Nonuniform Expansion of the Youngest Galactic Supernova Remnant G1.9+0.3
K. J. Borkowski, S. P. Reynolds, D. A. Green, U. Hwang, R. Petre, K. Krishnamurthy, R. Willett
aa r X i v : . [ a s t r o - ph . H E ] J u l A CCEPTED FOR PUBLICATION IN A P J L
ETTERS
Preprint typeset using L A TEX style emulateapj v. 5/2/11
NONUNIFORM EXPANSION OF THE YOUNGEST GALACTIC SUPERNOVA REMNANT G1.9+0.3 K AZIMIERZ
J. B
ORKOWSKI , S TEPHEN
P. R
EYNOLDS , D AVID
A. G
REEN , U NA H WANG , R OBERT P ETRE , K ALYANI K RISHNAMURTHY , & R EBECCA W ILLETT Accepted for publication in ApJ Letters
ABSTRACTWe report measurements of X-ray expansion of the youngest Galactic supernova remnant, G1.9+0.3, usingChandra observations in 2007, 2009, and 2011. The measured rates strongly deviate from uniform expansion,decreasing radially by about 60% along the X-ray bright SE-NW axis from 0 . ± .
06% yr - to 0 . ± .
03% yr - . This corresponds to undecelerated ages of 120 – 190 yr, confirming the young age of G1.9+0.3,and implying a significant deceleration of the blast wave. The synchrotron-dominated X-ray emission brightensat a rate of 1 . ± .
4% yr - . We identify bright outer and inner rims with the blast wave and reverse shock,respectively. Sharp density gradients in either ejecta or ambient medium are required to produce the suddendeceleration of the reverse shock or the blast wave implied by the large spread in expansion ages. The blastwave could have been decelerated recently by an encounter with a modest density discontinuity in the ambientmedium, such as found at a wind termination shock, requiring strong mass loss in the progenitor. Alternatively,the reverse shock might have encountered an order-of-magnitude density discontinuity within the ejecta, suchas found in pulsating delayed-detonation Type Ia models. We demonstrate that the blast wave is much moredecelerated than the reverse shock in these models for remnants at ages similar to G1.9+0.3. Similar effects mayalso be produced by dense shells possibly associated with high-velocity features in Type Ia spectra. Accountingfor the asymmetry of G1.9+0.3 will require more realistic 3D Type Ia models. Subject headings:
ISM: individual objects (G1.9+0.3) — ISM: supernova remnants — X-rays: ISM INTRODUCTION
G1.9+0.3 is the remnant of the Galaxy’s most recent su-pernova (Reynolds et al. 2008, Paper I). Expansion between2007 and 2009 of about 1.6% measured with Chandra(Carlton et al. 2011, Paper V) gives an expansion (undecel-erated) age of about 160 yr, but the estimated mean expan-sion index m ( R ∝ t m ) of about m = 0 . ∼ - (Borkowski et al. 2010, Paper IV),consistent with the proper motions for a distance of order10 kpc (Roy & Pal 2014). For an assumed location near theGalactic Center ( d = 8 . - (Paper V). Observ-ing the rapid evolution of G1.9+0.3 in morphology and bright-ness can provide unprecedented information on the dynamics Department of Physics, North Carolina State University, Raleigh, NC27695-8202; [email protected] Cavendish Laboratory; 19 J.J. Thomson Ave., Cambridge CB3 0HE,UK Department of Astronomy, University of Maryland, College Park, MD20742 NASA/GSFC, Code 660, Greenbelt, MD 20771 Department of Electrical and Computer Engineering, Duke University,Durham, NC 27708 Department of Electrical and Computing Engineering, University ofWisconsin-Madison, Madison, WI 53706 of SN ejecta and on particle acceleration. OBSERVATIONS
Chandra observed G1.9+0.3 with the ACIS S3 chip onthree epochs: (I) 2007 February and March, (II) 2009 July,and (III) 2011 May and July (details in Papers I and III –VI). All observations were done in Very Faint Mode, andreprocessed with CIAO v4.6 and CALDB 4.5.9. The corre-sponding effective exposure times are 49.6 ks, 237 ks, and977 ks. Exposure-weighted time intervals between the deepEpoch III and shorter Epoch I and II observations are 4.274 yrand 1.861 yr. Alignment of observations at different epochsis performed simultaneously with expansion measurements.Images, 512 pixels in size, were extracted from the mergedevent files by binning event positions to half the ACIS pixelsize, so one image pixel is 0 . ′′ × . ′′ . We also extracteddata cubes, 512 ×
64 in size, using the same spatial pixel size,while spectral channels from 84 to 595 (1.2–8.7 keV energyrange) were binned by a factor of 8.X-ray spectra were extracted from individual rather thanmerged event files, and then summed (response files wereaveraged). Spectral analysis was performed with XSPECv12.8.1 (Arnaud 1996), using C-statistics (Cash 1979). Back-ground was modeled rather than subtracted. Spectra ofG1.9+0.3 were modeled with an absorbed power law, usingsolar abundances of Grevesse & Sauval (1998) in the phabs absorption model. Thermal emission contributes negligibly tobroadband fluxes. EXPANSION
We first measured the overall expansion using the methoddescribed in Paper V. Briefly, we smoothed the 2011Chandra data cube with the spectro-spatial method ofKrishnamurthy et al. (2010), and summed the smoothedcubes in the spectral dimension to arrive at smoothed images
10 20 30 40 50 600 0.002 0.004 0.006 0.008 0.01
Figure 1.
