Carbon-Deficient Giants (CDGs)

• Carbon-deficient giants are evolved stars defined by a depressed [C/Fe] (typically below -0.6 dex) and enhanced nitrogen, indicating extensive CN-cycle processing.
• Large spectroscopic surveys reveal that CDGs span various populations with context‐dependent abundance thresholds, implicating diverse evolutionary channels including mergers and core mixing.
• Asteroseismic analyses show CDGs are predominantly low-mass red clump stars, often linked with Li-rich phenomena, challenging conventional RGB mixing models.

Carbon-deficient giants (CDGs) are evolved stars whose photospheres exhibit depressed carbon abundances, usually accompanied by nitrogen enhancement and low ${}^{12}\mathrm{C}/{}^{13}\mathrm{C}$, indicating that CN-cycle hydrogen-burning products have reached the surface. In the literature, the term spans both the rare peculiar field red giants historically called “weak G-band,” “weak-CH,” or carbon-deficient red giants (CDRGs), and the broader phenomenon of carbon-depleted giants identified in globular clusters, dwarf galaxies, and large spectroscopic surveys (Bond, 2019, Maben et al., 2023). Across these usages, the common signature is a carbon-poor envelope, but the numerical thresholds, evolutionary states, and inferred origins are context-dependent (Maben et al., 2023).

1. Terminology and observational definition

The classical field objects are the weak G-band or weak-CH stars: rare red giants with abnormally weak or absent CH absorption near the G band at $\sim 4300$ Å. Their spectroscopic peculiarity is sufficiently strong that photometry can also identify them. In Strömgren $uvby$ photometry, barium stars with enhanced CH show the Bond-Neff effect and unusually low $c_1$, whereas CDRGs, lacking CH absorption, show an “anti-Bond-Neff effect,” with higher $c_1$ and lower $m_1$ than normal red giants at a given color (Bond, 2019).

Operational abundance cuts vary by survey and stellar population. In the APOGEE search for field CDGs, a conservative criterion of $[\mathrm{C}/\mathrm{Fe}] < -0.6$ dex was adopted, corresponding to a $6\sigma$ excursion from the main giant population; in that framework, standard red-giant evolution was described as depleting carbon to about $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$ dex, or down to $-0.6$ for the most metal-poor or low-mass cases (Maben et al., 2023). In the GALAH analysis of lithium-enriched giants, $\sim 4300$0 was treated as carbon-deficient in the context of first dredge-up abundances (Deepak et al., 2020). In M10, the historical literature threshold for “carbon-deficient giants” was summarized as $\sim 4300$1 to $\sim 4300$2 (Maas et al., 2019). These differences show that “CDG” is not tied to a single universal abundance boundary.

Large surveys have made such identifications scalable. Low-resolution LAMOST DR2 spectra at $\sim 4300$3 were used to measure carbon and nitrogen abundances to $\sim 4300$4 dex for 450,000 giant stars, with cross-validation showing agreement with APOGEE labels to within $\sim 4300$5 dex in $\sim 4300$6, $\sim 4300$7, $\sim 4300$8, $\sim 4300$9, and $uvby$0 for high-SNR objects (Ho et al., 2016). This established that precise carbon-depletion diagnostics can be extended to very large samples, even when the spectra are low-resolution.

2. Chemical phenomenology and abundance diagnostics

The defining abundance pattern is carbon depletion combined with nitrogen enhancement. In the rare field CDRGs, carbon is depleted relative to iron by factors of 10 to 30, corresponding to $uvby$1 to $uvby$2 dex; $uvby$3 declines to $uvby$4; nitrogen is enhanced by factors up to 4, with $uvby$5 up to $uvby$6 dex; and metallicity is typically near-solar or slightly subsolar, $uvby$7 to $uvby$8 (Bond, 2019). Lithium is usually drastically depleted, but some CDRGs show normal lithium (Bond, 2019).

