Using APOGEE Stellar Abundances

This page attempts to address some common questions about APOGEE stellar parameters that are determined from the APOGEE Stellar Parameters and Chemical Abundances Pipeline (ASPCAP). Additional details are given in Holtzman et al (2021).

The APOGEE survey extracts the chemical abundances of multiple elements for the entire stellar sample. In DR17, we attempt to measure abundances for 23 species: C, C I, N, O, Na, Mg, Al, Si, P, S, K, Ca, Ti, Ti II, V, Cr, Mn, Fe, Co, Ni, Cu, Ce, and Nd. The accuracy of an individual element varies with the element and stellar type; certain parts of the element-star parameter space are not feasible to explore with the APOGEE data, as is discussed below. If you are interested in learning more about APOGEE’s abundances and their derivation see the DR17 ASPCAP Description or Holtzman et al. 2021.

Overview of APOGEE Stellar Abundances

In DR17, we attempt to measure abundances for 23 species: C (measured from molecules), C I (measured from neutral carbon lines), N, O, Na, Mg, Al, Si, P, S, K, Ca, Ti (measured from neutral titanium lines), Ti II (measured from singly ionized titanium lines), V, Cr, Mn, Fe, Co, Ni, Cu, Ce, and Nd. These abundances are reported in the bracket notation, for instance:

$[Fe/H] = log_{10}(N_{Fe} / N_{H}) - log_{10}(N_{Fe} / N_{H})_{Sun}$.

APOGEE’s solar scale should be close to the Grevasse et al. (2007) solar values . However, we find that abundances of some elements for stars of solar metallicity in the solar neighborhood come out with non-solar abundance ratios, which is not expected based on previous studies, so, in addition to the raw spectroscopic values, we provide calibrated values that have been adjusted by addition of a zeropoint to yield [X/M]=0 for these stars. Users interested in the details of APOGEE’s abundance scale should read the DR17 ASPCAP description.

These abundances are provided in various formats relative to hydrogen (H), total metallicity (M), or iron (Fe). The abundances that most will want to use are the metallicity reported by [Fe/H] and the individual elemental abundances relative to iron [X/Fe].

As described on the DR17 ASPCAP Description, DR17 uses a new set of synthetic spectral libraries to determine parameters and abundances. This library includes non-local thermal equilibrium (NLTE) treatment for four elements: Na, Mg, K, and Ca.

Multiple columns with stellar abundances

We provide multiple columns that record APOGEE’s stellar parameters. The raw abundances as measured by FERRE are saved in the FELEM array. For a uniform presentation relative to hydrogen or metals, we populate the X_H_SPEC and X_M_SPEC arrays. Zeropoint calibrations to bring solar metallicity stars in the solar neighborhood to [X/M]=0 are used to populate X_H and X_M. Finally, "named" tags giving abundances relative to iron (e.g., MG_FE_SPEC, MG_FE, etc.) are populated from the arrays, but only for objects which have not been flagged as problematic.

If you are unfamiliar with APOGEE data we recommend using the “named tags”: FE_H, and individual elemental abundances measured relative to iron, e.g., C_FE, or MG_FE. These named tags are the most conservative in how they are populated. Stars with the most suspect abundances or whose abundances are known to be wrong do not have any data in these named tags.

For more complete data, use the X_H and X_M arrays, but be aware that there may be significant issues with some of these. Consult the ASPCAPFLAG and ELEMFLAG bitmasks if you use these.


Uncertainties on abundances are estimated from repeat observations of stars and can be found in named tags such as C_FE_ERR.

In addition to the named uncertainty tags such as C_FE_ERR, we also provide the raw uncertainties from ASPCAP’s abundance fitting procedure in the FELEM_ERR array, but these
generally seem to be significantly underestimated.

Since the abundances are determined in a separate fit after the parameters have been determined, covariances between abundances and parameters (or other abundances) are not provided.

Quality Flags

Each element has an associated bitmask , e.g., C_FE_FLAG, that contains descriptive information about potential issues with the elemental abundance for each star; this is also saved in the ELEMFLAG following the order of elements in FELEM.

We note that some abundances appear to be unreliable and/or unmeasurable in some regions of parameter space. From visual inspection of trends among solar neighborhood stars, abundances become particularly challenging at low effective temperatures. Based on the results, we have selected effective temperature cuts for some elements, and for stars outside of the acceptable range, we have set a TEFF_CUT bit in the abundance bitmask. Stars with this bit set still have the abundances populated in the abundance arrays, but they are not populated in the named tags.

