5. Case Studies
5.1. Crystal structure prediction of Li-N-H (fake formula)
(base) ➜ local_slurm_toml_vasp git:(devel) ✗ ls
INCAR_1 INCAR_3 POTCAR_H POTCAR_N machine-1.json ref_ene.txt
INCAR_2 INCAR_4 POTCAR_Li input.toml machine-2.json run.calypso.sh
input.toml example:
[CALYPSO]
SystemName = "Crystal-VASP CALYPSO config"
Seed = -1 # positive int number to set random seed for REPRODUCIBILITY,
# negetive to do not set it.
IType = 1 # or "CRYSTAL".
ICode = 1 # or "VASP".
IAlgo = 2 # or "LPSO".
IFit = 1 # or "ENTHALPY".
IDisp = 1 # or "CORE" means the build-in dispatcher by CALYPSO
# rather than third party libraries.
IRunner = 1 # normal mode of calypso.
ISim = 1 # or "BCM".
BlockMode = true # block mode of calypso.
PickUp = false # pick up the calculation whenever you want.
[CALYPSO.EVO]
NBest = 4
PSOratio = 0.4
PopSize = 10
MaxStep = 2
[CALYPSO.GENERATOR]
FormulaUnit = ["(LiH4)1-2(NH3)1-2",]
VolumeUnit = {Li=10, H=10, N=10}
DistanceOfIon = [["X", "Li", "H", "N"],
["Li", 1.0, 1.0, 1.0],
["H", 1.0, 1.0, 1.0],
["N", 1.0, 1.0, 1.0],]
SpaceGroup = [1, "2-30", 31, 32, "33-50", "51:231:1"]
[CALYPSO.OPT]
DFTInputPath = "."
JobFlow = ["opt", "opt", "opt", "scf"]
PpMap = {Li="POTCAR_Li", N="POTCAR_N", H="POTCAR_H"}
[CALYPSO.DISPATCHER]
# list your resoures here
MachineList = ["./machine-1.json", "./machine-2.json"]
TimeInternal = 3
[CALYPSO.DESCRIPTOR]
SimThreshold = 0.01
There are several ways to control and utilize the computational resources, by defining different combinations of executor and scheduler in machine.json.
5.1.1. 1. Running CALYPSO in cluster with Slurm system in two queue using VASP
machine-1.json:
{
"name": "amd7513",
"host": null,
"port": null,
"username": null,
"password": null,
"key_filename": null,
"numb_cpu_per_node": 64,
"numb_node": 1,
"command": null,
"vasp_command": "mpirun -n 64 /data/software/apps/vasp.6.1.0/bin/vasp_std",
"python_path": null,
"max_run_time": 100,
"executor": "local",
"scheduler": "slurm",
"scheduler_env": [],
"remote_root": null,
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": ["source /data/env/compiler_intel-2021.3.0.sh", "source /data/env/mpi_intelmpi-2021.3.0.sh"],
"module": [],
"additional_head_setting": ["#SBATCH --exclude=node[49-50]"],
"queue": "amd7513"
}
machine-2.json:
{
"name": "amd9654_1t",
"host": null,
"port": null,
"username": null,
"password": null,
"key_filename": null,
"numb_cpu_per_node": 64,
"numb_node": 1,
"command": null,
"vasp_command": "mpirun -n 64 /data/software/apps/vasp.6.1.0/bin/vasp_std",
"python_path": null,
"max_run_time": 100,
"executor": "local",
"scheduler": "slurm",
"scheduler_env": [],
"remote_root": null,
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": ["source /data/env/compiler_intel-2021.3.0.sh", "source /data/env/mpi_intelmpi-2021.3.0.sh"],
"module": [],
"additional_head_setting": ["#SBATCH --exclude=node[49-50]"],
"queue": "amd9654_1t"
}
5.1.2. 2. Running CALYPSO in local laptop or WorkStation using VASP
machine-1.json:
{
"name": "laptop",
"host": null,
"port": null,
"username": null,
"password": null,
"key_filename": null,
"numb_cpu_per_node": null,
"numb_node": null,
"command": null,
"vasp_command": "mpirun -n 64 /data/software/apps/vasp.6.1.0/bin/vasp_std",
"python_path": null,
"max_run_time": 100,
"executor": "local",
"scheduler": "shell",
"scheduler_env": [],
"remote_root": "/home/user/calypso/work_dir",
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": ["source /data/env/compiler_intel-2021.3.0.sh", "source /data/env/mpi_intelmpi-2021.3.0.sh"],
"module": [],
"additional_head_setting": [],
"queue": null
}
5.1.3. 3. Running CALYPSO in local laptop but send tasks into remote WorkStation using VASP
machine-1.json:
{
"name": "WorkStation",
"host": "xxx.xxx.xxx.xxx",
"port": 10086,
"username": "username",
"password": null,
"key_filename": "key_file",
"numb_cpu_per_node": null,
"numb_node": null,
"command": null,
"vasp_command": "mpirun -n 64 /data/software/apps/vasp.6.1.0/bin/vasp_std",
"python_path": null,
"max_run_time": 100,
"executor": "ssh",
"scheduler": "shell",
"scheduler_env": [],
"remote_root": "/home/remote/user/calypso/work_dir",
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": ["source /data/env/compiler_intel-2021.3.0.sh", "source /data/env/mpi_intelmpi-2021.3.0.sh"],
"module": [],
"additional_head_setting": [],
"queue": null
}
5.1.4. 4. Running CALYPSO in local laptop but send tasks into remote WorkStation and remote LSF cluster using VASP
machine-1.json:
{
"name": "WorkStation",
"host": "xxx.xxx.xxx.xxx",
"port": 10086,
"username": "username",
"password": null,
"key_filename": "key_file",
"numb_cpu_per_node": null,
"numb_node": null,
"command": null,
"vasp_command": "mpirun -n 64 /data/software/apps/vasp.6.1.0/bin/vasp_std",
"python_path": null,
"max_run_time": 100,
"executor": "ssh",
"scheduler": "shell",
"scheduler_env": [],
"remote_root": "/home/remote/user/calypso/work_dir",
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": ["source /data/env/compiler_intel-2021.3.0.sh", "source /data/env/mpi_intelmpi-2021.3.0.sh"],
"module": [],
"additional_head_setting": [],
"queue": null
}
machine-2.json:
{
"name": "amd9654_1t",
"host": "xxx.xxx.xxx.xxx",
"port": 10086,
"username": "username",
"password": "xxxxxxxx",
"key_filename": null,
"numb_cpu_per_node": 64,
"numb_node": 1,
"command": null,
"vasp_command": "mpirun -n 64 /data/software/apps/vasp.6.1.0/bin/vasp_std",
"python_path": null,
"max_run_time": 100,
"executor": "ssh",
"scheduler": "lsf",
"scheduler_env": [],
"remote_root": "/remote/path/dir",
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": ["source /data/env/compiler_intel-2021.3.0.sh", "source /data/env/mpi_intelmpi-2021.3.0.sh"],
"module": [],
"additional_head_setting": ["#BSUB --exclude=node[49-50]"],
"queue": "amd9654_1t"
}
5.1.5. 5. Running CALYPSO in local laptop or WorkStation using MLP
Please note that ICode in input.toml must be 15 if one intends to use MLP.
machine-1.json:
{
"name": "laptop",
"host": null,
"port": null,
"username": null,
"password": null,
"key_filename": null,
"numb_cpu_per_node": null,
"numb_node": null,
"command": null,
"vasp_command": null,
"python_path": "/home/wangzy/softwares/miniconda3/envs/chgnet/bin/python",
"max_run_time": 100,
"executor": "local",
"scheduler": "shell",
"scheduler_env": [],
"remote_root": "/home/user/calypso/work_dir",
"machine_capacity": 4,
"group_size": 1,
"envs": [],
"source": [],
"module": [],
"additional_head_setting": [],
"queue": null
}
ref_ene.txt:
formula enthalpy_per_atom label
Li -1 element_Li
N -2 element_N
H -1.5 element_H
run.calypso.sh:
#!/bin/bash
source /data/env/compiler_intel-2021.3.0.sh
source /data/env/mpi_intelmpi-2021.3.0.sh
source /data/home/wangzhenyu/softwares/miniconda3/bin/activate /data/home/wangzhenyu/softwares/miniconda3/envs/calypso
calypso.x > caly.log 2>&1
As you can see, CALYPSO will be submitted in login node by running run.calypso.sh. If you are not allowed to run process in login node, you can submit the run.calypso.sh to Slurm or PBS.
