The PDE is the same as before, but now we consider non-homogenous Dirichlet boundary conditions. That is, boundary conditions may be nonzero constants or functions of time. For simplicity, we take a one-dimensional rod of length 1.
Examples:
We assume the solution splits into steady-state and transient parts:
where \(S(x,t)\) satisfies the boundary conditions (“homogenizer”), and \(U(x,t)\) is the transient part.
Constant (Non-Zero) Boundary Conditions
Suppose \(u(0,t)=C_1\), \(u(1,t)=C_2\). As \(t\to\infty\), the rod tends to a linear steady state between \(C_1\) and \(C_2\).
For example, with \(u(0,t)=1\), \(u(1,t)=5\), choose
Enforcing boundary values:
So \(S(x,t)=4x+1\). Substituting \(u=S+U\) into the PDE:
Since \(S_t=0,\; S_{xx}=0\), we obtain
The boundary conditions for \(U\) become homogeneous:
Thus \(U\) satisfies the standard heat equation with homogeneous BCs. Its solution is
where \(\phi(x)\) is the initial condition.
Therefore the full solution is
Time-Varying Boundary Conditions
For time-dependent BCs, use a linear-in-\(x\) interpolant:
Example: \(u(t,0)=3t,\; u(t,1)=1/t.\) Then
So
Substitute into the PDE. With \(u=S+U\):
Since \(S_{xx}=0,\; S_t=3-3x-\tfrac{1}{t^2}x\) (for this particular example), the PDE becomes
Thus \(U\) has homogeneous boundary conditions but satisfies a non-homogenous heat equation. This cannot be solved by simple separation of variables; instead, one uses eigenfunction expansion (covered in the next section).