18.175: lecture 37 .1in more brownian motionmath.mit.edu/~sheffield/175/lecture37.pdf ·...
TRANSCRIPT
![Page 1: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/1.jpg)
18.175: Lecture 37
More Brownian motion
Scott Sheffield
MIT
18.175 Lecture 37
![Page 2: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/2.jpg)
Outline
Brownian motion properties and construction
Markov property, Blumenthal’s 0-1 law
18.175 Lecture 37
![Page 3: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/3.jpg)
Outline
Brownian motion properties and construction
Markov property, Blumenthal’s 0-1 law
18.175 Lecture 37
![Page 4: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/4.jpg)
Basic properties
I Brownian motion is real-valued process Bt , t ≥ 0.
I Independent increments: If t0 < t1 < t2 . . . thenB(t0),B(t1 − t0),B(t2 − t1), . . . are independent.
I Gaussian increments: If s, t ≥ 0 then B(s + t)− B(s) isnormal with variance t.
I Continuity: With probability one, t → Bt is continuous.
I Hmm... does this mean we need to use a σ-algebra in whichthe event “Bt is continuous” is a measurable?
I Suppose Ω is set of all functions of t, and we use smallestσ-field that makes each Bt a measurable random variable...does that fail?
18.175 Lecture 37
![Page 5: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/5.jpg)
Basic properties
I Brownian motion is real-valued process Bt , t ≥ 0.
I Independent increments: If t0 < t1 < t2 . . . thenB(t0),B(t1 − t0),B(t2 − t1), . . . are independent.
I Gaussian increments: If s, t ≥ 0 then B(s + t)− B(s) isnormal with variance t.
I Continuity: With probability one, t → Bt is continuous.
I Hmm... does this mean we need to use a σ-algebra in whichthe event “Bt is continuous” is a measurable?
I Suppose Ω is set of all functions of t, and we use smallestσ-field that makes each Bt a measurable random variable...does that fail?
18.175 Lecture 37
![Page 6: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/6.jpg)
Basic properties
I Brownian motion is real-valued process Bt , t ≥ 0.
I Independent increments: If t0 < t1 < t2 . . . thenB(t0),B(t1 − t0),B(t2 − t1), . . . are independent.
I Gaussian increments: If s, t ≥ 0 then B(s + t)− B(s) isnormal with variance t.
I Continuity: With probability one, t → Bt is continuous.
I Hmm... does this mean we need to use a σ-algebra in whichthe event “Bt is continuous” is a measurable?
I Suppose Ω is set of all functions of t, and we use smallestσ-field that makes each Bt a measurable random variable...does that fail?
18.175 Lecture 37
![Page 7: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/7.jpg)
Basic properties
I Brownian motion is real-valued process Bt , t ≥ 0.
I Independent increments: If t0 < t1 < t2 . . . thenB(t0),B(t1 − t0),B(t2 − t1), . . . are independent.
I Gaussian increments: If s, t ≥ 0 then B(s + t)− B(s) isnormal with variance t.
I Continuity: With probability one, t → Bt is continuous.
I Hmm... does this mean we need to use a σ-algebra in whichthe event “Bt is continuous” is a measurable?
I Suppose Ω is set of all functions of t, and we use smallestσ-field that makes each Bt a measurable random variable...does that fail?
18.175 Lecture 37
![Page 8: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/8.jpg)
Basic properties
I Brownian motion is real-valued process Bt , t ≥ 0.
I Independent increments: If t0 < t1 < t2 . . . thenB(t0),B(t1 − t0),B(t2 − t1), . . . are independent.
I Gaussian increments: If s, t ≥ 0 then B(s + t)− B(s) isnormal with variance t.
I Continuity: With probability one, t → Bt is continuous.
I Hmm... does this mean we need to use a σ-algebra in whichthe event “Bt is continuous” is a measurable?
I Suppose Ω is set of all functions of t, and we use smallestσ-field that makes each Bt a measurable random variable...does that fail?
18.175 Lecture 37
![Page 9: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/9.jpg)
Basic properties
I Brownian motion is real-valued process Bt , t ≥ 0.
I Independent increments: If t0 < t1 < t2 . . . thenB(t0),B(t1 − t0),B(t2 − t1), . . . are independent.
