This page relays a list of known issues in GALAH DR4. Please also see the GALAH Survey Fourth Data Release paper (Buder et al. 2024).

We caution that it is not possible to inspect the millions of measurements that make up GALAH DR4. There are likely to be some unexpected correlations and problems with the data.

Flags

We recommend only using stars with flag_sp == 0, flag_X_fe == 0, and snr_px_ccd3 > 30.

There are three types of flags for the data.

  • flag_sp – reliability of the stellar parameters
  • flag_X_fe – reliability of the abundance of [X/Fe]
  • snr_px_ccdX - median signal to noise per pixel in HERMES camera X We leave it up to the user which flags they apply but stress that ignoring the flags is only advisable when necessary for the science case.

The bits in flag_sp do not all indicate fatal problems, and you should set your choice of flags to match your science case. For example, if the elemental abundances you need would not be affected by chromospheric emission or lack of data in CCD4, stars with a flag_sp value of 1, 2 or 3 should not present a problem.

Due to a bug in the code used to calculate flag_fe_h, it is not actually useful for indicating unreliable metallicity values. This flag should not be used.

For more information, see the Best Practices for using GALAH DR4 and the Flags in GALAH DR4 pages.

Binarity

- Our method for identifying binary stars is not perfectly accurate, and particularly for young stars and other rapid rotators there may be misidentifications as binaries.
- A full analysis mathod for GALAH binary stars is still under development.

Globular clusters

- Fixing the distances has removed the temperature trends seen in GALAH DR3, so the stellar parameters are improved.
- There continue to be issues with abundance zeropoints and scatter, such that the light-element abundance anomalies are not as clearly visible as they should be.

Intrinsic correlations and the parameter space

Quoting from the DR4 paper:

One of the primary challenges in creating an optimal training set for spectrum interpolation lies in the choice of parameter sampling. A common caveat is the use of randomized, uncorrelated parameter sampling, which can lead to unrealistic combinations of elemental abundances. Elements that share a similar nucleosynthesis channel often exhibit correlated behavior, for instance, stars with high abundances of Mg are typically also enhanced in Si, Ca, and Ti, while Na and Al tend to be elevated together. Similarly, neutron-capture elements like Y and Ba often follow similar trends (e.g. Ting et al. 2012; Kobayashi et al. 2020; Buder et al. 2021). To better capture this behavior in the training set, the use of scaled linear functions or normalizing flows could be advantageous. These approaches would help minimize the occurrence of unlikely parameter combinations and yield a more representative sample.