Abstract
Let \(M\) be a closed Riemannian manifold with a Riemannian metric \(g_{ij}(t)\) evolving by a geometric flow \(\partial _{t}g_{ij} = -2{S}_{ij}\), where \(S_{ij}(t)\) is a symmetric two-tensor on \((M, g(t))\). Suppose that \(S_{ij}\) satisfies the tensor inequality \(2{\mathcal H}(S, X)+{\mathcal E}(S,X) \ge 0\) for all vector fields \(X\) on \(M\), where \({\mathcal H}(S, X)\) and \({\mathcal E}(S,X)\) are introduced in Definition 1 below. Then, we shall prove differential Harnack estimates for positive solutions to time-dependent forward heat equations with potentials. In the case where \(S_{ij} = R_{ij}\), the Ricci tensor of \(M\), our results correspond to the results proved by Cao and Hamilton (Geom Funct Anal 19:983–989, 2009). Moreover, in the case where the Ricci flow coupled with harmonic map heat flow introduced by Müller (Ann Sci Ec Norm Super 45(4):101–142, 2012), our results derive new differential Harnack estimates. We shall also find new entropies which are monotone under the above geometric flow.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The main purpose of the current article is to study time-dependent heat equations with potentials on closed Riemannian \(n\)-manifolds \(M\) evolving by the geometric flow
where \(S_{ij}(t)\) is a symmetric two-tensor on \((M, g(t))\). A typical example would be the case where \(S_{ij} = R_{ij}\) is the Ricci tensor and \(g(t)\) is a solution of the Ricci flow introduced by Hamilton [8]. We shall derive Li–Yau type differential Harnack inequalities [13] for positive solutions to the following heat equation with a potential term:
where the symbol \(\Delta \) stands for the Laplacian of the evolving metric \(g(t)\) and \(S ={g}^{ij}S_{ij}\) is the trace of \(S_{ij}\). In the current article, we shall use the Einstein summation convention. For simplicity, we omit \(g(t)\) in the above notation. All geometric operators are with respect to the evolving metric \(g(t)\). Notice also that we have
The main results of the current article are Theorems A, B, C, D and E stated below, which can be seen as natural generalizations of results proved by Cao and Hamilton [4]. See also Sect. 2 below.
The study of differential Harnack inequalities for parabolic equations originated with the work of Li and Yau [13]. They first proved a gradient estimate for the heat equation using the maximal principle. By integrating the gradient estimate along a space-time path, a classical Harnack inequality was derived. Therefore, Li–Yau type gradient estimate is often called differential Harnack inequality. Hamilton adapted similar techniques to prove Harnack inequalities for the Ricci flow [10] and the mean curvature flow [11]. Many authors used similar techniques to prove Harnack inequalities for geometric flows. For instance, see [1–3, 5–7, 12, 17–19].
To state the main results of the current article, we shall introduce evolving tensor quantities associated with the tensor \({S}_{ij}\).
Definition 1
Suppose that \(g(t)\) evolves by the geometric flow (1) and let \(X=X^{i}\frac{\partial }{\partial x^{i}} \in \Gamma (TX)\) be a vector field on \(M\). We define
where \(R^{ij} = {g}^{ik}g^{j \ell }R_{k \ell }\), \(S^{ij} = {g}^{ik}g^{j \ell }S_{k \ell }\), \(S = g^{ij}S_{ij}\), \({\nabla }^{i} = g^{ij}{\nabla }_{j}\) and \(X_{k} = {g}_{ik}X^{i}\).
We notice that these quantities are also introduced by Müller [15] to prove the monotonicity of Perelman type reduced volume under (1).
The first main result of the current article is as follows:
Theorem A
Suppose that \(g(t)\) evolves by the geometric flow (1) on a closed oriented smooth \(n\)-manifold \(M\) and
holds for all vector fields \(X\) and all time \(t \in [0, T)\) for which the flow exists. Let \(f\) be a positive solution to the heat equation (2), \(u = -\log f\), and
Then, \(H_{S} \le 0\) for all time \(t \in (0, T)\).
In the case of the heat equation without the potential term, we shall also prove
Theorem B
Suppose that \(g(t)\) evolves by the geometric flow (1) on a closed oriented smooth \(n\)-manifold \(M\) and
holds for all vector fields \(X\) and all time \(t \in [0, T)\) for which the flow exists. Let \(f \ (< 1)\) be a positive solution to the heat equation \(\frac{\partial f}{\partial t} = \Delta f\), \(u = -\log f\) and
Then, \(H_{S} \le 0\) holds for all time \(t \in (0, T)\). Hence we have the following on \((0, T)\),
On the other hand, a similar technique with the proof of Theorem A also enables us to prove
Theorem C
Suppose that \(g(t)\) evolves by the geometric flow (1) on a closed oriented smooth \(n\)-manifold \(M\) and \(2{\mathcal H}(S, X) + {\mathcal E}(S,X) \ge 0\) holds for all vector fields \(X\) and all time \(t \in [0, T)\) for which the flow exists. Let \(f\) be a positive solution to the heat equation (2), \(v = -\log f- \frac{n}{2}\log (4{\pi }t)\) and
where \(d\) is any constant. Then for all time \(t \in (0, T)\),
and \(\max (t{P}_{S})\) is non-increasing.
Moreover, inspired by the works of Perelman [17] and Cao and Hamilton [4], we shall introduce two entropies which are associated with the above Harnack quantities. The first one is associated with \(H_{S}\).
Theorem D
Under the same assumption with Theorem A, we define
Then for all time \(t \in (0, T)\), we have \({\mathcal F}_{S} \le 0\) and
Moreover, suppose that \({\mathcal H}(S, X) \ge 0\) and \({\mathcal E}(S, X) \ge 0\) holds for all vector fields \(X\) and all time \(t \in [0, T)\) for which the flow exists. If \(\frac{\mathrm{d}}{{\mathrm{d}}t} {\mathcal F}_{S} = 0\) holds for some time \(t\), then the following holds:
The second one is associated with \(P_{S}\) as follows.
Theorem E
Under the same assumption with Theorem C, we define
Then for all time \(t \in (0, T)\), we have
Moreover, suppose that \({\mathcal H}(S, X) \ge 0\) and \({\mathcal E}(S,X) \ge 0\) holds for all vector fields \(X\) and all time \(t \in [0, T)\) for which the flow exists. If \(\frac{\mathrm{d}}{{\mathrm{d}}t} {\mathcal W}_{S} = 0\) holds for some time \(t\), then the following holds:
2 Examples
2.1 The Ricci flow
Let \(g(t)\) be a solution to the Ricci flow:
Namely, we have \(S_{ij} = {R}_{ij}\) and \(S = R\) the scalar curvature. Notice that it is known that the scalar curvature \(R\) evolves by \(\frac{\partial R}{\partial t} - \Delta R - 2|R_{ij}|^{2} = 0\). Moreover, we have the twice contracted second Bianchi identity \(2{\nabla }^{i}R_{i\ell } - {\nabla }_{\ell }R=0\). Hence, we have \({\mathcal E}(S, X) =0\) in this case. Therefore, (3) is equivalent to
This tells us that \(g(t)\) has weakly positive curvature operator (see also [9]). Moreover, we have \({\mathcal I}(S, X) =0\). Hence, (4) holds. Therefore, Theorems A, B, C, D and E in the case where \(S_{ij} = {R}_{ij}\) just correspond to the results proved by Cao and Hamilton [4]. Notice also that (5) particularly tells us that \(R_{ij} = -\frac{1}{t}g_{ij}\), i.e., \(g(t)\) is Einstein. Similarly, (6) implies \(R_{ij} +{\nabla }_{i}{\nabla }_{j}(-v) + \frac{1}{2t}g_{ij}=0\). This tells us that \(g(t)\) is an expanding gradient Ricci soliton. Since it is known [17] that any expanding Ricci soliton on a closed manifold must be Einstein, \(g(t)\) is Einstein.
2.2 Bernhard List’s flow
List [14] introduced a geometric flow closely related to the Ricci flow:
where \(\psi : M \rightarrow {\mathbb R}\) is a smooth function. If we set \(S_{ij} = {R}_{ij}-2{\nabla }_i \psi {\nabla }_j \psi \), it is clear that the first of List’s flow has the form (1). Notice also that we have \(S =R -2|\nabla \psi |^{2}\). List [14] pointed out that \(S\) satisfies the following evolution equation:
On the other hand, we get \(2{\nabla }^{i}S_{i\ell } - {\nabla }_{\ell }S = 2{\nabla }^{i}({R}_{i\ell }-2{\nabla }_i \psi {\nabla }_\ell \psi )-{\nabla }_{\ell }(R -2|\nabla \psi |^{2}) = -4{\nabla }^{i}({\nabla }_i \psi {\nabla }_\ell \psi ) - 2{\nabla }_{\ell }(\nabla ^{k} \psi \nabla _{k} \psi ) = -4{\Delta }\psi \nabla _{\ell } \psi \), where notice the twice contracted second Bianchi identity \(2{\nabla }^{i}R_{i\ell } = {\nabla }_{\ell }R\). Therefore, we have
In particular, (3) is particularly satisfied if
On the other hand, we have \({\mathcal I}(S, X) = {\nabla }^i \psi {\nabla }^j \psi X_{i}X_{j} = ({\nabla }_{X}\psi )^{2} \ge 0\). Hence, (4) holds. Therefore, Theorems A, B, C, D and E just correspond to the result proved by Fang [6]. Notice also that, under the situation on Theorem D, we particularly have the following by (5) and (7):
Since it follows that \(\Delta \psi =0\), \(\psi \) must be a harmonic function on the closed manifold \(M\). This implies that \(\psi \) is a constant for the time \(t\). Therefore, we have \({R}_{ij}= -\frac{1}{t}g_{ij}\), i.e., \(M\) is Einstein.
2.3 Rent Müller’s flow
Let \((Y, h)\) be a fixed Riemannian manifold. Let \((g(t), \phi (t))\) be the couple consisting of a family of metric \(g(t)\) on \(M\) and a family of maps \(\phi (t)\) from \(M\) to \(Y\). We call \((g(t), \phi (t))\) a solution of Rent Müller’s flow [16] (also known as the Ricci flow coupled with harmonic map heat flow) with coupling function \(\alpha (t) \ge 0\) if
where \( {\tau } _g \phi \) is the tension field of the map \(\phi \) with respect to the metric \(g(t)\). List’s flow is a special case of this flow. If we set \(S_{ij} = {R}_{ij}-\alpha (t){\nabla }_i \phi {\nabla }_j \phi \), the first of Müller’s flow has the form (1). Notice that \(S =R -2\alpha (t)|\nabla \phi |^{2}\) holds. Müller [16] proved that \(S\) satisfies
Since we are able to get \(2{\nabla }^{i}S_{i\ell } - {\nabla }_{\ell }S = -2\alpha (t){\tau }_{g}\phi \nabla _{\ell } \phi \), the following holds:
Therefore, \({\mathcal E}(S, X) \ge 0\) holds if \(\alpha (t) \ge 0\) is non-increasing. In this case, (3) is particularly satisfied if
Notice also that \({\mathcal I}(S, X) = \alpha (t){\nabla }^i \phi {\nabla }^j \phi X_{i}X_{j} = \alpha (t)({\nabla }_{X}\phi )^{2} \ge 0\). Hence, (4) holds. To the best of our knowledge, Theorems A, B, C, D and E in the case where \(S_{ij} = {R}_{ij}-\alpha (t){\nabla }_i \phi {\nabla }_j \phi \) are new.
On the other hand, under the situation on Theorem D, we have the following by (5) and the above computation if \(\alpha (t)\) is constant:
Therefore, we have \({R}_{ij}-2{\nabla }_i \phi {\nabla }_j \phi = -\frac{1}{t}g_{ij}\) and \(\tau _{g} \phi = 0\). In particular, \(\phi \) must be a harmonic map.
3 Proofs of Theorems A and B
Let \(f\) be a positive solution of the following heat equation with potential:
where \(c\) is a constant. In what follows, let \(u = - \log f\). By a direct computation, we are able to see that \(u\) satisfies
Let us introduce the following:
Definition 2
Suppose that \(g(t)\) evolves by (1) and let \(S\) be the trace of \(S_{ij}\). Let \(X=X^{i}\frac{\partial }{\partial x^{i}} \in \Gamma (TX)\) be a vector field on \(M\). We define
where \(a, \alpha \) and \(\beta \) are constants.
Notice that we have \({\mathcal E}(S, {X}) = {\mathcal D}_{(1, -2, 1)}(S, X)\).
Lemma 1
Let \(g(t)\) be a solution to the geometric flow (1) and \(u\) satisfies (9). Let
where \(\alpha , \beta , a, b\) and \(d\) are constants. Then, \(H_{S}\) satisfies
Proof
First of all, notice that we have the following three evolution equations, which follow from standard computation:
By (9), (10) and these equations, we are able to obtain
On the other hand, we also have the following by (10):
Therefore, we get
Using this, we are able to obtain
On the other hand, we also have the following Bochner–Weitzenbock type formula:
Using this formula, we get
where notice that Definition 2. \(\square \)
In particular, we shall use Lemma 1 to prove Theorem B. The following result is used to prove Theorem A.
Proposition 1
The evolution equation in Lemma 1 can be rewritten as follows:
where \(\lambda \) is a constant, \(\alpha \not = 0\) and \(\alpha \not = \beta \).
Proof
A direct computation implies
Therefore, we get
By this and Lemma 1, we obtain
The desired result now follows from the above equation and the following:
This equation also follows from a direct computation. \(\square \)
As a corollary of the above proposition, we obtain the following result which is a key to prove Theorem A:
Corollary 1
Suppose that \(g(t)\) evolves by the geometric flow (1) on a closed oriented smooth \(n\)-manifold \(M\). Let \(f\) be a positive solution to the heat equation (8) with \(c=-1\), \(u = -\log f\) and
Then,
Proof
By Proposition 1 in the case where \(\alpha = 2\), \(\beta = 1\), \(a = -3\), \(c=-1\), \(\lambda = 2\), \(b = 0\) and \(d= 2\), we get the desired result as follows:
where we used Definition 2. \(\square \)
We are now in a position to prove Theorem A. First of all, notice that, for \(t\) small enough, we get \(H_{S} < 0\). Since we assumed that (3) holds, the maximal principle and Corollary 1 tell us that
for all time \(t \in (0, T)\). Hence, we have proved Theorem A.
By Theorem A and integrating along a space-time path, we are able to get a classical Harnack inequality as follows:
Corollary 2
Suppose that \(g(t)\) evolves by the geometric flow (1) on a closed oriented smooth \(n\)-manifold \(M\) and \(2{\mathcal H}(S, X)+{\mathcal E}(S,X) \ge 0\) holds for all vector fields \(X\) and all time \(t \in [0, T)\) for which the flow exists. Let \(f\) be a positive solution to the heat equation (2). Assume that \((x_{1}, t_{1})\) and \((x_{2}, t_{2})\) are two points in \(M \times (0, T)\), where \(0 < t_{1} < {t}_{2}\). Let
where \(\ell \) is any space-time path joining \((x_{1}, t_{1})\) and \((x_{2}, t_{2})\). Then,
Proof
The strategy of the proof is now standard. For the reader, let us include the proof. First of all, we have \(H_{S} \le 0\) by Theorem A. And \(u = -\log f\) satisfies (9) with \(c=-1\), i.e.,
Therefore, we get
Pick a space-time path \(\ell (x, t)\) joining \((x_{1}, t_{1})\) and \((x_{2}, t_{2})\). Then, we obtain the following along the path \(\ell (x, t)\) using (12):
This implies
This tells us that (11) holds. \(\square \)
Let us close this section with the proof of Theorem B. Let \(f\) be a positive solution to linear heat equation \(\frac{\partial f}{\partial t} = \Delta f\). Then, we may assume that \(f< 1\) by the linearity. Then, \(u = -\log f\) satisfies (9) with \(c=0\). Therefore, by taking \(\alpha = 0\), \(\beta = -1\), \(a = c =0\), \(\lambda = 2\), \(b = 1\) and \(d= 0\) in Lemma 1, we have
and
Notice that as \(t\) small enough, \(H_{S}<0\). By the maximal principle and this evolution equation, Theorem B follows as desired.
4 Proof of Theorem C
Let \(f\) be a positive solution of (8). In what follows, let \(v = - \log f-\frac{n}{2}\log (4{\pi }t)\). By a direct computation, we see that \(v\) satisfies
Then, we have
Proposition 2
Let \(g(t)\) be a solution to the geometric flow (1) and \(v\) satisfies (13). Let
where \(\alpha , \beta , a, b\) and \(d\) are constants. Then, \(P_{S}\) satisfies
where \(\lambda \) is a constant, \(\alpha \not =0\) and \(\alpha \not = \beta \).
Proof
A similar computation with Proposition 1 enables us to prove this result. In fact, notice that we have \(v = u - \frac{n}{2}\log (4{\pi }t)\). Therefore, we get \(\nabla u = \nabla v\) and \(\Delta u = \Delta v\). We also have \(P_{S} =H_{S} + \frac{bn}{2t}\log (4{\pi }t)\). Then, Proposition 1 and a direct computation imply
Hence we obtained the desired result. \(\square \)
As a special case of Proposition 2, we get
Corollary 3
Suppose that \(g(t)\) evolves by the geometric flow (1) on a closed oriented smooth \(n\)-manifold \(M\). Let \(f\) be a positive solution to the heat equation (8) with \(c= -1\), \(v = - \log f-\frac{n}{2}\log (4{\pi }t)\) and
Then,
\(\square \)
Proof
By Proposition 2 in the case where \(\alpha = 2\), \(\beta = 1\), \(a = -3\), \(b = -1\), \(c=-1\), \(\lambda = 1\), we obtain
Therefore, the desired result follows. \(\square \)
We shall prove Theorem C as follows. In fact, we are able to obtain the following by Corollary 3:
Furthermore, the monotonicity of \(\max (t{P}_{S})\) follows from this equation and the maximal principle.
5 Proof of Theorem D
First of all, notice that \({\mathcal F}_{S} \le 0\) follows from the definition of \({\mathcal F}_{S}\) and \(H_{S} \le 0\), where we used Theorem A. Let us consider the following quantity:
On the other hand, a direct computation tells us that the following holds:
By (9) with \(c=-1\), Corollary 1 and (15), we get
On the other hand, notice that we have
Therefore, the following holds:
Assume moreover that \({\mathcal H}(S, X) \ge 0\) and \({\mathcal E}(S, X) \ge 0\) holds. Suppose also that \(\frac{\mathrm{d}}{{\mathrm{d}}t} {\mathcal F}_{S} = 0\) holds for some time \(t\). Then, we obtain
These imply (5) as desired. We proved Theorem D.
6 Proof of Theorem E
Let us consider the following quantity:
A direct computation tells us that the following holds:
By (13) with \(c=-1\), Corollary 3 and (16), we obtain the following:
By a direct computation, we also have
Therefore, we obtain
Assume moreover that \({\mathcal H}(S, X) \ge 0\) and \({\mathcal E}(S, X) \ge 0\). Suppose also that \(\frac{\mathrm{d}}{{\mathrm{d}}t} {\mathcal W}_{S} =0\) for some time \(t\). Then, we obtain
Hence, we have proved Theorem E.
References
Andrews, B.: Harnack inequalities for evolving hypersurfaces. Math. Z. 217, 179–197 (1994)
Cao, H.-D.: On Harnack’s inequalities for the Kahler–Ricci flow. Invent. Math. 109, 247–263 (1992)
Cao, H.-D., Ni, L.: Matrix Li–Yau–Hamilton estimates for the heat equation on Kahler manifolds. Math. Ann. 331, 795–807 (2005)
Cao, H.-D., Hamilton, R.: Differential Harnack estimates for time-dependent heat equations with potentials. Geom. Funct. Anal. 19, 983–989 (2009)
Chow, B.: On Harnack’s inequality and entropy for the Gaussian curvature flow. Comm. Pure Appl. Math. 44, 469–483 (1991)
Fang, S.: Differential Harnack inequalities for heat equations with potentials under the Bernhard List’s flow. Geom. Dedicata. 161, 11–22 (2012)
Guenther, C.M.: The fundamental solution on manifolds with time-dependent metrics. J. Geom. Anal. 12, 425–436 (2002)
Hamilton, R.: Three manifolds with positive Ricci curvature. J. Differ. Geom. 17, 255–306 (1982)
Hamilton, R.: Four-manifolds with positive curvature operator. J. Differ. Geom. 24, 153–179 (1986)
Hamilton, R.: The Harnack estimate for the Ricci flow. J. Differ. Geom. 37, 225–243 (1993)
Hamilton, R.: Harnack estimate for the mean curvature flow. J. Differ. Geom. 41, 215–226 (1995)
Ni, L.: Monotonicity and Li–Yau–Hamilton inequalities. In: Surveys in differential geometry. Geometric flows, pp. 251–301, Surv. Differ. Geom., Vol. 12. Int. Press, Somerville (2008)
Li, P., Yau, S.-T.: On the parabolic kernel of the Schrodinger operator. Acta Math. 156, 153–201 (1986)
List, B.: Evolution of an extended Ricci flow system. PhD thesis, AEI Potsdam (2005)
Müller, R.: Monotone volume formulas for geometric flows. J. Reine Angew. Math. 643, 39–57 (2010)
Müller, R.: Ricci flow coupled with harmonic map flow. Ann. Sci. Ec. Norm. Super. 45(4), 101–142 (2012)
Perelman, G.: The entropy formula for the Ricci flow and its geometric applications, math.DG/0211159 (2002)
Sun, J.: Gradient estimates for positive solutions of the heat equation under geometric flow. Pac. J. Math. 253, 489–510 (2011)
Qi Zhang, S.: Some gradient estimates for the heat equation on domains and for an equation by Perelman. Int. Math. Res. Not. 2006, Art. ID 92314
Acknowledgments
The author would like to thank the Max-Plank-Institut für Mathematik in Bonn for its hospitality during the inception of this work. This work was partially supported by the Grant-in-Aid for Scientific Research (C), Japan Society for the Promotion of Science, No. 20540090.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Ishida, M. Geometric flows and differential Harnack estimates for heat equations with potentials. Ann Glob Anal Geom 45, 287–302 (2014). https://doi.org/10.1007/s10455-013-9401-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10455-013-9401-1