Deja Vu All Over Again: The Reappearance of Supernova Refsdal

TL;DR

This paper reports the first successful prediction and observation of a supernova's reappearance due to gravitational lensing, testing lens models and constraining systematic uncertainties in predictions.

Contribution

It demonstrates the first successful prediction of a supernova reappearance in a gravitational lens system, validating lens models and improving understanding of systematic uncertainties.

Findings

The reappearance matched some model predictions.

Timing and brightness were consistent with certain models.

The study constrains systematic uncertainties in lens modeling.

Abstract

In Hubble Space Telescope (HST) imaging taken on 10 November 2014, four images of supernova (SN) "Refsdal" (redshift z=1.49) appeared in an Einstein-cross--like configuration (images S1-S4) around an early-type galaxy in the cluster MACS J1149.5+2223 (z=0.54). Almost all lens models of the cluster have predicted that the SN should reappear within a year in a second host-galaxy image created by the cluster's potential. In HST observations taken on 11 December 2015, we find a new source at the predicted position of the new image of SN Refsdal approximately 8" from the previous images S1-S4. This marks the first time the appearance of a SN at a particular time and location in the sky was successfully predicted in advance! We use these data and the light curve from the first four observed images of SN Refsdal to place constraints on the relative time delay and magnification of the new image (SX), compared to images S1-S4. This enables us, for the first time, to test "blind" lens model predictions of both magnifications and time delays for a lensed SN. We find that the timing and brightness of the new image are consistent with the blind predictions of a fraction of the models. The reappearance illustrates the discriminatory power of this blind test and its utility to uncover sources of systematic uncertainty. From planned HST photometry, we expect to reach a precision of 1-2% on the time delay between S1-S4 and SX.