Appendix A: Classification scheme of the sources with IRAS counterparts

Palagi et al. (1993) describe a procedure which classifies the H₂O maser sources into two classes: those arising in star forming regions (SFR) and those found around late-type stars (STAR), according to the four FIR fluxes of the IRAS point source associated to the maser.

The classification criteria are derived applying a multivariate analysis (principal component analysis and linear discriminant analysis, Murtagh & Heck 1987) to the subsample of the Arcetri H₂O maser catalog that satisfies the following conditions:

 

 

Table A.1: Boundary definition of SFR and STAR regions

	STAR	SFR
u1	+0.1 to +1.4	-2.4 to -0.3
u2	-1.8 to +0.8	-1.2 to +1.2
u3	-1.4 to +0.5	-1.0 to +1.3
u4	+0.0 to +1.2	-2.0 to -0.1
u5	-0.2 to +1.0	-3.6 to -0.4

a)

the H₂O maser source has a good morphological classification as SFR or STAR.

b)

the associated IRAS point source has no upper limit in any of the four bands.

The results of the multivariate analysis can be summarized as follows:

• the principal component analysis provides three new observables u₁ (Eq. A1), u₂ (Eq. A2) and u₃ (Eq. A3) as combinations of the logarithms of the IRAS fluxes, whose distributions show that the subsample is composed of two main populations well correlated with the morphological classification of the H₂O masers;
• the linear discriminant analysis provides a fourth combination of the IRAS colors, u₄ (Eq. A4), which maximizes the separation of the two populations;
• since a relevant fraction of the IRAS sources associated with the H₂O masers have an upper limit at $\lambda=100~\mu$m we derived a combination of IRAS colors, $u_5 = u_4 \,-\,0.480 \cdot u_2$ (Eq. A5), that is independent of the $100~\mu$m flux, but preserves most of the discriminating properties of u₄;
• from the distributions of each IRAS color combination we derive the boundaries of the STAR and SFR classes, as reported in Table A.1.

 

 

Table A.2: Classification criteria

u4	Color quality			u5
	Good	Upper	Lower
u4<-2.0	STRN	STRN	UNKN	u5<-3.6
-2.0<u4<-0.1	SFR	SFR	UNKN	-3.6<u5<-0.4
-0.1<u4<0.0	UNKN	UNKN	UNKN	-0.4<u5<-0.2
0.0<u4<+1.2	STAR	UNKN	STAR	-0.2<u5<+1.0
u4>+1.2	STRN	UNKN	STRN	u5>+1.0

The five combinations of the logarithms of the FIR fluxes determined through the multivariate analysis are:

u₁ = +0.510 [f₁₂] -0.137 [f₂₅]

+0.403 [f₆₀] -0.776 [f₁₀₀] (1)

u₂ = -0.533 [f₁₂] +0.522 [f₂₅]

+0.857 [f₆₀] -0.846 [f₁₀₀] (2)

u₃ = +0.473 [f₁₂] -1.057 [f₂₅]

+1.244 [f₆₀] -0.660 [f₁₀₀] (3)

u₄ = +0.226 [f₁₂] +0.501 [f₂₅]

-0.324 [f₆₀] -0.403 [f₁₀₀] (4)

u₅ = +0.482 [f₁₂] +0.250 [f₂₅]

-0.735 [f₆₀] (5)

where $[f_u]=\log_{10}(f_u)$, with $u=12,\,25,\,60,\,100$. Based on the criteria determined from this subset of the Arcetri Catalog, for the remainder of the sources the classification proceeds as follows (see Fig. A.1 for the flow-chart representation):

Figure A.1: Flowchart of the classification procedure

Figure A.2: Spectra of all the detections listed in Table 1. Each source is identified by the scan number (upper left) and source name (upper right). The spectra are ordered in right ascension from top left to bottom right

Figure A.2: continued

Figure A.2: continued

Figure A.2: continued

Figure A.2: continued

1.

If there are no upper limits in flux density, u₄ is considered a Good value and is compared with the class boundaries of Table A.1. The outcome is given by the Good column of Table A.2, where UNKN stands for unknown and STRN stands for strange;

2.

If the combination of upper limits in the FIR fluxes is such that u₄ can be assigned an Upper or Lower limit, then the corresponding column of Table A.2 gives the outcome of the classification;

3.

If the outcome of the classification based on u₄ is UNKN, u₅ is computed and the procedure is iterated once more starting from point 2 and using the appropriate class boundaries;

4.

In any case, when the outcome is either SFR or STAR, it is checked against u₂ and u₃ for consistency. If the check is not passed, the source is classified as STRN.