at elevated temperature (˜120 °C) and pressures (˜800-1000 psi) to
form aldehydes. For example, 1-hexene in the reaction produced a mixture of about
70% heptanal and 30% 2-methylhexanal in high yields. The catalyst for the reaction
was generally added as cobalt oxide, carbonate, acetate or similar compounds. It
was known that the cobalt compounds were converted into the well-known dicobalt octacarbonyl
under the reaction conditions. Studies carried out by Professor Walter Hieber of the
Technische Hochschule in Munich at about this time also suggested that the very unstable
(at room temperature and atmospheric pressure) cobalt hydrocarbonyl, HCo(CO)
4 , would be present in the catalytic reaction as well.
4
</P><P>A brief report appeared in the literature in 1953 by a research group under
Irving Wender
[
2 ]
at the U.S. Government Bureau of Mines Labs in Pittsburgh, which suggested that cobalt
hydrocarbonyl was the actual catalyst in the hydroformylation reaction. Since the
decomposition of HCo(CO)
4 was probably second order in cobalt, reactions of the compound could be studied at
25 °C in dilute solution if they were relatively fast. The Wender group added 1-hexene
dropwise to an excess of a solution of HCo(CO)
4 at room temperature and found the same 70:30 mixture of heptanal and 2-methylhexanal
as was obtained catalytically. They did not offer a reasonable explanation for their
results. During my previous work with transition-metal-catalyzed alkene polymerization,
there was the idea by some that transition-metal hydrides might be involved as initiators
of polymer formation by first undergoing addition to the alkenes to form alkyl transition
metal complexes presumably bonded also to aluminum chloride or other metal halides
employed as catalyst components. This step would then be followed by insertion of
multiple alkene units to form long-chain metal alkyls, which ultimately eliminate
polyalkene and reformed the metal hydride catalyst. There was no experimental evidence
to support this idea at that time but it was known that some non-transition hydrides,
such as those of boron and aluminum, would add to alkenes.</P><P>This proposal led
me to believe that the hydroformylation probably also began with a hydride addition
of HCo(CO)
4 to the alkene presumably in both possible directions to explain the formation of
two isomeric aldehydes in the reaction. The next step was not obvious at the time.
As it happened, I had noticed a paper by a group working at the Ethyl Corporation
on some hydrocarbon-soluble manganese carbonyl derivatives, probably made to test
as anti-knock agents. The group, Coffield, Koziekowski and Closson,
[
3 ]
for reasons not given, placed methylmanganese pentacarbonyl in solution in an autoclave,
pressured it up with carbon monoxide, shook it for a while and isolated the product.
The product turned out to be acetylmanganesepentacarbonyl (Equation 1).</P>
Equation 1
<P>To my knowledge, this was the first recognized carbon monoxide insertion reaction.
Of course, its generality was not known or even suggested at this time. It appeared
to me, however, as the probable next step in the hydroformylation reaction, presuming
cobalt would react as manganese did. If this were the case, then all that would be
needed to complete a reasonable mechanism for the hydroformylation reaction would
be a reduction of the acyl-cobalt bond by either hydrogen or the other potential reductant
known to be present, HCo(CO)
4 . Thus, the following mechanism was formulated with two possible final steps (shown
with only one of the two isomers expected, Scheme 1).</P>
Scheme 1
<P>My next move was to show that the proposed steps would actually occur as shown
with isolated, pure, cobalt carbonyl derivatives at low temperatures where they were
stable enough to be isolated and identified.</P><P>Fortunately, at about this time,
Walter Hieber had just reported the preparation of the very unstable methyl cobalttetracarbonyl.
[
4 ]
The compound was prepared by the reaction of sodium cobalttetracarbonyl with methyl
iodide. The yellow-orange compound was not only very readily oxidized by air, but
in the pure state, it decomposed rapidly well below 0 °C. Hieber established the formula
of the compound but did no further chemistry with it or any other cobalt alkyls.</P><P>I
prepared methyl cobalttetracarbonyl as Hieber reported and found that it reacted rapidly
with carbon monoxide even at 0 °C and one atmosphere to form acetyl cobalttetracarbonyl
as predicted by analogy with the manganese reaction. Another important discovery was
that the commonly employed ligand, triphenylphosphine, would react easily with alkyl
and acyl cobalttetracarbonyls to give much more stable (though still oxidizable in
air) and less soluble monophosphine derivatives. This allowed me to isolate and characterize
organocobalt complexes more easily. The alkyl cobalttetracarbonyls turned out to react
with triphenylphosphine to form acyl cobalttricarbonyl-triphenylphosphine complexes
as products and the acyl cobalttricarbonyltetracarbonyls gave the same products with
loss of carbon monoxide (Scheme 2). The kinetics of the PPh
3 reaction with acyl cobalttetracarbonyls were subsequently studied and found to be
first order in the cobalt compound but zero order in PPh
3 , indicating that the loss of a coordinated carbonyl group from the acyl cobalttetracarbonyl
is the rate-determining step. Thus, a 16-electron acyl cobalttricarbonyl species is
an intermediate in the PPh
3 reaction and probably in the CO reaction as well.</P>
Scheme 2
<P>When a solution of HCo(CO)
4 was allowed to react at 0 °C with a large excess of 1-pentene for 15 minutes and
an excess of triphenylphosphine was added, evaporation of the solution left a yellow
solid whose spectrum indicated that it was a mixture of alkyl cobalttricarbonyltriphenylphosphines.
