Some preliminary definitions and notation:

(1) Given a vector space $\mathbb{V}$, we denote by $G_k(\mathbb{V})$ the $k$-grassmannian of $\mathbb{V}$, i.e. the set of all $k$-dimensional vector subspaces of $\mathbb{V}$;

(2) Given a differential (or riemannian) manifold $M$, we denote by $G_k(M)$ the $k$-grasmannian bundle on $M$, i.e. the bundle with fibers $G_k(T_pM)$;

(3) A $k$-dimensional varifold on $M$ is any Radon measure $V$ on $G_k(M)$;

(4) We denote by $\mathcal{V}_k(M)$ the set of all $k$-dimensional varifolds on $M$;

(5) The mass of a varifold $V\in \mathcal{V}_k(M)$ is defined by $\|V\|=\int_{G_k(M)}dV$.

One can define a topology - the weak topology - on $\mathcal{V}_k(M)$ as follows. Let $J:=\mathcal{C}_c(G_k(M))$ be the set of compactly supported continuous functions $f:G_k(M)\to \mathbb{R}$ and consider the family of functions $\{\varphi_f\,:\, f\in J\}$ where $$\begin{matrix} \varphi_f:&\mathcal{V}_k(M)&\to&\mathbb{R}\\ &V&\mapsto&\int_{G_k(M)}f\,dV. \end{matrix}$$ The weak topology on $\mathcal{V}_k(M)$ is the least topology on $\mathcal{V}_k(M)$ for which every $\varphi_f$ is continuous or, in other words, is the topology on $\mathcal{V}_k(M)$ generated by the subbasis $$\beta=\{\varphi_f^{-1}(\Omega)\,:\, f\in J,\,\Omega\subset_{op} \mathbb{R}\}.$$

One can ask if the weak topology on $\mathcal{V}_k(M)$ is metrizable. Well, define $\mathbf{F}:\mathcal{V}_k(M)\times \mathcal{V}_k(M)\to [0,+\infty]$ by $$\mathbf{F}(V,W)=\sup E(V,W),$$ with $$E(V,W)=\left\{\int_{G_k(M)}f\,dV-\int_{G_k(M)}f\,dW\,:\, f\in J,\,|f|\leq 1,\,\mathrm{Lip}(f)\leq 1\right\}.$$

One verifies that $\mathbf{F}$ has the three axioms of a metric, although it can be $+\infty$. But given any $c\geq 0$, $\mathbf{F}$ is always finite on $$\mathcal{V}_{k,c}(M):=\{V\in \mathcal{V}_k(M)\,:\, \|V\|\leq c\}.$$

Indeed, for any $f\in J$, $|f|\leq 1$, $\mathrm{Lip}(f)\leq 1$, $$\int_{G_k(M)}f\,dV-\int_{G_k(M)}f\,dW\leq \int_{G_k(M)}\,dV+\int_{G_k(M)}\,dW=\|V\|+\|W\|\leq 2c.$$

Therefore, $\mathbf{F}$ is a finite genuine metric on $\mathcal{V}_{k,c}(M)$.

I want to prove that:

($\ast$) The weak topology and the $\mathbf{F}$-induced topology coincide on $V_{k,c}(M)$.

This $\mathbf{F}$-metric appears in several papers in the internet (generally talking about minimal surfaces) but always with its definition and the claim ($\ast$) above, followed by some sort of "it is easy to prove" or "one easily verifies".

Is it really that easy? I guess I could prove that $\tau_w\subset \tau_{\mathbf{F}}$ but, given an open $\mathbf{F}$-ball $B_\epsilon(V)$, how do I find a weak-basic open $U$ with $V\in U\subset B_\epsilon(V)$? (such open must be a finite intersection of elements of $\beta$. Which $f_i$'s and real opens $I_i$'s to choose?)

  • $\begingroup$ Just a curiosity: where did you learn varifolds from? $\endgroup$ – Eduardo Longa Apr 5 '18 at 1:09
  • $\begingroup$ @EduardoLonga I am still a beginner on the subject. :) $\endgroup$ – Ders Apr 5 '18 at 9:49
  • $\begingroup$ Are either of these obviously equivalent to the notion of weak (*?) convergence of measures? $\endgroup$ – Ryan Unger Apr 5 '18 at 18:47
  • 1
    $\begingroup$ @0celo7 In this case, weak convergence of Radon measures. Are weak and weak* convergence equal for measures? I find both names but the definitions seem to be always the same... $\endgroup$ – Ders Apr 10 '18 at 18:01

I'm pretty sure this is false in general. Consider the degenerate case where $M = \mathbb{R}$ and $k=0$, then $\mathcal{V}_0(\mathbb{R})$ is the space of Radon measures on $\mathbb{R}$, which includes the dirac delta measures $\delta_x : f \mapsto f(x)$. Now note that for any $f$ with compact support:

$$ \lim_{x \to \infty} f(x) = 0 $$

hence $\lim_{x\to\infty} \delta_x = 0$ in the weak*-topology. And yet:

$$ \lim_{x \to \infty} \mathbf{F}(\delta_x, 0) = 1 $$

so the $\delta_x$ do not converge in the $\mathbf{F}$ topology (or equivalently $\mathbf{F}$ is not continuous in the weak*-topology). Restricting to a subspace of measures $V$ with $\|V\| \leq c$ does not help since $\|c \delta_x \| = c$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.