7 Conclusions

The threshold behaviour of neutral atom and molecule collision strengths is not as clearly or simply determined as the high energy behaviour, with out access to detailed cross-section calculations. Thus the present extension of the Burgess-Tully method should be viewed as less rigorous than the original work and more as an operational convenience. Extrapolation and interpolation of collision data remain problematic in the case of very few data values and there is no real guarantee of precision. In practical terms, the B-C spline becomes markedly sensitive to small variations of the parameters when there are few data values, but then the strict interpolative approach based on approximate forms is both stable and safe. The method can be applied without change to electron impact ionisation collision strengths. However, it has been the practice to fit ionisation collision strengths by synthesising a superposition of direct and autoionisation components with adjustable threshold energies and equivalent electron multipliers. The latter approach is philosophically incompatible with the present method.

Acknowledgements

One of us (HPS) wishes to thank the National Institute of Fusion Science in Japan for their kind sponsorship of a visiting position. Dr. D. Brooks and Dr. A. Lanzafame prepared the ADAS Fortran version of the original Burgess-Tully C-plot code for ions.