Cleavage of the products with iodine in methanol gave a solution containing roughly
equal amounts of methyl hexanoate and methyl 2-methylpentanoate, demonstrating that
HCo(CO)
4 adds rapidly to alkenes. This was the first experimental proof of this reaction.
The rapid reduction of acyl cobalttetracarbonyls with cobalt hydrocarbonyl and with
hydrogen at above 500 psi at room temperature was also found to occur, providing evidence
for the proposed mechanism (Scheme 2).</P><P>The next task was to use this information
to find profitable applications of this new chemistry. My approach was to see what
else, besides hydrogen and HCo(CO)
4 , the acyl and alkyl cobaltcarbonyls would react with.</P><P>Ethylene reacted rapidly
with acetyl cobalttetracarbonyl at 0 °C and one atmosphere but a non-stoichiometric
amount was absorbed, as I recall about 2.5 moles of ethylene were taken up per mole
of acetyl cobalttetracarbonyl. The IR spectrum of the reaction mixture showed several
new carbonyl absorptions. I did not pursue this apparently complex reaction further,
since, in the meantime, I had discovered, while exploring the chemistry of the alkyl
cobaltcarbonyls, that 1,3-butadiene reacted selectively with the acetyl cobalttetracarbonyl
to give a single product, π-1-acetylmethylallyl cobalttricarbonyl.
[
5 ]
</P><P>This reaction turned out to be very general and nearly all conjugated dienes
and alkyl or acyl cobalttetracarbonyls would undergo the reaction. It was then found
that the organic ligand could be eliminated by reacting the complexes with a tertiary
amine under carbon monoxide at room temperature. Furthermore, the reaction became
catalytic in cobalttetracarbonyl anion with a stoichiometric amount of a tertiary
amine. Dicyclohexylethylamine was used to avoid any possible coordination of the amine
with cobalt intermediates. Thus, acyl chlorides or alkyl halides and CO reacted with
conjugated dienes in the presence of a few mol% of the cobalt carbonyl anion and a
stoichiometric quantity of the tertiary amine to form dienones in good yields (Scheme
3).
[
6 ]
</P>
Scheme 3
<P>Similarly, it was found that acyl cobalttetracarbonyls would add to alkynes at
25 °C or less to form 1,2-adducts which immediately cyclized by way of a 1-acyl-2-tricarbonylcobaltacyl
intermediate, ultimately forming a π-allylic cobalt tricarbonyl complex of a γ-lactone
(Scheme 4).
[
7 ]
</P>
Scheme 4
<P>In an analogous reaction using methyl bromoacetate and CO, 3-hexyne underwent reaction
with NaCo(CO)
4 and dicyclohexylethylamine, but the expected γ-lactone-cobalt intermediate underwent
base-catalyzed elimination of HCo(CO)
4 to form a γ-dienolactone as a major product (Scheme 5).
[
6 ]
</P>
Scheme 5
<P>A similar addition reaction was found with α,β-unsaturated carbonyl compounds to
give 1-acyloxy-π-allylic cobalt complexes and CO (Equation 2).
[
8 ]
</P>
Equation 2
<P>In spite of the discovery of the above chemistry, I had no immediate ideas on how
to employ it profitably for Hercules. Therefore, my research then took a new direction.</P><P>In
the laboratory across the hall from me at Hercules worked another chemist, named Pat
Henry. He had been studying the mechanism of the commercially important ”Wacker Process",
the palladium chloride catalyzed oxidation of ethylene with oxygen (air) to acetaldehyde.
