5 Conclusion

The present experiment showed results with a considerable range of density-contrast. While the total size of the studied region was about 40 pc, and the average density about 100 particles cm^-3 in the experiments, structures as small as 10^-1 pc, and densities of $\sim{10}^{4}$ cm^-3 were observed.

This performance was achieved with the aid of a gather-scatter smoothing length, a scatter variable softening of the gravity field, and an adaptive integration-scheme according to the locally required time-resolution. The adoption of the scatter softening length was done in detriment of energy conservation. However, this energy error was negligible in comparison to the high amount of energy loss due to the efficient cooling.

In the first steps of the experiment, the gravity was almost irrelevant on the formation of the shock. In this phase, the Dalgarno & McCray model was decisive on the fast cooling of the shock front and its kinetic energy dissipation. In the subsequent steps, as the shock front cooled considerably, the molecular cooling was preponderant. The strong contraction of both clouds onto almost a surface in the head-on experiment was then merely a conversion of kinetic energy into radiation due to the cooling efficiency, This efficiency increased as the density increased (shown in Fig. 1), which consequently allowed thermal instability along with the critical phases of the shock. Just in this phase density fluctuations were amplified, resulting in high-density peaks in the clumped configuration of Figs. 4 and 5.

The experiments demonstrate that clumps formation are strongly stimulated by supersonic cloud-cloud collisions, mainly due to the thermal instability. The main reason for the differences between our results and previous works is that the efficient cooling due to molecules at temperatures below 3000 K was not usually taken into account (e.g., [Lattanzio et al. 1985]; [Lattanzio & Henriksen 1988]). This cooling mechanism for higher temperatures was relevant in the first two steps of the collision, where the hotter parts of the shock reached temperatures of the order of 1000 K and then dropped fastly down to temperatures comparable to 100 K.

During the collision, the shocked layer becomes cooler than its surroundings. Consequently, the gas continuously accretes on both sides of the shocked layer. The time-scale for re-expansion of the gas after the end of the shock, in the head-on case, is found to be larger than the one obtained by Lattanzio & Henriksen (1988) in a similar situation; the efficient energy dissipation prevents re-expansion.

It seems that the dense molecular layer, produced by the shock, survives as long as it is maintained by the extra external pressure of the accumulated gas, and starts to dissipate as soon as the shock ends. In this sense, the dense molecular phase is possibly not an equilibrium phase of the interstellar medium at normal pressures. Encounters between molecular clouds seem to be an efficient mechanism to build up enough interstellar matter to reach densities of $\sim{10}^{4}$ cm^-3. If indeed molecular clouds are produced and maintained by collisions, we suggest that most of them could be geometrically thin, like the flat dense layers observed in our experiments.

Acknowledgements

E.P. Marinho is grateful to Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), SP - Brazil, by a financial support under contract number 97/6157-4 and to Carmen Maria Andreazza by useful suggestions and a critical reading of the manuscripts. We would like to thank our referee James Michael Owen for his constructive criticisms and advices on an earlier version of this paper.