Masatoshi Fukushima: Selecta 9783110215250, 9783110215243

Masatoshi Fukushima is one of the most influential probabilists of our times. His fundamental work on Dirichlet forms an

211 52 48MB

English Pages 559 Year 2010

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Frontmatter
Contents
Professor Masatoshi Fukushima – Scholar and Mentor
A construction of reflecting barrier Brownian motions for bounded domains
On boundary conditions for multi-dimensional Brownian motions with symmetric resolvent densities
Regular representations of Dirichlet spaces
Dirichlet spaces and strong Markov processes
On the generation of Markov processes by symmetric forms
Almost polar sets and ergodic theorem
On an LP-estimate of resolvents of Markov processes
Dirichlet spaces and additive functionals of finite energy
A note on irreducibility and ergodicity of symmetric Markov processes
Capacitary maximal inequalities and an ergodic theorem
Basic properties of Brownian motion and capacity on the Wiener space
A transformation of a symmetric Markov process and the Donsker-Varadhan theory
(r,p)-capacities for general Markov semigroups
On the continuity of plurisubharmonic functions along conformal diffusions
On Dirichlet forms for plurisubharmonic functions
On quasi-supports of smooth measures and closability of pre-Dirichlet forms
Dirichlet forms, diffusion processes and spectral dimensions for nested fractals
On a spectral analysis for the Sierpinski gasket
On a strict decomposition of additive functionals for symmetric diffusion processed
On a decomposition of additive functionals in the strict sense for a symmetric Markov process
Construction and decomposition of reflecting diffusions on Lipschitz domains with Hölder cusps
On semi-martingale characterizations of functions of symmetric Markov processes
On regular Dirichlet subspaces of H1(I) and associated linear diffusions
Entrance law, exit system and Lévy system of time changed processes
Extending Markov processes in weak duality by Poisson point processes of excursions
Backmatter
Recommend Papers

Masatoshi Fukushima: Selecta
 9783110215250, 9783110215243

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Masatoshi Fukushima Selecta

Masatoshi Fukushima

Masatoshi Fukushima Selecta Edited by

Niels Jacob Yoichi Oshima Masayoshi Takeda

De Gruyter

Editors Niels Jacob Department of Mathematics Swansea University Singleton Park Swansea SA2 8PP, Wales, United Kingdom E-mail: [email protected] Yoichi Oshima Faculty of Engineering Kumamoto University 2-39-1 Kurokami Kumamoto 860, Japan E-mail: [email protected] Masayoshi Takeda Mathematical Institute Tohoku University Aoba, Sendai, 980-8578, Japan E-mail: [email protected]

ISBN 978-3-11-021524-3 e-ISBN 978-3-11-021525-0 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2010 Walter de Gruyter GmbH & Co. KG, Berlin/New York Printing and binding: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany www.degruyter.com

Preface

Professor Masatoshi Fukushima is one of the most influential probabilists of our times. His fundamental work on Dirichlet forms and Markov processes made Hilbert space methods a tool in stochastic analysis and by this he opened the way to several new developments. His impact on a new generation of probabilists in his native country as well as in many other countries can hardly be overstated. In publishing a selection of his seminal papers we aim to serve the community, and at the same time we want to express our appreciation of a highly respected, humane scholar. All owners of copyrights of papers being included in the Selecta (see the list at the end of this volume) followed the old and good tradition to grant permission for a reproduction in a Selecta without any charge. Unfortunately this is no longer a policy adopted by all publishers. Therefore we are especially grateful to those who by their generosity continue to support the mathematical community in keeping the tradition of publishing selected or collected works alive. The editors’ thanks go to all who supported us in our enterprise, in particular we want to mention Professor T. Uemura (Kansai University), Dr. K. P. Evans (Swansea University) as well as S. Albroscheit and Dr. R. Plato (Walter de Gruyter). Swansea, Kumanoto and Sendai Fall 2009

Niels Jacob ¯ Y¯oichi Oshima Masayoshi Takeda

V

VI

Curriculum vitae

1935, August 23 1959 1961 1962 1963 1966 1967 1969–1971 1972 1977 1980 1990 1998 1998 2003 2006 2007

Born in Osaka, Japan Graduated from Kyoto University, Faculty of Science Master degree from Kyoto University, Faculty of Science Research Associate of Nagoya University, Faculty of Science Lecturer of Kyoto University, College of General Education Associate Professor of Tokyo University of Education, Faculty of Science Doctor of Science from Osaka University Post Doctral Fellow, University of Illinois, Department of Mathematics, Urbana-Champaign Associate Professor of Osaka University, Faculty of Science Professor of Osaka University, College of General Education Visiting Professor Bielefeld University, Faculty of Physics Professor of Osaka University, Faculty of Engineering Sciences Professor Emeritus, Osaka University Professor of Kansai University, Faculty of Engineering Awarded Analysis Prize of Mathematical Society of Japan Retired from Kansai University Honorary Professor Swansea University

VII

VIII

Contents

Entries in square brackets refer to the bibliography on pages 543–549.

Preface                                        

v

Curriculum vitae                                  

vii

Professor Masatoshi Fukushima – Scholar and Mentor by N. Jacob      

1

A construction of reflecting barrier Brownian motions for bounded domains [R5]                                         

3

On boundary conditions for multi-dimensional Brownian motions with symmetric resolvent densities [R6]                        

36

Regular representations of Dirichlet spaces [R8]                

72

Dirichlet spaces and strong Markov processes [R9]               

91

On the generation of Markov processes by symmetric forms [R11]      

131

Almost polar sets and ergodic theorem [R12]                  

165

On an  -estimate of resolvents of Markov processes [R19]         

181

Dirichlet spaces and additive functionals of finite energy [R22]        

189

A note on irreducibility and ergodicity of symmetric Markov processes [R29]

196

Capacitary maximal inequalities and an ergodic theorem [R31]        

204

Basic properties of Brownian motion and capacity on the Wiener space [R32]

211

A transformation of a symmetric Markov process and the Donsker-Varadhan theory (with M. Takeda) [R34]                          

226

  -capacities for general Markov semigroups (with H. Kaneko) [R37]  

242

IX

Contents

On the continuity of plurisubharmonic functions along conformal diffusions [R38]                                        

249

On Dirichlet forms for plurisubharmonic functions (with M. Okada) [R43] 

256

On quasi-supports of smooth measures and closability of pre-Dirichlet forms (with Y. LeJan) [R49]                                

299

Dirichlet forms, diffusion processes and spectral dimensions for nested fractals [R53]                                    

308

On a spectral analysis for the Sierpinski gasket (with T. Shima) [R55]    

319

On a strict decomposition of additive functionals for symmetric diffusion processed [R58]                                  

354

On a decomposition of additive functionals in the strict sense for a symmetric Markov process [R59]                               

359

Construction and decomposition of reflecting diffusions on Lipschitz domains with Hölder cusps (with M. Tomisakai) [R61]                  

374

On semi-martingale characterizations of functions of symmetric Markov processes [R66]                                   

411

On regular Dirichlet subspaces of  1   and associated linear diffusions (with X. Fang and Y. Ying) [R78]                         

443

Entrance law, exit system and Lévy system of time changed processes (with Z.-Q. Chen and J. Ying) [R81]                       

458

Extending Markov processes in weak duality by Poisson point processes of excursions (with Z.-Q. Chen and J. Ying) [R82]                 

497

Acknowledgments                                 

541

Bibliography                                    

543

X

Professor Masatoshi Fukushima – Scholar and Mentor

Let me start with some remarks on the scientific achievement of Professor Fukushima. First investigations of Professor Fukushima were related to diffusions under boundary conditions and this led him to consider related Hilbert spaces. From here the road was open to Dirichlet forms, Beurling’s and Deny’s axiomatisation of the notion of energy in potential theory. The breakthrough was the 1971 paper in the Transactions of the American Mathematical Society where a Hunt process was constructed associated to a given regular Dirichlet form. This result was immediately recognised by experts as an outstanding one, already 1978 Professor Fukushima was an invited speaker in the session on Probability Theory in the ICM in Helsinki. Adding to this I would like to mention two other honours Professor Fukushima received: The Analysis Prize from the Japanese Mathematical Society and being an Invited Lecturer of the London Mathematical Society. The 1971 paper and the book “Dirichlet forms and Markov Processes” published in 1980 in English changed the landscape of modern probability theory. Of course there are many other contributions of Professor Fukushima’s worth mentioning: 

exceptional sets and refinements



plurisubharmonic functions (especially the Acta Mathematica Paper with M. Okada)



stochastic analysis on fractals



boundary behaviour and traces of Markov processes

and many more. Of particular importance was and is the influence of his work to mathematical physics. The construction of diffusion processes on infinite dimensional state spaces highly depend on his seminal contribution. The impact of the “new” book “Dirichlet Forms and Symmetric Markov Processes” written jointly with Y. Oshima and M. Takeda can hardly be overstated. In 1990 when participating in a conference organised by Professor Kunita in Nagoya, I once had a coffee with Professor Shinzo Watanabe. In our discussion Professor Watanabe stated that for him in the 1970’s there had been two major breakthroughs in probability theory: Malliavin calculus and Fukushima’s theory of Hunt processes  This is a slightly modified version of a banquet speech given during a meeting to celebrate Professor Fukushima’s 70th birthday.

1

Professor Masatoshi Fukushima – Scholar and Mentor

associated with Dirichlet forms – Professor Watanabe is a very modest person – he should have added his own contributions too. However there is no doubt, Professor Fukushima’s work is of lasting impact. Mathematics is an international subject and Professor Fukushima was and is acting on the international stage. This is natural to all the outstanding Japanese probabilists raised in Professor K. Itô’s school. Professor Fukushima’s work on the 1971 paper was partly done when being in the U.S.A. with the late Professor Doob. Here he also established contacts to Martin Silverstein and he, Professor Fukushima, always emphasised Professor Silverstein’s contributions to our subject. Professor Fukushima was very engaged in the series of Japanese-Russian seminars on probability theory and a frequent visitor to European countries. In addition he was a great help and supporter of many young non-Japanese mathematicians, Y. Lejan, J. Kim, Z.-M. Ma, Z.-Q. Chen, J. Ying, . . . , and of course I have to mention myself. Being a world-open mathematician, a scholar who has visited (partly for longer periods) many countries is one aspect of Professor Fukushima. There is another one: As many cosmopolitans he is deeply rooted in his own culture, i.e. in the traditional Japanese culture. This makes any encounter with him also an encounter with Japan. Through him I myself as well as quite a few of my students and colleagues learnt to appreciate Japan’s great culture. Professor Fukushima belongs to the generation whose childhood was in war-time – I recommend everyone to read Kappa Senoh’s “A Boy Called H” to get a feeling of what this meant to his generation. He as many other Japanese scientists, writers and artists of his generation took on the difficult task to assure his country a respected place in the modern post-war world – and they were rather successful. A final more personal word. Due to his relations to the late Professor Heinz Bauer we met first in Erlangen. I am very happy that I could build on Professor Bauer’s contacts and could even extend them. This refers to the Oberwolfach meetings on Dirichlet forms or a German-Japanese exchange programme supported by DFG and JSPS. Moreover I am grateful that several of my own (former) PhD students could not only visit Japan but could start to build up their own contacts bringing the collaboration to the next generation. You, Professor Fukushima, have made lasting contributions to Mathematics and you have been over the years of great support to many of us. For this we are grateful and we are looking forward to many further years to come with your company. Niels Jacob

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

Probab. Theory Relat. Fields 106, 521–557 (1996)

c Springer-Verlag 1996

Construction and decomposition of reflecting diffusions on Lipschitz domains with H¨older cusps Masatoshi Fukushima1 , Matsuyo Tomisaki2 1 Department of Mathematical Science, Faculty of Engineering Science, Osaka University, Toyonaka, Osaka, Japan (e-mail: [email protected]) 2 Department of Mathematics, Faculty of Education, Yamaguchi University, Yamaguchi, Japan (e-mail: [email protected]. yamaguchi-u.ac.jp)

Received: 4 October 1995

Summary. We consider a d -dimensional Euclidean domain D whose boundary is Lipschitz continuous but admits locally finite number of outward or inward H¨older cusp points. Using a method of Stampacchia and Moser for PDE, we first construct a conservative diffusion process on the Euclidean closure of D possessing a strong Feller resolvent and associated with a second order uniformly elliptic differential operator of divergence form with measurable coefficients aij . The sample path of the constructed diffusion can be uniquely decomposed as a sum of a martingale additive functional and an additive functional locally of zero energy. The second additive functional will be proved to be of bounded variation with a Skorohod type expression whenever aij is weakly differentiable and the H¨older exponent at each outward cusp boundary point is greater than 1 2 regardless the dimension d . Mathematics Subject Classification (1991): 60J60, 60J55, 35J25, 31C25

1 Introduction Let D be a domain in the d -dimensional Euclidean space R d and D = D ! 'D  R d be its closure. The d -dimensional Lebesgue measure is denoted by m = m(dx ) or simply by dx . Given measurable functions aij (x ) 1 i  j d  on D such that aij = aji 

1 2



aij (x )i j 2  x  D   R d 

(1.1)

1i j d

for some constant   1, we consider a Dirichlet form  on L2 (D) = L2 (D; m) defined by 374

522

M. Fukushima, M. Tomisaki

 1

[ ] = H (D)  (u ) =



aij (x )'i u(x )'j (x )dx  u   H 1 (D)

D 1i j d

(1.2) where H 1 (D) = #u  L2 (D) : 'i u  L2 (D) 1 i d  the Sobolev space of order 1. Let #Tt  t 0 be the strongly continuous semigroup of Markovian symmetric operators on L2 (D) associated with the Dirichlet form  . We denote by C0 (D) [resp. B0 (D)] the space of continuous functions [resp. bounded measurable functions] on D with compact support [resp. vanishing outside a bounded set]. We further denote by C (D) [resp. C0 (D)] the space of bounded continuous functions on D [resp. the restrictions to D of functions in C0 (R d )]. Suppose that the Dirichlet form  is regular on L2 (D) rather than on L2 (D) in the sense that H 1 (D)  C0 (D) is dense in the space H 1 (D). This is the case for instance when the domain D is of class C in the sense that 'D is locally expressible as a graph of a continuous function of d  1 variables ([16]). According to general theorems ([13]), there exists then a conservative diffusion process M = (Xt  Px ) on D associated with the Dirichlet form  in the sense that the transition probability pt (x  E ) = Px (Xt  E ) of M satisfies that pt f is a version of Tt f for any f  B0 (D)

(1.3)

However we are now concerned with a highly non-trivial problem of constructing the process M on D with a strong Feller resolvent: G (B0 (D))  C (D)

(1.4)

which particularly implies the absolute continuity of the transition probability: pt (x  ) " m

for any t 0 and x  D

(1.5)

If both the conditions (1.3) and (1.5) are fulfilled, then we can invoke a general decomposition theorem in [13] of additive functionals (AFʼs in abbreviation) in the strict sense to conclude that the sample path Xt = (Xt1      Xtd ) of M admits the unique decomposition Xti  X0i = Mti + Nti 

1 i d  Px as for any x  D

(1.6)

where Mti are martingale additive functionals (MAFʼs in abbreviation) in the strict sense with covariations  M i  M j t = 2

t

aij (Xs )ds 1 i  j d  Px as for any x  D

(1.7)

0

and Nti are continuous additive functionals (CAFʼs in abbreviation) in the strict sense locally of zero energy. Nti are not necessarily of bounded variation (on each finite time interval) but locally of zero quadratic variation in a certain sense ([13]). Natural questions arise:

375

Reflecting diffusions on H¨older domains

523

(I) Under what condition on the domain D, there exists a conservative diffusion process M = (Xt  Px ) on D satisfying (1.3) and (1.5) ? (II) Under what additional conditions on the domain D and coefficients aij , the second terms Nti of Xti are of bounded variation Px a.s. for any x  D ? When D is a general bounded Lipschitz domain and aij = 12 ij  1 i  j d , Bass and Hsu gave affirmative answer to the both questions (I) and (II) in [2] and [3] respectively. Actually [2] and its refinement [13; Example 5.2.2] gave an explicit expression of Nti as Nti =

1 2



t

ni (Xs )dLs  1 i d 

Px as for any x  D

(1.8)

0

where n = (n1      nd ) is the inward unit normal vector at the boundary 'D and Lt is a positive continuous additive functional (a PCAF in abbreviation) in the strict sense associated with the surface measure on 'D; the local time of Xt on the boundary. In this case, the diffusion M is called the (normally) reflecting Brownian motion on D and the decomposition (1.6) with (1.8) is called its Skorohod representation. The first process (Mt1      Mtd ) appearing in (1.6) is the standard d -dimensional Brownian motion starting at the origin in this case. On the other hand, by extending a work of S.R.S.Varadhan and R.J.Williams on an infinite two-dimensional wedge [22], DeBlassie and Toby [7] have formulated under a submartingale problem a normally reflecting Brownian motion on a two-dimensional standard outward cusp domain 

C = (x  y)  R 2 : y  x   0    1 and constructed it from the normally reflecting Brownian motion on the upper half plane by means of a conformal map and a random time change. They have also shown in [8] that the constructed process admits the Skorohod representation if  12 but otherwise the process starting at the origin fails to be a semimartingale. By thinking of the direct product of the DeBlassie-Toby reflecting Brownian motion on C with the standard d  2-dimensional Brownian motion, we see that 1 older exponent for the semi-martingale property 2 is still the critical value of the H¨ of the reflecting Brownian motion on the special H¨older domain C  R d 2  R d . It is therefore tempting to consider the problem (I) for a general H¨older domain D and further look for a critical value of the H¨older exponent  with regard to the question (II). In this paper, we do not deal with a most general H¨older domain. However we assume that D is a general (not necessarily bounded) Lipschitz domain allowing locally finite number of outward or inward cusp boundary points with H¨older exponents uniformly bounded away from zero. Our first aim is to give an affirmative answer to the problem (I) (Theorem 2.1 and Theorem 2.2) by employing the PDE methods of Stampacchia and Moser. We then assume that (1.9) 'j aij  L

loc (D) 1 i  j d 

376

524

M. Fukushima, M. Tomisaki

and give an affirmative answer to the question (II) under the condition that the H¨older exponent at each outward cusp boundary point is greater than 12 regardless the dimension d . Actually an explicit expression of Nti using the boundary local time Lt will be derived in this case by invoking an extended version of a general theorem in [13] to characterize Nti and by combining the Sobolev inequalities obtained in Sect. 3 with the upper bounds of transition functions due to CarlenKusuoka-Stroock [5] (Theorem 2.3). Furthermore, we shall see that the diffusion process constructed in Theorem 2.2 can be, under the condition that 'j aij  L (D), related to a submartingale problem (Theorem 2.4), and accordingly, identified in law with VaradhanWilliamsʼs [resp. DeBlassie-Tobyʼs] normally reflecting Brownian motion when aij (x ) = 12 ij and D is a wedge [resp. a cusp C ] in R 2 . The present paper is an essential improvement of the previous one [14] where we gave affirmative answers to questions (I) and (II) only under the restriction that the H¨older exponents at cusps are uniformly greater than d 1 d , which was technically required in getting a modified Sobolev inequality of Moserʼs type - a key inequality in our construction of a strong Feller resolvent. This requirement now turns out to be unnecessary thanks to a specific transformation of a standard cusp domain onto a rectangular set exhibited in the last section. In the next section, we shall formulate a precise condition on the domain D and state main theorems answering the questions (I) and (II). Their proof will be carried out in the subsequent sections.

2 Statement of main theorems  Let F be a real valued function defined on a set E  R k including the origin such that F (x ) = x  +f (x ), where 0    1  R, and f is a k -dimensional Lipschitz continuous function vanishing at the origin. Here    denotes the Euclidean norm. In this paper we call such F a H¨older function and we denote its H¨older exponent, H¨older constant and Lipschitz constant respectively by Exp(F ) = 

H¨ol(F ) = 

Lip(F ) = Lip(f ) = min #K 0 : f (x )  f (y) K x  y x  y  E   For x = (x1      xd )  R d , we let x = (x1      xd 1 ) so that x = (x  xd ). Let us now consider the following condition (H) on a domain D  R d with d  2: (H) There are four constants   (0 1)  0 A  1 M 0 and a locally finite open covering #Uj j J of 'D satisfying the following properties : (i)

For each j  J , there are a H¨older function Fj of d  1 variables and a constant rj  such that Fj is defined on the d  1-dimensional ball centered at the origin with radius rj , Exp(Fj )  ,

377

Reflecting diffusions on H¨older domains

525

H¨ol(Fj ) = 0, or 1 A H¨ol(Fj ) A, or A H¨ol(Fj ) 1 A, Lip(Fj ) M , Uj  D = # = (  d ) :   rj  Fj ( )  d , for some Cartesian coordinate  system = (  d ).   U (ii) 'D  j  , where Uj  = #x  Uj : dist(x  'Uj ) . j J

When D is bounded, condition (H) reduces to a simple one that every point x of 'D has a neighbourhood Ux such that 'D  Ux is the graph of a H¨older function of d  1 variables. For later convenience, we let J+ = #j  J : H¨ol(Fj ) 0  J0 = #j  J : H¨ol(Fj ) = 0  J = #j  J : H¨ol(Fj )  0  For j  J , denote by aj ( 'D) the origin of Uj with respect to the coordinate system . aj is called an outward [resp. inward ] cusp boundary point of D if j  J+ [resp. j  J ].  In what follows, we work with the Dirichlet form   H 1 (D) on L2 (D) given by (1.1) and (1.2). Let #G  0 be the associated resolvent on L2 (D). It is then Markovian in the sense that 0 G f 1 whenever 0 f 1 and p it is well defined for any p  [1 ].  as a bounded linear operator on L (D) Denote by C D the space of those functions in C D vanishing at infinity. Theorem 2.1 Assume that a domain D  R d satisfies condition (H). Then G enjoys the following properties : p  L (D)  C D  p 1+ (d (i) G L2 (D)  1) .  (ii) G C D is a dense subspace of C D . (iii) There is a function G (x  y) continuous on D  D off diagonal such that   G f (x ) = G (x  y)f (y) dy x  D f  C D  (2.1) D

As will be seen in Sect. 4, Theorem 2.1 is still valid under condition (A) stated in Sect. 3. Condition (A) is weaker but less concrete than (H) so that we employ (H) in formulating main theorems. Theorem 2.1 (i) means that G has a strong Feller property. By virtue of Theorem 2.1 (ii) and the Hille-Yosida theorem, there  exists a strongly  t

0 on C D such that G f = continuous Markovian semigroup #T t

  t e T f dt f  C D . We have then a Feller transition function by t

0  Tt f (x ) = D pt (x  dy) f (y), which gives rise to a Hunt process (cf. [13; Theorem A.2.2]) M = (Xt  Px ) on D such that  Px (Xt  A) = pt (x  A) t 0 x  D A   D   M is associated withthe Dirichlet form   H 1 (D) of (1.2) since the re solvent G is. Since G C0 D is dense in the Dirichlet space, Theorem 2.1

378

526

M. Fukushima, M. Tomisaki

(ii) implies that the Dirichlet form  is regular. Therefore we can apply general theorems in [13] to the associated pair  and M. In particular, pt (x  ) is absolutely continuous because G (x  ) is ([13; Theorem 4.2.4]). Since  has the strong local property and aij are uniformly bounded, we can invoke [13; Theorem 4.5.4]) and [13; Theorem 5.7.2, Example 5.7.1] to conclude that M is a conservative diffusion process on D. Summing up what has been mentioned, we get Theorem 2.2 Under condition (H), there exists a conservative diffusion process M = (Xt  Px ) on D with resolvent G of Theorem 2.1. M is associated with the Dirichlet form given by (1.1) and (1.2), and the transition function pt (x  ) of M satisfies (1.3) and (1.5). We next formulate a decomposition of the sample path of M and its Skorohod representation. Theorem 2.3 Consider a domain D  R d satisfying condition (H). (i) The sample path Xt = (Xt1      Xtd ) of the conservative diffusion process M on D constructed in Theorem 2.2 admits a unique decomposition (1.6) with MAFʼs Mti in the strict sense satisfying (1.7) and CAFʼs Nti in the strict sense locally of zero energy. (ii) Assume condition (1.9) for aij . We also require the condition that Exp(Fj )

1  2

j  J+ 

(2.2)

for the domain D. Then Nti has the following representation : Nti

=

d  t  



'j aij (Xs ) ds +

j =1

0

d   j =1

t

aij (Xs ) nj (Xs ) dLs 

0

(2.3)

1 i d  t  0 Px  a.s. for any x  D where Lt is a unique PCAF in the strict sense with Revuz measure being the surface measure on 'D. Note that (2.3) reduces to (1.8) when aij = 12 ij . The above three theorems extend those results of R.F.Bass and P.Hsu in [2] and [3] formulated for a general bounded Lipschitz domain D and for aij = 12 ij .  Let us denote by + the set of all outward cusp boundary points. Cb2 D will stand for the set of twice continuously differentiable functions on R d that are together with their first and second partial derivatives bounded on D. Theorem 2.4 Under condition (H) for the domain and the assumption that 'i aij  L (D)

1 i j d

(2.4)

the conservative diffusion process M = (Xt  Px ) of Theorem 2.2 enjoys the following properties : for each x  D,

379

Reflecting diffusions on H¨older domains

1. 2.

527

Px (X0 = x ) = 1,  t d #  f (Xt )  'i aij 'j f (Xs ) ds

is a Px -submartingale,

0 i j =1

 whenever f  Cb2 D , f is constant in a neighbourhood of + and d 

'i f (x ) aij (x ) nj (x )  0

# -a.e. on 'D

(2.5)

i j =1

, 3. Ex



 I + (Xs ) ds = 0.

0

When d = 2, aij (x ) = 12 ij and D = C the standard outward cusp domain, R.D. DeBlassie and E.H. Toby [7] have shown the existence and the uniqueness of the corresponding submartingale problem for a probability measure Px on  = #  :  is a continuous function from [0 ) into C  : for each fixed x  C, 1 Px ((0) = x ) = 1   1 t 2 f ((t))  f (s) ds is a Px -submartingale 2 0 whenever f  Cb2 (C ) f is constant in a neighbourhood of the origin and,f (x )  n(x )  0 on 'C 

 I0 (s) ds = 0 3 Ex 0

Hence, by virtue of Theorem 2.4, the diffusion process of Theorem 2.2 coincides in law with DeBlassie-Tobyʼs one in [7] in this special case. In exactly the same way, we see that, when d = 2, aij (x ) = 12 ij and D is a wedge # : 0    R 2 for a fixed   (0 2), the diffusion process of Theorem 2.2 is identical in law with Varadhan-R.Williamsʼs normally reflecting Brownian motion [22]. 3 Lp -estimate, local estimate and Harnack inequality In this and the next sections, we shall work under another condition (A) on a domain D  R d which will be seen to be more general than (H) (Proposition 4.1). In this section, we derive some estimates for harmonic solutions of equations associated with   H 1 (D) under condition (A). In order to state condition (A), we employ the following notations: B (a ) = #x  R d : x  a   B () = B (0 ) B+ () = #(x  xd )  B () : xd 0 

C () = #(x  xd )  B () : x   xd  

Q () = #(x  xd )  B () :  x   xd 

380

528

M. Fukushima, M. Tomisaki

for a  R d   0   (0 1). For a Lipschitz mapping % from a set E  R k into R l such that %(x )  %(y) K x  y x  y  E , for some constant K 0, we also denote by Lip(%) the smallest constant K of this property. We now state condition (A) on a domain D  R d : (A) The following properties hold for an at most countable index set I , a constant    (0 1) and positive constants   r   M  : There are a point ak  'D and its neighbourhood Vk associated with each k  I such that

l; (i-1) D  Vk  Vl =  k  l  I  k = (i-2) there are a constant k  [   1) and a one to one mapping  %k from B ( ) onto Vk with %k (0) = ak , Vk  D equals either %k Ck ( ) or   %k Qk ( ) , Lip(%k ) M  , Lip(%1 k )M .  (ii) For any a  'D k I Vk , there are its neighbourhood Wa and a one to one mapping "a from B (r  ) onto Wa such that "a (0) = a "a (B+ (r  )) = Wa  D Lip("a ) M   Lip("a1 ) M  .

(i)

In our previous paper [14], we considered the same condition as above, but we assumed   (d  1) d . Further we did not consider the case Vk  D = %k Qk ( ) , namely, we assumed in [14] that every ak is an outward cusp boundary point but not an inward one. Under those assumptions, we got the same estimates as in this section following the PDE argument due to Stampacchia [18] and Moser [17]. The PDE argument is based on a Sobolev inequality of Moserʼs type formulated in Proposition 3.1 below. As will be proved in the last section, we need not the previous assumption   (d  1) d for the validity of Proposition 3.1. Once it is established, we can follow the PDE argument developed in [14] without any change so that we shall state the results of this section omitting the proof and only referring to the corresponding results in [14]. In the rest of this section, we assume (A). First of all, we note the following easily verifiable observation : if is a d d oneto one mapping from an open set U  R onto an open set in R with 1 M and if B (a  r)  U and (a ) = a, then Lip  B (a r M )  B (a  r) 

(3.1)

This observation particularly leads us to the following property of the domain D (see [14; Lemma 3.1]). We set   k  I : Vk  D = %k Ck ( ) 

  = k  I : Vk  D = %k Qk ( ) 

IC = IQ



ak for k  IC [resp. IQ ] may be called an outward [resp. inward] cusp boundary point. A collection of open sets is said to have a finite intersection property if there exists an integer M such that any subcollection of cardinality greater than M has an empty intersection.

381

Reflecting diffusions on H¨older domains

529

Lemma 3.1 For any   (0  ], there exist positive constantsr and m for which D satisfies the following : For every a  'D #  'D k I %k (B ()),  there are a neighbourhood V a of a and a one to one mapping a from B r = V a  D Lip( a ) onto V a such that a (0) = a a B+ r 1 ) m . m  Lip( a Furthermore, for any r  (0 r ] we have a subset A  'D # and a positive constant  = (r) such that # a (B+ (r))aA has a finite intersection property, and for every b  'D, the set B(b )  D is contained in one of the following  sets: %k Ck () for k  IC , %k Qk () for k  IQ , a (B+ (r)) for a  A. In the following we set  Ck () = %k Ck ()  Qk () = %k Qk ()  Ba (r) = a (B+ (r)) 

k  IC  k  IQ  a  'D # 

for 0     0  r r . Since a Sobolev inequality of Moserʼs type formulated in Lemma 2 in [17] is valid for u  H 1 (B+ (r)), we immediately obtain by means of the map a in Lemma 3.1 that for any   (0  ] !  (0 1] q  [2 2d (d  2)] ( q  [2 ) if d = 2 ) there is a positive constant C1 = C1 ( ! q) such that 

1 q uq dx

C1 r

d

1

1 q 2

% u2 dx + r 2

Ba (r)

N

1 2

d   i =1

'i u2 dx



Ba (r)

(3.2)  for u  H 1 Ba (r)  N  Ba (r) with N   !Ba (r) 0  r r , and a  'D # . Here E  denotes the Lebesgue measure for measurable sets E . (C1 and the other constants C2  C3 etc. below also depend on d    r   M  and them.) in some cases on   and . However we omit indicating  Actually (3.2) also holds for u  H 1 Ck () and u  H 1 Qk () : Proposition 3.1 (i) For any !  (0 1] there is a positive constant C2 = C2 (!) such that 1 q

 q

u dx Ck ( )

C2 

d 1+k k

1 q

12



% 2

2

u dx + 

 N

d   i =1

Ck ( )

1 2 2

'i u dx

 (3.3)

 for u  H 1 Ck ()  N  Ck () with N   !Ck (), 0     2 q 2(d  1 + k ) (d  1  k ), and k  IC . (ii) Let 2 q   in case d = 2 or 2 q 2d (d  2) in case d  3. Then for any !  (0 1] there is a positive constant C3 = C3 (! q) such that

382

530

M. Fukushima, M. Tomisaki

1 q

 uq dx

C3 

Qk ( )

d

1

1 q 2

% u2 dx + 2

d  

N

i =1

Qk ( )

1 2 'i u2 dx



(3.4)  for u  H 1 Qk ()  N  Qk () with N   !Qk (), 0    and k  IQ . The proof of Proposition 3.1 will be carried out in the last section by employing a specific transformation of a standard cusp domain onto a rectangular set. The following Sobolev inequality in an ordinary sense follows from (3.2), (3.3), (3.4) and Lemma 3.1 ([14; Proposition 3.2 (ii)]). (i) There is a positive constant C4 such that 1 2 % ( 1 q d   uq dx C4 u2 dx + 'i u2 dx 

Proposition 3.2

D

D

i =1

(3.5)

D

for u  H 1 (D) 2 q 2(d  1 +   ) (d  1    ). (ii) Assume the absence of outward cusp boundary point : IC = . Then the above statement is valid for 2 q 2d (d 2) in case d  3 and for 2 q   in case d = 2. We denote the norm of the Sobolev space H 1 (E ) by  H 1 (E ) . For an open set E  D, let us consider the following spaces :

 & (E ) = u  C 1 (E ) : u H 1 (E )   u = 0 on 'E  D  C (3.6) & (E ) = the completion of C & (E ) with respect to the norm  H 1 (E )  (3.7) H & (E ) coincides with H 1 (E ) if E  D. When E = C  () Q  () or Note that H 0 k k Ba (r), we can derive the following Sobolev inequalities from Proposition 3.1 ([14; Proposition 3.3]). Proposition 3.3 (i) For any   (0 1), there is a positive constant C5 = C5 () such that  1 q 1 2  d   q 2 u dx C5 'i u dx  (3.8) Ck ( )

i =1

Ck ( )

& (C  ()) 0     2 q 2(d  1 + k ) (d  1  k ) k  IC . for u  H k (ii) For any   (0 1) and for any 2 q 2d (d  2) (2 q   if d = 2), there is a positive constant C6 = C6 ( q) such that   d  1 q 1 2  q 2 u dx C6 'i u dx  (3.9) Qk ( )

i =1

Qk ( )

& (Q  ()) 0     k  IQ . for u  H k (iii) Let 0     0    1 and 2 q 2d (d  2) (2 q   if d = 2). Then there is a positive constant C7 = C7 (  q) satisfying

383

Reflecting diffusions on H¨older domains

531

1 q

 q

u dx

C7

1 2

 d  

Ba (r)

2

'i u dx



(3.10)

Ba (r)

i =1

& (Ba (r)) 0  r r  a  'D # . for u  H  We now turn to the Dirichlet form   H1 (D) given by (1.1) and (1.2). We

& (E ) for an open set E  D. also consider the following form E  H E (u ) =

d  

'i u(x )'j (x )aij (x ) dx 

& (E ) u   H

(3.11)

E

i j =1

 & (E ) is a Dirichlet form on with aij  1 i  j d satisfying (1.1). Since E  H L2 (E ), we have the associated Markovian resolvent #GE   0 on L2 (E ). Let T be a functional defined by  f0 $ dx +

T  $ = D

d   i =1

fi 'i $ dx 

$  H 1 (D)

(3.12)

D

for fi  L2 (D) i = 0 1     d . Since T is a continuous linear functional on H 1 (D), there is for each 0 a unique element u  H 1 (D) such that $  H 1 (D)

 (u $) = T  $

(3.13)

Here  (  ) =  (  ) + (  )L2 (D) . We denote this function u by G T . If T is & (E ) respectively and defined by (3.12) with D and H 1 (D) replaced by E and H & (E ) if every fi belongs to L2 (E ), then we have for each 0 a unique u  H denoted by GE  T such that & (E ) $H

E  (u $) = T  $

(3.14)

where E  (  ) = E (  ) + (  )L2 (E ) . Obviously G T [resp. GE  T ] coincides with G f [resp. GE  f ] in the case where T  $ = (f  $) for f  L2 (D) [resp. f  L2 (E )]. If E = Ck () Qk () or Ba (r), then the norm  H 1 (E ) is equivalent to  H& (E )  E ( )1 2 in view & (E ) satisfying (3.14) with of Proposition 3.3. Therefore there exists GE 0 T  H = 0. Using Sobolev inequalities (3.5), (3.8), (3.9), (3.10) and following a standard argument as in [18; Theorem 4.1] (see also [10]), we can get the following Lp estimates. Theorem 3.1 (i) Let p (d  1)   + 1 and 0. Then it holds that, for some C8 = C8 (p ) 0, G T

L (D)

C8

d  

fi L2 (D) + fi Lp (D) 

i =0

384

(3.15)

532

M. Fukushima, M. Tomisaki

where T is given by (3.12) with fi  L2 (D)  Lp (D) i = 0 1     d . (ii) Let k  IC  p (d  1) k + 1 and 0    1. Then there is a positive constant C9 = C9 (p ) such that GE  T

L (E )

C9 



1 p

p1 d1

 d

k

fi Lp (E ) 

(3.16)

i =0

where E = Ck () 0      0, and T is a continuous linear functional & (E ). given by (3.12) with fi  Lp (E ) and $  H (iii) Let p d and 0    1. Then there is a positive constant C10 = C10 (p ) such that d  (pd ) p  fi Lp (E )  (3.17) GE  T L (E ) C10  i =0

Qk ()



k  IQ  0      0, and T is a continuous linear where E = & (E ). functional given by (3.12) with fi  Lp (E ) and $  H  (iv) Let 0     p d and 0    1. Then there is a positive constant C11 = C11 (p ) such that GE  T L (E ) C11 r (pd ) p

d 

fi Lp (E ) 

(3.18)

i =0

for  0 E = Ba (r) 0  r r  a  'D # , and for T defined by (3.12) with & (E ) fi  Lp (E ) instead of D H 1 (D) fi  Lp (D) respectively. E H We are next concerned with local estimates for subsolutions of the equations associated with E . A function u  H 1 (E ) is called a subsolution if & (E ) $  0 $  H

E (u $) 0

(3.19)

In the same way as in [18; Theorem 5.1] or in [17; Theorem 1], we obtain the following local estimates from Proposition 3.3 or (3.2) ([14; Theorem 3.2]). Theorem 3.2 (i) Let 0    for some   (0 1) and E = Ck () with k  IC . Then every nonnegative subsolution u  H 1 (E ) of (3.19) satisfies ess sup u C12 (  s)

d 1+k 2k

1 2

 2



u dx

Ck (s)

0  s  

(3.20)

Ck ( )

for some C12 = C12 () 0. (ii) Let 0    for some   (0 1) and E = Qk () with k  IQ . Then every nonnegative subsolution u  H 1 (E ) of (3.19) satisfies  ess sup u C13 (  s)d 2 Qk (s)

385

1 2 u 2 dx

Qk ( )



0  s  

(3.21)

Reflecting diffusions on H¨older domains

533

for some C13 = C13 () 0. (iii) Let 0     0    1 0  r r  a  'D # and E = Ba (r). Then every nonnegative subsolution u  H 1 (E ) of (3.19) satisfies 1 2

 ess sup u C14 (r  s)

d 2

2

u dx

Ba (s)



0  s  r

(3.22)

Ba (r)

for some C14 = C14 ( ) 0. If u  H 1 (E ) satisfies E  (u $) = 0

& (E ) $H

(3.23)

for some  0, then u  0 and (u)  0 are both nonnegative subsolutions of (3.19). Therefore as an immediate consequence of Theorem 3.2 we get the following result.   Corollary Let 0    3.1 (i) # for some   (0 1) and E 1= Ck () with  k  IC resp. Qk () with k  IQ . Then every solution u  H (E ) of (3.23) satisfies (3.20) [ resp. (3.21) ] with u being replaced by u. (ii) Let 0     0    1 0  r r  a  'D # and E = Ba (r). Then every solution u  H 1 (E ) of (3.23) satisfies (3.22) with u being replaced by u.

Finally, by means of Proposition 3.1 and Theorem 3.2, we can get the following Harnack inequality for solutions u  H 1 (E ) of the equation (3.23) with = 0 ([14; Theorem 3.3]). # Theorem 3.3 (i) Let k  IC # resp k  IQ  0     0  !  1 and E = Ck () resp. Qk () . If u  H 1 (E ) is a nonnegative solution of #x : u(x )  1#  Ck ( 2)  ! Ck ( 2) (3.23) with = 0 and satisfies  resp. #x : u(x )  1  Qk ( 2)  ! Qk ( 2) , then there is a positive constant C15 = C15 (!) such that , ess inf u  C15

Ck ( 4)

 resp.

ess inf u  C15

Qk ( 4)



(3.24)

(ii) Let 0     0  r r  a  'D #  0  !  1 and set E = Ba (r). If u  H 1 (E ) is a nonnegative solution of (3.23) with = 0 and satisfies #x : u(x )  1  Ba (r 2)  ! Ba (r), then there is a positive constant C16 = C16 ( !) such that ess inf u  C16  Ba (r 4)

386

(3.25)

534

M. Fukushima, M. Tomisaki

4 Strong Feller resolvent In thissection, we will show that the resolvent #G  associated with the Dirichlet form   H 1 (D) has the same properties as those of Theorem 2.1 under condition (A). At the end of this section, we will show that condition (H) reduces to (A) and hence Theorem 2.1 follows. Theorem 4.1 Under condition (A), G satisfies the same properties as in Theorem 2.1. Namely,   p  p 1+ (d (i) G L2 (D) L (D)  C D   1)  .   (ii) G C D is a dense subspace of C D . (iii) There is a function G (x  y) continuous on D  D off diagonal such that   G (x  y)f (y) dy x  D f  C D  (4.1) G f (x ) = D

Theorem 4.1 is obtained essentially by the same argument as in [14; Sect. 4] but we give the proof here for completeness. Theorem 4.1 (i) is an immediate consequence of the following theorem. Theorem 4.2 Assume condition (A). Let p (d  1)   + 1 T be a functional given by (3.12) with fi  L2 (D)  Lp (D) i = 0 1 2     d , and 0. Then G T is uniformly continuous in D and accordingly G T can be extended to a continuous function on D. Proof Put u = G T . Fix a k  IC and an s  (0  2] arbitrarily. Set E = Ck (s). & (E ) be the solution of the equation (3.14) with = 0 Let   GE 0 (T  u)  H and T = T  u. We see by means of Theorem 3.1 (i), (ii),   % d  d 1 1 p1 k fi Lp (E ) f0  u Lp (E ) +  L (E ) C9 (p 1 2)s p i =1

c1 s

1 p



p1 d1

 d



fi L2 (D) + fi Lp (D) 

(4.2)

i =0

for some positive c1 independent of s and k . Since   u   belongs to H 1 (E ) and satisfies (3.23) with = 0, following the same argument as in [17] we get by means of Theorem 3.3 (i) (

  1  Osc ; Ck (s 4) 1  C15 (1 2) Osc ; Ck (s) 2  c2 Osc ; Ck (s)  for c2  (0 1) independent of s and k . Here Osc(; F ) denotes the oscillation of a function  over a set F : Osc(; F ) = ess supF   ess infF . Hence    Osc u; Ck (s 4) Osc ; Ck (s 4) + Osc ; Ck (s 4)   2  L (E ) + c2 Osc ; Ck (s) 4  L (E ) + c2 Osc u; Ck (s) 

387

Reflecting diffusions on H¨older domains

535

Combining this with (4.2) and using [18; Lemma 7.3], we get  0  s  4 k  IC  Osc u; Ck (s) c3 s 1 

(4.3)

for some constants c3 0 and 1  (0 1). In the same way we also get  Osc u; Qk (s) c4 s 2  0  s  4 k  IQ 

(4.4)

for some constants c4 0 and 2  (0 1). Recall r and 'D # appearing in Lemma 3.1. Similarly, for any   (0  2], there then exist a c5 0 and a 3  (0 1) such that  Osc u; Ba (s) c5 s 3  0  s r 4 a  'D #  (4.5) The estimate for oscillations on open balls with closures contained in D, which is due to Stampacchia [18], asserts that Osc (u; B (a s)) c6 s 4 

0  s  4 a  D D 

(4.6)

Here  is a positive number fixed arbitrarily, D = #x  D : dist(x  'D)  , and constants c6 0 and 4  (0 1) depend on  but are independent of a  D D . For an  0 fixed arbitrarily, we see by virtue of (4.3) and (4.4) that there exists an s1 = s1 ()  (0  4] such that  k  IC  (4.7) Osc u; Ck (s1 )     k  IQ  (4.8) Osc u; Qk (s1 )   By means of (4.5), we further find an s2 = s2 ( s1 )  (0 rs1 4] such that  Osc u; Ba (s2 )   a  'Ds#1 

(4.9)

In view of Lemma 3.1, we can find an o 0 such that every pair x  y  Do with x  y  o is simultaneously contained in one of sets Ck (s1 ) with some k  IC  Qk (s1 ) with some k  IQ  Ba (s2 ) with some a  'Ds#1 

(4.10)

(4.6) with this o leads us to Osc (u; B (a s3 ))  

a  D Do 2 

(4.11)

for some s3 = s3 ( o )  (0 o 8]. We now set  = (o 2)  s3 . Let x  y  D with x  y  . If x or y belongs to Do 2 , then u(x )  u(y)   by (4.10), (4.7), (4.8), (4.9). Otherwise, u(x )  u(y)   by (4.11).   Employing Corollary 3.1 in place of Theorem 3.1 (i) in getting (4.2), we obtain the following in the same way as above :

388

536

M. Fukushima, M. Tomisaki

Theorem 4.3 Let W be an open set of R d and E = W  D. Every solution u  H 1 (E ) of (3.23) for some  0 is uniformly continuous in W1  D for every open set W1 satisfying W1  W . We next give Proof of Theorem 4.1 (ii) We first follow an argument in [20; Proposition 5.1] to show    (4.12) G C D  C D    Since C0 D is dense in C D , it suffices to show that    G C0 D  C D (4.13) in the case where  D is unbounded. Let   C0 D and  0. Choose an R1 0 such that Supp[]  B (R1 )  D

c1 G  L2 (DB (R1 ))  

(4.14)

c1 being a positive constant specified later. We next take an R2 R1 satisfying the following : Ck ( )  D B (R1 ) Qk ( )  D B (R1 ) Ba (r  )  D B (R1 )

for k  IC with ak  'D B (R2 ) for k  IQ with ak  'D B (R2 ) for a  'D # B (R2 )

We set JC = #k  IC : ak  'D B (R2 )  JQ = #k  IQ : ak  'D B (R2 ), and A = 'D # 2 B (R2 ). Then, on account of (3.1), there is a constant   (0 R2 R1 ) depending on  but not on R1  R2 such that (   (  

r     2 D2 B (R2 )   ! ! Ck Qk Ba 2 2 2 aA k JC

k JQ

 # #a  ! A ! D D2 B (R2 ) and, for We consider the set K = k k JC JQ each a  K  we define a constant s and a set Ea (s) as follows :   s =   if a = ak  k  JC  Ck (s)     Q  (s)  s=  if a = ak  k  JQ  k Ea (s) =   s = r 2  if a  A Ba (s)    B (a s) s =  if a  D D2 B (R2 ) *

Note that D B (R2 ) 



Ea (s 2) 

aK



Ea (s)  D B (R1 )

aK

and Ea (s) (G  $) = ( $) = 0 By virtue of Corollary 3.1, we then have

389

& (Ea (s))  $H

(4.15)

Reflecting diffusions on H¨older domains

537

G  L (Ea (s 2)) c1 G  L2 (Ea (s)) 

(4.16)

where we used a local estimate due to Stampacchia [18] or Moser [17] in the case that a  D D2 B (R2 ). It should be noted that c1 is a positive constant independent of a R1 and R2 . By (4.14), (4.15) and (4.16), we find that G  L (DB (R2 )) c1 G  L2 (DB (R1 ))  

(4.17)

which along with Theorem 4.2 proves (4.13).  We next adopt Kunitaʼs argument [15]. Let denote by C0 D the space of differentiable functions on R d the restrictions to D of all infinitely continuously 

with compact support. For each u  C0 D , we define a functional Lu by   d d    Lu $ =  aij 'i u 'j $ dx  $  H 1 (D) j =1

D

i =1

Then T = u  Lu satisfies the condition of Theorem 3.1 (i) and u = G T for each 0. By virtue of Theorem 3.1 (i), there is for any  0 a   C0 D such that u  G  L (D)      

Since C  0 D is dense in C D , we thus obtain the denseness of G C D   in C D . Proof of Theorem 4.1 (iii) Since G is Markovian, there exists a function G (x  y) satisfying (4.1) by virtue of Theorem 3.1 (i) and Theorem 4.2. Hence it is enough to show that G (x  ) belongs to H 1 (U ) and is continuous on U 

(4.18)

for any open set U with U  D #x , where x  D and 0.  Let us denote the dual space of H 1 (E ) by H 1 (E ) . There exists for each  1 0 and T  H (D) a unique element u  H 1 (D) such that  (u $) = T  $

$  H 1 (D)

We denote this function u by G T . (We already usedthis notation for T given by (3.12) which is actually a general expression of T  H 1 (D) (cf. [16; 1.1.14]).) For a while we fix an x  D arbitrarily. We define a set Ex (s) according as three different cases. (Case 1) x is a cusp point, that is, x = ak for some k  I . In this case we take an s  (0  ].  (Case 2) x is a boundary point but not a cusp point, that is, x  'D k I #ak .  Choose   (0  ] such that x  'D k I %k (B ()). Then for an r given in Lemma 3.1 we take an s  (0 r ]. (Case 3) x is an interior point of D. In this case we take an s  (0 dx 2], where dx = dist(x  'D).

390

538

M. Fukushima, M. Tomisaki

Let us put Ex (s) =

  Ck (s)     Q  (s) k

Bx (s)     B (x  s)

if k  IC  in Case 1 if k  IQ  in Case 1 in Case 2 in Case 3

Then there exists a unique element sx   H 1 (D) such that   1 $(y) dy $  H 1 (D)  sx   $ = Ex (s) Ex (s)

(4.19)

Here we note the following lemma which is obtained by the same method as in [14; Lemma 4.3]. Lemma 4.1 Let U be an open set such that U  D #x . Then G (x  )U  H 1 (U ) and sx  U converges to G (x  )U weakly in H 1 (U ) as s  0. Take an open set V of R d such that U  V and x  V and set E = V  D. On account of Lemma 4.1, G (x  )E  H 1 (E ) and sx  E  G (x  )E weakly & (E ) and extend it to D by putting $ = 0 on in H 1 (E ) as s  0. Take any $  C D E . Then  E  (G (x  )E  $) = lim E  sx  E  $ s0   1 = lim  sx   $ = lim $(y) dy = 0 s0 s0 Ex (s) E (s) x This implies that G (x  )E  H 1 (E ) is a solution of (3.23) and hence, in view   of Theorem 4.3, G (x  ) is continuous in U . We finally note the following proposition which along with Theorem 4.1 implies Theorem 2.1. Proposition 4.1

Condition (H) reduces to condition (A).

Proof For each j  J , a H¨older function Fj in (H) (i) is given by Fj (x ) = j x j + fj (x ) where  j  1 j = 0 or 1 A j A or A j 1 A according to j  J0 or j  J+ or j  J , and fj is

a Lipschitz continuous  function defined on the d  1-dimensional closed ball x  R d 1 : x  rj with fj (0) = 0 and Lip(fj ) M . Then 

 + ! Uj  D = (j )  d(j )  B (rj ) : Fj (j )  d(j ) 

 

for some Cartesian coordinate system (j ) = (j )  d(j ) = 1(j )  2(j )      d(j ) . Let us put I = J+ ! J . For k  I , ak is the point of 'D corresponding to the origin in (k ) -coordinate system.   #ak : k  I  is then the totality of cusp

391

Reflecting diffusions on H¨older domains

539

boundary points. For each k  I , the neighbourhood Uk of ak contains no cusp

l k l  I . boundary point other than ak , and hence ak  al    k = For each k  I , we define a mapping %k from Ek  #(x  xd ) : x   rk  xd  R into (k ) -space by



%k (x  xd ) = (k )  d(k ) 

(k ) = x  d(k ) =  k xd + fk (x )   1 + A + AM . Put  = We then have Lip %k 1 + A + M and Lip %1 k    2(1 + A + M ) Vk = %k (B ( )) and M1 = (1 + A)(1 + M ). Then we see from (A) (i) with  =  and M  = M1 . In particular, (3.1) that #Vk k I satisfies  Vk  D = %k Ck ( ) if k 0 = %k Qk ( ) if k  0. We next show (A) (ii). Let o be the positive solution of the equation  2 + 2 =  2 o for o =   1. We then take a constant R 1 satisfying % 1  ( o R1 o   1+M +A R R will play the role of r  in (A) (ii). and put r  = o R. This r   Let us fix a p  'D k I Vk arbitrarily. By means of (H) (ii), there is a (j ) (j )  j  J such that p  U  pd(j ) ). We shall j  . Denote the -coordinate of p by (p define a mapping "p and a neighbourhood Wp in two cases j  I and j  I separately.

p  p-d ) = In the case that j  I , (p (j )  pd(j ) ) belongs to %j (Ej ). Putting (

 %1 p (j )  pd(j ) , we have that (p  p-d )  Ej B ( ) and p-d =  p  j in j accordance to the sign of j . We then define a mapping "p from the set Gp  #(x  xd ) : x + p-   rj  xd  R into (j ) -space as follows: 

"p (x  xd ) = (j )  d(j ) 

(j ) = x + p- 

  d(j ) = j x + p-  j + xd + fj x + p- 

 Notice that, on the region #x  Gp : x   o S  for S 1, Lip "p 1 + M +   j 1 A S 1 . Since the distance of p = "p (0) from ' "p (Gp ) is greater than S o , we can conclude from (3.1) that B (r  )  Gp for the above chosen r  = o R. In the case that j  I , we define a mapping "p from the set Gp  #(x  xd ) :

x + p (j )   rj  xd  R into (j ) -space by 

"p (x  xd ) = (j )  d(j ) 



(j ) = x + p (j )  d(j ) = xd + fj x + p (j )  In both cases, B (r  )  Gp . Accordingly we put Wp = "p (B (r  )). It is easy to see that "p is one-to-one, "p (0) = p, "p (B+ (r  )) = Wp  D, and Lip("p )  1 . M2  Lip("p1 ) M2 where M2 = 1 + M + A R1 R o

392

540

M. Fukushima, M. Tomisaki

Thus (H) reduces to (A) with I       r  as above and M  = M1  M2 .  

5 Decomposition of the sample path and additive functionals This section is devoted to the proof of Theorem 2.3 and Theorem 2.4. To this end, we first prepare an extended version of a general theorem [13; Theorem 5.5.5] to characterize the second term in the decomposition (1.6). Let X be a locally compact separable metric space, m be a positive Radon measure on X with full support and (   ) be a strongly local regular Dirichlet form on L2 (X ; m). We assume that there exists a conservative diffusion process M = (Xt  Px ) on X associated with the form  whose transition function pt (x  ) is absolutely continuous with respect to m for any t 0 and x  X . Then the resolvent of M admits a symmetric density G (x  y) with respect to m which is -excessive  in two variables x  y. The potential of a measure  is denoted by G (x ) = X G (x  y) (dy). The integral of a function f against a measure  is denoted by  f  or f  . A positive Radon measure  on X is said to be of finite energy integral if there exists a constant C17 such that   (x )(dx ) C17 1 ( )     (5.1) X

for some special standard core  of  . The totality of such measures is denoted by S0 . It is known that   S0 if and only if  G  is finite and that, in this case, G  is a -excessive and quasi-continuous version of the potential U  considered in [13; Sect. 2.2]. We further introduce two classes of positive Radon measures on X by S00 = # : (X )   sup G (x )   x X

S01 = # :   S0  G (x )   x  X  Obviously S00  S01  S0 . In our later application, the family S01 turns out to be more useful than S00 . An increasing  sequence #E  of finely open sets is said to be an exhaustive sequence if =1 E = X  A positive Borel measure  on X is called smooth in the strict sense if there exists an exhaustive sequence #E  of finely open sets such that IE   S00   = 1 2     Let S1 be the totality of smooth measures in the strict sense. S1 is known to be in one to one correspondence with the (equivalence classes of) positive continuous additive functionals (PCAFʼs in abbreviation) in the strict sense of M under the Revuz correspondence ([13; Theorem 5.1.7]). Lemma 5.1   S1 if and only if there exists an exhaustive sequence #E  of finely open sets such that IE   S01   = 1 2    

393

Reflecting diffusions on H¨older domains

541

Proof It suffices to show that any   S01 admits an exhaustive sequence #E  of finely open sets such that IE   S00   = 1 2     We may choose E as follows: E = #x  O : G (x )  

 = 1 2    

where #O  is an exhaustive sequence of relatively compact open sets. Then, (IE )(X ) = (E ) is finite, and further G (IE )(x )  for m-a.e. x  X by the maximum principle ([13; Lemma 2.2.4]) and hence for every x  X by the absolute continuity of the transition function.   We denote by  u the energy measure of u  loc . The Dirichlet form  is expressible as 1  (u ) =  u (X ) u    2 by using the co-energy measure  u . The second assertion of the next proposition replaces S00 in [13; Theorem 5.5.5] by S01 . Proposition 5.1 (i) Suppose that a function u satisfies the following conditions: 1. u is finite valued, finely continuous and u  loc . 2. IG  u  S00 for any relatively compact open set G. Then we have the unique decomposition u(Xt )  u(X0 ) = Mt[u] + Nt[u]  t  0 Px  ae x  X 

(5.2)

where M [u] is a CAF in the strict sense such that, for any relatively compact open set G,

  [u] 2   [u] Ex Mt = 0 E M = Ex BtG  x  G (5.3) x tG G B being the PCAF in the strict sense with Revuz measure  u and (G being the first leaving time from G. Nt[u] is a CAF in the strict sense locally of zero energy. (ii) Assume further the following property of u:

= (1)  (2) with IG  (1)  IG  (2)  S01 for any relatively compact open set G and  (u ) =       

(5.4)

for some special standard core  of  . Then N [u] = A(1) + A(2)  Px  as x  X  where A(1) and A(2) are PCAFʼs in the strict sense with Revuz measures (2) respectively.

(5.5) (1)

and

Proof The first assertion is a consequence of [11; Theorem 2]. Since (1)  (2) in (ii) are in the class S1 by the preceding lemma, we can see the validity of identity (5.5) on account of [12; Theorem 3.3, Corollary 3.1].  

394

542

M. Fukushima, M. Tomisaki

We are in a position to prove Theorem 2.3. We fix a domain D  R d possessing the property (H) and a data aij satisfying (1.1). We now apply the general theory prepared above to the specific Dirichlet form (1.2) on L2 (D; m), the resolvent G of Theorem 2.1 and the conservative diffusion M of Theorem 2.2. Here m denotes the d dimensional Lebesgue measure. M will be called the reflecting diffusion on D (associated with aij ). Accordingly G will be called the resolvent of the reflecting diffusion on D. Notice that, under the condition (H) for the domain D, the surface measure # on 'D is well defined with a local expression  " 1 + Fj ( )2 d  E  Uj  'D (5.6) #(E ) = E

where E = # : (  Fj ( ))  E . Further, with the unit inward normal vector n() = (n1 ()     nd ()) making sense #-a.e. on 'D according as " 1 + Fj ( )2   Uj  'D n() = (Fj ( ) 1) we have the divergence theorem   ' dm =  ni d # D 'xi D

1 i d    C0 (D)

This formula extends to a wider class of functions  and in particular the condition (1.9) for aij guarantees the identity   'j (aij  ) dm =   aij nj d #   C0 (D) (5.7) D

D

Denote by i the coordinate functions: i (x ) = xi  1 i d  Then, i  1 (D), and the co-energy measures  i j with respect to the Dirichlet form Hloc (1.2) are given by (cf. [13; Example 5.2.1])  i j = 2aij m

1 i j d

(5.8)

Let us denote by B an arbitrary ball in R d . Since aij are bounded, we have IB D   i  S00  The PCAF in the strict sense with Revuz measure m is just a constant functional t. Therefore Proposition 5.1 (i) implies the decomposition (1.6) where M i  1 i d  are CAFʼs in the strict sense satisfying (1.7) with t being replaced by t  (B D . Owing to the boundedness of aij however, we can let B R d to get (1.7), proving Theorem 2.3 (i). Turning to the proof of Theorem 2.3 (ii), we have under the condition (1.9)   (i  ) =

(x ) (dx )

with

D

=

d  j =1

395

('j aij )m 

d  j =1

aij nj #

(5.9)

Reflecting diffusions on H¨older domains

543

holding for any   C0 (D) because  (i  ) =

d   j =1

= 

aij (x )'j (x ) m(dx )

D

d   j =1

'j aij   dm +

D

d   j =1

'j (aij  ) dm

D

which equals to the right hand side of (5.9) by virtue of (5.7). Suppose that the surface measure # satisfies IB D #  S01

for any ball B  R d 

(5.10)

We then have from the boundedness of aij and condition (1.9) IB D    S01  Hence (5.9) and Proposition 5.1 (ii) lead us to the expression (2.3) in terms of a PCAF L in the strict sense with Revuz measure being the surface measure #, proving Theorem 2.3 (ii). It only remains to show (5.10) for the proof of Theorem 2.3. Theorem 5.1 If (2.2) holds, namely, each outward cusp boundary point is of H¨older exponent greater than 12 , then the surface measure # on 'D satisfies condition (5.10). For any ball B , the compact set B  'D can be covered by finite number  of open sets U j  appearing in the condition (H) (ii) for the domain D. Besides G (x  y) is jointly continuous off diagonal by Theorem 2.1 (iii). For the proof of (5.10), it is therefore sufficient to show I   #  S0

and G I  #(x )   x  

(5.11)

where  = # = (  d ) :    d = Fj ( )  Uj  'D for each fixed j  J and   rj . Exp(Fj ) will be denoted by j . c1  c2     will denote some positive constants. We further let  = # : (  Fj ( ))   ( # :    ) Lemma 5.2 Let j  J+ ! J . (i) I  #  S0 whenever d  3. When d = 2, this is true if j 12 . (ii) I  #  S0 for any  0 where  = # :   .

396

544

M. Fukushima, M. Tomisaki

Proof (i) In view of (1.1) and (5.1), it suffices to prove the inequality   u(  Fj ( ))#(d ) c1 D(u u) + (u u)L2 (D)  u  C0 (D) (5.12) 

where D(u u) denotes the Dirichlet integral of u on D. On account of (5.6), the surface measure # has a density #( ) with respect to d satisfying #( ) c2  j 1  Hence the square of the left hand side of (5.12) is dominated by   u(  Fj ( ))2 d   2j 2 d  c22

(5.13)

(5.14)

  



 The second factor equals 0 r 2j +d 4 dr, which is finite under the stated condition. Consider a function  C0 (U ) taking value 1 on the set  . Then from the expression  r 2   2 '

# (  d )u(  d )d d     u(  Fj ( )) =  '  d Fj ( ) we see that the first factor of (5.14) is dominated by  (u 2 + u2 )d c3 U D

arriving at (5.12). (ii) Since #( ) is bounded on  = # : (  Fj ( ))   , (5.12) with  being replaced by  holds for any  0.   In order to complete the proof of (5.11), we prepare a lemma on a comparison of resolvent densities. Lemma 5.3 Let K be a compact subset of D and U be a bounded domain containing K such that the domain D1 = D  U possesses the property (H). Denote by G1 (x  y) x  y  D 1 , the resolvent density of the reflecting diffusion on D 1 . Then, x y  K  x =

y (5.15) G (x  y) G1 (x  y) + C18  for some positive constant C18 depending on the set K . Proof Consider the set F = D  U and the resolvent density G0 (x  y) x  y  F , of the part MF of M on the set F . MF is obtained from M by killing the sample paths upon leaving the set F . Then by Dynkinʼs formula  G (x  y) = G0 (x  y) + Ex e  G (X  y)  x  y  F  where ( denotes the leaving time from the set F . Take an open set W such that K  W  W  U . The second term of the right side of the above identity with y being restricted to D  W is dominated by

397

Reflecting diffusions on H¨older domains

C18 =

545

G (x  y)

sup x DU  yDW

which is finite owing to the off diagonal continuity Theorem 2.1 (iii). Let M1 be the reflecting diffusion on D 1 and M1F be its part on the set F ( D 1 ). On account of [13; Theorem 4.4.3], MF and M1F share a common Dirichlet form F on L2 (F ) given by F (u ) =  (u ) & (D1 ) [F ] = H

u   [F ]

& (D1 ) is defined by (3.7). Therefore we have the inequality where H G0 (x  y) G1 (x  y) holding for m  m-a.e. (x  y)  F  F . In view of the continuity of G and G1 ,   we get (5.15) for every x  F and every y  D  W . We return to the set   Uj  'D specified before Lemma 5.2. Lemma 5.4 Following inequalities hold for x  y   x =

y and a positive constant C19 depending on the set Uj  D : (i) If j  J+ and d  2, then

G (x  y) C19 x  y

d 1j j



(5.16)

(ii) If j  J0 ! J and d  3, then G (x  y) C19 x  yd +2 

(5.17)

(iii) If j  J0 ! J and d = 2, then G (x  y) C19 x  y

for any  0

(5.18)

Proof (i) Notice that the Sobolev inequality in the statement of Proposition 3.2 (i) holds with D and   being replaced by Dj = Uj  D and j respectively. Since the domain Dj is bounded, we can invoke Carlen-Kusuoka-Stroock [5] to conclude in the same way as in [3; Sect. 2] that the resolvent density G1 (x  y) of the reflecting diffusion on Dj admits the estimate G1 (x  y) c1 x  y 

x  y  Dj 

(5.19)

for & = 4 (q  2). In particular, by taking q = 2(d  1 + j ) (d  1  j ) we see that (5.19) is valid for & = (d  1  j ) j . We can then use Lemma 5.3 to get (5.16). (ii), (iii) In these cases, Proposition 3.2 (ii) is applicable to the domain Dj = Uj  D and we see the validity of (5.19) for & = d  2 [resp. & =  0 ] by taking q = 2d (d  2) [resp. q = 4  + 2 ]. We again use Lemma 5.3 to get (5.17) [resp. (5.18)].  

398

546

M. Fukushima, M. Tomisaki

Lemma 5.5 Let j  J+ ! J . Assume that j #()     .

1 2

in case j  J+ . Then G I 

Proof Keeping the expression  G ( (  F ( ))#( )d 

G I  #() = 

and the bound (5.13) of # in mind, we first prove the finiteness of the potential for = 0 in case that j  J+ and j 12 . From (5.16), we have the bound  G I  #(0) c1





2

2

  + F ( )

j d 1 2 j

.1

1j

d  

 

(5.20)



Since

 d 1j  2 j   + F ( )2 2j  2

d 1j j

 d 1j 

   

for some  0, where j = H¨ol(Fj ), we obtain  G I #(0) c2



r d +j 3(d 1j ) dr + c3 

d 1j j





r d +j 3 dr  



0

In the case that j  J , we get the finiteness of G I  #(0) from (5.17) and (5.18) in a similar manner to the above. Next take a   =

0 We can choose a neighbourhood V of such that 0

V  V  Uj and D1 = V  D is a Lipschitz domain. Let ˜ =   V  Then the same reasoning as the proof of Lemma 5.3 works to see that G ( )   ˜  is dominated by c4   d +2 in case that d  3 and by c5      0, in case that d = 2. Since #() is bounded on ˜ , we see the finiteness of G I˜  #()   and hence of G I  #(). Proof of Theorem 5.1

We divide the situation into four cases :

(I) j  J+  j 1 2 (III) j  J  d = 2

(II) j  J  d  3 (IV) j  J0

In view of Lemma 5.2 and Lemma 5.5, we see that (5.11) holds in cases (I) and (II). Hence it remains to prove (5.11) in cases (III) and (IV). We can instead prove a stronger property sup G I  #(x )  

(5.21)

x 

in these cases. Indeed, when j  J and d = 2, we have the bound (5.18) of G (x  y) for any  0, and we can proceed in a similar manner to the proof of Theorem 6.1 in our preceding paper [14] in getting (5.21) by choosing  smaller than j . When

399

Reflecting diffusions on H¨older domains

547

j  J0 , then we have the bound (5.17) or (5.18) of G (x  y) which, together with the uniform boundedness on  of the density function # of the surface measure, readily leads us to (5.21).   Proof of Theorem 2.4 Take any function f as is stated in the theorem and denote by W a neighbourhood of + on which f is constant. Then IB (DW )  #  S01

for any ball B  R d 

(5.22)

To see this, it suffices to show I W  #  S0

and G I W  #(x )   x  

(5.23)

for the set   Uj 'D appearing in (5.11) and exclusively for j  J+ . Since  W   for some  0, the first assertion in (5.23) follows from Lemma 5.2 (ii). The second one for x = 0 [resp. for x =

0] is immediate from the continuity of G (0 y) [resp. from Lemma 5.5]. Now just as computations made in (5.8) and (5.9), we have  ' d  aij  'i f  'j f  m (5.24)  f f = 2  i j =1

and

  (f  ) =

   C0 D 

(x ) (dx ) D

with =

d 

d   'i aij 'j f  m  'i f  aij nj IDW  #

i j =1

(5.25)

i j =1

IDW can be inserted in the last expression because 'i f vanishes on W . In view of (2.4), (5.22), (5.24) and (5.25), Proposition 5.1 applies and we get f (Xt )  f (X0 ) = Mt[f ] + Nt[f ] 

Px -a.s. x  D

where M [f ] is a MAF in the strict sense with '   t  d  aij 'i f  'j f (Xs ) ds M [f ] t = 2 0

and  Nt[f ] = 0

 t



d 

i j =1



'i aij 'j f

'

(5.26)

(5.27)

i j =1





t

(Xs ) ds +



0

d 

' 'i f  aij nj

Ls  (5.28) (Xs ) d -

i j =1

Here Lt is a PCAF in the strict sense with Revuz measure IDW  #. (5.27) and (5.28) are valid Px -a.e. for every x  D. Under the condition (2.5) for f , the second functional in the right hand side of (5.28) is a PCAF in the strict sense. Therefore the desired conclusion follows from (5.26) and (5.28).  

400

548

M. Fukushima, M. Tomisaki

6 Sobolev inequality of Moser type In this section we will show Proposition 3.1. Throughout this section we assume condition (A). Proposition 3.1 is immediate from the following two propositions. Proposition 6.1 For any !  (0 1] there is a positive constant C20 = C20 (!   ,   M   d ) such that  %  d   2 2 2 2 u dx C20 u dx +  'i u dx  (6.1) Ek ( )

N



i =1

#

Ek ( )

for Ek () = Ck () resp. Qk () , u  H Ek ()  N being a Borel subset of # Ek () with N   !Ek (), 0    and k  IC resp. k  IQ .  1

Proposition 6.2 (i)  Let k  IC and 1 p (d  1 + k ) k . Take a q satisfying p q p d  1+ k (d  1 + k  k p) if p  (d  1 + k ) k , or p q   if p = (d  1 + k ) k . Then there is a positive constant C21 = C21 (p q      M   d ) such that  1 q  uq dx Ck ( )

C21 

d 1+k k

1 1 q p

% p

u dx + 



p

Ck ( )

1 p

d   i =1

p

Ck ( )

'i u dx



 for u  H 1 Ck ()  0    . (ii) Let 1 p d and take a q satisfying p q pd (d  p) if p  d , or p q   if p = d . Then there is a positive constant C22 = C22 (p q      M   d ) such that 1 q   uq dx Qk ( )

C22 

d

1 1 q p

% p

u dx + 

 Qk ( )

p

d   i =1

Qk ( )

1 p p

'i u dx



 for u  H 1 Qk ()  0     k  IQ . The part (i) of Proposition 6.2 is obtained by the same method as in [1; pp.128– 135] if we employ a transformation " (x ) defined below in place of the transformation rk (x ) in [1; p.130]. The part (ii) is also obtained following argument in [1; pp.103–104]. So we omit the proof of Proposition 6.2. In order to show Proposition 6.1 with Ek () = Ck (), we make use of a mapping " (x ) from a cusp C () onto a direct product set (). We begin with the definition of " . Let R+d = #(x  xd )  R d : xd 0 and d be the following product space:

401

Reflecting diffusions on H¨older domains

549



#(r t) : 0  r     t   if d = 2 if d  3 #(r t ) : 0  r   0 t     d 2   where d 2 = # 1  2      d 2 : 0 j  (j = 1     d  3) 0 d 2  2. Given   (0 1), we define the mapping " : R+d  d as follows : When d = 2, we set for x = (x1  x2 )  R+2 d =

" (x ) = (r t)

1 

r = x  ( 0)

t = x1 x2

( R);

When d  3, we set for x = (x  xd )  R+d " (x ) = (r t ) 1 

( 0) r = x  ( 0) t = x xd  (0     0) ( d 2 )  if x  = 0 =

if x  0  ( d 2 ) satisfying 1 ()     d 1 () = x x   Here 1 ()     d 1 () is the spherical polar coordinate of S d 2 , that is, for  = 1      d 2  d 2 , 1 () = cos 1  2 () = sin 1 cos 2  .. . d 2 () = sin 1 sin 2    sin d 3 cos d 2  d 1 () = sin 1 sin 2    sin d 3 sin d 2  For each r 0, we identify the points (r 0 )   d 2 by regarding them as a same point. Under this identification, " is one to one from R+d onto d and the inverse mapping "1 is as follows : For (r t)  2 , let  = (r t) be the (unique) positive solution of the equation t 2  2  +  2 = r 2 

(6.2)

Then "1 (r t) = (x1  x2 )

x1 = t 1  ( R)

x2 =  ( 0)

(6.3)

For (r t )  d with d  3, let  = (r t) be the positive solution of the equation (6.2). Then "1 (r t ) = (x1  x2      xd ) xi = t 1  i () ( R) i = 1 2     d  1

xd =  ( 0)

(6.4)

We next note that " : C ()  ()

one to one, onto

where () is a subset of d given by  #(r t) : 0  r   1  t  1 () = #(r t ) : 0  r   0 t  1   d 2 

402

if d = 2 if d  3

(6.5)

550

M. Fukushima, M. Tomisaki

For the sake of convenience, we write (r t )  () in case d = 2 too, and use (6.4) with the convention that 1 ()  1. Since  = (r t) is the solution of the equation (6.2), (x1      xd ) = "1 (r t ) satisfies the following relations : r 'xd =  'r  + (1 )t 2  2 1

(6.6)

t 2  'xd =  't  + (1 )t 2  2 1 'xd = 0 (j = 1 2     d  2) 'j and for i = 1 2     d  1,

(6.7) (6.8)

1 'xi 'xd = t 1 1 i () 'r  'r (

1 'xi 'xd =  1  + t 1 1 i () 't  't 'i () 'xi = t 1  (j = 1 2     d  2) 'j 'j

(6.9) (6.10) (6.11)

In the following, A1  A2     denote positive constants depending only on     and d . Let     1 0    and (x1  x2      xd ) = "1 (r t ) (r t )  (). Then we get the following estimates by means of (6.2), (6.3), (6.4), (6.6), (6.7) : (6.12) A1 r xd r 'xd A3  'r    'xd     't  A4 r

A2

(6.13) (6.14)

Further we see that the Jacobian determinant is given by J (r t ) =

'(x1      xd ) (d 1)  'xd d 2 = (1)d +1 xd t Sd () '(r t 1      d 2 ) 'r

where

 Sd () =

1 sind 3 1 sind 4 2    sin d 3

if d = 2 if d  3

(6.15)

(6.16)

 for  = 1      d 2  d 2 . By means of (6.12) – (6.16), we readily get J (r t ) A3 r (d 1)  

(6.17)

 A5 (d 1) +1 C () =

J (r t ) dr dt d  A6 (d 1) +1  ( )

We next note the following fact:

403

(6.18)

Reflecting diffusions on H¨older domains

551

 (i ) (0) Lemma 6.1 Let    

 (i1) 0 (i )  (i ) and x (i )  C () i = 0 1 x = (1) (i ) (i ) (i ) = " x   = 1      d 2  i = 0 1. For 0 s x . Set r  t   1, put  r (s) = r (0) + s r (1)  r (0)   t (s) = t (0) + s t (1)  t (0)   (6.19) (s) = 1(s)      d(s)2   j(s) = j(0) + s j(1)  j(0)  j = 1 2     d  2  Then r (s)  t (s)  (s) belongs to () for 0 s 1. Moreover the following estimate holds :     J r (0)  t (0)  (0) J r (1)  t (1)  (1)     (6.20)  A7 (d 1)      J r (s)  t (s)  (s)

for 0 s 1 if d = 2 and for 0  s  1 if d = 3 and

3   (0) (1) d) sin j(0) + sin j(1) =

0 t +t j =1

 Proof In view of (6.5) it is obvious that r (s)  t (s)  (s)  () 0 s 1.We  d 3  (0) (1) now assume that t (0) + t (1) =

0 in case d  3. Let 0 j =1 sin j + sin j    s 1 in case d = 2, and 0  s  1 in case d  3. Then J r (s)  t (s)  (s)  0 by virtue of (6.12), (6.13), (6.15) and (6.16). Moreover     J r (0)  t (0)  (0) J r (1)  t (1)  (1)         J r (s)  t (s)  (s) ( (0) (1) (d 1)  2 ( (0) (1) d 2  (0)  (1) Sd  Sd  A3 t t r r  

(s) (s) A1 r A2 t Sd (s) Note that r (s)  r (0)  r (1) 

and hence r (0) r (1) r (s) 

and if d  3, then t (s)  t (0)  t (1)  and for j = 1 2     d  3,  (0)   sin j (s) sin j    sin (1) j consequently

and hence t (0) t (1) t (s) 1         if j(0)   2  j(1)   2          if j(0)   2 j(1)   2 

  Sd (0) Sd (1)  1 0 Sd (s)  

We thus get (6.20).

404

552

M. Fukushima, M. Tomisaki

 For an open set E  R d , denote by C 1 E the restrictions to E of all continuously differentiable functions on R d . c1  c2 , etc. appearing in what follows denote positive constants depending only on     and d . Lemma 6.2 For any !  (0 1], there is a positive constant C23 = C23 (!      d ) such that  %  d   2 2 2 2 u dx C23 u dx +  'i u dx  (6.21) C ( )

N

i =1

C ( )

 for u  C 1 C () , a Borel subset N  C () with N   !C ()     1 and 0    .  Proof Let     1 and 0    . For u  C 1 C () , we set x = "1 (r t )  C () u-(r t ) = u  "1 (r t ) = u(x )  d       'u   1   'u(r t ) =  'xi  " (r t )  i =1

By virtue of (6.8) – (6.11), %  d 1 'u 1 1 1  'u 'xd 'u (r t ) = tx  (x ) i () + (x ) 'r  d 'xi 'xd 'r i =1

d 1  'u 'u 1  (r t ) = xd (x ) i () 't 'xi i =1  % d 1 'u 'xd 1 1 1  'u t xd  (x ) i () + (x ) +  'xi 'xd 't i =1

d 1  'i 'u 'u 1  (r t ) = t xd (x ) () 'j 'x 'j i i =j

j = 1 2     d  2

Combining these with (6.12), (6.13), (6.14), we find that      '  t )   u (r t ) A8 'u(r  'r       '  t )   u (r t ) A9 r 'u(r  't       '  t )  j = 1 2     d  2  u (r t ) A10 rt 'u(r  'j 

(6.22) (6.23) (6.24)

(0) (1) Let 0  ! 1 N  C ()  with N   !C () x  C () and x  N . (i ) (i ) (i ) (i ) (i ) = " x  i = 0 1. Then Put # = r  t  

405

Reflecting diffusions on H¨older domains

553

    u x (0)  u x (1) = u- # (0)  u- # (1) = 



1

0



'  (s) u- # ds 's

(s)

where # (s) = r (s)  t (s)   , and r (s)  t (s)  (s) are those given by (6.19). Integrating over x (1)  N , we find that    N  u x (0) 





  (1)  J #  d # (1)

u dx +  (1) ( )

N

 0

1

   '  (s)   u- #  ds  's 

Here d # (i ) denotes the product measure dr (i ) dt (i ) d (i ) for each i = 1 2. From this  2 u2 dx N  C ( )

(

2 u dx

2C () N

    (0)  (0) J #  d # 

 +2  (0) ( )

 (1) ( ) 0 s 1

   '  (s)    (1)   u- #  J #  d # (1) ds   's

 2(I + II )

'2

(6.25)

Obviously,

 I C ()2

u2 dx 

(6.26)

N

On account of Lemma 6.1 and (6.17), 

 d#

II

(0)  (1) ( ) 0 s 1

 (0) ( )

 

 (1) ( ) 0 s 1

     J # (0) J # (1)      d # (1) ds    J # (s) 

2(d 1) 

A3 A7 

 (0) ( )  (1) ( ) 0 s 1

 

 (1) ( ) 0 s 1

   '  (s) 2   (s)   u- #  J #  d # (0) d # (1) ds   's

d # (1) ds 

= c1 2(d 1) +1

   '  (s) 2   (s)  (1)   u- #  J # J #  d # (1) ds   's

 (0) ( )  (1) ( ) 0 s 1

   '  (s) 2   (s)   u- #  J #  d # (0) d # (1) ds   's

On the other hand, we get from (6.22)–(6.24)

406

(6.27)

554

M. Fukushima, M. Tomisaki

       u  (s)   (0)  '  (s)    u  (s)   (0)  (1)   '   u- #   '#  t  t (1)  +   'r #  r  r  's 't  d 2     '   u # (s)  (0)  (1)  + j j   'j j =1

 d    'u  (s)  1   c2   'xi  " #  

(6.28)

i =1

Denoting

'u  "1 by i and substituting (6.28) into (6.27), we arrive at 'xi

II c3 

2(d 1) +3

d   i =1

 c3 2(d 1) +3

d 

 (0) ( ) (1)

 2    i # (s) J # (s)  d # (0) d # (1) ds

 ( ) 0 s 1

IIi ()

(6.29)

i =1

 By fixing # (0) = r (0)  t (0) (0)  ()  1, we make use and 0(s)  s (s) (1) (1) (1) (1)    # = r  t   = r  t (s)  (s) . Putting of the transformation #  (s) (s) (s) is given by # = (r t ) = r  t   , we find that the Jacobian determinant  '# (1) '# = s d . Moreover # = (r t s) exhausts a set  # (0)  s specified by (1  s)r (0)  r  s + (1  s)r (0)  ad + (1  s)t (0)  t  s + (1  s)t (0)  (1  s)j(0)  j  j s + (1  s)j(0) 

j = 1 2     d  2

where ad = 1 if d = 2, = 0 if d  3, and j =  (j = 1 2     d 3) d 2 = 2. So we get, for each i = 1 2     d ,    2    (0) d # ds i # (s) J # (s)  d # (1) IIi () = (0)  =

 ( ) 0 s 1

 (0) ( ) 0 s 1

 (1) ( )

 d # (0) ds

 ( (0) s )

i (#)2 J (#) s d d #

By exchanging the order of integration,

(  r  s r 2 d IIi () =  0 i (#) J (#)s ( ) 1s 1s 0 s 1

(

d) 2 ( j  j s j t s t 1  ad  j   0 d #ds 1s 1s 1s 1s j =1  2d +1  d 2  i (#)2 J (#) d # 

( )

407

Reflecting diffusions on H¨older domains

555

Combining this with (6.29) and (6.18), we have II c4 2(d 1) +4

d   i =1

c5 2 C ()2

i (#)2 J (#) d #

( )

d  

2

'i u dx 

(6.30)

C ( )

i =1

 

Since N   !C (), (6.25), (6.26) and (6.30) lead us to (6.21).

Lemma 6.3 Let 0    and E () be the following subset of R d with d  2. E () = #(x  xd )  B () : xd (x )  where  is a continuous function on #x  R d 1 : x     such that (0) = 0 and (x ) 0 x    . Then the statement of Lemma 6.2 with C () replaced by E () above holds.  Proof Let u  C 1 E () and N ( E ()) be a Borel subset satisfying N   !E (). For x = (x  xd )  E (), we set x = (x  xd ). We also use the polar coordinate with center x : y = x + r r = y  x   = (y  x ) r  S d 1 . The following inequality is obvious :     u(y) dy + N  u(x )  u x  N  u(x )  N        u(y)  u y  dy + u x  u y  dy + N

N

from which we obtain N 2 u(x )2  ( 2 2 c1 u dy + N  N



+

'd u(x  s) ds

c1 E ()

'd u(y  s) ds yd

d (  i =1

2

yd 

dy yE ( )yd 0



xd



+

2

xd 



r

dy

yE ( )y=x +r

   'i u x + s  ds

0



 2

2 .

2

u dy + 2E ()  N  2 'd u dy +2E ()2

xd 

2

'd u(x  s) ds

xd

E ( )

+4E ()d +1

d   i =1

'i u(z )2

z E ( )zd 0

408

d 1

x  z 

. dz 



556

M. Fukushima, M. Tomisaki

where we used the following estimate for the last term. ( 2  r  dy 'i u x + s  ds yE ( )y=x +r

0





4E ()

r

dy



yE ( )y=x +r

 'i u x + s 2 ds

0



r d 1 drd 

= 4E () yE ( )y=x +r



4E ()d +1 z E ( )zd 0

r

 'i u x + s 2 ds

0

'i u(z )2 dz  x  z d 1

Therefore we have  u(x )2 dx N 2 E ( )





c2 E ()2

 u2 dy + E ()2 2

N

E ( )

+E ()d +1 



c3 E ()2

'd u2 dx

d   i =1

 'i u(z )2 dz

z E ( )zd 0



u2 dy + E ()2 2 N

+E ()d +2

E ( )

dx x  z d 1

.

'd u2 dx E ( )

. 'i u2 dx 

d   i =1

E ( )

Noting that c4 d E () c5 d , we get the conclusion.

 

Proposition 6.1 now follows from Lemmas 6.3 and 6.4. Proof of Proposition 6.1 Let k  IC  0    , and N be a Borel subset  N   !Ck (). In view of [16; Theorem 1.1.7], u  %k  of C  k () satisfying H 1 Ck () provided u  H 1 Ck () . By means  of [1 : Theorem 3.18], u  %k is approximated by functions belonging to C 1 Ck () in H 1 -norm. Noting that  1 

  %1 k (N ) is a Borel subset of Ck () and satisfies %k (N )  ! Ck () for some

  !  (0 1], we obtain (6.1) with Ek () = Ck () from Lemma 6.3. Similarly (6.1)   with Ek () = Qk () follows from Lemma 6.4. Acknowledgements. The second named author is very grateful to Professor Y. Ogura for valuable discussions on the transformation in Sect. 6.

References 1. R. A. Adams: Sobolev space. Academic Press, New York San Francisco London, 1975 2. R. F. Bass, P. Hsu: The semimartingale structure of reflecting Brownian motion. Proc. A.M.S. 108 (1990) 1007–1010

409

Reflecting diffusions on H¨older domains

557

3. R. F. Bass, P. Hsu: Some potential theory for reflecting Brownian motion in H¨older and Lipschitz domains. Ann. Probab. 19 (1991) 486–506 4. M. Biroli, U. Mosco: A Saint-Venant principle for Dirichlet forms on discontinuous media. Preprint Series Univ. Bonn SFB 224, 1992 5. E. A. Carlen, S. Kusuoka, D. W. Stroock: Upper bounds for symmetric Markov transition functions. Ann. Inst. H. Poincar´e Probab. Statist. 23 (1987) 245–287 6. Z. Q. Chen, P. J. Fitzsimmons, R. J. Williams: Quasimartingales and strong Caccioppoli set. Potential Analysis. 2 (1993) 219–243 7. R. D. DeBlassie, E. H. Toby: Reflecting Brownian motion in a cusp. Trans. Am. Math. Soc. 339 (1993) 297–321 8. R. D. DeBlassie, E. H. Toby: On the semimartingale representation of reflecting Brownian motion in a cusp. Probab. Theory Relat. Fields 94 (1993) 505–524 9. M. Fukushima: A construction of reflecting barrier Brownian motions for bounded domains. Osaka J. Math. 4 (1967) 183–215 10. M. Fukushima: On an Lp -estimate of resolvents of Markov processes. Publ. RIMS, Kyoto Univ. 13 (1977) 277–284 11. M. Fukushima: On a decomposition of additive functionals in the strict sense for a symmetric Markov processes. Proc. International Conference on ʻDirichlet Forms and Stochastic Processesʼ, Beijing, 1993, eds. Z. Ma, M. R¨ockner, J, Yan, Walter de Gruyter, Berlin 1995 12. M. Fukushima: On a strict decomposition of additive functionals for symmetric diffusion processes. Proc. Japan Acad. 70 Ser. A (1994) 277–281 13. M. Fukushima, Y.Oshima, M. Takeda: Dirichlet forms and symmetric Markov processes. Walter de Gruyter, Berlin 1994 14. M. Fukushima, M. Tomisaki: Reflecting diffusions on Lipshitz domains with cusps-analytic construction and Skorohod representation. Potential Analysis 4 (1995) 377–408 15. H. Kunita: General boundary conditions for multi-dimensional diffusion process. J. Math. Kyoto Univ. 10 (1970) 273–335 16. V. G. Mazʼja: Sobolev spaces. Springer-Verlag, Berlin Heidelberg New York 1985 17. J. Moser: A new proof of de Giorgiʼs theorem concerning the regularity problem for elliptic differential equations. Comm. Pure Appl. Math. 13 (1960) 457–468 18. G. Stampacchia: Equations elliptiques du second ordre a` coefficients discontinus. S´eminar sur les equations aux de´riv´ees partielles, Coll`ege de France, 1963 19. H. Tanaka: Stochastic differential equations with reflecting boundary condition in convex regions. Hiroshima Math. J. 9 (1979) 163–177 20. M. Tomisaki: Superposition of diffusion processes. J. Math. Soc. Japan 32 (1980) 671–696 21. M. Tomisaki: A construction of diffusion processed with singular product measures. Z. Wahrscheinlichkeitstheorie verw. Gebiete 52 (1980) 51–70 22. S. R. S. Varadhan, R. J. Williams: Brownian motion in a wedge with oblique reflection. Comm. Pure Appl. Math. 38 (1985) 405–443

This article was processed by the author using the LTEX style file pljour1m from Springer-Verlag.

410

  

 







 



 





  

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

411



!'%"('"!

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

412

@A?KILKOA@ =O  )   ) 

)





SDANA  ) EO = I=NPEJC=HA $ KB JEPA AJANCU =J@ ) EO = ?KJPEJQKQO $ KB VANK AJANCU ) JAA@O JKP >A KB >KQJ@A@ R=NE=PEKJ 'P EO KB >KQJ@A@ R=NE=PEKJ EB =J@ KJHU EB PDANA ATEOPO = OAP BQJ?PEKJ  KJ ' =J@ =J JAOP $  OQ?D PD=P   EO = JEPA OECJA@ IA=OQNA BKN A=?D 3 =J@  ?D=NCAO JK AT?ALPEKJ=H OAP OQ?D = OAP BQJ?PEKJ  EO ?=HHA@ = 0*,,1& 0'%+#" *#02/# 4'1& + 11!&#" +#01 $  =J@ BQNPDAN   8 9   9  ; + ;  9    $     

'J PDEO ?=OA IKNAKRAN ATLNAOOEJC  =O = @EANAJ?A      KB PSK LKOEPERA OIKKPD IA=OQNAO =J@ @AJKPEJC >U ! PDA LKOEPERA ?KJPEJQKQO $ =OOK?E=PA@ SEPD  ! >U PDA 0ARQV ?KNNAOLKJ@AJ?A 1    EP DKH@O PD=P

)

    



2DA =>KRA IAJPEKJA@ B=?PO BKN = NACQH=N "ENE?DHAP BKNI =J@ =J =OOK?E=PA@ &QJP LNK ?AOO SEHH >A OUOPAI=PE?=HHU ATPAJ@A@ EJ $ PK = CAJAN=H MQ=OENACQH=N "ENE?DHAP OL=?A =J@ =J =OOK?E=PA@ OP=J@=N@ LNK?AOO &ANA SA SEHH I=GA QOA KB = NACQH=N NALNAOAJP= PEKJ =J@ =J =OOK?E=PA@ MQ=OEDKIAKIKNLDEOI KB PDA QJ@ANHUEJC OL=?AO =O SEHH >A BKNIQ H=PA@ EJ $ BKHHKSEJC 8$ =9 8$ >9 8$-2 9 8!+0 9 1Q?D = IAPDK@ S=O =HNA=@U =@KLPA@ EJ 8+0 9 EJ PDAEN OLA?E? ?KJPATP KB = HK?=H ?KIL=?PE?=PEKJ SDANA = MQ=OE DKIAKIKNLDEOI EO NA=HEVA@ >U =J AI>A@@EJC =J@ PDAU ?=HHA@ EP = 1/+0$#/ *#1&," QP SA SKQH@ HEGA PK QOA PDEO PANI EJ PDA LNAOAJP IKNA CAJAN=H ?KJPATP  "ENE?DHAP OL=?A EO ?=HHA@ OPNKJCHU NACQH=N EB PDA =OOK?E=PA@ NAOKHRAJP =@IEPO = RANOEKJ LKOOAOOEJC = 0=U LNKLANPU OAA $ BKN LNA?EOA @AJEPEKJ  'J $ SA ?NQ?E=HHU JAA@ = NAJA@ PN=JOBAN IAPDK@ EJRKHREJC = OPNKJCHU NACQH=N NALNAOAJP=PEKJ 2DQO SA >=OE?=HHU JAA@ PK ODKS BKN 8   PDA AMQER=HAJ?A >APSAAJ PDA EJAMQ=HEPU  =J@ PDA ATEOPAJ?A KB = OECJA@ OIKKPD IA=OQNA  O=PEOBUEJC AMQ=PEKJ   2DEO SEHH >A @KJA EJ $ =J@ $ 5DEHA PDA EJAMQ=HEPU  BKHHKSO NA=@EHU BNKI PDA AMQ=PEKJ   PDA LNKKB KB PDA ?KJRANOA EILHE?=PEKJ AOLA?E=HHU PDA @ANER=PEKJ KB PDA OIKKPDJAOO KB = OECJA@ IA=OQNA EO = RANU @AHE?=PA I=PPAN 'P D=O >AAJ GJKSJ DKSARAN PD=P = GEJ@ KB OPNKJC $AHHAN LNKLANPU KB PDA NAOKHRAJP KB PDA =OOK?E=PA@ +=NGKR LNK?AOO  UEAH@O PDA @AOENA@ OIKKPDJAOO 8!$5 9 8$-2  2DAKNAI 9 8$ =9  'J $ SA OD=HH SKNG SEPD = OPNKJCHU NACQH=N "ENE?DHAP OL=?A PK ODKS PD=P PDEO NAMQENAIAJP ?=J >A SA=GAJA@ PK = 0=U LNKLANPU KB PDA NAOKHRAJP KB PDA =OOK?E=PA@ LNK?AOO 5A OD=HH PDAJ AILHKU EJ $ = PN=JOBAN IAPDK@ EJRKHREJC = OPNKJCHU NACQH=N NALNAOAJP=PEKJ KB PDA "ENE?DHAP OL=?A LNAL=NA@ EJ $ PK ?KILHAPA PDA LNKKB KB PDA @AOENA@ AMQER=HAJ?A 2DA ?KJ@EPEKJ  EO =??KN@EJCHU IKNA A=OEHU RANE=>HA PD=J PDA ATEOPAJ?A KB = OECJA@ OIKKPD IA=OQNA O=PEOBUEJC AMQ=PEKJ   ?PQ=HHU EP EO AJKQCD PK NAMQENA EJAMQ=HEPU  DKH@EJC BKN 9 EJ = IKNA PN=?P=>HA @AJOA OQ>OL=?A KB    $KN EJOP=J?A ?KJOE@AN PDA OEILHA ?=OA PD=P PDA JAOP EO PNERE=H $  ' 3        2DAJ  EO NA@Q?A@ PK PDA ?KJ@EPEKJ PD=P BKN 8   PDANA ATEOPO = LKOEPERA ?KJOP=JP  OQ?D PD=P PDA EJAMQ=HEPU  8 9   !9!  

413



DKH@O BKN =JU 9 EJ PDA OL=?A      ' 2  O SEHH >A OAAJ EJ $ PDA =>KRA ?KJ@EPEKJ EO AMQER=HAJP PK PDA KJA K>P=EJA@ >U NALH=?EJC PDA OL=?A  SEPD EPO OQ>OL=?A  O=PEOBUEJC   EO =J  @AJOA HEJA=N OQ>OL=?A KB   =J@ BKN =JU    PDANA ATEOPO = NA=H BQJ?PEKJ  7 OQ?D PD=P  7

    7  "

 7  7 7  8 9

   7   6  7  6 6  7 6 7  " =J@     5A OD=HH BQNPDAN OAA EJ $ PD=P EJAMQ=HEPU  DKH@EJC BKN 9   EO JKP KJHU OQ?EAJP >QP =HOK

) JA?AOO=NU BKN PDA $ ( PK >A KB >KQJ@A@ R=NE=PEKJ SEPD =J =@@EPEKJ=H LNKLANPU PD=P  (   HEI # + ' )    ( 7  SDANA PDA EJPACN=H EJOE@A PDA >N=?AO @AJKPAO PDA PKP=H R=NE=PEKJ KB LNKLANPU  O=UO PD=P PDEO .!$ D=O = JEPA 0ARQV IA=OQNA

)

KJ PDA EJPANR=H 8 79 2DA

5DAJ ' EO =J EJJEPA @EIAJOEKJ=H RA?PKN OL=?A =J@  EO K>P=EJA@ >U ?HKOEJC = LNA"ENE?DHAP BKNI @AJA@ BKN OIKKPD ?UHEJ@NE?=H BQJ?PEKJO KJ ' SA I=U P=GA =O  PDA OAP KB =HH OIKKPD ?UHEJ@NE?=H BQJ?PEKJO PK ?DA?G EJAMQ=HEPU   5DAJ PDA "ENE?DHAP OL=?A EO NACQH=N = J=PQN=H ?DKE?A KB  EO = @AJOA OQ>OL=?A KB  ' O=PEBUEJC PDA ?KJ@EPEKJ   'J PDEO ?=OA EJAMQ=HEPU  BKN  EO ARE@AJPHU AMQER=HAJP PK PDA ATEOPAJ?A KB = QJEMQA JEPA OECJA@ IA=OQNA  O=PEOBUEJC PDA AMQ=PEKJ  BKN =HH 9   -QN CAJAN=H PDAKNAI EJ $ =OOQNAO PD=P PDEO  EO =QPKI=PE?=HHU OIKKPD J=IAHU EP ?D=NCAO JK OAP KB VANK  ?=L=?EPU 5DAJ PDA "ENE?DHAP OL=?A EO JKP KJHU NACQH=N >QP =HOK OPNKJCHU HK?=H SA OD=HH ATPAJ@ EJ $ PDA H=OP OP=PAIAJP PK = BQJ?PEKJ 8  "% O=PEOBUEJC AMQ=PEKJ  BKN = OECJA@ 0=@KJ IA=OQNA  =J@ BKN =HH 9 >AHKJCEJC PK = IKNA OLA?E? OQ>OL=?A  KB    '  5A SEHH OAA PD=P PDEO LNKLANPU KB

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

414

= OAIEI=NPEJC=HA 8 '( QJ@AN !+ BKN ARANU ;  ' EB =J@ KJHU EB PDANA ATEOP JAHU KLAJ OAPO $ SEPD  $  $  '

!+ HEI  

  ;  ' $

OQ?D PD=P 8 EO = @EANAJ?A KB PSK AT?AOOERA BQJ?PEKJO KJ A=?D OAP $  'J PDA LNAOAJP L=LAN SA NAOPNE?P KQNOAHRAO PK PDA ?=OA PD=P  EO OUIIAPNE? =J@ PDA OAIEI=NPEJC=HA LNKLANPU KB 8 '( EO NAMQENA@ PK DKH@ KJHU BKN MA OP=NPEJC LKEJP ;  ' O EO ?HA=N BNKI PDA =>KRA ATLH=J=PEKJO KQN JA?AOO=NU =J@ OQ?EAJP ?KJ@EPEKJO KB PDA PULA  =NA IKNA A=OEHU RANE=>HA EJ I=JU ?=OAO SDANA ' =NA KB DECDAN @EIAJOEKJO



#%&!''"! ! $(&" " "%#&

5A O=U PD=P = MQ=@NQLHAP ' 2   EO = '/'!&)#1 0-!# EB ' EO = &=QO@KN PKLKHKCE?=H OL=?A SEPD = ?KQJP=>HA >=OA 2 EO = JEPA LKOEPERA KNAH IA=OQNA KJ ' =J@  SEPD @KI=EJ  EO = +=NGKRE=J ?HKOA@ OUIIAPNE? BKNI KJ  ' 2  2DA EJJAN LNK@Q?P EJ  ' 2 EO @AJKPA@ >U    =J@ SA HAP            

  

5A JKPA PD=P PDA OL=?A      ' 2 EO =J =HCA>N= %ERAJ PSK "ENE?DHAP OL=?AO X 2 X X  ' X 

' 2   

SA ?=HH PDAI #.2'3)#+1 EB PDANA EO =J =HCA>N=E? EOKIKNLDEOI  BNKI  KJPK X LNAOANREJC PDNAA GEJ@O KB IAPNE?O BKN 8   X !8!  !8!  8 8   8 8    8 8   8 8  -JA KB PDA PSK AMQER=HAJP "ENE?DHAP OL=?AO EO ?=HHA@ = /#-/#0#+11',+ KB PDA KPDAN X =NA O=E@ PK >A .20'&,*#,*,/-&'! EB PDANA ATEOP JAOP $  2DA QJ@ANHUEJC OL=?AO ' '  X X X JAOP X$  =J@ = KJA PK KJA I=LLEJC 5 BNKI '   $  $ KJPK '  $  $ OQ?D PD=P EPO NAOPNE?EPEKJ PK A=?D $ EO DKIAKIKNLDE? PK X$  5A O=U PD=P PDA AMQER=HAJ?A =O =>KRA EO '+"2!#" 6  .20'&,*#,*,/-&'0* EB PDANA ATEOPO = I=LLEJC 5 =O =>KRA OQ?D PD=P 8 X ;  8 5  X ;

; X  ' 

X  X EO PDA EI=CA KB   >U 5 $QNPDANIKNA 5 2DAJ 2 X EO PDA EI=CA IA=OQNA KB 2 >U 5 =J@  EO MQ=OEJKPEKJ LNAOANREJC 8!+0  !KN9   *AP $  >A =J EJ?NA=OEJC OAMQAJ?A KB ?HKOA@ OQ>OAPO KB ' 'P EO =J JAOP EB =J@ KJHU EB X 5 $  $  EO =J JAOP 

 ' EO AT?ALPEKJ=H EB =J@ KJHU EB 5 '  

415

X

EO AT?ALPEKJ=H

X   BQJ?PEKJ -  MA @AJA@ KJ ' EO MQ=OE?KJPEJQKQO EB =J@ KJHU EB -  5  EO  MQ=OE?KJPEJQKQO *AP QO JKS NA?=HH PDNAA GEJ@O KB NACQH=NEPU KB "ENE?DHAP OL=?AO 5A ?=HH = "ENE?DHAP OL=?A ' 2   .20'/#%2)/ EB PDANA ATEOPO =J JAOP ?KJOEOPEJC KB ?KIL=?P OAPO A=?D AHAIAJP EJ = ?ANP=EJ  @AJOA OQ>OL=?A KB  =@IEPO EPO MQ=OE?KJPEJQKQO RANOEKJ =J@ PDANA ATEOPO =J JAOP $  OQ?D PD=P PDA LKEJPO KB  $  $ =NA OAL=N=PA@ >U = ?ANP=EJ ?KQP=>HA B=IEHEU KB  MQ=OE?KJPEJQKQO BQJ?PEKJO >AHKJCEJC PK  #RANU AHAIAJP KB  PDAJ =@IEPO = MQ=OE?KJPEJQKQO RANOEKJ 5DAJ ' EO HK?=HHU ?KIL=?P SA @AJKPA >U  ' NAOL  '

PDA OL=?A KB ?KJPEJQKQO BQJ?PEKJO KJ ' SEPD ?KIL=?P OQLLKNP NAOL R=JEODEJC =P EJJEPU  5A ?=HH = "ENE?DHAP OL=?A ' 2   /#%2)/ EB ' EO = HK?=HHU ?KIL=?P OAL=N=>HA IAPNE? OL=?A 2 EO = LKOEPERA 0=@KJ IA=OQNA KJ ' SEPD BQHH OQLLKNP =J@ PDA OL=?A    ' EO  @AJOA EJ  =J@ QJEBKNIHU @AJOA EJ  '   OQ>I=NGKRE=J NAOKHRAJP GANJAH " ;  EO O=E@ PK >A = 6 /#0,)3#+1 EB "  '

  '



=J@ PDANA EO = ?KQJP=>HA B=IEHU  KB JKJJAC=PERA BQJ?PEKJ EJ  ' OAL=N=PEJC LKEJPO KB ' OQ?D PD=P " 8  8

8  

  

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

.20'/#%2)/ '$ +" ,+)6 /#-/#0#+11',+0 '0 '+"2!#"

'$ 6

0,*# +"  .20'

5A LNAL=NA = HAII= =>KQP E@AJPE?=PEKJO KB MQ=OEJKPEKJO EJ PDA NACQH=N ?=OA 1QLLKOA ' 2   EO = NACQH=N "ENE?DHAP OL=?A 2DA =OOK?E=PA@ ?=L=?EPU )4 EO @AJA@ BKN =JU KLAJ OAP   ' >U )4   EJB 8 8  8   8   

EJB 

=J@ BKN =JU OAP   ' >U )4   EJB)4      KLAJ  

416

'J 8$-2 9 A .20'!,+1'+2,20 EB BKN =JU    PDANA ATEOPO =J KLAJ OAP  SEPD )4    OQ?D PD=P 8  EO ?KJPEJQKQO #**  2--,0# ' 2   '0 /#%2)/ 1&#+ E  $*')6 ,$ '+!/#0'+% !),0#" 0#10 '0 + +#01 ' '1 '0  %#+#/)'7#" +#01 '+ 1&# 0#+0# ,$ 8$-2 9 EE

 ' '0 #5!#-1',+) ' )4  

EEE  $2+!1',+ ,+ ' '0 .20'!,+1'+2,20 ' '1 '0 .20'!,+1'+2,20 '+ 1&# 0#+0# ,$ 8$-2 9 /,,$ *AP   '(  !+ >A = &QJP LNK?AOO KJ ' SDE?D EO 00,!'1#" 4'1& 1&# $,/*  EJ PDA OAJOA PD=P PDA PN=JOEPEKJ OAIECNKQL 4( - KB PDA LNK?AOO  EO = RANOEKJ KB PDA  OAIECNKQL $( =OOK?E=PA@ SEPD  BKN =JU JKJJAC=PERA KNAH BQJ?PEKJ -   ' 2  E "AJKPA >U  PDA DEPPEJC PEIA KB = OAP    EJB7    '(   EJB 

 'J REAS KB *AII=  KB 8$-2 9 SA GJKS PD=P =J EJ?NA=OEJC OAMQAJ?A $  KB ?HKOA@ OAPO EO = CAJAN=HEVA@ JAOP E !+ HEI      MA ;  ' $



SDANA  @AJKPAO PDA HEBA PEIA KB  KNAH    ' 2  PDA BQJ?PEKJ     $ "  ;  + ,(  '( +7  ;  '

'B  EO PNQA PDAJ BKN =JU >KQJ@A@





>AHKJCO PK PDA OL=?A  =J@ ?KJRANCAO =O 3 PK PDA NAOKHRAJP "  KB  2=A =J@ EJ  IAPNE? =O SAHH &ANA SA OAP PDA R=HQA KB  =P PDA ?AIAPANU  PK >A VANK 1EJ?A PDA B=IEHU "  EO @AJOA EJ  $  EO =J JAOP 'B ?KJRANOAHU $  EO =J JAOP PDAJ BKN  HEI$   PDA BQJ?PEKJ 



8 ;  +

 ,(  '( +7



 

IQOP R=JEOD 2=A >A?=QOA 8   EO  KNPDKCKJ=H PK  $    1EJ?A 8 EO MQ=OE?KJPEJQKQO EP R=JEODAO MA !DKKOEJC  EJ  PK >A OPNE?PHU LKOEPERA KJ ' SA =NNERA =P PDA LNKLANPU   

417

EE 'B

EO AT?ALPEKJ=H PDAJ   $  '  $



BKN OKIA CAJAN=HEVA@ JAOP $  >U RENPQA KB E  2DAJ )4     BKN =JU ?KIL=?P OAP  =J@ DAJ?A )4   !KJRANOAHU =JU OAP KB VANK ?=L=?EPU O=PEOAO PDA EJ?HQOEKJ  BKN = ?ANP=EJ JAOP $  EJ PDA OAJOA KB 8$-2 9 SDE?D EO =J JAOP >U E  &AJ?A EO AT?ALPEKJ=H 2K OAA PDA AMQER=HAJ?A EEE  HAP 8 >A MQ=OE?KJPEJQKQO SEPD =OOK?E=PA@ JAOP $  1EJ?A $  EO = CAJAN=HEVA@ JAOP >U E  8 EO MQ=OE?KJPEJQKQO EJ PDA OAJOA KB 8$-2 9 KJ A=?D NAH=PERAHU ?KIL=?P KLAJ OQ>OAP KB ' SDE?D EJ PQNJ NA=@EHU EILHEAO PD=P EP EO MQ=OE?KJPEJQKQO KJ ' EJ PDA OAJOA KB 8$-2 9 2DA ?KJRANOA EILHE?=PEKJ EO PNERE=H =O EJ PDA LNKKB KB EE  

$KN = NACQH=N "ENE?DHAP OL=?A PDA JKPEKJ A?KIAO = OUJKIUI BKN U A?KIAO = OPK?D=OPE? ?D=N=?PANEV=PEKJ KB =J JAOP .,,$ ,$ %#,.#*  %ERAJ = CAJAN=H "ENE?DHAP OL=?A ' 2    = OQ>=HCA>N=  KB  ' 2 EO O=E@ PK O=PEOBU ?KJ@EPEKJ * EB *  EO = ?KQJP=>HU CAJAN=PA@ ?HKOA@ OQ>=HCA>N= KB  ' 2  *    EO @AJOA >KPD EJ   =J@ EJ  !  !  *  ' 2   EO @AJOA EJ  !  !    EJ  =J@ >U   PDA ATPAJOEKJ KB PDA  NAOKHRAJP "AJKPA >U   ' 2 PDA ?HKOQNA KB    KLAN=PKN  =OOK?E=PA@ SEPD  BNKI    PK    ?HKOA@ OQ>=HCA>N=  KB    ' 2 EO O=E@ PK O=PEOBU ?KJ@EPEKJ 0 EB     BKN =JU    0  0  EO CAJAN=PA@ >U = ?KQJP=>HA OQ>OAP  KB    OQ?D PD=P A=?D 8   EO JKJJAC=PERA =J@ O=PEOAO  8  8 

  

X >A EPO ?D=N=?PAN OL=?A *AP  >A = ?HKOA@ OQ>=HCA>N= KB  ' 2 O=PEOBUEJC ?KJ@EPEKJ * =J@ ' U RENPQA KB 8$-2  2D9 PDANA ATEOPO PDAJ = NACQH=N "ENE?DHAP OL=?A SEPD QJ@ANHUEJC X SDE?D EO AMQER=HAJP PK PDA CERAJ "ENE?DHAP OL=?A 1Q?D = NACQH=N "ENE?DHAP OL=?A EO ?=HHA@ OL=?A ' = NACQH=N NALNAOAJP=PEKJ SEPD NAOLA?P PK  8$ = 2D9 SAJP BQNPDAN =OOANPEJC PD=P PDANA ATEOPO = OQ>=HCA>N=  O=PEOBUEJC JKP KJHU * >QP =HOK 0  =J@ PDA NACQH=N NALNAOAJP=PEKJ SEPD NAOLA?P PK PDEO  >A?KIAO OPNKJCHU NACQH=N 8$-2  2D9 EO = NABKNIQH=PEKJ KB 8$ = 2D9 FQOP >U NAIKREJC PDA ENNAHARAJP =OOQILPEKJ PD=P ' EO HK?=HHU ?KIL=?P =J@ 2 EO 0=@KJ 'J PDA O=IA S=U 8$ = 2D9 ?=J >A NABKNIQH=PA@  .,,$ ,$ %#,.#*  'J REAS KB 8$ > 2D9 ?B 8$-2  2D9  SA OAA PD=P SDAJARAN PSK NACQH=N "ENE?DHAP OL=?AO =NA AMQER=HAJP PDAJ PDA AMQER=HAJ?A EO EJ@Q?A@ >U = MQ=OEDKIAKIKNLDEOI &ANA PDA MQ=OEDKIAKIKNLDEOI S=O BKNIQH=PA@ >U PDA JAOP @AJA@ >U PDA =OOK?E=PA@ ?=L=?EPEAO >QP EP EO = MQ=OEDKIAKIKNLDEOI EJ PDA LNAOAJP OAJOA >A?=QOA KB *AII=  E  

418

-J PDA KPDAN D=J@ 8!+0  2D9 ODKSO PD=P = "ENE?DHAP OL=?A EO MQ=OENACQH=N EB =J@ KJHU EB EP EO AMQER=HAJP PK OKIA NACQH=N "ENE?DHAP OL=?A >U IA=JO KB = MQ=OEDKIAKIKNLDEOI 5A =NNERA =P 2DAKNAI  >U ?KI>EJEJC PDKOA PSK B=?PO  ,.,)).3  $ 14, .20'/#%2)/ '/'!&)#1 0-!#0 /# #.2'3)#+1 1&#+ 1&# #.2'3)#+!# '0 '+ "2!#" 6  .20'&,*#,*,/-&'0* ,.,)).3  +6 .20'/#%2)/ '/'!&)#1 0-!# '0 #.2'3)#+1 1,  01/,+%)6 /#%2)/ '/'!&)#1 0-!# 6 *#+0 ,$  .20'&,*#,*,/-&0'*



 ""'!&& " &! ' &#&

&(%& "%  &'%"!) %(% %

'J PDEO OA?PEKJ SA SKNG SEPD = TA@ OPNKJCHU NACQH=N "ENE?DHAP OL=?A ' 2    $KN PDA =OOK?E=PA@ 0=U NAOKHRAJP " ;   PDANA ATEOPO = OQ>OPK?D=OPE? GAN JAH  ;  OQ?D PD=P  HEI " - ;  - <  ; +< -   '  



 LKEJP ;  ' EO ?=HHA@ = /+!&'+% -,'+1 EB PDA IA=OQNA  ;  EO JKP ?KJ?AJPN=PA@ KJ ; 2DA OAP ' KB =HH >N=J?DEJC LKEJPO EO ?=HHA@ PDA /+!& 0#1 -JA ?=J PDAJ ?KJOPNQ?P = +=NGKR LNK?AOO   '(  !+ KJ ' ?=HHA@ = 6 -/,!#00 SEPD NAOKHRAJP " ;   SDE?D EO GJKSJ PK AJFKU PDA BKHHKSEJC OLA?E? LNKLANPEAO 80 9 8)5 9  5A @AJKPA >U ' PDA KJA LKEJP ?KIL=?PE?=PEKJ KB ' =J@ LQP 

 EJB7    '(

   !+ '  ;  

;  '  ' 

 2DA O=ILHA L=PD '(  '(

EO ?=@H=C '(

 ' EO NECDP ?KJPEJQKQO BKN =HH 7   =J@ D=RA PDA HABP HEIEP '(

 ' BKN =HH 7   '(

  BKN =JU 7  

  !+ '(  '  '  

;  ' 7  

  EO OPNKJC +=NGKR  MQ=OEHABP ?KJPEJQEPU EJ = NAOPNE?PA@ OAJOA 'B OPKLLEJC PEIAO $ EJ?NA=OA PK  PDAJ BKN =JU ;  '   !+ HEI '  '   HEI '  '  '   $

$

 JKJJAC=PERA QJERANO=HHU IA=OQN=>HA BQJ?PEKJ - KJ ' EO O=E@ PK >A 02-#/*#"'+ EB " - ;  - ;    ;  ' =J@ PK >A #5!#00'3# EB " - ; - ;   ;  ' $KN = OQLANIA@E=J BQJ?PEKJ -  -: @AJKPAO EPO AT?AOOERA NACQH=NEV=PEKJ  -: ;  HEI " - ;  



419

?PQ=HHU -: EO PDAJ AT?AOOERA =J@ -: ;  - ; ;  ' 5A OD=HH QOA PDA BKHHKSEJC @AO?NELPEKJ KB PDA >N=J?D OAP ?B 8)5 9  $KN PDA B=IEHU      '

KB OQLANIA@E=J BQJ?PEKJO =PP=?DA@ PK PDA 0=U NAOKHRAJP SA OAP   -  *  -    * LKOEPERA N=PEKJ=H  2DAJ ' 



;  '  . ;  .: ; 



 

$KN =

KNAH OAP   ' SA HAP

 EJB7    '(  

;  EJB7    '(  

SEPD PDA ?KJRAJPEKJ PD=P EJB  =J@ SA BQNPDAN HAP     4 ;  + , 4; ;  + ,   5DAJ  EO KLAJ 4  4; =J@ EP EO = AT?AOOERA

KNAH IA=OQN=>HA BQJ?PEKJ

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

&#+  '0 0*,,1& +*#)6  !&/%#0 +, 0#1 ,$ 7#/, !-!'16 ,/#,3#/ '1 &,)"0 1&1   8 9:   9  + 9    

4&#/# 9  '0 +6 .20'!,+1'+2,20 3#/0',+ ,$ 9 

420



#**  002*# 1&# !,+"'1',+ ,$ /,-,0'1',+ E 2--,0# 9$ 01'0#0 1&# #.21',+    9$  9$ '0 ,2+"#" 9$ ; '0 2+'$,/*)6 ,2+"#" +" 9$ !,+3#/%#0 1,  $2+!1',+ 9   -,'+14'0# +" '+  ' 2 &#+ 9 01'0#0  EE &# #.21',+  &,)"0 $,/ 1&# -/,"2!1 9  9  9 $,/ +6 9     ' +" $,/ +6 ,2+"#" ,/#) #5!#00'3# $2+!1',+ 9   /,,$ E :  9$ PDAJ SA=GHU ?KJRANCAJP PK :  9 ?B8$-2  2D9  EE $ET =J =N>EPN=NU JKJJAC=PERA 9     ' =J@    =J@ HAP   -   ' 2  >KQJ@A@

KNAH  DKH@O BKN 9  9  " - 

1EJ?A "  '

    '   '   'B -  -   * -  * -   BKN OKIA ?KJOP=JPO *  *  PDAJ ?HA=NHU  -  * -   'B -$   EJ?NA=OAO PK = >KQJ@A@ BQJ?PEKJ -   ' 2  PDAJ 9  " -$ EO >KQJ@A@ QJEBKNIHU >KQJ@A@ =J@ ?KJRANCAJP PK 9  " - LKEJPSEOA =J@ EJ   &AJ?A -   >U RENPQA KB E  U PDA IKJKPKJA HAII= SA OAA PD=P AMQ=PEKJ  DKH@O BKN 9  9  " - BKN =JU JKJJAC=PERA >KQJ@A@ KNAH -    ,ATP P=GA =JU >KQJ@A@ KNAH AT?AOOERA BQJ?PEKJ 9   1EJ?A " 9 EO  ?KJRANCAJP PK 9 =O   9  " 9 EO >KQJ@A@ QJEBKNIHU >KQJ@A@ =J@ ?KJRANCAJP PK 9  9 LKEJPSEOA =J@ EJ   &AJ?A  DKH@O BKN 9  9  9 >U E =C=EJ  #**  002*# 1&# !,+"'1',+ ,$ /,-,0'1',+ +" "#+,1# 6  1&# 1,1) 3/'1',+ ,$ 1&# +'1# 0'%+#" *#02/#  E  '   EE ! '(  '

7  

 



/,,$ E 5A QOA PDA @AO?NELPEKJ  KB PDA >N=J?D OAP 2=GA .    $KN =JU /     '  /.     ' =J@   8 :/.   /.+ 

-J PDA KPDAN D=J@ .: EO = >KQJ@A@ KNAH AT?AOOERA BQJ?PEKJ =J@ @AJAO PDA O=IA AHAIAJP KB  =O . >A?=QOA " . EO  ?KJRANCAJP PK .   2DANABKNA >U *AII=  E    8 :/.   /: . + 

=J@ ?KJOAMQAJPHU  / .  .: +   /     '  

SDE?D EILHEAO PD=P  ;  '  . ;  .: ;   



421

EE A?=QOA KB HAII=  E  PDANA ATEOPO =J @A?NAOEJC KLAJ OAPO $ EJ?HQ@EJC ' OQ?D PD=P HEI$ )4 $   "QA PK *AII=  EE =J@ *AII=  EE  SA PDAJ D=RA   8 :/4    /4  + /     '   

'J REAS KB )4 $   4   4   4   4    4  EO >KQJ@A@ QJEBKNIHU >KQJ@A@ =J@  ?KJRANCAJP PK VANK 2DANABKNA BNKI  =J@ *AII=  E  SA D=RA BKN 4 ;  HEI 4  ; $

PDA E@AJPEPU  / ; 4 ; + ;  

/     ' 



SDE?D EILHEAO PD=P    ;  '  4 ;    !   SDANA     HEI 

  $

1EJ?A PDA OAP EJ PDA >N=?AO KB PDA HABP D=J@ OE@A KB  EO ?KJP=EJA@ EJ PDA IA=OQN=>HA OAP  SA CAP PK   

.,,$ ,$ ,.-,/&0&,+  2=GA =JU ?KIL=?P OAP  SEPD )4    $KN PDA NOP =OOANPEKJ KB PDA LNKLKOEPEKJ EP OQ?AO PK ODKS PD=P    



!DKKOA = OAMQAJ?A $  KB NAH=PERAHU ?KIL=?P KLAJ OAPO OQ?D PD=P $   $  



$  

$ 

1EJ?A )4 EO = !DKMQAP ?=L=?EPU HEI )4 $  HEI )4  $  )4   

$

$

-J PDA KPDAN D=J@  EO MQ=OEHABP ?KJPEJQKQO QJ@AN ! >U RENPQA KB  =J@ *AII=  EE  BKN =JU OPKLLEJC PEIAO $ EJ?NA=OEJC PK    ! HEI '  '    ! 

 $



422

&AJ?A  !

 HEI   ;   

$

=J@ =??KN@EJCHU HEI 4  ;  4; ;

$

BKN   ), ;  '

5A JKS D=RA PDA AMQ=PEKJ  BKN 4  >U *AII=  E =J@ *AII=  EE  'J PDA O=IA S=U =O EJ PDA LNKKB KB PDA LNA?A@EJC HAII= SA OAA PD=P PDA HABP D=J@ OE@A KB PDEO AMQ=PEKJ PAJ@O PK VANK =J@ ?KJOAMQAJPHU  / ; 4; ; + ;   /     '  

SDE?D IA=JO PD=P  4;     QP 4; ;   BKN ;    ' =J@ DAJ?A    '   SDE?D ?KI>EJA@ SEPD *AII=  E LNKRAO PDA @AOENA@   $KN PDA OA?KJ@ =OOANPEKJ P=GA =JU 9   =J@ HAP !9!   5A ?=J PDAJ J@ = OAMQAJ?A KB BQJ?PEKJO 9$     '  ?KJRANCAJP PK 9 OQ?D PD=P OQL+  9$ ;   =J@ BQNPDAN 9$ ?KJRANCAO MA PK =J MQ=OE?KJPEJQKQO RANOEKJ 9  KB 9 O HAII=  E  PDA @AOENA@ E@AJPEPU  JKS BKHHKSO BNKI  BKN 9$   5A =NA JKS EJ = LKOEPEKJ PK LNKRA PDA I=EJ PDAKNAI KB PDEO OA?PEKJ %#,.#*  ,/ 8   1&# +#51 14, !,+"'1',+0 /# #.2'3)#+1 E &#/# #5'010 + +#01 $  $,/ 4&'!& 1&# '+#.2)'16  '0 3)'" $,/ 0,*# -,0'1'3# !,+01+10 $  EE &#/# #5'010  0'%+#" 0*,,1& *#02/#  4'1& 0,*# 11!&#" +#01 $  $,/ 4&'!& 1&# #.21',+  '0 3)'" .,,$ ,$ 0%# &*-)&!0&,+ EE E  1QLLKOA EE EO BQHHHA@ 2=GA =JU 9    =J@ HAP   !9!  2DAJ 9      5, =J@ BQNPDAN 9      5, KJ ' " $  2DANABKNA PDA =>OKHQPA R=HQA KB PDA NECDP D=J@ OE@A KB  EO @KIKJ=PA@ >U $   SEPD $ >AEJC PDA PKP=H R=NE=PEKJ KB  KJ PDA OAP $   2DA ?KJRANOA EILHE?=PEKJ E EE SEHH >A LNKRA@ EJ PDA BKHHKSEJC IKNA OLA?E? BKNI .,-,/&0&,+  2--,0# $,/ 8   1&# '+#.2)'16  '0 3)'" $,/ 0,*# +#01 $  &#+ 1&#/# #5'010  0*,,1& 0'%+#" *#02/#  4'1& + 11!&#" +#01 $  4'1& $  $  3        $,/ 4&'!& 1&# #.21',+  '0 3)'" 

423

'J PDA NAOP KB PDEO OA?PEKJ SA =OOQIA PD=P 8   O=PEOAO  BKN =J JAOP $  5A OD=HH ?KJOPNQ?P  =J@ $  KB .NKLKOEPEKJ  >U = OANEAO KB HAII=O $ENOP T =J =N>EPN=NU : >AHKJCEJC PK PDA OL=?A   BKN OKIA 3 2DAJ 9  :    BKN =JU 9     8 9  :  $ !9  :!  $ !:!  !9!  9    1EJ?A PDEO EJAMQ=HEPU DKH@O BKN =JU 9 EJ PDA OL=?A   ' SDE?D EO QJEBKNIHU @AJOA EJ  '  PDANA ATEOPO = QJEMQA JEPA OECJA@ IA=OQNA   * BKN SDE?D PDA AMQ=PEKJ  EO R=HE@ 'J RENPQA KB .NKLKOEPEKJ  * ?D=NCAO JK OAP KB VANK ?=L=?EPU =J@   8 9  :   9  ; +* ; 9     

#**  : +*  : +* /,,$ $KN =JU 9   

:  :   $    



 8 9: :   

 9  : +*  



9  : +*  

2DEO HAII= O=UO PD=P NKQCDHU OLA=GEJC :  +* EO EJ@ALAJ@AJP KB : =J@ = ?=J@E@=PA KB PDA IA=OQNA  SA S=JP PK ?KJOPNQ?P 2K I=GA = NECKNKQO ?KJOPNQ?PEKJ SA JAA@ PK ?KJOE@AN = &QJP LNK?AOO =OOK?E=PA@ SEPD PDA CERAJ OPNKJCHU NACQH=N "ENE?DHAP OL=?A EJ KN@AN PK QOA = CAJAN=H PDAKNU EJ 8$-2 9 1Q?D = LNK?AOO ?=J >A EIIA@E=PAHU ?KJOPNQ?PA@ BNKI PDA 0=U LNK?AOO   '(  !+ =HNA=@U >AEJC ?KJOE@ANA@  'J B=?P *AII=  EE NA=@EHU EILHEAO PDA BKHHKSEJC 8$ > 2D9  BKN =JU OAP   ' SEPD )4    PDANA ATEOPO = KNAH OAP   SEPD )4   OQ?D PD=P '  EO '+3/'+1 EJ PDA OAJOA PD=P !+ '(  ' 

 '(  ' 

7     ;  ' 



1EJ?A )4 '   >U *AII=  E  PDA >N=J?D OAP EO EJ?HQ@A@ EJ = OAP KB PDA =>KRA PULA -J =??KQJP KB PDA LNKLANPEAO    KB PDA 0=U LNK?AOO  SA ?=J CAP = &QJP LNK?AOO KJ ' OPEHH =OOK?E=PA@ SEPD PDA BKNI  NOP >U NAOPNE?PEJC PDA OP=PA OL=?A KB  PK '  =J@ PDAJ >U I=GEJC A=?D LKEJP KB PK >A = PN=L OAA 8$-2  2D9 BKN PDKOA LNK?A@QNAO  5A I=U ?=HH PDA NAOQHPEJC &QJP LNK?AOO = 2+1 *,"'!1',+ ,$ 1&# 6 -/,!#00  'J SD=P BKHHKSO EJ PDEO OA?PEKJ   '(  !+ @AJKPAO = &QJP LNK?AOO KJ ' =OOK?E=PA@ SEPD PDA BKNI  $KN = OPNE?PHU LKOEPERA >KQJ@A@ KNAH BQJ?PEKJ  KJ ' SEPD    ' 2  SA LQP $

$ ;  "  ;  ;  ' SDANA PDA NECDP D=J@ OE@A EO @AJA@ >U   &ANA SA OAP     U 8$-2  2D9 SA GJKS PD=P $ EO =J MQ=OE?KJPEJQKQO KNAH BQJ?PEKJ EJ    5A PDAJ EJPNK@Q?A PDA OAPO $  ;  '  $ ; 

  3        3 

424



 '   $  $ 



$ EO = MQ=OE?HKOA@ KNAH OAP EJ?NA=OEJC EJ 3 =J@ $  $  3        1EJ?A HEI$ $ ;   SA OAA PD=P )4







 ;  '  

KSEJC PK PDA OPK?D=OPE? ?D=N?PANEV=PEKJ  KB PDA JAOP $  5A @AJA  >U  +; 

  +; KJ $  3         $ ; 



 



$KN 2  3 SA D=RA BNKI *AII=     +;   +; KJ $ # ;  $ ;  SDE?D IA=JO PD=P PDA =>KRA @AJEPEKJ I=GAO OAJOA  3 =J@  ?D=NCAO JK OAP KB VANK ?=L=?EPU +KNAKRAN



EO PDAJ = JEPA OECJA@ IA=OQNA BKN A=?D

  8 9  :   

9  :  +; 9    :   $    



'J B=?P PDA =>KRA @AJEPEKJ KB  =J@ *AII=  EILHU PD=P * +;  :  +;

:   $    

=J@ SA =NA HA@ PK  BNKI   'J KN@AN PK ?KJOPNQ?P =J =LLNKLNE=PA JAOP BNKI $  SA LNAL=NA = HAII= #**  2--,0# $  3        /# .20'!),0#" "#!/#0'+% '+ 3 +" )4  $  $   &#+ HEI )4 $    

$

BKN =JU ?KIL=?P OAP 

/,,$ 'J REAS KB PDA @AJEPEKJ KB PDA MQ=OE?HKOA@ OAP 8$-2  LL9  SA ?=J J@ BKN =JU    =J KLAJ OAP SEPD )4

  OQ?D PD=P $  =NA ?HKOA@ BKN =HH 3 $KN =JU ?KIL=?P  $ 

  =NA @A?NA=OEJC ?KIL=?P OAPO 1EJ?A !=L EO = !DKMQAP ?=L=?EPU  HEI )4 $ 

   )4  $  $ 

   )4 $  $  

$

5A PDAJ CAP HEI$ )4 $     BNKI )4 $    )4 $ 

   )4

  

425

.,,$ ,$ .,-,/&0&,+  2=GA = OAMQAJ?A   1EJ?A PDA OAPO $ @AJA@ >U  =NA MQ=OE ?HKOA@ SA ?=J J@ @A?NAOEJC KLAJ OAPO  SEPD )4    =J@ $   =NA ?HKOA@ BKN =HH 3 =J@ ( *AP QO @AJA EJ?NA=OEJC ?HKOA@ OAPO $ >U $  $  $

3      



=J@ LNKRA PD=P HEI )4   $   BKN =JU ?KIL=?P 

$



1EJ?A   $    $    $  SA D=RA )4   $  )4   $ 

  3

5A HAP $  $ 

  3

$ =NA PDAJ MQ=OE?HKOA@ =J@ $  $ >A?=QOA

$

 $ 

 $ 

   $  $  $  $ 



-J =??KQJP KB PDA LNA?A@EJC *AII= =J@   SA ?KJ?HQ@A PD=P )4   $  )4   $  3 

 EO LNKRAJ =J@ $  EO =J JAOP >U *AII=  +KNAKRAN $  $  $  3        $KN PDA IA=OQNA  @AJA@ >U     EO PDANABKNA = JEPA OECJA@ IA=OQNA BKN A=?D 3 1EJ?A  ?D=NCAO JK OAP KB VANK ?=L=?EPU EP >A?KIAO = OIKKPD IA=OQNA BKN SDE?D PDA LNAOAJP JAOP $  EO =PP=?DA@ $EJ=HHU P=GA =JU : >AHKJCEJC PK PDA OL=?A   BKN OKIA 3 *AP 9  3$   2DAJ 9   KJ $  $ =J@ 9  :  : 2DQO PDA AMQ=PEKJ  HA=@O QO PK   8 :  : ;  +; :    $     

?KILHAPEJC PDA LNKKB KB .NKLKOEPEKJ 



2DAKNAI  EO LNKRA@ &ANA SA =@@ = PDAKNAI ?KNNAOLKJ@EJC PK = OLA?E=H ?=OA KB 2DAKNAI  SDANA PDA JAOPO =NA PNERE=H 

426

%#,.#*  ,/ 8   1&# +#51 14, !,+"'1',+0 /# #.2'3)#+1 E &# '+#.2)'16  &,)"0 $,/ 0,*# !,+01+1    EE &#/# #5'010  +'1# 0'%+#" *#02/#  !&/%'+% +, #5!#-1',+) 0#1 02!& 1&1 1&# #.21',+  &,)"0 $,/ +6 9    /,,$ 2DA LNKKB KB PDA EILHE?=PEKJ E EE EO PDA O=IA =O PDA ?KNNAOLKJ@EJC LNKKB KB 2DAKNAI  2DA ?KJRANOA ?=J >A REASA@ =O = OLA?E=H ?=OA KB .NKLKOEPEKJ  PDA ?=OA SDANA :    



%!&%& " !)' '"% & '"  $(&%(% %' &#

$ENOP SA PN=JOBAN 2DAKNAI  BNKI = OPNKJCHU NACQH=N "ENE?DHAP OL=?A PK = MQ=OENACQH=N "ENE?DHAP OL=?A %#,.#*  #1 ' 2   #  .20'/#%2)/ '/'!&)#1 0-!# ,/ 8   1&# +#51 14, !,+"'1',+0 /# #.2'3)#+1 E &#/# #5'010 + +#01 $  $,/ 4&'!& 1&# '+#.2)'16  '0 3)'" $,/ 0,*# -,0'1'3# !,+01+10 $ EE &#/# #5'010  0'%+#" 0*,,1& *#02/#  4'1& 0,*# 11!&#" +#01 $  $,/ 4&'!& 1&# #.21',+  '0 3)'" /,,$ U !KNKHH=NU  PDA MQ=OENACQH=N "ENE?DHAP OL=?A ' 2   EO AMQ=ER=HAJP PK = X 2 X X >U = MQ=OEDKIAKILNLDEOI 5 PDANA ATEOP ?ANP=EJ OPNKJCHU NACQH=N "ENE?DHAP OL=?A ' X   X X X JAOP $  JAOP X$  5 EO = KJA PK KJA I=LLEJC BNKI '   $  $ KJPK '  $  $  EPO X X NAOPNE?PEKJ PK A=?D $ EO DKIAKILNLDE? PK $  =J@ PDA I=L BNKI  PK  @AJA@ >U X 8 X ;  8 5  X ;

; X' O=PEOAO X !8!  !8!  8 8   8 8    8 8   8 8  5 EO MQ=OE JKPEKJO LNAOANREJC =O EO ATLH=EJA@ EJ $ 1QLLKOA PD=P PDA ?KJ@EPEKJ E EO BQHHHA@ *AP X 8 X   8   

X$  5 $  $ 3       

X$  EO =J JAOP X 2DAJ  =J@ X 8 9X  $ !X  X 9 ! 

X 9  X   3      

BKN PDA O=IA ?KJOP=JP $ =O EJ E 



427

U RENPQA KB 2DAKNAI  PDANA ATEOPO = OIKKPD OECJA@ IA=OQNA X KJ ' SEPD =J =PP=?DA@ JAOP X$  BKN SDE?D PDA AMQ=PEKJ  X 8 9X   X  X 9X X ; +X  X ;  X 9   $     

 

DKH@O *AP    X 5   '



$  5  X$  X$ 3      

2DAJ SA ?=J A=OEHU OAA PD=P $  EO =J JAOP  EO = OIKKPD OECJA@ IA=OQNA KJ ' SEPD PDA =PP=?DA@ JAOP $  U NASEPEJC PDA =>KRA AMQ=PEKJ =J@ JKPEJC PD=P 9  ;  9X 5; CERAO =J MQ=OE?KJPEJQKQO RANOEKJ KB 9   X 9  SA =NNERA =P PDA AMQ=PEKJ  DKH@EJC BKN 8  X$  CAPPEJC PDA ?KJ?HQOEKJ EE  2DA ?KJRANOA EILHE?=PEKJ EE E ?=J >A @ENA?PHU ODKSJ =O PDA ?KNNAOLKJ@EJC LNKKB KB 2DAKNAI  

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

9   

&#+ 1&#/# #5'010 2+'.2#)6  +'1# 0'%+#" 0*,,1& *#02/#  $,/ 4&'!& 1&# #.21',+  '0 3)'" /,,$ 'J PDA O=IA S=U =O EJ PDA LNKKB KB 2DAKNAI  PDA =>KRA EJAMQ=HEPU EO DKJAOPHU EJDANEPA@ PK = OPNKJCHU NACQH=N NALNAOAJP=PEKJ SDANA SA ?=J QOA .NKLKOEPEKJ  PK K>P=EJ PDA ?KJ?HQOEKJ =O =>KRA SDE?D ?=J >A PN=JOBAN >=?G PK PDA MQ=OENACQH=N "ENE?DHAP OL=?A =HOK EJ PDA O=IA S=U =O EJ PDA LNKKB KB 2DAKNAI  



%!&%& " &'"&' "!'!'& '"  $(&%(% %' &#

2DEO OA?PEKJ EO @ARKPA@ PK PN=JOBANO KB LNK>=>EHEOPE? JKPEKJO =J@ PDAKNAIO BNKI = NACQH=N "ENE?DHAP OL=?A PK = MQ=OENACQH=N "ENE?DHAP OL=?A 'J L=NPE?QH=N SA BKNIQH=PA PDA OPK?D=OPE? OECJE?=J?A KB PDA ?KJ@EPEKJ EE EJ 2DAKNAI  IKNA LNA?EOAHU PD=J SD=P S=O IAJPEKJA@ EJ $ 

428

2K PDEO AJ@ HAP QO NOP ?KJOE@AN = NACQH=N "ENE?DHAP OL=?A ' 2   =J@ =J =OOK?E=PA@ &QJP LNK?AOO   '(  !+ KJ ' 'J PDEO ?KJPATP SA OP=NP SEPD IAJPEKJEJC OKIA LNK>=>EHEOPE? ?D=N=?PANEV=PEKJO KB AT?ALPEKJ=H OAPO EJ = ?KJRAJEAJP S=U BKN H=PAN QOA  OAP  ' EO ?=HHA@ #5!#-1',+) EB EO ?KJP=EJA@ EJ = JA=NHU KNAH OAP : OQ?D PD=P EO O=E@ PK >A  -/,-#/)6 #5!#-1',+) EB EP EO = JA=NHU KNAH !#  

   OAP IA=OQN=>HA 2JACHECE>HA OAP =J@ EPO ?KILHAIAJP '  EO EJR=NE=JP 5A ?=HH  -/,-#/)6 #5!#-1',+) '+ 1&# 01+"/" 0#+0# EB EJ PDA =>KRA @AJEPEKJ KB PDA LNKLAN AT?ALPEKJ=HEPU PDA EJR=N=J?A KB '  EO SA=GAJA@ PK PDA EJR=NE=J?A QL PK PDA HEBA PEIA EJ PDA BKHHKSEJC OAJOA !+ '(  '(  ' 

7  8 

  ;  ' 



#*.(  2DA H=OP JKPEKJ KB PDA AT?ALPEKJ=HEPU I=GAO OAJOA JKP KJHU BKN PDA LNAOAJP &QJP LNK?AOO >QP =HOK BKN = OP=J@=N@ LNK?AOO =J@ SA OD=HH H=PAN QPEHEVA EP BKN = OP=J@=N@ LNK?AOO =OOK?E=PA@ SEPD = MQ=OENACQH=N "ENE?DHAP OL=?A #**  &# $,)),4'+% !,+"'1',+0 $,/  0#1

 ' /# #.2'3)#+1 #!& ,1&#/

E

'0 #5!#-1',+)

EE

'0 #5!#-1',+)

EEE

'0 !,+1'+#" '+  -/,-#/)6 #5!#-1',+) 0#1

ER

'0 !,+1'+#" '+  -/,-#/)6 #5!#-1',+) 0#1 '+ 1&# 01+"/" 0#+0#

/,,$ 2DA AMQER=HAJ?A KB PDA NOP PDNAA ?KJ@EPEKJO =NA LNKRAJ EJ 8$-2 9 2DA EILHE?=PEKJ EEE ER EE EO =HOK K>REKQO  #*.(  E 'J OPQ@UEJC = &QJP LNK?AOO  =OOK?E=PA@ SEPD = NACQH=N "ENE?DHAP OL=?A ' 2    PDA OP=PA OL=?A KB  JAA@O JKP PK >A PDA AJPENA OL=?A ' 'J@AA@ SA I=U SAHH ?KJOE@AN = KNAH OQ>OAP  KB ' SEPD 2    =J@ = &QJP LNK?AOO  SEPD OP=PA OL=?A '   OQ?D PD=P EP EO =OOK?E=PA@ SEPD = "ENE?DHAP BKNI  KJ  ' 2 EJ PDA OAJOA PD=P EPO PN=JOEPEKJ BQJ?PEKJ KJ '   CAJAN=PAO PDA  OAIECNKQL ?KNNAOLKJ@EJC PK  2DAJ PDA OAP  >A?KIAO =QPKI=PE?=HHU AT?ALPEKJ=H >A?=QOA EP EO LNKLANHU AT?ALPEKJ=H SEPD NAOLA?P PK PDA PNERE=H ATPAJOEKJ  KB  PK '  EO K>P=EJA@ BNKI  >U FKEJEJC ARANU LKEJP KB  =O = PN=L OAA 8$-2  2D9   EO OPEHH =OOK?E=PA@ SEPD  =J@ DAJ?A *AII=  =LLHEAO EE %ERAJ = &QJP LNK?AOO  KJ '   =O =>KRA SA ?=HH = OAP LNKLANHU AT?ALPEKJ=H NAOL LNKLANHU AT?ALPEKJ=H EJ PDA OP=J@=N@ OAJOA EB   EO JA=NHU KNAH 2   =J@ '   '   EO EJR=NE=JP NAOL EJR=NE=JP QL PK PDA HEBA PEIA  5EPD PDEO OHECDP IK@E?=PEKJ KB PDA JKPEKJ KB LNKLAN AT?ALPEKJ=HEPU JKP KJHU *AII=  >QP =HOK SD=P SEHH >A OP=PA@ >AHKS =>KQP =@@EPERA BQJ?PEKJ=HO NAI=EJ R=HE@ BKN = &QJP LNK?AOO  KJ '   =O =>KRA EEE 5A ?=J =J@ SA OD=HH =HHKS =J =J=HKCKQO BNAA@KI KB ?DKE?A KB PDA OP=PA OL=?A =>KQP = OP=J@=N@ LNK?AOO =OOK?E=PA@ SEPD = MQ=OENACQH=N "ENE?DHAP OL=?A =??KIL=JUA@ >U =J =J=HKCKQO IK@E?=PEKJ KB PDA LNKLAN AT?ALPEKJ=HEPU EJ PDA OP=J@=N@ OAJOA 5A MQKPA BNKI 8$-2 9 PDKOA >=OE? JKPEKJO =J@ PDAKNAIO ?KJ?ANJEJC =@@EPERA BQJ?PEKJ=HO KB PDA &QJP LNK?AOO  KJ ' U ?KJRAJPEKJ =JU JQIANE?=H BQJ?PEKJ - KJ ' EO ATPAJ@A@ PK PDA 

429

KJALKEJP ?KIL=?PE?=PEKJ ' >U OAPPEJC -    *AP  ( (     ( >A PDA O=ILHA OL=?A PDA IEJEIQI ?KILHAPA@ =@IEOOE>HA HPN=PEKJ PDA HEBA PEIA =J@ PDA ODEBP KLAN=PKN NAOLA?PERAHU =PP=?DA@ PK PDA &QJP LNK?AOO  J ATPAJ@A@ NA=H R=HQA@ BQJ?PEKJ (

KB 7   =J@   EO ?=HHA@ =J ""'1'3# $2+!1',+) $ EJ =>>NARE=PEKJ EB EP EO ( =@=LPA@ =J@ PDANA ATEOP    SEPD (    7   =J@ = LNKLANHU AT?ALPEKJ=H OAP  ' SEPD !+    ;  '   OQ?D PD=P BKN A=?D   

  (

EO ?=@H=C =J@ JEPA KJ 8 

 (

 

 7  

 =J@ ('

 '

 ( '



6 7  

 NAOL EJ PDA =>KRA @AJEPEKJ EO ?=HHA@ = "#+'+% NAOL #5!#-1',+) 0#1 BKN PDA $  5A NAC=N@ PSK $ O PK >A #.2'3)#+1 EB     !+ (  (   7     MA ;  ' 2DAJ SA ?=J J@ = ?KIIKJ @AJEJC OAP  =J@ = ?KIIKJ LNKLANHU AT?ALPEKJ=H OAP KB     =J@  OQ?D PD=P (

 (

 7      &ANA SA D=RA PK QOA *AII=  PKCAPDAN SEPD PDA B=?P PD=P PDA OAP       '(

 '(

 ' 

7  

EO  IA=OQN=>HA J $ (

EO O=E@ PK >A +'1# !")% =J@ !,+1'+2,20 NAOLA?PERAHU EB EP O=PEOAO PDA NAOLA?PERA LNKLANPU =P ARANU 7  8

BKN A=?D EJ EPO @AJEJC OAP  8 9R=HQA@ ?KJPEJQKQO $ EO ?=HHA@ = -,0''13# !,+1'+2,20 $ .!$ EJ =>>NARE=PEKJ  5A @AJKPA >U   PDA OAP KB =HH .!$ O 5A OD=HH ?=HH =J $ (

,$ ,2+"#" 3/'1',+ EB EP EO KB >KQJ@A@ R=NE=PEKJ EJ 7 KJ A=?D ?KIL=?P OQ>EJPANR=H KB 8 

BKN ARANU TA@ EJ = @AJEJC OAP KB   LKOEPERA KNAH IA=OQNA  EO ?=HHA@ = 0*,,1& *#02/# EB  ?D=NCAO JK AT?ALPEKJ=H OAP =J@ PDANA EO =J JAOP $  OQ?D PD=P  $ EO JEPA BKN A=?D 3 2DA PKP=HEPU KB OIKKPD IA=OQNAO EO @AJKPA@ >U # 2DANA EO = KJA PK KJA ?KNNAOLKJ@AJ?A >APSAAJ PDA AMQER=HAJ?A ?H=OOAO KB   =J@ # >U PDA #327 !,//#0-,+"#+!#  (    HEI # - '( +(  /  - +         # ( 7   DKH@EJC BKN =JU JKJJAC=PERA KNAH BQJ?PEKJ - =J@ AT?AOOERA BQJ?PEKJ /    2DA OIKKPD IA=OQNA ?KNNAOLKJ@EJC PK = .!$  EJ PDA =>KRA S=U EO O=E@ PK >A PDA #327 *#02/# ,$  $KN =JU 8   PDANA ATEOPO = QJEMQA JEPA OIKKPD IA=OQNA  ) KJ ' SDE?D O=PEOAO PDA BKHHKSEJC AMQ=PEKJ EJ PDA ?=OA PD=P 8     -  ;  ) +;   8  - 8   8 - -     

 ) EO ?=HHA@ PDA #+#/%6 *#02/# ,$ 8   2DA #+#/%6 ,  KB =J $  EO @AJA@ >U ,   HEI (

 # (  7



430

)

$KN 8   ( @AJA@ >U  EO QL PK PDA AMQER=HAJ?A OLA?EA@ =>KRA = JEPA ?=@H=C $ KB JEPA AJANCU 2DA B=IEHEAO KB */1'+%)# 0 ,$ +'1# #+#/%6 =J@  0 ,$ 7#/, #+#/%6 =NA @AJA@ NAOLA?PERAHU >U 

   JEPA ?=@H=C $ + (   + (   MA =J@ ,      

  +

(

 MA =J@ ,  

$KN =JU 8   PDA $  ) =@IEPO = @A?KILKOEPEKJ  QJEMQAHU QL PK PDA AMQER=HAJ?A OLA?EA@ 

=>KRA SDANA  )  =J@ )   8$-2  2D9  2DA AJANCU IA=OQNA  ) KB 8

) ?KEJ?E@AO SEPD PDA 0ARQV IA=OQNA KB PDA MQ=@N=PE? R=NE=PEKJ  )     KB PDA $  8$-2  2D 9  $QNPDANIKNA 8$-2  2D9 =OOANPO PDA BKHHKSEJC PDA !$ ) EO KB >KQJ@A@ R=NE=PEKJ EB =J@ KJHU EB PDA ?KJ@EPEKJ EE KB 2DAKNAI  EO R=HE@ BKN OKIA OECJA@ OIKKPD IA=OQNA  'J PDEO ?=OA IKNAKRAN ) =@IEPO =J ATLNAOOEKJ  BKN OKIA .!$ O ! ?KNNAOLKJ@EJC PK OIKKPD IA=OQNAO  !  1    >U PDA 0ARQV ?KNNAOLKJ@AJ?A =J@ EP DKH@O PD=P        X 2 X X >A EPO ,KS SA PQNJ PK = CAJAN=H MQ=OENACQH=N "ENE?DHAP OL=?A ' 2    *AP ' X  NACQH=N NALNAOAJP=PEKJ =J@ X !X+ X   X X( (    ' X(    X =OOK?E=PA@ SEPD PDA H=PPAN KJA I=U P=GA BKN EJOP=J?A = OPNKJCHU NACQH=N >A = &QJP LNK?AOO KJ ' NALNAOAJP=PEKJ =J@ = &QJP IK@E?=PEKJ KB PDA =OOK?E=PA@ 0=U LNK?AOO  U 2DAKNAI  PDA PSK "ENE?DHAP OL=?AO =NA NAH=PA@ A=?D KPDAN FQOP =O EJ PDA NOP L=N=CN=LD KB PDA LNKKB KB 2DAKNAI X  5  =LLA=NEJC EJ PD=P L=N=CN=LD SEPDKQP  5A OD=HH QOA PDA JKP=PEKJO $  '  X$  ' NALA=PEJC PDA ATLH=J=PEKJ 5A =NA NA=@U PK ?KJOPNQ?P = OP=J@=N@ LNK?AOO KJ ' =OOK?E=PA@ SEPD X >U 5  EJ = OEIEH=N S=U PK 8$ >9  =O =J EI=CA KB  X X LLHUEJC PDA OPK?D=OPE? ?D=N=?PANEV=PEKJ  PK PDA JAOP X$  SA ?=J J@ =J LNKLANHU X X X AT?ALPEKJ=H KNAH OAP  EJ?HQ@EJC '  ' OQ?D PD=P X !X+ HEI        $

X  X  X ;'



'J KPDAN SKN@O SA D=RA X    !X+ 

X  X X ;'

X   X    X  SDANA BKN PDA OAP  X   X X ' X( ' X (  ' X   X 7   

  X  X   X X  HEI     X 

  (

@FKEJ = LKEJP  PK ' =O =J ATPN= LKEJP =O = LKEJP =P EJJEPU EB ' EO HK?=HHU ?KIL=?P =J@ X   X >U OAPPEJC 5    X 5A @AJA =J ATPAJ@ 5 PK = KJA PK KJA I=LLEJC BNKI '   KJPK ' AT?ALPEKJ=H KNAH OAP   ' >U '



X  X   5  ' 

431



5A HAP X 

X  7  8 9 (  X(  



X $EJ=HHU HAP QO 2DA AHAIAJP KB  NAOL (  EO @AJKPA@ >U NAOL  EJOPA=@ KB X NAOL   @AJA '(   !+ >U X(

  7   '(

 5  ' !+   !X&+ 

; ' 

X 

 



  

   

 

5EPD PDAOA @AJEPEKJO KB AHAIAJPO SA LQP X !X+  X    ( (    ' X(       ( (    '(   !+  X  EO = &QJP LNK?AOO KJ ' X  X SDE?D EO =OOK?E=PA@ SEPD PDA NACQH=N "ENE?DHAP BKNI  X 5A OD=HH  X  6 1&# .20'&,*#,*,/-&'0* 5   ?=HH  1&# '*%# ,$ 1&# 2+1 -/,!#00  %#,.#*   "#+#" 6    '0  01+"/" -/,!#00 ,+ '  .20'/#%2)/ '/'!&)#1 $,/*  2/1&#/  '0 0-#!') +" 1'%&1



00,!'1#" 4'1& 1&#

/,,$ 2DA NOP =OOANPEKJ ?=J >A LNKRA@ EJ PDA O=IA S=U =O EJ 8$ > $9 SDANA X 2 X X =J@  X SANA = NACQH=N "ENE?DHAP OL=?A EPO OPNKJCHU NACQH=N NALNA ' 2    ' X  OAJP=PEKJ =J@ = &QJP IK@E?=PEKJ KB PDA 0=U LNK?AOO =OOK?E=PA@ SEPD PDA H=PPAN NAOLA?PERAHU =J@ PDA LNK?AOO  @AJA@ >U    S=O ODKSJ PK >A = &QJP LNK?AOO KJ '    2DA KJHU @AANAJ?A BNKI PDA LNAOAJP OEPQ=PEKJ S=O EJ PD=P $  X$  SANA JAOPO EJ PDA OAJOA KB X 8$-2 9 N=PDAN PD=J JAOP =J@ JAOP =J@ =??KN@EJCHU SA D=@ PDA BKHHKSEJC OPNKJCAN LNKLANPU PD=J   X  X  !X+ HEI  ;'   

  X $



'J PDA LNAOAJP ?=OA SA ?=J =HOK OAA =O EJ 8$ >9 PD=P ( (   @AJA@ >U  EO PDA IEJEIQI ?KILHAPA@ =@IEOOE>HA HPN=PEKJ BKN '( @AJA@ >U   2DQO  EO OLA?E=H J=IAHU (  EO MQ=OE X $ >A ?KIL=?P OAPO EJ?NA=OEJC PK ' X OQ?D PD=P  X $ EO HABP ?KJPEJQKQO >A?=QOA OK EO X(  *AP  X $ =J@ LQP $  5   X $  X$  2DAJ $  =NA EJ?NA=OEJC ?KIL=?P EJ?HQ@A@ EJ PDA EJPANEKN KB  OQ>OAPO KB '   =J@ SA CAP K>REKQOHU PDA PECDPJAOO KB   !+ HEI      ;  '  $



 2DA OP=PA OL=?A KB  ?=J >A AJH=NCA@ PK ' EB JA?AOO=NU >U =J PNERE=H ATPAJOEKJ J=IAHU >U I=GEJC A=?D LKEJP KB  PK >A = PN=L 8+0  2D 9 D=O CERAJ =JKPDAN ?KJOPNQ?PEKJ KB =J 2PECDP OLA?E=H OP=J@=N@ LNK?AOO =OOK?E=PA@ SEPD = MQ=OENACQH=N "ENE?DHAP OL=?A 

432

#*.(  ??KN@EJC PK 8+0  2D  2D9 DKSARAN PSK OP=J@=N@ LNK?AOOAO =OOK?E =PA@ SEPD PDA O=IA MQ=OENACQH=N "ENE?DHAP OL=?A =@IEP = ?KIIKJ LNKLANHU AT?ALPEKJ=H OAP EJ PDA OP=J@=N@ OAJOA OQ?D PD=P PDAEN NAOPNE?PEKJO PK PDA ?KILHAIAJP KB PDEO OAP OD=NA = ?KIIKJ PN=J OEPEKJ BQJ?PEKJ 2DANABKNA EJ @A=HEJC SEPD = OP=J@=N@ LNK?AOO =OOK?E=PA@ SEPD = MQ=OENACQH=N "ENE?DHAP OL=?A SA I=U =OOQIA SEPDKQP HKOO KB CAJAN=HEPU PD=P EP EO =J EI=CA KB = &QJP LNK?AOO >U = MQ=OEDKIAKIKNLDEOI 2DEO REASLKEJP EO RANU ?KJRAJEAJP EJ QPEHEVEJC PDA PN=JOBAN IAPDK@ 2DQO EJ PDA NAOP KB PDEO OA?PEKJ SA ?KJPEJQA PK SKNG SEPD = MQ=OENACQH=N "ENE?DHAP OL=?A ' 2   =J@ =J =OOK?E=PA@ OP=J@=N@ LNK?AOO    (  '(   !+ KJ '   @AJA@ >U     5A ?=J PN=JOBAN *AII=  =O BKHHKSO #**  &# $,)),4'+% !,+"'1',+0 $,/  0#1

 ' /# #.2'3)#+1 #!& ,1&#/

E

'0  #5!#-1',+)

EE

'0  #5!#-1',+)

EEE

'0 !,+1'+#" '+  -/,-#/)6 #5!#-1',+) 0#1 '+ 1&# 01+"/" 0#+0#

#*.(  1EJ?A PDA OP=PA OL=?A KB  EO '    =JU  LNKLANHU AT?ALPEKJ=H OAP EJ PDA OP=J@=N@ OAJOA EO NAMQENA@ PK ?KJP=EJ PDA OAP  OAA 0AI=NG   X  X  -B /,,$ $KN OEILHE?EPU HAP >A = KNAH OQ>OAP KB '   =J@ LQP X  5  ' X ?KQNOA EO AT?ALPEKJ=H E X EO AT?ALPEKJ=H U KQN ?KJOPNQ?PEKJ KB   SA OAA PD=P PDA  AT?ALPEKJ=HEPU KB NAOL  LNKLAN AT?ALPEKJ=HEPU EJ PDA OP=J@=N@ OAJOA KB   EO X  AT?ALPEKJ=HEPU KB X NAOL X  LNKLAN AT?ALPEKJ=HEPU KB X  X  QP EJ AMQER=HAJP PK PDA  REAS KB *AII=  =J@ 0AI=NG  PDNAA ?KJ@EPEKJO KB PDA LNAOAJP HAII= =NA AMQER=HAJP BKN X  X  PDA NACQH=N "ENE?DHAP BKNI X =J@ PDA &QJP LNK?AOO   #*.(  'J NAH=PEKJ PK PDA OAP EJRKHRA@ EJ = LNKLANHU AT?ALPEKJ=H OAP EJ PDA OP=J@=N@ OAJOA SA I=GA PDA BKHHKSEJC NAI=NG  BKN = KNAH OAP   ' PDA OAP       '(

 '(

  7  8 

 EO  IA=OQN=>HA >A?=QOA BKN   '   =J@ A=?D $   PDA OAP   '(     7  8 $  

 EO ?KJP=EJA@ EJ   '( ATEOPO =J@ EO EJ    7  8 $ 9 =J@ >U 8 %  .NKL9 SA ?=J OAA PD=P  "    

2DA JKPEKJ KB PDA =@@EPERA BQJ?PEKJ=H (

KB PDA LNAOAJP OP=J@=N@ LNK?AOO  KJ '   EO @AJA@ AT=?PHU EJ PDA O=IA S=U =O BKN =J &QJP LNK?AOO AT?ALP PD=P SA JKS =@KLP = LNKLANHU AT?ALPEKJ=H OAP EJ PDA OP=J@=N@ OAJOA EJOPA=@ KB = LNKLANHU AT?ALPEKJ=H OAP =O =J AT?ALPEKJ=H OAP   ' KB PDA =@@EPERA BQJ?PEKJ=H  2DA AMQER=HAJ?A KB PSK $ O    KB 

433

 EO @AJA@ EJ PDA O=IA S=U =O BKN =J &QJP LNK?AOO -J =??KQJP KB *AII=  =J@ 0AI=NG  SA ?=J PDAJ J@ = ?KIIKJ @AJEJC OAP    =J@ = ?KIIKJ AT?ALPEKJ=H OAP   OQ?D PD=P 



(

 (

 7      5A ?=J A=OEHU K>OANRA PD=P EB (

EO =J $ KB  SEPD = @AJEJC OAP  =J@ =J AT?ALPEKJ=H X  SEPD OKIA @AJEJC OAP ?KJP=EJA@ EJ  =J@ OKIA OAP    PDAJ (

EO =J $ KB  X  SEPD = @AJEJC OAP  =J@ X AT?ALPEKJ=H OAP ?KJP=EJEJC 5    !KJRANOAHU =JU $ KB  =J AT?ALPEKJ=H OAP X  X ?=J >A REASA@ =O =J $ KB  SEPD PDA O=IA @AJEJC OAP =J@ PDA AT?ALPEKJ=H OAP 5  X

 2SK $ O =NA AMQER=HAJP SEPD NAOLA?P PK  E OK PDAU =NA SEPD X  NAOLA?P PK  4=NEKQO ?H=OOAO KB $ O KB  =NA @AJA@ EJ PDA O=IA S=U =O BKN =J &QJP LNK?AOO 'J L=NPE?HQN 

SA D=RA PDA ?H=OOAO   KB .!$ O  KB I=NCEJC=HA $ O KB JEPA AJANCU =J@  KB ?KJPEJQKQO $ O KB VANK AJANCU BKN PDA LNK?AOO   $KN 8   =J@ EPO MQ=OE?KJPEJQKQO RANOEKJ 8  SA LQP X  X  8 X X ;  8 5  X ;

; X' X 2DAJ 8 X EO =J MQ=OE?KJPEJQKQO RANOEKJ KB 8 X  8 =J@ X (

 8 X

  8 '(

 8 '

 8 X ' X ' &AJ?A  ) PDA HABP D=J@ OE@A EO QL PK PDA AMQER=HAJ?A = JEPA ?=@H=C $ =J@ QJEMQAHU X  ATLNAOOE>HA =O  BKN  >A?=QOA OK EO  ) PDA NECDP D=J@ OE@A BKN  $KN PDA LNAOAJP MQ=OENACQH=N "ENE?DHAP BKNI  PDA ?H=OO # KB OIKKPD IA=OQNAO =J@ PDA JKPEKJ KB PDA AJANCU IA=OQNA  ) KB 8   =NA @AJA@ =HOK EJ PDA O=IA S=U =O BKN = NACQH=N "ENE?DHAP BKNI 5 LNAOANRAO PDA JKPEKJ KB PDA OIKKPDJAOO KB LKOEPERA IA=OQNAO  EO PDA 0ARQV IA=OQNA KB X   2DA AJANCU IA=OQNA KB = .!$  KB  EB 5 EO PDA 0ARQV IA=OQNA KB  =O = .!$ KB  8   ?D=N=?PANEVA@ >U PDA AMQ=PEKJ  ?=J >A ?KJOPNQ?PA@ =O PDA EI=CA >U 5  KB PDA AJANCU X IA=OQNA KB 8  X SEPD NAOLA?P PK PDA NACQH=N "ENE?DHAP BKNI  1QIIEJC QL SD=P D=O >AAJ IAJPEKJA@ SA CAP %#,.#*  E &# #.2'3)#+!# !)00#0 ,$  0   ,$  +" 1&# 0*,,1& *#02/#0 # ,$  /# '+ ,+# 1, ,+# #327 !,//#0-,+"#+!# 6  EE +6 8   "*'10  2+'.2# +'1# 0*,,1& *#02/#  ) 01'0$6'+% 1&# #.21',+  '+ 1&# !0# 1&1 8    EEE ,/ +6 8   1&#   "*'10 1&# "#!,*-,0'1',+  2+'.2#)6 2- 1, 1&# #.2'3)#+!# 

4&#/#  )  )     ) "*'10 '10 .2"/1'! 3/'1',+ '+   4&,0# #327 *#02/# '0 1&# #+#/%6 *#02/# ,$ 8 $EJ=HHU SA PN=JOBAN 2DAKNAI  KB 8$-2 9 0A?=HH PD=P =J $ (

EO O=E@ PK >A KB >KQJ@A@ R=NE=PEKJ EB EP EO KB >KQJ@A@ R=NE=PEKJ EJ 7 KJ A=?D ?KIL=?P EJPANR=H KB 8 

BKN ARANU TA@ EJ EPO @AJEJC OAP



434

%#,.#*  &# $,)),4'+% !,+"'1',+0 /# #.2'3)#+1 $,/ 8   E &#   ) "#+#" 6  '0  0#*'*/1'+%)# '+ 1&# 0#+0# 1&1 1&#   -,0'1',+  '0 ,$ ,2+"#" 3/'1',+

)

'+ '10 "#!,*

EE &#/# #5'010  0'%+#" 0*,,1& *#02/#  4'1& 0,*# 11!&#" +#01 $  $,/ 4&'!& 1&# #.21',+  &,)"0 $ !,+"'1',+ EE &,)"0 1&#+  !  1    4&#/#

)

"*'10 + #5-/#00',+  $,/ !    4'1& #327 *#02/#     



'0 1&# ,/"+ "#!,*-,0'1',+ ,$ 1&# 0*,,1& 0'%+#" *#02/#  &#+!#  !  1    /# 2

) "*'10 + #5-/#00',+  6 0,*# 1,*1'!))6 0*,,1& $ !,+"'1',+ E &,)"0 1&#+ !    1    +" !,+"'1',+ EE '0 $2)))#" $,/ 1&# 0'%+#" 0*,,1& *#02/#  ,$  "#+#"  6 1&# #327 *#02/#  ! ,$ !  1   



  %' %'%*'"!& " &

*AP ' 2   >A = MQ=OENACQH=N "ENE?DHAP OL=?A =J@   '(  !+ >A =J =OOK?E=PA@ OP=J@=N@ LNK?AOO OLA?EA@ EJ 2DAKNAI  =O =J EI=CA KB = &QJP LNK?AOO >U = MQ=OEDKIAKIKNLDEOI U 2DAKNAI  =J@ 2DAKNAI  SA D=RA %#,.#*  &# $,)),4'+% 14, !,+"'1',+0 /# #.2'3)#+1 $,/ 8   E &#/# #5'010 + +#01 $  $,/ 4&'!& 1&# '+#.2)'16  '0 3)'" $,/ 0,*# -,0'1'3# !,+01+1 $  EE &#   ) "#+#" 6  '0  0#*'*/1'+%)# '+ 1&# 0#+0# 1&1 1&#   "#!,*-,0'1',+  '0 ,$ ,2+"#" 3/'1',+

)

'+ '10

&#+ ,+# ,$ 1�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



435

=J@    !DKKOA 9$   SDE?D =NA ?KJRANCAJP PK 9 2DAJ 9$   9$ EO  ?KJRANCAJP PK 9  9 $NKI PDA R=HE@EPU KB  BKN BQJ?PEKJO EJ  SA CAP  8 9$

 ! 9$ !     'P OQ?AO PK HAP 3 =J@ PDAJ  



%#,.#*  ,/ 8   1&# $,)),4'+% !,+"'1',+0 /# #.2'3)#+1 E &# '+#.2)'16  &,)"0 $,/ +6 9 '+ 1&# 0-!#  01'0$6'+% !,+"'1',+  EE &#/# #5'010  2+'.2# +'1# 0'%+#" 0*,,1& *#02/#  $,/ 4&'!& 1&# #.21',+  '0 3)'" $,/ +6 9    EEE &# !,+1'+2,20  01'0#0 1&# -/,-#/16 

) (

'+ 1&# "#!,*-,0'1',+  ,$ 1&#   '0 ,$ ,2+"#" 3/'1',+ +"

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

)

 (  HEI # (  (    '    '   ( 7

=NNEREJC =P EEE  !KJRANOAHU =OOQIA EEE =J@ HAP  (   

)

( 

) ( 

 (   

)

( 

) ( 

-J =??KQJP KB 2DAKNAI  ?KJ@EPEKJ EE DKH@O BKN PDA OECJA@ OIKKPD IA=OQNA  SDANA  ! EO @AJA@ PK >A PDA 0ARQV IA=OQNA KB !  1    2DAJ  EO JEPA >A?=QOA >U ?KJ@EPEKJ      '    '  HEI # (  (  HEI #  ( 7 ( 7

)

(   

#*.(  .NKLANPU  O=UO PD=P PDA 0ARQV IA=OQNA KB PDA .!$  ) ( D=O = JEPA PKP=H I=OO U 8$-2  2D9 =JU .!$  =J@ EPO 0ARQV IA=OQNA  =NA NAH=PA@ >U  (   ( # - '' +'  -  4'  +6 -   

%

&AJ?A LNKLANPU  EILHEAO PDA !# EJPACN=>EHEPU # 

)

(  

7  



'B PDA LNK?AOO EO ?KJOANR=PERA EJ PDA OAJOA PD=P 4'    6   PDAJ PDA EJPACN=>EHEPU  EILHEAO LNKLANPU   

436

'J PDA NAOP KB PDEO OA?PEKJ SA =OOQIA PD=P ' 2   EO = NACQH=N "ENE?DHAP OL=?A =J@   '(  !+ EO =JU =OOK?E=PA@ &QJP LNK?AOO -B ?KQNOA PDA LNA?A@EJC PSK PDAKNAIO NAI=EJ R=HE@ QJ@AN PDA LNAOAJP =OOQILPEKJ *AP QO BQNPDAN =OOQIA PD=P  EO OPNKJCHU HK?=H EB 8 9   =NA KB ?KIL=?P OQLLKNP =J@ 9 EO ?KJOP=JP KJ = JAECD>KQNDKK@ KB PDA OQLLKNP KB 8 PDAJ  8 9   2DAJ PDA =OOK?E=PA@ &QJP LNK?AOO  ?=J >A P=GAJ PK >A = @EQOEKJ J=IAHU KB ?KJPEJQKQO O=ILHA L=PDO KJ 8 

SEPD JK GEHHEJC EJOE@A  BQJ?PEKJ 8 EO O=E@ PK >A ),!))6 '+  8  "% EJ JKP=PEKJ EB BKN =JU NAH=PERAHU ?KIL=?P KLAJ OAP  PDANA EO = BQJ?PEKJ :   OQ?D PD=P 8  : 2  ), KJ  *AP 8  "%  8 =@IEPO =J MQ=OE?KJPEJQKQO RANOEKJ 8  2DA AJANCU IA=OQNA  ) KB 8 EO OPEHH SAHH @AJA@ 2DA $  ) @AJA@ >U  =@IEPO = @A?KILKOEPEKJ  

SDANA  ) "% =J@ )   "% 2DA @A?KILKOEPEKJ EO QJEMQA QL PK PDA AMQER=HAJ?A KB

) D=O =O EPO 0ARQV IA=OQNA PDA AJANCU HK?=H $ O 2DA MQ=@N=PE? R=NE=PEKJ  )     KB  IA=OQNA  ) KB 8 1AA 8$-2  $9 BKN @AP=EHO KB PDA =>KRA JKPEKJO JKP=PEKJO =J@ OP=PAIAJPO O BKN PDA OAIEI=NPEJC=HA ?D=N=?PANEV=PEKJ KB )   "% BKN 8  "%  =HH OP=PAIAJPO EJ 2DAKNAI  NAI=EJ PNQA AT?ALP PD=P PDA JAOP $  =LLA=NEJC EJ PDA ?KJ@EPEKJ EE PDANA EO JKS NAMQENA@ PK ?KJOEOP KB ?KIL=?P OQ>OAPO KB ' 8$-2  PD9  2DA JATP ?KJ@EPEKJ BKN = OQ>OAP  KB  ' EO P=GAJ BNKI 8$-2 9 ?KJ@EPEKJ  KB 8$-2 9    EO = @AJOA OQ>=HCA>N= KB  '  $KN =JU ?KIL=?P OAP  =J@ NAH=PERAHU ?KIL=?P KLAJ OAP  SEPD     =@IEPO =J AHAIAJP 8 OQ?D PD=P 8   8   KJ  =J@ 8   KJ '   *AP  >A = OQ>OAP KB    ' O=PEOBUEJC >KPD ?KJ@EPEKJO  =J@  1Q?D = OQ>OAP EO OEIEH=N PK = OLA?E=H OP=J@=N@ ?KNA EJ PDA OAJOA KB 8$-2 9 =J@ PDA KJHU @EANAJ?A HEAO EJ PDA @AJEPEKJ KB = NA=H BQJ?PEKJ  7 =LLA=NEJC EJ ?KJ@EPEKJ  $KN =JU KLAJ OAP   ' SA LQP   8    8     5, KJ '     8    #844889    EO QJEBKNIHU @AJOA EJ   =J@  @AJOA EJ  8$-2  *AI9  %#,.#*  &# +#51 !,+"'1',+0 /# #.2'3)#+1 $,/ 8  "% E ,/ +6 /#)1'3#)6 !,*-!1 ,-#+ 0#1   ' 1&#/# '0  -,0'1'3# !,+01+1  02!& 1&1  8 9   !9!

9   



EE &#/# #5'010  0'%+#" ",+ *#02/#  ,+ ' !&/%'+% +, 0#1 ,$ 7#/, !-!'16 02!& 1&1   8 9   9 ;  +; 9    

EEE

) (

'0 ,$ ,2+"#" 3/'1',+ +" 01'0#0 -/,-#/16 

+ 1&'0 !0# ) "*'10 + #5-/#00',+  6  0 !  1    4&,0# #320 *#02/#0  !  1    /# 0*,,1& ",+ *#02/#0 /#)1#" 1, 1&# 0'%+#" *#02/#  ,$ EE 6 



437

/,,$ 2DA EILHE?=PEKJ EE E EO PNERE=H !KJRANOAHU OQLLKOA ?KJ@EPEKJ E EO BQHHHA@ 2DAJ PDANA ATEOPO = QJEMQA OECJA@ 0=@KJ IA=OQNA KJ ' BKN SDE?D PDA AMQ=PEKJ  DKH@O 5A D=RA PK ODKS PD=P  ?D=NCAO JK OAP KB VANK ?=L=?EPU 2K PDEO AJ@ T :   =J@ ?DKKOA = NAH=PERAHU ?KIL=?P KLAJ OAP  ?KJP=EJEJC PDA OQLLKNP KB : 2=GA 8   O=PEOBUEJC 8  8 KJ  PDAJ  8 :9   8  :9  9   =J@ SA D=RA BNKI E   8  :  9    !9!

9  



SDANA      !:!  1EJ?A  EO =J =HCA>N= O=PEOBUEJC ?KJ@EPEKJ   SA ?=J ATPAJ@ EJAMQ=HEPU  BNKI  PK   'J B=?P GAALEJC PDA JKP=PEKJO EJ PDA LNKKB KB *AII=  SA ?=J ODKS PD=P :  9$ EO SA=GHU ?KJRANCAJP PK :  9 U RENPQA KB .NKLKOEPEKJ  PDA EJAMQ=HEPU  DKH@EJC BKN 8 =J@ BKN =JU 9   EILHEAO PD=P PDA AMQ=PEKJ  EO R=HE@ BKN 8 =J@ BKN = JEPA OECJA@ IA=OQNA * ?D=NCEJC JK OAP KB VANK ?=L=?EPU  ?KIL=NEOKJ SEPD  UEAH@O *  :   =J@ SA ?KJ?HQ@A PD=P  ?D=NCAO JK OAP KB VANK ?=L=?EPU >A?=QOA : EO =J =N>EPN=NU AHAIAJP KB  2DA LNKKB KB PDA AMQER=HA?A KB E =J@ EE EO ?KILHAPA !KJOE@AN NAH=PERAHU ?KIL=?P KLAJ OAPO !  OQ?D PD=P !  !  

 !  !  '

(QOP =O EJ PDA LNKKB KB 8$-2  !KN9 SA ?=J OAA PD=P ?KJ@EPEKJ EE EO AMQER=HAJP PK PDA R=HE@EPU KB 2DAKNAI  EE BKN 8  "%  = OECJA@ 0=@KJ IA=OQNA  ?D=NCEJC JK OAP KB VANK $  SDE?D EJ PQNJ ?=J >AAJ OAAJ PK >A AMQER=HAJP PK PDA LNK>=>EHEOPE? ?=L=?EPU =J@ PDA JAOP  ?KJ@EPEKJ EEE EJ PDA O=IA S=U =O PDA ?KNNAOLKJ@EJC LNKKB KB 2DAKNAI  >A?=QOA 2DAKNAI  EO =LLHE?=>HA PK PDA LNAOAJP OEPQ=PEKJ EJ REAS KB PDA NAI=NG I=@A EJ PDA L=NL=CN=LD LNA?A@EJC PDA EJPNK@Q?EPEKJ KB PDA OL=?A   #*.(  .NKLANPU  O=UO PD=P PDA 0ARQV IA=OQNA KB PDA .!$  IA=OQNA  EILHEAO PDA EJPACN=>EHEPU  ( 

) #  '' + '    ?KIL=?P 7   

!KJRANOAHU  EILHEAO  EB PDA LNK?AOO EO ?KJOANR=PERA



438

)  (

EO = 0=@KJ





'"&' %'%*'"!& "  (!'"!& ! "##" &'&

*AP " >A PDA +@EIAJOEKJ=H #Q?HE@A=J OL=?A =J@ 2 >A PDA *A>AOCQA IA=OQNA KJ EP 5A ?KJOE@AN = JKJJAC=PERA HK?=HHU EJPACN=>HA BQJ?PEKJ  KJ " =J@ PDA =OOK?E=PA@ #+#/%6 $,/* @AJA@ >U     8 9  #8 ;  #9 ;  ; 2 +;  8 9   "     2DNKQCDKQP PDEO OA?PEKJ HAP QO =OOQIA PDA &=IV= PULA ?KJ@EPEKJ KJ   &    SDANA

2=A KJ #    #   ;  " 

 <  2 +<  % ;   +

PDA 0'+%2)/ 0#1 KB  2DA ?KILHAIAJP "  EO ?=HHA@ PDA /#%2)/ 0#1 KB  3J@AN &  PDA OQLLKNP KB PDA IA=OQNA +2 AMQ=HO "   $QNPDAN PDA BKNI   EO ?HKO=>HA KJ  "    2 =J@ PDA ?HKOQNA      EO = OPNKJCHU HK?=H NACQH=N "ENE?DHAP BKNI KJ  "     2 805 9 8$ >9  2DA =OOK?E=PA@ @EQOEKJ   '(  !+ KJ "  EO ?=HHA@ = "'01,/1#" /,4+'+ *,1',+ 2DA NA=OKJ KB PDEO J=IEJC EO EJ PD=P EB SA =LLHU PDA @A?KILKOEPEKJ  PK ?KKN@EJ=PA BQJ?PEKJO  ;  ;  "%    0  +

PDAJ SA CAP PDA ATLNAOOEKJ KB PDA O=ILHA L=PD '(  '  (  SDANA ( EO = +@EIAJOEKJ=H  (



 (



NKSJE=J IKPEKJ =J@  ( 



 ( 

(



 (

  "%   0  +

!KJ@EPEKJ & EO BQHHHA@ EJ PDA BKHHKSEJC PSK EILKNP=JP ?=OAO '  EO JKJJAC=PERA ?KJPEJQKQO 2 =A KJ "  ''  ;   ;  ;  "  BKN =J KLAJ OAP   "  'J PDA OA?KJ@ ?=OA "    =J@ PDA @EOPKNPA@ NKSJE=J IKPEKJ  NA@Q?AO PK PDA *,"'#" /##!1'+% /,4+'+ *,1',+ KJ  =OOK?E=PA@ SEPD PDA OPNKJCHU HK?=H NACQH=N "ENE?DHAP BKNI   :     8 9  #8 ;  #9 ; 2 +;        :   @AJKPAO PDA ?HKOQNA KB PDA OL=?A KJ     2    2

OPQ@EA@ EJ 8$ =9 &ANA  :   ?KQH@ >A =  "  EJ PDA 1K>KHAR OL=?A     2DA PANI A?=QOA   LNKLAN OQ>OAP KB    

439

 BQJ?PEKJ   "% " EO ?=HHA@  @AJKPA@ >U   &"% EB BKN =JU >KQJ@A@ KLAJ OAP &  "  PDANA ATEOPO = LKOEPERA ?KJOP=JP  OQ?D PD=P  +09 9 2 +;   !9! 9   &  "   

%#,.#*  2--,0#  +,++#%1'3# $2+!1',+   "% " 01'0#0 1&# !,+"'1',+ & &#+ 1&# $,)),4'+% !,+"'1',+0 /# #.2'3)#+1 E   &"%  EE &# "'01,/1#" /,4+'+ -1& '( '0  0#*'*/1'+%)# '+ 1&# 0#+0# 1&1 #!& !,*-,+#+1 (    0  +  '+ 1&# "#!,*-,0'1',+  '0 ,$ ,2+"#" 3/'1',+ +" ""'1',+))6 '1 01'0#0 1&1  (     HEI #  '

+    0  +   ' '  ( 7  $,/ #3#/6 !,*-!1 0#1   "   /,,$ 5A =LLHU 2DAKNAI  PK PDA OPNKJCHU HK?=H NACQH=N "ENE?DHAP BKNI      KJ  ' 2 =J@ =J =OOK?E=PA@ @EQOEKJ   '(  !+ KJ ' SDANA '  "  2    2  5A P=GA    "   SDE?D K>REKQOHU D=O PDA LNKLANPEAO  =J@   2=GEJC =O 8 PDA ?KK@EJ=PA BQJ?PEKJ  "%  PDA ?KJ@EPEKJ E KB 2DAKNAI  NA=@O =O BKHHKSO BKN =JU NAH=PERAHU ?KIL=?P KLAJ OAP   "  PDANA EO = LKOEPERA ?KJOP=JP  OQ?D PD=P   9 ;  ; 2 +;   OQL 9 ;  9     

+ 

'P EO A=OU PK OAA PD=P   &"% EB =J@ KJHU EB PDA =>KRA ?KJ@EPEKJ EO O=PEOA@ BKN A=?D 0        + 2DQO SA CAP 2DAKNAI  BNKI 2DAKNAI    IA=OQN=>HA OAP   " EO ?=HHA@ !!',--,)' EB   &"%  5A GJKS 8#% 9 PD=P  EO !=??EKLLKHE EB =J@ KJHU EB PDANA ATEOPO = LKOEPERA 0=@KJ IA=OQNA KJ  =J@ = IA=OQN=>HA RA?PKN +   " SEPD +   =A OQ?D PD=P   +09 9 2 +;   

9  + +



U OLA?EBUEJC 2DAKNAI  PK     SA CAP 

440

9   "  " 



%#,.#*  &# $,)),4'+% !,+"'1',+0 /# #.2'3)#+1 $,/ + ,-#+ 0#1   " '  '0 !!',--,)' '' &# *,"'#" /##!1'+% /,4+'+ -1& '(  !+  '(  !+ ,+  '0  0#*'*/1'+%)# '+ 1&# 0#+0# 1&1 #!& !,*-,+#+1 ( ,$ 1&# 0#!,+" 1#/* ( '+ '10 "#!,*-,0'1',+  '0 ,$ ,2+"#" 3/'1',+ +" 01'0#0 1&# ""'1',+) -/,-#/16 1&1  (   HEI  #  '' + '  ( 7  + 1&'0 !0# 1&# *,"'#" /##!1'+% /,4+'+ *,1',+ "*'10 + #5-#00',+ ,$ (,/,&," 16-#  ( '(  '  (  + '' +'  

$,/    ( 4'1& #327 *#02/# 

%!& 8+ 9

1 H>ARANEK =J@ 7+ += %#+#/) !,//#0-,+"#+!# #14##+ '/'!&)#1 $,/*0 +" /'%&1 -/,!#00#0 QHH IAN +=PD 1K?    W

8 % 9

0 + HQIAJPD=H =J@ 0 ) %APKKN +=NGKR LNK?AOOAO =J@ LKPAJPE=H PDAKNU ?= @AIE? .NAOO ,AS 6KNG =J@ *KJ@KJ 

8!$5 9 7 / !DAJ . ( $EPVOEIIKJO =J@ 0 ( 5EHHE=IO ##!1'+% /,4+'+ *,1',+0 .20'*/1'+%)#0 +" 01/,+% !!',--,)' 0#10 .KPAJPE=H J=HUOEO    W 8!+0 9 7 / !DAJ 7+ += =J@ + 0Y K?GJAN 20'&,*#,*,/-&'0*0 ,$ '/'!&)#1 $,/*0 ,=CKU= +=PD (    W 8!(.1 9 # !

EJH=N ( (=?K@ .D .NKPPAN =J@ + ( 1D=NLA #*'*/1'+%)# +" /(,3 -/,!#00#0 7 5=DNO?DAEJHE?DGEAPOPDAKNEA RANS %A>EAPA    W 8#% 9

* ! #R=JO =J@ 0 $ %=NEALU +A=OQNA PDAKNU =J@ JA LNKLANPEAO KB BQJ?PEKJO !0! .NAOO K?= 0=PKJ ,AS 6KNG 2KGUK 

8$E 9

. ( $EPVOEIIKJO + 1&# .20'/#%2)/'16 ,$ 0#*''/'!&)#1 $,/*0 .KPAJPE=H J=HU OEO PK =LLA=N

8$% 9

. ( $EPVOEIIKJO =J@ 0 ) %APKKN *,,1& *#02/#0 +" !,+1'+2,20 ""'1'3# $2+! 1',+)0 ,$ /'%&1 /(,3 -/,!#00#0 'P: K O 1PK?D=OPE? !=H?QHQO =J@ .NK>=>EHEPU 2DAKNU , 'GA@= 1 5=JP=J=>A + $QGQODEI= & )QJEP= A@O 1LNEJCAN  W

8$ 9

+ $QGQODEI= !,+01/2!1',+ ,$ /##!1'+% //'#/ /,4+'+ *,1',+0 $,/ ,2+"#" ",*'+0 -O=G= ( +=PD    W

8$ =9

+ $QGQODEI= #%2)/ /#-/#0#+11',+0 ,$ '/'!&)#1 0-!#0 2N=JO I +=PD 1K?    W 

441

8$ >9

+ $QGQODEI= '/'!&)#1 0-!#0 +" 01/,+% /(,3 -/,!#00#0 2N=JO I +=PD 1K?    W

8$ 9

+ $QGQODEI= "ENE?DHAP BKNIO =J@ +=NGKR LNK?AOOAO )K@=JOD= =J@ ,KNPD &KHH=J@ 

8$ =9

+$QGQODEI= '/'!&)#1 $,/*0 !!',--,)' 0#10 +" 1&# (,/,&," #.21',+0 1PK?D=O PE? "EANAJPE=H =J@ "EANAJ?A #MQ=PEKJO !OEOV=N +E?D=HAPVGU A@O ENGDY=QOAN  W

8$ >9

+ $QGQODEI= '01,/1#" /,4+'+ *,1',+0 +"  $2+!1',+0 2NAJ@O EJ .NK>=>EHEPU =J@ J=HUOEO , ): KJK ,0 1DEAD A@O 5KNH@ 1?EAJPE?  W

8$ 9

+$QGQODEI= + "#!,*-,0'1',+ ,$ 06**#1/'! "'20',+ -/,!#00#0 +" /#)1#" 1,-'!0 '+ +)60'0 1QC=GQ    W EJ (=L=JAOA  #JCHEOD PN=JOH=PEKJ PK =LLA=N BNKI +1

8$-2 9 + $QGQODEI= 6 -ODEI= =J@ + 2=GA@= "ENE?DHAP BKNIO =J@ OUIIAPNE? +=NGKR LNK?AOOAO 5=HPAN @A %NQUPAN  8)5 9

& )QJEP= =J@ 2 5=P=J=>A ,*# 1&#,/#*0 !,+!#/+'+% /#0,)3#+10 ,3#/ ),!))6 !,* -!1 0-!#0 .NK? $EBPD ANGAHAU 1UILKO +=PD 1P=PEOP =J@ .NK>=>EHEPU RKH '' L=NP  3JER KB !=HBKNJE= .NAOO ANGAHAU !=HEB  W

8+-0 9 7+ += * -RAN>A?G =J@ + 0Y K?GJAN /(,3 -/,!#00#0 00!,'1#" 4'1& 0#*' '/'!&)#1 $,/*0 -O=G= ( +=PD    W 8+0 9

7+ += =J@ + 0Y K?GJAN 'JPNK@Q?PEKJ PK PDA PDAKNU KB JKJOUIIAPNE? "ENE?DHAP BKNIO 1LNEJCAN4ANH=C 

8+ 9

4 % +=V F= 1K>KHAR OL=?AO 1LNEJCAN4ANH=C 

80 9

" 0=U #0,)3#+10 1/+0'1',+ $2+!1',+0 +" 01/,+%)6 /(,3'+ -/,!#00#0 JJ KB +=PD    W

805 9

+ 0Y K?GJAN =J@ , 5EAHAJO '/'!&)#1 $,/*0!),0 ')'16 +" !&+%# ,$ 0-##" *#02/# 'JJEPA @EIAJOEKJ=H =J=HUOEO =J@ OPK?D=OPE? LNK?AOOAO 0AOA=N?D ,KPAO EJ +=PD 1 H>ARANEK A@ .EPI=J    W



442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

JF7@6;@9 &3D=AH )DA57EE7E ;@ 073= G3>;FK 4K )A;EEA@ )A;@F )DA57EE7E A8 J5GDE;A@E JXU^AY^W 3XU^  =QcQd_cXY 6e[ecXY]Q  Q^T :YQ^WQ^W IY^W 1

2

3

*JUFWYRJSY TK 3FYMJRFYNHX ;SN[JWXNY^ TK =FXMNSLYTS 9JFYYQJ =' "!" ;9' [email protected] :MJ WJXJFWHM TK YMNX FZYMTW NX XZUUTWYJI NS UFWY G^ 49, -WFSY *39 *JUFWYRJSY TK 3FYMJRFYNHX 1FSXFN ;SN[JWXNY^ 9ZNYF 5XFPF !! 0FUFS [email protected] :MJ WJXJFWHM TK YMNX FZYMTW NX XZUUTWYJI NS UFWY G^ -WFSYNS'NI KTW 9HNJSYNH 8JXJFWHM TK 3+>: 4T  *JUFWYRJSY TK 3FYMJRFYNHX ,ZIFS ;SN[JWXNY^ 9MFSLMFN )MNSF jgying@fudan. edu.cn :MJ WJXJFWHM TK YMNX FZYMTW NX XZUUTWYJI NS UFWY G^ 49,) 4T  "

!A@E?=PA@ PK -NKBAOOKN (EUKOE &PN V KJ PDA K??=OEKJ KB DEO PD >ENPD@=U Summary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

 "@FDA6G5F;A@ U =  # B  T . U K[% $ ) 

*KNAKRAN U 0  EO JEPA BKN A=?D [ ) "% YY / KB 1 M . 7 # B  T & # P #% =  L1U U M K[% DXU^ V_b M ! ,C 0 Q^T V_b -NKKB Y GU `ed  M  . 

] ! .C 0 fQ^YcXY^W Qd H gU XQfU ecY^W !  $ ( Q^T dXU BUfej V_b]e\Q K$# $ #%M B  M H . 2 B M P 2 B  M& ] . M& 2 ] P 2 ] B \ & ]2   =   ; = T% 5 1U 1U . #G N L ]CU #1 ' [K[ . #G N L ]CU K[  . #G N

%

;

= L1T

T1

%



T1

T

=

L1U ]CU K[ T .  ]#G N 



B  M H% .  ]2

501



 L1T KGT

!

@7 )MJS JY FQ

8U^SU \ B  T .  &

$ *

Vb_] gXYSX $ ) V_\\_gc Ri \UddY^W  "% CY^SU # B  T Yc Q `ebU\i UhSUccYfU ]UQcebU _V C   dXU e^YaeU^Ucc V_\\_gc SV K$"M DXU ^YdU^Ucc _V U V_\\_gc Vb_] $ & B  T & ] . \Y] U ] V_b Q^i C  UhSUccYfU YY 2i Y Q^T K' _dU dXQd FKT  , Z ! .p G  B Q^T FKT  , Z ! .p & T ' G  B  Yc dXU S_\\USdY_^c _V UhSebcY_^c Q^T _V dXU bUdeb^Y^W UhSebcY_^c QgQi Vb_] H bUc`USdYfU\i gXY\U KA . 0  ! B 1 B F+G Yc dXU ^_^bUdeb^Y^W _b UaeYfQ\U^d\i 7 ' % UhSebcY_^ gXU^UfUb G  ' DXU S_e^dY^W ]UQcebU _V K Yc TU^UT Ri ; # KV &  ! ,B & $ #$ Up Z& [M&  . Vp $T U

Q^T Up [&  . Up "& [M&  Yc dXU^ 06 QTQ`dUT Qc Q `b_SUcc Y^ [  "% EcY^W $ #" gU ^_g ]Q[U dXU dY]U ceRcdYdedU Y^ dXU bU\QdY_^ $ ( d_ _RdQY^   =  A ; E5  c P5 . / Kc M  #B E5 KZ % $ #% #B 

T.$

9^cUbdY^W dXU `bUTYSdQR\U `b_SUcc EV . # 6  \ gU QbbYfU Qd dXU V_b]e\Q X_\TY^W V_b dXU S_e^dY^W ]UQcebU _V dXU `_Y^d `b_SUcc K Qcc_SYQdUT gYdX G, #B KUp [& M . / KM  #B K[ D G M

V_b UfUbi [  " Q^T  ! ,B % $ #&

DXYc V_b]e\Q gY\\ RU edY\YjUT Y^ dXU ^Uhd cUSdY_^

 :3D35F7D;EF;5 &73EGD7 A8 4EAD476 )A;EEA@ )A;@F )DA57EE 9^ dXYc cUSdY_^ gU S_^dY^eU d_ g_b[ gYdX dXU cUddY^W Y^ §$ Q^T Y^fUcdYWQdU `b_`UbdYUc _V dXU `_Y^d `b_SUcc KT & .p  TU^UT Ri $ ## V_b dXU \_SQ\ dY]U

503



@7 )MJS JY FQ

G . FGU & [  "G Qd dXU `_Y^d H% 2i dXU _RcUbfQdY_^ ]QTU QVdUb $ ## Yd dXU^ X_\Tc dXQd G . ?

gXUbU ? . Y^VFZ ) " , KT ! B 1 B F+GG%

% #

9^ fYUg _V @b_`_cYdY_^ $ $  ? Yc Uh`_^U^dYQ\\i TYcdbYRedUT gYdX `QbQ]UdUb

. . B . % *@HH<  2J@AN IA=OQNA .B K EO =J =>OKN>A@ -KEOOKJ LKEJP SEPD =>OKNLPEKJ PEIA ? EJ *AUAN O OAJOA K=M PD=P EO

.B Up Y  Z & Y  [ M&   ! 3 &    & Up Y  ZO & Y  [O M& O  ! 3O

LNK?AOO    04

. #'8 #S( .B Up Z & [ M&   ! 3 &    & Up ZO & [O M& O  ! 3O  % $ #'8 S( #,1 "    #,1 "& BKN =JU Z ' [ &    & ZO ' [O & 3 &    & 3O   & Y ) "&  &    & O ! ,B % -NKKB DXU `b__V Yc dXU cQ]U Qc Y^ K= §$M Q\dX_eWX K=M S_^cYTUbUT _^\i dXU S_^cUbfQdYfU SQcU 9^ VQSd dXU YTU^dYdi SV . S  V 4 Y]`\YUc Up Y  Z& Y  [M&  . Up Z& [M&  4 Q^T S_^cUaeU^d\i gU cUU Vb_] % # Q^T dXU cdb_^W =Qb[_f `b_`Ubdi _V C dXQd dXU \UVd XQ^T cYTU _V % $ gYdX U . # UaeQ\c .KRA ?KJ@EPEKJO    =J@  DKH@ 1DAJ E 1DA IA=OQNA T @KAO JKP ?D=NCEJC KJ FHG TFHG . " EE C =@IEPO JK FQILEJC BNKI 0 PK PDA LKEJP H BKN ARANU _ ! 0 .Y CU1 ! 0 & CU . H BKN OKIA [ ! "&  . "&

& &

EEE C =@IEPO JK FQIL BNKI PDA LKEJP H PK 0 EJ PDA BKHHKSEJC OAJOA .Y CU1 . H& CU ! 0 BKN OKIA [ ! "&  . "

V_b a U _ ! 0% & '

%ANA a U IA=JO AT?ALP KJ =J TLKH=N OAP BKN C ER 1DA KJA LKEJP OAP FHG EO JKP TLKH=N BKN C )AP BQJ?PEKJO # =J@ \ >A @AJA@ =O EJ & % 1DAJ   BKN _ ! 0 % & ( #_ . .Y B '  =J@ \ _ . #Y L1% R \ & \ B ! 7 0 & T  BKN ARANU ) " -NKKB Y DXYc Yc Y]]UTYQdU Vb_]    V_b C Qc = B  M HTFHG . 2 M _2 #'B( _TK_ . " V_b UfUbi M ! ,  0% )

YY 9d V_\\_gc Vb_] & # Q^T  dXQd .Y C% 1 ! 0 & B '

."

