<div dir="ltr">Hi, all, Recently I am trying to utilize AIMD in CP2K to equilibrate liquid water and water clusters for further analysis. According to the literature, after classical MD equilibration, 2 ps in AIMD would be long enough to equilibrate the system under the NVT condition with Nose-Hoover thermostat (see Fig.1. Input). But under my test, for liquid water (32 water molecules composing cluster with PBC, TIP5P equilibrated), even after 5ps, the temperature still changes significantly (Fig.1.). So I change the input parameters (including decreasing time steps from 1fs to 0.5 fs, decreasing the inner scf max cycles while increasing the outer scf cycles, changing DFT from BLYP to PBE, increasing EPS_SCF from 1E-5 to 1E-6, changing thermostat from Nose to GLE) with the same input configuration, it turns out that the most influencing parameter for temperature equilibration is thermostat (see Fig.2. for GLE results). It seems like that GLE could equilibrate the liquid water faster. But why isn’t Nose-Hoover thermostat working in my test? A lot of literature use Nose-Hoover thermostat for equilibration and NVE for production. For simulating SO2 interacting with small water cluster surface to test, the equilibration time is much longer as the temperature oscillates significantly (Fig.3.) even though it is after almost 18 ps equilibration! The small water cluster is composed of 24 water molecules and equilibrated under TIP3P for 1ns, then the SO2 molecule is put around it, then do the geometry optimization with BLYP-D3 till convergence as the initial configuration for AIMD. So I am curious about the same question as that in simulating the liquid water: why Nose-Hoover thermostat cannot equilibrate my system in several ps as indicated in the literature? The faster equilibration under GLE thermostat rules out the issue of initial configuration, so are there some parameters wrongly setting? Thanks in advance.</div>