Abstract
We investigate the uniqueness, in suitable weighted Lebesgue spaces, of solutions to a class of elliptic equations with a drift posed on a complete, noncompact, Riemannian manifold M of infinite volume and dimension \(N\ge 2\). Furthermore, in the special case of a model manifold with polynomial volume growth, we show that the conditions on the drift term are sharp.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
We are concerned with uniqueness of solutions, in suitable weighted Lebesgue spaces, to the following linear elliptic equation
where M is an N-dimensional, \(N\ge 2\), complete, noncompact, Riemannian manifold of infinite volume, endowed with a metric tensor g. Here, \(\Delta \) and \(\nabla \) denote, respectively, the Laplace-Beltrami operator and the gradient on M with respect to g; \(b:M\rightarrow M\) is a given vector field and c is a nonnegative function defined in M. The huge interest around the Laplace-Beltrami operator with a drift comes from very different fields ranging from probability to analysis and geometry. A classical assumption that ensures the uniqueness property for harmonic functions
is to restrict u to lie in some \(L^p-\)space. In this setting the uniqueness result for (1.2) is the following: if \(u\in L^p(M)\), for some \(1\le p\le \infty \), then u is identically constant. In two celebrated papers Yau proved the uniqueness result for \(p=\infty \), under the assumption that the Ricci curvature of M is non-negative, and for \(p\in (1,\infty )\) without additional assumptions on M; we refer to [30] and to [29], respectively. These results have been extended to the case \(p\in (0,1)\) and to the case \(p=1\) under sharp lower bounds for the Ricci curvature of M, we refer to [11] and to [10], respectively. The simplest example of Laplace-Beltrami operator with a drift is the so-called weighted Laplacian
i.e. when \(b=-\nabla f\), for some smooth real-valued f on M. In this setting of metric measure spaces \((M,g,e^{-f}d\mu )\), where \(d\mu \) denotes the usual volume form induced by g, the aforementioned uniqueness results have been extended, e.g., in [14, 16, 17, 28] and we refer to these papers and to the references therein for further details. In this paper we consider the general equation (1.1) and we prove uniqueness results in weighted Lebesgue spaces, for the literature related to the weighted uniqueness results we refer to Subsection 1.2. Before presenting our main assumptions and results we need some notation; for all vector fields X, Y belonging to the tangent space of M, we set
Moreover, for any \(x_0\in M\) and \(R > 0\), let \(B_R(x_0) := \{x\in M:\,\, {\text {dist}}(x, x_0) < R\}\) denote the geodesic ball of radius R and centred at \(x_0\), where \({\text {dist}}(x, x_0)\) is the geodesic distance between x and \(x_0\). Furthermore, let \(V(x_0, R)\) denote the Riemannian volume of \(B_R(x_0)\). In what follows, we set
Let us define the following subsets of M
Then, concerning the vector field b, we will assume either that:
or that, for some \(0<\theta <1\) that will be specified later,
or that
Here, for a given function v, the negative part \([\cdot ]_{-}\) is defined as
Moreover, the coefficient c is such that
for some \(c_0>0\).
We now summarize our main results and we refer to Section 2 for the precise statements.
1.1 Outline of Our Results
The main results of this paper will be given in detail in the forthcoming Theorems 2.2, 2.3 and 2.4. We give here a sketchy outline of these results, describing motivations and techniques of proofs.
For any \(\phi \in C(M)\), \(\phi >0\), \(p> 1\), set
We shall prove the following
-
Let M be a complete, noncompact, Riemannian manifold of infinite volume such that, for some \(x_0\in M\),
$$\begin{aligned} V(x_0,r)\le e^{\alpha r},\quad \text {for some}\,\,\,\alpha >0. \end{aligned}$$(1.5)If \(b:M\rightarrow M\) satisfies assumption (\(H_0\)), then the solution to equation (1.1) is unique in the class \(L^p_{\psi }(M)\) with \(p>1\) and
$$\begin{aligned} \psi (x)=\psi (r):=e^{-\beta \,r}\;\, (r\in (0,+\infty ))\,, \end{aligned}$$(1.6)for properly chosen \(\beta >\alpha \) (see Theorem 2.2). Observe that, in general, our uniqueness class includes unbounded solutions. Thus, in particular, we get uniqueness of bounded solutions. More precisely, bounded solutions always belong to this class of uniqueness \(L^p_{\psi }(M)\). Furthermore, also unbounded solutions can be considered if such solutions satisfy
$$\begin{aligned} u(x)\le Ce^{\gamma r}\quad \text {for}\,\,\,r>1,\,\,\gamma >0\,, \end{aligned}$$provided that \(\beta >\alpha +\gamma p\).
-
Let M be a complete, noncompact, Riemannian manifold of infinite volume such that, for some \(x_0\in M\),
$$\begin{aligned} V(x_0,r)\le e^{\alpha r^{\theta }},\quad \text {for some}\,\,\,\alpha >0,\quad \text {for some}\,\,\,0<\theta <1. \end{aligned}$$(1.7)If \(b:M\rightarrow M\) satisfies assumption (\(H_1\)) with \(\theta \) as in (1.7), then the solution to equation (1.1) is unique in the class \(L^p_{\eta }(M)\) with \(p>1\) and
$$\begin{aligned} \eta (r):=e^{-\beta \,r^{\theta }} \quad \text {for some}\,\,\,\beta >0, \end{aligned}$$(1.8)for properly chosen \(\beta >\alpha \) (see Theorem 2.3). Observe that, in particular, bounded solutions always belong to this class of uniqueness \(L^p_{\eta }(M)\). Moreover, also unbounded solutions can be considered if such solutions satisfy
$$\begin{aligned} u(x)\le Ce^{\gamma r^{\theta }}\quad \text {for}\,\,\,r>1,\,\,\gamma >0\,, \end{aligned}$$provided that \(\beta >\alpha +\gamma p\).
-
Let M be a complete, noncompact, Riemannian manifold of infinite volume such that, for some \(x_0\in M\),
$$\begin{aligned} V(x_0,r)\le r^{\alpha },\quad \text {for some}\,\,\,\alpha >0. \end{aligned}$$(1.9)If \(b:M\rightarrow M\) satisfies assumption (\(H_2\)), then the solution to equation (1.1) is unique in the class \(L^p_{\xi }(M)\) with \(p>1\) and
$$\begin{aligned} \xi (x)=\xi (r):=(1+r)^{-\tau }\;\, (r\in (0,+\infty ))\,, \end{aligned}$$(1.10)for properly chosen \(\tau >\alpha +N-1\) (see Theorem 2.4). Similarly to the previous case, we observe that, our uniqueness class includes unbounded solutions. Thus, in particular, we get uniqueness of bounded solutions. More precisely, bounded solutions always belong to this class of uniqueness \(L^p_{\xi }(M)\). Moreover, also unbounded solutions can be considered if such solutions satisfy
$$\begin{aligned} u(x)\le Cr^{\gamma }\quad \text {for}\,\,\,r>1,\,\,\gamma >0\,, \end{aligned}$$provided that \(\tau >\alpha +\gamma p+N-1\).
Observe that hypotheses (\(H_1\)) and (\(H_2\)) allow b to be unbounded. In addition, such assumptions require a growth condition on b, when \(\langle b(x), x\rangle > 0\). However, when \(\langle b(x), x\rangle \le 0\) no further conditions on b are imposed.
-
Furthermore, we show sharpness of the result of Theorem 2.4 in the special case of a model manifold M such that (2.1) holds. More precisely, we show that, if the drift term b violetes, in an appropriate sense, assumption (\(H_2\)) (that is, if it satisfies inequality (2.2) below), then infinitely many bounded solutions to problem (1.1) exist (see Proposition 2.5 and Corollary 2.6 below).
We shall now briefly recall some known results related to equation (1.1).
1.2 Summary of Known Results
Let us firstly consider the Euclidean setting. In [2] uniqueness of solutions to the Cauchy problem
is investigated. Here \(T>0\) and
the coefficients which appear in L, together with all their derivatives, are locally bounded functions and the matrix \(A\equiv (a_{ij})\) is assumed to be positive semidefinite in \({\mathbb {R}}^N\times (0,T)\). Moreover the authors assume that
for almost every \((x,t)\in {\mathbb {R}}^N\times (0,T)\), for some constants \(\lambda \ge 0\) and \(K_i>0\), for \(i=1,2,3\). Given \(\phi \in C({\mathbb {R}}^N\times (0,T))\), \(\phi >0\), set
In [2, Theorem 1] it is shown that if u is a solution to problem (1.11) and \(u\in L^1_{\phi }({\mathbb {R}}^N\times (0,T))\), with
or
for some \(\alpha _0>0\), then
The result in [2] has attracted a lot of interests and analogue results has been proved in [3, 8, 9, 20] (we refer to these papers for further discussions and details) and generalized to the fractional heat equation in [12, 25, 26]. We also mention that similar results has been considered in the case of degenerate elliptic and parabolic problems in \(\mathbb {R}^N\) and also in bounded domains of \(\mathbb {R}^N\) (see e.g. [13, 19, 20, 23, 24] and the references therein for previous results).
Let us now consider the Riemannian setting. The study of uniqueness of solutions to the heat equation
where M is a complete Riemannian manifolds, has been largely investigated in the literature. Indeed, it is well-known that uniqueness of solutions to problem (1.14) is equivalent to the stochastic completeness of the Riemannian manifold M (see e.g. [7, Theorem 6.2]). We refer the interested reader to [7] for a complete and exhaustive picture about these kind of results. More precisely, in [7, Theorem 9.2] it is shown that, if u is a classical solution to (1.14) with \(u_0\equiv 0\), then
provided that \(u\in L^2_\phi (M\times (0,T))\), where
with
where r is given in (1.3) and f is a positive increasing continuous function defined in \((0,\infty )\) such that
for some \(R_0>0\). The previous result in [7] has been generalized in [22] in the following sense: if u is a classical solution to (1.14) with \(u_0\equiv 0\), then
provided \(u\in L^p_\phi (M\times (0,T))\), for \(1<p\le 2\), where
with \(\phi \) as before.
Concerning the elliptic case, in [6] (see also [7, Section 13.2]) the author proves that the only bounded solution to equation (1.1), with \(b\equiv 0\) and \(c(x)\ge 0\), is the trivial one provided that
where \(k>0\) and
Similar results have also been obtained for more general linear, elliptic, degenerate equations such as
where
with possibly unbounded coefficients \(a_{ij}\) , \(b_i\), c and function f. Well-posedness of problem (1.15) has been intensively investigated both in the case of \(\Omega ={\mathbb {R}}^N\) and \(\Omega \) being a bounded domain where the problem is completed with suitable boundary conditions and the coefficients can be degenerate or singular at the boundary of the domain (see e.g. [21, 24] and the references therein). Similar uniqueness results have been studied also in the framework of nonlocal diffusion, we refer to [12, 25, 26].
