1: %%%%%%%%%DISCONTINUOUS FINITE ELEMNT APPROXIMATION%%%%%%%%%%%%%%%%%%%
2: \documentclass[11pt]{amsart}
3: \usepackage{amssymb,amsmath,amsthm}
4: \usepackage{mathrsfs,a4wide}
5: %\usepackage[scrtime]{prelim2e}
6: %\usepackage[notref,notcite]{showkeys}
7: \theoremstyle{plain}
8: \newtheorem{theorem}{Theorem}[section]
9: \newtheorem{lemma}[theorem]{Lemma}
10: \newtheorem{proposition}[theorem]{Proposition}
11: \newtheorem{corollary}[theorem]{Corollary}
12: \newtheorem{remark}[theorem]{Remark}
13: \newtheorem{definition}[theorem]{Definition}
14: \theoremstyle{definition}
15: \theoremstyle{remark}
16: \numberwithin{equation}{section}
17:
18: %Lettere calligrafiche
19: \newcommand{\as}{{\mathcal A}}
20: \newcommand{\hs}{{\mathcal H}}
21: \newcommand{\ks}{{\mathcal K}}
22: \newcommand{\cs}{{\mathcal C}}
23: \newcommand{\gs}{{\mathcal G}}
24: \newcommand{\fs}{{\mathcal F}}
25: \newcommand{\leb}{{\mathcal L}}
26: \newcommand{\ds}{{\mathcal D}}
27: \newcommand{\ms}{{\mathcal M}}
28: \newcommand{\ns}{{\mathcal N}}
29: \newcommand{\bs}{{\mathcal B}}
30: \newcommand{\eub}{{\mathcal E}}
31: \newcommand{\vub}{{\mathcal V}}
32: \newcommand{\Ps}{{\mathcal P}}
33: \newcommand{\dis}{{\mathcal D}}
34: \newcommand{\Es}{{\mathcal E}}
35: \newcommand{\ts}{{\mathcal T}}
36: \newcommand{\rs}{{\mathcal R}}
37:
38:
39: %lettere in neretto
40: \newcommand{\tb}{{\bf T}}
41: \newcommand{\rb}{{\bf R}}
42: \newcommand{\seub}{{\bf S}}
43:
44:
45: %Insiemi numerici
46: \newcommand{\R}{{\mathbb R}}
47: \newcommand{\N}{{\mathbb N}}
48:
49: %Matrici
50: \newcommand{\msim}{{\rm M}^{n\times n}_{\rm sym}}
51:
52: %Spazi metrici
53: \newcommand{\Ba}{B_1(0)}
54: \newcommand{\Bb}{\overline{B}_1(0)}
55: \newcommand{\Bbg}{\overline{B}_h(0)}
56: \newcommand{\tint}[1]{{\rm int}(#1)}
57:
58: %Spazi funzionali
59: \newcommand{\Om}{\Omega}
60: \newcommand{\Omb}{\overline{\Omega}}
61: \newcommand{\bv}{BV(\Omega)}
62: \newcommand{\Cc}{C_c^1(\Omega)}
63: \newcommand{\deli}[2]{L^{1,2}(#1 \setminus #2)}
64: \newcommand{\LD}[2]{LD(#1 \setminus #2)}
65: \newcommand{\Ni}{\Gamma}
66:
67: %norme
68: \newcommand{\norm}[1]{||#1||}
69: \newcommand{\nL}[1]{||\nabla #1||_a}
70: \newcommand{\nA}[1]{||E #1||_A}
71:
72:
73: %Convergenze
74: \newcommand{\weakst}{\stackrel{\ast}{\rightharpoonup}}
75: \newcommand{\weakstloc}{{\stackrel{\ast}{\rightharpoonup}}_{loc}}
76: \newcommand{\weak}{\rightharpoonup}
77:
78: %Teoria geometrica della misura
79: \newcommand{\esbd}[1]{\partial^* \!#1}
80: \newcommand{\per}[2]{{\rm P}(#1,#2)}
81:
82: %Definizioni ad hoc
83: \newcommand{\eps}{\varepsilon}
84: \newcommand{\afeaom}{\as_{\varepsilon,a}(\Om)}
85: \newcommand{\afenaom}{\as_{\varepsilon_n,a}(\Om)}
86: \newcommand{\afeom}{\as \fs_{\varepsilon}(\Om)}
87: \newcommand{\afenom}{\as \fs_{\varepsilon_n}(\Om)}
88: \newcommand{\afea}{\as_{\varepsilon,a}(\Om')}
89: \newcommand{\afeaomg}[1]{\as_{\varepsilon,a}(#1)}
90: \newcommand{\afeag}[1]{\as_{\varepsilon,a}^{#1}(\Om')}
91: \newcommand{\aft}[1]{\as(#1)}
92: \newcommand{\treaom}{\ts_{\varepsilon,a}(\Om)}
93: \newcommand{\trea}{\ts_{\varepsilon,a}(\Om')}
94: \newcommand{\ueps}{u_\varepsilon}
95: \newcommand{\veps}{v_\varepsilon}
96: \newcommand{\uepsn}{u_{\varepsilon_n}}
97: \newcommand{\vepsn}{v_{\varepsilon_n}}
98: \newcommand{\ind}[1]{1_{\{#1\}}}
99: \newcommand{\utilde}{\tilde{u}}
100: \newcommand{\supt}[1]{{\rm supt}(#1)}
101: \newcommand{\Sg}[2]{S^{#1}(#2)}
102: \newcommand{\vuat}{u_a(t)}
103: \newcommand{\vuas}{u_a(s)}
104: \newcommand{\zi}{z^i}
105: \newcommand{\zr}{z^r}
106: \newcommand{\uat}{z_a(t)}
107: \newcommand{\uas}{z_a(s)}
108: \newcommand{\ur}{u^r}
109: \newcommand{\uni}{u^i_n}
110: \newcommand{\unr}{u^r_n}
111: \newcommand{\zni}{z^i_n}
112: \newcommand{\znr}{z^r_n}
113: \newcommand{\uant}{z^i_{n}}
114: \newcommand{\ul}{z^l}
115: \newcommand{\ulp}{(z^l)^+}
116: \newcommand{\ulm}{(z^l)^-}
117: \newcommand{\umk}{z^{r(k)}}
118: \newcommand{\umkp}{(z^{r(k)})^+}
119: \newcommand{\umkm}{(z^{r(k)})^-}
120: \newcommand{\umkn}{z^{r(k)}_n}
121: \newcommand{\gat}{\Gamma_a(t)}
122: \newcommand{\cint}{K}
123: \newcommand{\Enk}{E_n^{k}}
124: \newcommand{\Enkp}{E^{k,+}_n}
125: \newcommand{\enkpt}{\tilde{E}^{k,+}_n}
126: \newcommand{\enkmt}{\tilde{E}^{k,-}_n}
127: \newcommand{\Enkm}{E^{k,-}_n}
128: \newcommand{\rskn}{\rs_n(Q_k)}
129: \newcommand{\hkn}{H_{k}}
130: \newcommand{\hknp}{H^{k,+}_n}
131: \newcommand{\hknm}{H^{k,-}_n}
132: \newcommand{\vpk}{v^+}
133: \newcommand{\vmk}{v^-}
134: \newcommand{\cks}{c_s(r(k))}
135: \newcommand{\cksu}{c_1(r(k))}
136: \newcommand{\cksd}{c_2(r(k))}
137: \newcommand{\rbepsd}{\rb_\eps^D}
138: \newcommand{\wsig}{w_\sigma}
139: \newcommand{\twn}{\tilde{w}_n}
140: \newcommand{\hwn}{\hat{w}_n}
141:
142: \input psfig.sty
143:
144: %Varie
145: \newcommand{\parent}[3]{\left #1 {#3} \right #2}
146: % racchiude il testo tra il
147: % primo e secondo simbolo
148: \newcommand{\graffe}[1]{\parent \{ \}{#1}}
149: % racchiude il testo tra graffe adeg.
150: \newcommand{\res}{\mathop{\hbox{\vrule height 7pt width .5pt depth 0pt
151: \vrule height .5pt width 6pt depth 0pt}}\nolimits}
152:
153:
154: \title
155: [A discontinuous finite element approximation
156: %of quasi-static growth of brittle fracture
157: ]
158: {A discontinuous finite element approximation of \\
159: quasi-static growth of brittle fractures}
160: \author[A. Giacomini]
161: {Alessandro Giacomini}
162: \address[Alessandro Giacomini]{S.I.S.S.A., Via Beirut 2-4, 34014, Trieste,
163: Italy}
164: \email[A. Giacomini]{giacomin@sissa.it}
165: \author[M. Ponsiglione]
166: {Marcello Ponsiglione}
167: \address[Marcello Ponsiglione]{S.I.S.S.A., Via Beirut 2-4, 34014, Trieste,
168: Italy}
169: \email[M. Ponsiglione]{ponsigli@sissa.it}
170: \begin{document}
171: \vskip .2truecm
172: \begin{abstract}
173: \small{
174: We propose a discontinuous finite element approximation for a model
175: of quasi-static growth of brittle fractures in linearly elastic bodies formulated
176: by Francfort and Marigo, and based on the classical Griffith's criterion. We restrict our
177: analysis to the case of anti-planar shear and we consider discontinuous displacements
178: which are piecewise affine with respect to a regular triangulation.
179: \vskip .3truecm
180: \noindent Keywords : variational models, energy minimization, free discontinuity
181: problems, crack propagation, quasi-static evolution, brittle fracture,
182: finite elements.
183: \vskip.1truecm
184: \noindent 2000 Mathematics Subject Classification: 35R35, 35J25, 74R10, 35A35, 65L60.}
185: \end{abstract}
186: \maketitle
187: {\small \tableofcontents}
188:
189: \section{Introduction}
190: \label{intr}
191:
192: In this paper we formulate a discontinuous finite element approximation
193: for a model of quasi-static growth of brittle fractures in linearly
194: elastic bodies proposed by Francfort and Marigo \cite{FM}. Their
195: model is based on the classical Griffith's criterion which involves a
196: competition between {\it bulk} and {\it surface} energies. To be precise,
197: let $\Om \subseteq \R^3$ be an elastic body, $\partial_D \Om$ be a part of its
198: boundary and let $g: \partial_D \Om \to \R^3$ be the spatial displacement of $\Om$ at
199: the points of $\partial_D \Om$. According to Griffith's theory,
200: given a preexisting crack $\Gamma_1 \subseteq \Omb$,
201: the new crack $\Gamma$ and the displacement $u: \Om \setminus \Gamma \to \R^3$
202: associated to $g$ at the equilibrium minimizes the following
203: energy
204: \begin{equation}
205: \label{griffithenergy}
206: \Es(v,g,\Gamma):= \int_\Om \mu |Ev|^2 +\frac{\lambda}{2}\,|{\rm div}\,v|^2 \,dx+ \hs^2(\Gamma),
207: \end{equation}
208: among all cracks $\Gamma$ with $\Gamma_1 \subseteq \Gamma$ and all displacements
209: $v: \Om \setminus \Gamma \to \R^3$ with $v=g$ on $\partial_D \Om \setminus \Gamma$. Here $Ev$
210: denotes the symmetric part of the gradient of $v$, and $\hs^2$ denotes the two
211: dimensional Hausdorff measure, while $\mu$ and $\lambda$ are the Lam\'e coefficients.
212: The boundary condition is required only on $\partial_D \Om \setminus \Gamma$
213: because the displacement in a fractured region is assumed not to be transmitted.
214: Let us indicate by $\Es(g,\Gamma)$ the minimum value of
215: (\ref{griffithenergy}) among all $v: \Om \setminus \Gamma \to \R^3$
216: with $v=g$ on $\partial_D \Om \setminus \Gamma$.
217: \par
218: Supposing that the boundary displacement $g$ varies with the time $t \in [0,1]$, the quasi-static
219: evolution $t \to \Gamma(t)$ proposed in \cite{FM} requires that:
220: \begin{itemize}
221: \item[(1)] $\Gamma(t)$ is increasing in time, i.e., $\Gamma(t_1) \subseteq \Gamma(t_2)$
222: for all $0 \le t_1 \le t_2 \le 1$ (irreversibility of the process);
223: \item[{}]
224: \item[(2)] $\Es(g(t),\Gamma(t)) \le \Es(g(t),\Gamma)$ for all cracks $\Gamma$ such that
225: $\cup_{s<t} \Gamma(s) \subseteq \Gamma$ (equilibrium condition);
226: \item[{}]
227: \item[(3)] the total energy $\Es(g(t),\Gamma(t))$ is absolutely continuous in time
228: and (conservation of energy)
229: $$
230: \frac{d}{dt} \Es(g(t),\Gamma(t))= 2 \mu \int_\Om Eu(t) E\dot{g}(t)\,dx
231: +\lambda \int_\Om {\rm div}\,u(t) \,{\rm div}\,\dot{g}(t) \,dx.
232: $$
233: \end{itemize}
234: A precise mathematical formulation of this model has been given by
235: Dal Maso and Toader \cite{DMT} in the case of {\it anti-planar shear} in dimension two assuming
236: that the fractures are compact sets with a finite number of connected components.
237: Recently Francfort and Larsen \cite{FL}, using the framework of $SBV$-functions,
238: proved the existence of a quasi-static growth of brittle fractures in the case
239: of {\it generalized anti-planar shear} and without assumptions on the structure of the fractures
240: which are dealt with the set of jumps of the displacements.
241: To be precise, they consider as elastic body an infinite cylinder whose section $\Om \subseteq \R^N$
242: is subject to a displacement $u \in SBV(\Om)$ in the direction orthogonal to $\Om$.
243: The crack at time $t$ on the section $\Om$ is defined as
244: $$
245: \Gamma(t):=\bigcup_{s<t} \left[ S_{u(s)} \cup
246: (\partial_D \Om \cap \{u(s) \not= g(s)\})
247: \right],
248: $$
249: where $S_u$ denotes the set of jumps of $u$. Moreover the pair $(u(t), \Gamma(t))$
250: is such that:
251: \begin{itemize}
252: \item[(a)]
253: for all $v \in SBV(\Om)$
254: \begin{equation}
255: \label{intrfl1}
256: \int_\Om |\nabla u(t)|^2\,dx + \hs^{N-1}(\Gamma(t)) \le
257: \int_\Om |\nabla v|^2\,dx + \hs^{N-1}(S_v \cup
258: (\partial_D \Om \cap \{v \not= g(t)\})
259: \cup \Gamma(t));
260: \end{equation}
261: \item[{}]
262: \item[(b)]
263: the total energy $\Es(t):= \int_\Om |\nabla u(t)|^2\,dx+ \hs^{N-1}(\Gamma(t))$ is absolutely
264: continuous and
265: \begin{equation}
266: \label{intrfl2}
267: \Es(t)=\Es(0) + 2\int_0^t \int_\Om \nabla u(\tau) \nabla \dot{g}(\tau) \,dx \,d\tau.
268: \end{equation}
269: \end{itemize}
270: The aim of this paper is to discretize the model using a suitable finite
271: element method and to a give a rigorous proof of its convergence to a
272: quasi-static evolution in the sense of Francfort and Larsen.
273: We restrict our analysis to a two dimensional setting considering only a
274: polygonal reference configuration $\Om \subseteq \R^2$.
275: \par
276: The discretization of the domain $\Om$ is carried out, following \cite{N},
277: considering two parameters $\eps>0$ and $a \in ]0,\frac{1}{2}[$ .
278: We consider a regular triangulation $\rb_\eps$ of size $\eps$
279: of $\Om$, i.e. we assume that there exist two constants $c_1$ and $c_2$ so that
280: every triangle $T \in \rb_\eps$ contains a ball of radius $c_1 \eps$
281: and is contained in a ball of radius $c_2 \eps$. In order to treat the boundary data,
282: we assume also that $\partial_D \Om$ is composed of edges of $\rb_\eps$.
283: On each edge $[x,y]$ of $\rb_\eps$ we consider a point $z$ such that $z=tx+(1-t)y$ with
284: $t \in [a, 1-a]$. These points are called {\it adaptive vertices}. Connecting
285: together the adaptive vertices, we divide every $T \in \rb_\eps$ into four triangles.
286: We take the new triangulation $\tb$ obtained after this division as the discretization
287: of $\Om$. The family of all such triangulations is denoted by $\ts_{\eps,a}(\Om)$.
288: \par
289: The discretization of the energy functional is obtained restricting
290: the total energy to the family of functions $u$
291: which are affine on the triangles of some triangulation $\tb(u) \in \ts_{\eps,a}(\Om)$
292: and are allowed to jump across the edges of $\tb(u)$. We indicate this
293: space by ${\mathcal A}_{\eps,a}(\Om)$. The boundary data is assumed to belong
294: to the space $\afeom$ of continuous functions which are affine on every triangle $T \in
295: \rb_\eps$.
296: \par
297: Given the boundary data $g \in W^{1,1}([0,1],H^1(\Om))$ with $g(t) \in \afeom$ for all
298: $t \in [0,1]$,
299: we divide $[0,1]$ into subintervals $[t^\delta_i,t^\delta_{i+1}]$
300: of size $\delta>0$ for $i=0, \ldots, N_\delta$, we set
301: $g^\delta_i=g(t^\delta_i)$, and
302: for all $u \in \afeaom$ we indicate by $S_D^{g^\delta_i}(u)$ the edges
303: of the triangulation $\tb(u)$ contained in $\partial_D \Om$ on which $u \not= g^\delta_i$.
304: Using a variational argument we construct a {\it discrete evolution}
305: $\{u^{\delta,i}_{\eps,a}\,:\,i=0, \ldots,N_\delta\}$ such that
306: $u^{\delta,i}_{\eps,a} \in \afeaom$ for all $i=0, \ldots, N_\delta$, and
307: such that considering the {\it discrete fracture}
308: \begin{equation*}
309: \Gamma^{\delta,i}_{\eps,a}:=
310: \bigcup_{r=0}^i \big[ S_{u^{\delta,r}_{\eps,a}} \cup
311: S_D^{g^\delta_r}(u^{\delta,r}_{\eps,a}) \big],
312: \end{equation*}
313: the following {\it unilateral minimality property} holds:
314: \begin{equation}
315: \label{pieceminintr}
316: \int_\Om |\nabla u^{\delta,i}_{\eps,a}|^2\,dx \le \int_\Om |\nabla v|^2\,dx +
317: \hs^1\left( \big( S_v \cup S_D^{g^\delta_i}(v) \big) \setminus
318: \Gamma^{\delta,i-1}_{\eps,a}\right).
319: \end{equation}
320: Moreover we get suitable estimates for the discrete total energy
321: $$
322: \Es^{\delta,i}_{\eps,a}:=\|\nabla u^{\delta,i}_{\eps,a}\|^2_{L^2(\Om;\R^2)}+
323: \hs^1 \left( \Gamma^{\delta,i}_{\eps,a}\right).
324: $$
325: The definition of the discrete fracture ensures that
326: $\Gamma^{\delta,i}_{\eps,a} \subseteq \Gamma^{\delta,j}_{\eps,a}$ for all
327: $i \le j$, recovering in this discrete setting the irreversibilty of the growth
328: given in $(1)$.
329: The minimality property \eqref{pieceminintr} is the reformulation in the finite element
330: space of the equilibrium condition $(2)$.
331: \par
332: In order to perform the asymptotic analysis of the {\it discrete evolution}
333: $\{u^{\delta,i}_{\eps,a}\,:\,i=0, \ldots,N_\delta\}$, we make
334: the piecewise constant interpolation in time
335: $u^{\delta}_{\eps,a}(t)=u^{\delta,i}_{\eps,a}$ and
336: $\Gamma^{\delta}_{\eps,a}(t)=\Gamma^{\delta,i}_{\eps,a}$ for all $t^\delta_i \le t <t^\delta_{i+1}$.
337: The main result of the paper is the following theorem.
338:
339: \begin{theorem}
340: \label{mainthm}
341: Let $g \in W^{1,1}([0,1], H^1(\Om))$ be such that $\|g(t)\|_\infty \le C$
342: for all $t \in [0,1]$ and let $g_\eps \in W^{1,1}([0,1], H^1(\Om))$ be
343: such that $\|g_\eps(t)\|_\infty \le C$, $g_\eps(t) \in \afeom$ for all $t \in [0,1]$
344: and
345: \begin{equation}
346: \label{bdryconv}
347: g_\eps \to g \quad \mbox { strongly in }W^{1,1}([0,1], H^1(\Om)).
348: \end{equation}
349: Given the discrete evolution $\{t \to u^{\delta}_{\varepsilon,a}(t)\}$
350: relative to the boundary data $g_\eps$, let $\Gamma^\delta_{\eps,a}$ and
351: $\Es^\delta_{\eps,a}$ be the associated fracture and total energy.
352: \par
353: Then there exist $\delta_n \to 0$, $\varepsilon_n \to 0$, $a_n \to 0$,
354: and a quasi-static evolution
355: $\{t \to (u(t),\Gamma(t)),\,t \in [0,1]\}$ relative to the boundary data $g$,
356: satisfying \eqref{intrfl1} and \eqref{intrfl2}, and such that setting
357: $u_n:=u^{\delta_n}_{\varepsilon_n, a_n}$,
358: $\Gamma_n:=\Gamma^{\delta_n}_{\varepsilon_n, a_n}$,
359: $\Es_n:=\Es^{\delta_n}_{\varepsilon_n, a_n}$, the following hold:
360: \begin{itemize}
361: \item[(a)] if $\ns$ is the set of discontinuities of $\hs^1(\Gamma(\cdot))$,
362: for all $t \in [0,1] \setminus \ns$ we have
363: \begin{equation}
364: \label{mainconvgradt}
365: \nabla u_n(t) \to \nabla u(t) \quad \mbox{ strongly in }L^2(\Om;\R^2)
366: \end{equation}
367: and
368: \begin{equation}
369: \label{mainconvjumpt}
370: \lim_n \hs^1(\Gamma_n(t))= \hs^1(\Gamma(t));
371: \end{equation}
372: \item[{}]
373: \item[(b)] for all $t \in [0,1]$ we have
374: \begin{equation}
375: \label{mainconvent}
376: \lim_n \Es_n(t)=\Es(t).
377: \end{equation}
378: \end{itemize}
379: \end{theorem}
380:
381: We conclude that we have the convergence of the total energy at each time
382: $t \in [0,1]$, and the separate convergence of bulk and surface energy
383: for all $t \in [0,1]$ except a countable set.
