A Detailed Study of the Molecular and Atomic Gas Toward the γ-ray SNR RX J1713.7-3946: Spatial TeV γ-ray and ISM Gas Correspondence
Y. Fukui, H. Sano, J. Sato, K. Torii, H. Horachi, T. Hayakawa, N. M. McClure-Griffiths, G. Rowell, T. Inoue, S. Inutsuka, A. Kawamura, H. Yamamoto, T. Okuda, N. Mizuno, T. Onishi, A. Mizuno, H. Ogawa
aa r X i v : . [ a s t r o - ph . GA ] O c t Accepted version October 13, 2011
Preprint typeset using L A TEX style emulateapj v. 5/2/11
A DETAILED STUDY OF THE MOLECULAR AND ATOMIC GAS TOWARD THE γ -RAY SNR RX J1713.7 − γ -RAY AND ISM GAS CORRESPONDENCE Y. Fukui , H. Sano , J. Sato , K. Torii , H. Horachi , T. Hayakawa , N. M. McClure-Griffiths , G. Rowell , T.Inoue , S. Inutsuka , A. Kawamura , H. Yamamoto , T. Okuda , N. Mizuno , T. Onishi , A. Mizuno and H.Ogawa Department of Physics and Astrophysics, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Aichi 464-8601, Japan;[email protected] CSIRO Astronomy and Space Science, PO Box 76, Epping NSW 1710, Australia School of Chemistry and Physics, University of Adelaide, Adelaide 5005, Australia Department of Physics and Mathematics, Aoyama Gakuin University, Fuchinobe, Chuou-ku, Sagamihara, Kanagawa 252-5258, Japan National Astronomical Observatory of Japan, Mitaka, Tokyo 181-8588, Japan Department of Astrophysics, Graduate School of Science, Osaka Prefecture University, 1-1 Gakuen-cho, Naka-ku, Sakai, Osaka599-8531, Japan and Solar-Terrestrial Environment Laboratory, Nagoya University, Furo-cho, Chikusa-ku, Nagoya, Aichi 464-8601, Japan
Accepted version October 13, 2011
ABSTRACTRX J1713.7 − γ -ray SNR which emits γ -rays in the highest energyrange. We made a new combined analysis of CO and H I in the SNR and derived the total protons inthe interstellar medium (ISM). We have found that the inclusion of the H I gas provides a significantlybetter spatial match between the TeV γ -rays and ISM protons than the H gas alone. In particular,the southeastern rim of the γ -ray shell has a counterpart only in the H I . The finding shows that theISM proton distribution is consistent with the hadronic scenario that comic ray (CR) protons reactwith ISM protons to produce the γ -rays. This provides another step forward for the hadronic origin ofthe γ -rays by offering one of the necessary conditions missing in the previous hadronic interpretations.We argue that the highly inhomogeneous distribution of the ISM protons is crucial in the origin ofthe γ -rays. Most of the neutral gas was likely swept up by the stellar wind of an OB star prior tothe SNe to form a low-density cavity and a swept-up dense wall. The cavity explains the low-densitysite where the diffusive shock acceleration of charged particles takes place with suppressed thermalX-rays, whereas the CR protons can reach the target protons in the wall to produce the γ -rays. Thepresent finding allows us to estimate the total CR proton energy to be ∼ ergs, 0.1 % of the totalenergy of a SNe. Subject headings: cosmic rays – gamma rays: ISM – ISM: individual objects (RX J1713.7–3946) –ISM: clouds – ISM: molecules – ISM: H
I1.
INTRODUCTIONIt is a long-standing question how the cosmic ray (CR)protons, the major constituent of cosmic rays, are ac-celerated in the interstellar space. Supernova remnants(SNRs) are the most likely candidate for the accelerationbecause the high-speed shock waves offer an ideal sitefor diffusive shock acceleration (DSA) (e.g., Bell 1978;Blandford & Ostriker 1978). The principal site of CRproton acceleration is, however, not yet identified obser-vationally in spite of a number of efforts to address thisissue.RX J1713.7 − γ -ray SNR detected in the Galactic planesurvey with H.E.S.S. (Aharonian et al. 2006a) and isa primary candidate where the origin of the γ -raysmay be established. Discovery of the SNR was madein X-rays with ROSAT (Pfeffermann & Aschenbach1996) and
ASCA showed that the X-rays are non-thermal synchrotron emission with no thermal features(Koyama et al. 1997). TeV γ -rays were first detectedby CANGAROO (Enomoto et al. 2002) and, subse-quently, H.E.S.S. resolved the shell-like TeV γ -raydistribution with a ∼ . ◦ γ -raysare emitted via two mechanisms, either leptonic or hadronic, closely connected to the CR particles and it isimportant to understand which mechanism is workingin the SNR. The leptonic process includes the inverseCompton effect of CR electrons which energize the lowenergy photons of the cosmic microwave backgroundand additional soft photon fields (e.g., infrared). Thehadronic process includes the neutral pion decay into γ -rays following the reaction of CR protons with the lowenergy target protons in the interstellar medium (ISM).Considerable work has been devoted to explainingthe γ -ray emission in the framework of leptonic andhadronic scenarios (Aharonian et al. 2006b; Porter et al.2006; Katz & Waxman 2008; Berezhko & V¨olk 2008;Ellison & Vladimirov 2008; Tanaka et al. 2008;Morlino et al. 2009; Acero et al. 2009; Ellison et al.2010; Patnaude et al. 2010; Zirakashvili & Aharonian2010; Abdo et al. 2011; Fang et al. 2011).The molecular gas interacting with the SNR was dis-covered in V LSR , the velocity with respect to the lo-cal standard of rest, around − − at 2.6 arcminresolution based on NANTEN Galactic plane CO sur-vey and the distance of the SNR was determined to be1 kpc by using the flat rotation curve of the Galaxy(Fukui et al. 2003). This determination offered a ro-bust verification for a small distance of 1 kpc and re- FUKUI ET AL.vised an old value 6 kpc derived from lower resolutionCO observations (Slane et al. 1999). Studies of X-rayabsorption suggested a similar distance 1 kpc underan assumption of uniform foreground gas distribution(Koyama et al. 1997), but the local bubble of H I locatedby chance toward the SNR makes the X-ray absorptionuncertain in estimating the distance (Slane et al. 1999;Matsunaga et al. 2001). A subsequent careful analy-sis of the X-ray absorption favors the smaller distance(Cassam-Chena¨ı et al. 2004). At 1 kpc the SNR has aradius of 9 pc and an age of 1600 yrs (Fukui et al. 2003;Wang et al. 1997) and the expanding shock front has aspeed of 3000 km s − (Zirakashvili & Aharonian 2007;Uchiyama et al. 2003, 2007). Moriguchi et al. (2005)showed further details of the NANTEN CO distributionand confirmed the identification of the interacting molec-ular gas at − − by Fukui et al. (2003). Mostrecently, Sano et al. (2010) showed that the SNR harborsthe star forming dense cloud core named peak C at sim-ilar V LSR and argued that X-rays are bright around thecore, reinforcing the association of the molecular gas.Importantly, the associated molecular gas with theSNR opened a possibility to identify target protons wherethe hadronic process is working. If the cosmic ray densityis nearly uniform, we expect the γ -ray distribution mim-ics that of the interstellar target protons. Some nearbymolecular clouds show good spatial correlation with rela-tively high resolution γ -ray images, clearly verifying thatthe hadronic process is working to produce γ -rays in thecosmic ray sea (e.g., Bertsch et al. 1993 and the refer-ences therein.). A detailed comparison between the ISMprotons and the recent high-resolution γ -ray images ofH.E.S.S. sources is useful to test the correlation, althoughsuch a test was not possible until recently in the preced-ing low resolution γ -ray observations at degree-scale res-olution. Aharonian et al. (2006b) compared the NAN-TEN CO distribution with the H.E.S.S. TeV γ -ray im-age and examined both leptonic and hadronic scenariosas the origin of the γ -rays. By adopting annular averag-ing of TeV γ -rays and CO in the shell (see Figure 17 ofAharonian et al. 2006b), these authors found that TeV γ -rays are fairly well correlated with CO, whereas thecorrelation is not complete in the sense that the south-eastern rim of the TeV γ -ray shell has no counterpart inCO. The complete identification of target ISM protonsthus remained unsettled in the hadronic scenario.The γ -rays and X-rays are significantly enhancedtoward the clumpy molecular gas at a pc scale asfirst shown by Fukui et al. (2003). A strong con-nection among the molecular gas, the γ -rays, theX-rays, and perhaps cosmic rays is therefore sug-gested (Fukui et al. 2003; Fukui 2008; Sano et al. 2010;Zirakashvili & Aharonian 2010). Most of the previousmodels of the γ - and X-rays cited earlier assume more orless uniform density distribution of the ISM in the SNR,whereas the observations of the ISM indicate that theactual distribution is highly inhomogeneous, with den-sity varying by a factor 100 or more around the SNR.In addition, Galactic-scale studies of γ -rays suggest thatthere is ”dark gas” which is not detectable in CO or inH I but still contributes to the γ -rays and visual extinc-tion (Grenier et al. 2005; Ade et al. 2011). Such gas maybe either cold H I or H with no detectable CO. So, it is important to consider H I carefully in order to havea comprehensive understanding of the ISM protons. Inthe present paper, we shall use the term ”dark H I ” forobserved H I with significantly lower brightness than thesurroundings. So, ”dark H I ” does not necessarily mean”dark gas” above, while they might be linked.We here present a combined analysis of both the CO( J =1–0) and H I datasets in order to clarify thedistribution of the ISM protons in RX J1713.7 − CO( J =1–0) and H I data here and a comparativestudy of the CO( J =1–0, 2–1) transitions is a subjectof a forthcoming paper. A theoretical work of magneto-hydrodynamics which incorporates fully the inhomoge-neous ISM in RX J1713.7 − I datasets. Section 3 consists of five Sub-sections; Sub-section 3.1 gives a combined data set of CO and H I ,Sub-sections 3.2–3.4 the distribution of total ISM pro-tons in the SNR with an emphasis on dark H I , the coldand dense atomic phase of ISM protons, and Sub-section3.5 comparison with the γ -rays. Section 4 consists of twoSub-sections; Sub-section 4.1 describes the initial distri-bution of the ISM before the stellar explosion, and Sub-section 4.2 the scheme of particle acceleration and itsrelationship with the γ -rays. The conclusions are givenin Section 5. DATASETS OF CO, H I AND TEV γ -RAYSThe CO( J =1–0) data at 2.6 mm wavelength weretaken with NANTEN 4m telescope in 2003 April and areidentical with those published by Moriguchi et al. (2005).The system temperature of the SIS receiver was ∼
250 Kin the single side band including the atmosphere towardthe zenith. The beam size of the telescope was 2 . ′ . ′ in these obser-vations. We adopt hereafter the brightness temperature(K) as the spectral line intensity scale. The velocity res-olution and rms noise fluctuations are 0.65 km s − and0.3 K, respectively.The CO( J =2–1) data at 1.3 mm wavelength weretaken with NANTEN2 4m telescope in the period fromAugust to November in 2008 and part of the dataset waspublished by Sano et al. (2010). The frontend was a 4K cooled Nb SIS mixer receiver and the single-side-band(SSB) system temperature was ∼
250 K, including the at-mosphere toward the zenith. The telescope had a beamsize of 90 ′′ at 230 GHz. We used an acoustic optical spec-trometers (AOS) with 2048 channels having a bandwidthof 390 km s − and resolution per channel of 0.38 km s − .Observations in CO( J =2–1) were carried out in the on-the-fly (OTF) mode, scanning with an integration timeof 1.0 to 2.0 sec per point. The chopper wheel methodwas employed for the intensity calibration and the rmsnoise fluctuations were better than 0.66 and 0.51 K perchannel with 1.0 and 2.0 sec integrations, respectively.An area of 2.25 square degrees in a region of 346.7 deg ≤ l ≤ − . ≤ b ≤ I data at 21 cm wavelength are fromthe Southern Galactic Plane Survey (SGPS; Fig. 1.— (a) The H.E.S.S. TeV γ -ray distribution of RX J1713.7 − CO( J =1–0) emission in a velocity range of V LSR = −
20 km s − to 0 km s − is shown in color (Fukui et al.2003; Moriguchi et al. 2005). White contours show the H.E.S.S. TeV γ -ray distribution and are plotted every 20 smoothed counts from 20smoothed counts. (c) Averaged brightness temperature distribution of H I emission obtained by ATCA and Parkes in a velocity range from V LSR = − − to − − (McClure-Griffiths et al. 2005) is shown in color. White contours show the CO( J =1–0) brightnesstemperature integrated in the same velocity range every 1.0 K km s − ( ∼ σ ). McClure-Griffiths et al. 2005) and combined fromthe Australia Telescope Compact Array and the ParkesRadio Telescope. The beam size of the dataset was 2.2arcmin and we adopted a grid spacing of 40 ′′ toward theRX J1713.7 − − and 1.9 K, respectively.Moriguchi et al. (2005) showed an analysis of the CO( J =1–0) distribution over 100 km s − with a coarsevelocity window of 10 km s − in order to test associationwith the SNR. These authors showed that the velocityrange V LSR = −
20 to 0 km s − has convincing signs ofassociation with the SNR. We adopt in the present workthe velocity interval, − − , for the associatedISM and present detailed CO ( J =1–0 and 2–1) and H I data every 1 km s − (see Figure A in Appendix A).For the H.E.S.S. γ -ray data we used the combinedH.E.S.S. image shown in Figure 2 of Aharonian et al.(2007). Data of 2004 and 2005 are used for thissmoothed, acceptance-corrected gamma-ray excess im-age. The TeV image utilizes minimum 3 H.E.S.S. tele-scopes in event reconstruction to obtain a Gaussian stan-dard deviation of 0 . ◦
06 or FWHM of 0 . ◦
14 (8.3 ′ ). COMBINED ANALYSIS OF THE CO AND H I DATA3.1.
Distribution of CO and H I Figure 1(a) shows TeV γ -ray distribution toward RXJ1713.7 − CO( J =1–0) overlayed on the TeV γ -ray distribution. The CO( J =1–0) intensity becomes larger in the north tothe Galactic plane than in the south and the most promi-nent features above 0.7 K are located in the north-west. The general CO( J =1–0) distribution is shell- like associated with the γ -ray shell, showing weaker orno CO emission in part of the south. There are tworegions where CO( J =1–0) delineates particularly wellthe outer boundary of the shell in the southwest and east.In addition, we see some of the CO( J =1–0) features arelocated within the shell including the prominent peak Cat ( l , b ) = (347 . ◦
07, -0 . ◦ CO( J =2–1) distribution is qualitatively similarto the CO( J =1–0) distribution. A typical ratio of the[ J =1-0]/[ J =2-1] line intensities is ∼ CO( J =2–1) data by takingan advantage of higher angular resolution.Figure 1(c) shows an overlay of the H I distributionsuperposed on the CO( J =1–0) intensity in a velocityrange of − . − . − . The average H I brightnesstemperature ranges from 60 to 150 K and becomes highertoward the Galactic plane. The brightest H I of ∼ l , b )=(347 . ◦ − . ◦ CO( J =1–0) is seen. We find dark H I clouds of around 60 Kin the west (W cloud) and in the southeast (SE cloud).These dark H I clouds are not due to absorption of theradio continuum radiation which is very weak toward theSNR (Lazendic et al. 2004). The dark H I W cloud wellcorresponds to the CO( J =1–0) distribution, showingsharp edges both toward the east and west. The darkH I SE cloud has almost no counterpart in CO. Therelatively bright H I emission is seen in the north of theSNR (N cloud). The N cloud tends to be located toward CO( J =1–0) peaks, whereas the H I brightness shows a FUKUI ET AL. Fig. 2.— (a) Schematic of the identified CO( J =2–1) clouds is shown in colored contours. The image and white contours show the TeV γ -ray distribution (Figure 1(a)). The integration velocity ranges are follows; − − − (contour level: 2.3 K km s − ) for the SWcloud (CO), − − − (contour level: 4.9 K km s − ) for the W and N clouds (CO) and − − (contour level: 2.7 K km s − )for the NE, E and SE clouds (CO). (b) The locations of the identified H I clouds are shown in colored contours. The gray scale image andwhite contours show the TeV γ -ray distribution. The solid contours are for H I emission and the dashed contours are for dark H I . Theintegration velocity ranges are as follows; − − − (contour level: 214 K km s − ) for the W cloud (H I ), − − − (contourlevel: 780 K km s − ) for the central and SE clouds (H I ), and − −
11 km s − (contour level: 399 K km s − ) for the N cloud (H I ). non-monotonic, more complicated behavior than in theW cloud. In the northeast, we find a rim of relativelylower H I brightness of ∼
100 K toward ( l , b )=(347 . ◦ − . ◦
25) that lies along the γ -ray shell. A schematic ofthe four main H I clouds is given in Figure 2(b). Thegood correspondence of the H I clouds with the CO andthe γ -rays supports that the H I is physically associatedwith the SNR.In Figure 3 we show typical H I and CO profiles in thefour main H I clouds. Figure 3 indicates that the H I emission is generally peaked at −
10 km s − with smallhints of saturation, confirming that the H I is associatedwith the SNR and is generally not optically thick. Wefind that narrow H I dips having depths of 20–30 K oftencorrespond to CO( J =1–0) emission features in the Nand W clouds. The linewidths of the narrow H I dipsare as small as a few km s − . It is likely that these H I dips represent residual H I in cold CO gas seen as self-absorption. The broad H I dip in the SE cloud is alsoascribed to self-absorption as argued into detail in Sub-section 3.3. We show H I expected profiles of the back-ground H I emission with a straight-line approximationas dashed areas in Figure 3 (e.g., Sato & Fukui 1978).3.2. Molecular protons
In order to covert the CO( J =1–0) intensity intothe total molecular column density, we use an X fac-tor, which is defined as X (cm − /K km s − ) = N (H )(cm − )/W( CO) (K km s − ). In order to derive anX factor, the CO( J =1–0) intensity is compared withthe cloud dynamical mass (virial mass), or with the γ -rays produced via interaction of cosmic ray protons withmolecular clouds. An X factor therefore accounts for thetotal hadronic mass and is observationally uniform in theGalactic disk (e.g. Fukui & Kawamura 2010). We here adopt an X factor of 2.0 × W ( CO) (cm − /K kms − ) derived from the γ -rays and CO( J =1–0) inten-sity in the Galaxy (Bertsch et al. 1993). We double theH column density to derive the ISM protons in molec-ular form as shown in Figure 7(a). Compared with the CO( J =1–0) line, the CO ( J =2–1) line is not a com-mon probe of the molecular mass. This is in part be-cause the CO( J =2–1) emission samples a smaller por-tion, having a higher excitation condition, of a molecularcloud than traced by the CO( J =1–0) emission. We es-timate for instance that a typical fraction in area of the CO( J =2–1) emission to the CO( J =1–0) emission isabout 70–80 % at the half-intensity level convolved tothe same beam size in the present region from the COdata in Figure A. 3.3. Atomic protons
Optically thin case
We use the 21cm H I transition to estimate the atomicproton column density. A usual assumption is that theH I emission is optically thin and the following relation-ship is used to calculate the H I column density; N p (H I ) = 1 . × Z T L ( V ) dV (cm − ) (1)where T L ( V ) is the observed H I brightness temperature(K) (Dickey & Lockman 1990). We note that this sim-ple assumption is usually valid and apply equation (1)to the regions where no H I dips are seen. It is certainthat the narrow H I dips in the W and N clouds repre-sent self-absorption by cold residual H I in CO gas fromtheir exact coincidence with CO in velocity. The mostprominent dark H I cloud, the SE cloud, shows largelinewidths, not so common as self-absorption. We shall Fig. 3.— CO( J =1–0) and H I profiles at the four H I clouds; the N cloud ( l , b ) = (347 . ◦ − . ◦ l , b ) = (347 . ◦ − . ◦ l , b ) = (347 . ◦ − . ◦ l , b ) = (347 . ◦ − . ◦ γ -ray distribution. The shaded area shows expected profiles behind the self-absorption. Fig. 4.— (a)The H.E.S.S. TeV γ -ray distribution toward the SE cloud (Aharonian et al. 2007). Red contours show averaged H I brightnesstemperature distribution in a velocity range from −
15 km s − to − − (McClure-Griffiths et al. 2005). (b) The H I and CO( J =1–0)spectra at ( l , b ) =(347 . ◦ − . ◦ I profile. examine if the SE cloud represents self-absorption in thefollowings. 3.3.2. The dark H I SE cloud
We first show the integrated intensity image of the SEcloud in Figure 4(a). The H I contours are every 3.9 σ noise level and shows significant details not apparent inFigure 1(c), where a coarser color code is used. We findthe H I brightness variation is generally well correlatedwith the shell of γ -rays in gray scale in the sense that H I brightness decreases toward the enhanced γ -rays. Thistrend lends a support for physical connection of the SEcloud with the γ -ray shell and may be interpreted as due to decrease in spin temperature with density increase inthe self-absorbing H I gas (see Sub-section 3.3.3). No CO( J =1–0) emission is seen toward the SE cloud, ex-pect for a possible small counterpart at ( l , b ) = (347 . ◦ . ◦
72) and V LSR = − − (Figure 2(a) and FigureA), suggesting that density of the SE cloud is lower thanthe CO clouds.Figure 4(b) shows a typical H I profile in the SE cloudhaving a deep and broad dip. The large velocity spanof 20 km s − is not so common as a self-absorptionfeature; in nearby dark clouds H I self-absorption isgenerally narrow with a few km s − in linewidth (e.g. FUKUI ET AL. Fig. 5.— (a) A V distribution (Dobashi et al. 2005) is shown in color. Contours are the same as in Figure 1(a). (b) Distribution of columndensity of the total ISM protons estimated from both CO and H I in a velocity range from V LSR = −
20 to 10 km s − . Here the H I self-absorption is taken into account. Contours are the same as in Figure 1(a). Krˇco & Goldsmith 2010), whereas H I self-absorption asbroad as 10 km s − is seen in giant molecular clouds (e.g.Sato & Fukui 1978). The SE cloud delineates the γ -rayshell (Figure 4(a)) and is possibly compressed gas by thewind of a high-mass star, the SN projenitor. We haveinvestigated the velocity distribution of the SE cloud asgiven in Appendix A. We find that the SE cloud shows astrong velocity gradient which matches the blue-shiftedpart of an expanding swept-up shell. Such a shell is anatural outcome of the stellar-wind compression by theSN projenitor, supporting that the broad H I dip is as-cribed to the acceleration of H I gas by the wind. A H I stellar-wind shell in Pegasus driven by an early B starindeed shows a linewidth as large as 15 km s − (Sub-section 4.1, Yamamoto et al. 2006), similar to that ofthe SE cloud. The difference from the narrow H I dipsin the W and N clouds may be due to density; the SEcloud has lower density and is subject to stronger accel-eration than the CO clouds with narrow H I dips (seefor further discussion Sub-section 4.1), whereas the COclouds having higher density are less accelerated by thewind, making a systematic velocity gradient less clear inCO than in H I (see Figure B3).Figure 5(a) shows the distribution of the extinction A V toward RX J1713.7 − N p of ∼ cm − in Figure 5(b)corresponds to extinction of ∼ N p (cm − ) = 2.5 × · A V (magnitude)(Jenkins & Savage 1974). The extinction toward the SEcloud is 2–3 magnitude in Figure 5(a) and is consistent with the H I self-absorption by considering the contami-nation by the foreground stars which tends to reduce A V toward the Galactic plane.In summary, we find it a reasonable interpretation thatthe SE cloud represents H I self-absorption associatedwith the SNR shell.3.3.3. Analysis of the H I self-absorption dips We shall briefly review some basic properties of H I gas in order to understand the behavior of H I bright-ness (e.g. Sato & Fukui 1978). The spin temperature, T s , of H I is ∼
100 K or higher in warm neutral mediumat particle density less than 10 cm − . T s decreases withdensity from 100 K down to 10 K in a density range of100–1000 cm − (e.g., Figure 2 in Goldsmith et al. 2007).The temperature decrease is mainly due to higher shield-ing of stellar radiation and increased line cooling.It is well established that H I is converted into H on dust surfaces with increasing of the gas column den-sity and UV shielding and that H is dissociated by cos-mic rays and UV photons (e.g., Allen & Robinson 1977).The equilibrium H I abundance is determined by the bal-ance between formation and destruction of H and theresidual density of H I is about 10 − that of H in typi-cal interstellar molecular clouds (Allen & Robinson 1977;Sato & Fukui 1978). We also note that the H abun-dance should be time dependent since the formation ofH is a slow process in the order of 10 Myrs for densityaround 100 cm − (e.g., Allen & Robinson 1977).Based on the H I -H transition, we interpret the darkH I in Figure 2(b) as representing the H I with lower T s . The CO W cloud shows a good spatial coincidencewith the dark H I W cloud as is consistent with theinterpretation. The other prominent dark H I region, theSE cloud, shows no CO and we suggest that its densityis lower and its T s is higher than in the CO W cloud. H I brightness T L ( V ) is expressed as follows (e.g., Sato and Fig. 6.— (a) Distribution of peak optical depth of the H I self-absorption. (b) Distribution of atomic proton column density, N p (H I ),estimated for the H I emission and self-absorption. The velocity range in the both figures is from −
20 km s − to 0 km s − . Contours showthe H.E.S.S. TeV γ -rays distribution (Aharonian et al. 2007) and are plotted at 20 smoothed counts. We assume spin temperatures T s of40 K and 10 K, inside and outside the dotted box toward the SE cloud, respectively. Fukui 1978); T L ( V ) = T s [1 − e − τ ( V ) ] + T FG L ( V )+[ T BG L ( V ) + T BG C ] e − τ ( V ) − ( T FG C + T BG C ) (2)where T L ( V ), T s , τ ( V ), T FG L ( V ) and T BG L ( V ) are theobserved H I brightness temperature, the spin temper-ature, the optical depth of cold H I in the cloud, andthe foreground and background H I brightness temper-ature, respectively, at velocity V . T FG C and T BG C arethe continuum brightness temperature at 21-cm wave-length in the foreground and background of the cloud,respectively. The radio continuum emission is weak inRX J1713.7 − T FG C and T BG C are nearly zero as compared with T L ( V ).We are then able to estimate the H I column densityof dark H I clouds. Figure 4(b) shows the H I self-absorption dip with the background H I emission inter-polated by a straight line connecting the two shouldersat 0 and 20 km s − . This gives a conservative estimatebecause the actual background H I shape perhaps has amore intense peak at −
10 km s − as seen in the northernarea of the SNR. The spin temperature T s of the dark H I gas is an unknown parameter. We estimate T s to be lessthan ∼
55 K from the lowest H I brightness at the bottomof the dip in Figure 4(b) and higher than ∼
20 K, wherethe temperature of the CO clouds is ∼
10 K (Sano et al.2010). We estimate the absorbing dark H I column den-sity to be N p (H I ) = 1.0 × cm − (optical depth =0.8), 1.8 × cm − (optical depth = 1.1) and 3.1 × cm − (optical depth =1.5) for assumed three cases T s = 30, 40 and 50 K, respectively, for the half-power linewidth ∆ v =10 km s − , where the H I optical depth ¯ τ isestimated by equation (2) and N p (H I ) by the followingrelationship; N p (H I) (cm − ) = 1 . × T s (K) ∆ v (km s − )¯ τ (3) We shall here adopt T s = 40 K and a corresponding darkH I optical depth of 1.1. A higher T s gives a higheroptical depth and vice versa. The relatively large opticaldepth around 1 is consistent with the fairly flat H I dipin Figure 4(b), which suggests weak saturation. We alsotested the effects of elevating the background H I by 15K and found a small change of 5 × cm − . The erroris mainly introduced by the straight-line approximationand uncertainty in T s of ∼
10 K. We infer the dark H I column density is accurate within a systematic error of ∼ × cm − .The average H I density in the SE cloud is roughly esti-mated to be 150 cm − by dividing 1 . × cm − by ∼ J =1–0) tran-sition, ∼ − , consistent with no CO emission fromthe SE cloud and with low spin temperature around 40K.We also extended such an analysis to the regions withnarrow H I dips associated with CO emission, where weadopt T s = 10 K, the kinetic temperature of the COgas. The small dips in these regions indicate that theH I optical depth is generally as low as ∼ I in CO gas. We showthe distributions of the peak optical depth of the H I self-absorption in Figure 6(a), and the derived total H I column density distribution, the sum of the H I in emis-sion and self-absorption in Figure 6(b), where the SEcloud is significant. We shall hereafter refer to the darkH I of T s = 40 K as ”cool H I ” and that of T s = 10 K as”cold H I ”. 3.4. Total ISM protons
The number of the total ISM protons in the SNR isgiven by summing up the three components in a velocity FUKUI ET AL.
Fig. 7.— (a) Distributions of column density of ISM protons N p estimated from CO( J =1–0) N p (H ), (b) H I emission with correctionfor the H I self-absorption N p (H I ) and (c) sum of N p (H ) and N p (H I ). All the datasets used here are smoothed to a HPBW of TeV γ -ray distribution with a Gaussian function. (d) TeV γ -ray distribution. Contours are plotted every 50 smoothed counts from 20 smoothedcounts. range from −
20 to 0 km s − ; H derived from CO( J =1–0), dark H I (dips) and warm H I (emissions). The re-sults are shown as spatial distributions in Figure 7. Fig-ure 7(a)–(d) show N p (H ), N p (H I ), N p (H +H I ) andTeV γ -rays, respectively. We see the total ISM protons N p (H +H I ) shows a shell-like shape similar to the TeV γ -rays which significantly improves the correlation withthe γ -rays as compared with the case of molecular gasonly. We therefore conclude that the contribution of H I is critical as well as H in counting the ISM protons. Wefind that in the south the total ISM proton is dominatedby the atomic gas, whereas in the north the molecularand atomic protons are both important. A more quan-titative comparison will be given in Sub-section 3.5.2.Similar diagrams of the total ISM protons to Figure 7 are presented for the optically-thin case for reference Fig-ure C1 in Appendix C, where the shell-like distributiontoward the SE cloud is missing.3.5. The γ -rays and the ISM protons Gamma-ray distribution
The TeV γ -ray distribution obtained by H.E.S.S. isa nearly circular-symmetric shell with some ellipticityelongated in the north-south direction. In order to gainan insight into the distribution of the γ -ray emissivitywe undertake a simple analysis of the γ -ray distribution.We first adopt an elliptical annular ring in the analysis,while Aharonian et al. (2006b) made a similar analysisby using a circular annular ring in correlating γ -rays andNANTEN CO intensity (see their Figure 17). Fig. 8.— (a) Distributions of column density of the total ISM protons N p (H +H I ) in a velocity range from −
20 km s − to 0 km s − .Contours are the same as in Figure 1(a). (b) Azimuthal distributions of N p (H ), N p (H I ), N p (H +H I ) and TeV γ -ray smoothed countsper beam between the two elliptical rings shown in Figure 8(a). The proton column densities are averaged values between the rings (seetext). Semi-major and semi-minor radii of the outer ring are 0.46 degrees and 0.42 degrees, respectively, and the radii of the inner ring arehalf of them. The same plots inside the inner ring are shown on the right side of Figure 8(b). Fig. 9.—
Radial distribution of TeV γ -rays radiation. Small dotsshow the distributions of all the H.E.S.S. data points and largefilled circles with error bars show averaged values at each radius.We assume a 3D spherical shell with a Gaussian-like intensity dis-tribution along its radius to approximate the TeV γ -ray distribu-tion (see text). The green line shows the estimated 3D Gaussiandistribution and the red line shows its projected distribution. Thepeak radius and the full width at half maximum of the green lineare estimated to be 0.46 deg ( ∼ ∼ We estimated the radius of the γ -ray shell as defined ata half-intensity level of the peak γ -ray smoothed countevery 15 degrees for an assumed center. We averagedthe radii in angle and minimized the sum of the squaresof the deviation from the average. This process gives acentral position to be ( l , b ) = (347 . ◦ − . ◦ γ -ray smoothedcounts and an averaged value shown by a step functionin radius r every 0.05 degrees. Here we also adoptedthe elliptical shape and normalized the radius to thatof the major axis with the elliptical modification. Afterseveral trials of different functional forms, we found aGaussian radial distribution of the γ -ray emissivity pervolume reproduces well the projected radial distributionin Figure 9. In the fitting we have two free parametersof the Gaussian shape, the peak radius r and the sigma σ expressed as follows; F ( r ) = A × e − (cid:0) r − r (cid:1) / σ (4)where A is a normalization coefficient. By requiring thatthe error in the fitting becomes minimum in the pro-jected distribution shown by the step function, we found r = 0.46 degrees and σ = 0.10 degrees give the best fitas shown in Figure 9. This distribution shows that theobserved shell is consistent with a shell of a half-intensitythickness ∼ γ -raysare mainly emitted in a thick shell of 8.0 pc radius and4.2 pc width at the half-intensity level with nearly zeroemission from the inner part. A similar thick-shell modelwas also obtained by Aharonian et al. (2006b). Numeri-cal modeling of the γ -ray emission has been undertakenby several authors and indicates that the γ -ray emis-sion has a rather steep gradient beyond the peak of theshell in either of the leptonic or hadronic scenario (e.g.Jun & Norman 1996; Zirakashvili & Aharonian 2010).The fitting to the H.E.S.S. data above shows that thegradient in the γ -ray distribution is not so steep towardthe outside, which may be due to smearing in space byaveraging. We shall not try a further elaborated analysis0 FUKUI ET AL. Fig. 10.—
Radial distributions of averaged values of TeV γ -raysradiation, N p (H ), N p (H I ) and N p (H +H I ). N p (H ) and N p (H I )show column densities estimated from CO( J =1–0) and H I , re-spectively, and N p (H +H I ) shows the total ISM proton columndensity. here due to the limiting angular resolution of H.E.S.S.which is 0.14 deg (FWHM).Figure 9 shows that the projected radial distributionof ISM protons follows a fairly similar distribution to the γ -rays inside the SNR. This is consistent with that theISM distribution is also shell-like with an inner cavityas is consistent with the stellar wind shell discussed inSub-section 4.1; if the ISM has no cavity in the innerpart, the projected distribution of the ISM should in-crease toward the center. We shall assume hereafter thatthe ISM distribution is also approximated by the sameGaussian shape as the γ -rays with a radius of 8.0 pc witha thickness of 4.2 pc at the half-intensity level.3.5.2. Comparison between the γ -rays and the ISM protons In the hadronic scenario, the target distribution shouldbe correlated with the γ -ray distribution for a uniformCR distribution. This correlation should be seen insidethe shell of the SN shock which has a sharp gradient be-yond its outer radius. We expect that the ISM protonsare distributed beyond the outermost edge of the shellwhere CR protons cannot reach by diffusion. Beyondthe SNR shock, the γ -ray emission profile may be influ-enced by components from the diffuse cosmic-ray back-ground and by the energy dependent transport of escap-ing cosmic-rays from RX J1713.7 − γ -rays in the position angleshown in Figure 8(a), where the vertical scale is adjustedso that the correspondence with the TeV γ -rays becomesoptimum. Here, the error in the TeV γ -ray emission fromthe publicly available H.E.S.S. image is approximately(smoothed counts) . . In Figure 8(b) the uncertainty inthe dark H I in the SE cloud, 1 × cm − , is in the or- der of 10–20 % of the total. The total ISM proton densityshows a good agreement with the TeV γ -ray angular dis-tribution and also the central part in the inner ring. Werecall that CO alone showed marked deficiency towardthe SE cloud as compared with the γ -rays (see Figure17 of Aharonian et al. 2006b). The present analysis in-dicates the deficiency is recovered by including H I andhas shown that the total gas of both atomic and molec-ular components have a good correlation with the TeV γ -rays in the annular ring. The total mass of the ISMprotons responsible for the γ -rays is 2.0 × M ⊙ overthe whole SNR (radius 0.65 deg); the mass of molecu-lar protons is 0.9 × M ⊙ and that of atomic protonsis 1.1 × M ⊙ , where we assume the ISM protons inter-acting with the CR protons is proportional to the TeV γ -rays (Sub-section 3.5.1., Figure 10).There are two points in Figure 8(b), for which addi-tional remarks may be appropriate. One is the point atan azimuth angle of 115 degrees which may be estimatedtoo low due to lack of correction for the self-absorptionbecause of the large velocity shift in the expanding shell(see Figure B1). Another is the point at an azimuth an-gle of 165 degrees where the strong CO emission (peakA after Fukui et al. 2003) increases the proton columndensity, although the increased protons may not be in-teracting with the CR protons beyond the SNR shock,leading to less γ -rays.An independent test is made by the radial distributionof the ISM protons given in Figure 10, where an averagetaken over the same binning as the γ -rays in Figure 9is shown by a step function and the total ISM protonsand γ -rays are superposed with the same proportionalfactor as adopted in Figure 8. Here, the error in the TeV γ -ray emission is approximately (oversampling-correctedtotal smoothed count) . normalized to 1 (arcmin) . Wesee the N p (H +H I ) and γ -rays show a good agreementinside the shell and the γ -rays sharply decrease outsidethe shell. This offers another presentation of the goodcorrelation between the γ -rays and the ISM protons.We argue that the apparent anti-correlation betweenthe H I brightness at the bottom of the dips and the γ -rays in the SE cloud (Figure 4) is consistent with thatthe H I dips are due to the cool and dense H I gas. Theanti-correlation is interpreted that the spin temperature T s of H I decreases with density (Sub-section 3.3) andthat the γ -rays increase with the ISM proton density lo-cally in the SE cloud, demonstrating detailed correspon-dence between the γ -rays and the ISM protons which ismainly atomic. The small and narrow H I dips in the Wand N clouds have the H I column density less than 10 cm − , significantly lower than the typical molecular col-umn density by two orders of magnitude. So, in most ofthe regions except for the SE cloud the H I column den-sity is dominated by emission but not by self-absorption.For the sake of reference, we show a set of similar dia-grams of ISM proton distributions for the optically-thincase in Figures C2 and C3 in Appendix C, correspondingto Figures 8 and 10, respectively.Before concluding this Sub-section, we cautiously notethat the cool/cold H I could not be estimated accurately,if the cool/cold H I is optically thick, if the cool/coldH I lies behind optically thick foreground H I in the lineof sight, or if the background H I profile has a differ-1 TABLE 1A Comparison between RX J1713.7 − RX J1713.7 − ∗ Pegasus Loop † Distance (kpc)...................................................................... 1 0.1Diameter (pc)....................................................................... 17.4 25Total mass of the ISM ( M ⊙ )................................................ ∼ ‡ ∼ ∼ ∼ I (K).................................................... ∼ ∼ I (km s − ).................................................... ∼ ∼ − ).................... ∼ ∼ ∗∗ B2 IV
Note . — ∗ Fukui et al. (2003), Moriguchi et al. (2005), † Yamamoto et al. (2006), ∗∗ Cassam-Chena¨ı et al. (2004), ‡ present work. The Pegasus loop may consist of two shells and the themass should be regarded as an upper limit (Yamamoto et al. 2006). ent shape from its neighbors. Such effects, while posingintrinsic limits for probing cool/cold H I , are relativelyunimportant for nearby objects at a distance of 1 kpc orless where foreground H I is not important. The dark H I W and SE clouds are probably good examples where thecool/cold H I is well traced by the low H I brightness,whereas the N cloud with higher H I brightness may bepartially affected by the foreground H I in the line ofsight. DISCUSSION4.1.
The evacuated cavity by the stellar wind
It is likely that the CO shell in Figure 1(b) was formedover a timescale of Myrs by the stellar wind of the pro-jenitor, an OB star that exploded as a supernova (SN)1600 yr ago. The total velocity span of the CO shell, ∼
20 km s − , is much smaller than the SN shock speedand indicates that it takes Myr to form the shell of theISM as roughly estimated by dividing the radius 9 pc by10 km s − . Molecular gas expanding at 10 km s − canmove only 0.01 pc in 1000 yrs. Therefore, the current COdistribution has little been affected by the supernova ex-plosion (SNe) and holds the initial condition before theshock interaction.While a stellar-wind shell with a known central staris not often observed elsewhere, one such example is thePegasus loop found in CO( J =1–0), H I and dust emis-sion at ( l , b ) = (109 ◦ , − ◦ ) centered on a run-way starHD886 (B2 IV) (Yamamoto et al. 2006). The Pegasusloop is located at ∼
100 pc in a relatively uncontami-nated environment outside the Galactic plane. No SNeoccurred yet in this shell. A comparison between RXJ1713.7 − ∼ ∼
18 pc and a total mass of ∼ ⊙ . The shell is mostly atomic and consists of 78 smaller CO( J =1–0) clumps (see Figure 10 in Yamamoto et al.2006). The clumped CO is a natural outcome of ther-mal/gravitational instability and seems common in sucha shell. The shell is expanding at a total velocity spanof 15 km s − . The H I density inside the shell is ∼ − in the north, where the stellar wind evacuated theISM over 1 Myrs. The Pegasus loop is located in a some-what lower-density environment than RX J1713.7 − − − − − (e.g. Zirakashvili & Aharonian 2010; Morlino et al.2009; Berezhko & V¨olk 2008) and the dense shell withCO clumps remaining more or less as they were priorto the SNe. The SN shock front moves almost freelyat ≥ − in the cavity in the early phase of ∼ γ -ray shell is not stronglydeformed, while we see some deviations of a pc scalefrom a perfect circular shell, suggesting effects of recentdynamical interaction.The interaction between molecular clumps and theshock is observed as the X-ray enhancement arounddense molecular clumps at a spatial resolution higherthan 0.5 pc. Sano et al. (2010) showed that the molecu-lar clump peak C is rim-brightened in X-rays, suggestingthat it is a dense clump overtaken by the shock, andpeak A (Fukui et al. 2003) is also X-ray brightened onlytoward its inner edge, indicating the shock interactionat the inner boundary of peak A. IYIF2011 showed thatthe initial magnetic field B of 1 µ G is amplified to 0.1to 1 mG near dense clumps by the enhanced turbulencedriven by the shock. The stronger magnetic field ex-plains the X-ray enhancement as due to the enhancedsynchrotron emission that is proportional to B , or, dueto increased acceleration. IYIF2011 also showed thatthe shock speed v s is significantly reduced locally withdensity n (cm − ) such that v s ∼ − / p n/n ,where n =1 cm − . This dependence of v s on densitycan explain the absence of thermal X-rays in the SNRbecause the molecular gas is too dense to be affected bythe shock to emit thermal X-rays (IYIF2011). A uni-form lower-density case with significant thermal X-raysby shock heating is presented by Ellison et al. (2010) butsuch a model is not applicable to the highly inhomoge-neous ISM of RX J1713.7 − cm − , has survived without erosion(Sano et al. 2010).2 FUKUI ET AL.4.2. The γ -ray emission mechanism TeV γ -rays are emitted via two mechanisms, eitherleptonic or hadronic processes. The leptonic processexplains γ -rays via the inverse Compton effect betweenCR electrons and low energy photons. In the hadronicscenario γ -rays are emitted by the decay of neutralpions which are produced in the high energy reactionsbetween CR protons and ISM protons. Diffusive shockacceleration (DSA) is the most widely accepted schemeof particle acceleration (Bell 1978; Blandford & Ostriker1978; Jones & Ellison 1991; Malkov & Drury 2001).The previous works on RX J1713.7 − γ -rays andX-rays is explained by either of the leptonic and/orhadronic mechanisms if DSA works to acceleratethe particles (Aharonian et al. 2006b; Porter et al.2006; Katz & Waxman 2008; Berezhko & V¨olk 2008;Ellison & Vladimirov 2008; Tanaka et al. 2008;Morlino et al. 2009; Acero et al. 2009; Ellison et al.2010; Patnaude et al. 2010; Zirakashvili & Aharonian2010; Abdo et al. 2011; Fang et al. 2011)In the hadronic scenario, where the neutral pion decaydetermines the γ -rays via proton-proton reactions, theaverage density of the target protons is constrained bythe total energy of CR protons; the average target den-sity greater than 0.1 cm − is required to produce CR pro-tons having the total energy of 10 erg, for the maximumenergy of a SNe, while higher target density is requiredfor less CR proton energy. In the leptonic scenario, wherethe inverse Compton process produces γ -rays, the criticalparameter is the magnetic field which constrains the syn-chrotron loss timescale of CR electrons; a magnetic fieldof order of 10 µ G is usually required (e.g. Tanaka et al.2008).We here argue that the highly inhomogeneous dis-tribution of the ISM, the cavity and the dense andclumpy wall opens a possibility to accommodate thelow-density site for DSA and the high-density target si-multaneously as discussed into detail by IYIF2011. Asimilar argument on the hadronic interaction betweenCR protons with the ambient dense clouds has beenpresented by Zirakashvili & Aharonian (2010). In thispicture, first, the cosmic rays are accelerated via DSAin the low density cavity, and second, the CR protonsreach and react with the target protons in the densewall to produce γ -rays. The main energy range of theCR protons required for hadronic TeV γ -rays is 10–800 TeV (Zirakashvili & Aharonian 2010). The penetra-tion depth, l pd , of cosmic rays is expressed as follows(IYIF2011); l pd ∼ . η / (cid:18) E
10 TeV (cid:19) / (cid:18) B µ G (cid:19) − / (cid:18) t age yr (cid:19) / (pc) (5)where E , B and t age are the particle energy, the magneticfield and the age of the SNR. The parameter η is theso-called ”gyro-factor” and has some ambiguity. In theSNR, it is reasonable to consider η ∼ µ G and 0.1–0.9 pc for 100 µ Gin a typical timescale of ∼ yr. The penetration depth of the CR electrons is determined by taking t age equal tothe synchrotron loss timescale (e.g. Tanaka et al. 2008)in equation (6) and becomes energy-independent for theX-ray emitting electrons of 1–40 TeV as follows; l = 0 . η / (cid:18) B µ G (cid:19) − / (pc) (6)We estimate l to be from 0.8 pc for 10 µ G to 0.026 pcfor 100 µ G if η = 1. CR protons can therefore reachand penetrate into the dense gas within pc-scale of theacceleration site to produce TeV γ -rays, while the CRelectrons stay relatively closer to the acceleration site,in particular, near the dense gas having strong magneticfield. This offers an explanation on the hadronic γ -rayproduction and the correlation between the γ -rays andtarget protons in Figures 4, 8 and 10 is a natural outcomein the scenario (IYIF2011).Gabici et al. (2007) discussed the importance of theenergy-dependent interaction between CR protons andmolecular clouds and Zirakashvili & Aharonian (2010)discussed that the γ -ray spectrum may not distinguishthe leptonic and hadronic scenarios in case of RXJ1713.7 − − γ -ray spectrum becomessimilar both for the leptonic and hadronic scenarios,not usable to distinguish the two scenarios, as noted byZirakashvili & Aharonian (2010) and concluded that thehadronic origin is testable only by comparing γ -rays withthe ISM target distribution. The present results havedemonstrated that the ISM proton distribution show in-deed a good spatial correspondence with the γ -rays bytaking into account the contribution of the H I and matchwith the prediction by Zirakashvili & Aharonian (2010)and IYIF2011.The total energy of CR protons is estimated by the re-lationship between the total target protons and the ob-served γ -rays (2–400 TeV) after extrapolating the protonspectrum to 1 GeV as follows (Aharonian et al. 2006b); W tot ∼ − × (cid:18) d (cid:19) (cid:18) n − (cid:19) − (erg) (7)where the distance to the source d ∼ n . The average density of ISMprotons is calculated to be ∼
130 cm − for the total massof the ISM protons 2.0 × M ⊙ over the whole SNR (ra-dius 0.65 degrees) as modeled in Figure 10 and the totalCR proton energy to be ∼ × ergs by usingequation (7). This corresponds to ∼ –10 ergs (Abdo et al. 2010; Giuliani et al. 2010). Wemay speculate that the CR protons become accumulated3in a few times 10000 yrs to reach more than 10 % of theSNe energy. This issue is to be further tested by exam-ining cosmic ray escaping from SNRs (e.g. Gabici et al.2009; Casanova et al. 2010a,b).To summarize the discussion, we have shown that acombined analysis of CO and H I provides a reasonablecandidate for the target ISM protons and thereby lendsa new support for the hadronic scenario. We should notethat the present analysis offers one of the necessary con-ditions for the hadronic scenario for uniform CR pro-ton distribution, but it is not a full verification of thehadronic scenario and does not rule out leptonic compo-nents. We need to acquire additional observations beforefully establishing the hadronic scenario, including betterdetermination of the magnetic field and higher angularresolution images of γ -rays at least comparable to thatof the ISM. Cherenkov Telescope Array will provide suchimages in future. We discussed that the observed highlyinhomogeneous distribution of the ISM plays an essen-tial role in the γ -ray production; DSA works in highlyevacuated cavity and the accelerated CR protons travelover a pc to interact with the surrounding dense ISMprotons. It is important to develop a similar analysis ofboth H I and CO in the other similar objects like RXJ0852.0-4622 (Vela Jr.), RCW86 and HESS J1731-347.Such works are in progress based on the NANTEN2 ob-servations and high-resolution H I interferometry. CONCLUSIONSWe summarize the main conclusions as follows;1. A new analysis of CO and H I has revealed thatthe TeV γ -ray SNR RX J1713.7 − I gas without H derived from CO. This H I gas is relatively denseand cold and detectable mainly as H I emission. Wehave also identified regions where H I is observedas dark H I in self-absorption dips and derived thetotal ISM proton column density over the SNR.The H I plus H , the total ISM protons, providesone of the necessary conditions, target protons, inthe hadronic origin of the γ -rays. Such target ISMprotons have not been identified in the previousstudy that took into account only H , although thepresent finding alone does not exclude the leptonicorigin.2. For an annular pattern around the TeV γ -ray shell,we compared the total ISM proton distributionwith the TeV γ -ray distribution and found thatthey show reasonably good correspondence, vary-ing by similar factors. The inclusion of the atomicprotons observed as the H I self-absorption dips isessential particularly in the southeast of the γ -rayshell. The interpretation of H I self-absorption dipsis also supported by the enhanced optical extinc-tion toward the southeast rim. 3. The cavity surrounding the SNR was created by thestellar wind of the SN projenitor. The inside of thecavity is of low density with < − while the cav-ity wall consists of the dense and clumpy atomic ormolecular target protons of ≥ − . Thediffusive shock acceleration in the highly inhomo-geneous ISM offers a reasonable mechanism of par-ticle acceleration in the low-density cavity and thedense wall acts as the target for γ -ray productionby the CR protons. Hydro-dynamical numericalsimulations of the interaction have shown detailedphysical processes involved (IYIF2011).4. By considering the other pieces of the observationaland theoretical works accumulated thus far, thepresent results make the hadronic interpretationmuch more comfortable in RX J1713.7 − ∼ ergs, 0.1 % of the total energy of SNe,if we assume the γ -rays are all produced by thehadronic process.We are grateful to Felix Aharonian for the lively andfruitful discussion on the subject, without which the workwould not have been completed. The NANTEN projectis based on a mutual agreement between Nagoya Univer-sity and the Carnegie Institution of Washington (CIW).We greatly appreciate the hospitality of all the staff mem-bers of the Las Campanas Observatory of CIW. We arethankful to many Japanese public donors and compa-nies who contributed to the realization of the project.NANTEN2 is an international collaboration of ten uni-versities, Nagoya University, Osaka Prefecture Univer-sity, University of Cologne, University of Bonn, SeoulNational University, University of Chile, University ofNew South Wales, Macquarie University, University ofSydney and ETH Zurich. The work is financially sup-ported by a grant-in-aid for Scientific Research (KAK-ENHI, no. 15071203, no. 21253003, no. 20244014, no.23403001, no. 22540250, and no. 22244014) from MEXT(the Ministry of Education, Culture, Sports, Science andTechnology of Japan) and JSPS (Japan Society for thePromotion of Science) as well as JSPS core-to-core pro-gram (no. 17004). We also acknowledge the support ofthe Mitsubishi Foundation and the Sumitomo Founda-tion. This research was supported by the grant-in-aidfor Nagoya University Global COE Program, ”Quest forFundamental Principles in the Universe: from Particlesto the Solar System and the Cosmos”, from MEXT. Thesatellite internet connection for NANTEN2 was providedby the Australian Research Council.APPENDIX APPENDIX AVELOCITY CHANNEL DISTRIBUTIONS IN RX J1713.7 − We show velocity channel distributions of CO( J =1–0, 2–1) and H I every 1 km s − from −
20 km s − to 0 km s − superposed on the TeV γ -ray distribution in Figure A.4 FUKUI ET AL. Fig. A.—
Velocity channel distributions of CO( J =1–0, 2–1) emission and H I brightness temperature overlayed on the TeV γ -raydistribution. F irst - row panels ( top ): H I image and CO( J =1–0) contours. Second - row panels : H I image superposed on the TeV γ -ray contours. T hird - row panels : CO( J =1–0) image superposed on the TeV γ -ray contours. F ourth - row panels : CO( J =2–1) imagesuperposed on the TeV γ -ray contours. Each panel shows CO and H I distributions every 1 km s − in a velocity range from −
20 to 0 kms − . The lowest contour levels of CO and TeV γ -rays are 0.73 K ( ∼ σ ) and 20 smoothed counts and contour intervals of CO and TeV γ -rays are 0.73 K ( ∼ σ ) and 10 smoothed counts, respectively. APPENDIX BEXPANDING MOTION OF THE DARK H I SE CLOUD
Figure B1 left shows schematically an expanding spherical shell of radius R = 9 pc and uniform expansion velocity V = 10 km s − and Figure B1 right a position-velocity diagram of the shell, where the ellipsoidal nature of the shell isnot taken into account for simplicity. Figure B2 shows three representative velocity-channel distributions of the darkH I SE cloud for a velocity range from −
20 to −
10 km s − and shows that the SE cloud is extended to the north.The extension shifts toward the northwest with velocity decrease from −
10 to −
20 km s − as is consistent with theiso-velocity contours expected from the shell model in Figure B1. Figure B3 shows another presentation of kinematicaldetails of the SE cloud in position-velocity diagrams. We choose a line AB passing through the center of the SNR5 Fig. A.—
Continued and the SE cloud, and another line CD passing through the SE cloud in the north-south (Figure B3(a)). We show aposition-velocity distribution of H I along the line AB (Figure B3(b)) and H I profiles along the two lines AB and CD(Figure B3(c)). We find the SE cloud is extended to the northwest with a large velocity gradient of 10 km s − per0.5 degrees, or ∼ − pc − . The H I profiles in Figure B3(c) shows that the dips are deep and clear at b lessthan − . b = − . b higher than − . I probably suffersfrom self-absorption to some extent as suggested by the weaker H I brightness at −
12 km s − toward b = − . ◦
52 thantoward b = − . ◦
35 (line AB). We note that the strong velocity gradient in Figures B2 and B3 is consistent with theblue-shifted part of an expanding shell. The strong velocity gradient is interpreted in terms of the expanding shell asdepicted by a white circle in the position-velocity diagram (Figure B3(b)). The blue shift by 10 km s − toward thecenter of the SNR indicates this part of the shell is in the foreground. This is consistent with that the dips are due toself-absorption against the background H I emission. We also infer that the swept-up shell is highly non-uniform sincethe broad H I dips are seen only in a quarter of the shell.6 FUKUI ET AL. Fig. A.—
Continued
APPENDIX CANALYSIS OF THE H I EMISSION; THE OPTICALLY THIN CASE
The present analysis has shown that the H I is self-absorbed in part of the SNR as indicated by the H I dips and theH I column density is estimated by taking into account the self-absorption (Figure 7). In order to see the effects ofthe self-absorption quantitatively, we here show for comparison the ISM proton distribution in the optically thin case,which does not take into account the self-absorption. Figure C1, equivalent to the self-absorption case in Figure 7,includes the H I column density distribution for the optically-thin assumption smoothed to the HESS resolution ((b)and (c)), where the SE cloud is not seen. Figures C1(a) and (d) are the same with those in Figure 7. Figure C2 isequivalent to Figure 8. Figure C2(a) is the total ISM proton column density for the optically thin H I at NANTENresolution overlayed on the TeV γ -ray distribution. Figure C2(b) is the corresponding azimuthal distribution of ISMprotons and TeV γ -rays, where the ISM protons is deficient in azimuthal angle from −
90 to 0 degrees as compared toFigure 8(b). Figure C3 is equivalent to Figure 10, and shows the radial distribution of ISM protons for the optically thinH I without correction for the H I self-absorption. In the smoothed radial distribution, the effect of the self-absorption7 Fig. A.—
Continued is not so obvious.8 FUKUI ET AL.
Fig. B1.—
Schematic image of a uniformly expanding shell and its velocity distribution in the position-velocity plane. Here we assumea radius of the shell R and an expansion velocity V exp of 9 pc and 10 km s − , respectively.REFERENCESAbdo, A. A., Ackermann, M., Ajello, M., et al. 2010, Science, 327,1103Abdo, A. A., Ackermann, M., Ajello, M., et al. 2011, ApJ, 734, 28Acero, F., Ballet, J., Decourchelle, A., et al. 2009, ApJ, 505, 157Ade, P. A. R., Aghanim, N., Arnaud, M., et al. 2011,arXiv:1101.2029Aharonian, F. A., Akhperjanian, A. G., Aye, K.-M., et al. 2004,Nature, 432, 75Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al.2006a, ApJ, 636, 777Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al.2006b, A&A, 449, 223Aharonian, F., Akhperjanian, A. G., Bazer-Bachi, A. R., et al.2007, ApJ, 464, 235Allen, M., & Robinson, G. W. 1977, ApJ, 212, 396Bell, A. R. 1978, MNRAS, 182, 147Berezhko, E. G., & V¨olk, H. J. 2008, A&A, 492, 695Bertsch, D. L., Dame, T. M., Fichtel, C. E., et al. 1993, ApJ, 416,587Blandford, R. D. & Ostriker, J. P. 1978, ApJ, 221, L29Casanova, S., Aharonian, F. A., Fukui, Y., et al. 2010a, PASJ, 62,769Casanova, S., Jones, D. I., Aharonian, F. A., et al. 2010b, PASJ,62, 1127Cassam-Chena¨ı, G., Decourchelle, A., Ballet, J., et al. 2004,A&A, 427, 199Dickey, J. M. & Lockman, F. J. 1990, ARA&A, 28, 215Dobashi, K., Uehara, H., Kandori, R., et al. 2005, PASJ, 57, 1Enomoto, R., Tanimori, T., Naito, T., et al. 2002, Nature, 416,823Ellison, D. C., & Vladimirov, A. 2008, ApJ, 673, L47Ellison, D. C., Patnaude, D. J., Slane, P., & Raymond, J. 2010,ApJ, 712, 287Fang, J., Tang, Y., & Zhang, L. 2011, ApJ, 731, 32Fukui, Y., Moriguchi, Y., Tamura, K., et al. 2003, PASJ, 55, 61Fukui, Y. 2008, in AIP Conf. Proc., Vol. 1085, Proc. of 4thInternational Meeting on High-Energy Gamma-RayAstronomy, ed. F. A. Aharonian, W. Hofmann, & F. Rieger(Melville, NY: AIP), 104Fukui, Y., & Kawamura, A. 2010, ARA&A, 48, 547Gabici, S., Aharonian, F. A., & Blasi, P. 2007, Ap&SS, 309, 365Gabici, S., Aharonian, F. A., & Casanova, S. 2009, MNRAS, 396,1629Giuliani, A., Tavani, M., Bulgarelli, A., et al. 2010, A&A, 516,L11Goldsmith, P. F., Li, D., & Krˇco, M. 2007, ApJ, 654, 273 Grenier, I. A., Casandjian, J.-M., & Terrier, R. 2005, Science,307, 1292Inoue, T., Yamazaki, R., & Inutsuka, S. 2009, ApJ, 695, 825Inoue, T., Yamazaki, R., Inutsuka, S., & Fukui, Y. 2011,arXiv:1106.3080Jenkins, E. B., & Savage, B. D. 1974, ApJ, 187, 243Jones, F. C., & Ellison, D. C. 1991, SSRv., 58, 259Jun, B.-I., & Norman, M. L, 1996, ApJ, 465, 800Katz, B., & Waxman, E. 2008, JCAP., 01, 018Koyama, K., Kinugasa, K., Matsuzaki, K., et al. 1997, PASJ, 49, 7Krˇco, M., & Goldsmith, P. F. 2010, ApJ, 724, 1402Lazendic, J. S., Slane, P. O., Gaensler, B. M., et al. 2004, ApJ,602, 271McClure-Griffiths, N. M., Dickey, John M., Gaensler, B. M., et al.2005, ApJS, 158, 178Malkov, M. A., & O’C Drury, L. 2001, RPPh., 64, 429Matsunaga, K., Mizuno, N., Moriguchi, Y., et al. 2001, PASJ, 53,1003Moriguchi, Y., Tamura, K., Tawara, Y., et al. 2005, ApJ, 631, 947Morlino, G., Amato, E., & Blasi, P. 2009, MNRAS, 392, 240Ohama, A., Dawson, J. R., Furukawa, N., et al. 2010, ApJ, 709,975Patnaude, D. J., Slane, P., Raymond, J. C., & Ellison, D. C.2010, ApJ, 725, 1476Pfeffermann, E., & Aschenbach, B. 1996, in Proc.R¨ontgenstrahlung from the Universe, ed. H. U. Zimmermann,J. Tr¨umper, & H. Yorke (MPE Rep. 263; Garching: MPE), 267Porter, T. A., Moskalenko, I. V., & Strong, A. W. 2006, ApJ, 648,L29Sano, H., Sato, J., Horachi, H., et al. 2010, ApJ, 724, 59Sato, F., & Fukui, Y. 1978, AJ, 83, 1607Slane, P., Gaensler, B. M., Dame, T. M., et al. 1999, ApJ, 525,357Tanaka, T., Uchiyama, Y., Aharonian, F. A., et al. 2008, ApJ,685, 988Torii, K., Enokiya, R., Sano, H., et al. 2011, ApJ, 738, 46Uchiyama, Y., Aharonian, F. A., & Takahashi, T. 2003, A&A,400, 567Uchiyama, Y., Aharonian, F. A., Tanaka, T., Takahashi, T., &Maeda, Y. 2007, Nature, 449, 576Wang, Z. R., Qu, Q.-Y., & Chen, Y. 1997, A&A, 318, 59Yamamoto, H., Kawamura, A., Tachihara, K., et al. 2006, ApJ,642, 307Zirakashvili, V. N., & Aharonian, F. A. 2007, A&A, 465, 695Zirakashvili, V. N., & Aharonian, F. A. 2010, ApJ, 708, 965 Fig. B2.— (Left) Velocity channel distributions of H I integrated intensity toward the SE cloud superposed on the TeV γ -ray contours.TeV γ -ray contours are plotted every 10 smoothed counts from 20 smoothed counts. The faded area in the upper panel is a componentunrelated to the SNR. (Right) Model velocity distributions of an expanding shell shown in Figure B1. Iso-velocity lines are shown here,and blue areas show the corresponding velocity range shown in the left panels. Fig. B3.— (a) Averaged brightness temperature distribution of H I in a velocity range from −
20 km s − to 0 km s − . Contours show theH.E.S.S. TeV γ -rays (Aharonian et al. 2007) and are plotted every 10 smoothed counts from 10 smoothed counts. The line AB is inclinedby 60 degrees to the Galactic plane and the line CD passes the center of the SNR. (b) Position-velocity distribution of H I along the lineAB in Figure B3(a). The velocity resolution is smoothed to 1 km s − and the integration interval is 200 arcsec. The white circle shows aschematic image of an expanding spherical shell (Figure B1). (c) H I spectra along the lines AB and CD in Figure B3(a). Expected profilesof H I self-absorption are shown by straight lines in the spectra with significant H I dips. Fig. C1.— (a) Distributions of column density of the ISM protons N p estimated from CO( J =1–0) N p (H ), (b) H I emission withoutcorrection for the self-absorption N p (H I ) and (c) sum of N p (H ) and N p (H I ). Here we assume for reference that the H I emission isoptically thin and the H I self-absorption is not taken into account. All datasets used here are smoothed to a HPBW of the TeV γ -raydistribution with a Gaussian function. (d) TeV γ -ray distribution. Contours are plotted every 50 smoothed counts from 20 smoothedcounts. Fig. C2.— (a) Distribution of column density of ISM protons N p (H +H I ) in a velocity range from −
20 km s − to 0 km s − , wherethe H I is assumed to be optically thin and without self-absorption. Contours and two elliptical rings are the same as in Figure 8(a). (b)Azimuthal distributions of N p (H ), N p (H I ), N p (H +H I ) and TeV γ -ray smoothed counts per beam in the two elliptical rings in FigureC2(a). The same plots inside of the inner ring are shown on the right side in Figure C2(b). Fig. C3.—