Chapter 7

Brownian motion and transition semigroups

[21/100%]
Show
LaTeX
Problem 7.1

Show that C∞:={u:Rd→R continuous and lim⁡∣x∣→∞u(x)=0}\mathcal{C}_{\infty }: = \left\{ u: \mathbb {R}^{d} \rightarrow \mathbb {R} \text{ continuous and } \lim_{\left|x\right| \rightarrow \infty } u(x) = 0\right\} equipped with the uniform topology is a Banach space. Show that in this topology C∞\mathcal{C}_{\infty } is the closure of Cc\mathcal{C}_{c}, i.e. the family of continuous functions with compact support.

?
Problem 7.2

Let (Bt)t⩾0\left(B_{t}\right)_{t \geqslant 0} be a BMd\mathrm{BM}^{d} with transition semigroup (Pt)t⩾0\left(P_{t}\right)_{t \geqslant 0}. Show, using arguments similar to those in the proof of Proposition 7.3.f), that (t,x,u)↦Ptu(x)(t, x, u) \mapsto P_{t} u(x) is continuous. As usual, we equip [0,∞)×Rd×C∞(Rd)[0, \infty ) \times \mathbb {R}^{d} \times \mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right) with the norm ∣t∣+∣x∣+∥u∥∞\left|t\right|+\left|x\right|+\left\| u\right\|_{\infty }.

?
Problem 7.3

Let (Xt)t⩾0\left(X_{t}\right)_{t \geqslant 0} be a dd - dimensional Feller process and let f,g∈C∞(Rd)f, g \in \mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right). Show that the function x↦Ex(f(Xt)g(Xt+s))x \mapsto \mathbb {E}^{x}\left(f\left(X_{t}\right) g\left(X_{t+s}\right)\right) is also in C∞(Rd)\mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right).

?
Problem 7.4

Let (Bt)t⩾0\left(B_{t}\right)_{t \geqslant 0} be a BMd\mathrm{BM}^{d} and set u(t,x):=Ptu(x)u(t, x): = P_{t} u(x). Adapt the proof of Proposition 7.3.g) and show that for u∈Bb(Rd)u \in \mathcal{B}_{b}\left(\mathbb {R}^{d}\right) the function u(t,⋅)∈C∞u(t, \cdot ) \in \mathcal{C}^{\infty } and u∈C1,2u \in \mathcal{C}^{1,2} (i.e. once continuously differentiable in tt, twice continuously differentiable in xx ).

?
Problem 7.5

Let (Pt)t⩾0\left(P_{t}\right)_{t \geqslant 0} be the transition semigroup of a BMd\mathrm{BM}^{d} and denote by Lp,1⩽p<∞L^{p}, 1 \leqslant p < \infty, the space of pp th power integrable functions with respect to Lebesgue measure on Rd\mathbb {R}^{d}. Set un:=(−n)∨u∧nu_{n}: = ( - n) \vee u \wedge n for every u∈Lpu \in L^{p}. Verify that

?
(a)

un∈Bb(Rd)∩Lp(Rd)u_{n} \in \mathcal{B}_{b}\left(\mathbb {R}^{d}\right) \cap L^{p}\left(\mathbb {R}^{d}\right) and Lp−lim⁡n→∞un=uL^{p} - \lim_{n \rightarrow \infty } u_{n} = u.

(b)

P~tu:=lim⁡n→∞Ptun\widetilde{P}_{t} u: = \lim_{n \rightarrow \infty } P_{t} u_{n} extends PtP_{t} and gives a contraction semigroup on LpL^{p}.

(c)

P~t\widetilde{P}_{t} is sub - Markovian, i.e. if u∈Lpu \in L^{p} and 0⩽u⩽10 \leqslant u \leqslant 1 a.e., then 0⩽P~tu⩽10 \leqslant \widetilde{P}_{t} u \leqslant 1 a. e.

(d)

P~t\widetilde{P}_{t} is strongly continuous, i.e. lim⁡t→0∥P~tu−u∥Lp=0\lim_{t \rightarrow 0}\left\| \widetilde{P}_{t} u - u\right\|_{L^{p}} = 0 for all u∈Lpu \in L^{p}.

