A Stochastic Heat Equation

104
A Stochastic Heat Equation Recall that F : R ! R is Lipschitz continuous if Lip(F) := sup 1 <x6 =y<1 F(x) F(y) x y <1: July 28, 2016 1 / 17

Transcript of A Stochastic Heat Equation

Page 1: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.

Let b, σ : R→ R be nonrandom and Lipschitz continuous.We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:

I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;I u(t , 0) = u(t , 1) = 0 ∀t > 0.

Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 2: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.Let b, σ : R→ R be nonrandom and Lipschitz continuous.

We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:

I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;I u(t , 0) = u(t , 1) = 0 ∀t > 0.

Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 3: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.Let b, σ : R→ R be nonrandom and Lipschitz continuous.We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:

I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;I u(t , 0) = u(t , 1) = 0 ∀t > 0.Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 4: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.Let b, σ : R→ R be nonrandom and Lipschitz continuous.We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;

I u(t , 0) = u(t , 1) = 0 ∀t > 0.Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 5: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.Let b, σ : R→ R be nonrandom and Lipschitz continuous.We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;I u(t , 0) = u(t , 1) = 0 ∀t > 0.

Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 6: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.Let b, σ : R→ R be nonrandom and Lipschitz continuous.We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;I u(t , 0) = u(t , 1) = 0 ∀t > 0.Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].

First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 7: A Stochastic Heat Equation

A Stochastic Heat EquationRecall that F : R→ R is Lipschitz continuous ifLip(F ) := sup

−∞<x 6=y<∞

∣∣∣∣F (x)− F (y)x − y

∣∣∣∣ <∞.Let b, σ : R→ R be nonrandom and Lipschitz continuous.We wish to solve the stochastic PDE [SPDE],u(t ,x) = u′′(t ,x) + b(u(t ,x)) + σ (u(t ,x))W (t ,x) on R+ × [0 , 1],

subject to:I u(0) = u0 ∈ L2[0 , 1] where u0 is nonrandom;I u(t , 0) = u(t , 1) = 0 ∀t > 0.Describes heat flow in a random environment [W ] with localinteraction/feedback [σ (u)].First, we need to give rigorous meaning to our SPDE.

July 28, 2016 1 / 17

Page 8: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + b(u) + σ (u)W ; u(0) = u0 ∈ L2[0 , 1]; u(t , 0) = u(t , 1) = 0 ∀t > 0First pretend that W is a smooth function. Then ∃! solution ufrom general theory.

Apply Duhamel’s principle: u uniquely solves the integralequationu(t ,x) = (Ptu0)(x) + ∫[0,t]×[0,1] pt−s(x , y)b(u(s , y)) ds dy

+ ∫[0,t]×[0,1] pt−s(x , y)σ (u(s , y))W (s , y) ds dy.

Mild form of the solution.For us, the mild solution continues to make sense and is in facta weak solution.We care only about mild solutions in general.Let us study the simplest case first where b ≡ 0 and σ ≡ 1.

July 28, 2016 2 / 17

Page 9: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + b(u) + σ (u)W ; u(0) = u0 ∈ L2[0 , 1]; u(t , 0) = u(t , 1) = 0 ∀t > 0First pretend that W is a smooth function. Then ∃! solution ufrom general theory.Apply Duhamel’s principle: u uniquely solves the integralequation

u(t ,x) = (Ptu0)(x) + ∫[0,t]×[0,1] pt−s(x , y)b(u(s , y)) ds dy

+ ∫[0,t]×[0,1] pt−s(x , y)σ (u(s , y))W (s , y) ds dy.

Mild form of the solution.For us, the mild solution continues to make sense and is in facta weak solution.We care only about mild solutions in general.Let us study the simplest case first where b ≡ 0 and σ ≡ 1.

July 28, 2016 2 / 17

Page 10: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + b(u) + σ (u)W ; u(0) = u0 ∈ L2[0 , 1]; u(t , 0) = u(t , 1) = 0 ∀t > 0First pretend that W is a smooth function. Then ∃! solution ufrom general theory.Apply Duhamel’s principle: u uniquely solves the integralequation

u(t ,x) = (Ptu0)(x) + ∫[0,t]×[0,1] pt−s(x , y)b(u(s , y)) ds dy

+ ∫[0,t]×[0,1] pt−s(x , y)σ (u(s , y))W (s , y) ds dy.

Mild form of the solution.

For us, the mild solution continues to make sense and is in facta weak solution.We care only about mild solutions in general.Let us study the simplest case first where b ≡ 0 and σ ≡ 1.

July 28, 2016 2 / 17

Page 11: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + b(u) + σ (u)W ; u(0) = u0 ∈ L2[0 , 1]; u(t , 0) = u(t , 1) = 0 ∀t > 0First pretend that W is a smooth function. Then ∃! solution ufrom general theory.Apply Duhamel’s principle: u uniquely solves the integralequation

u(t ,x) = (Ptu0)(x) + ∫[0,t]×[0,1] pt−s(x , y)b(u(s , y)) ds dy

+ ∫[0,t]×[0,1] pt−s(x , y)σ (u(s , y))W (s , y) ds dy.

