<div dir="ltr">Dear CP2K group, before the MD simulation of the interfacial arrangement of water molecules near the ferroelectric lithium niobate single crystal (here the positive polar surface site), I had performed the geometry optimization of the LiNbO3 slab (consisting of 8 unit cells) with fixing all atoms of the lower unit cells to get reconstructed LiNbO3 surface (Nb-O2 surface termination). (the comparison between surface reconstruction with only the lowest atom layers fixed and that with all atoms of lower unit cells fixed shows no remarkable difference, only small energy difference of about 0.014 a.u.) For the next step I build a large system consisting of 24 LiNbO3 unit cells (6x6x2) and 478 water molecules (solvated by VMD package solvate 1.5) in the volume cell box of size 30.888 x 26.7498 x 100.0 Angstrom^3. Like the geometry optimization above I set all atoms of the lower unit cells of LiNbO3 fixed and I want to start the geometry optimization of the large system (of about 3600 atoms), but short time after submitting the simulation files I got the error message that the required memory exceeds the memory size available (of a small part of super computer system at our institute). After searching for the alternate way I found some interesting articles about new linear scaling methods: a) J. Chem. Theory Comput. 9 (2013), 4421 [http://pubs.acs.org/doi/abs/10.1021/ct400595k] b) J. Chem. Theory Comput. 8 (2012), 3565 [http://pubs.acs.org/doi/abs/10.1021/ct200897x] Now I would like to rewrite the input code but because I am beginner and I don't really know how to write correct codes into the input file for geometry optimization and then MD simulation (one input file for geom.-opt. and one for MD sim.) on the basis of the linear scaling SCF method. Is it possible to write an efficient method for DFT calculation on a part of super computer system in the order of 10^2 cores? As additional information of the energies after geometry optimization of the LiNbO3 slabs an additional file is attached here also. For a better overview of the input code (this is for geometry optimization for LiNbO3 slab) and of the large system the files are attached here. Kind regards, Phil </div>