The weak G-band (wGb) stars represent the most chemically extreme subset among field giants. High-resolution spectroscopy of 19 wGb stars gave typical $uvby$9 between $c_1$0 and $c_1$1 dex, $c_1$2 between $c_1$3 and $c_1$4 dex, $c_1$5 close to $c_1$6–5, oxygen near solar, modest sodium enhancement up to $c_1$7 dex, and no s-process enrichment in Sr or Ba (Palacios et al., 2015). Their abundance pattern is that of material “fully processed through the CNO cycle to an extent not known in other evolved intermediate-mass stars,” while their lithium abundances are diverse, with about half having $c_1$8 dex (Palacios et al., 2015).

For A-, F-, and G-type supergiants and bright giants, the carbon anomaly is milder but systematic. Most of the 36 stars analyzed in non-LTE show $c_1$9 to $c_1$0 dex, with a minimum around $c_1$1 dex, and $c_1$2 spans mostly $c_1$3 to $c_1$4 dex (Lyubimkov et al., 2014). The diagnostic relation

$c_1$5

encodes the CNO-cycle anti-correlation directly: larger nitrogen excess tracks stronger carbon depletion (Lyubimkov et al., 2014).

Another important discriminator is the CNO sum. In the APOGEE field sample, the newly identified CDGs generally have $c_1$6, consistent with scaled-solar initial composition subsequently reprocessed by hydrogen burning, whereas the previously known CDGs generally have $c_1$7, implying that some He-burning products were added to their envelopes before or during later processing (Maben et al., 2023). A later asteroseismic synthesis argued that two chemically distinct groups are sufficient, because one of the apparently separate high-pollution groups can be interpreted mainly as an $c_1$8-rich initial-composition analogue rather than a separate processing channel (Maben et al., 27 Aug 2025).

3. Carbon depletion along giant branches in clusters and dwarf galaxies

In metal-poor globular-cluster red giant branch (RGB) stars, carbon depletion is a standard evolutionary signature rather than an exceptional peculiarity. Photometric carbon-abundance measurements in M22, M5, M3, and M92 were used to define the surface carbon depletion rate

$c_1$9

fitted only for stars brighter than the RGB bump (Lee, 2023). The least-squares relation with metallicity is

$m_1$0

which expresses faster carbon depletion in more metal-poor clusters (Lee, 2023).

The cluster-to-cluster amplitudes are large. In M5, the depletion rates are $m_1$1 for first-generation (FG) stars and $m_1$2 for second-generation (SG) stars; in the more metal-poor M92, they rise to $m_1$3 and $m_1$4, respectively (Lee, 2023). The same study found that SG stars generally have slightly larger depletion rates than FG stars, most likely due to different internal temperature profiles associated with different initial helium abundances (Lee, 2023). This places many bright globular-cluster RGB stars within the phenomenological domain often labeled “carbon-deficient.”

M10 provides a spectroscopic benchmark for the isotopic signature. Infrared CO-band measurements gave average $m_1$5 for CN-normal stars and $m_1$6 for CN-enhanced stars, with no statistically significant difference between the populations over the observed magnitude range (Maas et al., 2019). Many M10 giants, especially above the luminosity-function bump, have $m_1$7 well below $m_1$8, showing that “CDG-like” surface chemistry can be a generic RGB outcome in metal-poor clusters rather than a rare anomaly (Maas et al., 2019).

The Draco dwarf spheroidal galaxy shows a related but not identical pattern. Among 35 giants with $m_1$9, nearly all have $[\mathrm{C}/\mathrm{Fe}] < -0.6$0, with most values between $[\mathrm{C}/\mathrm{Fe}] < -0.6$1 and $[\mathrm{C}/\mathrm{Fe}] < -0.6$2 dex, except for one candidate carbon-rich star with $[\mathrm{C}/\mathrm{Fe}] < -0.6$3 (Shetrone et al., 2013). For stars with $[\mathrm{C}/\mathrm{Fe}] < -0.6$4, $[\mathrm{C}/\mathrm{Fe}] < -0.6$5 declines with increasing luminosity on the RGB, consistent with deep mixing; for the six stars with $[\mathrm{C}/\mathrm{Fe}] < -0.6$6, no such decline is seen, and the absence of a trend was attributed either to small-number statistics or to primordial carbon inhomogeneities (Shetrone et al., 2013).

