= Strict monotonicity without coercivity on H1
{title2=$\langle A(v),h\rangle=\int(\nabla v\cdot\nabla h+\tanh(v)h)$}
Let $\Omega$ be a nonempty bounded <Lipschitz domain>, and define $A:H^1(\Omega)\to H^1(\Omega)^*$ by
$$
\langle A(v),h\rangle=\int_\Omega\bigl(\nabla v\cdot\nabla h+\tanh(v)h\bigr).
$$
This operator is strictly monotone: the difference pairing is the sum of $\|\nabla(v-w)\|_2^2$ and the nonnegative integral $(\tanh v-\tanh w)(v-w)$, which is strictly positive wherever $v\ne w$. The bounded-domain hypothesis also makes nonzero constant <functions> admissible in $H^1$.
For constant $v=t$,
$$
\frac{\langle A(v),v\rangle}{\|v\|_{H^1}}=|\Omega|^{1/2}\operatorname{sign}(t)\tanh t\longrightarrow|\Omega|^{1/2}\quad(|t|\to\infty).
$$
Thus $A$ is not a <nonlinear coercive operator> in the large-norm sense. On $H_0^1(\Omega)$, the <Poincaré inequality> instead gives
$$
\langle A(v),v\rangle\geq\|\nabla v\|_2^2\geq\frac{\|v\|_{H^1}^2}{1+C_P^2},
$$
which proves coercivity on that zero-boundary space. The example separates being a <strictly monotone operator> from coercivity and identifies the uncontrolled constant direction.
Back to article page