#!/bin/bash
#SBATCH --job-name=VaspTest
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=4
#SBATCH --cpus-per-task=1
#SBATCH --partition=amd9654
source /data/env/compiler_intel-2021.3.0.sh
source /data/env/mpi_intelmpi-2021.3.0.sh
source /data/home/wangzhenyu/softwares/miniconda3/bin/activate /data/home/wangzhenyu/softwares/miniconda3/envs/calypso
calypso.x > caly.log 2>&1
In this way, calypso will run in this compute node, but if your cluster is not allowed you to submit tasks in compute node with sbatch command, you may need to change the scheduler back to shell for normal mode.
5.2. Molecule Crystal Prediction of \(H_2O\)
In this section, we use CALYPSO to predict the structure of \(H_2O\) VIII. The following files are required:
.
├── H2O.xyz
├── input.toml
└── machine-1.json
Molecular Structure Input (XYZ File)
For molecular crystal prediction, an XYZ file is needed to store the molecular structure. Here, we use H2O.xyz to define the H₂O molecule structure:
3
H2O
O 0.0 0.0 0.0
H -0.895696 0.0 -0.454015
H 0.895696 0.0 -0.454015
Task Configuration (input.toml)
Next, we configure the prediction task in input.toml:
[CALYPSO]
# ...
IType = 3 # For molecular crystal prediction
# ...
[CALYPSO.GENERATOR]
FormulaUnit = "{H2O}4" # Curly brackets define a molecule unit
VolumeUnit = {H2O=25}
DistanceOfIon = [["X","H2O"],
["H2O",1.2]]
SpaceGroup = [141,141]
GenerateMethod = "random"
MoleculesPath = {'H2O'='./H2O.xyz'}
LatticeMaxAttempts = 10
OrbitalMaxAttempts = 10
MaxTryPointEachOrb = 10
[CALYPSO.EVO]
NBest = 4
PSOratio = 0 # Must be set to 0, as PSO evaluation is not supported
Explanation of Key Parameters
IType: Must be 3 to run molecular crystal prediction.
Molecular Labeling: The label
H2Orepresents the water molecule.Formula Unit Definition:
FormulaUnit = "{H2O}4"specifies four \(H_2O\) molecules in curly brackets.Volume and Distance Configuration: Using the label, you can set parameters like volume, inter-molecular distance, and molecule paths.
PSO Ratio: PSOratio must be set to 0 because PSO evaluation is not supported.
This setup ensures the correct execution of the molecular crystal prediction for H₂O VIII using CALYPSO.
5.3. Molecule Crystal Prediction of Fe-O-OH
Molecules can also be defined with other elements. In this example, we input OH as a unit along with the elements Fe and O.
[CALYPSO]
# ...
IType = 3 # For molecular crystal prediction
# ...
[CALYPSO.GENERATOR]
FormulaUnit = "(Fe)4(O)4{OH}4"
VolumeUnit = {Fe=15,O=15,OH=15}
DistanceOfIon = [["X", "Fe","O","OH"],
["Fe", 1.2,1.2,1.2],
["O",1.2,1.2,1.2],
["OH",1.2,1.2,1.2]]
SpaceGroup = [62,62]
GenerateMethod = "random"
MoleculesPath = {'OH'='./OH.xyz'}
LatticeMaxAttempts = 10
OrbitalMaxAttempts = 10
MaxTryPointEachOrb = 10
[CALYPSO.EVO]
NBest = 4
PSOratio = 0 # Must be set to 0, as PSO evaluation is not supported
and here is the OH.xyz file:
2
HO
O 0 0 0
H 0 0 0.9