4. Demography, masses, and evolutionary state

Systematic survey work has shown that field CDGs are genuinely rare. The APOGEE DR17 search identified 103 new CDGs, increasing the known sample by more than a factor of 3, and found that they represent only $[\mathrm{C}/\mathrm{Fe}] < -0.6$7 per cent of giants in the high-quality sample (Maben et al., 2023). In that dataset they appear as an extended tail of the normal carbon distribution and are found in all major Galactic components: 48% in the thin disk, 44% in the thick disk, and 8% in the halo (Maben et al., 2023).

HR-diagram placement initially suggested that the bulk of these APOGEE CDGs were intermediate-mass stars, typically $[\mathrm{C}/\mathrm{Fe}] < -0.6$8–$[\mathrm{C}/\mathrm{Fe}] < -0.6$9, with the majority likely in the subgiant-branch or red-clump phases and others possibly on the red giant branch or early asymptotic giant branch (Maben et al., 2023). Earlier Gaia-based work on the classical weak G-band stars had also placed most CDRGs in a fairly tight clump in the CMD, consistent with initial masses of roughly $6\sigma$0 to $6\sigma$1 if they evolved as single stars, with about 10% at higher masses of roughly $6\sigma$2 to $6\sigma$3 and unusually high rotational velocities (Bond, 2019).

Asteroseismology substantially revised this picture. In the Kepler field, 15 new CDGs were identified, and 14 of 15 were unambiguously in the core-He-burning red clump; nearly all had asteroseismic masses $6\sigma$4, spanning approximately $6\sigma$5 (Maben et al., 2023). The same study found a bimodal luminosity distribution, with one red-clump group at $6\sigma$6 and another at $6\sigma$7, the latter being roughly a factor of two more luminous than expected for their masses (Maben et al., 2023).

A larger asteroseismic synthesis of the known population extended this revision. Among 129 CDGs observed by Kepler, K2, and TESS, solar-like oscillations were detected in 43 stars, and 79% of the seismically characterized sample were low-mass, $6\sigma$8 (Maben et al., 27 Aug 2025). That study confirmed the bimodal luminosity distribution, with peaks at $6\sigma$9 and $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$0, and showed that spectroscopic $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$1 is systematically offset from seismic values by about 0.06 dex, and in some cases by as much as 0.22 dex (Maben et al., 27 Aug 2025). A major implication is that HR-diagram placement and spectroscopy alone can misidentify both the mass scale and the evolutionary phase of CDGs.

5. Mixing physics and formation scenarios

For ordinary metal-poor RGB stars, extra mixing after first dredge-up provides a concrete physical route to carbon depletion. Models invoking thermohaline ($[\mathrm{C}/\mathrm{Fe}] \approx -0.2$2) mixing triggered by $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$3 burning reproduce the observed behavior of carbon-normal metal-poor giants: for a $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$4 model at $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$5, $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$6 falls by about 0.73 dex, $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$7 drops below 10 by the RGB tip, and the mixing phase lasts $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$8 yr (0904.2393). The same mechanism naturally predicts reduced carbon depletion in carbon-rich stars that have accreted AGB material, because their altered structure shifts the onset of mixing to lower core mass and cooler temperatures, where CN cycling is less efficient (0904.2393).

However, the most chemically extreme CDGs are not explained by a single standard mixing prescription. In M10, thermohaline models required an efficiency factor of $[\mathrm{C}/\mathrm{Fe}] \approx -0.2$9 to reproduce the observed $-0.6$0 decline, but a factor of $-0.6$1 to reproduce $-0.6$2, and no single parameter fit both observables simultaneously (Maas et al., 2019). In the wGb stars, detailed modeling concluded that the observed carbon depletion, nitrogen enrichment, low isotopic ratios, and in some cases preserved lithium are “very unlikely due to self-enrichment,” while no solid self-consistent pollution scenario could be presented either (Palacios et al., 2015).

