292
J. N. Harvey
within which molecules come together to react. The benefit in terms of supporting
researchers’ intuitions is considerable.
All of these factors have led to an explosion in the number of scientists studying
reaction mechanisms computationally, with an accordingly large rise in the number
of publications. Like most changes, the huge benefits brought about by this shift
in emphasis within mechanistic studies carry some problems with them also. These
negative aspects show up in a variety of ways, as does the awareness that the field has
of these difficulties. Here, I would like to cite a withering criticism of computational
studies published by the editors of a major journal in the field of catalysis, as part of
an editorial about ‘ethical conduct.’ In their perspective, the authors write:
A second example is that researchers often overly rely on computation in interpreting a
reaction mechanism. The conclusions derived from an unrealistic assumption are far from
truth, and confuse and mislead the community. It is worse than nothing. The problem is not
the quality of the calculations, but arises from the abuse of a powerful tool by inexperienced
researchers. We all know that organic synthesis with wrong starting materials will never
result in the correct target product. We should appreciate solid experimental evidence more
than frivolous computations. This prevailing trend is harmful to the community [6].
As can be seen, the authors explicitly argue both that computational studies can
be wrong and misleading and that experimental studies should be preferred. Such
an attitude can be expected to arise whenever a field such as mechanistic studies
undergoes such a large shift in its research methodology in a relatively short period.
As such, some of the language in the above extract can appear exaggerated or beside
the point. Nevertheless, in my view, the arguments expressed are valuable and find
many echoes in contemporary evaluations of mechanistic computational chemistry.
In the following sections, I will attempt to explore some of the issues that lead to
such attitudes, and how the field can respond to these challenges in future.
3 Reactivity Studies in Organic and Organometallic
Chemistry
The work in our group on mechanistic organic and organometallic chemistry was
initiated shortly after I started to work as an independent researcher in computational
chemistry in 1999, and was facilitated by the fact that my Ph.D. studies were in
the field of experimental organic chemistry, with a focus on reaction mechanisms
involving organosulfur species [7, 8]. The first organic [9] and organometallic [10]
studies carried out in my group were both published in 2002. I had not initially
intended to perform such work, but it became obvious that there was a large demand
for such studies and that the existing methods had the scope to provide significant
mechanistic input. Already at the time, many experimental groups were performing
their own computational studies, and the key density functional theory (DFT) and
continuum solvation (or polarizable continuum model, PCM) techniques that still
form the core of computational studies were available. This led to a perception that
theoretical work in mechanistic organic chemistry was a very ‘applied’ subfield of
J. N. Harvey
within which molecules come together to react. The benefit in terms of supporting
researchers’ intuitions is considerable.
All of these factors have led to an explosion in the number of scientists studying
reaction mechanisms computationally, with an accordingly large rise in the number
of publications. Like most changes, the huge benefits brought about by this shift
in emphasis within mechanistic studies carry some problems with them also. These
negative aspects show up in a variety of ways, as does the awareness that the field has
of these difficulties. Here, I would like to cite a withering criticism of computational
studies published by the editors of a major journal in the field of catalysis, as part of
an editorial about ‘ethical conduct.’ In their perspective, the authors write:
A second example is that researchers often overly rely on computation in interpreting a
reaction mechanism. The conclusions derived from an unrealistic assumption are far from
truth, and confuse and mislead the community. It is worse than nothing. The problem is not
the quality of the calculations, but arises from the abuse of a powerful tool by inexperienced
researchers. We all know that organic synthesis with wrong starting materials will never
result in the correct target product. We should appreciate solid experimental evidence more
than frivolous computations. This prevailing trend is harmful to the community [6].
As can be seen, the authors explicitly argue both that computational studies can
be wrong and misleading and that experimental studies should be preferred. Such
an attitude can be expected to arise whenever a field such as mechanistic studies
undergoes such a large shift in its research methodology in a relatively short period.
As such, some of the language in the above extract can appear exaggerated or beside
the point. Nevertheless, in my view, the arguments expressed are valuable and find
many echoes in contemporary evaluations of mechanistic computational chemistry.
In the following sections, I will attempt to explore some of the issues that lead to
such attitudes, and how the field can respond to these challenges in future.
3 Reactivity Studies in Organic and Organometallic
Chemistry
The work in our group on mechanistic organic and organometallic chemistry was
initiated shortly after I started to work as an independent researcher in computational
chemistry in 1999, and was facilitated by the fact that my Ph.D. studies were in
the field of experimental organic chemistry, with a focus on reaction mechanisms
involving organosulfur species [7, 8]. The first organic [9] and organometallic [10]
studies carried out in my group were both published in 2002. I had not initially
intended to perform such work, but it became obvious that there was a large demand
for such studies and that the existing methods had the scope to provide significant
mechanistic input. Already at the time, many experimental groups were performing
their own computational studies, and the key density functional theory (DFT) and
continuum solvation (or polarizable continuum model, PCM) techniques that still
form the core of computational studies were available. This led to a perception that
theoretical work in mechanistic organic chemistry was a very ‘applied’ subfield of