lévy processes and stochastic partial differential...
TRANSCRIPT
![Page 1: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/1.jpg)
Levy Processes and Stochastic Partial DifferentialEquations
Davar Khoshnevisanwith M. Foondun and E. Nualart
Department of MathematicsUniversity of Utah
http://www.math.utah.edu/˜davar
Levy Processes: Theory and ApplicationsAugust 13–17, 2007
Copenhagen, Denmark
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 1 / 21
![Page 2: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/2.jpg)
Problem 1
I The stochastic heat equation:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x) ∀t ∈ 0, x ∈ Rd .
I Question: Why ∃ function solutions only when d = 1?(Walsh, Dalang–Frangos, Dalang, Pesat–Zabczyk)
I Answer: BM has local times only in d = 1.In fact, ∃ function solutions in dimension 2− α for all α ∈ (0 ,2].
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 2 / 21
![Page 3: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/3.jpg)
Problem 1
I The stochastic heat equation:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x) ∀t ∈ 0, x ∈ Rd .
I Question: Why ∃ function solutions only when d = 1?(Walsh, Dalang–Frangos, Dalang, Pesat–Zabczyk)
I Answer: BM has local times only in d = 1.In fact, ∃ function solutions in dimension 2− α for all α ∈ (0 ,2].
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 2 / 21
![Page 4: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/4.jpg)
Problem 1
I The stochastic heat equation:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x) ∀t ∈ 0, x ∈ Rd .
I Question: Why ∃ function solutions only when d = 1?(Walsh, Dalang–Frangos, Dalang, Pesat–Zabczyk)
I Answer: BM has local times only in d = 1.In fact, ∃ function solutions in dimension 2− α for all α ∈ (0 ,2].
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 2 / 21
![Page 5: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/5.jpg)
Problem 1
I The stochastic heat equation:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x) ∀t ∈ 0, x ∈ Rd .
I Question: Why ∃ function solutions only when d = 1?(Walsh, Dalang–Frangos, Dalang, Pesat–Zabczyk)
I Answer: BM has local times only in d = 1.
In fact, ∃ function solutions in dimension 2− α for all α ∈ (0 ,2].
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 2 / 21
![Page 6: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/6.jpg)
Problem 1
I The stochastic heat equation:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x) ∀t ∈ 0, x ∈ Rd .
I Question: Why ∃ function solutions only when d = 1?(Walsh, Dalang–Frangos, Dalang, Pesat–Zabczyk)
I Answer: BM has local times only in d = 1.
In fact, ∃ function solutions in dimension 2− α for all α ∈ (0 ,2].
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 2 / 21
![Page 7: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/7.jpg)
Problem 1
I The stochastic heat equation:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x) ∀t ∈ 0, x ∈ Rd .
I Question: Why ∃ function solutions only when d = 1?(Walsh, Dalang–Frangos, Dalang, Pesat–Zabczyk)
I Answer: BM has local times only in d = 1.In fact, ∃ function solutions in dimension 2− α for all α ∈ (0 ,2].
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 2 / 21
![Page 8: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/8.jpg)
Problem 2
I Weakly interacting system of stochastic wave equations:∂ttui(t ,x) = (∂xxui)(t ,x) +d
∑j=1
QijW j(t ,x) ∀x ∈ R, t ≥ 0,
ui(0 ,x) = ∂tui(0 ,x) = 0,
W 1, . . . , W d := i.i.d. white noises; Q = (Qij)di ,j=1 invert.
I Question: When is u(t ,x) = 0 for some t > 0 and x ∈ R?I Answer: Iff d < 4.
(Orey–Pruitt, K, Dalang–Nualart; closely-related: LeGall)
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 3 / 21
![Page 9: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/9.jpg)
Problem 2
I Weakly interacting system of stochastic wave equations:∂ttui(t ,x) = (∂xxui)(t ,x) +d
∑j=1
QijW j(t ,x) ∀x ∈ R, t ≥ 0,
ui(0 ,x) = ∂tui(0 ,x) = 0,
W 1, . . . , W d := i.i.d. white noises; Q = (Qij)di ,j=1 invert.
I Question: When is u(t ,x) = 0 for some t > 0 and x ∈ R?
I Answer: Iff d < 4.(Orey–Pruitt, K, Dalang–Nualart; closely-related: LeGall)
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 3 / 21
![Page 10: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/10.jpg)
Problem 2
I Weakly interacting system of stochastic wave equations:∂ttui(t ,x) = (∂xxui)(t ,x) +d
∑j=1
QijW j(t ,x) ∀x ∈ R, t ≥ 0,
ui(0 ,x) = ∂tui(0 ,x) = 0,
W 1, . . . , W d := i.i.d. white noises; Q = (Qij)di ,j=1 invert.
I Question: When is u(t ,x) = 0 for some t > 0 and x ∈ R?I Answer: Iff d < 4.
(Orey–Pruitt, K, Dalang–Nualart; closely-related: LeGall)
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 3 / 21
![Page 11: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/11.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:
I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .I Identification via Wiener integrals:Z
φdW 'Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 12: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/12.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.
I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .I Identification via Wiener integrals:Z
φdW 'Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 13: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/13.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .
I Identification via Wiener integrals:ZφdW '
Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 14: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/14.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .I Identification via Wiener integrals:Z
φdW 'Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 15: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/15.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .I Identification via Wiener integrals:Z
φdW 'Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 16: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/16.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .I Identification via Wiener integrals:Z
φdW 'Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.
I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 17: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/17.jpg)
The standard equation
I W := {W (t ,x)}t≥0,x∈Rd space-time white noise:I {W (A)}A∈B(R+×Rd ) := a centered gaussian process.I Cov(W (A) , W (B)) = |A∩B| ∀A,B ⊂ R+ ×Rd .I Identification via Wiener integrals:Z
φdW 'Z ∞
0
ZRdφ(t ,x)W (t ,x)dx dt .
I The stochastic heat equation: ∃(?)u := u(t ,x)[t ≥ 0, x ∈ Rd ]:
∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Fact 1: Function-valued solution ∃ iff d = 1.I Rough explanation: ∆x smooths; W makes rough.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 4 / 21
![Page 18: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/18.jpg)
The heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Consider ∂tu(t ,x) = (∆xu)(t ,x) + f (t ,x)[f nice]
I Let
Qt(y) :=1
(4πt)d/2exp
(−‖y‖2
4t
).
I Solution:
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)f (s ,y)dy ds.
I Apply to “f := W .” [Mild solution; Walsh, 1986]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 5 / 21
![Page 19: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/19.jpg)
The heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Consider ∂tu(t ,x) = (∆xu)(t ,x) + f (t ,x)[f nice]
I Let
Qt(y) :=1
(4πt)d/2exp
(−‖y‖2
4t
).
I Solution:
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)f (s ,y)dy ds.
I Apply to “f := W .” [Mild solution; Walsh, 1986]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 5 / 21
![Page 20: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/20.jpg)
The heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Consider ∂tu(t ,x) = (∆xu)(t ,x) + f (t ,x)[f nice]
I Let
Qt(y) :=1
(4πt)d/2exp
(−‖y‖2
4t
).
I Solution:
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)f (s ,y)dy ds.
I Apply to “f := W .” [Mild solution; Walsh, 1986]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 5 / 21
![Page 21: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/21.jpg)
The heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Consider ∂tu(t ,x) = (∆xu)(t ,x) + f (t ,x)[f nice]
I Let
Qt(y) :=1
(4πt)d/2exp
(−‖y‖2
4t
).
I Solution:
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)f (s ,y)dy ds.
I Apply to “f := W .” [Mild solution; Walsh, 1986]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 5 / 21
![Page 22: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/22.jpg)
The heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Consider ∂tu(t ,x) = (∆xu)(t ,x) + f (t ,x)[f nice]
I Let
Qt(y) :=1
(4πt)d/2exp
(−‖y‖2
4t
).
I Solution:
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)f (s ,y)dy ds.
I Apply to “f := W .” [Mild solution; Walsh, 1986]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 5 / 21
![Page 23: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/23.jpg)
The stochastic heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).
I Mild solution: If ∃, then
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)W (dy ds).
I Weak solution: A family u(t , ϕ), for nicely tempered ϕ:
u(t , ϕ) :=Z t
0
ZRd
(Qt−s ∗ ϕ)(y)W (dy ds).
“u(t , ϕ) =R
Rd u(t ,x)ϕ(x)dx .”I ϕ 7→ u(t , ϕ) is a linear gaussian distribution [Ito; Menshov].I Need:
E(∣∣∣u(t , ϕ)
∣∣∣2) =Z t
0
ZRd
∣∣∣(Qt−s ∗ϕ)(y)∣∣∣2 dy ds <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 6 / 21
![Page 24: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/24.jpg)
The stochastic heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).I Mild solution: If ∃, then
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)W (dy ds).
I Weak solution: A family u(t , ϕ), for nicely tempered ϕ:
u(t , ϕ) :=Z t
0
ZRd
(Qt−s ∗ ϕ)(y)W (dy ds).
“u(t , ϕ) =R
Rd u(t ,x)ϕ(x)dx .”I ϕ 7→ u(t , ϕ) is a linear gaussian distribution [Ito; Menshov].I Need:
E(∣∣∣u(t , ϕ)
∣∣∣2) =Z t
0
ZRd
∣∣∣(Qt−s ∗ϕ)(y)∣∣∣2 dy ds <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 6 / 21
![Page 25: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/25.jpg)
The stochastic heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).I Mild solution: If ∃, then
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)W (dy ds).
I Weak solution: A family u(t , ϕ), for nicely tempered ϕ:
u(t , ϕ) :=Z t
0
ZRd
(Qt−s ∗ ϕ)(y)W (dy ds).
“u(t , ϕ) =R
Rd u(t ,x)ϕ(x)dx .”
I ϕ 7→ u(t , ϕ) is a linear gaussian distribution [Ito; Menshov].I Need:
E(∣∣∣u(t , ϕ)
∣∣∣2) =Z t
0
ZRd
∣∣∣(Qt−s ∗ϕ)(y)∣∣∣2 dy ds <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 6 / 21
![Page 26: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/26.jpg)
The stochastic heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).I Mild solution: If ∃, then
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)W (dy ds).
I Weak solution: A family u(t , ϕ), for nicely tempered ϕ:
u(t , ϕ) :=Z t
0
ZRd
(Qt−s ∗ ϕ)(y)W (dy ds).
“u(t , ϕ) =R
Rd u(t ,x)ϕ(x)dx .”I ϕ 7→ u(t , ϕ) is a linear gaussian distribution [Ito; Menshov].
I Need:
E(∣∣∣u(t , ϕ)
∣∣∣2) =Z t
0
ZRd
∣∣∣(Qt−s ∗ϕ)(y)∣∣∣2 dy ds <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 6 / 21
![Page 27: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/27.jpg)
The stochastic heat equation
I ∂tu(t ,x) = (∆xu)(t ,x) + W (t ,x).I Mild solution: If ∃, then
u(t ,x) =Z t
0
ZRd
Qt−s(x − y)W (dy ds).
I Weak solution: A family u(t , ϕ), for nicely tempered ϕ:
u(t , ϕ) :=Z t
0
ZRd
(Qt−s ∗ ϕ)(y)W (dy ds).
“u(t , ϕ) =R
Rd u(t ,x)ϕ(x)dx .”I ϕ 7→ u(t , ϕ) is a linear gaussian distribution [Ito; Menshov].I Need:
E(∣∣∣u(t , ϕ)
∣∣∣2) =Z t
0
ZRd
∣∣∣(Qt−s ∗ϕ)(y)∣∣∣2 dy ds <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 6 / 21
![Page 28: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/28.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 29: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/29.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 30: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/30.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 31: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/31.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 32: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/32.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd
⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 33: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/33.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd
⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 34: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/34.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd
⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 35: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/35.jpg)
The stochastic heat equation
I Plancherel’s theorem:Z t
0
ZRd
∣∣∣(Qt−s ∗ ϕ)(y)∣∣∣2 dy ds
=1
(2π)d
Z t
0
ZRd
e−2‖ξ‖2(t−s) |ϕ(ξ)|2 dξ ds
=1
(2π)d
ZRd|ϕ(ξ)|2 1− e−2t‖ξ‖2
2‖ξ‖2 dξ.
I ∴ weak solution ∃ iff ϕ ∈ H−2(Rd). In fact,
E(∣∣∣u(t , ϕ)
∣∣∣2) � tZ
Rd
|ϕ(ξ)|2
1 + t2‖ξ‖2 dξ.
I ∴ mild solution ∃ iff δx ∈ H−2(Rd) ∀x ∈ Rd⇔ d = 1. �
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 7 / 21
![Page 36: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/36.jpg)
An explanation
I (Dalang–Frangos) Replace W (t ,x) by F (t ,x), where F is acentered gaussian noise with Cov(
RφdF ,
Rψ dF ) =ZZ
φ(t ,x)(s ∧ t)κ(‖x − y‖)︸ ︷︷ ︸Σ(s ,t ,‖x−y‖)
ψ(s ,y)dt dx ds dy .
I (Dalang–Frangos, Pesat–Zabczyk, Dalang) There can ∃ solutionsfor d > 1, depending on κ [NASC].
I Explains the roughening effect of white noise.[analytic]
I We propose to explain the smoothing effect of ∆x .[probabilistic]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 8 / 21
![Page 37: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/37.jpg)
An explanation
I (Dalang–Frangos) Replace W (t ,x) by F (t ,x), where F is acentered gaussian noise with Cov(
RφdF ,
Rψ dF ) =ZZ
φ(t ,x)(s ∧ t)κ(‖x − y‖)︸ ︷︷ ︸Σ(s ,t ,‖x−y‖)
ψ(s ,y)dt dx ds dy .
I (Dalang–Frangos, Pesat–Zabczyk, Dalang) There can ∃ solutionsfor d > 1, depending on κ [NASC].
I Explains the roughening effect of white noise.[analytic]
I We propose to explain the smoothing effect of ∆x .[probabilistic]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 8 / 21
![Page 38: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/38.jpg)
An explanation
I (Dalang–Frangos) Replace W (t ,x) by F (t ,x), where F is acentered gaussian noise with Cov(
RφdF ,
Rψ dF ) =ZZ
φ(t ,x)(s ∧ t)κ(‖x − y‖)︸ ︷︷ ︸Σ(s ,t ,‖x−y‖)
ψ(s ,y)dt dx ds dy .
I (Dalang–Frangos, Pesat–Zabczyk, Dalang) There can ∃ solutionsfor d > 1, depending on κ [NASC].
I Explains the roughening effect of white noise.[analytic]
I We propose to explain the smoothing effect of ∆x .[probabilistic]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 8 / 21
![Page 39: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/39.jpg)
An explanation
I (Dalang–Frangos) Replace W (t ,x) by F (t ,x), where F is acentered gaussian noise with Cov(
RφdF ,
Rψ dF ) =ZZ
φ(t ,x)(s ∧ t)κ(‖x − y‖)︸ ︷︷ ︸Σ(s ,t ,‖x−y‖)
ψ(s ,y)dt dx ds dy .
I (Dalang–Frangos, Pesat–Zabczyk, Dalang) There can ∃ solutionsfor d > 1, depending on κ [NASC].
I Explains the roughening effect of white noise.[analytic]
I We propose to explain the smoothing effect of ∆x .[probabilistic]
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 8 / 21
![Page 40: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/40.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .
I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).That is, Z
Rdf (x)(Lg)(x)dx = −
ZRd
f (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.I X ′ := an independent copy of X ;
X (t) := X (t)−X ′(t).[Levy]
I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 41: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/41.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).
That is, ZRd
f (x)(Lg)(x)dx = −Z
Rdf (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.I X ′ := an independent copy of X ;
X (t) := X (t)−X ′(t).[Levy]
I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 42: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/42.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).
That is, ZRd
f (x)(Lg)(x)dx = −Z
Rdf (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.
I X ′ := an independent copy of X ;
X (t) := X (t)−X ′(t).[Levy]
I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 43: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/43.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).
That is, ZRd
f (x)(Lg)(x)dx = −Z
Rdf (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.I X ′ := an independent copy of X ;
X (t) := X (t)−X ′(t).[Levy]
I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 44: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/44.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).
That is, ZRd
f (x)(Lg)(x)dx = −Z
Rdf (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.I X ′ := an independent copy of X ;
X (t) := X (t)−X ′(t).[Levy]
I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 45: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/45.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).
That is, ZRd
f (x)(Lg)(x)dx = −Z
Rdf (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.I X ′ := an independent copy of X ; X (t) := X (t)−X ′(t).
[Levy]
I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 46: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/46.jpg)
Levy processes
I L := L2-generator of a Levy process X in Rd .I Normalization: E exp(iξ ·X (t)) = exp(−tΨ(ξ)), L(ξ) = −Ψ(ξ).
That is, ZRd
f (x)(Lg)(x)dx = −Z
Rdf (ξ) g(ξ)Ψ(ξ)dξ.
I Dom(L) := {f ∈ L2(Rd) :R
Rd |f (ξ)|2ReΨ(ξ)dξ <∞}.I X ′ := an independent copy of X ; X (t) := X (t)−X ′(t).
[Levy]I X is a Levy process with char. exponent 2ReΨ.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 9 / 21
![Page 47: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/47.jpg)
Local times
I λxt := local time of X , at place x and time t , when it exists.
I Defining property: ∀ Borel meas. f : Rd → R+, t > 0,Z t
0f (X (s))ds =
ZRd
f (x)λxt dx a.s.
I λxt = Ot(dx)/dx , where Ot(E) :=
R t0 1E(X (s))ds.
Theorem (Hawkes){λx
t }t≥0,x∈Rd exists iff ZRd
dξ1 + ReΨ(ξ)
<∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 10 / 21
![Page 48: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/48.jpg)
Local times
I λxt := local time of X , at place x and time t , when it exists.
I Defining property: ∀ Borel meas. f : Rd → R+, t > 0,Z t
0f (X (s))ds =
ZRd
f (x)λxt dx a.s.
I λxt = Ot(dx)/dx , where Ot(E) :=
R t0 1E(X (s))ds.
Theorem (Hawkes){λx
t }t≥0,x∈Rd exists iff ZRd
dξ1 + ReΨ(ξ)
<∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 10 / 21
![Page 49: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/49.jpg)
Local times
I λxt := local time of X , at place x and time t , when it exists.
I Defining property: ∀ Borel meas. f : Rd → R+, t > 0,Z t
0f (X (s))ds =
ZRd
f (x)λxt dx a.s.
I λxt = Ot(dx)/dx , where Ot(E) :=
R t0 1E(X (s))ds.
Theorem (Hawkes){λx
t }t≥0,x∈Rd exists iff ZRd
dξ1 + ReΨ(ξ)
<∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 10 / 21
![Page 50: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/50.jpg)
Local times
I λxt := local time of X , at place x and time t , when it exists.
I Defining property: ∀ Borel meas. f : Rd → R+, t > 0,Z t
0f (X (s))ds =
ZRd
f (x)λxt dx a.s.
I λxt = Ot(dx)/dx , where Ot(E) :=
R t0 1E(X (s))ds.
Theorem (Hawkes){λx
t }t≥0,x∈Rd exists iff ZRd
dξ1 + ReΨ(ξ)
<∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 10 / 21
![Page 51: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/51.jpg)
Examples and remarks
I Hawkes’ condition:R
Rd (1 + ReΨ(ξ))−1 dξ <∞.
I Ψ(ξ) = O(‖ξ‖2) [Bochner]. Therefore, local times can ∃ only whend = 1, if at all.
I If d = 1 and Ψ(ξ) = |ξ|α then local times exist iff α > 1.I For general (nonsymmetric) Levy processes, the Hawkes
condition is that Z ∞
−∞Re
(1
1 + Ψ(ξ)
)dξ <∞.
I Suppose d = 1 and Ψ(ξ) = |ξ|(1 + icsgn|ξ| log |ξ|) for0 ≤ |c| ≤ 2/π. Then local times exist iff c 6= 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 11 / 21
![Page 52: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/52.jpg)
Examples and remarks
I Hawkes’ condition:R
Rd (1 + ReΨ(ξ))−1 dξ <∞.I Ψ(ξ) = O(‖ξ‖2) [Bochner]. Therefore, local times can ∃ only when
d = 1, if at all.
I If d = 1 and Ψ(ξ) = |ξ|α then local times exist iff α > 1.I For general (nonsymmetric) Levy processes, the Hawkes
condition is that Z ∞
−∞Re
(1
1 + Ψ(ξ)
)dξ <∞.
I Suppose d = 1 and Ψ(ξ) = |ξ|(1 + icsgn|ξ| log |ξ|) for0 ≤ |c| ≤ 2/π. Then local times exist iff c 6= 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 11 / 21
![Page 53: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/53.jpg)
Examples and remarks
I Hawkes’ condition:R
Rd (1 + ReΨ(ξ))−1 dξ <∞.I Ψ(ξ) = O(‖ξ‖2) [Bochner]. Therefore, local times can ∃ only when
d = 1, if at all.I If d = 1 and Ψ(ξ) = |ξ|α then local times exist iff α > 1.
I For general (nonsymmetric) Levy processes, the Hawkescondition is that Z ∞
−∞Re
(1
1 + Ψ(ξ)
)dξ <∞.
I Suppose d = 1 and Ψ(ξ) = |ξ|(1 + icsgn|ξ| log |ξ|) for0 ≤ |c| ≤ 2/π. Then local times exist iff c 6= 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 11 / 21
![Page 54: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/54.jpg)
Examples and remarks
I Hawkes’ condition:R
Rd (1 + ReΨ(ξ))−1 dξ <∞.I Ψ(ξ) = O(‖ξ‖2) [Bochner]. Therefore, local times can ∃ only when
d = 1, if at all.I If d = 1 and Ψ(ξ) = |ξ|α then local times exist iff α > 1.I For general (nonsymmetric) Levy processes, the Hawkes
condition is that Z ∞
−∞Re
(1
1 + Ψ(ξ)
)dξ <∞.
I Suppose d = 1 and Ψ(ξ) = |ξ|(1 + icsgn|ξ| log |ξ|) for0 ≤ |c| ≤ 2/π. Then local times exist iff c 6= 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 11 / 21
![Page 55: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/55.jpg)
Examples and remarks
I Hawkes’ condition:R
Rd (1 + ReΨ(ξ))−1 dξ <∞.I Ψ(ξ) = O(‖ξ‖2) [Bochner]. Therefore, local times can ∃ only when
d = 1, if at all.I If d = 1 and Ψ(ξ) = |ξ|α then local times exist iff α > 1.I For general (nonsymmetric) Levy processes, the Hawkes
condition is that Z ∞
−∞Re
(1
1 + Ψ(ξ)
)dξ <∞.
I Suppose d = 1 and Ψ(ξ) = |ξ|(1 + icsgn|ξ| log |ξ|) for0 ≤ |c| ≤ 2/π. Then local times exist iff c 6= 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 11 / 21
![Page 56: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/56.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?I When is there a Holder-continuous solution?I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 57: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/57.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.
I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?I When is there a Holder-continuous solution?I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 58: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/58.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?I When is there a Holder-continuous solution?I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 59: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/59.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?I When is there a Holder-continuous solution?I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 60: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/60.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?
I When is there a Holder-continuous solution?I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 61: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/61.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?I When is there a Holder-continuous solution?
I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 62: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/62.jpg)
A heat equation
I Consider the heat equation
∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x),
where Lx is the generator of a Levy process on Rd , acting on thevariable x .
I We call this the heat equation for L.I Fundamental questions:
I When is there a function-valued solution?[General answer by Dalang]
I When is there a continuous solution?I When is there a Holder-continuous solution?I Etc.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 12 / 21
![Page 63: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/63.jpg)
A heat equation
Theorem (K–Foondun–Nualart)Let u denote the weak solution to the heat equation for L. Then, for alltempered functions ϕ and all t , λ > 0,
1− e−2λt
2Eλ(ϕ ,ϕ) ≤ E
(|u(t , ϕ)|2
)≤ e2λt
2Eλ(ϕ ,ϕ),
where
Eλ(ϕ ,ψ) :=1
(2π)d
ZRd
ϕ(ξ)ψ(ξ)
λ+ ReΨ(ξ)dξ.
Corollary∃ function-valued solutions iff X has local times. The solution iscontinuous iff x 7→ λx
t is.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 13 / 21
![Page 64: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/64.jpg)
A heat equation
Theorem (K–Foondun–Nualart)Let u denote the weak solution to the heat equation for L. Then, for alltempered functions ϕ and all t , λ > 0,
1− e−2λt
2Eλ(ϕ ,ϕ) ≤ E
(|u(t , ϕ)|2
)≤ e2λt
2Eλ(ϕ ,ϕ),
where
Eλ(ϕ ,ψ) :=1
(2π)d
ZRd
ϕ(ξ)ψ(ξ)
λ+ ReΨ(ξ)dξ.
Corollary∃ function-valued solutions iff X has local times. The solution iscontinuous iff x 7→ λx
t is.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 13 / 21
![Page 65: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/65.jpg)
A heat equation
I In fact, many [virtually all] of the properties of x 7→ u(t ,x) areinherited from x 7→ λx
t and vice versa:
I Existence and continuity.I Holder continuity.I p-variation of the paths . . . .
I ∃ an embedding of the isomorphism theorem? [Dynkin;Brydges–Frohlich–Spencer]
I A final Theorem (K–Foondun–Nualart): t 7→ u(t , ϕ) has acontinuous version iffZ ∞
1
Eλ(ϕ ,ϕ)
λ√| logλ|
dλ <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 14 / 21
![Page 66: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/66.jpg)
A heat equation
I In fact, many [virtually all] of the properties of x 7→ u(t ,x) areinherited from x 7→ λx
t and vice versa:I Existence and continuity.
I Holder continuity.I p-variation of the paths . . . .
I ∃ an embedding of the isomorphism theorem? [Dynkin;Brydges–Frohlich–Spencer]
I A final Theorem (K–Foondun–Nualart): t 7→ u(t , ϕ) has acontinuous version iffZ ∞
1
Eλ(ϕ ,ϕ)
λ√| logλ|
dλ <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 14 / 21
![Page 67: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/67.jpg)
A heat equation
I In fact, many [virtually all] of the properties of x 7→ u(t ,x) areinherited from x 7→ λx
t and vice versa:I Existence and continuity.I Holder continuity.
I p-variation of the paths . . . .
I ∃ an embedding of the isomorphism theorem? [Dynkin;Brydges–Frohlich–Spencer]
I A final Theorem (K–Foondun–Nualart): t 7→ u(t , ϕ) has acontinuous version iffZ ∞
1
Eλ(ϕ ,ϕ)
λ√| logλ|
dλ <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 14 / 21
![Page 68: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/68.jpg)
A heat equation
I In fact, many [virtually all] of the properties of x 7→ u(t ,x) areinherited from x 7→ λx
t and vice versa:I Existence and continuity.I Holder continuity.I p-variation of the paths . . . .
I ∃ an embedding of the isomorphism theorem? [Dynkin;Brydges–Frohlich–Spencer]
I A final Theorem (K–Foondun–Nualart): t 7→ u(t , ϕ) has acontinuous version iffZ ∞
1
Eλ(ϕ ,ϕ)
λ√| logλ|
dλ <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 14 / 21
![Page 69: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/69.jpg)
A heat equation
I In fact, many [virtually all] of the properties of x 7→ u(t ,x) areinherited from x 7→ λx
t and vice versa:I Existence and continuity.I Holder continuity.I p-variation of the paths . . . .
I ∃ an embedding of the isomorphism theorem? [Dynkin;Brydges–Frohlich–Spencer]
I A final Theorem (K–Foondun–Nualart): t 7→ u(t , ϕ) has acontinuous version iffZ ∞
1
Eλ(ϕ ,ϕ)
λ√| logλ|
dλ <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 14 / 21
![Page 70: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/70.jpg)
A heat equation
I In fact, many [virtually all] of the properties of x 7→ u(t ,x) areinherited from x 7→ λx
t and vice versa:I Existence and continuity.I Holder continuity.I p-variation of the paths . . . .
I ∃ an embedding of the isomorphism theorem? [Dynkin;Brydges–Frohlich–Spencer]
I A final Theorem (K–Foondun–Nualart): t 7→ u(t , ϕ) has acontinuous version iffZ ∞
1
Eλ(ϕ ,ϕ)
λ√| logλ|
dλ <∞.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 14 / 21
![Page 71: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/71.jpg)
Solutions in dimension 2− ε
I Recall:∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x). (HE)
I We chose L to be the generator of a Levy process only becausethere we have NASC.
I Could have L := generator of a nice Markov process.
Theorem (K–Foondun–Nualart)Let L := Laplacian on a “nice” fractal of dimH = 2− α for α ∈ (0 ,2].Then (HE) has function solutions that are in fact Holder continuous.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 15 / 21
![Page 72: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/72.jpg)
Solutions in dimension 2− ε
I Recall:∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x). (HE)
I We chose L to be the generator of a Levy process only becausethere we have NASC.
I Could have L := generator of a nice Markov process.
Theorem (K–Foondun–Nualart)Let L := Laplacian on a “nice” fractal of dimH = 2− α for α ∈ (0 ,2].Then (HE) has function solutions that are in fact Holder continuous.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 15 / 21
![Page 73: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/73.jpg)
Solutions in dimension 2− ε
I Recall:∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x). (HE)
I We chose L to be the generator of a Levy process only becausethere we have NASC.
I Could have L := generator of a nice Markov process.
Theorem (K–Foondun–Nualart)Let L := Laplacian on a “nice” fractal of dimH = 2− α for α ∈ (0 ,2].Then (HE) has function solutions that are in fact Holder continuous.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 15 / 21
![Page 74: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/74.jpg)
Solutions in dimension 2− ε
I Recall:∂tu(t ,x) = (Lxu)(t ,x) + W (t ,x). (HE)
I We chose L to be the generator of a Levy process only becausethere we have NASC.
I Could have L := generator of a nice Markov process.
Theorem (K–Foondun–Nualart)Let L := Laplacian on a “nice” fractal of dimH = 2− α for α ∈ (0 ,2].Then (HE) has function solutions that are in fact Holder continuous.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 15 / 21
![Page 75: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/75.jpg)
System of wave equations
I Let L := be space–time Levy noise with values in Rd .
I E exp(iξ · L(A)) = exp(−|A|Ψ(ξ)) for ξ ∈ Rd and A ∈B(R+ ×R).I Assume:
I Ψ is nonnegative real.I ∀a > 0∃Aa > 0 such that
Ψ(aξ) ≥ AaΨ(ξ).
[stable like]
I Gauge function ∃ and is finite, where
Φ(λ) :=Z
Rde−λΨ(ξ) dξ ∀λ > 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 16 / 21
![Page 76: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/76.jpg)
System of wave equations
I Let L := be space–time Levy noise with values in Rd .I E exp(iξ · L(A)) = exp(−|A|Ψ(ξ)) for ξ ∈ Rd and A ∈B(R+ ×R).
I Assume:
I Ψ is nonnegative real.I ∀a > 0∃Aa > 0 such that
Ψ(aξ) ≥ AaΨ(ξ).
[stable like]
I Gauge function ∃ and is finite, where
Φ(λ) :=Z
Rde−λΨ(ξ) dξ ∀λ > 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 16 / 21
![Page 77: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/77.jpg)
System of wave equations
I Let L := be space–time Levy noise with values in Rd .I E exp(iξ · L(A)) = exp(−|A|Ψ(ξ)) for ξ ∈ Rd and A ∈B(R+ ×R).I Assume:
I Ψ is nonnegative real.I ∀a > 0∃Aa > 0 such that
Ψ(aξ) ≥ AaΨ(ξ).
[stable like]
I Gauge function ∃ and is finite, where
Φ(λ) :=Z
Rde−λΨ(ξ) dξ ∀λ > 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 16 / 21
![Page 78: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/78.jpg)
System of wave equations
I Let L := be space–time Levy noise with values in Rd .I E exp(iξ · L(A)) = exp(−|A|Ψ(ξ)) for ξ ∈ Rd and A ∈B(R+ ×R).I Assume:
I Ψ is nonnegative real.
I ∀a > 0∃Aa > 0 such that
Ψ(aξ) ≥ AaΨ(ξ).
[stable like]
I Gauge function ∃ and is finite, where
Φ(λ) :=Z
Rde−λΨ(ξ) dξ ∀λ > 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 16 / 21
![Page 79: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/79.jpg)
System of wave equations
I Let L := be space–time Levy noise with values in Rd .I E exp(iξ · L(A)) = exp(−|A|Ψ(ξ)) for ξ ∈ Rd and A ∈B(R+ ×R).I Assume:
I Ψ is nonnegative real.I ∀a > 0∃Aa > 0 such that
Ψ(aξ) ≥ AaΨ(ξ).
[stable like]
I Gauge function ∃ and is finite, where
Φ(λ) :=Z
Rde−λΨ(ξ) dξ ∀λ > 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 16 / 21
![Page 80: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/80.jpg)
System of wave equations
I Let L := be space–time Levy noise with values in Rd .I E exp(iξ · L(A)) = exp(−|A|Ψ(ξ)) for ξ ∈ Rd and A ∈B(R+ ×R).I Assume:
I Ψ is nonnegative real.I ∀a > 0∃Aa > 0 such that
Ψ(aξ) ≥ AaΨ(ξ).
[stable like]
I Gauge function ∃ and is finite, where
Φ(λ) :=Z
Rde−λΨ(ξ) dξ ∀λ > 0.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 16 / 21
![Page 81: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/81.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
Theorem (K–Nualart)TFAE:
1. With positive probab. u(t ,x) = 0 for some t > 0 and x ∈ R.
2. Almost surely, u(t ,x) = 0 for some t > 0 and x ∈ R.
3.R 1
0 λΦ(λ)dλ <∞.
4. IfR 1
0 λΦ(λ)dλ <∞, then a.s.,
dimH u−1{0} = 2− lim supλ↓0
logΦ(λ)
log(1/λ).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 17 / 21
![Page 82: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/82.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
Theorem (K–Nualart)TFAE:
1. With positive probab. u(t ,x) = 0 for some t > 0 and x ∈ R.
2. Almost surely, u(t ,x) = 0 for some t > 0 and x ∈ R.
3.R 1
0 λΦ(λ)dλ <∞.
4. IfR 1
0 λΦ(λ)dλ <∞, then a.s.,
dimH u−1{0} = 2− lim supλ↓0
logΦ(λ)
log(1/λ).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 17 / 21
![Page 83: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/83.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
Theorem (K–Nualart)TFAE:
1. With positive probab. u(t ,x) = 0 for some t > 0 and x ∈ R.
2. Almost surely, u(t ,x) = 0 for some t > 0 and x ∈ R.
3.R 1
0 λΦ(λ)dλ <∞.
4. IfR 1
0 λΦ(λ)dλ <∞, then a.s.,
dimH u−1{0} = 2− lim supλ↓0
logΦ(λ)
log(1/λ).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 17 / 21
![Page 84: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/84.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
Theorem (K–Nualart)TFAE:
1. With positive probab. u(t ,x) = 0 for some t > 0 and x ∈ R.
2. Almost surely, u(t ,x) = 0 for some t > 0 and x ∈ R.
3.R 1
0 λΦ(λ)dλ <∞.
4. IfR 1
0 λΦ(λ)dλ <∞, then a.s.,
dimH u−1{0} = 2− lim supλ↓0
logΦ(λ)
log(1/λ).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 17 / 21
![Page 85: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/85.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
Theorem (K–Nualart)TFAE:
1. With positive probab. u(t ,x) = 0 for some t > 0 and x ∈ R.
2. Almost surely, u(t ,x) = 0 for some t > 0 and x ∈ R.
3.R 1
0 λΦ(λ)dλ <∞.
4. IfR 1
0 λΦ(λ)dλ <∞, then a.s.,
dimH u−1{0} = 2− lim supλ↓0
logΦ(λ)
log(1/λ).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 17 / 21
![Page 86: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/86.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
Theorem (K–Nualart)TFAE:
1. With positive probab. u(t ,x) = 0 for some t > 0 and x ∈ R.
2. Almost surely, u(t ,x) = 0 for some t > 0 and x ∈ R.
3.R 1
0 λΦ(λ)dλ <∞.
4. IfR 1
0 λΦ(λ)dλ <∞, then a.s.,
dimH u−1{0} = 2− lim supλ↓0
logΦ(λ)
log(1/λ).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 17 / 21
![Page 87: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/87.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
ExampleSuppose L1, . . . , Ld are independent, Lj = stable(αj). Then:
1. u has zeros iff ∑dj=1(1/αj) < 2.
2. If ∑dj=1(1/αj) < 2, then dimH u−1{0} = 2−∑d
j=1(1/αj).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 18 / 21
![Page 88: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/88.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
ExampleSuppose L1, . . . , Ld are independent, Lj = stable(αj). Then:
1. u has zeros iff ∑dj=1(1/αj) < 2.
2. If ∑dj=1(1/αj) < 2, then dimH u−1{0} = 2−∑d
j=1(1/αj).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 18 / 21
![Page 89: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/89.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
ExampleSuppose L1, . . . , Ld are independent, Lj = stable(αj). Then:
1. u has zeros iff ∑dj=1(1/αj) < 2.
2. If ∑dj=1(1/αj) < 2, then dimH u−1{0} = 2−∑d
j=1(1/αj).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 18 / 21
![Page 90: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/90.jpg)
Zeros of the solution
[∂ttui(t ,x) = ∂xxui(x , t) + Li(t ,x),
ui(0 ,x) = ∂tui(0 ,x) = 0.
ExampleSuppose L1, . . . , Ld are independent, Lj = stable(αj). Then:
1. u has zeros iff ∑dj=1(1/αj) < 2.
2. If ∑dj=1(1/αj) < 2, then dimH u−1{0} = 2−∑d
j=1(1/αj).
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 18 / 21
![Page 91: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/91.jpg)
Idea of proof [existence part]
I WLOG consider u(t ,x) for t ,x ≥ 0.
I P{0 ∈ u(G)} � P{0 ∈ X (G)}, where
X (t ,x) := X1(t) + X2(x),
where X1,X2 are i.i.d. Levy processes, exponent Ψ.[additive Levy process]
I Appeal to K–Shieh–Xiao.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 19 / 21
![Page 92: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/92.jpg)
Idea of proof [existence part]
I WLOG consider u(t ,x) for t ,x ≥ 0.I P{0 ∈ u(G)} � P{0 ∈ X (G)}, where
X (t ,x) := X1(t) + X2(x),
where X1,X2 are i.i.d. Levy processes, exponent Ψ.[additive Levy process]
I Appeal to K–Shieh–Xiao.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 19 / 21
![Page 93: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/93.jpg)
Idea of proof [existence part]
I WLOG consider u(t ,x) for t ,x ≥ 0.I P{0 ∈ u(G)} � P{0 ∈ X (G)}, where
X (t ,x) := X1(t) + X2(x),
where X1,X2 are i.i.d. Levy processes, exponent Ψ.[additive Levy process]
I Appeal to K–Shieh–Xiao.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 19 / 21
![Page 94: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/94.jpg)
Idea of proof [second part]
u(t ,x) = 12 L(C (t ,x)), where C (t ,x) is the “light cone” emanating from
(t ,x).
x x+tx−ty
s
t
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 20 / 21
![Page 95: Lévy Processes and Stochastic Partial Differential Equationsdavar/ps-pdf-files/Denmark07/Research... · 2007-07-21 · Levy Processes and Stochastic Partial Differential´ Equations](https://reader033.vdocuments.mx/reader033/viewer/2022042802/5f418af5b17f6462287bd0d5/html5/thumbnails/95.jpg)
Idea of proof [zero-one law part]
The zero set in the black triangle depends on the noise through its“backward light cone,” shaded black/pink.
Therefore, P{u−1({0}) 6= ∅} is zero or one.
Davar Khoshnevisan (Salt Lake City, Utah) Levy processes and SPDEs Copenhagen 2007 21 / 21