Top: 2011 Chandra image of G1.9+0.3. Scale is in counts perACIS pixel in the 1.2–8 keV energy range (image was smoothed with themultiscale partitioning method of Krishnamurthy et al. 2010). Bottom: 1.4GHz VLA radio image from 2008 December. Scale is in Jy beam - . Reso-lution 2 . ′′ × . ′′ . N is up and E is to the left. Intensities shown with thecubehelix color scheme of Green (2011). Pairs of regions chosen for expan-sion studies are overlaid: outer (green), middle (magenta), and inner (yellow)pairs along the SE-NW axis, and NS (red) pair. Image size 123 ′′ × ′′ . in the 1.2–8 keV energy range. (Softer X-rays are absorbedby the intervening interstellar medium, while the backgrounddominates at the highest energies.) We used the smoothedimage as a model for the surface brightness of G1.9+0.3 atEpoch III (Figure 1). This model image was background-subtracted then fit to the unsmoothed 1.2–8 keV images fromearlier epochs (i.e., shrunk to fit) using C-statistics. Sevenpoint sources within G1.9+0.3 were masked out. There arefour free parameters in this model: a physical scaling factor,a surface-brightness scaling factor, and expansion center co-ordinates. Independent fits to Epoch I and II images wereconsistent with the constant expansion. We then assumed thesame expansion rate while fitting jointly to the Epoch I and II images, but allowed for independent surface-brightness scal-ing factors and expansion center coordinates. Results of allmeasurements are listed in Table 1. The measured expansionrate is 0 . ± . - (all errors are 90% confidence),in reasonable agreement with our previous measurement of0 . ± . h m . s ± . s
004 (17 h m . s ± . s - ◦ ′ . ”85 ± . ”06 ( - ◦ ′ . ”96 ± . ”03) forEpoch I (II). A small (0 . ′′
17) but significant difference be-tween these centers suggests that the coordinates of the cor-responding reference observations ID6708 and ID10112 areslightly misaligned. The magnitude of this shift is consistentwith the Chandra external astrometric errors (mean error of0 . ′′
16; Rots 2009).The measured expansion rates strongly deviate from uni-form expansion (Table 1). Expansion rates increase inwardby about 60% along the bright SE-NW axis, ranging from0 . ± .
03% yr - for the outer ears to 0 . ± . - for the inner rims. The bright rims in the middle ex-pand slightly faster (0 . ± . - ) than the ears,but even this small difference is statistically highly signifi-cant. The brightness-weighted linear displacement is 0 . ′′ - for all three rims. The N–S expansion is intermediate(0 . ± .
04% yr - ) between the bright middle rims and theears, with the average displacement of 0 . ′′
23 yr - . We demon-strate this differential expansion for two representative pro-files in SE and NW based on the Epoch II observations (Fig-ures 2 and 3). As is seen most clearly in the close-up inserts,the expansion rate (green) that matches the bright middle rimsis too slow (too little shrinkage) for the inner rims, while thefaster expansion rate (red) that matches the inner rims is toofast for the middle and outer rims.Contributions to C-statistic along the profiles quantify thefit quality. Systematic deviations are present even for the best-fit models, perhaps due to spatial variation of the Chandrapoint-spread function, smaller-scale spatial variations in ex-pansion rate, or projection effects. Strong counting noise lim-its the accuracy of the expansion measurements. However,these profiles are merely illustrative; the results in Table 1 arederived from the expansion of the entire regions shown in Fig-ure 1. FLUX INCREASE
The spatially-integrated spectra of G1.9+0.3 from eachepoch were jointly fit, together with their background spectra,with an absorbed power law in the 1 – 9 keV energy range.XPANSION OF YOUNGEST GALACTIC SNR G1.9+0.3 3
Table 1
Expansion Rates and FluxesRegion Expansion Rate ˙ S ˙ S N H b Γ c F F F ˙ F e (% yr - ) (% yr - ) (10 cm - ) (10 - ergs cm - s - ) (% yr - )Total 0.589 -0.2 0.4 7.25 2.40 27.34 28.79 29.73 1.9(0.573, 0.605) (-0.7, 0.2) (-0.1, 0.9) (7.16, 7.34) (2.37, 2.43) (26.75, 27.93) (28.50, 29.08) (29.57, 29.99) (1.5, 2.3)Outer SE-NW pair 0.523 0.0 1.9 7.90 2.13 2.72 2.87 2.89 0.9(0.487, 0.560) (-1.5, 1.7) (0.2, 3.6) (7.60, 8.21) (2.04, 2.23) (2.61, 2.83) (2.81, 2.92) (2.86, 2.91) (-0.3, 2.2)Middle SE-NW pair 0.616 -0.4 1.0 7.58 2.34 9.16 9.75 10.05 1.9(0.592, 0.640) (-1.1, 0.4) (0.2, 1.8) (7.45, 7.72) (2.29, 2.38) (8.97, 9.35) (9.66, 9.84) (10.00, 10.09) (1.3, 2.5)Inner SE-NW pair 0.842 0.3 -0.3 7.02 2.38 3.56 3.66 3.91 2.8(0.783, 0.898) (-0.9, 1.6) (-1.6, 1.0) (6.82, 7.23) (2.31, 2.45) (3.45, 3.70) (3.60, 3.71) (3.88, 3.94) (1.8, 3.8)N-S pair 0.576 -0.2 -0.2 7.09 2.51 8.23 8.65 9.00 2.1(0.544, 0.609) (-1.0, 0.7) (-1.1, 0.6) (6.96, 7.23) (2.46, 2.56) (8.05, 8.41) (8.56, 8.74) (8.95, 9.05) (1.4, 2.8) Note . — Expansion rates and fluxes in odd rows, 90% confidence limits in even rows. a Surface brightness change. b Hydrogen column density. c Power-law photon index. d Absorbed flux in the 1–7 keV energy range. e Flux rate increase.
The power-law index and absorbing column density were as-sumed constant in time, but the fluxes at each epoch were leftfree. There is good agreement with previous measurementsfor Epoch I, but the newly-determined Epoch II flux is 2.5%larger, in disagreement with the previous measurement alsoreported in Paper V, presumably due to updates in the ACISS3 calibration that were applied to all three datasets.Spatially-integrated fluxes are clearly increasing with time(Table 1). A likelihood ratio test reveals that the linear fluxincrease is consistent with the individual flux measurements.The measured flux increase is 1 . ± .
4% yr - , in agreementwith the previous, less accurate value of 1 . ± .
0% yr - (Paper V).We also measured fluxes and rates of flux increase for theregion pairs shown in Figure 1, in the same way as for thespatially-integrated fluxes except that the background con-tribution was scaled down (rather than fit again) by the re-gion/total area ratios from the global fit. For each regionpair, the measured rate of flux increase is consistent with thespatially-integrated rate. There is an apparent trend in bright-ening rate with radius, but uncertainties are large, due to un-certainties in expanding the regions and to dust scattering ofemission from brighter regions (Paper III), as well as to largemeasurement errors. We conclude that evidence for spatialvariations in the rate of flux increase is weak, although itshould be the subject of future investigations. DISCUSSION
The nonuniform expansion we observe for our three pairsof regions can be rephrased as large differences in expansionage ( t exp ≡ ∆ t R / ∆ R ), in the sense that t exp is largest (great-est deceleration) for the slower-expanding outermost material(see Figure 1 and Table 1). The outermost ears, bright rimemission in the middle, and distinct interior rims all have mea-surably different expansion ages: 190 yr, 160 yr, and 120 yr.We define expansion indices m ≡ d ln r / d ln t so that a featureat radius r obeys r ∝ t m (note that d ln r / d ln t = d ln R / d ln t ,where R is the projection of the true radius r onto the planeof the sky). Then the true remnant age t = m fw t exp , where m fw is the forward-shock expansion index. Since we only have anestimate of 100 yr for the true remnant age (Paper V), we canonly determine relative m values.The large spread in expansion ages between the inner rimsand the ears implies a large deceleration of the forward shock, m fw . .
6, significantly stronger than expected ( m fw ∼ . m in these models typically varies only slightly (ifat all) from the reverse shock to the blast wave, and it in-creases instead of decreasing with radius (Chevalier 1982;Dwarkadas & Chevalier 1998, hereafter DC98). Either theejecta or the ambient medium density distribution (or both)must be very different from the slowly-varying density distri-butions considered so far. Various possibilities include sub-stantial clumping or sudden density jumps within the ejectaor the ambient medium.Fine-scale clumping within the ejecta is unlikely to explainfaster than expected expansion of the inner rims, however.Much like the outer rims, the inner rims consist of cohesiveand continuous filaments in contrast to the much more clumpynorth rim (Figure 1) where the ejecta emit most strongly inthermal X-rays (Paper VI). Another explanation involves arapid deceleration of the blast wave upon a recent encounterwith a moderate density jump in the ambient medium. In or-der to account for about 60% – 70% slower expansion of themiddle and outer rims relative to the inner rims (Table 1), afactor of 3 – 6 density jump is required (from eqs. 4 and 6 inBorkowski et al. 1997). A density jump of this magnitude isconsistent with a wind termination shock. If this were true,the SN progenitor must have been losing mass in a strongand asymmetric stellar wind. We estimate the wind parameter( ≡ ˙ M / v w , where ˙ M is the mass-loss rate and v w is the windspeed) at 4 × - M ⊙ yr - / - , using the analyticwind thin-shell solution for an exponential ejecta density dis-tribution from Paper V with the remnant’s radius of 2 pc andundecelerated age of 120 yr, and assuming a standard ther-monuclear explosion with kinetic energy 10 ergs and ejectedmass equal to the Chandrasekhar mass. Such a strong windfavors a single-degenerate progenitor (Hachisu et al. 1996).The deceleration is m = 0 .
88 in this model, suggesting that theexplosion occurred sometime in the first decade of the 20thcentury. In this scenario, it is difficult to understand the ori-gin of the strong north-south asymmetry seen at radio wave-lengths (Figure 1) and in the spatial distribution of thermalX-ray emission. As discussed in Paper VI, a strongly asym-metric Type Ia explosion provides the best explanation for this B r i g h t n e ss ( c o u n t s )
27 28 29 30 31 32010203040506070 ( arcseconds ) − − s i g n ( d a t a − m o d e l ) ∆ C Figure 2.
Top: SE profile from 2009 (blue) along position shown on 2009image (left inset), together with model profiles corresponding to best-fit ex-pansion rates of 0 .
62% yr - (green) and 0 .
84% yr - (red) for the middle andinner rims (see Figure 1 and Table 1). Right inset: Zoomed view of the innerrim. Bottom: Contribution to C-statistic. Horizontal scale is distance alongthe profile, measured from E to W. asymmetry.Unlike for smooth ejecta density profiles (such as result-ing from delayed-detonation explosions), it is possible to ob-tain less deceleration inward ( m fw < m rev ) in 1-D numeri-cal models for SNe Ia expanding into a uniform ambientmedium for ejecta profiles with substantial structure (e.g.,the PDDe model in Badenes et al. 2003). Prominent den-sity structures in the outer ejecta layers are present in thedeflagration models (e.g., the W7 model of Nomoto et al.1984), sub-Chandrasekhar explosions (DC98; Badenes et al.2003), and in pulsating delayed-detonation (PDD) models(e.g., Dessart et al. 2014, see their PDDEL1 model plotted inFigure 4). This density profile can be satisfactorily approxi-mated by a power law ρ ∝ r - n with n = 5 . v tr = 14 ,
800 km s - , and by the exponential model ofDC98 at lower velocities. At v tr , there is a large (11 .
4) densityjump.In order to investigate how a density jump affects speeds ofthe reverse and forward shocks, we performed 1-D hydrody- B r i g h t n e ss ( c o u n t s ) ( arcseconds ) − − − s i g n ( d a t a − m o d e l ) ∆ C Figure 3.
NW profile from 2009. See caption to Figure 2 for explanation. namical simulations with the VH-1 hydrocode (for a recentdescription of this code, see Warren & Blondin 2013) for thiscomposite power-law-exponential ejecta model, We discusshere results for a preshock density n of 0 . - , matchingthe preshock density found in a young Type Ia SNR 0509-67.5 in the Large Magellanic Cloud (Williams et al. 2011),but they can be scaled to any value of n (DC98). Prior to theimpact of the reverse shock with the density jump, the den-sity profile can be described by the self-similar solutions ofChevalier (1982) with m fw = m rev = ( n - / n = 0 .
48 (see thedensity profile at 50 yr in Figure 4). The reverse shock ar-rives at the density jump at t = 83( n / . - ) - / yr withvelocity v s = 3 v tr / n = 0 . v tr (in the frame of reference mov-ing with the ejecta), and then splits into transmitted and re-flected shocks. The transmitted shock is the decelerated re-verse shock (inward-facing). Its velocity is v t = ( β/δ ) / v s where β is the pressure enhancement that depends only onthe density jump δ and varies between 1 and 6 (see equa-tion 6 in Borkowski et al. 1997). In the rest frame of theexplosion, deceleration of the transmitted reverse shock be-comes m rev = 1 - β/δ ) / / n . With δ = 11 .
4, the overpressureis β = 2 .
76, and m rev increases to 0 .
75, in good agreementwith hydrodynamical simulations (Figure 4). The reflectedXPANSION OF YOUNGEST GALACTIC SNR G1.9+0.3 5 − − − − − − − D e n s i t y ( g c m − ) D e c e l e r a t i o n P a r a m e t e r m
50 100 150 200 250 300Time (yr)0.51.01.52.02.53.03.54.0 R a d i u s ( p c ) Figure 4.
Top: Density vs. radius in 1-D hydrodynamical simulations witha composite power-law-exponential ejecta model and the uniform ambi-ent medium with n = 0 . - at 50 yr (solid blue line), 100 yr (green),and 150 yr (magenta). Undecelerated (freely expanding) ejecta at 50 yrare also shown: composite model (dashed line), and the PDDEL1 model(Dessart et al. 2014) ( + signs). Bottom: Radii of the reverse and forwardshocks (in blue and green solid lines), and their deceleration parameters m (dashed lines). The reverse shock expansion is faster than the blast waveexpansion during a long ( ∼
100 yr) period of time. shock propagates first back into the shocked ejecta, and theninto the shocked ambient medium. At 100 yr, it has alreadypassed through the low-density contact discontinuity that sep-arates the shocked ejecta from the shocked ambient medium,and can be seen in Figure 4 as a small density discontinuitynear the contact discontinuity. The reflected shock strength-ens with time (see density profile at 150 yr), and eventuallymerges with the blast wave at 175 yr, resulting in an abrupt in-crease of m fw from 0.48 to 0.62. By this time, m rev has alreadydecreased to about the same value after a transient phase fol-lowing the sudden deceleration of the reverse shock. This de-crease continues in the subsequent evolution, while m fw staysabout constant, so that m fw > m rev .During an extended period of time comparable with ayoung Type Ia remnant’s age, PDD explosions show a morerapid expansion of the reverse shock than the forward shock(Figure 4). While this still might be a reasonable interpre-tation for G1.9+0.3, published PDD models fail to matchits properties in detail. Assuming the ejecta structure ofthe composite model just described and identifying the blastwave with the bright middle rims at r = 2 . d / . - , the inferred preshock density n and free expansion ejecta velocity at the reverse shockare 0 . d / . - . cm - and 18 , d / . - . At 8.5 kpc, n is several times higher than estimatedin Paper V, but the ejecta velocity is consistent with pre-vious estimates. G1.9+0.3 is dynamically too young, how-ever, for the reverse shock to have reached the density jumpat v tr = 14 ,
800 km s - . No published PDD model has v tr as high as 18 ,
000 km s - , with the highest (16 , - ) in the PDDa model (Badenes et al. 2003). Amongmodels considered by Badenes et al. (2003), only the sub-Chandrasekhar model SCH has higher v tr at 18 ,
500 km s - .Furthermore, sub-Chandrasekhar explosions produce large(up to several × - M ⊙ ) amounts of radioactive Ti (e.g.,Woosley & Kasen 2011), while the strength of the (possible) Sc line implies that at most ∼ - M ⊙ of Ti was expelledby the SN that produced G1.9+0.3 (Borkowski et al. 2013a).Only the outermost ( v & , d / . - ) ejectahave been shocked so far in G1.9+0.3 (velocities might be ashigh as v & ,
000 km s - if d ≥