ebook img

Quasilinear Lane-Emden equations with absorption and measure data PDF

0.31 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Quasilinear Lane-Emden equations with absorption and measure data

Quasilinear Lane-Emden equations with absorption and measure data 3 1 0 2 Marie-Fran¸coise Bidaut-V´eron∗ n Nguyen Quoc Hung† a J Laurent V´eron‡ 5 1 Laboratoire de Math´ematiques et Physique Th´eorique, Universit´e Franc¸ois Rabelais, Tours, FRANCE ] P A . h t Contents a m [ 1 Introduction 2 2 2 Lorentz spaces and capacities 4 v 4 2.1 Lorentz spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1 2.2 Wolff potentials, fractional and η-fractional maximal operators . . . 5 3 6 2.3 Estimates on potentials . . . . . . . . . . . . . . . . . . . . . . . . . 5 . 2.4 Approximation of measures . . . . . . . . . . . . . . . . . . . . . . . 16 2 1 2 3 Renormalized solutions 19 1 3.1 Classical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 : v 3.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 i X 4 Equations with absorption terms 23 r a 4.1 The general case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 4.2 Proofs of Theorem 1.1 and Theorem 1.2 . . . . . . . . . . . . . . . . 26 2010 Mathematics Subject Classification. 35J92,35R06,46E30. Key words: quasilinear elliptic equations, Wolff potential, maximal functions, Borel measures, Lorentz spaces,Lorentz-Besselcapacities. ∗E-mail address: [email protected] †E-mail address: [email protected] ‡E-mail address: [email protected] 1 Abstract We study the existence of solutions to the equation −∆ u+g(x,u) = µ when p g(x,.) is a nondecreasing function and µ a measure. We characterize the good measures, i.e. the ones for which the problem has a renormalized solution. We study particularly the caseswhereg(x,u)=|x|−β|u|q−1uandg(x,u)=sgn(u)(eτ|u|λ−1). Theresultsstatethata measure is good if it is absolutely continuous with respect to an appropriate Lorentz-Bessel capacities. 1 Introduction Let Ω ⊂ RN be a bounded domain containing 0 and g : Ω×R → R be a Caratheodory function. We assume that for almost all x ∈ Ω, r 7→ g(x,r) is nondecreasing and odd. In this article we consider the following problem −∆ u+g(x,u)=µ in Ω p (1.1) u=0 in ∂Ω where∆ u=div |∇u|p−2∇u ,(1<p<N),isthe p-Laplacianandµaboundedmeasure. p A measureforwh(cid:16)ichthe probl(cid:17)emadmits asolution,in anappropriateclass,is calleda good measure. When p = 2 and g(x,u) = g(u) the problem has been considered by Benilan and Brezis [4] in the subcritical case that is when any bounded measure is good. They prove that such is the case if N ≥3 and g satisfies ∞ g(s)s−NN−−12ds<∞. (1.2) Z1 Thesupercriticalcase,alwayswithp=2,hasbeenconsideredbyBarasandPierre[3]when g(u) = |u|q−1u and q > 1. They prove that the corresponding problem to (1.1 ) admits a solution (always unique in that case) if and only if the measure µ is absolutely continuous with respect to the Bessel capacity C (q′ = q/(q−1)). In the case p 6= 2 it is shown 2,q′ by Bidaut-V´eron [6] that if problem (1.1 ) with β = 0 and g(s) = |s|q−1s (q > p−1 > 0) admits a solution, then µ is absolutely continuous with respect to any capacity Cp, q +ǫ q+1−p for any ǫ>0. InthisarticleweintroduceanewclassofBesselcapacitieswhicharemodelledonLorentz spaces Ls,q instead of Lq spaces. If G is the Bessel kernel of order α > 0, we denote by α Lα,s,q(RN)theBesovspacewhichisthespaceoffunctionsφ=G ∗f forsomef ∈Ls,q(RN) α and we set kφk = kfk (a norm which is defined by using rearrangements). Then we α,s,q s,q set C (E)=inf{kfk : f ≥0, G ∗f ≥1 on E} (1.3) α,s,q s,q α for any Borel set E. We say that a measure µ in Ω is absolutely continuous with respect to the capacity C if , α,s,q ∀E ⊂Ω, E Borel , C (E)=0=⇒|µ|(E)=0. (1.4) α,s,q We also introduce the Wolff potential of a positive measure µ∈M (RN) by + W [µ](x)= ∞ µ(Bt(x)) s−11 dt (1.5) α,s tN−αs t Z0 (cid:18) (cid:19) 2 if α > 0, 1 < s < α−1N. When we are dealing with bounded domains Ω ⊂ B and R µ∈M (Ω), it is useful to introduce truncated Wolff potentials. + WR [µ](x)= R µ(Bt(x)) s−11 dt (1.6) α,s tN−αs t Z0 (cid:18) (cid:19) We prove the following existence results concerning −∆ u+|x|−βg(u)=µ in Ω p (1.7) u=0 in ∂Ω Theorem 1.1 Assume 1 < p < N, q > p−1 and 0 ≤ β < N and µ is a bounded Radon measure in Ω. 1- If g(s)=|s|q−1s, then (1.7 ) admits a renormalized solution if µ is absolutely continuous with respect to the capacity C . p, Nq , q Nq−(p−1)(N−β)) q+1−p 2- If g satisfies ∞ g(s)s−q−1ds<∞ (1.8) Z1 then (1.7 ) admits a renormalized solution if µ is absolutely continuous with respect to the capacity C . p, Nq ,1 Nq−(p−1)(N−β)) Furthermore, in both case there holds −cW2diam(Ω)[µ−](x)≤u(x)≤cW2diam(Ω)[µ+](x) for almost all x∈Ω. (1.9) 1,p 1,p where c is a positive constant depending on p and N. Inordertodealwithexponentialnonlinearitiesweintroducefor0<α<N thefractional maximal operator (resp. the truncated fractional maximal operator), defined for a positive measure µ by µ(B (x)) µ(B (x)) M [µ](x)=sup t , resp M [µ](x)= sup t , (1.10) α tN−α α,R tN−α t>0 (cid:18) 0<t<R (cid:19) and the η-fractional maximal operator (resp. the truncated η-fractional maximal operator) µ(B (x)) µ(B (x)) Mη[µ](x)=sup t , resp Mη [µ](x)= sup t , (1.11) α tN−αh (t) α,R tN−αh (t) t>0 η (cid:18) 0<t<R η (cid:19) where η ≥0 and (−lnt)−η if 0<t< 1 h (t)= 2 (1.12) η ( (ln2)−η if t≥ 21 Theorem 1.2 Assume 1 < p < N, τ > 0 and λ ≥ 1. Then there exists M > 0 depending on N,p,τ and λ such that if a measure in Ω, µ=µ+−µ− can be decomposed as follows µ+ =f +ν and µ− =f +ν , (1.13) 1 1 2 2 3 where f ∈L1(Ω) and ν ∈Mb (Ω) (j =1,2), and if j + j + (p−1)(λ−1) Mp,2diaλm(Ω)[νj] <M, (1.14) (cid:13) (cid:13)L∞(Ω) (cid:13) (cid:13) (cid:13) (cid:13) there exists a renormalized solu(cid:13)tion to (cid:13) −∆ u+sign(u) eτ|u|λ −1 =µ in Ω p (1.15) (cid:16) u(cid:17)=0 in ∂Ω. and satisfies (1.9 ). OurstudyisbasedupondelicateestimatesonWolffpotentialsandη-fractionalmaximal operators which are developed in the first part of this paper. 2 Lorentz spaces and capacities 2.1 Lorentz spaces Let (X,Σ,α) be a measured space. If f : X →R is a measurable function, we set S (t) := f {x∈X :|f|(x)>t} andλ (t)=α(S (t)). The decreasingrearrangementf∗ off is defined f f by f∗(t)=inf{s>0:λ (s)≤t}. f It is well known that (Φ(f))∗ = Φ(f∗) for any continuous and nondecreasing function Φ : R →R . We set + + 1 t f∗∗(t)= f∗(τ)dτ ∀t>0. t Z0 and, for 1≤s<∞ and 1<q ≤∞, ∞tqs(f∗∗(t))qdt 1q if q <∞ t kfkLs,q = (cid:18)Z0 (cid:19) (2.1)  supessts1f∗∗(t) if q =∞ t>0 It is known that Ls,q(X,α) isa Banach space when endowed with the norm k.k . Fur- Ls,q thermore there holds (see e.g. [12]) s t1sf∗ Lq(R+,dt) ≤kfkLs,q ≤ s−1 t1sf∗ Lq(R+,dt), (2.2) (cid:13) (cid:13) t (cid:13) (cid:13) t (cid:13) (cid:13) (cid:13) (cid:13) the left-hand side i(cid:13)nequa(cid:13)lity being valid only if s >(cid:13)1. Fi(cid:13)nally, if f ∈ Ls,q(RN) (with 1 ≤ q,s < ∞ and α being the Lebesgue measure) and if {ρ } ⊂ C∞(RN) is a sequence of n c mollifiers, f∗ρ →f and(fχ )∗ρ →f inLs,q(RN), whereχ is the indicatorfunction oftheballB cnenteredattheoBrniginonfradiusn. InparticularC∞B(nRN)isdenseinLs,q(RN). n c 4 2.2 Wolff potentials, fractional and η-fractional maximal operators If D is either a bounded domain or whole RN, we denote by M(D) (resp Mb(D)) the set of Radon measure (resp. bounded Radon measures) in D. Their positive cones are M (D) + and Mb (D) respectively. If 0<R≤ ∞ and µ ∈M (D) and R ≥diam(D), we define, for + + α>0 and 1<s<α−1N, the R-truncated Wolff-potential by WR [µ](x)= R µ(Bt(x)) s−11 dt for a.e. x∈RN. (2.3) α,s tN−αs t Z0 (cid:18) (cid:19) If h (t) = min{(−lnt)−η,(ln2)−η} and 0 < α < N, the truncated η-fractional maximal η operator is µ(B (x)) Mη [µ](x)= sup t for a.e. x∈RN. (2.4) α,R tN−αh (t) 0<t<R η If R=∞, we drop it in expressions (2.3 ) and (2.4 ). In particular µ(B (x))≤tN−αh (t)Mη [µ](x). (2.5) t η α,R We also define G the Bessel potential of a measure µ by α G [µ](x)= G (x−y)dµ(y) ∀x∈RN, (2.6) α α ZRN where G is the Bessel kernel of order α in RN. α Definition 2.1 We denote by Lα,s,q(RN)the Besov space the space of functions φ=G ∗f α for some f ∈Ls,q(RN) and we set kφk =kfk . If we set α,s,q s,q C (E)=inf{kfk : f ≥0, G ∗f ≥1 on E}, (2.7) α,s,q s,q α then C is a capacity, see [1]. α,s,q 2.3 Estimates on potentials In the sequel, we denote by |A| the N-dimensional Lebesgue measure of a measurable set A and, if F,G are functions defined in RN, we set {F >a} := {x ∈ RN : F(x) > a}, {G≤b}:={x∈RN :G(x)≤b} and {F >a,G≤b}:={F >a}∩{G≤b}. The following result is an extension of [14, Th 1.1] Proposition 2.2 Let 0≤η <p−1, 0<αp<N and r >0. There exist c >0 depending 0 on N,α,p,η and ǫ > 0 depending on N,α,p,η,r such that, for all µ ∈ M (RN) with 0 + diam(supp(µ))≤r and R∈(0,∞], ǫ∈(0,ǫ ], λ> µ(RN) p−11 l(r,R) there holds, 0 WαR,p[µ]>3λ,(Mηαp,R[µ])p−11 ≤ǫλ (cid:0) (cid:1) (cid:12)n p−1o(cid:12) (2.8) (cid:12)(cid:12) ≤c0exp − p4−(p1−−1η) p−1−η(cid:12)(cid:12)αpln2ǫ−p−p−1−1η {WαR,p[µ]>λ} . (cid:18) (cid:16) (cid:17) (cid:19)(cid:12) (cid:12) where l(r,R) = N−αp min{r,R}−Np−−α1p −R−Np−−α1p if R < ∞(cid:12), l(r,R) = N−(cid:12)αpr−Np−−α1p if p−1 p−1 R=∞. Furthermore,(cid:16)if η =0, ǫ is independent of(cid:17)r and (2.8 ) holds for all µ∈M (RN) 0 + with compact support in RN and R∈(0,∞], ǫ∈(0,ǫ ], λ>0. 0 5 Proof. Case R=∞. Let λ>0; since W [µ] is lower semicontinuous, the set α,p D :={W [µ]>λ} λ α,p is open. By Whitney covering lemma, there exists a countable set of closed cubes {Q } i i o o such that D =∪ Q , Q ∩Q =∅ for i6=j and λ i i i j diam(Q)≤dist(Q,Dc)≤4diam(Q). i i λ i Let ǫ > 0 and Fǫ,λ = Wα,p[µ]>3λ,(Mηαp[µ])p−11 ≤ǫλ . We claim that there exist c0 = c0(N,α,p,η) > 0 and nǫ0 = ǫ0(N,α,p,η,r) > 0 such thatofor any Q ∈ {Qi}i, ǫ ∈ (0,ǫ0] and λ> µ(RN) p−11 l(r,∞) there holds (cid:0) (cid:1) p−1 |Fǫ,λ∩Q|≤c0exp − p−1−η p−1−η ǫ−p−p−1−1ηαpln2 |Q|. (2.9) 4(p−1) (cid:18) (cid:19) ! The first we show that there exists c > 0 depending on N,α,p and η such that for any 1 Q∈{Q } there holds i i F ∩Q⊂E ∀ǫ∈(0,c ],λ>0 (2.10) ǫ,λ ǫ,λ 1 where Eǫ,λ = x∈Q:Wα5,dpiam(Q)[µ](x)>λ,(Mαηp[µ](x))p−11 ≤ǫλ . (2.11) Infact, take Q ∈ {Q }n such that Q∩F 6= ∅ and let x ∈ Dc such thoat dist(x ,Q) ≤ i i ǫ,λ Q λ Q 4diam(Q) and W [µ](x )≤λ. For k ∈N, r =5diam(Q) and x∈F ∩Q, we have α,p Q 0 ǫ,λ 2k+1r0 µ(Bt(x)) p−11 dt =A+B tN−αp t Z2kr0 (cid:18) (cid:19) where 2k1+1+2k2+k1r0 µ(Bt(x)) p−11 dt 2k+1r0 µ(Bt(x)) p−11 dt A= and B = . Z2kr0 (cid:18) tN−αp (cid:19) t Z2k1+1+2k2+k1r0(cid:18) tN−αp (cid:19) t Since µ(B (x))≤tN−αph (t)Mη [µ](x)≤tN−αph (t)(ǫλ)p−1. (2.12) t η αp η Then 2k+1r0 tN−αphη(t)(ǫλ)p−1 p−11 dt 2k+1r0 1 dt B ≤Z2k1+1+2k2+k1r0(cid:18) tN−αp (cid:19) t =ǫλZ2k1+1+2k2+k1r0(hη(t))p−1 t Replacing h (t) by its value we obtain B ≤ c ǫλ2−k after a lengthy computation where c η 2 2 depends onlyonpandη. Since δ :=(2k2+k1)Np−−α1p,then1−δ ≤c32−k where c3 depends only on N−αp, thus p−1 (1−δ)A≤c 2−k 2k+1r0 µ(Bt(x)) p−11 dt 3 tN−αp t Z2kr0 (cid:18) (cid:19) ≤c32−kǫλ 2k+1r0(hη(t))p−11 dt t Z2kr0 ≤c 2−kǫλ, 4 6 where c =c (N,α,p,η)>0. 4 4 Byachangeofvariablesandusingthatforanyx∈F ∩Qandt∈[r (1+2k),r (1+2k+1)], ǫ,λ 0 0 B (x)⊂B (x ), we get 2kt t Q 1+2k 1 δA= r0(1+2k+1) µ(B12+k2tk)(x) p−1 dt ≤ r0(1+2k+1) µ(Bt(xQ)) p−11 dt.  tN−αp  t tN−αp t Zr0(1+2k) Zr0(1+2k) (cid:18) (cid:19)   Therefore 2k+1r0 µ(Bt(x)) p−11 dt ≤c 2−kǫλ+ r0(1+2k+1) µ(Bt(xQ)) p−11 dt, tN−αp t 5 tN−αp t Z2kr0 (cid:18) (cid:19) Zr0(1+2k) (cid:18) (cid:19) with c =c (N,α,p,η)>0. This implies 5 5 ∞ µ(Bt(x)) p−11 dt ∞ µ(Bt(xQ)) p−11 dt ≤2c ǫλ+ ≤(1+2c ǫ)λ, (2.13) tN−αp t 5 tN−αp t 5 Zr0 (cid:18) (cid:19) Z2r0 (cid:18) (cid:19) since W [µ](x )≤λ. If ǫ∈(0,c ] with c =(2c )−1 then α,p Q 1 1 5 ∞ µ(Bt(x) p−11 dt ≤2λ tN−αp t Zr0 (cid:18) (cid:19) which implies (2.10 ). Now,weletλ> µ(RN) p−11 l(r,∞). LetB beaballwithradiusrsuchthatsupp(µ)⊂B . 1 1 We denote B by the ball concentric to B with radius 2r. Since x∈/ B , 2 1 2 (cid:0) (cid:1) W [µ](x)= ∞ µ(Bt(x)) p−11 dt ≤ µ(RN) p−11 l(r,∞). α,p tN−αp t Zr (cid:18) (cid:19) (cid:0) (cid:1) Thus, we obtain D ⊂B . In particular, r =5diam(Q)≤20r. λ 2 0 Next we set m = max(1,ln(40r)), so that 2−mr ≤2−1 if m≥m . Then for any x∈E 0 ln2 0 0 ǫ,λ r0 µt(NB−t(αxp)) p−11 dtt ≤ǫλ r0 (hη(t))p−11dtt Z2−mr0(cid:16) (cid:17) Z2−mr0 2−m0r0 −η dt r0 −η dt ≤ǫλ (−lnt)p−1 +ǫλ (ln2)p−1 t t Z2−mr0 Z2−m0r0 ≤m ǫλ+ (p−1)((m−m0)ln2)1−p−η1ǫλ. 0 p−1−η Forthelastinequalitywehaveuseda1−p−η1−b1−p−η1 ≤(a−b)1−p−η1 validforanya≥b≥0. Therefore, r0 µ(Bt(x)) p−11 dt ≤ 2(p−1) m1−p−η1ǫλ ∀m∈N,m>mp−p−1−1η. (2.14) tN−αp t p−1−η 0 Z2−mr0(cid:18) (cid:19) 7 Set 2−i+1r0 µ(Bt(x)) p−11 dt g (x)= , i tN−αp t Z2−ir0 (cid:18) (cid:19) then Wαr0,p[µ](x)≤ p2−(p1−−1)ηm1−p−η1ǫλ+Wα2−,pmr0[µ](x) ∞ ≤ 2(p−1) m1−p−η1ǫλ+ gi(x) p−1−η i=m+1 X p−1 for all m>mp−1−η. We deduce that, for β >0, 0 ∞ |Eǫ,λ|≤ x∈Q: gi(x)> 1− 2(p−1) m1−p−η1ǫ λ (cid:12)(cid:12)( i=Xm+1 (cid:18) p−1−η (cid:19) )(cid:12)(cid:12) (cid:12) ∞ (cid:12) ≤ (cid:12)(cid:12)x∈Q: gi(x)>2−β(i−m−1)(1−2−β) 1− 2(p(cid:12)(cid:12)−1) m1−p−η1ǫ λ (cid:12)(cid:12)( i=Xm+1 (cid:18) p−1−η (cid:19) )(cid:12)(cid:12) (cid:12) ∞ (cid:12) ≤(cid:12)(cid:12) x∈Q:gi(x)>2−β(i−m−1)(1−2−β) 1− 2(p−1) m1−p−η1ǫ λ (cid:12)(cid:12). p−1−η i=m+1(cid:12)(cid:26) (cid:18) (cid:19) (cid:27)(cid:12) X (cid:12) (cid:12) (cid:12) ((cid:12)2.15) (cid:12) (cid:12) Next we claim that c (N,η) |{x∈Q:g (x)>s}|≤ 6 2−iαp|Q|(ǫλ)p−1. (2.16) i sp−1 To see that, we pick x ∈E and we use the Chebyshev’s inequality 0 ǫ,λ 1 |{x∈Q:g (x)>s}|≤ |g |p−1dx i sp−1 i ZQ 1 r02−i+1 µ(Bt(x)) p−11 dt p−1 = dx sp−1 tN−αp t ZQ Zr02−i (cid:18) (cid:19) ! 1 µ(B (x)) ≤ r02−i+1 :=A. sp−1 (r 2−i)N−αp ZQ 0 Thanks to Fubini’s theorem, the last term A of the above inequality can be rewritten as 1 1 A= χ (y)dµ(y)dx sp−1(r02−i)N−αpZQZRN Br02−i+1(x) 1 1 = χ (x)dxdµ(y) sp−1(r02−i)N−αpZQ+Br02−i+1(0)ZQ Br02−i+1(y) 1 1 ≤ |B (y)|dµ(y) sp−1(r 2−i)N−αp r02−i+1 0 ZQ+Br02−i+1(0) 1 ≤c (N) 2−iαprαpµ(Q+B (0)) 7 sp−1 0 r02−i+1 1 ≤c (N) 2−iαprαpµ(B (x )), 7 sp−1 0 r0(1+2−i+1) 0 8 sinceQ+B (0)⊂B (x ). Usingthefactthatµ(B (x ))≤(ln2)−ηtN−αp(ǫλ)p−1 r02−i+1 r0(1+2−i+1) 0 t 0 for all t>0 and r =5diam(Q), we obtain 0 1 1 A≤c (N,η) 2−iαprαp(r (1+2−i+1))N−αp(ǫλ)p−1 ≤c (N,η) 2−iαp|Q|(ǫλ)p−1, 8 sp−1 0 0 9 sp−1 which is (2.16 ). Consequently, (2.15 ) can be rewritten as ∞ c (N,η) |E |≤ 6 2−iαp(ǫλ)p−1|Q| ǫ,λ i=m+1 2−β(i−m−1)(1−2−β) 1− 2(p−1)m1−p−η1ǫ λ p−1 X p−1−η (cid:16) (cid:16) p−1 (cid:17) (cid:17) ∞ ≤c (N,η)2−(m+1)αp ǫ |Q| 1−2−β −p+1 2(β(p−1)−αp)(i−m−1). 6 1− 2(p−1)m1−p−η1ǫ! p−1−η (cid:0) (cid:1) i=Xm+1 (2.17) If we choose β =β(α,p) so that β(p−1)−αp<0, we obtain p−1 ǫ p−1 |E |≤c 2−mαp |Q| ∀m>mp−1−η (2.18) ǫ,λ 10 1− 2(p−1)m1−p−η1ǫ! 0 p−1−η where c = c (N,α,p,η) > 0. Put ǫ = min 1 ,c . For any ǫ ∈ (0,ǫ ] we choose m10∈N s1u0ch that 0 (cid:26)p4(−p1−−1η)m0+1 1(cid:27) 0 p−1 p−1 p−1 p−1 p−1−η p−1−η 1 p−1−η p−1−η p−1−η 1 p−1−η −1 −1<m≤ −1 . 2(p−1) ǫ 2(p−1) ǫ (cid:18) (cid:19) (cid:18) (cid:19) (cid:18) (cid:19) (cid:18) (cid:19) Then p−1 ǫ ≤1 1− 2(p−1)m1−p−η1ǫ! p−1−η and p−1 2−mαp ≤2αp−αp(p2(−p1−−1η))p−p−1−1η(1ǫ−1)p−p−1−1η ≤2αpexp −αpln2 p−1−η p−1−η ǫ−p−p−1−1η . 4(p−1) (cid:18) (cid:19) ! Combining these inequalities with (2.18 ) and (2.10 ), we get (2.9 ). In the case η =0 we still have for any m∈N, λ,ǫ>0 and x∈E ǫ,λ ∞ Wr0 [µ](x)≤mǫλ+ g (x) α,p i i=m+1 X Accordingly (2.18 ) reads as p−1 ǫ |E |≤c 2−mαp |Q| ∀m∈N,λ,ǫ>0 with mǫ<1. ǫ,λ 10 1−mǫ (cid:18) (cid:19) 9 Put ǫ = min{1,c }. For any ǫ ∈ (0,ǫ ] and m ∈ N satisfies ǫ−1−2 < m ≤ ǫ−1−1, we 0 2 1 0 finally get from (2.10 ) |F ∩Q|≤|E |≤c 22αpexp −αpǫ−1ln2 |Q|, (2.19) ǫ,λ ǫ,λ 10 which ends the proof in the case R=∞. (cid:0) (cid:1) Case R < ∞. For λ > 0, D = {WR > λ} is open. Using again Whitney covering λ α,p lemma, there exists a countable set of closed cubes Q := {Q } such that ∪ Q = D , i i i λ o o Q ∩Q =∅ fori6=j anddist(Q ,Dc)≤4diam(Q). IfQ∈Q:is suchthatdiam(Q)> R, i j i λ i 8 there exists a finite number n of closed dyadic cubes {P }nQ such that ∪nQ P = Q, Q j,Q j=1 j=1 j,Q o o P ∩P = ∅ if i 6= j and R < diam(P )≤ R. We set Q′ = Q∈Q:diam(Q)≤ R , i,Q j,Q 16 j,Q 8 8 Q′′ = P :1≤i≤n ,Q∈Q,diam(Q)> R and F =Q′∪Q′′. i,Q Q 8 (cid:8) (cid:9) For ǫ >(cid:8) 0 we denote again Fǫ,λ = WαR,p[µ]>(cid:9)3λ,(Mηαp,R[µ])p−11 ≤ǫλ . Let Q ∈ F such that Fǫ,λ∩Q6=∅ and r0 =5diam(nQ). o If dist(Dc,Q) ≤ 4diam(Q), that is if there exists x ∈ Dc such that dist(x ,Q) ≤ λ Q λ Q 4diam(Q) and WR [µ](x ) ≤ λ, we find, by the same argument as in the case R = ∞, α,p Q (2.13 ), that for any x∈F ∩Q there holds ǫ,λ R µ(Bt(x)) p−11 dt ≤(1+c ǫ)λ. (2.20) tN−αp t 11 Zr0 (cid:18) (cid:19) where c =c (N,α,p,η)>0. 11 11 If dist(Dc,Q) > 4diam(Q), we have R < diam(Q) ≤ R since Q ∈ Q′′. Then, for all λ 16 8 x∈F ∩Q, there holds ǫ,λ R µ(Bt(x)) p−11 dt R tN−αp(ln2)−η(ǫλ)p−1 p−11 dt ≤ tN−αp t tN−αp t Zr0 (cid:18) (cid:19) Z51R6 (cid:18) (cid:19) =(ln2)−p−η1 ln16 ǫλ (2.21) 5 ≤2ǫλ. Thus, if we take ǫ∈(0,c ] with c =min{1,c−1}, we derive 12 12 11 F ∩Q⊂E , (2.22) ǫ,λ ǫ,λ where 1 E = Wr0 [µ]>λ, Mη [µ] p−1 ≤ǫλ . ǫ,λ α,p αp,R (cid:26) (cid:16) (cid:17) (cid:27) Furthermore, since x∈/ B , 2 WR [µ](x)= R µ(Bt(x)) p−11 dt ≤ µ(RN) p−11 l(r,R). α,p tN−αp t Zmin{r,R}(cid:18) (cid:19) (cid:0) (cid:1) 10

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.