[
9 ]
</P><P>One step of his mechanism interested me, particularly. In this step, Pat proposed
that an intermediate β-hydroxyethylpalladium chloride species decomposed by the elimination
of dichloropalladium hydride anion (Cl
2 PdH
- ) forming acetaldehyde. There was the widespread belief at this time that transition-metal-carbon
bonds were quite weak and easily decomposed by homolytic cleavage. This belief arose
because attempts to prepare compounds such as dimethylpalladium, dimethylnickel, methylcobalt
or trimethyliron all failed. Henry’s proposal of the decomposition occurring by a
β-hydrogen elimination raised the doubt in my mind that maybe the metal-carbon bond
wasn’t so weak but that decomposition may occur preferably by other mechanisms. I
decided to try an experiment to see what would happen if an organopalladium compound
without β-hydrogens was prepared in the presence of something else - perhaps it could
react with something other than another group on the metal. The first experiment I
tried was to add phenylmercuric acetate to a stirred solution of lithium tetrachloropalladate
in acetonitrile at 0 °C under an atmosphere of ethylene. An immediate reaction occurred.
Palladium metal precipitated and ethylene gas was rapidly absorbed (about one mole
per mole of palladium). Analysis of the reaction mixture showed that about an 80%
yield of styrene and 10% of
trans -stilbene had been formed (Scheme 6)!</P>
Scheme 6
<P>This was a very exciting result! A new method of carbon-carbon bond formation under
very mild conditions had been found. Further study showed the reaction to be very
general and tolerant of all important functional groups.
[
10 ]
Unfortunately, there were obvious problems that would keep the reaction from ever
becoming generally useful. While a wide variety of arylmercurials with nearly all
functional groups could be made by direct electrophilic mercuration and used successfully
stoichiometrically, this reaction was not only expensive, but the use of highly toxic
mercurials was a severe deterrent. We found that aryltin compounds could also be used,
but most substituted derivatives were not readily available. The need for a stoichiometric
amount of a palladium salt was also prohibitive. A solution to the problem was suggested
by a literature report.
[
11 ]
</P><P>This report, and subsequent ones, showed that air-stable arylpalladium-bistriphenylphosphine
halides could be prepared by warming tetrakistriphenylphosphinepalladium(0) with iodobenzene,
bromobenzenes or even chlorobenzenes (Equation 3).</P>
Equation 3
<P>The next question was, would finely divided palladium metal do the same and give
the reactant needed for reaction with alkenes? While separately prepared palladium
powder did work with iodobenzene, it was found more convenient to simply add a palladium(II)
salt, which was reduced by the excess alkene in the reaction mixture. The reaction
went poorly, however, until it was found that addition of a base such as triethylamine
in a stoichiometric amount gave a high yield of the expected product, catalytically
in palladium.
[
12 ]
This catalytic form of the reaction is shown in Scheme 7.</P>
Scheme 7
<P>This reaction was discovered independently in Japan at about the same time by Mizoroki;
[
13 ]
however, he did not appear to appreciate the significance of it since he did not report
any further work on it. We looked at several examples of the reaction but soon found,
as Mizoroki apparently had as well, that the reaction was mainly limited to aryl
iodides. Since it was known, as noted in Equation 3, that tetrakistriphenylphosphinepalladium
would react with bromobenzene to give the phenylpalladiumbistriphenylphosphine bromide,
a solution to getting bromobenzene to react well could be to simply add triphenylphosphine
to the reaction mixture. Indeed, this worked such that many aryl bromides reacted,
catalytically, with a variety of mono- and disubstituted alkenes. However, it was
soon found that aryl bromides with strongly electron-donating substituents such as
p -bromophenol reacted very poorly.</P><P>After considerable study, we found that the
problem was due to aryl bromides having strongly electron-donating substituents undergoing
a palladium-catalyzed quaternerization reaction with the triphenylphosphine ligands.
This ultimately removed the triphenylphosphine ligands from the metal and palladium
metal alone did not react well with the aryl bromide under these conditions.</P><P>We
soon found a solution to this problem. We needed only to substitute tri-
o -tolylphosphine (or other
o -substituted triarylphosphines) for the triphenylphosphine to prevent the migration
of the aryl group on palladium to an unhindered arylphosphine ligand and stopping
the reaction. We were thus able to get a wide variety of aryl and vinyl iodides and
bromides to react with any mono- or di- and even some trisubstituted alkenes normally.
Aryl chlorides, however, generally reacted poorly or not at all. Still, the reaction
has proved to be very convenient and useful for preparing a wide variety of aryl alkenes
and dienes.
[
14 ]
Recently, Fu
[
15 ]
found it was possible to get arylchlorides to react well by the use of the more basic
phosphines, tri-
tert -butylphosphine as a ligand.</P><P>The reaction occurs to add the vinyl or aryl group
from the bromide or iodide exclusively, or at least mainly, to the less substituted
(hindered) carbon of the double bond. The only known exception occurs with unsubstituted
vinyl alkyl ethers or amides where the reverse direction of addition is favored.</P><P>The
catalyzed addition of aryl bromides to
cis - and
trans -alkenes is stereospecific. The major products formed are explainable on the basis
of a mechanism involving a
syn -addition of the aryl palladium bromide followed by a
syn -elimination of the hydrido-palladium bromide.