Mild form of the solution.For us, the mild solution continues to make sense and is in facta weak solution.

We care only about mild solutions in general.Let us study the simplest case first where b ≡ 0 and σ ≡ 1.

July 28, 2016 2 / 17

Page 12: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + b(u) + σ (u)W ; u(0) = u0 ∈ L2[0 , 1]; u(t , 0) = u(t , 1) = 0 ∀t > 0First pretend that W is a smooth function. Then ∃! solution ufrom general theory.Apply Duhamel’s principle: u uniquely solves the integralequation

u(t ,x) = (Ptu0)(x) + ∫[0,t]×[0,1] pt−s(x , y)b(u(s , y)) ds dy

+ ∫[0,t]×[0,1] pt−s(x , y)σ (u(s , y))W (s , y) ds dy.

Mild form of the solution.For us, the mild solution continues to make sense and is in facta weak solution.We care only about mild solutions in general.

Let us study the simplest case first where b ≡ 0 and σ ≡ 1.

July 28, 2016 2 / 17

Page 13: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + b(u) + σ (u)W ; u(0) = u0 ∈ L2[0 , 1]; u(t , 0) = u(t , 1) = 0 ∀t > 0First pretend that W is a smooth function. Then ∃! solution ufrom general theory.Apply Duhamel’s principle: u uniquely solves the integralequation

u(t ,x) = (Ptu0)(x) + ∫[0,t]×[0,1] pt−s(x , y)b(u(s , y)) ds dy

+ ∫[0,t]×[0,1] pt−s(x , y)σ (u(s , y))W (s , y) ds dy.

Mild form of the solution.For us, the mild solution continues to make sense and is in facta weak solution.We care only about mild solutions in general.Let us study the simplest case first where b ≡ 0 and σ ≡ 1.July 28, 2016 2 / 17

Page 14: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0

The mild solution is, by definition,u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y),

provided that the Wiener integral exists; i.e.,∫ t0 ds∫ 10 dy |pt−s(x , y)|2 <∞.

Because pt−s(x , y) =∑∞n=1 en(x)en(y) exp{−n2π2(t − s)},

∫ t

0 ds∫ 1

0 dy |pt−s(x , y)|2 = ∫ t

0 ds∞∑

n=1 |en(x)|2e−2n2π2(t−s)

6 2 ∫ t

0 ds∞∑

n=1 e−2n2π2(t−s) - ∞∑n=1 n−2 <∞.

July 28, 2016 3 / 17

Page 15: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0

The mild solution is, by definition,u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y),

provided that the Wiener integral exists; i.e.,∫ t0 ds∫ 10 dy |pt−s(x , y)|2 <∞.Because pt−s(x , y) =∑∞n=1 en(x)en(y) exp{−n2π2(t − s)},

∫ t

0 ds∫ 1

0 dy |pt−s(x , y)|2 = ∫ t

0 ds∞∑

n=1 |en(x)|2e−2n2π2(t−s)

6 2 ∫ t

0 ds∞∑

n=1 e−2n2π2(t−s) - ∞∑n=1 n−2 <∞.

July 28, 2016 3 / 17

Page 16: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0

The mild solution is, by definition,u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y),

provided that the Wiener integral exists; i.e.,∫ t0 ds∫ 10 dy |pt−s(x , y)|2 <∞.Because pt−s(x , y) =∑∞n=1 en(x)en(y) exp{−n2π2(t − s)},∫ t

0 ds∫ 1

0 dy |pt−s(x , y)|2 = ∫ t

0 ds∞∑

n=1 |en(x)|2e−2n2π2(t−s)

6 2 ∫ t

0 ds∞∑

n=1 e−2n2π2(t−s) - ∞∑n=1 n−2 <∞.

July 28, 2016 3 / 17

Page 17: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0

The mild solution is, by definition,u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y),

provided that the Wiener integral exists; i.e.,∫ t0 ds∫ 10 dy |pt−s(x , y)|2 <∞.Because pt−s(x , y) =∑∞n=1 en(x)en(y) exp{−n2π2(t − s)},∫ t

0 ds∫ 1

0 dy |pt−s(x , y)|2 = ∫ t

0 ds∞∑

n=1 |en(x)|2e−2n2π2(t−s)

6 2 ∫ t

0 ds∞∑

n=1 e−2n2π2(t−s)

-∞∑

n=1 n−2 <∞.

July 28, 2016 3 / 17

Page 18: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0

The mild solution is, by definition,u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y),

provided that the Wiener integral exists; i.e.,∫ t0 ds∫ 10 dy |pt−s(x , y)|2 <∞.Because pt−s(x , y) =∑∞n=1 en(x)en(y) exp{−n2π2(t − s)},∫ t

0 ds∫ 1

0 dy |pt−s(x , y)|2 = ∫ t

0 ds∞∑

n=1 |en(x)|2e−2n2π2(t−s)

6 2 ∫ t

0 ds∞∑

n=1 e−2n2π2(t−s) - ∞∑n=1 n−2 <∞.

