IX.D. Classification

IRAS Explanatory Supplement
IX. The Low-Resolution Spectra
D. Classification

Introduction
Classification Scheme
Performance of the Classification Scheme

D.1 Introduction

Visual inspection of spectra shows that they fall naturally into a number of classes. A characterization program was used to return a two-digit code for each spectrum. The first digit characterizes the overall shape of the spectrum (main class) and the second digit gives quantitative information on the dominant feature in the spectrum (subclass).

This classification code appears to be an adequate description of the spectra of more than 95% of the sources in the catalog. Although the classification is based on spectrometer data only, and therefore independent of other astrophysical classification schemes, most classes are dominated by a well-known, well-defined type of sources.

D.2 Classification Scheme

D.2.a Overview

There are five general types of spectra distributed in nine main classes. The five types are: Blue continuous energy distributions, Red continuous energy distributions, spectra with spectral lines including the 11.3 µm line, spectra with lines but not the 11.3 µm line and "others" which fit into none of the preceding classes.

As a first step blue and red spectra were separated using the long-wavelength part of the spectrum. For red sources the flux density per octave (f ) rises from 14 to 25 µm, and for blue spectra it decreases.

The second step was the determination of the relative excesses or deficiencies in narrow bands (0.5 µm wide) centered on the two broad features often seen in the spectra: the 12 µm silicate band and the 12 µm SiC band. The local "continuum" level was estimated by a logarithmic interpolation between narrow bands just outside the features. Also determined were the excess fluxes in 7 narrow bands (0.5 to 1 µm wide) centered on the emission lines at 9.0 µm [Ar III], 10.5 µm [S IV], 11.3 µm (unidentified), 12.8 µm [Ne II], 14.5 µm [Ne V], 15.5 µm [Ne III], and 18.8 µm [S III]. Those were the only lines that were seen in the spectra with the possible exception of the [Ne VI] line at 7.5 µm, just on the edge of the pass-band. The local continuum was estimated by a linear interpolation between two similar bands on both sides of the line.

Table IX.D.1 summarizes the classification scheme described below and lists the number of sources of each type. Figures IX.D.2.1,2 & 3 give examples of some of the different kinds of spectra.

Sources with Blue Energy Distributions: Classes 1n,2n,3n,4n

Sources with blue energy distributions without the 11.3 µm or other narrow spectral lines were classified according to the relative strengths of the 10 and 12 µm emission or absorption features. Spectra with no emission or absorption features at all were assigned to class 1n, where the subclass n is equal to twice the absolute value of the spectral index between 8.0 and 13.0 µm. The spectral index, , is defined according to

(IX.D.1)

Normal K stars have no emission or absorption and show a Rayleigh-Jeans tail at these wavelengths with = -4. Such stars are thus assigned code 18.

Sources with blue energy distributions showing 12 µm silicate band emission were assigned to class 2n, sources with silicate absorption to class 3n. for class 2n the subclass n was defined according to the strength of the emission band with respect to the adjacent continuum following the equation:

n = 10 × [ln f(9.8 µm) - (0.589 ln f(7.9µm) + 0.411 ln f(13.3µm)]

(IX.D.2)

for class 3n, the subclass n was derived on the basis of the strength of the absorption band:

n = -5 × [ln f(9.8µm) - (0.589 ln f(7.9µm) + 0.411 ln f (13.3µm)]

(IX.D.3)

Sources with blue spectra showing 12 µm SiC band emission were assigned to class 4n with the subclass n derived from the strength of the band according to

n = 10 × [ln f(11.4µm) - (0.506 ln f(9.8µm) + 0.494 ln f(13.3µm)]

(IX.D.4)

Figure IX.D.1 The classification scheme for blue spectra.
larger largest

Figure IX.D.1 shows how the classification depends on the ratios of the total flux to the narrow band fluxes at 10.0 and 11.3 µm. In the figure the uncertainties in the positions in the points is less than 0.01, approximately the size of the plotting symbols. The cluster of sources at the center of the figure shows objects with no significant excesses or deficits at either wavelength (Class 1n). The loci of the other blue main classes and subclasses are indicated on the figure.

D.2.c Sources with Red Energy Distributions: Classes 5n, 6n, 7n

Sources with red energy distributions (as defined above), but without the 11.3 µm line or other narrow spectral features, were classified according to the presence or absence of 12 µm silicate emission or absorption. No 12 µm emission or absorption features due to SiC were seen in these red sources.

