Supernova Polarimetry with the VLT: Lessons from Asymmetry L

Supernova Polarimetry with the VLT: Lessons from Asymmetry L. WANG1, D. BAADE 2, P. HÖFLICH 3, J.C. WHEELER 3

1Lawrence Berkeley National Laboratory, California, USA 2European Southern Observatory, Garching, Germany 3Department of Astronomy and McDonald Observatory, The University of Texas at Austin, Austin, USA

1. Introduction Fritz Zwicky shortly after the discovery in this sort of supernova resulting from of the neutron in 1932. When the first the spin of the white dwarf, the motion Picture a 55-year-old telescope of neutron stars were discovered by of the orbit, a surrounding accretion modest two-metre aperture and the Jocelyn Bell they were manifested as disk, or the presence of the companion need to take multiple spectral expo- rapidly rotating pulsars with intense star. As for the case of core-collapse sures of the same object to elicit any magnetic fields. Pulsars have space supernovae, this was known, but there signal from the noise with an inexpen- velocities that average several hundred was no compelling observational rea- sive spectrograph jury rigged to do po- kilometres per second. This indicates son to consider departures from spher- larimetry, an effort requiring integra- that they are somehow “kicked” at birth ical symmetry. This, too, is changing. tions on a single object lasting a whole in a manner that requires a departure long winter night with a single observer from both spherical and up/down sym- 2. Polarization of Supernovae and no night assistant, if any transient metry. More recently, NTT and then target is available during the scheduled Hubble Space Telescope images of The question of the shape of super- time. Contrast that with ordering up Supernova 1987A in the Large Magel- novae has undergone a revolution in queue-scheduled target-of-opportunity lanic Cloud showed rings of gas that the last decade. The driving force has observations by a dedicated profes- had been ejected by the progenitor star been a new type of observation: meas- sional staff on an 8 metre telescope before it exploded. This means either urement of the polarization of the light with a state-of-the-art spectrograph and the progenitor star or its surroundings from supernovae. polarimeter. That is the leap that has possessed some sort of asymmetry. Light consists of oscillating electric occurred in our programme to obtain Further observations showed that the and magnetic fields. An ordinary beam spectropolarimetry of all accessible su- debris of the explosion were also asym- of light is a mix of photons with all ori- pernovae. The spectropolarimetry ob- metric. The supernova remnant Cas- entations equally present, a state that tained in a brief exposure on the VLT siopeia A shows signs of a jet and coun- leaves the light unpolarized on aver- with the FORS1 spectrograph is com- terjet that have punched holes in the age. Some processes of producing or parable to the total flux spectrum ob- expanding shell of debris and there are scattering light favour certain orienta- tained on those long nights on the 2.1- numerous other asymmetric supernova tions of the electric and magnetic fields m Struve Telescope at McDonald Ob- remnants. Each of these things has over others. One such process is the servatory where this programme be- been known. The question has been: reflection of light. When light scatters gan. While the data obtained at McDon- are they merely incidental or a vital clue through the expanding debris of a su- ald pointed the way to a revolution to how supernovae work? pernova, it retains information about in the way we think about super- The previous discussion pertained to the orientation of the scattering layers. novae, it is the quality of the data from core-collapse supernovae. These If the supernova is spherically symmet- the VLT that has led the programme to come in a variety of spectral classifica- ric, all orientations will be present flourish. tions, Type II, Type Ib or Type Ic, de- equally and will average out, so there Supernovae have been studied with pending on whether there is abundant will be no net polarization. If, however, modern scientific methods for nearly a hydrogen, helium or neither in the out- the gas shell is not round, a slight net century. During this time, it has been er layers. There is another kind of su- polarization will be imprinted on the traditional to assume that these cata- pernova known as Type Ia. These are light. strophic stellar explosions are, for all thought to be the result of the ther- Since we cannot spatially resolve the practical purposes, spherically sym- monuclear explosion of a white-dwarf average extragalactic supernova, po- metric. There were observational rea- star composed of carbon and oxygen. larization is the most powerful tool we sons for this. The Sun is essentially When