Fresenius Z. Anal. Chem. 300, 6-7 (1980)

Trends in Analytical Chemistry D. Betteridge Chemistry Department, University Collegeof Swansea, Swansea SA2 8PP, U.K.

Thirty years ago analytical science was on the verge of a massive expansion. Spectrophotometers, pH-meters, infra-red spectrometers, radiochemical methods, chromatography and ion-exchange resins were coming into wide-spread use and the emphasis was changing from the analysis of major constituents to trace analysis. We are now at the end of that period and can look back on some remarkable developments; labs have been transformed from fume-filled rooms full of people using a limited amount of glassware to rooms full of instruments, computer terminals and purified air but nearly devoid of people; the unit " p p m " has changed from a measure of insignificance to an indication of unacceptably high concentration levels; the size of samples has decreased by a factor 106; it has become routine to expect in less than 1 min multielement analysis and in a few hours the multicomponent analysis of extremely complicated organic samples e.g. drugs in the blood of race horses; and apparatus is expected to work under most inhospitable conditions. The undoubted peak of non-routine analysis was to take the analytical lab to the sample and send back the results 108 miles through space. The development of routine analysis is epitonised by the Laboratories of the Wessex Water Board, in which, under the supervision of Dr. J. G. Jones, 30,000 samples each of which require 11 chemical determinations, and 10,000 samples requiring 3 microbiological tests are analysed each year by a team of 12 scientists and at a total cost of s 160,360. It is reasonable to expect that the search for new, more sensitive methods will continue, especially as there seems to be no shortage of people publically proclaiming that some substance, hitherto undetectable, is extremely hazardous to the human species. However, there are signs of a surfeit of techniques, for trace analysis, and the problem does not represent to the analyst of the 1980s the challenge that it did to his counterpart of the 1950s.

Great developments have taken place in other sciences and these will have a great influence on the further development of analytical science. Most notable of these are (i) large scale integrated (LSI) circuits, which result in cheap, easily assembled, robust electronic devices such as microprocessors and (ii) the establishment of novel numerical methods such as pattern recognition (or cluster) analysis and systems analysis. The former of these is giving rise to powerful computers which are not only within the budget of every laboratory, but are cheap enough to dedicate to the simplest of analytical instruments. Electronic components are compact, robust and easy to assemble. They are well suited for instrumentation, which must function reliably over long periods of time with little attention. The rate of development in this area is indicated by the growth of the microprocessor industry since the introduction of the first chip in 1971. Whilst most attention is given to computing power, the facility with which these devices can control instruments and collect data is also of the utmost significance. The practical position now is that the analyst can measure more and more in less and less and to do it with greater frequency; and the pace of things is accelerating. The analyst is thus generating huge numbers of figures. For what purpose? The questions for the 80s are whether the maximum of analytical information is being obtained from this immense output, and whether it is being put to good use. At present it is common to examine the results of a multielement analysis, element by element to see if any one of them exceeds some specified limit. Would it not be more interesting to look at the pattern of the total analysis? It would then be possible for example, to establish whether a particular source of pollution was varying its contribution to the general level. This is one of the capabilities of pattern recognition analysis.

The cost and effort of establishing an analytical system for performing the sort of analysis which is now considered to be common place, e.g. the scanning of blood samples of patients in large hospitals, justifies consideration of the total system, not just the analytical techniques available. This is the province of systems analysis and/or operational research, which are well established management techniques. The application of Fourier analysis via the fast Fourier transform, has revolutionised NMR and infrared spectroscopy. Are there any other mathematical methods which will enable us to extract useful analytical information from an apparently non-descript signal? The answer is unquestionably, yes. The analyst, as always, must be aware of the latest mathematical methods, but he is not required to transform himself into a mathematician. There is a growing availability of relevant programs in computer library and the production of the most important ones on chips. Because of the advances in electronics and computational methods complicated mathematical procedures are now available to the analyst at the press

of a button. Moreover, a lot of the groundwork for analytical application has been laid by de Clerc, Deming, Denton, Isenhour, Jurs, Kowalski, Massert, and Perone to name but a few. In our laboratory, we have found that improved computational facilities have enabled us to extract more and better information from spectra, have simplified automated methods and are helping to develop a novel method of analysis based on acoustic emissions. They are leading to improvements in existing methods and opening up possibilities for new ones. Of greater importance is the opportunity of using the total information provided by the analysis, because the power of diagnosis, the province of the analyst throughout history, is greatly enhanced. Bearing in mind the interdisciplinary growth of the subject, and the scale of the problems which are being submitted for analysis the next thirty years should be just as exciting as the last, but despite (or because of?) the developments in instrumentation the difficult part in an analysis will still be getting the chemistry right.