1: \documentclass{emulateapj}

2: 

3: %%\documentclass[12pt,preprint]{aastex}

4: 

5: %% manuscript produces a one-column, double-spaced document:

6: 

7: %%\documentclass[manuscript]{aastex}

8: 

9: %% preprint2 produces a double-column, single-spaced document:

10: 

11: %%\documentclass[preprint2]{aastex}

12: 

13: %% Sometimes a paper's abstract is too long to fit on the

14: %% title page in preprint2 mode. When that is the case,

15: %% use the longabstract style option.

16: 

17: %% \documentclass[preprint2,longabstract]{aastex}

18: 

19: %% If you want to create your own macros, you can do so

20: %% using \newcommand. Your macros should appear before

21: %% the \begin{document} command.

22: %%

23: %% If you are submitting to a journal that translates manuscripts

24: %% into SGML, you need to follow certain guidelines when preparing

25: %% your macros. See the AASTeX v5.x Author Guide

26: %% for information.

27: 

28: \newcommand{\vdag}{(v)^\dagger}

29: \newcommand{\myemail}{yashiro@ssedmail.gsfc.nasa.gov}

30: 

31: %% You can insert a short comment on the title page using the command below.

32: 

33: %\slugcomment{To be submitted to {\it ApJL}}

34: 

35: 

36: \shorttitle{Spatial Relationship between Flares and CMEs}

37: \shortauthors{Yashiro et~al.}

38: 

39: \begin{document}

40: 

41: 

42: \title{Spatial Relationship between Solar Flares and Coronal Mass

43: Ejections}

44: 

45: 

46: \author{S. Yashiro,\altaffilmark{1,2} G. Michalek,\altaffilmark{1,2,3}

47: S. Akiyama,\altaffilmark{1,2} N. Gopalswamy,\altaffilmark{2} and

48: R. A. Howard\altaffilmark{4}}

49: 

50: \altaffiltext{1}{Catholic University of America, Washington, DC 20064}

51: % yashiro@ssedmail.gsfc.nasa.gov, akiyama@ssedmail.gsfc.nasa.gov

52: 

53: \altaffiltext{2}{NASA Goddard Space Flight Center, Greenbelt, MD 20771}

54: % gopals@ssedmail.gsfc.nasa.gov

55: 

56: \altaffiltext{3}{Astronomical Observatory of Jagiellonian University,

57: Krakow, Poland}

58: 

59: \altaffiltext{4}{Naval Research Laboratory, Washington, DC 20375}

60: % russ.howard@nrl.navy.mil

61: 

62: 

63: \begin{abstract}

64: 

65: We report on the spatial relationship between solar flares and coronal

66: mass ejections (CMEs) observed during 1996-2005 inclusive.  We

67: identified 496 flare-CME pairs considering limb flares (distance from

68: central meridian $\ge 45^\circ$) with soft X-ray flare size $\ge$ C3

69: level. The CMEs were detected by the Large Angle and Spectrometric

70: Coronagraph (LASCO) on board the {\it Solar and Heliospheric

71: Observatory} ({\it SOHO}). We investigated the flare positions with

72: respect to the CME span for the events with X-class, M-class, and

73: C-class flares separately. It is found that the most frequent flare

74: site is at the center of the CME span for all the three classes, but

75: that frequency is different for the different classes. Many X-class

76: flares often lie at the center of the associated CME, while C-class

77: flares widely spread to the outside of the CME span. The former is

78: different from previous studies, which concluded that no preferred

79: flare site exists. We compared our result with the previous studies

80: and conclude that the long-term LASCO observation enabled us to obtain

81: the detailed spatial relation between flares and CMEs. Our finding

82: calls for a closer flare-CME relationship and supports eruption models

83: typified by the CSHKP magnetic reconnection model.

84: 

85: \end{abstract}

86: 

87: \keywords{Sun: flares --- Sun: CMEs}

88: 

89: 

90: \section{Introduction}

91: 

92: A solar flare is sudden flash of electromagnetic radiation (suggesting

93: plasma heating) in the solar atmosphere, and a coronal mass ejection

94: (CME) is an eruption of the atmospheric plasma into interplanetary

95: space. Both phenomena are thought to be different manifestations of

96: the same process which releases magnetic free energy stored in the

97: solar atmosphere. The spatial relation between flares and CMEs

98: contains information on the magnetic field configurations involved in

99: the eruptive process and hence is important for modeling them. Many

100: flare-CME models are based on the CSHKP (Carmichael, Sturrock,

101: Hirayama, Kopp \& Pneuman) magnetic reconnection model. The model

102: requires that a flare occurs just underneath of an erupting filament

103: which eventually becomes the core of the CME associated with the

104: flare. Normally the core corresponds to the center of the CME, thus

105: the CSHKP model requires that the flare occurs near the center of the

106: CME span.

