scieee Open visual document viewer

On the regularity of very weak solutions for linear elliptic equations in divergence form

La Manna, Domenico Angelo,Leone, Chiara,Schiattarella, Roberta

Full text

This is a sel -a chi ed e sion o an o iginal a icle. This e sion may di e om he o iginal in pagina ion and ypog aphic de ails. Au ho (s): Ti le: Yea : Ve sion: Copy igh : Righ s: Righ s u l: Please ci e he o iginal e sion: CC BY 4.0 h ps://c ea i ecommons.o g/licenses/by/4.0/ On he egula i y o e y weak solu ions o linea ellip ic equa ions in di e gence o m © 2020 he Au ho s Published e sion La Manna, Domenico Angelo; Leone, Chia a; Schia a ella, Robe a La Manna, D. A., Leone, C., & Schia a ella, R. (2020). On he egula i y o e y weak solu ions o linea ellip ic equa ions in di e gence o m. Nodea: Nonlinea Di e en ial Equa ions and Applica ions, 27(5), A icle 43. h ps://doi.o g/10.1007/s00030-020-00646-8 2020 Nonlinea Diffe . Equ. Appl. (2020) 27:43 c 2020 The Au ho (s) h ps://doi.o g/10.1007/s00030-020-00646-8 Nonlinea Diffe en ial Equa ions and Applica ions NoDEA On he egula i y o e y weak solu ions o linea ellip ic equa ions in di e gence o m Domenico Angelo La Manna, Chia a Leone and Robe a Schia a ella Abs ac . In his pape we conside a linea ellip ic equa ion in di e gence o m  i,j Dj(aij (x)Diu)=0 inΩ.(0.1) Assuming he coefficien s aij in W1,n(Ω) wi h a modulus o con inui y sa is ying a ce ain Dini- ype con inui y condi ion, we p o e ha any e y weak solu ion u∈Ln loc(Ω) o (0.1) is ac ually a weak solu ion in W1,2 loc (Ω). Ma hema ics Subjec Classifica ion. P ima y 35-02, Seconda y 35B65. Keywo ds. Ellip ic equa ions, Ve y weak solu ions. 1. In oduc ion Le n≥2 and Ω ⊂Rnbe a bounded open se . In his pape we s udy egula i y p ope ies o e y weak solu ions o he linea ellip ic equa ion  i,j Dj(aij(x)Diu)=0 inΩ,(1.1) whe e he ma ix-field A:Ω→Rn×n,A(x)=(aij(x))i,j, is ellip ic and belongs o W1,n(Ω,Rn×n)∩L∞(Ω,Rn×n), i.e. sup i,j=1,...,n aijW1,n(Ω) ≤M(1.2) and λ|ξ|2≤ i,j aij(x)ξiξj≤Λ|ξ|2∀ξ=(ξ1,...,ξ n)∈Rn,a.e. in Ω,(1.3) o some posi i e cons an s λ, Λ,and M. Mo eo e , he ma ix Ais symme ic, ha is aij =aji a.e. in Ω o all i, j ∈{1, ..., n}. 0123456789().: V,- ol 43 Page 2 o 23 D. A. L. Manna e al. NoDEA Finally we assume ha he coefficien s (aij(x))i,j a e double-Dini con in- uous in Ω, i.e. aij ∈C0(Ω) and ¯ AΩ( ):= i,j sup x,y∈Ω |x−y|≤ |aij(x)−aij(y)|, >0, sa isfies ˆdiam(Ω) 0 1 ˆ 0 ¯ AΩ(s) sds d < ∞.(1.4) A common ype o double-Dini con inuous unc ions a e, o cou se, ω( )= α, 0<α≤1, hus an example o a ma ix-field Asa is ying (1.2) and (1.4)is A∈W1,p(Ω,Rn×n), wi h p>n. On he o he hand, condi ion (1.4) occu s no only o ω( )= α, bu mo e gene ally o ω( ) = logβ1 ,β<−2. Gi en a measu able ma ix A(x)=(aij(x))i,j sa is ying (1.3), a unc ion u∈W1,2 loc (Ω) is called a weak solu ion o (1.1)i  i,j ˆaij(x)DiuDjϕdx=0,∀ϕ∈C∞ c(Ω). The celeb a ed esul by De Gio gi in [5] s a es ha i uis a weak solu ion o (1.1) hen uis locally H¨olde con inuous. Subsequen ly, J. Se in p oduced in [14] a amous example, cons uc ing an equa ion o he o m (1.1) which has a solu ion u∈W1,p(Ω), wi h 1 < p<2, and u/∈L∞ loc(Ω). Se in conjec u ed ha i he coefficien s aij a e locally H¨olde con inuous, hen any solu ion (in he sense o dis ibu ions) u∈W1,1 loc (Ω) o (1.1) mus be a (usual) weak solu ion, i.e. u∈W1,2 loc (Ω). Se in’s conjec u e was es ablished by Hage and Ross [9], and hen in ull gene ali y by B ezis [2](seealso[1] o a ull p oo ) s a ing wi h u∈W1,1 loc (Ω), o e en wi h u∈BVloc(Ω), i.e., u∈L1 loc(Ω) and i s de i a i es (in he sense o dis ibu ions) being Radon measu es. Le us ema k ha in B ezis’s esul he coefficien s aij, sa is ying (1.3), a e Dini con inuous unc ions in Ω. The Dini con inui y o he coefficien s is op imal in some sense: o he uni ball B1 and con inuous coefficien s, Jin e al. [10] cons uc ed a solu ion (in he sense o dis ibu ions) u∈W1,1 loc (B1) W1,p loc (B1) o e e y p>1. Fo A(x)=(aij(x))i,j sa is ying (1.2) and (1.3), we will conside a e y weak solu ion u∈Ln loc(Ω) o (1.1), namely  i,j ˆu(x)Di(aij(x)Djϕ(x)) dx =0,∀ϕ∈C∞ c(Ω),(1.5) wi h n=n n−1. Rema k 1.1. I is no difficul o p o e ha he es unc ions ϕin (1.5)can be aken in W2,n(Ω) ∩W1,∞(Ω), wi h supp ϕΩ. Indeed, one can a gue by densi y o show ha gi en a unc ion ϕ∈W2,n(Ω) ∩W1,∞(Ω) wi h compac suppo , we may find a sequence ϕk∈C∞ c(Ω) such ha ϕk→ϕs ongly in W2,n and supkϕk1,∞<∞(so ha Dϕkcon e ges o Dϕ weakly* in L∞) and hen aking he limi as kgoes o infini e in he Eq. (1.5) o ϕk. NoDEA On he egula i y o e y weak solu ions Page 3 o 23 43 The main esul o he pape is he ollowing. Theo em 1.2. Le ube a e y weak solu ion o (1.1),wi hA(x)=(aij(x))i,j sa is ying (1.2),(1.3)and (1.4), hen ubelongs o W1,2 loc (Ω) and hus i is a weak solu ion. Rema k 1.3. I is wo h no ing ha , unde hypo heses (1.2) and (1.3), one can conside a e y weak solu ion u∈Ln loc(Ω) o (1.1), bu when dealing wi h he egula i y p ope ies o usome ex a condi ions on he coefficien s aij mus be conside ed. The coun e example cons uc ed in [10] p o ides in ac con inuous coefficien s aij which belong also o W1,n(B1), showing ha one can no expec a e y weak solu ion u∈Ln loc(Ω) o be a weak solu ion in W1,2 loc (Ω) unde jus condi ions (1.2) and (1.3). Fo he sake o comple eness, we will p opose he example gi en in [10] in he “Appendix B”, unde lining ha he cons uc ed coefficien s belong also o W1,n(B1). On he o he hand, in Sec . 4we p opose an al e na i e o double Dini con inuous coefficien s which again bypasses he coun e example. In pa icula , unde hypo heses (1.2) and (1.3) we conside a e y weak solu ion in Lq loc(Ω), wi h q>n . Rema k 1.4. In [15] Zhang and Bao deal wi h he case o e y weak solu ions u∈L1 loc(Ω) o (1.5), in e p e ing he coefficien s as Lipschi z unc ions, due o he assump ion made on he solu ions. Thus ou esul ep esen s a na u al ex ension om hei esea ch. 2. No a ion and p elimina y esul s We collec he e he main defini ions and no a ion and some use ul esul s ha will be needed in he sequel. 2.1. No a ion In he ollowing, we deno e by B (x)={y∈Rn:|y−x|< } he ball o adius cen e ed a x. We indica e by {e1,...e n} he canonical basis o Rn.Gi enh∈R {0}, o a measu able unc ion ψ:Rn→Rand o =1,...,n, we in oduce he no a ion Δ hψ:= ψ(x+he)−ψ(x) h o he inc emen al quo ien in he - h di ec ion. We ecall ha o e e y pai o unc ions ϕ, ψ,weha e Δ h(ϕψ)=Δ  hϕψ+ϕ(x+he)Δ  hψ. (2.1) The ollowing esul pe aining o diffe ence quo ien s o unc ions in Sobole spaces is well known (see [8, P oposi ion 4.8] o example). 43 Page 4 o 23 D. A. L. Manna e al. NoDEA Theo em 2.1. Le p>1;i ψ∈W1,p(Ω), hen Δ hψ∈Lp(Ω) o any ΩΩ sa is ying h<dis (Ω,∂Ω) 2, and we ha e Δ hψLp(Ω)≤DψLp(Ω). I ψ∈Lp(Ω) and he e exis s L≥0such ha , o e e y h<dis (Ω,∂Ω), =1,...,n, we ha e Δ hψLp(Ω)≤L, hen ψ∈W1,p(Ω),DψLp(Ω)≤Land Δ hψ→Dψin Lp(Ω)as h→0. Finally, gi en p>1, we deno e by p=p p−1 he conjuga e exponen o p. 2.2. Dini con inuous unc ions We say ha a con inuous unc ion on Ω is Dini con inuous i he modulus o con inui y ¯ Ω:[0,diam(Ω)] →R+defined by ¯ Ω( ):= sup x,y∈Ω |x−y|≤ | (x)− (y)| sa isfies ˆdiam(Ω) 0 ¯ Ω( ) d < ∞. We also deno e by CD(Ω) he space o Dini con inuous unc ions; i u ns ou o be a Banach space equipped wi h he ollowing no m:  CD(Ω) :=  ∞+ˆdiam(Ω) 0 ¯ Ω( ) d , whe e · ∞is he usual uni o m no m. Le us ema k ha by he uni o m con inui y, any unc ion in CD(Ω) may be ex ended up o he bounda y o Ω wi h he same modulus o con inui y. Mo eo e , C0,α(Ω) ⊆CD(Ω), o any 0 <α≤1, whe e C0,α(Ω) deno es he space o H¨olde con inuous unc ions. The space CD c(Ω) will deno e he se o unc ions in CD(Ω) wi h compac suppo in Ω. Lemma 2.2. The space C∞ c(Ω) is dense in CD c(Ω). P oo . Le ∈CD c(Ω) ha we ex end o ze o on Rn Ω and se ε(x)= (ρε∗ )(x), whe e ρεis a s anda d mollifie . Then, i εis sufficien ly small, ε∈C∞ c(Ω); we will p o e ha ε→ in CD(Ω).(2.2) I is easily seen ha εuni o mly con e ges o in Ω, hus in o de o p o e (2.2) we will jus show ha ˆdiam(Ω) 0 ( − ε)Ω( ) d →0, NoDEA On he egula i y o e y weak solu ions Page 5 o 23 43 as ε ends o 0. Obse e ha ( − ε)Ω( )= sup x,y∈Ω |x−y|< {| ε(x)− (x)− ε(y)+ (y)|} ≤ ¯ Ω( )+(¯ ε)Ω( ) and (¯ ε)Ω( )= sup x,y∈Ω |x−y|< {| ε(x)− ε(y)|} =sup x,y∈Ω |x−y|< ˆρε(z)( (x−z)− (y−z)) dz ≤ˆρε(z)¯ Ω( )dz =¯ Ω( ), which oge he yield ( ε− )Ω( )≤2¯ Ω( ). On he o he hand, since ( ε− )Ω→0 poin wise, he domina ed con e gence heo em implies ˆdiam(Ω) 0 ( ε− )Ω( ) →0, which concludes he p oo o (2.2).  Rema k 2.3. The p e ious esul ensu es ha CD c(Ω) is a sepa able space, no ing ha C1 c(Ω) is sepa able wi h espec o he usual no m  1,∞:= |α|≤1Dα ∞,C1 c(Ω) ⊆CD c(Ω) and ¯ Ω( )≤ D ∞, o e e y ∈C1 c(Ω). Lemma 2.4. Le , ε,and gbelonging o CD(Ω) such ha εcon e ges o in CD; hen g εcon e ges o g in CD. P oo . As be o e, i is enough o p o e he con e gence o he semino m since he uni o m con e gence is immedia e. Then, w i ing he defini ion o he modulus o con inui y, we ha e g( ε− ) Ω( )= sup x,y∈Ω |x−y|< {|g(x)( ε(x)− (x)) −g(y)( ε(y)− (y))|} ≤sup x,y∈Ω |x−y|< {|g(x)||( ε(x)− (x)) −( ε(y)− (y))|} +sup x,y∈Ω |x−y|< {|g(x)−g(y)|| (y)− ε(y)|} ≤g∞( − ε)Ω( )+¯gΩ( ) − ε∞. (2.3) Hence, ˆdiam(Ω) 0 [g( ε− )]Ω( ) d ≤g∞ˆdiam(Ω) 0 ( − ε)Ω( ) d + − ε∞ˆdiam(Ω) 0 ¯gΩ( ) d , 43 Page 6 o 23 D. A. L. Manna e al. NoDEA which goes o ze o as ε ends o ze o.  2.3. C1-Dini egula i y o solu ions o di e gence o m ellip ic equa ions wi h Dini-con inuous coefficien s Fo he p oo o ou esul , we will need he ollowing ex ension o he Schaude egula i y heo y o ellip ic equa ions in di e gence o m wi h Dini con inuous coefficien s (see [12, Theo em 1.1] and [6, Theo em 1.3], see also [11]whichis inclusi e o he pa abolic case.). Fo he Lp- egula i y heo y we e e o [7], whe e he gene al case o VMO coefficien s is ea ed (see also [13, Theo em 5.5.3 (a)] o [3, Theo em 2.2. Chap e 10] o he case o con inuous coeffi- cien s). Theo em 2.5. Fo Ω⊂Rn,le aij sa is y (1.3)and (1.4); we conside = ( 1, 2,..., n)wi h j∈C∞ c(Ω) o all j∈{1,...,n}. Assume ha u∈H1(Ω) is a weak solu ion o he equa ion  i,j Dj(aijDiu)= j Dj jin Ω(2.4) Then u∈C1,D(Ω), o any bounded open se Ω,ΩΩ. Mo eo e , le ΩaC1,1bounded open subse o Rn,le aij sa is y (1.2) and (1.3),andle j∈Lp(Ω), o e e y j∈{1,...,n},wi h1<p<∞, hen he e exis s a unique solu ion u∈W1,p 0(Ω) o he p oblem  i,j ˆΩ aijDiuDjϕdx= jˆΩ jDjϕdx ∀ϕ∈W1,p 0(Ω), and uW1,p(Ω) ≤C j  jLp(Ω) (2.5) holds, whe e Cdepends on n, λ, Λ,p,∂Ω,AW1,n(Ω,Rn×n). Rema k 2.6. The fi s conclusion o Theo em 2.5 comes wi h an es ima e o he Dini modulus o con inui y o Du in ol ing he Dini modulus o con i- nui y o aij and j. Ac ually, in [12, Theo em 1.1] and in [6, Theo em 1.3] only he con inui y o Du is p o ed and hese esul s a e ob ained wi h a weake assump ion on he coefficien s aij . Assuming (1.4) o he coefficien s we a e able o p o e also he Dini con inui y o he g adien o he solu ion. In “Appendix A” we will esume in b oad e ms he p oo o [12, Theo em 1.1], de eloping i in o de o ge he needed Dini con inui y esul . 2.4. C2- egula i y o solu ions o non di e gence o m ellip ic equa ions wi h Dini-con inuous coefficien s Le us fi s ecall he W2,p-sol abili y o he Di ichle p oblem o non di e - gence ellip ic equa ions wi h discon inuous coefficien s (see [4, Theo em 4.2 and Theo em 4.4]). Theo em 2.7. The Di ichle p oblem ⎧ ⎨ ⎩  i,j aij(x)Diju= a.e.in Ω u=0 on ∂Ω (2.6) NoDEA On he egula i y o e y weak solu ions Page 7 o 23 43 whe e Ωis a C1,1smoo h and bounded subse o Rn, ∈Lp(Ω) wi h 1<p< ∞,andaij sa is ies (1.2)and (1.3), admi s a unique solu ion u∈W2,p(Ω) ∩ W1,p 0(Ω) and ||u||W2,p(Ω) ≤C||u||Lp(Ω) +|| ||Lp(Ω),(2.7) whe e he cons an Cdepends on n, p, λ, Λ,∂Ω,AW1,n(Ω,Rn×n). The nex esul specifies es ima e (2.7); i s p oo is qui e s anda d bu we p e e o w i e i o he sake o comple eness. P oposi ion 2.8. Suppose uis a solu ion o he ellip ic Di ichle p oblem (2.6) wi h aij, ,p and Ωas abo e. Then ||u||W2,p(Ω) ≤C|| ||Lp(Ω).(2.8) P oo . Le L=L= i,j aijDij ,sup i,j ||aij||W1,n(Ω) ≤2M, λ|ξ|2≤ i,j aij(x)ξiξj≤Λ|ξ|2; ha ing in mind Theo em 2.7, i we p o e ha o any ope a o L∈Land o any ∈Lp(Ω), he solu ion uo Lu = a.e. in Ω u=0 on∂Ω, sa isfies ||u||Lp(Ω) ≤C|| ||Lp(Ω), we a e done. Suppose i is no he case, hen his is equi alen o say ha o e e y N∈N, he e exis s an ope a o LN=i,j aN ij Dij ∈Land a unc ion N∈Lp(Ω) such ha he co esponding solu ion uN o he Di ichle p oblem LNuN= Na.e. in Ω uN=0 on∂Ω, sa isfies ||uN||Lp(Ω) >N|| N||Lp(Ω).(2.9) Le us define N=uN/uNLp(Ω) and gN= N/uNLp(Ω),so ha N sol es (2.6) wi h LNand gN.By heW2,p es ima e (2.7), || N||W2,p(Ω) ≤C|| N||Lp(Ω) +||gN||Lp(Ω)<C1+ 1 N, whe e Cdoes no depend on Nand hence, || N||W2,p(Ω) ≤C. (2.10) Thus Nis a p ecompac sequence: up o a non elabeled subsequence, we can suppose Nu ∗weakly in W2,p(Ω), o some u∗∈W2,p(Ω), mo eo e u∗∈ W2,p(Ω) ∩W1,p 0(Ω). Simila ly, we can also say ha , o e e y i, j =1,...,n, 43 Page 8 o 23 D. A. L. Manna e al. NoDEA aN ij a ∗ ij weakly in W1,n(Ω) and aN ij →a∗ ij s ongly in Lq(Ω) ∀1≤q<∞. Thus, he ope a o L∗=i,j a∗ ijDij belongs o Land o ϕ∈Lp(Ω) we ha e ˆΩ (LN N−L∗u∗)ϕdx  ≤ n  i,j=1 ˆΩ (aN ij −a∗ ij)∂2 N ∂xi∂xj ϕ dx +ˆΩ a∗ ijϕ∂2 N ∂xi∂xj −∂2u∗ ∂xi∂xjdx ≤C n  i,j=1 (aN ij −a∗ ij)ϕLp(Ω) + n  i,j=1 ˆΩ a∗ ijϕ∂2 N ∂xi∂xj −∂2u∗ ∂xi∂xjdx. The e o e, LN Ncon e ges weakly in Lp(Ω) o L∗u∗. On he o he hand, using (2.9), we ha e ||gN||Lp(Ω) <1 N. Passing o he limi in he equa ion sa isfied by N, we disco e ha he limi u∗∈W2,p(Ω) ∩W1,p 0(Ω) sa isfies L∗u∗= 0 a.e. in Ω. By he uniqueness p ope ies o he solu ions o (2.6), i ollows ha u∗=0.Thus Ncon e ges o ze o and he a gumen becomes con adic o y since  NLp(Ω) =1.  In [6, Theo em 1.5] i is shown ha solu ions o ellip ic equa ions in non di e gence o m wi h ze o Di ichle bounda y condi ions a e C2up o he bounda y when he leading coefficien s a e Dini con inuous unc ions. Theo em 2.9. Assume ha Ωis a C2,1smoo h and bounded open subse o Rn, ∈CD(Ω) and aij sa is ies (1.2),(1.3),and(1.4).Le u∈W2,2(Ω)∩W1,2 0(Ω) be a solu ion o he Di ichle p oblem ⎧ ⎨ ⎩  i,j aij(x)Diju= a.e. in Ω u=0 on ∂Ω, (2.11) hen u∈C2(Ω). Rema k 2.10. The assump ion in [6] abou he coefficien s is weake hen (1.4), since hey assume ha he modulus o con inui y ˜ AΩ( ):= i,j sup x∈Ω − ˆB (x)∩Ω |aij(y)−(aij)B (x)∩Ω|dy wi h (aij)B (x)∩Ω=− ˆB (x)∩Ω aij, sa isfies ˆ0 ˜ AΩ( ) d < ∞. NoDEA On he egula i y o e y weak solu ions Page 15 o 23 43 A he end, he es ima es p o ed o I1,I 2and I3lead o  j μj,ξ j≤C Lp(Ω,Rn), as well  j μjξ, j≤C Lp(Ω,Rn). Since is an a bi a y smoo h unc ion in Lp(Ω,Rn), we conclude  j μjξLp(Ω)≤C, which means, using a fini e co e ing a gumen , ha μjis a unc ion in Lp loc(Ω) and hen u∈W1,p loc (Ω), since, o e e y ϕ∈C∞ c(Ω) and o hsmall enough, we ha e ˆΔj huϕdx=ˆuΔj −hϕdx; passing o he limi as h→0, we de i e μj,ϕ=ˆϕμjdx =−ˆuDjϕdx. Since u∈W1,p loc (Ω), B ezis’s esul implies ha uis a weak solu ion o he equa ion (1.5), i.e. ou s a emen .  4. Sobole coefficien s As poin ed ou in he In oduc ion, e y weak solu ions in Ln loc(Ω) associa ed o coefficien s in W1,n(Ω) a e no weak solu ions, due o he coun e example ound in [10]. The quo ed e e ences on his p oblem ha e sugges ed us o conside Sobole coefficien s wi h a modulus o con inui y sa is ying he double Dini condi ion. On he o he hand, ano he way o ge a ound he coun e example is o deal wi h e y weak solu ions in Lq loc(Ω), wi h q>n . The esul is he ollowing. Theo em 4.1. Le u∈Lq loc(Ω),q>n , be a e y weak solu ion o (1.1),wi h A(x)=(aij (x))i,j sa is ying (1.2)and (1.3), hen ubelongs o W1,2 loc (Ω) and hus i is a weak solu ion. P oo . The p oo es s on a duali y and a boo s ap a gumen . S ep 1 We claim ha u∈W1,qn q−n loc (Ω). We p oceed as in he S ep 1 o he p oo o Theo em 1.2 o a i e o (3.2). Now we es ima e he six e ms Im.WeuseH¨olde ’s inequali y and P oposi ion 2.8 o ge 43 Page 16 o 23 D. A. L. Manna e al. NoDEA |I1|≤DηL∞(Rn)AW1,n(Ω,Rn×n)uLq(Ω0)D  L qn q−n(Ω0,Rn) ≤Cw L qn q−n(Ω0) , |I2|≤AW1,n (Ω,Rn×n)uLq(Ω0)D2  L qn q−n(Ω0,Rn×n) ≤Cw L qn q−n(Ω0) , |I3|,|I4|≤ΛuLq(Ω0)DηL∞(Rn,Rn)D2 Lq(Ω0,Rn×n)≤CwLq(Ω)≤Cw L qn q−n(Ω) , |I5|≤ΛuLq(Ω0)D2ηL∞(Rn,Rn×n)D Lq(Ω0,Rn)≤CwLq(Ω)≤Cw L qn q−n(Ω) , and finally, as o (3.3), |I6|≤Cw L qn q−n(Ω) +CwLq(Ω)≤Cw L qn q−n(Ω) . So, a guing as in he S ep 1 o Theo em 1.2, we deduce ˆΩ wΔ hudx  ≤Cw L qn q−n(Ω) , which in u n implies, hanks also o Theo em 2.1, ha u∈W1,qn q−n loc (Ω). Le us no e ha hanks o his, he equa ion sa isfied by umay be ew i en as  i,j ˆaij(x)DiuDjϕdx=0,(4.1) whe e he es unc ions ϕcan be aken in W1,qn q−n(Ω) wi h compac suppo . On he o he hand, he summabili y o he solu ion uis no imp o ed by i s belonging o his Sobole space, since qn q−n=qn qn−q+nand he Sobole conjuga e o qn qn−q+nis q. S ep 2 We p o e ha u∈W1,q loc (Ω). As in he S ep 2 o he p oo o Theo em 1.2, o j∈{1,...,n}le =( 1,..., n) wi h j∈C∞ c(Ω) be such ha  j || j||Lq(Ω)≤1. Fo e e y p>1, le ∈W1,p 0(Ω) be he weak solu ion o he p oblem  i,j ˆaijDi Djϕdx= jˆDjϕ jdx ∀ϕ∈W1,p 0(Ω).(4.2) By Theo em 2.5 we ha e in pa icula ha || ||W1,q(Ω)≤C|| ||Lq(Ω,Rn). NoDEA On he egula i y o e y weak solu ions Page 17 o 23 43 As be o e, we ake BR/2⊂BR⊂Ωa pai o concen ic balls cen e ed a x0∈Ωandweconside ξ(x)=ξ(|x−x0|) a smoo h unc ion such ha ξ( )=1 o ∈[0,R/2] and ξ( )=0 o ≥R.Wecanchooseϕ= ξ in (4.1)andϕ=uξ as es unc ion in (4.2), so ha  i,j ˆaijDiuDj ξdx+ i,j ˆaijDiuDjξ dx =0, and  i,j ˆaijDi Djuξdx+ i,j ˆaijDi Djξudx = jˆ jDjuξdx+ jˆ jDjξudx. Sub ac ing he wo equa ions and using he symme y o aij we ge  jˆ jDjuξdx=− jˆ jDjξudx + i,j ˆaijDi Djξudx − i,j ˆaijDiuDjξ dx =I1+I2+I3. We es ima e he h ee e ms Im.Weha e |I1|≤uLq(Ω)DξL∞(BR,Rn) Lq(Ω,Rn)≤C Lq(Ω,Rn), |I2|≤ΛDξL∞(BR,Rn)uLq(Ω)D Lq(Ω,Rn)≤C Lq(Ω,Rn), and finally |I3|≤Λu W1,(qn q−n) (Ω)   L qn q−n(Ω) ≤C W1,q(Ω), whe e he las inequali y de i es om he ac ha he Sobole conjuga e o qis qn q−n. To sum up we ha e ob ained  jˆ jξDjudx ≤C Lq(Ω,Rn), as well ξDuLq(Ω,Rn)≤C, and, using a fini e co e ing a gumen , his implies ha u∈W1,q loc (Ω). Le us obse e ha his Sobole egula i y imp o es he summabili y o u.In pa icula , u∈Lq∗ loc(Ω), whe e q∗is he Sobole conjuga e o q. S ep 3 We claim ha i q>n hen uis a weak solu ion. By he p e ious s ep, we deduce ha i q>n hen he solu ion uis in L∞ loc(Ω). A his poin , i is no difficul o p o e, a guing as in S ep 1, ha u∈W1,n loc (Ω). S ep 4 We p o e ha u∈L∞ loc(Ω). 43 Page 18 o 23 D. A. L. Manna e al. NoDEA We jus obse ed ha i q>nwe a e done. Le us conside now q≤n. The solu ion uis in W1,q loc (Ω) and by he Sobole ’s embedding u∈Lq∗ loc(Ω), whe e q∗=qn n−qi q<nand any numbe g ea e hen 1 i q=n. A guing exac ly as in he S ep 2 we de i e ha u∈W1,q∗ loc (Ω), which in u n implies ha uis in L∞ loc(Ω) i q∗>n. We al eady no iced in S ep 3 ha his gi es he desi ed esul . Le us obse e ha i q=n,q∗is any numbe g ea e hen 1 and so his can be chosen g ea e hen n, while i q<n,q∗>nis equi alen o q>n 2. We can i e a e his p ocedu e. Gi en q>n =n n−1a e (a mos ) n−1 imes we deduce ha uis locally bounded. By S ep 3 he locally boundedness o he solu ion gi es he desi ed esul .  Acknowledgemen s Open access unding p o ided by Uni e si `a degli S udi di Napoli Fede ico II wi hin he CRUI-CARE Ag eemen . Open access unding p o ided by Uni- e si `a degli S udi di Napoli Fede ico II wi hin he CRUI-CARE Ag eemen . The au ho s a e membe s o G uppo Nazionale pe l’Analisi Ma ema ica, la P obabili `a e le lo o Applicazioni (GNAMPA) o INdAM. The fi s au ho was pa ially suppo ed by he Academy o Finland G an 314227. The esea ch o R. S. has been unded by PRIN P ojec 2017JFFHSH. Open Access. This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License, which pe mi s use, sha ing, adap a ion, dis ibu ion and ep o- duc ion in any medium o o ma , as long as you gi e app op ia e c edi o he o ig- inal au ho (s) and he sou ce, p o ide a link o he C ea i e Commons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his a icle a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line o he ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended use is no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission di ec ly om he copy igh holde . To iew a copy o his licence, isi h p://c ea i ecommons.o g/ licenses/by/4.0/. Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic- ional claims in published maps and ins i u ional affilia ions. Appendix A: The C1-Dini egula i y o solu ions o di e gence o m ellip ic equa ions wi h Dini-con inuous coefficien s As announced in Rema k 2.6, we will speci y he modulus o con inui y o he g adien o solu ions o (2.4) in he p oo o [12, Theo em 1.1]. We will conside only he main poin s o he p oo e e ing o he es o [12]. The se Ω is supposed o be he ball B4cen e ed a 0 and Ω=B1. The imp o emen ega ds P oposi ion 1.1 in [12]: o he sake o comple eness we will ske ch he p oo , modi ying he o iginal when needed. NoDEA On he egula i y o e y weak solu ions Page 19 o 23 43 P oposi ion A.1. Fo B4⊂Rn,n≥1,le aij, de ined on B4, sa is y (1.3) and (1.4)and le =( 1, 2,..., n)wi h j∈C∞ c(B4) o all j∈{1,...,n}. Assume ha u∈H1(B4)is a weak solu ion o (2.4), hen he e exis a∈R and b∈Rnsuch ha − ˆB |u(x)−(a+b·x)|dx ≤ δ( )[uL2(B2)+ C1(B2)],∀ ∈(0,1),(A.1) whe e δ( ), depending on n, λ, Λ,and on he modulus o con inui y o aij and , is a mono onically inc easing posi i e unc ion de ined on (0,1) sa is ying ˆ1 0 δ( ) d < +∞. Rema k A.2. As shown in [12, P oposi ion 1.2], δ( ) will be he modulus o con inui y o Du. P oo . The p oo is ca ied ou o = 0. We use he same no a ion o [12], deno ing by ϕ he modulus o con inui y such ha − ˆB |A−A(0)|21 2 ≤ϕ( ), whe e A=(aij)i,j. Obse e ha in ou case, assuming (1.4), ϕ( ) has he ollowing o m − ˆB |A−A(0)|21 2 ≤− ˆB ¯ AB4(|x|)2dx1 2 ≤C1 nˆ 0 ¯ AB4(ρ)2ρn−1dρ1 2 =: ϕ( ), which is double-Dini con inuous since ϕ( )≤¯ AB4( ), and sa isfies max /2≤s≤ ϕ(s)≤μϕ( ), wi h μ>1. As in [12], by induc ion, one will find, o k≥0, wk∈H1(B3/4k+1 ) such ha  i,j Dj(aij(0)Diwk)=0 inB3/4k+1 ,(A.2) wkL2(B2/4k+1 )≤C4−k(n+2) 2ϕ(4−k),DwkL∞(B1/4k+1 ,Rn)Cϕ(4−k), (A.3) D2wkL∞(B1/4k+1 ,Rn×n)≤C4kϕ(4−k),(A.4) u− k  j=0 wjL2(B1/4k+1 )≤4−(k+1)(n+2) 2ϕ(4−(k+1)),(A.5) and wkL∞(B1/4k+1 )≤C4−kϕ(4−k),(A.6) see [12, (14), (15), (16), (17), and (18) o P oposi ion 1.1]. He e and in he sequel Cwill deno e a uni e sal cons an . 43 Page 20 o 23 D. A. L. Manna e al. NoDEA Fo x∈B1/4k+1 , using (A.3), (A.4), (A.6) and Taylo expansion, | k  j=0 wj(x)− ∞  j=0 wj(0) − ∞  j=0 Dwj(0) ·x| ≤ ∞  j=k+1 (|wj(0)|+|Dwj(0)||x|)+ k  j=0 D2wjL∞(B1/4k+1 ,Rn×n)|x|2 ≤C ∞  j=k+1 (4−jϕ(4−j)+ϕ(4−j)|x|)+C k  j=0 4jϕ(4−j)|x|2 ≤C4−(k+1) ˆ4−k 0 ϕ( ) d +C|x|2ˆ1 |x| 2 ϕ( ) 2d . (A.7) We hen de i e om (A.5)and heabo e,usingH¨olde ’s inequali y, ha ˆB1/4k+1 |u(x)− ∞  j=0 wj(0) − ∞  j=0 Dwj(0) ·x|dx ≤ k  j=0 wj(x)− ∞  j=0 wj(0) − ∞  j=0 Dwj(0) ·xL1(B1/4k+1 ) +u− k  j=0 wj(x)L1(B1/4k+1 ) ≤C4−(k+1)(n+1) ˆ4−k 0 ϕ( ) d +Cˆ1/4k+1 0 ρn+1 ˆ1 ρ 2 ϕ( ) 2d dρ +C4−(k+1)(n+1)ϕ(1/4k+1). (A.8) P oposi ion A.1 ollows om he abo e wi h a= ∞  j=0 wj(0), b= ∞  j=0 Dwj(0) ·x, and δ( )≃ˆ 0 ϕ(s) sds +1 n+1 ˆ 0 ρn+1 ˆ1 ρ 2 ϕ(s) s2ds dρ +ϕ( ), he symbol ≃s anding o = up o a cons an . I emains o p o e ha δ( ) is a Dini modulus o con inui y. Thanks o assump ion (1.4), i occu s i we show he Dini con inui y o he second e m in he p e ious sum. I yields 1 n+1 ˆ 0 ρn+1 ˆ1 ρ 2 ϕ(s) s2ds dρ ≤ ˆ1 2 ϕ(s) s2ds +ˆ 0 ϕ(ρ) ρdρ, so ha , in eg a ing by pa s, ˆ0ˆ1 2 ϕ(s) s2ds d = ˆ1 2 ϕ(s) s2ds0 +ˆ0 ϕ( 2) /4d . NoDEA On he egula i y o e y weak solu ions Page 21 o 23 43 I is easy o see ha lim →0 ´1 2 ϕ(s) s2ds = 0, and hus he hesis ollows by he Dini con inui y o ϕ. Appendix B: The coun e example To cons uc he example, one fi s conside s, o ∈(0,1) and o β>1, he unc ion α( )= −βn (n−1) log  0 +β(β+1) (n−1) log2 0 , o some 0>1. One akes hen A(x)=(aij(x))i,j defined by aij(x)=δij +α(|x|)δij −xixj |x|2, wi h 0la ge enough so ha α≥− 1 2,Abeing hen uni o mly ellip ic. Le us check now ha A∈W1,n(B1,Rn×n). Simple compu a ion gi es  ∂aij ∂x |α(|x|)|+|α(|x|)|1 |x|, o e e y i, j,  =1,...,n ( he symbol s and o ≤up o a cons an ). On he o he hand |α(|x|)|≃ 1 |x|log2 0 |x|+1 |x|log3 0 |x|, which in u n implies  ∂aij ∂x 1 |x|log  0 |x|+1 |x|log2 0 |x|+1 |x|log3 0 |x|1 |x|log  0 |x|, i 0is big enough. Thus, he belonging o A o W1,n(B1,Rn×n) is p o ided by he es ima e ˆB1 ∂aij ∂x n dx ˆB1 1 |x|nlogn 0 |x|dx ≃ˆ1 0 1 logn 0 d < +∞. Wi h such an A, which is con inuous bu no Dini-con inuous, in [10] he au ho s cons uc a solu ion o (1.1) (in he sense o dis ibu ions) u∈ W1,1 loc (B1) W1,p loc (B1) o e e y p>1. In pa icula , le us obse e ha such a solu ion belongs o Ln loc(B1). Re e ences [1] Ancona, A.: Ellip ic ope a o s, cono mal de i a i es and posi i e pa s o unc- ions (wi h an appendix by Ha¨ım B ezis). J. Func . Anal. 257, 2124–2158 (2009) [2] B ezis, H.: On a conjec u e o J. Se in. Rend. Lincei Ma . Appl. 19, 335–338 (2008) 43 Page 22 o 23 D. A. L. Manna e al. NoDEA [3] Chen, Y.Z., Wu, L.C.: Second O de Ellip ic Equa ions and Ellip ic Sys ems. T ansla ions o Ma hema ical Monog aphs, ol. 174. Ame ican Ma hema ical Socie y, P o idence (1998) [4] Chia enza, F., F asca, M., Longo, P.: W2,p- sol abili y o he Di ichle p oblem o non di e gence ellip ic equa ions wi h VMO coefficien s. T ans. Am. Ma h. Soc. 336, 841–853 (1993) [5] De Gio gi, E.: Sulla diffe enziabili `a e l’anali ici `a delle es emali degli in eg ali mul ipli egola i. Mem. Accad. Sci. To ino Cl. Sci. Fis. Ma . Na . 3, 25–43 (1957) [6] Dong, H., Escau iaza, L., Kim, S.: On C1,C2, and weak ype-(1,1) es ima es o linea ellip ic ope a o s, pa II. Ma h. Ann. 370, 447–489 (2018) [7] Di Fazio, G.: Lpes ima es o di e gence o m ellip ic equa ions wi h discon in- uous coefficien s. Boll. Un. Ma . I al. A (7) 10, 409–420 (1996) [8] Giaquin a, M., Ma inazzi, L.: An In oduc ion o he Regula i y Theo y o Ellip ic Sys ems. Ha monic Maps and Minimal G aphs. Edizioni della No male, Pisa (2012) [9] Hage , R.A., Ross, J.: A egula i y heo em o linea second o de ellip ic di e - gence equa ions. Ann. Sc. No m. Sup. Pisa 23, 283–290 (1971) [10] Jin, T., Maz’ya, V., Scha ingen, J.V.: Pa hological solu ions o ellip ic p oblems in di e gence o m wi h con inuous coefficien s. C. R. Acad. Sci. Pa is Se . I 347, 773–778 (2009) [11] Kuusi, T., Mingione, G.: New pe u ba ion me hods o nonlinea pa abolic p oblems. J. de Ma h. Pu es e Appl. 98(4), 390–427 (2012) [12] Li, Y.: On he C1 egula i y o solu ions o di e gence o m ellip ic sys ems wi h Dini-con inuous coefficien s. Chin. Ann. Ma h. Se . B 38(2), 489–496 (2017) [13] Mo ey, C.B.: Mul iple in eg als in he Calculus o Va ia ions. G undleh en de ma hema ischen Wissenscha en, ol. 130. Sp inge , New Yo k (1996) [14] Se in, J.: Pa hological solu ions o ellip ic diffe en ial equa ions. Ann. Sc. No m. Supe . Pisa (3) 18, 385–387 (1964) [15] Zhang, W., Bao, J.: Regula i y o e y weak solu ions o ellip ic equa ion o 601 di e gence o m. J. Func . Anal. 262, 1867–1878 (2012) NoDEA On he egula i y o e y weak solu ions Page 23 o 23 43 Domenico Angelo La Manna Depa men o Ma hema ics and S a is ics Uni e si y o Jy askyl¨a P.O. Box 35 (MaD)40014 Jy askyla Finland e-mail: [email p o ec ed] Chia a Leone and Robe a Schia a ella Dipa imen o di Ma ema ica e Applicazioni “R. Caccioppoli” Uni e si `a di Napoli Fede ico II Via Cin ia 80126 Naples I aly e-mail: c[email p o ec ed] Robe a Schia a ella e-mail: obe a.sc[email p o ec ed] Recei ed: 10 Ma ch 2020. Accep ed: 3 July 2020.