Once-in-a-lifetime event

Astronomers successfully predict a distant supernova replay in advance

Claudio Grillo


Credits: NASA & ESA

1 Basic concepts and historical remarks

Light, just like massive particles, is deflected when it propagates through a gravitational field. Einstein’s theory of General Relativity (GR, 1915), the theory which we believe describes most accurately gravity, predicts that the deflection α of light rays passing a point mass M with an impact parameter ξ is given, in the limits of weak fields and of velocity of the deflecting mass small compared to c, by

$ \alpha = \frac{4GM}{c^{2}\xi} $ ,

where G and c are Newton constant of gravity and the velocity of light, respectively. A measurement of this deflection was performed by Sir Arthur Eddington in 1919, during a total solar eclipse, from the shift of the apparent positions of stars close to the solar surface (see fig. 1). The agreement between Einstein’s predictions and observations provided the first real test of GR and established its success. The deflection of light by gravity is usually called gravitational lensing effect (box 1).

In 1936, Albert Einstein considered several aspects of the lensing effect produced by a star. He also discussed the possibility of observing two images of a background source coalingned with the foreground lens star, one on either side of it. He concluded that due to the small angular separation between the two images (of order milli-arcsecond) “there is no great chance of observing this phenomenon”. One year later, Fritz Zwicky pointed out that by looking at lensing by “extragalactic nebulae” (nowadays called galaxies), instead of considering stars as lenses, the expected image separation of a pair of images of a background source could be resolvable with telescopes. He noted that such an observation would furnish another test of GR, make more distant galaxies detectable through the magnification effect, and allow one to measure the masses of the lenses. Moreover, he estimated that 1 over 400 distant sources should be lensed by a nebula, and hence claimed that “the probability that nebulae which act as gravitational lenses will be found becomes practically a certainty”.

In spite of that, it took about forty more years before the first lens system was observed and identified as such. In fact, only in 1979 Walsh and collaborators discovered in the radio band a pair of quasars with an angular separation of about 6 arcseconds, located at the same distance from us and with same colours and spectra. The following year, a massive elliptical galaxy of a small cluster was optically identified between the two images. These two images were then observed in all wavelength ranges and showed always the same spectral features, leaving no doubt to their lensed nature. That was the beginning of one of the most fruitful and rapidly growing branches of astrophysics: strong gravitational lensing.

In gravitational lensing, as in standard optics, the Fermat’s principle is valid: (multiple) images of a source form following ray paths for which the light propagation time is stationary. When two (or more) images of a single source are observed, the travel time that light takes to reach the observer along the different light paths is, in general, different. This is the result of two effects: a geometric and a potential time delay. The first term can be ascribed to the fact that light rays bent by a lens (“curved”) are geometrically longer than unperturbed (“straight”) ones. The second term, also known as Shapiro delay, is a relativistic effect related to the traversing of the gravitational field of the lens.

In 1964, Sjur Refsdal speculated on the possibility of observing two multiple images of a single distant supernova (SN) (box 2), lensed by a foreground massive galaxy, and using the difference in the light travel time for the two different paths to determine both the value of the Hubble parameter and the total mass of the lens. Since then, astronomers have steadily tried to identify multiply imaged SNe, recognising the important implications of detecting such rare events.

For about 50 years, a very small number of singly imaged and not significantly magnified SNe have been found. In 2013, a Type Ia SN approximately 30 times more luminous than a normal SN Ia at that distance was discovered in the Panoramic Survey Telescope and Rapid Response System 1 (Pan-STARRS) survey. Less than a year later, the foreground, magnifying lens galaxy was detected and it was estimated that the bright SN had very likely been split into multiple images, but they could not be resolved in the available observations, taken only from the ground.

2 SN Refsdal – Act I

The first multiply imaged, highly magnified and spatially resolved SN was finally identified in November 2014 in the Hubble Space Telescope (HST) images of the galaxy cluster MACS J1149.6+2223 (hereafter MACS 1149). This object was nicknamed “SN Refsdal” in recognition of the astronomer’s original idea. Four multiple images of SN Refsdal (labelled as S1-S4) were found, in a so-called Einstein cross configuration, around an elliptical galaxy, member of MACS 1149 (see fig. 4). The SN exploded in one arm of an almost face-on background spiral galaxy, magnified by the gravitational potential of the foreground galaxy cluster.

The SN Refsdal host galaxy is clearly visible in the HST data, lensed into three main multiple images showing angular distances of several arcseconds, typical of cluster-scale strong lensing. The four multiple images of SN Refsdal were discovered in one of these three images with an average separation of approximately 2 arcseconds, as usually observed in galaxy-scale lensing. All of the first published strong lensing models of MACS 1149 predicted that SN Refsdal had already appeared in another image of the spiral host galaxy and that it would be visible in the near future in yet another image of the spiral close to the cluster core (see fig. 4).