I Gaussian increments: If s, t ≥ 0 then B(s + t)− B(s) isnormal with variance t.
I Continuity: With probability one, t → Bt is continuous.
I Hmm... does this mean we need to use a σ-algebra in whichthe event “Bt is continuous” is a measurable?
I Suppose Ω is set of all functions of t, and we use smallestσ-field that makes each Bt a measurable random variable...does that fail?
18.175 Lecture 37
![Page 10: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/10.jpg)
Basic properties
I Translation invariance: is Bt0+t − Bt0 a Brownian motion?
I Brownian scaling: fix c , then Bct agrees in law with c1/2Bt .
I Another characterization: B is jointly Gaussian, EBs = 0,EBsBt = s ∧ t, and t → Bt a.s. continuous.
18.175 Lecture 37
![Page 11: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/11.jpg)
Basic properties
I Translation invariance: is Bt0+t − Bt0 a Brownian motion?
I Brownian scaling: fix c , then Bct agrees in law with c1/2Bt .
I Another characterization: B is jointly Gaussian, EBs = 0,EBsBt = s ∧ t, and t → Bt a.s. continuous.
18.175 Lecture 37
![Page 12: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/12.jpg)
Basic properties
I Translation invariance: is Bt0+t − Bt0 a Brownian motion?
I Brownian scaling: fix c , then Bct agrees in law with c1/2Bt .
I Another characterization: B is jointly Gaussian, EBs = 0,EBsBt = s ∧ t, and t → Bt a.s. continuous.
18.175 Lecture 37
![Page 13: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/13.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 14: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/14.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 15: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/15.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 16: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/16.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 17: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/17.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 18: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/18.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 19: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/19.jpg)
Defining Brownian motion
I Can define joint law of Bt values for any finite collection ofvalues.
I Can observe consistency and extend to countable set byKolmogorov. This gives us measure in σ-field F0 generated bycylinder sets.
I But not enough to get a.s. continuity.
I Can define Brownian motion jointly on diadic rationals prettyeasily. And claim that this a.s. extends to continuous path inunique way.
I We can use the Kolmogorov continuity theorem (next slide).
I Can prove Holder continuity using similar estimates (seeproblem set).
I Can extend to higher dimensions: make each coordinateindependent Brownian motion.
18.175 Lecture 37
![Page 20: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/20.jpg)
Continuity theorem
I Kolmogorov continuity theorem: SupposeE |Xs − Xt |β ≤ K |t − s|1+α where α, β > 0. If γ < α/β thenwith probability one there is a constant C (ω) so that|X (q)− X (r)| ≤ C |q − r |γ for all q, r ∈ Q2 ∩ [0, 1].
I Proof idea: First look at values at all multiples of 2−0, thenat all multiples of 2−1, then multiples of 2−2, etc.
I At each stage we can draw a nice piecewise linearapproximation of the process. How much does theapproximation change in supremum norm (or some otherHolder norm) on the ith step? Can we say it probably doesn’tchange very much? Can we say the sequence ofapproximations is a.s. Cauchy in the appropriate normedspaced?
18.175 Lecture 37
![Page 21: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/21.jpg)
Continuity theorem
I Kolmogorov continuity theorem: SupposeE |Xs − Xt |β ≤ K |t − s|1+α where α, β > 0. If γ < α/β thenwith probability one there is a constant C (ω) so that|X (q)− X (r)| ≤ C |q − r |γ for all q, r ∈ Q2 ∩ [0, 1].
I Proof idea: First look at values at all multiples of 2−0, thenat all multiples of 2−1, then multiples of 2−2, etc.
I At each stage we can draw a nice piecewise linearapproximation of the process. How much does theapproximation change in supremum norm (or some otherHolder norm) on the ith step? Can we say it probably doesn’tchange very much? Can we say the sequence ofapproximations is a.s. Cauchy in the appropriate normedspaced?
18.175 Lecture 37
![Page 22: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/22.jpg)
Continuity theorem
I Kolmogorov continuity theorem: SupposeE |Xs − Xt |β ≤ K |t − s|1+α where α, β > 0. If γ < α/β thenwith probability one there is a constant C (ω) so that|X (q)− X (r)| ≤ C |q − r |γ for all q, r ∈ Q2 ∩ [0, 1].
I Proof idea: First look at values at all multiples of 2−0, thenat all multiples of 2−1, then multiples of 2−2, etc.
I At each stage we can draw a nice piecewise linearapproximation of the process. How much does theapproximation change in supremum norm (or some otherHolder norm) on the ith step? Can we say it probably doesn’tchange very much? Can we say the sequence ofapproximations is a.s. Cauchy in the appropriate normedspaced?
18.175 Lecture 37
![Page 23: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/23.jpg)
Continuity theorem proof
I Kolmogorov continuity theorem: SupposeE |Xs − Xt |β ≤ K |t − s|1+α where α, β > 0. If γ < α/β thenwith probability one there is a constant C (ω) so that|X (q)− X (r)| ≤ C |q − r |γ for all q, r ∈ Q2 ∩ [0, 1].
I Argument from Durrett (Pemantle): Write
Gn = |X (i/2n)−X ((i−1)/2n)| ≤ C |q− r |λ for 0 < i ≤ 2n.
I Chebyshev implies P(|Y | > a) ≤ a−βE |Y |β, so ifλ = α− βγ > 0 then
P(G cn ) ≤ 2n · 2nβγ · E |X (j2−n)|β = K2−nλ.
18.175 Lecture 37
![Page 24: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/24.jpg)
Continuity theorem proof
I Kolmogorov continuity theorem: SupposeE |Xs − Xt |β ≤ K |t − s|1+α where α, β > 0. If γ < α/β thenwith probability one there is a constant C (ω) so that|X (q)− X (r)| ≤ C |q − r |γ for all q, r ∈ Q2 ∩ [0, 1].
I Argument from Durrett (Pemantle): Write
Gn = |X (i/2n)−X ((i−1)/2n)| ≤ C |q− r |λ for 0 < i ≤ 2n.
I Chebyshev implies P(|Y | > a) ≤ a−βE |Y |β, so ifλ = α− βγ > 0 then
P(G cn ) ≤ 2n · 2nβγ · E |X (j2−n)|β = K2−nλ.
18.175 Lecture 37
![Page 25: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/25.jpg)
Continuity theorem proof
I Kolmogorov continuity theorem: SupposeE |Xs − Xt |β ≤ K |t − s|1+α where α, β > 0. If γ < α/β thenwith probability one there is a constant C (ω) so that|X (q)− X (r)| ≤ C |q − r |γ for all q, r ∈ Q2 ∩ [0, 1].
I Argument from Durrett (Pemantle): Write
Gn = |X (i/2n)−X ((i−1)/2n)| ≤ C |q− r |λ for 0 < i ≤ 2n.
I Chebyshev implies P(|Y | > a) ≤ a−βE |Y |β, so ifλ = α− βγ > 0 then
P(G cn ) ≤ 2n · 2nβγ · E |X (j2−n)|β = K2−nλ.
18.175 Lecture 37
![Page 26: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/26.jpg)
Easy observations
I Brownian motion is Holder continuous for any γ < 1/2 (applytheorem with β = 2m, α = m − 1).
I Brownian motion is almost surely not differentiable.
I Brownian motion is almost surely not Lipschitz.
I Kolmogorov-Centsov theorem applies to higher dimensions(with adjusted exponents). One can construct a.s. continuousfunctions from Rn to R.
18.175 Lecture 37
![Page 27: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/27.jpg)
Easy observations
I Brownian motion is Holder continuous for any γ < 1/2 (applytheorem with β = 2m, α = m − 1).
I Brownian motion is almost surely not differentiable.
I Brownian motion is almost surely not Lipschitz.
I Kolmogorov-Centsov theorem applies to higher dimensions(with adjusted exponents). One can construct a.s. continuousfunctions from Rn to R.
18.175 Lecture 37
![Page 28: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/28.jpg)
Easy observations
I Brownian motion is Holder continuous for any γ < 1/2 (applytheorem with β = 2m, α = m − 1).
I Brownian motion is almost surely not differentiable.
I Brownian motion is almost surely not Lipschitz.
I Kolmogorov-Centsov theorem applies to higher dimensions(with adjusted exponents). One can construct a.s. continuousfunctions from Rn to R.
18.175 Lecture 37
![Page 29: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/29.jpg)
Easy observations
I Brownian motion is Holder continuous for any γ < 1/2 (applytheorem with β = 2m, α = m − 1).
I Brownian motion is almost surely not differentiable.
I Brownian motion is almost surely not Lipschitz.
I Kolmogorov-Centsov theorem applies to higher dimensions(with adjusted exponents). One can construct a.s. continuousfunctions from Rn to R.
18.175 Lecture 37
![Page 30: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/30.jpg)
Outline
Brownian motion properties and construction
Markov property, Blumenthal’s 0-1 law
18.175 Lecture 37
![Page 31: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/31.jpg)
Outline
Brownian motion properties and construction
Markov property, Blumenthal’s 0-1 law
18.175 Lecture 37
![Page 32: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/32.jpg)
More σ-algebra thoughts
I Write Fos = σ(Br : r ≤ s).
I Write F+s = ∩t>sFo
t
I Note right continuity: ∩t>sF+t = F+
s .
I F+s allows an “infinitesimal peek at future”
18.175 Lecture 37
![Page 33: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/33.jpg)
More σ-algebra thoughts
I Write Fos = σ(Br : r ≤ s).
I Write F+s = ∩t>sFo
t
I Note right continuity: ∩t>sF+t = F+
s .
I F+s allows an “infinitesimal peek at future”
18.175 Lecture 37
![Page 34: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/34.jpg)
More σ-algebra thoughts
I Write Fos = σ(Br : r ≤ s).
I Write F+s = ∩t>sFo
t
I Note right continuity: ∩t>sF+t = F+
s .
I F+s allows an “infinitesimal peek at future”
18.175 Lecture 37
![Page 35: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/35.jpg)
More σ-algebra thoughts
I Write Fos = σ(Br : r ≤ s).
I Write F+s = ∩t>sFo
t
I Note right continuity: ∩t>sF+t = F+
s .
I F+s allows an “infinitesimal peek at future”
18.175 Lecture 37
![Page 36: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/36.jpg)
Markov property
I If s ≥ 0 and Y is bounded and C-measurable, then for allx ∈ Rd , we have
Ex(Y θs |F+s ) = EBsY ,
where the RHS is function φ(x) = ExY evaluated at x = Bs .
I Proof idea: First establish this for some simple functions Y(depending on finitely many time values) and then usemeasure theory (monotone class theorem) to extend togeneral case.
18.175 Lecture 37
![Page 37: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/37.jpg)
Markov property
I If s ≥ 0 and Y is bounded and C-measurable, then for allx ∈ Rd , we have
Ex(Y θs |F+s ) = EBsY ,
where the RHS is function φ(x) = ExY evaluated at x = Bs .
I Proof idea: First establish this for some simple functions Y(depending on finitely many time values) and then usemeasure theory (monotone class theorem) to extend togeneral case.
18.175 Lecture 37
![Page 38: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/38.jpg)
Looking ahead
I Theorem: If Z is bounded, measurable then for s ≥ 0 have
Ex(A|F+s ) = Ex(Z |F0
s ).
18.175 Lecture 37
![Page 39: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/39.jpg)
Looking ahead
II Theorem: If Z is bounded, measurable then for s ≥ 0 have
Ex(A|F+s ) = Ex(Z |F0
s ).
18.175 Lecture 37
![Page 40: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/40.jpg)
Blumenthal’s 0-1 law
II If A ∈ F+0 , then P(A) ∈ 0, 1 (if P is probability law for
Brownian motion started at fixed value x at time 0).
I There’s nothing you can learn from infinitesimal neighborhoodof future.
18.175 Lecture 37
![Page 41: 18.175: Lecture 37 .1in More Brownian motionmath.mit.edu/~sheffield/175/Lecture37.pdf · 2014-02-11 · 18.175: Lecture 37 More Brownian motion Scott She eld MIT 18.175 Lecture 37](https://reader036.vdocuments.us/reader036/viewer/2022070917/5fb78739eae9ad6f9012283b/html5/thumbnails/41.jpg)
Blumenthal’s 0-1 law
I If A ∈ F+0 , then P(A) ∈ 0, 1 (if P is probability law for
Brownian motion started at fixed value x at time 0).
I There’s nothing you can learn from infinitesimal neighborhoodof future.
18.175 Lecture 37