Citation: Luca Capogna, Giovanna Citti, Nicola Garofalo. Regularity for a class of quasilinear degenerate parabolic equations in the Heisenberg group[J]. Mathematics in Engineering, 2021, 3(1): 1-31. doi: 10.3934/mine.2021008
[1] | Mattia Fogagnolo, Andrea Pinamonti . Strict starshapedness of solutions to the horizontal p-Laplacian in the Heisenberg group. Mathematics in Engineering, 2021, 3(6): 1-15. doi: 10.3934/mine.2021046 |
[2] | Mikyoung Lee, Jihoon Ok . Local Calderón-Zygmund estimates for parabolic equations in weighted Lebesgue spaces. Mathematics in Engineering, 2023, 5(3): 1-20. doi: 10.3934/mine.2023062 |
[3] | Nicolai Krylov . On parabolic Adams's, the Chiarenza-Frasca theorems, and some other results related to parabolic Morrey spaces. Mathematics in Engineering, 2023, 5(2): 1-20. doi: 10.3934/mine.2023038 |
[4] | Marco Cirant, Kevin R. Payne . Comparison principles for viscosity solutions of elliptic branches of fully nonlinear equations independent of the gradient. Mathematics in Engineering, 2021, 3(4): 1-45. doi: 10.3934/mine.2021030 |
[5] | Fausto Ferrari, Nicolò Forcillo . A new glance to the Alt-Caffarelli-Friedman monotonicity formula. Mathematics in Engineering, 2020, 2(4): 657-679. doi: 10.3934/mine.2020030 |
[6] | Youchan Kim, Seungjin Ryu, Pilsoo Shin . Approximation of elliptic and parabolic equations with Dirichlet boundary conditions. Mathematics in Engineering, 2023, 5(4): 1-43. doi: 10.3934/mine.2023079 |
[7] | Patrizia Pucci, Letizia Temperini . On the concentration–compactness principle for Folland–Stein spaces and for fractional horizontal Sobolev spaces. Mathematics in Engineering, 2023, 5(1): 1-21. doi: 10.3934/mine.2023007 |
[8] | Boumediene Abdellaoui, Ireneo Peral, Ana Primo . A note on the Fujita exponent in fractional heat equation involving the Hardy potential. Mathematics in Engineering, 2020, 2(4): 639-656. doi: 10.3934/mine.2020029 |
[9] | Arthur. J. Vromans, Fons van de Ven, Adrian Muntean . Homogenization of a pseudo-parabolic system via a spatial-temporal decoupling: Upscaling and corrector estimates for perforated domains. Mathematics in Engineering, 2019, 1(3): 548-582. doi: 10.3934/mine.2019.3.548 |
[10] | David Cruz-Uribe, Michael Penrod, Scott Rodney . Poincaré inequalities and Neumann problems for the variable exponent setting. Mathematics in Engineering, 2022, 4(5): 1-22. doi: 10.3934/mine.2022036 |
Questa nota è dedicata a Sandro Salsa, con profondo affetto e ammirazione.
In this paper we establish the C∞ smoothness of solutions of a certain class of quasilinear parabolic equations in the Heisenberg group Hn (see [16,18,19,29] for a review of the literature on subelliptic, and degenerate parabolic PDE in the Heisenberg group). In a cylinder Q=Ω×(0,T), where Ω⊂Hn is an open set and T>0, we consider the equation
∂tu=2n∑i=1XiAi(x,∇0u) in Q=Ω×(0,T), | (1.1) |
modeled on the regularized parabolic p-Laplacian
∂tu=2n∑i=1Xi((1+|∇0u|2)p−22Xiu), | (1.2) |
where p≥2. The term regularized here refers to the fact that the non-linearity (1+|∇0u|2)p−22 affects the ellipticity of the right hand side only when the gradient blows up, and not when it vanishes, thus presenting a weaker version of the singularity in the p−Laplacian. Here, we indicate with x=(x1,...,x2n,x2n+1) the variable point in Hn. We alert the reader that, although it is customary to denote the variable x2n+1 in the center of the group with the letter t, we will be using z instead, since we have reserved the letter t for the time variable. Consequently, we will indicate with ∂i partial differentiation with respect to the variable xi, i=1,...,2n, and use the notation Z=∂z for the partial derivative ∂x2n+1. The notation ∇0u≅(X1u,...,X2nu) represents the so-called horizontal gradient of the function u, where
Xi=∂i−xn+i2∂z, Xn+i=∂n+i+xi2∂z, i=1,...,n. |
As it is well-known, the 2n+1 vector fields X1,...,X2n,Z are connected by the following commutation relation: for every couple of index i,j, if i≤j, then i≤n and [Xi,Xj]=δi+njZ, all other commutators being trivial.
We now introduce the relevant structural assumptions on the vector-valued function (x,ξ)→A(x,ξ)=(A1(x,ξ),...,A2n(x,ξ)): there exist p≥2, δ>0 and 0<λ≤Λ<∞ such that for a.e. x∈Ω,ξ∈R2n and for all η∈R2n, one has
{λ(δ+|ξ|2)p−22|η|2≤∂ξjAi(x,ξ)ηiηj≤Λ(δ+|ξ|2)p−22|η|2,|Ai(x,ξ)|+|∂xjAi(x,ξ)|≤Λ(δ+|ξ|2)p−12. | (1.3) |
Given an open set Ω⊂Hn, we indicate with W1,p(Ω) the Sobolev space associated with the p-energy EΩ,p(u)=1p∫Ω|∇0u|p, i.e., the space of all functions u∈Lp(Ω) such that their distributional derivatives Xiu, i=1,...,2n, are also in Lp(Ω). The corresponding norm is ||u||pW1,p(Ω)=||u||Lp(Ω)+||∇0u||Lp(Ω). We denote by W1,p0(Ω) the completion of C∞0(Ω) with respect to such norm. A function u∈Lp((0,T),W1,p0(Ω)) is a weak solution of (1.1) if
∫T0∫Ωuϕt dxdt−∫T0∫Ω2n∑i=1Ai(x,∇0u)Xiϕ dxdt=0, | (1.4) |
for every ϕ∈C∞0(Q). Our main result is the following.
Theorem 1.1. Let Ai satisfy the structure conditions (1.3) for some p≥2 and δ>0. We also assume that (1.1) can be approximated as in (1.6)-(1.8) below. Let u∈Lp((0,T),W1,p0(Ω)) be a weak solution of (1.1) in Q=Ω×(0,T). For any open ball B⊂⊂Ω and T>t2≥t1≥0, there exist constants C=C(n,p,λ,Λ,d(B,∂Ω),T−t2,δ)>0 and α=α(n,p,λ,Λ,d(B,∂Ω),T−t2,δ)∈(0,1) such that
||∇0u||Cα(B×(t1,t2))+||Zu||Cα(B×(t1,t2))≤C(∫T0∫Ω(δ+|∇0u|2)p2dxdt)1p. | (1.5) |
Besides the structural hypothesis (1.3), Theorem 1.1 will be established under an additional technical approximating assumption. Namely, for ε≥0 we consider the left-invariant Riemannian metric gε in Hn in which the frame defined by Xε1=X1,...,Xε2n=X2n,Xε2n+1=εZ is orthonormal, and denote by ∇ε the gradient in such metric. We will adopt the unconventional notation W1,p,ε(Ω) to indicate the Sobolev space associated with the p-energy EΩ,p,ε(u)=1p∫Ω|∇εu|p. We assume that one can approximate Ai by a 1-parameter family of regularized approximants Aε(x,ξ)=(Aε1(x,ξ),...,Aε2n+1(x,ξ)) defined for a.e. x∈Ω and every ξ∈R2n+1, and such that for a.e. x∈Ω, and for all ξ=(ξ1,...,ξ2n,ξ2n+1)∈R2n+1 one has uniformly on compact subsets of Ω×(0,T),
(Aε1(x,ξ),...,Aε2n+1(x,ξ)) ⟶ε→0+ (A1(x,ξ1,...,ξ2n),...,A2n(x,ξ1,...,ξ2n),0), | (1.6) |
and furthermore
{λ(δ+|ξ|2)p−22|η|2≤∂ξjAεi(x,ξ)ηiηj≤Λ(δ+|ξ|2)p−22|η|2,|Aεi(x,ξ)|+|∂xjAεi(x,ξ)|≤Λ(δ+|ξ|2)p−12, | (1.7) |
for all η∈R2n+1, and for some 0<λ≤Λ<∞ independent of ε. The proof of the C1,α regularity in Theorem 1.1 is based on a priori estimates for solutions of the one-parameter family of regularized partial differential equations which approximate (1.1) as the parameter ε→0. The key will be in establishing estimates that do not degenerate as ε→0. Specifically, for any ε>0 we will consider a weak solution uε to the equation
∂tuε=2n+1∑i=1XεiAεi(x,∇εuε) | (1.8) |
in a cylinder Q0=B(x0,R0)×(t0,t1), with B(x0,R0)⊂Ω and (t0,t1)⊂(0,T), and with (parabolic) boundary data uε=u. Since (1.8) is strongly parabolic for every ε>0, the solutions uε are smooth in every compact subset K⊂Q0 and, in view of the comparison principle, and of the uniform Harnack inequality established in [2], converge uniformly on compact subsets to a function u0. The bulk of the paper consists in establishing higher regularity estimates for uε that are uniform in ε>0, to show that u0 inherits such higher regularity and is a solution of (1.1), thus it coincides with u. Here is our main result in this direction.
Theorem 1.2. In the hypothesis (1.6), (1.7), consider for each ε>0 a weak solution uε∈Lp((0,T),W1,p,ε(Ω))∩C2(Q) of the approximating equation (1.8) in Q. For any open ball B⊂⊂Ω and T>t2≥t1≥0 there exists a constant C=C(n,p,λ,Λ,d(B,∂Ω),T−t2,δ)>0, such that
||∇εuε||pL∞(B×(t1,t2))+∫t2t1∫B(δ+|∇εuε|2)p−222n∑i,j=1|XεiXεjuε|2dxdt≤C∫T0∫Ω(δ+|∇εuε|2)p2dxdt. | (1.9) |
Moreover, for any open ball B⊂⊂Ω and T>t2≥t1≥0, there exist constants C>0 and α∈(0,1), which depend on n,p,λ,Λ,d(B,∂Ω),T−t2,δ, such that
||∇εuε||Cα(B×(t1,t2))+||Zuε||Cα(B×(t1,t2))≤C(∫T0∫Ω(δ+|∇εuε|2)p2dxdt)1p. | (1.10) |
We emphasise that the constants in (1.9) and (1.10) are independent of ε.
