levels of research. Furthermore, this level of resolution enables accurate quantum
mechanical calculations based on experimental observations. So-called quantum
crystallography [206] can calculate a range of properties and in particular quantitative energy calculations on interactions in the solid state. Such energetic calculations
provide a new dimension for crystallography, as traditionally intermolecular interactions can only be inferred between atomic centres and quantified by a distance, but
with this approach they can be directly observed. Furthermore, it is also possible to
derive and deconvolute whole molecule-molecule interaction energies which begins
to provide some of the information required to understand assembly (and disassembly) in the solid state.
Current state-of-the-art instrumentation is essentially at the point where this goal
would be achievable. However, there are two further factors which would need to be
addressed to achieve it. Firstly, there would have to be a significant input into
developing accessible and sustainable software to process and refine this higherresolution data and multipole models. Perhaps more of a challenge would be the
need to effect a cultural change towards conducting such higher-level experiments,
along with the significant amount of retraining and education that would be necessary. Furthermore, if there were to be a transition to collecting higher-resolution
structures, then a question would be raised as to how to treat the one million
structures already amassed in order to have comparable data.
4.1.3 Crystal Structure Prediction (CSP)
It is not the purpose of this review to comprehensively cover CSP; however, we note
here the impact of inclusion of experimental data. In 2000 the CCDC ran its first
blind test [207] to evaluate the state of methods for predicting crystal structures. The
ability to predict more complex structures with greater accuracy has improved
greatly in the subsequent blind test exercises. However, in the most recent, sixth
blind test [208], CCDC entered itself for the first time and used the known structures
in the CSD to predict unknown structures. The method used shape and packing
similarity to generate potential crystallographic lattices. Although not a complete
CSP solution, it proved valuable as a complementary technique reducing the chemical space needed to be searched using relatively cheap computer power and aid in
structure ranking – CSP Speculator [209].
Recent developments from the CSP community are using machine learning
approaches to explore the energy landscape of ensembles for predicted crystal
structures [210]. While these approaches are currently more confined to understanding conformations of lowest energy, it would be entirely possible to combine
experimental data and thereby use the power of both approaches. While CSP
contributes significantly to the corpus of knowledge on crystal structures, it is also
noteworthy in that it calculates a range of solid-state properties while doing so
(or can be combined with computational property calculation), and therefore this
complementary approach is destined to become increasingly powerful in the future.
Leading Edge Chemical Crystallography Service Provision and Its Impact on. . .
123
mechanical calculations based on experimental observations. So-called quantum
crystallography [206] can calculate a range of properties and in particular quantitative energy calculations on interactions in the solid state. Such energetic calculations
provide a new dimension for crystallography, as traditionally intermolecular interactions can only be inferred between atomic centres and quantified by a distance, but
with this approach they can be directly observed. Furthermore, it is also possible to
derive and deconvolute whole molecule-molecule interaction energies which begins
to provide some of the information required to understand assembly (and disassembly) in the solid state.
Current state-of-the-art instrumentation is essentially at the point where this goal
would be achievable. However, there are two further factors which would need to be
addressed to achieve it. Firstly, there would have to be a significant input into
developing accessible and sustainable software to process and refine this higherresolution data and multipole models. Perhaps more of a challenge would be the
need to effect a cultural change towards conducting such higher-level experiments,
along with the significant amount of retraining and education that would be necessary. Furthermore, if there were to be a transition to collecting higher-resolution
structures, then a question would be raised as to how to treat the one million
structures already amassed in order to have comparable data.
4.1.3 Crystal Structure Prediction (CSP)
It is not the purpose of this review to comprehensively cover CSP; however, we note
here the impact of inclusion of experimental data. In 2000 the CCDC ran its first
blind test [207] to evaluate the state of methods for predicting crystal structures. The
ability to predict more complex structures with greater accuracy has improved
greatly in the subsequent blind test exercises. However, in the most recent, sixth
blind test [208], CCDC entered itself for the first time and used the known structures
in the CSD to predict unknown structures. The method used shape and packing
similarity to generate potential crystallographic lattices. Although not a complete
CSP solution, it proved valuable as a complementary technique reducing the chemical space needed to be searched using relatively cheap computer power and aid in
structure ranking – CSP Speculator [209].
Recent developments from the CSP community are using machine learning
approaches to explore the energy landscape of ensembles for predicted crystal
structures [210]. While these approaches are currently more confined to understanding conformations of lowest energy, it would be entirely possible to combine
experimental data and thereby use the power of both approaches. While CSP
contributes significantly to the corpus of knowledge on crystal structures, it is also
noteworthy in that it calculates a range of solid-state properties while doing so
(or can be combined with computational property calculation), and therefore this
complementary approach is destined to become increasingly powerful in the future.
Leading Edge Chemical Crystallography Service Provision and Its Impact on. . .
123