19. Exercises#
19.1. Friedrichs’ inequality#
19.1.1. Friedrichs’ inequality in 1D#
Prove Friedrichs’ inequality on the interval \(I = (a,b)\) with elementary tools. Show that there exists a constant \(c_F\) such that
how does your constant \(c_F\) depend on \(a\) and \(b\) ?
19.1.2. Friedrichs’ inequality on the square#
Define \(\Omega = (0,a) \times (0,b)\). Prove Friedrichs’ inequality
How does \(c_F\) depend on \(a\) and \(b\) ?
19.2. Poincaré inequality#
19.2.1. Eqivalent versions of the Poincaré inequality#
Prove that the following inequalities are equivalent
There exists a constant \(c_p\) such that
Here, \(\overline u\) is the mean value of \(u\), which is \(\frac{1}{|\Omega|} \int_\Omega u(x) dx\), understood as a constant function.
There exists a constant \(c_p\) such that
\[ \| u \|_{L_2}^2 \leq c_P^2 \, \| \nabla u \|_{L_2}^2 + \frac{1}{|\Omega|} \left( \int_\Omega u \right)^2 \]
Hint: Pythagoras. First, observe that \(\overline u\) and \(u - \overline u\) are orthogonal with respect to the \(L_2\) inner product.
19.2.2. Poincaré inequality in 1D#
Prove the Poincaré inequality on the interval \(I = (a,b)\) with elementary tools:
How does the Poincaré constant depend on \(a\) and \(b\) ?
Hint: Bring the left hand side to something similar to \(\int \left( \int u(y) - u(x) dx\right) ^2 dy\). Then, use the fundamental theorem of calculus: \(u(y)-u(x) = \int_x^y u^\prime(s) \, ds\). finally: Cauchy-Schwarz
19.3. Bramble-Hilbert Lemma#
19.3.1. Mean-value interpolation#
Let \(\Omega \subset {\mathbb R}^d\), open, bounded and connected. Prove that there exists a constant \(c\) such that
where \(\overline u := | \Omega |^{-1} \int_\Omega u \, dx\) is the mean value.
Use the Bramble-Hilbert theorem
19.3.2. Scaled domain#
Let \(\Omega = B_r(p)\) a ball with center \(p\) and radius \(r\). Prove that there exists a constant \(c\), independent of \(r\) and \(p\), such that
Hint: Prove the estimate for the unit ball \(B_1(0)\). Define a function \(\Phi\) mapping the unit-ball to the arbitrary ball \(B_r(p)\). For \(u \in H^1(B_r(p))\), define the pull-back \(u \circ \Phi \in H^1(B_1(0))\). Does mean-value and pull-back commute, i.e. does there hold \(\overline u \circ \Phi = \overline{ u \circ \Phi}\) ?
19.4. Fractional Sobolev spaces#
19.4.1. Step function#
Consider the function \(u : (-\pi, \pi) \rightarrow {\mathbb R}\) defined as
Find all \(\alpha \in [0,1]\) such that \(u \in H^\alpha ( (-\pi, \pi) )\).
Hint: compute Fourier coefficients \(a_k\) and \(b_k\) of \(u(x) = \sum a_k \cos(k x) + b_k \sin(k x)\), and find all \(\alpha\) such that \(\sum_{k=0}^\infty k^{2\alpha} (a_k^2 + b_k^2) < \infty\)
19.4.2. Point evaluation functional#
Consider the linear functional
Find all \(\alpha \in (0,1)\) such that \(f\) can be continuously extended onto \(H^\alpha((-\pi,\pi))\), i.e. find all \(\alpha\) such that
Hint: Expand \(v\) as a Fourier series \(v(x) = \sum a_k \cos(k x) + b_k \sin(k x)\), and express \(f(v)\) by means of the \(a_k\) and \(b_k\).