LETTERS https://doi.org/10.1038/s41550-017-0293-z © 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved. 1 Physical Research Laboratory, Ahmedabad, Gujarat 380009, India. 2 Tata Institute of Fundamental Research, Mumbai, Maharashtra 400005, India. 3 Inter-University Center for Astronomy and Astrophysics, Pune, Maharashtra 411007, India. 4 Indian Institute of Technology Bombay, Mumbai, Maharashtra 400076, India. 5 Raman Research Institute, Bengaluru, Karnataka 560080, India. 6 National Centre for Radio Astrophysics, Pune, Maharashtra 411007, India. 7 Vikram Sarabhai Space Center, Thiruvananthapuram, Kerala 695002, India. 8 Indian Space Research Organization, Bengaluru, Karnataka 560231, India. *e-mail: santoshv@prl.res.in The Crab pulsar is a typical example of a young, rapidly spin- ning, strongly magnetized neutron star that generates broad- band electromagnetic radiation by accelerating charged particles to near light speeds in its magnetosphere 1 . Details of this emission process so far remain poorly understood. Measurement of polarization in X-rays, particularly as a func- tion of pulse phase, is thought to be a key element necessary to unravel the mystery of pulsar radiation 2–4 . Such measure- ments are extremely difficult, however: to date, Crab is the only pulsar to have been detected in polarized X-rays 5–8 and the measurements have not been sensitive enough to ade- quately reveal the variation of polarization characteristics across the pulse 7 . Here, we present the most sensitive mea- surement to date of polarized hard X-ray emission from the Crab pulsar and nebula in the 100–380 keV band, using the Cadmium–Zinc–Telluride Imager 9 instrument on-board the Indian astronomy satellite AstroSat 10 . We confirm with high significance the earlier indication 6,7 of a strongly polarized off-pulse emission. However, we also find a variation in polar- ization properties within the off-pulse region. In addition, our data hint at a swing of the polarization angle across the pulse peaks. This behaviour cannot be fully explained by the exist- ing theoretical models of high-energy emission from pulsars. The Cadmium–Zinc–Telluride Imager (CZTI) on-board the first dedicated Indian astronomy satellite, AstroSat, is a coded aperture telescope designed for hard X-ray observations, including polariza- tion. The hard X-ray polarization measurement capability of CZTI has been securely established by extensive ground calibrations using partially polarized and, more importantly, unpolarized X-rays 11 . Since the launch of AstroSat on 2015 September 28, the Crab nebula has been extensively observed on 21 different occasions during its first 18 months of operation and we use all observations in our analy- sis. Since hard X-ray polarization measurements almost always have limited statistical significance and can thus be affected by large sys- tematic uncertainties, we have carefully validated all the steps in the analysis procedure to extract the polarization signal (see Methods for details). These steps include the selection of Compton events for polarization analysis, selection and subtraction of the appro- priate background component, normalization of the instrumental modulation, co-addition of individual observations, and so on. One very important validation step is the demonstration of the pulsed signal in the Compton events selected for the polarization analysis (as shown in Supplementary Fig. 1). We find that the pulse profile of the Compton events closely matches the known double-peaked pulse intensity profile in this energy range 12 , with the second peak stronger than the first. Subsequent to this validation, we carried out a polar- ization analysis following the procedure reported by Vadawale et al. 11 . We found that all CZTI observations with a net exposure greater than 50 ks yield a clear detection of polarization. In a few cases where the individual exposures were lower, we co-added exposures from a given month, resulting in eight independent datasets. The polariza- tion fraction and angle derived from these eight datasets (Fig. 1a,b, respectively) show a consistent behaviour. Two representative modu- lation curves are shown in the insets of Fig. 1b (all modulation curves are shown in Supplementary Fig. 5). When all Crab observations are co-added, the resultant exposure time is ~800 ks and the measured polarization fraction is 32.7 ± 5.8%, yielding detection at more than 5σ significance. For comparison, the minimum detectable polariza- tion at the 99% confidence level for an 800 ks observation of a Crab- like source with CZTI is around 10%. The measured position angle is 143.5 ± 2.8° eastwards from the celestial north. For the off-pulse region alone (a phase range of 0.82 to 1.18 as per our phase defini- tion, which has an offset of 0.3 phase with respect to the phase defini- tion used by Kuiper et al. 12 ), our estimates of the polarization fraction and polarization angle are 39.0 ± 10.0% and 140.9 ± 3.7°, respectively, providing close to 4σ detection in the off-pulse region itself. Our measured polarization fraction value is consistent with that obtained by INTEGRAL (46 ± 10%) in the energy range of 0.1 to 1 MeV; how- ever, the polarization angle is slightly higher than that reported by INTEGRAL (123 ± 11°) 6 . Considering the fact that, at soft X-ray energies, the polarization fraction is even lower (19.5 ± 2.8%) and the polarization angle is farther away (152.6° ± 4.0°) 5 , this suggests a systematic energy dependence of polarization properties. The co- added modulation curves are shown in Fig. 1c,d for the total Crab emission and the off-pulse emission alone, respectively. The confi- dence regions shown in Fig. 1e (total Crab) and Fig. 1f (off-pulse) demonstrate that these are the most precise hard X-ray polarization measurements of the Crab so far. The high significance of our polarization detection enables us to examine the dependence of the polarization characteristics on the pulse phase. Here, we are guided by the detailed phase-resolved polarization study made at radio 13 and optical 14 wavelengths. In particular, studies in the optical band 14 show that the off-pulse and ‘bridge emission’ (connecting the two peaks) have relatively steady Phase-resolved X-ray polarimetry of the Crab pulsar with the AstroSat CZT Imager S. V. Vadawale  1 *, T. Chattopadhyay 1 , N. P. S. Mithun 1 , A. R. Rao 2 , D. Bhattacharya  3 , A. Vibhute 3 , V. B. Bhalerao  4 , G. C. Dewangan 3 , R. Misra 3 , B. Paul 5 , A. Basu 6 , B. C. Joshi  6 , S. Sreekumar 7 , E. Samuel 7 , P. Priya 7 , P. Vinod 7 and S. Seetha 8 NATURE ASTRONOMY | www.nature.com/natureastronomy