The \(\square_b\)-heat equation on pseudoconvex manifolds of finite type in \(\mathbb{C}^2\) (Q5949442)

Let \(M\subset \mathbb{C}^2\) be a connected \(\mathcal C^\infty\) pseudoconvex hypersurface of finite type and assume that either \(M\) is the boundary of a bounded pseudoconvex domain in \(\mathbb{C}^2\) of finite type \(m\), or that \(M\) is an unbounded polynomial model of type \(m\) in \(\mathbb{C}^2\). Let \(\square_b\) be the Kohn-Laplacian operator on \(M\) and let \(\mathcal H=\partial_s+\square_b\) be the heat operator on \(R\times M\). In this paper, the authors study the following problems : (IVP) The initial value problem, for a given function or (0,1)-form \(f\) on \(M\), to find a smooth function or (0,1)-form \(F\) on \((0,\infty)\times M\) such that \(\mathcal H(F)=0\) on \((0,\infty)\times M\) and \(\lim_{s\to 0} F(s,\cdot)=f(\cdot)\) with convergence in appropriate norm. (HE) To establish the local regularity of solutions \(u\) to the heat equation \(\mathcal H[u]=g\) in terms of the local regularity of \(g\). Since the \(\square_b\) is not hypoelliptic here, both (IVP) and (HE) have different property from the classical case. In general, the solution of the problem (IVP) on the space \(L^2(M)\) is given by a semigroup of operators \(e^{-s\square_b}\). The heat kernel for \(\square_b\) is the distribution \(H(s,x,y)\) defined on \((0,\infty)\times M\times M\) such that the solution is given by \[ F(s,x)=e^{-s\square_b}[f](x)= \int_M f(y)H(s,x,y)\,dy. \] In this paper, (0,1)-forms are identified with function and the operator \(\partial_b\) is identified with the vector field \(\overline L=X_1+iX_2\). There is a non-isotropic control metric \(d_M(x,y)\) associated with \(X_1\) and \(X_2\). We write \(B_M(x,R)=\{y\in M: d_M(x,y)< R\}\) for the ball in this metric centered at \(x\) of radius \(R\). Let \(V_M(x,y)\) denote the volume of the ball centered at \(x\) of radius \(d_M(x,y)\). For the local regularity of (HE), the authors establish the following estimate. Theorem: Let \(V=W\times U\subset R\times M\) be a relatively compact open set. Let \(\zeta_j\in \mathcal C^\infty_0(V), j=0,1,2\) with the property that \(\zeta_j\equiv 1\) on the support of \(\zeta_{j+1} , j=0,1\). Let \(b\in R\) and let \(N\) be a positive integer. There is an NIS constant \(C\) (for the definition NIS see the paper) with the following property. Let \(a\) be the smallest positive integer greater than or equal to \(b+1\). Let \(F\) be any distribution on \(R\times M\). Then \[ \| \zeta_2F\| _b\leq C \left[ \sum_{k=1}^a \| \zeta_1\mathcal H^k(F)\| _a +||| \zeta_1S[\zeta_0F]||| _{(0,a)} +\| \zeta_0F\| _{-N} \right] \] in the sense that if the expression on the right hand side is finite, the left hand side is also finite, and the stated inequality holds. Here \(\| \cdot \| _b\) are the standard Sobolev norms on \(R\times M\), and \(||| \cdot||| _{(a,b)}\) is a variant which measures \(a\) derivatives in time and \(b\) derivatives in space. As a consquence that if \(F\) is a distribution on \(R\times M\) ,if \(\zeta\in \mathcal C^\infty_0(R\times M)\) and if \(\mathcal H(F)\) and \(S[\zeta F]\) are \(\mathcal C^\infty\) on the support of \(\zeta\) , then \(F\) is also \(\mathcal C^\infty\) over there. Applying the estimates above, the authors obtain the following result for the (IVP). Theorem: There is a function \(G\in {\mathcal C}^\infty((0,\infty)\times M\times M)\) so that the heat kernel for \(\square_b\) on functions is given by the distribution \[ H(s,x,y)=G(s,x,y)+S(x,y), \] which is thus given by integration against a smooth function away from the diagonal . Thus the singularities of \(H\) are exactly the same as those of \(S\). The functions \(G\) and \(H\) satisfy the following estimates for \(x\neq y\). There is a constant \(C\) such that \[ | G(s,x,y)| \leq \begin{cases} CV_M(x,y)^{-1} &\text{if }s\leq d_M(x,y)^2,\\ C| B_M(x,\sqrt s)| ^{-1} &\text{if }s\geq d_M(x,y)^2.\end{cases} \] For every non-negative integer \(N\) there is a constant \(C_N\) such that \[ | H(s,x,y)| \leq C_N{1\over V_M(x,y)}\left [{s^N\over s^N+d_M(x,y)^{2N}}\right]. \] They also have estimates on derivatives of \(G\) and \(H\) with respect to \(s\) and to the vector fields \(X_1\) and \(X_2\). For every integer \(j\) and all multi-indices \(\alpha\) and \(\beta\), there is a constant \(C_{\alpha,\beta,j}\) such that \[ | \partial_s^j X_x^\alpha X_y^\beta G(s,x,y)| \leq \begin{cases} C_{\alpha,\beta,j}d_M(x,y)^{-2j-| \alpha| -| \beta| } V_M(x,y)^{-1} &\text{if } s\leq d_M(x,y)^2,\\ C_{\alpha,\beta,j} s^{-j-(1/2)(| \alpha| +| \beta| )} | B_M(x,\sqrt s)| ^{-1} &\text{if } s\geq d_M(x,y)^2.\end{cases} \] Moreover, for every non-negative integer \(N\) there is a constant \(C_{N,\alpha,\beta,j}\) such that \[ | \partial_s^j X_x^\alpha X_y^\beta H(s,x,y)| \leq C_{N,\alpha,\beta,j}{d_M(x,y)^{-2j-| \alpha| -| \beta| } \over V_M(x,y)}\left [{s^N\over s^N+d_M(x,y)^{2N}}\right]. \]