V_b UfUbi _ ! 0 %

& )

6_b Q^i _`U^ cUd : dXQd XQc Q `_cYdYfU TYcdQ^SU Vb_] FHG \Ud FBO & U  "G& F O & U  "G RU dXU cd_``Y^W dY]Uc TU^UT Ri   . "& B . B &  O . BO1  3 %11 & BO .  O  B 1

& *

gYdX Q^ _RfY_ec ]_TYSQdY_^ QVdUb _^U _V dXU] RUS_]Uc Y^^Ydi 3\UQb\i dXU dY]U cUd F[ ! "&  , CU1  ! :& CU  . HG  FBO - U . "& #& $&    G% DXec Yd V_\\_gc Vb_] dXU cdb_^W =Qb[_f `b_`Ubdi _V C Q^T & ) dXQd V_b UfUbi _ ! 0  .Y dXUbU Yc c_]U [ ) " ceSX dXQd CU1 ! :& CU . H . "% _dU dXQd C d_ K%# )% # &) #"M dXU cdQ^TQbT `b_SUcc C  _^ 0 XQc Q < Ufi cicdU] 9& 3 _^ 0 DXQd Yc 9 _& K` Yc Q [Ub^U\ _^ 0 & ,0  Q^T 3 Yc Q @316 _V C  Y^ dXU cdbYSd cU^cU gYdX R_e^TUT #`_dU^dYQ\ ceSX dXQd V_b Q^i ^_^^UWQdYfU 2_bU\ Ve^SdY_^ M _^ 0  0 B F G dXQd fQ^YcXUc _^ dXU TYQW_^Q\ Q^T Yc UhdU^TUT d_ RU jUb_ _edcYTU 0  0   = U =  ;  #Y ! ' & M CT1 & CT " . #Y M CT & `9 CT & K`K3T 

T U

)0

B  XQc Q < Ufi V_b UfUbi _ ! 0 Q^T [  " CY]Y\Qb\i dXU cdQ^TQbT `b_SUcc C B & 3 B KNDKK@ @ KB H EJ 0 @AJA ( @G 9 ^ . Y^VF[ ) " , ^[ !

BKN ^ ! B%

1DAJ I F9 ' G '

B F9 ' G ' I

=J@

%

-NKKB GU _^\i WYfU Q `b__V V_b I "DPNDJIN DI #P>GD?@ 65 "   `   6 0 ,NY_XNRRTSX FSI 8 1 -JYTTW 8J[Z_ RJFXZWJX FSI YNRJ HMFSLJX (7J> 4 199 "!! `  6 0 ,NY_XNRRTSX FSI 8 1 -JYTTW 9RTTYM RJFXZWJX FSI HTSYNSZTZX FIINYN[J KZSHYNTSFQX TK WNLMY 3FWPT[ UWTHJXXJX /S $JD ; I .JE9>7IJ?9 7B9KBKI 7D: +HE87 8?B?JO />;EHO +IX 4 /PJIF 9 =FYFSFGJ 3 ,ZPZXMNRF FSI . 1ZSNYF UU `" 9UWNSLJW ""  6 0 ,NY_XNRRTSX FSI 8 - -JYTTW +]HZWXNTS YMJTW^ WJ[NXNYJI YT FUUJFW NS $BB?DE?I % (7J> 50   3 ,ZPZXMNRF 5S GTZSIFW^ HTSINYNTSX KTW RZQYNINRJSXNTSFQ (WT\SNFS RTYNTSX \NYM X^RRJYWNH WJXTQ[JSY IJSXNYNJX % (7J> .E9 %7F7D 21 "" !`"  3 ,ZPZXMNRF 6 .J FSI 0 ?NSL :NRJ HMFSLJX TK X^RRJYWNH INZXNTSX FSI ,JQQJW RJFXZWJX DD +HE878 32  !`  3 ,ZPZXMNRF ? 5XMNRF FSI 3 :FPJIF ?H?9>B;J !EHCI 7D: .OCC;JH?9 (7HAEL +HE9;II;I =FQYJW IJ -WZ^YJW (JWQNS ""

539

"

@7 )MJS JY FQ

 3 ,ZPZXMNRF FSI . :FSFPF 6TNXXTS UTNSY UWTHJXXJX FYYFHMJI YT X^RRJYWNH INZXNTSX DD $DIJ # +E?D97H; +HE878 .J7J?IJ 41  "`"   8 1 -JYTTW 3FWPT[ UWTHJXXJX# 8F^ UWTHJXXJX FSI WNLMY UWTHJXXJX ';9J )EJ;I ?D (7J> 440 9UWNSLJW "  ! 8 1 -JYTTW *ZFQNY^ TK 2J[^ X^XYJR 4 27>HI9> L;HM ";8?;J; 19 "   `  " 8 1 -JYTTW +]HZWXNTSX TK F 3FWPT[ UWTHJXX DD +HE878 7 " " `  8 1 -JYTTW N9;II?L; (;7IKH;I (NWPMc FZXJW ""  8 1 -JYTTW FSI 3 0 9MFWUJ 4FYZWFQNY^ XYFSIFWISJXX FSI \JFP IZFQNY^ KTW 3FWPT[ UWTHJXXJX 4 27>HI9> L;HM ";8?;J; 67 "! `  6 .XZ# 5S J]HZWXNTSX TK WJJHYNSL (WT\SNFS RTYNTS /H7DI C;H (7J> .E9 296 "! "`  4 /PJIF FSI 9 =FYFSFGJ .JE9>7IJ?9 ?;H;DJ?7B GK7J?EDI 7D: ?KI?ED +HE9;II;I 9JHTSI +INYNTS 4TWYM.TQQFSI1TIFSXMF "!"  1 /YD T 6TNXXTS UTNSY UWTHJXXJX FYYFHMJI YT 3FWPT[ UWTHJXXJX NS# +HE9 .?NJ> ;HA;B;O .OCF (7J> .J7J +HE878 III "  `"  . 1ZSNYF -JSJWFQ GTZSIFW^ HTSINYNTSX KTW RZQYNINRJSXNTSFQ INZXNTS UWTHJXXJX % (7J> &OEJE 0D?L 10 "   `  ( 3FNXTSSJZ[J +]NY X^XYJRX DD +HE878 3 "  ""`   6 ' 3J^JW 6WTHJXXZX IJ 6TNXXTS UTSHYZJQX IFUWJX 1 /YD T .;C?D7?H; :; +HE878 ;EH -;B7J !?;B:I 73 "! "`  : 9 9FQNXGZW^ )TSXYWZHYNTS TK WNLMY UWTHJXXJX KWTR J]HZWXNTSX +HE878 />;EH -;B7J !?;B:I 73 "! `  3 0 9MFWUJ ";D;H7B />;EHO E< (7HAEL +HE9;II;I 'HFIJRNH UWJXX "!!  9 =FYFSFGJ 5S XYFGQJ UWTHJXXJX \NYM GTZSIFW^ HTSINYNTSX % (7J> .E9 %7F7D 14"  `"!

540

Acknowledgments

The editors and the publisher wish to thank the following for granting permission to reprint in this volume the papers listed below. Abbreviations in brackets refer to the Bibliography of Masatoshi Fukushima. Academiæ Scientarium Fennicæ [R22] American Mathematical Society [R8], [R9] Cambridge University Press [R53] Kyoto University. Research Institute for Mathematical Sciences [R19] Mathematical Society of Japan [R6], [R12], [R32] Mittag-Leffler Institute [R43] Osaka University (Grad. School of Sciences, Dept. of Mathematics) [R5], [R34], [R38], [R49], [R78] Springer [R11], [R29], [R31], [R55], [R61], [R82] The Japan Academy [R58] University of Illinois (Dept. Mathematics) [R81] We would also like to add that the copyright for [R37] and [R66] is with M. Fukushima, and that for [R59] with the publisher.

541

542

Bibliography

Research Papers (English) [R1]

On Feller’s kernel and the Dirichlet norm, Nagoya Math. J. 24 (1964) 167–175.

[R2]

Resolvent kernels on a Martin space, Proc. Japan Acad. 41 (1965), 260–263.

[R3]

On spectral functions related to birth and death processes, J. Math. Kyoto Univ. 5 (1966), 151–161.

[R4]

On a class of Markov processes taking values on lines and the central limit theorem, (with M. Hitsuda), Nagoya Math. J. 30 (1967), 47–56.

[R5]

A construction of reflecting barrier Brownian motions for bounded domains, Osaka J. Math. 4 (1967), 183–215.

[R6]

On boundary conditions for multi-dimensional Brownian motions with symmetric resolvent densities, J. Math. Soc. Japan 21 (1969), 58–93.

[R7]

On Dirichlet spaces and Dirichlet rings, Proc. Japan Acad. 45 (1969), 433–436.

[R8]

Regular representations of Dirichlet spaces, Trans. Amer. Math. Soc. 155 (1971), 455– 473.

[R9]

Dirichlet spaces and strong Markov processes, Trans. Amer. Math. Soc. 162 (1971), 185–224.

[R10]

On transition probabilities of symmetric strong Markov processes, J. Math. Kyoto Univ. 12 (1972), 431–450.

[R11]

On the generation of Markov processes by symmetric forms, in Proceedings of the Second Japan-USSR Symposium on Probability Theory, (Kyoto, 1972), 46–79, Lecture Notes in Math. 330, Springer, Berlin, 1973.

[R12]

Almost polar sets and an ergodic theorem, J. Math. Soc. Japan 26 (1974), 17–32.

[R13]

On the spectral distribution of a disordered system and the range of a random walk, Osaka J. Math. 11 (1974), 73–85.

[R14]

Local property of Dirichlet forms and continuity of sample paths, Z. Wahrscheinlichkeitstheorie und Verw. Gebiete 29 (1974), 1–6.

[R15]

On an asymptotic property of spectra of a random difference operator (with H. Nagai and S. Nakao), Proc. Japan Acad. 51 (1975), 100–102.

543

Bibliography [R16]

Asymptotic properties of the spectral distributions of disordered systems, in International Symposium on Mathematical Problems in Theoretical Physics, (Kyoto Univ., Kyoto, 1975), 224–227, Lecture Notes in Phys. 39, Springer, Berlin, 1975.

[R17]

On spectra of the Schrödinger operator with a white Gaussian noise potential (with S. Nakao), Z. Wahrscheinlichkeitstheorie und Verw. Gebiete 37 (1976/77), 267–274.

[R18]

Potential theory of symmetric Markov processes and its applications, in Proceedings of the Third Japan-USSR Symposium on Probability Theory, (Tashkent, 1975), 119– 133. Lecture Notes in Math. 550, Springer, Berlin, 1976.

[R19]

On an  -estimate of resolvents of Markov processes, Publ. Res. Inst. Math. Sci. 13 (1977/78), 277–284.

[R20]

A decomposition of additive functionals of finite energy, Nagoya Math. J. 74 (1979), 137–168.

[R21]

On additive functionals admitting exceptional sets, J. Math. Kyoto Univ. 19 (1979), 191–202.

[R22]

Dirichlet spaces and additive functionals of finite energy, in Proceedings of the International Congress of Mathematicians, (Helsinki, 1978), 741–747, Acad. Sci. Fennicæ, Helsinki, 1980.

[R23]

A generalized stochastic calculus in homogenization, in Quantum fields–algebras, processes, (Proc. Sympos., Univ. Bielefeld, Bielefeld, 1978), 41–51, Springer, Vienna, 1980.

[R24]

On a stochastic calculus related to Dirichlet forms and distorted Brownian motions, in New stochasitic methods in physics, Phys. Rep. 77 (1981), 255–262.

[R25]

On a representation of local martingale additive functionals of symmetric diffusions, in Stochastic integrals, (Proc. Sympos. Durham, 1980), 110–118, Lecture Notes in Math. 851, Springer, Berlin, 1981.

[R26]

Capacity and quantum mechanical tunneling (with S. Albeverio, W. Karwowski and L. Streit), Comm. Math. Phys. 81 (1981), 501–513.

[R27]

On asymptotics of spectra of Schrödinger operators, in Statistical and physical aspects of Gaussian processes, (Saint-Flour, 1980), 335–347, Colloq. Internat. CNRS 307, CNRS, Paris, 1981.

[R28]

On absolute continuity of multidimensional symmetrizable diffusions, in Functional analysis in Markov processes, (Katata/Kyoto, 1981), 146–176, Lecture Notes in Math. 923, Springer, Berlin-New York, 1982.

[R29]

A note on irreducibility and ergodicity of symmetric Markov processes, in Stochastic processes in quantum theory and statistical physics, (Marseille, 1981), 200–207, Lecture Notes in Phys. 173, Springer, Berlin, 1982.

[R30]

Markov processes and functional analysis, in Proc. International Math. Conf. Singapore, 187–202, Eds. L.H. Chen, Y.B. Ng, M.J. Wicks, North Holland, 1982.

544

Bibliography [R31]

Capacitary maximal inequalities and an ergodic theorem, in Probability theory and mathematical statistics, (Tbilisi, 1982), 130–136, Lecture Notes in Math. 1021, Springer, Berlin, 1983.

[R32]

Basic properties of Brownian motion and a capacity on the Wiener space, J. Math. Soc. Japan 36(1984), 161–176.

[R33]

On conformal martingale diffusions and pluripolar sets (with M. Okada), J. Funct. Anal. 55 (1984), 377–388.

[R34]

A transformation of a symmetric Markov process and the Donsker-Varadhan theory (with M. Takeda), Osaka J. Math. 21 (1984), 311–326.

[R35]

A Dirichlet form on the Wiener space and properties on Brownian motion, in Théorie du potentiel, (Orsay, 1983), 290–300, Lecture Notes in Math. 1096, Springer, Berlin, 1984.

[R36]

Energy forms and diffusion processes, in Mathematics + Physics. Vol. 1, 65–97, World Sci. Publishing, Singapore, 1985.

[R37]

  -capacities for general Markovian semigroups (with H. Kaneko), in Infinitedimensional analysis and stochastic processes, (Bielefeld, 1983), 41–47, Res. Notes in Math. 124, Pitman, Boston, MA, 1985.

[R38]

On the continuity of plurisubharmonic functions along conformal diffusions, Osaka J. Math. 23 (1986), 69–75.

[R39]

A stochastic approach to the minimum principle for the complex Monge-Ampère operator, in Stochastic processes and their applications, (Nagoya, 1985), 38–50, Lecture Notes in Math. 1203, Springer, Berlin, 1986.

[R40]

Reversibility of solutions to martingale problems (with D. W. Stroock), in Probability, statistical mechanics, and number theory, 107–123, Adv. Math. Suppl. Stud. 9, Academic Press, Orlando, FL, 1986.

[R41]

On recurrence criteria in the Dirichlet space theory, in From local times to global geometry, control and physics, (Coventry, 1984/85), 100–110, Pitman Res. Notes Math. Ser. 150, Longman Sci. Tech., Harlow, 1986.

[R42]

On Dirichlet forms with random data—recurrence and homogenization (with S. Nakao and M. Takeda), in Stochastic processes—mathematics and physics, II, (Bielefeld, 1985), 87–97, Lecture Notes in Math. 1250, Springer, Berlin, 1987.

[R43]

On Dirichlet forms for plurisubharmonic functions (with M. Okada), Acta Math. 159 (1987), 171–213.

[R44]

A note on capacities in infinite dimensions, in Probability theory and mathematical statistics, (Kyoto, 1986), 80–85, Lecture Notes in Math. 1299, Springer, Berlin, 1988.

[R45]

On holomorphic diffusions and plurisubharmonic functions, in Geometry of random motion, (Ithaca, N.Y., 1987), 65–78, Contemp. Math. 73, Amer. Math. Soc., Providence, RI, 1988.

545

Bibliography [R46]

On two classes of smooth measures for symmetric Markov processes, in Stochastic analysis, (Paris, 1987), 17–27, Lecture Notes in Math. 1322, Springer, Berlin, 1988.

[R47]

On the skew product of symmetric diffusion processes (with Y. Oshima), Forum Math. 1 (1989), 103–142.

[R48]

Capacities on Wiener space: tightness and invariance (with S. Albeverio, W. Hansen, Z.-M. Ma and M. Röckner), C. R. Acad. Sci. Paris Se’r. I Math. 312 (1991), 931–935.

[R49]

On quasi-supports of smooth measures and closability of pre-Dirichlet forms (with Y. LeJan), Osaka J. Math. 28 (1991), no. 4, 837–845.

[R50]

On the closable parts of pre-Dirichlet forms and the fine supports of underlying measures (with K. Sato and S. Taniguchi), Osaka J. Math. 28 (1991), 517–535.

[R51]

An invariance result for capacities on Wiener space (with S. Albeverio, W. Hansen, Z.-M. Ma and M. Röckner), J. Functional Analysis 106 (1992), 35–49.

[R52]

On   2-capacities for a class of elliptic pseudo differential operators (with N. Jacob and H. Kaneko), Math. Ann. 293 (1992), 343–348.

[R53]

Dirichlet forms, diffusion processes and spectral dimensions for nested fractals, in Ideas and methods in mathematical analysis, stochastics, and applications, (Oslo, 1988), 151–161, Cambridge Univ. Press, Cambridge, 1992.

[R54]

  -capacities and Hunt processes in infinite dimensions, in Probability theory and mathematical statistics, (Kiev, 1991), 96–103, World Sci. Publishing, River Edge, NJ, 1992.

[R55]

On a spectral analysis for the Sierpi´nski gasket (with T. Shima), Potential Analysis 1 (1992), 1–35.

[R56]

Two topics related to Dirichlet forms: quasi-everywhere convergences and additive functionals, in Dirichlet forms (Varenna, 1992), 21–53, G. Dell’Antonio and U. Mosco (Eds), Lecture Notes in Math. 1563, Springer, Berlin, 1993.

[R57]

On discontinuity and tail behaviours of the integrated density of states for nested prefractals (with T. Shima), Comm. Math. Phys. 163 (1994), 461–471.

[R58]

On a strict decomposition of additive functionals for symmetric diffusion processes, Proc. Japan Acad. Ser. A Math. Sci. 70 (1994), 277–281.

[R59]

On a decomposition of additive functionals in the strict sense for a symmetric Markov process, in Dirichlet forms and stochastic processes, Z. Ma, M. Röckner, J. Yan (Eds), Walter de Gruyter, (1995), 155–169.

[R60]

Reflecting diffusions on Lipschitz domains with cusps—analytic construction and Skorohod representation (with M. Tomisaki), in Potential theory and degenerate partial differential operators (Parma), Potential Analysis 4 (1995), 377–408.

[R61]

Construction and decomposition of reflecting diffusions on Lipschitz domains with Hölder cusps (with M. Tomisaki), Probab. Theory Related Fields 106 (1996), 521– 557.

546

Bibliography [R62]

On decomposition of additive functionals of reflecting Brownian motions (with M. Tomisaki), in Ito’s stochastic calculus and probability theory, N. Ikeda, S. Watanabe, M. Fukushima, H. Kunita (Eds), Springer, Tokyo, 1996, 51–61.

[R63]

Dirichlet forms, Caccioppoli sets and the Skorohod equation, in Stochastic Differential and Difference equations, Csiszar Michaletzky (Eds.), Birkhäuser 1997, 56–66.

[R64]

Distorted Brownian motions and BV functions, in Trends in probability and related analysis, (Taipei, 1996), 143–150, N-R.Shieh (Ed), World Sci. Publishing, River Edge, NJ, 1997.

[R65]

Large deviations and related LIL’s for Brownian motions on nested fractals (with T. Shima and M. Takeda), Osaka J. Math. 36 (1999), 497–537.

[R66]

On semi-martingale characterizations of functionals of symmetric Markov processes, Electron. J. Probab. 4 (1999), Paper 18, 1–32, www.math.washington.edu/ ~ejpecp.

[R67]

 functions and distorted Ornstein Uhlenbeck processes over the abstract Wiener space, J. Funct. Analysis. 174 (2000), 227–249.

[R68]

On limit theorems for Brownian motions on unbounded fractal sets, in Fractal geometry and stochastics, II (Greifswald/Koserow, 1998), 227–237, Progr. Probab., S. Graf (Ed) 46, Birkhäuser, Basel, 2000.

[R69]

On Ito’s formulae for additive functionals of symmetric diffusion processes, in Stochastic processes, physics and geometry: new interplays, I (Leipzig, 1999), 201– 211, F. Getsztesy (Ed), CMS Conf. Proc. 28, Amer. Math. Soc., Providence, RI, 2000.

[R70]

On the space of BV functions and a related stochastic calculus in infinite dimensions (with M. Hino), J. Funct. Analysis 183 (2001), 245–268.

[R71]

Dynkin games via Dirichlet forms and singular control of one-dimensional diffusions (with M. Taksar), SIAM J. Control Optim. 41 (2002), 682–699.

[R72]

On Sobolev and capacitary inequalities for contractive Besov spaces over -sets (with T. Uemura), Potential Analysis 18 (2003), 59–77.

[R73]

Capacitary bounds of measures and ultracontractivity of time changed processes (with T. Uemura), J. Math. Pures Appl. 82 (2003), 553–572.

[R74]

A note on regular Dirichlet subspaces (with J. Ying), Proc. Am. Math. Soc. 131 (2003) 1607-1610 (2003); erratum ibid. 132 (2004), 1559.

[R75]

On spectral synthesis for contractive -norms and Besov spaces (with T. Uemura), Potential Analysis 20 (2004), 195–206.

[R76]

Function spaces and symmetric Markov processes, in Stochastic analysis and related topics in Kyoto, Advanced Studies in Pure Mathematics 41 (2004), 75–89, Math. Soc. Japan.

[R77]

Time changes of symmetric diffusions and Feller measures (with P. He and J. Ying), Ann. Probab. 32 (2004) 3138–3166.

547

Bibliography [R78]

On regular Dirichlet subspaces of  1   and associated linear diffusions (with X. Fang and J. Ying), Osaka J. Math. 42 (2005), 27–41.

[R79]

Poisson point processes attached to symmetric diffusions (with H. Tanaka), Ann. Inst. Henri Poincaré Probab. Stat. 41 (2005), 419–459.

[R80]

Traces of symmetric Markov processes and their characterizations (with Z.-Q. Chen and J. Ying), Ann. Probab. 34 (2006), 1052–1102.

[R81]

Entrance law, exit system and Lévy system of time changed processes (with Z.-Q. Chen and J. Ying), Illinois J. Math. 50 (2006), 269–312.

[R82]

Extending Markov processes in weak duality by Poisson point processes of excursions, (with Z.-Q. Chen and J. Ying), in The Abel Symposium 2005, Stochastic Analysis and Applications, A Symposium in Honor of Kiyosi Itô (Oslo), F.E. Benth, G.Di Nunno, T. Lindstrom, B. Oksendal, T. Zhang (Eds), 153–196, Springer, 2007.

[R83]

On Feller’s boundary problem for Markov processes in weak duality (with Z.-Q. Chen), J. Funct. Anal. 252 (2007), 710–733.

[R84]

One-point extensions of Markov processes by darning (with Z.-Q. Chen), Probab. Th. Rel. Fields 141 (2008), 61–112.

[R85]

Flux and lateral conditions for symmetric Markov processes (with Z.-Q. Chen), Potential Anal. 29 (2008), 241–269.

[R86]

On unique extension of time changed reflecting Brownian motions (with Z.-Q. Chen), Ann. Inst. Henri Poincare, Probab. Statist. 45 (2009), 864–875.

[R87]

On extended Dirichlet spaces and the space of BL functions, Potential Theory and Stochastics in Albac, Aurel Cornea Memorial Volume, Eds. D. Bakry, L.Beznea, N. Boboc, G. Bucur, M. Röckner, Theta Foundation, Bucharest, AMS distribution, 2009, 101–110.

Expository Writing [E1]

Boundary problems of Brownian motions and the Dirichlet spaces, S¯ugaku 20 (1968), 211–221 (in Japanese).

[E2]

On the Theory of Markov Processes, Butsuri (The Physical Society of Japan) 25 (1970), 37–41 (in Japanese).

[E3]

On Random Spectra, Butsuri (The Physical Society of Japan) 34 (1979), 153–159 (in Japanese).

[E4]

On Spectral Analysis on Fractal, Manufacturing and Technology 48 no. 1 (1996), 55–59 (in Japanese).

[E5]

Fractal and Random walk – Toward the function and the shape of nature, in: Shizennoshikumi to Ningennochie, Osaka University Press, 1997, 211–222 (in Japanese).

548

Bibliography [E6]

Decompositions of symmetric diffusion processes and related topics in analysis, S¯ugaku Expositions 14 (2001), 1–13, AMS.

[E7]

Stochastic Control Problem and Dynkin’s Game Theory, Gien (Organization for Research and Development of Innovative Science and Technology (ORDIST)), Kansai University 111 (2001), 35–39 (in Japanese).

[E8]

Refined solutions of optimal stopping games for symmetric Markov processes, Technology Reports of Kansai University (with K. Menda), 48 (2006), 101–110.

[E9]

On the works of Kiyosi Itô and stochastic analysis, Japanese J. Math. 2 (2007), 45– 53.

Seminar on Probability (in Japanese) [S1]

(with Kiyosi Itô and Shinzo Watanabe) On Diffusion processes, Seminar on Probability 3, 1960.

[S2]

(with Ken-ichi Sato and Masao Nagasawa) Transformation of Markov processes and boundary problems, Seminar on Probability 16, 1960.

[S3]

Dirichlet space and its representations, Seminar on Probability 31, 1969.

[S4]

(with Hiroshi Kunita) Studies on Markov processes, Seminar on Probability 40, 1969.

[S5]

(with Shintaro Nakao and Sin-ichi Kotani) On Random spectra, Seminar on Probability 45, 1977.

Monographs and Textbooks [MT1] Dirichlet Forms and Markov Processes (Japanese), Kinokuniya Co. Ltd., 1975. [MT2] (with Kazunari Ishii) Natural Phenomena and Stochastic Processes (Japanese), Nippon-Hyoron-Sha Co., Ltd., 1980, (enlarged ed.) 1996. [MT3] Dirichlet Forms and Markov Processes, North-Holland Mathematical Library 23, North-Holland, Amsterdam-New York/ Kodansha, Tokyo, 1980. ¯ [MT4] (with Y¯oichi Oshima and Masayoshi Takeda) Dirichlet Forms and Symmetric Markov Processes, de Gruyter Studies in Mathematics 19, de Gruyter, Berlin, 1994. [MT5] Probability Theory (Japanese), Shokabo Pub. Co., Ltd., 1998. [MT6] (with Masayoshi Takeda) Markov Processes (Japanese), Baifukan Pub. Co., Ltd, 2008.

549