6. Output Files Format¶
RANK
means the rank of process (replica) (0, 1, ...
).
6.1. RANK/structure.XXX.vasp
¶
The atomic coordinates for each step are saved in the POSCAR file format of VASP.
XXX
in the filename corresponds to the index of the step.
Example:
Mg8 Al16 O32
1.0
8.113600 0.000000 0.000000
0.000000 8.113600 0.000000
0.000000 0.000000 8.113600
Al Mg O
16 8 32
direct
0.011208 0.995214 0.998158 Al
0.758187 0.240787 0.499981 Al
... skipped ...
0.746308 0.744706 0.233021 O
0.257199 0.255424 0.771040 O
6.2. RANK/minE.vasp
¶
The lowest-energy structure among the samples in this replica.
6.3. RANK/obs.dat
¶
The temperature and the total energy for each step in units of eV.
Example:
0 0.1034076 -41690.28269769395
1 0.1034076 -41692.06763035158
2 0.1034076 -41692.06763035158
3 0.1034076 -41691.98205990787
4 0.1034076 -41692.74143710456
6.4. RANK/obs_save.npy
¶
The total energy for each step in units of eV in the Numpy binary format.
Users can load it as darray
by using numpy.load('obs_save.npy')
.
Example:
$ python -c "import numpy; print(numpy.load('obs_save.npy'))"
[[-41690.28269769]
[-41692.06763035]
[-41692.06763035]
[-41691.98205991]
[-41692.7414371 ]]
6.5. RANK/kT_hist.npy
¶
The temperature for each step in units of eV in the Numpy binary format.
Users can load it as darray
by using numpy.load('kT_hist.npy')
.
Example:
$ python -c "import numpy; print(numpy.load('kT_hist.npy'))"
[0.1034076 0.1034076 0.1034076 0.1034076 0.1034076]
6.6. RANK/Trank_hist.npy
¶
(ONLY for RXMC)
The rank (index) of the temperature for each step in the Numpy binary format.
Users can load it as darray
by using numpy.load('Trank_hist.npy')
.
Example:
$ python -c "import numpy; print(numpy.load('Trank_hist.npy'))"
[1 1 1 1 1]
6.7. RANK/logweight_hist.npy
¶
(ONLY for PAMC) The logarithm of the Neal-Jarzynski weight for each step in the Numpy binary format.
Example:
$ python -c "import numpy; print(numpy.load('logweight_hist.npy'))"
[0 0 0 0 0]
6.8. RANK/acceptance_ratio.dat
¶
Acceptance ratio of Monte Carlo steps for each temperature. The first column is temperature and the second column is acceptance ratio (number of accepted / number of trials).
6.9. logZ.dat
¶
The logarithm of the partition function, \(\log Z_i/Z_0\) where \(i\) is the index of temperature.
The 1st column is temperature \(T_i\).
The 2nd and 3rd columns are \(\log Z_i/Z_0\) and its error.
The 4th and 5th columms are \(\log Z_i/Z_{i-1}\) and its error.
6.10. <name>.dat
¶
Canonical expectation value \(\langle O \rangle\) and statistical error \(\sigma[O]\) of an observable \(O\) for each temperature.
<name>
is the name of the observable which is specified by name
keyword in [[observer.solver]]
section of the input file.
The 1st column is temperature \(T_i\).
The 2nd and 3rd columns are \(\langle O \rangle\) and its error.
The 4th and 5th columns are \(\langle O^2 \rangle\) and its error.
The 6th and 7th columns are fluctuation, \(\langle O^2 \rangle - \langle O \rangle^2\) and its error.
Note that the heat capacity \(C\) is related to the fluctuation of energy as \(k_B T^2 C = \left[ \langle E^2 \rangle - \langle E \rangle^2 \right]\).