It is worth mentioning here that the prototype for the class of equations (1.1) and for their parabolic approximation comes from considering the regularized p−Laplacian operator Lpu=divg0,μ0((δ+|∇0u|2g0)p−22∇0u) in a sub-Riemannian contact manifold (M,ω,g0), where M is the underlying differentiable manifold, ω is the contact form and g0 is a Riemannian metric on the contact distribution. The measure μ0 is the corresponding Popp measure. The approximants are constructed through Darboux coordinates, considering the p−Laplacians associated to a family of Riemannian metrics gε that tame g0 and such that the metric structure of the spaces (M,gε) converge in the Gromov-Hausdorff sense to the metric structure of (M,ω,g0). For a more detailed description, see [8,Section 6.1]. As an immediate corollary of Theorem 1.1 one has the following.
Theorem 1.3. Let (M,ω,g0) be a contact, sub-Riemannian manifold and let Ω⊂M be an open set. For p≥2, consider u∈Lp((0,T),W1,p0(Ω)) be a weak solution of
∂tu=divg0,μ0((δ+|∇0u|2g0)p−22∇0u), |
in Q=Ω×(0,T). For any open ball B⊂⊂Ω and T>t2≥t1≥0, there exist constants C=C(n,p,d(B,∂Ω),T−t2,δ)>0 and α=α(n,p,d(B,∂Ω),T−t2,δ)∈(0,1) such that
||∇0u||Cα(B×(t1,t2))+||Zu||Cα(B×(t1,t2))≤C(∫T0∫Ω(δ+|∇0u|2)p2dxdt)1p. | (1.11) |
The C1,α estimates in (1.10) in Theorem 1.2 allow us to apply the Schauder theory developed in [5,30], and finally deduce the following result.
Theorem 1.4. Let k∈N and α∈(0,1). If Ai(x,ξ),∂xkAi(x,ξ),∂ξjAi(x,ξ)∈Ck,αloc satisfy the structure conditions (1.3) for some p≥2 and δ>0, then any weak solution u∈Lp((0,T),W1,p0(Ω)) is Ck+1,α on compact subsets of Q.
The present paper is the first study of higher regularity of weak solutions for the non stationary p-Laplacian type in the sub-Riemannian setting, and it is based on the techniques introduced by Zhong in [31]. The stationary case has been developed so far essentially only in the Heisenberg group case thanks to the work of Domokos, [14], Manfredi, Mingione [21], Mingione, Zatorska-Goldstein and Zhong [22], Ricciotti [26,27] and Zhong [31]. Regularity in more general contact sub-Riemannian manifolds, including the rototraslation group, has been recently established by the two of the authors and coauthors [8] and independently by Mukherjee [25] based on an extension of the techniques in [31]. Domokos and Manfredi [15] are rapidly making substantial progress in higher steps groups and in some special non-group structures, using the Riemannian approximation approach as in the work [8].
The plan of the paper is as follows. In Section 2 we collect some preparatory material that will be used in the main body of the paper. Section 3 is devoted to proving the first part of Theorem 1.2, which establishes the Lipschitz regularity of the approximating solutions uε. In Section 4 we prove the Hölder regularity of derivatives of uε in Theorem 1.2. Finally, in Section 5 we use the comparison principle and Theorem 1.2 to establish Theorem 1.1.
Some final comments are in order. The non-degeneracy hypothesis δ>0 in (1.3) (see also (1.7)) is not needed in the Euclidean setting and, in the stationary regime, it is not needed in the Heisenberg group either. We suspect the C1,α regularity of weak solutions for (1.1) still holds without this hypothesis, but at the moment we are unable to prove it. In this note we use δ>0 notably in (3.17) and in Theorem 3.13.
In order to extend the parabolic regularity theory to the sub-Riemannian setting one has to find a way to implement, in this non-Euclidean framework, some of the techniques introduced by Di Benedetto [13] which rely on non-isotropic cylinders in space-time. The key idea is to work with cylinders whose dimensions are suitably rescaled to reflect the degeneracy exhibited by the partial differential equation. To give an example, if one sets x∈Ω, R,μ>0, one can define the intrinsic cylinder
QR(μ):=B(x,R)×(−μ2−pR2,0), with supQR(μ)|∇0u|≤μ. |
In contrast with the usual parabolic cylinders of the linear theory, the shape of the QR(μ) cylinders is stretched in the time dimension by a factor of the order |∇0u|2−p.
The use of such non-isotropic cylinders seems necessary in order to make-up for the different homogeneity of the time derivative and the space derivatives in the degenerate regime δ=0. In a future study we plan to return to the problem of extending Di Benedetto's Caccioppoli inequalities on non-isotropic cylinders to the Heisenberg group and beyond.
In this section we collect a few definitions and preliminary results that will be used throughout the rest of the paper. As indicated in the introduction, for each ε∈(0,1) we define gε to be the Riemannian metric in Hn such that X1,...,X2n,εZ is an orthonormal frame, and denote such frame as Xε1,...,Xε2n+1. The corresponding gradient operator will be denoted by ∇ε.
Definition 2.1. For x0∈Ω⊂Hn, we define a parabolic cylinder Qε,r(x0,t0)⊂Q to be a set of the form Qε,r(x0,t0)=Bε(x0,r)×(t0−r2,t0). where r>0, Bε(x0,r)⊂Ω denotes the gε-Riemannian ball of center x0 and t0∈(0,T). We call parabolic boundary of the cylinder Qε,r(x0,t0)⊂Q the set Bε(x0,r)×{t0−r2}∪∂Bε(x0,r)×[t0−r2,t0).
First of all we recall the Hölder regularity, and local boundedness of weak solutions of (1.1) and (1.8) from [2].
Lemma 2.2. Let Q=Ω×(0,T)⊂Hn×R+, and δ≥0. For ε≥0 and p≥2, consider a weak solution uε∈Lp((0,T),W1,p,ε(Ω))∩C2(Q) of the approximating equation (1.8) in Q. For any open ball B⊂⊂Ω and T>t2≥t1≥0 there exist constants C=C(n,p,λ,Λ,d(B,∂Ω),T−t2)>0, and α=α(n,p,λ,Λ,d(B,∂Ω),T−t2)∈(0,1), such that
||uε||Cα(B×(t1,t2))≤C(∫T0∫Ω(δ+|∇εuε|2)p2dxdt)1p. | (2.1) |
When ε>0 and δ>0, classical regularity results (e.g., [20]) yield that weak solutions have bounded gradient, and hence (1.8) is strongly parabolic, thus leading to weak solutions being smooth. Clearly such smoothness may degenerate as ε→0, and the main point of this paper is to show that this does not happen.
Let Ω⊂Rn be a bounded open set and let Q=Ω×(0,T). For a function u:Q→R, and 1≤p,q we define the Lebesgue spaces Lp,q(Q)=Lq([0,T],Lp(Ω)), endowed with the norms
||u||Lp,q(Q)=(∫T0(∫Ω|u|pdx)qpdt)1q. | (2.2) |
When p=q, we will refer to Lp,p(Q) as Lp(Q). One has the following useful reformulation of the Sobolev embedding theorem [16] in terms of Lp,q spaces. In the next statement, we will indicate with N=2n+2 the homogenous dimension of Hn with respect to the non-isotropic group dilations, δλ(x)=(λx1,...,λ2z), and we will denote by
N1=N+2=2n+4 | (2.3) |
the corresponding parabolic dimension with respect to the dilations (x,t)→(δλx,λ2t).
Lemma 2.3. Let v be a Lipschitz function in Q, and assume that for all 0<t<T, v(⋅,t) has compact support in Ω×{t}.
(i) There exists C=C(n)>0 such that for any ε∈[0,1] one has
||v||L2NN−2,2(Q)≤C||∇εv||L2,2(Q). |
(ii) If v∈L2,∞(Q), then v∈L2N1N1−2,2N1N1−2(Q), and there exists C>0, depending on n, such that for any ε∈[0,1] one has
||v||2L2N1N1−2,2N1N1−2(Q)≤C(||v||2L2,∞(Q)+||∇εv||2L2,2(Q)). |
We note that as ε decreases to zero, the background geometry shifts from Riemannian to sub-Riemannian. The stability with respect to ε of the constant C in the Lemma 2.3 is not trivial, see [10,7].
In the sequel we will use an interpolation inequality that will take the place of the Sobolev inequality in a Moser type iteration, just as, for example, in [11,Proposition 4.2]. Although the result does not use the equation at all, we state it in terms that will make it immediately applicable later on. Henceforth, to simplify the notation, we will routinely omit the indication of dx, dxdt, etc. in all integrals involved, unless there is risk of confusion.
Lemma 2.4. Let uε be a weak solution of (1.8) in Q. If β≥0, and η∈C1([0,T],C∞0(Ω)) vanishes on the parabolic boundary of Q, then there is a constant C>0, depending only on ||uε||L∞(Q), such that
∫t2t1∫Ω(δ+|∇εuε|2)(β+p+2)/2|η|β+2≤C(β+p+1)2∫t2t1∫Ω(δ+|∇εuε|2)p+β−222n+1∑i,j=1|XεjXεiuε|2|η|β+2+Cβ2∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2|η|β(|η|2+|∇εη|2). |
Proof. Writing (δ+|∇εuε|2)(β+p+2)/2=(δ+|∇εuε|2)(β+p)/2(δ+|∇εuε|2), one has
∫t2t1∫Ω(δ+|∇εuε|2)(β+p+2)/2|η|β+2=δ∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2|η|β+2+2n+1∑i=1∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2XεiuεXεiuε|η|β+2=δ∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2|η|β+2−2n+1∑i=1∫t2t1∫ΩXεi((δ+|∇εuε|2)(β+p)/2Xεiuε)uε|η|β+2−(β+2)∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2Xεiuεuε|η|β+1Xεiη≤δ∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2|η|β+2+(β+p+1)∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/22n+1∑i,j=1|XεiXεjuε||uε||η|β+2+C(β+2)∫t2t1∫Ω(δ+|∇εuε|2)(β+p+1)/2|η|β+1|∇εη|. |
To conclude the argument, it suffices to apply Young's inequality.
In this section we establish Lipschitz regularity for the derivatives of the solutions uε. The main results of this section are summarized in the following estimates, which are unform in ε>0.
Theorem 3.1. Let Aεi satisfy the structure conditions (1.3) for some p≥2 and δ>0. Consider an open set Ω⊂Hn and T>0, and let uε be a weak solution of (1.8) in Q=Ω×(0,T). For any open ball B⊂⊂Ω and T>t2≥t1≥0, there exists a constant C>0, depending on n,p,λ,Λ,d(B,∂Ω),T−t2,δ, such that
||∇εuε||pL∞(B×(t1,t2))+∫t2t1∫B(δ+|∇εu|2)p−22(2n∑i,j=1|XεiXεjuε|2+|∇εZuε|)≤C∫T0∫Ω(δ+|∇εuε|2)p2. | (3.1) |
The proof of Theorem 3.1 will follow from combining the results in Theorem 3.11, Lemma 3.12, Proposition 3.13 and Proposition 4.1, that are all established later in the section. The Caccioppoli inequalities needed to prove Theorem 3.1 will take up most of the section, and they all apply to a solution uε of the approximating equation (1.8) in a cylinder Q=Ω×(0,T). We begin with two lemmas in which we explicitly detail the PDE satisfied by the smooth approximants Zuε and Xεℓuε.
Lemma 3.2. Let uε be a solution of (1.8) in Q. If we set vεℓ=Xεℓuε, with ℓ=1,...,2n+1, and sℓ=(−1)[ℓ/n] for ℓ≤2n, s2n+1=0, then the function vεℓ solves the equation
∂tvεℓ=2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεℓXεjuε)+2n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))+sℓZ(Aεℓ+sℓn(x,∇εuε)). | (3.2) |
Proof. Differentiating (1.8) with respect to Xεℓ, when ℓ≤n, we find
∂tvεℓ=2n+1∑i=1XεℓXεiAεi(x,∇εuε)=2n+1∑i=1Xεi(XεℓAεi(x,∇εuε))+2n+1∑i=1[Xεℓ,Xεi]Aεi(x,∇εuε)=2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεℓXεjuε)+2n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)−xℓ+n2Aεi,x2n+1(x,∇εuε))+Z(Aεℓ+n(x,∇εuε)). |
Taking the derivative with respect to Xεℓ when n+1≤ℓ≤2n, we obtain
∂tvεℓ=2n+1∑i=1XεℓXεiAεi(x,∇εuε)=2n+1∑i=1Xεi(XεℓAεi(x,∇εuε))+2n+1∑i=1[Xεℓ,Xεi]Aεi(x,∇εuε)=2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεlXεjuε)+2n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)+xℓ−n2Aεi,x2n+1(x,∇εuε))−Z(Aεℓ−n(x,∇εuε)). |
Since for ℓ=2n+1 the vector field Xεℓ commutes with the others, taking the derivatives with respect to Xε2n+1 we obtain the thesis.
Lemma 3.3. Let uε be a solution of (1.8) in Q. Then, the function Zuε is a solution of the equation
∂tZuε=2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεjZuε)+2n+1∑i=1Xεi(Aεi,x2n+1(x,∇εuε)). |
Proof. The assertion immediately follows from Lemma 3.2, with ℓ=2n+1, since Xε2n+1=εZ.
Lemma 3.4. Let uε be a solution of (1.8) in Q. For any β≥0 and for all η∈C1([0,T],C∞0(Ω)), one has
1β+2∫Ω|Zuε|β+2η2|t2t1+λ(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β|η|2,≤Λ(β+1)(16Λλ+2)∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εη|2|Zuε|β+2+2β+2∫t2t1∫Ω|Zuε|β+2η∂tη+Λ(β+1)(16Λλ+2)∫t2t1∫Ω(δ+|∇εuε|2)p2η2|Zuε|β. |
Proof. We use ϕ=η2|Zuε|βZuε as a test function in the equation satisfied by Zuε, see Lemma 3.3, to obtain
∫t2t1∫Ω∂tZuεη2|Zuε|βZuε=∫t2t1∫Ω2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεjZuε)η2|Zuε|βZuε+∫t2t1∫Ω2n+1∑i,j=1Xεi(Aεi,x2n+1(x,∇εuε))η2|Zuε|βZuε. |
The left-hand side of the latter equation can be expressed as follows:
∫t2t1∫Ω∂tZuεη2|Zuε|βZuε=1β+2∫t2t1∫Ω∂t|Zuε|β+2η2. |
Considering the first term in the right-hand side, we obtain
∫t2t1∫Ω2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεjZuε)η2|Zuε|βZuε=−∫t2t1∫Ω2n+1∑i,j=1Aεi,ξj(x,∇εuε)XεjZuεXεi(η2|Zuε|βZuε)=−2∫t2t1∫Ω2n+1∑i,j=1Aεi,ξj(x,∇εuε)XεjZuεηXεiη|Zuε|βZuε−(β+1)∫t2t1∫Ω2n+1∑i,j=1Aεi,ξj(x,∇εuε)XεjZuεη2|Zuε|βXεiZuε. |
As for the second term in the right-hand side, we have
∫t2t1∫Ω2n+1∑i,j=1Xεi(Aεi,x2n+1(x,∇εuε))η2|Zuε|βZuε=−2∫t2t1∫Ω2n+1∑i,j=1Aεi,x2n+1(x,∇εuε)ηXεiη|Zuε|βZuε−(β+1)∫t2t1∫Ω2n+1∑i,j=1Aεi,x2n+1(x,∇εuε)η2|Zuε|βXεiZuε. |
Combining the latter three equations, we find
1β+2∫t2t1∫Ω∂t|Zuε|β+2η2+(β+1)∫t2t1∫Ω2n+1∑i,j=1∂ξjAεi(x,∇εuε)XεjZuεη2|Zuε|βXεiZuε=−2∫t2t1∫Ω2n+1∑i,j=1∂ξjAεi(x,∇εuε)XεjZuεXεiηη|Zuε|βZuε−2∫t2t1∫Ω2n+1∑i,j=1Aεi,x2n+1(x,∇εuε)ηXεiη|Zuε|βZuε−(β+1)∫t2t1∫Ω2n+1∑i,j=1Aεi,x2n+1(x,∇εuε)η2|Zuε|βXεiZuε. |
The structure conditions (1.3) yield
1β+2∫Ω|Zuε|β+2η2|t2t1+λ(β+1)∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β|η|2≤1β+2∫Ω|Zuε|β+2η2|t2t1+(β+1)∫t2t1∫Ω2n+1∑i,j=1∂ξjAεi(x,∇εuε)XεjZuεXεiZuεη2|Zuε|β=−2∫t2t1∫Ω2n+1∑i,j=1∂ξjAεi(x,∇εuε)XεjZuεXεiηη|Zuε|βZuε+2β+2∫t2t1∫Ω|Zuε|β+2η∂tη−2∫t2t1∫Ω2n+1∑i=1Aεi,x2n+1(x,∇εuε)ηXεiη|Zuε|βZuε−(β+1)∫t2t1∫Ω2n+1∑i=1Aεi,x2n+1(x,∇εuε)η2|Zuε|βXεiZuε≤2Λ∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZu|η|∇εη||Zuε|β+1+2β+2∫t2t1∫Ω|Zuε|β+2η∂tη+2Λ∫t2t1∫Ω(δ+|∇εuε|2)p−12η|∇εη||Zuε|β+1+(β+1)Λ∫t2t1∫Ω(δ+|∇εuε|2)p−12η2|Zuε|β|∇εZuε|, |
thus concluding the proof.
Lemma 3.5. Let uε be a weak solution of (1.8) in Q. There exists C0=C0(n,p,λ,Λ)>0. For any t2≥t1≥0, β≥0 and all η∈C∞0(Ω), we have
1β+2∫Ωη2[(δ+|∇εuε|2)(β+2)/2]|t2t1+∫t2t1∫Ωη2(δ+|∇εuε|2)(p−2+β)/22n+1∑i,j=1|XεiXεjuε|2≤C0∫t2t1∫Ω(η2+|∇εη|2+η|Zη|)(δ+|∇εuε|2)(p+β)/2+C0(β+1)4∫t2t1∫Ωη2(δ+|∇εuε|2)(p+β−2)/2|Zuε|2. | (3.3) |
Proof. In view of Lemma 3.2 we know that, if uε∈C∞(Q) is a solution of ∂tuε=∑2n+1i=1XεiAεi(x,∇εuε), then vεℓ=Xεℓuε solves (3.2). If in the first term in the right-hand side of (3.2) we use the fact that XεℓXεjuε=XεjXεℓuε+[Xεℓ,Xεj]uε=Xεjvεℓ+sℓZvεℓ, we find
∂tvεℓ=2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)Xεjvεℓ)+sℓ2n+1∑i=1Xεi(Aεi,ξℓ+sℓn(x,∇εuε)Zuε)+ | (3.4) |
+2n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))+sℓZ(Aεℓ+sℓn(x,∇εuε)). |
Fix η∈C∞0(Ω) and let ϕ=η2(δ+|∇εuε|2)β/2Xεℓuε. Taking such ϕ as the test-function in the weak form of (3.4), and integrating by parts the terms in divergence form, one has
12∫t2t1∫Ω(δ+|∇εuε|2)β2∂t[Xεℓuε]2η2+2n+1∑i,j=1∫t2t1∫ΩAεiξj(x,∇εuε)XεjvεℓXεi(η2(δ+|∇εuε|2)β/2Xεℓuε)=−sℓ2n+1∑i=1∫t2t1∫ΩAεiξℓ+sℓn(x,∇εuε)ZuεXεi(η2(δ+|∇εuε|2)β/2Xεℓuε)+∫t2t1∫Ω2n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)−slxℓ+sℓn2Aεi,x2n+1(x,∇εuε))η2(δ+|∇εuε|2)β/2Xεℓuε+sℓ∫t2t1∫ΩslZ(Aεℓ+sℓn(x,∇εuε))η2(δ+|∇εuε|2)β/2Xεℓuε. |
The latter equation implies that for every ℓ=1,...,2n+1 one has
12(β+2)∫t2t1∫Ω(δ+|∇εuε|2)β2∂t[Xεℓuε]2η2+2n+1∑i,j=1∫t2t1∫ΩAεiξj(x,∇εuε)XεjXεℓuεXεiXεℓuε η2(δ+|∇εuε|2)β/2+2n+1∑i,j=1β2∫t2t1∫ΩAεiξj(x,∇εuε)XεjXεℓuεXεℓuεXεi(|∇εuε|2) η2(δ+|∇εuε|2)β−22=−2n+1∑i,j=1∫t2t1∫ΩAεiξj(x,∇εuε)XεjXεℓuεXεℓuεXεi(η2)(δ+|∇εu|2)β/2−sℓ2n+1∑i=1∫t2t1∫ΩAεiξℓ+sℓn(x,∇εuε)ZuXεi(η2(δ+|∇εuε|2)β/2Xεℓuε)−∫t2t1∫Ω2n+1∑i=1(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))Xεi(η2(δ+|∇εuε|2)β/2Xεℓuε)+sℓ2n+1∑j=1∫t2t1∫ΩZ(Aεℓ+sℓn(x,∇εuε))η2(δ+|∇εuε|2)β/2Xεℓuε=I1ℓ+I2ℓ+I3ℓ+I4ℓ. |
Summing over ℓ=1,...,2n+1, by a simple application of the chain rule, and using the structural assumption (1.7), we see that the left-hand side can be bounded from below by
1β+2∫t2t1∫Ω∂t[(δ+|∇εuε|2)β2+1]η2+2n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ωη2Aεiξj(x,∇εuε)XεjXεℓuεXεiXεℓuε(δ+|∇εuε|2)β/2+2n+1∑ℓ=12n+1∑i,j=1β2∫t2t1∫Ωη2Aεiξj(x,∇εuε)XεjXεℓuεXεℓuεXεi(|∇εuε|2)(δ+|∇εuε|2)β−22≥1β+2∫t2t1∫Ω∂t[(δ+|∇εuε|2)β2+1]η2+λ∫t2t1∫Ωη2(δ+|∇εuε|2)p−2+β22n+1∑i,j=1|XεiXεjuε|2+λβ4∫t2t1∫Ωη2(δ+|∇εuε|2)p+β−42|∇ε(|∇εuε|2)|2. |
Since the last term in the right-hand side of this estimate is nonnegative, we obtain from this bound
1β+2∫Ω[(δ+|∇εuε|2)β2+1η2]|t2t1+λ∫t2t1∫Ωη2(δ+|∇εuε|2)p−2+β22n+1∑i,j=1|XεiXεjuε|2≤2n+1∑ℓ=1(I1ℓ+I2ℓ+I3ℓ+I4ℓ). | (3.5) |
Next, we estimate each of the terms in the right-hand side separately. Recalling that from (1.7) one has |Aiξj(x,η)|=|∂ξjAεi(x,η)|≤C(δ+|η|2)p−22, one has that for any α>0 there exists Cα>0 depending only on α,p,n and the structure constants, such that
2n+1∑ℓ=1I1ℓ=−2n+1∑ℓ=12n+1∑i,j=1∫t2t1∫ΩAεiξj(x,∇εuε)XεjXεℓuεXεℓuεXεi(η2)(δ+|∇εuε|2)β/2≤22n+1∑i,j=1∫t2t1∫Ω|η|(δ+|∇εuε|2)(p−2)/2|XεjXεiuε||∇εuε||∇εη|(δ+|∇εuε|2)β2≤α2n+1∑i,j=1∫t2t1∫Ωη2(δ+|∇εuε|2)(p+β−2)/2|XεjXεiuε|2+Cα∫t2t1∫Ω(δ+|∇εuε|2)(p+β)/2|∇εη|2. | (3.6) |
Analogously, we find
2n+1∑ℓ=1I2ℓ≤α2n+1∑i,j=1∫t2t1∫Ωη2(δ+|∇εuε|2)(p+β−2)/2|XεiXεjuε|2+C∫t2t1∫Ω(δ+|∇εuε|2)(p+β)/2|∇εη|2+Cα(β+1)2∫t2t1∫Ωη2(δ+|∇εuε|2)(p+β−2)/2|Zuε|2. | (3.7) |
In a similar fashion, we obtain
2n∑ℓ=1I3ℓ≤α2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)(p+β−2)/2|XεiXεjuε|2η2+Cα(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)(p+β)/2(|∇εη|2+|η|2). | (3.8) |
Finally, integrating by parts twice, and using the structural assumptions, one has
2n+1∑ℓ=1I4ℓ=−2n+1∑ℓ=1∫t2t1∫ΩZ(Aℓ+sℓn(x,∇εuε))η2(δ+|∇εuε|2)β/2Xεℓuε=22n+1∑ℓ=1∫t2t1∫ΩAℓ+sℓn(x,∇εuε)ηZη(δ+|∇εuε|2)β/2Xεℓuε+β2n+1∑ℓ=12n+1∑j=1∫t2t1∫ΩAℓ+sℓn(x,∇εuε)η2(δ+|∇εuε|2)β−22XjuεXjZuεXεℓuε+2n+1∑ℓ=1∫t2t1∫ΩAℓ+sℓn(x,∇εuε)η2(δ+|∇εuε|2)β/2XεℓZuε=22n+1∑ℓ=1∫t2t1∫ΩAℓ+sℓn(x,∇εuε)ηZη(δ+|∇εuε|2)β/2Xεℓuε−β2n+1∑ℓ=12n+1∑j=1∫t2t1∫ΩXj(Aℓ+sℓn(x,∇εuε)η2(δ+|∇εuε|2)β−22XjuεXεℓuε)Zuε−2n+1∑ℓ=1∫t2t1∫ΩXεℓ(Aℓ+sℓn(x,∇εuε)η2(δ+|∇εuε|2)β/2)Zuε≤α2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)(p+β−2)/2|XεiXεjuε|2η2+C(β+1)∫t2t1∫Ω(δ+|∇εuε|2)(p+β)/2(η2+|∇εη|2+|ηZη|)+Cα(β+1)4∫t2t1∫Ω(δ+|∇εuε|2)(p+β−2)/2|Zuε|2. | (3.9) |
Combining (3.6)–(3.9) with (3.5), we reach the desired conclusion (3.3).
In the case β=0 we obtain the following stronger estimate, which we will need in the sequel. We denote by ||⋅|| the L∞ norm of a function on the parabolic cylinder Q.
Lemma 3.6. Let uε be a weak solution of (1.8) in Q, let t2≥t1≥0, and η∈C1([0,T],C∞0(Ω)) be such that 0≤η≤1, and for which ||∂tη||≤C||∇εη||2, where C>0 is a universal constant. For every α>0 there exists Cα>0 such that
12∫Ω((δ+|∇εuε|2)η2)|t2t1+λ∫t2t1∫Ω(δ+|∇εuε|2)p−222n+1∑i,j=1|XεiXεjuε|2η2≤α∫t2t1∫Ω|Zuε|2η3+Cα∫t2t1∫Ω(δ+|∇εuε|2)p/2(η2+|∇εη|2+|ηZη|). |
Proof. In view of Lemma 3.2 we notice that, if uε∈C∞(Q) is a solution of ∂tuε=∑2n+1i=1XεiAεi(x,∇εuε), then vεℓ=Xεℓuε solves
∂tvεℓ=2n+1∑i,j=1Xεi(Xεℓ(Aεi(x,∇εuε)))+2n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))+sℓZ(Aεℓ+sℓn(x,∇εuε)). | (3.10) |
With η as in the statement of the lemma, we take ϕ=η2Xεℓuε as a test function in the weak form of (3.10). Integrating by parts the terms in divergence form, one has
12∫t2t1∫Ωη2∂t(Xεℓuε)2+2n+1∑i=1∫t2t1∫ΩXℓ(Aεi(x,∇εuε))Xεi(η2Xεℓuε)=∫t2t1∫Ωη22n+1∑i=1Xεi(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))Xεℓuε+sℓ∫t2t1∫Ωη2Z(Aεℓ+sℓn(x,∇εuε))Xεℓuε. |
This gives
12∫t2t1∫Ωη2∂t(Xεℓuε)2+2n+1∑i,j=1∫t2t1∫Ωη2Aεi,ξj(x,∇εuε)XεℓXεjuεXεℓXεiuε=−2n+1∑i,j=1∫t2t1∫Ωη2Xℓ(Aεi(x,∇εuε))Zuε−2n+1∑i,j=1∫t2t1∫ΩXℓ(Aεi(x,∇εuε))ηXεiηXεℓuε−∫t2t1∫Ω2n+1∑i=1(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))Xεi(η2Xεℓuε)+sℓ2n+1∑j=1∫t2t1∫Ωη2Z(Aεℓ+sℓn(x,∇εuε))Xεℓuε=I1ℓ+I2ℓ+I3ℓ+I4ℓ. |
Summing over ℓ=1,...,2n+1, in view of the structural hypothesis (1.7), after an integration by parts in the first term in the left-hand side we obtain the following bound
12∫Ω(δ+|∇εuε|2)η2|t2t1+λ∫t2t1∫Ω(δ+|∇εuε|2)p−222n+1∑i,j=1|XεiXεjuε|2η2≤I1ℓ+I2ℓ+I3ℓ+I4ℓ+∫t2t1∫Ω(δ+|∇εuε|2)η∂tη. |
Next, we estimate each of the terms in the right-hand side separately. Recalling that |Aεiξj(x,η)|≤C(δ+|η|2)p−22, we find that for any α1,α2>0 there exist Cα1,Cα2>0, depending only on α1,α2,p,n and the structure constants, such that
2n+1∑ℓ=1I1ℓ=2n∑ℓ=12n+1∑i,j=1∫t2t1∫Ωη2Xℓ(Aεi(x,∇εuε))Zuε=−22n+1∑ℓ=12n+1∑i=1∫t2t1∫ΩAεi(x,∇εuε)ηXℓηZuε−2n+1∑ℓ=12n+1∑i=1∫t2t1∫Ωη2Aεi(x,∇εuε)XεℓZuε=−22n+1∑ℓ=12n+1∑i=1∫t2t1∫ΩAεi(x,∇εuε)ηXℓηZuε+2n+1∑ℓ=12n+1∑i=1∫t2t1∫ΩAεi(x,∇εuε)2ηZηXεℓuε+2n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ωη2Aεiξj(x,∇εuε)XεjZuεXεℓuε≤∫t2t1∫Ω(δ+|∇εuε|2)(p−1)/2η|∇εη||Zuε|+2n+1∑ℓ=12n+1∑i=1∫t2t1∫Ω(δ+|∇εuε|2)p/2|2ηZη|+2n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ωη2(δ+|∇εuε|2)(p−1)/2|∇εZuε|≤α1∫t2t1∫Ω(δ+|∇εuε|2)(p−2)/2η2|Zuε|2+Cα1∫t2t1∫Ω(δ+|∇εuε|2)p/2|∇εη|2+2n+1∑ℓ=12n+1∑i=1∫t2t1∫Ω(δ+|∇εuε|2)p/2|2ηZη|+α2||∇εη||22n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ωη4(δ+|∇εuε|2)(p−2)/2|∇εZuε|2+Cα2||∇εη||2∫t2t1∫supp(η)(δ+|∇εuε|2)p/2. |
Now, we apply Lemma 3.4 to find, for any α>0,
α||∇εη||22n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)(p−2)/2|∇εZuε|2η4≤αC∫t2t1∫Ω(δ+|∇εuε|2)p−22|Zuε|2η2+α||∂tη||||∇εη||2∫t2t1∫Ω|Zuε|2η3+α∫t2t1∫Ω(δ+|∇εuε|2)p2η4. |
Analogously,
2n+1∑ℓ=1I2ℓ+2n+1∑ℓ=1I3ℓ≤α2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)(p−2)/2|XεiXεjuε|2η2+C∫t2t1∫Ω(δ+|∇εuε|2)p/2|∇εη|2. |
Using the structure conditions, one has
2n+1∑ℓ=1I4ℓ≤2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)(p−1)/2|∇εZuε|η2≤α2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)(p−2)/2|∇εZuε|2η2+Cα∫t2t1∫Ω(δ+|∇εuε|2)p/2(η2+|∇εη|2+|ηZη|), |
thus concluding the proof.
Next, we need to establish mixed type Caccioppoli inequalities, where the left-hand side includes terms with both horizontal derivatives and derivatives along the second layer of the stratified Lie algebra of Hn.