In particular, in [21, Section 1.2] the author consider the problem (1.15) in a bounded and regular domain \(\Omega \) of \(\mathbb {R}^N\) where the coefficients \(a_{ij}\) and \(b_i\) satisfy suitable assumptions (see [21, \((H_1)-(H_3)-(H_4)\)]), c satisfies (\(H_3\)) and \(f\in C(\Omega )\). Under these hypothesis in [21, Theorem 1.5] it is shown that there exist at most one solution \(u\in L^p_{\phi }(\Omega )\) of problem (1.15), for some suitable \(p>1\). In this case,
where \(\alpha >0\) and \(d(x)=\textrm{dist}(x,\partial \Omega )\). Of course, if \(f\equiv 0\) then one obtains a uniqueness result in the weighted Lebesgue space in the same spirit of the results that we are going to present.
Organisation of the paper The paper is organized as follows. In Sect. 2 we state our main results and we recall some basic facts from Riemannian geometry that will be useful in the sequel. Sections 3, 4 and 5 are devoted to the proof of Theorem 2.2, Theorem 2.3 and Theorem 2.4, respectively. Finally, in Sect. 6, we show Proposition 2.5 and Corollary 2.6.
2 Mathematical Framework and Results
Throughout the paper we deal with classical solutions to equation (1.1), for this reason we recall their definition.
Definition 2.1
We say that a function u is a classical solution to equation (1.1) if
-
(i)
\(u\in C^{2}(M)\);
-
(ii)
\(\Delta u + \left\langle b(x),\,\nabla u \right\rangle - c(x)u \,=\,0\) for all \(x\in M\).
Furthermore, we say that u is a supersolution (subsolution) to equation (1.1), if in (ii) instead of “\(=\)” we have “\(\ge \)” “(\(\le )\)”.
We now state the main result where we consider those Riemannian manifolds M such that (1.5) is satisfied and we assume the vector field b to be bounded, see (\(H_0\)).
Theorem 2.2
Let M be a complete, noncompact, Riemannian manifold such that (1.5) is satisfied. Let assumptions (\(H_0\)) and (\(H_3\)) be satisfied. Let u be a classical solution to equation (1.1). Let \(\psi \) be as in (1.6). Let \(p>1\), with \(p c_0>C=C(\beta ,K_0)\) where C can be explicitly computed, see (3.24). If \(u\in L^p_{\psi }(M)\), then
The next two results show that, if one restricts the assumptions on the Riemannian manifold, thus (1.5) does not follows anymore, then b can also be unbounded in an appropriate sense that, somehow, is related to the volume growth condition made on M.
Theorem 2.3
Let M be a complete, noncompact, Riemannian manifold such that (1.7) is satisfied. Let assumptions (\(H_1\)) and (\(H_3\)) be satisfied. Let u be a classical solution to equation (1.1). Let \(\eta \) be as in (1.8). Let \(p>1\), with \(p c_0>C=C(\beta ,K_1)\), where C can be explicitly computed, see (4.6). If \(u\in L^p_{\eta }(M)\), then
Theorem 2.4
Let M be a complete, noncompact, Riemannian manifold such that (1.9) is satisfied. Let assumptions (\(H_2\)) and (\(H_3\)) be satisfied. Let u be a classical solution to equation (1.1). Let \(\xi \) be defined as in (1.10). Let \(p>1\), with \(p c_0>C=C(\beta ,K_2)\) where C can be explicitly computed, see (5.18). If \(u\in L^p_{\xi }(M)\), then
In the special case of \(b=0\), our Theorems apply and agree with the already mentioned Yau’s results, see [29] and [30], concerning harmonic and positive subharmonic functions. Similarly, when b is the gradient of some function similar results have been proved in the already mentioned [14, 16, 17, 28]. Our results cover also the case when b is not a gradient, where the situation is different and more involved.
In the following we collect some observations regarding the hypothesis of the theorems:
-
(i)
The hypothesis \(p \,c_0 \) large enough made in Theorem 2.2, Theorem 2.3 and Theorem 2.4 will be specified in their proofs.
-
(ii)
Observe that assumption (1.7) in Theorem 2.3 is stronger than assumption (1.5) in Theorem 2.2. Although, this hypothesis enables us to relax the assumption on the vector field b, thus b in Theorem 2.3 might be unbounded, see (\(H_1\)).
-
(iii)
We mention that thanks to the Laplacian comparison Theorems (see e.g. [5, Section 2] or [7, Section 15]) we have that assumptions (1.5), (1.7) and (1.9) are satisfied if the Ricci curvature of the Riemannian manifold M is bounded from below. In particular, (1.5) and (1.7) are guaranteed if
$$\begin{aligned} \textrm{Ric}\ge -k\quad \text {for some}\,\,\, k>0\,. \end{aligned}$$Similarly, (1.9) is true if
$$\begin{aligned} \textrm{Ric}\ge 0\,. \end{aligned}$$
Finally, in the special case of a model manifold \(\mathbb {M}^N_{\varphi }\) (see Definition 2.7) satisfying (1.9) with
we show that the hypothesis (\(H_2\)) on b, is sharp in the sense that, if it fails, then infinitely many bounded solutions to problem (1.1) exist; this is the content of the next proposition. In particular, we observe that due to (2.8), assumption (2.1) is compatible with (1.9), provided \(\lambda <\frac{\alpha -1}{N-1}\).
Proposition 2.5
Let \(\mathbb {M}^N_{\varphi }\) be a model manifold such that (2.1) and (1.9) are satisfied. Suppose that assumption (\(H_3\)) holds and let \(b\in C^1(\mathbb {M}^N_{\varphi })\) be such that
for some
Then equation (1.1) admits infinitely many bounded solutions in the sense of Definition 2.1.
Observe that, provided that \(\lambda <\frac{\alpha -1}{N-1}\), then it is always true that \(\lambda <\frac{\tau -1}{N-1}\), where \(\tau \) is introduced in (1.10). Consequently any bounded solution to problem (1.1) given by Proposition 2.5 belongs also to the space \(L^p_{\xi }(\mathbb {M}^N_\varphi )\) with \(\xi \) as in (1.10).
In the following corollary we summarize the picture of uniqueness and non-uniqueness results in the special case of model manifolds.
Corollary 2.6
Let \(\mathbb {M}^N_{\varphi }\) be a model manifold such that (2.1) and (1.9) are satisfied. Suppose that assumption (\(H_3\)) holds and let \(b\in C^1(\mathbb {M}^N_{\varphi })\).
-
(i)
Assume that b satisfies assumption (\(H_2\)). Let u be a classical solution to equation (1.1). Let \(p>1\) and assume that \(pc_0\) large enough. If \(u\in L^p_{\xi }(\mathbb {M}^N_{\varphi })\), then \(u\equiv 0\) in \(\mathbb {M}^N_{\varphi }\).
-
(ii)
Assume that b satisfies assumption (2.2). Then equation (1.1) admits infinitely many classical solutions.
2.1 Notation from Riemannian Geometry
In this subsection we recall and fix some basic notations and useful definitions from Riemannian geometry that we will use in the rest of the paper (we refer e.g. to [1, 4, 7]).
Throughout the paper, M will denote an \(N-\)dimensional, complete, connected, noncompact, smooth Riemannian manifold without boundary and of infinite volume, endowed with a smooth Riemannian metric \(g=\lbrace g_{ij}\rbrace \). As usual, we consider the volume form, denoted by \(d\mu \), induced by g and we denote by \(\textrm{div}(X)\) the divergence of a smooth vector field X on M, that is, in local coordinates
where \(\vert g\vert =\textrm{det}(g_{ij})(\ge 0)\).
We also denote by \(\nabla \) and \(\Delta \) the Riemannian gradient and the Laplace-Beltrami operator on M with respect to g, that is, in local coordinates,
and
for any \(C^2-\)function \(u:M\rightarrow \mathbb {R}\), where \(\lbrace g^{ij}\rbrace \) denotes the inverse of the metric tensor g.
Due to the divergence structure of the Laplace-Beltrami operator we have the following integration by parts formula
for any \(C^2-\)functions \(v,u:M\rightarrow \mathbb {R}\), with either v or u compactly supported and where \(\langle \cdot ,\cdot \rangle \) is the scalar product induced by g.
Moreover, given a vector field X, we denote
Observe that, for any \(C^2-\)functions \(u:M\rightarrow \mathbb {R}\) and \(\psi :\mathbb {R}\rightarrow \mathbb {R}\) we have
Hence if u is a classical solution of (1.1) then
Lastly, we recall the definition of model manifold (see e.g. [5, 15, 18])
Definition 2.7
An \(N-\)dimensional model manifold with warping function \(\varphi \) is a Riemannian manifold \(\mathbb {M}^N_{\varphi }\) diffeomorphic to \(\mathbb {R}^N\) and endowed with a rotationally symmetric Riemannian metric. The model \(\mathbb {M}^N_{\varphi }\) is realized as the quotient space \([0,+\infty )\times \mathbb {S}^{N-1}/\backsim \), where \(\backsim \) identifies \(\lbrace 0\rbrace \times \mathbb {S}^{N-1}\) with the pole o of the space, and the Riemannian metric has the expression
where \(r(x)=\textrm{dist}(x,o)\), \(d\theta ^2\) denotes the standard metric of \(\mathbb {S}^{N-1}\) and \(\varphi :[0,+\infty )\rightarrow [0,+\infty )\) is a smooth function satisfying the following conditions:
-
\(\varphi (r)>0\) for all \(r>0\),
-
\(\varphi '(0)=1\),
-
\(\varphi ^{(2k)}(0)=0\) for every \(k\ge 0\).
Observe that the Euclidean space \(\mathbb {R}^N\) and the (generalized) hyperbolic space \(\mathbb {H}^N\) of constant curvature \(-c<0\) can be seen as model manifolds (take \(\varphi (r)=r\) and \(\varphi (r)=\frac{\sinh (\sqrt{c}r)}{\sqrt{c}}\), respectively). Moreover, due to their structure the Laplace-Beltrami operator on \(\mathbb {M}^N_{\varphi }\) can be written in the following way
where \(\Delta _\theta \) denotes the Laplace-Beltrami operator on \(\mathbb {S}^{N-1}\). In particular,
Moreover, being the volume form on \(\mathbb {M}^N_{\varphi }\) given by
one has the following:
where, \(c_N\) is the area of the \((N-1)-\)dimensional unit sphere and o is the pole of \(\mathbb {M}^N_{\varphi }\).
3 Proof of Theorem 2.2
Let M be a complete, noncompact, Riemannian manifold such that (1.5) is satisfied. Firstly, we state a general criterion for uniqueness of solutions to equation (1.1) in \(L^p_{\psi }(M)\) for \(\psi \) as in (1.6). We suppose that there exists a positive function \(\zeta \in C^1(M)\), which solves the following
for some \(\delta >0\) sufficiently large and for some \(p>1\). We mention that, \(\delta =\delta (\varepsilon )\) where \(\varepsilon >0\) will be specified in the proof of Theorem 2.2. Such inequality is meant in the sense of Definition 2.1-(ii), where instead of \(``="\) we have \(``<"\).
Proposition 3.1
Let M be a complete, noncompact, Riemannian manifold such that (1.5) is satisfied. Moreover, assume that (\(H_0\)) and (\(H_3\)) hold. Let u be a solution to equation (1.1). Assume that there exists a positive function \(\zeta \in C^1(M)\) such that \(\phi =e^{\zeta }\) satisfies
for some constant \(C>0\) and \(\psi \) as in (1.6). Moreover, assume that \(\zeta \) solves (3.1). If \(u\in L^p_\psi (M)\), then
3.1 Proof of Proposition 3.1
In order to show Proposition 3.1 we need the following
Lemma 3.2
Let M be a complete, noncompact, Riemannian manifold. Assume that (\(H_0\)) and (\(H_3\)) hold. Let \(\phi \in C^1(M)\), \(\phi >0\) be such that (3.2) holds. Let \(v\in L^1_{\psi }(M)\). Then
and
for a suitable family of cut-off functions \(\gamma _R\in C_c^{\infty }(B_R)\) and where \(D_+\) defined as in (1.4).
Remark 3.3
Observe that the quantity into the brackets of formula (3.4) is negative due to (1.4).
Proof of Lemma 3.2
Let us consider a cut-off function \(\gamma \in C^{\infty }([0,+\infty ))\), \(0\le \gamma \le 1\) such that
where r is the geodesic distance given by (1.3). Then, let us define
To show (3.3), let us observe that, for any \(x\in B_R\setminus B_{R/2}\), for some \(\tilde{C}>0\)
Therefore, by combining (3.2) with (3.6), we obtain
Then, since \(v\in L^1_{\psi }(M)\), from the latter we obtain (3.3) by letting \(R\rightarrow \infty \). On the other hand, to show (3.4), let us observe that, due to (\(H_0\)), for any \(x\in D_+\cap (B_R\setminus B_{R/2})\), for some \(\bar{C}>0\)
where, \(D_+\) has been defined in (1.4). Then, due to (3.2) together with (3.7), we get
Then, since \(v\in L^1_{\psi }(M)\), we obtain from (3.8) that
This completes the proof of the Lemma.
Proof of Proposition 3.1
For any \(\alpha >0\) small enough, we define
Then we take \(v\in C^1(M)\) such that
where \(\gamma _R\) has been defined in (3.5). Now, we consider left-hand side of equation (1.1) with \(u=G_{\alpha }(u)\), i.e.
Then, we multiply it by v, we integrate over M and we apply the integration by parts formula (2.3). We obtain
By using Young’s inequality, for every \(\varepsilon >0\), (3.11) reduces to
On the other hand, formula (3.10) can be reduced as follows
where we used (2.4), (2.5) and (1.1). From (3.12) and (3.13),
By rearranging the terms in (3.14), we get
On the left-hand side of (3.15), by substituting the definition of \(G_{\alpha }\) given in (3.9), we have
Similarly, on the right-hand side of (3.16), by exploiting the definition of \(G_{\alpha }\) given in (3.9) and by choosing \(\varepsilon >0\) sufficiently small, for any \(p>1\), we have that
since, by choosing \(0<\varepsilon \le 2-\frac{2}{p}\),
Hence, (3.15), combined with (3.16) and (3.17), reduces to
Letting \(\alpha \rightarrow 0^+\)
and
Hence, by the dominated convergence theorem, (3.18) becomes
which, defining \(\delta :=\frac{1}{\varepsilon }\), is equivalent to
Due to (3.19), and the definitions of \(D_+\) and \(D_-\) in (1.4), we obtain
Observing that, for all \(x\in D_-\)
we can rewrite (3.20) as follows
We now set \(\phi (x):=e^{\zeta (x)}\) for any \(x\in M\). Observe that, \(\phi \) satisfies assumption (3.2). Hence, we can apply Lemma 3.2 with this choice of \(\phi \). From Lemma 3.2 and the monotone convergence theorem, sending \(R\rightarrow \infty \) in (3.21) we get
From (3.22) and (3.1), since \(|u|^p\ge 0\), we can infer that \(u\equiv 0\) in M. This completes the proof. \(\square \)
We are now ready to prove Theorem 2.2, due to Proposition 3.1 it suffices to construct a subsolution to (3.1).
Proof of Theorem 2.2
Let \(\phi =\phi (r):=\psi (r)\) for any \(r>0\), where \(\psi \) has been defined in (1.6), with \(\beta >\alpha \). Moreover, we set \(\zeta (r):=-\beta r\). Hence, \(\phi (r)\equiv e^{\zeta (r)}\). At first observe that \(\phi \) satisfies (3.2). Moreover, \(\zeta \) solves (3.1), for properly chosen \(\beta >\alpha \). Indeed, due to (\(H_0\)) and (\(H_3\)), we have
Then, by (3.23), one has
provided that
Thus, by Proposition 3.1 the conclusion follows. \(\square \)
4 Proof of Theorem 2.3
Similarly to Theorem 2.2, we want to state a general criterion for uniqueness by means of assumption (\(H_1\)). Therefore, we state the next
Proposition 4.1
Let M be a complete, noncompact, Riemannian manifold such that (1.7) is satisfied. Moreover, assume that (\(H_1\)) and (\(H_3\)) hold. Let u be a solution to equation (1.1) and \(\eta \) be as in (1.8). Assume that there exists a positive function \(\zeta \in C^1(M)\) such that \(\phi =e^{\zeta }\) satisfies
and
Moreover, assume that \(\zeta \) solves (3.1). If \(u\in L^p_\eta (M)\), then
Analogously to Lemma 3.2 the next lemma can be shown. The proof is similar and for this reason we only sketch it.
Lemma 4.2
Let M be a complete, noncompact, Riemannian manifold such that (1.7) holds. Assume (\(H_1\)) and (\(H_3\)). Let \(\phi \in C^1(M)\), \(\phi >0\) be such that (4.1) and (4.2) hold. Let \(v\in L^1_{\eta }(M)\). Then
and
for a suitable family of cut-off functions \(\gamma _R\in C_c^{\infty }(B_R)\) and where \(D_+\) defined as in (1.4).
Proof
Equality (4.3) can be obtained as (3.3) by using assumption (4.1). To show (4.4) it is sufficient to argue as in the proof of Lemma 3.2 by replacing assumptions (\(H_0\)) and (3.2) with (\(H_1\)) and (4.2) respectively. \( \square \)
Proof of Proposition 4.1
To prove Proposition 4.1 it is sufficient to argue as in the proof of Proposition 3.1 by using Lemma 4.2 instead of Lemma 3.2.
Proof of Theorem 2.3
Let \(\phi =\phi (r):=\eta (r)\) for any \(r>0\), where \(\eta \) has been defined in (1.8). Moreover, we set \(\zeta (r):=-\beta r^{\theta }\), \(0<\theta <1\) as in (\(H_1\)). Hence, \(\phi (r)\equiv e^{\zeta (r)}\). At first observe that \(\phi \) satisfies (4.1) and (4.2). Moreover, \(\zeta \) solves (3.1), for properly chosen \(\beta >\alpha \). Indeed, due to (\(H_1\)) and (\(H_3\)), we have
Then, by (4.5), one has
provided that
Thus, by Proposition 4.1 the conclusion follows.
5 Proof of Theorem 2.4
Let M be a complete, noncompact, Riemannian manifold such that (1.9) is satisfied. Firstly, we state a general criterion for uniqueness of solutions to equation (1.1) in \(L^p_{\xi }(M)\) for \(\xi \) as in (1.10). We suppose that there exists a positive function \(\phi \in C^1(M)\), which solves, the following
for some \(\hat{\delta }>0\) sufficiently large and for some \(p>1\). We mention that, \(\hat{\delta }=\hat{\delta }(\varepsilon )\) wher \(\varepsilon >0\) will be specified in the proof of Theorem 2.4. Such inequality is meant in the sense of Definition 2.1-(ii), where instead of \(``="\) we have \(``<"\).
Proposition 5.1
Let M be a complete, noncompact, Riemannian manifold such that (1.9) is satisfied. Moreover, assume that (\(H_2\)) and (\(H_3\)) hold. Let u be a solution to equation (1.1) and \(\xi \) be as in (1.10). Assume that there exists a positive function \(\phi \in C^1(M)\), which solves (5.1) and such that
and
If \(u\in L^p_\xi (M)\), then
5.1 Proof of Proposition 5.1
Analogously to Lemma 3.2 and Lemma 4.2 the next lemma can be shown.
Lemma 5.2
Let M be a complete, noncompact, Riemannian manifold such that (1.9) holds. Assume (\(H_2\)) and (\(H_3\)). Let \(\phi \in C^1(M)\), \(\phi >0\) be such that (5.2) and (5.3) hold. Let \(v\in L^1_{\xi }(M)\). Then
and
for a suitable family of cut-off functions \(\gamma _R\in C_c^{\infty }(B_R)\) and where \(D_+\) defined as in (1.4).
Proof
Equality (5.4) can be obtained as (3.3) by using assumption (5.4). To show (5.5) it is sufficient to argue as in the proof of Lemma 3.2 by replacing assumptions (\(H_0\)) and (3.2) with (\(H_2\)) and (5.3) respectively. \( \square \)
Proof of Proposition 5.1
For any \(\alpha >0\) small enough, Let \(G_{\alpha }\) be defined as in (3.9). Then we take \(v\in C^1(M)\) such that
where \(\gamma _R\) has been defined in (3.5). Now, we consider the left-hand side of equation (1.1) with \(u=G_{\alpha }(u)\), i.e.
Then, we multiply it by v, we integrate over M and we use the integration by parts formula (2.3)
By Young’s inequality we obtain
On the other hand, arguing as in (3.13), (5.6) can be reduced as follows
where we have used (2.4), (2.5) and (1.1). Similarly to the proof of Proposition 3.1, by combining together (5.7) and (5.8), due to (3.9), we get
On the left-hand side of (5.9), by substituting the definition of \(G_{\alpha }\) given in (3.9), we have
Similarly, on the right-hand side of (5.9), by exploiting the definition of \(G_{\alpha }\) given in (3.9), arguing as in (3.17), we can choose \(\varepsilon >0\) sufficiently small such that, for any \(p>1\),
Hence, (5.9), combined with (5.10) and (5.11), reduces to
Letting \(\alpha \rightarrow 0^+\) in (5.12), by the dominated convergence and by choosing \(\hat{\delta }=\frac{2}{\varepsilon }\), we obtain the analogue of (3.19), i.e.
Due to (5.13), and the definitions of \(D_+\) and \(D_-\) in (1.4), we obtain
Observing that, for all \(x\in D_-\)
we rewrite (5.14) as follows
We now set \(\phi (x):=e^{\zeta (x)}\) for any \(x\in M\). Observe that, \(\phi \) satisfies assumptions (5.2) and (5.3). Hence, we can apply Lemma 5.2 with this choice of \(\phi \). From Lemma 5.2 and the monotone convergence theorem, sending \(R\rightarrow \infty \) in (5.15) we get
From (5.16) and (5.1), since \(|u|^p\ge 0\), we can infer that \(u\equiv 0\) in M. This completes the proof. \(\square \)
Proof of Theorem 2.4
Let \(\phi =\phi (r):=\xi (r)^{\frac{1}{2}}\) for any \(r>0\), where \(\xi \) has been defined in (1.10), with \(\tau >0\). At first observe that \(\phi ^2\) satisfies (5.2) and (5.3). Moreover, \(\phi \) solves (5.1). Indeed, due to (\(H_2\)) and (\(H_3\)), we have
Then, by (5.17), one has
provided that
Thus, by Proposition 5.1 the conclusion follows. \(\square \)
6 Proof of Proposition 2.5 and Corollary 2.6
We now show nonuniqueness for problem (1.1) with b as in (2.2) in the special case of M being a model manifold satisfying (2.1). From standard results, see e.g. [24, Theorem 2.5 and Proposition 2.7], the following Proposition can be proved.
Proposition 6.1
Let \(\mathbb {M}^N_\varphi \) be a model manifold such that (2.1) and (1.9) are satisfied. Let \(b:\mathbb {M}^N_\varphi \rightarrow \mathbb {R}^N\), \(b\in C^1(\mathbb {M}^N_\varphi )\) be such that assumption (2.2) holds. Moreover assume (\(H_3\)). Suppose that there exist
-
(i)
a bounded supersolution h of equation
$$\begin{aligned} \Delta h+\left\langle b,\,\nabla h\right\rangle - c h =-1 \quad \quad \text {in}\,\,\, \mathbb {M}^N_\varphi \setminus B_{R_0}\,, \end{aligned}$$(6.1)for some \(R_0>0\), such that
$$\begin{aligned} h>0 \quad \text { in }\,\, M, \quad \lim _{r\rightarrow +\infty } h(x)=0\,. \end{aligned}$$ -
(ii)
a positive bounded supersolution W of the equation
$$\begin{aligned} \Delta W+\left\langle b,\,\nabla W\right\rangle - c W =-1 \quad \quad \text {in}\,\,\, \mathbb {M}^N_\varphi \,. \end{aligned}$$(6.2)
Then there exist infinitely many bounded solutions u of problem (1.1). In particular, for any \(\gamma \in \mathbb R\), there exists a solution u to equation (1.1) such that
Proposition 6.1 shows that infinitely many solutions to equation (1.1) exist, if a supersolution h to (6.1) exists and if a superolution W to (6.2) exists. Therefore the following existence results, combined with the above proposition, imply nonuniqueness.
Lemma 6.2
Let assumptions (2.1), (1.9), (\(H_3\)) and (2.2) hold. Then there exists a supersolution \(h>0\) of equation (6.1).
Proof
We consider the function
where \(C>0\) and \(\beta >0\) have to be chosen. Note that, due to (2.6) and (2.1)
Thus, in view of (2.2), we have, for some \(C_{\beta }>0\).
for all \(x\in \mathbb {M}^N_\varphi \setminus B_{R_0}\), with \(R_0\) as in (2.2). Now, from (6.3) it follows that
provided that
Hence, by the assumption \(\sigma >1\), for \(\beta \) small enough, we get the thesis. \(\square \)
Lemma 6.3
Let assumption (\(H_3\)) holds. Then there exists a bounded supersolution \(W>0\) of equation (6.2).
Proof
We define the function
for \(c_0\) as in (\(H_3\)). Then, due to (\(H_3\))
for all \(x\in \mathbb {M}^N_\varphi \). The Lemma is proved. \(\square \)
Proof of Proposition 2.5
By simply combining Proposition 6.1, Lemmas 6.2 and 6.3, the result follows. \(\square \)
Proof of Corollary 2.6
(i) Due to (2.1), it follows that, for any \(B_R\subset M\),
Consequently, M satisfies assumption (1.9) for \(\alpha =(N-1)\lambda +1\). Therefore, since b satisfies (\(H_2\)), the thesis follows by Theorem 2.4.
(ii) The thesis directly follows by applying Proposition 2.5. \(\square \)
Data Availability
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
References
Alias, L.J., Mastrolia, P., Rigoli, M.: Maximum principles and geometric applications. Springer Monographs in Mathematics. Springer International Publishing, Berlin (2016)
Aronson, D.G., Besala, P.: Uniqueness of solutions to the Cauchy problem for parabolic equations. J. Math. Anal. Appl. 13, 516–526 (1966)
Eidelman, S.D., Kamin, S., Porper, F.: Uniqueness of solutions of the Cauchy problem for parabolic equations degenerating at infinity. Asympt. Anal. 22, 349–358 (2000)
Gallot, S., Hulin, D., Lafontaine, J.: Riemannian Geometry, 2nd edn. Universitext. Springer, Berlin (1990)
Greene, R., Wu, H.H.: Function theory on manifolds which possess a pole. Lecture Notes in Mathematics, vol. 699. Springer, Berlin (1979)
Grigoryan, A.: Bounded solutions of the Schrödinger equation on non-compact Riemannian manifolds (in Russian) Trudy seminaria I.G. Petrovskogo 14 (1989), 66–77. Engl. transl: J. Soviet Math. 51 no. 3 (1990), 2340-2349
Grigoryan, A.: Analytic and geometric background of recurrence and non-explosion of the Brownian motion on Riemannian manifolds. Bull. Am. Math. Soc. 36, 134–249 (1999)
Il’in, A.M., Kalashnikov, A.S., Oleinik, O.A.: Linear equations of the second order of parabolic type. Russian Math. Surveys 17, 1–144 (1962)
Kamin, S., Pozio, M.A., Tesei, A.: Admissible conditions for parabolic equations degenerating at infinity. St. Petersburg Math. J. 19, 239–251 (2008)
Li, P.: Uniqueness of \(L^1\) solutions for the Laplace equation and the heat equation on Riemannian manifolds. J. Diff. Geom. 20, 447–457 (1984)
Li, P., Schoen, R.: \(L^p\) and mean value properties of subharmonic functions on Riemannian manifolds. Acta Math. 153(3–4), 279–301 (1984)
Meglioli, G., Punzo, F.: Uniqueness for fractional parabolic and elliptic equations with drift. Preprint (2022)
Monticelli, D.D., Punzo, F.: Distance from submanifolds with boundary and applications to Poincaré inequalities and to elliptic and parabolic problems. J. Diff. Equ. 267, 4274–4292 (2019)
Naber, A.: Noncompact shrinking four solitons with nonnegative curvature. J. Reine Angew. Math. 645, 125–153 (2010)
Petersen, P.: “Riemannian Geometry”, 3rd edition, Graduate Texts in Mathematics 171. Springer (2016)
Petersen, P., Wylie, W.: Rigidity of gradient Ricci solitons. Pacific J. Math. 241(2), 329–345 (2009)
Pigola, S., Rigoli, M., Setti, A.G.: Vanishing theorems on Riemannian manifolds, and geometric applications. J. Funct. Anal. 229(2), 424–461 (2005)
Pigola, S., Setti, A.G.: Global divergence theorems in nonlinear PDEs and geometry. En. Mat. 26, 1–77 (2014)
Pozio, M.A., Punzo, F., Tesei, A.: Criteria for well-posedness of degenerate elliptic and parabolic problems. J. Math. Appl. 90, 353–386 (2008)
Pozio, M.A., Tesei, A.: On the uniqueness of bounded solutions to singular parabolic problems. Discrete Contin. Dyn. Syst. Ser. A 13, 117–137 (2005)
Punzo, F.: Uniqueness of solutions to degenerate parabolic and elliptic equations in weighted Lebesgue spaces. Math. Nachr. 286, 1043–1054 (2013)
Punzo, F.: Uniqueness for the heat equation in Riemannian manifolds. J. Math. Anal. Appl. 424, 402–422 (2015)
Punzo, F.: Integral conditions for uniqueness for solutions to degenerate parabolic equations. J. Diff. Equ. 267, 6555–6573 (2019)
Punzo, F., Tesei, A.: Uniqueness of solutions to degenerate elliptic problems with unbounded coefficients. Annal. Inst. Henri Poincaré (C) Non Linear Anal. 26, 2001–2024 (2009)
Punzo, F., Valdinoci, E.: Uniqueness in weighted Lebesgue spaces for a class of fractional parabolic and elliptic equations. J. Diff. Equ. 258, 555–587 (2015)
Punzo, F., Valdinoci, E.: Prescribed conditions at infinity for fractional parabolic and elliptic equations with unbounded coefficients. ESAIM: COCV 24, 105–127 (2018)
Tihonov, A.N.: Théorèmes d’unicité pour l’équation de la chaleur. Mat. Sb. 42, 199–215 (1935)
Wu, J.Y.: \(L^p-\)Liouville theorems on complete smooth metric measure spaces. Bull. Sci. Math. 138, 510–539 (2014)
Yau, S.T.: Harmonic functions on complete Riemannian manifolds. Comm. Pure Appl. Math. 28, 201–228 (1975)
Yau, S.T.: Some function-theoretic properties of complete Riemannian manifold and their applications to geometry. Indiana Univ. Math. J. 25(7), 659–670 (1976)
Acknowledgements
The authors have been partially supported by the “Gruppo Nazionale per l’Analisi Matematica, la Probabilitá e le loro Applicazioni” (GNAMPA) of the “Istituto Nazionale di Alta Matematica” (INdAM, Italy). The first author is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - SFB 1283/2 2021 - 317210226. The second author has been partially supported by the GNAMPA project 2023 “Equazioni di diffusione e analisi geometrica”.
Funding
Open access funding provided by Politecnico di Milano within the CRUI-CARE Agreement.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Meglioli, G., Roncoroni, A. Uniqueness in Weighted Lebesgue Spaces for an Elliptic Equation with Drift on Manifolds. J Geom Anal 33, 320 (2023). https://doi.org/10.1007/s12220-023-01378-8
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s12220-023-01378-8