Non-LTE inversions of a confined X2.2 flare

Context. Obtaining an accurate measurement of magnetic field vector in the solar atmosphere is essential for studying changes in field topology during flares and reliably modelling space weather. Aims. We tackle this problem by applying various inversion methods to a confined X2.2 flare that occurred in NOAA AR 12673 on 6 September 2017 and comparing the photospheric and chromospheric magnetic field vector with the results of two numerical models of this event. Methods. We obtained the photospheric magnetic field from Milne-Eddington and (non-)local thermal equilibrium (non-LTE) inversions of Hinode SOT/SP Fe I 6301.5 Å and 6302.5 Å. The chromospheric field was obtained from a spatially regularised weak-field approximation (WFA) and non-LTE inversions of Ca II 8542 Å observed with CRISP at the Swedish 1 m Solar Telescope. We investigated the field strengths and photosphere-to-chromosphere shear in the field vector. Results. The LTE- and non-LTE-inferred photospheric magnetic field components are strongly correlated across several optical depths in the atmosphere, with a tendency towards a stronger field and higher temperatures in the non-LTE inversions. For the chromospheric field, the non-LTE inversions correlate well with the spatially regularised WFA, especially in terms of the line-of-sight field strength and field vector orientation. The photosphere exhibits coherent strong-field patches of over 4.5 kG, co-located with similar concentrations exceeding 3 kG in the chromosphere. The obtained field strengths are up to two to three times higher than in the numerical models, while the photosphere-to-chromosphere shear close to the polarity inversion line is more concentrated and structured. Conclusions. In the photosphere, the assumption of LTE for Fe I line formation does not yield significantly different magnetic field results in comparison to the non-LTE case, while Milne-Eddington inversions fail to reproduce the magnetic field vector orientation where Fe I is in emission. In the chromosphere, the non-LTE-inferred field is excellently approximated by the spatially regularised WFA. Our inversions confirm the locations of flux rope footpoints that have been predicted by numerical models. However, pre-processing and lower spatial resolution lead to weaker and smoother field in the models than what our data indicate. This highlights the need for higher spatial resolution in the models to better constrain pre-eruptive flux ropes.