Chapter 14

Local times

[10/0%]
LaTeX
Problem 14.1

Suppose ff is C2C^{2} with compact support and

g(x)=∫f(y)∣x−y∣dy g(x)=\int f(y)|x-y| d y

Show that gg is C2C^{2} and g′′=2fg^{\prime \prime }=2 f.

?
Problem 14.2

Let LtyL_{t}^{y} be the jointly continuous local times of a Brownian motion WW. Show

12ε∫0t  1[y−ε,y+ε](Ws)ds→Lty, a.s.  \frac{1}{2 \varepsilon } \int _{0}^{t} \; \mathbb {1}_{[y-\varepsilon , y+\varepsilon ]}\left(W_{s}\right) d s \rightarrow L_{t}^{y}, \quad \text{ a.s. }

Show the null set can be taken to be independent of yy. Thus there is no need to take a subsequence εn\varepsilon_{n} to get almost sure convergence to LtyL_{t}^{y}.

?
Problem 14.3

Let WW be a Brownian motion and fix tt. Show that the function x→∫0t  1(−∞,x](Ws)dsx \rightarrow \int_{0}^{t} \; \mathbb {1}_{(-\infty , x]}\left(W_{s}\right) d s is continuous, a.s., but that the function x→  1(−∞,x](Wt)x \rightarrow \; \mathbb {1}_{(-\infty , x]}\left(W_{t}\right) is not continuous.

?
Problem 14.4

Let {Ft}\left\{ \mathcal{F}_{t}\right\} be a filtration satisfying the usual conditions. Suppose WtW_{t} is a Brownian motion and Xt=Wt+AtX_{t}=W_{t}+A_{t}, where Xt≥0X_{t} \geq 0 for all tt, a.s., and AtA_{t} is an increasing continuous adapted process such that AA increases only at those times when Xt=0X_{t}=0. Suppose also that Xt′=Wt+At′X_{t}^{\prime }=W_{t}+A_{t}^{\prime }, where Xt′≥0X_{t}^{\prime } \geq 0 for all tt, a.s., and At′A_{t}^{\prime } is an increasing continuous adapted process that increases only when Xt′=0X_{t}^{\prime }=0. Show that Xt′=XtX_{t}^{\prime }=X_{t} and At=At′A_{t}=A_{t}^{\prime }, a.s., for all t≥0t \geq 0.

?
Problem 14.5

Let WW be a Brownian motion and Lt0L_{t}^{0} the local time at 0 . Since Lt0L_{t}^{0} is increasing, for each ω\omega there is a Lebesgue-Stieltjes measure dLt0d L_{t}^{0}. Show that the support of dLt0d L_{t}^{0} is equal to {t:Wt=0}\left\{ t: W_{t}=0\right\}.

Since Theorem 14.1(2) states that Lt0L_{t}^{0} does not increase when WtW_{t} is not equal to 0 , what you need to show is that with probability one, if Wu(ω)=0W_{u}(\omega )=0 and t<u<vt<u<v, then Lv0(ω)>Lt0(ω)L_{v}^{0}(\omega )>L_{t}^{0}(\omega ).

?
Problem 14.6

Use Tanaka's formula to show that if LtyL_{t}^{y} is the local time of Brownian motion at level yy, a≤x≤y≤ba \leq x \leq y \leq b, and T=inf⁡{t>0:Wt∉[a,b]}T=\inf \left\{ t>0: W_{t} \notin [a, b]\right\}, then

Ex[LTy]=2(x−a)(b−y)b−a \mathbb {E}_{x}\left[L_{T}^{y}\right]=\frac{2(x-a)(b-y)}{b-a}
?
Problem 14.7

If Lt0L_{t}^{0} is the local time of a Brownian motion at 0 , show that Lat0L_{a t}^{0} has the same law as aLt0\sqrt{a} L_{t}^{0}.

?
Problem 14.8

Let WW be a Brownian motion with local times LtyL_{t}^{y}. Set Lt∗=sup⁡yLtyL_{t}^{*}=\sup_{y} L_{t}^{y}. Let p>0p>0. Prove that there exist constants c1,c2c_{1}, c_{2} such that if TT is any finite stopping time,

c1E[Tp/2]≤E[LT∗]≤c2E[Tp/2] c_{1} \mathbb {E}\left[T^{p / 2}\right] \leq \mathbb {E}\left[L_{T}^{*}\right] \leq c_{2} \mathbb {E}\left[T^{p / 2}\right]

The constants c1,c2c_{1}, c_{2} can depend on pp, but not on TT.

Hint: Use Exercise 12.11.

?
Problem 14.9

This exercise defines the local time of a continuous martingale. If MM is a continuous martingale, then Mt2M_{t}^{2} is a submartingale and so equals a martingale plus an increasing process. The increasing process Lt0L_{t}^{0} is called the local time of MM at 0 .

?
(1)

Prove the analog of Tanaka's formula.

(2)

Define the local time LtaL_{t}^{a} of MM at aa. Prove that LtaL_{t}^{a} is jointly continuous in tt and aa.

(3)

Prove that

∫0tf(Ms)d⟨M⟩s=∫RLtaf(a)da, a.s.  \int _{0}^{t} f\left(M_{s}\right) d\langle M\rangle _{s}=\int _{\mathbb {R}} L_{t}^{a} f(a) d a, \quad \text{ a.s. }

if ff is non-negative and measurable.

Problem 14.10

This exercise is a complement to Exercise 7.8. Let WW be a Brownian motion and let us define Z={t∈[0,1]:Wt=0}Z=\left\{ t \in [0,1]: W_{t}=0\right\}, the zero set. Let ε∈(0,1/2)\varepsilon \in (0,1 / 2) and let δ>0\delta >0. Fix ω\omega and let {Bi}\left\{ B_{i}\right\} be any countable covering of Z(ω)Z(\omega ) by closed intervals such that the interiors of the BiB_{i} 's are pairwise disjoint and the length of each BiB_{i} is less than or equal to δ\delta. We write Bi=[ai,bi]B_{i}=\left[a_{i}, b_{i}\right].

Let ε>0\varepsilon >0. Since L0L^{0} has the same law of the maximum of Brownian motion, there exists a cc (depending on ω\omega ) such that

Lt0−Ls0≤c(t−s)12−ε2 L_{t}^{0}-L_{s}^{0} \leq c(t-s)^{\frac{1}{2}-\frac{\varepsilon }{2}}

for each 0≤s≤t≤00 \leq s \leq t \leq 0. Write

∑i∣bi−ai∣12−ε≥δ−ε/2cc∑i∣bi−ai∣12−ε2≥δ−ε/2c∑i(Lbi0−Lai0)=δ−ε/2c[L10−L00]. \begin{aligned} \sum _{i}\left|b_{i}-a_{i}\right|^{\frac{1}{2}-\varepsilon } & \geq \frac{\delta ^{-\varepsilon / 2}}{c} c \sum _{i}\left|b_{i}-a_{i}\right|^{\frac{1}{2}-\frac{\varepsilon }{2}} \\ & \geq \frac{\delta ^{-\varepsilon / 2}}{c} \sum _{i}\left(L_{b_{i}}^{0}-L_{a_{i}}^{0}\right) \\ & =\frac{\delta ^{-\varepsilon / 2}}{c}\left[L_{1}^{0}-L_{0}^{0}\right] . \end{aligned}

Show that this implies that the Hausdorff dimension of ZZ is at least 1/21 / 2.

?