1: \documentclass[11pt, keyword, address]{article}

2: \usepackage{graphicx, comment}

3: \usepackage{times}

4: \usepackage{amsmath}

5: \usepackage{amssymb}

6: \usepackage{latexsym}

7: \usepackage{amsthm}

8: \newtheorem{theorem}{Theorem}[section]

9: \newtheorem{lemma}[theorem]{Lemma}

10: \newtheorem{proposition}[theorem]{Proposition}

11: \newtheorem{example}[theorem]{Example} 

12: \newtheorem{definition}[theorem]{Definition}

13: \newtheorem{corollary}[theorem]{Corollary}

14: \newtheorem{remark}[theorem]{Remark}

15: \newcommand{\As}{\mathcal{A}_{\mathrm{s}}}

16: \newcommand{\Aw}{\mathcal{A}_{\mathrm{w}}}

17: \newcommand{\A}{\mathcal{A}}

18: \newcommand{\Areg}{\mathcal{A}_{\mathrm{reg}}}

19: \newcommand{\ds}{\mathrm{d}_{\mathrm{s}}}

20: \newcommand{\dw}{\mathrm{d}_{\mathrm{w}}}

21: \newcommand{\dd}{\mathrm{d}}

22: %\newcommand{\Ec}{\mathcal{E}}

23: \newcommand{\Dc}{\mathcal{E}}

24: \newcommand{\Fc}{\mathcal{F}}

25: \newcommand{\s}{\mathrm{s}}

26: \newcommand{\w}{\mathrm{w}}

27: \newcommand{\ww}{\omega_{\mathrm{w}}}

28: \newcommand{\ws}{\omega_{\mathrm{s}}}

29: \newcommand{\Xw}{X_{\mathrm{w}}}

30: \newcommand{\Hw}{H_{\mathrm{w}}}

31: \newcommand{\Ab}{\mathcal{A}_{\bullet}}

32: \newcommand{\db}{\mathrm{d}_{\bullet}}

33: \newcommand{\I}{\mathcal{I}}

34: \newcommand{\Bb}{B_{\bullet}}

35: \newcommand{\wb}{{\omega}_{\bullet}}

36: \newcommand{\ddt}{\frac{d}{dt}}

37: \newcommand{\glim}{\widetilde{\lim} \, }

38: \newcommand{\ts}{\textstyle}

39: \def\setmarsing{

40: \oddsidemargin-0in

41: \evensidemargin-0in

42: \textwidth5.5in

43: \textheight8in

44: }

45: \setmarsing

46: 

47: \title{On global attractors of the 3D Navier-Stokes equations}

48: \author{A. Cheskidov\thanks{University of Michigan, Department of Mathematics,

49: Ann Arbor, MI 48109 (acheskid@umich.edu)} \ and

50: C. Foias\thanks{Texas A\&M University, Department of Mathematics,

51: College Station, TX 77843}}

52: \begin{document}

53: \maketitle

54: \abstract{

55: In view of the possibility that the 3D Navier-Stokes equations (NSE)  might

56: not always have regular solutions, we introduce an abstract

57: framework for studying the asymptotic behavior

58: of multi-valued dissipative evolutionary systems with respect to two

59: topologies---weak and strong. Each such system

60: possesses a global attractor in the weak topology, but not necessarily

61: in the strong. In case the latter exists and is weakly closed, it coincides

62: with the weak global attractor. We give a sufficient condition for

63: the existence of the strong global attractor, which is verified for the

64: 3D NSE when all solutions on the weak global attractor are strongly continuous.

65: We also introduce and study a two-parameter family of models for the

66: Navier-Stokes equations, with similar properties and open problems.

67: These models always possess weak global attractors, but on some of them

68: every solution blows up (in a norm stronger than the standard energy one) in finite time.

69: \\

70: \\

71: {\bf Keywords:} Navier-Stokes equations, global attractor, blow-up in finite time.

72: }

73: 

74: 

75: 

76: \section{Introduction}

77: A remarkable feature of many dissipative partial differential equations (PDEs)

78: is the existence of a global attractor to which all the solutions converge as

79: time goes to infinity \cite{SY,T2}. The global attractor $\A$

80: is the minimal closed set in

81: a phase space $H$ (i.e., the functional space, usually a Banach space,

82: in which the solutions exist) that uniformly attracts the trajectories starting

83: from any a priori given

84: bounded set in $H$. When the topology on $H$ is referred as the strong

85: (weak) topology of $H$, we will call $\A$ the strong (respectively weak)

86: global attractor.

87: 

88: It is possible that a dissipative PDE does not have a strong global attractor.

89: For instance, the 2D Navier-Stokes equations (NSE) on a bounded domain

90: $\Omega \subset \mathbb{R}$, when supplemented with appropriate boundary

91: conditions, possess a strong global attractor in $H$ (a certain subspace of

92: $L^2(\Omega)^3$) \cite{L,T2}, but it is not yet known

93: whether this holds for the 3D NSE.

94: 

95: Nevertheless, even for the 3D NSE one can prove that there exists

96: %a minimal weakly closed set in $H$ for which the attracting property holds

97: %with respect to the weak topology in $H$

98: a weak global attractor \cite{FT85}.

99: % Therefore, this attractor is referred to as the weak global attractor.

100: %For convenience, the first attractor mentioned above will be called the strong

101: %global attractor.

102: When the strong global attractor is strongly compact in $H$ (e.g., in the 2D NSE),

103: then it is also the weak global attractor. But, in any case, the weak global

104: attractor is an

105: appropriate generalization of the strong global attractor since it captures

106: the long-time behavior of the solutions. In particular, the support of any

107: time-average measure of the 3D NSE is included in the weak global attractor

108: (see \cite{FMRT}).

109: One should note that Sell \cite{S} introduced a related notion of a trajectory

110: attractor $\mathfrak{A}$ in the space of all trajectories, which was further

111: studied in \cite{CV, FS, S, SY}. The weak global attractor coincides with the

112: set of values of all trajectories in $\mathfrak{A}$ at any fixed time $t$.

113: 

114: The aim of this study is to present a general abstract framework which

115: is applicable to the 3D NSE even in the case where they do not

116: possess a strong global attractor. This framework may be also useful in

117: the study of other PDEs for which the existence of the strong global

118: attractor is in limbo. This aim forces us to consider multi-valued

119: evolutionary systems.

120: A number of papers have

121: been published concerning attractors of multi-valued semiflows. See

122: \cite{CMR} for a comparison of two canonical abstract frameworks by

123: Melnik and Valero \cite{MV} and Ball \cite{B1}. The main difference between

124: our evolutionary system $\Dc$ and Ball's generalized semiflow is that

125: we do not include

126: the hypotheses of concatenation and upper semicontinuity  with respect to

127: the initial data. This allows us to consider an evolutionary system whose

128: trajectories are all Leray-Hopf weak solutions of the 3D NSE.

129: 

130: Our definition of the evolutionary system $\Dc$ already exploits the effect of

131: dissipativity, namely the existence of an absorbing ball.

132: In fact, the space $X$ in which the trajectories of $\Dc$ live is, in applications,

133: precisely such an absorbing ball. Since in most applications the phase

134: space is a separable reflexive Banach space, both the strong and the

135: weak topologies on $X$ are metrizable. This is the motivation for us

136: to define strong and weak topologies on $X$ to be the ones induced by

137: appropriate metrics.

138: 

139:  We show

140: that every evolutionary system always possesses a weak global

141: attractor;  moreover, if the strong global attractor exists and is weakly closed,

142: then it has to coincide with the weak global attractor.

143: Note that some classical definitions (see, e.g., \cite{T2}) require a global

144: attractor to be an invariant set. We will see that under

145: a condition, which is, for example, satisfied by the Leray-Hopf weak solutions

146: of the 3D NSE, the weak global attractor is also the maximal bounded

147: invariant set. We recall that those solutions are always weakly continuous

148: in $L^2(\Omega)^3$.

149: 

150: It is known that if a weak global attractor for the 3D NSE is bounded in $V$,

151: then it is in fact strong \cite{FT85}. Moreover,

152: Ball \cite{B1} showed that if a generalized semiflow for a dissipative evolutionary

153: system is  asymptotically compact, then a strong global attractor exists.

154: This generalizes corresponding results for semiflows (see {\cite{H,HLS,L1}})

155: and implies that the strong global attractor for the 3D NSE exists

156: under the condition that all weak solutions are strongly continuous from $(0, \infty)$

157: to $L^2(\Omega)^3$ (see \cite{B1}). In this paper we show that even without the assumptions of concatenation and upper

158: semicontinuity  with respect to the initial data, the asymptotic compactness implies

159: that the weak global attractor is the minimal compact attracting set in the strong

160: metric, i.e., the weak global attractor is in fact the strong compact global attractor

161: (Theorem~\ref{t:asymptoticcompact}).

162: Applied to the 3D NSE, this result implies the existence of a strong compact global

163: attractor in the case when the solutions on the weak global attractor are

164: continuous in $L^2(\Omega)^3$ (Theorem~\ref{thm:main}).

165: 

166: The convergence of Leray-Hopf weak solutions was also studied by Rosa \cite{R}, 

167: namely, he introduced an asymptotic regularity condition

168: to insure the strong convergence of a weak solution towards its weak

169: $\omega$-limit. This condition requires the limit solutions to be strongly 

170: continuous in $L^2(\Omega)^3$, and implies that the weak global attractor is a

171: strongly compact strongly attracting set if the weak solutions on the weak global

172: attractor are strongly continuous in $L^2(\Omega)^3$.

173: Moreover, since a trajectory of the 3D NSE that is not strongly continuous

174: in  $L^2(\Omega)^3$ also, obviously, is not relatively strongly compact in $L^2(\Omega)^3$,

175: the strong continuity of weak solutions on the weak global

176: attractor $\Aw$ is also a necessary condition for $\Aw$ to be strongly compact

177: (see \cite{R}).

178: 

179: Recall that we define a global attractor as the minimal closed (uniformly)

180: attracting set in the corresponding metric, and hence, allow the possibility

181: for the solutions on a strong global attractor to be discontinuous.

182: We address this issue by

183: studying the weak convergence of weak solutions towards a weak solution strongly continuous from the right in $L^2(\Omega)^3$.  

184: We show that the weak convergence is strong under the

185: condition that the "energy jumps" of solutions converge to the "energy jumps"

186: of the limit solution (Theorem~\ref{t:deltat}). However, at this stage, no

187: necessary condition for the strong convergence is known.

188: 

189: 

190: Finally, we provide an example of a dissipative evolutionary system for

191: which all solutions on the weak global attractor blow-up in finite time.

192: We introduce a two-parameter family of simple infinite-dimensional

193: systems of differential equations.

194: These systems, called {\it tridiagonal models for the NSE (TNS models)},

195: display basic features of the NSE.

196: In particular, they are examples of dissipative systems that possess a weak

197: global attractor, but the existence of a strong global attractor is not

198: known.

199: 

200: TNS models have similar form and some similar properties to shell

201: models, specifically dyadic models studied in \cite{FP,KP}.

202: Moreover, a similar analysis of the dyadic models also results in a finite

203: time  blow-up (see \cite{C}).

204: In the TNS models though, the coefficients in the equations are chosen

205: to yield NSE-like scaling properties.

206: More precisely, we first mimic the Stokes operator in 3D via

207: the choice a positive definite operator on $l^2$ whose eigenvalues grow with the same speed as the eigenvalues of the Stokes operator. 

208: Second, we mimic the nonlinear term of the NSE via the choice of a bilinear

209: form, which scales like the Sobolev estimate for the NSE.

210: Then we obtain the following system of differential equations:

211: \begin{equation} \label{model-intro}

212: \ddt u + \nu Au + B(u,u) = g,

213: \end{equation}

214: where $u=(u_1,u_2,\dots)$,

215: \[

216: (Au)_n = n^{\alpha} u_n,

217: \]

218: and

219: \[

220: (B(u,v))_n=- n^\beta u_{n-1} v_{n-1} + (n+1)^\beta u_n v_{n+1},

221: \qquad n =1, 2, \dots,

222: \]

223: with $u_0=0$.

224: Here $\alpha$ and $\beta$ are two positive parameters. Note that

225: the orthogonality property in $l^2$ holds for $B$, which implies the existence

226: of an absorbing ball and a weak global attractor. Moreover, when

227: $\alpha = 2/3$, which corresponds to the speed with which the eigenvalues

228: of the Stokes operator grow in three-dimensional space, and $\beta=11/6$,

229: we have the following sharp estimate:

230: \[

231: | (B(u,u),Au) | \lesssim |Au |^{3/2} |A^{1/2}u|^{3/2},

232: \]

233: where

234: $

235: |v|^2=\sum v_n^2. 

236: $

237: This estimate is exactly the same as the the estimate based on the Sobolev

238: inequalities for the nonlinear term of the 3D NSE, with $|\cdot|$ being the $L^2$-norm.

239: It is an open question whether solutions of (\ref{model-intro})

240: can lose regularity in the case $(\alpha,\beta)=(2/3,11/6)$. However,

241: we show that in the nonviscous case $\nu=0$, for every $\alpha>0$, $\beta>0$,

242: $\gamma>0$,

243: and $g_n\geq 0$ for all $n\in\mathbb{N}$, the norm

244: $|A^{(\beta+\gamma-1)/(3\alpha)} u|$ of every solution

245: with $u_n(0)\geq 0$ and $u(0)\ne0$

246: blows up (Theorem~\ref{t:Eblow}).

247: 

248: When the viscosity is not zero, the model always possesses a weak global

249: attractor $\Aw$. Moreover, we prove that if the force $g$ is

250: large enough, then all solutions on $\Aw$  blow up in finite time in

251: an appropriate norm when $2\beta>3\alpha+3$ (Remark~\ref{rem:bl}).

252: The question whether $\Aw$ is a strong global attractor 

253: remains open in the case where $\beta > \alpha+1$.

254: 

255: 

256: 

257: \section{Evolutionary system and global attractors}

258: 

259: Let $(X,\ds(\cdot,\cdot))$ be a metric space endowed with

260: a metric $\ds$, which will be referred to as a strong metric.

261: Let $\dw(\cdot, \cdot)$ be another metric on $X$ satisfying

262: the following conditions:

263: \begin{enumerate}

264: \item $X$ is $\dw$-compact.

265: \item If $\ds(u_n, v_n) \to 0$ as $n \to \infty$ for some

266: $u_n, v_n \in X$, then $\dw(u_n, v_n) \to 0$ as $n \to \infty$,

267: that is, the identity map $(X, \ds) \mapsto (X,\dw)$ is uniformly continuous.

268: \end{enumerate}

269: Due to the latter property, $\dw$ will be referred to as the weak metric on $X$.

270: Note that any strongly compact ($\ds$-compact) set is weakly compact

271: ($\dw$-compact), and any weakly closed set is strongly closed.

272: Also it will be convenient to denote by $\overline{A}^{\bullet}$ the

273: closure of the set $A\subset X$ in the topology generated by $\db$; here and

274: throughout $\bullet$ stands for either $\mathrm{s}$ or $\mathrm{w}$.

275: 

276: To define an evolutionary system, first let

277: \[

278: \mathcal{T} := \{ I: \ I=[T,\infty) \mbox{ for some } T\in \mathbb{R}, \mbox{ or } 

279: I=(-\infty, \infty) \},

280: \]

281: and for each $I \subset \mathcal{T}$ let $\mathcal{F}(I)$ denote

282: the set of all $X$-valued functions on $I$.

283: Now we define an evolutionary system $\Dc$ as follows.

284: \begin{definition} \label{d:E}

285: A map $\Dc$ that associates to each $I\in \mathcal{T}$ a subset

286: $\Dc(I) \subset \mathcal{F}$ will be called an evolutionary system if

287: the following conditions are satisfied:

288: \begin{enumerate}

289: \item $\Dc([0,\infty))\ne \emptyset$.

290: \item

291: $\Dc(I+s)=\{u(\cdot): \ u(\cdot -s) \in \Dc(I) \}$ for

292: all $s \in \mathbb{R}$.

293: \item $\{u(\cdot)|_{I_2} : u(\cdot) \in \Dc(I_1)\}

294: \subset \Dc(I_2)$ for all

295: pairs of $I_1,I_2 \in \mathcal{T}$, such that $I_2 \subset I_1$.

296: \item

297: $\Dc((-\infty , \infty)) = \{u(\cdot) : \ u(\cdot)|_{[T,\infty)}

298: \in \Dc([T, \infty)) \ \forall T \in \mathbb{R} \}.$

299: \end{enumerate}

300: \end{definition}

301: We will refer to $\Dc(I)$ as the set of all trajectories (solutions)

302: on the time interval $I$. Let $P(X)$ be the set of all subsets of $X$.

303: For every $t \geq 0$ define a map

304: \begin{eqnarray*}

305: &R(t):P(X) \to P(X),&\\

306: &R(t)K := \{u(t): u(0)\in K, u(\cdot) \in \Dc([0,\infty))\}, \qquad

307: K \subset X.&

308: \end{eqnarray*}

309: Note that the assumptions on $\Dc$ imply that $R(t)$ enjoys

310: the following property:

311: \begin{equation} \label{eq:propR(T)}

312: R(t+s)K \subset R(t)R(s)K, \qquad K \subset X,\quad t,s \geq 0.

313: \end{equation}

314: 

315: Let us first point out that the trajectories are not required to be continuous

316: (even with respect to $\dw$) nor are they uniquely determined by their

317: starting points, i.e.,  it is possible to have two different

318: trajectories $u,v \in \Dc([0,\infty))$, such that $u(0)=v(0)$. Second, there is no assumption of concatenation. If $u \in \Dc([0,\infty))$ and

319: $v \in \Dc([T,\infty))$ for some $T>0$, so that

320: $u(T)=v(T)$, then the following function

321: \[

322: w(t)= \left\{

323: \begin{aligned}

324: u(t), \qquad &\mbox{if } t\in[0,T],\\

325: v(t), \qquad &\mbox{if } t\in(T,\infty).

326: \end{aligned}

327: \right.

328: \]

329: need not be in $\Dc([0,\infty))$.

330: We avoid the assumptions of the continuity,

331: uniqueness, and concatenation in order to be able to consider

332: an evolutionary system consisting of Leray-Hopf weak solutions to the

333: 3D Navier-Stokes equations.

334: 

335: Often an evolution of a dynamical system can be described by

336: a semigroup of continuous mappings acting on some metric space $H$:

337: \[

338: S(t): H \to H, \quad t \geq 0.

339: \]

340: The semigroup properties are the following:

341: \begin{equation} \label{eq:semigroup}

342: S(t+s) = S(t)S(s), \quad t,s \geq 0, \qquad S(0)= \mbox{Identity operator.}

343: \end{equation}

344: Then, if $u(t)\in H$ represents a state of the dynamical system at 

345: time $t$, we have

346: \[

347: u(t+s) = S(t)u(s), \qquad t,s \geq 0.

