We investigate the emergence of magnetic flux in the quiet Sun at very small spatial scales, focusing on the magnetic connection between the photosphere and chromosphere. The observational data consist of spectropolarimetric measurements and filtergrams taken with the Hinode satellite and the Dutch Open Telescope. We find that a significant fraction of the magnetic flux present in internetwork regions appears in the form of Ωshaped loops. The emergence rate is 0.02 loops per hour and arcsec −2 , which brings 1.1x10 12 Mx s −1 arcsec −2 of new flux to the solar surface. Initially, the loops are observed as small patches of linear polarization above a granular cell. Shortly afterwards, two footpoints of opposite polarity become visible in circular polarization within or at the edges of the granule and start to move toward the adjacent intergranular space. The orientation of the footpoints does not seem to obey Hale's polarity rules. The loops are continuously buffeted by convective motions, but they always retain a high degree of coherence. Interestingly, 23% of the loops that emerge in the photosphere reach the chromosphere (16 cases out of 69). They are first detected in Fe I 630 nm magnetograms and 5 minutes later in Mg I b 517.3 nm magnetograms. After about 8 minutes, some of them are also observed in Ca II H line-core images, where the footpoints produce small brightness enhancements.

Aims. We study the contradictory magnetic field strength distributions retrieved from independent analyses of spectropolarimetric observations in the near-infrared (1.56 µm) and in the visible (630 nm) spectral ranges in internetwork regions. Methods. To solve this apparent controversy, we present simultaneous and co-spatial 1.56 µm and 630 nm observations of an internetwork area. The properties of the circular and linear polarization signals, as well as the Stokes V area and amplitude asymmetries, are discussed. As a complement, we also used inversion techniques to infer the physical parameters of the solar atmosphere. As a first step, the infrared and visible observations are analysed separately to check their compatibility. Finally, the simultaneous inversion of the two data sets is performed. Results. The magnetic flux densities retrieved from the individual analysis of the infrared and visible data sets are strongly correlated. The polarity of the Stokes V profiles is the same at co-spatial pixels in both wavelength ranges. This indicates that both 1.56 µm and 630 nm observations trace the same magnetic structures on the solar surface. The simultaneous inversion of the two pairs of lines reveals an internetwork full of sub-kG structures that fill only 2% of the resolution element. A correlation is found between the magnetic field strength and the continuum intensity: equipartition fields (B ∼ 500 G) tend to be located in dark intergranular lanes, whereas weaker field structures are found inside granules. The most probable unsigned magnetic flux density is 10 Mx/cm 2 . The net magnetic flux density in the whole field of view is nearly zero. This means that both polarities cancel out almost exactly in our observed internetwork area.

Aims. The purpose of this work is to analyze the reliability of the magnetic field strengths inferred from the 630 nm pair of Fe i lines in internetwork quiet Sun regions. Methods. Some numerical experiments have been performed that demonstrate the inability of these lines to recover the magnetic field strength in such low flux solar regions. Results. It is shown how different model atmospheres, with magnetic field strengths ranging from a few hundred Gauss to kiloGauss, give rise to Stokes profiles that cannot be distinguished. The reasons for this degeneracy are discussed.

Aims. We study the structure of the magnetic field vector in the internetwork and search for the presence of small-scale loops. Methods. We invert 1.56 µm spectropolarimetric observations of internetwork regions at disc centre by applying the SIR code. This allows us to recover the atmospheric parameters that play a role in the formation of these spectral lines. We are mainly interested in the structure of the magnetic field vector. Results. We find that many opposite polarity elements of the internetwork are connected by short (2−6 ), low-lying (photospheric) loops. These loops connect at least the 10−20% of the internetwork flux visible in our data. Also we have some evidence that points towards a dynamic scenario that can be produced by the emergence of internetwork magnetic flux.