It was Winston Churchill in his July 8th, 1920 House of Commons speech condemning General Dyer's role in the 1919 Amritsar Massacre (1) who noted that "it was compassion and its absence which marked the difference between Englishmen and Bolsheviks" (2); just as it could be said that is it software and its absence, that marks the difference between being able to perform and not to perform modern crystallography. Modern crystallographic software is fundamental to present routine structure determination and the continuation of crystallography as a dynamic science. If new or updated computer algorithms are unavailable in a readily useable form, new areas of crystallography grind down to a snail's pace . It can be argued that lack of new, "available" algorithms is affecting to a great extent the possibility of routinely using powder diffraction data for ab-initio structure solution (3) (4). Structure solution from powder diffraction data does not come easy compare to single crystal diffraction; where readily available structure solving programs such as Shelxs (5), Patsee (6), Sir (7) and Dirdif (8) are well adapted to a wide range of single crystal systems.
A present threat and disincentive to the development of leading edge academic crystallographic software in Britain is the indifference this arduous research activity encounters in performance reviews. It is difficult to imagine any sound logic or rationale that denies academic software developers full and proper acknowledgment for their research and work, especially when it also holds such major benefits to the scientific community. Modern crystallography and crystallographers are reliant on those who undertake intellectually challenging and demanding research endeavours to develop useable, practical algorithms and software. This is but a one issue, though the major one, revolving around the field of new cutting edge crystallographic software development.
A not uncommon utterance is that many "new" algorithmic techniques have apparently re-invented the wheel, "supposedly" replicating work performed before in some form or another (9) (10). However, one must respect that many of these "supposed" re-inventions of the wheel differ greatly from their predecessors by having crossed over the bridge into being a practical scientific method incorporated into useable software. Contrast this with the algorithmic methods published in the literature that have yet not crossed the bridge between "not science" and "science". It is open, routine, continuous and ever present validation of software, and its absence that marks the difference between what is scientific software, and what is not. This is not to infer or imply that many of these developing, but unrealised methods published in the literature are not exciting and "cracking hot stuff". However, in the spirit of the Royal Society's motto of "Nullius in verba" (11), the scientific method involves never ending, routine validation and testing. Software and algorithms that are not openly accessible circumvents what is a basic requirement for "scientific software"; its availability for routine validation through continuous and open usage.
Of many exciting claims have been made in the literature over new crystallographic algorithms, some include "routine" methods for solving molecular structures from powder diffraction data. In the spirit of extraordinary claims requiring extraordinary evidence, the Internet means there is no technical obstacle in making new algorithms and software available to put through the rigours of falsification and testing. Indeed a quality measure of a scientific claim or theory is how much effort the authors have gone through to make their work as easily understandable, testable and verifiable as possible. Le Bail extraction is a notable example in powder diffraction where effort was undertaken by its creator, Armel Le Bail, to provide relevant information to the scientific community, such that it could be openly applied and tested in a time effective manner (12). It is the open availability of the Le Bail extraction in a variety of freely available software packages that has helped enable the "relative explosion" in the number of ab-initio structure solutions from powder diffraction data.
Another trend in academic software development is the push to commercialise in the hope of realising extra sources of external income. As this can lock new, innovative software away from the bulk of crystallographers, it should be noted that strategies exist that allow both options of commercialised software that is still freely available to academics (7) (5). There are obvious advantages of providing freely available software to academics. These includes enabling a critical mass of users and testers and obtaining a dynamic resource pool of new ideas to keep the software ahead of rivals. If necessary, providing compiled binaries can protect the integrity of the source code. Many, if not most users of crystallographic software prefer ready to use binary executables instead of source code. Is it also just coincidence that some of the most respected names in crystallography also supply freely available "benchmark" software for academics?
Despite the above statements of gloom, there is still excellent software being authored and freely distributed by British academics. Programs include the Crystals Single Crystal Suite (13); WinGX(14) and GX (15) single crystal suites; PROFIL Powder Diffraction Suite (16); CRYS2RUN Powder Indexing Suite (17); UNITCELL powder cell refinement (18); XHYDEX for locating hydrogens (19); CRUSH for modelling Polyhedra Frameworks (20), etc. Though it is debateable how long even this can continue when the pivotal role of academic software development is not appropriately acknowledged by existing research review processes.
What is to be the future? Science does not doddle on in a linear and predictable fashion. Who is to know what developments may render diffraction based crystallography hopelessly obsolete in the future? However, for the present, crystallography needs a continuing stream of modern software and algorithms to cope with more difficult and challenging scientific problems. Current trends could imply that progress will not be limited by availability of crystallographic data; but the ability to analyse the majority of this data in a routine manner. This will most likely be mainly due to the current disincentive to develop new algorithms and software, convoluted with the non-availability of what has already been published in the scientific literature.
Lachlan M. D. Cranswick
CCP14 Secretary
Collaborative Computational Project Number 14 for Single Crystallography and Powder Diffraction,
Daresbury Laboratory, Warrington, Cheshire, WA4 4AD U.K
E-mail: [email protected]
WWW: http://www.ccp14.ac.uk