348: \]

349: A ball $B\subset H$ is called an absorbing ball, if for any bounded

350: set $K \subset H$ there exists $t_0$, such that

351: \[

352: S(t)K \subset B, \qquad \forall t \geq t_0.

353: \]

354: Hence, if we are interested in the long-time behavior of the dynamical system,

355: it is enough to consider a restriction of the system to an absorbing ball.

356: So, we let $X$ be a closed absorbing ball, and

357: define the map $\Dc$

358: in the following way:

359: \[

360: \Dc(I) := \{u(\cdot): u(t+s)=S(t)u(s) \mbox{ and }

361: u(s) \in X \ \forall s\in I, \ t \geq 0 \}.

362: \]

363: Note that conditions 1--4 for the evolutionary system $\Dc$ automatically follow

364: from the semigroup properties (\ref{eq:semigroup}) of $S(t)$.

365: In addition, let $T$ be such that

366: \[

367: S(t)X \subset X \qquad \forall t \geq T.

368: \]

369: Then we have

370: \[

371: R(t)K = S(t)K, \qquad \forall K \subset S(T)X, \ t \geq 0.

372: \]

373: The 3D Navier-Stokes equations will serve as an instructive illustration and 

374: application  of our consideration. 

375: As yet in our knlowledge of the 3D NSE, the time evolution of the 3D NSE cannot

376: be described by a semigroup of maps. Therefore, for the 3D NSE we will have a

377: more involved definition of $\Dc$ (see Section~\ref{3DNSE}).

378: 

379: 

380: Having defined the evolutionary system $\Dc$, we proceed to define attracting

381: sets and global attractors.

382: For $A \subset X$ and $r>0$, denote

383: $

384: B_{\bullet}(A,r) = \{u: \ \db(u, A) < r\},

385: $

386: where

387: \[

388: \db(u, A):=\inf_{x\in A}\db(u,x).

389: \]

390: 

391: 

392: \begin{definition}

393: A set $A \subset X$ is a $\mathrm{d}_{\bullet}$-attracting set

394: ($\bullet = \mathrm{s,w}$) if it uniformly

395: attracts $X$ in $\mathrm{d}_{\bullet}$-metric, i.e.,

396: for any $\epsilon>0$, there exists

397: $t_0$, such that

398: \[

399: R(t)X \subset B_{\bullet}(A, \epsilon), \qquad \forall t \geq t_0.

400: \]

401: \end{definition}

402: 

403: \begin{definition}

404: $\mathcal{A}_{\bullet}\subset X$ is a

405: $\db$-global attractor ($\bullet = \mathrm{s,w}$) if

406: $\mathcal{A}_{\bullet}$ is a minimal $\db$-closed

407: $\db$-attracting  set, i.e., $\mathcal{A}_{\bullet}$

408: is $\db$-closed $\db$-attracting

409: and every subset $A \subset \Ab$ that is also $\db$-closed and $\db$-attracting

410: satisfies $A=\Ab$.

411: \end{definition}

412: Note that the empty set is never an attracting set.

413: Note also that since $X$ is not strongly compact, the intersection of two 

414: $\ds$-closed $\ds$-attracting sets might not be $\ds$-attracting.

415: Nevertheless, the uniqueness of a global attractor is a direct consequence

416: of the following lemma.

417: 

418: 

419: \begin{lemma}

420: If $\mathcal{A}_{\bullet}$ exists and $A$ is a $\db$-closed $\db$-attracting set,

421: then $\Ab \subset A$ ($\bullet = \mathrm{s,w}$).

422: \end{lemma}

423: \begin{proof}

424: Let $A$ be an arbitrary $\db$-closed $\db$-attracting set.

425: Take any point $a \in \Ab$. Let $\epsilon>0$. If there exists $t_\epsilon>0$, such that 

426: \[

427: R(t)X \cap \Bb(a,\epsilon) = \emptyset, \qquad \forall t\geq t_\epsilon,

428: \]

429: then $\Ab \setminus \Bb(a,\epsilon/2)$ is a $\db$-closed $\db$-attracting

430: set contradicting the minimality of $\Ab$. 

431: So, there exists a sequence $t_n \to \infty$ as $n\to \infty$,

432: such that

433: \[

434: R(t_n)X \cap \Bb(a,\epsilon) \ne \emptyset, \qquad \forall n.

435: \]

436: On the other hand, since $A$ is $\db$-attracting, we infer that

437: \[

438: R(t_n)X \subset \Bb(A,\epsilon),

439: \]

440: for $n$ large enough. It follows that

441: \[

442: A \cap \Bb(a,2\epsilon) \ne \emptyset.

443: \]

444: Since $A$ is $\db$-closed, we have that $a \in A$.  Thus,

445: $\Ab \subset A$.

446: \end{proof}

447: As a direct consequence of this lemma we have the following.

448: 

449: \begin{corollary}

450: If $\mathcal{A}_{\bullet}$ exists, then it is unique

451: ($\bullet = \mathrm{s,w}$). 

452: \end{corollary}

453: Assume now that the weak global attractor $\Aw$ exists. If $\Aw$ is

454: a strongly attracting set, does it follow that the strong global attractor

455: exists? In general, this may not be true. However, if the strong global attractor

456: exists, then clearly it is a $\dw$-attracting set. Moreover, we have the following.

457: 

458: \begin{theorem} \label{thm:coinside}

459: If $\As$ exists, then $\Aw$ exists and

460: \[

461: \Aw=\overline{\As}^{\w}.

462: \]

463: \end{theorem}

464: \begin{proof}

465: If there exists

466: a $\dw$-closed $\dw$-attracting set $A\subset \overline{\As}^{\mathrm{w}}$ and

467: $A\ne \overline{\As}^{\mathrm{w}}$, then there exists

468: $u_0 \in \As$, such that

469: \[

470: d=\dw(u_0, A)>0.

471: \]

472: By the definition of an attracting set, there exists a time $t_0>0$, such that

473: \begin{equation} \label{e:111}

474: R(t)X \subset B_{\mathrm{w}}(A, d/2) \qquad \forall t\geq t_0.

475: \end{equation}

476: Note that

477: \[

478: \dw(u_0, B_{\mathrm{w}}(A, d/2)) \geq d/2.

479: \]

480: Therefore, by virtue of Property 2 in the definition of $\dw$, there

481: exists $\delta>0$, such that

482: \[

483: \ds(u_0, B_{\mathrm{w}}(A, d/2)) > \delta,

484: \]

485: whence,

486: \[

487: B_{\mathrm{s}}(u_0, \delta) \cap B_{\mathrm{w}}(A, d/2) = \emptyset.

488: \]

489: Now from \eqref{e:111} it follows that

490: \[

491: B_{\mathrm{s}}(u_0, \delta)\cap R(t)X =\emptyset \qquad \forall t \geq t_0.

492: \]

493: Consequently, $\As \setminus B_{\mathrm{s}}(u_0, \delta/2)$ is a $\ds$-closed 

494: $\ds$-attracting set strictly included in $\As$, a contradiction.

495: Hence, $\overline{\As}^{\mathrm{w}}$ is the weak global attractor.

496: \end{proof}

497: 

498: 

499: The following are two simple examples of evolutionary systems

500: that possess a weak global attractor $\Aw$, but not a strong global

501: attractor $\As$.

502: 

503: \noindent

504: \textbf{Example 1.}

505: Let

506: \[

507: X=\left\{ u \in L^2(-\infty,\infty): \int_{-\infty}^\infty u(x)^2 \, dx \leq 1,

508: u(x)=0 \mbox{ for } x> 0\right\},

509: \]

510: and define on $X$ the distances

511: \[

512: \ds(u,v):=\left( \int_{-\infty}^\infty(u(x)-v(x))^2 \, dx\right)^{1/2}, \qquad

513: \dw(u,v)=\int_{-\infty}^\infty \frac{1}{2^{|x|}}

514: \frac{|u(x)-v(x)|}{1 + |u(x)-v(x)|} \, dx.

515: \]

516: Consider the following partial differential equation:

517: \[

518: \frac{\partial u}{\partial t} = \frac{\partial u}{\partial x}.

519: \]

520: The trajectories of the evolutionary system $\Dc$

521: will be solutions of this equation, i.e.,

522: \[

523: \begin{aligned}

524: &\Dc([s,\infty)) =\{u\in \Fc([s,\infty)) : u(t) =u_0(\cdot+t-s), t\in[s,\infty), u_0 \in X\}, \qquad

525: \forall s \in \mathbb{R},\\

526: & \Dc ((-\infty,\infty))=\{0\}.

527: \end{aligned}

528: \]

529: Then it is easy to show that $\Aw=\{0\}$ (see also Theorem~\ref{c:weakA}). However, no trajectory except the trivial one $u=0$ strongly

530: converges to $0$ as $t\to \infty$.

531: 

532: \noindent

533: \textbf{Example 2.}

534: Take

535: \[

536: X=\left\{ u\in l^2: \sum_{n=1}^\infty u_n^2  \leq 1\right\}.

537: \]

538: where $u=(u_n)$, $u_n \in \mathbb{R}$ for all $n$.  For $u, v \in X$, let

539: \[

540: \ds(u,v):=\left( \sum_{n=1}^\infty(u_n-v_n)^2 \right)^{1/2}, \qquad

541: \dw(u,v)= \sum_{n=1}^\infty \frac{1}{2^n}

542: \frac{|u_n-v_n|}{1 + |u_n-v_n|}.

543: \]

544: Consider the following differential equation:

545: \[

546: \ddt u_n = -\frac{1}{n} u_n, \qquad n \in \mathbb{N}

547: \]

548: The trajectories of the evolutionary system $\Dc$

549: will be solutions of this equation, i.e.,

550: \[

551: \Dc([s,\infty)) =\{u\in\Fc([s,\infty)): (u_n(t))=(u_n^0e^{(s-t)/n}), t\in[s,\infty),

552: u^0 \in X\},

553: \]

554: for $s \in \mathbb{R}$, and

555: \[

556: \Dc((-\infty,\infty))=\{0\}.

557: \]

558: Take any $u^0 \in X$ and consider the trajectory $u \in \Dc([0,\infty))$ starting

559: at $u^0$, i.e., $u(0)=u^0$. Then we have

560: \[

561: \ds(u(t),0)^2=\sum_{n=1}^\infty u_n(t)^2 = \sum_{n=1}^\infty(u^0_n)^2 e^{-\frac{2t}{n}}

562: \to 0 \qquad \mbox{as} \qquad t \to \infty.

563: \]

564: However, the convergence is not uniform in the $\ds$-metric,

565: although it is uniform in the $\dw$-metric. So

566: again $\Aw=\{0\}$, but $\As$ does not exist.

567: 

568: Note that the nonexistence of $\As$ in the two examples is due to

569: two different behaviors of the trajectories.

570: In the first example all the nontrivial trajectories converge

571: to $\Aw$ weakly, but not strongly. In the second example all the nontrivial

572: trajectories converge to $\Aw$ strongly, but not uniformly.

573: 

574: \begin{definition} \label{d:ucomp}

575: The map $R(t)$ is uniformly $\db$-compact

576: ($\bullet= \mathrm{s, w}$) if

577: there exists $t_0\geq 0$, such that

578: \[

579: \bigcup_{t \geq t_0} R(t) X

580: \]

581: is relatively $\db$-compact.

582: \end{definition}

583: 

584: Note that since $X$ is $\dw$-compact, 

585: $R(t)$ is automatically uniformly $\dw$-compact.

586: 

587: \begin{definition} The $\wb$-limit ($\bullet= \mathrm{s, w}$)

588: of a set $K \subset X$ is

589: \[

590: \wb (K):=\bigcap_{T\geq 0} \overline{\bigcup_{t \geq T} R(t) K}^{\bullet},

591: \]

592: where the closure is taken in $\db$-metric.

593: \end{definition}

594: 

595: \begin{lemma} \label{l:wbinA}

596: Let $A$ be a $\db$-closed $\db$-attracting set. Then

597: \[

598: \wb(X) \subset A.

599: \]

600: \end{lemma}

601: \begin{proof}

602: Suppose that there exists $a \in \wb(X) \setminus A$.

603: Since $A$ is $\db$-closed, there exists $\epsilon>0$, such that

604: \[

605: A \cap \Bb(a,\epsilon) = \emptyset.

606: \]

607: By the definition of the $\wb$-limit,

608: there exist a sequence $t_n \to \infty$ as $n \to \infty$ and

609: a sequence $x_n \in R(t_n)X$, such that $\db(x_n,a) \to 0$

610: as $n \to \infty$. Hence, there exists $N>0$, such that 

611: \[

612: x_n \notin \Bb(A, \epsilon/2), \qquad \forall n\geq N.

613: \]

614: This means that $A$ is not $\db$-attracting, a contradiction.

615: \end{proof}

616: 

617: \begin{lemma} \label{l:oncompactsemigroup}

618: If the map $R(t)$ is uniformly $\db$-compact

619: ($\bullet= \mathrm{s, w}$), then 

620: $\wb(X)$ is a nonempty $\db$-compact $\db$-attracting set.

621: \end{lemma}

622: \begin{proof}

623: By Definition~\ref{d:ucomp} and the fact that $\Dc([0,\infty)) \ne \emptyset$,

624: there exists $t_0$, such that

625: \[

626: W(T):=\overline{\bigcup_{t \geq T} R(t) X}^{\bullet}

627: \]

628: is a nonempty $\db$-compact set for all $T\geq t_0$.

629: In addition, $W(s) \subset W(t)$ for all $s\geq t \geq 0$. Thus,

630: \[

631: \wb(X)= \bigcap_{T\geq t_0} W(T)

632: \]

633: is a nonempty  $\db$-compact set.

634: 

635: We will now prove that $\wb (X)$ uniformly $\db$-attracts $X$. Assume it

636: does not. Then there exists $\epsilon>0$, such that

637: \[

638: V(t):=W(t)\cap(X\setminus \Bb(\wb(X), \epsilon)) \ne \emptyset, \qquad \forall t\geq 0.

639: \]

640: Since $V(t)$ is $\db$-compact for $t\geq t_0$ and $V(s) \subset V(t)$ for

641: $s\geq t \geq 0$, we have that there exists

642: \[

643: x \in \bigcap_{t\geq t_0} V(t).

644: \]

645: Hence, $x \in \wb(X)$. However, this implies that $x\notin V(t)$, $t\geq 0$,

646: a contradiction.

647: \end{proof}

648: 

649: 

650: 

651: 

652: \begin{theorem} \label{thm:exofA}

653: If the map $R(t)$ is uniformly $\db$-compact

654: ($\bullet= \mathrm{s, w}$), then the  $\db$-global

655: attractor exists and satisfies the following additional properties:

656: \begin{enumerate}

657: \item[(a)]

658: $\Ab = \wb(X)$.

659: \item[(b)]

660: $\Ab$ is $\db$-compact.

661: \end{enumerate}

662: \end{theorem}

663: \begin{proof}

664: By Lemma~\ref{l:oncompactsemigroup},

665: $\wb(X)$ is a nonempty $\db$-compact $\db$-attracting set.

666: Moreover, $\wb(X)$ is the

667: minimal $\db$-closed $\db$-attracting set due to Lemma~\ref{l:wbinA}.

668: Therefore, $\wb(X)$ is the $\db$-global attractor.

669: \end{proof}

670: 

671: Note that since $X$ is $\dw$-compact (see the definition of $\dw$), $R(t)$ is uniformly weakly compact.

672: Hence we have the following.

673: \begin{corollary}

674: The evolutionary system $\Dc$ always possesses a weak global attractor $\Aw$.

675: \end{corollary}

676: Our next goal is to investigate whether $\Aw$ is an invariant set in the following

677: sense.

678: 

679: \begin{definition} The set $A \subset X$ is invariant, if

680: \[

681: \{u(t): u\in \Dc((-\infty,\infty)), u(0) \in A \}= A, \qquad \forall t\geq 0.

682: \]

683: \end{definition}

684: Assume that $u_0$ belongs to some invariant set. Then for all $t>0$

685: we have that $u_0 \in R(t)X$.

686: Hence, $u_0 \in \omega_{\mathrm{w}}(X)=\Aw$. Therefore, $\Aw$ contains every invariant set. Moreover, we will show that $\Aw$ is invariant under some compactness

687: property that is for instance satisfied by the family of all 

688: Leray-Hopf solutions of the 3D NSE (see Section~\ref{3DNSE}).

689: 

690: Let $C([a, b];X_\bullet)$ be the space of $\db$-continuous $X$-valued

691: functions on $[a, b]$ endowed with the metric

692: \[

693: \dd_{C([a, b]; X_\bullet)}(u,v) = \sup_{t\in[a,b]}\db(u(t),v(t)). 

694: \]

695: Let also $C([a, \infty);X_\bullet)$ be the space of $\db$-continuous

696: $X$-valued functions on $[a, \infty)$

697: endowed with the metric

698: \[

699: \dd_{C([a, \infty); X_\bullet)}(u,v) = \sum_{T\in \mathbb{N}} \frac{1}{2^T} \frac{\sup\{\db(u(t),v(t)):a\leq t\leq a+T\}}

700: {1+\sup\{\db(u(t),v(t)):a\leq t\leq a+T\}}.

701: \]

702: 

703: \begin{theorem} \label{c:weakA}

704: If $\Dc([0,\infty))$ is compact in $C([0, \infty);\Xw)$, then

705: \begin{enumerate}

706: \item[(a)]

707: $

708: \Aw = \mathcal{I}:=\{ u_0: \ u_0=u(0) \mbox{ for some }

709: u \in \Dc((-\infty, \infty))\}.

710: $

711: \item[(b)] $\Aw$ is the maximal invariant set.

712: \end{enumerate}

713: \end{theorem}

714: \begin{proof}

715: Since obviously $\I$ is the maximal invariant set, we have that $\I \subset \Aw$.

716: It remains to prove that $\Aw \subset \I$. Take any $a \in \Aw$.

717: Since $\Aw = \omega_{\mathrm{w}}(X)$,  there exist

718: $t_n \to \infty$, as $n\to \infty$ and $a_n\in R(t_n)X$, such that

719: $a_n \to a$ weakly as $n \to \infty$. 

720: Using Property 2 in Definition~\ref{d:E}, there exist

721: $u_n \in \Dc([-t_n, \infty))$, such that

722: $u_n(0) =a_n$. Also, Properties 2 and 3 in Definition~\ref{d:E}

723: of $\Dc$ imply that

724: $\Dc([-t_n,\infty))$ is compact in $C([-t_n, \infty);\Xw)$

725: and

726: \[

727: \{u|_{[-t_1,\infty)} : u \in \Dc([-t_{n},\infty))\}

728: \subset \Dc([-t_1,\infty)),

729: \]

730: for every $n$.

731: Now, passing to a subsequence and dropping a subindex, we can assume

732: that $u_n|_{[-t_1,\infty)} \to u^1\in \Dc([-t_1, \infty))$ in  $C([-t_1, \infty);\Xw)$

733: as $n\to \infty$. 