107: 

108: Full-scale studies on the flare-CME relationship started in the 70s

109: and 80s with the CME observations obtained by the {\it Solwind}

110: coronagraph on board {\it P78-1} and the Coronagraph/Polarimeter

111: telescope on board the {\it Solar Maximum Mission} ({\it

112: SMM}). \citet{harri86} carried out a detailed analysis of three

113: flare-CME events observed by {\it SMM} and reported that flares

114: occurred near one foot of an X-ray arch, which is supposed to become a

115: CME. He also analyzed 48 flare-CME events observed by {\it SMM} and

116: {\it Solwind} and reported that many flares occurred near one leg of

117: the associated CMEs. This result, called the flare-ejection asymmetry,

118: is inconsistent with the CSHKP flare-CME model. \citet{kahle89}

119: examined 35 events observed by the {\it Solwind} and reported that

120: flare positions did not peak neither at the center nor at one leg of

121: the CMEs. They concurred with Harrison at the point that the

122: observations do not match with the CSHKP model, while disagreeing with

123: the result that flares are likely to occur at one leg of CMEs.  They

124: pointed out that the parameter employed by Harrison was biased, and

125: concluded that both observations are compatible with the fact that

126: there is no preferred flare site with respect to the CME span.  It

127: should be noted that the two studies applied different criteria for

128: the event selection. Harrison did not apply any criteria on flare

129: X-ray intensity, flare location, and CME span, while Kahler et

130: al. used only strong limb flares ($\ge$ M1 level; central meridian

131: distance (CMD) $\ge 40^\circ$) and wide CMEs (angular span $\ge

132: 40^\circ$). Different criteria might produce different spatial

133: distributions, but the results in both the studies were inconsistent

134: with the schematic view of the CSHKP type flare-CME model.

135: 

136: The Large Angle and Spectrometric Coronagraph

137: \citep[LASCO;][]{bruec95} on board the {\it Solar and Heliospheric

138: Observatory} ({\it SOHO}) mission has observed more than 11,000 CMEs

139: from 1996, which provides a great opportunity to investigate the flare-CME

140: relationship. \citet{harri06} reviewed several flare-CME studies and

141: stated that "the pre-SOHO conclusions about relative flare-CME

142: locations and asymmetry are consistent with many recent studies."

143: However, systematic statistical study is needed before reaching a firm

144: conclusion. In this paper we revisit this issue using the large CME

145: data obtained by {\it SOHO} LASCO.

146: 

147: \begin{figure*} 

148: \epsscale{0.90} \plotone{f1.eps} 

149: \caption{Three CMEs observed by {\it SOHO} LASCO to illustrate the

150: measurement of CME span. The top row shows direct images used to

151: measure the main CME body, and the bottom row shows corresponding

152: running difference images used to measure the whole CMEs. $\phi_1$ and

153: $\phi_2$ indicate the PAs of side edges of the main CME body, and

154: $\phi_A$ and $\phi_B$ indicate those of the whole CME. Arrows point to

155: the position of the flares associated with the CMEs.}

156: \end{figure*}

157: 

158: 

159: \section{Data and Analysis}

160: 

161: Solar flares are continuously monitored by the X-ray Sensor (XRS) on

162: board the {\it Geostationary Operational Environmental Satellite}

163: ({\it GOES}) missions. The XRS observes the whole-sun X-ray flux in

164: the 0.1-0.8 nm wavelength band to detect solar flares. The flare

165: location has been determined by H$\alpha$ images obtained by

166: ground-base observatories and X-ray images obtained by Soft X-ray

167: Imager (SXI) on {\it GOES}. All flares have been listed in the Solar

168: Geophysical Data (SGD) and the online solar event

169: report\footnote{\url{http://www.sec.noaa.gov/ftpmenu/indices.html}}

170: compiled by NOAA Space Environment Center. From the online report we

171: selected limb flares (CMD $\ge 45^\circ$) with soft X-ray flare size

172: $\ge$ C3 level.

173: 

174: We used the {\it SOHO} LASCO CME

175: Catalog\footnote{\url{http://cdaw.gsfc.nasa.gov/CME\_list/index.html}}

176: \citep{yashi04} to investigate the CME associations. The CME

177: candidates associated with a given flare were searched within a 3 hr

178: time window (90 minutes before and 90 minutes after the onset of the

179: flare).  However, because the time window analysis by itself could

180: produce false flare-CME pairs, we checked the consistency of the

181: associations by viewing both flare and CME movies in the Catalog. We

182: played movies obtained by the Extreme ultraviolet Imaging Telescope

183: (EIT) on {\it SOHO} and Soft X-ray Telescope (SXT) on {\it Yohkoh} to

184: look for any eruptive surface activities (e.g., filament eruptions,

185: dimmings, and arcade formations) associated with the flares. All

186: flares can be divided into those with and without CMEs except for some

187: in which the eruptive signatures were obscure. We excluded such

188: uncertain flare-CME pairs from this analysis. From 1996 to 2005, we

189: found 496 definitive flare-CME pairs.

