CLOSING REMARKS BY H. CHRISTOPHER LONGUET-HIGGINS In commenting on this discussion I feel rather like a man who has just stepped out of a time machine which he boarded ten years ago. Since that time the Faraday Society has got married to the Chemical Society, but her character has been unspoiled by the change, and Faraday Discussions are as well run and as scientifically exciting as they ever were. It is impossible in five or ten minutes to do justice to all the contributions, so I may be forgiven for merely mentioning one or two points which have crossed my mind in listening to this excellent series of papers and the remarks which have followed them. The variety of topics which have been discussed bears witness to the central role of the potential energy surface as a unifying concept in physical chemistry.Its importance has long been recognised in spectroscopy and chemical kinetics, but ten years ago we knew very little about the potential surfaces of polyatomic molecules except in the immediate neighbourhood of their energy minima. One striking advance in this area has been in the experimental and theoretical investigation of non-rigid, or " floppy ", molecules about which Mills told us in his paper. I was particularly intrigued by the phenomena which he reported on carbon suboxide and related molecules. I hope it will soon be possible to suggest, in descriptive terms, why this molecule is not linear in its ground state and why the effective potential for bending is so markedly affected by excitation of the skeletal stretching vibrations. The reliable determination of potential energy surfaces from vibrational spectra is a mathematically challenging problem, and so is the inverse problem of determining vibrational energy levels from non-separable potentials, where there have been some very significant advances, in the hands of Miller, Handy and others.The use of semi- classical approximations obviously has a rosy future, and one may expect that such methods will also prove valuable in molecular dynamics, where one is anxious to understand scattering cross-sections and product state distributions in terms of pre- sumed potential energy surfaces. This brings one to a consideration of the very considerable experimental advances about which we have heard from Herschbach and many other authors.Herschbach reminded me of an experienced batsman attempting to infer the action of a spin bowler from the way in which the ball bounces from the crease; his lecture revealed the richness of the information which could be obtained from an experimental measurement of all the angular momenta involved in both elastic and inelastic scatter- ing processes. Polanyi set an example of how one should set about interpreting the translational, rotational and vibrational excitation of the products of a simple chemical reaction, at least according to classical molecular dynamics. One presumes that the next stage in such work will be a quantitative improvement in the accuracy with which one can determine the precise form of the energy surface. A particularly intriguing reaction, of which we were given both experimental and theoretical analyses, is that between Nf and HZ.(In passing one may note the con- fidence with which chemists can now handle non-hear as well as linear configurations348 POTENTIAL ENERGY SURFACES of atoms, in discussing reaction mechanisms.) A nice feature of this reaction is the crossing which occurs between the 3B, and 3A2 states when the N+ ion approaches the H2 molecule from the side. As Dewar pointed out, this crossing is directly interpret- able in qualitative terms, and one would like to see more such simple explanations for the rather forbidding results of computations on polyatomic systems. It has always seemed to me that there are three kinds of Chemistry: experimental, theoretical and computational.Most chemists tend to think of molecular computa- tions as belonging to theoretical chemistry, but it could be argued that such computa- tions are really experiments. Conventional experiments are carried out on real atoms and molecules; computational experiments are performed on more or less ‘‘ modest ” and unreliable models of the real thing. So the computational chemist has even more of an obligation than the experimentalist to interpret his results in an intelligible fashion. Unless he can offer a convincing explanation why his numbers come out as they do, there is always room for serious doubt as to whether they may not be artifacts of his basic approximations. This is the substance of most current objections to heavy computations of molecular properties by ab initio methods.If such methods have been well attested for a given class of problems, then it is not unreasonable to attach weight to the computational solution of a further problem in that particular class. Unfortunately, the most interesting problems are usually those with some element of novelty. I cannot close without referring to two matters in particular which have generated much comment during the discussion: the interaction between three alkali metal atoms, and the phenomenon of laser-induced fluorescence. It is delightful to see both theorists and experimentalists hunting down the conical intersections whose existence was predicted by Teller so long ago, and one may hope that before very long it may be possible to obtain concrete evidence for the phase change which is confidently expected to occur in the electronic wave function when one circles round such an intersection. The phenomenon of laser-induced fluorescence is, as George rightly remarked, a new challenge to theory as well as to experiment. Plainly there is no shortage of interest- ing experimental phenomena for the theoretician to get his teeth into, nor of theoretical issues inviting resolution by ingenious experiment,