384: \par
385: In order to prove Theorem \ref{mainthm}, we proceed in two steps.
386: Firstly, we fix $a$ and let
387: $\delta \to 0$ and $\eps \to 0$. We obtain an evolution
388: $\{t \to u_a(t)\,:\,t \in [0,1]\}$ such that
389: $\nabla u^{\delta}_{\eps,a}(t) \to \nabla u_a(t)$ strongly in $L^2(\Om;\R^2)$
390: for all $t$ up to a countable set and such that the following minimality
391: property holds: for all $v \in SBV(\Om)$
392: \begin{equation}
393: \label{aminimality}
394: \int_\Om |\nabla \vuat|^2\,dx \le \int_\Om |\nabla v|^2\,dx+ \mu(a)
395: \hs^1 \left( \big( S_v \cup (\partial_D \Om \cap \{v \not= g(t)\}) \big)
396: \setminus \Gamma_a(t) \right),
397: \end{equation}
398: where $\mu:]0,\frac{1}{2}[ \to ]0,+\infty[$ is a function independent of $\eps$ and
399: $\delta$, such that $\mu \ge 1$, $\lim_{a \to 0} \mu(a)=1$ and
400: $\Gamma_a(t):=\bigcup_{s \le t, s \in D} S_{\vuas} \cup (\partial_D \Om \cap \{u_a(s) \not= g(s)\})$.
401: The minimality property \eqref{aminimality} takes into account possible
402: anisotropies that could be generated as $\delta$ and $\eps \to 0$:
403: in fact, since $a$ is fixed, we have that the angles of the triangles in $\ts_{\eps,a}(\Om)$
404: are between fixed values (determined by $a$), and so fractures with
405: certain directions cannot be approximated in length.
406: In the second step, we let $a \to 0$ and determine from $\{t \to u_a(t)\,:\,t \in [0,1]\}$
407: a quasi-static evolution $\{t \to u(t)\,:\,t \in [0,1]\}$ in the sense of Francfort
408: and Larsen. Then, using a diagonal argument, we find sequences $\delta_n \to 0$,
409: $\eps_n \to 0$, and $a_n \to 0$ satisfying Theorem \ref{mainthm}.
410: \par
411: The main difficulties arise in the first part of our analysis, namely when
412: $\delta,\eps \to 0$. The convergence $u^\delta_{\eps,a}(t) \to u_a(t)$ in
413: $SBV(\Om)$ for $t \in D \subseteq [0,1]$ countable and dense is easily
414: obtained by means of Ambrosio's Compactness Theorem. The minimality property
415: \eqref{aminimality} derives from its discrete version \eqref{pieceminintr}
416: using a variant of Lemma 1.2 of \cite{FL}: given $v \in SBV(\Om)$, we
417: construct $v^\delta_{\eps,a} \in {\mathcal A}_{\eps,a}(\Om)$ such that
418: \begin{equation}
419: \label{intrconv1}
420: \nabla v^\delta_{\eps,a} \to \nabla v
421: \quad \mbox{ strongly in }L^2(\Om;\R^2)
422: \end{equation}
423: and
424: \begin{multline}
425: \label{intrconv2}
426: \limsup_{\delta,\eps \to 0}
427: \hs^1 \left[ \big( S_{v^\delta_{\eps,a}} \cup
428: S_D^{g_\eps^\delta(t)}(v^\delta_{\eps,a}) \big)
429: \setminus \Gamma^\delta_{\eps,a}(t) \right] \le \\
430: \le \mu(a)
431: \hs^1 \left[ \big( S_v \cup
432: \big( \partial_D \Om \cap \{v \not= g(t) \} \big) \big)
433: \setminus \Gamma_a(t) \right],
434: \end{multline}
435: where $g_\eps^\delta(t):=g_\eps(t^\delta_i)$ for $t^\delta_i \le t <t^\delta_{i+1}$.
436: The main difference with respect to Lemma 1.2 of \cite{FL} is that we have
437: to find the approximating functions $v^\delta_{\eps,a}$ in the finite
438: element space ${\mathcal A}_{\eps,a}(\Om)$. This can be regarded
439: as an interpolation problem, so we try to construct
440: triangulations $\tb_\eps \in \ts_{\eps,a}(\Om)$ adapted to $v$ in order to obtain
441: \eqref{intrconv1} and \eqref{intrconv2}.
442: In all the geometric operations involved, we need to avoid
443: degeneration of the triangles of $\tb(u^\delta_{\eps,a}(t))$ which is
444: guaranteed from the fact that $a$ is constant: this is the principal reason
445: to keep $a$ fixed in the first step.
446: A second difficulty arises when $u_a(\cdot)$ is extended from $D$ to the entire
447: interval $[0,1]$: indeed it is no longer clear whether
448: $\nabla u^\delta_{\eps,a}(t) \to \nabla u_a(t)$ for $t \not\in D$. Since
449: the space ${\mathcal A}_{\eps,a}(\Om)$ is not a vector space, we cannot
450: provide an estimate on
451: $\|\nabla u^\delta_{\eps,a}(t)-\nabla u^\delta_{\eps,a}(s)\|$ with $s \in D$
452: and $s<t$: we thus cannot expect to recover the convergence at time $t$ from
453: the convergence at time $s$. We overcome this difficulty observing that
454: $\nabla u^\delta_{\eps,a}(t) \to \nabla \tilde{u}_a$ with $\tilde{u}_a$ satisfying a minimality
455: property similar to \eqref{aminimality} and then proving
456: $\nabla \tilde{u}_a=\nabla u_a(t)$ by a uniqueness argument for the gradients of the
457: solutions.
458: \par
459: The plan of the paper is the following. In Section \ref{prel} we give the basic
460: definitions and prove some auxiliary results. In Section \ref{devol}, we
461: prove the existence of a discrete evolution. In Section
462: \ref{convres} we prove the convergence of the discrete evolution to a quasi-static
463: evolution of brittle fractures in the sense of Francfort and Larsen. The proof
464: of minimality property \eqref{aminimality} requires a careful analysis to which is
465: dedicated Section \ref{secmin}. In Section \ref{remark} we show that the arguments
466: of Section \ref{convres} can be used to improve the convergence results
467: for the discrete in time approximation considered in \cite{FL}.
468:
469: \section{Preliminaries}
470: \label{prel}
471:
472: In this section we state the notation and prove some preliminaries employed in
473: the rest of the paper.
474:
475: \vskip10pt
476: {\it Basic notation.}
477: We will employ the following basic notation:
478: \begin{itemize}
479: \item[-] $\Om$ is a polygonal open subset of $\R^2$;
480: \item[-] $L^p(\Om;\R^m)$
481: with $1 \le p < +\infty$ and $m \ge 1$ is the usual Lebesgue space of
482: $p$-summable $\R^m$-valued functions, and $L^p(\Om):=L^p(\Om;\R)$;
483: \item[-] for all $k \ge 1$ and $1 \le p \le +\infty$,
484: $W^{k,p}(\Om)$ is the usual Sobolev space of functions
485: in $L^p(\Om)$ with distributional derivatives of order $1, \ldots, k$ in
486: $L^p(\Om)$; we will write $H^k(\Om)$ for $W^{k,2}(\Om)$;
487: \item[-] if $u \in W^{k,p}(\Om)$, $\nabla u$ is its gradient;
488: \item[-] $\hs^{1}$ is the one-dimensional Hausdorff measure;
489: \item[-] $\|\cdot\|_{\infty}$ denotes the sup-norm;
490: \item[-] if $f \in L^2(\Om;\R^m)$, $\|f\|$ denotes the $L^2$-norm
491: of $f$;
492: \item[-] for all $A \subseteq \R^2$, $|A|$ denotes the Lebesgue measure of $A$;
493: \item[-] if $\mu$ is a measure on $\R^2$ and $A$ is a Borel subset of $\R^2$,
494: $\mu \res A$ denotes the restriction of $\mu$ to $A$, i.e.
495: $(\mu \res A)(B):=\mu(B \cap A)$ for all Borel sets $B \subseteq \R^2$;
496: \item[-] if $\sigma \in ]0,+\infty[$, $o_\sigma$ is such that
497: $\lim_{\sigma \to 0^+} o_\sigma=0$.
498: \end{itemize}
499:
500: \vskip20pt
501: {\it Special functions of bounded variation.}
502: For the general theory of functions of bounded variation, we refer to \cite{AFP}; here
503: we recall some basic definitions and theorems we need in the sequel.
504: Let $A$ be an open subset of $\R^N$, and let $u: A \to \R^n$.
505: We say that $u \in BV(A;\R^n)$ if $u \in L^1(A;\R^n)$, and its distributional
506: derivative is a vector-valued Radon measure on $A$.
507: We say that $u \in SBV(A;\R^n)$ if $u \in BV(A;\R^n)$ and its distributional
508: derivative can be represented as
509: $$
510: Du(A)= \int_A \nabla u(x) \,dx+ \int_{A \cap S_u} (u^+(x)-u^-(x)) \otimes \nu_x
511: \,d\hs^{N-1}(x),
512: $$
513: where $\nabla u$ denotes the approximate gradient of $u$, $S_u$ denotes the set of
514: approximate jumps of $u$, $u^+$ and $u^-$ are the traces of $u$ on $S_u$, $\nu_x$
515: is the normal to $S_u$ at $x$, and $\hs^{N-1}$ is the $(N-1)$-dimensional Hausdorff measure.
516: The space $SBV(A;\R^n)$ is called the space of
517: {\it special functions of bounded variation}. Note that if $u \in SBV(A;\R^n)$, then
518: the singular part of $Du$ is concentrated on $S_u$ which turns out to be countably
519: $\hs^{N-1}$-rectifiable.
520: \par
521: The space $SBV$ is very useful when dealing with variational problems involving
522: volume and surface energies because of the following compactness and lower
523: semicontinuity result due to L.Ambrosio (see \cite{A1}, \cite{A2}, \cite{A3}, \cite{AFP}).
524:
525: \begin{theorem}
526: \label{SBVcompact}
527: Let $A$ be an open and bounded subset of $\R^N$, and let $(u_k)$ be a sequence in
528: $SBV(A;\R^n)$. Assume that there exists $q>1$ and $c \ge 0$ such that
529: $$
530: \int_A |\nabla u_k|^q \,dx+ \hs^{N-1}(S_{u_k})+ ||u_k||_\infty \le c
531: $$
532: for every $k \in \N$. Then there exists a subsequence $(u_{k_h})$ and a function
533: $u \in SBV(A;\R^n)$ such that
534: \begin{align}
535: \label{sbvconv}
536: \nonumber
537: u_{k_h} \to u \quad {strongly \; in}\; L^1(A;\R^n); \\
538: \nabla u_{k_h} \weak \nabla u \quad {weakly \; in}\; L^1(A;M^{N \times n}); \\
539: \nonumber
540: \hs^{N-1}(S_u) \le \liminf_h \hs^{N-1}(S_{u_{k_h}}).
541: \end{align}
542: \end{theorem}
543:
544: In the rest of the paper, we will say that $u_k \to u$ in $SBV(A;\R^n)$ if
545: $u_k$ and $u$ satisfy \eqref{sbvconv}.
546: It will also be useful the following fact which can be derived from
547: Ambrosio's Theorem: if $u_k \to u$ in $SBV(A;\R^n)$ and if
548: $\hs^{N-1} \res S_{u_k} \weakst \mu$ weakly-star in the
549: sense of measures, then $\hs^{N-1} \res S_u \le \mu$ as measures.
550: We will set $SBV(A):=SBV(A;\R)$.
551:
552: \vskip20pt
553: {\it Quasi-static evolution of brittle fracture.}
554: Let $\Om$ be an open bounded subset of $\R^N$ with Lipschitz boundary, and let
555: $\partial_D \Om$ be a subset of $\partial \Om$ open in the relative topology.
556: Let $g:[0,1] \to H^1(\Om)$ be absolutely continuous; we indicate
557: the gradient of $g$ at time $t$ by $\nabla g(t)$, and the time
558: derivative of $g$ at time $t$ by $\dot{g}(t)$. The main result of \cite{FL} is
559: the following theorem.
560:
561: \begin{theorem}
562: \label{qse}
563: There exists a crack $\Gamma(t) \subseteq \Omb$ and a field $u(t) \in SBV(\Om)$ such
564: that:
565: \begin{itemize}
566: \item[(a)] $\Gamma(t)$ increases with $t$;
567: \item[{}]
568: \item[(b)] $u(0)$ minimizes
569: $$
570: \int_\Om |\nabla v|^2 \,dx +
571: \hs^{N-1}( S_v \cup
572: \{x \in \partial_D \Om: v(x) \not= g(0)(x)\})
573: $$
574: among all $v \in SBV(\Om)$ (inequalities on $\partial_D \Om$ are intended for the
575: traces of $v$ and $g$);
576: \item[{}]
577: \item[(c)]
578: for $t>0$, $u(t)$ minimizes
579: $$
580: \int_\Om |\nabla v|^2 \,dx
581: +\hs^{N-1} \left( \left[ S_v \cup
582: \{x \in \partial_D \Om: v(x) \not= g(t)(x)\} \right] \setminus \Gamma(t)
583: \right)
584: $$
585: among all $v \in SBV(\Om)$;
586: \item[{}]
587: \item[(d)]
588: $S_{u(t)} \cup \{x \in \partial_D \Om\,:\, u(t)(x) \not= g(t)(x) \}
589: \subseteq \Gamma(t)$, up to a set of $\hs^{N-1}$-measure $0$.
590: \end{itemize}
591: Furthermore, the total energy
592: $$
593: \Es(t):= \int_\Om |\nabla u(t)|^2 \,dx +\hs^{N-1}( \Gamma(t))
594: $$
595: is absolutely continuous and satisfies
596: $$
597: \Es(t)=\Es(0)+2 \int_0^t \int_\Om \nabla u(\tau) \nabla \dot{g}(\tau) \,dx \,d\tau
598: $$
599: for every $t \in [0,1]$. Finally, for any countable, dense set $I \subseteq [0,1]$,
600: the crack $\Gamma(t)$ and the displacement $u(t)$ can be chosen so that
601: $$
602: \Gamma(t)= \bigcup_{\tau \in I, \tau \le t}
603: \left( S_{u(\tau)} \cup
604: \{x \in \partial_D \Om\,:\, u(\tau)(x) \not= g(\tau)(x) \} \right).
605: $$
606: \end{theorem}
607:
608: The main tool in the proof of Theorem \ref{qse} is the following result
609: \cite[Theorem 2.1]{FL}, which is useful also in our analysis.
610:
611:
612: \begin{theorem}
613: \label{jumptransfer}
614: Let $\Omb \subseteq \Om'$, with $\partial \Om$ Lipschitz, and let for $r=1,\dots,i$
615: $(u^r_n)$ be a sequence in $SBV(\Om')$ such that
616: \begin{itemize}
617: \item[(1)] $S_{u^r_n} \subseteq \Omb$;
618: \item[{}]
619: \item[(2)] $|\nabla u^r_n|$ weakly converges in $L^1(\Om')$; and
620: \item[{}]
621: \item[(3)] $u^r_n \to u^r$ strongly in $L^1(\Om')$,
622: \end{itemize}
623: where $u^r \in BV(\Om')$ with $\hs^{N-1}(S_{u^r}) <\infty$. Then for every $\phi \in
624: SBV(\Om')$ with $\hs^{N-1}(S_\phi) <\infty$ and $\nabla \phi \in L^q(\Om;\R^N)$ for some $q \in [1,+\infty[$,
625: there exists a sequence $(\phi_n)$ in
626: $SBV(\Om')$ with $\phi_n \equiv \phi$ on $\Om' \setminus \Omb$ such that
627: \begin{itemize}
628: \item[(a)] $\phi_n \to \phi$ strongly in $L^1(\Om')$;
629: \item[{}]
630: \item[(b)] $\nabla \phi_n \to \nabla \phi$ strongly in $L^q(\Om')$; and
631: \item[{}]
632: \item[(c)]
633: $\hs^{N-1} \left( [S_{\phi_n} \setminus \bigcup_{r=1}^i S_{u^r_n}] \setminus
634: [S_{\phi} \setminus \bigcup_{r=1}^i S_{u^r}] \right) \to 0$.
635: \end{itemize}
636: In particular
637: \begin{equation}
638: \label{jt}
639: \limsup_n \hs^{N-1} \left( S_{\phi_n} \setminus \bigcup_{r=1}^i S_{u^r_n} \right) \le
640: \hs^{N-1} \left( S_{\phi} \setminus \bigcup_{r=1}^i S_{u^r} \right).
641: \end{equation}
642: \end{theorem}
643:
644: \vskip 20pt
645: {\it Hausdorff metric on compact sets.}
646: Let $A \subseteq \R^2$ be open and bounded, and let $\ks(\overline{A})$
647: be the set of all compact subsets of $\overline{A}$.
648: $\ks(\overline{A})$ can be endowed by the
649: Hausdorff metric $d_H$ defined by
650: $$
651: d_H(K_1,K_2) := \max \left\{ \sup_{x \in K_1} {\rm dist}(x,K_2), \sup_{y \in
652: K_2} {\rm dist}(y,K_1)\right\},
653: $$
654: with the conventions ${\rm dist}(x, \emptyset)= {\rm diam}(A)$ and $\sup
655: \emptyset=0$, so that $d_H(\emptyset, K)=0$ if $K=\emptyset$ and
656: $d_H(\emptyset,K)={\rm diam}(A)$ if $K \not=\emptyset$. It turns out that
657: $\ks(\overline{A})$ endowed with the Hausdorff metric is a compact space
658: (see e.g. \cite{Ro}).
659:
660:
661: \vskip20pt
662: {\it Triangulations.}
663: Let $\Om \subseteq \R^2$ be a polygonal set and let us fix two positive
664: constants $0<c_1<c_2$. By a {\it regular triangulation} of $\Om$ of
665: size $\eps$ we intend a finite family of (closed) triangles $T_i$ such that $\Omb=\bigcup_i T_i$,
666: $T_i \cap T_j$ is either empty or equal to a common edge or to a common vertex,
667: and each $T_i$ contains a ball of diameter $c_1 \eps$ and is contained in a ball of
668: diameter $c_2 \eps$.
669: \par
670: We indicate by $\rs_\eps(\Om)$ the family of all regular triangulations of
671: $\Om$ of size $\eps$.
672: It turns out that there exist $0<\vartheta_1 < \vartheta_2 <\pi$
673: such that for all $T$ belonging to a triangulation
674: $\tb \in \rs_\eps(\Om)$, the inner angles of $T$ are between
675: $\vartheta_1$ and $\vartheta_2$. Moreover, every edge of $T$ has length greater than
676: $c_1 \eps$ and lower than $c_2 \eps$.
677: \par
678: Let us fix a triangulation $\rb_\eps \in \rs_\eps(\Om)$ for all $\eps>0$
679: and let $a \in ]0,\frac{1}{2}[$.
680: Let us consider a new triangulation $\tb$ nested in $\rb_\eps$ obtained
681: dividing each $T \in \rb_\eps$ into four triangles taking over every edge $[x,y]$ of $T$
682: a knot $z$ which satisfies
683: $$
684: z=tx+(1-t)y, \quad \quad t \in [a,1-a].
685: $$
686: We will call these new vertices {\it adaptive}, the triangles obtained joining these points
687: {\it adaptive triangles}, and their edges {\it adaptive edges} (see Fig.1).
688: \begin{center}
689: \psfig{figure=fig1.eps}
690: \end{center}
691: We denote by $\ts_{\eps,a}(\Om)$ the set of all triangulations $\tb$ constructed in this way.
692: Note that for all $\tb \in \ts_{\eps,a}(\Om)$ there exists $0<c_1^a<c_2^a<+\infty$ such that
693: every $T_i \in \tb$ contains a ball of diameter $c_1^a \eps$ and is contained in a ball
694: of diameter $c_2^a \eps$. Then there exist $0<\vartheta_1^a < \vartheta_2^a <\pi$
695: such that for all triangles $T$ belonging to a triangulation
696: $\tb \in \ts_{\eps,a}(\Om)$, the inner angles of $T$ are between
697: $\vartheta_1^a$ and $\vartheta_2^a$. Moreover, every edge of $T$ has length greater than
698: $c_1^a \eps$ and lower than $c_2^a \eps$.
699: \par
700: We will often use the following {\it interpolation estimate} (see \cite[Theorem 3.1.5]{Cia}).
701: If $u \in W^{2,2}(\Om)$ and $T \in \rb_\eps$, let $u_T$ denote the affine interpolation
702: of $u$ on $T$. We have that there exists $\cint$ depending only on $c_1,c_2$ such that
703: \begin{equation}
704: \label{interp1}
705: \|u_T -u\|_{W^{1,2}(T)} \le \cint \eps \|u\|_{W^{2,2}(T)}.
706: \end{equation}
707: Estimate \eqref{interp1} holds also for
708: $\tb \in \ts_{\eps,a}(\Om)$: in this case $K$ depends on $a$.
709:
710:
711: \vskip20pt
712: {\it Some elementary constructions.}
713: The following lemmas will be used in Section 4.
714:
715: \begin{lemma}
716: \label{simplecurve}
717: Let $\tb \in \ts_{\eps,a}(\Om)$, and let $l \subseteq \Om$
718: be a segment with extremes $p,\,q$ belonging to edges of $\tb$.
719: There exists a polyhedral curve $\Gamma$ with extremal points $p$ and $q$
720: (see Fig.2) such that $\Gamma$ is contained in the union of the edges of those
721: $T \in \tb$ with $T \cap l \neq \emptyset$, and such that the following
722: properties hold:
723: \begin{itemize}
724: \item[(1)]
725: $\Gamma = \gamma_p \cup \gamma \cup \gamma_q$, where $\gamma$ is union of edges
726: of $\tb$ and $\gamma_p$, $\gamma_q$ are segments containing $p$ and $q$ respectively,
727: and each one is contained in an edge of $\tb$;
728: \item[]
729: \item[(2)]
730: there exists a constant $c$ independent of $\epsilon$ (but depending on $a$) such that
731: $$
732: \hs^1(\Gamma)\le c \, \hs^1(l).
733: $$
734: \end{itemize}
735: \end{lemma}
736:
737: \begin{center}
738: \psfig{figure=fig2.eps}
739: \end{center}
740:
741: \begin{proof}
742: Let $\{T_1, \ldots T_k\}$ be the family of
743: triangles in $\tb$ such that the intersection with $l$ is a segment with positive
744: length. For every integer $1\le i \le k$,
745: let $l_i:=T_i\cap l.$ If $l_i$ is an edge of $T_i$, we set $D_i= T_i$. Otherwise
746: let $D_i$ be a connected component of $T_i \setminus l_i$ such that
747: $|D_i| \leq \frac{1}{2} |T_i|.$
748: We claim that there exists a constant $c>0$ independent of $\epsilon$ such that
749: \begin{equation}
750: \label{inelung}
751: \hs^1 (\partial D_i) \leq c\, \hs^1(l_i).
752: \end{equation}
753: We have to analyze two possibilities, namely $D_i$ is a triangle, or $D_i$ is a
754: trapezoid. Suppose that $D_i$ is a triangle and that $m_i$ is an edge of $D_i$.
755: Let $\alpha$ be the angle of $D_i$ opposite to
756: $l_i$. It is easy to prove that $\hs^1(l_i) \ge \hs^1(m_i) \sin\alpha$,
757: and so
758: $$
759: \hs^1(l_i) \ge \frac{1}{3} \sin\alpha \hs^1(\partial D_i).
760: $$
761: Since $\vartheta_1^a\le\alpha\le\vartheta_2^a$, $\sin\alpha$ is uniformly bounded
762: from below, and hence inequality (\ref{inelung}) follows.
763: If $D_i$ is a trapezoid, since $|D_i| \leq \frac{1}{2} |T_i|$,
764: it follows that $T_i\setminus D_i$ is a triangle such that
765: its edges different from $l_i$ have length greater than
766: $\frac{1}{2} c_1^a \eps$. Let $\alpha$ be the inner angle of $T_i\setminus D_i$
767: opposite to $l_i$. We have that
768: $$
769: \hs^1(l_i) \ge \frac{1}{2}\sin\alpha c_1^a {\eps} \ge
770: \frac{1}{2} \sin\alpha \frac {c_1^a}{c_2^a}\frac{1}{4} \hs^1(\partial D_i).
771: $$
772: Since $\vartheta_1^a\le\alpha\le\vartheta_2^a$, inequality (\ref{inelung}) follows.
773: \par
774: By \eqref{inelung}, we deduce that
775: \begin{equation*}
776: \hs^1 \left( \bigcup_{i=1}^k\partial D_i \right) \leq c\, \hs^1(l);
777: \end{equation*}
778: moreover, since $\bigcup_{i=1}^k\partial D_i$ is arcwise connected and contains
779: $p,q$, we conclude that there exists a curve $\Gamma \subseteq
780: \bigcup_{i=1}^k\partial D_i$ which satisfies the thesis.
781: \end{proof}
782:
783: \begin{lemma}
784: \label{guscio}
785: There exists a constant $c>0$ such that for
786: every segment $l \subseteq \Om$ there exists $\eps_0$ with the following property:
787: for every $\varepsilon \leq \varepsilon_0$, setting
788: $\rs(l):=\{T\in \rb_\eps:\, T \cap l \neq \emptyset\}$, we have
789: $$
790: \hs^1(\partial \rs(l))\leq c \hs^1(l).
791: $$
792: \end{lemma}
793:
794: \begin{proof}
795: Let $ \ns_{\eps}(l):=\{ x \in \Om:\, {\rm dist}(x,l) \leq c_2 \eps\}$.
796: We have that
797: $|\ns_{\eps}(l)| = \hs^1(l)c_2\eps+ \pi c_2^2\eps^2$,
798: and hence there exists a positive constant $\eps_0$ such that,
799: for every $\eps \leq \eps_0$, we have that
800: $$
801: |\ns_{\eps}(l)| \leq 2 \hs^1(l) c_2 \eps.
802: $$
803: We have that $\rs(l) \subseteq \ns_{\eps}(l),$ and
804: $$
805: \sharp \rs(l) \le \frac{4}{c_1^2\pi^2} \frac{|\ns_{\eps}(l)|}{\eps^2},
806: $$
807: where $\sharp \rs(l)$ denotes the number of triangles of $\rs(l)$.
808: Then, we have
809: $$
810: \hs^1(\partial \rs(l)) \leq 3 c_2 \eps\, \sharp\rs(l) \leq
811: 3 c_2 \eps \frac{4}{c_1^2\pi^2}
812: \frac{|\ns_{\eps}(l)|}{\eps^2} \leq
813: 3 c_2^2 \frac{4}{c_1^2\pi^2} 2\hs^1(l),
814: $$
815: and so the proof is concluded.
816: \end{proof}
817:
818: \vskip20pt
819: {\it A density result.}
820: Let $A \subseteq \R^2$ be open.
821: We say that $K \subseteq A$ is polygonal (with respect to $A$),
822: if it is the intersection of $A$ with the union of a finite number of closed segments.
823: The following density result is proved in \cite{C}.
824:
825: \begin{theorem}
826: \label{piecedensity}
827: Assume that $\partial A$ is locally Lipschitz, and let $u \in SBV(A)$ such that
828: $u \in L^2(A)$, $\nabla u \in L^2(\Om;\R^2)$, and $\hs^{1}(S_u)<+\infty$. For
829: every $\eps>0$, there exists a function $v \in SBV(A)$ such that
830: \begin{itemize}
831: \item[(a)]
832: $S_v$ is essentially closed, i.e., $\hs^{1}( \overline{S_v} \setminus S_v)=0$;
833: \item[{}]
834: \item[(b)]
835: $\overline{S_v}$ is a polyhedral set;
836: \item[{}]
837: \item[(c)]
838: $v \in W^{k, \infty}(A \setminus \overline{S_v})$ for every $k \in \N$;
839: \item[{}]
840: \item[(d)]
841: $||v-u||_{L^2(A)} < \eps$;
842: \item[{}]
843: \item[(e)]
844: $||\nabla v- \nabla u||_{L^2(A; \R^2)} < \eps$;
845: \item[{}]
846: \item[(f)]
847: $|\hs^{1}(S_v)-\hs^{1}(S_u)| <\eps$.
848: \end{itemize}
849: \end{theorem}
850:
851: Let $\partial_D \Om$ be a relatively open subset of $\partial \Om$ composed of edges lying in
852: $\partial \Om$. Let us consider $\Om_D$ polygonal open bounded subset of $\R^2$ such that
853: $\Om_D \cap \Om=\emptyset$ and $\partial \Om \cap \partial \Om_D= \partial_D \Om$ up to a finite
854: number of vertices. We set $\Om':= \Om \cup \Om_D \cup \partial_D \Om$.
855: In Section 4, we will use the following result.
856:
857: \begin{proposition}
858: \label{regularization}
859: Given $u \in SBV(\Om')$ with $u=0$ on $\Om' \setminus \Omb$ and $\hs^{N-1}(S_u)<+\infty$,
860: there exists $u_h \in SBV(\Om')$ such that
861: \begin{itemize}
862: \item[(a)] $u_h=0$ in $\Om' \setminus \Omb$;
863: \item[{}]
864: \item[(b)] $S_{u_h}$ is polyhedral,
865: $\overline{S_{u_h}} \subseteq \Om$ and $u_h \in
866: W^{k,\infty}(\Om' \setminus \overline{S_{u_h}})$ for all $k$;
867: \item[{}]
868: \item[(c)] $u_h \to u$ strongly in $L^2(\Om')$ and
869: $\nabla u_h \to \nabla u$ strongly in $L^2(\Om';\R^2)$;
870: \item[{}]
871: \item[(d)] for all $A$ open subset of $\Om'$ with $\hs^1(\partial A \cap S_u)=0$,
872: we have
873: $$
874: \lim_h \hs^1(A \cap S_{u_h})=\hs^1(A \cap S_u).
875: $$
876: \end{itemize}
877: \end{proposition}
878:
879: \begin{proof}
880: Using a partition of unity, we may prove the result in the case $\Om:=]-1,1[ \times ]0,1[$,
881: $\Om':=]-1,1[ \times ]-1,1[$, and $\partial_D \Om:=]-1,1[ \times \{0\}$.
882: We set $w_h(x,y):=u(x,y-h)$, and let $\varphi_h$ be a cut off function with $\varphi_h=1$ on
883: $]-1,1[ \times ]-1,\frac{h}{3}[$, $\varphi_h=0$ on $]-1,1[ \times ]\frac{h}{2},1[$, and
884: $||\nabla \varphi_h||_\infty \le \frac{7}{h}$. Let us set $v_h:=(1-\varphi_h)w_h$.
885: We have that $v_h=0$ in $\Om' \setminus \Omb$; moreover we have
886: $$
887: \nabla v_h=(1-\varphi_h) \nabla w_h-\nabla \varphi_h w_h.
888: $$
889: Since $\nabla \varphi_h w_h \to 0$ strongly in $L^2(\Om';\R^2)$, we have
890: $\nabla v_h \to \nabla u$ strongly in $L^2(\Om';\R^N)$. Finally, for all
891: $A$ open subset of $\Om'$ with $\hs^1(\partial A \cap S_u)=0$,
892: we have
893: $$
894: \lim_h \hs^1(A \cap S_{v_h})=\hs^1(A \cap S_u).
895: $$
896: In order to conclude the proof, let us apply Theorem \ref{piecedensity}
897: obtaining $\tilde{v}_h$ with polyhedral jumps in $\Om$ such that $\tilde{v}_h
898: \in W^{k,\infty}(\Om' \setminus \overline{S_{\tilde{v}_h}})$,
899: $||w_h-\tilde{v}_h||_{L^2(\Om)}+
900: ||\nabla w_h -\nabla \tilde{v}_h||_{L^2(\Om;\R^2)} \le h^2$ and
901: $|\hs^{N-1}(S_{w_h}) -\hs^{N-1}(S_{\tilde{v}_h})| \le h$. If we set
902: $u_h:= \varphi_h g+(1-\varphi_h) \tilde{v}_h$, we obtain the thesis.
903: \end{proof}
904:
905: \section{The discontinuous finite element approximation}
906: \label{devol}
907:
908: In this section we construct a discrete approximation of quasi-static evolution
909: of brittle fractures in linearly elastic bodies: the discretization is done both in
910: space and time.
911: \par
912: From now on we suppose that $\Om$ is a polygonal open bounded subset
913: of $\R^2$, and that $\partial_D \Om \subseteq \partial \Om$ is open in the relative
914: topology.
915: For all $\varepsilon>0$, we fix a triangulation $\rb_\varepsilon \in \rs_\varepsilon(\Om)$,
916: and suppose that $\partial_D \Om$ is composed of edges of $\rb_\eps$ for
917: all $\eps$; we indicate the family of these edges by $\seub_\eps$.
918: \par
919: We consider the following discontinuous finite element space.
920: We indicate by $\afeaom$ the set of all $u$ such that there exists a triangulation
921: $\tb(u) \in \treaom$ nested in $\rb_\varepsilon$ with $u$ affine on every
922: $T \in \tb(u)$. For every $u \in \afeaom$, we write $\|\nabla u\|$
923: for the $L^2$-norm of $\nabla u$ and we indicate by $S_u$ the family of edges of
924: $\tb(u)$ inside $\Om$ across which $u$ is discontinuous. Notice that $u \in SBV(\Om)$
925: and that the notation is consistent with the usual one employed in the theory of
926: functions with bounded variation. Let us also denote by $\afeom$ the set of affine
927: functions in $\Om$ with respect to the triangulation $\rb_\eps$.
928: Finally, given any $g \in \afeom$, for all $u \in \afeaom$ we set
929: \begin{equation}
930: \label{jump*}
931: S_D^g(u):= \{\zeta \in \seub_\eps\,:\, u \not=g \mbox{ on } \zeta\},
932: \end{equation}
933: that is $S_D^g(u)$ denotes the edges at which the boundary condition is not
934: satisfied. Moreover we set
935: \begin{equation}
936: \label{jumpbis}
937: \Sg{g}{u}:= S_u \cup S_D^g(u)
938: \end{equation}
939: Let now consider $g \in W^{1,1}([0,1];H^1(\Om))$ with $g(t) \in \afeom$ for all
940: $t \in [0,1]$. Let $\delta>0$ and let $N_\delta$ be the largest integer such that
941: $\delta (N_\delta-1) < 1$; for $0\le i \le N_\delta-1$ we set $t_i^\delta:=i\delta$,
942: $t^\delta_{N_\delta}:=1$ and $g_i^\delta:=g(t_i^\delta)$.
943: The following proposition holds.
944:
945: \begin{proposition}
946: \label{discrevol}
947: Let $\eps>0$, $a \in ]0,\frac{1}{2}]$ and $\delta>0$ be fixed. Then for all
948: $i=0, \ldots, N_\delta$ there exists $u^{\delta,i}_{\eps,a} \in \afeaom$
949: such that, setting
950: \begin{equation*}
951: \Gamma^{\delta,i}_{\eps,a}:=
952: \bigcup_{r=0}^i \Sg{g^\delta_r}{u^{\delta,r}_{\eps,a}},
953: \end{equation*}
954: the following hold:
955: \begin{itemize}
956: \item[]
957: \item[(a)] $\|u^{\delta,i}_{\eps,a}\|_\infty \le \|g^\delta_i\|_\infty$;
958: \item[{}]
959: \item[(b)] for all $v \in \afeaom$ we have
960: \begin{equation}
961: \label{MSdiscr}
962: \|\nabla u^{\delta,0}_{\eps,a}\|^2+
963: \hs^1 \left(
964: \Sg{g_0^\delta}{u^{\delta,0}_{\eps,a}} \right)
965: \le \|\nabla v\|^2 +
966: \hs^1\left( \Sg{g_0^\delta}{v} \right),
967: \end{equation}
968: and
969: \begin{equation}
970: \label{piecemin}
971: \|\nabla u^{\delta,i}_{\eps,a}\|^2 \le \|\nabla v\|^2 +
972: \hs^1\left( \Sg{g^\delta_i}{v} \setminus
973: \Gamma^{\delta,i-1}_{\eps,a}\right).
974: \end{equation}
975: \end{itemize}
976: \end{proposition}
977:
978: \begin{proof}
979: The proof is carried out through a variational argument.
980: Let $u^{\delta,0}_{\eps,a}$ be a minimum of the following problem
981: \begin{equation}
982: \label{step0}
983: \min
984: \left\{ \|\nabla u\|^2+ \hs^1(\Sg{g^\delta_0}{u}) \right\}.
985: \end{equation}
986: We set $\Gamma^{\delta,0}_{\eps,a}:= \Sg{g^\delta_0}{u^{\delta,0}_{\eps,a}}$.
987: Recursively, supposing to have constructed $u^{\delta,i-1}_{\eps,a}$ and
988: $\Gamma^{\delta,i-1}_{\eps,a}$, let $u^{\delta,i}_{\eps,a}$ be a minimum for
989: \begin{equation}
990: \label{stepj}
991: \min \left\{ \|\nabla u\|^2 +
992: \hs^1 \left( \Sg{g^\delta_i}{u} \setminus
993: \Gamma^{\delta,i-1}_{\eps,a} \right) \right\}.
994: \end{equation}
995: We set $\Gamma^{\delta,i}_{\eps,a}:=
996: \Sg{g^\delta_i}{u^{\delta,i}_{\eps,a}} \cup \Gamma^{\delta,i-1}_{\eps,a}$.
997: We claim that problems \eqref{step0} and \eqref{stepj} admit a solution
998: $u^{\delta,i}_{\eps,a}$ such that
999: $\|u^{\delta,i}_{\eps,a}\|_{\infty} \le \|g^\delta_i\|_\infty$ for all
1000: $i=0, \ldots, N_\delta$.
1001: We prove the claim for problem \eqref{stepj}, the other case being similar.
1002: Let $(u_n)$ be a minimizing sequence for problem \eqref{stepj}: since
1003: $g^\delta_i$ is an admissible test function, we deduce that
1004: for $n$ large
1005: \begin{equation*}
1006: \|\nabla u_n\|^2+ \hs^1 \left( \Sg{g^\delta_i}{u_n}
1007: \setminus \Gamma^{\delta,i-1}_{\eps,a} \right) \le
1008: \|\nabla g^\delta_i\|^2.
1009: \end{equation*}
1010: Moreover, we may modify $u_n$ in the following way.
1011: If $\pi$ denotes the projection in
1012: $\R$ over the interval $I:=[-\|g^\delta_i\|_\infty, \|g^\delta_i\|_\infty]$, let
1013: $\tilde{u}_n \in \afeaom$ be defined on each $T \in \tb(u_n)$ as the affine
1014: interpolation of the values $(\pi(u_n(x_1)),\; \pi(u_n(x_2)),\; \pi(u_n(x_3))$, where
1015: $x_1, \;x_2$ and $x_3$ are the vertices of $T$. Note that by construction we have
1016: for all $n$
1017: \begin{equation*}
1018: \|\tilde{u}_n\|_\infty \le \|g^\delta_i\|_\infty, \quad \quad
1019: \|\nabla \tilde{u}_n\| \le \|\nabla u_n\|, \quad \quad
1020: \Sg{g^\delta_i}{\tilde{u}_n} \subseteq
1021: \Sg{g^\delta_i}{u_n},
1022: \end{equation*}
1023: so that $(\tilde{u}_n)$ is a minimizing sequence for problem
1024: \eqref{stepj}. We conclude that it is not restrictive to assume
1025: $\|u_n\|_\infty \le \|g^\delta_i\|_\infty$.
1026: \par
1027: Since $\tb(u_n) \in \treaom$, we have that the number of elements of
1028: $\tb(u_n)$ is uniformly bounded. Up to a subsequence,
1029: we may suppose that there exists an integer $k$ such that $\tb(u_n)$ has exactly $k$
1030: elements $T_n^1, \ldots, T_n^k$. Using a diagonal argument we may suppose that, up
1031: to a further subsequence, there exists $\tb=\{T^1, \ldots, T^k\} \in \treaom$ such that
1032: $T_n^i \to T^i$ in the Hausdorff metric for all $i=1, \ldots, k$. Let us consider
1033: $T^i \in \tb$, and let $\tilde{T}^i$ be contained in the interior of $T^i$. For $n$
1034: large enough, $\tilde{T}^i$ is contained in the interior of $T^i_n$ and
1035: $(u_n)_{|\tilde{T}^i}$ is affine with
1036: $\int_{\tilde{T}^i} |\nabla u_n|^2 \,dx \le C$ with $C$ independent of $n$.
1037: We deduce that there exists a function $u^i$ affine on $\tilde{T}^i$ such that up to a
1038: subsequence $u_n \to u$ uniformly on $\tilde{T}^i$. Since $\tilde{T}^i$ is arbitrary,
1039: it turns out that $u^i$ is actually defined on $T^i$ and
1040: \begin{equation*}
1041: \int_{T^i} |\nabla u^i|^2 \,dx \le \liminf_n \int_{T_n^i} |\nabla u_n|^2 \,dx.
1042: \end{equation*}
1043: Let $u \in \afeaom$ such that $u=u^i$ on $T^i$ for every $i=1,\dots,k$: we have
1044: \begin{equation*}
1045: \|\nabla u\|^2 \le \liminf_n \|\nabla u_n\|^2.
1046: \end{equation*}
1047: On the other hand, it is easy to see that
1048: $\Sg{g^\delta_i}{u}$ is contained in the Hausdorff limit
1049: of $\Sg{g^\delta_i}{u_n}$, and that
1050: \begin{equation*}
1051: \hs^1 \left(
1052: \Sg{g^\delta_i}{u} \setminus \Gamma^{\delta,i-1}_{\eps,a}
1053: \right)
1054: \le
1055: \liminf_n \hs^1 \left(
1056: \Sg{g^\delta_i}{u_n}
1057: \setminus \Gamma^{\delta,i-1}_{\eps,a}
1058: \right).
1059: \end{equation*}
1060: We conclude that $u$ is a minimum point for the problem \eqref{stepj} with
1061: $\|u\|_\infty \le \|g^\delta_i\|_\infty$. We have that point $(a)$ is proved.
1062: \par
1063: Concerning point $(b)$, by construction we get
1064: \eqref{MSdiscr}; for $i \ge 1$ we have
1065: \begin{equation*}
1066: \|\nabla u^{\delta,i}_{\eps,a}\|^2 +
1067: \hs^1 \left(
1068: \Sg{g^\delta_i}{u^{\delta,i}_{\eps,a}}
1069: \setminus \Gamma^{\delta,i-1}_{\eps,a} \right) \le
1070: \|\nabla v\|^2 +
1071: \hs^1 \left( \Sg{g^\delta_i}{v}
1072: \setminus \Gamma^{\delta,i-1}_{\eps,a} \right)
1073: \end{equation*}
1074: for all $v \in \afeaom$, so that
1075: \begin{equation*}
1076: \|\nabla u^{\delta,i}_{\eps,a}\|^2 \le
1077: \|\nabla v\|^2 +
1078: \hs^1 \left( \Sg{g^\delta_i}{v} \setminus
1079: \Gamma^{\delta,i-1}_{\eps,a} \right),
1080: \end{equation*}
1081: and this proves point $(b)$.
1082: \end{proof}
1083:
1084: \begin{remark}
1085: \label{oss}
1086: {\rm
1087: For technical reasons due to the asymptotic analysis of the
1088: discrete evolution $u^{\delta,i}_{\eps,a}$ when $\delta \to 0$,
1089: $\eps \to 0$ and $a \to 0$, we define $u^{\delta,i}_{\eps,a}$
1090: from $u^{\delta,i-1}_{\eps,a}$ through problem \eqref{stepj}
1091: without requiring that the adaptive vertices determining
1092: $\Gamma^{\delta,i-1}_{\eps,a}$ remain fixed. We just
1093: penalize their possible changes if they are used to create new
1094: fracture: in fact in this case, the surface energy increases at each change
1095: of a quantity at least of order $a\eps$.
1096: As a consequence, during the step by step minimization, it could
1097: happen that some triangles $T \in \ts_{\eps,a}(\Om)$ contain the
1098: fracture $\Gamma^{\delta,i}_{\eps,a}$ in their interior.
1099: This is in contrast with the interpretation of the triangles as elementary
1100: blocks for the elasticity problem, but being this situation penalized in the
1101: minimization process, we expect that it occurs rarely.
1102: }
1103: \end{remark}
1104:
1105: The following estimate is essential for the study of asymptotic behavior of
1106: the discrete evolution.
1107:
1108: \begin{proposition}
1109: \label{dener}
1110: If $(u^{\delta,i}_{\eps,a},\Gamma^{\delta,i}_{\eps,a})$ for $i=0,\dots,N_\delta$ satisfies condition
1111: $(b)$ of Proposition \ref{discrevol}, setting $\Es^{\delta,i}_{\eps,a}:=\|\nabla u^{\delta,i}_{\eps,a}\|^2+
1112: \hs^1 \left( \Gamma^{\delta,i}_{\eps,a}\right)$, we have
1113: for $0 \le j \le i \le N_{\delta}$
1114: \begin{equation}
1115: \label{discrenergy}
1116: \Es^{\delta,i}_{\eps,a} \le \Es^{\delta,j}_{\eps,a}+
1117: 2 \sum_{r=j}^{i-1} \int_{t^\delta_{r}}^{t^\delta_{r+1}}
1118: \int_{\Om} \nabla u^{\delta,r}_{\eps,a} \nabla\dot{g}(\tau)
1119: \,dx\,d\tau +o^\delta,
1120: \end{equation}
1121: where
1122: \begin{equation}
1123: \label{odelta}
1124: o^\delta:= \left[
1125: \max_{r=0,\ldots,N_\delta-1}
1126: \int_{t^\delta_{r}}^{t^\delta_{r+1}} \|\dot{g}(\tau)\|_{H^1(\Om)} \,d\tau
1127: \right]
1128: \int_0^1 \|\dot{g}(\tau)\|_{H^1(\Om)}\,d\tau.
1129: \end{equation}
1130: \end{proposition}
1131:
1132: \begin{proof}
1133: For all $0 \le j \le N_\delta-1$,
1134: by construction of $u^{\delta,j+1}_{\eps,a}$ we have that
1135: \begin{multline*}
1136: \|\nabla u^{\delta,j+1}_{\eps,a}\|^2 +
1137: \hs^1 \left(
1138: \Sg{g^\delta_{j+1}}{u^{\delta,j+1}_{\eps,a}}
1139: \setminus \Gamma^{\delta,j}_{\eps,a} \right)
1140: \le
1141: \|\nabla u^{\delta,j}_{\eps,a}+ \nabla (g^\delta_{j+1} - g^\delta_{j})\|^2 =\\
1142: =
1143: \|\nabla u^{\delta,j}_{\eps,a}\|^2 +
1144: 2 \int_{\Om}
1145: \nabla u^{\delta,j}_{\eps,a} \nabla (g^\delta_{j+1} - g^\delta_{j})
1146: \,dx+||\nabla (g^\delta_{j+1}-g^\delta_{j})||^2.
1147: \end{multline*}
1148: Notice that
1149: \begin{equation*}
1150: \nabla (g^\delta_{j+1}-g^\delta_{j}) =
1151: \int_{t^\delta_{j}}^{t^\delta_{j+1}}
1152: \nabla \dot{g}(\tau) \,d\tau,
1153: \end{equation*}
1154: so that
1155: \begin{multline}
1156: \label{ineqi}
1157: \|\nabla u^{\delta,j+1}_{\eps,a}\|^2 +
1158: \hs^1 \left(
1159: \Sg{g^\delta_{j+1}}{u^{\delta,j+1}_{\eps,a}}
1160: \setminus \Gamma^{\delta,j}_{\eps,a} \right) \le \\
1161: \le \|\nabla u^{\delta,j}_{\eps,a}\|^2 +
1162: 2 \int_{t^\delta_{j}}^{t_{j+1}^\delta} \int_{\Om}
1163: \nabla u^{\delta,j}_{\eps,a} \nabla \dot{g}(\tau)\,dx \,d\tau
1164: +e(\delta)\int_{t^\delta_{j}}^{t^\delta_{j+1}} \|\dot{g}(\tau)\|_{H^1(\Om)}
1165: \,d\tau,
1166: \end{multline}
1167: where
1168: \begin{equation*}
1169: e(\delta):= \max_{r=0,\ldots,N_\delta-1}
1170: \int_{t^\delta_{r}}^{t^\delta_{r+1}} \|\dot{g}(\tau)\|_{H^1(\Om)} \,d\tau.
1171: \end{equation*}
1172: From \eqref{ineqi}, we obtain that for all $0 \le j \le i \le N_\delta$
1173: \begin{multline*}
1174: ||\nabla u^{\delta,i}_{\eps,a}||^2+ \hs^1(\Gamma^{\delta,i}_{\eps,a}) \le
1175: ||\nabla u^{\delta,j}_{\eps,a}||^2 + \hs^1(\Gamma^{\delta,j}_{\eps,a})+ \\
1176: +2 \sum_{r=j}^{i-1} \int_{t^\delta_{r}}^{t^\delta_{r+1}} \int_{\Om}
1177: \nabla u^{\delta,r}_{\eps,a} \nabla \dot{g}(\tau) \,dx\,d\tau
1178: +e(\delta) \int_{t^\delta_{j}}^{t^\delta_{i}}
1179: ||\dot{g}(\tau)||_{H^1(\Om)} \,d\tau,
1180: \end{multline*}
1181: and so the proof of point $(c)$ is complete choosing
1182: $$
1183: o^\delta:=e(\delta) \int_0^1 \|\dot{g}(\tau)\|_{H^1(\Om)}\,d\tau.
1184: $$
1185: \end{proof}
1186:
1187: \section{The convergence result}
1188: \label{convres}
1189: This section is devoted to the proof of Theorem \ref{mainthm}.
1190: As in Section \ref{devol}, let $\Om$ be a polygonal open bounded subset of $\R^2$, and
1191: let $\partial_D \Om \subseteq \partial \Om$ be open in the relative topology.
1192: For all $\varepsilon>0$, let $\rb_\varepsilon \in \rs_\varepsilon(\Om)$ be
1193: a regular triangulation of $\Om$ such that $\partial_D \Om$ is composed of edges of $\rb_\eps$.
1194: As in the previous section, let $\afeom$ be the family of continuous piecewise affine
1195: functions with respect to $\rb_\eps$, and let $\afeaom$ be the family of functions which are affine
1196: on the triangles of some triangulation $\tb \in \treaom$ nested in $\rb_\eps$ and can jump across
1197: the edges of $\tb$.
1198: \par
1199: In the following, it will be useful to treat points at which the boundary condition
1200: is violated (see \eqref{jump*}) as internal jumps. Thus we
1201: consider $\Om_D$ polygonal open bounded subset of $\R^2$ such that
1202: $\Om_D \cap \Om=\emptyset$ and $\partial \Om \cap \partial \Om_D= \partial_D \Om$ up to a finite number of
1203: points;
1204: we set $\Om':= \Om \cup \Om_D \cup \partial_D \Om$.
1205: Given $u \in \afeaom$ and $g \in \afeom$, we may extend
1206: $g$ to a function of $H^1(\Om')$ and $u$ to a function
1207: $\tilde{u} \in SBV(\Om')$ setting $\tilde{u}=g$ on $\Om_D$.
1208: In this way, recalling \eqref{jumpbis}, we have
1209: \begin{equation*}
1210: \Sg{g}{u}=S_{\tilde{u}},
1211: \end{equation*}
1212: so that the violation of the boundary condition of $u$ can be read in the
1213: set of jumps of $\tilde{u}$.
1214: Analogously, given $u \in SBV(\Om)$ and $g \in H^1(\Om)$, we set
1215: \begin{equation}
1216: \label{jumptris}
1217: \Sg{g}{u}:=S_u \cup \{x \in \partial_D \Om\,:\, \gamma(u)(x) \not= \gamma(g)(x)\}
1218: \end{equation}
1219: where $\gamma$ denotes the trace operator on $\partial \Om$.
1220: We may assume $g \in H^1(\Om')$ using an extension operator.
1221: We can then consider $\tilde{u} \in SBV(\Om')$
1222: such that $\tilde{u}=u$ on $\Om$, and $\tilde{u}=g$ on $\Om_D$.
1223: In this way we have
1224: \begin{equation*}
1225: \Sg{g}{u}=S_{\tilde{u}} \quad \mbox{ up to a set of $\hs^1${-}measure $0$}.
1226: \end{equation*}
1227: Let us consider $g \in W^{1,1}([0,1],H^1(\Om))$ such that $\|g(t)\|_\infty \le C$
1228: for all $t \in [0,1]$ and let $g_\eps \in W^{1,1}([0,1],H^1(\Om))$ be such that
1229: $g_\eps(t) \in \afeom$ for all $t \in [0,1]$,
1230: \begin{equation}
1231: \label{lib}
1232: \|g_\eps(t)\|_\infty \le C
1233: \end{equation}
1234: for all $t \in [0,1]$, and for $\eps \to 0$
1235: \begin{equation}
1236: \label{strconv}
1237: g_\eps \to g
1238: \quad \mbox{ strongly in }W^{1,1}([0,1],H^1(\Om)).
1239: \end{equation}
1240: We indicate by $\{u^{\delta,i}_{\eps,a},\,i=0, \ldots,N_\delta\}$ the
1241: discrete evolution relative to the boundary data $g_\eps$ given by Proposition
1242: \ref{discrevol}, and we denote by $\Es^{\delta,i}_{\eps,a}$ its total energy as in
1243: Proposition \ref{dener}.
1244: \par
1245: We assume that $g(\cdot)$ and $g_\eps(\cdot)$ are defined in
1246: $H^1(\Om')$ (we still denote these extensions by $g(\cdot)$ and $g_h(\cdot)$), in
1247: such a way that \eqref{lib} and \eqref{strconv} hold in $\Om'$.
1248: Let us moreover set $g^\delta_\eps(t):=g_\eps(t^\delta_i)$
1249: for all $t^\delta_i \le t <t^\delta_{i+1}$ with $i=0,\dots,N_\delta-1$ and
1250: $g^\delta_\eps(1):=g_\eps(1)$.
1251: \par
1252: Let us make the following piecewise constant interpolation in time:
1253: \begin{equation*}
1254: %\label{piecediscrevol}
1255: u^\delta_{\eps,a}(t):=u^{\delta,i}_{\eps,a}
1256: \quad \mbox{ for } t^\delta_i \le t <t^\delta_{i+1}
1257: \quad i=0,\dots,N_\delta-1,
1258: \end{equation*}
1259: and $u^\delta_{\eps,a}(1):=u^{\delta,N_\delta}_{\eps,a}$.
1260: For all $t \in [0,1]$ we define the {\it discrete fracture} at time $t$ as
1261: \begin{equation*}
1262: %\label{pdf}
1263: \Gamma^\delta_{\eps,a}(t):= \bigcup_{s \le t}
1264: \Sg{g^\delta_\eps(s)}{u^\delta_{\eps,a}(s)},
1265: \end{equation*}
1266: and the {\it discrete total energy} at time $t$ as
1267: \begin{equation*}
1268: %\label{pde}
1269: \Es^\delta_{\eps,a}(t):=
1270: \|\nabla u^\delta_{\eps,a}(t)\|^2+
1271: \hs^1 \left( \Gamma^\delta_{\eps,a}(t) \right).
1272: \end{equation*}
1273: We have for all $t \in [0,1]$
1274: \begin{equation}
1275: \label{boundonu}
1276: \|u^{\delta}_{\eps,a}(t)\|_\infty \le \|g_\eps^\delta(t)\|_\infty.
1277: \end{equation}
1278: Moreover for all $v \in \afeaom$ we have
1279: \begin{equation}
1280: \label{MSdiscr2}
1281: \|\nabla u^{\delta}_{\eps,a}(0)\|^2
1282: +\hs^1\left( \Sg{g^\delta_\eps(0)}{u^{\delta}_{\eps,a}(0)} \right)
1283: \le \|\nabla v\|^2 +
1284: \hs^1\left( \Sg{g^\delta_\eps(0)}{v} \right),
1285: \end{equation}
1286: and for all $t \in ]0,1]$ and for all $v \in \afeaom$
1287: \begin{equation}
1288: \label{piecemin2}
1289: \|\nabla u^{\delta}_{\eps,a}(t)\|^2 \le \|\nabla v\|^2 +
1290: \hs^1 \left( \Sg{g^\delta_\eps(t)}{v} \setminus
1291: \Gamma^{\delta}_{\eps,a}(t) \right).
1292: \end{equation}
1293: Finally for all $0 \le s \le t \le 1$ we have
1294: \begin{equation}
1295: \label{discrenergybists}
1296: \Es^{\delta}_{\eps,a}(t) \le \Es^{\delta}_{\eps,a}(s)+
1297: 2\int_{s^\delta_{i}}^{t^\delta_{i}}
1298: \int_{\Om} \nabla u^{\delta}_{\eps,a}(\tau) \nabla
1299: \dot{g}_\eps(\tau) \,dx\,d\tau +o^\delta_{\eps},
1300: \end{equation}
1301: where $t^\delta_i \le t <t^\delta_{i+1}$,
1302: $s^\delta_i \le s <s^\delta_{i+1}$ and
1303: \begin{equation}
1304: \label{opiccolo}
1305: o^\delta_{\eps}:=
1306: \left[
1307: \max_{r=0,\ldots,N_\delta-1}
1308: \int_{t^\delta_{r}}^{t^\delta_{r+1}} \|\dot{g}_\eps(\tau)\|_{H^1(\Om)} \,d\tau
1309: \right] \int_0^1 \|\dot{g}_\eps(\tau)\|_{H^1(\Om)}.
1310: \end{equation}
1311: For $s=0$ we obtain the following estimate from above for the discrete total energy
1312: \begin{equation}
1313: \label{discrenergybis}
1314: \Es^{\delta}_{\eps,a}(t) \le \Es^{\delta}_{\eps,a}(0)+
1315: 2\int_{0}^{t^\delta_{i}}
1316: \int_{\Om} \nabla u^{\delta}_{\eps,a}(\tau) \nabla
1317: \dot{g}_\eps(\tau) \,dx\,d\tau +o^\delta_{\eps},
1318: \end{equation}
1319: where $t^\delta_i \le t <t^\delta_{i+1}$.
1320: \par
1321: We study the behavior of the evolution
1322: $\{t \to u^\delta_{\varepsilon,a}(t),\,t \in [0,1]\}$ varying the parameters
1323: in the following way. We let firstly $\eps \to 0$ and $\delta \to 0$
1324: obtaining an evolution $\{t \to u_a(t),\,t \in [0,1]\}$
1325: relative to the boundary data $g$ with the minimality property \eqref{amint}; then we
1326: let $a \to 0$ obtaining a quasi-static evolution of brittle fractures
1327: $\{t \to u(t)\,,t \in [0,1]\}$ relative to the boundary data $g$.
1328: Finally, by a diagonal argument we deal with $(\delta,\eps,a)$ at the
1329: same time.
1330: \par
1331: In order to develop this program, we need some compactness, and so
1332: we derive a bound for the total energy $\Es^\delta_{\eps,a}$.
1333: By \eqref{piecemin}, we have that for all $t \in [0,1]$
1334: \begin{equation*}
1335: \|\nabla u^{\delta}_{\varepsilon,a}(t)\| \le \|\nabla g^\delta_\eps(t)\| \le \tilde{C}
1336: \end{equation*}
1337: with $\tilde C$ independent of $\delta$, $\eps$ and $t$.
1338: We deduce for all $t \in [0,1]$
1339: \begin{equation*}
1340: \Es^{\delta}_{\varepsilon,a}(t) \le
1341: \Es^{\delta}_{\varepsilon,a}(0)+
1342: 2 \tilde{C}^2 +o^\delta_\eps
1343: \end{equation*}
1344: Notice that $\Es^\delta_{\eps,a}(0)$ is uniformly bounded as
1345: $\delta,\eps$ vary. Moreover, by \eqref{boundonu} and since
1346: $\|g_\eps(t)\|_\infty \le C$ for all $t \in [0,1]$, we have that
1347: $u^{\delta}_{\varepsilon,a}(t)$
1348: is uniformly bounded in $L^\infty(\Om)$ independently of $\delta,\eps$ and $a$.
1349: Taking into account \eqref{strconv}, we conclude
1350: that there exists $C'$ independent of $\delta,\varepsilon,a$ such that for all
1351: $t\in [0,1]$
1352: \begin{equation}
1353: \label{unifenergybound}
1354: \Es^{\delta}_{\varepsilon,a}(t) + \|u^{\delta}_{\varepsilon,a}(t)
1355: \|_\infty \le C'.
1356: \end{equation}
1357: Formula \eqref{unifenergybound} gives the desired compactness in order to perform
1358: the asymptotic analysis of the discrete evolution.
1359: \par
1360: Let now consider
1361: $\delta_n \to 0$ and $\eps_n \to 0$: by \eqref{strconv} we have
1362: \begin{equation}
1363: \label{Cto0}
1364: o^{\delta_n}_{\eps_n} \to 0,
1365: \end{equation}
1366: where $o^{\delta_n}_{\eps_n}$ is defined in \eqref{opiccolo}.
1367: By Helly's theorem on monotone functions, we may suppose that there exists an
1368: increasing function $\lambda_a$ such that (up
1369: to a subsequence) for all $t \in [0,1]$
1370: \begin{equation}
1371: \label{lambn}
1372: \lambda_{n,a}(t):=
1373: \hs^1 \left( \bigcup_{s \le t}\Sg{g^{\delta_n}_{\eps_n}(s)}
1374: {u^{\delta_n}_{\eps_n,a}(s)} \right)
1375: \to \lambda_a(t).
1376: \end{equation}
1377: Let us fix $D \subseteq [0,1]$ countable and dense with $0 \in D$.
1378:
1379: \begin{lemma}
1380: \label{eps-to-0}
1381: For all $t \in D$ there exists $\vuat \in SBV(\Om)$
1382: such that up to a subsequence independent of $t$
1383: \begin{equation*}
1384: u^{\delta_n}_{\eps_n,a}(t) \to \vuat \quad \mbox{ in } SBV(\Om).
1385: \end{equation*}
1386: Moreover for all $t \in D$ we have
1387: \begin{equation}
1388: \label{ueb2}
1389: \|\nabla \vuat \|^2+ \hs^1 \big( \Sg{g(t)}{\vuat} \big) +
1390: \|\vuat\|_\infty \le C'.
1391: \end{equation}
1392: \end{lemma}
1393:
1394: \begin{proof}
1395: Let us consider $t \in D$. By \eqref{unifenergybound}, we can apply
1396: Ambrosio's Compactness Theorem \ref{SBVcompact} obtaining $u \in SBV(\Om)$ such
1397: that, up to a subsequence, $u^{\delta_n}_{\eps_n,a}(t) \to u$ in $SBV(\Om)$.
1398: Let us set $\vuat:=u$. Using a diagonal argument, we deduce that there exists a
1399: subsequence of $(\delta_n, \eps_n)$ (which we still denote by
1400: $(\delta_n,\eps_n)$) such that $u^{\delta_n}_{\eps_n,a}(t) \to
1401: \vuat$ in $SBV(\Om)$ for all $t \in D$.
1402: In order to obtain inequality \eqref{ueb2}, we extend
1403: $u^{\delta_n}_{\eps_n,a}(t)$ and $\vuat$ to $\Om'$ setting
1404: $u^{\delta_n}_{\eps_n,a}(t):= g^{\delta_n}_{\eps_n}(t)$ and
1405: $u_a(t):=g(t)$ on $\Om_D$; since $g^{\delta_n}_{\eps_n}(t) \to g(t)$ on
1406: $\Om_D$ strongly in $H^1(\Om_D)$, we have that
1407: $u^{\delta_n}_{\eps_n,a}(t) \to u_a(t)$ in $SBV(\Om')$, so that
1408: we can apply Ambrosio's Theorem, and derive \eqref{ueb2} from
1409: \eqref{unifenergybound}.
1410: \end{proof}
1411:
1412: The following result is essential for the sequel: its proof is postponed to
1413: Section \ref{secmin}.
1414:
1415: \begin{proposition}
1416: \label{pminpropD}
1417: Let $t \in D$. For all $v \in SBV(\Om)$ we have
1418: \begin{equation}
1419: \label{minpropD}
1420: \|\nabla \vuat\|^2 \le \|\nabla v\|^2+ \mu(a) \hs^1(\Sg{g(t)}{v} \setminus
1421: \bigcup_{s \le t, s \in D} \Sg{g(s)}{\vuas}),
1422: \end{equation}
1423: where $\mu:]0,\frac{1}{2}[ \to ]0,+\infty[$ is such that
1424: $\lim_{a \to 0} \mu(a)=1$.
1425: Moreover, $\nabla u^{\delta_n}_{\eps_n,a}(t) \to \nabla \vuat$ strongly in
1426: $L^2(\Om;\R^2)$.
1427: \end{proposition}
1428:
1429: We now extend the evolution $\{t \to u_a(t)\,:\,t \in D\}$ to the entire interval
1430: $[0,1]$. Let us set for all $t \in [0,1]$
1431: \begin{equation*}
1432: %\label{afracture2}
1433: \Gamma_a(t):= \bigcup_{s \le t, s \in D} \Sg{g(s)}{u_a(s)}.
1434: \end{equation*}
1435:
1436: \begin{lemma}
1437: \label{aextension}
1438: For every $t \in [0,1]$ there exists $u_a(t) \in SBV(\Om)$
1439: such that the following hold:
1440: \begin{itemize}
1441: \item[(a)] for all $t \in [0,1]$
1442: \begin{equation}
1443: \label{ajumpt}
1444: \Sg{g(t)}{u_a(t)} \subseteq
1445: \Gamma_a(t) \;\mbox{ up to a set of }\hs^1\mbox{-measure }0,
1446: \end{equation}
1447: and
1448: \begin{equation}
1449: \label{aueb}
1450: \|\nabla \vuat \|^2+ \hs^1 \big( \Sg{g(t)}{\vuat} \big) +
1451: \|\vuat\|_\infty \le C';
1452: \end{equation}
1453: \item[{}]
1454: \item[(b)] for all $v \in SBV(\Om)$
1455: \begin{equation}
1456: \label{amint}
1457: \|\nabla u_a(t)\|^2 \le \|\nabla v\|^2 +
1458: \mu(a)\hs^1 \left( \Sg{g(t)}{v} \setminus \Gamma_a(t) \right);
1459: \end{equation}
1460: \item[{}]
1461: \item[(c)]
1462: $\nabla u_a$ is left continuous in $[0,1] \setminus D$ with respect to the
1463: strong topology of $L^2(\Om;\R^2)$;
1464: \item[{}]
1465: \item[(d)]
1466: for all $t \in [0,1] \setminus \ns_a$ we have that
1467: \begin{equation*}
1468: \nabla u^{\delta_n}_{\eps_n,a}(t) \to \nabla u_a(t)
1469: \quad \mbox{ strongly in }L^2(\Om, \R^2),
1470: \end{equation*}
1471: where $\ns_a$ is the set of discontinuities of the function $\lambda_a$
1472: defined in \eqref{lambn}.
1473: \end{itemize}
1474: \end{lemma}
1475:
1476: \begin{proof}
1477: Let $t \in [0,1] \setminus D$ and let $t_n \in D $ with $t_n \nearrow t$.
1478: By \eqref{ueb2}, we can apply Ambrosio's Theorem to the sequence $(u_a(t_n))$
1479: obtaining $u \in SBV(\Om)$ such that, up to a subsequence, $u_a(t_n) \to u$ in
1480: $SBV(\Om)$. Let us set $\vuat:=u$.
1481: Let us extend $u_a(t_n)$ and $\vuat$ to $\Om'$ setting $u_a(t_n):=g(t_n)$ and
1482: $\vuat:=g(t)$ on $\Om_D$: we have $u_a(t_n) \to \vuat$ in $SBV(\Om')$.
1483: Since $\hs^1 \res S_{u_a(t_n)} \le \hs^1 \res \Gamma_a(t)$ for all $n$,
1484: as a consequence of Ambrosio's Theorem, we deduce that
1485: $\hs^1 \res S_{\vuat} \le \hs^1 \res \Gamma_a(t)$. This means
1486: $\hs^1 \res \Sg{g(t)}{\vuat} \le \hs^1 \res \Gamma_a(t)$, so that
1487: \eqref{ajumpt} holds.
1488: Moreover, for all $v \in SBV(\Om)$, by \eqref{minpropD} we may write
1489: \begin{multline}
1490: \label{minbis}
1491: \|\nabla u_a(t_n)\|^2 \le
1492: \|\nabla v-\nabla g(t)+\nabla g(t_n)\|^2+
1493: \mu(a) \hs^1 \left( \Sg{g(t)}{v} \setminus \Gamma_a(t_n) \right) \le \\
1494: \le \|\nabla v-\nabla g(t)+\nabla g(t_n)\|^2+
1495: \mu(a) \hs^1 \left( \Sg{g(t)}{v} \setminus \Gamma_a(t) \right) +
1496: \mu(a) \hs^1 \left( \Gamma_a(t) \setminus \Gamma_a(t_n) \right),
1497: \end{multline}
1498: so that, since by definition of $\Gamma_a(t)$ we have
1499: $\hs^1(\Gamma_a(t) \setminus \Gamma_a(t_n)) \to 0$, we obtain that
1500: \eqref{amint} holds; choosing $v=u_a(t)$ and taking the limsup in \eqref{minbis},
1501: we obtain that
1502: $$
1503: \limsup_n \|\nabla u_{a}(t_n)\|^2 \le \|\nabla u_a(t)\|^2,
1504: $$
1505: and so the convergence $\nabla u_a(t_{n}) \to \nabla u_a(t)$ is strong in
1506: $L^2(\Om, \R^2)$.
1507: Notice that $\nabla u_a(t)$ is uniquely determined by (\ref{ajumpt}) and (\ref{amint})
1508: since the gradient of the solutions of the minimum problem
1509: \begin{equation*}
1510: \min \left\{ \|\nabla u\|^2 \,:\, \Sg{g(t)}{u} \subseteq \Gamma_a(t)
1511: \mbox{ up to a set of }\hs^1\mbox{-measure }0 \right\}
1512: \end{equation*}
1513: is unique by the strict convexity of the functional: we conclude that $\nabla u_a(t)$
1514: is well defined.
1515: The same arguments prove that $\nabla u_a$ is left continuous
1516: at all $t \in [0,1] \setminus D$.
1517: Finally \eqref{aueb} is a direct
1518: consequence of \eqref{ueb2} and of Ambrosio's Theorem, and so points $(a)$, $(b)$, $(c)$ are
1519: proved.
1520: \par
1521: Let us come to point $(d)$.
1522: Let us consider $u^{\delta_n}_{\eps_n,a}(t)$ with $t \not\in \ns_a$; we
1523: may suppose that $t \not \in D$, since otherwise the result has already been
1524: established. By Proposition \ref{pminpropD} with $D':=D \cup\{t\}$ in place of $D$,
1525: we have that, up to
1526: a subsequence, $u^{\delta_n}_{\eps_n,a}(t) \to u$ in $SBV(\Om)$
1527: such that
1528: \begin{equation*}
1529: %\label{minpropua}
1530: \|\nabla u\|^2 \le \|\nabla v\|^2 +\mu(a) \hs^1 \left( \Sg{g(t)}{v}
1531: \setminus (\Gamma_a(t) \cup \Sg{g(t)}{u}) \right)
1532: \end{equation*}
1533: for all $v \in SBV(\Om)$ and
1534: $\nabla u^{\delta_n}_{\eps_n,a}(t) \to \nabla u$ strongly in $L^2(\Om;\R^2)$.
1535: Let $s<t$ with $ s\in D$; by the minimality of $u^{\delta_n}_{\eps_n,a}(s)$
1536: and by \eqref{unifenergybound} we have
1537: \begin{multline*}
1538: \|\nabla u^{\delta_n}_{\eps_n,a}(s)\|^2 \le
1539: \|\nabla u^{\delta_n}_{\eps_n,a}(t)-\nabla g^{\delta_n}_{\eps_n}(t)+
1540: \nabla g^{\delta_n}_{\eps_n}(s)\|^2
1541: +\lambda_{n,a}(t)-\lambda_{n,a}(s) \le \\
1542: \le \|\nabla u^{\delta_n}_{\eps_n,a}(t)\|^2
1543: +2 \sqrt{C'}\|\nabla g^{\delta_n}_{\eps_n}(t)-\nabla g^{\delta_n}_{\eps_n}(s)\|+ \\
1544: +\|\nabla g^{\delta_n}_{\eps_n}(t)-\nabla g^{\delta_n}_{\eps_n}(s)\|^2+
1545: \lambda_{n,a}(t)-\lambda_{n,a}(s).
1546: \end{multline*}
1547: Passing to the limit for $n \to +\infty$, recalling that
1548: $g^{\delta_n}_{\eps_n}(\tau) \to g(\tau)$
1549: strongly in $H^1(\Om)$ for all $\tau \in [0,1]$, we deduce
1550: \begin{multline*}
1551: \|\nabla u_a(s)\|^2 \le \|\nabla u\|^2+
1552: 2\sqrt{C'}\|\nabla g(t)-\nabla g(s)\|+
1553: \|\nabla g(t)-\nabla g(s)\|^2+
1554: \lambda_a(t)-\lambda_a(s),
1555: \end{multline*}
1556: so that, since $t$ is a point of continuity for $\lambda_a$, $\nabla u_a$
1557: is left continuous at $t$, and $g$ is absolutely continuous, we get
1558: for $s \to t$
1559: \begin{equation*}
1560: \|\nabla u_a(t)\|^2 \le \|\nabla u\|^2.
1561: \end{equation*}
1562: We conclude that $u_a(t)$ is a solution of
1563: \begin{equation*}
1564: \min \{\|\nabla v\|^2\,:\,
1565: \Sg{g(t)}{v} \subseteq \Gamma_a(t) \cup \Sg{g(t)}{u}
1566: \mbox{ up to a set of }\hs^1\mbox{-measure }0\},
1567: \end{equation*}
1568: so that $\nabla u=\nabla u_a(t)$ by uniqueness of the gradient of the solution.
1569: We deduce that $\nabla u^{\delta_n}_{\eps_n,a}(t) \to \nabla u_a(t)$
1570: strongly in $L^2(\Om;\R^2)$, and so the proof is complete.
1571: \end{proof}
1572:
1573: We can now let $a \to 0$.
1574:
1575: \begin{lemma}
1576: \label{azero}
1577: There exists $a_n \to 0$ such that, for all $t \in D$,
1578: $u_{a_n}(t) \to u(t)$ in $SBV(\Om)$ for some
1579: $u(t) \in SBV(\Om)$ such that for all $v \in SBV(\Om)$
1580: we have
1581: \begin{equation}
1582: \label{alessminpropD}
1583: \|\nabla u(t)\|^2 \le \|\nabla v\|^2+ \hs^1(\Sg{g(t)}{v} \setminus
1584: \bigcup_{s \le t, s \in D} \Sg{g(s)}{u(s)}).
1585: \end{equation}
1586: Moreover, $\nabla u_{a_n}(t) \to \nabla u(t)$ strongly in $L^2(\Om;\R^2)$ and
1587: \begin{equation}
1588: \label{ueb3}
1589: \|\nabla u(t) \|^2+ \hs^1(\Sg{g(t)}{u(t)}) +
1590: \|u(t)\|_\infty \le C'.
1591: \end{equation}
1592: \end{lemma}
1593:
1594: \begin{proof}
1595: By \eqref{aueb}, applying Ambrosio's Theorem to the extensions of
1596: $\vuat$ to $\Om'$ by setting $\vuat:=g(t)$ on $\Om_D$, and using a diagonal argument,
1597: we find a sequence $a_n \to 0$ such that, for all $t \in D$,
1598: $u_{a_n}(t) \to u(t)$ in $SBV(\Om)$ for some
1599: $u(t) \in SBV(\Om)$ such that \eqref{ueb3} holds.
1600: \par
1601: We now prove that $u(t)$ satisfies property (\ref{alessminpropD}).
1602: Let $v \in SBV(\Om)$.
1603: Let us fix $t_1 \le t_2 \le \ldots \le t_k=t$ with $t_i \in D$.
1604: We extend $v$ and $u_{a_n}(t_i)$ to $\Om'$ setting $v:=g(t)$ and
1605: $u_{a_n}(t_i):=g(t_i)$ on $\Om_D$ respectively.
1606: Since $u_{a_n}(t_i) \to u(t_i)$ in $SBV(\Om')$ for all $i=1, \ldots, k$, by
1607: Theorem \ref{jumptransfer} there exists $v_n \in SBV(\Om')$ with $v_n=g(t)$ on
1608: $\Om_D$ such that $\nabla v_n \to \nabla v$ strongly in $L^2(\Om';\R^2)$ and
1609: \begin{equation}
1610: \label{convlength}
1611: \limsup_n \hs^1 \left( S_{v_n} \setminus \bigcup_{i=1}^k S_{u_{a_n}(t_i)} \right)
1612: \le \hs^1 \left( S_v \setminus \bigcup_{i=1}^k S_{u(t_i)} \right).
1613: \end{equation}
1614: By \eqref{minpropD} we obtain
1615: \begin{equation}
1616: \label{minan}
1617: \|\nabla u_{a_n}(t)\|^2 \le \|\nabla v_n\|^2+ \mu(a_n)
1618: \hs^1 \left( S_{v_n} \setminus \bigcup_{i=1}^k S_{u_{a_n}(t_i)} \right),
1619: \end{equation}
1620: so that passing to the limit for $n \to +\infty$ and recalling that
1621: $\mu(a) \to 1$ as $a \to 0$, we obtain
1622: $$
1623: \|\nabla u(t)\|^2 \le \|\nabla v\|^2+
1624: \hs^1 \left( S_v \setminus \bigcup_{i=1}^k S_{u(t_i)} \right).
1625: $$
1626: Thus we get
1627: $$
1628: \|\nabla u(t)\|^2 \le \|\nabla v\|^2+
1629: \hs^1 \left( \Sg{g(t)}{v} \setminus \bigcup_{i=1}^k \Sg{g(t_i)}{u(t_i)} \right).
1630: $$
1631: Since $t_1, \ldots, t_k$ are arbitrary, we obtain \eqref{alessminpropD}.
1632: Choosing $v=u(t)$, taking the limsup in \eqref{minan} and using
1633: \eqref{convlength}, we obtain $\nabla u_{a_n}(t) \to \nabla u(t)$ strongly in
1634: $L^2(\Om;\R^2)$.
1635: \end{proof}
1636:
1637: In order to deal with $\delta,\eps$ and $a$ at the same time,
1638: we need the following lemma.
1639:
1640: \begin{lemma}
1641: \label{convD}
1642: Let $\{u(t)\,:\,t \in D\}$ be as in Lemma \ref{azero}.
1643: There exist $\delta_n \to 0$, $\varepsilon_n \to 0$, and $a_n \to 0$
1644: such for all $t \in D$ we have
1645: \begin{equation*}
1646: u^{\delta_n}_{\varepsilon_n,a_n}(t) \to u(t)
1647: \quad \mbox{ in $SBV(\Om)$ }.
1648: \end{equation*}
1649: Moreover, for all $n$ there exists $\bs_n \subseteq [0,1]$
1650: with $|\bs_n|<2^{-n}$ such that for all $t \in [0,1] \setminus \bs_n$
1651: \begin{equation}
1652: \label{graduanconv}
1653: \|\nabla u^{\delta_n}_{\varepsilon_n,a_n}(t)-\nabla u_{a_n}(t)\|
1654: \le \frac{1}{n}.
1655: \end{equation}
1656: Finally, we have that for all $v \in SBV(\Om)$
1657: \begin{equation}
1658: \label{min0}
1659: \| \nabla u(0)\|^2 +\hs^1 \left( \Sg{g(0)}{u(0)} \right)
1660: \le \|\nabla v\|^2+\hs^1 \left( \Sg{g(0)}{v} \right)
1661: \end{equation}
1662: and
1663: \begin{equation}
1664: \label{conv0}
1665: \Es^{\delta_n}_{\eps_n,a_n}(0) \to \|\nabla u(0)\|^2+
1666: \hs^1 \left( \Sg{g(0)}{u(0)} \right).
1667: \end{equation}
1668: \end{lemma}
1669:
1670: \begin{proof}
1671: Let $(a_n)$ be the sequence determined by Lemma \ref{azero}.
1672: By Lemma \ref{eps-to-0}, for all $n$ there exists
1673: $(\delta^n_m,\varepsilon^n_{m})$
1674: such that for all $t \in D$ and $m \to +\infty$ we have
1675: \begin{equation*}
1676: u^{\delta^n_m}_{\varepsilon^n_m,a_n}(t) \to u_{a_n}(t)
1677: \quad \mbox{ in }SBV(\Om),
1678: \end{equation*}
1679: and
1680: \begin{equation*}
1681: \nabla u^{\delta^n_m}_{\varepsilon^n_m,a_n}(t)
1682: \to \nabla u_{a_n}(t)
1683: \quad \mbox{ strongly in }L^2(\Om;\R^2).
1684: \end{equation*}
1685: Moreover by Lemma \ref{aextension} we have that
1686: $\nabla u^{\delta^n_m}_{\varepsilon^n_m,a_n} \to \nabla u_{a_n}$
1687: quasi-uniformly on $[0,1]$ as $m \to +\infty$.
1688: Let $\bs_n \subseteq [0,1]$ with $|\bs_n| <2^{-n}$ such that
1689: $\nabla u^{\delta^n_m}_{\varepsilon^n_m,a_n} \to \nabla u_{a_n}$
1690: uniformly on $[0,1] \setminus \bs_n$ as $m \to +\infty$.
1691: We now perform the following diagonal argument. Let $D=\{t_n,\,n \ge 1\}$.
1692: Choose $m_1$ such that
1693: \begin{equation*}
1694: \|\nabla u^{\delta^1_{m_1}}_{\varepsilon^1_{m_1},a_1}(t_1)-
1695: \nabla u_{a_1}(t_1)\|+
1696: \|u^{\delta^1_{m_1}}_{\varepsilon^1_{m_1},a_1}(t_1)-u_{a_1}(t_1)\|
1697: \le 1,
1698: \end{equation*}
1699: and
1700: \begin{equation*}
1701: \|\nabla u^{\delta^1_{m_1}}_{\varepsilon^1_{m_1},a_1}(t)-
1702: \nabla u_{a_1}(t)\| \le 1
1703: \quad \mbox{ for all $t \in [0,1] \setminus \bs_1$ }.
1704: \end{equation*}
1705: Let $m_n$ be such that
1706: \begin{equation*}
1707: \|\nabla u^{\delta^n_{m_n}}_{\varepsilon^n_{m_n},a_n}(t_j)
1708: -\nabla u_{a_n}(t_j)\|+
1709: \|u^{\delta^n_{m_n}}_{\varepsilon^n_{m_n},a_n}(t_j)-u_{a_n}(t_j)\|
1710: \le \frac{1}{n}
1711: \;\;\mbox{ for all $j=1, \ldots, n$}
1712: \end{equation*}
1713: and
1714: \begin{equation*}
1715: \|\nabla u^{\delta^n_{m_n}}_{\varepsilon^n_{m_n},a_n}(t)-
1716: \nabla u_{a_n}(t)\| \le \frac{1}{n}
1717: \quad \mbox{ for all $t \in [0,1] \setminus \bs_n$ }.
1718: \end{equation*}
1719: We may suppose that $\delta^n_{m_n} \to 0$, $\eps^n_{m_n} \to 0$.
1720: Then $(\delta^n_{m_n},\eps^n_{m_n},a_n)$ is the sequence which satisfies
1721: the thesis. In fact by construction and taking into account \eqref{unifenergybound},
1722: for all $t \in D$ we have
1723: $u^{\delta^n_{m_n}}_{\varepsilon^n_{m_n},a_n}(t) \to u(t)$ in $SBV(\Om)$;
1724: moreover the set $\bs_n$ satisfies \eqref{graduanconv}.
1725: Notice that $u^{\delta^n_{m_n}}_{\eps^n_{m_n},a^n_{m_n}}(0)$
1726: satisfies \eqref{MSdiscr2} and so \eqref{min0} and \eqref{conv0} follow by
1727: the $\Gamma${-}convergence result of \cite{N}.
1728: \end{proof}
1729:
1730: Let $(\delta_n,\eps_n,a_n)$ be the sequence determined by Lemma
1731: \ref{convD}. For all $t \in [0,1]$ let us set
1732: \begin{equation*}
1733: %\label{ln}
1734: \lambda_n(t):=
1735: \hs^1 \left( \Gamma^{\delta_n}_{\eps_n,a_n}(t) \right).
1736: \end{equation*}
1737: By Helly's theorem, we may suppose that there exist two increasing functions
1738: $\lambda$ and $\eta$ such that up to a subsequence
1739: \begin{equation*}
1740: %\label{lambdan}
1741: \lambda_n \to \lambda \quad \mbox{ pointwise in }[0,1],
1742: \end{equation*}
1743: and
1744: \begin{equation}
1745: \label{eta}
1746: \lambda_{a_n} \to \eta \quad \mbox{ pointwise in }[0,1],
1747: \end{equation}
1748: where $\lambda_{a_n}$ is defined as in \eqref{lambn}.
1749: We now extend the evolution $\{t \to u(t)\,:\,t \in D\}$ to the entire interval
1750: $[0,1]$. Let us set for all $t \in [0,1]$
1751: \begin{equation*}
1752: %\label{fracture2}
1753: \Gamma(t):= \bigcup_{s \le t, s \in D} \Sg{g(s)}{u(s)},
1754: \end{equation*}
1755: and let $\ns$ be the set of discontinuities of $\hs^1(\Gamma(\cdot))$.
1756: Notice that for all $t \in [0,1]$
1757: \begin{equation}
1758: \label{size2}
1759: \hs^1(\Gamma(t)) \le \lambda(t).
1760: \end{equation}
1761: In fact if $t \in D$, let $t_1 \le t_2 \le \ldots \le t_k=t$ with $t_i \in D$,
1762: consider $w_n \in SBV(\Om';\R^k)$ defined as
1763: \begin{equation*}
1764: w_n(x):=(u^{\delta_n}_{\varepsilon_n,a_n}(t_1)(x), \ldots,
1765: u^{\delta_n}_{\varepsilon_n,a_n}(t_k)(x)),
1766: \end{equation*}
1767: where we assume that $u^{\delta_n}_{\varepsilon_n,a_n}(t_i)=g^{\delta_n}_{\eps_n}(t_i)$
1768: on $\Om_D$. We have $w_n \to w:=(u(t_1), \ldots, u(t_k))$ in $SBV(\Om';\R^k)$,
1769: where $u(t_i)=g(t_i)$ on $\Om_D$.
1770: Note that for all $n$ we have $S_{w_n}=
1771: \bigcup_{i=1}^k S_{u^{\delta_n}_{\varepsilon_n,a_n}(t_i)}$ so that
1772: \begin{equation*}
1773: \hs^1(S_{w_n}) \le \lambda_n(t).
1774: \end{equation*}
1775: Passing to the limit for $n \to +\infty$ and applying Ambrosio's Theorem we get
1776: \begin{equation*}
1777: \hs^1 \left( \bigcup_{i=1}^k S_{u(t_i)} \right)
1778: =\hs^1(S_w) \le \liminf_n \hs^1(S_{w_n}) \le
1779: \lambda(t);
1780: \end{equation*}
1781: we thus have
1782: \begin{equation*}
1783: \hs^1 \left( \bigcup_{i=1}^k \Sg{g(t_i)}{u(t_i)} \right)
1784: =\hs^1(S_w) \le \lambda(t)
1785: \end{equation*}
1786: and taking the sup over all $t_1, \ldots, t_k$, we obtain \eqref{size2} in $D$.
1787: The case $t \not \in D$ follows since $\hs^1(\Gamma(\cdot))$ is left continuous
1788: by definition.