190: 

191: 

192: A typical CME consists of a bright frontal structure (leading edge),

193: followed by a dark cavity, and a bright core. This configuration is

194: called the CME three-part structure \citep{illin85,webb88}.  The

195: bright core corresponds to the erupting filament

196: \citep{webb87,gopal03a}. There is an issue whether narrow CMEs have

197: the three-part structure or not \citep[e.g.][]{gilbe01}, but at least

198: for large CMEs, this structure is fundamental. In this paper we refer

199: to the three-part structure as the main CME body. Some CMEs possess a

200: faint envelope outside of the main CME body. The envelope might be a

201: shock wave driven by the CME \citep{sheel00,vourl03,ciara05}, thus

202: there is a problem whether the envelope is a part of the CMEs or

203: not \citep[see][]{stcyr05}. However, since there is no established way

204: to identify a shock by coronagraph observation itself, we have

205: included the envelope structures as a part of the CME and refer to all

206: the CME features as the whole CME.

207: 

208: 

209: For the comparison between flare and CME positions, it is ideal if we

210: could measure the position angles\footnote{PA is measured

211: counterclockwise from Solar North in degrees} (PAs) of the CME edges

212: on the solar limb. The innermost coronagraph C1 is the best, but its

213: data are not available for the most of the CMEs. Since several CMEs

214: did show non-radial motion \citep{gopal00,zhang04}, we measured PAs of

215: the CME edges in C2 images as close to the occulting disk as possible.

216: 

217: \begin{figure*} 

218: \epsscale{1.0}\plotone{f2.eps} 

219: 

220: \caption{

221: Distributions of flare positions with respect to the CPA of the

222: CME. The first and second columns show the PA difference in degree for

223: the main CME body and for the whole CME, respectively. The standard

224: deviation ($\sigma$) obtained by Gaussian fit is shown in each plot. The

225: third and forth columns show the distributions of PA differences

226: normalized by the half CME span. The vertical dashed lines mark the

227: two side edges of the CMEs. $P$ is the percentage of the flares lying

228: inside of the CME span. The second, third, and forth rows correspond

229: to the events with X-class, M-class, and C-class flares, respectively.}

230: 

231: 

232: \end{figure*}

233: 

234: 

235: 

236: Figure~1 illustrates how we measured the PAs of the main CME body and

237: the whole CME.  Top panels are LASCO C2 images for three CMEs with the

238: corresponding running difference images (previous images are

239: subtracted to enhance the faint structure of the CMEs) in the bottom

240: panels. The side edges of the main CME body (the whole CME) are

241: denoted by $\phi_1$ and $\phi_2$ ($\phi_A$ and $\phi_B$). The CME on

242: 1997 November 14 (Figs.~1a and 1b) did not have an envelope, thus the

243: $\phi_1$ ($\phi_2$) and $\phi_A$ ($\phi_B$) are identical. On the

244: other hand, the CME on 2000 June 25 had a faint envelope to the north

245: of the main CME body (Figs.~1c and 1d). Since it is hard to see it in

246: print, we traced out the edge of the envelope by a dotted curve on

247: Fig.~1d. The northern edge of the envelope denoted by $\phi_B$ is used

248: for the edge of the whole CME. The black-and-white radial features

249: (denoted by $S$) at both sides of the CME are a signature of streamer

250: shift caused by the expansion of the CME. We should note that we did

251: not use them for the determination of the edges of the whole

252: CME. Since we can not see an envelope to the south of the CME, the

253: southern edges of the main CME body and whole CME ($\phi_1$ and

254: $\phi_A$) are almost identical. The CME on 2005 July 14 appeared in

255: the C2 FOV at 10:54 UT (Figs.~1e and 1f). The CME had a clear

256: three-part structure with a faint envelope. The envelope covered the

257: occulting disk at 11:54 UT, thus the CME is listed as a halo

258: \citep{howar82} in the CME catalog. In this case $\phi_A$ and $\phi_B$

259: cannot be defined.

260: 

261: The location of a CME is represented by the central position angle

262: (CPA), which is defined as the mid-angle of the two side edges of the

263: CME in the sky plane. We define the CPA of the main CME body as

264: $\phi_3=(\phi_1+\phi_2)/2$ and that of the whole CME as

265: $\phi_C=(\phi_A+\phi_B)/2$. The PAs of flares ($\phi_F$) are computed

266: from their location in heliographic coordinates listed in NOAA SGD.

267: The angular span of the main CME body ($\omega_3$) and the whole CME

268: ($\omega_C$) is defined as the difference between the two side edges

269: [$\omega_3=\phi_2-\phi_1$; $\omega_C=\phi_B-\phi_A$].

