如何求解下述线性PDE？演化方程转热方程的移动边界求解咨询

Hey there! Let's tackle your two PDE questions head-on—first covering general linear PDE solving strategies, then digging into that tricky moving boundary heat equation scenario you mentioned.

Linear PDEs fall into three core categories (elliptic, parabolic, hyperbolic), and your choice of method depends on the equation type, domain, and boundary/initial conditions. Here are the most reliable techniques:

• Separation of Variables
  This is the go-to for bounded domains with homogeneous boundary conditions. Assume a solution of the form u(x,t) = X(x)T(t) (for time-dependent equations) or u(x,y) = X(x)Y(y) (steady-state), substitute into the PDE, and split it into ordinary differential equations (ODEs) for each variable. Solve the ODEs, then combine solutions via superposition (linearity lets you add valid solutions). Perfect for equations like the heat equation u_t = αu_xx or wave equation u_tt = c²u_xx on fixed intervals.

• Method of Characteristics
  Ideal for hyperbolic PDEs (e.g., transport equation u_t + c u_x = 0 or linear wave equations). Characteristics are curves along which the PDE reduces to an ODE—trace these curves to carry initial/boundary data across the domain and build your solution. For linear cases, characteristics are straight lines, making this method straightforward.

• Green's Functions
  Use this for nonhomogeneous linear PDEs (with a source term). A Green's function G(x,t;x',t') represents the solution at (x,t) caused by a point source at (x',t'). You can express the full solution as an integral of the Green's function convolved with the source term, plus contributions from initial/boundary conditions. Each PDE and domain has its own unique Green's function, so you'll need to derive or look up the right one for your problem.

• Transform Methods (Fourier/Laplace)
  Great for unbounded domains or problems with specific initial conditions. Apply a Fourier transform in space (for u_t = αu_xx on -∞ < x < ∞) or Laplace transform in time, convert the PDE into an algebraic equation or ODE, solve that, then invert the transform to get your solution. Fourier transforms work well for periodic/decaying boundary conditions, while Laplace transforms handle initial value problems nicely.

When your evolution equation maps to a heat equation but with moving boundaries, the key is either fixing the boundary via coordinate changes or handling the motion explicitly. Here are the most effective approaches:

• Coordinate Transformation to Fix Boundaries
  If your moving boundary is defined by x = b(t) (e.g., a boundary moving at a time-dependent speed), create a new coordinate ξ = x - b(t) to shift the boundary to a fixed position (like ξ = 0). Substitute this into the heat equation—you'll end up with a convection term alongside the diffusion term. Here's a concrete example:

  Start with the standard heat equation: u_t = αu_xx
  Let ξ = x - vt (where v is a constant boundary speed). Using the chain rule, we rewrite the derivatives:
  u_t = -v u_ξ + u_t (holding ξ constant), and u_xx = u_{ξξ}
  Substitute back: u_t = αu_{ξξ} + v u_ξ
  Now you have a heat equation with a linear convection term. Eliminate this with another substitution: let u(ξ,t) = e^{-vξ/(2α) - v²t/(4α)} w(ξ,t). Plugging this in converts the equation to the standard heat equation w_t = αw_{ξξ}, which you can solve with separation of variables or Green's functions, then convert back to the original coordinates.

• Front Tracking (For Free Boundaries)
  If the moving boundary is a free boundary (e.g., in phase change problems, where the solid-liquid interface moves as heat transfers), you need to couple the domain's heat equation with an ODE governing the boundary's motion. For example, the boundary x = s(t) might follow an energy balance: ρL ds/dt = k(u_x|_{x=s(t)^+} - u_x|_{x=s(t)^-}), where ρ is density, L is latent heat, and k is thermal conductivity. Solve the heat equation in each region (solid/liquid) with appropriate boundary conditions, then update the boundary position using the ODE, iterating until convergence.

• Numerical Methods for Moving Grids
  If an analytical solution is out of reach, numerical methods are your friend:

  • Arbitrary Lagrangian-Eulerian (ALE) Methods: The computational grid moves with the boundary, so boundary conditions apply to fixed grid nodes, simplifying discretization.
  • Fixed Grid Methods: Use a static grid, track the boundary's position over time, and apply boundary conditions via interpolation (e.g., finite differences with ghost nodes near the boundary).

内容的提问来源于stack exchange，提问作者will_cheuk