The problem of missing atoms has been bothering me for a year. I need some help. Thank you.
Excluding other causes such as model errors, I believe that the root of the problem lies in the potential parameters.
The data file is in the comments section. Both methods share one.
Here are my two proposals:
Method 1
Just use buckhaming. Except for hydrogen and oxygen, the other potential functions come from the same literature.The literature on the potential function of hydrogen and oxygen contains silicon hydroxide, except hydrogen and oxygen, the other potential functions are the same as the other literature, so I choose to use together. This is my in file code:

log                         log.txt
units               metal
dimension           3
boundary                    p p p
atom_style          charge

read_data               J0.66.data

# atom charges
set type 1 charge 1.2 # Ca
set type 2 charge 2.4 # Si
set type 3 charge 0.6 # Na
set type 4 charge -0.6 # F
set type 5 charge 0.256 # H
set type 6 charge -1.2 #O


group Ca type 1
group Si type 2
group Na type 3
group F type 4
group H type 5
group O type 6

          
##################################################################
pair_style          buck/coul/long 10.0 10.0          
kspace_style        ewald 1.0e-5

pair_coeff      * * 0.00000000 1.000 0.00000000 # others
pair_coeff      1 6  7747.1834 0.252623 93.109
pair_coeff      2 6  13702.905 0.193817 54.681
pair_coeff      3 6  4383.75555 0.243838 30.7
pair_coeff      4 6  1863.6049 0.328812 141.27 
pair_coeff      5 6  100 0.25 0
pair_coeff      6 6  1844.7458 0.343645 192.58 


neighbor                       2.0 bin
neigh_modify        every 1 delay 0 check yes
timestep            0.001       ## 1.0 fs, metal time unit is ps
##################################################################
# outputs
thermo 100
thermo_style custom step temp pe press lx density

# mixing2
minimize 1.0e-8 1.0e-8 10000 10000
reset_timestep  0
velocity all create 3000.0 5812775 dist gaussian         #initial temp 3000k
fix 1 all nvt temp 3000 3000 0.001
run 400000                                                                   #400ps

The error is ERROR: Lost atoms: original 1080 current 1060 (src/thermo.cpp:494)
Last command: run 400000. The error screen is in the comment area

Method 2
Use the mixed potential of buckhaming and Morse. Both methods report atom loss, the same error.

This is my in file code:

log                         log.txt
units               metal
dimension           3
boundary                    p p p
atom_style          charge

read_data               J0.66.data

# atom charges
set type 1 charge 1.2 # Ca
set type 2 charge 2.4 # Si
set type 3 charge 0.6 # Na
set type 4 charge -0.6 # F
set type 5 charge 0.256 # H
set type 6 charge -1.2 #O


group Ca type 1
group Si type 2
group Na type 3
group F type 4
group H type 5
group O type 6

          
##################################################################
kspace_style        ewald 1.0e-5
pair_style          hybrid/overlay  morse 10.0 buck/coul/long 10.0 10.0

pair_coeff      * * morse 1.0 1.0 1.0
pair_coeff      5 6  morse 7.0525 0.19 0.9485
pair_coeff      * *   buck/coul/long 1.0 1.0 1.0
pair_coeff      1 6  buck/coul/long 7747.1834 0.252623 93.109
pair_coeff      2 6  buck/coul/long 13702.905 0.193817 54.681
pair_coeff      3 6  buck/coul/long 4383.75555 0.243838 30.7
pair_coeff      4 6  buck/coul/long 1863.6049 0.328812 141.27
pair_coeff      6 6  buck/coul/long  1844.7458 0.343645 192.58



neighbor                       2.0 bin
neigh_modify        every 1 delay 0 check yes
timestep            0.001       ## 1.0 fs, metal time unit is ps
##################################################################
# outputs
thermo 100
thermo_style custom step temp pe press lx density

# mixing2
minimize 1.0e-8 1.0e-8 10000 10000
reset_timestep  0
velocity all create 3000.0 5812775 dist gaussian         #initial temp 3000k
fix 1 all nvt temp 3000 3000 0.001
run 400000                                                                   #400ps

The root of the problem is that for the buckhaming potential of hydrogen and oxygen, only one 100 0.25 0 is found.I found three sets of buckhaming potentials without hydrogen, and they all tried to be used with hydrogen, but they all didn’t fit, and they all lost atoms.
1.I want to get a buckhaming potential parameters that also contains the HOCaSiNaF element.
2.Or I want to get a potential parameters that allows my in file to run without losing atoms, LJ Morse, buckhaming.
3.Or I want to get a mixed potential that allows my in file to run without losing atoms.
In a word, I want to find a potential function that fits my system.
24 hours online waiting for a savior. Thank you!