270: 

271: 

272: 

273: \section{Results}

274: 

275: \subsection{PA Difference}

276: 

277: The distributions of differences between flare PAs and CME CPAs for

278: the main CME body and for the whole CME are shown in Figures~2a and

279: 2e, respectively. Fifty four halo CMEs are not used in Figure 2e since

280: their $\phi_C$ cannot be defined. Both the distributions are very

281: similar and are well represented by Gaussians. The standard deviation

282: is 17.1$^\circ$ for the difference between the flares and main CME

283: bodies and 17.5$^\circ$ for the difference between the flares and

284: whole CMEs. The average (median) angular span is 52.2$^\circ$

285: (43.8$^\circ$) for the main body of CMEs and 89.5$^\circ$

286: (75.0$^\circ$) for the whole CMEs. Thus the PA differences between

287: flares and CMEs are smaller than the angular span of the CMEs. One

288: might think that the flare site is inherently close to the center of

289: the CME and the non-radial motion below the C2 occulting disk produces

290: the PA differences. In order to explain a difference of 17$^\circ$, a

291: CME needs to erupt 30$^\circ$ away from the radial direction.

292: 

293: We separated the events into three groups according to their flare

294: intensity and made the same plots for each group. The second, third,

295: and forth rows in Figure~2 correspond to the events with X-class,

296: M-class, and C-class flares, respectively. The standard deviation is

297: shown in each plot, which ranges from 16.6$^\circ$-17.7$^\circ$ for

298: the main CME body and 16.6$^\circ$-21.1$^\circ$ for the whole CME. The

299: distributions of PA differences in the three groups are almost

300: identical, suggesting that the events with weak flares (below C3

301: level) have a similar distribution.

302: 

303: \subsection{Relative PA Difference}

304: 

305: In order to investigate the flare position with respect to the main

306: CME body (frontal structure), we normalize the PA differences by the

307: half angular span. We define relative flare location

308: $r_3=(\phi_F-\phi_3)/0.5\omega_3$. The $r_3=\pm1$ indicates the flare

309: is located at either leg of the CME frontal structure. The $r_3=0$

310: indicates that the flare is located at the center of the CME span. The

311: distribution of $r_3$ is shown in Figure~2i. It is clear that most of

312: the flares are located under the span of the main CME body. Out of the

313: 496 flares, 350 (or 71\%) resided under the span of the main CME

314: body. Figure~2m is the same as the Figure~2i, but for flare locations

315: with respect to the edges of the whole CMEs

316: [$r_C=(\phi_F-\phi_C)/0.5\omega_C$].  Again, we exclude 54 full halos

317: ($\omega_C=360^\circ$). Out of the 442 flares, 379 (or 86\%) resided

318: under the angular span of the whole CMEs. Both the distributions are

319: well represented by Gaussians with standard deviations of 0.59 for the

320: main CME body and 0.37 for the whole CME. In both the cases, the peak

321: of the Gaussian is around zero, meaning that the flares frequently

322: occur under the center of the CME span, not near one leg (outer edge)

323: of the CMEs.

324: 

325: As we did for the PA difference distributions, we separated the events

326: into three groups according to their flare intensity. The second,

327: third, and forth rows in Figure~2 correspond to events with X-class,

328: M-class, and C-class flares, respectively. $\sigma$ is the standard

329: deviation and $P$ is the percentage of the flares occurring under the

330: CME span. We found that all distributions have a peak around zero,

331: while the width of the distributions is different for different flare

332: levels. The flare-CME events with X-class flares (hereafter X-class

333: events\footnote{Similarly we labeled flare-CME events with M-class

334: (C-class) flares as M-class (C-class) events.}) have a narrower

335: distribution suggesting that many X-class flares lie under the center

336: of the CME span. On the other hand, the C-class events have a broader

337: distribution and a significant number of events occurred outside of

338: the CME span.

339: 

340: We do not see a significant distinction in the PA difference

341: distributions among the three flare levels, but we do see a difference

342: in the relative position distributions.  Since the relative position

343: is defined by the PA difference normalized by the CME half span, the

344: distinction shown in Figure 2 results from the difference in CME

345: span. The average angular span of the main CME body (the whole CME) is

346: 87$^\circ$ (224$^\circ$), 54$^\circ$ (124$^\circ$), and 38$^\circ$

347: (75$^\circ$) for X-, M-, and C-class events, respectively. As reported

348: by the previous studies \citep[e.g.,][]{yashi05}, by means of

349: statistics, CMEs associated with stronger flares have larger angular

350: span. 

351: 

352: \begin{figure}[t] \epsscale{1.00}\plotone{f3.eps} 

353: \caption{Solar cycle variation ({\it left}) and distribution ({\it

354: right}) of the latitude difference between flares and

355: CMEs. $\lambda_F$, $\lambda_3$, and $\lambda_C$ are latitudes of the

356: flares, main CME body, and whole CME, respectively. The flare-CME

357: pairs occuring in different hemispheres (e.g. a flare in the northern

358: hemisphere and a CME in the southern) were excluded.}

359: \end{figure}

360: 

361: 

362: \subsection{Latitude Difference}

363: 

364: Even though the flare PAs tend to be close to those of CMEs, we need

365: to point out some systematic offsets between flares and CMEs stemming

366: from the varying influence of the global solar magnetic field as a

367: function of the solar cycle. \citet{gopal03a} examined the

368: relationship between prominence eruptions (PEs) and CMEs and found

369: that, during solar minimum, the CPA of CMEs tend to be closer to the

370: equator compared to those of PEs, while no such effect was seen during

371: solar maximum. A similar relation is expected between flares and

372: CMEs. In order to examine whether flare positions are equatorward or

373: polarward with respect to the CMEs, we have shown their latitudinal

374: differences in Figures~3.  The CME latitudes ($\lambda_3$ and

375: $\lambda_C$) were calculated from CPAs of the main CME body ($\phi_3$)

376: and whole CME ($\phi_C$), respectively. The flare-CME pairs occurring

377: in different hemispheres were excluded.