Problem 7.6

Let (Tt)t⩾0\left(T_{t}\right)_{t \geqslant 0} be a Markov semigroup given by Ttu(x)=∫Rdu(y)pt(x,dy)T_{t} u(x) = \int_{\mathbb {R}^{d}} u(y) p_{t}(x, d y) where pt(x,C)p_{t}(x, C) is a kernel in the sense of Remark 7.6. Show that the semigroup property of (Tt)t⩾0\left(T_{t}\right)_{t \geqslant 0} entails the Chapman - Kolmogorov equations

pt+s(x,C)=∫pt(y,C)ps(x,dy) for all x∈Rd and C∈B(Rd) p_{t+s}(x, C) = \int p_{t}(y, C) p_{s}(x, d y) \quad \text{ for all } \quad x \in \mathbb {R}^{d} \quad \text{ and } \quad C \in \mathscr {B}\left(\mathbb {R}^{d}\right)
?
Problem 7.7

Show that pt1,…,tnx(C1×⋯×Cn)p_{t_{1}, \ldots , t_{n}}^{x}\left(C_{1} \times \cdots \times C_{n}\right) of Remark 7.6 define measures on B(Rd⋅n)\mathscr {B}\left(\mathbb {R}^{d \cdot n}\right).

?
Problem 7.8
?
(a)

Let d(x,A):=inf⁡a∈A∣x−a∣d(x, A): = \inf_{a \in A}\left|x - a\right| be the distance between the point x∈Rdx \in \mathbb {R}^{d} and the set A∈B(Rd)A \in \mathscr {B}\left(\mathbb {R}^{d}\right). Show that x↦d(x,A)x \mapsto d(x, A) is continuous.

(b)

Let unu_{n} be as in (7.7). Show that un∈Cc(Rd)u_{n} \in \mathcal{C}_{c}\left(\mathbb {R}^{d}\right) and inf⁡nun(x)=  1K(x)\inf_{n} u_{n}(x) = \; \mathbb {1}_{K}(x). Draw a picture if d=1d = 1 and K=[0,1]K = [0,1].

(c)

Let χn\chi_{n} be a sequence of type δ\delta, i.e. χn∈Cc∞(Rd)\chi_{n} \in \mathcal{C}_{c}^{\infty }\left(\mathbb {R}^{d}\right), supp⁡χn⊂B(0,1/n),χn⩾0\operatorname {supp} \chi_{n} \subset \mathbb {B}(0,1 / n), \chi_{n} \geqslant 0 and ∫χn(x)dx=1\int \chi_{n}(x) d x = 1. Show that vn:=χn⋆un∈Cc∞(Rd)v_{n}: = \chi_{n} \star u_{n} \in \mathcal{C}_{c}^{\infty }\left(\mathbb {R}^{d}\right) are smooth functions such that lim⁡n→∞vn=  1K\lim_{n \rightarrow \infty } v_{n} = \; \mathbb {1}_{K}.

Problem 7.9

Let (Tt)t⩾0\left(T_{t}\right)_{t \geqslant 0} be a Feller semigroup, i.e. a strongly continuous, positivity preserving sub - Markovian semigroup on (C∞(Rd),∥⋅∥∞)\left(\mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right),\left\| \cdot \right\|_{\infty }\right), cf. Remark 7.4.

?
(a)

Show that (t,x)↦Ttu(x)(t, x) \mapsto T_{t} u(x) is jointly continuous for t⩾0t \geqslant 0 and x∈Rdx \in \mathbb {R}^{d}.

(b)

Conclude that for every compact set K⊂RdK \subset \mathbb {R}^{d} the function (t,x)↦Tt  1K(x)(t, x) \mapsto T_{t} \; \mathbb {1}_{K}(x) is measurable.

(c)

Show that (t,x,u)↦Ttu(x)(t, x, u) \mapsto T_{t} u(x) is jointly continuous for all t⩾0,x∈Rdt \geqslant 0, x \in \mathbb {R}^{d} and u∈C∞(Rd)u \in \mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right).

Problem 7.10

Let A,B∈Rd×dA, B \in \mathbb {R}^{d \times d} and set Pt:=exp⁡(tA):=∑j=0∞(tA)j/jP_{t}: = \exp (t A): = \sum_{j = 0}^{\infty }(t A)^{j} / j !.

?
(a)

Show that PtP_{t} is a strongly continuous semigroup. Is it contractive?

(b)

Show that ddtetA\frac{d}{d t} e^{t A} exists and that ddtetA=AetA=etAA\frac{d}{d t} e^{t A} = A e^{t A} = e^{t A} A.

(c)

Show that etAetB=etBetA∀t⟺AB=BAe^{t A} e^{t B} = e^{t B} e^{t A} \forall t \Longleftrightarrow A B = B A.

(d)

(Trotter product formula) Show that eA+B=lim⁡k→∞(eA/keB/k)ke^{A+B} = \lim_{k \rightarrow \infty }\left(e^{A / k} e^{B / k}\right)^{k}.

