Context. Polarimetric imaging is one of the most effective techniques for high-contrast imaging and for the characterization of protoplanetary disks, and it has the potential of becoming instrumental in the characterization of exoplanets. The Spectro-Polarimetric High-contrast Exoplanet REsearch (SPHERE) instrument installed on the Very Large Telescope (VLT) contains the InfraRed Dual-band Imager and Spectrograph (IRDIS) with a dual-beam polarimetric imaging (DPI) mode, which offers the capability of obtaining linear polarization images at high contrast and resolution. Aims. We aim to provide an overview of the polarimetric imaging mode of VLT/SPHERE/IRDIS and study its optical design to improve observing strategies and data reduction. Methods. For H-band observations of TW Hydrae, we compared two data reduction methods that correct for instrumental polarization effects in different ways: a minimization of the "noise"image (Uφ), and a correction method based on a polarimetric model that we have developed, as presented in Paper II of this study. Results. We use observations of TW Hydrae to illustrate the data reduction. In the images of the protoplanetary disk around this star, we detect variability in the polarized intensity and angle of linear polarization that depend on the pointing-dependent instrument configuration. We explain these variations as instrumental polarization effects and correct for these effects using our model-based correction method. Conclusions. The polarimetric imaging mode of IRDIS has proven to be a very successful and productive high-contrast polarimetric imaging system. However, the instrument performance is strongly dependent on the specific instrument configuration. We suggest adjustments to future observing strategies to optimize polarimetric efficiency in field-tracking mode by avoiding unfavorable derotator angles. We recommend reducing on-sky data with the pipeline called IRDAP, which includes the model-based correction method (described in Paper II) to optimally account for the remaining telescope and instrumental polarization effects and to retrieve the true polarization state of the incident light.

@en

Young giant exoplanets emit infrared radiation that can be linearly polarized up to several percent. This linear polarization can trace: 1) the presence of atmospheric cloud and haze layers, 2) spatial structure, e.g. cloud bands
and rotational attening, 3) the spin axis orientation and 4) particle sizes and cloud top pressure. We introduce a novel high-contrast imaging scheme that combines angular dierential imaging (ADI) and accurate near-infrared
polarimetry to characterize self-luminous giant exoplanets. We implemented this technique at VLT/SPHEREIRDIS and developed the corresponding observing strategies, the polarization calibration and the data-reduction approaches. The combination of ADI and polarimetry is challenging, because the eld rotation required for ADI negatively aects the polarimetric performance. By combining ADI and polarimetry we can characterize planets that can be directly imaged with a very high signal-to-noise ratio. We use the IRDIS pupil-tracking mode and
combine ADI and principal component analysis to reduce speckle noise. We take advantage of IRDIS' dual-beam polarimetric mode to eliminate dierential eects that severely limit the polarimetric sensitivity (at-elding errors, dierential aberrations and seeing), and thus further suppress speckle noise. To correct for instrumental polarization eects, we apply a detailed Mueller matrix model that describes the telescope and instrument and that has an absolute polarimetric accuracy 0:1%. Using this technique we have observed the planets of HR 8799 and the (sub-stellar) companion PZ Tel B. Unfortunately, we do not detect a polarization signal in a rst analysis. We estimate preliminary 1 upper limits on the degree of linear polarization of 1% and 0:1% for the planets of HR 8799 and PZ Tel B, respectively. The achieved sub-percent sensitivity and accuracy show that our technique has great promise for characterizing exoplanets through direct-imaging polarimetry.@en