1789:
1790: \begin{lemma}
1791: \label{extension}
1792: For every $t \in [0,1]$ there exists $u(t) \in SBV(\Om)$
1793: such that the following hold:
1794: \begin{itemize}
1795: \item[(a)] for all $t \in [0,1]$
1796: \begin{equation}
1797: \label{jumpt}
1798: \Sg{g(t)}{u(t)} \subseteq
1799: \Gamma(t) \; \mbox{ up to a set of }\hs^1\mbox{-measure }0,
1800: \end{equation}
1801: and for all $t \in [0,1]$ and for all $v \in SBV(\Om)$
1802: \begin{equation}
1803: \label{mint}
1804: \|\nabla u(t)\|^2 \le
1805: \|\nabla v\|^2+ \hs^1 \left( \Sg{g(t)}{v} \setminus \Gamma(t) \right);
1806: \end{equation}
1807: \item[{}]
1808: \item[(b)] $\nabla u$ is continuous in $[0,1] \setminus (D \cup \ns)$ with respect to
1809: the strong topology of $L^2(\Om;\R^2)$;
1810: \item[{}]
1811: \item[(c)] if $\tilde{\ns}$ is the set of discontinuities of the function $\eta$ defined in
1812: \eqref{eta}, for all $t \in [0,1] \setminus \tilde{\ns}$
1813: we have that
1814: \begin{equation*}
1815: \nabla u_{a_n}(t) \to \nabla u(t)
1816: \quad \mbox{ strongly in }L^2(\Om, \R^2).
1817: \end{equation*}
1818: \end{itemize}
1819: Finally
1820: \begin{equation}
1821: \label{energybelow}
1822: \Es(t) \ge \Es(0)+
1823: 2\int_0^t \int_{\Om} \nabla u(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau,
1824: \end{equation}
1825: where
1826: \begin{equation*}
1827: \label{defenergy}
1828: \Es(t):= \|\nabla u(t)\|^2+\hs^1(\Gamma(t)).
1829: \end{equation*}
1830: \end{lemma}
1831:
1832: \begin{proof}
1833: The definition of $u(t)$ is carried out as in Lemma \ref{aextension}
1834: considering $t \in [0,1] \setminus D$, $t_n \in D$ with $t_n \nearrow t$,
1835: and the limit (up to a subsequence) of $u(t_n)$ in $SBV(\Om)$:
1836: \eqref{jumpt} and \eqref{mint} hold, so that point $(a)$ is proved.
1837: It turns out that $\nabla u(t)$ is uniquely determined and that it is left continuous
1838: in $[0,1] \setminus D$.
1839: Let us consider $t \in [0,1] \setminus (D \cup \ns)$, and let $t_n \searrow t$.
1840: By Ambrosio's Theorem, we have that there exists $u \in SBV(\Om)$ with
1841: such that, up to a subsequence, $u(t_n) \to u$ in $SBV(\Om)$.
1842: Since $t$ is a continuity point of $\hs^1(\Gamma(\cdot))$, we deduce that
1843: $\Sg{g(t)}{u} \subseteq \Gamma(t)$ up to a set of $\hs^1$-measure $0$.
1844: Moreover by the minimality property for $u(t_n)$ and the fact
1845: $\Gamma(t) \subseteq \Gamma(t_n)$, we have that for all $v \in SBV(\Om)$ with
1846: \begin{multline*}
1847: \|\nabla u(t_n)\|^2 \le
1848: \|\nabla v-\nabla g(t)+\nabla g(t_n)\|^2+
1849: \hs^1 \left( \Sg{g(t)}{v} \setminus \Gamma(t_n) \right) \le \\
1850: \le \|\nabla v-\nabla g(t)+\nabla g(t_n)\|^2+
1851: \hs^1 \left( \Sg{g(t)}{v} \setminus \Gamma(t) \right),
1852: \end{multline*}
1853: and so we deduce that \eqref{mint} holds with $u$ in place of $u(t)$, and that
1854: $\nabla u(t_n) \to \nabla u$ strongly in $L^2(\Om;\R^2)$.
1855: We obtain by uniqueness that $\nabla u=\nabla u(t)$, and so $\nabla u(\cdot)$ is continuous in
1856: $[0,1] \setminus (D \cup \ns)$ and this proves point $(b)$.
1857: Point $(c)$ follows in the same way of point $(d)$ of Lemma \ref{aextension}.
1858: \par
1859: Let us come to the proof of \eqref{energybelow}. Given $t \in [0,1]$ and $k>0$,
1860: let $s_i^k:= \frac{i}{k}t$ for all $i=0, \ldots, k$. Let us set
1861: $u^k(s):=u(s_{i+1}^k)$ for $s_i^k< s \le s_{i+1}^k$. By (\ref{mint}), comparing
1862: $u(s^k_i)$ with $u(s^k_{i+1})-g(s^k_{i+1})+g(s^k_{i})$, it is easy to see that
1863: \begin{equation*}
1864: %\label{energybelowk}
1865: \Es(t) \ge \Es(0)+
1866: 2\int_0^t \int_{\Om} \nabla u^k(\tau) \nabla \dot{g}(\tau) \,d\tau \,dx+o_k,
1867: \end{equation*}
1868: where $o_k \to 0$ as $k \to +\infty$.
1869: Since $\nabla u$ is continuous with respect to the strong topology of $L^2(\Om;\R^2)$
1870: in $[0,1]$ up to a countable set, passing to the limit for $k \to +\infty$ we deduce
1871: \eqref{energybelow}.
1872: \end{proof}
1873:
1874: We are now ready to prove the main result of the paper.
1875: \begin{proof}[\rm \underline{PROOF OF THEOREM \ref{mainthm}}]
1876: Let $D$ be a countable and dense set in $[0,1]$ such that
1877: $0 \in D$, and let $(\delta_n,\eps_n,a_n)$ and
1878: $\{t \to u(t) \in SBV(\Om)\,:\, t \in [0,1]\}$
1879: be the sequence and the evolution determined in Lemma \ref{convD} and
1880: Lemma \ref{extension}.
1881: Let us set
1882: $$
1883: u_n:=u^{\delta_n}_{\varepsilon_n, a_n}, \quad \quad
1884: \Gamma_n:=\Gamma^{\delta_n}_{\varepsilon_n, a_n}, \quad \quad
1885: \Es_n:=\Es^{\delta_n}_{\varepsilon_n, a_n}.
1886: $$
1887: Let $\overline{\ns}$ be the union of the sets of discontinuities of
1888: $\eta$ and $\hs^1(\Gamma(\cdot))$,
1889: where $\eta$ is defined in \eqref{eta}.
1890: Let $\bs:= \bigcap_{k=1}^{+\infty} \bigcup_{h=k}^\infty \bs_h$, where
1891: $\bs_h$ are as in Lemma \ref{convD}; since
1892: $|\bigcup_{h=k}^\infty \bs_h| <2^{-k+1}$, we have $|\bs|=0$.
1893: For all $t \in [0,1] \setminus (\bs \cup \overline{\ns})$ we claim that
1894: \begin{equation}
1895: \label{convgrad2}
1896: \nabla u_n(t) \to \nabla u(t) \quad \mbox{ strongly in }L^2(\Om;\R^2).
1897: \end{equation}
1898: In fact,
1899: since $t \not \in \bigcup_{h=k}^\infty \bs_h$ for some $k$,
1900: by Lemma \ref{convD} we have
1901: \begin{equation*}
1902: \lim_n \|\nabla u^{\delta_n}_{\varepsilon_n,a_n}(t)-\nabla u_{a_n}(t)\|=0;
1903: \end{equation*}
1904: for $t \not \in \overline{\ns}$, by Lemma \ref{extension} we have that
1905: $\nabla u_{a_n}(t) \to \nabla u(t)$ strongly in $L^2(\Om; \R^2)$
1906: and so \eqref{convgrad2} holds.
1907: \par
1908: Since $g_{\eps_n} \to g$ strongly in $W^{1,1}([0,1];H^1(\Om))$,
1909: we deduce that for a.e. $\tau \in [0,1]$
1910: \begin{equation*}
1911: %\label{regdotgconv}
1912: \nabla \dot{g}_{\eps_n}(\tau) \to \nabla \dot{g}(\tau)
1913: \quad \mbox{ strongly in }L^2(\Om;\R^2).
1914: \end{equation*}
1915: Since $\Es_n(0) \to \Es(0)$ by \eqref{conv0} and
1916: $o^{\delta_n}_{\eps_n} \to 0$, by semicontinuity of the energy and by
1917: \eqref{discrenergybis} we have that for all
1918: $t \in D$
1919: \begin{equation}
1920: \label{energyconv*}
1921: \Es(t) \le \liminf_n \Es_n(t) \le \limsup_n \Es_n(t) \le
1922: \Es(0)+2 \int_0^t \int_{\Om}
1923: \nabla u(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau.
1924: \end{equation}
1925: In view of \eqref{energybelow}, we conclude that for all $t \in D$
1926: \begin{equation*}
1927: \Es(t)=
1928: \Es(0)+2 \int_0^t \int_{\Om}
1929: \nabla u(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau,
1930: \end{equation*}
1931: and since $\nabla u(\cdot)$ and $\hs^1(\Gamma(\cdot))$ are left continuous at $t \not \in D$
1932: and so $\Es(\cdot)$ is,
1933: we conclude that the equality holds for all $t \in [0,1]$. As a consequence
1934: $\{t \to u(t)\,, t \in [0,1]\}$ is a quasi-static evolution of brittle fractures.
1935: Let us prove that \eqref{energyconv*} is indeed true for all $t \in [0,1]$. In fact, if $t \notin D$,
1936: it is sufficient to prove
1937: \begin{equation}
1938: \label{truelsc}
1939: \liminf_n \Es_n(t) \ge \Es(t).
1940: \end{equation}
1941: Considering $s \ge t$ with $s \in D$, by \eqref{discrenergybists}
1942: we have
1943: $$
1944: \Es_n(s) \le \Es_n(t)+
1945: \int_{t^{\delta_n}_{j_n}}^{s^{\delta_n}_{j_n}} \int_{\Om} \nabla u_n(\tau)
1946: \nabla \dot{g}_{\eps_n}(\tau)\,dx\,d\tau
1947: +o^{\delta_n}_{\eps_n}
1948: \quad
1949: t^{\delta_n}_{j_n} \le t <t^{\delta_n}_{j_n+1},\;
1950: s^{\delta_n}_{j_n} \le s <s^{\delta_n}_{j_n+1},
1951: $$
1952: so that
1953: $$
1954: \liminf_n \Es_n(t) \ge \Es(s)-
1955: \int_t^s \int_{\Om} \nabla u(\tau) \nabla \dot{g}(\tau)\,dx\,d\tau.
1956: $$
1957: Letting $s \searrow t$, since $\Es(\cdot)$ is continuous, we have
1958: \eqref{truelsc} holds. By \eqref{energyconv*} we deduce that $\Es_n(t) \to \Es(t)$ for all
1959: $t \in [0,1]$, so that point $(b)$ is proved.
1960: \par
1961: We now come to point $(a)$.
1962: Since $\lambda(t) \ge \hs^1(\Gamma(t))$ for all $t \in [0,1]$,
1963: by \eqref{convgrad2} and point $(b)$, we deduce that $\lambda=\hs^1(\Gamma(\cdot))$ in $[0,1]$
1964: up to a set of measure $0$. Since they are increasing functions,
1965: we conclude that $\lambda$ and $\hs^1(\Gamma(\cdot))$
1966: share the same set of continuity points $[0,1] \setminus \ns$, and that
1967: $\lambda=\hs^1(\Gamma(\cdot))$ on $[0,1] \setminus \ns$. In view of \eqref{convgrad2},
1968: point $(a)$ is thus established for all $t$ except $t \in (\bs \cup \overline{\ns}) \setminus \ns$.
1969: In order to treat this case, we use the following argument.
1970: Considering the measures $\mu_n:=\hs^1 \res \Gamma_n(t)$, we have
1971: that, up to a subsequence, $\mu_n \weakst \mu$ weakly-star in the sense of measures,
1972: and as a consequence of Ambrosio's Theorem we have $\hs^1 \res \Gamma(t) \le \mu$ as
1973: measures. Since $t \notin \ns$ we have $\mu_n(\R^2) \to \hs^1(\Gamma(t))$, and so
1974: we deduce
1975: $\hs^1 \res \Gamma(t) =\mu$. Let us consider now $u_n(t)$; we have up to a subsequence
1976: $u_n(t) \to u$ in $SBV(\Om)$ for some $u \in SBV(\Om)$. Setting $u_n(t):=g^{\delta_n}_{\eps_n}(t)$
1977: and $u:=g(t)$ on $\Om_D$, we have $u_n(t) \to u$ in $SBV(\Om')$, and
1978: as a consequence of Ambrosio's Theorem, we get that $\hs^1 \res \Sg{g(t)}{u}
1979: \le \mu=\hs^1 \res \Gamma(t)$, that is
1980: $\Sg{g(t)}{u} \subseteq \Gamma(t)$. By Theorem \ref{jumptransfer}, we deduce
1981: that $u$ is a minimum for
1982: $$
1983: \min \{ \|\nabla v\|^2\,:\, \Sg{g(t)}{v} \subseteq \Gamma(t)
1984: \mbox{ up to a set of $\hs^1$-measure $0$ } \},
1985: $$
1986: and by uniqueness of the gradient we get that $\nabla u=\nabla u(t)$, so that
1987: the proof is concluded.
1988: \end{proof}
1989:
1990: \section{Piecewise Affine Transfer of Jump and Proof of Proposition \ref{pminpropD}}
1991: \label{secmin}
1992:
1993: The proof of Proposition \ref{pminpropD} is based on the following proposition,
1994: which is a variant of Theorem \ref{jumptransfer} in the context of piecewise affine
1995: approximation.
1996:
1997: \begin{proposition}
1998: \label{pctransfer}
1999: Given $\eps_n \to 0$, let $g^r_n\in H^1(\Om)$ be such that $g^r_n \in \afenom$ and
2000: $g^r_n \to g^r$ strongly in $H^1(\Om)$ for all $r=0,\dots,i$.
2001: If $\unr \in \afenaom$ is such that $\unr \to \ur$ in $SBV(\Om)$ for $r=0,\dots,i$,
2002: then for all $v \in SBV(\Om)$ with $\hs^1 \left( \Sg{g^i}{v} \right) <+\infty$ and $\nabla v \in
2003: L^2(\Om;\R^2)$, there exists $v_n \in \afenaom$ such that $v_n \to v$ strongly in $L^1(\Om)$,
2004: $\nabla v_n \to \nabla v$ strongly
2005: in $L^2(\Om;\R^2)$ and
2006: \begin{equation}
2007: \label{pctransfjump}
2008: \limsup_n \hs^1 \left( \Sg{g^i_n}{v_n} \setminus
2009: \bigcup_{r=0}^i \Sg{g^r_n}{u^r_n}\right) \le
2010: \mu(a) \hs^1 \left( \Sg{g(t)}{v} \setminus
2011: \bigcup_{r=0}^i \Sg{g^r}{u^r} \right),
2012: \end{equation}
2013: where $\mu\,:\,]0;\frac{1}{2}[ \to \R$ with $\lim_{a \to 0^+} \mu(a)=1$.
2014: \end{proposition}
2015:
2016: In view of Proposition \ref{pctransfer}, we can now prove Proposition \ref{pminpropD}.
2017:
2018: \begin{proof}[Proof of Proposition \ref{pminpropD}.]
2019: Notice that, in order to prove \eqref{minpropD},
2020: it is sufficient to prove the existence of $\mu\,:\,]0;\frac{1}{2}[ \to \R$ with
2021: $\lim_{a \to 0^+} \mu(a)=1$ such that,
2022: given
2023: $t \in D$, for every $0=t_0 \le \ldots \le t_r \le \ldots \le t_i=t$, $t_r \in D$,
2024: for all $v \in SBV(\Om)$ we have
2025: \begin{equation}
2026: \label{mineqk*}
2027: \|\nabla u_a(t)\|^2 \le
2028: \|\nabla v\|^2 +
2029: \mu(a) \hs^1 \left( \Sg{g(t)}{v} \setminus
2030: \bigcup_{r=0}^i \Sg{g(t_r)}{u_a(t_r)} \right).
2031: \end{equation}
2032: In fact, taking the sup over all
2033: possible $t_0, \ldots, t_i$, we get \eqref{minpropD}.
2034: \par
2035: We apply Proposition \ref{pctransfer} considering $g^r_n:=g^{\delta_n}_{\eps_n}(t_r)$, $g^r:=g(t_r)$,
2036: $u^r_n:=u^{\delta_n}_{\eps_n,a}(t_r)$, and $u^r:=u_a(t_r)$ for $r=0,\dots,i$. There exists
2037: $\mu\,:\,]0;\frac{1}{2}[ \to \R$ with $\lim_{a \to 0^+} \mu(a)=1$ such that
2038: for $v \in SBV(\Om)$, there exists $v_n \in \afenaom$ with $\nabla v_n \to \nabla v$ strongly
2039: in $L^2(\Om;\R^2)$ and
2040: \begin{multline*}
2041: \limsup_n \hs^1 \left( \Sg{g^{\delta_n}_{\eps_n}(t)}{v_n} \setminus
2042: \bigcup_{r=0}^i \Sg{g^{\delta_n}_{\eps_n}(t_r)}{u^{\delta_n}_{\eps_n,a}(t_r)} \right) \le \\
2043: \le
2044: \mu(a) \hs^1 \left( \Sg{g(t)}{v} \setminus
2045: \bigcup_{r=0}^i \Sg{g(t_r)}{u_a(t_r)} \right),
2046: \end{multline*}
2047: Comparing $u^{\delta_n}_{\eps_n,a}(t)$ and $v_n$ by means of \eqref{piecemin2}, we obtain
2048: \begin{multline}
2049: \label{mminimality}
2050: \|\nabla u^{\delta_n}_{\eps_n,a}(t) \|^2 \le \|\nabla v_n \|^2+
2051: \hs^1 \left( \Sg{g^{\delta_n}_{\eps_n}(t)}{v_n} \setminus \Gamma^{\delta_n}_{\eps_n,a}(t) \right) \le \\
2052: \le
2053: \|\nabla v_n \|^2+
2054: \hs^1 \left( \Sg{g^{\delta_n}_{\eps_n}(t)}{v_n} \setminus
2055: \bigcup_{r=0}^i \Sg{g^{\delta_n}_{\eps_n}(t_r)}{u^{\delta_n}_{\eps_n,a}(t_r)} \right),
2056: \end{multline}
2057: so that, passing to the limit for $n \to +\infty$, we obtain that \eqref{mineqk*} holds. Moreover, we have that
2058: choosing $v=u_a(t)$, and taking the limsup in \eqref{mminimality}, we get that $\nabla u^{\delta_n}_{\eps_n,a}(t)
2059: \to \nabla u_a(t)$ strongly in $L^2(\Om;\R^2)$.
2060: \end{proof}
2061:
2062: The rest of the section is devoted to the proof of Proposition \ref{pctransfer}.
2063: It will be convenient, as in Section \ref{convres}, to consider
2064: $\Om_D$ polygonal open bounded subset of $\R^2$ such that
2065: $\Om_D \cap \Om=\emptyset$ and $\partial \Om \cap \partial \Om_D= \partial_D \Om$ up to a finite number of
2066: vertices; we set $\Om':= \Om \cup \Om_D \cup \partial_D \Om$. We suppose that $\rb_\eps$ can
2067: be extended to a regular triangulation of $\Om'$ which we still indicate by
2068: $\rb_\eps$.
2069: \par
2070: We need several preliminary results.
2071: Let us set $\znr:=u^r_n-g^r_n$, and let
2072: us extend $\znr$ to zero on $\Om_D$. Similarly, we set $\zr:=u^r-g^r$, and
2073: we extend $\zr$ to zero on $\Om_D$.
2074: \par
2075: Let $\sigma>0$, and let
2076: $C$ be the set of corners of $\partial_D \Om$.
2077: Let us fix $G \subseteq \R$ countable and dense:
2078: we recall that for all $r=0,\dots,i$ we have up to a set of $\hs^{1}$-measure zero
2079: $$
2080: S_{\zr}= \bigcup_{c_1,c_2 \in G} \partial^* E_{c_1}(r) \cap \partial^*E_{c_2}(r),
2081: $$
2082: where $E_c(r):=\{x \in \Om'\,:\, \zr(x) > c\}$ and $\partial^*$ denotes the
2083: essential boundary (see \cite{AFP}). Let us consider
2084: $$
2085: J_j:=\{x \in \bigcup_{r=0}^i S_{\zr} \setminus C\,:\,
2086: \ulp (x)-\ulm (x)>\frac{1}{j} \mbox{ for some } l=0, \ldots, i\},
2087: $$
2088: with $j$ so large that $\hs^1(\bigcup_{r=0}^i S_{\zr} \setminus J_j) \le \sigma$.
2089: Let $U$ be a neighborhood of $\bigcup_{r=0}^i S_{\zr}$
2090: such that $|U| \le \frac{\sigma}{j^2}$.
2091: Following \cite[Theorem 2.1]{FL} (see Fig.3), we can find a finite
2092: disjoint collection of closed cubes $\{Q_k\}_{k=1, \ldots,K}$
2093: with center $x_k \in J_j$, edge of length $2r_k$ and
2094: oriented as the normal $\nu(x_k)$ to $S_{\umk}$ at $x_k$,
2095: such that $\bigcup_{k=1}^K Q_k \subseteq U$ and
2096: $\hs^{1}(J_j \setminus \bigcup_{k=1}^K Q_k) \le \sigma$. Moreover
2097: for all $k=1, \ldots,K$ there exists $r(k) \in \{0, \ldots, i\}$
2098: and $\cksu,\cksd>0$ such that
2099: $$
2100: \hs^1 \left( \left[ \bigcup_{r=0}^i S_{\zr} \setminus S_{\umk}
2101: \right] \cap Q_k \right) \le \sigma r_k,
2102: $$
2103: and the following hold
2104:
2105: \begin{center}
2106: \psfig{figure=fig3.eps}
2107: \end{center}
2108:
2109: \begin{itemize}
2110: \item[(a)] if $x_k \in \Om$ then $Q_k \subseteq \Om$, and if $x_k \in \partial_D \Om$ then
2111: $Q_k \cap \partial_D \Om=H_k$, where
2112: $H_k$ denotes the intersection of $Q_k$ with the straight line
2113: through $x_k$ orthogonal to $\nu(x_k)$;
2114: \item[{}]
2115: \item[(b)] $\hs^{1} (S_{\umk} \cap \partial Q_k)=0$;
2116: \item[{}]
2117: \item[(c)] $r_k \le c \hs^{1} (S_{\umk} \cap Q_k)$ for some $c>0$;
2118: \item[{}]
2119: \item[(d)] $\umkm(x) < \cksu < \cksd < \umkp(x)$ and
2120: $\cksd-\cksu \ge \frac{1}{2j}$;
2121: \item[{}]
2122: \item[(e)] $\hs^{1} ([S_{\umk} \setminus
2123: \partial^* E_{\cks}(r(k))] \cap Q_k)
2124: \le \sigma r_k$ for $s=1,2$;
2125: \item[{}]
2126: \item[(f)] if $s=1,2$, $\hs^{1} (\{y \in \partial^* E_{\cks}(r(k)) \cap Q_k\,:\,
2127: {\rm dist}(y,H_k) \ge \frac{\sigma}{2} r_k\}) <\sigma r_k$;
2128: \item[{}]
2129: \item[(g)] if $Q^+_k:= \{x\in Q_k \,|\, x \cdot \nu(x_k) >0\}$ and $s=1,2$
2130: \begin{equation}
2131: \label{density1}
2132: \| 1_{E_{\cks}(r(k)) \cap Q_k} -1_{Q^+_k}\|_{L^1(\Om')}
2133: \le \sigma^2 r_k^2;
2134: \end{equation}
2135: \item[{}]
2136: \item[(h)] $\hs^{1} ((S_v \setminus S_{\umk}) \cap Q_k) < \sigma r_k$ and
2137: $\hs^{1} (S_v \cap \partial Q_k)=0$.
