Stability of solutions of nonlinear diffusion problems (Q2523631)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Stability of solutions of nonlinear diffusion problems
scientific article

    Statements

    Stability of solutions of nonlinear diffusion problems (English)
    0 references
    0 references
    0 references
    1966
    0 references
    The authors consider stability theorems for solutions of the boundary-value problem \[ (\text{I})\qquad u_t = F(x, u, u_x, u_{xx}),\quad x\in [a, b],\ t > 0, \] \[ (\text{II})\quad u_x(a, t) = f_1(u(a, t))\text{ and }u_x(b, t) = f _2(u(b, t)),\quad\text{for all }t > 0. \tag{D} \] \[ (\text{III})\qquad u(x, 0) = \varphi (x). \] The differentiability assumptions on \(F, f_1, f_2, \varphi\), and \(u\) are, \(F\in C^1\) and \(F\) is non-decreasing in \(u_{xx}\), \(\vert f_i\vert + \vert f'_i\vert \le M\), \(i = 1,2\), \(\varphi\in C^2([a, b])\), \(u, u_x, u_{xx}, u_t\in C((a, b)\times (0,\infty))\), and \(u, u_x \in C^1([a, b] \times (0, \infty))\). The notation \(u(\varphi; x, t)\) is used to denote a solution to problem (D). Definition 1. Suppose for every \(\varepsilon > 0\) there is a \(\delta > 0\) such that if \(\psi(x)\) satisfies \(\max_{x\in (a,b)} \vert\psi(x) - \varphi(x)\vert < \delta\) then it is true that \(\max_{x\in (a,b)} \vert u(\varphi; x, t) - u(\psi; x, t)\vert < \varepsilon\). Then \(u(\varphi; x, t)\) is called a stable solution to problem (D). Theorem. Consider problem (D) and assume there is a one-parameter family \(v(x, \lambda)\), \(\lambda\in [\lambda_1, \lambda_2]\), of solutions of \(F(x, v, v_x, v_{xx}) = 0\) satisfying the following four conditions: (i) There is a number \(\lambda'\in (\lambda_1, \lambda_2)\) such that \(v_x(a, \lambda') = f_1(v(a, \lambda'))\) and \(v_x(b, \lambda') = f_2(v (b, \lambda'))\). (ii) \(v_\lambda(x, \lambda) > 0\) for all \(x\in [a, b]\) and \(\lambda\in [\lambda_1, \lambda_2]\). (iii) \(v_x(a, \lambda) > f_1(v(a, \lambda))\) and \(v_x(b, \lambda) < f_2(v(b, \lambda))\) for \(\lambda\in [\lambda_1, \lambda']\). (iv) \(v_x(a, \lambda) < f_1(v(a, \lambda))\) and \(v_x(b, \lambda) > f_2(v(b, \lambda))\) for \(\lambda\in [\lambda', \lambda_2]\). Then, if \(\varphi(x) = v(x, \lambda')\), \(u(\varphi; x, t)\equiv \varphi(x)\) is a solution of problem (D), i. e., there is a steady-state solution to problem (D). Corollary. Suppose there exists a solution, \(v(x)\), of \(F(x, v, v_x, v_{xx}) = 0\) satisfying \(v_x(a) < f_1(v(a))\) and \(v_x(b) > f_2(v(b))\). Then if \(u(\psi; x, t)\) is a solution to (D) where \(\psi(x)<v(x)\) for all \(x\in [a, b]\), we have \(u(x,t) < v(x)\) for all \(x\in [a, b]\) and \(t > 0\). The authors also state a theorem giving sufficient conditions for the existence of an unstable, steady-state solution to (D). The conditions amount to reversing some of the inequalities in the previous theorem.
    0 references
    stability of solutions
    0 references
    boundary value problem
    0 references
    nonlinear diffusion problems
    0 references
    steady-state solution
    0 references
    existence of unstable, steady-state solution
    0 references

    Identifiers