Abstract. The ionisation probability for target L subshells by light ion impact is calculated in the semiclassical approximation, including the couplings between the subshells during the collision. The inclusion of these couplings is found to modify the subshell occupation numbers only slightly for proton and alpha particle impact on heavy target atoms, but a marked effect is found for heavier projectiles. Numerical results for the systems H e + Pt, Ne+ Pt and N e + % are presented.There is in general quiie satisfactory agreement between perturbation theory and experiments for inner-shell ionisation in very asymmetric ion-atom collisions (see e.g. the contributions in Paul 1982). However, when one considers L-subshell ionisation, some significant discrepancies have persisted since the first cross section measurements were reported, in particular at very low collision energies (Daiz er a1 1974, Chang et a1 1975). Introducing a relativistic electron description, using united-atom ( C A ) binding energies a n d wavefunctions and corrections for projectile Coulomb deflection only marginally improves the situation (Amundsen 1977, Aashamar andAmundsen 1982). It was unclear, however, for a long time to what extent these discrepancies are spurious, in the sense that they arise from errors in the values of the fluorescence yields and Coster-Kronig ratios used t o convert measured x-ray intensities into subshell ionisation cross sections. Recent accurate measurements of the radiative and radiationless yields for Au (Jitschin et a1 1985) have indeed given results significantly different from those recommended by Krause (1979), but, as these authors note, it does not explain all the differences between theory and experiment for ionisation. Such discrepancies also appear in some o f the recent measurements of impact-parameter-dependent ionisation probabilities, P ( b ) (Dexheimer et a1 1986, Zehender et a1 1986).A widely accepted explanation for the disagreements between theory and experiment is that perturbation theory is inadequate even in rather asymmetric collisions. This effect was first studied quantitatively by Sarkadi and Mukoyama (1981) for systems like C + A u in a simple two-step model, treating ionisation and subshell couplings as independent processes. This two-step model has subsequently been refined (Fink et ai 1983, Sarkadi a n d Mukoyama 1984, Sarkadi 1986a