378: 

379: 

380: The gray lines in Figs~3a and 3b show annual average of the latitude

381: difference.  In spite of the small data set during solar minimum, we

382: see a positive offset in 1997, meaning that CMEs occurred in lower

383: latitudes as compared to flares. This is consistent with the result of

384: Gopalswamy et al. On the other hand, during solar maximum, we see a

385: negative offset, indicating that flares occurred in lower latitudes as

386: compared to CMEs.  This is different from Gopalswamy et al. who

387: reported that no systematic offset exists between PE and CME

388: latitudes. The difference may result from the exclusion of

389: high-latitude flares from our analysis.  Since we did not include

390: small flares (below C3 level), the high-latitude flares (e.g. X-ray

391: arcade formations associated with the eruption of polar crown

392: filaments) were excluded.  Therefore the sampled flares mainly

393: occurred in active regions, i.e. in low latitudes. We should note that

394: high-latitude CMEs are not associated with active-region flares, but

395: appear frequently during solar maximum \citep{gopal03b}. During the

396: declining phase of solar cycle 23, CMEs have gradually started

397: clustering around the equator, while active regions have remained

398: around the equator. This is why we still see the negative offset in

399: 2004 and 2005. The positive offset between flare and CME latitude is

400: likely to resume with the start of solar cycle 24.

401: 

402: One would think that the existence of the latitude offset is

403: inconsistent with the result that flares frequently occur under the

404: center of the CME span. In order to check this, we have shown the

405: distributions of latitude differences in figures~3c and 3d. Because of

406: the exclusion of the flare-CME events from different hemispheres, the

407: distributions became narrower since the excluded events have

408: relatively larger PA differences. We can see that both the

409: distributions are asymmetric with a broad tail on the left. However,

410: the most frerqent bin stays as 0$^\circ$ and the average (median)

411: difference is -5.7$^\circ$ (-3.9$^\circ$) for the main CME body and

412: -5.2$^\circ$ (-2.0$^\circ$) for the whole CME.  We conclude that there

413: are systematic offsets between flares and CMEs, but such an offset is

414: only a small fraction ($\sim 10\%$) of the CME span.

415: 

416: \section{Summary \& Discussion}

417: 

418: We investigated the spatial relationship between solar flares and CMEs

419: for 496 pairs occurring from 1996 to 2005. It is found that the

420: distribution of the difference between flare PA and CME CPA can be

421: represented by Gaussian centered at zero with a standard deviation of

422: $\sim17^\circ$, and the distribution does not change with the flare

423: level. We examined the flare positions with respect to the CME span

424: and found that the most probable flare site is the center of the CME

425: span for all flare levels, but the width of the distributions is

426: different for different flare levels. For C-class events, the flare

427: positions widely scattered with respect to the CME span, while for

428: X-class events, most of the flares lie under the center of the CME

429: span. The result is suitable for flare-CME models typified by the

430: CSHKP reconnection model. 

431: 

432: \begin{figure}[bt] 

433: \epsscale{1.0}\plotone{f4.eps} 

434: \caption{

435: Flare position ($R$) with respect to the CME span. $R=0$ for a flare

436: centered under the CME, $R=1$ for a flare at one leg of the CME, and

437: $R>1$ for a flare lying outside of the CME span.}

438: \end{figure}

439: 

440: \input{tab1}

441: 

442: \subsection{Comparison with Previous Studies}

443: 

444: Let us compare our results with four previous studies,

445: \citet[hereafter H86]{harri86}, \citet[hereafter K89]{kahle89},

446: \citet[hereafter H91]{harri91}, and \citet[hereafter

447: H95]{harri95}. All previous studies examined flare positions with

448: respect to the CME span using data obtained by {\it Solwind} or {\it

449: SMM}. Since the observational capability of the pre-SOHO coronagraphs

450: is thought to be lower, it is possible that they did not detect the

451: faint envelope around the three-part CME structure. Therefore, for the

452: proper comparison, we used $r_3$, the flare positions with respect to

453: the main CME body.

454: 

455: H86 employed the parameter $\alpha=(\phi_2-\phi_F)/(\phi_F-\phi_1)$

456: (In Figure~4 of H86, $\phi_2$, $\phi_F$, and $\phi_1$ correspond to

457: letters $A$, $B$, and $C$). He found a significant peak at

458: $\alpha=0-0.19$, meaning that flares often occur at the one leg of the

459: CMEs. However, K89 pointed out that the bin size in equal $\alpha$ is

460: biased; the smallest bin ($\alpha=0-0.19$) is about 3 times more

461: probable than the largest bin ($\alpha=0.8-1.0$) for the random

462: distribution of flare positions. They employed the parameter

463: $R=|\phi_F-\phi_3|/0.5\omega_3$, which is the absolute of $r_3$. The

464: relation between parameters $\alpha$ and $R$ is

465: $R=(1-\alpha)/(1+\alpha)$.