Quality of derived abundances

While some quality cuts have been applied to the values in the named abundance tags, not all elements have the same quality data. For those who are unfamiliar with APOGEE elemental abundances, we provide a general guideline for the reliability of individual elements below. However, these descriptions are very general and the quality of a given element will vary by metallicity, temperature and $S/N$ (for instance at the lowest metallicities only a few elements remain measurable) and we encourage users to explore the data and make their own judgements.

Challenges with cooler stars

Abundances of cool stars (Teff<4000 K, and especially Teff<3500 K) are particularly challenging, perhaps due to the significant presence of molecular absorption and challenges in interpolating between synthetic spectra in this regime. Shallow minima in the fitting space may also contribute.

For giant stars, several elements appear to be measured in narrow sequences of abundances, sometimes multi-modal, in the coolest stars.

For dwarf stars, abundances seem systematically low, by as much as several tenths of a dex for stars with Teff < 3500 K. In addition, many abundances seem to show an anonymously low dip in abundance between 4000 < Teff < 5000 K that may be related to the presence of very strong lines in this range of effective temperature.

Challenges with warmer stars

At warmer effective temperatures, lines from many elements become weak or disappear. Above 7000 K, it is difficult to determine any abundances, and no calibrated abundances are populated.

Elements in Giants

Best (precise, measurable over a wide range of stellar parameters, match literature chemical trends): C, N, O, Mg, Al, Si, Mn, Fe, Ni
Good (less precise, measurable over narrower range of stellar parameters, apparent chemical abundance trends): C I, Na, K, Ca, Co, Ce
Okay (moderate scatter, measurable over limited range of stellar parameters, coherent chemical abundance trends only apparent in narrow parameter ranges): S, V, Cr
Deviate from literature Galactic trends: Ti, Ti II
Unreliable (no stars have populated named tags): P, Cu, Nd
Not attempted in DR17: Ge, Rb, Yb

Elements in Dwarfs

APOGEE’s abundances for dwarfs are typically not as precise as for giants, so the overall quality of these abundances may be slightly lower. Additionally some elements are more difficult to measure in dwarfs and have therefore have not been populated.

Best (low scatter, clear chemical abundance patterns that match literature patterns): C, Mg, Si, Fe, Ni
Good (moderate scatter, chemical abundance patterns generally match literature): C I, O, Al, K, Ca, Mn
Okay (moderate scatter, vague chemical abundance patterns): N, S
Large scatter, no coherent chemical abundance patterns: Na, Ti, V, Cr
Unreliable (no stars have populated named tags): P, Ti II, Co, Cu, Ce, Nd
Not attempted in DR17: Ge, Rb, Yb

Systematics in abundances of stars across the HR Diagram

Measuring chemical abundances is a challenging endeavor, and it is difficult to measure them consistently across the HR diagram. While APOGEE applies some basic quality cuts and zero-point calibrations to dwarf and giant abundances, the strength and measurability of elements vary across the HR diagram, so some systematic trends and features may still be found in APOGEE’s chemical abundances. Users should exercise caution when comparing the abundances of stars across the HR diagram. To minimize systematic trends, you may not want to compare samples that span a wide range of stellar parameters, or you may need to apply corrections. The only calibrations we have applied are simple zero-point corrections, although we do adopt different zero-point corrections for giants (log g < 3.8) and dwarfs.

Particularly difficult parts of the HR diagram are cool stars (with temperatures below $\sim 3500$ K) due to frequent line blending with strong molecular lines and in “hot” stars (with temperatures above $\sim 6000$ K) whose atomic and molecular lines are very weak. There are also several elements that have systematic temperature trends in dwarfs cooler than $\sim 4500$ K.

What else should I watch out for?

Users should be aware that the precision of abundance measurements decreases with decreasing signal-to-noise ratios. APOGEE’s target for precision chemical abundances is a SNR of 100, however most elements can be well measured down to a SNR of 70 although the ideal SNR differs from element to element depending on the strength of their lines.

Be aware that outliers may not be astrophysical and should be checked to assess the quality of their abundances. Users interested in outliers should consult the ASPCAPFLAG bitmask and STARFLAG bitmask for those stars to see if they have any quality warnings and may be interested in investigating the spectra of those stars to inspect the lines of their element of interest (see the Using Spectra for details about investigating APOGEE spectra). See the Tutorials for a detailed walk-through of how to investigate a star that appears to be an outlier.