It was immediately recognised that SN Refsdal could offer a rare opportunity to test different strong lensing models used to describe the total mass distribution of MACS 1149 and to predict the positions, magnifications and time delays of the multiple images of the lensed supernova. In particular, it became obvious that the “future” SN Refsdal image could enable astronomers to perform for the first time a true blind test of model predictions on extragalactic scales of an event that had not happened yet, thus free from experimenter bias.

SN Refsdal was discovered in near-infrared (near-IR) exposures obtained using the HST Wide Filed Camera 3 (WFC3), taken as part of the Grism Lens Amplified Survey from Space (GLASS). The same field was then the target of an intensive optical and near-IR imaging campaign that had already been planned within the Hubble Frontier Fields (HFF) initiative This was a Director Discretionary Time programme consisting of 140 HST orbits and aimed at obtaining deep, multi-colour observations in six massive galaxy clusters to study, through strong lensing analyses, their dark-matter haloes and to detect the highest redshift galaxies, magnified by the clusters. After the discovery of this exceptional event, a number of additional imaging and spectroscopic follow-up HST programmes have been triggered to classify the SN, as well as to measure the relative time delays of the multiple images. Extensive spectroscopy from the ground was also collected with the Multi-Unit Spectroscopic Explorer (MUSE) and the X-shooter echelle spectrograph, both mounted on the Very Large Telescope (VLT) at the Paranal observatory and the Multi-Object Spectrometer for Infrared Exploration (MOSFIRE) and the Deep Imaging Multi-Object Spectrograph (DEIMOS), mounted on the 10 m Keck-I and Keck-II telescopes, respectively.

Taking advantage of this exceptional and highly complementary data sets in the MACS 1149 field of view, it was possible to measure reliable spectroscopic redshifts (i.e., distances) for 429 stars and galaxies, and to identify 170 member galaxies of the cluster and 23 multiple images of 10 different background, lensed galaxies.

In a coordinated collaborative effort, five different international groups (including a significant fraction of Italian scientists) worked together to merge all available follow-up data of MACS 1149 and decide which information to use in their improved strong lensing models. The astronomers first carefully examined and agreed on a new set of multiple images to include in their models (some based on a number of physically motivated mass components, others on more flexible mass parameterisations) and then independently updated them, punctually providing their falsifiable predictions for the possible reappearance of SN Refsdal. They managed to submit a summary comparison work in the second half of October 2015, before the field could be targeted again with HST (between 2015 July 21 and 2015 October 30 the Sun was too close to the field for safe observations). With some scatter, the different models predicted that a “past” image of SN Refsdal had appeared between 1994 and 2004, but would have been too faint to be visible in HST archival images and that a “future” image of SN Refsdal could be detected in single-orbit HST images between the end of October 2015 and the beginning of 2016, with its peak luminosity occurring in the first half of 2016.

3 SN Refsdal – Act II

At the end of 2015, some traces of nervouseness could be perceived among the lensing experts, owing to the lack of any new HST detection in the MACS 1149 field, and they started discussing possible systematic effects (such as mass structure along the line of sight and uncertainties on the values of the cosmological parameters), not considered in the models, which could have significantly affected their predictions. Suddenly, great excitement spread in December 2015, when a new source at the model-predicted position of the “future” image of SN Refsdal was found (see fig. 5, fig. 6). That signed an important moment in astronomy: the appearance of a distant supernova at a specific sky position and time had been correctly predicted in advance! It should be remarked that significant differences between the observed and model-predicted location, luminosity and time of the SN Refsdal reappearance would be expected if our understanding of how (luminous and dark) matter is distributed in galaxy clusters or the values of the cosmological parameters, that are used to describe the geometry of the Universe, or the theory of GR were incorrect.

Since the first detection of SN Refsdal in November 2014, several HST programmes have constantly been monitoring the MACS 1149 cluster field to measure the light curves of all multiple images. From them, it is possible to get robust estimates of the relative time delays and magnification ratios of the different SN images. These quantities have already been published for the first four multiple images, while an on-going programme should soon provide them for the most recent image with only a few per cent errors.

The timing and brightness definitive information of the first 4 multiple images of SN Refsdal was not available at the time when the strong lensing models for the prediction of its reappearance were developed and thus was not included there. A posteriori, this information has allowed a first fair comparison of the different models, in the effort of understanding whether any particular choice in the cluster mass parameterisation or in the employed photometric and spectroscopic data could match better the SN observations.

