Meudon PDR code emulator predictions

This notebook shows how to use the MeudonPDR class, which provides a fast, memory-light and accurate emulator of the Meudon PDR code.

[1]:

import os
import sys

import matplotlib.pyplot as plt
import pandas as pd

sys.path.insert(0, os.path.join(".."))

from infobs.model import MeudonPDR
from infobs.graphics import PDRPlotter

%load_ext autoreload
%autoreload 2

Predictions from a grid of parameters

[2]:

lines = ["13c_o_j1__j0", "13c_o_j2__j1", "13c_o_j3__j2"]

Grid setup

[3]:

df_params = pd.DataFrame(
    {
        "Av": [1, 5, 10, 15, 20],
        "G0": [1e2, 1e2, 1e2, 1e2, 1e2],
        "Pth": [1e5, 1e5, 1e5, 1e5, 1e5],
        "angle": [0, 0, 0, 0, 0],
    }
)

df_params

[3]:
Av G0 Pth angle
0 1 100.0 100000.0 0
1 5 100.0 100000.0 0
2 10 100.0 100000.0 0
3 15 100.0 100000.0 0
4 20 100.0 100000.0 0

Predictions of integrated line intensities

[4]:

meudonpdr = MeudonPDR()
meudonpdr.predict(df_params, lines)

[4]:
Av G0 Pth angle kappa 13c_o_j1__j0 13c_o_j2__j1 13c_o_j3__j2
0 1.0 100.0 100000.0 0.0 1.0 0.000276 0.000318 0.000128
1 5.0 100.0 100000.0 0.0 1.0 10.176692 11.708745 3.919423
2 10.0 100.0 100000.0 0.0 1.0 22.873276 21.915024 8.102161
3 15.0 100.0 100000.0 0.0 1.0 29.498773 25.964494 10.321277
4 20.0 100.0 100000.0 0.0 1.0 34.912525 29.605632 12.635616

Integrated intensity unit

Set the kelvin parameter to False if you want integrated line intensities in erg cm-2 s-1 sr-1

[5]:

meudonpdr = MeudonPDR(kelvin=False)
meudonpdr.predict(df_params, lines)

[5]:
Av G0 Pth angle kappa 13c_o_j1__j0 13c_o_j2__j1 13c_o_j3__j2
0 1.0 100.0 100000.0 0.0 1.0 3.785540e-13 3.483762e-12 4.746177e-12
1 5.0 100.0 100000.0 0.0 1.0 1.395919e-08 1.284490e-07 1.451301e-07
2 10.0 100.0 100000.0 0.0 1.0 3.137487e-08 2.404155e-07 3.000103e-07
3 15.0 100.0 100000.0 0.0 1.0 4.046295e-08 2.848396e-07 3.821807e-07
4 20.0 100.0 100000.0 0.0 1.0 4.788890e-08 3.247841e-07 4.678770e-07

Adding a scaling parameter to integrate uncertainty on geometry and beam dilution

In this example, we assume the logarithm of the scaling parameter \(\kappa\) to be uniformly distributed:

\[\log_{10}\kappa \in \left[-\frac12, \frac12\right]\]

When not specified, the kappa parameter is constant equal to 1 and therefore has no impact.

While the predictions of the PDR code emulator are deterministic, adding this parameter to the model introduces randomness. This parameter is sampled from a Sampler, see the samplers.ipynb notebook for more details.

[6]:

from infobs.sampling import LogUniform

kappa = LogUniform(10**-0.5, 10**0.5)

meudonpdr.predict(df_params, lines, kappa=kappa)

[6]:
Av G0 Pth angle kappa 13c_o_j1__j0 13c_o_j2__j1 13c_o_j3__j2
0 1.0 100.0 100000.0 0.0 0.362598 1.372631e-13 1.263207e-12 1.720956e-12
1 5.0 100.0 100000.0 0.0 1.245769 1.738993e-08 1.600178e-07 1.807985e-07
2 10.0 100.0 100000.0 0.0 1.030183 3.232187e-08 2.476720e-07 3.090656e-07
3 15.0 100.0 100000.0 0.0 2.247830 9.095385e-08 6.402709e-07 8.590772e-07
4 20.0 100.0 100000.0 0.0 0.408667 1.957059e-08 1.327284e-07 1.912057e-07

Prediction plots

[7]:

plotter = PDRPlotter(kelvin=True)
plotter.print_parameters_space()

Parameters space
Av: [1, 40] mag
G0: [0.6393, 63930.0]
Pth: [100000.0, 1000000000.0] K.cm-3
angle: [0, 60] deg

Profiles

[8]:

lines = ["13c_o_j1__j0", "13c_o_j2__j1", "13c_o_j3__j2"]

Av = 6
G0 = 1e2
Pth = 1e5

[9]:

plt.figure(figsize=(6.4, 0.7 * 4.8), dpi=125)

plotter.plot_profile(
    lines,
    Av=None,
    G0=1e2,
    Pth=1e5,
    angle=0,
    logy=False,
)

plt.show()
_images/meudonpdr_17_0.png 
[10]:

plt.figure(figsize=(6.4, 0.7 * 4.8), dpi=125)

plotter.plot_profile(
    lines,
    Av=None,
    G0=1e2,
    Pth=1e5,
    angle=0,
    logy=True,
)

plt.show()
_images/meudonpdr_18_0.png

Slice (color plot)

[11]:

line = "13c_o_j1__j0"

[12]:

plt.figure(figsize=(6.4, 0.7 * 4.8), dpi=125)

plotter.plot_slice(
    line,
    Av=None,
    G0=None,
    Pth=1e5,
    angle=0,
    logz=False,
)

plt.show()
_images/meudonpdr_21_0.png 
[13]:

plt.figure(figsize=(6.4, 0.7 * 4.8), dpi=125)

plotter.plot_slice(
    line,
    Av=None,
    G0=None,
    Pth=1e5,
    angle=0,
    logz=True,
)

plt.show()
_images/meudonpdr_22_0.png

Slice (contour plot)

[14]:

line = "13c_o_j1__j0"

[15]:

# TODO
# plt.figure(figsize=(6.4, 0.7*4.8), dpi=125)

# plotter.plot_slice(
#     line,
#     Av=None,
#     G0=None,
#     Pth=1e5,
#     angle=0,
#     logz=True,
#     contour=True
# )

# plt.show()

Plots from CSV

It is also possible to automatically plot and save a series of figures (profiles or slices) based on instructions contained in a .csv file.

Profiles

[16]:

csv_file = os.path.join("meudonpdr", "profiles.csv")
path_outputs = os.path.join("meudonpdr", "profiles")
contour = False

plotter.save_profiles_from_csv(csv_file, path_outputs, legend=True, latex=True, dpi=150)

Slices

[17]:

csv_file = os.path.join("meudonpdr", "slices.csv")
path_outputs = os.path.join("meudonpdr", "slices")
contour = False

plotter.save_slices_from_csv(
    csv_file, path_outputs, contour=contour, legend=True, latex=True, dpi=150
)