734: By a standard diagonalization process we obtain that there exist

735: $u \in  \Fc((-\infty,\infty))$ and a subsequence of $u_n$, still denoted by

736: $u_n$, such that 

737: $u_n|_{[-T,\infty)} \to u_{[-T,\infty)}$ in $C([-T, \infty);\Xw)$ for all $T>0$.

738: Thus, by the compactness we have that $u|_{[-T,\infty)} \in \Dc([-T, \infty))$

739: for all $T>0$, and hence $u \in \Dc((-\infty, \infty))$.

740: Finally, since $u(0)=a$. we have $a \in \mathcal{I}$.

741: Hence, $\Aw\subset \I$.

742: \end{proof}

743: 

744: Now that we know that the weak global attractor always exists, we can

745: weaken the condition on the existence of the strong global attractor. 

746: 

747: \begin{definition}

748: The map $R(t)$ is asymptotically $\db$-compact

749: ($\bullet= \mathrm{s, w}$) if for any $t_n \to \infty$ and

750: any $x_n \in R(t_n) X$, the  sequence

751: $\{x_n\}$ is relatively $\db$-compact.

752: \end{definition}

753: 

754: 

755: \begin{theorem} \label{t:asymptoticcompact}

756: If the map $R(t)$ is asymptotically  $\ds$-compact, then

757: $\Aw$ is $\ds$-compact strong global attractor.

758: \end{theorem}

759: \begin{proof}

760: First note that $\ws(X)\subset \ww(X)=\Aw$. On the other hand,

761: let $a \in \Aw=\ww(X)$. By

762: the definition of $\ww$-limit, there exist $t_n \to \infty$ as $n \to \infty$ and

763: $x_n \in R(t_n)X$, such that 

764: \[

765: \dw(x_n, a) \to 0 \qquad \mbox{ as } \qquad n \to \infty.

766: \]

767: Thanks to the asymptotic compactness of $R(t)$, this convergence

768: is in fact strong. Therefore, $a\in \ws(X)$. Hence, $\ws(X)= \Aw$.

769: 

770: Now let us show that $\ws(X)$ is a $\ds$-attracting set. Assume that it is not. Then there

771: exist $\epsilon>0$, $x_n \in X$, and $t_n \to \infty$ as

772: $n\to \infty$, such that

773: \[

774: x_n \in R(t_n)X \setminus B_{\s}(\ws(X), \epsilon), \qquad

775: \forall n \in \mathbb{N}.

776: \]

777: Since  $R(t)$ is asymptotically $\ds$-compact, then

778: $\{x_n\}$ is relatively $\ds$-compact.

779: Passing to a subsequence and dropping a subindex, we may assume that

780: \[

781: x_n \to x \in X \qquad \mbox{strongly},

782:  \qquad \mbox{as} \ n \to \infty.

783: \]

784: Therefore, we have that $x  \in \omega_{\mathrm{s}}(X)$,

785: a contradiction.

786: 

787: Now note that $\ws(X)$ is the

788: minimal $\ds$-closed $\ds$-attracting set due to Lemma~\ref{l:wbinA}.

789: Therefore, $\ws(X)$ is the strong global attractor $\As$.

790: Finally, let us show that $\ws(X)$

791: is strongly compact. Take any sequence $a_n \in \ws(X)$. By the definition of

792: $\ws$-limit, there exist $t_n \to \infty$ and

793: $x_n \in R(t_n)X$, such that 

794: \[

795: \ds(x_n, a_n) \to 0 \qquad \mbox{ as } \qquad n \to \infty.

796: \]

797: Note that $\{x_n\}$ is relatively $\ds$-compact due to the asymptotic compactness

798: of $R(t)$. Hence, $\{a_n\}$ is relatively $\ds$-compact, which concludes the proof.

799: 

800: \end{proof}

801: 

802: Finally, in the following example we show that the asymptotic compactness

803: is a weaker condition than the uniform strong compactness.

804: 

805: \noindent

806: \textbf{Example.}

807: Take

808: \[

809: X=\left\{ u\in l^2: \sum_{n=1}^\infty u_n^2  \leq 1\right\}.

810: \]

811: where $u=(u_n)$, $u_n \in \mathbb{R}$ for all $n$.  For $u, v \in X$, let

812: \[

813: \ds(u,v):=\left( \sum_{n=1}^\infty(u_n-v_n)^2 \right)^{1/2}, \qquad

814: \dw(u,v)= \sum_{n=1}^\infty \frac{1}{2^n}

815: \frac{|u_n-v_n|}{1 + |u_n-v_n|}.

816: \]

817: Consider the following differential equations:

818: \[

819: \ddt u_1=0,

820: \]

821: and

822: \[

823: \ddt u_n = - u_n, \qquad n \in \mathbb{N}.

824: \]

825: The trajectories of the evolutionary system $\Dc$ will be solutions of this

826: equation, i.e.,

827: \[

828: \begin{split}

829: \Dc([s,\infty)) =\{&u\in\Fc([s,\infty)): u_1(t)=u^0_1, u_n(t)=u_n^0e^{(s-t)}

830: \mbox{ for } n\geq 2,\\ & t\in[s,\infty), u^0 \in X\},

831: \end{split}

832: \]

833: for $s \in \mathbb{R}$, and

834: \[

835: \Dc((-\infty,\infty))=\{u: u_1\in [-1,1], u_n=0, n\geq 2\}.

836: \]

837: Clearly, $R(t)$ is not uniformly strongly compact, but it is

838: asymptotically compact. Hence, the strong global attractor

839: exists and

840: \[

841: \As = \{u: u_1\in [-1,1], u_n=0, n\geq 2\}.

842: \]

843: 

844: 

845: Finally,

846: Theorems~\ref{c:weakA} and \ref{t:asymptoticcompact}

847: imply the following

848: \begin{remark} 

849: If $\Dc([0,\infty))$ is compact in $C([0, \infty);\Xw)$ and $R(t)$ is asymptotically  

850: $\ds$-compact, then  the strong global attractor $\As$ exists, and is the strongly compact

851: maximal invariant set. Consequently, $\As$ is a compact global attractor in the

852: conventional sense.

853: \end{remark}

854: 

855: 

856: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

857: \section{3D Navier-Stokes equations} \label{3DNSE}

858: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

859: 

860: Here we apply the results from the previous section 

861: to the space periodic 3D Navier-Stokes equations (NSE)

862: \begin{equation} \label{NSE1}

863: \left\{

864: \begin{aligned}

865: &\ddt u - \nu \Delta u + (u \cdot \nabla)u + \nabla p = f,\\

866: &\nabla \cdot u =0,\\

867: &u, p, f \mbox{ are periodic with period } L \mbox{ in each space variable,}\\

868: &u, f \mbox{ are in } L^2_{\mathrm{loc}}(\mathbb{R}^3)^3,\\

869: &u|_{t=0}=u_0,

870: \end{aligned}

871: \right.

872: \end{equation}

873: where $u$, the velocity, and $p$, the pressure, are unknowns, $f$ is

874: a given driving force, and $\nu>0$ is the kinematic  viscosity coefficient

875: of the fluid. By a Galilean change of

876: variables, we can assume that the space average of $u$ is

877: zero, i.e.,

878: \[

879: \int_\Omega u(x,t) \, dx =0, \qquad \forall t,

880: \]

881: where $\Omega=[0,L]^3$ is a periodic box.

882: 

883: In this section we will apply our general results from the previous section

884: to the study of the assymptotical behaviour of weak solutions to \eqref{NSE1}.

885: 

886: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

887: \subsection{Functional setting} \label{funcset}

888: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

889: 

890: First, let us introduce some notations and functional setting for \eqref{NSE1}.

891: %Let $\Omega:=[0,L]^3$ be a periodic box.

892: We denote by

893: $(\cdot,\cdot)$ and $|\cdot|$ the $L^2(\Omega)^3$-inner product and the

894: corresponding $L^2(\Omega)^3$-norm.

895: Let $\mathcal{V}$ be the space of all $\mathbb{R}^3$ trigonometric polynomials of

896: period $L$ in each variable satisfying 

897: $\nabla \cdot u =0$ and $\int_\Omega u(x) \, dx =0$.

898: Let $H$ and $V$ to be the

899: closures of $\mathcal{V}$ in $L^2(\Omega)^3$ and $H^1(\Omega)^3$, respectively.

900: Also, define the distances $\db$ by

901: \[

902: \ds(u,v):=|u-v|, \qquad

903: \dw(u,v)= \sum_{\kappa \in \mathbb{Z}^3} \frac{1}{2^{|\kappa|}}

904: \frac{|u_{\kappa}-v_{\kappa}|}{1 + |u_{\kappa}-v_{\kappa}|},

905: \qquad u,v \in H,

906: \]

907: where $u_{\kappa}$ and $v_{\kappa}$ are Fourier coefficients of $u$

908: and $v$ respectively.

909: 

910: We denote by $P_{\sigma} : L^2(\Omega)^3 \to H$ the $L^2$-orthogonal

911: projection, referred to as the Leray projector, and by

912: $A=-P_{\sigma}\Delta = -\Delta$ the Stokes operator with the domain

913: $D(A)=(H^2(\Omega))^3 \cap V$. The Stokes operator is a self-adjoint

914: positive operator with a compact inverse.

915: Denote

916: \[

917: \|u\| := |A^{1/2} u| = \left(\int_\Omega \sum_{i,j=1}^3\left|\frac{\partial u_i}{\partial x_j}\right|^2 \, dx \right)^{1/2}.

918: \]

919: Note that $\|u\|$ is equivalent to the $H^1$-norm of $u$ for $u\in D(A^{1/2})$.

920: 

921: For a rigorous mathematical study of  the equation \eqref{NSE1} we need

922: a few more concepts from functional analysis. Namely, let $V'$ be the set

923: of all distributions of the form $v=\Delta u$, with $u\in V$. The $V'$-norm of

924: this $v$ is by definition $\|u\|$. Endowed with this norm, $V'$ becomes

925: the dual space of $V$, and if $v\in H$, then the value of $v$ at a point $w\in H$

926: equals to the usual scalar product $(v,w)$ in $H$. Now for $u$ and $v$ in $V$,

927: let $B(u,v):=P_{\sigma}(u \cdot \nabla v)$, which is an element of $V'$. If

928: $v \in \mathcal{D}(A)$, then $B(u,v) \in H$. Moreover, 

929: \[

930: \langle B(u,v),w \rangle =- \langle B(u,w),v \rangle, \qquad u,v,w \in V,

931: \]

932: in particular, $\langle B(u,v),v \rangle=0$ for all $u,v \in V$.

933: 

934: Equations (\ref{NSE1}) now can be condensed in the functional

935: differential equation

936: \begin{equation} \label{NSE}

937: \ddt u + \nu Au +B(u,u) = g \qquad \mbox{in} \qquad  V',

938: \end{equation}

939: where $u$ is a $V$-valued function of time and $g = P_{\sigma} f$.

940: Throughout, we will assume that  $g$ is time independent and $g\in H$.

941: 

942: \begin{definition}

943: A weak solution  of \eqref{NSE} on $[T,\infty)$ is an $H$-valued

944: function $u(t)$ defined for $t \in [T, \infty)$, such that

945: \[

946: u(\cdot) \in C([T, \infty); \Hw) \cap 

947: L_{\mathrm{loc}}^2([T, \infty); V),

948: \]

949: and

950: \begin{equation} \label{333}

951: w(\cdot):=g-\nu A u(\cdot) - B(u(\cdot),u(\cdot))\in L_{\mathrm{loc}}^1([T, \infty); V'),

952: \end{equation}

953: \begin{equation} \label{444}

954: \begin{split}

955: (u(t)-u(T),v)&= \langle u(t)-u(T),v \rangle\\

956: &=\int_T^t\langle w(s),v\rangle \, ds \qquad

957: \forall v \in V, t\geq T.

958: \end{split}

959: \end{equation}

960: \end{definition}

961: The relations \eqref{333} and \eqref{444} imply that

962: $\ddt u$ exists in $V'$ a.e. in $[T,\infty)$. Therefore, often in

963: the literature \eqref{444} and \eqref{333} are written as \eqref{NSE} and

964: \[

965: \ddt u \in L_{\mathrm{loc}}^1([T, \infty); V'),

966: \] 

967: respectively.

968: 

969: The classical fundamental result concerning \eqref{NSE} is the following.

970: 

971: \begin{theorem}[Leray, Hopf] \label{thm:Leray}

972: For every $u_0 \in H$,

973: there exists a weak solution $u(t)$ of (\ref{NSE}) on $[T,\infty)$ with $u(T)=u_0$

974: satisfying the following energy inequality: 

975: \begin{equation} \label{EI}

976: |u(t)|^2 + 2\nu \int_{t_0}^t \|u(s)\|^2 \, ds \leq

977: |u(t_0)|^2 + 2\int_{t_0}^t (g(s), u(s)) \, ds

978: \end{equation}

979: for all $t \geq t_0$, $t_0$ a.e. in $[T,\infty)$.

980: \end{theorem}

981: See \cite{K} for a hypothesis under which the energy equality holds. However,

982: in general, the existence of weak solutions satisfying the energy equality is

983: not known.  Therefore, we introduce the following definition. 

984: \begin{definition} \label{d:ex}

985: A Leray-Hopf solution of the \eqref{NSE} on the interval $[T, \infty)$

986: is a weak solution on $[T,\infty)$ satisfying the

987: energy inequality (\ref{EI}) for all $T \leq t_0 \leq t$,

988: $t_0$ a.e. in $[T,\infty)$. The set $Ex$ of measure $0$ of points $t_0$ for

989: which the energy

990: inequality does not hold will be called the exceptional set (of the solution).

991: In addition, the solution $u(t)$ will be called regular on an interval $(\alpha, \beta) \subset [T,\infty)$ if $u(t)\in V$ and $\|u(t)\|$ is continuous on $(\alpha,\beta)$.

992: \end{definition}

993: 

994: Note that the uniqueness of Leray-Hopf solutions of the Initial Value Problem is

995: not known.

996: 

997: 

998: \begin{theorem}[Leray] \label{existence}

999: For every $u_0 \in V$, there exists a strong solution

1000: $u(t)$ of \eqref{NSE} on some interval $[0,T)$, $T>0$, with $u(0)=u_0$.

1001: \end{theorem}

1002: 

1003: \begin{theorem}[Leray] \label{structure}

1004: Let $u(t)$ be a Leray-Hopf solution of \eqref{NSE} on $[T,\infty)$. Then there are

1005: at most countably many distinct open intervals $I_j$, such that

1006: \[

1007: [T,\infty) = \bigcup_j \overline{I}_j,

1008: \]

1009: $u(t)$ is regular on each $I_j$, and the measure of $[T,\infty)\setminus \cup_j I_j$

1010: is zero.

1011: \end{theorem}

1012: 

1013: \begin{theorem} \label{uniqueness}

1014: Let $u(t)$ be a regular solution on $[0,T)$. Then every Leray-Hopf solution

1015: $v(t)$ on $[0,\infty)$ with $v(0)=u(0)$ coincides with $u(t)$ in $[0,T)$.

1016: \end{theorem}

1017: 

1018: Finally, we recall several well-known supplementary facts.

1019: \begin{remark} \label{r:ex}

1020: The complement of the exceptional set $Ex$ coincides with the set of points of

1021: strong continuity from the right.

1022: \end{remark}

1023: 

1024: \begin{theorem} \label{t:doubleconv}

1025: Let $u(t)$, $u_n(t)$ be Leray-Hopf solutions on the interval $[0,\infty)$,

1026: such that 

1027: \[

1028: u_n \to u \qquad \mbox{in} \qquad C([0,T];\Hw),

1029: \]

1030: as $n\to \infty$, for some $T>0$. Let $(\alpha,\beta) \subset (0, T)$ be an interval of regularity

1031: of $u(t)$. Then for every $0 <\delta <(\beta-\alpha)/2$,

1032: \[

1033: \|u_n(t)-u(t)\|\to 0 \qquad \mbox{uniformly on} \qquad [\alpha +\delta, \beta -\delta],

1034: \]

1035:  as $n \to \infty$.

1036: \end{theorem}

1037: 

1038: \begin{remark}

1039: Let $u(t)$ be a Leray-Hopf solution. As a $V$-valued function, $u(t)$ is analytic in time on every interval of regularity.

1040: \end{remark}

1041: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1042: \subsection{The weak global attractor for the 3D NSE} \label{s:weakattractor}

1043: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1044: A ball $B\subset H$ is called an absorbing ball for the equation \eqref{NSE}

1045: if for any bounded set $K \subset H$, there exists $t_0$, such that

1046: \[

1047: u(t) \in B, \qquad \forall t \geq t_0,

1048: \]

1049: for all Leray-Hopf solutions $u(t)$ of \eqref{NSE} on $[0,\infty)$ with $u(0)\in K$.

1050: It is well known that there exists an absorbing ball in $H$ for the 3D NSE.

1051: In fact, one has the following.

1052: \begin{proposition} \label{p:aset}

1053: The 3D Navier-Stokes equations possess an absorbing ball

1054: \[

1055: B = B_{\mathrm{s}}(0, R),

1056: \]

1057: where $R$ is any number larger than

1058: $|g|\nu^{-1} L/(2\pi)$ (see, e.g., \cite{CF}).

1059: \end{proposition}

1060: 

1061: Fix $R>|g|\nu^{-1} L/(2\pi)$ and let $X$ be the closed absorbing ball

1062: \[

1063: X= \{u\in H: |u| \leq R\},

1064: \]

1065: which is, clearly, weakly compact. Then for any bounded set $K \subset H$,

1066: there exists a time $t_0$, such that

1067: \[

1068: u(t) \in X, \qquad \forall t\geq t_0,

1069: \]

1070: for every Leray-Hopf solution $u(t)$ with the initial data $u(0) \in K$.

1071: Classical NSE estimates (see \cite{CF}) imply that for any sequence of Leray--Hopf solutions $u_n(t)$ (not only for the ones guaranteed by Theorem~\ref{thm:Leray}),

1072: the following result holds.

1073: 

1074: \begin{lemma} \label{l:convergenceofLH}

1075: Let $u_n(t)$ be a sequence of Leray-Hopf solutions,

1076: such that $u_n(t) \in X$ for all $t\geq t_0$. Then 

1077: \[

1078: \begin{aligned}

1079: u_n \ \ &\mbox{is bounded in} \ \ L^2([t_0,T];V),\\

1080: \ddt u_n \ \  &\mbox{is bounded in} \ \ L^{4/3}([t_0,T];V'),

1081: \end{aligned}

1082: \]

1083: for all $T>t_0$.

1084: Moreover, there exists a subsequence $u_{n_j}$ of $u_n$, which converges

1085: in $C([t_0, T]; \Hw)$ to some Leray-Hopf solution $u(t)$, i.e.,

1086: \[

1087: (u_{n_j},v) \to (u,v) \qquad

1088: \mbox{uniformly on} \qquad  [t_0,T],

1089: \]

1090: as $n_j\to \infty$, for all $v \in H$.