Lemma 3.7. Set T>t2>t1>0. Let uε be a weak solution of (1.8) in Q=Ω×(0,T). Let β≥2 and let η∈C1((0,T),C∞0(Ω)), with 0≤η≤1. For all α≤1 there exist constants CΛ, Cα=C(α,λ,Λ)>0 such that
∫t2t1∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β2n+1∑i,j=1|XεiXεjuε|2+∫Ωηβ+2|Zuε|β|∇εuε|2|t2t1+(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2ηβ+2|∇εuε|2≤Cα(β+1)2(1+|∇εη||2L∞)∫t2t1∫Ω(ηβ+ηβ+4)(δ+|∇εuε|2)p2|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2+2α(1+||∇εη||2)(β+2)∫t2t1∫Ω|Zuε|β+2ηβ+3|∂tη|dx+α(β+2)2∫Ω|Zuε|β+2ηβ+4|t=t1+CΛ(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p+22|Zuε|β−2ηβ+2+∫t2t1∫Ω|Zuε|β|∇εuε|2∂t(ηβ+2). | (3.11) |
Proof. Let η∈C∞0(Ω×(0,T)) be a nonnegative cutoff function. Fix β≥2 and ℓ∈{1,...,2n}. Note that
∂t(|Xεℓuε|2|Zuε|β)=2Xεℓuε∂tXεℓuε|Zuε|β+β|Xεℓuε|2|Zuε|β−2Zuε∂tZuε, |
which suggests to use 2Xεℓuε|Zuε|β as a test function in the Eq (3.2) satisfied by Xεℓuε and to choose β|Xεℓuε|2|Zuε|β−2Zuε as a test function in the Eq (3.3) satisfied by Zuε. Eq (3.2) becomes in weak form
∫t2t1∫Ω∂tXεℓuεϕ=−2n+1∑i,j=1∫t2t1∫Ω(Aεi,ξj(x,∇εuε)XεℓXεjuε)Xεiϕ+sℓZ(Aεℓ+sℓn(x,∇εuε))ϕ−2n+1∑i=1∫t2t1∫Ω(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))Xεiϕ. |
Consequently, if we substitute the test function ϕ=2ηβ+2|Zu|βXεℓu, we obtain
2∫t2t1∫Ω∂tXεℓuεηβ+2|Zuε|βXεℓuε+22n+1∑i,j=1∫t2t1∫ΩAεi,ξj(x,∇εuε)XεℓXεjuεηβ+2|Zuε|βXεℓXεiuε=−2n+1∑i,j=12∫t2t1∫ΩAεi,ξj(x,∇εuε)XεℓXεjuεXεi(ηβ+2|Zuε|β)Xεℓuε−22n+1∑i,j=1∫t2t1∫ΩAεi,ξj(x,∇εuε)XεℓXεjuεηβ+2|Zuε|β[Xεi,Xεℓ]uε−2sl∫t2t1∫ΩZ(Aℓ+sℓn(x,∇εuε))ηβ+2|Zuε|βXεℓuε−22n+1∑i=1∫t2t1∫Ω(Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε))Xεi(ηβ+2|Zuε|βXεℓuε)=I1ℓ+I2ℓ+I3ℓ+I4ℓ. | (3.12) |
We will show that these terms satisfy the following estimate
4∑k=12n+1∑ℓ=1|Ikℓ|≤α∫t2t1∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β2n+1∑i,j=1|XεiXεjuε|2+Cα(β+1)2(1+||∇εη||2L∞∫t2t1∫Ω(ηβ+ηβ+4)(δ+|∇εuε|2)p2|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2+2α(1+||∇εη||2)(β+2)∫t2t1∫Ω|Zuε|β+2ηβ+3|∂tη|+α(β+2)2∫Ω|Zuε|β+2ηβ+4|t=t1+α(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22ηβ+4|Zuε|β−2|∇εZuε|2|∇εuε|2. | (3.13) |
We first note that
2n+1∑ℓ=1|I1ℓ|≤22n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ω|Aεi,ξj(x,∇εuε)XεℓXεjuεXi(ηβ+2|Zuε|β)Xεℓuε|≤2nΛ(β+2)2n+1∑ℓ=12n+1∑j=1∫t2t1∫Ω(δ+|∇εuε|2)p−12|XεℓXεjuε|ηβ+1|∇εη||∇εuε||Zuε|β+2nβ2n+1∑ℓ=12n+1∑j=1∫t2t1∫Ω(δ+|∇εuε|2)p−12|XεℓXεjuε|ηβ+2|Zuε|β−1|∇εZuε|≤α∫t2t1∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β2n+1∑i,j=1|XεiXεjuε|2+Cα(β+1)2∫t2t1∫Ωηβ|∇εη|2(δ+|∇εuε|2)p2|Zuε|β+Cα(β+1)2(1+||∇εη||2)∫t2t1∫Ωηβ(δ+|∇εuε|2)p2|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2+α1+||∇εη||2∫t2t1∫Ωηβ+4(δ+|∇εuε|2)p−22|Zuε|β|∇εZuε|2. |
The last term can be estimated, as follows, using Lemma 3.4:
α∫t2t1∫Ωηβ+4(δ+|∇εuε|2)p−22|Zuε|β|∇εZuε|2≤αCΛ,λ∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εη|2ηβ+2|Zuε|β+2+2αβ+2∫t2t1∫Ω|Zuε|β+2ηβ+3∂tη+α(β+1)2∫Ω|Zuε|β+2ηβ+4|t=t1+αCΛ,λ∫t2t1∫Ω(δ+|∇εuε|2)p2ηβ+4|Zuε|β≤αCΛ,λ∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εη|2ηβ+2|Zuε|β∑ij|XiXju|2+2αβ+2∫t2t1∫Ω|Zuε|β+2ηβ+3∂tη+α(β+1)2∫Ω|Zuε|β+2ηβ+4|t=t1+αCΛ,λ∫t2t1∫Ω(δ+|∇εuε|2)p2ηβ+4|Zuε|β. | (3.14) |
From here estimate (3.13) holds. Integrating by parts we have
2n+1∑ℓ=1|I2ℓ|=−22n+1∑ℓ=12n+1∑i,j=1∫t2t1∫Ω|Aεi(x,∇εuε)Xεl(ηβ+2|Zuε|β[Xεi,Xεℓ]uε)|≤2(β+2)Λ∫t2t1∫Ω(δ+|∇εuε|2)p−12ηβ+1|∇εη||Zuε|β+1++2(β+1)∫t2t1∫Ω(δ+|∇εuε|2)p−12ηβ+2|Zuε|β|∇εZuε|≤Cα(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p2ηβ|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2+α∫t2t1∫Ω(δ+|∇εuε|2)p−22ηβ+2|Zuε|β2n+1∑i,j=1|XεiXεjuε|2+α∫t2t1∫Ω(δ+|∇εuε|2)p−22ηβ+4|Zuε|β|∇εZuε|2+Cα(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p2ηβ|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2. |
From here, using inequality (3.14), we deduce that I2ℓ satisfies inequality (3.13). The estimate of I3ℓ can be made as follows:
|I3ℓ|≤α∫t2t1∫Ω(δ+|∇εuε|2)p−22ηβ+4|Zuε|β|∇εZuε|2+Cα∫t2t1∫Ω(δ+|∇εuε|2)p2ηβ|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2 |
From here and (3.14) the inequality (3.13) follows. The estimate of I4ℓ is analogous:
|I4ℓ|≤2(β+1)Λ∫t2t1∫Ω(δ+|∇εuε|2)p−12ηβ+1|∇εη||Zuε|β|∇εuε|+2(β+1)Λ∫t2t1∫Ω(δ+|∇εuε|2)p−12ηβ+2|Zuε|β−1|∇εZuε||∇εuε|+Λ2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)p−12ηβ+2|Zuε|β|XεiXεjuε|≤α(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22ηβ+4|Zuε|β−2|∇εZuε|2|∇εuε|2+α2n+1∑i,j=1∫t2t1∫Ω(δ+|∇εuε|2)p−22ηβ+2|Zuε|β|XεiXεjuε|2+Cα(β+1)(1+||∇εη||2L∞)∫t2t1∫Ω(δ+|∇εuε|2)p2|Zuε|β(ηβ+ηβ+2). |
We now recall the following pde from Lemma 3.3
∂tZuε=2n+1∑i,j=1Xεi(Aεi,ξj(x,∇εuε)XεjZuε)+2n+1∑i=1Xεi(Aεi,x2n+1(x,∇εuε)). |
Substituting in this equation the test function ϕ=β|Zu|β−2Zuηβ+2|∇εuε|2, one obtains
β∫t2t1∫Ω∂tZuε|Zuε|β−2Zuεηβ+2|∇εuε|2+β(β−1)2n+1∑i,j=1∫t2t1∫ΩAiξj(x,∇εuε)XεjZuεXεiZuε|Zuε|β−2ηβ+2|∇εuε|2=−β2n+1∑i,j=1∫t2t1∫ΩAiξj(x,∇εuε)XεjZuε|Zuε|β−2ZuεXεi(ηβ+2|∇εuε|2)−β2n+1∑i=1∫t2t1∫ΩAεi,x2n+1(x,∇εuε)Xεi(|Zuε|β−2Zuεηβ+2|∇εuε|2)=−β(β+2)2n+1∑i,j=1∫t2t1∫ΩAiξj(x,∇εuε)XεjZuε|Zuε|β−2ZuεXεiηηβ+1|∇εuε|2−2β2n+1∑ℓ,i,j=1∫t2t1∫ΩAiξj(x,∇εuε)XεjZuε|Zuε|β−2Zuεηβ+2XεℓuεXεiXεℓuε−β(β−1)2n+1∑i=1∫t2t1∫ΩAεi,x2n+1(x,∇εuε)|Zuε|β−2XεiZuεηβ+2|∇εuε|2−β(β+1)2n+1∑i=1∫t2t1∫ΩAεi,x2n+1(x,∇εuε)|Zuε|β−2Zuεηβ+1Xεiη|∇εuε|2−β2n+1∑i,ℓ=1∫t2t1∫ΩAεi,x2n+1(x,∇εuε)|Zuε|β−2Zuεηβ+2XεℓuεXεiXεℓuε=I5+⋯+I9. | (3.15) |
We observe that the ellipticity condition yields
β(β−1)2n+1∑i,j=1∫t2t1∫ΩAiξj(∇εuε)XεjZuXεiZuε|Zuε|β−2ηβ+2|∇εuε|2≥(β+1)2Cλ∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2ηβ+2|∇εuε|2. |
Let us now consider I5:
I5=−β(β+2)2n+1∑i,j=1∫t2t1∫ΩAiξj(x,∇εuε)XεjZuε|Zuε|β−2ZuεXεiηηβ+1|∇εuε|2≤2(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε||Zuε|β−1|∇εη|ηβ+1|∇εuε|2≤α(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2ηβ+2|∇εuε|2+Cα(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p2|Zuε|β|∇εη|ηβ. |
The estimate of I6 is identical to that I1ℓ and we thus omit it. Let us consider I7. One has
I7≤(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−12|Zuε|β−2|∇εZuε|ηβ+2|∇εuε|2≤α(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22|Zuε|β−2|∇εZuε|2ηβ+2|∇εuε|2+Cα(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p2|Zuε|β−2ηβ+2|∇εuε|2. |
Similar consideration holds for I8
I8≤(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−12|Zuε|β−1ηβ+1|∇εη||∇εuε|2≤CΛ(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p2|Zuε|βηβ|∇εη|2+CΛ(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p+22|Zuε|β−2ηβ+2. |
Finally, we estimate I9.
I9≤CΛ(β+1)2n+1∑ℓ,i=1∫t2t1∫Ω(δ+|∇εuε|2)p−12|Zuε|β−1ηβ+2|∇εuε||XεiXεℓuε|≤CΛ(β+1)2n+1∑ℓ,i=1∫t2t1∫Ω(δ+|∇εuε|2)p2|Zuε|β−2ηβ+2|XεiXεℓuε|2. |
It follows that
4∑k=12n+1∑ℓ=1Ikℓ+9∑k=5Ik≤α∫t2t1∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β2n+1∑i,j=1|XεiXεjuε|2+Cα(β+1)2(1+|∇εη||2L∞)∫t2t1∫Ω(ηβ+ηβ+4)(δ+|∇εuε|2)p2|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2+2α(1+||∇εη||2)(β+2)∫t2t1∫Ω|Zuε|β+2ηβ+3|∂tη|dx+α(β+2)2∫Ω|Zuε|β+2ηβ+4|t=t1+α(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2(ηβ+2+ηβ+4)|∇εuε|2+CΛ(β+1)2∫t2t1∫Ω(δ+|∇εuε|2)p+22|Zuε|β−2ηβ+2. | (3.16) |
Summing up Eqs (3.12) and (3.15), we obtain
2n+1∑i,j=1∫t2t1∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β|XεiXεjuε|2+∫Ω(ηβ+2|Zuε|β|∇εuε|2)|t2t1+∫t2t1∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2ηβ+2|∇εuε|2=∫t2t1∫Ω|Zuε|β|∇εuε|2∂t(ηβ+2)+4∑k=12n+1∑ℓ=1Ikℓ+9∑k=5Ik. |
Applying (3.16), the proof is completed.
At this point we make use of the non-degeneracy condition δ>0, and recalling that Z is obtained as a commutator of the horizontal vector fields and that η≤1, we estimate
∫t2t1∫Ω|Zuε|2η3dxdt≤Cδ∫t2t1∫Ω(δ+|∇εuε|2)p−222n+1∑i,j=1|XεiXεjuε|2η2dxdt. | (3.17) |
Lemma 3.6 and (3.17) yield the following
Corollary 3.8. Let uε be a weak solution of (1.8) in Q. For any t2≥t1≥0, and all η∈C∞0(Ω), such that η≤1, ||∂tη||≤C||∇εη||2. For every fixed value of δ there exists Cδ depending on δ,p,n and on the structure constants, such that
12∫Ω((δ+|∇εuε|2)η2)|t2t1+λ∫t2t1∫Ω(δ+|∇εuε|2)p−222n+1∑i,j=1|XεiXεjuε|2η2≤Cδ∫t2t1∫Ω(δ+|∇εuε|2)p/2(η2+|∇εη|2+|ηZη|). |
Corollary 3.9. Let uε be a solution of (1.8) in Ω×(0,T) and Bε(x0,r)×(t0−r2,t0) a parabolic cylinder. Let η∈C∞(Bε(x0,r)×(t0−r2,t0)) be a non-negative test function η≤1, which vanishes on the parabolic boundary and such that there exists a constant Cλ,Λ>1 for which ||∂tη||L∞≤Cλ,Λ(1+||∇εη||2L∞). Set t1=t0−r2. There exists a constant Cδ,λ,Λ, also depending on δ, such that for all β≥2 one has
∫t0t0−r2∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β2n+1∑i,j=1|XεiXεjuε|2+maxt∈(t0−r2,t0]∫Ωηβ+2|Zuε|β|∇εuε|2+(β+1)2∫t0t0−r2∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2ηβ+2|∇εuε|2≤Cλ,Λ(β+1)2(1+|∇εη||2L∞)∫t0t0−r2∫Ωηβ(δ+|∇εuε|2)p2|Zuε|β−22n+1∑i,j=1|XεiXεjuε|2+Cλ,Λ(β+1)2∫t0t0−r2∫Ω(δ+|∇εuε|2)p+22|Zuε|β−2ηβ+2. | (3.18) |
Proof. The statement follows at once by standard parabolic pde arguments, after choosing α appropriately small in (3.11) and applying (3.17), once one notes that |Zuε|≤∑2n+1i,j=1|XεiXεjuε|.
Corollary 3.10. In the hypotheses of the previous corollary we have
2n+1∑i,j=1∫t0t0−r2∫Ωηβ+2(δ+|∇εuε|2)p−22|Zuε|β|XεiXεjuε|2+maxt∈(t0−r2,t0]∫Ωηβ+2|Zuε|β|∇εuε|2+(β+1)2∫t0t0−r2∫Ω(δ+|∇εuε|2)p−22|∇εZuε|2|Zuε|β−2ηβ+2|∇εuε|2≤Cβ/2(β+1)β(||∇εη||2L∞+1))β/22n+1∑i,j=1∫t0t0−r2∫Ωηβ(δ+|∇εuε|2)p−2+β2|XεiXεjuε|2, |
where c=c(n,p,L)>0.
Proof. In order to handle the first term in the right-hand side of the sought for conclusion, it suffices to observe that
C(β+1)2(||∇εη||2L∞+1)ηβ(δ+|∇εuε|2)p/2|Zu|β−22n+1∑i,j=1|XεiXεjuε|2==ηβ−2(δ+|∇εu|2)(p−2)(β−2)/2β|Zu|β−2(|2n+1∑i,j=1XεiXεjuε|2)(β−2)/β+η2(δ+|∇εuε|2)(p+β−2)/β(2n+1∑i,j=1|XεiXεjuε|2)2/βC(β+1)2(||∇εη||2L∞+1). |
The conclusion then follows from Hölder's inequality. We can handle the second term in the same way
C(β+1)2(||∇εη||2L∞)ηβ(δ+|∇εuε|2)(p+2)/2|Zu|β−42n+1∑i,j=1|XεiXεjuε|2=ηβ−2(δ+|∇εu|2)(p−2)(β−4)2β|Zu|β−4(|2n+1∑i,j=1XεiXεjuε|2)(β−4)/β×η2(δ+|∇εuε|2)2(p+β−2)/β(2n+1∑i,j=1|XεiXεjuε|2)4/βC(β+1)2(||∇εη||2L∞+1). |
The key step in the proof of the Lipschitz regularity of solutions is the following Caccioppoli type inequality which is a parabolic analogue of [31,Theorem 3.1].
Theorem 3.11. Let uε be a solution of (1.8) in Ω×(0,T) and Bε(x0,r)×(t0−r2,t0) a parabolic cylinder. Let η∈C∞(Bε(x0,r)×(t0−r2,t0]) be a non-negative test function η≤1, which vanishes on the parabolic boundary such that there exists a constant Cλ,Λ>1 for which ||∂tη||L∞≤Cλ,Λ(1+||∇εη||2L∞). Set t1=t0−r2,t2=t0. There exists a constant C>0 depending on δ p, and Λ such that for all β≥2 one has
∫t2t1∫Ωη2(δ+|∇εuε|2)(p−2+β)/22n+1∑i,j=1|XεiXεjuε|2dxdt+1β+2maxt∈(t0−r2,t0]∫Ω(δ+|∇εuε|2)β2+1η2≤C(β+1)5(||∇εη||2L∞+||ηZη||L∞+1)∫t2t1∫spt(η)(δ+|∇εuε|2)(p+β)/2. |
Proof. In view of Lemma 3.5, the conclusion will follow once we provide an appropriate estimate of the term
∫t2t1∫Ωη2(δ+|∇εuε|2)(p−2+β)/2|Zuε|2. |
The first step is to apply Hölder's inequality to obtain
∫t2t1∫Ωη2(δ+|∇εuε|2)(p−2+β)/2|Zuε|2≤(∫t2t1∫ηβ+2(δ+|∇εuε|2)p−22|Zuε|β+2dxdt)2β+2(∫t2t1∫spt(η)(δ+|∇εuε|2)p+β2)ββ+2(since |Zuε|≤n∑i,j=1|XεiXεjuε|)≤(∫t2t1∫ηβ+2(δ+|∇εuε|2)p−22|Zuε|βn∑i,j=1|XεiXεjuε|2)2β+2(∫t2t1∫spt(η)(δ+|∇εuε|2)p+β2)ββ+2(the first integral in the right-hand side can be bounded by applying Corollary 3.10, resulting in the estimate)≤Cββ+2(β+1)2ββ+2(||∇εη||2L∞+1)ββ+2(∫t2t1∫Ωηβ(δ+|∇εuε|2)p−2+β22n+1∑i,j=1|XεiXεju|2)2β+2×(∫t2t1∫spt(η)(δ+|∇εuε|2)p+β2)ββ+2(by Young' s inequality, recalling C0 from the statement of Lemma 3.5)≤Cββ+2(4C0(β+1)4(β+2))2β(β+1)2(||∇εη||2L∞+1)∫t2t1∫spt(η)(δ+|∇εuε|2)p+β2+12C0(β+1)4∫t2t1∫Ωηβ(δ+|∇εuε|2)p−2+β22n+1∑i,j=1|XεiXεjuε|2. |
Now we note that
ββ+2(4C0(β+1)4(β+2))2β(β+1)2≤Cλ,Λ(β+1)5. |
Substituting the previous estimate in Lemma 3.5, we conclude
∫t2t1∫Ωη2(δ+|∇εuε|2)p−2+β2|XεiXεjuε|2≤Cλ,Λ(β+1)5(||∇εη||2L∞+||ηZη||L∞+1)∫t2t1∫spt(η)(δ+|∇εuε|2)p+β2. |
This completes the proof of the theorem.
In the next result, from Lemma 2.4 and Theorem 3.11 we will establish local integrability of ∇εuε in Lq for every q≥p.
Lemma 3.12. Let uε be a solution of (1.8) in Q. For any open ball B⊂⊂Ω and T>t2≥t1≥0, consider a test function η∈C∞([0,T]×B), vanishing on the parabolic boundary, such that η≤1, ||∂tη||≤C||∇εη||2. For every β≥0, there exists a constant C=C(n,p,λ,Λ,d(B,∂Ω),T−t2,δ)>0, such that
∫t2t1∫Ω(δ+|∇εuε|2)(β+p+2)/2|η|β+2≤Cβ(β+1)β∫t2t1∫B(δ+|∇εuε|2)p/2. |
Proof. We begin by examining the case β=0. Applying Lemma 2.4 and Corollary 3.8 one can find positive constants C1,C2,C3, depending on n,p,λ,Λ,d(B,∂Ω),T−t2,δ, such that
∫t2t1∫Ω(δ+|∇εuε|2)(p+2)/2|η|2≤C1(p+1)2∫t2t1∫Ω(δ+|∇εuε|2)p−22∑i,j|XεjXεiuε|2|η|2+C2β2∫t2t1∫Ω(δ+|∇εuε|2)p/2(|η|2+|∇εη|2)≤C3∫t2t1∫Ω(δ+|∇εuε|2)p/2(η2+|∇εη|2+|ηZη|), |
concluding the proof in the case β=0. Next, we consider the range β≥2. The interpolation inequality Lemma 2.4 and Theorem 3.11 imply the existence of positive constants C4,...,C7, depending on n,p,λ,Λ,d(B,∂Ω),T−t2, and δ, such that
∫t2t1∫Ω(δ+|∇εuε|2)(β+p+2)/2|η|β+2≤C4(β+p+1)2∫t2t1∫Ω(δ+|∇εuε|2)p+β−22∑i,j|XεjXεiuε|2|η|β+2+C5β2∫t2t1∫Ω(δ+|∇εuε|2)(β+p)/2|η|β(|η|2+|∇εη|2)≤C6(β+p+1)7∫t2t1∫Ω(δ+|∇εuε|2)(p+β)/2(η2+|∇εη|2+|ηZη|)≤C7(β+1)7(||∇εη||2L∞+||ηZη||L∞+1)∫B(δ+|∇εuε|2)(p+β)/2. | (3.19) |
Iterating the latter [β]/2 times, the conclusion follows.
In the next result we establish Lipschitz bounds that are uniform in ε. The argument consists in implementing Moser iterations, and rests on the observation that the quantity δ+|∇εuε|2 is bounded from below by δ>0, and that for every β≥0 it is bounded in Lp+β in a parabolic cylinder, uniformly in ε.
In the iteration itself, we will consider metric balls Bε defined through the Carnot-Caratheodory metric associated to the Riemannian structure gε defined by the orthonormal frame Xε1,...,Xε2n+1. We recall here that gε converges to the sub-Riemannian structure of the Heisenberg group in the Gromov-Hausdorff sense [19], and in particular Bε→B0 in terms of Hausdorff distance. These considerations should make it clear that the estimates in the following theorem are stable as ε→0.
Theorem 3.13. Let uε be a solution of (1.8) in Ω×(0,T) and Qε0=Bε(x0,r)×(t0−r2,t0) a parabolic cylinder contained in Ω×(0,T). For given σ∈(0,1), there exists a constant C=C(p,σ,β0,λ,Λ,δ)>0 such that
supB(x0,σr)×(t0−(σr)2,t0)(δ+|∇εuε|2)p2≤C−∫Bνεt0t0−r2−∫BνεB(x0,r)(δ+|∇εuε|2)p2. | (3.20) |
Proof. We recall the main steps. Let us consider a family of cylinders Qεi=Bε(x0,ri)×(t0−r2i,t0)⊂⊂Qε0 and with ri<ri−1. Applying (ii) in Lemma 2.3 to the function wβ=(δ+|∇εuε|2)β+24, one obtains
(∫t0t0−r2i∫Bε(x0,ri)(δ+|∇εuε|2)(β+2)N12(N1−2))N1−2N1=||wβ||22N1N1−2,2N1N1−2,Qεi≤||wβ||22,∞,Qεi+||∇εwβ||22,2,Qεi≤∫t0t0−r2i∫Bε(x0,ri)η2(δ+|∇εuε|2)β/22n+1∑i,j=1|XεiXεjuε|2+1β+2maxt∈(t0−r2,t0]∫Bε(x0,ri)(δ+|∇εuε|2)β+22η2. |
Next, we set g=(δ+|∇εuε|2)(p−2)/2. Using Theorem 3.11, along with the fact that (δ+|∇εuε|)≥δ>0, we obtain
(∫t0t0−r2i∫Bε(x0,ri)(δ+|∇εuε|2)(β+2)N12(N1−2))N1−2N1≤C(β+p)6(ri−ri−1)2∫t0t0−r2i∫Bε(x0,ri)g(δ+|∇εuε|2)(β+2)/2. |
Setting q=(β+2)N1N1−1 and k=N1−1N1−2 in the latter inequality, we deduce
(−∫Bνεt0t0−r2i−∫BνεB(x0,ri)(√δ+|∇εuε|2)qk)1qk≤C1β+2(β+p)6β+2(r2+Ni−1rNi(ri−ri−1)2)12+β(−∫Bνεt0t0−r2i−1−∫BνεB(x0,ri−1)g(√δ+|∇εuε|2)β+2)1β+2≤C1β+2(β+p)6β+2(r2+Ni−1rNi(ri−ri−1)2)12+β(−∫Bνεt0t0−r2i−1−∫BνεB(x0,ri−1)(√δ+|∇εuε|2)q)1q. |
The classical Moser iteration scheme in see [24] now applies, leading to the sought for conclusion.
This section focuses on the proof of the second part of Theorem 1.2. Namely, we want to prove that for each δ,ε>0 a weak solution
uε∈Lp((0,T),W1,p,ε(Ω))∩C2(Q) |
of the approximating PDE (1.8) in Q=Ω×(0,T) satisfies the Hölder estimates
||∇εuε||Cα(B×(t1,t2))+||Zuε||Cα(B×(t1,t2))≤C(∫T0∫Ω(δ+|∇εuε|2)p2dxdt)1p, |
for any open ball B⊂⊂Ω and T>t2≥t1≥0, and for some constants C=C(n,p,λ,Λ,d(B,∂Ω),T−t2,δ)>0 and α=α(n,p,λ,Λ,d(B,∂Ω),T−t2,δ)∈(0,1) independent of ε. It is clear that the above estimate represents the ε-version of (1.11).
We begin by studying the regularity of the derivatives of uε. In view of Lemma (3.2) and (3.3), for each ε>0, ℓ=1,⋯,2n+1 all derivatives Xεℓuε and Zuε of uε satisfy the PDE
∂twε=2n+1∑i=1Xεi(2n+1∑j=1aεij(x,t)Xεjwε+aεi(x,t))+aε(x,t), | (4.1) |
where
aεij(x,t)=Aεi,ξj(x,∇εuε), |
aεi(x,t)=Aεi,xℓ(x,∇εuε)−sℓxℓ+sℓn2Aεi,x2n+1(x,∇εuε), |
aε(x,t)=sℓZ(Aεℓ+sℓn(x,∇εuε)). |
By Lemma 3.3, Zuε satisfies the same equation, for sℓ=0. For every K⊂⊂Q, by Theorem 3.13, |∇εuε| is bounded in K uniformly in ε. Hence ai,j and ai are locally bounded in K, with ellipticity constants uniform in ε>0, but dependent on δ. Precisely, there exists a constant C0, and constants λδ=δ and Λδ=Λ(δ2+C21) such that for every η∈R2n+1 and for a.e. (x,t)∈K, ε>0, i,j=1,⋯,2n+1
||aεij(x,t)||L∞(K)+||aεi(x,t)||L∞(K)≤C0λδ|η|2≤2n+1∑i,j=1aεij(x,t)ηiηj≤Λδ|η|2. | (4.2) |
Since aε=0 in the equation satisfied by Zuε, we will then start with studying the regularity of derivatives of the solution along the center of the group.
Proposition 4.1. Let uε be a solution of (1.8) in Ω×(0,T) and Qr=B(x0,r)×(t0−r2,t0) a parabolic cylinder contained in Ω×(0,T). There exists constants C=C(p,σ,λ,Λ,δ)>0 and α=α(p,σ,λ,Λ,δ)∈(0,1) such that
||Zuε||Cα(Qr/2)+||∇εZuε||L2(Qr/2)≤C(||uε||Lp(Qr)+||∇εuε||Lp(Qr)). |
Proof. First of all, we observe that since δ>0 is fixed, Lemma 3.12 and Theorem 3.11 imply that for all i,j=1,...,2n, one has |XiXjuε| is bounded in L2 uniformly in ε>0. It follows that Zuε∈L2loc(Q) uniformily in ε>0. Since Zuε is a solution of (4.1), with aε=0, the Caccioppoli inequality implies that ∇εZuε is in L2loc(Q), uniformly in ε>0. The stable Harnack inequality established in [9] and [2] (see also [1,17,28] for the Riemannian case) yields interior Hölder estimates for wε in Qr, which are stable as ε→0.
Actually, we will prove a stronger result, in parabolic Morrey spaces. Mq,α(Q) denotes the space of all functions f∈Lq(Q) such that
||f||Mq,λ(Q)=supr∈S1rα−1(r−N1∫t0min(t0−r2,0)∫B∩Ω|f|qdxdt)1/q<∞, | (4.3) |
where S is the set of positive radius r such that B=B(x0,r)⊂Ω, and r2<t0<T.
We also recall that the parabolic Campanato spaces Lq,α(Q) is the collection of all f∈Lq(Q) such that
||f||Lq,α(Q)=supr∈S1rα(r−N1∫t0min(t0−r2,0)∫B∩Ω|f−f(x0,t0),r|qdxdt)1/q<+∞. | (4.4) |
Here, we have set
f(x0,t0)=r−N1∫t0min(t0−r2,0)∫B∩Ωf(x,t)dxdt. |
Remark 4.2. Let α∈(0,1) denote the Hölder exponent of Zuε (which is uniform in ε>0). By observing that wε−wε(x0,t0) is also a solution of (4.1), then a standard Caccioppoli type argument yields
∫t0t0−r2∫B|∇εZu|2dxdt≤C1r2∫t0t0−(2r)2∫2B|Zuε−Zuε(x0,t0)|2dxdt≤Cr2α−2rN1, | (4.5) |
where N1=2n+4 is the parabolic dimension, defined in (2.3).
This shows, in particular, that for every compact K cointained in Q there is a constant C>0 independent of ε such that ||∇εZu||M2,α(K)≤C, so that the coefficient aε in Eq (4.1) satisfies
||aε||L2(K)+||aε||M2,α(K)≤C0. | (4.6) |
A standard argument, see for instance [12], shows that the Campanato space is isomorphic to the space of Hölder continuous functions. In particular, we rely on the following instance of this general result.
Lemma 4.3. Let K⊂⊂Q. There exists M,r0>0 such that for any (x0,t0)∈K and 0<r<r0, if f∈Lq,α(B(x0,r)×(t0−r2,t0)) then f∈Cαε(B(x0,r/M)×(t0−r2/M2,t0)).
We need to invoke a standard result from the theory of Morrey-Campanato which adapts immediately to the Heisenberg group setting, see [6,23].
Lemma 4.4. For each ε≥0, let wε be a weak solution in a cylinder Q=Ω×(0,T) to the Eq (4.1) with smooth coefficients. Assume that for every compact K⊂⊂Q there are constants C0,Λδ,λδ,>0, α∈(0,1) such that (4.2) and (4.6) are satisfied. Also assume that
||wε||L∞(K)+||∇εwε||L2(K)≤C0. | (4.7) |
Then for every K⊂⊂Ω, there exists a constant C>0 depending on C0,Λδ,λδ,α such that ||∇εw||M2,α(K)≤C.
Proof. Choose r>0 such that the cylinder Qr⊂K, and denote by zε the unique solution of the linear PDE, (where we omit the term a):
∂tzε=2n+1∑i=1Xεi(2n+1∑j=1aεij(x,t)Xεjzε),zε=wε on the parabolic boundary of Qr. |
From the maximum principle ||zε||L∞(Qr)≤||wε||L∞(Qr)≤C0, by assumption. Arguing as in Remark 4.2, we see that ||∇εzε||M2,λ(Qr)≤C. Choosing the test function φ=wε−zε in the weak formulation of (4.1) we obtain
∫t0t0−r2∫B|∇ε(wε−ze)|2dxdt≤∫t0t0−r2∫BaiXεi(wε−zε)dxdt+∫t0t0−r2∫Ba(wε−zε)dxdt |
From the hypothesis (4.7), (4.6), and using Young inequality, it immediately follows that
∫t0t0−r2∫B|∇ε(wε−ze)|2dxdt≤CrN1+CrN1/2(∫t0t0−r2∫Ba2dxdt)12≤CrN1+α−1. | (4.8) |
The thesis follows from the fact that
||∇εwε||M2,α(Qr)≤||∇ε(zε−wε)||M2,α(Qr)+||∇εzε||M2,α(Qr)≤2C |
and the right hand side is bounded independently of ε.
Remark 4.5. If uε be a solution of (1.8) in Ω×(0,T), the derivative ∂tuε satisfies the same equation as Zuε. As a result, arguing as in Remark 4.2 we deduce that for every compact K contained in Q there is a constant C>0 independent of ε such that
||∇ε∂tuε||M2,α(K)≤C |
Proof of Theorem 1.2. For every K⊂⊂Q, by Theorem 3.13, there exists a constant C0 independent of ε such that |∇εuε|≤C0 and Theorem 3.11 imply that for all i,j=1,...,2n, one has ||XiXjuε||L2≤C0. Hence the function wℓ=Xεℓuε for every ℓ=1,...,2n satisfies (4.7). Furthermore we already noted that it is a solution of equation (4.1) with smooth coefficients satisfying (4.2) and (4.6) uniformly in ε>0. One can apply Lemma 4.4 and Remark 4.5 to conclude that ||∂tXεluε||M2,α+||∇εXεluε||M2,α≤C, for a suitable constant C. In view of the Poincaré inequality, and recalling that its costant is independent of ε (see [7,9]), one then has that ∇εuε belongs to the Campanato spaces L2,α. Finally, by virtue of Lemma 4.3 it follows that ∇εuε is Hölder continuous, with norm independent of ε, thus concluding the proof.
We will need a simple form of the comparison principle, see [3] and [4].
Lemma 5.1. Let u,w be weak solutions of (1.1) in a cylinder B×(t1,t2). If on the parabolic boundary B×{t1}∪∂B×(t1,t2) one has that u≥w, then u≥w in B×(t1,t2).
We now show how Theorem 1.1 follows from the comparison principle and from Theorem 1.2.
Proof of Theorem 1.1. Recall from Lemma 2.1 that u is Hölder continuous in any compact subdomain of Q, in particular in the closure of B×(t1,t2). For each ε>0 consider uε, the unique smooth solution of the quasilinear parabolic problem
{∂tuε=∑2n+1i=1XεiAεi(x,∇εuε), in B×(t1,t2)uε=u in B×{t1}∪∂B×(t1,t2), | (5.1) |
where Aεi(x,ξ) satisfies the structure conditions (1.7). By virtue of Theorem 3.1 and of the Hölder regularity from Theorem 1.2, one has that for every K⊂⊂Q, and q≥1, there exist M=M(p,q,λ,Λ,n,δ)>0 and α=α(p,q,λ,Λ,n,δ)∈(0,1), such that for every ε>0, (x0,t0)∈K and B(x0,r)×(t0−r2,t0)⊂Q,
||∇ε|∇εuε|q||L2(B(x0,r)×(t0−r2,t0))≤M,||Z|∇εuε|q||L2(B(x0,r)×(t0−r2,t0))≤M||∇εuε||Cαε(B(x0,r)×(t0−r2,t0))+||Zuε||Cαε(B(x0,r)×(t0−r2,t0))≤M. |
By the theorem of Ascoli-Arzelà, one can find u0∈C1,αloc(Q) and a sequence εk→0 such that
uεk→u0 and ∇εkuεk→∇0u0 uniformly on compact subsets of Q. |
The latter implies that u0 is a weak solution of (1.1), in B(x0,r)×(t0−r2,t0), which agrees with the function u on the parabolic boundary of B(x0,r)×(t0−r2,t0). By the comparison principle, the solution to this boundary values problem is unique, and hence we conclude that u∈C1,αloc(B(x0,r)×(t0−r2,t0)).
The first author was partially funded by NSF awards DMS 1101478, and by a Simons collaboration grant for mathematicians 585688. The second author was partially funded by Horizon 2020 Project ref. 777822: GHAIA, and by PRIN 2015 Variational and perturbative aspects of nonlinear differential problems. The third author was supported in part by a Progetto SID (Investimento Strategico di Dipartimento) "Non-local operators in geometry and in free boundary problems, and their connection with the applied sciences", University of Padova, 2017.
We thank Vira A. Markasheva, who collaborated with us on an earlier version of this project.
The authors declare no conflict of interest.
[1] | Aronson DG, Serrin J (1967) Local behavior of solutions of quasilinear parabolic equations. Arch Ration Mech Anal 25: 81-122. |
[2] | Avelin B, Capogna L, Citti G, et al. (2014) Harnack estimates for degenerate parabolic equations modeled on the subelliptic p-Laplacian. Adv Math 257: 25-65. |
[3] | Bieske T (2005) Comparison principle for parabolic equations in the Heisenberg group. Electron J Differ Eq 95: 1-11. |
[4] | Bieske T (2006) Equivalence of weak and viscosity solutions to the p-Laplace equation in the Heisenberg group. Ann Acad Sci Fenn Math 31: 363-379. |
[5] | Bramanti M, Brandolini L (2007) Schauder estimates for parabolic nondivergence operators of Hörmander type. J Differ Equations 234: 177-245. |
[6] | Capogna L (1999) Regularity for quasilinear equations and 1-quasiconformal mappings in Carnot groups. Math Ann 313: 263-295. |
[7] | Capogna L, Citti G (2016) Regularity for subelliptic PDE through uniform estimates in multiscale geometries. B Math Sci 6: 173-230. |
[8] | Capogna L, Citti G, Ottazzi A, et al. (2019) Conformality and Q-harmonicity in sub-Riemannian manifolds. J Math Pure Appl 122: 67-124. |
[9] | Capogna L, Citti G, Rea G (2013) A subelliptic analogue of Aronson-Serrin's Harnack inequality. Math Ann 357: 1175-1198. |
[10] | Citti G, Manfredini M (2006) Uniform estimates of the fundamental solution for a family of hypoelliptic operators. Potential Anal 25: 147-164. |
[11] | Citti G, Lanconelli E, Montanari A (2002) Smoothness of Lipschitz-continuous graphs with nonvanishing Levi curvature. Acta Math 188: 87-128. |
[12] | Da Prato G (1964) Spazi Lp, θ(?, δ) e loro proprietà. Annali di Matematica 69: 383-392. |
[13] | Di Benedetto E (1993) Degenerate Parabolic Equations, Springer, Universitext. |
[14] | Domokos A (2004) Differentiability of solutions for the non-degenerate p-Laplacian in the Heisenberg group. J Differ Equations 204: 439-470. |
[15] | Domokos A, Manfredi JJ (2020) C1, α-subelliptic regularity on S U(3) and compact, semi-simple Lie groups. Anal Math Phys 10: 4. |
[16] | Folland GB, Stein EM (1982) Hardy Spaces on Homogeneous Groups, Princeton: Princeton University Press. |
[17] | Grigor'yan AA (1992) The heat equation on noncompact Riemannian manifolds. Sb Math 72: 47-77. |
[18] | Hörmander L (1967) Hypoelliptic second order differential equations. Acta Math 119: 147-171. |
[19] | Koranyi A (1983) Geometric aspects of analysis on the Heisenberg group, In: Topics in Modern Harmonic Analysis, Vol. I, II (Turin/Milan, 1982), Francesco Severi, Rome: Ist. Naz. Alta Mat., 209-258. |
[20] | Ladyzhenskaya O, Solonnikov VA, Ural'tseva NN (1967) Linear and Quasilinear Parabolic Equations, Moskow: Nauka. |
[21] | Manfredi JJ, Mingione G (2007) Regularity results for quasilinear elliptic equations in the Heisenberg group. Math Ann 339: 485-544. |
[22] | Mingione G, Zatorska-Goldstein A, Zhong X (2009) Gradient regularity for elliptic equations in the Heisenberg group. Adv Math 222: 62-129. |
[23] | Morrey Jr CB (1959) Second order elliptic equations in several variables and Hölder continuity. Math Z 72: 146-164. |
[24] | Moser J (1964) A Harnack inequality for parabolic differential equations. Commun Pure Appl Math 17: 101-134. |
[25] | Mukherjee S (2019) On local Lipschitz regularity for quasilinear equations in the Heisenberg group. arXiv: 1804.00751. |
[26] | Ricciotti D (2015) p-Laplace Equation in the Heisenberg Group, Cham: Springer. |
[27] | Ricciotti D (2018) On the C1, α regularity of p-harmonic functions in the Heisenberg group. P Am Math Soc 146: 2937-2952. |
[28] | Saloff-Coste L (1992) A note on Poincaré, Sobolev, and Harnack inequalities. Int Math Res Notices 1992: 27-38. |
[29] | Stein EM (1993) Harmonic Analysis: Real-Variable Methods, Orthogonality, and Oscillatory Integrals, Princeton: Princeton University Press. |
[30] | Xu CJ (1992) Regularity for quasilinear second-order subelliptic equations. Commun Pure Appl Math 45: 77-96. |
[31] | Zhong X (2009) Regularity for variational problems in the Heisenberg group. arXiv: 1711.03284. |
1. | Chengwei Yu, Huiying Wang, Kunpeng Cui, Zijing Zhao, Second-Order Regularity for Degenerate Parabolic Quasi-Linear Equations in the Heisenberg Group, 2024, 12, 2227-7390, 3494, 10.3390/math12223494 | |
2. | Nicola Garofalo, Dimiter Vassilev, Overdetermined problems in groups of Heisenberg type: Conjectures and partial results, 2024, 287, 00221236, 110588, 10.1016/j.jfa.2024.110588 | |
3. | Luca Capogna, Giovanna Citti, Xiao Zhong, Regularity Theory of Quasilinear Elliptic and Parabolic Equations in the Heisenberg Group, 2024, 52, 2305-221X, 807, 10.1007/s10013-023-00644-0 | |
4. | Yongming He, Chengwei Yu, Hongqing Wang, Local C0,1-Regularity for the Parabolic p-Laplacian Equation on the Group SU(3), 2024, 12, 2227-7390, 1288, 10.3390/math12091288 |