466: 

467: 

468: In order to compare the previous studies properly, we converted the

469: distributions in equal $\alpha$ into those in equal $R$. For each

470: $\alpha$ bin, the frequency ($f$) in 0.01 $R$ intervals in percentage

471: is computed by dividing the number of events ($n$) by both the total

472: number of events ($N$) and the interval of $R$ ($dR$),

473: i.e. $f=n/N/dR$. For example, H86 examined 48 flare-CME events and

474: found that 3 of them are in the bin of $\alpha=0.8-1.0$, which

475: corresponds to $R=0-0.11$ ($N=48$, $n=3$, and $dR=0.12$). Then we

476: obtained the frequency for $R=0-0.11$ bin to be 0.52 (=3/48/0.12). We

477: carried out the same conversion for other bins. However, treatment of

478: the $\alpha=0-0.19$ bin is not easy. Harrison used 30 {\it Solwind}

479: events compiled by Sheeley et al. (1984) and 18 {\it SMM} events

480: compiled by Sawyer. We examined the Sheeley et al.'s list and found 7

481: events lying outside the CME span. Such flares should have negative

482: $\alpha$, but there is no corresponding bin in Fig.~5 of H86. Thus we

483: supposed that the 7 events were included in $\alpha=0-0.19$ bin, and

484: determined their $R$ using the Sheeley et al's list. Unfortunately we

485: could not locate Sawyer's 18 events. Thus we assumed that the same

486: fraction of the events in the bin inherently lied outside of the CME

487: span. The result of this analysis is shown in Figure~4a. By the

488: conversion of equal $\alpha$ to equal $R$ distribution and the special

489: treatment of $\alpha=0-0.19$ bin, the peak at the $\alpha=0-0.19$ bin

490: in Fig.~5 of H86 disappeared. For K89, we obtained the $R$

491: distribution from their Table~3. They reported 7 out of the 35 flares

492: occurring outside the CME span. However we could not find out how far

493: apart the flares were located from the nearest CME leg.  Thus we

494: assumed that the 7 events resided in $R=1.01-2.00$. The $R$

495: distribution of K89 is shown in Figure~4b. H91 has a histogram of the

496: $\alpha$ distribution and their conversion to the $R$ distribution was

497: straightforward (Fig.~4c). H95 does not have a histogram of spatial

498: distribution, but has the scatter plots of $R$ vs. flare intensity and

499: $R$ vs. flare duration. We read the $R$ value from the plots and made

500: a histogram of the $R$ distribution (Fig.~4d). For the present study,

501: we plotted the $R$ distribution for the X-class, M-class, and C-class

502: events in Figure~4e, 4f, and 4g, respectively.

503: 

504: As we showed in Section~3.2, the $R$ distribution varies according to

505: the scale of the flare-CME events. Therefore, in the comparison with

506: previous studies, we should pay attention to their data source and

507: selection criteria, which are summarized in Table~1. The second and

508: third columns show that the number of events and satellites used in

509: each study. The forth column shows the study period. H86 and K89 used

510: data during solar maximum, while H91 and H95 used data during solar

511: minimum. The present study covers the almost whole of the solar cycle

512: 23, but many events were obtained during solar maximum. The last

513: column is for event selection criteria in each study. K89 and the

514: present study used only strong flares ($\ge$ M1 and $\ge$ C3,

515: respectively), but other studies by Harrison did not eliminate weak

516: flares. K89, H95, and the present study used limb events only, which

517: reduces the projection effects.

518: 

519: Except for H95 (Fig.~4d), the three previous studies show a trend that

520: the flare occurred near the center of CMEs rather than at the

521: edges. The majority of the events in H95 were C-class flares, thus we

522: should compare the result of H95 with our C-class events

523: (Fig.~4g). The C-class events in our data show a trend that flare

524: occurred near the CME center, but the trend is very weak. Therefore it

525: is not surprising that the examination of the 25 events can not see

526: such weak trend. On the other hand, H91 shows a strong peak at the

527: center of the CME even though the data were obtained during solar

528: minimum (1984-87).  We could not find a statement of exclusion of disk

529: events, thus the projection effects might produce the peak (Apparent

530: CME span becomes larger than inherent span if the distance from the

531: limb is farther). Except for the lack of flares under the center of

532: the CME span ($R<0.25$), the $R$ distribution of K89 (Fig.~4b) is

533: similar to that of our M-class events (Fig.~4f), which can be

534: explained by their selection criterion about flare intensity ($\ge$ M1

535: level). By the same token, because 30\% and 57\% of the H86 events

536: were X-class and M-class flares, respectively, Fig.~4a should be the

537: similar to Fig.~4e or Fig.~4f. However the distribution is similar to

538: that of C-class events (Fig.~4g). Additionally H86 also lacks flares

539: under the center of the CME span ($R<0.25$). The average CME span of

540: H86 events (64$^\circ$) is larger than that of our M-class events

541: (56$^\circ$), suggesting that the different capability between SOHO

542: LASCO and previous coronagraphs does not explain the H86 lack of

543: flares under the center of the CME span. Except for this discrepancy,

544: all the flare position distributions reported in the previous studies

545: are consistent with our results. The long-term LASCO observation

546: enabled us obtain a large number of flare-CME pairs from small to

547: large events for the first time that revealed the detailed spatial

548: relation between flares and CMEs.

549: 

550: \subsection{Flare-CME Geometry}

551: 

552: The flare-CME asymmetry found by H86 has been the basis of the claim

553: that flare-CME observations are inconsistent with the schematic

554: picture of the CSHKP type flare-CME models. In this paper we show that

555: most of the X-class flares are located at the center of the CME span

556: while a significant number of C-class flares reside near the edge or

557: even outside of the CME span. The CSHKP type flare-CME models are well

558: suited for strong events, but may not be applicable for the many weak

559: events.  The other extreme is the non-eruptive (or compact) flares

560: which do not involve any mass motion, and hence their geometry may not

561: be appropriate for CSHKP models.

562: 

563: 

564: The flare-CME geometry is possibly different between weak (narrow) and

565: strong (wide) events. This is related to the issue whether narrow CMEs

566: are physically distinct from general CMEs

567: \citep{kahle89,kahle01}. \citet{reames02} presented two types of

568: flare-CME geometry which are responsible for two types of solar

569: energetic particle (SEP) events, i.e. impulsive and gradual

570: events. The gradual SEP events are associated with large CMEs, which

571: fit CSHKP type models, while the impulsive SEPs are associated with

572: narrow CMEs that fit the X-ray jet model

573: \citep{shimo00}. \citet{bempo05} reported that blob-like narrow CMEs

574: in a streamer (called "streamer puffs") differ from general CMEs

575: \citep[see also][]{moore07}. The streamer puffs are associated with

576: weak flares (below C4 level) and their schematic picture clearly

577: explains the flare-CME asymmetry.

578: 

579: Even X-class events, some of them showed the clear flare-CME

580: asymmetry. A good example is the event on 2002 May 20. The X2.1 flare

581: at 15:21 UT located at one edge of the CME at 15:50 UT. Another

582: example is the event on 2003 November 3 (The X2.7 flare at 01:09 UT

583: and the CME at 01:59 UT). In both cases EIT dimmings were clearly

584: observed only on the CME side of the flares. It is important to

585: investigate the flare-dimming asymmetry for understanding the origin

586: of such flare-CME asymmetry.

587: 

588: We examined the flare positions with respect to CME spans using LASCO

589: data and found that most frequent flare site is the center of the CME

590: span. However, since we examined the spatial relationship using limb

591: events, our finding can apply only in the latitudinal direction. The

592: flare-CME geometry in the longitudinal direction has never been

593: examined. The {\it Solar TErrestrial RElations Observatory} ({\it

594: STEREO}) mission started to observe CMEs in stereoscopic

595: view. Three-dimensional structure of the CMEs and their relation to

596: the associated flares should be tested again using {\it STEREO} data.

597: 

598: 

599: \acknowledgments

600: 

601: The authors would like to thank to the referee whose suggestions and

602: comments led to improvement of the manuscript. {\it SOHO} is a project

603: of international cooperation between ESA and NASA. The LASCO data used

604: here are produced by a consortium of the Naval Research Laboratory

605: (USA), Max-Planck-Institut fuer Aeronomie (Germany), Laboratoire

606: d'Astronomie (France), and the University of Birmingham (UK). Part of

607: this effort was supported by NASA (NNG05GR03G). Work done by G. M. was

608: partly supported by MNiSW through the grant N203 023 31/3055.

609: 

610: 

611: %\clearpage

612: 

613: \begin{thebibliography}{}

614: 

615: \bibitem[Bemporad et~al.(2005)]{bempo05}

616: Bemporad, A., Sterling, A. C., Moore, R. L., \& Poletto, G. 2005, \apj, 635, L189

617: 

618: \bibitem[Brueckner et~al.(1995)]{bruec95}

619: Brueckner, G. E., et~al. 1995, \solphys, 162, 357

620: 

621: \bibitem[Ciaravella et~al.(2005)]{ciara05}

622: Ciaravella, A., Raymond, J. C., \& Kahler, S. W. 2005, \apj, 621,

623: 1121

624: 

625: \bibitem[Gilbert et~al.(2001)]{gilbe01}

626: Gilbert, H R., Serex, E. C., Holzer, T. E., MacQueen, R. M., \&

627: McIntosh, P. S. 2001, \apj, 550, 1093

628: 