1091: \end{lemma}

1092: 

1093: Consider an evolutionary system for which

1094: a family of all trajectories consists

1095: of all Leray-Hopf solutions of the 3D Navier-Stokes equations

1096: in $X$. More precisely, define

1097: \[

1098: \begin{split}

1099: \Dc([T,\infty)) := \{&u(\cdot): u(\cdot)

1100: \mbox{ is a Leray-Hopf solution on } [T,\infty)\\

1101: & \mbox{and } u(t) \in X \ \forall t \in [T,\infty)\},

1102: \qquad T \in \mathbb{R},

1103: \end{split}

1104: \]

1105: \[

1106: \begin{split}

1107: \Dc((\infty,\infty)) :=\{&u(\cdot): u(\cdot)

1108: \mbox{ is a Leray-Hopf solution on } (-\infty,\infty)\\

1109: & \mbox{and } u(t) \in X \ \forall t \in (-\infty,\infty)\}.

1110: \end{split}

1111: \]

1112: Since $X$ is weakly compact, the existence of the weak

1113: global attractor is a direct consequence of Theorem~\ref{c:weakA}.

1114: Moreover, we have the following.

1115: \begin{lemma} \label{l:compact}

1116: $\Dc([0,\infty))$ is compact in $C([0, \infty);\Hw)$.

1117: \end{lemma}

1118: \begin{proof}

1119: Take any sequence

1120: $u_n \in \Dc([0,\infty))$, $n\in \mathbb{N}$.

1121: Thanks to Lemma~\ref{l:convergenceofLH}, there exists

1122: a subsequence, still denoted by  $u_n$, that converges

1123: to some $u^{1} \in \Dc([0,\infty))$ in $C([0, 1];\Hw)$ as $n \to \infty$.

1124: Now, passing to a subsequence and dropping a subindex, we obtain that

1125: there exists $u^{2}\in \Dc([0,\infty))$, such that

1126: $u_n \to u^2$ in $C([0, 2];\Hw)$ as $n \to \infty$.

1127: Note that $u^1(t)=u^2(t)$ on $[0, 1]$.

1128: Continuing

1129: this diagonalization process, we obtain a subsequence $u_{n_j}$

1130: of $u_n$ that converges

1131: to some $u \in \Dc([0,\infty))$ in $C([0, \infty);\Hw)$ as $n_j \to \infty$, which

1132: concludes the proof.

1133: \end{proof}

1134: 

1135: Now Theorem~\ref{c:weakA} yields the following.

1136: \begin{theorem} \label{t:AweakNSE}

1137: The weak global attractor $\Aw$ for the 3D

1138: Navier-Stokes equations exists and satisfies

1139: \begin{enumerate}

1140: \item[(a)]

1141: $

1142: \Aw = \{u(0): u \in \Dc((-\infty, \infty))\}.

1143: $

1144: \item[(b)] $\Aw$ is the maximal invariant set.

1145: \end{enumerate}

1146: \end{theorem}

1147: 

1148: \begin{lemma}

1149: If $u(t)$, a Leray-Hopf solution of the 3D NSE, satisfies

1150: \[

1151: \limsup_{t \to \infty} \|u(t)\| < \infty,

1152: \]

1153: then $u(t)$ converges strongly in $H$ to the weak global

1154: attractor $\Aw$.

1155: \end{lemma}

1156: \begin{proof}

1157: Suppose that $u(t)$ does not converge strongly in $H$ to $\Aw$. 

1158: Then there exist $M>0$ and a sequence $t_n \to \infty$ as

1159: $n \to \infty$, such that

1160: \begin{equation} \label{eq:farfromAw}

1161: \ds(u(t_n), \Aw) > M, \qquad n \in \mathbb{N}.

1162: \end{equation} 

1163: Note that there exists a time $T>0$, such that

1164: $\{u(t): t\geq T\}$ is relatively compact in $H$. Therefore, passing

1165: to a subsequence, we may assume that $u(t_n)$ converges strongly

1166: (and hence weakly) in $H$ to some $a \in H$. Therefore, $a \in \Aw$, which

1167: contradicts

1168: (\ref{eq:farfromAw}).

1169: \end{proof}

1170: 

1171: 

1172: \subsection{Strong convergence of Leray-Hopf solutions}

1173: 

1174: The aim of this subsection is to give sufficient conditions for a

1175: sequence of Leray-Hopf solutions on $[T,\infty)$ to converge

1176: in $C([T,\infty);H)$, provided it converges in $C([T,\infty);\Hw)$.

1177: We start with preliminary properties, some of which may have

1178: an intrinsic interest.

1179: \begin{theorem} \label{t:lim}

1180: Let $u(t)$ be a Leray-Hopf solution of \eqref{NSE} on $[T,\infty)$.

1181: Let $Ex$ be the exceptional set for this

1182: solution (see Def.~\ref{d:ex}). Then for any time $t_0 >T$,

1183: there exist $A_-$ and $A_+$, such that

1184: \begin{enumerate}

1185: \item[(a)]

1186: For every sequence $\{t_n\} \subset [T,\infty) \setminus Ex$,

1187: such that $t_n \to t_0$, $t_n < t_0$, it follows that  $|u(t_n)| \to A_-$

1188: as $n \to \infty$.

1189: \item[(b)]

1190: For every sequence $\{t_n\} \subset [T,\infty) \setminus Ex$,

1191: such that $t_n \to t_0$, $t_n > t_0$, it follows that $|u(t_n)| \to A_+$

1192: as $n \to \infty$.

1193: \end{enumerate}

1194: For these $A_-$ and $A_+$ we will use the following notations:

1195: \[

1196: \widetilde{\lim_{t \to t_0 -}} |u(t)| := A_- \qquad \mbox{and}

1197: \qquad \widetilde{\lim_{t \to t_0 +}} |u(t)| := A_+.

1198: \]

1199: \end{theorem}

1200: \begin{proof}

1201: For $\{t_n\}$ as in (a), the energy inequality (\ref{EI}) on

1202: $[t_n, t_{n+k}]$ is 

1203: \[

1204: |u(t_{n+k})|^2 + 2\nu \int_{t_n}^{t_{n+k}}\|u(s)\|^2 \, ds \leq

1205: |u(t_n)|^2 + 2\int_{t_n}^{t_{n+k}} (g, u(s)) \, ds,

1206: \]

1207: provided $t_{n+k} \geq t_n$.

1208: Taking the upper limit as $k \to \infty$, we obtain

1209: \[

1210: \limsup_{n \to \infty} |u(t_n)|^2 + 2\nu \int_{t_n}^{t_0}\|u(s)\|^2 \, ds \leq

1211: |u(t_n)|^2 + 2\int_{t_n}^{t_0} (g, u(s)) \, ds.

1212: \]

1213: Taking the lower limit as $n \to \infty$, we arrive at

1214: \[

1215: \limsup_{n \to \infty} |u(t_n)|^2 \leq

1216: \liminf_{n \to \infty} |u(t_n)|^2,

1217: \]

1218: i.e., $\lim_{n \to \infty} |u(t_n)|$ exists. Since the limit exists for any sequence

1219: $t_n$, it does not depend on the choice of a sequence.

1220: 

1221: For $\{t_n\}$ as in (b), the energy inequality (\ref{EI}) on

1222: $[t_{n+k}, t_n]$ is

1223: \[

1224: |u(t_{n})|^2 + 2\nu \int_{t_{n+k}}^{t_n}\|u(s)\|^2 \, ds \leq

1225: |u(t_{n+k})|^2 + 2\int_{t_{n+k}}^{t_n} (g, u(s)) \, ds,

1226: \]

1227: provided $t_{n+k} \leq t_n$.

1228: Taking the lower limit as $k \to \infty$, we obtain

1229: \[

1230: |u(t_n)|^2 + 2\nu \int_{t_0}^{t_n}\|u(s)\|^2 \, ds \leq

1231: \liminf_{n \to \infty} |u(t_n)|^2 + 2\int_{t_0}^{t_n} (g, u(s)) \, ds.

1232: \]

1233: Finally, taking the upper limit as $n \to \infty$, we arrive at

1234: \[

1235: \limsup_{n \to \infty} |u(t_n)|^2 \leq

1236: \liminf_{n \to \infty} |u(t_n)|^2,

1237: \]

1238: i.e., $\lim_{n \to \infty} |u(t_n)|$ exists. Since the limit exists for any sequence

1239: $t_n$, it does not depend on the choice of a sequence.

1240: \end{proof}

1241: 

1242: 

1243: \begin{lemma} \label{l:simple}

1244: Let $u(t)$ be a Leray-Hopf solution of \eqref{NSE}

1245: on $[T,\infty)$. Then

1246: \[

1247: \begin{split}

1248: \widetilde{\lim_{t \to t_0 -}} |u(t)| =

1249: \limsup_{t \to t_0 -} |u(t)|,\\

1250: \widetilde{\lim_{t \to t_0 +}} |u(t)| =

1251: \limsup_{t \to t_0 +} |u(t)|,

1252: \end{split}

1253: \]

1254: and,

1255: \begin{equation} \label{eq:orderoflim}

1256: \widetilde{\lim_{t \to t_0 -}} |u(t)| \geq 

1257: \widetilde{\lim_{t \to t_0 +}} |u(t)| \geq |u(t_0)|.

1258: \end{equation}

1259: for all $t_0 >T$.

1260: \end{lemma}

1261: \begin{proof}

1262: Take any $t_0 > T$. Obviously, we have

1263: \[

1264: \begin{split}

1265: \widetilde{\lim_{t \to t_0 -}} |u(t)| \leq

1266: \limsup_{t \to t_0 -} |u(t)|,\\

1267: \widetilde{\lim_{t \to t_0 +}} |u(t)| \leq

1268: \limsup_{t \to t_0 +} |u(t)|.

1269: \end{split}

1270: \]

1271: To show the opposite inequalities, note that for any

1272: $t_1 \in [T,\infty) \setminus Ex$ and $t_2>t_1$,

1273: the energy inequality (\ref{EI}) on $[t_1, t_2]$ is

1274: \begin{equation*}

1275: |u(t_2)|^2 + 2\nu \int_{t_1}^{t_2}\|u(s)\|^2 \, ds \leq

1276: |u(t_1)|^2 + 2\int_{t_1}^{t_2} (g, u(s)) \, ds.

1277: \end{equation*}

1278: First, we fix $t_1<t_0$ and take the upper limit

1279: as $t_2 \to t_0-$, obtaining

1280: \[

1281: \limsup_{t \to t_0 -} |u(t)|^2 + 2\nu \int_{t_1}^{t_0}\|u(s)\|^2 \, ds \leq

1282: |u(t_1)|^2 + 2\int_{t_1}^{t_0} (g, u(s)) \, ds.

1283: \]

1284: Now we take the limit as $t_1 \to t_0 -$ avoiding the exceptional set

1285: (see Theorem~\ref{t:lim}). We get

1286: \[

1287: \limsup_{t \to t_0 -} |u(t)|^2 \leq

1288: \widetilde{\lim_{t \to t_0 -}} |u(t)|.

1289: \]

1290: 

1291: Second, we fix $t_2>t_0$ and take the limit

1292: as $t_1 \to t_0+$ avoiding the exceptional set. We arrive at

1293: \[

1294: |u(t_2)|^2 + 2\nu \int_{t_0}^{t_2}\|u(s)\|^2 \, ds \leq

1295: \widetilde{\lim_{t \to t_0+}} |u(t)|^2 + 2\int_{t_0}^{t_2} (g, u(s)) \, ds.

1296: \]

1297: Taking the upper limit as $t_2 \to t_0 +$, we get

1298: \[

1299: \limsup_{t \to t_0 +} |u(t)|^2 \leq

1300: \widetilde{\lim_{t \to t_0 +}} |u(t)|.

1301: \]

1302: 

1303: Third, we fix $t_2>t_0$ and take the limit

1304: as $t_1 \to t_0-$ avoiding the exceptional set. We obtain

1305: \[

1306: |u(t_2)|^2 + 2\nu \int_{t_0}^{t_2}\|u(s)\|^2 \, ds \leq

1307: \widetilde{\lim_{t \to t_0-}} |u(t)|^2 + 2\int_{t_0}^{t_2} (g, u(s)) \, ds.

1308: \]

1309: Taking the limit as $t_2 \to t_0 +$ avoiding the exceptional set, we get

1310: \[

1311: \widetilde{\lim_{t \to t_0 +}} |u(t)|^2 \leq

1312: \widetilde{\lim_{t \to t_0 -}} |u(t)|.

1313: \]

1314: Finally the weak continuity of $u(t)$ yields

1315: \[

1316: \widetilde{\lim_{t \to t_0 +}} |u(t)| \geq |u(t_0)|,

1317: \]

1318: which concludes the proof.

1319: \end{proof}

1320: 

1321: \begin{remark} We can now rewrite the energy

1322: inequality for a Leray-Hopf solution $u(t)$ in the following form:

1323: \begin{equation} \label{e:gen}

1324: |u(t)|^2 + 2\nu \int_{t_0}^{t}

1325: \|u(s)\|^2 \, ds \leq

1326: \widetilde{\lim_{t\to t_0+}} |u(t)|^2 + 2\int_{t_0}^{t} (g, u(s)) \, ds,

1327: \end{equation}

1328: for all $0\leq t_0 \leq t$.

1329: \end{remark}

1330: 

1331: Recall that if the energy norm $|u(t)|$ of a Leray-Hopf solution 

1332: is continuous from the right at some $t=t_0$, then $t_0$ does not belong to

1333: the exceptional set for $u(t)$, i.e. the energy inequality holds

1334: for $t_0$ (see Remark~\ref{r:ex}).

1335: 

1336: \begin{lemma} \label{l:contfromtheright}

1337: Let $u(t)$ be a Leray-Hopf solution of \eqref{NSE} on $[T,\infty)$. Then

1338: $|u(t)|$ is continuous from the right at $t=t_0 \geq T$

1339: if and only if

1340: \[

1341: \widetilde{\lim_{t \to t_0 +}} |u(t)| = |u(t_0)|.

1342: \]

1343: \end{lemma}

1344: \begin{proof}

1345: If $|u(t)|$ is continuous from the right

1346: at $t=t_0 \geq T$, then, thanks to Lemma~\ref{l:simple}, we have that

1347: \begin{equation} \label{eq:tempcontr}

1348: \widetilde{\lim_{t \to t_0 +}} |u(t)| = \limsup_{t \to t_0 +} |u(t)| = |u(t_0)|. 

1349: \end{equation}

1350: Assume now that (\ref{eq:tempcontr}) holds.

1351: Due to the weak continuity of $u(t)$, we have

1352: \[

1353: \liminf_{t \to t_0 +} |u(t)|^2 \geq |u(t_0)|.

1354: \]

1355: Hence, 

1356: \[

1357: \lim_{t \to t_0 +} |u(t)|^2 = |u(t_0)|.

1358: \]

1359: \end{proof}

1360: 

1361: 

1362: 

1363: Now we will show that the strong

1364: continuity of a Leray-Hopf solution is equivalent to the strong

1365: continuity from the left (avoiding the exceptional set).

1366: 

1367: 

1368: \begin{lemma} \label{l:simplecontin}

1369: Let $u(t)$ be a Leray-Hopf solution of \eqref{NSE} on $[T, \infty)$. Then

1370: $|u(t)|$ is continuous at $t=t_0>T$ if and only

1371: if

1372: \[

1373: \widetilde{\lim_{t \to t_0 -}} |u(t)| = |u(t_0)|.

1374: \]

1375: \end{lemma}

1376: \begin{proof}

1377: Clearly, if $|u(t)|$ is continuous at

1378: $t=t_0 >T$, then

1379: \begin{equation} \label{eq:tempcont}

1380: \widetilde{\lim_{t \to t_0 -}} |u(t)| =\limsup_{t \to t_0 -} |u(t)| =  |u(t_0)|. 

1381: \end{equation}

1382: Assume now that (\ref{eq:tempcont}) holds. 

1383: Then due to the weak continuity of $u(t)$, we have

1384: \[

1385: \lim_{t \to t_0 -} |u(t)| = |u(t_0)|.

1386: \]

1387: In addition, Lemma~\ref{l:simple} (equation \eqref{eq:orderoflim}) implies that

1388: \[

1389: \widetilde{\lim_{t \to t_0 +}} |u(t)| = |u(t_0)|.

1390: \]

1391: Finally, thanks to Lemma~\ref{l:contfromtheright}, we have

1392: \[

1393: \lim_{t \to t_0 +} |u(t)| = |u(t_0)|.

1394: \]

1395: Therefore, $|u(t)|$ is continuous at $t=t_0$.

1396: \end{proof}

1397: 

1398: 

1399: 

1400: 

1401: 

1402: 

1403: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1404: %       LEFT Lemma

1405: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1406: 

1407: We will now study a weak convergence of Leray-Hopf solutions. Our goal

1408: is to obtain sufficient conditions for a strong convergence.

1409: 

1410: \begin{lemma} \label{l:left}

1411: Let $\{u_n(t)\}$, $u(t)$ be Leray-Hopf solutions of \eqref{NSE} on $[T_1, \infty)$.

1412: If $u_n \to u$ in $C([T_1, T_2]; \Hw)$, then

1413: \[

1414: \limsup_{n \to \infty} \widetilde{\lim_{t \to t_0 -}} |u_n(t)| \leq

1415: \widetilde{\lim_{t \to t_0 -}} |u(t)|,

1416: \]

1417: for all $t_0 \in (T_1, T_2]$.

1418: \end{lemma}

1419: \begin{proof}

1420: Suppose this is not true for some $t_0 \in (T_1, T_2]$. Then passing to a subsequence and dropping the subindexes, we can assume that

1421: \[

1422: \widetilde{\lim_{t \to t_0-}} |u_n(t)| - 

1423: \widetilde{\lim_{t \to t_0-}} |u(t)| \geq  \delta > 0, \qquad \forall n

1424: \]

1425: and $|u_n(t)| \to |u(t)|$ on $[T_1, T_2]\setminus S$, where $S$ is

1426: a set of zero measure, which includes the exceptional set for $u(t)$.

1427: 

1428: Let $S':= S\cup \left( \bigcup_n Ex_n \right)$, where 

1429: $Ex_n$ is the exceptional set for $u_n(t)$.

1430: The energy inequality for $u_n(t)$ implies

1431: \[

1432: \widetilde{\lim_{\tau \to t_0 - }} |u_n(\tau)|^2 \leq |u_n(t)|^2 +

1433: 2\int_{t}^{t_0}\left(g, u_n(s)\right) \, ds,

1434: \]

1435: for all $t\in [T_1,t_0] \setminus S'$.

1436: Taking the upper limit as $n \to \infty$ and using the strong convergence

1437: of $u_n(t)$ to $u(t)$ on $[T_1,T_2] \setminus S'$, we obtain

1438: \[

1439: \limsup_{n \to \infty} \widetilde{\lim_{\tau \to t_0 - }} |u_n(\tau)|^2 \leq

1440: |u(t)|^2 + 2\int_{t}^{t_0}\left(g, u(s)\right) \, ds,

1441: \qquad t\in [T_1, t_0] \setminus S'.

1442: \]

1443: Finally, letting $t \to t_0 -$, we get

1444: \[

1445: \limsup_{n \to \infty} \widetilde{\lim_{t \to t_0 - }} |u_n(t)|^2

1446: \leq \widetilde{\lim_{t \to t_0-}} |u(t)|^2,

1447: \]

1448: which is in contradiction with the definition of $\delta$.

1449: \end{proof}

1450: 

1451: 

1452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1453: %       RIGHT Lemma

1454: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1455: 

1456: \begin{lemma} \label{l:right}

1457: Let $\{u_n(t)\}$, $u(t)$ be Leray-Hopf solutions of \eqref{NSE} on $[T_1, \infty)$.

1458: If $u_n \to u$ in $C([T_1, T_2]; \Hw)$, then

1459: \[

1460: \liminf_{n \to \infty} \widetilde{\lim_{t \to t_0 +}} |u_n(t)| \geq

1461: \widetilde{\lim_{t \to t_0 +}} |u(t)|,

1462: \]

1463: for all $t_0 \in [T_1, T_2)$.

1464: \end{lemma}

1465: \begin{proof}

1466: Suppose this is not true for some $t_0 \in [T_1,T_2)$. Then passing to a

1467: subsequence and dropping the subindexes, we can assume that

1468: \[

1469: \widetilde{\lim_{t \to t_0+}} |u_n(t)| - 

1470: \widetilde{\lim_{t \to t_0+}} |u(t)| \leq \delta < 0,

1471: \qquad \forall n

1472: \]

1473: and

1474: $|u_n(t)| \to |u(t)|$ on $[T_1, T_2]\setminus S$, where $S$ is a zero

1475: measure set, which includes the exceptional set for $u(t)$.

1476: The energy inequality (\ref{e:gen}) for $u_n(t)$ implies

1477: \[

1478: |u_n(t)|^2  \leq \widetilde{\lim_{\tau \to t_0 + }} |u_n(\tau)|^2 +

1479: 2\int_{t_0}^{t}\left(g, u_n(s)\right) \, ds,

1480: \]

1481: for all $t\in [t_0, T_2]$.

1482: Taking the lower limit as $n \to \infty$ and using the strong convergence

1483: of $u_n(t)$ to $u(t)$ on $[T_1, T_2] \setminus S$, we obtain

1484: \[

1485: |u(t)|^2  \leq \liminf_{n \to \infty} \widetilde{\lim_{\tau \to t_0 + }}

1486: |u_n(\tau)|^2 +

1487: 2\int_{t_0}^{t}\left(g, u(s) \right) \, ds,

1488: \qquad t\in [t_0, T_2] \setminus S.

1489: \]

1490: Finally, letting $t \to t_0 +$, we get

1491: \[

1492: \widetilde{\lim_{t \to t_0+}} |u(t)|^2  \leq

1493: \liminf_{n \to \infty} \widetilde{\lim_{t \to t_0 + }} |u_n(t)|^2,

1494: \]

1495: which is in contradiction with the definition of $\delta$.

1496: \end{proof}

1497: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1498: 

1499: 

1500: 

1501: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1502: \begin{definition}

1503: For $u(t)$, a Leray-Hopf solution of \eqref{NSE}, denote

1504: \[

1505: [u(t_0)] := \widetilde{\lim_{t \to t_0 -}} |u(t)| -

1506: \widetilde{\lim_{t \to t_0 +}} |u(t)|,

1507: \]

1508: which we will call the energy norm jump (loss) at $t=t_0$. 

1509: \end{definition}

1510: 

1511: Note that due to (\ref{eq:orderoflim}), the energy norm jumps of the Leray-Hopf

1512: solutions are never negative, i.e.,

1513: \[

1514: [u(t)] \geq 0, \qquad \forall t,

1515: \]

1516: for any Leray-Hopf solution $u(t)$. Moreover, we have the following result.

1517: 

1518: 

1519: \begin{theorem} \label{thm:weak-jumps}

1520: Let $\{u_n(t)\}$, $u(t)$ be Leray-Hopf solutions of \eqref{NSE} on $[T_1, \infty)$.

1521: If $u_n \to u$ in $C([T_1, T_2], \Hw)$, then

1522: \[

1523: \limsup_{n \to \infty} [u_n(t)] \leq [u(t)],

1524: \]

1525: for all $t \in (T_1, T_2)$.

1526: \end{theorem}

1527: \begin{proof}

1528: Indeed, Lemmas \ref{l:left} and \ref{l:right}  yield

1529: \[

1530: \begin{split}

1531: \limsup_{n \to \infty} [u_n(t_0)] &= \limsup_{n \to \infty}

1532: \left(\widetilde{\lim_{t \to t_0-}}|u_n(t)| -

1533: \widetilde{\lim_{t \to t_0+}}|u_n(t)| \right) \\

1534: &\leq \limsup_{n \to \infty} \widetilde{\lim_{t \to t_0-}}|u_n(t)| -

1535: \liminf_{n \to \infty} \widetilde{\lim_{t \to t_0+}}|u_n(t)| \\

1536: &\leq \widetilde{\lim_{t \to t_0-}}|u(t)| -

1537: \widetilde{\lim_{t \to t_0+}}|u(t)|\\

1538: &= [u(t_0)].

1539: \end{split}

1540: \]

1541: \end{proof}

1542: 

1543: \begin{comment}

1544: For our strong convergence criteria we will need the following "uniform"

1545: version of Lemma~\ref{l:left}

1546: \begin{theorem} \label{t:left}

1547: Let $\{u_n\}$ be a sequence of Leray-Hopf solutions on $[T_1, \infty)$.

1548: If $u_n \to u$ in $C([T_1, T_2]; \Hw)$, then for any $\epsilon>0$

1549: there exists $N$, such that

1550: \[

1551: \widetilde{\lim_{t \to t_0 -}} |u_n(t)| \leq \widetilde{\lim_{t \to t_0 -}} |u(t)| + \epsilon

1552: \qquad \forall n\geq N, \forall t_0 \in (T_1, T_2].

1553: \]

1554: \end{theorem}

1555: \begin{proof}

1556: Suppose this is not true for some $t_0 \in (T_1, T_2]$. Then passing to a subsequence and dropping the subindexes, we can assume that

1557: \[

1558: \widetilde{\lim_{t \to t_0-}} |u_n(t)| - 

1559: \widetilde{\lim_{t \to t_0-}} |u(t)| \geq  \delta > 0, \qquad \forall n

1560: \]

1561: and $|u_n(t)| \to |u(t)|$ on $[T_1, T_2]\setminus S$, where $S$ is

1562: a set of zero measure, which includes the exceptional set.

1563: 

1564: Let $S':= S\cup \left( \bigcup_n Ex_n \right)$, where 

1565: $Ex_n$ is the exceptional set for $u_n$.

1566: The energy inequality for $u_n$ implies

1567: \[

1568: \widetilde{\lim_{\tau \to t_0 - }} |u_n(\tau)|^2 \leq |u_n(t)|^2 +

1569: 2\int_{t}^{t_0}\left(g, u_n(\tau)\right) \, d \tau.

1570: \]

1571: for all $t\in [T_1,t_0] \setminus S'$.

1572: Taking the upper limit as $n \to \infty$ and using the strong convergence

1573: of $u_n(t)$ on $[T_1,T_2] \setminus S'$, we obtain

1574: \[

1575: \limsup_{n \to \infty} \widetilde{\lim_{\tau \to t_0 - }} |u_n(\tau)|^2 \leq

1576: |u(t)|^2 + 2\int_{t}^{t_0}\left(g, u(\tau)\right) \, d \tau.

1577: \qquad t\in [T_1, t_0] \setminus S'.

1578: \]

1579: Finally, letting $t \to t_0 -$, we get

1580: \[

1581: \limsup_{n \to \infty} \widetilde{\lim_{t \to t_0 - }} |u_n(t)|^2

1582: \leq \widetilde{\lim_{t \to t_0-}} |u(t)|^2,

1583: \]

1584: which is in contradiction with the definition of $\delta$.

1585: \end{proof}

1586: \end{comment}

1587: 

1588: 

1589: 

1590: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1591: % Main Theorem!!!

1592: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1593: 

1594: Assume now that Leray-Hopf solutions converge weakly to a strongly

1595: continuous from the right in $H$ Leray-Hopf solution. We will show that the weak 

1596: convergence is strong if the energy jumps of solutions converge to the

1597: energy jumps of the limit solution.

1598: 

1599: \begin{theorem} \label{t:deltat}

1600: Let $\{u_n(t)\}$, $u(t)$ be Leray-Hopf solutions of \eqref{NSE} on $[T_1, \infty)$.

1601: If $u_n \to u$ in $C([T_1, T_2]; \Hw)$, 

1602: \[

1603: \widetilde{\lim_{t \to t_0 +}}|u(t)|=|u(t_0)|, \qquad \mbox{and} \qquad

1604: \liminf_{n \to \infty} [u_n(t_0)] \geq [u(t_0)],

1605: \]

1606: for some $t_0 \in (T_1, T_2)$, then

1607: \[

1608: \lim_{n\to \infty} |u_n(t_0)| = |u(t_0)|.

1609: \]

1610: \end{theorem}

1611: \begin{proof}

1612: Thanks to \eqref{eq:orderoflim} and Lemma~\ref{l:left}, we have

1613: \[

1614: \begin{split}

1615: \limsup_{n \to \infty} |u_n(t_0)| &\leq \limsup_{n \to \infty}

1616: \widetilde{\lim_{t\to t_0+}} |u_n(t)|\\

1617: &\leq \liminf_{n \to \infty} \widetilde{\lim_{t\to t_0-}} |u_n(t)|

1618: - \liminf_{n\to \infty}[u_n(t_0)]\\

1619: &\leq\limsup_{n \to \infty} \widetilde{\lim_{t\to t_0-}} |u_n(t)| - [u(t_0)]\\

1620: &\leq \widetilde{\lim_{t\to t_0-}} |u(t)| - [u(t_0)]\\

1621: &= \widetilde{\lim_{t\to t_0+}} |u(t)|\\

1622: &= |u(t_0)|,

1623: \end{split}

1624: \]

1625: which concludes the proof.

1626: \end{proof}

1627: 

1628: Note that if a Leray-Hopf solution $u(t)$ is continuous at $t=t_0$, then

1629: $[u(t_0)]=0$. Therefore, in particular, Theorem~\ref{t:deltat}

1630: immediately implies the following.

1631: 

1632: \begin{corollary} \label{c:gaex}

1633: Let $\{u_n\}$ be a sequence of Leray-Hopf solutions of \eqref{NSE} on

1634: $[T_1, \infty)$.

1635: If $u_n \to u$ in $C([T_1, T_2]; \Hw)$, and

1636: $|u(t)|$ is continuous at some $t=t_0 \in (T_1, T_2)$,

1637: then $u_n(t_0) \to u(t_0)$ strongly in $H$.

1638: \end{corollary}

1639: So, if a solution $u(t)$ on the weak global atractor is continuous

1640: in $H$, then it attracts its basin strongly.

1641: Note that this is exactly Rosa's asymptotic regularity condition (see \cite{R}).

1642: 

1643: \subsection{The strong global attractor for the 3D NSE}

1644: In this subsection we will see that if

1645: all solutions on the weak global attractor $\Aw$ are continuous in $H$, then

1646: $\Aw$ is the strong global attractor.

1647: First, we will show that if a solution

1648: belongs to $\Aw$ on some open time-interval $I$, then on any closed subinterval

1649: of $I$ it has to coincide  with a solution that stays on $\Aw$ for all time.

1650: 

1651: \begin{lemma} \label{l:glue1}

1652: Let $u \in \Dc([T,\infty))$ and $\tilde{u} \in \Dc([0,\infty))$, such that

1653: $u(T)=\tilde{u}(T)$ and

1654: \[

1655: \widetilde{\lim_{t \to T-}} |\tilde{u}(t)| \geq 

1656: \widetilde{\lim_{t \to T+}} |u(t)|,

1657: \]

1658: for some $T >0$. Let

1659: \[

1660: v(t)=\left\{

1661: \begin{aligned}

1662: &\tilde{u}(t), &0 \leq t \leq T, \\

1663: &u(t), &t > T.

1664: \end{aligned}

1665: \right.

1666: \]

1667: Then $v \in \Dc([0,\infty))$.

1668: \end{lemma}

1669: \begin{proof}

1670: Obviously, $v(t)$ is a weak solution of the 3D NSE. To show that it satisfies

1671: the energy inequality, take any $t\geq T$ and $t_0 \in (0,T)$,

1672: $t_0 \notin Ex$, where $Ex$ is the exceptional set for $\tilde{u}$. Then

1673: we have

1674: \[

1675: |v(t)|^2 + 2\nu \int_{T}^t \|v(s)\|^2 \, ds \leq

1676: \widetilde{\lim_{s \to T+}} |v(s)|^2 + 2\int_{T}^t (g, v(s)) \, ds,

1677: \]

1678: and

1679: \[

1680: \widetilde{\lim_{s \to T-}}|v(s)|^2 + 2

1681: \nu \int_{t_0}^{T} \|v(s)\|^2 \, ds \leq

1682: |v(t_0)|^2 + 2\int_{t_0}^{T} (g, v(s)) \, ds.

1683: \]

1684: Adding these inequalities, we obtain

1685: \[

1686: |v(t)|^2 + 2\nu \int_{t_0}^t \|v(s)\|^2 \, ds \leq

1687: |v(t_0)|^2 + 2\int_{t_0}^t (g, v(s)) \, ds,

1688: \]

1689: which concludes the proof.

1690: \end{proof}

1691: 

1692: \begin{corollary}

1693: \label{l:glue}

1694: Let $u\in \Dc([T,\infty))$ and $\tilde{u} \in \Dc((-\infty,\infty))$, such that

1695: $u(T)=\tilde{u}(T)$ and $u(t)$ is strongly continuous from the

1696: right at $t=T$. Let

1697: \[

1698: v(t)=\left\{

1699: \begin{aligned}

1700: &\tilde{u}(t), &t \leq T, \\

1701: &u(t), &t > T.

1702: \end{aligned}

1703: \right.

1704: \]

1705: Then $v \in \Dc((-\infty,\infty))$.

1706: \end{corollary}

1707: 

1708: \begin{lemma} \label{l:contona}

1709: Let $u(t)$ be a Leray-Hopf solution $u\in \Dc ([T_1,\infty))$,

1710: such that $u(t) \in \Aw$ for all

1711: $t \in (T_1, T_2)$. Then for any $T_0\in(T_1,T_2)$, there

1712: exists $v\in \Dc ((-\infty, \infty))$,

1713: such that $u(t) = v(t)$ on $[T_0,\infty)$. In particular, $u(t) \in \Aw$ for all

1714: $t>T_1$.

1715: \end{lemma}

1716: \begin{proof}

1717: First note that $(T_1, T_0)$ contains

1718: some interval of regularity of $u(t)$ and take $T$ in the interior of this

1719: interval. Since $u(T) \in \Aw$, there exists a solution

1720: $\tilde{u} \in \Dc ((-\infty, \infty))$, such that $\tilde{u}(T) =u(T)$.

1721: We will now glue them at point $t=T$, obtaining

1722: \[

1723: v(t)=\left\{

1724: \begin{aligned}

1725: &\tilde{u}(t), &t \leq T, \\

1726: &u(t), &t > T.

1727: \end{aligned}

1728: \right.

1729: \]

1730: Since $u(t)$ is strongly continuous at $t=T$,

1731: Corolary~\ref{l:glue} implies that $v \in \Dc((-\infty, \infty))$.

1732: \end{proof}

1733: 

1734: Assume now that all solutions that stay on

1735: the weak global attractor are strongly continuous in $H$. We will prove that

1736: in this case the weak global attractor is strong. 

1737: This generalizes a well-known result in \cite{FT85}.

1738: Indeed, if $\Aw$ is bounded in $V$, then all the solutions,

1739: as long as they stay on the weak global attractor, are regular, and

1740: are therefore continuous in $H$. This also weakens the condition

1741: of Ball \cite{B1} -- the continuity of all Leray-Hopf solutions from $(0,\infty)$

1742: to $H$.

1743: 

1744: \begin{lemma} \label{l:cont-comp}

1745: If  all solutions on the weak global attractor are strongly continuous

1746: in $H$, i.e., if $\Dc ((-\infty, \infty)) \subset C((-\infty, \infty); H)$,

1747: then $R(t)$ is asymptotically compact.

1748: \end{lemma} 

1749: \begin{proof}

1750: Take

1751: any $\{t_n\}$, such that $t_n \to \infty$ as $n \to \infty$, and $x_n \in R(t_n)X$.

1752: Then there exists

1753: a sequence of solutions $v_n\in \Dc ([0,\infty))$,

1754: such that $v_n(t_n) =x_n$.

1755: We will show that $\{x_n\}$ has a convergent subsequence.

1756: Without loss of generality, there exists $T>0$, such that $t_n \geq 2T$ for all $n$. Consider a sequence $u_n(t)=v_n(t+t_n -T)$, where $t\geq0$.

1757: Due to Lemma~\ref{l:compact}, $\Dc([0,\infty))$ is compact in

1758: $C([0, \infty);\Hw)$. Hence,

1759: passing to a subsequence and dropping a subindex,

1760: we can assume that $u_n$ coverges

1761: to some $u\in \Dc ([0,\infty))$ in $C([0, \infty); \Hw)$ as $n\to \infty$.

1762: By the definition of the weak global attractor, $u(t) \in \Aw$ for all $t \in [0, \infty)$.

1763: Applying Lemma~\ref{l:contona} with $(T_1,T_2)=(0,T)$, we obtain that

1764: there exists a complete trajectory

1765: $v\in \Dc((-\infty, \infty))$, such that $u(t)=v(t)$ on $[T/2, \infty)$. Therefore,

1766: \[

1767: u \in C([T/2, \infty); H).

1768: \]

1769: Hence, Corollary~\ref{c:gaex} yields that $u_n(T) \to u(T)$ strongly in $H$,

1770: i.e., $x_n  \to u(T)$  strongly in $H$.

1771: \end{proof}

1772: 

1773: We can now conclude with the two main results of this section. First, as

1774: a direct consequence of Theorem~\ref{t:asymptoticcompact} and

1775: Lemma~\ref{l:cont-comp} we obtain

1776: 

1777: \begin{theorem} \label{thm:main}

1778: If  all solutions on the weak global attractor are strongly continuous

1779: in $H$, then the

1780: strong global attractor $\As$ exists, is strongly compact, and coincides with $\Aw$.

1781: \end{theorem} 

1782: \begin{proof}

1783: Due to Lemma~\ref{l:cont-comp} we have that $R(t)$ is asymptotically compact.  Then Theorem~\ref{t:asymptoticcompact} implies that

1784: $\As$ exists, is strongly compact, and coincides with $\Aw$.

1785: \end{proof}

1786: 

1787: Second, we prove that the condition

1788: $\Dc((-\infty, \infty)) \subset C((-\infty, \infty);H)$

1789: is equivalent to a condition that all the ``energy jumps'' 

1790: uniformly converge to zero as time goes to infinity. More precisely,

1791: let

1792: \[

1793: [R(t)X]:= \sup \{[u(t)] : u \in \Dc([0,\infty))\}.

1794: \]

1795: Then we have the following.

1796: \begin{theorem}

1797: $\Dc((-\infty, \infty)) \subset C((-\infty, \infty); H)$ if and only

1798: if $[R(t)X] \to 0$ as $t \to \infty$.

1799: \end{theorem} 

1800: \begin{proof}

1801: It is obvious that if  $[R(t)X] \to 0$ as $t \to \infty$, then we have

1802: $\Dc((-\infty, \infty)) \subset C((-\infty, \infty); H)$.

1803: 

1804: Assume now that $\Dc((-\infty, \infty)) \subset C((-\infty, \infty); H)$,

1805: but $[R(t)X]$ does not converge to $0$ as $t \to \infty$.

1806: Then there exist a sequence of Leray-Hopf solutions

1807: $v_n \in \Dc([0,\infty))$

1808: and a time sequence $t_n \to \infty$ as $n\to \infty$, such that 

1809: \begin{equation} \label{a:faraway}

1810: \limsup_{n \to \infty}[v_n(t_n)] >0.

1811: \end{equation}

1812: Proceeding as in the proof of Lemma~\ref{l:cont-comp}, we can now

1813: assume that there exists some $T>0$, such that the sequence

1814: $u_n(t)=v_n(t+t_n-T)$ (where $t\geq 0$) converges in $C([0,\infty);\Hw)$

1815: to the restriction $u=v|_{[0,\infty)}$ of some complete trajectory

1816: $v \in \Dc((-\infty,\infty))$. Since $v(t)$ is continuous in $H$,

1817: by Theorem~\ref{thm:weak-jumps} we must have

1818: \[

1819: \lim_{n \to \infty} [v_n(t_n)] =\lim_{n \to \infty} [u_n(T)]=[u(T)]=[v(T)]=0, 

1820: \]

1821: a contradiction.

1822: 

1823: \end{proof}

1824: 

1825: \subsection{Regular part of the global attractor}

1826: We define the regular part of the weak global attractor, first introduced is

1827: \cite{FT85}, as follows.

1828: \[

1829: \begin{aligned}

1830: \Areg := \left\{ \right.&u_0:  \exists \tau>0, u \in \Dc((-\infty, \infty)) \mbox{ with }

1831: u(0)=u_0, \mbox{ such that }

1832: u(t) \mbox{ is regular} \\ &\mbox{on } (-\tau, \tau),

1833: \mbox{ and for each } \tilde{u} \in \Dc((-\infty, \infty))  \mbox{ with } \tilde{u}(0)=u(0) \mbox{ we have }\\

1834: & u(t)=\tilde{u}(t) \  \forall t \in(-\tau, \tau) \left. \right\}.

1835: \end{aligned}

1836: \]

1837: 

1838: The following result was proven in \cite{FT85}:

1839: \begin{theorem} The regular part of the global attractor satisfies the

1840: following properties:

1841: \begin{enumerate}

1842: \item[(a)] $\Areg$ is weakly open in $\Aw$,

1843: \item[(b)] $\Areg$ is weakly dense in $\Aw$,

1844: \item[(c)] If all solutions in  $\Aw$ are regular, then $\Aw$ is bounded in $V$

1845: (hence, $\Areg=\Aw$). 

1846: \end{enumerate}

1847: \end{theorem}

1848: Part $(c)$ was misstated in \cite{FT85} as

1849: \begin{enumerate}

1850: {\it \item[($c'$)] If $\Aw \subset V$, then $\Areg=\Aw$.}

1851: \end{enumerate}

1852: However, the proof provided in \cite{FT85} yields only $(c)$. As yet it is not

1853: known whether $(c')$ is true.

1854: 

1855: Now we will further study the case when all weak solutions on the

1856: weak global attractor of the 3D NSE are continuous in $H$. Under this

1857: assumption, Theorem~\ref{thm:main} implies that the strong compact

1858: global attractor $\As$ exists and $\As=\Aw$. Moreover,

1859: 

1860: \begin{theorem} If $\Dc((-\infty, \infty)) \subset C((-\infty, \infty); H)$,

1861: then the regular part of the global attractor $\Areg$

1862: is strongly dense in $\As$.

1863: \end{theorem}

1864: \begin{proof}

1865: Due to Theorem~\ref{thm:main} $\As$ exists and is strongly compact.

1866: Therefore, weak and strong topologies are equivalent on $\As$.

1867: Then since $\Areg$ is weakly dense in $\Aw$, it is also strongly dense

1868: in $\As=\Aw$.

1869: \end{proof}

1870: 

1871: We conclude this section with the following remark. In the case when the cubic

1872: box $\Omega=[0,L]^3$ is replaced by the cuboid $\Omega=[0,L]^2 \times [0,l]$

1873: with $0<l \ll L$, there exists a function $\alpha(g)>0$, such that if

1874: \begin{equation} \label{e:condonlL}

1875: \frac{l}{L} \leq \alpha(g),

1876: \end{equation}

1877: then $\Aw$ is the strong global attractor and is regular (see \cite{RS,TZ}). It would be interesting to

1878: show that for some values of $l/L$ larger than $\alpha(g)$, all the Leray-Hopf solutions

1879: on $\Aw$ are strongly continuous, i.e., $\Aw$ is also the strong global attractor.

1880: 

1881: \section{Tridiagonal models for the Navier-Stokes equations}

1882: In this section we introduce a two-parameter family of new simple models

1883: for the Navier-Stokes equations with a nonlinear term enjoying the

1884: same basic properties as the nonlinear term $B(u,u)$ in the NSE \eqref{NSE}.

1885: 

1886: The role of the space $H$ will be played by $l^2$ with the usual inner

1887: product and norm:

1888: \[

1889: (u,v)= \sum_{n=1}^{\infty} u_nv_n, \qquad |u|=\sqrt{(u,u)}.

1890: \]

1891: The norm $|u|$ will be called the energy norm.

1892: Let $A:D(A) \to H$ be the Laplace operator defined by

1893: \[

1894: (Au)_n = n^{\alpha} u_n, \qquad n\geq 1,

1895: \]

1896: for some $\alpha >0$. The domain $D(A)$ of this operator is

1897: \[

1898: \left\{u: \sum_{n=1}^\infty n^{2\alpha}u_n^2 <\infty \right\}.

1899: \]

1900: Clearly, $D(A)$ is dense in $H$ and 

1901: $A$ is a positive definite operator whose eigenvalues are

1902: \[

1903: 1, 2^{\alpha},  3^{\alpha}, \dots

1904: \]

1905: 

1906: Let $V=A^{-1/2}H$ endowed with the following inner product and norm:

1907: \[

1908: ((u,v))= \sum_{n=1}^{\infty} n^{\alpha}u_nv_n, \qquad \|u\|=\sqrt{((u,u))}.

1909: \]

1910: Here $\|u\|$ is an analog of $H^1$-norm of $u$ and we will call it the enstrophy

1911: norm. Let also

1912: \[

1913: \|u\|_\gamma = \left(\sum_{n=1}^{\infty} n^\gamma u_n^2 \right)^{1/2},

1914: \]

1915: which is an analog of $H^{\gamma/\alpha}$-norm of $u$.

1916: 

1917: Our models for the NSE are given by the

1918: following equations:

1919: \begin{equation} \label{model}

1920: \left\{

1921: \begin{aligned}

1922: &\ddt u_n + \nu n^{\alpha}u_n - n^\beta u_{n-1}^2 + (n+1)^\beta u_n u_{n+1}

1923: =g_n, \qquad n=1,2,3\dots\\

1924: &u_0=0.

1925: \end{aligned}

1926: \right.

1927: \end{equation}

1928: Here, $\nu>0$, $\alpha>0$, and $\beta>1$. Note that the value of $\ddt u_n$ is determined

1929: only by the values of $u_{n-1}$, $u_n$, and $u_{n+1}$. Therefore, we will

1930: refer to the equations \eqref{model} as the tridiagonal model for the Navier-Stokes

1931: equations or shortly TNS equations.

1932: For $u=(u_1,u_2,\dots)$ they can be written in a more condensed form as

1933: \begin{equation}

1934: \ddt u + \nu Au + B(u,u) = g,

1935: \end{equation}

1936: where

1937: \[

1938: (B(u,v))_n=- n^\beta u_{n-1} v_{n-1} + (n+1)^\beta u_n v_{n+1},

1939: \]

1940: and $u_0=0$. Note that the orthogonality property holds for $B$:

1941: \begin{eqnarray*}

1942: \left(B(u,v),v\right)&=&\sum_{n=1}^{\infty}\left(

1943: -n^{\beta}u_{n-1} v_{n-1} v_n + (n+1)^\beta u_n v_{n+1} v_n \right)\\

1944: &=&

1945: \sum_{n=1}^{\infty}\left(

1946: -n^{\beta}u_{n-1} v_{n-1} v_n +n^\beta u_{n-1} v_n v_{n-1} \right)\\

1947: &=&0.

1948: \end{eqnarray*}

1949: 

1950: In the case of TNS equations, 

1951: a weak solution on $[T,\infty)$ (or $(-\infty, \infty)$, if

1952: $T=-\infty$) of \eqref{model} is actually a locally bounded $H$-valued

1953: function $u(t)$ on $[T, \infty)$, such that $u_n \in C^1([T,\infty))$

1954: and $u_n(t)$ satisfies \eqref{model} for all $n$. From now on weak solutions will

1955: be called just solutions.

1956: 

1957: A solution $u(t)$ is

1958: strong (or regular) on some interval $[T_1,T_2]$, if $\|u(t)\|$ is bounded

1959: on $[T_1,T_2]$. A solution is strong on $[T_1,\infty)$, if it is strong on every

1960: interval $[T_1,T_2]$, $T_2\geq 0$.

1961: 

1962: A Leray-Hopf solution of \eqref{model} on the interval $[T, \infty)$

1963: is a solution of \eqref{model} on $[T,\infty)$ satisfying the

1964: energy inequality

1965: \[

1966: |u(t)|^2 + 2\nu \int_{t_0}^t \|u(\tau)\|^2 \, d\tau \leq

1967: |u(t_0)|^2 + 2\int_{t_0}^t (g, u(\tau)) \, d\tau,

1968: \]

1969: for all $T \leq t_0 \leq t$, $t_0$ a.e. in $[T,\infty)$.

1970: The set $Ex$ of those $t_0$ for which the energy

1971: inequality does not hold will be called the exceptional set.

1972: 

1973: 

1974: \subsection{A priori estimates and the existence of strong solutions}

1975: Taking a limit of the Galerkin approximation, the existence of

1976: Leray-Hopf solutions follows in exactly the same way as for the 3D NSE.

1977: In this paper we will show some {\it a priori} estimates, which

1978: can be obtained rigurously for the Leray-Hopf solutions.

1979: For simplicity, we assume that $g$ is independent of time, $g\in H$, and

1980: $g_n \geq 0$ for all $n$.

1981: \\

1982: 

1983: \noindent

1984: {\bf Energy estimates.} Formally taking a scalar product of (\ref{model}) with $u$,

1985: we obtain

1986: \[

1987: \begin{split}

1988: \frac{1}{2} \ddt |u|^2  &\leq  -\nu \|u\|^2 + |g||u|\\

1989: &\leq  -\nu |u |^2 + \frac{\nu}{2}|u|^2 + \frac{|g|^2}{2\nu}\\

1990: &= -\frac{\nu}{2}|u|^2 + \frac{|g|^2}{2\nu}.

1991: \end{split}

1992: \]

1993: Using Gronwall's inequality, we conclude that

1994: \begin{equation} \label{eq:defK}

1995: |u(t)|^2 \leq e^{- \nu t} |u(0)|^2 +

1996: \frac{|g|^2}{\nu^2}(1-e^{-\nu t}).

1997: \end{equation}

1998: Hence $B=\{u\in H: \ |u| \leq R\}$ is an absorbing ball for the

1999: Leray-Hopf solutions, where $R$ is any number larger that $|g|/\nu$.

2000: \\

2001: 

2002: \noindent

2003: {\bf Enstrophy estimates.} Let $v=A^{1/2}u$ and 

2004: \[

2005: c_\mathrm{b} := \left\{

2006: \begin{split}

2007: &\alpha 2^\beta, &0<\alpha \leq 1,\\

2008: &\alpha 2^{\alpha+\beta-1}, &\alpha > 1.

2009: \end{split}

2010: \right.

2011: \]

2012: Using H\"older's inequality, we obtain the following estimate

2013: for the nonlinear term:

2014: \[

2015: \begin{split}

2016: | (B(u,u),Au) | &\leq

2017: \left| \sum_{n=1}^\infty \left[(n+1)^{\alpha}-n^{\alpha} \right]

2018: (n+1)^{\beta }u_n^2 u_{n+1} \right|\\

2019: &\leq c_\mathrm{b} \sum_{n=1}^\infty n^{\beta-\alpha/2-1} |v_n|^2 |v_{n+1}|\\

2020: &\leq c_\mathrm{b} (\max_n |v_n|) \sum_{n=1}^{\infty}

2021: n^{\beta-\alpha/2-1} v_n^2 \\

2022: &\leq c_\mathrm{b} |v| |A^{1/2} v|^{2\beta/\alpha-2/\alpha-1} |v|^{-2\beta/\alpha+

2023: 2/\alpha+3}\\

2024: &= c_\mathrm{b} |A^{1/2} v |^{2\beta/\alpha-2/\alpha-1}

2025: |v|^{-2\beta/\alpha+2/\alpha+4}\\

2026: &= c_\mathrm{b} |Au |^{2\beta/\alpha-2/\alpha-1}

2027: \|u\|^{-2\beta/\alpha+2/\alpha+4},

2028: \end{split}

2029: \]

2030: whenever $\beta\in[\alpha/2+1,3\alpha/2+1]$.

2031: Choosing $u$ to have only two consecutive nonzero terms, it is easy to

2032: check that this estimate is sharp. Moreover, when $\alpha=2/3$ and

2033: $\beta=11/6$, we have

2034: \[

2035: | (B(u,u),Au) | \leq c_\mathrm{b} |Au |^{3/2}

2036: \|u\|^{3/2},

2037: \]

2038: which is exactly what Sobolev estimates give for the 3D NSE.

2039: Therefore, formally taking a scalar product of (\ref{model}) with $Au$,

2040: we obtain

2041: \[

2042: \begin{split}

2043: \frac{1}{2}\ddt \|u\|^2 &\leq -\nu|Au|^2 + c_\mathrm{b} |Au |^{3/2}

2044: \|u\|^{3/2} + (g, Au)\\

2045: &\leq -\nu|Au| + \frac{\nu}{3} |Au|^2 + \frac{3^6c_\mathrm{b}^4}{2^8\nu^3} \|u\|^6 +

2046: \frac{3}{4\nu} |g|^2 + \frac{\nu}{3} |Au|^2\\

2047: &\leq -\frac{\nu}{3}|Au|^2 + \frac{3^6c_\mathrm{b}^4}{2^8\nu^3} \|u\|^6 +

2048: \frac{3}{4\nu} |g|^2,

2049: \end{split}

2050: \]

2051: a Riccati-type equation for $\|u\|^2$. Hence, the model has the same

2052: enstrophy estimates as the 3D NSE, similar properties, and the same open question concerning the regularity of the solutions in the case

2053: $(\alpha, \beta) = (2/3, 11/6)$. In particular, we have a global existence of

2054: Leray-Hopf weak solutions (see Theorem~\ref{thm:Leray}), local existence of

2055: strong solutions (see Theorem~\ref{existence}), Leray's

2056: structure theorem (see Theorem~\ref{structure}),

2057: uniqueness of strong solutions in the class of

2058: Leray-Hopf solutions (see Theorem~\ref{uniqueness}), and existence of

2059: a weak global attractor (see Theorem~\ref{t:AweakNSE}).

2060: 

2061: In the case $(\alpha,\beta) = (1/2, 7/4)$, we have 

2062: \[

2063: | (B(u,u),Au) | \leq c_\mathrm{b} |Au |^{2}

2064: \|u\|,

2065: \]

2066: which corresponds to the 4D Navier-Stokes equations.

2067: In the case $(\alpha,\beta) = (2/5, 17/10)$, we have 

2068: \[

2069: | (B(u,u),Au) | \leq c_\mathrm{b} |Au |^{5/2} \|u\|^{1/2},

2070: \]

2071: which corresponds to the 5D Navier-Stokes equations.

2072: In general, the choice

2073: \[

2074: \alpha = \frac{2}{d}, \qquad \beta = \frac{3}{2} + \frac{1}{d}

2075: \]

2076: would correspond to the d-dimensional Navier-Stokes equations.

2077: 

2078: 

2079: The similarity of the TNS equations \eqref{model} with the NSE holds also

2080: for values of $d\ne 3$. Indeed,

2081: when $2\beta < 3\alpha +2$, the enstrophy estimate implies a local existence of strong solutions, i.e.,

2082: solutions whose enstrophy norms are continuous.

2083: More precisely, 

2084: if $2\beta < 3\alpha +2$, then for any initial data

2085: $u_0\in V$, there exists a

2086: strong solution $u(t)$ with $u(0)=u_0$ on some interval $[0,T]$. 

2087: In terms of the dimension,

2088: a sufficient condition for the local existence of strong solutions is

2089: $d<4.$

2090: In the case when $\beta \leq \alpha +1$, the enstrophy estimate implies

2091: a global regularity. In terms of the dimension,

2092: a sufficient condition for the global existence of strong solutions is

2093: $d\leq2$.

2094: 

2095: Now we will concentrate on the solutions with initial data $u_n(0)\geq0$

2096: for all $n$.

2097: 

2098: \begin{theorem} \label{t:poditofsol}

2099: Let $u(t)$ be a solution of \eqref{model} with $u_n(0)\geq 0$.

2100: Then $u_n(t) \geq 0$ for all $t>0$, and 

2101: $u(t)$ satisfies the energy inequality

2102: \begin{equation} \label{ee}

2103: |u(t)|^2 + 2\nu \int_{t_0}^t \|u(\tau)\|^2 \, d\tau \leq

2104: |u(t_0)|^2 + 2\int_{t_0}^t (g, u(\tau)) \, d\tau

2105: \end{equation}

2106: for all $0 \leq t_0 \leq t$.

