4.1.1 Data Science
There is still new ground to be covered and many grand challenges exist. For
example, it is not readily possible for engineering and manufacturing sectors to get
an immediate response to a demand for a material with a particular property or
function. The future of structural science will be in addressing these issues and data
is the key – not specifically collecting more data, although that is part of the solution
in some respects, but more being able to link structural data to property/function
data. Also, more complex hierarchical materials and multiple component or hybrid
materials require a greater understanding to generate, so-called Directed Assembly –
which is the subject of a ‘Grand Challenge’ supported by the UK Engineering and
Physical Sciences Council [205]. In fact, there is an equal challenge in Directed
Disassembly, which requires a deep level of understanding of structure, while we
still fundamentally understand very little about the processes of nucleation and
crystallisation. There is much data engineering to be done before the promised riches
of data science can be realised – however, once it is possible to extract data from
different databases on demand and run algorithms over them, then whole new
research opportunities are opened up. Discovery, recognition and utilisation of
patterns in data are fundamental in data science, and chemical/solid-state structure
would be a key element in driving these approaches in application to chemical
problems.
Structural similarity and structural informatics in the solid state are still relatively
unexplored, yet have huge potential. This way of thinking raises the question of how
big ‘achievable solid-state space’ actually is, i.e. if there were no barriers, how many
crystal structures could we actually collect? This in turn leads on to recognition that
there are currently significant ‘gaps’ in our databases. These gaps are mainly due to
the fact that the majority of data arises from the traditional literature and therefore,
e.g. collection of homologous series, that which is not deemed worthy of publication, ‘uninteresting chemistry’, etc. is work that is not undertaken. Data gathering
exercises need to be given more value and credit if they are to be incorporated into a
collection that can then be further utilised in many, many different ways. So, an
immediate question is therefore one of how to identify these gaps and which ones are
the most valuable to fill?
4.1.2 Higher-Resolution Structural Information
A logical progression that arises from the advances in instrumentation that are being
realised is one of the resolutions of structural information that is potentially achievable. It is conceivable that with the right developments, then the time and effort taken
to collect data and refine multipole models for charge density-level resolution will be
drastically reduced. So, what is to stop this from becoming the normal approach to
service crystallography? Databases that go beyond utilising atomic coordinates and
allow investigation of electron distribution would be very powerful and open up new
122
S. J. Coles et al.
There is still new ground to be covered and many grand challenges exist. For
example, it is not readily possible for engineering and manufacturing sectors to get
an immediate response to a demand for a material with a particular property or
function. The future of structural science will be in addressing these issues and data
is the key – not specifically collecting more data, although that is part of the solution
in some respects, but more being able to link structural data to property/function
data. Also, more complex hierarchical materials and multiple component or hybrid
materials require a greater understanding to generate, so-called Directed Assembly –
which is the subject of a ‘Grand Challenge’ supported by the UK Engineering and
Physical Sciences Council [205]. In fact, there is an equal challenge in Directed
Disassembly, which requires a deep level of understanding of structure, while we
still fundamentally understand very little about the processes of nucleation and
crystallisation. There is much data engineering to be done before the promised riches
of data science can be realised – however, once it is possible to extract data from
different databases on demand and run algorithms over them, then whole new
research opportunities are opened up. Discovery, recognition and utilisation of
patterns in data are fundamental in data science, and chemical/solid-state structure
would be a key element in driving these approaches in application to chemical
problems.
Structural similarity and structural informatics in the solid state are still relatively
unexplored, yet have huge potential. This way of thinking raises the question of how
big ‘achievable solid-state space’ actually is, i.e. if there were no barriers, how many
crystal structures could we actually collect? This in turn leads on to recognition that
there are currently significant ‘gaps’ in our databases. These gaps are mainly due to
the fact that the majority of data arises from the traditional literature and therefore,
e.g. collection of homologous series, that which is not deemed worthy of publication, ‘uninteresting chemistry’, etc. is work that is not undertaken. Data gathering
exercises need to be given more value and credit if they are to be incorporated into a
collection that can then be further utilised in many, many different ways. So, an
immediate question is therefore one of how to identify these gaps and which ones are
the most valuable to fill?
4.1.2 Higher-Resolution Structural Information
A logical progression that arises from the advances in instrumentation that are being
realised is one of the resolutions of structural information that is potentially achievable. It is conceivable that with the right developments, then the time and effort taken
to collect data and refine multipole models for charge density-level resolution will be
drastically reduced. So, what is to stop this from becoming the normal approach to
service crystallography? Databases that go beyond utilising atomic coordinates and
allow investigation of electron distribution would be very powerful and open up new
122
S. J. Coles et al.