629: \bibitem[Gopalswamy et~al.(2000)]{gopal00} 

630: Gopalswamy, N., Hanaoka, Y., \& Hudson, H. S. 2000, {Adv. Sp. Res.},

631: 25(9), 1851

632: 

633: \bibitem[Gopalswamy et~al.(2003a)]{gopal03a} 

634: Gopalswamy, N., Shimojo, M., Lu, W., Yashiro, S., Shibasaki, K., \&

635: Howard, R. A. 2003a, \apj, 586, 562

636: 

637: \bibitem[Gopalswamy et~al.(2003b)]{gopal03b} 

638: Gopalswamy, N., Lara, A., Yashiro, S., \& Howard, R. A. 2003b, \apj,

639: 598, L63

640: 

641: \bibitem[Harrison(1986)]{harri86}

642: Harrison, R. A. 1986, \aap, 162, 283

643: 

644: \bibitem[Harrison(1991)]{harri91}

645: Harrison, R. A. 1991, Adv. Space Res., 11, 25

646: 

647: \bibitem[Harrison(1995)]{harri95}

648: Harrison, R. A. 1995, \aa, 304, 585

649: 

650: \bibitem[Harrison(2006)]{harri06} 

651: Harrison, R. A. 2006, in Solar Eruptions and Energetic Particles, AGU

652: Monograph 165, 73

653: 

654: \bibitem[Howard et~al.(1982)]{howar82}

655: Howard, R. A., Michels, D. J., Sheeley, N. R., \& Koomen,

656: M. J. 1982, \apj, 263, L101

657: 

658: %\bibitem[Howard et~al.(1985)]{howar85} 

659: %Howard, R. A., Sheeley, N. R., Michels, D. J., \& Koomen, M. J. 1985,

660: %\jgr, 90, 8173

661: 

662: %\bibitem[Hundhausen(1993)]{hundh93} 

663: %Hundhausen, A. J. 1993, \jgr, 98, 113

664: 

665: \bibitem[Illing \& Hundhausen(1985)]{illin85}

666: Illing, R. M. E. \& Hundhausen, A. J. 1985, \jgr, 90, 275

667: 

668: \bibitem[Kahler et~al.(1989)]{kahle89}

669: Kahler, S. W., Sheeley, N. R., \& Ligget, M. 1989, \apj, 344, 1026

670: 

671: \bibitem[Kahler et~al.(2001)]{kahle01}

672: Kahler, S. W., Reames, D. V., \& Sheeley, N. R. 2001, \apj, 562, 558

673: 

674: \bibitem[Moore \& Sterling(2007)]{moore07}

675: Moore, R. L. \& Sterling, A. C. 2007, \apj, 661, 543

676: 

677: \bibitem[Reames(2002)]{reames02}

678: Reames, D. V. 2002, \apj, 571, L63

679: 

680: \bibitem[Sheeley et~al.(2000)]{sheel00}

681: Sheeley, N. R., Hakala, W. N., \& Wang, Y.-M. 2000, \jgr, 105, 5081

682: 

683: \bibitem[Shimojo \& Shibata(2000)]{shimo00}

684: Shimojo, M. \& Shibata, K. 2000, \apj, 542, 1100

685: 

686: 

687: \bibitem[St. Cyr(2005)]{stcyr05}

688: St. Cyr, O. C. 2005, Eos Trans. AGU, 86(30), 281

689: 

690: \bibitem[Vourlidas et~al.(2003)]{vourl03}

691: Vourlidas, A., Wu, S. T., Wang, A. H., Subramanian, P., \& Howard,

692: R. A. 2003, \apj, 598, 1392

693: 

694: \bibitem[Webb(1988)]{webb88}

695: Webb, D. F. 1988, \jgr, 93, 1749

696: 

697: \bibitem[Webb \& Hundhausen(1987)]{webb87}

698: Webb, D. F. \& Hundhausen, A. J. 1987, \solphys, 108, 383

699: 

700: \bibitem[Yashiro et~al.(2004)]{yashi04}

701: Yashiro, S., Gopalswamy, N., Michalek, G., St. Cyr, O. C., Plunkett, S.

702: P., Rich, N. B., \& Howard, R. A. 2004, \jgr,

703: 109, 7105

704: 

705: \bibitem[Yashiro et~al.(2005)]{yashi05} 

706: Yashiro, S., Gopalswamy, N., Akiyama, S., Michalek, G., \& Howard,

707: R. A. 2005, \jgr, 110, A12S05

708: 

709: %\bibitem[Yashiro et~al.(2006)]{yashi06} 

710: %Yashiro, S., Akiyama, S., Gopalswamy, N., \& Howard,

711: %R. A. 2006, \apj, 650, L143

712: 

713: \bibitem[Zhang et~al.(2004)]{zhang04} 

714: Zhang, J, Dere, K. P., Howard, R. A., \& Vourlidas, A. 2004, \apj,

715: 604, 420

716: 

717: \end{thebibliography}

718: 

719: 

720: 

721: 

722: 

723: 

724: 

725: 

726: 

727: 

728: 

729: 

730: \end{document}

731: