2025MNRAS.544.1255C

The redshifted 21-cm signal of neutral hydrogen from the epoch of reionization (EoR) can potentially be detected using low-frequency radio instruments such as the Low-Frequency Array (LOFAR). So far, LOFAR upper limits on the 21-cm signal power spectrum have been published using a single target field: the North Celestial Pole (NCP). In this work, we analyse and provide upper limits for the 3C 196 field, observed by LOFAR, with a strong ${\approx }80$ Jy source in the centre. This field offers advantages such as higher sensitivity due to zenith-crossing observations and reduced geostationary radio-frequency interference, but also poses challenges due to the presence of the bright central source. After constructing a wide-field sky model, we process a single 6-h night of 3C 196 observations using direction-independent and direction-dependent calibration, followed by a residual foreground subtraction with a machine learning Gaussian process regression (ML-GPR). A bias correction is necessary to account for signal suppression in the GPR step. Still, even after this correction, the upper limits are a factor of 2 lower than previous single-night NCP results, with a lowest $2\sigma$ upper limit of $(146.61\, \text{mK})^2$ at redshift $z = 9.16$ and wavenumber $k=0.078\, h\, \text{cMpc}^{-1}$ (with $\text{d}k/k\approx 0.3$). The results also reveal an excess power, different in behaviour from that observed in the NCP field, suggesting a potential residual foreground origin. In future work, the use of multiple nights of 3C 196 observations combined with improvements to sky modelling and ML-GPR to avoid the need for bias correction should provide tighter constraints per unit observing time than the NCP.