1: %%%%%%%%%%%%%%%%%%%%%

2: \def\autori{P.\ DALL'AGLIO}

3: \def\titolo{Stability results for solutions of obstacle problems with

4: measure data} 

5: 

6: 

7: \font\sixrm=cmr6

8: \newcount\tagno \tagno=0                        % equation #'s

9: \newcount\thmno \thmno=0                        % theorem numbers

10: \newcount\bibno \bibno=0                        % bibliography numbers

11: \newcount\chapno\chapno=0                       % chapters

12: \newcount\verno            % number of the version

13: \newif\ifproofmode

14: \proofmodetrue

15: \newif\ifwanted

16: \wantedfalse

17: \newif\ifindexed

18: \indexedfalse

19: % AUXILIARY MACROS

20: 

21: 

22: \def\ifundefined#1{\expandafter\ifx\csname+#1\endcsname\relax}

23: \def\Wanted#1{\ifundefined{#1} \wantedtrue 

24: \immediate\write0{Wanted #1  \the\chapno.\the\thmno}\fi}

25: \def\Increase#1{{\global\advance#1 by 1}}

26: 

27: %for our numbering

28: \def\Assign#1#2{\immediate

29: \write1{\noexpand\expandafter\noexpand\def

30:  \noexpand\csname+#1\endcsname{#2}}\relax

31:  \global\expandafter\edef\csname+#1\endcsname{#2}}

32: 

33: %for pages

34: \def\pAssign#1#2{\write1{\noexpand\expandafter\noexpand\def

35:  \noexpand\csname+#1\endcsname{#2}}}

36: \def\lPut#1{\ifproofmode\llap{\hbox{\sixrm #1\ \ \ }}\fi}

37: \def\rPut#1{\ifproofmode$^{\hbox{\sixrm #1}}$\fi}

38: 

39: 

40: % USER'S MACROS

41: % Number of a new chapter labelled as #1:

42: \def\chp#1{\global\tagno=0\global\thmno=0\Increase\chapno

43: \Assign{#1}

44: {\the\chapno}{\lPut{#1}\the\chapno}}

45: 

46: % Number of a  new theorem (remark, definition, example,...)

47: % labelled as #1:  

48: \def\thm#1{\Increase\thmno 

49: \Assign{#1}

50: {\the\chapno.\the\thmno}\the\chapno.\the\thmno\rPut{#1}}

51: 

52: % Number of a new formula labelled as #1:

53: \def\frm#1{\Increase\tagno

54: \Assign{#1}{\the\chapno.\the\tagno}\lPut{#1}

55: {\the\chapno.\the\tagno}}

56: 

57: % Number of a new bibliographical item labelled as #1:

58: \def\bib#1{\Increase\bibno

59: \Assign{#1}{\the\bibno}\lPut{#1}{\the\bibno}}

60: 

61: % Remembers the page number of the place labelled as #1

62: \def\pgp#1{\pAssign{#1/}{\the\pageno}}

63: 

64: % Creates the index item #3 labelled as #2: it is \ix2 for 

65: % index and \ix3 for  list of symbols.

66: \def\ix#1#2#3{\pAssign{#2}{\the\pageno}

67: \immediate\write#1{\noexpand\idxitem{#3}

68: {\noexpand\csname+#2\endcsname}}}

69: 

70: % Number of the item labelled as #1, for references:

71: \def\rf#1{\Wanted{#1}\csname+#1\endcsname\relax\rPut {#1}}

72: 

73: % Page number of the place labelled as #1, for references:

74: \def\rfp#1{\Wanted{#1}\csname+#1/\endcsname\relax\rPut{#1}}

75: 

76: \input \jobname.auxi

77: \Increase\verno

78: \immediate\openout1=\jobname.auxi

79: 

80: \immediate\write1{\noexpand\verno=\the\verno}

81: 

82: \ifindexed

83: \immediate\openout2=\jobname.idx

84: \immediate\openout3=\jobname.sym 

85: \fi

86: 

87: %To be set in front of the end of the job: \prebye\enddocument

88: \def\prebye{\ifwanted

89: \errmessage{Warning: Undefined references! Rerunning could help}\fi}

90: 

91: 

92: 

93: %FONTS

94: \font\twelverm=cmr12

95: \font\twelvei=cmmi12

96: \font\twelvesy=cmsy10

97: \font\twelvebf=cmbx12

98: \font\twelvett=cmtt12

99: \font\twelveit=cmti12

100: \font\twelvesl=cmsl12

101: 

102: \font\ninerm=cmr9

103: \font\ninei=cmmi9

104: \font\ninesy=cmsy9

105: \font\ninebf=cmbx9

106: \font\ninett=cmtt9

107: \font\nineit=cmti9

108: \font\ninesl=cmsl9

109: 

110: \font\eightrm=cmr8

111: \font\eighti=cmmi8

112: \font\eightsy=cmsy8

113: \font\eightbf=cmbx8

114: \font\eighttt=cmtt8

115: \font\eightit=cmti8

116: \font\eightsl=cmsl8

117: 

118: \font\sixrm=cmr6

119: \font\sixi=cmmi6

120: \font\sixsy=cmsy6

121: \font\sixbf=cmbx6

122: 

123: \catcode`@=11 % we will access private macros of plain TeX (carefully)

124: \newskip\ttglue

125: 

126: %MACRO TWELVEPOINT

127: \def\twelvepoint{\def\rm{\fam0\twelverm}% switch to 12-point type

128: \textfont0=\twelverm  \scriptfont0=\ninerm  

129: \scriptscriptfont0=\sevenrm

130: \textfont1=\twelvei  \scriptfont1=\ninei  \scriptscriptfont1=\seveni

131: \textfont2=\twelvesy  \scriptfont2=\ninesy  

132: \scriptscriptfont2=\sevensy

133: \textfont3=\tenex  \scriptfont3=\tenex  \scriptscriptfont3=\tenex

134: \textfont\itfam=\twelveit  \def\it{\fam\itfam\twelveit}%

135: \textfont\slfam=\twelvesl  \def\sl{\fam\slfam\twelvesl}%

136: \textfont\ttfam=\twelvett  \def\tt{\fam\ttfam\twelvett}%

137: \textfont\bffam=\twelvebf  \scriptfont\bffam=\ninebf

138: \scriptscriptfont\bffam=\sevenbf  \def\bf{\fam\bffam\twelvebf}%

139: \tt  \ttglue=.5em plus.25em minus.15em

140: \normalbaselineskip=15pt

141: \setbox\strutbox=\hbox{\vrule height10pt depth5pt width0pt}%

142: \let\sc=\tenrm  \let\big=\twelvebig  \normalbaselines\rm}

143: 

144: %MACRO TENPOINT

145: \def\tenpoint{\def\rm{\fam0\tenrm}% switch to 10-point type

146: \textfont0=\tenrm  \scriptfont0=\sevenrm  \scriptscriptfont0=\fiverm

147: \textfont1=\teni  \scriptfont1=\seveni  \scriptscriptfont1=\fivei

148: \textfont2=\tensy  \scriptfont2=\sevensy  \scriptscriptfont2=\fivesy

149: \textfont3=\tenex  \scriptfont3=\tenex  \scriptscriptfont3=\tenex

150: \textfont\itfam=\tenit  \def\it{\fam\itfam\tenit}%

151: \textfont\slfam=\tensl  \def\sl{\fam\slfam\tensl}%

152: \textfont\ttfam=\tentt  \def\tt{\fam\ttfam\tentt}%

153: \textfont\bffam=\tenbf  \scriptfont\bffam=\sevenbf

154: \scriptscriptfont\bffam=\fivebf  \def\bf{\fam\bffam\tenbf}%

155: \tt  \ttglue=.5em plus.25em minus.15em

156: \normalbaselineskip=12pt

157: \setbox\strutbox=\hbox{\vrule height8.5pt depth3.5pt width0pt}%

158: \let\sc=\eightrm  \let\big=\tenbig  \normalbaselines\rm}

159: 

160: %MACRO NINEPOINT

161: \def\ninepoint{\def\rm{\fam0\ninerm}% switch to 9-point type

162: \textfont0=\ninerm  \scriptfont0=\sixrm  \scriptscriptfont0=\fiverm

163: \textfont1=\ninei  \scriptfont1=\sixi  \scriptscriptfont1=\fivei

164: \textfont2=\ninesy  \scriptfont2=\sixsy  \scriptscriptfont2=\fivesy

165: \textfont3=\tenex  \scriptfont3=\tenex  \scriptscriptfont3=\tenex

166: \textfont\itfam=\nineit  \def\it{\fam\itfam\nineit}%

167: \textfont\slfam=\ninesl  \def\sl{\fam\slfam\ninesl}%

168: \textfont\ttfam=\ninett  \def\tt{\fam\ttfam\ninett}%

169: \textfont\bffam=\ninebf  \scriptfont\bffam=\sixbf

170: \scriptscriptfont\bffam=\fivebf  \def\bf{\fam\bffam\ninebf}%

171: \tt  \ttglue=.5em plus.25em minus.15em

172: \normalbaselineskip=11pt

173: \setbox\strutbox=\hbox{\vrule height8pt depth3pt width0pt}%

174: \let\sc=\sevenrm  \let\big=\ninebig  \normalbaselines\rm}

175: 

176: %MACRO EIGHTPOINT

177: \def\eightpoint{\def\rm{\fam0\eightrm}% switch to 8-point type

178: \textfont0=\eightrm  \scriptfont0=\sixrm  \scriptscriptfont0=\fiverm

179: \textfont1=\eighti  \scriptfont1=\sixi  \scriptscriptfont1=\fivei

180: \textfont2=\eightsy  \scriptfont2=\sixsy  \scriptscriptfont2=\fivesy

181: \textfont3=\tenex  \scriptfont3=\tenex  \scriptscriptfont3=\tenex

182: \textfont\itfam=\eightit  \def\it{\fam\itfam\eightit}%

183: \textfont\slfam=\eightsl  \def\sl{\fam\slfam\eightsl}%

184: \textfont\ttfam=\eighttt  \def\tt{\fam\ttfam\eighttt}%

185: \textfont\bffam=\eightbf  \scriptfont\bffam=\sixbf

186: \scriptscriptfont\bffam=\fivebf  \def\bf{\fam\bffam\eightbf}%

187: \tt  \ttglue=.5em plus.25em minus.15em

188: \normalbaselineskip=9pt

189: \setbox\strutbox=\hbox{\vrule height7pt depth2pt width0pt}%

190: \let\sc=\sixrm  \let\big=\eightbig  \normalbaselines\rm}

191: 

192: %MACRO BIG

193: \def\twelvebig#1{{\hbox{$\textfont0=\twelverm\textfont2=\twelvesy

194: 	\left#1\vbox to10pt{}\right.\n@space$}}}

195: \def\tenbig#1{{\hbox{$\left#1\vbox to8.5pt{}\right.\n@space$}}}

196: \def\ninebig#1{{\hbox{$\textfont0=\tenrm\textfont2=\tensy

197: 	\left#1\vbox to7.25pt{}\right.\n@space$}}}

198: \def\eightbig#1{{\hbox{$\textfont0=\ninerm\textfont2=\ninesy

199: 	\left#1\vbox to6.5pt{}\right.\n@space$}}}

200: \def\tenmath{\tenpoint\fam-1 } 

201: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

202: %%%%%%%%%%%%%

203: %%  \displayliness  acts like \displaylines, but one can put      

204: %%  \eqno  or  \leqno after it                            

205: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

206: %%%%%%%%%%%%%

207: \def\displayliness#1{\null\,\vcenter{\openup1\jot \m@th

208:   \ialign{\strut\hfil$\displaystyle{##}$\hfil

209:     \crcr#1\crcr}}\,}

210: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

211: %%%%%%%%%%%%%

212: %%  \displaylinesno  puts numbers to the right of each line,

213: %%              like \eqalignno

214: %%  SYNTAX    \displaylinesno{<formula>&<number>\cr etc. \cr}              

215: 	       

216: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

217: %%%%%%%%%%%%%

218: \def\displaylinesno#1{\displ@y \tabskip=\centering

219:    \halign to\displaywidth{ \hfil$\@lign \displaystyle{##}$ \hfil

220: 	\tabskip=\centering

221:      &\llap{$\@lign##$}\tabskip=0pt \crcr#1\crcr}}

222: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

223: %%%%%%%%%%%%%

224: %%  \ldisplaylinesno  puts numbers to the left of each line,

225: %%              like \leqalignno

226: %%  SYNTAX    \ldisplaylinesno{<formula>&<number>\cr etc. \cr}             

227: 		     

228: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

229: %%%%%%%%%%%%%

230: \def\ldisplaylinesno#1{\displ@y \tabskip=\centering

231:    \halign to\displaywidth{ \hfil$\@lign \displaystyle{##}$\hfil

232: 	\tabskip=\centering

233:      &\kern-\displaywidth

234:      \rlap{$\@lign##$}\tabskip=\displaywidth \crcr#1\crcr}}

235: 

236: 

237: 

238: \catcode`@=12 % at signs are no longer letters

239: 

240: 

241: 

242: \def\parag#1#2{\goodbreak\bigskip\bigskip\noindent

243: 		   {\bf #1.\ \ #2}

244: 		   \nobreak\bigskip} 

245: \def\intro#1{\goodbreak\bigskip\bigskip\goodbreak\noindent

246: 		   {\bf #1}\nobreak\bigskip\nobreak}

247: \long\def\th#1#2{\goodbreak\medskip\noindent

248: 		{\bf Theorem #1.\ \ \it #2}\smallskip}

249: \long\def\lemma#1#2{\goodbreak\bigskip\noindent

250: 		{\bf Lemma #1.\ \ \it #2}}

251: \long\def\prop#1#2{\goodbreak\bigskip\noindent

252: 		  {\bf Proposition #1.\ \ \it #2}}

253: \long\def\cor#1#2{\goodbreak\bigskip\noindent

254: 		{\bf Corollary #1.\ \ \it #2}}

255: \long\def\defin#1#2{\goodbreak\bigskip\noindent

256: 		  {\bf Definition #1.\ \ \rm #2}}

257: \long\def\rem#1#2{\goodbreak\medskip\noindent

258: 		 {\bf Remark #1.\ \ \rm #2}}

259: \long\def\ex#1#2{\goodbreak\medskip\noindent

260: 		 {\bf Example #1.\ \ \rm #2}}

261: \def\cond#1#2{\goodbreak\bigskip\noindent

262: 		{\bf Condition #1.\ \ \it #2}\goodbreak\bigskip}

263: \def\negbigskip{\vskip-\bigskipamount}

264: 