For more luminous evolved stars, rotationally induced mixing and first dredge-up can account for more moderate carbon deficiencies. In A-, F-, and G-type supergiants and bright giants, the observed N–C anti-correlation is consistent with post-main-sequence objects with initial rotational velocities $-0.6$3–300 km/s, or post-first-dredge-up objects with $-0.6$4–300 km/s (Lyubimkov et al., 2014). That framework fits the mild $-0.6$5 to $-0.6$6 regime, but not the extreme weak G-band stars.

Recent asteroseismic work favors multiple formation channels for the rare field CDGs. One group of normal-luminosity, low-mass red-clump CDGs is most consistent with core He-flash mixing, while the more luminous and more chemically processed groups are best explained by mergers involving helium white dwarfs and RGB stars (Maben et al., 2023, Maben et al., 27 Aug 2025). The later synthesis argued that the more processed groups are plausibly merger products in hierarchical triple systems, with some stars remaining bound to a wide tertiary after an inner merger, and that pollution from AGB stars is very unlikely because $-0.6$7 remains unchanged across the groups (Maben et al., 27 Aug 2025). This suggests that the extreme field CDGs are not a simple extension of ordinary RGB extra mixing, even though both phenomena share CN-cycle surface chemistry.

6. Links to lithium-rich giants, related classes, and unresolved issues

A strong empirical connection exists between CDGs and Li-rich giants. In the Kepler asteroseismic sample, 6 out of 12 CDGs with measurable lithium were Li-rich, corresponding to 50%, and among the normal-luminosity CDGs the Li-rich fraction was about 90% (Maben et al., 2023). The broader asteroseismic synthesis likewise found lithium enrichment to be common across all groups, linking CDGs to Li-rich giants and suggesting a shared evolutionary origin (Maben et al., 27 Aug 2025).

That connection is chemically selective rather than universal. In the GALAH survey, ordinary Li-rich giants with $-0.6$8 to 3.1 have carbon deficiencies very similar to normal giants, reflecting first dredge-up alone, whereas super Li-rich giants with $-0.6$9 show an additional depression in $\sim 4300$00 of about 0.1–0.2 dex, consistent with extra CN-cycle processing (Deepak et al., 2020). A plausible implication is that CDGs overlap especially with the Li-rich giants that experienced unusually strong or unusually late mixing episodes.

The term “carbon-deficient giant” should also not be confused with carbon-rich or hydrogen-deficient carbon stars. DY Persei, for example, was argued to be a cool R Coronae Borealis variable rather than a normal carbon giant because it shows $\sim 4300$01, a ratio characteristic of RCB/HdC stars and sharply different from the $\sim 4300$02–1600 values of regular cool carbon giants (Garcia-Hernandez et al., 2023). That is a different abundance class: carbon-rich and likely hydrogen-deficient, not weak-G-band carbon-deficient in the sense used for CDGs.

Several issues remain unresolved. The rarity of CDGs is striking—$\sim 4300$03 per cent in APOGEE and $\sim 4300$04 in the Kepler background sample—given how common ordinary dredge-up and extra mixing are (Maben et al., 2023, Maben et al., 2023). The apparent mass scale changed from intermediate-mass HR-diagram estimates to predominantly low-mass seismic determinations (Bond, 2019, Maben et al., 27 Aug 2025). The relative roles of core-flash mixing, mergers, and binary evolution remain debated, and direct evidence for binarity is mixed: the APOGEE sample found a binary fraction of about 41%, not significantly different from the field, whereas the hierarchical-triple interpretation emphasizes wide tertiaries in the merger-formed groups (Maben et al., 2023, Maben et al., 27 Aug 2025). CDGs therefore remain one of the clearest cases in which surface abundance anomalies, asteroseismic structure, and binary or merger evolution must be interpreted jointly rather than in isolation.