Mechanism and Kinetics in Homogeneous Catalysis …
311
demanding. In our study of ruthenium-catalyzed hydrogenation of ketones [18], we
made a mistake in our analysis of this type. This was pointed out to us by a referee
prior to publication, fortunately, and could be remedied by more careful thinking.
I argue that such errors occur quite frequently and that beyond a certain level of
mechanistic complexity, authors would be well advised to carry out explicit kinetic
simulation [28] of the network of reaction steps they have modeled, in order to check
that the emerging behavior meshes with that which they had expected.
A third recommendation is that when studying details of reaction mechanisms, one
should pay great attention to the fact that most interesting chemical transformations
do not have one single mechanism, but rather exist on a mechanistic manifold, with
the preferred route or rate-limiting step sometimes changing when substrates or
reaction conditions are changed.
A fourth point is that the ‘unknown unknowns’ in mechanistic exploration will
continue to tax our ingenuity as computational chemists and invite us to think carefully about possible mechanisms. The unknown unknowns here refer to species or
transition states that we simply have not thought about and that perhaps cannot even
be represented while using the atomistic model that we have chosen to use to study
the problem at hand. Occasionally, as in our study of palladium-catalyzed alkene
isomerization, these unknown unknowns, once they become known, will turn out to
lie low in free energy and to play an important role in the reaction.
One more conclusion emerging from this work is that statistical mechanics plays
an important role in mechanistic organic and organometallic chemistry. Because
the corresponding calculations typically performed, using simple ‘ideal-gas’ statistical mechanics with the rigid-rotor and harmonic oscillator approximations and the
Sackur–Tetrode equation for translational entropy, are very undemanding computationally, in comparison with the electronic structure calculations, there may be a
tendency to overlook their importance and to reach incorrect conclusions about their
accuracy. For example, we ourselves argued at one point [22] that adjustments to
solution-phase free energies may need to be applied, based on an incorrect analysis
of the theoretical framework. In fact, more recent work makes it clear that this suggestion is seriously incorrect and that the ‘standard’ ideal-gas-type expressions are
reasonably correct.
As an overall conclusion taken from this work, I would like to argue that the computing accurate free energy changes, rate constants and mechanisms for organic and
organometallic chemistry is still highly challenging. There has been huge progress,
so that computation can now provide valuable assistance in studying mechanisms,
and indeed in some cases, near-quantitative conclusions can be obtained. However,
for the foreseeable future at least, it does not appear likely that computational chemistry will become predictive enough to be able to completely replace experimental
approaches for studying mechanisms.
311
demanding. In our study of ruthenium-catalyzed hydrogenation of ketones [18], we
made a mistake in our analysis of this type. This was pointed out to us by a referee
prior to publication, fortunately, and could be remedied by more careful thinking.
I argue that such errors occur quite frequently and that beyond a certain level of
mechanistic complexity, authors would be well advised to carry out explicit kinetic
simulation [28] of the network of reaction steps they have modeled, in order to check
that the emerging behavior meshes with that which they had expected.
A third recommendation is that when studying details of reaction mechanisms, one
should pay great attention to the fact that most interesting chemical transformations
do not have one single mechanism, but rather exist on a mechanistic manifold, with
the preferred route or rate-limiting step sometimes changing when substrates or
reaction conditions are changed.
A fourth point is that the ‘unknown unknowns’ in mechanistic exploration will
continue to tax our ingenuity as computational chemists and invite us to think carefully about possible mechanisms. The unknown unknowns here refer to species or
transition states that we simply have not thought about and that perhaps cannot even
be represented while using the atomistic model that we have chosen to use to study
the problem at hand. Occasionally, as in our study of palladium-catalyzed alkene
isomerization, these unknown unknowns, once they become known, will turn out to
lie low in free energy and to play an important role in the reaction.
One more conclusion emerging from this work is that statistical mechanics plays
an important role in mechanistic organic and organometallic chemistry. Because
the corresponding calculations typically performed, using simple ‘ideal-gas’ statistical mechanics with the rigid-rotor and harmonic oscillator approximations and the
Sackur–Tetrode equation for translational entropy, are very undemanding computationally, in comparison with the electronic structure calculations, there may be a
tendency to overlook their importance and to reach incorrect conclusions about their
accuracy. For example, we ourselves argued at one point [22] that adjustments to
solution-phase free energies may need to be applied, based on an incorrect analysis
of the theoretical framework. In fact, more recent work makes it clear that this suggestion is seriously incorrect and that the ‘standard’ ideal-gas-type expressions are
reasonably correct.
As an overall conclusion taken from this work, I would like to argue that the computing accurate free energy changes, rate constants and mechanisms for organic and
organometallic chemistry is still highly challenging. There has been huge progress,
so that computation can now provide valuable assistance in studying mechanisms,
and indeed in some cases, near-quantitative conclusions can be obtained. However,
for the foreseeable future at least, it does not appear likely that computational chemistry will become predictive enough to be able to completely replace experimental
approaches for studying mechanisms.