Plot convergence

This example shows how to plot the convergence of scalar quantities wrt to one independent variable. For instance, the convergence of the energy per atom with the cutoff energy ecut.

Let’s assume we have the total energy per atom and the pressure in GPa as a function of ecut stored in three python lists:

ecut_list = [35., 40., 45., 50., 55., 60., 65.]

energy_per_atom_ev = [
-444.29611271, -444.3113239, -444.3124781,
-444.31251066, -444.31254606, -444.31256315, -444.31258609,
]

pressure_gpa =  [
-17.56316746, -16.00634717, -15.80295881,
-15.79975677, -15.79858973, -15.79811464, -15.79929941,
]

To plot energy vs ecut with a convergence window of 1e-3 centered around our best estimate i.e. the value obtained with the highest ecut, use the high-level API from_xy_label_vals:

from abipy.tools.plotting import ConvergenceAnalyzer

ca = ConvergenceAnalyzer.from_xy_label_vals("ecut (Ha)", ecut_list,
                                            "E/natom (eV)", energy_per_atom_ev, tols=1e-3)

print(ca)
ca.plot()
x: 41.1 for $\Delta$: 0.001
Number of points for x-axis: 7
ykey: E/natom (eV)
    ytol  ix         xx          ykey
0  0.001   2  41.135808  E/natom (eV)

To analyze the convergence of two or more quantities, use the low-level API that expects two dictionaries mapping the y-label used in the plot to (1) the list of values and (2) the associated absolute tolerance as in:

yvals_dict = {
    "E/natom (eV)": energy_per_atom_ev,
    "P (GPa)": pressure_gpa,
}

ytols_dict = {
    "E/natom (eV)": 1e-3,
    "P (GPa)": 1e-1,
}

ca = ConvergenceAnalyzer("ecut (Ha)", ecut_list, yvals_dict, ytols_dict)

#print(ca)
#ca.plot()

To specify multiple convergence criteria, replace scalars with tuples as in:

ytols_dict = {
    "E/natom (eV)": (1e-3, 1e-4),
    "P (GPa)": (1e-1, 1e-2),
}

ca = ConvergenceAnalyzer("ecut (Ha)", ecut_list, yvals_dict, ytols_dict)
#print(ca)
ca.plot()
x: 41.1 for $\Delta$: 0.001, x: 46.2 for $\Delta$: 0.0001, x: 42.6 for $\Delta$: 0.1, x: 44.8 for $\Delta$: 0.01

Finally, if you have a (xlsf or csv) file with columns named e.g.: ecut, energy_per_atom, pressure and you want to plot the results with ConvergenceAnalyzer, use the from_file method with a ytols_dict whose keys define the columns in tabular data:

ytols_dict = {
    "ecut": (1e-3, 1e-4),
    "pressure": (1e-1, 1e-2),
}
#ca = ConvergenceAnalyzer.from_file(filepath, ytols_dict)

# To change the x/y-labels in the plot, use:
#ca.set_label("ecut", "ecut (Ha)")
#ca.set_label("energy_per_atom", "E/natom (eV)")
#ca.set_label("pressure", "P (GPa)")

#print(ca)
#ca.plot()

Total running time of the script: (0 minutes 0.805 seconds)

Gallery generated by Sphinx-Gallery