Problem 7.11

Let (Pt)t⩾0\left(P_{t}\right)_{t \geqslant 0} and (Tt)t⩾0\left(T_{t}\right)_{t \geqslant 0} be two Feller semigroups with generators (A,D(A))(A, \mathfrak {D}(A)), resp. (B,D(B))(B, \mathfrak {D}(B)).

?
(a)

If D(B)⊂D(A)\mathfrak {D}(B) \subset \mathfrak {D}(A), then we have ddsPt−sTs=−Pt−sATs+Pt−sBTs\frac{d}{d s} P_{t - s} T_{s} = - P_{t - s} A T_{s}+P_{t - s} B T_{s} on D(B)\mathfrak {D}(B).

(b)

Is it possible that Ut:=TtPtU_{t}: = T_{t} P_{t} is again a Feller semigroup?

(c)

(Trotter product formula) Assume that Utf:=lim⁡n→∞(Tt/nPt/n)nfU_{t} f: = \lim_{n \rightarrow \infty }\left(T_{t / n} P_{t / n}\right)^{n} f exists strongly and locally uniformly for tt. Show that UtU_{t} is a Feller semigroup and determine its generator.

Problem 7.12

Complete the following alternative argument for Example 7.16. Assume that (un)n⩾1⊂C∞2(R)\left(u_{n}\right)_{n \geqslant 1} \subset \mathcal{C}_{\infty }^{2}(\mathbb {R}) such that (12un′′)n⩾1\left(\frac{1}{2} u_{n}^{\prime \prime }\right)_{n \geqslant 1} is a Cauchy sequence in C∞\mathcal{C}_{\infty } and that unu_{n} converges uniformly to uu.

?
(a)

Show that un(x)−un(0)−xun′(0)=∫0x∫0yun′′(z)dzdy→∫0x∫0y2g(z)dzdyu_{n}(x) - u_{n}(0) - x u_{n}^{\prime }(0) = \int_{0}^{x} \int_{0}^{y} u_{n}^{\prime \prime }(z) d z d y \rightarrow \int_{0}^{x} \int_{0}^{y} 2 g(z) d z d y uniformly. Conclude that c′:=lim⁡n→∞un′(0)c^{\prime }: = \lim_{n \rightarrow \infty } u_{n}^{\prime }(0) exists.

(b)

Show that un′(x)−un′(0)→∫0x2g(z)dzu_{n}^{\prime }(x) - u_{n}^{\prime }(0) \rightarrow \int_{0}^{x} 2 g(z) d z uniformly. Deduce from this and part a) that un′(x)u_{n}^{\prime }(x) converges uniformly to ∫0x2g(z)dz+c′\int_{0}^{x} 2 g(z) d z+c^{\prime } where c′=∫−∞02g(z)dzc^{\prime } = \int_{ - \infty }^{0} 2 g(z) d z.

(c)

Use a) and b) to show that un(x)−un(0)→∫0x∫−∞y2g(z)dzu_{n}(x) - u_{n}(0) \rightarrow \int_{0}^{x} \int_{ - \infty }^{y} 2 g(z) d z uniformly and argue that un(x)u_{n}(x) has the uniform limit ∫−∞x∫−∞y2g(z)dzdy\int_{ - \infty }^{x} \int_{ - \infty }^{y} 2 g(z) d z d y.

Problem 7.13

Let UαU_{\alpha } be the α\alpha - potential operator of a BMd\mathrm{BM}^{d}. Give a probabilistic interpretation of Uα  1CU_{\alpha } \; \mathbb {1}_{C} and lim⁡α→0Uα  1C\lim_{\alpha \rightarrow 0} U_{\alpha } \; \mathbb {1}_{C} if C⊂RdC \subset \mathbb {R}^{d} is a Borel set.

?
Problem 7.14

Let U0U_{0} be the potential operator of a BMd\mathrm{BM}^{d} in dimension d=1d = 1 or d=2d = 2. Show that every u∈D(U0)u \in \mathfrak {D}\left(U_{0}\right) such that u⩾0u \geqslant 0 is trivial, i.e. u=0u = 0.

?
Problem 7.15

Let (Uα)α>0\left(U_{\alpha }\right)_{\alpha > 0} be the α\alpha - potential operator of a BMd\mathrm{BM}^{d}. Use the resolvent equation to prove the following formulae for f∈Bbf \in \mathcal{B}_{b} and x∈Rdx \in \mathbb {R}^{d} :

dndαnUαf(x)=n!(−1)nUαn+1f(x) \frac{d^{n}}{d \alpha ^{n}} U_{\alpha } f(x) = n !( - 1)^{n} U_{\alpha }^{n+1} f(x)

and

dndαn(αUα)f(x)=n!(−1)n+1(id−αUα)Uαnf(x) \frac{d^{n}}{d \alpha ^{n}}\left(\alpha U_{\alpha }\right) f(x) = n !( - 1)^{n+1}\left(\mathrm{id} - \alpha U_{\alpha }\right) U_{\alpha }^{n} f(x)
?
Problem 7.16

Let (fn)n⩾1⊂C∞(Rd)\left(f_{n}\right)_{n \geqslant 1} \subset \mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right) be a sequence of functions such that 0⩽fn⩽fn+10 \leqslant f_{n} \leqslant f_{n+1} and f:=sup⁡nfn∈C∞(Rd)f: = \sup_{n} f_{n} \in \mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right). Show that lim⁡n→∞∥f−fn∥∞=0\lim_{n \rightarrow \infty }\left\| f - f_{n}\right\|_{\infty } = 0.

?
Remark.
?

Positivity is in fact not needed. If instead f1⩽f2⩽⋯⩽fn⩽⋯f_{1} \leqslant f_{2} \leqslant \cdots \leqslant f_{n} \leqslant \cdots, fn∈C∞(Rd)f_{n}\in \mathcal{C}_{\infty }\left(\mathbb {R}^{d}\right) of any sign, and f=sup⁡nfnf=\sup_{n}f_{n}, then by Dini's theorem lim⁡n→∞sup⁡x∈K∣f(x)−fn(x)∣=0\lim_{n\rightarrow \infty }\sup_{x\in K}\left|f(x)-f_{n}(x)\right|=0 for any compact K⊂RdK\subset \mathbb {R}^{d}; fixing ε>0\varepsilon >0 and a compact KεK_{\varepsilon } with −ε⩽f1(x)⩽fn(x)⩽f(x)⩽ε-\varepsilon \leqslant f_{1}(x)\leqslant f_{n}(x)\leqslant f(x)\leqslant \varepsilon on Rd∖Kε\mathbb {R}^{d}\setminus K_{\varepsilon } gives lim sup⁡nsup⁡x∣fn(x)−f(x)∣⩽lim sup⁡nsup⁡x∈Kε∣fn(x)−f(x)∣+lim sup⁡nsup⁡x∉Kε(f(x)−fn(x))⩽0+2ε⩽3ε\limsup_{n}\sup_{x}\left|f_{n}(x)-f(x)\right| \leqslant \limsup_{n}\sup_{x\in K_{\varepsilon }}\left|f_{n}(x)-f(x)\right|+\limsup_{n}\sup_{x\notin K_{\varepsilon }}(f(x)-f_{n}(x)) \leqslant 0+2\varepsilon \leqslant 3\varepsilon, i.e. the same conclusion holds without positivity. (Alternatively: apply the positive-case argument to gn:=fn−f1g_{n}:=f_{n}-f_{1}, using 0⩽gn⩽gn+10\leqslant g_{n}\leqslant g_{n+1} and sup⁡ngn=f−f1\sup_{n}g_{n}=f-f_{1}.)

Problem 7.17

Let (A,D(A))(A, \mathfrak {D}(A)) be the generator of a BMd\mathrm{BM}^{d}. Adapt the arguments of Example 7.25 and show that C∞2(Rd)⫋D(A)\mathcal{C}_{\infty }^{2}\left(\mathbb {R}^{d}\right) \varsubsetneqq \mathfrak {D}(A) for any dimension d⩾3d \geqslant 3. Instructions. Use in Example 7.25d7.25 d - dimensional polar coordinates: y1=rsin⁡ϕsin⁡θ1⋅…⋅sin⁡θd−2,y2=rcos⁡ϕsin⁡θ1⋅…⋅sin⁡θd−2,…yd−1=rcos⁡θd−3⋅sin⁡θd−2y_{1} = r \sin \phi \sin \theta_{1} \cdot \ldots \cdot \sin \theta_{d - 2}, y_{2} = r \cos \phi \sin \theta_{1} \cdot \ldots \cdot \sin \theta_{d - 2}, \ldots y_{d - 1} = r \cos \theta_{d - 3} \cdot \sin \theta_{d - 2}, yd=rcos⁡θd−2y_{d} = r \cos \theta_{d - 2}, where (r,ϕ,θ1,…,θd−2)∈(0,∞)×(0,2π)×(0,π)d−2\left(r, \phi , \theta_{1}, \ldots , \theta_{d - 2}\right) \in (0, \infty ) \times (0,2 \pi ) \times (0, \pi )^{d - 2}.

The functional determinant is

rd−1(sin⁡θ1)1(sin⁡θ2)2…(sin⁡θd−2)d−2 r^{d - 1}\left(\sin \theta _{1}\right)^{1}\left(\sin \theta _{2}\right)^{2} \ldots \left(\sin \theta _{d - 2}\right)^{d - 2}

The calculations will lead to the following integrals (∣x∣<1)(\left|x\right| < 1)

∫0π(sin⁡θ)d−2(1+x2−2xcos⁡θ)(d−2)/2dθ and ∫0π(sin⁡θ)d(1+x2−2xcos⁡θ)(d−2)/2dθ \int _{0}^{\pi } \frac{(\sin \theta )^{d - 2}}{\left(1+x^{2} - 2 x \cos \theta \right)^{(d - 2) / 2}} d \theta \text{ and } \int _{0}^{\pi } \frac{(\sin \theta )^{d}}{\left(1+x^{2} - 2 x \cos \theta \right)^{(d - 2) / 2}} d \theta

which have the values ada_{d} and bdx2+cdb_{d} x^{2}+c_{d}, respectively, with constants ad,bd,cd∈Ra_{d}, b_{d}, c_{d} \in \mathbb {R} depending on the dimension, cf. Gradshteyn - Ryzhik [101, 3.665.2, p. 384].

Remark. A simple evaluation of the integrals ⋆{ }^{\star } ) is possible only in d=3d = 3. For d>3d > 3 these integrals are most elegantly evaluated using Gegenbauer polynomials, cf. [101, 8.93, pp. 1029 - 1031] and the discussion on stack exchange http://math.stackexchange.com/questions/395336.

?
Problem 7.18

Let (Bt)t⩾0\left(B_{t}\right)_{t \geqslant 0} be a BM1\mathrm{BM}^{1} and consider the two - dimensional process Xt:=(t,Bt),t⩾0X_{t}: = \left(t, B_{t}\right), t \geqslant 0.

?
(a)

Show that (Xt)t⩾0\left(X_{t}\right)_{t \geqslant 0} is a Feller process.

(b)

Determine its transition semigroup, resolvent and generator.

(c)

State and prove Theorem 7.30 for this process and compare the result with Theorem 5.6

Problem 7.19

Show that Dynkin's formula (7.30) in Proposition 7.31 follows from Theorem 7.30.

?
Problem 7.20

Let t↦Xtt \mapsto X_{t} be a right continuous stochastic process. Show that for closed sets FF

P(Xt∈F∀t∈R+)=P(Xq∈F∀q∈Q+) \mathbb {P}\left(X_{t} \in F \quad \forall t \in \mathbb {R}^{+}\right) = \mathbb {P}\left(X_{q} \in F \quad \forall q \in \mathbb {Q}^{+}\right)
?
Problem 7.21

Assume that L:Cc∞(Rd)→C(Rd)L: \mathcal{C}_{c}^{\infty }\left(\mathbb {R}^{d}\right) \rightarrow \mathcal{C}\left(\mathbb {R}^{d}\right) is a local operator (cf. Definition 7.36) which is almost positive, i.e. for all ϕ∈Cc∞\phi \in \mathcal{C}_{c}^{\infty } such that ϕ⩾0\phi \geqslant 0 and ϕ(x0)=0\phi \left(x_{0}\right) = 0 we have Lϕ(x0)⩾0L \phi \left(x_{0}\right) \geqslant 0.

?
(a)

Show that an operator that satisfies the positive maximum principle is almost positive. (This does not require locality).

(b)

Show that supp⁡Lϕ⊂supp⁡ϕ\operatorname {supp} L \phi \subset \operatorname {supp} \phi.

(c)

Show that the operator LL is of order two, i.e. ∥Lϕ∥∞⩽C∥ϕ∥(2)\left\| L \phi \right\|_{\infty } \leqslant C\left\| \phi \right\|_{(2)} holds for all ϕ∈Cc∞(K)\phi \in \mathcal{C}_{c}^{\infty }(K) and any compact set K⊂RdK \subset \mathbb {R}^{d}. Here, ∥ϕ∥(2)=∑0⩽∣α∣⩽2∥∂αϕ∥∞\left\| \phi \right\|_{(2)} = \sum_{0 \leqslant \left|\alpha \right| \leqslant 2}\left\| \partial^{\alpha } \phi \right\|_{\infty } and the constant CC depends only on KK.

(d)

Use Theorem 7.37 to conclude that LL is a differential operator of order ⩽2\leqslant 2.