Convergence Problem

[potti-charles]

Dear Zhaoming,

Thanks for sharing this usefel tool.

I got one problem when I calculated the copper surface. The NELECT kept changing but not converged for long run.

This is the tmp-log-FCP.log

<style> </style>

loop	NELECT	Fermi(eV)	Fermishift(eV)	mu(eV)	Ucal(V)	conv(V)	Ewithoutentropy(eV)	Ewithoutentropy_grand(eV)	Etoten(eV)	Etoten_grand(eV)	Cpersurf(e/V/A^2)	time(s)
1	416.5	-7.58968	2.353605	-5.4	0.636077	0.163923	-100.321	-212.286	-111.965	0.0125	2709
2	416.3957	-7.70972	2.353665	-5.4	0.756054	0.043946	-123.691	-259.835	-136.143	0.0125	3080
3	416.3574	-7.44744	2.353669	-5.4	0.493771	0.306229	-115.005	-242.759	-127.755	0.017079	3120
4	416.0911	-7.62028	2.353673	-5.4	0.666611	0.133389	-113.434	-241.69	-128.256	0.017079	3096
5	412.4948	-9.58383	2.353642	-5.4	2.630192	-1.83019	-76.4291	-119.114	-76.4528	-119.138	0.529477	3172
6	416.0491	-7.72991	2.353669	-5.4	0.776243	0.023757	-108.167	-231.476	-123.31	0.038139	3094
7	416.003	-7.79553	2.353669	-5.4	0.841859	-0.04186	-111.813	-239.125	-127.313	0.038108	3088
8	416.0844	-7.57615	2.353605	-5.4	0.622548	0.177452	-113.066	-241.007	-127.94	0.038167	3183
9	415.7423	-8.00652	2.353605	-5.4	1.052917	-0.25292	-113.793	-245.112	-131.319	0.037864	3086
10	416.2299	-7.43499	2.353605	-5.4	0.481385	0.318615	-114.495	-242.731	-128.237	0.037861	3095
11	415.1292	-8.36489	2.353605	-5.4	1.41128	-0.61128	-105.62	-233.522	-127.902	0.067844	3080

we can see that the Ucal changes, but not converged to the set value of 0.8 V. How can I fix it?

P.S. This is my script

from ase.calculators.FCPelectrochem import FCP

from ase.calculators.vasp import Vasp
from ase.io import read
from ase.optimize import LBFGS

cal_sol=Vasp(xc='PBE', #functional
pp='PBE', #type of pseudopotential
kpts=(3, 3, 1), #kpoint
ncore=4,
ispin=2,lasph=True,ismear=0, sigma=0.1, algo='Fast', ediff=1E-5, prec='Accurate', encut=400, nelm=500 , addgrid='Ture',lreal='Auto',lorbit=11, ldau_luj={'Cu': {'L': 2, 'U': 5.0, 'J': 0.46}}, lmaxmix=4, #parameters for SCF
tau=0, lrhoion=False, lsol=True, eb_k=78.4, lambda_d_k=3.0, #parameters for vaspsol
lwave=True, lcharg = False, #write WAVECAR to speed up the SCF of the next ionic step
)
cal_sol.set(label='sol', directory='sol')

cal_FP=FCP(innercalc=cal_sol,
fcptxt='log-fcp.txt',
U=0.8,
NELECT = 416.5,
C = 1/80, #1/k capacitance per A^2
FCPmethod = 'Newton-fitting',
FCPconv=0.01,
NELECT0=418,
adaptive_lr=False,
work_ref=4.6,
max_FCP_iter=10000
)

atoms=read('POSCAR')
atoms.calc=cal_FP
dyn=LBFGS(atoms, trajectory='fp.traj')
dyn.run(fmax=0.01)

This is my POSCAR
CONTCAR
1.0
6.640630 -3.833971 0.000000
0.000000 7.667940 0.000000
0.000000 0.000000 16.260851
H C O Cu Ag
2 2 2 33 3
Selective Dynamics
Direct
0.835122 0.729924 0.606477 T T T
0.192244 -0.032920 0.602237 T T T
0.829531 0.725088 0.489406 T T T
0.033103 0.896433 0.497212 T T T
0.736450 0.649656 0.563126 T T T
0.080419 0.950958 0.566204 T T T
0.946910 0.978576 0.008843 F F F
0.940488 0.987888 0.393301 T T T
0.167276 0.088266 0.264884 T T T
0.058644 0.201401 0.134865 F F F
0.280228 0.978880 0.009939 F F F
0.258950 0.972238 0.401684 T T T
0.497780 0.089135 0.264434 T T T
0.390635 0.201232 0.135802 F F F
0.613327 0.978685 0.008384 F F F
0.617769 0.980433 0.391651 T T T
0.831692 0.087360 0.261974 T T T
0.724487 0.200199 0.133988 F F F
0.947145 0.312759 0.009600 F F F
0.944833 0.311880 0.386815 T T T
0.169872 0.425090 0.263198 T T T
0.058967 0.534527 0.134288 F F F
0.280658 0.312533 0.010239 F F F
0.502950 0.423431 0.267105 T T T
0.391766 0.534532 0.134958 F F F
0.613326 0.312687 0.009926 F F F
0.831455 0.421761 0.262207 T T T
0.724174 0.534967 0.133898 F F F
0.946623 0.645636 0.009342 F F F
0.939715 0.635500 0.391270 T T T
0.166353 0.756119 0.265096 T T T
0.057968 0.869642 0.134726 F F F
0.280875 0.645952 0.009577 F F F
0.501393 0.753946 0.265366 T T T
0.392576 0.869471 0.135091 F F F
0.613685 0.645050 0.008338 F F F
0.629998 0.658682 0.396681 T T T
0.835556 0.757300 0.261589 T T T
0.725275 0.868654 0.133221 F F F
0.616943 0.306214 0.418494 T T T
0.281567 0.655506 0.412118 T T T
0.276890 0.314533 0.406836 T T T

[IMPLICIT-NONE]

Hi, this issue may be caused by insufficient kpoints resulting in the Fermi energy levels constantly jumping between two k-points with large energy level differences during the calculation. Try using a larger kpoint, or a larger SIGMA value.