the mass of the white dwarf have to judge the shape of the ejecta. spherically symmetric and most stars closely approaches the Chandrasekhar The method used is called spectropo- are thought to be. The assumption that limit of about 1.4 solar masses, the car- larimetry. This technique both spreads stars are round is quite reasonable. bon ignites. The resulting thermonu- the light out into its spectrum of colours Self-gravity will tend to pull any large clear explosion is thought to complete- and determines the net orientation of body into a sphere which is the mini- ly disrupt the star, leaving no compact the electric field at each wavelength. mum-energy configuration. There were remnant. This sort of explosion has re- This way both the overall shape of the also practical reasons. Theoretical ceived prominence recently because emitting region and the shape of re- study of stellar explosions has been dif- they have been the tool to discover the gions composed of particular chemical ficult enough even with the assumption accelerating Universe and the dark en- elements can be determined. We note of spherical symmetry. There has been ergy that drives the expansion. The that the effective spatial resolution at- no commanding observational need to progenitor white dwarf has long been tained by polarimetry of a supernova of abandon that simplifying assumption. treated as basically spherically sym- radius 1015 cm at 10 Mpc is 10 mi- Now there is. metric even though the popular model croarcsec. This is a factor of 100 better The evidence that supernovae may is that the explosion must take place in resolution than VLTI or other compara- depart a little or even drastically from a binary system where the white dwarf ble optical interferometer installations – spherical symmetry has been growing grows to the critical mass by accretion and at a tiny fraction of the cost. for years. Neutron stars were predicted of mass and, inevitably, angular mo- There were systematic and stimulat- to form and to power supernovae by mentum. There could be asymmetries ing observations of the polarization of

47 the light of SN 1987A that are still being studied and interpreted. Another event that was modestly well studied was the hydrogen-depleted event SN 1993J in M81. These two events just illustrated how poor the overall data base of supernova spectropolarimetry was. In 1994 we began a programme to obtain spectropolarimetry of as many supernovae as possible that were visible from McDonald Observatory. At the time, only a handful of events had been examined at all and there were virtually no statistics. The data reduction was tricky, if only because the intervening interstellar medium can impose a polarization signal that has nothing to do with the supernova. The data were also difficult to interpret. There are, in principle, many reasons why the light from a supernova could be polarized. The supernova could be aspherical, it could be spherical but have off-centre sources of light, or other matter in the vicinity could be asymmetrically distributed, blocking part of the scattering surface and yield- ing a net polarization signal from even a spherical surface. To make matters worse, the first few supernovae our group studied (and those in the previous sparse record like SN 1987A and SN 1993J) were classified as “peculiar” in some way, so we did not know whether we were seeing incidental pe- Figure 1: Polarimetry of the Type II SN 1999em on the Q-U plane. The wavelength of differ- culiarities or something truly significant. ent data points are colour encoded. The data points are rebinned to 100 Å for clarity. The 1999 Nov 8 data (clustered in the lower-right quadrant) are clearly separated from the 2000 As data accumulated, however, this Jan. 9 data (clustered in the upper-right quadrant), suggesting strong polarization evolution. uncertainty was removed, and signifi- The data points of both epochs fall roughly on the line denoted AB. Line AB defines the axis cant new insights were revealed. With of symmetry if the object is axially symmetric. Note that the blue and red dots are well sepa- more data and better statistics, we rated for each epoch with the blue points preferentially located at the lower-right of the data identified the first key trend. In 1996, we cluster of each observations. The circles are the upper limits to the interstellar polarization realized that the data were bi-modal. assuming E(B-V) toward the supernova to be 0.05 (inner circle) and 0.1 (outer circle). The Type Ia supernovae showed little or no approximate location of the component due to interstellar dust is shown as a solid circle. polarization signal (we will talk about some significant exceptions below). Supernovae thought to arise by core and allows us a view deeper inside. We this tendency to follow a fixed orienta- collapse in massive stars – Types II, Ib, found that even in a Type II supernova, tion in space. and Ic – were, by contrast, all signifi- the longer we watched and the deeper These observations of core-collapse cantly polarized. So far there has been inside we could see, the larger the po- supernovae taken together tell us that no exception. Every core-collapse su- larization became. This was illustrated the closer we see to the centre, the pernova for which we or other groups by our VLT observations of the classic larger is the asymmetry. The asymme- have obtained adequate data has been Type II supernova, SN 1999em, that try is not some incidental aspect of the substantially polarized. Core-collapse was characterized by an especially progenitor’s surrounding environment supernovae are definitely not spherical- long plateau, suggesting an especially that is causing the polarization, but ly symmetric. The question is, why? As massive outer hydrogen envelope. something deep in the heart of the ex- data continued to mount, new trends Observations were obtained around plosion. The implication is that the ex- appeared that give critical clues to ad- optical maximum and about 2 months plosion mechanism itself is asymmet- dress that question. past optical maximum. The polarization ric. Normal Type II supernovae explode rose from 0.1 per cent in the early ob- in red-giant stars that retain large, mas- servations to about 1 per cent in the lat- 3. The Cause of Asymmetry in sive outer envelopes of hydrogen. er data. Figure 1 shows the data for SN Core Collapse Types Ib and Ic are thought to happen 1999em on a “Q-U” plot where Q and U in stars that have already shed much or represent different projections of the The fact that the asymmetry of many all of their outer hydrogen layers, so polarization vector. The striking feature core-collapse supernovae is aligned in they allow us to see deeper into the of the SN 1999em data presented in one direction provides an important heart of the exploding star. We noticed this way is that all the data points fall on clue to the engine of the explosion. The that the Type II supernovae, with their a single line. This suggests a well-de- explosion mechanism must impose large blankets of hydrogen, showed rel- fined symmetry axis throughout the some axial symmetry. There must be atively less polarization. The Type Ib ejecta and independent of wavelength. some sustained bi-polar influence be- and Ic that allowed us to peer deeper The explosion of SN 1999em is asym- cause otherwise, as the supernova ex- into the exploding matter had higher metric, but aligned in a significant way. pands, pressure gradients will tend to polarization. In addition, as a given su- Figure 2 shows data from another Type heal irregularities and the ejecta will pernova expands, the debris thins out II event, SN 2001dh, that also shows tend to become more spherical rather

48 Figure 2: A total flux fast rotation and magnetic fields that spectrum (bottom) are intrinsic to pulsars. Progress in un- and polarization derstanding the origin of jets from mag- spectrum (top) of the netized disks around black holes Type II SN 2001dh, obtained on August makes us optimistic that similar 8, 2001 with the VLT. processes will form jets from a newborn In the top panel the pulsar in a stellar core. Recent work in locations of the Austin has shown that the magnetoro- prominent lines of tational instability may play a significant hydrogen and other role to produce strong magnetic fields elements can be in a fraction of a second after core identified by the bounce. These fields may, in turn, pro- emission peaks at mote the flow of energy up the rotation these wavelengths. To the short wave- axis by a combination of hoop stresses length side of the and other pressure anisotropies. emission peaks are Doppler blue-shifted 4. SN 2002ap – Not a absorption troughs Hypernova? corresponding to absorption in the ex- This work may also shed light on the panding atmosphere supernova/gamma-ray burst connec- of the supernova. The average polarization at the epoch of these data (a week or so after dis- tion. Several supernovae have been covery) was about 1 to 2 per cent. The peaks and troughs in the polarization spectrum prove that the polarization comes from the supernova, not the intervening interstellar matter. Some identified as “hypernovae” since they of the features correspond to those in the total flux spectrum, but others correspond to fea- show excessive velocities and luminos- tures that are only clearly revealed by the polarization. ity. The most famous example, SN 1998bw, was apparently associated with the gamma-ray burst of April 25, 1998. We have been concerned that than less. To do what we see, the alone explain the explosion or whether asymmetric explosions could mimic mechanism that drives the supernova it merely supplements the standard some of