A number of research groups have data-mined the CSD and ICSD for relevant
structures in their area of interest [194], including ferroelectrics [195] and non-linear
optical materials [196]. As an example, one study extracted unbound 2,2
0 -bipyridine
ligands and used atom distances to predict if the complex would exhibit spin
crossover behaviour [197]. Another study performed property descriptor calculations on structures taken from the CSD and used this data to identify potential new
organic semiconductors [198, 199]. A similar methodology was also used to find
new a class of molecules for dye-sensitised solar cells [200].
It is not immediately and intuitively obvious that solid-state crystal structures
should be relevant in developing soft matter such as gels. However, two different
approaches have been used in this field. Firstly, a crystal engineering approach
mined the CSD for hydrogen-bonded synthons for the design of new low molecular
weight gelators [201]. Secondly, crystal morphology prediction has been used as an
indication of the directionality of strong intermolecular interactions to develop a gel
sensor to detect lead in paint [202].
Co-crystal design, as mentioned in an earlier section, is not only used in pharmaceutical and agrochemical research but can also be used in the rational design of
energetic materials. It has additionally been shown, as a proof of concept, that it is
possible to predict some key properties for energetic materials by machine learning
[203]. All the structures in the CSD, and indeed crystallography itself, rely on being
able to obtain a crystalline sample of the compound. This can be challenging for
some areas of chemistry. Using compounds known to crystallise, due to a crystal
structure being in the CSD, machine learning algorithms have been applied to
determine if a molecule will crystallise at all [204].
4 Closing the Loop and Future Prospects
4.1 How Is Data Now Driving the Scientific Process and What
Is the Future?
Many of the ‘solid-state rules’ have been worked out from amassing large volumes
of data. This has resulted in the fact that molecular geometry prediction and
assessment can readily be performed with a high degree of confidence,
e.g. Mogul. In turn, in recent years there has been an explosion of crystal engineering
resulting in a depth of understanding of intermolecular interactions that readily
allows for design and control of relatively large solid-state structures – and this
has led to fields such as supramolecular chemistry where complicated architectures
and molecular interactions can be controlled so that these systems can now exhibit
quite advanced functionality.
Leading Edge Chemical Crystallography Service Provision and Its Impact on. . .
121
structures in their area of interest [194], including ferroelectrics [195] and non-linear
optical materials [196]. As an example, one study extracted unbound 2,2
0 -bipyridine
ligands and used atom distances to predict if the complex would exhibit spin
crossover behaviour [197]. Another study performed property descriptor calculations on structures taken from the CSD and used this data to identify potential new
organic semiconductors [198, 199]. A similar methodology was also used to find
new a class of molecules for dye-sensitised solar cells [200].
It is not immediately and intuitively obvious that solid-state crystal structures
should be relevant in developing soft matter such as gels. However, two different
approaches have been used in this field. Firstly, a crystal engineering approach
mined the CSD for hydrogen-bonded synthons for the design of new low molecular
weight gelators [201]. Secondly, crystal morphology prediction has been used as an
indication of the directionality of strong intermolecular interactions to develop a gel
sensor to detect lead in paint [202].
Co-crystal design, as mentioned in an earlier section, is not only used in pharmaceutical and agrochemical research but can also be used in the rational design of
energetic materials. It has additionally been shown, as a proof of concept, that it is
possible to predict some key properties for energetic materials by machine learning
[203]. All the structures in the CSD, and indeed crystallography itself, rely on being
able to obtain a crystalline sample of the compound. This can be challenging for
some areas of chemistry. Using compounds known to crystallise, due to a crystal
structure being in the CSD, machine learning algorithms have been applied to
determine if a molecule will crystallise at all [204].
4 Closing the Loop and Future Prospects
4.1 How Is Data Now Driving the Scientific Process and What
Is the Future?
Many of the ‘solid-state rules’ have been worked out from amassing large volumes
of data. This has resulted in the fact that molecular geometry prediction and
assessment can readily be performed with a high degree of confidence,
e.g. Mogul. In turn, in recent years there has been an explosion of crystal engineering
resulting in a depth of understanding of intermolecular interactions that readily
allows for design and control of relatively large solid-state structures – and this
has led to fields such as supramolecular chemistry where complicated architectures
and molecular interactions can be controlled so that these systems can now exhibit
quite advanced functionality.
Leading Edge Chemical Crystallography Service Provision and Its Impact on. . .
121