2138: \end{itemize}
2139:
2140: Let us indicate by $R_k$ the intersection of $Q_k$
2141: with the strip centered in $H_k$ with width $2\sigma r_k$, and let
2142: us set $V_k^\pm:=\{x_k\pm r_k e(x_k)+s \nu(x_k):\, s \in \R\} \cap R_k$, where
2143: $e(x_k)$ is such that $\{e(x_k), \nu(x_k)\}$ is an orthonormal base of $\R^2$ with the
2144: same orientation of the canonical one.
2145: \par
2146: For all $B \subseteq \Om'$, let us set
2147: \begin{equation*}
2148: %\label{regtouch}
2149: \rs_n(B):=\{T \in \rb_{\eps_n} \,:\, T \cap B \not= \emptyset\}, \quad \quad
2150: \ts^k_n(B):=\{T \in \tb(\umkn)\,:\, T \cap B \not= \emptyset\}.
2151: \end{equation*}
2152: In the following, we will often indicate with the same symbol a family of
2153: triangles and their support in $\R^2$, being clear from the context
2154: in which sense has to be intended.
2155: We will consider $\umkn$ defined pointwise in
2156: $\Om' \setminus \overline{S}_{\umkn}$ and so the upper levels of $\umkn$ are
2157: intended as subsets of $\Om' \setminus \overline{S}_{\umkn}$.
2158:
2159: \begin{lemma}
2160: \label{upperlevels}
2161: For all $k=1, \ldots,K$ there exists
2162: $c_n^k \in [\cksu,\cksd]$
2163: such that, setting $\Enk:=\{x \in \rskn\,:\,
2164: \umkn(x) >c^k_n\}$, we have
2165: \begin{equation}
2166: \label{coarea1}
2167: \limsup_n \sum_{k=1}^K
2168: \hs^1 \left( \big( \partial_{\rskn} \Enk
2169: \big) \setminus S_{\umkn} \right)=o_\sigma,
2170: \end{equation}
2171: and
2172: \begin{equation}
2173: \label{nearset1}
2174: \limsup_n \|1_{\Enk}-1_{Q_k^+}\|_{L^1(\Om')} \le
2175: \sigma^2 r_k^2,
2176: \end{equation}
2177: where $\partial_{\rskn}$ denotes the boundary operator in
2178: $\rskn$, and $o_\sigma \to 0$ as $\sigma \to 0$.
2179: \end{lemma}
2180:
2181: \begin{proof}
2182: Note that for $n$ large we have $\bigcup_{k=1}^K \rskn \subseteq U$, so that
2183: $|\bigcup_{k=1}^K \rskn| \le \frac{\sigma}{j^2}$. By H\"older inequality and since
2184: $\|\nabla \znr \| \le C'$ for all $r=0,\dots,i$, it follows that
2185: \begin{equation*}
2186: \sum_{r=0}^i \int_{\{\cup_k \rskn: r(k)=r\}} |\nabla \znr| \,dx \le
2187: \sum_{r=0}^i \|\nabla \znr\| \frac{\sqrt{\sigma}}{j} \le (i+1)C'\frac{\sqrt{\sigma}}{j}.
2188: \end{equation*}
2189: Following \cite[Theorem 2.1]{FL}, we can apply coarea-formula for BV-functions (see \cite{AFP})
2190: taking into account that $\umkn$ belongs to $SBV(\Om')$ so that the singular
2191: part of the derivative is carried only by $S_{\umkn}$:
2192: since for $n$ large the $\rskn$'s are disjoint, we obtain
2193: \begin{equation}
2194: \label{coarea}
2195: \sum_{k=1}^K \int_\R
2196: \hs^1 \Big( \big( \partial E_{c,n}(r(k))
2197: \cap \rskn \big) \setminus S_{\umkn} \Big)\, dc \le (i+1)C' \frac{\sqrt{\sigma}}{j},
2198: \end{equation}
2199: where $E_{c,n}(r(k)):=\{x \in \Om' \setminus \overline{S}_{\umkn}\,:\,
2200: \umkn(x) >c\}$, and so
2201: \begin{equation*}
2202: \sum_{k=1}^K \int_{\cksu}^{\cksd}
2203: \hs^1 \Big( \big( \partial E_{c,n}(r(k))
2204: \cap \rskn \big) \setminus S_{\umkn} \Big) \, dc
2205: \le (i+1)C' \frac{\sqrt{\sigma}}{j}.
2206: \end{equation*}
2207: Notice that we can use the topological boundary instead of the reduced boundary of $E_{c,n}(r(k))$
2208: in \eqref{coarea} since $\umkn$ is piecewise affine, and so $\partial E_{c,n}(r(k)) \setminus
2209: \partial^* E_{c,n}(r(k)) \not= \emptyset$ just for a finite number of $c$'s.
2210: By the Mean Value Theorem we have that there exist
2211: $c_n^k \in [\cksu,\cksd]$ such that
2212: \begin{equation*}
2213: \sum_{k=1}^K
2214: \hs^1 \Big( \big( \partial E_{c_n^k,n}(r(k))
2215: \cap \rskn \big) \setminus S_{\umkn} \Big) \, \le 2iC' \sqrt{\sigma},
2216: \end{equation*}
2217: and taking the limsup for $n \to +\infty$, we get \eqref{coarea1}.
2218: Let us come to \eqref{nearset1}. Since
2219: $$E_{\cksd,n}(r(k)) \subseteq E_{c_n^k,n}(r(k)) \subseteq E_{\cksu,n}(r(k)),$$ by
2220: \eqref{density1} we have that for $n$ large
2221: $$
2222: \|1_{E_{c_n^k,n}(r(k)) \cap Q_k} -1_{Q_k^+}\|_{L^1(\Om')} \le \sigma^2 r_k^2,
2223: $$
2224: and so, since $|\rskn \setminus Q_k| \to 0$, we conclude that
2225: \eqref{nearset1} holds.
2226: \end{proof}
2227:
2228: Fix $k \in \{1, \ldots, K\}$, and let us consider the family
2229: $\ts^k_n(\Enk)$. Let us modify this family in the following way.
2230: Let $T \in \ts^k_n(\Enk)$; we keep it if
2231: $|T \cap \Enk|> \frac{1}{2}|T|$, and we erase it
2232: otherwise. Let $\Enkp$ be this new family of triangles, and let
2233: $\Enkm$ be its complement in $\ts^k_n(\rskn)$.
2234:
2235: \begin{lemma}
2236: \label{en+}
2237: For all $k=1, \ldots,K$ we have
2238: \begin{equation}
2239: \label{estper}
2240: \limsup_n \sum_{k=1}^K
2241: \hs^1 \left( \partial_{\rskn} \Enkp \setminus S_{\umkn} \right)
2242: =o_\sigma,
2243: \end{equation}
2244: and
2245: \begin{equation}
2246: \label{estarea}
2247: \limsup_n \|1_{\Enkp}-1_{Q_k^+}\|_1 \le 4\sigma^2 r_k^2,
2248: \end{equation}
2249: where $o_\sigma \to 0$ as $\sigma \to 0$.
2250: \end{lemma}
2251:
2252: \begin{proof}
2253: Let $T \in \ts^k_n(\Enk)$. Since $\umkn$ is affine on
2254: $T$, it follows that $T \cap \Enk$ is either a triangle with at least two edges
2255: contained in the edges of $T$ or a trapezoid with three edges contained
2256: in the edges of $T$. Let $l(T)$ be the edge inside $T$ where $\umkn=c_n^k$,
2257: where $c^k_n$ is the value determining $\Enk$ (we consider $l(T)=\emptyset$ if
2258: ${\rm int}(T) \subseteq \Enk$).
2259: In the case $T \in \Enkp$ as in the case $T \in \Enkm$, since the angles
2260: of the triangles of $\tb(\umkn)$ are uniformly bounded away from $0$ and from
2261: $\pi$, arguing as in Lemma \ref{simplecurve}, we deduce that keeping or erasing $T$,
2262: we increase $\partial_{\rskn} \Enk$ of a quantity which is less than
2263: $c\hs^1(l(T))$ with $c$ independent of $\varepsilon_n$. Then we have
2264: $$
2265: \sum_{k=1}^K \hs^1(\partial_{\rskn} \Enkp \setminus \partial_{\rskn} \Enk)
2266: \le \sum_{k=1}^K
2267: \sum_{T \in \ts^k_n(\Enk)} c\hs^1(l(T))
2268: \le c \sum_{k=1}^K \hs^1( \partial_{\rskn} \Enk \setminus S_{\umkn}),
2269: $$
2270: so that taking the limsup for $n \to +\infty$ and in view of \eqref{coarea1}
2271: we deduce that \eqref{estper} holds.
2272: \par
2273: Let us come to \eqref{estarea}. Note that $|\ts^k_n(\partial Q_k^+)| \to 0$
2274: as $n \to +\infty$. Then if
2275: $A^{k,+}_n:=\{T \in \tb(\umkn)\,:\, T \subseteq \tint{Q_k^+}\}$, for $n$
2276: large we have
2277: $$
2278: |Q_k^+ \setminus \Enkp| \le |A^{k,+}_n \setminus \Enkp|+
2279: |\ts^k_n(\partial Q_k^+)| \le 2|Q_k^+ \setminus \Enk|+
2280: |\ts^k_n(\partial Q_k^+)|,
2281: $$
2282: where the last inequality follows by construction of $\Enkp$.
2283: Taking the limsup for $n \to +\infty$, in view of \eqref{nearset1} we get
2284: $$
2285: \limsup_n |Q_k^+ \setminus \Enkp| \le 2\sigma^2 r_k^2.
2286: $$
2287: The inequality $\limsup_n |\Enkp \setminus Q_k^+| \le 2\sigma^2 r_k^2$ follows
2288: analogously.
2289: \end{proof}
2290:
2291: For all $k=1, \ldots, K$ and $s \in \R$, let us set
2292: $$
2293: \hkn(s) := \{x + s \nu(x_k),\, x\in H_k\}.
2294: $$
2295:
2296: \begin{lemma}
2297: \label{fubini}
2298: There exist $s^+_n \in ]\frac{\sigma}{4}r_k, \frac{\sigma}{2} r_k[$ and
2299: $s^-_n \in ]-\frac{\sigma}{2} r_k, -\frac{\sigma}{4}r_k[$ such that, setting
2300: $\hknp:=\hkn(s^+_n)$ and $\hknm:=\hkn(s^-_n)$ we have for $n$ large enough
2301: \begin{equation*}
2302: %\label{cut}
2303: \hs^1(\hknp \setminus \Enkp) \le 20\sigma r_k,
2304: \quad
2305: \hs^1(\hknm \cap \Enkp) \le 20\sigma r_k.
2306: \end{equation*}
2307: \end{lemma}
2308:
2309: \begin{proof}
2310: By (\ref{estarea}) we can write for $n$ large
2311: $$
2312: \int_{\frac{\sigma}{4}r_k}^{\frac{\sigma}{2} r_k}
2313: \hs^1(H_k(s) \setminus \Enkp) \,ds \le
2314: 5 \sigma^2 r_k^2,
2315: $$
2316: so that we get $s^+_n \in ]\frac{\sigma}{4}r_k, \frac{\sigma}{2} r_k[$
2317: with
2318: $$
2319: \hs^1(H_k(s^+_n) \setminus \Enkp) \le 20\sigma r_k.
2320: $$
2321: Similarly we can reason for $s^-_n$.
2322: \end{proof}
2323:
2324: Let $S^{k,+}_n$ be the straight line containing
2325: $\hknp$: up to replacing $\hknp$ by
2326: the connected component of $S^{k,+}_n \cap \rskn$
2327: to which it belongs, we may suppose that
2328: $\hknp \setminus \Enkp$ is a finite union of segments
2329: $l_j^+$ with extremes $A_j$ and $B_j$ belonging to
2330: the edges of the triangles of $\ts_n^k(\rskn)$ such that for $n$ large
2331: $$
2332: \hs^1(\hknp \setminus \Enkp)=\hs^1 \left( \bigcup_{j=1}^{m} l_j^+ \right) \le 20\sigma r_k.
2333: $$
2334: By Lemma \ref{simplecurve},
2335: for all $j$ there exists a curve $L_j^+$
2336: inside the edges of the triangles of $\ts_n^k(\rskn)$ joining $A_j$ and $B_j$
2337: and such that
2338: \begin{equation}
2339: \label{estsimplecurve}
2340: \hs^1(L_j^+) \le c\hs^1(l_j^+),
2341: \end{equation}
2342: with $c$ independent of $\varepsilon_n$.
2343: Let us set
2344: $$
2345: \gamma^{k,+}_n:= L_1^+ \cup B_1 A_2 \cup L_2^+ \cup \cdots \cup B_{m-1}A_m \cup
2346: L_{m}^+.
2347: $$
2348: Similarly, let us construct $\gamma^{k,-}_n$ relative to $\hknm \cap E_n^+$.
2349: Note that for $n$ large enough $\gamma^{k,+}_n \cap H_k(\sigma)=\emptyset$,
2350: $\gamma^{k,-}_n \cap H_k(-\sigma)=\emptyset$, and
2351: $\gamma^{k,+}_n \cap \gamma^{k,-}_n=\emptyset$.
2352: Let us consider the connected component
2353: $\cs_k^+$ of $\rskn \setminus \gamma^{k,+}_n$
2354: containing $H_k(\sigma)$.
2355: Similarly, let us consider the connected component $\cs_k^-$ of
2356: $\rskn \setminus \gamma^{k,-}_n$ containing $H_k(-\sigma)$.
2357: For $n$ large enough, by \eqref{estsimplecurve}
2358: \begin{equation}
2359: \label{estsimplecurve2}
2360: \hs^1 \left( \partial_{\rskn} \cs_k^+
2361: \setminus
2362: \bigcup_{i=1}^{m-1} B_i A_{i+1}\right)
2363: \le c \sum_{j=1}^{m} \hs^1(l_j^+) \le 20c\sigma r_k.
2364: \end{equation}
2365: A similar estimate holds for $\partial_{\rskn} \cs_k^-$.
2366: \par
2367: Let $\enkpt$ be the family of triangles obtained adding to $\Enkp$
2368: those $T \in \Enkm$ such that $T \subseteq \cs_k^+$, and subtracting
2369: those $T \in \Enkp$ such that $T \subseteq \cs_k^-$. Let $\enkmt$
2370: be the complement of $\enkpt$ in $\ts_n^k(\rskn)$.
2371: \par
2372: We claim that there exists
2373: $C>0$ independent of $n$ such that for all $k=1, \ldots, K$ and for $n$ large
2374: \begin{equation}
2375: \label{estper3}
2376: \hs^1 \left( \partial_{\rskn} \enkpt \setminus \partial_{\rskn} \Enkp \right) \le
2377: C\sigma r_k.
2378: \end{equation}
2379: In fact, let $\zeta$ be an edge of $\partial_{\rskn} \enkpt \setminus
2380: \partial_{\rskn} \Enkp$, that is $\zeta$ belongs to a triangle $T$ that has
2381: been changed in the operation above described.
2382: Let us assume for instance that $T \in \Enkm$ and $T \subseteq \cs_k^+$.
2383: If $T'$ is such that $T \cap T'=\zeta$, then $T' \in \Enkm$: in fact
2384: if by contradiction $T' \in \Enkp$, then $T' \in \enkpt$ and so
2385: we would have $\zeta \not\in \partial_{\rskn} \enkpt$ which is absurd.
2386: Similarly we get $T' \not \subseteq \cs_k^+$. This means that
2387: $\zeta \subseteq \partial_{\rskn} \cs_k^+$, and
2388: since the horizontal edges of $\gamma^{k,+}_n$ intersect by construction only elements of $\Enkp$,
2389: we deduce that $\zeta \subseteq \partial_{\rskn} \cs_k^+ \setminus
2390: \left( \cup_{i=1}^{m} A_iB_i \right)$,
2391: and by \eqref{estsimplecurve2} we conclude that \eqref{estper3} holds.
2392: \par
2393: We can summarize the previous results as follows.
2394:
2395: \begin{lemma}
2396: \label{families}
2397: For all $k=1, \ldots, K$ there exist two families $\enkpt$ and $\enkmt$
2398: of triangles with $\ts_n^k(\rskn)=\enkpt \cup \enkmt$,
2399: $Q_k^+ \setminus R_k \subseteq \enkpt$ and $Q_k^- \setminus R_k \subseteq \enkmt$,
2400: and such that
2401: \begin{equation}
2402: \label{estper2}
2403: \limsup_n \sum_{k=1}^K \hs^1 \left( \partial_{\rskn} \enkpt \setminus S_{\umkn} \right)=
2404: o_\sigma,
2405: \end{equation}
2406: where $o_\sigma \to 0$ as $\sigma \to 0$.
2407: Moreover, in the case $x_k \in \partial_D \Om$, we can modify
2408: $\enkpt$ or $\enkmt$ in such a way that $\enkpt \subseteq \Om$ or $\enkmt \subseteq \Om$.
2409: \end{lemma}
2410:
2411: \begin{proof}
2412: We have that \eqref{estper2} follows from \eqref{estper} and
2413: \eqref{estper3}, and the fact that $\sum_{k=1}^K r_k \le c$, with
2414: $c$ independent of $\sigma$.
2415: Let us consider the case $x_k \in \partial_D \Om$
2416: with $Q_k^+ \setminus R_k \subseteq \Om$ (the other case being similar).
2417: From \eqref{estper2} we have that for $n$ large
2418: $\sum_{k=1}^K \hs^1 \left( \partial_{\rskn} \enkpt \cap Q_k^- \right) \le o_\sigma$
2419: because $\umkn=R_{\eps_n} g_{h_n}(r(k))$ on $Q_k^-$ and so there are no jumps
2420: in $Q_k^-$. We can thus redefine $\enkpt$ subtracting those triangles that are
2421: in $Q_k^-$ obtaining again \eqref{estper2}.
2422: \end{proof}
2423:
2424: We are now in position to prove Proposition \ref{pctransfer}.
2425:
2426: \begin{proof}[Proof of Proposition \ref{pctransfer}.]
2427: We work in the context of $\Om'$. For all $v \in SBV(\Om')$ with $v=g^i$ on $\Om_D$,
2428: $\hs^1(S_v) <+\infty$ and $\nabla v \in L^2(\Om';\R^2)$,
2429: we have to construct $v_n \in SBV(\Om')$
2430: such that $v_n=g^i_n$ on $\Om_D$, $(v_n)_{|\Om} \in \afenaom$, $v_n \to v$ strongly in $L^1(\Om')$,
2431: $\nabla v_n \to \nabla v$ strongly in
2432: $L^2(\Om';\R^2)$ and
2433: \begin{equation}
2434: \label{mineqk}
2435: \limsup_n \hs^1 \left( S_{v_n} \setminus
2436: \bigcup_{r=0}^i S_{u^r_n}\right) \le
2437: \mu(a) \hs^1 \left( S_{v} \setminus
2438: \bigcup_{r=0}^i S_{u^r} \right),
2439: \end{equation}
2440: where we suppose that
2441: $u^r_n$ and $u^r$ are extended to $\Om'$ setting $u^r_n:=g^r_n$, and $u^r:=g^r$ on $\Om_D$ respectively.
2442: \par
2443: We set $v=g^i+w$, where $w \in SBV(\Om')$ with $w=0$ on $\Om_D$. By density, it is sufficient
2444: to consider the case $w \in L^\infty(\Om')$.
2445: Up to reducing $U$, we may assume that $\|\nabla g^i\|_{L^2(U;\R^2)} < \sigma$
2446: and $\|\nabla w\|_{L^2(U;\R^2)} <\sigma$. Let $R'_k$ be a rectangle centered in $x_k$,
2447: oriented as $R_k$, and such that $\overline{R'_k} \subset {\rm int}R_k$ and
2448: $\hs^1(S_w \cap (R_k \setminus R'_k)) < \sigma r_k$.
2449: We claim that there exists $\wsig \in SBV(\Om')$ with $\wsig=w$ on $\bigcup_{k=1}^K R'_k$ and
2450: $\wsig=0$ in $\Om_D$ such that
2451: \begin{itemize}
2452: \item[]
2453: \item[(1)] $\|w -\wsig\|+\|\nabla w-\nabla \wsig\| \le \sigma$;
2454: %$\|\nabla \wsig\|_{L^2(U;\R^2)} \le \sigma$,
2455: \item[]
2456: \item[(2)] $\hs^1(S_{\wsig} \cap (Q_k \setminus R'_k)) \le o_\sigma r_k$, with $o_\sigma \to 0$ as
2457: $\sigma \to 0$;
2458: \item[]
2459: \item[(3)] $\hs^1(S_{\wsig} \setminus \bigcup_{k=1}^K R_k) \le \hs^1(S_w \setminus \bigcup_{k=1}^K R_k)+\sigma$;
2460: \item[]
2461: \item[(4)] $S_{\wsig} \setminus \bigcup_{k=1}^K R_k$ is union of disjoint segments
2462: with closure contained in $\Om \setminus \bigcup_{k=1}^K R_k$;
2463: \item[]
2464: \item[(5)] $\wsig$ is of class $W^{2,\infty}$ on $\Om \setminus \left( \bigcup_{k=1}^K R_k \cup
2465: \overline{S_{\wsig}} \right)$.
2466: \end{itemize}
2467: In fact, by Proposition \ref{regularization}, there exists
2468: $w_m \in SBV(\Om')$ with $w_m=0$ in $\Om' \setminus \Omb$
2469: such that $w_m \to w$
2470: strongly in $L^2(\Om')$, $\nabla w_m \to \nabla w$ strongly in
2471: $L^2(\Om';\R^2)$, $S_{w_m}$ is polyhedral with
2472: $\overline{S_{w_m}} \subseteq \Om$,
2473: $w_m$ is of class $W^{2,\infty}$ on $\Om \setminus \left( \bigcup_{k=1}^K R_k \cup
2474: \overline{S_{w_m}} \right)$, and
2475: $\lim_m \hs^{1}(A \cap S_{w_m})=\hs^{1}(A \cap S_w)$
2476: for all $A$ open subset of $\Om'$ with $\hs^{1}(\partial A \cap S_{w})=0$.
2477: It is not restrictive to assume that $\hs^1(S_{w} \cap \partial R_k)=0$ and
2478: $\hs^1(S_{w_m} \cap \partial R_k)=0$ for all $m$.
2479: Let $\psi_k$ be a smooth function such that $0 \le \psi_k \le 1$, $\psi_k=1$ on
2480: $R'_k$ and $\psi_k=0$ outside $R_k$. Setting $\psi:=\sum_{k=1}^K \psi_k$, let
2481: us consider $\tilde{w}_m:= \psi w+ (1-\psi) w_m$. Note that
2482: $\tilde{w}_m \to w$ strongly in $L^2(\Om')$, $\nabla \tilde{w}_m \to \nabla w$ strongly in
2483: $L^2(\Om';\R^2)$, $\tilde{w}_m=0$ in $\Om_D$. Moreover,
2484: by capacity arguments, we may assume that $S_{\tilde{w}_m} \setminus \bigcup_{k=1}^K R_k$
2485: is a finite union of disjoint segments with closure contained in
2486: $\Om \setminus \bigcup_{k=1}^K R_k$.
2487: Finally, for $m \to +\infty$, we have
2488: $$
2489: \hs^1(S_{\tilde{w}_m} \setminus \bigcup_{k=1}^K R_k) \to
2490: \hs^1(S_{w} \setminus \bigcup_{k=1}^K R_k)),
2491: $$
2492: $$
2493: \hs^1(S_{\tilde{w}_m} \cap \bigcup_{k=1}^K (Q_k \setminus R_k)) \to
2494: \hs^1(S_{w} \cap \bigcup_{k=1}^K (Q_k \setminus R_k))
2495: $$
2496: and
2497: $
2498: \limsup_m \hs^1(S_{\tilde{w}_m} \cap (R_k \setminus R_k')) \le
2499: 2\hs^1(S_w \cap (R_k \setminus R_k')) \le 2\sigma r_k.
2500: $
2501: Then we can take $\wsig:=\tilde{w}_m$ for $m$ large enough.
2502: \par
2503: Let $S_{\wsig} \setminus \bigcup_{k=1}^K Q_k:= \bigcup_{j=1}^m l_j$, where, by capacity arguments,
2504: we can always assume that $l_j$ are
2505: disjoint segments with closure contained in $\Om \setminus \bigcup_{k=1}^K Q_k$. We define a triangulation
2506: $\tb_n \in \ts_{\varepsilon_n,a}(\Om')$ specifying its adaptive vertices
2507: as follows.
2508: Let us consider the families $\rskn$ and $\rs_n(l_j)$ for $k=1, \ldots, K$
2509: and $j=1, \ldots, m$. Note that for $n$ large enough,
2510: $\rs_n(Q_{k_1}) \cap \rs_n(Q_{k_2})= \emptyset$ for
2511: $k_1 \not= k_2$, $\rs_n(l_{j_1}) \cap \rs_n(l_{j_2})= \emptyset$ for $j_1 \not= j_2$,
2512: and $\rs_n(Q_k) \cap \rs_n(l_j)= \emptyset$ for every $k,j$.
2513: We consider inside the triangles of $\rs_n(Q_k)$ the adaptive vertices
2514: of $\tb(\umkn)$.
2515: Passing to $\rs_n(l_j)$, by density arguments it is not restrictive to assume that $l_j$
2516: does not pass through the vertices of $\rb_{\eps_n}$ and that its extremes belong to the
2517: edges of $\rb_{\eps_n}$.
2518: Let $\zeta:=[x,y]$ be an edge of $\rs_n(l_j)$ such that $l_j \cap \zeta =\{P\}$.
2519: Proceeding as in \cite{N}, we take as adaptive vertex of $\zeta$
2520: the projection of $P$ on $\{tx+(1-t)y:\, t \in [a,(1-a)]\}$.
2521: Connecting these adaptive vertices, we obtain
2522: an {\it interpolating} polyhedral curve $\tilde{l}_j$ with
2523: \begin{equation}
2524: \label{estinterp}
2525: \hs^1(\tilde{l}_j) \le \mu(a) \hs^1(l_j),
2526: \end{equation}
2527: where $\mu$ is an increasing function such that $\lim_{a \to 0} \mu(a)=1$.
2528: Finally, in the remaining edges, we can consider any admissible adaptive vertex,
2529: for example the middle point.
2530: \par
2531: Let us define $w_n \in SBV(\Om')$ in the following way. For all
2532: $Q_k$, let $w_n$ be equal to $\wsig$ on $\rs_n(Q_k) \setminus R_k$,
2533: equal to the reflection of ${\wsig}_{|Q_k^+ \setminus R_k}$ with respect to $H_k(\sigma)$ on $\enkpt \cap R_k$ and
2534: equal to the reflection of ${\wsig}_{|Q_k^- \setminus R_k}$ with respect to $H_k(-\sigma)$ on $\enkmt \cap R_k$, where
2535: $\tilde{E}^{k,\pm}_n$ are defined as in Lemma \ref{families}.
2536: On the other elements of $\tb_n$, let us set $w_n=\wsig$.
2537: Notice that $w_n=0$ on $\Om_D$ and that inside each $\rskn$, all the discontinuities of
2538: $w_n$ are contained in $\partial_{\rskn} \enkpt \cup V_k \cup P_{\wsig}^k$, where $P_{\wsig}^k$ is
2539: the union of the polyhedral jumps of $\wsig$ in $\rs_n(Q_k)$ and of their reflected version
2540: with respect to $H_k(\pm \sigma)$.
2541: By Lemma \ref{families} and since $\sum_{k=1}^K \hs^1(V_k \cup P_{\wsig}^k) \le o_\sigma$ with
2542: $o_\sigma \to 0$ as $\sigma \to 0$,
2543: and $\hs^1 \left( \bigcup_{r=0}^i S_{\zr} \setminus \bigcup_{k=1}^K Q_k \right)
2544: \le 2\sigma$, we have that
2545: \begin{multline*}
2546: %\label{estimatejumps}
2547: \limsup_n \hs^1 \left( S_{w_n} \setminus \bigcup_{r=0}^i S_{\znr} \right) \le \\
2548: \le \hs^1 \left( S_{\wsig} \setminus \bigcup_{k=1}^K Q_k \right) +
2549: \limsup_n \hs^1 \left( (S_{w_n} \setminus \bigcup_{r=0}^i S_{\znr}) \cap \rskn \right) \le \\
2550: \le \hs^1 \left( S_{\wsig} \setminus \bigcup_{r=0}^i S_{\zr} \right)+
2551: \hs^1 \left( \bigcup_{r=0}^i S_{\zr} \setminus \bigcup_{k=1}^K Q_k \right)+
2552: o_\sigma \le \\
2553: \le \hs^1 \left( S_{\wsig} \setminus \bigcup_{r=0}^i S_{\zr} \right)+o_\sigma,
2554: \end{multline*}
2555: and since $\|\nabla \wsig\|_{L^2(U;\R^2)} \le o_\sigma$ we get for $n$ large
2556: \begin{equation}
2557: \label{estimategrad}
2558: \|\nabla w_n\|^2_{L^2(\bigcup_{k=1}^K \rskn)} \le o_\sigma.
2559: \end{equation}
2560: We now want to define an interpolation $\twn$ of $w_n$ on $\tb_n$.
2561: Firstly, we set $\twn=0$ on all regular triangles of $\Om_D$.
2562: Passing to the triangles in $\rskn$ (see fig.4), by Lemma \ref{guscio},
2563: we know that for $n$ large enough, we have
2564: $$
2565: \hs^1(\partial \rs_n(V_k)) \le c \hs^1(V_k), \quad
2566: \hs^1(\partial \rs_n(P_{\wsig}^k)) \le c \hs^1(P_{\wsig}^k),
2567: $$
2568: with $c$ independent of $n$.
2569: If $T \in \rs_n(V_k) \cup \rs_n(P_{\wsig}^k)$, we set $\twn=0$ on $T$;
2570: otherwise, we define $\twn$ on $T$ as the affine interpolation of $w_n$.
2571: \par
2572: Since $\nabla \twn$ is uniformly bounded on $\rs_n(H_k(\pm \sigma))$,
2573: $|\rs_n(H_k(\pm \sigma))| \to 0$ and since $w_n$ is uniformly bounded in $W^{2,\infty}$ on
2574: the triangles contained in $\rs_n(Q_k) \setminus \rs_n(V_k \cup P_{\wsig}^k \cup H_k(\pm \sigma))$
2575: we have by the interpolation estimate \eqref{interp1} and by \eqref{estimategrad}
2576: \begin{equation}
2577: \label{estgradqint}
2578: \limsup_n \|\nabla \twn\|^2_{L^2(\bigcup_{k=1}^K \rskn)} \le o_\sigma.
2579: \end{equation}
2580: Moreover we have
2581: \begin{equation}
2582: \label{estjumpqint}
2583: \limsup_n \sum_{k=1}^K
2584: \hs^1 \left( \Big( S_{\twn} \setminus \bigcup_{r=0}^k S_{\znr} \Big) \cap \rskn \right) \le o_\sigma.
2585: \end{equation}
2586: Let us come to the triangles not belonging to $\rskn$ for $k=1, \ldots, K$.
2587: For all $j=1, \ldots, m$, we denote by $\hat{\rs}_n(l_j)$ the family of regular
2588: triangles that have edges in common with triangles of $\rs_n(l_j)$.
2589: For $n$ large we have that $\hat{\rs}_n(l_{j_1}) \cap \hat{\rs}_n(l_{j_2})= \emptyset$ for
2590: $j_1 \not= j_2$. On every regular triangle $T \not \in \bigcup_{k=1}^K \rskn \cup
2591: \bigcup_{j=1}^m \hat{\rs}_n(l_j)$,
2592: we define $\twn$ as the affine interpolation of $\wsig$.
2593: Since $\wsig$ is of class $W^{2,\infty}$ on $T$ and $T$ is regular,
2594: we obtain by the interpolation estimate \eqref{interp1}
2595: \begin{equation}
2596: \label{estgradinside}
2597: \|\twn - \wsig\|^2_{W^{1,2}(T)} \le \cint \varepsilon_n \|\wsig\|_{W^{2,\infty}}.
2598: \end{equation}
2599: Let us consider now those triangles that are contained in the elements
2600: of $\bigcup_{j=1}^m \hat{\rs}_n(l_j)$. Following \cite{N}, we can define $\twn$
2601: on every $T$ in such a way that $\twn$ admits discontinuities
2602: only on $\tilde{l}_j$, and $\|\nabla \twn\|_{L^\infty(T)} \le \|\nabla \wsig\|_\infty$.
2603: Since $|\hat{\rs}_n(l_j)| \to 0$ as $n \to \infty$, we deduce that
2604: \begin{equation}
2605: \label{estgradjumps}
2606: \lim_n \|\nabla \twn\|^2_{L^2(\hat{\rs}_n(l_j))}=0.
2607: \end{equation}
2608: Moreover by \eqref{estinterp} and since
2609: $\hs^1 \left( \bigcup_{r=0}^i S_{\zr} \setminus \bigcup_{k=1}^K Q_k \right) \le 2\sigma$,
2610: we have
2611: \begin{equation}
2612: \label{estjumpinside}
2613: \hs^1 \left( S_{\twn} \cap \bigcup_{j=1}^m \hat{\rs}_n(l_j) \right) \le \mu(a)
2614: \hs^1 \left( S_{\wsig} \setminus \bigcup_{i=1}^k S_{\zr} \right) +o_\sigma,
2615: \end{equation}
2616: where $o_\sigma \to 0$ as $\sigma \to 0$.
2617:
2618: \begin{center}
2619: \psfig{figure=fig4.eps}
2620: \end{center}
2621:
2622: We are now ready to conclude. Let us consider $\hwn \in \afenaom$ defined
2623: as $\hwn:=g^i_n+\twn$. We have $\hwn \to g^i+\wsig$ strongly in $L^2(\Om')$.
2624: By \eqref{estgradqint}, \eqref{estgradinside}, \eqref{estgradjumps} we get
2625: $$
2626: \limsup_n \|\nabla \hwn\|^2 \le \|\nabla g^i+\nabla \wsig\|^2 +o_\sigma,
2627: $$
2628: while by \eqref{estjumpqint} and \eqref{estjumpinside} we have
2629: $$
2630: \limsup_n \hs^1 \left( S_{\hwn} \setminus \bigcup_{r=0}^i S_{\znr} \right) \le \mu(a)
2631: \hs^1 \left( S_{\wsig} \setminus \bigcup_{r=0}^i S_{\zr} \right)+o_\sigma.
2632: $$
2633: Letting now $\sigma \to 0$, using a diagonal argument, we conclude that Proposition \ref{pctransfer} holds.
2634: \end{proof}
2635:
2636: \section{Revisiting the approximation by Francfort and Larsen}
2637: \label{remark}
2638:
2639: In this section we show how the arguments of Section \ref{convres} may be used
2640: to deal with the discrete in time approximation of quasi-static growth of
2641: brittle fractures proposed by Francfort and Larsen in \cite{FL}.
2642: More precisely, we prove that there is strong convergence of the gradient of the
2643: displacement
2644: (in particular convergence of the bulk energy) and convergence of
2645: the surface energy at all times of continuity of the length of the crack;
2646: moreover there is convergence of the total energy at any time.
2647: \par
2648: We briefly recall the notation employed in \cite{FL}. Let $I_\infty$ be countable and
2649: dense in $[0,1]$, and let $I_n:=\{0=t^n_0 \le \ldots \le t^n_n=1\}$ such that
2650: $(I_n)$ is an increasing sequence of sets whose union is $I_\infty$. Let $\Om \subseteq
2651: \R^N$ be a Lipschitz bounded domain, and let $\partial \Om=\partial \Om^c_f \cup \partial \Om_f$,
2652: where $\partial \Om^c_f$ is open in the relative topology. Let $\Om' \subseteq \R^N$ be open and
2653: such that $\overline{\Om} \subseteq \Om'$, and let $g \in W^{1,1}([0,1];H^1(\Om'))$. At any
2654: time $t^n_k$, Francfort and Larsen consider $u^n_k$ minimizer of
2655: $$
2656: \int_{\Om} |\nabla v|^2 \,dx+ \hs^{N-1} \left( S_v \setminus \left[
2657: \bigcup_{0 \le j \le k-1} S_{u^n_j} \cup \partial \Om_f \right] \right)
2658: $$
2659: in $\{v \in SBV(\Om'):v=g(t^n_k)\mbox{ in }\Om'\setminus \overline{\Om}\}$.
2660: Setting $u^n(t):=u^n_k$ for $t \in [t^n_k,t^n_{k+1}[$, and $\Gamma^n(t):=
2661: \bigcup_{s \le t, s \in I_n} S_{u^n(s)} \cup \partial \Om_f$, they prove that
2662: \begin{equation}
2663: \label{estabovec}
2664: \Es^n(t) \le \Es^n(0)+2\int_0^{t^n_k} \int_{\Om} \nabla u^n(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau
2665: +o_n, \quad\quad t\in [t^n_k,t^n_{k+1}[,
2666: \end{equation}
2667: where $\Es^n(t):=\int_{\Om} |\nabla u^n(t)|^2 \,dx+ \hs^{N-1}\left( \Gamma^n(t) \right)$ and
2668: $o_n \to 0$ as $n \to +\infty$.
2669: Using Theorem \ref{jumptransfer}, they obtain a subsequence of $(u^n(\cdot))$, still denoted by the same symbol,
2670: such that $u^n(t) \to u(t)$ in $SBV(\Om')$ and $\nabla u^n(t) \to \nabla u(t)$ strongly in $L^2(\Om';\R^N)$
2671: for all $t \in I_\infty$, with $u(t)$ a minimizer of
2672: $$
2673: \int_{\Om} |\nabla v|^2 \,dx+ \hs^{N-1} \left( S_v \setminus \Gamma(t) \right),
2674: $$
2675: where $\Gamma(t):= \bigcup_{s \in I_{\infty}, s \le t} S_{u(s)} \cup \partial \Om_f$.
2676: The evolution $\{t \to u(t),\,t \in I_\infty\}$ is extended to the whole $[0,1]$ using the
2677: approximation from the left in time.
2678: \par
2679: We can now use the arguments of Section \ref{convres}. Following Lemma \ref{extension},
2680: it turns out that for all $t \in [0,1]$
2681: \begin{equation}
2682: \label{estbelowc}
2683: \Es(t) \ge \Es(0)+2\int_0^t \int_\Om \nabla u(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau.
2684: \end{equation}
2685: Moreover, by the Transfer of Jump and the uniqueness argument of Lemma \ref{aextension},
2686: we have that $\nabla u^n(t) \to \nabla u(t)$ strongly
2687: in $L^2(\Om';\R^N)$ for all $t \not\in \ns$, where $\ns$ is the (at most countable)
2688: set of discontinuities
2689: of the pointwise limit $\lambda$ of $\hs^{N-1}(\Gamma(\cdot))$ (which exists up to a further subsequence
2690: by Helly's Theorem). Then we pass to the limit in \eqref{estabovec} obtaining
2691: $$
2692: \Es(t) \le \Es(0)+2\int_0^t \int_{\Om} \nabla u(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau;
2693: $$
2694: moreover, following the proof of Theorem \ref{mainthm}, we have that
2695: for all $t \in [0,1]$
2696: $$
2697: \Es(t) \le \liminf_n \Es_n(t) \le \limsup_n \Es_n(t)=\Es(0)
2698: +2\int_0^t \int_{\Om} \nabla u(\tau) \nabla \dot{g}(\tau) \,dx\,d\tau,
2699: $$
2700: and taking into account \eqref{estbelowc} we get the convergence of the total
2701: energy at any time. Since $\nabla u^n(t) \to \nabla u(t)$ strongly in
2702: $L^2(\Om';\R^N)$ for every $t \in I_\infty$, we deduce that $\lambda=\hs^{N-1}(\Gamma(\cdot))$
2703: on $I_\infty$, so that the convergence of the surface energy holds in $I_\infty$.
2704: The extension to the continuity times for $\hs^{N-1}(\Gamma(\cdot))$ follows like in the final
2705: part of the proof of Theorem \ref{mainthm}.
2706:
2707:
2708:
2709: \bigskip
2710: \bigskip
2711: \centerline{ACKNOWLEDGMENTS}
2712: \bigskip\noindent
2713: The authors wish to thank Gianni Dal Maso for having proposed them the problem,
2714: and Gianni Dal Maso and Gilles A. Francfort for many helpful and interesting discussions.
2715:
2716: \begin{thebibliography}{99}
2717:
2718: \bibitem{A1}{Ambrosio L.}: A compactness theorem for a new class of functions
2719: of bounded variations. {\it Boll. Un. Mat. Ital.}\,{\bf 3-B} (1989), 857-881.
2720:
2721: \bibitem{A2}{Ambrosio L.}: Existence theory for a new class of variational problems.
2722: {\it Arch. Ration. Mech. Anal.}\,{\bf 111} (1990), 291-322.
2723:
2724: \bibitem{A3}{Ambrosio L.}: A new proof of the SBV compactness theorem.
2725: {\it Calc. Var.}\,{\bf 3} (1995), 127-137.
2726:
2727: \bibitem{AFP}{Ambrosio L., Fusco N., Pallara D.}: {\it Functions of bounded
2728: variations and Free Discontinuity Problems}. Clarendon Press, Oxford, 2000.
2729:
2730: \bibitem{BFM}{Bourdin B., Francfort G.A., Marigo J.J.}: Numerical experiments in
2731: revisited brittle fracture. {\it J. Mech. Phys. Solids},\,{\bf 48-4} (2000), 797-826.
2732:
2733: \bibitem{Br1}{Brezis, H.}: {\it Op\'erateurs maximaux monotones et semi-groupes de
2734: contractions dans les espaces de Hilbert}. North-Holland, Amsterdam, 1973.
2735:
2736: \bibitem{Cia}{Ciarlet P.G.}: {\it The Finite Element Method for Elliptic Problems},
2737: North Holland, Amsterdam (1978).
2738:
2739: \bibitem{C}{Cortesani G.}: Strong approximation of $GSBV$ functions by piecewise smooth
2740: functions. {\it Ann. Univ. Ferrara - Sez VII - Sc. Mat.}\,{\bf 43} (1997), 27-49.
2741:
2742: \bibitem{DMT}{Dal Maso G., Toader R.}: A model for the quasi-static growth of
2743: brittle fractures: existence and approximation results.
2744: {\it Arch. Ration. Mech. Anal.}\,{\bf 162} (2002), 101-135.
2745:
2746: \bibitem{FL}{Francfort G.A., Larsen C.J.}: Existence and convergence for quasi-static
2747: evolution in brittle fracture. {\it To appear.}
2748:
2749: \bibitem{FM}{Francfort G.A., Marigo J.-J.}: Revisiting brittle fractures as an energy
2750: minimization problem.
2751: {\it J. Mech. Phys. Solids}\,{\bf 46} (1998), 1319-1342.
2752:
2753: \bibitem{N}{Negri M.}: Numerical methods for free-discontinuity problems based on approximations
2754: by $\Gamma${-}convergence. PhD Thesis, SISSA/ISAS, (2001).
2755:
2756: \bibitem{Ro}{Rogers C.A.}:{\it Hausdorff Measures}. Cambridge University Press,
2757: Cambridge, 1970.
2758:
2759: \end{thebibliography}
2760:
2761: \end{document}
2762: %e vai!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
2763: