Unbiased Seismic Forward Modelling

Seismic forward modelling, i.e. the search for a match between some observed and model frequencies in order to identify a representative stellar model, is an important piece in the big puzzle of stellar astrophysics. It allows one to determine otherwise unknown stellar properties, to evaluate our stellar models, and even to test the underlying physics. However, there are know deficiencies in the stellar models that cause erratic shifts of the model frequencies and therefore hamper a direct comparison with the observations. This so-called surface effect is commonly treated by heuristic corrections calibrated to the Sun. Even though it is meanwhile expected that such corrections tend to fail, especially for stars very different from the Sun, it is still widely in use. In this talk, I present a new approach that uses probabilistic methods to marginalise the influence of the surface effect and therefore allows for an unbiased search of a best-fit model by using only the observed frequencies. The approach was tested for a wide range of solar-type oscillating stars using Kepler and TESS data. It thereby gives accurate results that are consistent with independent measurements, from e.g. eclipsing binary systems. As a byproduct the method returns the actual model frequency shifts, which are quite different from what is presently expected. This at least questions many results based on commonly used surface effect corrections.