2107: \end{theorem}

2108: \begin{proof}

2109: A general solution for $u_n(t)$ can be written as

2110: \begin{multline*}

2111: u_n(t)=u_n(0)\exp\left(-

2112: \int_{0}^t\left[\nu n^\alpha + (n+1)^\beta u_{n+1}(\tau)\right] \, d\tau\right)\\

2113: + \int_{0}^t(g_n+n^\beta u_{n-1}^2(s) )\exp\left(-\int_{s}^{t}

2114: \left[ \nu n^\alpha+(n+1)^\beta u_{n+1}(\tau)\right] \, d\tau \right) \, ds.

2115: \end{multline*}

2116: Since $u_n(0) \geq 0$ for all $n$, then  $u_n(t)\geq 0$ for all $n$, $t>0$.

2117: Hence, multiplying \eqref{model} by $u_n$,

2118: taking a sum from $1$ to $N$, and integrating between $t_0$ and $t$,

2119: we obtain

2120: \[

2121: \begin{split}

2122: \sum_{n=1}^N u_n(t)^2  -& \sum_{n=1}^N u_n(t_0)^2

2123:  +2\nu \int_{t_0}^t \sum_{n=1}^N n^\alpha u_n^2 \, d\tau \\

2124: &= - 2\int_{t_0}^t  (N+1)^\beta u_N^2 u_{N+1} \, d\tau

2125: +2\int_{t_0}^t\sum_{n=1}^N  g_n u_n \, d\tau\\

2126: &\leq 2\int_{t_0}^t\sum_{n=1}^N g_n u_n \, d\tau.

2127: \end{split}

2128: \]

2129: Taking the limit as $N \to \infty$, we obtain \eqref{ee}.

2130: 

2131: \end{proof}

2132: 

2133: \subsection{Blow-up in finite time} \label{sub:finitetime}

2134: Here, when $2\beta -3\alpha -3 >0$ and $g_1>0$ is large enough,

2135: we will show that every solution $u(t)$ of \eqref{model} with $u_n(0) \geq 0$

2136: blows up in finite time in an appropriate norm.

2137: First, we need the following two lemmas.

2138: 

2139: %%%%%%%%%%%% LEMMA %%%%%%%%%%%%%%

2140: \begin{lemma} \label{l:big}

2141: Let $u(t)$ be a solution to \eqref{model} on $[0,\infty)$ with $u_n(0) \geq 0$

2142: for all $n$. Assume that

2143: $\|u(t)\|_{2(\beta+\gamma-1)/3} \in L^3_{\mathrm{loc}}([0,\infty);\mathbb{R})$.

2144: Then

2145: \begin{equation} \label{e:tmp777}

2146: \int_{t_0}^t \sum_{n=1}^\infty n^{\beta+\gamma-1}u_n^2u_{n+1} \, d\tau < \infty,

2147: \qquad \int_{t_0}^t \sum_{n=1}^\infty n^{\beta+\gamma-1}u_n^3

2148: \, d\tau < \infty,

2149: \end{equation}

2150: and

2151: \begin{equation} \label{e:tmp888}

2152: \|u(t)\|_\gamma^2 - \|u(t_0)\|_\gamma^2 + 2\nu \int_{t_0}^{t} \|u\|_{\alpha+\gamma}^2 \, d\tau \geq

2153: 2\gamma \int_{t_0}^{t} \sum_{n=1}^{\infty} (n+1)^{\beta+\gamma-1}u_n^2 u_{n+1}

2154: \, d\tau 

2155: \end{equation}

2156: for all $0\leq t_0 \leq t$, $0<\gamma\leq1$.

2157: \end{lemma}

2158: \begin{proof}

2159: Thanks to Theorem~\ref{t:poditofsol}, $u_n(t) \geq 0$ for all $n, t>0$.

2160: Since $\|u(t)\|^3_{2(\beta+\gamma-1)/3}$ is integrable on $[t_0,t]$  for all

2161: $0\leq t_0\leq t$, we obtain

2162: \begin{equation*} 

2163: \begin{split}

2164: \int_{t_0}^t \sum_{n=1}^\infty n^{\beta+\gamma-1}u_n^2u_{n+1} \, d\tau &\leq

2165: 2\int_{t_0}^t \sum_{n=1}^\infty n^{\beta+\gamma-1}u_n^3 \, d\tau\\

2166: &\leq 2\int_{t_0}^t \left(\sum_{n=1}^\infty

2167: n^{\frac{2}{3}(\beta+\gamma-1)}u_n^2\right)^{3/2} \, d\tau\\

2168: &= 2 \int_{t_0}^{t} \|u\|_{2(\beta+\gamma-1)/3}^3 \, d\tau\\

2169: &< \infty.

2170: \end{split}

2171: \end{equation*}

2172: Hence, the relations in \eqref{e:tmp777} hold. In particular,

2173: \begin{equation} \label{integrq}

2174: \liminf_{n \to \infty} \int_{t_0}^t n^{\beta+\gamma} u_n^2 u_{n+1} \, d\tau =0.

2175: \end{equation}

2176: Now multiplying \eqref{model} by

2177: $n^\gamma u_n$, taking a sum from $1$ to $N$, and integrating from

2178: $t_0$ to $t$, we obtain

2179: \begin{equation*} 

2180: \begin{split}

2181: \sum_{n=1}^N & n^{\gamma}u_n(t)^2  - \sum_{n=1}^N n^{\gamma} u_n(t_0)^2

2182: +2\nu \int_{t_0}^{t} \sum_{n=1}^N n^{\alpha+\gamma} u_n^2 \, d\tau \\

2183: &= 2\int_{t_0}^{t} \sum_{n=1}^{N-1} (n+1)^\beta((n+1)^\gamma -n^\gamma)u_n^2 u_{n+1}

2184: \, d\tau\\

2185: & \quad - 2\int_{t_0}^{t} (N+1)^\beta N^\gamma u_N^2 u_{N+1} \, d\tau

2186: + 2\int_{t_0}^{t}\sum_{n=1}^N n^\gamma g_n u_n \, d\tau\\

2187: &\geq 2\gamma \int_{t_0}^{t} \sum_{n=1}^{N-1} (n+1)^{\beta+\gamma-1}u_n^2 u_{n+1}

2188: \, d\tau

2189:  - 2\int_{t_0}^{t} (N+1)^\beta N^\gamma u_N^2 u_{N+1} \, d\tau

2190: \end{split}

2191: \end{equation*}

2192: Thanks to \eqref{integrq}, taking the lower limit as $N\to \infty$, we get

2193: \eqref{e:tmp888}.

2194: 

2195: \end{proof}

2196: 

2197: 

2198: \begin{lemma} \label{l:doublelarge}

2199: For any $c>0$, there exists $g_1>0$, such that 

2200: \[

2201: \int_t^{t+1} |u|^2 \, d\tau > c, \qquad \forall t\geq0,

2202: \]

2203: for all solutions $u(t)$ with $g=(g_1,g_2,\dots)$.

2204: \end{lemma}

2205: \begin{proof}

2206: Take any $c>0$. Thanks to Theorem~\ref{t:poditofsol}, $u_n(t) \geq 0$ for all $n, t>0$.

2207: Therefore,

2208: \[

2209: u_1(t+1/2) \geq u_1(t) - \nu \int_t^{t+1/2} u_1 \, d\tau -

2210: 2^\beta \int_t^{t+1/2} u_1u_2 \, d\tau +g_1.

2211: \]

2212: Now integrating this inequality over $[t,t+1/2]$, we obtain

2213: \[

2214: \int_t^{t+1} u_1 \, d\tau + \nu \int_t^{t+1} u_1 \, d\tau +

2215: 2^{\beta-1} \int_t^{t+1} u_1^2 \, d\tau + 2^{\beta-1}

2216: \int_t^{t+1} u_2^2 \, d\tau  \geq \frac{g_1}{2}. 

2217: \]

2218: Hence, using Cauchy-Schwarz inequality, we get

2219: \[

2220: (\nu+1) \left(\int_t^{t+1} (u_1^2 +u_2^2)\, d\tau \right)^{1/2} + 2^{\beta-1}

2221: \int_t^{t+1} (u_1^2 + u_2^2) \, d\tau \geq \frac{g_1}{2}. 

2222: \]

2223: Obviously, for $g_1$ large enough, we have that

2224: \[

2225: \int_t^{t+1} |u|^2 \, d\tau \geq \int_t^{t+1} (u_1^2+u_2^2) \, d\tau > c.

2226: \]

2227: \begin{comment}

2228: Note that

2229: \begin{eqnarray*}

2230: u_n(t+1) \geq u_n(t) \exp\left(-2\nu n^{\alpha} - (n+1)^\beta

2231: \int_t^{t+1} u_{n+1}(\tau) \, d\tau  \right)\\

2232: +\int_t^{t+1} g_n \exp\left(-2\nu n^{\alpha}(t+2-s) - (n+1)^\beta

2233: \int_t^{t+1-s} u_{n+1}(\tau) \, d\tau  \right) \, ds

2234: \end{eqnarray*}

2235: \end{comment}

2236: \end{proof}

2237: 

2238: Now we proceed to our main result in this section.

2239: 

2240: \begin{theorem} \label{thm:blowup}

2241: For every solution $u(t)$ to equation (\ref{model}) with $u_n(0)\geq0$, $\nu>0$, $2\beta - 3\alpha -3>0$, and $g_1$ large enough,

2242: $\|u(t)\|^3_{2(\beta+\gamma-1)/3}$ is not locally integrable for all $\gamma>0$.

2243: \end{theorem}

2244: \begin{proof}

2245: Thanks to Theorem~\ref{t:poditofsol}, $u_n(t) \geq 0$ for all $n, t>0$.

2246: Assume that there exist 

2247: \[

2248: \gamma\in(0,\min\{2\beta -3\alpha -3, 1\})

2249: \]

2250: and a solution $u(t)$ to \eqref{model}, such that

2251: $\|u(t)\|_{2(\beta+\gamma-1)/3} \in L^3_{\mathrm{loc}}([0,\infty);\mathbb{R})$.

2252: Then Lemma~\ref{l:big} implies that

2253: \[

2254:  \int_0^T \sum_{n=1}^\infty n^{\beta+\gamma-1}u_n^2u_{n+1} \, d\tau < \infty

2255: \qquad \text{and} \qquad  \int_0^T \sum_{n=1}^\infty n^{\beta+\gamma-1}u_n^3

2256: \, d\tau < \infty,

2257: \]

2258: for all $T>0$.

2259: Moreover, since $\alpha + \gamma < 2(\beta + \gamma -1)/3$, we have that

2260: $\|u(t)\|^2_{\alpha +\gamma}$ is locally integrable on $[0,\infty)$.

2261: 

2262: Now note that if $u_{n+1} \geq 2u_n$, then $u_nu_{n+1}^2 \leq \frac{1}{2}u_{n+1}^3$.

2263: Otherwise, $u_nu_{n+1}^2 \leq 2 u_n^2 u_{n+1}$. Hence,

2264: \begin{equation} \label{eq:ineq1}

2265: u_nu_{n+1}^2 \leq {\ts \frac{1}{2}}u_{n+1}^3 + 2 u_n^2 u_{n+1}, \qquad n \in \mathbb{N}.

2266: \end{equation}

2267: This also implies

2268: \begin{equation} \label{eq:ineq2}

2269: \begin{split}

2270: u_nu_{n+1} u_{n+2} &\leq {\ts \frac{1}{2}}u_n^2u_{n+1} + {\ts \frac{1}{2}}u_{n+1}u_{n+2}^2

2271: \\

2272: &\leq {\ts \frac{1}{2}}u_n^2u_{n+1} + {\ts \frac{1}{4}}u_{n+2}^3 + u_{n+1}^2u_{n+2},

2273: \end{split}

2274: \end{equation}

2275: for all $n \in \mathbb{N}$. From \eqref{model} we have

2276: \begin{multline*}

2277: \ddt(u_nu_{n+1}) =

2278:   -\nu (n^\alpha + (n+1)^\alpha)

2279: u_nu_{n+1}

2280:  +n^\beta u_{n-1}^2u_{n+1}- (n+1)^\beta u_nu_{n+1}^2\\

2281:   +(n+1)^\beta u_n^3- (n+2)^\beta u_nu_{n+1}u_{n+2} +u_ng_{n+1}+u_{n+1}g_n.

2282: \end{multline*}

2283: From this, using inequalities (\ref{eq:ineq1}) and (\ref{eq:ineq2}), we obtain

2284: \begin{multline} \label{e:ineq11}

2285: \sum_{n=1}^\infty (n+1)^{\gamma-1}(u_nu_{n+1})(t+1)-\sum_{n=1}^\infty (n+1)^{\gamma-1}(u_nu_{n+1})(t) \\

2286: + 2\nu \int_t^{t+1} \sum_{n=1}^\infty (n+1)^{\alpha+\gamma-1}  u_nu_{n+1} \, d\tau

2287: \\

2288: + (3+(3/2)^\beta) \int_t^{t+1}\sum_{n=1}^\infty (n+1)^{\beta+\gamma-1}  u_n^2u_{n+1} \, d\tau \\

2289: \geq \frac{1}{4}\int_t^{t+1}\sum_{n=1}^\infty n^{\beta+\gamma -1} u_n^3 \, d\tau, 

2290: \end{multline}

2291: for all $t>0$.

2292: On the other hand, Lemma~\ref{l:big} yields

2293: \begin{equation} \label{e:ineq22}

2294: \|u(t+1)\|^2_\gamma - \|u(t)\|^2_\gamma + 2\nu \int_t^{t+1} \|u\|_{\alpha+\gamma}^2 \, d\tau \geq

2295: 2\gamma \int_t^{t+1} \sum_{n=1}^{\infty} (n+1)^{\beta+\gamma-1}u_n^2 u_{n+1}

2296: \, d\tau 

2297: \end{equation}

2298: for all $t>0$.

2299: Denote

2300: \[

2301: \Theta(t) = \int_t^{t+1}\|u(\tau)\|_{\gamma}^2 \, d\tau + 

2302: \frac{2\gamma}{3+(3/2)^\beta} \int_t^{t+1}

2303: \sum_{n=1}^{\infty} (n+1)^{\gamma-1}(u_nu_{n+1})(\tau) \, d\tau.

2304: \]

2305: Note that $\Theta(t)$ is absolutely continuous on $[0,\infty)$.

2306: We will show that $\Theta(t)$ is a Lyapunov function for the equation, i.e.,

2307: $\Theta(t)$ is always increasing. Indeed, multiplying the inequality

2308: (\ref{e:ineq11}) by $2\gamma/(3+(3/2)^\beta)$ and adding

2309: (\ref{e:ineq22}), we obtain

2310: \begin{multline*}

2311: \ddt \Theta(t) \geq -2\nu \int_t^{t+1} \|u(\tau)\|_{\alpha + \gamma}^2 \, d\tau

2312: - \frac{4\gamma \nu}{3+(3/2)^\beta}

2313: \int_t^{t+1}\sum_{n=1}^{\infty} (n+1)^{\alpha+\gamma-1}  u_nu_{n+1} \, d\tau \\

2314: + \frac{\gamma}{6+2(3/2)^\beta}

2315: \int_t^{t+1}\sum_{n=1}^{\infty}  n^{\beta+\gamma -1}u_n^3 \, d\tau,

2316: \end{multline*}

2317: a.e. on $(0,\infty)$.

2318: Since $\gamma$ is such that

2319: \[

2320: \epsilon:= 2\beta -3\alpha -\gamma -3 >0,

2321: \]

2322: let

2323: \[

2324: A:=\left(\sum_{n=1}^{\infty} n^{-1-\epsilon} \right)^{-1/2}.

2325: \]

2326: Now H\"older's inequality yields

2327: \[

2328: \begin{split}

2329: \int_t^{t+1} \sum_{n=1}^{\infty} n^{\alpha + \gamma} u_n^2 \, d\tau &\leq 

2330: \left(\int_t^{t+1} \sum_{n=1}^{\infty} n^{-1-\epsilon} \, d\tau\right)^{1/3}

2331: \left(\int_t^{t+1} \sum_{n=1}^{\infty} n^{\beta + \gamma -1} u_n^3 \, d\tau\right)^{2/3}

2332: \\

2333: &=

2334: A^{-2/3}\left(\int_t^{t+1} \sum_{n=1}^{\infty} n^{\beta + \gamma -1} u_n^3 \, d\tau\right)^{2/3}.

2335: \end{split}

2336: \]

2337: Hence,

2338: \[

2339: \int_t^{t+1} \sum_{n=1}^{\infty} n^{\beta + \gamma -1} u_n^3 \, d\tau \geq

2340: A \left(\int_t^{t+1} \sum_{n=1}^{\infty} n^{\alpha + \gamma} u_n^2 \, d\tau \right)^{3/2}.

2341: \]

2342: Finally, we obtain

2343: \begin{multline*}

2344: \ddt \Theta(t)

2345: \geq -2\nu\left( 1+ 2\gamma \frac{(3/2)^{\alpha+\gamma}}{3+(3/2)^\beta} \right)

2346: \int_t^{t+1} \|u(\tau)\|_{\alpha + \gamma}^2 \, d\tau\\

2347: + \frac{\gamma A}{6+2(3/2)^\beta}

2348: \left(\int_t^{t+1} \|u(\tau)\|_{\alpha + \gamma}^2 \, d\tau\right)^{3/2},

2349: \end{multline*}

2350: a.e. on $(0,\infty)$.

2351: Due to Lemma~\ref{l:doublelarge}, if $g_1$ is large enough, then there

2352: exists a positive constant $c$, such that

2353: \[

2354: \ddt \Theta(t) \geq c\Theta(t)^{3/2}, \qquad \text{a.e. on } (0,\infty).

2355: \]

2356: This is a Riccati-type equation. Hence, $\Theta(t)$ blows up in finite time,

2357: which contradicts the fact that it is continuous on $[0,\infty)$.

2358: \end{proof}

2359: 

2360: Figure~\ref{fig:regions} shows three regions, the ones where we were able

2361: to prove local regularity, global regularity, and  and blow-up in finite time.

2362: The labels $2D$, $3D$, and $4D$ show the dimensions of the 

2363: Navier-Stokes systems corresponding to the models at those points.

2364: 

2365: 

2366: \begin{figure}

2367: \center

2368: \includegraphics[width=4in]{regions.ps}

2369: \caption{Regions of local regularity, global regularity, and blow-up.} 

2370: \label{fig:regions}

2371: \end{figure}

2372: 

2373: 

2374: 

2375: \subsection{Non-regular weak global attractor}

2376: As in Subsection~\ref{s:weakattractor}, we can define an evolutionary

2377: system $\Dc$ whose trajectories are all Leray-Hopf solutions of the

2378: TNS equations.

2379: The weak global attractor for this system is

2380: \[

2381: \begin{split}

2382: \Aw = \{&u_0 \in H: \mbox{ there exists a Leray-Hopf solution } u(t) \mbox{ on }

2383: (-\infty,\infty),\\

2384: &\mbox{ such that } u(0)=u_0 \mbox{ and } |u(t)| \mbox{ is  bounded on }  (-\infty, \infty)\}.

2385: \end{split}

2386: \]

2387: 

2388: Recall that $g_n\geq 0$ for all $n\in \mathbb{N}$. Obviously, if $g=0$, then $\Aw=\{0\}$.

2389: Henceforth we will assume that $g \ne 0$.

2390: 

2391: \begin{theorem} \label{t:posA}

2392: If $g_n=0$ for all $n \geq N_g$, then

2393: every $u=(u_1,u_2,\dots) \in \Aw$ satisfies

2394: \[

2395: u_n \geq 0, \qquad n=1,2,\dots

2396: \]

2397: \end{theorem}

2398: \begin{proof}

2399: A general solution for $u_n(t)$ can be written as

2400: \begin{multline} \label{eq:gensolution}

2401: u_n(t)=u_n(t_0)\exp\left(-

2402: \int_{t_0}^t\nu n^\alpha + (n+1)^\beta u_{n+1}(\tau) \, d\tau\right)\\

2403: + \int_{t_0}^t\exp\left(-\int_{s}^{t}\nu n^\alpha+(n+1)^\beta

2404: u_{n+1}(\tau) \, d\tau\right) (g_n+n^\beta u_{n-1}^2(s) )\, ds.

2405: \end{multline}

2406: Clearly, this implies the following facts.

2407: \begin{enumerate}

2408: \item[(a)] If $u_n(t_0) \geq 0$ for some $n$ and $t_0$, then

2409: $u_n(t) \geq 0$ for all $t \geq t_0$.

2410: \item[(b)] If $|u(t)|$ is bounded for all $t\in\mathbb{R}$, then

2411: $u_n(t) \geq 0$ for all $t\in \mathbb{R}$, whenever

2412:  $u_{n+1}(t) \geq 0$ for all $t\in \mathbb{R}$.

2413: \end{enumerate}

2414: 

2415: Now assume that there exists  $u^0 \in \Aw$, such that $u^0_N < 0$

2416: for some $N \geq N_g$. Then there exists a Leray-Hopf solution $u(t)$,

2417: such that $u(0) = u^0$ and $|u(t)|$ is bounded on $(-\infty, \infty)$. For such a

2418: solution we have $u_N(t) <0$ for all $t \leq 0$.

2419: In addition, from the energy inequality for $u(t)$ we deduce that

2420: \[

2421: \begin{split}

2422: \sum_{n=N}^{\infty} u_n(t_1)^2 - \sum_{n=N}^{\infty} u_n(t_0)^2 &\leq

2423: 2\int_{t_0}^{t_1} \left[ N^\beta u_{N-1}(\tau)^2u_{N}(\tau)

2424:  -\nu \sum_{n=N}^\infty n^\alpha u_n(\tau)^2 \right] \, d\tau\\

2425:  &\leq  -2 \nu \int_{t_0}^{t_1} \sum_{n=N}^\infty u_n(\tau)^2 \, d\tau,

2426: \end{split}

2427: \]

2428: for all $t_0 \leq t_1 \leq 0$. Hence,

2429: \[

2430: \sum_{n=N}^{\infty} u_n(t_0)^2 \geq e^{-2\nu t_0}\sum_{n=N}^{\infty} u_n(0)^2,

2431: \]

2432: for all $t_0 \leq 0$. This implies that $|u(t)|^2$ is not bounded backwards

2433: in time, a contradiction.

2434: 

2435: Now take any $u^0 \in \Aw$. There exists a Leray-Hopf solution $u(t)$,

2436: such that $u(0) = u^0$ and $|u(t)|$ is bounded on $(-\infty, \infty)$.

2437: For such a solution we proved that

2438: \[

2439: u_n(t) \geq 0, \qquad \forall t\in\mathbb{R}, n \geq N_g.

2440: \]

2441: Now note that due to the remark (b) above, if $u_{n+1}(t) \geq 0$ for all

2442: $t\in \mathbb{R}$, then $u_n(t) \geq0$ for all $t\in \mathbb{R}$,

2443: which concludes the proof.

2444: \end{proof}

2445: 

2446: Now Theorem~\ref{t:posA} allows us to apply all the results in

2447: Subsection~\ref{sub:finitetime} to the solutions on the weak global

2448: attractor $\Aw$. Therefore, we have the following.

2449: \begin{remark} \label{rem:bl}

2450: Let $2\beta> 3\alpha +3$ and $g_1$ be large enough. Then

2451: $\|u(t)\|_{2(\beta+\gamma-1)/3}$ blows up in finite time

2452: for every solution $u(t)$ on $\Aw$, i.e., $\Aw$ is not bounded in

2453: $H^{2(\beta + \gamma-1)/(3\alpha)}$ for any $\gamma>0$.

2454: \end{remark}

2455: However, this does not mean that the weak global attractor cannot be strong.

2456: The question whether $\Aw$ is the strong global attractor remains open

2457: in the case $\beta > \alpha +1$.

2458: 

2459: 

2460: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2461: \subsection{Tridiagonal models for the Euler equations}

2462: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2463: In this section we consider the tridiagonal models for the Euler equations

2464: (TE), the equations \eqref{model} with $\nu=0$. First, let us show the

2465: global existence of the weak solutions to the TE equations.

2466: Take a sequence

2467: $\nu_j\to 0$ as $j\to \infty$. Given $u^0\in H$,

2468: let $u^j(t)$ be a solution of \eqref{model} with $\nu=\nu_j$ and $u^j(0)=u^0$.

2469: It is easy

2470: to show that the sequence $\{u^j\}$ is weakly equicontinuous. Therefore, thanks

2471: to Ascoli-Arzela theorem, passing to a subsequence and dropping

2472: a subindex, we obtain that there exists a function $u:[0,\infty) \to H$,

2473: such that $u^j \to u$ in $C([0,\infty);\Hw)$

2474: as $j\to \infty$. Clearly, $u(t)$ is a solution of the TE equations, in

2475: the sense that it is a locally bounded $H$-valued

2476: function on $[T, \infty)$, such that

2477: $u_n \in C^1([0,\infty))$ and $u_n(t)$ satisfies \eqref{model}

2478: for all $n$.

2479: 

2480: Now let us show that in the nonviscous case $\nu=0$, for every solution $u(t)$

2481: of \eqref{model}, the norm $\|u(t)\|_{2(\beta+\gamma-1)/3}$ blows up

2482: for any $\alpha>0$, $\beta>1$, and $\gamma>0$, reflecting the fact that there is no backward energy 

2483: transfer for this model.

2484: 

2485: \begin{theorem} \label{t:Eblow}

2486: Let $u(t)$ be a solution of \eqref{model} on $[0,\infty)$ with $\nu=0$,

2487: $g_n \geq 0$, $u_n(0) \geq 0$ for all $n$, and $u(0) \ne 0$.

2488: Then $\|u(t)\|_{2(\beta+\gamma-1)/3}$ is not bounded on $[0,\infty)$

2489: for every $\gamma >0$ .

2490: 

2491: \end{theorem}

2492: \begin{proof}

2493: Clearly, it is enough to prove the theorem in the case where

2494: $0<\gamma < \min\{1,2(\beta-1)\}$.

2495: Assume that $\|u(t)\|_{2(\beta+\gamma-1)/3}$ is bounded on $[0,\infty)$.

2496: Then Lemma~\ref{l:big} implies that

2497: \begin{equation} \label{eq:dobnorme}

2498: \|u(t)\|_\gamma^2 - \|u(t_0)\|_\gamma^2  \geq

2499: 2\gamma \int_{t_0}^{t} \sum_{n=1}^{\infty} (n+1)^{\beta+\gamma-1}u_n^2 u_{n+1}

2500: \, d\tau \geq 0, 

2501: \end{equation}

2502: for all $0\leq t_0 \leq t$. Thus, $\|u(t)\|^2_{\gamma}$ is non-decreasing. 

2503: Since $\gamma < 2(\beta+\gamma-1)/3$, $\|u(t)\|_{\gamma}$ is bounded on $[0,\infty)$. Then there

2504: exists $E_0>0$ such that

2505: \[

2506: \lim_{t \to \infty} \|u(t)\|_{\gamma}^2 = E_0.

2507: \]

2508: Then (\ref{eq:dobnorme}) implies that 

2509: \begin{equation} \label{eq:tempeuler}

2510: \lim_{t\to \infty} \int_t^\infty u_n(\tau)^2u_{n+1}(\tau) \, d\tau = 0, \qquad n\in \mathbb{N}. 

2511: \end{equation}

2512: Hence,

2513: \[

2514: \begin{split}

2515: u_n(t)^2 -u_n(0)^2&= 2n^\beta\int_0^tu_{n-1}^2u_n \, d\tau - 2(n+1)^\beta

2516: \int_0^t u_n^2u_{n+1} \, d\tau + 2\int_0^t g_n u_n \, d\tau \\

2517: &\to 2n^\beta\int_0^\infty u_{n-1}^2u_n \, d\tau - 2(n+1)^\beta

2518: \int_0^\infty u_n^2u_{n+1} \, d\tau + 2\int_0^\infty g_n u_n \, d\tau,

2519: \end{split}

2520: \]

2521: as $t\to \infty$. Hence $u_n(\infty) := \lim_{t\to \infty} u_n(t)$ exists for

2522: all $n$.

2523: Now (\ref{eq:tempeuler}) implies that $u_n(\infty) u_{n+1}(\infty)=0$ 

2524: for all $n$. Suppose that $u_k(\infty) \ne 0$ for some $k$. Then

2525: $u_{k+1}(\infty)=0$ and there exists $t_0>0$, such that

2526: \[

2527: (k+2)^{\beta}u_{k+1}(t)u_{k+2}(t) \leq  {\ts \frac{1}{3}}(k+1)^\beta u_k(\infty)^2 \qquad \mbox{and}

2528: \qquad u_k^2(t) \geq {\ts \frac{2}{3}}u_k(\infty)^2, 

2529: \]

2530: for all $t\geq t_0$. Thus,

2531: \[

2532: \begin{split}

2533: \ddt u_{k+1} &= (k+1)^\beta u_k(t)^2 - (k+2)^\beta u_{k+1}(t)u_{k+2}(t) +

2534: g_{k+1} \\

2535: &\geq {\ts \frac{1}{3}}(k+1)^\beta u_k(\infty)^2,

2536: \end{split}

2537: \]

2538: for all $n \geq t_0$. Therefore,

2539: \[

2540: \lim_{t \to \infty} u_{k+1}(t) =\infty,

2541: \]

2542: a contradiction.

2543: \end{proof}

2544: 

2545: \section*{Acknowledgment}

2546: We thank the referee for his constructive criticism and pertinent remarks.

2547: 

2548: \begin{comment}

2549: We conclude with the following result.

2550: \begin{theorem}

2551: If $2\beta > 3\alpha +3$,

2552: If $g_n=0$ for all $n \geq N_g$, and $g_1$ is large enough, then

2553: \[

2554: \sup_{t\in \mathbb{R}}u_n(t) \geq \frac{\nu n^{\beta-\alpha}}{(n+1)^{2(\beta-\alpha)}},

2555: \qquad \forall n\geq N_g,

2556: \]

2557: for each $u$, such that $u(t)\in \Aw$, $t\in\mathbb{R}$ .

2558: \end{theorem}

2559: \begin{proof}

2560: Take any weak solution with $|u(t)|$ bounded on $\mathbb{R}$. Let

2561: \[

2562: \bar{u}_n := \sup_{t\in\mathbb{R}} u_n(t)^2.

2563: \]

2564: Taking the limit as $t_0\to -\infty$ of (\ref{eq:gensolution}), we obtain

2565: \[

2566: u_n(t) \leq \frac{n^{\beta-\alpha}}{\nu}\bar{u}_{n-1}^2, \qquad n\geq N_g, \ t \in \mathbb{R}.

2567: \]

2568: Hence,

2569: \[

2570: \bar{u}_n \leq \frac{n^{\beta-\alpha}}{\nu}\bar{u}_{n-1}^2, \qquad n\geq N_g.

2571: \]

2572: Let

2573: \[

2574: v_n := \frac{(n+1)^{2(\beta-\alpha)}}{\nu n^{\beta-\alpha}}\bar{u}_n.

2575: \]

2576: Then we have

2577: \[

2578: v_n \leq v_{n-1}^2.

2579: \]

2580: Assume that $v_N <1$ for some $N \geq N_g$. Then

2581: \[

2582: v_n < v_N^{2(n-N)}, \qquad \mbox{and} \qquad \bar{u}_n \leq

2583: \frac{\nu n^{\beta-\alpha}}{(n+1)^{2(\beta-\alpha)}}v_N^{2(n-N)}.

2584: \]

2585: It is obvious that in this case $\|u(t)\|_{\theta}$  must be bounded on

2586: $\mathbb{R}$ for all $\theta>0$. On the other hand, 

2587: due to Theorem~\ref{thm:blowup}, 

2588: $\|u(t)\|_{2(\beta+\gamma-1)/3}$ blows up in finite time for any $\gamma>0$.

2589: Hence, $v_n\geq1$ for all $N \geq N_g$, i.e.,

2590: \[

2591: \bar{u}_n \geq \frac{\nu n^{\beta-\alpha}}{(n+1)^{2(\beta-\alpha)}}, \qquad n\geq N_g,

2592: \]

2593: which concludes the proof.

2594: \end{proof}

2595: \end{comment}

2596: \begin{thebibliography}{99}

2597: 

2598: 

2599: \bibitem{B1} J. M. Ball, Continuity properties and global attractors of generalized

2600: semiflows and the Navier-Stokes equations, {\it J. Nonlinear Sci.} {\bf7} (1997),

2601: 475Ð502. Erratum: {\it J. Nonlinear Sci.} {\bf 8} (1998),  233.

2602: 

2603: \bibitem{B2} J. M. Ball, Global attractors for damped semilinear wave equations,

2604: {\it Discr. Cont. Dyn. Sys.} {\bf10} (2004), 31Ð52.

2605: 

2606: \bibitem{CMR} T. Caraballo, P. Mar’n-Rubio, and J. C. Robinson,

2607: A comparison between two theories for multi-valued semiflows and their

2608: asymptotic behaviour, {\it Set-Valued Anal.} {\bf 11} (2003), 297--322.

2609: 

2610: \bibitem{CV} V. V. Chepyzhov and M. I. Vishik, {\it Attractors for Equations of

2611: Mathematical Physics}, American Mathematical Society Colloquium Publications

2612: {\bf 49}, American Mathematical Society, Providence, RI, 2002.

2613: 

2614: \bibitem{C} A. Cheskidov, Blow-up in finite time for the dyadic model of the

2615: Navier-Stokes equations, {\it  Trans. Amer. Math. Soc.}, to appear.

2616: 

2617: \bibitem{CF} P. Constantin and C. Foias, {\it Navier-Stokes Equation}, University of Chicago Press, Chicago, 1989. 

2618: 

2619: \bibitem{FS}F. Flandoli and B. Schmalfu\ss, 

2620: Weak solutions and attractors for three-dimensional Navier-Stokes equations with

2621: nonregular force, {\it J. Dynam. Differential Equations} {\bf 11} (1999),  355--398.

2622: 

2623: \bibitem{FMRT} C. Foias, O. P. Manley, R. Rosa, and R. Temam, {\it Navier-Stokes

2624: equatinon and Turbulence}, Encyclopedia of Mathematics and its Applications

2625: {\bf 83}, Cambridge University Press, Cambridge, 2001. 

2626: 

2627: \bibitem{FT85}  C. Foias and R. Temam, The connection between the Navier-Stokes

2628: equations, and turbulence theory, {\it Directions in Partial

2629: Differential Equations} (Madison, WI, 1985), Publ. Math. Res. Center

2630: Univ. Wisconsin, 55--73.

2631: 

2632: \bibitem{FP}

2633: S. Friedlander and N. Pavlovi\'c, Blowup in a three-dimensional vector

2634: model for the Euler equations, {\it Comm. Pure Appl. Math.} {\bf 57} (2004),

2635: 705--725.

2636: 

2637: \bibitem{H} J. K. Hale, {\it Asymptotic behavior of dissipative systems}, 

2638: Amer. Math. Soc., Providence, RI, 1988.

2639: 

2640: \bibitem{HLS}J. K. Hale,  J. P. LaSalle, and M. Slemrod,

2641: Theory of a general class of dissipative processes, {\it J. Math. Anal. Appl.}

2642: {\bf 39} (1972), 177--191.

2643: 

2644: 

2645: \bibitem{K} I. Kukavica, Role of the pressure for validity of the energy equality for solutions of

2646: the NavierÐStokes equation,

2647: {\it J. Dynam. Differential Equations} {\bf 18} (2006), 461--482.

2648: 

2649: \bibitem{KP}

2650: N. H. Katz and N. Pavlovi\'c, Finite time blow-up for a dyadic model of the Euler

2651: equations, {\it Trans. Amer. Math. Soc.} {\bf 357} (2005), 695--708.

2652: 

2653: \bibitem{L} O. A. Ladyzhenskaya, On the dynamical system generated by the Navier-Stokes

2654: equations, {\it J. Soviet Math.}  {\bf 3} (1975), 458--479.

2655: 

2656: \bibitem{L1} O. A. Ladyzhenskaya, {\it Attractors for Semigroups and Evolution

2657: Equations}, Cambridge Univ. Press, Cambridge, 1991.

2658: 

2659: \bibitem{MV} V. S. Melnik and J. Valero,  On attractors of multivalued semi-flows

2660: and differential inclusions, {\it Set-Valued Anal.} {\bf 6} (1998), 83--111.

2661: 

2662: \bibitem{RS} G. Raugel and G. R. Sell, 

2663: Navier-Stokes equations on thin 3D domains. I. Global attractors and global regularity of solutions.

2664: {\it J. Amer. Math. Soc.} {\bf 6} (1993), 503--568

2665: 

2666: \bibitem{R} R. M. S. Rosa,  Asymptotic regularity condition for the strong

2667: convergence towards weak limit sets and weak attractors of the 3D

2668: Navier-Stokes equations, {\it J. Diff. Eq.}, to appear.

2669: 

2670: \bibitem{S} G. R. Sell, Global attractors for the three-dimensional Navier-Stokes

2671: equations, {\it J. Dynam. Differential Equations} {\bf 8} (1996), 1--33. 

2672: 

2673: \bibitem{SY} G. R. Sell and Y. You, {\it Dynamics of Evolutionary Equations},

2674: Applied Mathematical Sciences {\bf 143}, Springer-Verlag, New York, 2002.

2675: 

2676: \bibitem{T2} R. Temam, {\it Infinite Dimensional Dynamical Systems in Mechanics and

2677: Physics}, Applied Mathematical Sciences {\bf 68}, (2nd Edition, 1997) Springer Verlag, New York, 1988.

2678: 

2679: \bibitem{TZ} R. Temam and M. Ziane, 

2680: Navier-Stokes equations in three-dimensional thin domains with various boundary conditions.

2681: {\it Adv. Differential Equations} {\bf 1} (1996), 499--546.

2682: \end{thebibliography}

2683: 

2684: 

2685: \end{document}

2686: 

2687: 

2688: