Dissecting the molecular structure of the Orion B cloud: Insight from Principal Component Analysis

TL;DR

This study applies Principal Component Analysis to widefield spectral data of the Orion B cloud, revealing key physical parameters like density and radiation field without complex modeling.

Contribution

It introduces a statistical approach combining data rescaling and PCA to analyze interstellar medium properties from spectral line data.

Findings

Identified 3 principal components correlating with physical parameters.

Demonstrated PCA's effectiveness in decomposing molecular emission.

Provided a new method for analyzing large spectral datasets.

Abstract

Context. The combination of wideband receivers and spectrometers currently available in (sub-)millimeter observatories deliver wide- field hyperspectral imaging of the interstellar medium. Tens of spectral lines can be observed over degree wide fields in about fifty hours. This wealth of data calls for restating the physical questions about the interstellar medium in statistical terms. Aims. We aim at gaining information on the physical structure of the interstellar medium from a statistical analysis of many lines from different species over a large field of view, without requiring detailed radiative transfer or astrochemical modeling. Methods. We coupled a nonlinear rescaling of the data with one of the simplest multivariate analysis methods, namely the Principal Component Analysis, to decompose the observed signal into components that we interpret first qualitatively and then quantitatively based on our deep knowledge of the observed region and of the astrochemistry at play. Results. We identify 3 principal components, linear compositions of line brightness temperatures, that are correlated at various levels with the column density, the volume density and the UV radiation field. Conclusions. When sampling a sufficiently diverse mixture of physical parameters, it is possible to decompose the molecular emission in order to gain physical insight on the observed interstellar medium. This opens a new avenue for future studies of the interstellar medium.

Tables

Table 1. Table 1: Properties of the observed spectral lines. The last six columns show the statistics of the data before and after asinh reparametrization.

				Original data				After asinh reparametrization
Molecule	Transitions	Frequency	Noise	Min.	Median	Max.	Std.	Min.	Median	Max.	Std.
		(MHz)	(K)	(K)	(K)	(K)	(K)	(K)	(K)	(K)	(K)
${}_{}{}^{12}\mathrm{CO}$	$J=1\to 0$	115271.202	$0.09$	$-0.39$	$13.40$	$57.11$	$10.18$	$-0.37$	$2.39$	$3.32$	$0.91$
${}_{}{}^{13}\mathrm{CO}$	$J=1\to 0$	110201.354	$0.04$	$-0.19$	$1.38$	$36.43$	$3.27$	$-0.19$	$0.97$	$3.03$	$0.74$
$\mathrm{CS}$	$J=2\to 1$	97980.953	$0.06$	$-0.36$	$0.06$	$15.53$	$0.48$	$-0.35$	$0.06$	$2.48$	$0.23$
$\mathrm{HCN}$	$J=1\to 0,F=2\to 1$	88631.848	$0.10$	$-0.58$	$0.15$	$10.32$	$0.39$	$-0.52$	$0.15$	$2.22$	$0.25$
${\mathrm{HCO}}^{+}$	$J=1\to 0$	89188.525	$0.09$	$-0.45$	$0.26$	$8.07$	$0.47$	$-0.42$	$0.25$	$2.07$	$0.30$
$\mathrm{SO}$	$N=3\to 2,J=2\to 1$	99299.870	$0.06$	$-0.43$	$0.04$	$6.46$	$0.24$	$-0.40$	$0.04$	$1.92$	$0.17$
$\mathrm{CN}$	$N=1\to 0,J=3/2\to 1/2,F=5/2\to 3/2$	113490.970	$0.09$	$-0.58$	$0.10$	$6.33$	$0.27$	$-0.52$	$0.09$	$1.91$	$0.20$
$\mathrm{HNC}$	$J=1\to 0$	90663.568	$0.08$	$-0.49$	$0.07$	$6.01$	$0.27$	$-0.45$	$0.07$	$1.88$	$0.19$
$\mathrm{CCH}$	$N=1\to 0,J=3/2\to 1/2,F=2\to 1$	87316.898	$0.12$	$-0.62$	$0.08$	$5.72$	$0.22$	$-0.55$	$0.08$	$1.85$	$0.18$
$C^{18}O$	$J=1\to 0$	109782.173	$0.06$	$-0.30$	$0.06$	$5.55$	$0.42$	$-0.29$	$0.06$	$1.83$	$0.26$
$N_2{H}^{+}$	$J=1\to 0,\mathrm{F1}=2\to 1,F=3\to 2$	93173.764	$0.08$	$-0.44$	$0.00$	$4.53$	$0.13$	$-0.41$	$0.00$	$1.70$	$0.10$
${\mathrm{CH}}_3\mathrm{OH}$	$J=2\to 1,K=0\to 0,\left(A+\right)$	96741.375	$0.06$	$-0.34$	$0.01$	$2.24$	$0.08$	$-0.32$	$0.01$	$1.26$	$0.08$

Equations

$$T(x)=a\mathrm{asinh}(x/a),$$

$$\mathrm{hue}=\mathrm{atan}(uv,density)$$

Full text

*Aims. *We aim at gaining information on the physical structure of the interstellar medium from a statistical analysis of many lines from different species over a large field of view, without requiring detailed radiative transfer or astrochemical modeling.

*Methods. *We coupled a nonlinear rescaling of the data with one of the simplest multivariate analysis methods, namely the Principal Component Analysis, to decompose the observed signal into components that we interpret first qualitatively and then quantitatively based on our deep knowledge of the observed region and of the astrochemistry at play.

*Results. *We identify 3 principal components, linear compositions of line brightness temperatures, that are correlated at various levels with the column density, the volume density and the UV radiation field.

*Conclusions. *When sampling a sufficiently diverse mixture of physical parameters, it is possible to decompose the molecular emission in order to gain physical insight on the observed interstellar medium. This opens a new avenue for future studies of the interstellar medium.

1 Introduction