the effects of “hypernova” ac- must produce energy and momentum neutrino-driven explosion remains to be tivity as interpreted by spherical mod- asymmetrically from the start, then hold seen. If the jets up and down the sym- els. In particular, asymmetric models that special orientation long enough for metry axes are somewhat unequal, could give especially high velocities in its imprint to be permanently frozen into they might also account for the run- some directions, the directions of axial the expanding matter. Appropriate out- away velocities of pulsars. jets, and be brighter in some directions flows might be caused by MHD jets, by If such jets produce the asymmetries, than others because of the resulting accretion flow around the central neu- the most likely cause of the jets are the asymmetric flux distribution. This could tron star, by asymmetric neutrino emission, or by some combination of those mechanisms. The light we see from a supernova comes substantially from the decay of short-lived radioactive elements, nickel-56, cobalt-56 and later titani- um-44 in the debris. If this material is ejected in a bi-polar fashion, then the overall debris shell could be nearly spherical, while the asymmetric source of illumination leads to a net polarization. This mechanism may be at work in the early phases of Type II supernovae such as SN 1999em and SN 2001dh. By injecting jets of mass and energy up and down along a common axis deep within a model of an evolved star, we have shown that typical asymmetric configurations emerge. As shown in Figure 3, bow shocks form at the heads of the jets as they plow through the core, and a significant portion of the star’s matter bursts through the core along the jet axis. The bow shocks also drive “transverse” shocks sideways through the star. These shocks proceed away from the axis, converge toward the star’s equator, and collide in the equatorial plane. From there, matter is compressed and ejected in an equato- Figure 3: Three-dimensional computation of a jet-induced supernova explosion in a helium rial torus perpendicular to the jets. core. The frames show the density in the plane parallel to the jet axis that passes through the centre of computational domain. The time since the beginning of the simulation is given in the These models have shown that suffi- upper left corner of each frame. The scale expands in each frame with the left frame being ciently energetic jets can both cause about 1010 cm, the middle frame about 5 × 1010 cm and the right frame about 1011 cm. Note the explosion and imprint the observed the transverse shock waves that converge on the equator. From Khokhlov et al. 1999, asymmetries. Whether this process can Astrophys. J., 524, L107.

49 be especially true for Type Ic events tion in terms of the impact of a bi-polar where the lack of hydrogen and helium flow from the core that is stopped with- envelopes give a close view of the in the outer envelope of a carbon/oxy- asymmetries of the inner explosion. gen core. Although the symmetry axis A particular case in point is the recent remains fixed, as the photosphere re- Type Ic event SN 2002ap. This event treats by different amounts in different showed high velocities, but none of the directions due to the asymmetric veloc- other characteristics of a “hypernova,” ity flow and density distribution, geo- neither a strong relativistic radio source, metrical blocking effects in deeper, nor excessive brightness. High-quality Ca-rich layers can lead to a different spectropolarimetric data of SN 2002ap dominant axis in the Q-U plane. The were obtained with the VLT Melipal and features that characterize SN 2002ap, the FORS1 spectrograph at 3 epochs specifically its high velocity, can be ac- that correspond to –6, –2, and +1 days counted for in an asymmetric model for a V maximum of 9 Feb 2002. A sam- with a larger ejecta mass than the ple of the data is presented in Figure 4. well-studied Type Ic SN 19941 such The polarization spectra show three that the photosphere remains longer in distinct broad (~ 100 nm) features at ~ higher velocity material. 400, 550, and 750 nm that evolve in We conclude that the characteristics shape, amplitude and orientation in the of “hypernovae” may be the result of Q-U plane. The continuum polarization orientation effects in a mildly inhomo- grows from nearly zero to ~ 0.2 per geneous set of progenitors, rather than cent. The 750 nm feature is polarized at requiring an excessive total energy or a level 1 per cent. We identify the luminosity. In the analysis of asymmet- 550 and 750 nm features as Na I D and ric events with spherically symmetric OI λ 777.4 moving at about 20,000 km models, it is probably advisable to refer s–1. The blue feature may be Fe II. to “isotropic equivalent” energy, lumi- We interpret the polarization evolu- nosity, ejected mass, and nickel mass.

Figure 4: Spectropolarimetry of the Type Ic SN 2002ap on 2002 3 Feb, 6 days before V maximum. The Stokes parameters Q and U are rebinned into 15 Å bins. An interstellar polarization component is subtracted from the observed Stokes parameters so that the data points represent intrinsic polarization due to the supernova. The assumed interstellar polarization is shown as the solid dot in the Q–U plot (top panel). Without subtraction of the interstellar component, the origin Figure 5: Spectropolarimetry of the Type Ia SN 2001el on 2001 Sept. 26, 7 days before max- of the coordinates would be centred at this imum. The Stokes parameters Q and U are rebinned into 15 Å bins. An interstellar polariza- solid dot. The solid line represents the axis tion component is subtracted from the observed Stokes parameters so that the data points from the origin through the value of the ISP. represent intrinsic polarization due to the supernova. The assumed interstellar polarization is The dashed line illustrates the locus of the shown as a solid dot in the Q–U plot (panel a, upper left). Without subtraction of the inter- OI feature in the Q–U plot. The polarization stellar component, the origin of the coordinates would be centred at this solid dot. The straight spectra (middle panel) and polarized flux line illustrates the dominant axis shifted to the origin of the Q–U plot. The Q (panel b, upper (bottom panel) show conspicuously polar- right) and U (panel d, lower right) spectra show conspicuously polarized spectral features. ized spectral features corresponding to Fe II, The degree of polarization is shown as the thin line in panel c (lower left) with the flux spec- Na I D, and O I 777.4 nm. The wavelength trum (panel c, lower left, thick line) overplotted to show the correlations of the degree of po- colour code is presented at the bottom of the larization and the spectral features. The wavelength colour code is presented at the bottom top panel. of panel a.

50 5. Asymmetries in Type Ia due to the Ca II IR triplet at very high that all core-collapse supernovae are Supernovae velocities (20,000–26,000 kms–1). The substantially asymmetric. They explode 800 nm feature is distinct in velocity by means of bi-polar flow associated Most Type Ia supernovae are not sub- space from the photospheric Ca II IR with the newborn neutron stars. This stantially polarized at the epochs that triplet and has a significantly higher de- discovery may, in turn, give new in- have been observed. This suggests that, gree of polarization (= 0.7%), and dif- sights into more exotic jet-induced despite occurring in binary systems, the ferent polarization angle than the con- events like gamma-ray bursts. While explosions are essentially spherically tinuum. Taken together, these aspects most Type Ia supernovae have been symmetric. There are some interesting suggest that this high velocity calcium found to be little polarized, the number exceptions to this, however. SN 1999by is a kinematically distinct feature with of exceptions is growing. The asym- was one of the class of subluminous, the matter distributed in a filament, metries observed in Type Ia may fi- rapidly declining Type Ia events. It was torus, or array of “blobs” almost nally yield direct observational evi- substantially polarized and hence edge-on to the line of sight. This feature dence that they occur in binary sys- asymmetric in some way. We do not yet could thus be an important clue to the tems, as long assumed, and clues to the know whether this was characteristic of binary nature of SN Ia, perhaps associ- combustion mechanism. Understand- subluminous Type Ia, or whether SN ated with an accretion disk, or to the na- ing these asymmetries may be neces- 1999by was odd in this regard. ture of the thermonuclear burning, per- sary to properly interpret future data on In this context, it is important to obtain haps representing a stream of material cosmologically distant Type Ia’s. spectropolarimetry of “normal” Type Ia ballistically ejected from the site of the The authors are grateful to the Euro- supernovae. A step in this direction was deflagration to detonation transition. pean Southern Observatory for the gen- taken with our observations of the Type If modelled in terms of an oblate erous allocation of observing time. We Ia SN 2001el. High-quality spectropo- spheroid, the continuum polarization are also anxious to acknowledge that, larimetry of the SN 2001el was also ob- implies a minor to major axis ratio of contrary to the impression perhaps giv- tained with VLT Melipal and FORS1 at 5 around 0.9 if seen equator-on; this lev- en in the Introduction, requests for epochs. Some of these data are shown el of asymmetry would produce an ab- service-mode observations with the in Figure 5. The spectra a week before solute luminosity dispersion of about 0.1 VLT are much different than orders to a and around maximum indicate photos- mag when viewed at different view- pizza home-delivery service: The Para- pheric expansion velocities of about ing angles. If typical for SNe Ia, this nal Science Operations staff and the 10,000 km s–1. Prior to optical maxi- would create an RMS scatter of sever- User Support Group in Garching have mum, the linear polarization of the con- al hundredths of a magnitude around gone to considerable effort to augment tinuum was 0.2–0.3% with a constant the mean brightness-decline relation. our proposal with their full range of position angle, showing that SN 2001el This scatter might have implications for expertise. We recognize that accom- has a well-defined axis of symmetry. the high precision measurements re- modating our target-of-opportunity ob- The polarization was nearly unde- quired to determine the cosmological servations in an already busy observing tectable a week after optical maximum. equation of state of the “dark energy.” and work schedule often poses a spe- The spectra of SN 2001el are similar cial extra challenge. Only this sym- to those of the normally-bright SN 6. Conclusions biosis enables the ongoing success of 1994D with the exception of a strong this project. We are especially grate- double-troughed absorption feature The acquisition of systematic super- ful for that. This work was supported in seen around 800 nm (FWHM about 22 nova polarization data has led to re- part by NASA Grant NAG5-7937 to PAH nm). The 800 nm feature is probably markable new insights. It seems likely and NSF Grant AST 0098644 to JCW.

OTHER ASTRONOMICAL NEWS An Exciting Working Session on Cataclysmic Variables at ESO/Santiago E. MASON (ESO/Chile, fellow) and S. HOWELL (ESO/Chile, visiting scientist) An intensive working session on We hope that the wealth of ideas and etry and (v) the CVs accretion disks and Cataclysmic Variables (CVs) was held projects discussed during the working current analysis of their emission lines. at ESO/Santiago on August 14, 2002. session, will trigger regular CV work- The afternoon talks were more The workshop was organized on the shops here in Chile, possibly involving specifically focused on the observation occasion of the presence in Santiago a larger number of participants and in- of particular objects or on some as- of Dr. S. Howell, from the Planetary vited speakers. The workshop was or- pects of theoretical modelling. Science Institute in Tucson, thanks to ganized in a morning review session on The participants really benefited from the ESO/Chile visiting scientist pro- CVs, both on observations and on the- being in a fairly small but highly moti- gramme. ory, and an afternoon discussion ses- vated group and could present, dis- The goal of the workshop was to sion. cuss, and confront various problems gather all astronomers in Chile working Reviews were about: (i) the photo- and results of their current research on CVs, for exchanges and fruitful dis- metric behaviour of dwarf novae (DNs) programmes. In particular, the discus- cussions. The participants were from during cycles and super-cycles, (ii) the sion of unsolved problems turned out to the University of Concepción and from spectroscopic characteristics of CVs in be important and fruitful as it triggered ESO/Chile, and we could also welcome the wavelength range UV-IR, (iii) the evo- the submission of new proposals (on Dr. N. Vogt, from Heidelberg, who had lution of CVs – theory vs. observations, ESO telescopes!), as well as the devel- organized the first workshop on CVs (iv) radial velocity measurements as a opment of new research projects and ever held in Chile (Viña del Mar, 1992). diagnostic for the binary system geom- collaborations.