265: \def\proof{\vskip.4cm\noindent{\it Proof.\ \ }}

266: \def\proofof#1{

267: 		\vskip.4cm\noindent{\it Proof of #1.\ \ }}

268: 

269: \def\sqr#1#2{\vbox{

270:    \hrule height .#2pt 

271:    \hbox{\vrule width .#2pt height #1pt \kern #1pt 

272:       \vrule width .#2pt}

273:    \hrule height .#2pt }}

274: \def\square{\sqr74}

275: 

276: \def\endproof{{\unskip\nobreak\hfill \penalty50

277: \hskip1em\hbox{}\nobreak\hfill $\square$ \goodbreak

278: \parfillskip=0pt  \finalhyphendemerits=0}}

279: 

280: 

281: \mathchardef\emptyset="001F

282: \mathchardef\hyphen="002D

283: 

284: 

285: 

286: \def\LM#1{\hbox{\vrule width.2pt \vbox to#1pt{\vfill \hrule 

287: width#1pt

288: height.2pt}}}

289: \def\LL{{\mathchoice {\>\LM7\>}{\>\LM7\>}{\,\LM5\,}{\,\LM{3.35}\,}}}

290: 

291: 

292: %Typographic symbols

293: \def\interior{\mathaccent'27}

294: \def\wto{\rightharpoonup}

295: 

296: 

297: 

298: 

299: 

300: %media integrale

301: %\def\ave{-\hskip-.4cm\int}

302: 

303: %HEADLINE FOR TEXT AND FRONT PAGE

304: \def\rightheadline{\eightpoint\hfil\titolo

305: \hfil\tenrm\folio} 

306: \def\leftheadline{\tenrm\folio\hfil\eightpoint

307: \autori \hfil}

308: \def\zeroheadline{\hfill} 

309: %

310: %\headline={\ifodd\pageno\rightheadline \else\leftheadline\fi}

311: \headline={\ifnum\pageno=0 \zeroheadline

312: \else\ifodd\pageno\rightheadline

313: \else\leftheadline\fi\fi}

314: 

315: 

316: 

317: %PARAMETERS FOR EVERY PAGE

318: \nopagenumbers

319: \magnification=1200

320: \baselineskip=14pt

321: \hfuzz=2pt

322: \parindent=2em

323: \mathsurround=1pt

324: \tolerance=1000

325: 

326: 

327: %FRONT PAGE (PER LA COPERTINA SISSA; non modificare)

328: 

329: \pageno=0

330: \hsize 14truecm

331: \vsize 25truecm

332: \hoffset=0.8truecm

333: \voffset=-1.55truecm

334: 

335: \null

336: \vskip 2.8truecm

337: {\twelvepoint

338: \baselineskip=1.7\baselineskip

339: \centerline{\bf STABILITY RESULTS FOR SOLUTIONS OF}

340: \centerline{\bf OBSTACLE PROBLEMS WITH MEASURE DATA}

341: }

342: \vskip2truecm

343: \centerline{Paolo DALL'AGLIO}

344: \vfil

345: 

346: {\eightpoint

347: \baselineskip=1.2\baselineskip

348: \centerline{\bf Abstract}

349: \bigskip

350: \noindent 

351: In this paper we study the continuous dependence with respect to

352: obstacles for obstacle problems with measure data. This is deeply

353: investigated introducing a suitable type of convergence, which gives

354: stability under very general hypotheses. Moreover stability with

355: respect to $\rm H^1$ and uniform convergent obstacles is proved.

356: \par

357: }

358: 

359: \vfil

360: \vskip 1truecm 

361: \centerline {Ref. S.I.S.S.A. 

362: 142/99/M (December 1999)}

363: \vskip 1truecm 

364: \eject

365: 

366: %\pageno=0

367: %{\phantom{a}}

368: %\vfil

369: %\eject

370: 

371: %PARAMETERS FOR THE TEXT

372: \topskip=15pt %default=10pt

373: \vsize 22.5truecm

374: \hsize 17truecm

375: \hoffset=0truecm

376: \voffset=0.5truecm

377: 

378: \overfullrule=0pt

379: \proofmodefalse % non stampa le etichette della numerazione

380: 		% deve essere messo dopo il gruppo di macro

381: 		% per la numerazione automatica

382: 

383: %definizioni

384: \def\R{{\rm I\! R}}

385: \def\Rn{{\rm I\! R}^N}

386: \def\Luno{{\rm L}^1(\Omega)}

387: \def\Linf{{\rm L}^\infty(\Omega)}

388: \def\Hunozero{{\rm H}^1_0(\Omega)}   

389: \def\Huno{{\rm H}^1(\Omega)}

390: \def\Hduale{{\rm H}^{\hbox{\kern 1truept \rm -}\kern -1truept1}\kern

391: 	-1truept(\Omega)}

392: \def\Wunoqzero{{\rm W}^{1,q}_0(\Omega)}

393: \def\Wunoq{{\rm W}^{1,q}(\Omega)} 

394: \def\Mb{{\cal M}_b(\Omega)}

395: \def\Mbp{{\cal M}^+_b(\Omega)}

396: \def\Mbo{{\cal M}^0_b(\Omega)}

397: \def\Mbop{{\cal M}^{0,+}_b(\Omega)}

398: \def\Inters{\Mb\cap\Hduale}

399: \def\intl{\int\limits}

400: \def\dualita#1#2{\langle#1,#2\rangle}

401: \def\Fpsimu{{\cal F}_\psi(\mu)}  

402: \def\Fpsimuj{{\cal F}^{\rm J}_\psi(\mu)}

403: \def\Fpsinmuj{{\cal F}^{\rm J}_{\psi_n}(\mu)}

404: \def\Fgpsimu{{\cal F}^g_\psi(\mu)}  

405: \def\Ggpsimu{{\cal G}^g_\psi(\mu)}  

406: \def\wto{\rightharpoonup}

407: \def\d{\,d\!}

408: \def\cc{\subset\subset}

409: \def\um{u_{\mu}}

410: \def\un{u_{\nu}}

411: \def\ul{u_{\lambda}}

412: \def\uln{u_{\lambda_n}}

413: \def\ugo{u^g_0}

414: \def\uho{u^h_0}

415: \def\weakq{\hbox{\ weakly in }\Wunoqzero}

416: \def\weak2{\hbox{\ weakly in }\Hunozero}

417: \def\strongq{\hbox{\ strongly in }\Wunoq}

418: \def\strong2{\hbox{\ strongly in }\Huno}

419: \def\weaks{\hbox{\ weakly-$\ast$ in }\Mb}

420: \def\mosco{\buildrel{\rm M}\over{-\!\!\!-\!\!\!\longrightarrow}}

421: \def\lev{\buildrel{\rm lev}\over{-\!\!\!-\!\!\!\longrightarrow}}

422: \def\ae{\hbox{\ a.e.\ in }\Omega}

423: \def\qe{\hbox{\ q.e.\ in }\Omega}

424: \def\Cqe{\hbox{ cap-q.e.\ in }\Omega}

425: \def\vuoto{{\scriptstyle\bigcirc\!\!\!}\!\lower.15ex\hbox{/}}

426: \def\f{\varphi}

427: \def\e{\varepsilon}

428: \def\fe{\varphi_\varepsilon}

429: \def\res{\mathop{\hbox{\vrule height 7pt width .5pt depth 0pt

430: 	\vrule height .5pt width 6pt depth 0pt}}\nolimits}

431: \def\ave{-\hskip-.58cm\intl}

432: \def\Kpsi{{\rm K}_{\psi}(\Omega)}

433: \def\Kpsin{{\rm K}_{\psi_n}(\Omega)}

434: \def\KpsiH{\Kpsi\cap\Hunozero}

435: \def\KpsinH{\Kpsin\cap\Hunozero}

436: \def\inters{\Kpsi\cap \Wunoqzero}

437: \def\crr{\cr\noalign{\smallskip}}% in "cases" per avere piu' spazio tra

438: 				 % riga e l'altra

439: 				 % non mettere all'ultimo \cr

440: 

441: \parag{\chp{intro}}{Introduction}

442: 

443: Given a regular bounded open set $\Omega$ of ${\R}^N$, $N\ge1$, and a

444: linear elliptic operator ${\cal A}$ of the form

445: $$

446: {\cal A}u =-\sum_{j,j=1}^N D_i ( a_{ij} D_j u ),

447: $$

448: with $a_{ij}\in L^\infty(\Omega)$, we study obstacle problems for the

449: operator ${\cal A}$ in

450: $\Omega$ with

451: homogeneous Dirichlet boundary conditions on $\partial\Omega$, when the

452: datum

453: $\mu$ is a bounded Radon measure on $\Omega$ and the obstacle

454: $\psi$ is an arbitrary function on $\Omega$.

455: 

456: According to~[\rf{DAL-LEO}], a function $u$ is a solution of this problem,

457: which will be denoted by $OP(\mu,\psi)$, if $u$ is the smallest

458: function with the following properties: $u\geq\psi$ in $\Omega$ and

459: $u$ is a solution in the sense of Stampacchia~[\rf{STA}] of a problem

460: of the form

461: $$

462: \cases{{\cal A}u=\mu+\lambda\quad&in $\Omega$\crr

463: 	u=0&on $\partial\Omega$,\cr}\eqno(\frm{stampintro})

464: $$

465: for some bounded Radon measure $\lambda\geq 0$. The measure $\lambda$

466: which corresponds to the solution of the obstacle problem is called

467: the obstacle reaction.

468: 

469: Existence and uniqueness of the solution of $OP(\mu,\psi)$ have been

470: proved in~[\rf{DAL-LEO}], provided that there exists a measure

471: $\lambda$ such that the solution of~(\rf{stampintro}) is greater than

472: or equal to $\psi$. These results have been extended to the non-linear

473: case in~[\rf{LEO}], when $\mu$ vanishes on all sets with capacity

474: zero. For a different approach to obstacle problems for non-linear

475: operators with measure data see~[\rf{BOC-GAL}], [\rf{BOC-CIR-1}], 

476: [\rf{BOC-CIR-2}], [\rf{OPP-ROS-1}] and~[\rf{OPP-ROS-2}].

477: 

478: If the measure $\mu$ belongs to the dual $\Hduale$ of the Sobolev

479: space $\Hunozero$, and if there exists a function $w\in\Hunozero$

480: above the obstacle $\psi$, then the solution of the obstacle problem 

481: $OP(\mu,\psi)$ according to the previous definition coincides with the

482: solution $u$ of the variational inequality

483: $$

484: \cases{u\in\Hunozero,\ u\geq\psi,\crr

485: 	\dualita{{\cal A}u}{v-u}\geq\dualita{\mu}{v-u},\crr

486: 	\forall v\in\Hunozero\hbox{ s.t. }v\geq\psi\cr}

487: $$

488: where $\langle\cdot,\cdot\rangle$ denotes the duality pairing between

489: $\Hduale$ and $\Hunozero$. In this case the obstacle reaction

490: $\lambda$ belongs to $\Hduale$. We also know it is concentrated on the contact set

491: $\{u=\psi\}$ if $\psi$ is continuous, or, more in general, quasi upper

492: semicontinuous.

493: 

494: An important role in this problem is played by the space $\Mbo$ of all

495: bounded Radon measures on $\Omega$ which are absolutely continuous

496: with respect to the harmonic capacity. If the datum $\mu$ belongs to

497: $\Mbo$ (it is actually enough that its negative part $\mu^-$ is such), 

498: so does the obstacle reaction, provided that there exists  a

499: measure $\lambda\in\Mbo$ such that the solution of~(\rf{stampintro})

500: is greater than or equal to $\psi$ (see~[\rf{DAL-LEO}], theorem~7.5). In

501: this case the obstacle reaction is concentrated on the contact set

502: $\{u=\psi\}$, whenever the obstacle $\psi$ is quasi upper

503: semicontinuous (see~[\rf{LEO}], theorem~2.9). 

504: 

505: It can be seen that in general this does not occur when

506: $\mu^-\not\in\Mbo$. This case was studied

507: in~[\rf{DAL-DAL}]. Remembering that $\mu^-$ can be decomposed as

508: $\mu^-=\mu^-_a+\mu^-_s$, where $\mu^-_a \in{\cal M}_b^0(\Omega)$ and

509: $\mu^-_s$ is concentrated on a set of capacity zero, it was proved

510: that under some natural assumptions on the obstacle, the singular part

511: $\mu^-_s$ can be neglected and the obstacle problems $OP(\mu,\psi)$

512: and $OP(\mu^+-\mu^-_a,\psi)$ have the same solutions.

513: 

514: The topic of continuous dependence with respect to data was already

515: treated in~[\rf{DAL-LEO}].

516: 

517: As for stability with respect to the right hand side, it was proved

518: that, if $\mu_n,\,\mu\in\Mb$ are such that $\mu_n\to\mu$ strongly in

519: $\Mb$, then $u_n\to u$ strongly in $\Wunoq$, where $u_n$ and $u$ are

520: the solutions of $OP(\mu_n,\psi)$ and $OP(\mu,\psi)$ respectively.

521: Trying to use weak-$*$ convergence, it was seen that in general

522: $\mu_n\wto\mu$ weakly-$*$ does not imply that $u_n\to u$, even with

523: the obstacle $\psi\equiv 0$, but we know only that for any measure

524: $\mu\in\Mb$, there exists a special sequence $\mu_k\wto\mu$ weakly-$*$

525: in $\Mb$, with $\mu_k\in\Inters$, such that $u_k\to u$ strongly in

526: $\Wunoq$.

527: 

528: In this paper we consider stability with respect to obstacles. To

529: study this question we introduce a kind of convergence of functions,

530: the level set convergence, which yields the convergence of solutions

531: under very mild assumptions.

532: 

533: The convergence of $\psi_n$ to $\psi$ in the sense of level sets, defined

534: precisely in definition~\rf{livelli}, is verified in particular when

535: $$

536: {\rm cap}(\{\psi>t\}\cap B)=\lim_{n\to+\infty}{\rm cap}(\{\psi_n>t\}\cap

537: B)

538: $$

539: for  all $t\in\R$ and for all $B\cc \Omega$ (see also

540: remark~\rf{semplice}).

541: 

542: We will see that without further hypothesis it can only be proved that,

543: calling $u_n$ and $u$ the solutions of $OP(\mu,\psi_n)$ and of

544: $OP(\mu,\psi)$ respectively, if $\psi_n\lev\psi$ then, up to a

545: subsequence, $u_n$ converges to some function $u^*$ which is always

546: greater than or equal to $u$ (proposition~\rf{ustar}).

547: 

548: Then we will obtain, from the level set convergence of the obstacles,

549: that $u_n$ converges to $u$, under some conditions: in particular, by

550: means of the Mosco convergence of convex sets, we obtain that

551: 

552: \item{(i)} if $\mu^-\in\Hduale$ then $u_n\to u\strong2$;

553: 

554: \item{(ii)} if $\mu^-\in\Mbo$ then $u_n\to u\strongq$;

555: 

556: \item{(iii)} if $\psi$ is suitably controlled below, then $u_n\to u\strongq$.

557: 

558: In section~\rf{teoremini} we consider the case $\psi_n\leq\psi$ and

559: show that we have the convergence of solutions for any datum $\mu\in\Mb$.

560: 

561: We conclude this study considering two cases in which the assumptions

562: that the obstacles converge in a stronger way allows to obtain a

563: stronger convergence also for the solutions. When the difference

564: $\psi_n-\psi$ belongs to $\Hunozero$ and tends to zero strongly in

565: this space, then we obtain the same type of convergence for the

566: solutions, for any $\mu\in\Mb$.

567: 

568: In section~\rf{elleinfinito}, we extend the theory so

569: far developed to the case of nonzero boundary values. For any function

570: $g\in\Huno$, we can define the function $u$ to be the solution of

571: $OP(\mu,g,\psi)$ if and only if $u-\ugo$ is the solution of

572: $OP(\mu,\psi)$,

573: where $\ugo$ is the solution of

574: $$

575: \cases{	{\cal A}\ugo=0\quad&in $\Hduale$\crr

576: 	\ugo-g\in\Hunozero.&\cr}

577: $$

578: All the results developed in the case of homogeneous boundary

579: conditions can be extended, thanks to the linearity of ${\cal A}$.

580: 

581: Using this extension we prove a new characterization: the solution of

582: $OP(\mu,g,\psi)$ is the minimum element among all the supersolutions of

583: ${\cal A}-\mu$ which are above the obstacle and greater than or equal to

584: $g$ on the boundary $\partial\Omega$. 

585: From this we easily prove that if the obstacles 

586: converge uniformly then so do the solutions of the corresponding

587: obstacle problems for any $\mu\in\Mb$.

588: 

589: 

590: 

591: \parag{\chp{fattinoti}}{Notations and basic results.}

592: 				 

593: Let $\Omega$ be an open bounded subset of $\Rn$, $N\geq1$, with

594: Lipschitz boundary.

595: 

596: Let ${\cal A}(u)=-{\rm div}(A(x)\nabla u)$ be a linear elliptic operator

597: with coefficients in $L^\infty(\Omega)$, that is $A(x)=(a_{ij}(x))$ is

598: an $N\times N$ matrix such that

599: $$

600: a_{ij}\in L^\infty(\Omega)\hbox{\ \ and \ }\sum

601: a_{ij}(x)\xi_i\xi_j\geq\gamma

602: |\xi|^2,\quad\forall\xi\in\R^N, \ae,

603: $$

604: with $\gamma>0$.

605: 

606: We want to consider the obstacle problem also in the case of thin

607: obstacles, so we will need the techniques of capacity theory. For this

608: theory we refer, for instance, to [\rf{HEI-KIL}].

609: 

610: We recall very briefly that, given a set $E\subseteq\Omega$ its

611: capacity with respect to $\Omega$ is given~by

612: $$

613: {\rm cap}(E)=\inf\{\|z\|^2_{\Huno}\,:\,z\in\Hunozero, z\geq 1\hbox{ a.e.\ in a

614: neighbourhood of }E\}.

615: $$

616: 

617: A property holds quasi everywhere (abbreviated as q.e.) when it holds up

618: to sets of capacity zero. 

619: 

620: A set $A$ is said to be quasi open (resp.\ quasi closed) if for any $\e>0$

621: there exists an open (resp.\ closed) set $V$ such that cap$(A\triangle

622: V)<\e$.

623: 

624: A function $v:\Omega\to{\overline{\R}}$ is quasi continuous (resp.

625: quasi upper semicontinuous) if, for every $\e>0$ there exists a set   

626: $E$ such that ${\rm cap}(E)<\e$ and $v|_{\Omega\setminus E}$ is continuous

627: (resp. upper semicontinuous) in

628: $\Omega\setminus E$.

629: 

630: We recall also that if $u$ and $v$ are quasi continuous functions and

631: $u\leq v$ a.e.\ then also $u\leq v$ q.e.\ in $\Omega$.

632: 

633: A function $u\in\Hunozero$ always has a quasi continuous

634: representative, that is there exists a quasi continuous function

635: $\tilde u$ which equals $u$ a.e.

636: 

637: Consider the function $\psi:\Omega\to{\overline{\R}}$, and let the convex

638: set be

639: $$

640: K_{\psi}(\Omega):=\{z\ \hbox{quasi continuous }  : z\geq \psi \hbox{

641: q.e.\ in }\Omega\}.

642: $$

643:   

644: Without loss of generality we may suppose that $\psi$ is quasi upper

645: semicontinuous thanks to the following proposition (it is a

646: consequence of proposition 1.5 in [\rf{DAL-4}]).

647: 

648: \prop{\thm{invilup}}{Let $\psi:\Omega\to{\overline{\R}}$.  Then 

649: there exists a quasi upper semicontinuous function $\hat

650: \psi:\Omega\to\overline{\R}$ such that:

651: 

652: \item{ 1.} $\hat \psi\geq\psi$ q.e.\ in $\Omega$;

653: 

654: \item{ 2.} if $\varphi:\Omega\to{\overline{\R}}$ is quasi upper

655: semicontinuous and $\varphi\geq\psi$ q.e.\ in $\Omega$ then

656: $\varphi\geq\hat\psi$ q.e.\ in~$\Omega$.}

657: 

658: Thus, in particular, $K_{\psi}(\Omega)=K_{\hat{\psi}}(\Omega)$.

659: 

660: 

661: In their natural setting, obstacle problems are part of

662: the theory of variational

663: inequalities (for which we refer to well known books such as

664: [\rf{KIN-STA}] and [\rf{TRO}]).

665: 

666: 

667: For any datum $F\in\Hduale$ the

668: variational inequality with obstacle $\psi$

669: $$

670: \cases{\dualita{{\cal A} u}{v-u}\geq\dualita{F}{v-u}

671: 		\quad\forall v\in\Kpsi\cap\Hunozero\crr

672: 	u\in\Kpsi\cap\Hunozero\cr}\eqno(\frm{vi})

673: $$

674: (which, for simplicity, will be indicated by

675: $

676: VI(F,\psi)

677: $),

678: has a unique solution, whenever the set $\Kpsi\cap \Hunozero$ is nonempty,

679: i.e.

680: $$

681: \exists z\in\Hunozero\ :\ z\geq\psi\quad\qe.\eqno(\frm{zeta})  

682: $$

683: In this case we will say that the obstacle is $VI$-admissible.

684: 

685: Let now $\Mb$ be the space of bounded Radon measures, with the norm

686: given by $\|\mu\|_{\Mb}=|\mu|(\Omega)$. $\Mbo$ is the subspace of

687: measures of $\Mb$ vanishing on sets of zero capacity. $\Mbp$ and

688: $\Mbop$ are the corresponding cones of non negative measures.  Recall

689: that $\Hduale\not\subseteq\Mb$ but $\Hduale\cap\Mb\subseteq\Mbo$.

690:   

691: Any measure $\mu\in\Mb$ can be decomposed as $\mu=\mu_a+\mu_s$

692: (see~[\rf{FUK}]), where $\mu_a\in\Mbo$ and $\mu_s$ is concentrated on

693: a set of capacity zero.

694: 

695: If $x\in\Omega$, we denote by $\delta_x$ the Dirac's delta centered at

696: $x$.

697: 

698: 

699: When the datum is a measure, equations and inequalities can not be studied

700: in the variational framework.

701: 

702: In [\rf{STA}] G. Stampacchia gave the following definition

703: 

704: \defin{\thm{stamp}}{}A function

705: $\um\in\Luno$ is a solution

706: in the sense of Stampacchia (also called

707: solution by duality) of the

708: equation

709: $$

710: \cases{{\cal A}\um=\mu\ \ & in $\Omega$\crr

711: 	\um=0& on $\partial\Omega$\cr}\eqno(\frm{stampacchia})

712: $$

713: if

714: $$

715: \intl_\Omega \um g\,dx\,=\,\intl_\Omega u^*_g\,d\mu,\quad\forall

716: g\in\Linf,

717: $$

718: where $u^*_g$ is the solution of

719: $$

720: \cases{{\cal A}^*u^*_g=g\quad \hbox{ in }\Hduale\crr

721: 	u^*_g\in\Hunozero\cr}

722: $$

723: and ${\cal A}^*$ is the adjoint of ${\cal A}$.

724: 

725: 

726: Throughout the paper $q$ will be any exponent satisfying

727: $1<q<{N\over{N-1}}$; Stampacchia proved that a solution

728: $\um$ exists, is unique, and belongs to $\Wunoqzero$; moreover if the

729: datum $\mu$ is more regular, namely belongs to $\Inters$, then the

730: solution coincides with the variational one. It is possible to prove

731: that, when the data converge weakly-$*$ in $\Mb$, the solutions

732: converge strongly in $\Wunoqzero$.

733:   

734: If $u$ is such a solution then $T_k(u)\in\Hunozero$, for any

735: $k\in\R^+$, where $T_k(s):=(-k)\vee(s\wedge k)$ is the usual

736: truncation function. Moreover

737: $$

738: \intl_\Omega|DT_k(u)|^2dx\leq k\,|\mu|(\Omega).\eqno(\frm{tronche})

739: $$

740: These facts imply that $u$ has a quasi continuous representative which is

741: finite$\qe$. In the rest of the paper we shall always identify $u$ with

742: its quasi continuous representative.

743: 

744: We will use the following notation: $\um$ denotes

745: the solution of the equation~(\rf{stampacchia}).

746: 

747: 

748: The following definition of solution for obstacle problems with measure

749: data was given in [\rf{DAL-LEO}].

750: 

751: \defin{\thm{disvar}}{}We say that the function

752: $u$ is a solution of the

753: obstacle problem with datum $\mu$ and obstacle $\psi$ (shortly

754: $OP(\mu,\psi)$) if

755: \item{1.} $u\in\Kpsi$ and there exists a positive bounded measure

756: $\lambda\in\Mbp$ such

757: that

758: $$

759: u=\um+\ul;

760: $$

761: \item{2.} for any $\nu\in\Mbp$, such that $v=\um+\un$ belongs to $\Kpsi$,

762: we have

763: $$

764: u\leq v\ae.

765: $$

766: 

767: For the problem to make sense let us assume that the obstacle $\psi$

768: satisfies a minimal

769: hypothesis, instead of (\rf{zeta}), namely

770: $$

771: \exists \rho\in\Mb\ :\ u_\rho\geq\psi\hbox{ q.e.\ in

772: }\Omega.\eqno(\frm{ipomin})

773: $$

774: In this case we will say that $\psi$ is $OP$-admissible.

775: 

776: 

777: \th{\thm{esist2}}{Given $\mu\in\Mb$ and $\psi$ $OP$-admissible

778: (namely $\psi\leq u_\rho$ q.e.\ in $\Omega$), the

779: obstacle problem $OP(\mu,\psi)$ has a unique solution $u=\um+\ul$.

780: The measure $\lambda\in\Mbp$ satisfies

781: $$

782: \|\lambda\|_{\Mb}\leq\left\|(\mu-\rho)^-\right\|_{\Mb}\eqno(\frm{mumenorho}).

783: $$

784: Moreover, if $\rho\in\Inters$, there exists a sequence $\mu_k={\cal

785: A}T_k(\um-u_\rho)+\rho

786: \in\Inters$ such that the solutions

787: $u_k$ of

788: $OP(\mu_k,\psi)$ converge strongly in $\Wunoq$. This sequence does not

789: depend on the obstacle but only on the measure $\rho$.}

790: 

791: Also here the positive measure $\lambda$ associated with the solution is

792: called the obstacle reaction.

793: 

794: We mention here a very simple and very useful result whose proof is

795: immediate, but it is worth stating it on its own.

796: 

797: \lemma{\thm{magico}}{Let $\mu,\,\nu\in\Mb$ and $\psi$ is $OP$-admissible. 

798: Then $u$ is the solution of $OP(\mu+\nu,\psi)$ if and only if $u-\un$ is

799: the solution of $OP(\mu,\psi-\un)$}.

800:   

801: The particular case in which the datum $\mu$ belongs to $\Mbo$ has

802: been investigated in detail. Then the obstacle reaction itself belongs

803: to $\Mbo$, provided the obstacle satisfies the following condition

804: $$

805: \exists \sigma\in\Mbo\ :\ u_\sigma\geq\psi\hbox{ q.e.\ in

806: }\Omega;\eqno(\frm{ipominzero})

807: $$

808: This will be shortened by saying that $\psi$ is $OP^0$-admissible.

809: 

810: Notice that if the datum $\mu$ is in $\Mbo$, but the obstacle is

811: only $OP$-admissible, then the reaction $\lambda$ in general does

812: not belong to $\Mbo$. This is shown in the following simple example.

813: 

814: \ex{\thm{lambda}}{Let $\mu=0$ and $\psi=u_{\delta_{x_0}}$, where

815: $\delta_{x_0}$ is the Dirac's delta centered at $x_0\in\Omega$. Then the

816: solution of $OP(0,\psi)$ is $u_{\delta_{x_0}}$ itself and hence

817: $\lambda=\delta_{x_0}\not\in\Mbo$.}

818: 

819: 

820: The interaction between obstacles and solutions is studied deeply in

821: [\rf{DAL-DAL}], where the following result was proved.

822: 

823: \th{\thm{lostesso}}{Let $\mu\in\Mb$ and let $\psi:\Omega\to\overline\R$ be

824: such that

825: $$

826: -u_{\tau}-u_{\sigma}-\f\leq\psi\leq u_{\sigma}\quad\qe\eqno(\frm{condizione})

827: $$

828: where $\f\in\Huno$, $\sigma\in\Mbo$ and $\tau\in\Mb$ such that

829: $\tau\perp\mu^-_s$.

830: Then the

831: solutions 

832: $$

833: u=\um+\ul \hbox{ of } OP(\mu,\psi)\ \ \ \hbox{ and }\ \ \ 

834: u_{\mu^+}-u_{\mu^-_a}+u_{\lambda_1}\hbox{ of }OP(\mu^+-\mu^-_a,\psi)

835: $$

836: are the same.  Moreover $\lambda=\lambda_1+

837: \mu^-_s$ with $\lambda_1\in\Mbop$.}

838: 

839: 

840: Here condition (\rf{condizione}) is given in its full generality, for

841: instance it is

842: satisfied by obstacles in $\Huno$ that are $OP$-admissible.

843: 

844: \rem{\thm{delte}}{If the obstacle $\psi$ satisfies~(\rf{condizione})

845: with $\tau=0$ then the conclusion holds for every $\mu\in\Mb$.

846: 

847: The presence of $\tau$, which deppends on $\mu$, in (\rf{condizione})

848: allows anyway to treat situations like the following one.  If ${\cal

849: A}=-\Delta$, $\Omega=B_1(0)$, the obstacle is $-u_{\delta_0}$ and the

850: datum is $-\delta_{x_0}$ for any $x_0\neq 0$, then the solution of the

851: obstacle problem is zero, because the theorem applies, and because the

852: solution must be less than or equal to zero.  }

853: 

854: In this paper we will be concerned with the continuous dependence of

855: the solutions with respect to various types of convergence of the

856: obstacles. To our knowledge the only result that was proved on this

857: problem for an arbitrary measure $\mu$ is the following (proved

858: in~[\rf{DAL-LEO}]) which deals with a very special case.

859: 

860: \prop{\thm{psienne}}{Let $\psi_n:\Omega\to\overline{\R}$ be obstacles

861: such

862: that

863: $$

864: \psi_n\leq\psi \quad\hbox{ and }\quad\psi_n\to\psi\quad\qe,

865: $$

866: $\psi$ $OP$-admissible, and let $u_n$ and $u$ be the solutions of

867: $OP(\mu,\psi_n)$ and $OP(\mu,\psi)$, respectively. Then

868: $$

869: u_n\to u\quad\hbox{ strongly in }\Wunoq.

870: $$}

871: {\vskip-1.5em}

872: 

873: This result will be generalized in section~\rf{teoremini}.

874: 

875: In order to study the problem in the most general way we introduce in section~\rf{mosco}

876: a notion of convergence that was used in~[\rf{DAL-4}]. It will be called

877: ``level set convergence'' and in most cases it is equivalent to the

878: convergence of convex sets introduced by U.  Mosco in~[\rf{MOS}], see

879: also~[\rf{ATT}]

880: 

881: Mosco proved that this type of convergence is the right one for the

882: stability of variational inequalities with respect to obstacles. This is

883: the main theorem of his theory.

884: 

885: \th{\thm{vimosco}}{Let $\psi_n$ and $\psi$ be $VI$-admissible. Then

886: $$

887: \Kpsin\cap\Hunozero\mosco\Kpsi\cap\Hunozero,

888: $$

889: if and only if, for any $f\in\Hduale$,

890: $$

891: u_n\to u\strong2

892: $$

893: where $u_n$ and $u$ are the solutions of $VI(f,\psi_n)$  and

894: $VI(f,\psi)$, respectively.}

895: 

896: 

897: \parag{\chp{mosco}}{The level set convergence, and the related stability

898: properties.}

899: 

900: In this section we will define a kind of convergence of functions

901: which will prove to be a good one for the obstacles in obstacle

902: problems with measure data: it is rather general and allows to obtain

903: the convergence of the solutions under very mild assumptions.

904: 

905: \defin{\thm{livelli}}{Let $\psi_n$ and $\psi$ be quasi upper semicontinuous

906: function from $\Omega$ to $\overline\R$. We say that $\psi_n$ tends to

907: $\psi$ in the sense of level sets and write

908: $$

909: \psi_n\lev\psi

910: $$

911: if 

912: $$

913: \quad\qquad {\rm cap}(\{\psi>t\}\cap B)\leq\liminf_{n\to+\infty}

914: {\rm cap}(\{\psi_n>s\}\cap B')\eqno(\frm{liminf})

915: $$

916: $$

917: {\hskip-1.5cm}\limsup_{n\to+\infty}{\rm cap}(\{\psi_n>t\}\cap B)\leq

918: {\rm cap}(\{\psi>s\}\cap B')\eqno(\frm{limsup})

919: $$

920: for all $s,t\in\R$, $s<t$, and for all $B\subset\subset B'\subset

921: \subset \Omega$.}

922: 

923: \rem{\thm{semplice}}{From the definition it is clear that the level set

924: convergence is implied~by 

925: $$

926: {\rm cap}(\{\psi>t\}\cap B)=\lim_{n\to+\infty}{\rm

927: cap}(\{\psi_n>t\}\cap B).

928: \eqno(\frm{limitecap})

929: $$

930: for all $t\in\R$ and for all $B\cc\Omega$}

931: \rem{\thm{cap}}{From the definition it follows that, if

932: $\psi_n$ converge to $\psi$ locally in capacity,~i.e.

933: $$

934: {\rm cap}(\{|\psi_n-\psi|>t\}\cap A)\to0,\quad\forall t\in\R^+,\;

935: \forall A\cc\Omega,

936: $$

937: then $\psi_n\lev\psi$.}

938: 

939: To prove that the limit in the sense of level set is unique we give the

940: following lemma from capacity theory which can be found in~[\rf{FUG}].

941: \lemma{\thm{capaci}}{Let $E$ and $F$ be quasi closed subsets of $\Omega$ 

942: such that

943: $$

944: {\rm cap}(E\cap A)\leq {\rm cap}(F\cap A),\quad\forall A\subseteq\Omega\hbox{ open },

945: \eqno(\frm{cap})

946: $$

947: then ${\rm cap}(E\setminus F)=0$ (we say also that $E$ is quasi contained

948: in $F$).}

949: 

950: \prop{\thm{benposto}}{Let $\psi_n$, $\psi$ and $\f$ be quasi upper

951: semicontinuous functions. If

952: $$

953: \psi_n\lev\psi\qquad\hbox{ and }\qquad\psi_n\lev\f,

954: $$

955: then $\psi=\f$.}

956: \proof 

957: Let us fix an open set $A\cc\Omega$ and two real numbers  $s<t$. Take now

958: two subsets $A'$ and $A''$ such that $A''\cc A'\cc A$ and real numbers

959: $t'$ and $t''$ such that $s<t'<t''<t$. Then

960: $$

961: \eqalign{

962: {\rm cap}(\{\psi>t''\}\cap A'')&

963: 	\leq \liminf_{n\to+\infty}{\rm cap}(\{\psi_n>t'\}\cap A')\crr

964: &\leq\limsup_{n\to+\infty}{\rm cap}(\{\psi_n>t'\}\cap A')

965: 		\leq {\rm cap}(\{\f\geq s\}\cap A).\cr}

966: $$

967: Hence, since $\{\psi\geq t\}\subseteq\{\psi>t''\}$, we have

968: $$

969: {\rm cap}(\{\psi\geq t\}\cap A'')\leq {\rm cap}(\{\f\geq s\}\cap A),

970: $$

971: from which, invading $A$ by means of $A''\cc A$,

972: $$

973: {\rm cap}(\{\psi\geq t\}\cap A)\leq {\rm cap}(\{\f\geq s\}\cap A).

974: $$

975: Using the fact that $\psi$ and $\f$ are quasi upper semicontinuous

976: and thanks to lemma~\rf{capaci} we deduce that $\{\psi\geq t\}$ is quasi

977: contained $\{\f\geq s\}$. Now, given $t$, consider two sequences

978: $t_k\searrow t$ and $s_k\searrow t$, with $t_k> s_k$, so that 

979: $$

980: \{\psi\geq t_k\}\nearrow \{\psi> t\}\hbox{\ \ and\ \ }

981: \{\f\geq s_k\}\nearrow \{\f> t\},

982: $$

983: and we get that 

984: $$

985: \{\psi> t\}\hbox{ is quasi contained in }\{\f>t\},

986: $$

987: for all $t\in\R$.

988: 

989: Exchanging the roles of $\psi$ and $\f$ we get the reverse inclusion so

990: that

991: $\{\psi> s\}$ and $\{\f>s\}$ coincide up to sets of capacity zero.

992: 

993: Now we recover the

994: values of $\psi$ and $\f$ at quasi every point $x\in\Omega$ thanks to the

995: well known formula

996: $$

997: \f(x)=\sup_{s\in Q}s\,\chi_{\{\f> s\}}(x).

998: $$

999: Since the level sets are the same, the two functions coincide quasi

1000: everywhere.

1001: \endproof

1002: 

1003: The main result on level sets convergence is the following theorem,

1004: which shows the connection with the Mosco convergence introduced

1005: in~[\rf{MOS}] (for the proof see theorem~5.9 in~[\rf{DAL-4}]).

1006: 

1007: 

1008: \th{\thm{equiva}}{Let 

1009: $\psi_n$ and $\psi$ be functions $\Omega\to\overline\R$.

1010: If  

1011: $$

1012: \Kpsin\cap\Hunozero\mosco\Kpsi\cap\Hunozero.

1013: $$ 

1014: then 

1015: $$

1016: \psi_n\lev\psi.

1017: $$

1018: If moreover the obstacles are equicontrolled

1019: from above, namely

1020: $$

1021: \psi_n,\ \psi\,\leq u_\rho\,,

1022: \quad\hbox{ with }\rho\in\Inters,

1023: $$

1024: then also the reverse implication holds.}

1025: 

1026: Notice that, though very similar to Mosco convergence, the level set

1027: convergence concerns also the case of obstacles that are not

1028: $VI$-admissible.

1029: 

1030: 

1031: 

1032: Another simple observation, which requires no proof, but which is

1033: useful to state separately is the following.

1034: \lemma{\thm{crescente}}{Let $\psi_n\lev\psi$ and let $\Phi:\R\to\R$ be a

1035: continuous non decreasing function. Then

1036: $$

1037: \Phi(\psi_n)\lev\Phi(\psi).

1038: $$}

1039: {\vskip-1.5em}

1040: 

1041: In the next lemmas we will denote the solution of $OP(\mu,\psi_n)$ and

1042: of $OP(\mu,\psi)$ by $u_n$ and $u$, respectively.

1043: 

1044: Let us show that in general the Mosco convergence (and so also the level

1045: set convergence) of the obstacles does not imply the convergence of the

1046: solutions for an arbitrary measure.

1047: 

1048: \ex{\thm{nomosco}} Let $\Omega=B_1(0)\subseteq\R^N$, 

1049: with $N>2$, ${\cal A}=-\Delta$ and $\mu= 

1050: -\delta_0$, the Dirac delta in the origin.

1051: 

1052: Let the obstacles $\psi_n=-n$, so that clearly 

1053: $$

1054: \Kpsin\cap\Hunozero\mosco\Hunozero

1055: $$

1056: and, by theorem~\rf{equiva}, also $\psi_n\lev -\infty$

1057: 

1058: It is immediate to see that the solutions $u_n=u_{-\delta_0}

1059: +u_{\lambda_n}$ of $OP(-\delta_0, -n)$ are less than or equal to zero

1060: since the latter satisfies condition 1 of definition~\rf{disvar}.  So

1061: $u_n=T_n(u_n)$ and hence is in $\Hunozero$. But then

1062: $-\delta_0+\lambda_n\in\Hduale\cap\Mb\subset\Mbo$, and it must be a

1063: positive measure and hence $u_n=0$ for each $n$.  On the other hand

1064: $u=u_{-\delta_0}$ and cannot be the limit of the~$u_n$.

1065: 

1066: What can be proved without further assumptions is the following result.

1067: 

1068: \prop{\thm{ustar}}{Let $\psi_n,\psi\leq u_\rho\qe$ with 

1069: $\rho\in\Mb$. Assume that 

1070: $$

1071: \psi_n\lev\psi

1072: $$

1073: Then there exists a subsequence $u_{n'}$ and a quasi continuous function

1074: $u^*\in\Wunoqzero$, such that

1075: $$

1076: u_{n'}\to u^\ast\strongq,

1077: $$

1078: and 

1079: $$

1080: u^\ast\geq u\quad\qe.

1081: $$}

1082: {\vskip-1cm}

1083: \proof By theorem~\rf{esist2}

1084: $$

1085: \|\lambda_n\|_{\Mb}\leq\left\|(\mu-\rho)^-\right\|_{\Mb}\eqno(\frm{lambdan})

1086: $$

1087: so that there exists a subsequence

1088: $\{\lambda_{n'}\}$ and a measure

1089: $\lambda^*\in\Mbp$ such that $\lambda_n\wto\lambda^\ast,

1090: \weaks$ and hence $u_{n'}=\um+u_{\lambda_{n'}}\to

1091: u^\ast=\um+u_{\lambda^\ast}$ strongly in $\Wunoq$.

1092: 

1093: If we show that $u^\ast\geq\psi\qe$ we will have that $u^\ast$

1094: satisfies condition~1 of definition~\rf{disvar} and get the thesis,

1095: by definition~\rf{disvar}.

1096: 

1097: Given $k>0$, observe that, thanks to (\rf{tronche}) and to (\rf{lambdan})

1098: $$

1099:  \intl_\Omega|DT_k(u_{n'})|^2dx\leq kc

1100: $$ 

1101: and hence $T_k(u_{n'})\wto T_k(u^\ast)\weak2$.

1102: 

1103: From lemma~\rf{crescente} it follows that

1104: $$

1105: T_k(\psi_n)\lev T_k(\psi),\quad\forall k\in\R^+;

1106: $$

1107: Since $T_k(u_n)\geq T_k(\psi_n)\qe$ for each $n$ and $k$, and using

1108: theorem~\rf{equiva} and 

1109: the definition of Mosco convergence,  we get

1110: $T_k(u^\ast)\geq T_k(\psi)\qe$.

1111: 

1112: Now we can pass to the limit as $k\to+\infty$ and obtain

1113: $u^\ast\geq \psi\qe$.

1114: \endproof

1115: 

1116: 

1117: We prove now the central lemma of this section.

1118: 

1119: \lemma{\thm{psimenow}}{Let $\psi_n$ and $\psi$ be quasi upper

1120: semicontinuous functions

1121: controlled above by $u_\rho$ with $\rho\in\Inters$, and let $w$ be a quasi

1122: continuous function. If 

1123: $$

1124: \psi_n\lev\psi

1125: $$

1126: then

1127: $$

1128: \psi_n-w\lev\psi-w.

1129: $$}

1130: To prove this lemma we need the following result.

1131:  

1132: \lemma{\thm{c2approx}}{Given a quasi continuous function $w$, for 

1133: each $A\subset\subset\Omega$ and for each $\varepsilon>0$, there exists

1134: $u\in C^{\infty}_0(\Omega)$ such that

1135: $$

1136: {\rm cap}\left({\left\{|u-w|>\varepsilon\right\}\cap A}\right)

1137: <\varepsilon.\eqno(\frm{umenow})

1138: $$}

1139: {\vskip -1 truecm}

1140: \proof By definition of quasi continuity there exists a relatively

1141: closed subset $C$ such that 

1142: ${\rm cap}(\Omega\setminus C)<\varepsilon$ and $w|_{C}$  is

1143: continuous. By Tietze's theorem there exists a continuous function

1144: $g$ which extends $w_{|_{C\cap\overline A}}$ it to $\Rn$.

1145: 

1146: Obviously, for any $A\cc\Omega$, we have 

1147: $\left\{|w-g|>0\right\}\cap A\subseteq\Omega\setminus C$ so

1148: that

1149: $${\rm cap}\left({\left\{|w-g|>0\right\}}\cap A\right)

1150: <\varepsilon

1151: $$. 

1152: {\vskip-1.5em}

1153: 

1154: On its turn $g$ can be approximated in $A$ with a  

1155: function $u\in C^\infty_0(\Omega)$ so

1156: that $\sup_A|u-g|<\varepsilon$, and again from

1157: $\left\{|w-u|>\varepsilon\right\}\cap A\subseteq

1158: \left\{|w-g|>0\right\}\cap A$ we get 

1159: ${\rm cap}\left({\left\{|w-u|>\varepsilon\right\}}\cap A\right)<

1160: \varepsilon$.

1161: \endproof

1162: 

1163: \proofof{\rf{psimenow}}

1164: It is immediate to observe that the thesis is true when, instead of 

1165: $w$, we have a function $u$ in $\Hunozero$.

1166: This is because in our hypotheses level set convergence is equivalent to

1167: Mosco convergence (see theorem~\rf{equiva}) and translating $\Kpsin$ and

1168: $\Kpsi$ by

1169: $u\in\Hunozero$ we get

1170: $$

1171: {\rm K}_{(\psi_n-u)}\cap\Hunozero\mosco{\rm K}_{(\psi-u)}\cap\Hunozero.

1172: $$

1173: or equivalently

1174: $$

1175: \psi_n-u\lev\psi-u

1176: $$

1177: 

1178: We want to show the inequalities (\rf{liminf}) and (\rf{limsup}) for

1179: $\psi_n-w$ and $\psi-w$.

1180: 

1181: Let us fix $B\cc \Omega$, $\e>0$ and

1182: a function $u\in C^{\infty}_0(\Omega)$ such that (\rf{umenow}) holds with

1183: respect to

1184: $B$.

1185: 

1186: Observe now that for any $t\in\R$ we have

1187: $$

1188: \left\{\psi_n-w>t\right\}\cap B\subseteq\left({\left\{

1189: \psi_n-u>t-\e\right\}\cap B}\right)\cup

1190: \left({\left\{u-w>\e\right\}\cap B}\right),

1191: $$

1192: hence, by subadditivity,

1193: $$

1194: {\rm cap}\left({\left\{\psi_n-w>t\right\}\cap B}\right)

1195: \leq {\rm cap}\left({\left\{

1196: \psi_n-u>t-\e\right\}\cap B}\right)+{\rm cap}

1197: \left({\left\{|u-w|>\e\right\}\cap B}\right).

1198: $$

1199: Passing to the limsup and using (\rf{umenow}), we obtain

1200: $$

1201: 	\limsup_{n\to+\infty} 

1202: 	{\rm cap}\left({\left\{\psi_n-w>t\right\}\cap B}\right)

1203: 	\leq \limsup_{n\to+\infty} {\rm cap}\left({\left\{

1204: 	\psi_n-u>t-\e\right\}\cap B}\right)+\e.

1205: $$

1206: We know that for $\psi-u$,~(\rf{liminf}) holds true,

1207: so we can use it with $t-\e$ and $t-2\e$

1208: instead of

1209: $t$ and $s$, so that, for $B\cc B'\cc\Omega$, we get

1210: $$

1211: \limsup_{n\to\infty} 

1212: {\rm cap}\left({\left\{\psi_n-w>t\right\}\cap B}\right)\leq

1213: {\rm cap}\left({\left\{

1214: \psi-u>t-2\e\right\}\cap B'}\right)+\e.\eqno(\frm{passaggio})

1215: $$

1216: With an argument similar to before, from

1217: $$

1218: \left\{\psi-u>t-2\e\right\}\cap B'\subseteq\left({\left\{

1219: \psi-u>t-3\e\right\}\cap B'}\right)\cup

1220: \left({\left\{w-u>\e\right\}\cap B'}\right),

1221: $$

1222: we obtain,

1223: $$

1224: {\rm cap}\left({\left\{\psi-u>t-2\e\right\}\cap B'}\right)

1225: \leq {\rm cap}\left({\left\{

1226: \psi-u>t-3\e\right\}\cap B'}\right)+\e,

1227: $$

1228: and substituting in (\rf{passaggio})

1229: we get

1230: $$

1231: \limsup_{n\to\infty} 

1232: {\rm cap}\left({\left\{\psi_n-w>t\right\}\cap B}\right)\leq

1233: {\rm cap}\left({\left\{

1234: \psi-w>t-3\e\right\}\cap B'}\right)+2\e.

1235: $$

1236: For any choice of $s$ and $t$, $\e$ can be taken sufficiently small so

1237: that $s<t-3\e$. Then we can let $\e\to0$ and conclude

1238: $$

1239: \limsup_{n\to\infty} 

1240: {\rm cap}\left({\left\{\psi_n-w>t\right\}\cap B}\right)\leq

1241: {\rm cap}\left({\left\{

1242: \psi-w>s\right\}\cap B'''}\right).

1243: $$

1244: Here nothing depends on $u$ so this holds for all $s,t\in\R$, $s<t$ and

1245: for all $B\cc B'\cc\Omega$.

1246: 

1247: Inequality (\rf{liminf}) is proved in a similar way and this concludes the 

1248: proof.

1249: \endproof

1250: 

1251: \th{\thm{psimenoumu}}{Let $\psi_n,\psi\leq u_\rho$ with $\rho\in

1252: \Mb\cap\Hduale$. Let $\mu\in\Mb$ with $\mu^-\in\Hduale$ 

1253: and let $u_n$ and $u$ be the solutions of $OP(\mu,\psi_n)$

1254: and of $OP(\mu,\psi)$ respectively.

1255: If

1256: $$

1257: \psi_n\lev\psi

1258: $$

1259: then $u_n-u$ belongs to $\Hunozero$ and tends to zero 

1260: strongly in $\Hunozero$.}

1261: \proof

1262: Thanks to the previous lemma we have that

1263: $$

1264: \psi_n-\um\lev\psi-\um,

1265: $$

1266: and then, since $\psi_n-\um,\,\psi-\um\leq 

1267: u_{\rho}+u_{\mu^-}$ q.e.\ in

1268: $\Omega$, by theorem~\rf{equiva}, 

1269: $$

1270: {\rm K}_{(\psi_n-\um)}\cap\Hunozero\mosco{\rm

1271: K}_{(\psi-\um)}\cap\Hunozero.

1272: $$

1273: Hence all solutions of variational inequalities converge. In particular

1274: if $v_n$ and $v$ are the solutions of $VI(0,\psi_n-\um)$ and 

1275: $VI(0,\psi-\um)$, respectively, then $v_n\to v$ strongly $\Hunozero$. 

1276: 

1277: By lemma~\rf{magico} we have $u_n=v_n+\um$ and $u=v+\um$. This implies

1278: that $u_n-u=v_n-v$ and the conclusion follows.

1279: \endproof

1280: 

1281: The minimal hypothesis on the obstacles $\psi_n$ and $\psi$ in order to

1282: have the solutions of $OP(\mu,\psi_n)$ and $OP(\mu,\psi)$ is that they are

1283: $OP$-admissible. Nevertheless if in this theorem we drop the

1284: request that they are controlled by a function which is also in

1285: $\Hunozero$, the conclusion fails. Ideed there is the following example

1286: which derives from example~\rf{nomosco}.

1287: 

1288: \ex{\thm{controllo}}{Let us consider the operator ${\cal A}=-\Delta$, the

1289: domain $\Omega=B_1(0)\subseteq\R^N$, with $N>2$, 

1290: and the datum $\mu=0$. 

1291: 

1292: Consider now as obstacles $\psi_n=u_{\delta_0}-n$, where ${\delta_0}$

1293: is the Dirac delta centred at zero. They are

1294: clearly

1295: $OP$-admissible, and also bounded by the same function $u_\rho$, but in

1296: this case $\rho=\delta_0\not\in\Hduale$.

1297: 

1298: Now for each $n$ the solution $u_n$ of $OP(0,\psi_n)$ is $u_{\delta_0}$

1299: itself. Indeed, according to lemma~\rf{magico}, $u_n-u_{\delta_0}$ is the

1300: solution of $OP(-\delta_0,-n)$, that, as seen in 

1301: example~\rf{nomosco}, is zero.

1302: 

1303: But then we have that $\psi_n\lev -\infty$ and

1304: $u_n\to u_{\delta_0}$, while the solution of $OP(0,-\infty)$ is $u=0$.

1305: }

1306: 

1307: 

1308: 

1309: When the negative part of the measure $\mu$ is only in $\Mbo$, we can not

1310: use the same trick because the the sets $K_{\psi_n-\um}$ might be empty,

1311: but anyway we do not fall into the pathology of the example \rf{nomosco}, and

1312: in fact we

1313: can prove the following theorem which gives the convergence of the

1314: solutions as well, though in a weaker sense.

1315: 

1316: \th{\thm{menoemmezero}}{Let $\psi_n,\psi\leq u_\rho$ with $\rho\in

1317: \Mb\cap\Hduale$, $\mu\in\Mb$ such that $\mu^-\in\Mbo$, 

1318: and let $u_n$ and $u$ be the solutions of $OP(\mu,\psi_n)$

1319: and of $OP(\mu,\psi)$ respectively.

1320: If

1321: $$

1322: \psi_n\lev\psi,

1323: $$

1324: %$$

1325: %\Kpsin\cap\Hunozero\mosco\Kpsi\cap\Hunozero.

1326: %$$

1327: then $u_n\to u\strongq$.}

1328: \proof

1329: From [\rf{DAL-3}] we know that $\mu^-$ can be written as $g\nu$ where

1330: $\nu\in\Mbp\cap\Hduale$ and $g\in

1331: L^1(\Omega,\nu)$, $g\geq 0$.

1332: Hence the measure $\mu^-_k:=(g\wedge k)\nu$ is in $\Hduale$, so that

1333: $\mu_k:=\mu^+-\mu^-_k$ satisfies the hypothesis of the previous theorem.

1334: 

1335: Call $u^k_n$ and $u^k$ the solutions of $OP(\mu^k,\psi_n)$ and 

1336: $OP(\mu^k,\psi)$, respectively. By theorem~\rf{psimenoumu},

1337: $$

1338: u^k_n\to u^k,\ \strong2,\quad\forall k>0.

1339: $$

1340: 

1341: Now, observing that $\mu^--\mu^-_k$ is a positive 

1342: measure, we easily obtain by comparison 

1343: that $u^k_n\geq u_n$ and $u^k\geq u,\qe$.

1344: 

1345: On the other hand

1346: $$

1347: u_n+u_{(\mu^--\mu^-_k)}=\um+u_{(\lambda_n+\mu^--\mu^-_k)},

1348: \eqno(\frm{quella})

1349: $$

1350: where $\lambda_n\geq 0$ is the obstacle reaction of $OP(\mu,\psi_n)$.

1351: Since also $u_n+u_{(\mu^--\mu^-_k)}\geq u_n\geq\psi_n$, by definition

1352: \rf{disvar} and eq.~(\rf{quella}) we have $u_n+u_{(\mu^--\mu^-_k)}

1353: \geq u^k_n\qe$. In the same way we prove that

1354: $u+u_{(\mu^--\mu^-_k)}\geq u^k\qe$.

1355: 

1356: Since $\mu^--\mu^-_k\to 0$ strongly in $\Mb$, we have, by

1357: proposition~4.2 in~[\rf{DAL-LEO}] $u_{(\mu^--\mu^-_k)}\to 0\strongq$,

1358: so, from

1359: $$

1360: u+u_{(\mu^--\mu^-_k)}\geq u^k\geq u\quad\qe

1361: $$

1362: letting $k\to\infty$ we get that $u^k\to u\ae$.

1363: 

1364: Recalling proposition~\rf{ustar}, let us fix a subsequence $\{u_{n'}\}$

1365: which converges to a function $u^*$ strongly in $\Wunoq$, so that from

1366: $$

1367: u_{n'}+u_{(\mu^--\mu^-_k)}\geq u^k_{n'} \geq u_{n'}\quad\qe

1368: $$

1369: letting first $n'\to\infty$ and then $k\to\infty$ we obtain $u^\ast

1370: \geq u\geq u^\ast\qe$. Therefore $u^k_n\to u$, since the limit does not

1371: depend on the subsequence.

1372: \endproof

1373: 

1374: 

1375: 

1376: As seen in example \rf{controllo} the request that the obstacles be well

1377: controlled can not be dropped, even if the datum is regular. On the other

1378: hand example \rf{nomosco} showed that the control from above can be not

1379: enough to have convergence for all data $\mu\in\Mb$.

1380: 

1381: In the following theorem we show how, provided we strengthen the

1382: assumptions on the obstacles in the way given by theorem~\rf{lostesso},

1383: we can give up any assumption on the data $\mu$.

1384: 

1385: Notice that in the examples is always the limit obstacle the one that

1386: gives troubles.  Indeed we see here that it is enough to require the

1387: control from below only for the limit.

1388: 

1389: \th{\thm{tutto}}{Let $\psi_n,\psi\leq u_\rho$ with $\rho\in

1390: \Mb\cap\Hduale$, let $\mu\in\Mb$  and let $\psi$be an obstacle.

1391: Let $u_n$ and $u$ be the solutions of $OP(\mu,\psi_n)$

1392: and of $OP(\mu,\psi)$, respectively.

1393: If $\psi$ and $\mu$ satisfy condition~(\rf{condizione}) and

1394: \vskip-1em

1395: $$

1396: \psi_n\lev\psi

1397: $$

1398: %\vskip-1em

1399: \noindent then $u_n\to u\strongq$.}

1400: %\vskip-1em

1401: \proof

1402: From proposition~\rf{ustar} we know that, up to a subsequence,  $u_n\to

1403: u^*$ strongly in $\Wunoq$, and

1404: $u^*\geq u\qe$.

1405: 

1406: Now consider the $v_n$'s solutions of $OP(\mu^+-\mu^-_a,\psi_n)$. These, by

1407: theorem

1408: \rf{menoemmezero}, converge to $v$, the solution of

1409: $OP(\mu^+-\mu^-_a,\psi)$, but, according to theorem~\rf{lostesso}, $v=u$.

1410: 

1411: On the other side $v_n=\um+\uln+u_{\mu^-_s}$, with $\lambda_n\in\Mbp$, and

1412: $v_n\geq\psi_n\qe$ and so, by definition~\rf{disvar}, we have

1413: $$

1414: v_n\geq u_n\quad\qe.

1415: $$

1416: Letting $n$ go to $+\infty$ we obtain $u\geq u^*\qe$. Therefore $u_n\to

1417: u\strongq$. The limit doesn't depend on the subsequence, so the whole

1418: sequence $u_n$ converges to~$u$.

1419: \endproof

1420: 

1421: Let us show now a further example, which clarifies more

1422: deeply in which cases there is not convergence of the solutions.

1423: 

1424: In particular we see that theorems~\rf{lostesso} and~\rf{tutto} do not

1425: hold for some obstacles $\psi$ which are too singular only at one point.

1426: 

1427: \ex{\thm{menodelta}}{}Let us choose ${\cal A}=-\Delta$,

1428: $\Omega=B_1(0)\subseteq\R^N$, with $N>2$, and $\mu=-\delta_0$. Let us

1429: consider the obstacles $\psi=-u_{\delta_0}$ and

1430: $$

1431: \psi_n(x)=\cases{- \displaystyle{1\over 2}

1432: 		u_{\delta_0}(x)\quad& if \ \ \ 

1433: 			$|x|<a_n$\crr

1434: 		-n& if \ $a_n<|x|<b_n$\crr

1435: 		-u_{\delta_0}& if \ $b_n<|x|$\cr}

1436: \eqno(\frm{biz})$$

1437: where $a_n$ and $b_n$ are appropriate constants, which tend to zero as

1438: $n\to+\infty$ (see picture).

1439: 

1440: 

1441: 

1442: 

1443: It is easy to verify that  $\psi_n\lev\psi$ and that the solution of the

1444: limit problem $OP(-\delta_0,\psi)$ is clearly

1445: $-u_{\delta_0}$  itself.

1446: 

1447: Let us prove that the solution of $OP(-\delta_0,\psi_n)$ 

1448: is $- {1\over 2}u_{\delta_0}$.

1449: 

1450: This function satisfies condition~1 of definition~\rf{disvar} because it is of the form $\um+u_{{1\over

1451: 2}\delta_0}$ and it is above the obstacle for each $n$.

1452: 

1453: 

1454: Fix $n$ and suppose $\nu_n\in\Mbp$ such that $\um+u_{\nu_n}$ is the

1455: solution.

1456: Then it is smaller than or equal to $\um+u_{{1\over 2}\delta_0}$, or also

1457: $$

1458: \un\leq u_{{1\over 2}\delta_0}\quad\qe.

1459: $$

1460: \eject

1461: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1462: \hbox{\hsize 17cm

1463: \vtop{\hsize 8truecm

1464: In the small circle $B_{a_n}(0)$ they must be equal. In

1465: $B_1(0)\setminus B_{a_n}(0)$, $\un$ is superharmonic and $u_{{1\over

1466: 2}\delta_0}$ is harmonic and they have the same boundary data. So

1467: $\un\geq u_{{1\over 2}\delta_0}$ and they must coincide.

1468: 

1469: This proves that the solution $u_n$ of $OP(-\delta_0,\psi_n)$ is

1470: $- {1\over 2}u_{\delta_0}$ independently of $n$,

1471: and that $u_n$ does not converge

1472: to the solution $u=u_{\delta_0}$ of $OP(-\delta_0,\psi)$. 

1473: 

1474: As of remark~\rf{delte} we point out that in the example it is crucial

1475: that the

1476: deltas involved are centered in the same point. If for instance, with the

1477: same obstacles, we had as datum $\mu=-\delta_{x_0}$ for any $x_0\neq 0$, we

1478: would obtain, thanks to theorem~\rf{lostesso} that the solutions of

1479: $OP(-\delta_{x_0},\psi_n)$ and of $OP(-\delta_{x_0},\psi)$ are all

1480: identically zero.

1481: 

1482: The last consideration of this section concerns the fact that passing

1483: from theorem~\rf{psimenoumu} to theorem~\rf{menoemmezero} we loose

1484: something on the convergence of the solutions. To see that this loss

1485: is not due to the technique of the proof we can consider the following

1486: example.}

1487: \qquad

1488: \vtop{\hsize 7truecm \vskip 11cm

1489: \special{psfile=figura2.eps}}}

1490: %%%%% IN CASE A CAPTION IS NEEDED: The obstacles~(\rf{biz}). 

1491: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1492: 

1493: \ex{\thm{orsina}}{}

1494: Let $\Omega=B_{1\over 2}(0)\subseteq\R^N$, with $N>2$, 

1495: and let $f\in\Luno$ be the function defined

1496: by

1497: $$

1498: f(x)=\cases{

1499: 	\displaystyle{1\over{|x|^N(-\log|x|)^\vartheta}}\quad& if $x\in

1500: 		\Omega,\, x\neq 0$\crr

1501: 	0& if $x=0$\cr}

1502: $$

1503: with $\vartheta>1$. L. Orsina in [\rf{ORS}] noticed that the solution

1504: $u_f$ of the equation 

1505: $$

1506: \cases{-\Delta u_f=f\quad &in $\Omega$\crr

1507: 	u_f=0&on $\partial\Omega$\cr }

1508: $$

1509: \vskip-.5em

1510: \noindent belongs to $\Wunoq$ for any $q< {N\over N-1}$ but does not

1511: belong  to ${\rm W}^{1,{N\over N-1}}(\Omega)$.

1512: 

1513: With this choice of ${\cal A}$ and $\Omega$, take as datum $\mu=-f$,

1514: which clearly belongs to $\Mbo$, and as limit obstacle $\psi=-u_f$, which

1515: satisfies

1516: condition (\rf{condizione}) with $\sigma=f$, so that the solution of

1517: $OP(-f,-u_f)$

1518: is $-u_f$ itself.

1519: 

1520: If we set $\psi_n=-(u_f\wedge n)$ then the solution $u_n$ of

1521: $OP(-f,-(u_f\wedge n))$ is between $0$ (because $f$ is positive and

1522: $u_{-f}+u_f$ is a supersolution) and $-n$. Hence $u_n=T_n(u_n)$ and

1523: this implies that $u_n\in\Hunozero$.

1524: 

1525: Now it is easy to see (use for instance remark~\rf{semplice}) that

1526: $\psi_n\lev\psi$ so that, by proposition~\rf{tutto}, $u_n\to

1527: u\strongq$. Nevertheless it is not possible to have this convergence

1528: in the norm of ${\rm W}^{1,p}(\Omega)$ with $p\geq {N\over

1529: N-1}$, because the fact that $u_n\in\Huno$ would imply also that

1530: $u\in{\rm W}^{1,p}(\Omega)$, which is false since $u=-u_f$.

1531: 

1532: 

1533: \vskip-1em

1534: 

1535: \parag{\chp{teoremini}}{The case $\psi_n\leq\psi$.}

1536: \vskip-.5em

1537: 

1538: In the previous section we have seen that the level sets convergence

1539: of the obstacles in general it is not enough to give the convergence

1540: of the solutions for any $\mu$.

1541: 

1542: In this section we want to generalize

1543: proposition~\rf{psienne}. Pointwise convergence is replaced by level

1544: sets convergence, the condition $\psi_n\leq\psi\quad\qe$ is strong enough

1545: to give the result for any measure $\mu\in\Mb$ with no control from below on

1546: the limit obstacle.

1547: 

1548: \vskip-.5em

1549: 

1550: \prop{\thm{dasotto}}{Let $\psi_n$ and $\psi$ be $OP$-admissible and such 

1551: that

1552: \vskip-.5em

1553: $$

1554: \psi_n\lev\psi.

1555: $$

1556: \vskip-.5em

1557: \noindent If in addition $\psi_n\leq\psi\qe$ then 

1558: \vskip-.5em

1559: $$

1560: u_n\to u\quad\qe.

1561: $$}

1562: \vskip-1.5em

1563: \proof

1564: First apply proposition~\rf{ustar} from which we know that, up to a

1565: subsequence, $u_n\to u^*\strongq$ and that $u^*\geq u$.  

1566: The reverse inequality is guaranteed by the fact that $u\geq u_n$, for

1567: all $n$ thanks to condition~1 of definition~\rf{disvar}.

1568: \endproof

1569: 

1570: Let us remark that this is a generalized version of proposition

1571: \rf{psienne}. Indeed under the assumption that $\psi_n\leq\psi$, we have

1572: that quasi everywhere convergence implies level sets convergence. 

1573: 

1574: If $\psi_n$ is monotone increasing, this is an easy consequence of the

1575: continuity of capacities on increasing sequences of sets.

1576: 

1577: The general case, $\psi_n\leq\psi$ but $\psi_n$ not necessarily

1578: increasing, is proved by considering the sequence

1579: $\f_n:=\inf_{k\geq n}\psi_k$, which is increasing, converge pointwise

1580: to $\psi$ and satisfies

1581: $$

1582: {\rm cap}(\{\f_n>t\}\cap B)\leq

1583: {\rm cap}(\{\psi_n>t\}\cap B)\leq

1584: {\rm cap}(\{\psi>t\}\cap B)

1585: $$

1586: for all $t\in\R$, and $B\cc\Omega$. Passing to the limit, thanks to the

1587: previous step, we conclude the proof.

1588: 

1589: 

1590: \parag{\chp{convhuno}} {Obstacles converging in the energy space.}

1591: 

1592: It is well known that in the case of variational inequalities

1593: the convergence of obstacles in the norm of $\Huno$ implies the 

1594: convergence of the corresponding solutions. In particular we have the

1595: following result

1596: 

1597: \th{\thm{vihuno}}{Let $\psi_1,\psi_2:\Omega\to\overline{\R}$ 

1598: be 

1599: $VI$-admissible. Suppose that $\psi_1-\psi_2\in\Hunozero$. Let $f\in\Hduale$  and let 

1600: $u_1$ and $u_2$ be the solutions of $VI(f,\psi_1)$ and $VI(f,\psi_2)$, 

1601: respectively. Then

1602: $$

1603: \|u_n-u\|_{\Hunozero}\leq {C\over\gamma}\,\left\|\psi_n-\psi\right\|_{\Hunozero},\eqno(\frm{acca})

1604: $$

1605: where $\gamma$ is the ellipticity constant and $C$ is such that

1606: $|\dualita{{\cal A}u}{v}|\leq C\|u\|_{\Hunozero}\|v\|_{\Hunozero}$, so they  

1607: depend only on the operator ${\cal A}$.} 

1608: 

1609: Also for the solutions of obstacle problems, we want to investigate

1610: the dependence on $\Huno$ convergence. In this frame, as we have seen

1611: with example~\rf{orsina}, this can not follow directly from Mosco

1612: convergence, as it was in the variational case.  The next theorem

1613: concerns the case in which the obstacles ``have the same boundary

1614: value''.

1615: 

1616: 

1617: \th{\thm{ophunozero}}{Let $\psi_n,\psi:\Omega\to

1618: \overline{\R}$ be such that $\psi_n,\psi\leq u_\rho$ 

1619: q.e.\ in $\Omega$, with $\rho\in\Mb$. 

1620: Assume  $\psi_n-\psi\in\Hunozero$, and let $\psi_n-\psi

1621: \to 0\strong2$. Let $\mu\in\Mb$ and let 

1622: $u_n$ and $u$ be the solutions of $OP(\mu,\psi_n)$ and 

1623: $OP(\mu,\psi)$, respectively. 

1624: Then

1625: $$

1626: u_n-u\in\Hunozero\hbox{ \ and \ }u_n-u\to 0\strong2,

1627: $$

1628: }

1629: \proof

1630: As a first step assume that $\rho$ belongs also to $\Hduale$,

1631: and consider the special sequence $\mu_k$ of 

1632: measures in $\Mb\cap\Hduale$ 

1633: $\mu_k\wto\mu,\weaks$, such that the solutions of 

1634: the corresponding obstacle

1635: problems converge (see theorem~\rf{esist2}). In particular $u^k_n\to

1636: u_n,\strongq$ for all

1637: $n$ and $u^k\to u,\strongq$.

1638: 

1639: Thanks to (\rf{acca}), for all $k$ we also have 

1640: $$

1641: \|u^k_n-u^k\|_{\Hunozero} \leq c\|\psi_n-\psi\|_{\Hunozero},

1642: $$

1643: so that the sequence $\{u_n^k-u^k\}_k$ is bounded in $\Hunozero$, 

1644: for each $n$ fixed. Thus, up to a subsequence, there is a limit 

1645: function $z$. But we already know that the sequence converges,

1646: strongly in $\Wunoq$, to $u_n-u$, so this must be also the weak limit

1647: in $\Huno$.

1648: 

1649: By lower semicontinuity of the norm we have

1650: $$

1651: \|u_n-u\|_{\Hunozero} \leq\liminf_{k\to\infty}

1652: \|u^k_n-u^k\|_{\Hunozero} \leq c\|\psi_n-\psi\|_{\Hunozero}.

1653: $$

1654: This says that $u_n-u$ belongs to $\Hunozero$ (while $u_n$ and $u$, in 

1655: general, do not) and gives the thesis in the first case.

1656: 

1657: Let now $\rho$ be only in $\Mb$.

1658: Set

1659: $\psi^h:=\psi\wedge h$ and $\psi^h_n:=\psi_n-\psi+\psi^h$. 

1660: These obstacles are equi $OP$-admissible, because $\psi_n^h\leq\psi_n$ and

1661: $\psi^h\leq\psi$ q.e.\ in $\Omega$.

1662: They are also equi  $VI$-admissible since, if $\psi\leq u_\rho$, then 

1663: $\psi^h\leq T_h(u_\rho)\in\Hunozero$ and $\psi^h_n\leq\psi_n-\psi+

1664: T_h(u_\rho)\in\Hunozero$, and we can find a 

1665: function $\f\in\Hunozero$ such

1666: that $\f\geq\psi_n-\psi$ for all $n$. 

1667: 

1668: It is easy to see that in this case there 

1669: exists $\rho_h\in\Inters$ such that $\psi^h_n,\,\psi^h\leq u_{\rho_h}$

1670: q.e.\ in $\Omega$. Hence we are in the hypothesis of the

1671: first step. 

1672: Moreover $\|\psi^h_n-\psi^h\|_{\Hunozero}=\|\psi_n-\psi\|_{\Hunozero}$.

1673: 

1674: So by the first step, for each $h$

1675: $$

1676: \|u^h_n-u^h\|_{\Hunozero} \leq\gamma\|\psi_n-\psi\|_{\Hunozero}.

1677: $$

1678: On the other side we know that, since $\psi_n^h\nearrow\psi_n$ and

1679: $\psi^h\nearrow\psi$, by proposition~\rf{psienne}  $u^h_n-u^h\to

1680: u_n-u\strongq$, so that

1681: we can conclude as in the first step.

1682: \endproof

1683: 

1684: We want to remark that if, more generally, the obstacles are such that

1685: $\psi_n-\psi\to 0$ in $\Huno$ then they also converge in the sense of

1686: level sets, so we can deduce the convergence of the solutions in all the

1687: situations given by theorems \rf{psimenoumu}, \rf{menoemmezero} and

1688: \rf{tutto}, but here we obtain a stronger convergence with no further

1689: assumptions on the obstacles and on the data.

1690: 

1691: We may now wonder what happens when the obstacles converge in the space

1692: $\Wunoq$, with $1<q< {N\over N-1}$. In general this is not

1693: enough to obtain the convergence of the

1694: solutions. Indeed reconsider example~\rf{menodelta}. 

1695: Let us prove that

1696: $\psi_n\to\psi$ strongly in $\Wunoqzero$. We have

1697: $$

1698: \psi_n-\psi=\cases{ {\displaystyle{1\over 2}} u_{\delta_0} 

1699: 		& in $\quad|x|<a_n$\crr

1700: 		u_{\delta_0}-n\qquad& in $a_n<|x|<b_n$\crr

1701: 		0& otherwise\cr}

1702: $$

1703: so that 

1704: $$

1705: \|\psi_n-\psi\|^q_{\Wunoq}=

1706: {1\over 2}\intl_{B_{a_n}(0)}|Du_{\delta_0}|^qdx+

1707: {\hskip-.6cm}\intl_{B_{b_n}(0)\setminus

1708: B_{a_n}(0)}{\hskip-.6cm}|Du_{\delta_0}|^qdx,

1709: $$

1710: which tends to zero, since $a_n$ and $b_n$ tend to zero and by

1711: the absolute continuity of the integral.

1712: But, as already seen in example~\rf{menodelta},

1713: the solutions of the obstacle problems do not converge.

1714: 

1715: Anyway it is possible to prove the following result.

1716: 

1717: \prop{\thm{rhoenne}}{Let $\mu$ be in $\Mb$ and let $\psi_n$ and $\psi$ be 

1718: $OP$-admissible and such that $\psi_n-\psi=u_{\rho_n}$ with

1719: $\rho_n\in\Mb$, $\|\rho_n\|_{\Mb}\to 0$. Then

1720: $$

1721: u_n\to u\strongq,

1722: $$

1723: where $u_n$ and $u$ are the solutions of $OP(\mu,\psi_n)$ and

1724: $OP(\mu,\psi)$, respectively.}

1725: \proof

1726: Since $\psi_n=\psi-u_{\rho_n}$, we have (using lemma \rf{magico}) that

1727: $u_n-u_{\rho_n}$ is the solution of $OP(\mu+\rho_n,\psi)$. So from

1728: theorem~4.2 in~[\rf{DAL-LEO}] we get that 

1729: $$

1730: u_n-u_{\rho_n}\to u\strongq.\eqno(\frm{urhon})

1731: $$

1732: 

1733: Then

1734: $$

1735: \|u_n-u\|_{\Wunoq}\leq

1736: \|u_n-u_{\rho_n}-u\|_{\Wunoq}+\|u_{\rho_n}\|_{\Wunoq};

1737: $$

1738: the first term goes to zero because of~(\rf{urhon}), 

1739: the second one by hypothesis, and we get the thesis.

1740: \endproof

1741: 

1742: \parag{\chp{elleinfinito}}{Problems with nonzero boundary data and

1743: uniform convergence}

1744: 

1745: In this section we extend the theory of obstacle problems with measure

1746: data developed in [\rf{DAL-LEO}] to problems with nonzero 

1747: boundary data. This is standard for variational inequalities

1748: and also in this case 

1749: this generalization is very simple; we will

1750: only point out what has to be settled.

1751: 

1752: Let $g\in\Huno$ we will denote by $u^g_0$ the solution of

1753: $$

1754: \cases{{\cal A}u^g_0=0\qquad& in $\Hduale$\crr

1755: 	u^g_0-g\in\Hunozero.\cr}

1756: $$

1757: We will look for solutions of obstacle problems which take the value $g$

1758: on the boundary $\partial \Omega$. So we have to change accordingly the

1759: notion of admissibility for the obstacles.

1760: 

1761: An obstacle

1762: $\psi:\Omega\to{\overline{\R}}$ is said to be $OP_g$-admissible if

1763: $$

1764: \exists \rho\in\Mbp\ \hbox{s.t. }\psi\leq u_\rho+u^g_0\quad\qe.

1765: $$

1766: 

1767: Given a measure $\mu\in\Mb$, a boundary datum $g\in\Huno$ and an

1768: $OP_g$-admissible obstcale $\psi$, the solution of the obstacle

1769: problem $OP(\mu,g,\psi)$, if it exists, is the minimum element of the

1770: set

1771: $$

1772: {\cal F}^g_\psi(\mu):=\left\{v\in\Wunoq\,:\,\exists\nu\in\Mbp,\,

1773: v=\um+u^g_0+\un;\,v\geq\psi\qe\right\}.

1774: $$

1775: 

1776: It is immediate to prove the following

1777: 

1778: \th{\thm{esistg}}{Let $\mu\in\Mb$ and let $\psi$ be $OP_g$-admissible.

1779: Then there exists a unique solution of $OP(\mu,g,\psi)$.}

1780: \proof

1781: Consider the obstacle $\psi-u^g_0$. It is $OP$-admissible. So there

1782: exists a unique solution $v$ of $OP(\mu,\psi-u^g_0)$. Then $v+\ugo$ is our

1783: solution: indeed it belongs to $\Fgpsimu$, and it is less than or equal to

1784: any $z\in\Fgpsimu$.

1785: \endproof

1786: 

1787: \rem{\thm{stima}}{From theorem~\rf{esistg} and (\rf{mumenorho}) it

1788: follows also that, if $\um+\ugo+\ul$ is the solution of $OP(\mu,g,\psi)$,

1789: then

1790: $$

1791: \|\lambda\|_{\Mb}\leq\left\|(\mu-\rho)^-\right\|_{\Mb},

1792: $$

1793: independently of $g$.}

1794: 

1795: \rem{\thm{convergenze}}{Since $\psi$ is $OP$-admissible if and only if

1796: $\psi-\ugo$ is $OP_g$-admissible and

1797: $u_n\to u\strongq$ if and only if $u_n-\ugo\to u-\ugo\strongq$, all

1798: the theorems on continuous dependence on the data hold 

1799: without modifications, in particular

1800: propositions \rf{psienne} which will be useful in the 

1801: following.}

1802: 

1803: \rem{\thm{coincide}}{When $\mu\in\Inters$ and $\psi\leq u_\rho+u^g_0\qe$

1804: with $\rho\in\Inters$ then the solution of $OP(\mu,g,\psi)$ coincides

1805: with the solution of $VI(\mu,g,\psi)$\footnote{$^\ast$}{\eightpoint There is no need to

1806: define explicitly what is $VI(\mu,g,\psi)$.}.}

1807: 

1808: 

1809: 

1810: We come now to discuss the continuous dependence of the solutions on the

1811: obstacles when these converge uniformly.

1812: 

1813: To do this we will use a characterization via supersolutions similar to

1814: the one that holds in the variational case (see~[\rf{KIN-STA}]). 

1815: 

1816: %Recall that writing $u\geq

1817: %g$ on $\partial\Omega$ we mean in the sense of $\Huno$, i.e.\

1818: %$(u-g)^-\in\Hunozero$, and we say that $\psi$ is $VI_g$-admissible if

1819: %there exists $z\in\Huno$ s.t.\

1820: %$$

1821: %\psi\leq z\ \quad\qe\quad\hbox{ with }\ z-g\in\Hunozero.

1822: %$$

1823: %

1824: %

1825: %

1826: %\prop{\thm{sopra}}{Let $F\in\Hduale$, $\psi$ be

1827: %$VI_g$-admissible and $g\in\Huno$. Then $u$ is the solution of

1828: %$VI(F,g,\psi)$ if and only if the two following conditions hold:

1829: %$$

1830: %\cases{{\cal A}u-F\geq 0& in ${\cal D}'(\Omega)$\crr

1831: %		 u\geq\psi& q.e.\ in $\Omega$\crr

1832: %		 u\geq g& on $\partial\Omega$\cr}

1833: %$$

1834: %and for all v such that 

1835: %$$

1836: %\cases{{\cal A}v-F\geq 0& in ${\cal D}'(\Omega)$\crr

1837: %		 v\geq\psi& q.e.\ in $\Omega$\crr

1838: %		 v\geq g& on $\partial\Omega$\cr}

1839: %$$

1840: %then $v\geq u,\quad\qe$.}

1841: %

1842: 

1843: To this aim, let us introduce the set ${\cal G}^g_\psi(\mu)$ of all

1844: the functions $v\in\Wunoq$ with $v\geq\psi\qe$, such that

1845: $v=\um+u^h_0+\un$, where $\nu\in\Mbp$ and $h\in\Huno$ such that $h\geq

1846: g$ on $\partial\Omega$, i.e.\ $(h-g)^-\in\Hunozero$.

1847: 

1848: 

1849: We see now that the solution of $OP(\mu,g,\psi)$ can be compared not only

1850: with the functions of $\Fgpsimu$, but also with all those

1851: that have boundary datum greater than or equal to $g$.

1852: 

1853: 

1854: \prop{\thm{Ggpsimu}}{Let $\mu\in\Mb$ and $\psi$ be $OP_g$-admissible.

1855: If $u$ is the solution of $OP(\mu,g,\psi)$ then it is the

1856: minimum element of $\Ggpsimu$.}

1857: \proof

1858: {\it Step 1.} First consider $\mu\in\Inters$ and $\psi$ both

1859: $VI$- and $OP$-admissible.

1860: 

1861: Let $v=\um+\uho+\un\in\Ggpsimu$. We approximate $\nu$ by means of the

1862: sequence $\nu_k:={\cal A}T_k(\un)$. We have that $\nu_k\in\Inters$ and

1863: that $\nu_k\wto\nu\weaks$. Moreover observe that $u_{\nu_k}=T_k(\un)$ tends

1864: to $\un$ q.e.\ in $\Omega$ and, since $\un$ is nonnegative it is an

1865: increasing sequence.

1866: 

1867: Hence if we define $v_k:=\um+\uho+u_{\nu_k}$, then $v_k\nearrow v\qe$, and

1868: setting $\psi_k:=\psi\wedge v_k$ also $\psi_k\nearrow \psi\qe$

1869: 

1870: Let now $u_k$ be the

1871: solutions of $VI(\mu,g,\psi_k)$. So, by theorem~II.6.4 in~[\rf{KIN-STA}],

1872: $v_k\geq u_k$.

1873: Using proposition~\rf{psienne} and remark~\rf{convergenze} 

1874: we know that $u_k\to u\ae$. Then $v\geq

1875: u\ae$ and then also$\qe$.

1876: 

1877: {\it Step 2.} Consider now $\mu\in\Mb$ and $\psi$ still 

1878: both $VI$- and $OP$-admissible. Take again $v\in\Ggpsimu$.

1879: 

1880: Let $\mu_k={\cal A}T_k(\um-u_\rho)+\rho$ be  the sequence of measures 

1881: given in theorem~\rf{esist2}, so that we know that if $u_k$ are the 

1882: solutions of $VI(\mu_k,g,\psi)$ then $u_k\to u\strongq$. 

1883: 

1884: Taking now $v_k=u_{\mu_k}+\uho+\un$ it is easy to verify that $v_k\geq

1885: \psi\qe$ for all $k>0$, and then, by definition~\rf{disvar}, $v_k

1886: \geq u_k\qe$.

1887: Also $v_k\to v\strongq$ so, passing to the limit, we obtain $v\geq u\ae$ 

1888: and then also $\qe$.

1889: 

1890: {\it Step 3.} Finally consider the general case $\mu\in\Mb$ and $\psi$ 

1891: $OP$-admissible.

1892: The obstacles $\psi_k:=\psi\wedge k$ are also $VI$-admissible and such

1893: that $\psi_k\nearrow\psi\qe$. So,  if 

1894: $u_k$ is the solution of $OP(\mu,g,\psi_k)$, 

1895: by proposition~\rf{psienne} and remark~\rf{convergenze}, 

1896: we have that $u_k\to u\strongq$.

1897: 

1898: Taken any $v\in\Ggpsimu$, then $v\geq \psi_k$, for all $k$. Hence, by 

1899: definition~\rf{disvar}, $v\geq u_k\qe$. Passing to the limit,

1900: we get $v\geq u\ae$ and also$\qe$.

1901: \endproof

1902: From this we point out that the sets $\Fgpsimu$ and $\Ggpsimu$ have the

1903: following lattice property.

1904:  

1905: \prop{\thm{reticolo}}{Let $\mu\in\Mb$, $g\in\Huno$ and $\psi$

1906: $OP_g$-admissible. Then

1907: \item{(i)}If $u,v\in\Fgpsimu$ then $u\wedge v\in\Fgpsimu$;

1908: \item{(ii)}If $u,v\in\Ggpsimu$ then $u\wedge v\in\Ggpsimu$.}

1909: \proof

1910: Let us prove only the first statement, the proof of the second being

1911: alike.

1912: 

1913: Set $w:=u\wedge v$ and let $z$ be the solution of $OP(\mu,g,w)$. Then

1914: $u,v\in {\cal F}^g_w(\mu)$ and hence also $w\geq z$.

1915: 

1916: On the other hand $z\geq w$ and hence they are equal. So $u\wedge v$ is of

1917: the form $\um+\ugo+\un$ and is above $\psi$, and hence belongs to

1918: $\Fgpsimu$.

1919: \endproof

1920: 

1921: We can prove now the following continuity result

1922: 

1923: \th{\thm{uniforme}}{Let $\mu\in\Mb$, $g\in\Huno$, and $\psi_n$ and

1924: $\psi$ be

1925: $OP$-admissible 

1926: and let $u_n$ and $u$ be the solutions of $OP(\mu,\psi_n)$ and $OP(\mu,

1927: \psi)$, respectively.

1928: Assume that $\psi_n-\psi\in\Linf$ and

1929: $\psi_n-\psi\to0$ in $\Linf$.

1930: Then

1931: $$

1932: u_n-u\in\Linf\quad\hbox{ and }\quad u_n-u\to0\hbox{ in }\Linf.

1933: $$}

1934: \negbigskip

1935: \proof

1936: Set $c_n:=\|\psi_n-\psi\|_{\Linf}$. Obviously $c_n=u^{c_n}_0$, so that

1937: $$

1938: u+c_n=\um+u^{c_n+g}_0+\ul\ \hbox{ and }\ u+c_n\geq\psi_n\quad\qe

1939: $$

1940: hence $u+c_n\in{\cal G}^g_{\psi_n}(\mu)$ and hence $u+c_n\geq u_n$.

1941: 

1942: The same can be done the other way round to obtain that $u_n+c_n\geq u$.

1943: In the end we get $|u_n- u|\leq c_n$, and, taking the sup over $x\in\Omega$, the

1944: thesis.

1945: \endproof

1946: 

1947: 

1948: \rem{\thm{implica}}{Also in this case we have to remark that the uniform

1949: convergence of the obstacles implies their level set convergence (via

1950: remark~\rf{cap}). But the result we have obtained in this section does

1951: not require that the obstacles be equicontrolled, and the convergence

1952: of the solutions is in a different norm.}

1953: 

1954: \bigskip

1955: 

1956: {\bf Acknowldgement.} The author would like to thank prof.~Gianni Dal Maso

1957: for helpful discussions and valuable suggestions.

1958: 

1959: %bibliografia

1960: \intro{References}

1961: %

1962: \def\interrefspace{\smallskip}  %spazio tra le voci della bibliografia   

1963: %

1964: {\ninepoint\frenchspacing

1965: 

1966: \item{[\bib{ATT}]}

1967: ATTOUCH  H.:

1968: {\it Variational convergence for 

1969: functions and operators}. Pitman, Boston, 1984

1970: 

1971: \item{[\bib{ATT-PIC-1}]}

1972: ATTOUCH  H., PICARD  C.:

1973: Probl\`emes variationnels et th\'eorie du potentiel non lin\'eaire.

1974: {\it Ann. Fac. Sci. Toulouse } {\bf 1} (1979), 89-136

1975: \interrefspace 

1976: 

1977: \item{[\bib{ATT-PIC-2}]}

1978: ATTOUCH  H., PICARD  C.:

1979: In\'equations variationnelles avec obstacles et espaces fonctionnels en

1980: th\'eorie du potentiel.

1981: {\it Applicable Analysis } {\bf 12} (1981), 287-306

1982: \interrefspace 

1983: 

1984: %\item{[\bib{B-B-G-G-P-V}]}

1985: %B\'ENILAIN  Ph., BOCCARDO L., GALLOU\"ET  T.,

1986: %GARIEPY  R., PIERRE  M., VAZQUEZ  J.L.:

1987: %An $L^1$-theory of existence and uniqueness of solutions of nonlinear

1988: %elliptic equations.

1989: %%{\it J. Funct. Anal.\/} {\bf 50} (1983), 329-386.

1990: %\interrefspace

1991: 

1992: \item{[\bib{BOC-CIR-1}]}

1993: BOCCARDO  L., CIRMI  G.R.:

1994: Nonsmooth unilateral problems. {\it Nonsmooth optimization: methods and

1995: applications (Erice, 1991)}, F. Giannessi ed., 1-10. Gordon and Breach,

1996: Amsterdam, 1992

1997: $L^1$-data.

1998: \interrefspace

1999: 

2000: \item{[\bib{BOC-CIR-2}]}

2001: BOCCARDO  L., CIRMI  G.R.:

2002: Existence and uniqueness of solution of unilateral problems with

2003: $L^1$-data.

2004: \interrefspace

2005: 

2006: \item{[\bib{BOC-GAL}]}

2007: BOCCARDO  L., GALLOU\"ET  T.:

2008: Probl\`emes unilat\'eraux avec donn\'ees dans $\hbox{L}^1$. {\it

2009: C. R. Acad. Sci. Paris S\'er. I Math.} {\bf 311} (1990), 617--655.

2010: \interrefspace

2011: 

2012: \item{[\bib{BOC-GAL-ORS}]}BOCCARDO  L., GALLOU\"ET  T., ORSINA  L.:

2013: Existence and uniqueness of entropy solutions for nonlinear elliptic

2014: equations with measure data. {\it Ann. Inst. Henri Poincar\'e}  {\bf 13}

2015: (1996), 539-551.

2016: \interrefspace

2017: 

2018: 

2019: \item{[\bib{BOC-MUR-1}]}

2020: BOCCARDO  L., MURAT  F.:

2021: Nouveaux r\'esultats de convergence dans des probl\`emes unilateraux.

2022: In ``Nonlinear partial differential equations and their applications''.

2023: Coll\`ege de France Seminar, Volume II. H. Br\'ezis and J.L. Lions

2024: editors. Pitman, London, 1982

2025: \interrefspace

2026: 

2027: \item{[\bib{DAL-3}]} 

2028: DAL MASO G.: 

2029: On the integral representation of certain local functionals. 

2030: {\it Ricerche Mat.}, {\bf 22} (1983), 85-113.

2031: \interrefspace

2032: 

2033: 

2034: \item{[\bib{DAL-4}]}

2035: DAL MASO  G.:

2036: Some necessary and sufficient conditions for the convergence of sequences

2037: of unilateral convex sets.

2038: {\it J. Funct. Anal.} {\bf 62,2} (1985), 119-159.

2039: \interrefspace

2040: 

2041: 

2042: \item{[\bib{DAL-DAL}]}

2043: DALL'AGLIO  P., DAL MASO G.:

2044: Some properties of solutions of obstacle problems with measure data.

2045: Preprint S.I.S.S.A., Trieste, 1998

2046: \interrefspace

2047: 

2048: \item{[\bib{DAL-LEO}]}

2049: DALL'AGLIO  P., LEONE  C.:

2050: Obstacle problems with measure data.

2051: Preprint S.I.S.S.A., Trieste, 1997.

2052: \interrefspace

2053: 

2054: \item{[\bib{FUG}]}

2055: FUGLEDE  B.:

2056: The quasi topology associated with a countably subadditive set function.

2057: {\it Ann. Inst. Fourier} {\bf 21,1} (1971), 123-169

2058: \interrefspace

2059: 

2060: \item{[\bib{FUK}]}

2061: FUKUSHIMA M., SATO K., TANIGUCHI S.:

2062: On the closable part of pre-Dirichlet forms and the fine supports of 

2063: underlying measures.

2064: {\it Osaka J. Math.} {\bf 28} (1991), 517-535.

2065: \interrefspace

2066: 

2067: %\item{[\bib{GRU-WID}]}

2068: %GR\"UTER  M., WIDMAN  K.O.:

2069: %The Green function for uniformly elliptic equations.

2070: %{\it Manuscripta Math.} {\bf 37} (1982), 303-342.

2071: %\interrefspace

2072: 

2073: \item{[\bib{HEI-KIL}]}HEINONEN  J., KILPEL\"AINEN  T., MARTIO  O.: {\it

2074: Nonlinear potential theory of degenerate elliptic equations.} Clarendon

2075: Press, Oxford, 1993.

2076: \interrefspace

2077: 

2078: \item{[\bib{KIN-STA}]}KINDERLEHRER  D., STAMPACCHIA  G.: {\it

2079: An introduction to variational inequalities and their applications.}

2080: Academic, New York, 1980.

2081: 

2082: 

2083: 

2084: %\item{[\bib{LAN}]}

2085: %LANDKOF  N.S.:

2086: %{\it Foundations of potential theory.} 

2087: %Springer Verlag, Berlin, 1972

2088: %\interrefspace

2089: 

2090: \item{[\bib{LEO}]}

2091: LEONE  C.:

2092: Existence and uniqueness of solutions for nonlinear obstacle problems

2093: with

2094: measure data.

2095: Preprint S.I.S.S.A., Trieste, 1998

2096: \interrefspace

2097: 

2098: 

2099: %\item{[\bib{LEW-STA}]}

2100: %LEWY  H., STAMPACCHIA  G.:

2101: %On the regularity of the solution of a variational inequality.

2102: %{\it Comm. Pure Appl. Math.\/} {\bf 22} (1969), 153-188.

2103: %\interrefspace

2104: 

2105: %\item{[\bib{LEW-STA-2}]}

2106: %LEWY  H., STAMPACCHIA  G.:

2107: %On the smoothness of superharmonics which solve a minimum problem.

2108: %{\it J. Analyse Math.\/} {\bf 23} (1970), 224-236. 

2109: %\interrefspace

2110: 

2111: \item{[\bib{MOS}]}

2112: MOSCO  U.:

2113: Convergence of convex sets and of solutions of variational inequalities. 

2114: {\it Adv. in Math.\/} {\bf 3} (1969), 510-585

2115: \interrefspace

2116: 

2117: \item{[\bib{OPP-ROS-1}]}

2118: OPPEZZI P., ROSSI A.M.:

2119: Existence and solutions for unilateral problems with multivalued

2120: operators. {\it J. Convex. Anal.} {\bf 2} (1995) 241--261.

2121: \interrefspace

2122: 

2123: \item{[\bib{OPP-ROS-2}]}

2124: OPPEZZI P., ROSSI A.M.:

2125: Esistenza di soluzioni per problemi unilaterali con dato misura o

2126: $\hbox{L}^1$. {\it Ricerche Mat.}, to appear.

2127: \interrefspace

2128: 

2129: \item{[\bib{ORS}]}

2130: ORSINA  L.:

2131: Solvability of linear and semilinear eigenvalue problems with $\Luno$

2132: data.

2133: {\it Rend. Sem. Mat. Padova } {\bf 90} (1993), 207-238

2134: \interrefspace

2135: 

2136: %\item{[\bib{ORS-PRI}]}

2137: %ORSINA  L., PRIGNET  A.:

2138: %Nonexistence of solutions for some nonlinear elliptic equations involving

2139: %measures.

2140: %\interrefspace

2141: 

2142: \item{[\bib{STA}]}

2143: STAMPACCHIA  G.:

2144: Le probl\`eme de Dirichlet pour les \'equations elliptiques du second

2145: ordre \`a coefficients discontinus.

2146: {\it Ann. Inst. Fourier Grenoble\/} {\bf 15} (1965), 189-258.

2147: \interrefspace

2148: 

2149: \item{[\bib{TRO}]}TROIANIELLO  G.M.: {\it

2150: Elliptic differential equations and obstacle problems.}

2151: Plenum Press, New York, 1987.

2152: 

2153: 

2154: \end

2155: 

2156: