Using a very simple mean field description for the dynamics of soap froth, we discovered that a continuous (one-parameter) family of stationary solutions exists [11] , [C3] . The one that corresponds to the physically observed universal scaling state is singled out by a selection mechanism. Next we studied various very special initial conditions of the froth, and found that indeed evolution can proceed to diferent asymptotic behaviors [27] . We studied temporal correlations by looking at the survivors [21] , the subset of soap cells that survive for long times. These studies revealed shortcomings of the existing mean field approaches, as well as discrepancies between various simulation techniques and experiments. The source of these discrepancies was identified and incorporated in various topological simulations, yielding excellent agreement with experiment [27] . We made a connection [35] between the survivors' areas and the problem of non-flip probabilities in various models.
The related experimental work motivated our studies of Ostwald ripening in a 2-d crystalline system. The dynamic equations were mapped onto a 2-d Coulomb problem, for which a numerically efficient perturbative solution was obtained [37] [38]
I am not working currently in this field. The most interesting long-term basic issue I may return to is to find an effective ensemble approach to such non-equilibrium problems that evolve to a scaling state.
In the context of non-equilibrium problems I got also involved in some theoretical models. We solved a model [3] , [C1] of domain growth in a 2-d Ising ferromagnet at T=0 and mapped it onto the 8-vertex model [6] and one of particles that hop on a line. When the hopping rate is biased one enters a class of asymmetric exclusion processes, for which we were able to obtain analytic solutions that exhibit interesting phase transitions [16] , [17] . I also worked on damage spreading in a variety of dynamic models. By considering entire families of possible models that implement a model's evolution, we were able to give an objective, observer - independent definition of damage spreading phases [33] . We discovered that by extending the family of dynamic procedures we can generate a damage spreading transition for the one-dimensional Ising model; furthermore, we identified the universality classes of these damage spreading transitions [34] ,. Finally, we proposed [52] , [53] a that experiments performed on sand flowing along an inclined plane may provide experimental realization for a class of generalized directed percolation models.