For accurate model predictions, a clear suggestion is that the quality of the lensing input constraints is more crucial than the quantity. For instance, a large fraction of spectroscopically confirmed (i.e., secure) multiple images of different background sources is more important than the absolute total number of photometrically identified (i.e., possible) multiple images.

4 SN Refsdal classification, its environment and host galaxy

By using HST and VLT near-IR spectra (clear hydrogen emission and absorption features detected) and images, SN Refsdal was classified as a 1987A-like SN, a not so frequent type of SNe II. SN 1987A exploded in the Large Magellanic Cloud and is the brightest and best-studied SN observed in the last 400 hundred years. Its progenitor was identified as a compact (radius on the order of 100 times that of the Sun or smaller), blue, massive (approximately 20 times more massive than the Sun) star. The size of such a progenitor was responsible for a light curve slowly rising. After discovery, SN Refsdal multiple images continued to rise in brightness for approximately 150 days, becoming incompatible with any normal SN Ia, Ib, Ic, and SN II and showing close similarities to the light curves of SN 1987A-like SNe.

It has been shown that the evolution of galaxies in the Universe is regulated by energetic feedback from active galactic nuclei (AGN) and SNe. These processes contribute to determine how much gas is able to condense and form stars. The investigation of the properties of the interstellar medium (ISM) surrounding SNe, especially in galaxies in the young Universe, is key to determining how efficient the feedback of SNe can be. Studies of this kind are currently possible only in gravitationally lensed galaxies, where the spatial resolution and luminosity are boosted by the lensing magnification effect. VLT optical data have provided the unique opportunity to study the SN Refsdal environment before, during and after the SN explosion (at the observed and model-predicted multiple image positions). Several oxygen, magnesium and iron lines measured in the spectra indicated that at all positions the ISM was optically thin, with a high degree of ionization and low metallicity, implying that the feedback of young and hot stars, possibly with the contribution of previous SNe, had already removed most of the gas from the environment before SN Refsdal exploded.

The same VLT observations, in particular the strong emission of an oxygen doublet, were used to explore the kinematics of the spiral galaxy hosting SN Refsdal. The estimated rotation curve showed a steeply rising part in the inner core, followed by a flat part out to large radii. The HST multi-colour photometry was modelled to measure the galaxy stellar mass and estimate that approximately one fifth of the total (dynamical) mass is in the form of stars (the rest being mainly gas and dark matter). In contrast to previous findings of highly turbulent discs in the young Universe, the kinematic analysis returned the picture of a regular, star-forming, spiral galaxy in the 4 billion year old Universe: a well-settled and rotation-dominated disc with lowly turbulent ionized gas orbiting the galaxy centre.

5 The future of lensed SNe

Refsdal’s vision was to measure the relative time delays of the multiple images of a single, lensed, time-variable source to infer the values of the cosmological parameters, in particular of the Hubble constant. Multiply-imaged SNe have eluded detection for 50 years, but distant quasars strongly lensed by foreground galaxies have been used for more than 20 years for time delay measurements and following cosmographic studies. They have provided accurate and precise estimates of cosmologically relevant quantities, complementary to those of more traditional probes, like the cosmic microwave background, Type Ia SNe, and baryonic acoustic oscillations. Despite rarer than strongly lensed quasars, strongly lensed SNe have the advantage of substantially simpler light curves and the possibility (for standard candle Type Ia SNe) of absolute magnification estimates.

Reasonably bright SNe multiply imaged by massive galaxy clusters are estimated to be on the order of a few per century. This would translate into very time-consuming, deep monitoring programmes of a large number of cluster fields in order to have a good probability of detecting at least one new SN Refsdal-like event per year. On-going and planned wide-field imaging surveys, like Pan-STARRS and the Large Synoptic Survey Telescope (LSST), together with the future Wide Field Infrared Survey Telescope (WFIRST) are expected to find more than a hundred SNe, strongly lensed by foreground galaxies. The first example of a Type Ia SN quadruply imaged by an intervening small lens galaxy has been recently announced. Compared to cluster-scale systems, galaxy-scale systems are easier to model, because of the less complex total mass distribution of a single lens galaxy, and have multiple-image time delays on the order of days or months (as opposed to years or decades). With an adequate investment in observational and modelling resources, it will be possible to develop time delays of lensed SNe into a competitive cosmological tool. We are not so far from turning Refsdal’s idea into reality!

Acknowledgments

Many of the results presented in this article have been obtained within medium-sized international collaborations. The author’s main publications would not have been possible without the great help of the CLASH-VLT and CLASH teams. The author thanks the ESO User Support group for their excellent work and acknowledges support by VILLUM FONDEN Young Investigator Programme through grant no. 10123 and “Programma per Giovani Ricercatori Rita Levi Montalcini (Bando 2013)”.