Chapter 38

Dirichlet forms

[16/0%]
LaTeX
Problem 38.1

Let F1={f∈C1[0,∞):fF_{1}=\left\{ f \in C^{1}[0, \infty ): f\right. has compact support }\} and F2=F1∩{f∈C1[0,∞)F_{2}=F_{1} \cap \left\{ f \in C^{1}[0, \infty )\right. : ff has compact support, f′(0)=0}\left.f^{\prime }(0)=0\right\}. Show that the closures of F1F_{1} and F2F_{2} with respect to the norm (∫(∣f(x)∣2+∣f′(x)∣2)dx)1/2\left(\int \left(|f(x)|^{2}+\left|f^{\prime }(x)\right|^{2}\right) d x\right)^{1 / 2} are the same.

?
Problem 38.2

If H1H^{1} is the completion of CK1C_{K}^{1}, the C1C^{1} functions on Rd\mathbb {R}^{d} with compact support, relative to the norm given by ( 38.11 ), show H1H^{1} is a Hilbert space.

?
Problem 38.3

Show that the resolvent operator RλR_{\lambda } defined in Theorem 38.1 is a symmetric operator, that is, if f,g∈Bf, g \in \mathcal{B}, then ⟨Rλf,g⟩=⟨f,Rλg⟩\left\langle R_{\lambda } f, g \right\rangle = \left\langle f, R_{\lambda } g \right\rangle.

?
Problem 38.4

Show that if the resolvent operator RλR_{\lambda } is a symmetric operator, then the transition operators PtP_{t} are also symmetric: if f,g∈Bf, g \in \mathcal{B}, then ⟨Ptf,g⟩=⟨f,Ptg⟩\left\langle P_{t} f, g \right\rangle = \left\langle f, P_{t} g \right\rangle.

?
Problem 38.5

To do the next few exercises, you will have to know some functional analysis, specifically, the spectral theorem for self-adjoint operators. See Lax (2002).

Let E\mathcal{E} be a Dirichlet form with domain D(E)\mathcal{D}(\mathcal{E}) and let L\mathcal{L} be the infinitesimal generator of the semigroup PtP_{t} that corresponds to L\mathcal{L}. Let E(dλ)E(d \lambda ) be a spectral resolution of the identity for −L-\mathcal{L}. (The operator L\mathcal{L} is a negative operator, so −L-\mathcal{L} is a positive one.) Then a consequence of the spectral theorem is that

Ptf=∫0∞e−λtE(dλ)f P_{t} f=\int _{0}^{\infty } e^{-\lambda t} E(d \lambda ) f

and

Raf=∫0∞1a+λE(dλ)f R_{a} f=\int _{0}^{\infty } \frac{1}{a+\lambda } E(d \lambda ) f

Also

⟨f,g⟩=∫0∞⟨E(dλ)f,g⟩ \left\langle f, g \right\rangle =\int _{0}^{\infty } \left\langle E(d \lambda ) f, g \right\rangle

Show that if f,g∈Df, g \in \mathcal{D}, then

E(f,g)=∫0∞λ⟨E(dλ)f,g⟩ \mathcal{E}(f, g)=\int _{0}^{\infty } \lambda \left\langle E(d \lambda ) f, g \right\rangle

Hint: First prove it for f=Rahf=R_{a} h. Write

E(Rah,g)=⟨h,g⟩−a⟨Rah,g⟩=∫0∞(1−aa+λ)⟨E(dλ)h,g⟩=∫0∞λa+λ⟨E(dλ)h,g⟩=∫0∞λ⟨E(dλ)(Rah),g⟩ \begin{aligned} \mathcal{E}(R_{a} h, g) & = \left\langle h, g \right\rangle -a \left\langle R_{a} h, g \right\rangle =\int _{0}^{\infty }\left(1-\frac{a}{a+\lambda }\right) \left\langle E(d \lambda ) h, g \right\rangle \\ & =\int _{0}^{\infty } \frac{\lambda }{a+\lambda } \left\langle E(d \lambda ) h, g \right\rangle =\int _{0}^{\infty } \lambda \left\langle E(d \lambda )\left(R_{a} h\right), g \right\rangle \end{aligned}

To extend this to all ff in the domain of E\mathcal{E}, use the fact that E\mathcal{E} is closed.

?
Problem 38.6

If L\mathcal{L} is the infinitesimal generator of the semigroup associated with the Dirichlet form E\mathcal{E}, show that D(−L)=D(E)\mathcal{D}(\sqrt{-\mathcal{L}})=\mathcal{D}(\mathcal{E}).

?
Problem 38.7

Show that if f∈D(E)f \in \mathcal{D}(\mathcal{E}), then aRafa R_{a} f converges to ff with respect to the norm induced by E1\mathcal{E}_{1}.

?
Problem 38.8

Show that if b>0b>0, then {Rbf:f∈L2}\left\{ R_{b} f: f \in L^{2}\right\} is a dense subset of D(E)\mathcal{D}(\mathcal{E}) with respect to the norm induced by E1\mathcal{E}_{1}.

?
Problem 38.9

Show that {Ptf:f∈L2,t>0}\left\{ P_{t} f: f \in L^{2}, t>0\right\} is a dense subset of D(E)\mathcal{D}(\mathcal{E}) with respect to the norm induced by E1\mathcal{E}_{1}.

?
Problem 38.10

This exercise shows how to approximate E\mathcal{E} by forms whose domain is all of B\mathcal{B}. Let

E(t)(f,g)=1t⟨f−Ptf,g⟩ \mathcal{E}^{(t)}(f, g)=\frac{1}{t} \left\langle f-P_{t} f, g \right\rangle

Show that if f∈D(E)f \in \mathcal{D}(\mathcal{E}), then E(t)(f,f)\mathcal{E}^{(t)}(f, f) increases to E(f,f)\mathcal{E}(f, f). Show that if f,g∈D(E)f, g \in \mathcal{D}(\mathcal{E}), then E(t)(f,g)\mathcal{E}^{(t)}(f, g) converges to E(f,g)\mathcal{E}(f, g).

?
Problem 38.11

Show that if u∈D(E)u \in \mathcal{D}(\mathcal{E}), then ∣u∣∈D(E)|u| \in \mathcal{D}(\mathcal{E}) and E(∣u∣,∣u∣)≤E(u,u)\mathcal{E}(|u|,|u|) \leq \mathcal{E}(u, u).

Hint: Use Exercise 38.10 .

?
Problem 38.12

Use Exericse 38.11 to show that if u∈D(E)u \in \mathcal{D}(\mathcal{E}), then E(u+,u−)≤0\mathcal{E}\left(u^{+}, u^{-}\right) \leq 0.

?
Problem 38.13

Suppose {Pt}\left\{ P_{t}\right\} are the transition probabilities corresponding to a Dirichlet form E\mathcal{E}. Suppose there exist functions pt(x,y)p_{t}(x, y) such that for each tt,

Ptf(x)=∫pt(x,y)m(dy) P_{t} f(x)=\int p_{t}(x, y) m(d y)

for almost every xx. Prove that for almost every pair (x,y)(x, y) with respect to the product measure m×m,pt(x,y)=pt(y,x)m \times m, p_{t}(x, y)=p_{t}(y, x).

?
Problem 38.14

Let f∈L2(m)f \in L^{2}(m) and define the functional

ψ(u)=E(u,u)+λ⟨u,u⟩−2⟨f,u⟩ \psi (u)=\mathcal{E}(u, u)+\lambda \left\langle u, u \right\rangle -2 \left\langle f, u \right\rangle

for uu in the domain of E\mathcal{E}. Prove that ψ\psi is minimized by u=Rλfu=R_{\lambda } f, and that this function is the unique minimizer.

?
Problem 38.15

Let PtP_{t} be the semigroup associated with a Dirichlet form and define

J(dx,dy)=Pt(x,dy)m(dx) J(d x, d y)=P_{t}(x, d y) m(d x)
?
(1)

Prove that if f,gf, g are continuous with compact support, then

∬f(x)g(y)J(dx,dy)=∬g(x)f(y)J(dx,dy) \iint f(x) g(y) J(d x, d y)=\iint g(x) f(y) J(d x, d y)
(2)

With ff and gg continuous with compact support, prove that

∫f(x)g(y)J(dx,dy)=⟨f,Ptg⟩ \int f(x) g(y) J(d x, d y)= \left\langle f, P_{t} g \right\rangle

and

∬f(x)g(x)J(dx,dy)=⟨fg,Pt1⟩ \iint f(x) g(x) J(d x, d y)= \left\langle f g, P_{t} 1 \right\rangle
(3)

Let k(x)=1−Pt1(x)k(x)=1-P_{t} 1(x). Prove that if E(t)\mathcal{E}^{(t)} is defined as in Exercise 38.10 , then

2tE(t)(f,g)=∬(f(x)−f(y))(g(x)−g(y))J(dx,dy)+∫f(x)g(x)k(x)m(dx) 2 t \mathcal{E}^{(t)}(f, g)=\iint (f(x)-f(y))(g(x)-g(y)) J(d x, d y)+\int f(x) g(x) k(x) m(d x)
(4)

Is E(t)\mathcal{E}^{(t)} a Dirichlet form? A regular Dirichlet form?

Problem 38.16

This is a continuation of the previous exercise. If ff is a function on the state space, we say that gg is a normal contraction of ff if ∣g(x)∣≤∣f(x)∣|g(x)| \leq |f(x)| for all xx and ∣g(x)−g(y)∣≤∣f(x)−f(y)∣|g(x)-g(y)| \leq |f(x)-f(y)| for all xx and yy. As an example, note that if g(x)=−1∨(f(x)∧1)g(x)=-1 \vee (f(x) \wedge 1), then gg is a normal contraction of ff. Prove that if f∈D(E)f \in \mathcal{D}(\mathcal{E}), where E\mathcal{E} is a Dirichlet form and gg is a normal contraction of ff, then for each t>0t>0,

E(t)(g,g)≤E(t)(f,f)≤E(f,f). \mathcal{E}^{(t)}(g, g) \leq \mathcal{E}^{(t)}(f, f) \leq \mathcal{E}(f, f) .
?