The spectral characteristics of these stars are quite distinctive, and they were first recognized by their spectra by Angelo Secchi in the 1860s — a pioneering time in astronomical spectroscopy. In "normal" stars (such as the Sun), the atmosphere is richer in oxygen than carbon.
In the classical carbon stars, the abundance of carbon is thought to be a product of helium fusion, specifically the triple-alpha process within a star, which giants reach near the end of their lives in the so called Asymptotic Giant Branch (AGB). These fusion products have been brought to the stellar surface by episodes of convection after the carbon and other products were made. Normally this kind of AGB carbon star fuses hydrogen in a hydrogen burning shell, but in episodes separated by 104-105 years, the star transforms to burning helium in a shell, while the hydrogen fusion temporarily ceases. In this phase, the star's luminosity rises, and material from the interior of the star (notably carbon) moves up. Since the luminosity rises, the star expands so that the helium fusion ceases, and the hydrogen shell burning restarts. During these shell helium flashes, the mass loss from the star is significant, and after many shell helium flashes, an AGB star is transformed into a hot white dwarf and its atmosphere becomes material for a planetary nebula.
The non-classical kinds of carbon stars are believed to be binary stars, where one star is observed to be a giant star (or occasionally a red dwarf) and the other a white dwarf. The star presently observed to be a giant star accreted carbon-rich material when it was still a main sequence star from its companion (that is, the star that is now the white dwarf) when the latter was still a classical carbon star. That phase of stellar evolution is relatively brief, and most such stars ultimately end up as white dwarfs. We are now seeing these systems a comparatively long time after the mass transfer event, so the extra carbon observed in the present red giant was not produced within that star. This scenario is also accepted as the origin of the barium stars, which are also characterized as having strong spectral features of carbon molecules and of barium (an s-process element). Sometimes the stars whose excess carbon came from this mass transfer are called "extrinsic" carbon stars to distinguish them from the "intrinsic" AGB stars which produce the carbon internally. Many of these extrinsic carbon stars are not luminous or cool enough to have made their own carbon, which was a puzzle until their binary nature was discovered.
Other less convincing theories, such as CNO cycle unbalancing and Core Helium Flash have also been proposed as mechanisms for carbon enrichment in the atmospheres of smaller carbon stars.
When astronomers developed the spectral classification of the carbon stars, they got into considerable hardships when trying to correlate the spectra to the stars' effective temperatures. The trouble was with all the atmospheric carbon hiding the absorption lines normally used as temperature indicators for the stars.
| MK-type | R0 | R3 | R5 | R8 | Na | Nb |
| giant equiv. | G7-G8 | K1-K2 | ~K2-K3 | K5-M0 | ~M2-M3 | M3-M4 |
| Teff | 4300 | 3900 | ~3700 | 3450 | --- | --- |
The later N classes correspond less well to the counterparting M types, because the Harvard classification was only partially based on temperature, but also carbon abundance; so it soon became clear that this kind of carbon star classification was incomplete. Instead a new dual number star class C was erected so to deal with temperature and carbon abundance. Such a spectrum measured for Y Canum Venaticorum, was determined to be C54, where 5 refers to temperature dependent features, and 4 to the strength of the C2 Swan bands in the spectrum. (C54 is very often alternatively written C5,4).
| MK-type | C0 | C1 | C2 | C3 | C4 | C5 | C6 | C7 |
| giant equiv. | G4-G6 | G7-G8 | G9-K0 | K1-K2 | K3-K4 | K5-M0 | M1-M2 | M3-M4 |
| Teff | 4500 | 4300 | 4100 | 3900 | 3650 | 3450 | --- | --- |
A new revised Morgan-Keenan classification was published in 1993 by Philip Keenan, defining the classes: C-N, C-R and C-H. Later the classes C-J and C-Hd were added. This constitutes the established classification system used today :
| class | spectrum | population | MV | theory | temperature range (K) | example(s) | # known |
|---|---|---|---|---|---|---|---|
| classical carbon stars | |||||||
| C-R: | the old Harvard class R reborn: are still visible at the blue end of the spectrum, strong isotopic bands, no enhanced Ba line | medium disc pop I | 0 | red giants? | 5100-2800 | S Camelopardalis | ~25 |
| C-N: | the old Harvard class N reborn: heavy diffuse blue absorption, sometimes invisible in blue, s-process elements enhanced over solar abundance, weak isotopic bands | thin disc pop I | -2.2 | AGB | 3100-2600 | R Leporis | ~90 |
| non-classical carbon stars | |||||||
| C-J: | very strong isotopic bands of C2 and CN | unknown | unknown | unknown | 3900-2800 | Y Canum Venaticorum | ~20 |
| C-H: | very strong CH absorption | halo pop II | -1.8 | bright giants, mass transfer (all C-H:s are binary ) | 5000-4100 | V Arietis, TT Canum Venaticorum | ~20 |
| C-Hd: | hydrogen lines and CH bands weak or absent | thin disc pop I | -3.5 | unknown | ? | HD 137613 | ~7 |