300
J. N. Harvey
the relative height of the two barriers was not consistent with observed behavior.
Why did we miss this point? A first explanation is simply that beyond a certain
mechanistic complexity, kinetic analysis based only on computed free energies or
indeed potential energies becomes difficult to perform and errors are made. This
system was not highly complex, so other people such as the referee could readily
detect the problem. Indeed, having learned from this case and others, we would
probably no longer make this mistake for a system of this relative simplicity. Still,
it is worthy of note that errors of this type can occur and are expected to become
more frequent as complexity increases. A second explanation is that our focus in the
second study [19] was enantioselectivity, which we simply assumed must be due to
the hydrogenation TSs, and accordingly paid too little attention to other points.
A second issue highlighted by this mistake is that chemical reaction mechanisms
are complicated, in ways that are sometimes hard to capture in computational studies.
For ketone hydrogenation by diamine dihydride ruthenium complexes, our first study
[10] had focused on understanding reactivity and kinetics in a non-chiral variant
of the reaction, carried out in a nonpolar solvent. Additionally, in that study, the
amidoamine intermediate shown on the top right in Fig. 5 (and bottom right in Fig. 3)
had been isolated and shown to be catalytically competent. The experimental rate
law was compatible with hydrogen splitting being turnover-limiting (although the
experimental activation enthalpy was somewhat smaller than the calculated one). All
of this suggests that the mechanism suggested in that paper, and shown in Figs. 3 and
4, is actually broadly correct under the reaction conditions used there (broadly correct
in that H 2 -splitting is indeed turnover-limiting, though the too-high barrier may mean
that the detailed mechanism for this step needs to be refined). In contrast, with protic
solvents, the variant (3) shown in Fig. 5—which is likely to be very unfavorable
under non-protic, nonpolar conditions—may dominate. The free energy landscape
underlying Fig. 5 is relatively flat, and its shape is sensitive to the environment.
Hence, mechanistic conclusions based on one set of reaction conditions may not
apply under all conditions.
3.3 Mechanism Discovery: Cis–Trans Isomerization
of Alkenes
In 2011, we published a study of a deceptively simple reaction: reversible cis–trans
isomerization of an alkene [21]; see Fig. 6. This isomerization was discovered by
our experimental collaborators in the group of Prof. Guy Lloyd-Jones, using the
simple Pd(II) catalyst PdCl 2 in solvents such as acetonitrile. The experimental studies
showed that this reaction could partly be caused by small amounts of palladium
hydride impurities formed under the reaction conditions, but conditions were found
where this side route could be inhibited, yet isomerization continued to be observed.
The question was, therefore, how does Pd(II) cause isomerization? Clearly, the alkene
could coordinate to Pd, which would weaken the double-bond character of the alkene
J. N. Harvey
the relative height of the two barriers was not consistent with observed behavior.
Why did we miss this point? A first explanation is simply that beyond a certain
mechanistic complexity, kinetic analysis based only on computed free energies or
indeed potential energies becomes difficult to perform and errors are made. This
system was not highly complex, so other people such as the referee could readily
detect the problem. Indeed, having learned from this case and others, we would
probably no longer make this mistake for a system of this relative simplicity. Still,
it is worthy of note that errors of this type can occur and are expected to become
more frequent as complexity increases. A second explanation is that our focus in the
second study [19] was enantioselectivity, which we simply assumed must be due to
the hydrogenation TSs, and accordingly paid too little attention to other points.
A second issue highlighted by this mistake is that chemical reaction mechanisms
are complicated, in ways that are sometimes hard to capture in computational studies.
For ketone hydrogenation by diamine dihydride ruthenium complexes, our first study
[10] had focused on understanding reactivity and kinetics in a non-chiral variant
of the reaction, carried out in a nonpolar solvent. Additionally, in that study, the
amidoamine intermediate shown on the top right in Fig. 5 (and bottom right in Fig. 3)
had been isolated and shown to be catalytically competent. The experimental rate
law was compatible with hydrogen splitting being turnover-limiting (although the
experimental activation enthalpy was somewhat smaller than the calculated one). All
of this suggests that the mechanism suggested in that paper, and shown in Figs. 3 and
4, is actually broadly correct under the reaction conditions used there (broadly correct
in that H 2 -splitting is indeed turnover-limiting, though the too-high barrier may mean
that the detailed mechanism for this step needs to be refined). In contrast, with protic
solvents, the variant (3) shown in Fig. 5—which is likely to be very unfavorable
under non-protic, nonpolar conditions—may dominate. The free energy landscape
underlying Fig. 5 is relatively flat, and its shape is sensitive to the environment.
Hence, mechanistic conclusions based on one set of reaction conditions may not
apply under all conditions.
3.3 Mechanism Discovery: Cis–Trans Isomerization
of Alkenes
In 2011, we published a study of a deceptively simple reaction: reversible cis–trans
isomerization of an alkene [21]; see Fig. 6. This isomerization was discovered by
our experimental collaborators in the group of Prof. Guy Lloyd-Jones, using the
simple Pd(II) catalyst PdCl 2 in solvents such as acetonitrile. The experimental studies
showed that this reaction could partly be caused by small amounts of palladium
hydride impurities formed under the reaction conditions, but conditions were found
where this side route could be inhibited, yet isomerization continued to be observed.
The question was, therefore, how does Pd(II) cause isomerization? Clearly, the alkene
could coordinate to Pd, which would weaken the double-bond character of the alkene