1: \begin{abstract}

2: We establish rates of convergence of solutions to scaling (or similarity) profiles in a coagulation type system modelling 

3: submonolayer deposition. We prove that, although all memory of the initial condition is lost in the similarity limit, information

4: about the large cluster tail of the initial condition  is preserved in the rate of approach to the similarity profile.

5: The proof relies in a change of variables that allows for the decoupling of the original infinite system of ordinary differential equations into a closed 

6: two-dimensional nonlinear system for the monomer--bulk dynamics and a lower triangular infinite dimensional linear one for the cluster

7: dynamics. The detailed knowledge of the long time  

8: monomer concentration, which was obtained earlier by Costin et al.

9: in \cite{costin} using asymptotic methods and is rederived here by center manifold arguments, is then used for the asymptotic

10: evaluation of an integral representation formula for the concentration of $j$-clusters. The use of higher order expressions, both

11: for the Stirling expansion and for the monomer evolution at large times allow us to obtain, not only the similarity limit, but also

12: the rate at which it is approached.

13: \end{abstract}

14: