Question on the proof of Theorem 2.3.2 in Partial Differential Equations by Evans

42 Views Asked by At

I'm working through Evans's proof of the solution for the nonhomogenous heat equation (Theorem 2.3.2) and I'm getting stuck on a certain part. He starts with the proposed solution: $$ u(x,t) = \int^t_0 \int_{\Bbb{R}^n} \Phi(y,s)f(x-y,t-s)dyds $$ In the middle of the proof, he calculates $$ u_t(x,t)-\Delta u(x,t) = \int^t_0 \int_{\Bbb{R}^n}\Phi(y,s)(\frac{\partial}{\partial t}-\Delta_x)f(x-y,t-s)dyds + \int_{\Bbb{R}^n}\Phi(y,t)f(x-y,0)dy = \quad \int^t_\varepsilon \int_{\Bbb{R}^n}\Phi(y,s)(-\frac{\partial}{\partial s}-\Delta_y)f(x-y,t-s)dyds + \int^\varepsilon_0\int_{\Bbb{R}^n}\Phi(y,s)(-\frac{\partial}{\partial s}-\Delta_y)f(x-y,t-s)dyds + \int_{\Bbb{R}^n}\Phi(y,t)f(x-y,0)dy$$. So far so good. Evans then denotes $$I_\varepsilon = \int^t_\varepsilon \int_{\Bbb{R}^n}\Phi(y,s)(-\frac{\partial}{\partial s}-\Delta_y)f(x-y,t-s)dyds $$ and then argues that, by using integration by parts, we obtain: $$ I_\varepsilon = \int^t_\varepsilon \int_{\Bbb{R}^n}(\frac{\partial}{\partial s}-\Delta_y)\Phi(y,s)f(x-y,t-s)dyds + \int_{\Bbb{R}^n}\Phi(y,\varepsilon)f(x-y,t-\varepsilon)dy - \int_{Bbb{R}^n}\Phi(y,t)f(x-y,0)dy$$

This is where I am confused. I tried working out the details of this integration by parts by first looking at the $\frac{\partial}{\partial s} $ part of the integral first. Just looking at that part separately, we have $$ \int^t_\varepsilon \int_{\Bbb{R}^n}\Phi(y,s)[-\frac{\partial}{\partial s}f(x-y,t-s)]dyds = \int_{\Bbb{R}^n}\int^t_\varepsilon\Phi(y,s)[-\frac{\partial}{\partial s}f(x-y,t-s)]dsdy = \quad \int_{\Bbb{R}^n}\int^t_\varepsilon\frac{\partial}{\partial s}\Phi(y,s)f(x-y,t-s)dsdy - \int_{\Bbb{R}^n} [\Phi(y,s)f(x-y,t-s)]^t_\varepsilon dy = \int^t_\varepsilon\int_{\Bbb{R}^n}\frac{\partial}{\partial s}\Phi(y,s)f(x-y,t-s)dsdy - \int_{\Bbb{R}^n} \Phi(y,t)f(x-y,0) dy + \int_{\Bbb{R}^n} \Phi(y,\varepsilon)f(x-y,t-\varepsilon)dy$$

So that explains where the last two terms come from. However, I'm confused about how to handle the Laplacian part of this integral. Using Green's third identity, we have $$ \int^t_\varepsilon \int_{\Bbb{R}^n}\Phi(y,s)(-\Delta_y f(x-y,t-s))dyds = \int^t_\varepsilon \int_{\Bbb{R}^n}-\Delta_y \Phi(y,s)f(x-y,t-s)dyds + \int^t_\varepsilon \int_{\partial \Bbb{R}^n} [\Phi(y,s)\nabla f(x-y,t-s) \cdot \nu - f(x-y,t-s) \nabla \Phi(y,s) \cdot \nu]dS $$

Here, $\nu$ is the outward facing unit normal vector. Based on the result that Evans gets, I assume that $$(\ast) \quad \int^t_\varepsilon \int_{\partial \Bbb{R}^n} [\Phi(y,s)\nabla f(x-y,t-s) \cdot \nu - f(x-y,t-s) \nabla \Phi(y,s) \cdot \nu]dS = 0 $$ but I'm not sure how to verify this. I'm also not sure what it means to integrate a function over $\partial \Bbb{R}^n$ (I don't know what the boundary of $\Bbb{R}^n$ would be. So my question is, is my work thus far correct, and if so, how do I verify that ($\ast$) is correct? Any help would be appreciated.