July 28, 2016 3 / 17

Page 19: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0Therefore the SPDE above has a mild solution u(t ,x).

That solution is the mean-zero Gaussian random fieldu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y).

Theorem(t ,x) 7Ï u(t ,x) is continuous a.s. [up to a modification]. In fact,u ∈ C 14−, 12− a.s.

Outline of Proof (in 2 steps).

1 One proves that for all x, y ∈ [0 , 1], s, t > 0, k > 2,E(|u(t ,x)− u(s , y)|k) 6 Ck

[|t − s|k/4 + |x − y|k/2] .

2 Appeal to a two-parameter version of the Kolmogorov continuitytheorem.

July 28, 2016 4 / 17

Page 20: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0Therefore the SPDE above has a mild solution u(t ,x).That solution is the mean-zero Gaussian random field

u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y).

Theorem(t ,x) 7Ï u(t ,x) is continuous a.s. [up to a modification]. In fact,u ∈ C 14−, 12− a.s.

Outline of Proof (in 2 steps).

1 One proves that for all x, y ∈ [0 , 1], s, t > 0, k > 2,E(|u(t ,x)− u(s , y)|k) 6 Ck

[|t − s|k/4 + |x − y|k/2] .

2 Appeal to a two-parameter version of the Kolmogorov continuitytheorem.

July 28, 2016 4 / 17

Page 21: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0Therefore the SPDE above has a mild solution u(t ,x).That solution is the mean-zero Gaussian random field

u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y).Theorem(t ,x) 7Ï u(t ,x) is continuous a.s. [up to a modification]. In fact,u ∈ C 14−, 12− a.s.

Outline of Proof (in 2 steps).

1 One proves that for all x, y ∈ [0 , 1], s, t > 0, k > 2,E(|u(t ,x)− u(s , y)|k) 6 Ck

[|t − s|k/4 + |x − y|k/2] .

2 Appeal to a two-parameter version of the Kolmogorov continuitytheorem.

July 28, 2016 4 / 17

Page 22: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0Therefore the SPDE above has a mild solution u(t ,x).That solution is the mean-zero Gaussian random field

u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y).Theorem(t ,x) 7Ï u(t ,x) is continuous a.s. [up to a modification]. In fact,u ∈ C 14−, 12− a.s.

Outline of Proof (in 2 steps).

1 One proves that for all x, y ∈ [0 , 1], s, t > 0, k > 2,E(|u(t ,x)− u(s , y)|k) 6 Ck

[|t − s|k/4 + |x − y|k/2] .

2 Appeal to a two-parameter version of the Kolmogorov continuitytheorem.

July 28, 2016 4 / 17

Page 23: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0Therefore the SPDE above has a mild solution u(t ,x).That solution is the mean-zero Gaussian random field

u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y).Theorem(t ,x) 7Ï u(t ,x) is continuous a.s. [up to a modification]. In fact,u ∈ C 14−, 12− a.s.

Outline of Proof (in 2 steps).1 One proves that for all x, y ∈ [0 , 1], s, t > 0, k > 2,

E(|u(t ,x)− u(s , y)|k) 6 Ck

[|t − s|k/4 + |x − y|k/2] .

2 Appeal to a two-parameter version of the Kolmogorov continuitytheorem.

July 28, 2016 4 / 17

Page 24: A Stochastic Heat Equation

A Stochastic Heat Equationu = u′′ + W ; u(0) ≡ 0; u(t , 0) = u(t , 1) = 0 ∀t > 0Therefore the SPDE above has a mild solution u(t ,x).That solution is the mean-zero Gaussian random field

u(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y).Theorem(t ,x) 7Ï u(t ,x) is continuous a.s. [up to a modification]. In fact,u ∈ C 14−, 12− a.s.

Outline of Proof (in 2 steps).1 One proves that for all x, y ∈ [0 , 1], s, t > 0, k > 2,

E(|u(t ,x)− u(s , y)|k) 6 Ck

[|t − s|k/4 + |x − y|k/2] .

2 Appeal to a two-parameter version of the Kolmogorov continuitytheorem.July 28, 2016 4 / 17

Page 25: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)A few more details: Of course,

E(|u(t ,x)− u(s , y)|k)6 2k E(|u(t ,x)− u(t , y)|k)︸ ︷︷ ︸

T1+2k E(|u(t , y)− u(s , y)|k)︸ ︷︷ ︸

T2.

Fact. T1 6 Ck|x − y|k/2 and T2 6 Ck|t − s|k/4.Let us begin with the bound on T1.Idea. If Z is Gaussian with mean zero and SD σ , thenE(|Z|k) = Akσk.Therefore, it suffices to prove thatE(|u(t ,x)− u(t , y)|2) - |x − y|.

July 28, 2016 5 / 17

Page 26: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)A few more details: Of course,

E(|u(t ,x)− u(s , y)|k)6 2k E(|u(t ,x)− u(t , y)|k)︸ ︷︷ ︸

T1+2k E(|u(t , y)− u(s , y)|k)︸ ︷︷ ︸

T2.

