The pre-shock gas of SN1006 from HST/ACS observations J.C. Raymond1, K.E. Korreck1, Q.C. Sedlacek1, W. P. Blair 2, P. 7 Ghavamian2, R. Sankrit3 0 0 2 n a J 0 ABSTRACT 1 We derive the pre-shock density and scale length along the line of sight for 1 the collisionless shock from a deep HST image that resolves the Hα filament in v 1 SN1006 and updated model calculations. The very deep ACS high-resolution 1 image of the Balmer line filament in the northwest (NW) quadrant shows that 3 1 0.25 ≤ n0 ≤ 0.4 cm−3 and that the scale along the line of sight is about 0 2×1018 cm, while bright features within the filament correspond to ripples with 7 radii of curvature less than 1/10 that size. The derived densities are within the 0 / broad range of earlier density estimates, and they agree well with the ionization h p time scale derived from the Chandra X-ray spectrum of a region just behind the - o opticalfilament. ThisprovidesatestforwidelyusedmodelsoftheX-rayemission r from SNR shocks. The scale and amplitude of the ripples are consistent with t s expectations for a shock propagating though interstellar gas with ∼ 20% density a : fluctuations on parsec scales as expected from studies of interstellar turbulence. v i One bulge in the filament corresponds to a knot of ejecta overtaking the blast X wave, however. The interaction results from the rapid deceleration of the blast r a wave as it encounters an interstellar cloud. Subject headings: ISM:individual(SN1006)–supernova remnants–shock waves– optical:ISM 1. Introduction 1Harvard-Smithsonian Center for Astro- physics, 60 Garden Street, Cambridge, MA SN1006(G327.6+14.6)isoneofthebest 02138 SNRs for studying the physics of colli- 2 The Johns Hopkins University, Baltimore, sionless astrophysical shocks, in particu- MD lar the acceleration of non-thermal parti- 3Space Sciences Laboratory,University of Cal- cles. It is a nearby Type Ia supernova ifornia, Berkeley, CA remnant at a distance of 2.1 kpc (the dis- tance we assume throughout) with a diam- 1 eter of ∼ 18 pc (Winkler, Gupta & Long nent profile (Chevalier & Raymond 1978). 2003) and a shock speed in the 2500 - The broad component is due to charge −1 2900kms range(Ghavamian et al.2002; exchange between neutrals and protons, Heng & McCray 2006). The remnant has which produces a population of neutrals a high Galactic latitude and modest fore- at nearly the post-shock proton tempera- ground reddening, E(B-V)=0.11 ± 0.02 ture. The narrow component is produced (Schweizer & Middleditch 1980). when cold ambient neutrals pass through SN1006 has been observed at radio the shock and emit line radiation before (Reynolds & Gilmore1993;Moffett, Goss & Reyncohldarsge transfer or ionization occurs. The 1993), optical (Ghavamian et al. 2002; ratio of the broad to narrow flux is sensi- Kirshner, Winkler & Chevalier1987;Smith et al.tive to the electron and ion temperatures. 1991),ultraviolet(Raymond, Blair & Long The FWHM of Hα broad component is 1995; Korreck et al. 2004) and X-ray 2290 ± 80 km s−1, and models imply a (Koyama et al.1995;Winkler, Gupta & Long shock speed of vshock = 2890±100 km s−1 2003; Long et al. 2003; Bamba et al. 2003; for a shock with little electron-ion equi- Dyer et al.2004)wavelengths. PureBalmer libration (Ghavamian et al. 2002). How- line filaments were found in the optical by ever, new models incorporating some addi- van den Bergh (1976). In the radio and tionalphysicsobtainalowershockspeedof X-ray, the remnant has a limb-brightened 2509±111 km s−1 (Heng & McCray 2006). shell structure with cylindrical symme- The pre-shock density, n0, is an im- try around a SE to NW axis probably portant parameter for understanding the aligned with the ambient galactic mag- evolution of SN1006, as well as for in- netic field (Reynolds 1996; Jones & Pye terpreting the X-ray spectra in terms 1989). The NE shock front of SN1006 of ionization time scale and determin- shows strong non-thermal X-ray emission ing the relative contributions of shocked (Koyama et al. 1995; Dyer et al. 2004) ISM and SN ejecta to the X-ray emis- while the NW shock shows very little sion. The density could also be im- non-thermal emission at radio or X-ray portant for attempts to understand the wavelengths. On the other hand, the pre- high ratio of non-thermal to thermal X- shock density is several times higher in the ray emission in this SNR. Estimates of NW than the NE (Korreck et al. 2004). n0 cover a wide range, from 0.05-0.1 −3 Knots of X-ray emission from shocked SN cm based on the global X-ray emission ejecta are scattered through the interior of (Hamilton, Sarazin & Szymkowiak 1986) −3 the remnant (Long et al. 2003; Vink et al. to 1 cm based on interpreting the scale 2003). over which the X-rays brighten as the A Balmer line filament defines the blast length scale for ionization of the shocked wave inthenorthwest quadrant ofSN1006, plasma (Winkler & Long 1997). A related and the Hα profile provides diagnostics for parameter is the length scale along the the shock speed and ion-electron thermal line of sight. The ripples in the SNR equilibration. The Hα emission from a blast wave could in principle result from Balmer-dominatedshockhasatwocompo- density inhomogeneities in the ambient 2 medium or from knots of SN ejecta over- α2000 =15h 2m 19.02s, δ2000 =-41o 44′ 48.4′′ taking the blast wave. ISM density fluc- on 9 orbits from 15-17 February 2006. Ex- tuations have been inferred for a section posures withdurations of2,746.0s, 2,848.0 of the non-radiative shock in the Cygnus s, and2,828.0swereobtainedforeach sub- Loop (Raymond 2003), while an ejecta set of 3 orbits, for a combined 25,266 sec- knot is clearly the cause of one bulge in ond exposure. The exposures were taken the Hα filament of SN1006 (Long et al. with the F658N H-alpha filter, which has a 2003; Vink et al. 2003). The issue is im- flat response at wavelengths above about portant for estimating the amplitude of in- 6558˚A, drops to half the peak transmis- terstellar turbulence on sub-parsec scales sion at about 6548 ˚A and 20% of the peak and for the interpretation of the distance transmission at 6540 ˚A. This means that it between the blast wave and the reverse passes the narrow component, all the red shock in terms of particle acceleration wing of the broad component, and about (Warren et al. 2005). half theblue wing of thebroadcomponent, In this paper we use an Hα image ob- or about 90% of the Hα emission. tained with the ACS imager on the Hub- The image was centered on the position ble Space Telescope to determine the pre- where Ghavamian et al. (2002) obtained shock density and the length scale along a low dispersion spectrum and Sollerman the line of sight. This is possible because et al. (2003) obtained a high dispersion the ACS images resolve the thickness of spectrum, so we are able to use the shock the narrow zone behind the shock where speed, neutral fraction and electron-ion hydrogen atoms are excited and ionized, equilibration derived from those observa- and that thickness scales inversely as the tions. The position was also observed in pre-shock density. We have computed new the ultraviolet with the Hopkins Ultra- models of the Hα emissivity as a function violet Telescope (Raymond, Blair & Long of distance behind the shock taking into 1995; Laming et al. 1996) and with FUSE account recent results by Heng & McCray (Korreck et al. 2004), so that we can use (2006). We discuss the observations in the information derived in those papers the next section, then compare with mod- about ion-ionthermal equilibration to con- els to derive the shock parameters. Sec- strain model parameters. The Hα filament tion 4 provides a limit on the brightness of also defines the outer boundary of posi- any shock precursor, compares the densi- tion NW-1 in the X-ray spectral analysis ties with other estimates and discusses the of Long et al. (2003), who found the spec- implications for other analyses of SN1006 trum to be consistent with thermal emis- observations. Section 5 summarizes the sion from shocked interstellar gas. conclusions. The raw data were combined and re- duced with the standard ACS calibration 2. Observations pipeline (CALACS), involving bias/dark- current subtraction, flat-fielding, image The Hubble Space Telescope’s ACS combination, and cosmic-ray rejection Wide Field Camera (WFC) imaged a full (Sirianni et. al. 2005), then ’drizzled’ 2 field of 202 x 202 arcsec at coordinates 3 (Fruchter and Hook 2002) onto a 0.03 arc- proper motion analysis. Figure 2 shows sec pixel scale with the task ’multidrizzle’ the boxes used to extract spatial profiles, to correct for geometric distortion and im- starting with profile 8 in the upper left prove the sampling of the point spread and ending with profile 29 in the lower function (Koekemoer et al. 2002). right. The profiles were extracted using ′′ ′′ Figure 1 shows the HST image overlaid boxes4.5 wideand9 long. Their positions on the Chandra X-ray image. The mor- and position angles are shown in Table 1. phology of the filament is clearly that of a Theprofilesextractedweresimilartothose rippled sheet seen edge-on, with the bright usedby(Winkler, Gupta & Long2003)for rims corresponding to tangencies to the measuring the proper motion of the fila- line of sight (Hester 1987). For the present ment; here, however, the goal was to de- purposes we are interested in the simplest termine the width, not the position, of the tangencies, since those are amenable to filament. modeling. The bulge near the SW cor- 3. Analysis and Results ner of the image is morphologically sim- ilar to a larger, brighter region farther to To interpret the images we compute the SW where a clump of ejecta is overtak- models of the Hα brightness behind a ingtheshock(Long et al.2003;Vink et al. curved shock, convolve the model intensity 2003), but it shows only a slight X-ray en- distribution with the ACS point spread hancement. The shock morphology in the function, and compare the models to the bulge is more complex, so this paper con- observations. In general, the pre-shock centrates on the smoother regions of the density controls the brightness drop off be- filament. hind the peak, because the thickness of the IDL was used to extract and plot the emission region is inversely proportionalto curve of the shock front spanning the driz- the density. The radius of curvature of the zled image and to find the approximate shock determines the fall off ahead of the direction perpendicular to the shock. At peak for the concave outward filaments each of 40 positions we extracted the spa- that we model here. The absolute inten- tial profiles across the shock for a range sity scales approximately as n2f R1/2, 0 neut of angles near the initial estimate and se- where n0 is the pre-shock density, fneut is lected the the profile showing the narrow- the pre-shock neutral fraction, and R is est Hα peak as the one closest to the shock the radius of curvature of the shock. normal direction. Many of the profiles suf- fer from low signal to noise or from com- 3.1. Model Calculations plexity due to several tangencies to the We start with model calculations simi- line of sight. We have selected 8 profiles lar to those of Laming et al. (1996). Fig- with bright, simple Hα peaks for further ure 3 shows the Hα emissivity as a func- analysis. In particular, we model sections tion of distance behind the shock for a of the trailing edge of the filament in re- −1 2900 km s shock model with pre-shock gions corresponding to sections F, G and −3 H of the Winkler, Gupta & Long (2003) density n0 = 0.25 cm , a neutral frac- tion of 0.1, and a ratio T /T = 0.05 e p 4 at the shock front. The model follows to the value 2890±100 km s−1 found by neutral hydrogen as it passes through the Ghavamian et al. Both values apply to shock and undergoes collisional excitation the case of little ion-electron thermal equi- and ionization by protons and electrons as libration in the shock. Strong equilibra- wellaschargetransferwithpost-shockpro- tion is ruled out by the X-ray spectrum tons. The atomic rates are described by of Long et al. (2003) and by the combi- Laming et al. (1996), and Coulomb colli- nation of broad to narrow intensity ra- sions slowly transfer energy from the ions tio given by Ghavamian et al. (2002) and to the electrons. The radiative transfer in- the Lyβ radiative transfer calculations of volved in the conversion of Lyβ photons Laming et al. (1996). The smaller shock to Hα photons, important for the narrow speed of Heng & McCray (2006) would component, is described in Laming et al. decrease the distance scale of the Hα (1996). A fraction of the Hα arises from emission behind the shock. However, it converted Lyβ photons, andthoseHαpho- would also imply a smaller distance for tons are produced over a scale of about 1 SN1006 based the combination of the Lyβ mean free path. That scale is about proper motion and the and shock speed ′′ 0.1 , which is small enough compared to (Winkler, Gupta & Long 2003), so the an- the observed filament widths that we ig- gular width of the emission region would nore it. be unchanged to first order. Evidence for a shock precursor has been There is another important implica- reported for a number of non-radiative tion of the work of Heng & McCray (2006) shocks from low ionization emission lines for the present study. The rapid drop in (Hester, Raymond & Blair1994;Sollerman et al. charge transfer cross section with increas- 2003), from the velocity widths of their ing velocity means that neutrals are more narrow components (Smith et al. 1994; likely to undergo charge transfer with pro- Hester, Raymond & Blair 1994), and from tons moving away from the shock than the spatial distribution of the narrow com- with protons moving toward the shock. ponent emission (Lee et al. 2006). The While the velocity dependence of the cross narrow component in SN1006 shows no section was included in earlier models, the broadening beyond that expected for the anisotropy of the resulting H I velocity dis- ambient ISM (Sollerman et al. 2003), so tribution was not. Thus the earlier models we do not include any precursor emission implicitly assumed that the broad compo- in the models. nent neutrals move away from the shock Recently, Heng & McCray (2006) inves- at Vs/4, the same speed as the post-shock tigated the consequences of the sharp de- protons. By integrating the product of ve- cline of the the charge transfer cross sec- locity times charge transfer cross section, tion at speeds above about 2000 km s−1 σcxv (Barnett 1990), over Maxwellian dis- for the Balmer line profiles. They esti- tributions at the post shock temperatures, mated a shock speed of 2509±111 km s−1 we find that after 1 charge transfer event from the Balmer line profile presented the average neutral is moving away from −1 by Ghavamian et al. (2002), as opposed the shock at 1500 km s behind a 3000 5 −1 −1 km s shock, rather than the 750 km s count for the pixel sampling in cuts made −1 ◦ of the ionized gas. For a 2500 km s at about 45 to the rows and columns −1 shock, the neutrals move at 1100 km s of the detector, we measured the Gaus- −1 instead of 625 km s . Thus for the inter- sian widths of stars near the filament (2.31 esting range of shock speeds, the neutrals pixel FWHM), then placed a dense series ◦ move away from the shock twice as fast as of circular Gaussians along a 45 line, ex- is assumed in the model shown in Figure tracted the profile perpendicular to that 3. The relative velocity is much smaller line, and measured a slightly broader 2.43 for the second charge transfer event, so pixel width. This Gaussian is convolved after two charge transfers the neutrals with the models for comparison with the have a speed closer to that of the down- observed spatial profiles. stream plasma. Kevin Heng (2007, private Figure 4 is a schematic diagram of the communication) has provided the average geometry we imagine in a plane that in- downstream speed of the broadcomponent cludes the line of sight (dashed line) and a neutrals computed by the Heng & McCray radial vector from the center of the SNR. (2006) model code. For the 2500 to 3000 The light line shows a large scale ripple in −1 km s velocity range it is 1.32 times the the shock front, and we have superposed a plasma speed. Therefore, for comparison smaller scale ripple to obtain the shape of to observations we stretch the spatial scale theheavier curvethatisshadedtowardthe of the broad component emission shown inside of SN1006. The shading indicates in Figure 3 by a factor of 1.32. Particle the Hα emissivity, which fades gradually conservation implies that the neutral den- from the shock front towards the inside of sity is decreased by the same factor, so the the SNR. The trace in the lower right indi- broad component emissivity at each point cates the Hα brightness obtained for a cut is reduced by a factor of 1.32. Finally, acrossthe tangent pointof theshock byin- for comparison with the observations we tegrating the Hα emissivity along different multiply the broad component emission lines of sight. by 0.75 to account for the drop of sensi- tivity in the blue wing resulting from the 3.2. Comparison of Models and Ob- transmission of the F658N filter. servations To model the geometry of the filament, Figure 5 compares a grid of models we assume a shock surface that is concave with different n0 and R to the spatial outward with radius of curvature R. The profile at position 10. The sharp spikes Hα emissivity from the planar model ex- ahead of the Hα peak are faint stars in tends behind the shock at each point. Nu- the extraction region. The models have merical integration then gives the bright- been scaled to match the peak Hα bright- ness as a function of position relative to ness of the filament by simply adjusting the tangent point of the shock. For com- the neutral fraction. Neutral fractions parison with the observations, we convolve above 1 are obviously unphysical, so mod- the model emissivity with the ACS point els with low n0 and small R are ruled out. spread function. In order to properly ac- Ghavamian et al. (2002) estimated a pre- 6 shockneutralfractionof0.1fromtheratios total brightness. Thus the upper limit on of He I and He II lines to Hα, so we take n0 and the lower limit on R are less secure the permitted range of neutral fractions to than in the case of position 10. be 0.05 ≤ fneut ≤ 0.2. Figure 5 shows Table 1 presents the ranges of n0 and R −3 thatmodelswithn0 belowabout0.30cm derived from other positions along the fila- fall off too slowly behind the shock, while ment. We did not attempt to model other −3 the models with n0 above about 0.35 cm positions where the emission peak is faint give too sharp a peak. The R=5 × 1016 or where complex morphology indicates a cm model comes closest to matching the more complex structure than can be ap- observed profile. Models with smaller R proximated by a simple curved sheet. The predict Hα emission fainter than observed. profiles of positions 8 through 11 are qual- We conclude that 0.3 ≤ n0 ≤ 0.35 and itatively similar to that of position 10, and 1016.5 ≤ R < 1016.8 cm. those of positions 26 through 29 are like Figure 6 shows the analogous plots for that of position 28. position 28. The peaks in this section of thefilament arebothbrighter andbroader, 4. Discussion with a fairly constant Hα intensity ahead 4.1. Pre-shock density of the brightness peak. None of the mod- els match exactly, probably because the −3 The densities near 0.3 cm that we de- rippled sheet does not follow the assumed rive fall in the middle of the range of previ- shape of an arc of a circle. From Figure 6 ousvalues. Hamilton, Sarazin & Szymkowiak we conclude that n0 must be greater than (1986) estimated 0.05-0.1 cm−3 from the −3 0.25 cm to avoid a long tail toward the global X-ray spectrum. Since the pre- inside of the remnant, but that densities shock density in the NE, and quite likely above 0.4 produce too sharp a peak. R the rest of the remnant, is about 2.5 must be less than 2 × 1017 cm, because (Long et al. 2003) to 4 (Korreck et al. larger radii of curvature predict a shoul- 2004) times smaller than in the region we der on the outer side of the spatial pro- observed, theHamilton, Sarazin & Szymkowiak file that exceeds the observations, while (1986) estimate is consistent with our re- R less than 0.7 × 1017 cm requires unac- sults, even though most of the X-ray emis- ceptably high values of f to match the neut sion was subsequently shown to be non- brightness. We conclude that the accept- thermal in nature. Winkler & Long (1997) able ranges are 0.25 ≤ n0 ≤ 0.35 and obtained a pre-shock density near 1.0 from 0.7×1017 ≤ R ≤ 1.5×1017 cm. Wenote, the spatial profile of the X-ray emission, however, that the agreement between the but Long et al. (2003) found n0 ≃ 0.25 model spatial profiles and the observations from the value of n t obtained from the e is not as good as at position 10. Based on Chandra spectrum of their region NW-1, Figure 2 , it seems possible that the bright which lies immediately behind the region filament at position 28 contains more than of the Balmer line filament we observed. one tangency to the line of sight, broad- TheChandra spectrum showed solarabun- ening the spatial profile and increasing the dances, which is consistent with the inter- 7 pretation of shocked interstellar gas, and isfarenoughfromtheplanethattheveloc- the temperature of 0.7 keV is consistent itycouldeasilydiffer. Dubner et al.(2002) with very inefficient thermal equilibration estimate a neutral hydrogen density of 0.5 −3 between ions and electrons. Thus the cm , which with the low neutral fraction present results agree well with the results from Ghavamian et al. (2002) would im- −3 of shock wave models of the X-ray spec- ply a total density of 5 cm , well above tra and confirm the parameters derived. the values we derive. We conclude that However, as Long et al. (2003) point out, either SN1006 is not interacting with the the shock models did not provide an ac- cloud identified by Dubner et al., or that ceptable χ-squared fit to the data, so some the shock has only reached the low density aspect of the physics remains to be under- outskirts of the cloud. stood. In particular, with n0 as an inde- pendently measured quantity rather than 4.2. Length scales a free parameter, reanalysis of the Chan- The length scale given by the radius of dra spectrum might be able to place better curvature is considerably shorter than the limits on the non-thermal emission in the length of the filament itself. The more or NW part of SN1006. less straight portion of the Hα filament AnotherdensityestimatefortheNWre- ′ extends for perhaps 7, or about 4 pc, gionofSN1006wasobtainedbyLaming et al. roughly 100 times typical radius of curva- (1996), who found that the relative inten- ture derived for the ripples. The actual sities of the UV lines could be explained line of sight scale of the filament is sev- if only about half the O VI emission fell eral times that of the bright rim at the within the aperture of the HUT telescope. trailing edge of the filament. The region This would require an ionization length for ahead of the trailing bright rim at position theOVIof about10′′, orn0 ∼ 0.04cm−3. 28 has a nearly constant surface brightness This interpretation is not consistent with of 2 − 3×10−5 photons cm−2 s−1 ′′−2. For thevalues ofn0 derived forthesame region n0 =0.3,fneut =0.1andVs =2900kms−1, from the HST image, so we conclude that this implies an angle of 6 to 10 degrees be- there is a problem either with the models tween the shock and the line of sight. The of Laming et al. (1996), or perhaps more roughly constant intensity region is about likely with the reddening correction in the 2.8×1017 cm in radial exent, so the range far UV. of angles implies a depth along the line of A density estimate that is independent sight between 1.5 and 2.3 × 1018 cm, or of SN1006 itself comes from H I 21 cm about a tenth the scale of the filament in observations by Dubner et al. (2002). An the plane of the sky. The 1018 cm scale H I feature with a column density of 7 × corresponds to the smoother curve in the 20 −2 −1 10 cm atavelocityof-6kms liesjust schematic diagram in Figure 4, while the outside the NW rim of SN1006. While the 1017 cm scale derived from the Hα bright- velocity does not correspond to the Galac- ness profiles corresponds to the smaller ticrotationvalueatthedistancetoSN1006 scale ripple superposed to obtain the shock (−25 ≤ V ≤ −16 km s−1), the SNR geometry shown in the schematic. LSR 8 Even the larger scale inferred for the di- scale length of the ripples to density fluc- rection along the line of sight is several tuations in the interstellar gas, we can times smaller than the length of the fila- estimate the amplitude of the density ment in the plane of the sky. Very faint fluctuation from the amplitude of the Hα emission, which is not apparent in Fig- ripple. The amplitude of the ripple is ure 2, is seen in the very deep Hα image of about one tenth the wavelength, and that Winkler, Gupta & Long (2003) extending should be about equal to δV/V. For con- 2 out ahead of the bright filament (their Fig- stant ram pressure, nV , this requires ure 5). This is probably a shock in lower density fluctuations of about 20%. As density gas that may be farther from tan- was found for similar ripples in the blast gency with the line of sight, and it prob- wave of the Cygnus Loop, this agrees ably extends for a distance comparable to reasonably well with the expectations the length of the filament. from the spectrum of interstellar turbu- Figure 2 shows that ripples on the scale lence (Minter & Spangler 1997; Raymond of 1018 cm can be seen in the direction 2003). However, the smaller scale ripples 17 along the filament, for instance between revealed by the 10 cm radius of curva- thetwoboxeswhereradialprofileswereex- ture also seem to have amplitudes of order tracted. Ripples on the 1017 cm scale are 1/10 the wavelength, and a Kolmogorov not apparent, perhaps because a 1017 cm spectrum of density fluctuations from the ripple with 10% amplitude would be only turbulent cascadewould leadoneto expect afewresolutionelements radiallyinFigure a smaller amplitude (Minter & Spangler 2. However, such small scale rippling un- 1997). Beresnyak & Lazarian (2006) show doubtedly contributes to the widths of the that the spectrum of density fluctuations Hα brightness peaks and prevents them may be considerably flatter than the Kol- from being as sharp as the model peaks, mogorov spectrum. as is particularly apparent in the position One feature in the filament is probably 28 profile. It is also quite likely that the not due to density fluctuations in the ISM, ripples are not isotropic. If the magnetic however. The bulge near the western edge field lies near the plane of the sky in the of the filament in Figure 2 is very simi- NE-SW direction, then density structures lar to one farther west along the filament would be elongated in that direction and that coincides closely with a bright knot smaller scales would appear along the line of X-ray emission with enhanced elemental of sight. There is some evidence that the abundances (Long et al. 2003; Vink et al. field does lie in this direction based on the 2003). The X-ray knot (position NW- cap-like morphology of the non-thermal X- 2 of Long et al. (2003)) is clearly a knot ray emission (Willingale et al. 1996), but of ejecta overtaking the blastwave. The there is also a suggestion that the field lies bulge at the SW corner of the ACS im- in the SE-NW direction based on the rel- age is probably a similar structure with ative temperatures of protons and oxygen lower X-ray contrast due to its smaller size ions (Korreck et al. 2004). and perhaps differences in density, ioniza- If we attribute the ∼ 2 × 1018 cm tiontimescaleandabundances. According 9 to hydrodynamic simulations of Type Ia 4.3. Shock Precursor SNRs, ejecta knots form at the Rayleigh- Shock wave precursors have been in- Taylor unstable contact discontinuity, but ferred from anomalously high line widths they reach at most 87% of the blast wave of the narrow component Hα emission in radius (Wang & Chevalier 2001). Thus a number of SNRs and attributed to heat- the ejecta knots should be over a par- ing in the precursor predicted by diffu- ′ sec, or about 2 behind the Balmer fila- sive shock acceleration models or to broad ment. Warren et al. (2005) have pointed component hydrogen atoms that over- out that if a substantial fraction of the en- take the shock and heat the upstream gas ergy dissipated by a shock goes into accel- (Smith et al.1994;Hester, Raymond & Blair erating cosmic rays, the ejecta knots can 1994). The small width of the narrow come much closer to the outer shock. In component in SN1006 means that there the case of the bulge in the southwestern is no evidence for such a precursor, but corner of the HST image and the X-ray it is nevertheless worthwhile to place on knot farther to the SW, the lack of syn- limit on the emission from such a precur- chrotronemissionatradioandX-raywave- sor. One expects a more or less exponen- lengths indicates that little energy goes tial falloff of brightness ahead of the shock into cosmic rays. Instead, since the blast peak on a scale given by κ/V , where κ wave encounted the dense gas in the NW s is the comic ray diffusion coefficient, or sector of SN1006 only about 180 years −1 by (n0σ) where σ is the charge transfer ago (Long et al. 2003), the ejecta knots cross section. The former would be about haveundoubtedlyovertakentheblastwave 1′′ for κ ∼ 1025 cm2s−1. For a speed because the blast wave decelerated when −1 near 3000 km s and the densities in Ta- it encountered denser gas. This has an ′′ ble 1, the latter would be about 0.6 . The important implication for analyses of the brightness could be a significant fraction fraction of shock energy that goes into cos- of the narrow component brightness, or up mic rays. Warren et al. (2005) show that to half the total brightness of the filament. ejecta knots in Tycho’s SNR come much The concave outward ripples we have closer to the blast wave than predicted analyzed so far are not appropriate for by Wang & Chevalier (2001), and they in- searching for a precursor, because the terpret this in terms of energy that goes emission from the curved part of the into accelerating particles to high energies. shock front could easily resemble emission That argument can be applied to most from a precursor. Instead the bright rims of Tycho’s SNR, but regions where the ′′ about 10 ahead are convex outward rip- SNR shell is flattened or where the Balmer ples, so that any emission ahead of the line emission is especially bright should be peak would be due to a precursor. Spa- avoided in the analysis, because those are tial profiles do show shoulders of order regions where the shock has probably been ′′ 1 wide ahead of the leading bright rims decelerated by higher density gas. in several places. However, there is ad- ditional faint emission out ahead of the main filament along the entire region im- 10