Estimates for solutions of Dirac equations and an application to a geometric elliptic-parabolic problem (Q1731777): Difference between revisions

Let $E$ be a Dirac bundle over a Riemannian manifold $M^m$ ($m \ge 2$) with boundary $\partial M$, and let $\nabla_0$ be a fixed smooth Dirac connection on $E$. Recall that a Dirac bundle is a Hermitian vector bundle of left Clifford modules over the Clifford bundle $\text{Cl}(M)$ such that multiplication by unit vectors in $TM$ is orthogonal and the covariant derivative is a module derivation. We define the Dirac connection space $\mathfrak{D}^p(E)$ by the completetion of subspace of $\Omega^1(\mathrm{Ad}(E))$ commuting with Clifford multiplication with respect to the norm $\Vert\Gamma\Vert_p := \Vert\Gamma\Vert_{L^{2p}(M)} + \Vert d\Gamma\Vert_{L^p(M)}$ so that $\nabla= \nabla_0 + \Gamma$ becomes another Dirac connection on $E$ for $\Gamma \in \mathfrak{D}^p(E)$. The Dirac operator associated with the Dirac connection $\nabla$ is defined by ${D\!\!\!\!/} := e_i \cdot \nabla_{e_i}$, where $e_i\cdot$ denotes Clifford multiplication, and $\{e_i\}$ is a local orthonormal frame on $M$. In this paper, the authors first consider the existence and uniqueness for Dirac operators under a class of local elliptic boundary value conditions $\mathcal{B}$ including chiral conditions, MIT bag boundary conditions and J-boundary conditions: \[ \begin{cases} {D\!\!\!\!/} \psi = \varphi \text{in }M\\ \mathcal{B}\psi = \mathcal{B}\psi_0 \text{on }\partial M, \end{cases} \] where $\varphi \in L^p(M)$ and $\mathcal{B}\psi_0\in W^{1-1/p, p}(E|_{\partial M})$. The Sobolev spaces of sections of $E$ are associated with the fixed smooth Dirac connection $\nabla_0,$ and assume $p^*1$ if $m=2$ and $p^*\ge (3m-2)/4$ if $m 2$. The authors prove that if $\Gamma \in \mathfrak{D}^{p^*}(E)$, then for any $1 pp^*$ the above Dirac equation with boundary condition admits a unique solution $\psi \in W^{1,p}(E)$, and $\psi$ satisfies the following estimate \[ \Vert \psi\Vert_{W^{1,p}(E)} \le c (\Vert \varphi\Vert_{L^p(E)} + \Vert\mathcal{B}\psi_0\Vert_{W^{1-1/p,p}(E|_{\partial M})}), \] where $c = c(p, \Vert \Gamma\Vert_{p^*})0$. This estimate is optimal in dimension 2 in the sense that the exponents cannot be improved. It also improves the known estimates in higher dimensions in [\textit{R. A. Bartnik} and \textit{P. T. Chruściel}, J. Reine Angew. Math. 579, 13--73 (2005; Zbl 1174.58305)] and [\textit{D. Jerison}, Adv. Math. 62, 118--134 (1986; Zbl 0627.35008)]. Applying this result, the authors also derive the existence and uniqueness for boundary value problems for Dirac operators along a map. Let $M$ be a compact Riemannian spin manifold with boundary $\partial M$, $N$ be a compact Riemannian manifold and $\Phi: M\to N$ be a smooth map. Given a fixed spin structure on $M$, let $\Sigma M$ be the spin bundle of $M$. On the twisted bundle $\Sigma M \otimes \Phi^{-1}TN$, the Dirac operator ${D\!\!\!\!/}$ along a map $\Phi$ is defined by ${D\!\!\!\!/}\Psi := {\partial\!\!\!/}\psi^\alpha \otimes \theta_\alpha + e_i\cdot \psi^\alpha \otimes \nabla_{\Phi_*(e_i)}^{TN} \theta_\alpha$, where $\Psi= \psi^\alpha\otimes \theta_\alpha$, $\{\theta_\alpha\}$ are local cross-section of $\Phi^{-1}(TN)$, ${\partial\!\!\!/} = e_i\cdot \nabla_{e_i}$ is the usual Dirac operator on the spin bundle over $M$. We say that $\Psi$ is a harmonic spinor along the map $\Phi$ if ${D\!\!\!\!/}\Psi = 0$. The second main result in this paper is the following. Under these situations, if $\Phi \in W^{1, 2p^*}(M;N)$, then for any $1pp^*, \eta \in L^p(M; \Sigma M\otimes \Phi^{-1}TN)$ and $\mathcal{B}\psi \in W^{1-1/p, p}(\partial M; \Sigma M \otimes \phi^{-1}TN)$, the boundary value problem for the Dirac equation \[ \begin{cases} {D\!\!\!\!/} \Psi = \eta \text{in }M\\ \mathcal{B}\Psi = \mathcal{B}\psi \text{on }\partial M, \end{cases} \] admits a unique solution $\Psi\in W^{1,p}(M; \Sigma M \otimes \Phi^{-1}TN)$, and \[ \Vert \Psi\Vert_{W^{1,p}(M)} \le c (\Vert \eta\Vert_{L^p(M)} + \Vert\mathcal{B}\psi\Vert_{W^{1-1/p,p}(\partial M)}). \] Finally, the authors introduce the notion of Dirac-harmonic map heat flow and obtian the local existence and uniqueness of this heat flow for Dirac-harmonic maps by applying the elliptic estimates for Dirac equations with boundary conditions mentioned above. More precisely, consider the functional \[ L(\Phi, \Psi) = \frac{1}{2}\int_M \left(\Vert d\Phi\Vert^2 + (\Psi, {D\!\!\!\!/}\Psi)\right), \] where $(\,\,,\,\,) = \mathrm{Re} \langle \,\,,\,\,\rangle$ is the real part of the Hermitian inner product on $\Sigma M \otimes \Phi^{-1}TN$. A Dirac-harmonic map is defined to be a critial point $(\Phi, \Psi)$ of $L$. The Euler-Lagrange equations for this functional are \[ \tau(\Phi) = \frac{1}{2}(\psi^\alpha, e_i \cdot \psi^\beta) R^N(\theta_\alpha, \theta_\beta)\Phi_*(e_i) =:\mathcal{R}(\Phi, \Psi) \quad \text{and}\quad, {D\!\!\!\!/}\Psi = 0. \] Consider the following flow for Dirac-harmonic maps: for $\Phi \in C^{2, 1.\alpha} (M \times (0, T]; N)$ and $\Psi \in C^{1, 0.\alpha}(M\times [0, T]; \Sigma M \times \Phi^{-1}TN)$, \[ \begin{cases} \partial_t \Phi = \tau(\Phi) - \mathcal{R}(\Phi, \Psi) \text{in }M\times (0, T]\\ {D\!\!\!\!/} \Psi =0 \text{in } M\times [0, T] \end{cases} \] with the boundary-initial data \[ \begin{cases} \Phi = \phi \text{in }M \times \{0\}\cup \partial M \times [0, T]\\ \mathcal{B}\Psi = \mathcal{B}\psi \text{on }\partial M\times [0, T], \end{cases} \] where $\phi \in C^{2, 1.\alpha}(M \times \{0\}\cup \partial M\times [0, T])$ and $\psi \in C^{1, 0.\alpha}(\partial M\times [0, T]; \Sigma M \times \phi^{-1}TN)$. Here $f \in C^{2, l.\alpha}$ means that $f(x, \cdot) \in C^{l+\alpha/2}$ and $f(\cdot, t) \in C^{k+\alpha}$. The authors show that this heat flow for Dirac-harmonic maps admit a unique solution provided that \[ \phi \in \bigcap_{T0} C^{2, 1.\alpha}(\bar M \times [0, T]; N) \] and \[ \mathcal{B}\psi \in \bigcap_{T0} C^{1, 0.\alpha}(\partial M \times [0, T]; \Sigma M \otimes \phi^{-1}TN) \] for some $0 \alpha 1$.