Fact. T1 6 Ck|x − y|k/2 and T2 6 Ck|t − s|k/4.

Let us begin with the bound on T1.Idea. If Z is Gaussian with mean zero and SD σ , thenE(|Z|k) = Akσk.Therefore, it suffices to prove thatE(|u(t ,x)− u(t , y)|2) - |x − y|.

July 28, 2016 5 / 17

Page 27: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)A few more details: Of course,

E(|u(t ,x)− u(s , y)|k)6 2k E(|u(t ,x)− u(t , y)|k)︸ ︷︷ ︸

T1+2k E(|u(t , y)− u(s , y)|k)︸ ︷︷ ︸

T2.

Fact. T1 6 Ck|x − y|k/2 and T2 6 Ck|t − s|k/4.Let us begin with the bound on T1.

Idea. If Z is Gaussian with mean zero and SD σ , thenE(|Z|k) = Akσk.Therefore, it suffices to prove thatE(|u(t ,x)− u(t , y)|2) - |x − y|.

July 28, 2016 5 / 17

Page 28: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)A few more details: Of course,

E(|u(t ,x)− u(s , y)|k)6 2k E(|u(t ,x)− u(t , y)|k)︸ ︷︷ ︸

T1+2k E(|u(t , y)− u(s , y)|k)︸ ︷︷ ︸

T2.

Fact. T1 6 Ck|x − y|k/2 and T2 6 Ck|t − s|k/4.Let us begin with the bound on T1.Idea. If Z is Gaussian with mean zero and SD σ , thenE(|Z|k) = Akσk.

Therefore, it suffices to prove thatE(|u(t ,x)− u(t , y)|2) - |x − y|.

July 28, 2016 5 / 17

Page 29: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)A few more details: Of course,

E(|u(t ,x)− u(s , y)|k)6 2k E(|u(t ,x)− u(t , y)|k)︸ ︷︷ ︸

T1+2k E(|u(t , y)− u(s , y)|k)︸ ︷︷ ︸

T2.

Fact. T1 6 Ck|x − y|k/2 and T2 6 Ck|t − s|k/4.Let us begin with the bound on T1.Idea. If Z is Gaussian with mean zero and SD σ , thenE(|Z|k) = Akσk.Therefore, it suffices to prove thatE(|u(t ,x)− u(t , y)|2) - |x − y|.

July 28, 2016 5 / 17

Page 30: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)

Now,E(|u(t ,x)− u(t , y)|2) = ∫[0,t]×[0,1] |pt−s(x ,a)− pt−s(y ,a)|2 ds da

= ∫[0,t]×[0,1] |ps(x ,a)− ps(y ,a)|2 ds da.

Since ps(x ,a) =∑∞n=1 en(x)en(a)e−n2π2t ,∫ 10 |ps(x ,a)− ps(y ,a)|2 da = ∞∑

n=1 |en(x)− en(y)|2 |en(a)|2e−2n2π2s.

|en(a)| 6 √2 and |en(x)− en(y)| 6 √2 min{2 ,nπ|x − y|}.Compute.

July 28, 2016 6 / 17

Page 31: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)

Now,E(|u(t ,x)− u(t , y)|2) = ∫[0,t]×[0,1] |pt−s(x ,a)− pt−s(y ,a)|2 ds da

= ∫[0,t]×[0,1] |ps(x ,a)− ps(y ,a)|2 ds da.

Since ps(x ,a) =∑∞n=1 en(x)en(a)e−n2π2t ,∫ 10 |ps(x ,a)− ps(y ,a)|2 da = ∞∑

n=1 |en(x)− en(y)|2 |en(a)|2e−2n2π2s.

|en(a)| 6 √2 and |en(x)− en(y)| 6 √2 min{2 ,nπ|x − y|}.Compute.

July 28, 2016 6 / 17

Page 32: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)

Now,E(|u(t ,x)− u(t , y)|2) = ∫[0,t]×[0,1] |pt−s(x ,a)− pt−s(y ,a)|2 ds da

= ∫[0,t]×[0,1] |ps(x ,a)− ps(y ,a)|2 ds da.

Since ps(x ,a) =∑∞n=1 en(x)en(a)e−n2π2t ,∫ 10 |ps(x ,a)− ps(y ,a)|2 da = ∞∑

n=1 |en(x)− en(y)|2 |en(a)|2e−2n2π2s.

|en(a)| 6 √2 and |en(x)− en(y)| 6 √2 min{2 ,nπ|x − y|}.

Compute.

July 28, 2016 6 / 17

Page 33: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)

Now,E(|u(t ,x)− u(t , y)|2) = ∫[0,t]×[0,1] |pt−s(x ,a)− pt−s(y ,a)|2 ds da

= ∫[0,t]×[0,1] |ps(x ,a)− ps(y ,a)|2 ds da.

Since ps(x ,a) =∑∞n=1 en(x)en(a)e−n2π2t ,∫ 10 |ps(x ,a)− ps(y ,a)|2 da = ∞∑

n=1 |en(x)− en(y)|2 |en(a)|2e−2n2π2s.

|en(a)| 6 √2 and |en(x)− en(y)| 6 √2 min{2 ,nπ|x − y|}.Compute.July 28, 2016 6 / 17

Page 34: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:

I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,

E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2

= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 35: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); and

I T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2

= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 36: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).

By Wiener’s isometry,E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2

= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 37: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,

E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2

= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 38: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,

E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y)

= ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 39: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,

E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 40: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,

E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr

6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 41: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)WLOG 0 < s < t. We next estimate T2 = E(|u(t , y)− u(s , y)|k)by writing

T2 6 2kT2,1 + 2kT2,2,where:I T2,1 := ∫[0,s]×[0,1] (pt−r(y ,a)− ps−r(r ,a)) dW (r ,a); andI T2,2 := ∫[s,t]×[0,1] pt−r(y ,a) dW (r ,a).By Wiener’s isometry,

E(T22,2) = ∫ t

sdr∫ 1

0 da |pt−r(y ,a)|2= ∫ t−s

0 dr∫ 1

0 da pr(y ,a)pr(a , y) = ∫ t−s

0 p2r(y , y) dr

= ∞∑n=1

∫ t−s

0 |en(y)|2e−2n2π2r dr 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr.

July 28, 2016 7 / 17

Page 42: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr

= ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2 · · · - (t − s)1/2.Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2= ∞∑

n=1∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 43: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr = ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2 · · · - (t − s)1/2.Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2= ∞∑

n=1∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 44: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr = ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2

· · · - (t − s)1/2.Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2= ∞∑

n=1∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 45: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr = ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2 · · · - (t − s)1/2.

Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2= ∞∑

n=1∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 46: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr = ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2 · · · - (t − s)1/2.Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2

= ∞∑n=1

∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 47: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr = ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2 · · · - (t − s)1/2.Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2= ∞∑

n=1∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 48: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Therefore,

E(T22,2) 6 2 ∞∑n=1

∫ t−s

0 e−2n2π2r dr = ∞∑n=1

1− e−2n2π2(t−s)n2π2

6∞∑

n=1min (1 , 2n2π2(t − s))

n2π2 · · · - (t − s)1/2.Similarly,

E(T22,1) = ∫ s

0 dr∫ 1

0 da [pt−r(y ,a)− ps−r(y ,a)]2= ∞∑

n=1∫ s

0 dr[e−n2π2(t−s+r) − e−n2π2r

]2|en(y)en(a)|2

6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr.

July 28, 2016 8 / 17

Page 49: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Thus,

E(T22,1) 6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr

= 2 ∞∑n=1

∫ s

0[1− e−n2π2(t−s)]2

e−2n2π2r dr

= ∞∑n=1

[1− e−n2π2(t−s)]2 1− e−2n2π2(t−s)n2π2

6∞∑

n=1 min(1 ,n4π4(t − s)2) min (1 , 2n2π2(t − s))n2π2

· · · - (t − s)1/2.Combine T2,1+T2,2 to see thatT2 = E(|u(t , y)− u(s , y)|2) - (t − s)1/2.

July 28, 2016 9 / 17

Page 50: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Thus,

E(T22,1) 6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr

= 2 ∞∑n=1

∫ s

0[1− e−n2π2(t−s)]2

e−2n2π2r dr

= ∞∑n=1

[1− e−n2π2(t−s)]2 1− e−2n2π2(t−s)n2π2

6∞∑

n=1 min(1 ,n4π4(t − s)2) min (1 , 2n2π2(t − s))n2π2

· · · - (t − s)1/2.Combine T2,1+T2,2 to see thatT2 = E(|u(t , y)− u(s , y)|2) - (t − s)1/2.

July 28, 2016 9 / 17

Page 51: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Thus,

E(T22,1) 6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr

= 2 ∞∑n=1

∫ s

0[1− e−n2π2(t−s)]2

e−2n2π2r dr

= ∞∑n=1

[1− e−n2π2(t−s)]2 1− e−2n2π2(t−s)n2π2

6∞∑

n=1 min(1 ,n4π4(t − s)2) min (1 , 2n2π2(t − s))n2π2

· · · - (t − s)1/2.Combine T2,1+T2,2 to see thatT2 = E(|u(t , y)− u(s , y)|2) - (t − s)1/2.

July 28, 2016 9 / 17

Page 52: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Thus,

E(T22,1) 6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr

= 2 ∞∑n=1

∫ s

0[1− e−n2π2(t−s)]2

e−2n2π2r dr

= ∞∑n=1

[1− e−n2π2(t−s)]2 1− e−2n2π2(t−s)n2π2

6∞∑

n=1 min(1 ,n4π4(t − s)2) min (1 , 2n2π2(t − s))n2π2

· · · - (t − s)1/2.Combine T2,1+T2,2 to see thatT2 = E(|u(t , y)− u(s , y)|2) - (t − s)1/2.

July 28, 2016 9 / 17

Page 53: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x ,a) dW (s ,a)Thus,

E(T22,1) 6 2 ∞∑n=1

∫ s

0[e−n2π2(t−s+r) − e−n2π2r

]2dr

= 2 ∞∑n=1

∫ s

0[1− e−n2π2(t−s)]2

e−2n2π2r dr

= ∞∑n=1

[1− e−n2π2(t−s)]2 1− e−2n2π2(t−s)n2π2

6∞∑

n=1 min(1 ,n4π4(t − s)2) min (1 , 2n2π2(t − s))n2π2

· · · - (t − s)1/2.Combine T2,1+T2,2 to see thatT2 = E(|u(t , y)− u(s , y)|2) - (t − s)1/2.

July 28, 2016 9 / 17

Page 54: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?

Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy

= ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s

t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.

July 28, 2016 10 / 17

Page 55: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy

= ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s

t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.

July 28, 2016 10 / 17

Page 56: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy

= ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s

t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.

July 28, 2016 10 / 17

Page 57: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy = ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s

t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.

July 28, 2016 10 / 17

Page 58: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy = ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s

t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.

July 28, 2016 10 / 17

Page 59: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy = ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.

July 28, 2016 10 / 17

Page 60: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Large-time behavior of x 7Ï u(t ,x)?Spatial covariance function: ∀t > 0 and 0 6 x, z 6 1,E [u(t ,x)u(t , z)] = ∫[0,t]×[0,1] pt−s(x , y)pt−s(z , y) ds dy

= ∫[0,t]×[0,1] ps(x , y)ps(y , z) ds dy = ∫ t

0 p2s(x , z) ds

= ∫ t

0 ds∞∑

n=1 en(x)en(z)e−2n2π2s t→∞−−−Ï∞∑

n=1en(x)en(z)2n2π2 .

We will simplify this sum next.July 28, 2016 10 / 17

Page 61: A Stochastic Heat Equation

A Stochastic Heat Equation∑∞n=1 en (x)en (z)2n2π2 =?Let c0 ≡ 1 and cn(a) := √2 cos(nπa), n > 1, a ∈ [0 , 1]. Then,(

1[0,z] , cn) = √2 ∫ z

0 cos(nπa) da = √2 sin(nπz)nπ = en(z)

nπ ,

∀n > 1 and z ∈ [0 , 1].

Therefore,∞∑

n=1en(x)en(z)2n2π2 = ∞∑

n=1(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)

= ∞∑n=0

(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)−(1[0,x]√2 , c0

)(1[0,z]√2 , c0

)= (1[0,x]√2 ,

1[0,z]√2)− xz2

= 12 [min(x , y)− xy] .

July 28, 2016 11 / 17

Page 62: A Stochastic Heat Equation

A Stochastic Heat Equation∑∞n=1 en (x)en (z)2n2π2 =?Let c0 ≡ 1 and cn(a) := √2 cos(nπa), n > 1, a ∈ [0 , 1]. Then,(

1[0,z] , cn) = √2 ∫ z

0 cos(nπa) da = √2 sin(nπz)nπ = en(z)

nπ ,

∀n > 1 and z ∈ [0 , 1].Therefore,∞∑

n=1en(x)en(z)2n2π2 = ∞∑

n=1(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)

= ∞∑n=0

(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)−(1[0,x]√2 , c0

)(1[0,z]√2 , c0

)= (1[0,x]√2 ,

1[0,z]√2)− xz2

= 12 [min(x , y)− xy] .

July 28, 2016 11 / 17

Page 63: A Stochastic Heat Equation

A Stochastic Heat Equation∑∞n=1 en (x)en (z)2n2π2 =?Let c0 ≡ 1 and cn(a) := √2 cos(nπa), n > 1, a ∈ [0 , 1]. Then,(

1[0,z] , cn) = √2 ∫ z

0 cos(nπa) da = √2 sin(nπz)nπ = en(z)

nπ ,

∀n > 1 and z ∈ [0 , 1].Therefore,∞∑

n=1en(x)en(z)2n2π2 = ∞∑

n=1(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)= ∞∑

n=0(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)−(1[0,x]√2 , c0

)(1[0,z]√2 , c0

)

= (1[0,x]√2 ,1[0,z]√2

)− xz2

= 12 [min(x , y)− xy] .

July 28, 2016 11 / 17

Page 64: A Stochastic Heat Equation

A Stochastic Heat Equation∑∞n=1 en (x)en (z)2n2π2 =?Let c0 ≡ 1 and cn(a) := √2 cos(nπa), n > 1, a ∈ [0 , 1]. Then,(

1[0,z] , cn) = √2 ∫ z

0 cos(nπa) da = √2 sin(nπz)nπ = en(z)

nπ ,

∀n > 1 and z ∈ [0 , 1].Therefore,∞∑

n=1en(x)en(z)2n2π2 = ∞∑

n=1(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)= ∞∑

n=0(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)−(1[0,x]√2 , c0

)(1[0,z]√2 , c0

)= (1[0,x]√2 ,

1[0,z]√2)− xz2

= 12 [min(x , y)− xy] .

July 28, 2016 11 / 17

Page 65: A Stochastic Heat Equation

A Stochastic Heat Equation∑∞n=1 en (x)en (z)2n2π2 =?Let c0 ≡ 1 and cn(a) := √2 cos(nπa), n > 1, a ∈ [0 , 1]. Then,(

1[0,z] , cn) = √2 ∫ z

0 cos(nπa) da = √2 sin(nπz)nπ = en(z)

nπ ,

∀n > 1 and z ∈ [0 , 1].Therefore,∞∑

n=1en(x)en(z)2n2π2 = ∞∑

n=1(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)= ∞∑

n=0(1[0,x]√2 , cn

)(1[0,z]√2 , cn

)−(1[0,x]√2 , c0

)(1[0,z]√2 , c0

)= (1[0,x]√2 ,

1[0,z]√2)− xz2 = 12 [min(x , y)− xy] .

July 28, 2016 11 / 17

Page 66: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)We have shown thus far that

limt→∞

E [u(t ,x)u(t , z)] = 12 [min(x , z)− xz] .

The Brownian bridge W0(x) := [W (x)− xW (1)]/√2 is amean-zero Gaussian process withE [W0(x)W0(z)] = 12 [min(x , z)− xz] .

Therefore, it follows that u(t) t→∞===ÑW0 in the sense offinite-dimensional distributions.Tightness follows from our earlier estimateE(|u(t ,x)− u(t , y)|2) - |x − y|1/2.

July 28, 2016 12 / 17

Page 67: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)We have shown thus far that

limt→∞

E [u(t ,x)u(t , z)] = 12 [min(x , z)− xz] .The Brownian bridge W0(x) := [W (x)− xW (1)]/√2 is amean-zero Gaussian process with

E [W0(x)W0(z)] = 12 [min(x , z)− xz] .

Therefore, it follows that u(t) t→∞===ÑW0 in the sense offinite-dimensional distributions.Tightness follows from our earlier estimateE(|u(t ,x)− u(t , y)|2) - |x − y|1/2.

July 28, 2016 12 / 17

Page 68: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)We have shown thus far that

limt→∞

E [u(t ,x)u(t , z)] = 12 [min(x , z)− xz] .The Brownian bridge W0(x) := [W (x)− xW (1)]/√2 is amean-zero Gaussian process with

E [W0(x)W0(z)] = 12 [min(x , z)− xz] .Therefore, it follows that u(t) t→∞===ÑW0 in the sense offinite-dimensional distributions.

Tightness follows from our earlier estimateE(|u(t ,x)− u(t , y)|2) - |x − y|1/2.

July 28, 2016 12 / 17

Page 69: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)We have shown thus far that

limt→∞

E [u(t ,x)u(t , z)] = 12 [min(x , z)− xz] .The Brownian bridge W0(x) := [W (x)− xW (1)]/√2 is amean-zero Gaussian process with

E [W0(x)W0(z)] = 12 [min(x , z)− xz] .Therefore, it follows that u(t) t→∞===ÑW0 in the sense offinite-dimensional distributions.Tightness follows from our earlier estimate

E(|u(t ,x)− u(t , y)|2) - |x − y|1/2.July 28, 2016 12 / 17

Page 70: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 71: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;

u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 72: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 73: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).

As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 74: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).

As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].

Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 75: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].

Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]

An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 76: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]

An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 77: A Stochastic Heat Equation

A Stochastic Heat EquationTheorem (Funaki)Let u denote the solution to the SPDE

u = u′′ + W on R+ × [0 , 1],subject to:

u(0) ≡ 0;u(t , 0) = u(t , 1) = 0 ∀t > 0.

Then, u(t) C[0 ,1]===ÑW0 as t →∞.

If v solves the same SPDE subject to v(0) = f ∈ L2[0 , 1], thenv(t ,x) = (Ptf )(x) + u(t ,x).As t →∞, (Ptf )(x) =∑∞n=1(f ,en)en(x)e−n2π2t → 0 in L2[0 , 1].Therefore, v(t) L2[0 ,1]====ÑW0 as t →∞ ∀v(0) ∈ L2[0 , 1]An infinite-dimensional ergodic theorem.

July 28, 2016 13 / 17

Page 78: A Stochastic Heat Equation

A Stochastic Heat Equation

July 28, 2016 14 / 17

Page 79: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Once again consider u = u′′ + W on (0 ,∞)× [0 , 1] withDirichlet zero boundary and initial value zero.

Since pt−s(x , y) =∑∞n=1 en(x)en(y)e−n2π2(t−s),u(t ,x) = ∞∑

n=1 en(x)∫[0,t]×[0,1] en(y)e−n2π2(t−s) dW (s , y)

= √2 ∞∑n=1 sin(nπx)∫ t

0 e−n2π2(t−s) dXn(s)︸ ︷︷ ︸=Yn(t).

dYn(t) = dXn(t)− n2π2Yn(t) dt.Yn ’s are Ornstein–Uhlenbeck processes.

July 28, 2016 15 / 17

Page 80: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Once again consider u = u′′ + W on (0 ,∞)× [0 , 1] withDirichlet zero boundary and initial value zero.Since pt−s(x , y) =∑∞n=1 en(x)en(y)e−n2π2(t−s),u(t ,x) = ∞∑

n=1 en(x) ∫[0,t]×[0,1] en(y)e−n2π2(t−s) dW (s , y)

= √2 ∞∑n=1 sin(nπx)∫ t

0 e−n2π2(t−s) dXn(s)︸ ︷︷ ︸=Yn(t).

dYn(t) = dXn(t)− n2π2Yn(t) dt.Yn ’s are Ornstein–Uhlenbeck processes.

July 28, 2016 15 / 17

Page 81: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Once again consider u = u′′ + W on (0 ,∞)× [0 , 1] withDirichlet zero boundary and initial value zero.Since pt−s(x , y) =∑∞n=1 en(x)en(y)e−n2π2(t−s),u(t ,x) = ∞∑

n=1 en(x) ∫[0,t]×[0,1] en(y)e−n2π2(t−s) dW (s , y)= √2 ∞∑

n=1 sin(nπx)∫ t

0 e−n2π2(t−s) dXn(s)︸ ︷︷ ︸=Yn(t).

dYn(t) = dXn(t)− n2π2Yn(t) dt.Yn ’s are Ornstein–Uhlenbeck processes.

July 28, 2016 15 / 17

Page 82: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Once again consider u = u′′ + W on (0 ,∞)× [0 , 1] withDirichlet zero boundary and initial value zero.Since pt−s(x , y) =∑∞n=1 en(x)en(y)e−n2π2(t−s),u(t ,x) = ∞∑

n=1 en(x) ∫[0,t]×[0,1] en(y)e−n2π2(t−s) dW (s , y)= √2 ∞∑

n=1 sin(nπx)∫ t

0 e−n2π2(t−s) dXn(s)︸ ︷︷ ︸=Yn(t).

dYn(t) = dXn(t)− n2π2Yn(t) dt.

Yn ’s are Ornstein–Uhlenbeck processes.

July 28, 2016 15 / 17

Page 83: A Stochastic Heat Equation

A Stochastic Heat Equationu(t ,x) = ∫[0,t]×[0,1] pt−s(x , y) dW (s , y)

Once again consider u = u′′ + W on (0 ,∞)× [0 , 1] withDirichlet zero boundary and initial value zero.Since pt−s(x , y) =∑∞n=1 en(x)en(y)e−n2π2(t−s),u(t ,x) = ∞∑

n=1 en(x) ∫[0,t]×[0,1] en(y)e−n2π2(t−s) dW (s , y)= √2 ∞∑

n=1 sin(nπx)∫ t

0 e−n2π2(t−s) dXn(s)︸ ︷︷ ︸=Yn(t).

dYn(t) = dXn(t)− n2π2Yn(t) dt.Yn ’s are Ornstein–Uhlenbeck processes.

July 28, 2016 15 / 17

Page 84: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 85: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].

I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 86: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.

I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 87: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].

I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 88: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.

I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theformu = u′′ + b(u) + σ (u)W .

First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 89: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .

Instead we return to the nonlinear/semilinear SPDEs of theformu = u′′ + b(u) + σ (u)W .

First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 90: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .

First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 91: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 92: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.

2 Uniqueness.3 Local regularity.Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 93: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.

3 Local regularity.Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 94: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.

Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 95: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + WOther constant-coefficient SPDEs of interest include:

I Different boundary conditions [Neumann, mixed, . . . ].I u = Au + W , where A = generator of a nice Markov process.I u = Au + W [Stoch. Wave Eq.].I u = Au + F for a different noise model.I Etc. . . .Instead we return to the nonlinear/semilinear SPDEs of theform

u = u′′ + b(u) + σ (u)W .First-order questions:

1 Existence.2 Uniqueness.3 Local regularity.Questions about the structure of the solution, if any ∃.

July 28, 2016 16 / 17

Page 96: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.

For example:

1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].

The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 97: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:

1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 98: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].

2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 99: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].

3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 100: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].

4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 101: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].

The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 102: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.

We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 103: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].

Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17

Page 104: A Stochastic Heat Equation

General Remarks on SPDEsu = u′′ + b(u) + σ (u)W

There is also interest in these equations when b and/or σ arenon-Lipschitz continuous.For example:1 b(u) ≡ 0 and σ (u) =√|u| [Population genetics].2 b(u) = u − u3 and σ (u) = 1 [QFT].3 b(u) = u − u3 and σ (u) = u [Statistical mechanics].4 b(u) ≡ 0 and σ (u) = u [KPZ+random polymers].The last Item #4 is included in the general theory to followsince b and σ are Lipschitz continuous.We will say a few things about Items #2 and #3 as well[Allen-Cahn, Φ41, Swift–Hohenberg, Ginzburg–Landau, . . . ].Item #1 is perhaps the best-studied case [super BM,Brownian density process]. Not covered here [Dawson, Dynkin,Mytnik, Perkins, Shiga, Watanabe, . . . ].

July 28, 2016 17 / 17