[
16 ]
</P><P>Vinyl halide additions to mono-, di- or trienes provide convenient routes to
conjugated polyenes. Cyclic products are formed if a vinyl bromide group can form
a five- or six-membered ring by reacting with an appropriately placed double bond
in the same molecule. In cases where conjugated diene-palladium hydrido-halide π-complexes
are formed as intermediate species, π-allylic palladium products are formed. These
are stable up to at least 100 °C and must be decomposed by using a nucleophilic secondary
amine rather than triethylamine as the base, in which case allylic amine products
are formed catalytically (Scheme 8).
[
14 ]
</P>
Scheme 8
<P>Terminal alkynes, also, were found to react with aryl and vinyl halides using a
palladium triarylphosphine catalyst to form aryl or vinyl alkynes (Scheme 9).
[
17 ]
</P>
Scheme 9
<P>The same reaction was found by Luigi Cassar at about the same time.
[
18 ]
Subsequently, Sonogashira fount that copper salts catalyzed some of these reactions.
[
19 ]
I then turned my attention to another palladium-catalyzed reaction. Many examples
of the dimerization of 1,3-butadiene to a mixture of isomeric octatrienes by numerous
palladium complexes were known, but I was more interested in reports by two independent
groups that the dimerization reaction led to a mixture of 1,6- and 1,7-octadienes
by employing a π-allylic palladium acetate dimer catalyst combined with a triarylphosphine,
formic acid and triethylamine.
[
20 ]
[
21 ]
Two things interested me about this reaction. Could isoprene be dimerized selectively
to a natural product structure with use of the proper phosphine co-catalyst, and
what else could be reduced besides the central double bond in octatriene with this
catalyst system?</P><P>We found that, indeed, isoprene could be dimerized catalytically
in about 80% yield to a mixture of 3,7-dimethyl-1,6- and -1,7-octadienes, monoterpene-type
structures employing tri-
o -tolylphosphine in the catalyst (Equation 4).
[
22 ]
</P>
Equation 4
<P>Unfortunately, even an 80% yield does not compete with Mother Nature because of
the problem of separating 20% of other dimethyloctadiene isomers, also formed.</P><P>We
then turned our attention to the use of the catalyst system as a hydrogenation catalyst.
We soon found that nearly any palladium compound or even palladium on carbon would
catalyze reductions with formic acid and triethylamine. In fact, we were able to catalytically
reduce just about anything with HCO
2 H/Et
3 N that could be reduced with low-pressure hydrogen (<˜2 atm) and the same catalyst.
Thus,
o -bromonitrobenzene was reduced with one molar equivalent of 98%, commercially available,
formic acid in triethylamine solution with a palladium on carbon catalyst to nitrobenzene
selectively, while with three equivalents of formic acid and a platinum on carbon
catalyst the nitro group was reduced selectively (Scheme 10).
[
23 ]
</P>
Scheme 10
<P>Similarly, α,β-double bonds to carbonyl groups were reduced selectively in the
presence of other unsaturation.</P><P>More useful was the finding that triple bonds
could be easily reduced to
cis -double bonds in high yield, employing normal, commercially available palladium on
carbon as catalyst (Equation 5).
[
24 ]
</P>
Equation 5
<P>Apparently, the triethylammonium formate is stable in the presence of the catalyst
unless there is a reducible material present, since quantitative reductions occur
with all of the formic acid added.</P><P>Carbonyl groups directly bound to aromatic
rings were also reduced by HCO
2
- Et
3 NH
+ analogously to the reduction with hydrogen gas.</P><P>The formic acid-amine reducing
system has two major advantages over hydrogen gas reduction. Most importantly, it
is safer because with air, flammable and explosive hydrogen is not involved. No free
hydrogen (or carbon monoxide) is produced in the formic acid reductions. Only carbon
dioxide is evolved. Secondly, quantitative, partial reductions are very easy to carry
out simply by adding the calculated amount of 98% liquid formic acid to the amine,
the catalyst, and the substance to be reduced. The reaction is complete when CO
2 evolution ceases. This procedure avoids the necessity of removing the hydrogen gas
as soon as the partial reduction is complete to prevent over-reduction. Of course,
there is no need to control temperature carefully either (usually) to correct for
the vapor pressure of the solvent.</P><P>This completes a brief summary of the most
important results I obtained during more than 32 years of research. Recent reviews
on palladium-catalyzed reactions are available.
[
25a ]
[
b ]
</P>