**Table IX.D.1 Spectral Classification Scheme**
Class	Characteristic	Number	Typical objects
0n	other class -subclasses n=0,2,3,4: see text -subclass n = l: blue, low S/N -subclass n = 5: red, low S/N	363	unknown
1n	blue, featureless -subclass: n=2 times spectral index	2246	stars with spectral type earlier than M5
2n	blue, 12 µm emission -subclass: n= band strength	1738	stars with not too thick oxygen-rich envelopes
3n	blue, 12 µm absorption -subclass: n= band strength	230	stars with thick oxygen-rich envelopes
4n	blue, 12 µm emission-subclass: n= band strength	542	stars with carbon-rich envelopes
5n	red, no line, no 12 µm band -subclass: n = 2 times spectral index	64	unknown
6n	red, 12 µm emission -subclass: n= band strength	78	stars with very thick oxygen-rich envelopes
7n	red, 12 µm absorption -subclass: n= band strength	67	stars with extremely thick oxygen-rich envelopes and hot spots in molecular clouds
8n	11.3µm emission line -subclass n=0: no atomic line -other subclasses: strongest line [Ne II],12.8 µm:n=1 [S III],18.8 µm:n=2 [Ar III],9.0 µm:n=3 [S IV],10.5 µm:n=4 [Ne III],15.5 µm:n=5 [Ne V],14.5 µm:n=6	71	compact HII regions and planetary nebulae
9n	same as class 8, but without 11.3 µm line	50	unknown


Figure IX.D.2.1 larger largest	Figure IX.D.2.2 larger largest	Figure IX.D.2.3 larger largest
Representative low resolution spectra. Numbers refer to the spectral classes defined in the text.

Class 5n objects showed no significant 12 µm excess or deficit. The subclass n is again equal to twice the spectral index between 14 and 25 µm.

Sources with red spectra showing 12 µm silicate band emission were assigned to class 6n; sources with silicate absorption were assigned to class 7n. In both cases the subclass n was derived from the strength of the emission or absorption with respect to the adjacent continuum according to (Eq. IX.D.2,3).

D.2.d Spectra with Spectral Lines:. Classes 8n,9n

For sources with spectral lines the classification proceeded independently of the red or blue shape of the overall energy distribution. The classification processor searched down to the 4

level for the seven spectral lines listed in Section IX.D.2.a. If the 11.3 µm line was among those found, then the source was assigned to class 8n. If the 11.3 µm line was not present, then the source was assigned to class 9n. The subclass n was based on which of the six remaining lines, in order of increasing excitation, was strongest (see Table IX.D.1).

D.2.e Miscellaneous Spectra: Class 0n

Spectra that did not fit into one of the above categories, or which were of lower quality, were assigned to class 0n. Class 01 spectra had such low signal-to-noise ratios that excesses or deficits greater than 0.5 (Eq. IX.D.2,3, 4) were not significant; spectra more than 5 sigma away from the boundaries of the 1n,2n,3n,4n spectra were given codes 00, 02, 03, and 04. These subclasses had only a few sources. Code 05 was given to those objects with low signal-to-noise red spectra without significant lines or bands.

D.3 Performance of the Classification Scheme

About 90% of the sources in the catalog fall in the blue classes 1n to 4n. Figure IX.D.1 shows the adequacy of the classification scheme, with most of the strong sources located in three well-defined areas.

The featureless stars (K and early M spectral types) cluster in the center, and their spectral indices in both the short- and long-wavelength parts of the spectrum are close to -4, as expected.

The class 2n objects (oxygen-rich stars with 12 µm emission bands) lie in a narrow strip indicating that the shape of the 12 µm band is fairly constant. Moreover, the spectral indices increase systematically as the 12 µm band becomes stronger. The class 4n objects (carbon-rich stars with 12 µm emission bands) are well separated from the class 2n objects except in the lowest subclasses, where confusion between the two types occurs.

The class 4n objects occupy a less clearly defined area in Fig. IX.D.1 because of variable structure shortward of the 11 µm band. A few objects with code 04 and in the class 4n seem to be oxygen-rich stars where the 12 µm band is self-absorbed. For these sources, simple classification schemes break down.

The class 3n sources (oxygen-rich stars with 12 µm absorption) suffers from the problem that the 12 µm band becomes so broad that it covers most of the short-wavelength half of the spectrum, leaving no part of the spectrum for an estimate of the continuum. The class 3n crosses show a large scatter in Fig. IX.D.1.

The red classification was tuned to detect as many emission-line sources as possible, and very liberal criteria were applied for significance. Consequently, the automatic classification program contaminates classes 8n and 9n with objects that belong to other categories. Deep 12 µm absorption bands combined with the instrumental short-wavelength band edge near 13.5 µm often produced spurious 12.8 µm lines; 10 and 12 µm emission bands gave rise to features that resembled 11.3 µm lines. All line spectra were inspected by eye and the classification was changed where needed.

For the strongest 300 sources, the individual spectra contributing to the coverage were classified independently by visual inspection. Apart from the expected problems around the boundaries between the various classes, the classifications were consistent within one subclass.

Another test of the classification scheme was done on the line sources. Spectra were extracted for more than a hundred planetary nebulae. Roughly twenty had acceptable line spectra according to the criteria given in Section IX.C.4. The classification processor correctly described those sources.

IRAS Explanatory Supplement IX. The Low-Resolution Spectra D. Classification