Elliptic boundary value problems with complex coefficients and - - PowerPoint PPT Presentation

elliptic boundary value problems with complex
SMART_READER_LITE
LIVE PREVIEW

Elliptic boundary value problems with complex coefficients and - - PowerPoint PPT Presentation

Elliptic boundary value problems with complex coefficients and fractional regularity data (the first order approach) Alex Amenta (joint work with Pascal Auscher) Delft University of Technology, Netherlands August 18, 2017 BVP / L


slide-1
SLIDE 1

Elliptic boundary value problems with complex coefficients and fractional regularity data

(the first order approach) Alex Amenta (joint work with Pascal Auscher)

Delft University of Technology, Netherlands

August 18, 2017

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 1 / 17

slide-2
SLIDE 2

The divergence form elliptic equation

div A∇u = 0, u: R1+n

+

→ C.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 2 / 17

slide-3
SLIDE 3

The divergence form elliptic equation

div A∇u = 0, u: R1+n

+

→ C. The coefficients A(t, x) = A(x) ∈ L∞(Rn : L(C1+n)) are t-independent,

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 2 / 17

slide-4
SLIDE 4

The divergence form elliptic equation

div A∇u = 0, u: R1+n

+

→ C. The coefficients A(t, x) = A(x) ∈ L∞(Rn : L(C1+n)) are t-independent, uniformly elliptic: there exists κ > 0 such that Re(A(x)v, v) ≥ κ|v|2 ∀v ∈ C1+n, x ∈ Rn,

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 2 / 17

slide-5
SLIDE 5

The divergence form elliptic equation

div A∇u = 0, u: R1+n

+

→ C. The coefficients A(t, x) = A(x) ∈ L∞(Rn : L(C1+n)) are t-independent, uniformly elliptic: there exists κ > 0 such that Re(A(x)v, v) ≥ κ|v|2 ∀v ∈ C1+n, x ∈ Rn, not assumed to be real, symmetric, or smooth in any way.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 2 / 17

slide-6
SLIDE 6

The divergence form elliptic equation

div A∇u = 0, u: R1+n

+

→ C. The coefficients A(t, x) = A(x) ∈ L∞(Rn : L(C1+n)) are t-independent, uniformly elliptic: there exists κ > 0 such that Re(A(x)v, v) ≥ κ|v|2 ∀v ∈ C1+n, x ∈ Rn, not assumed to be real, symmetric, or smooth in any way. Maximum principle, existence of fundamental solutions, local H¨

  • lder

regularity of solutions all fail.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 2 / 17

slide-7
SLIDE 7

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-8
SLIDE 8

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-9
SLIDE 9

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

T p

θ : weighted tent space (definition on next slide)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-10
SLIDE 10

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

T p

θ : weighted tent space (definition on next slide)

˙ Hp

θ : homogeneous Hardy–Sobolev space of order θ

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-11
SLIDE 11

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

T p

θ : weighted tent space (definition on next slide)

˙ Hp

θ : homogeneous Hardy–Sobolev space of order θ

tangential gradient: ∇u = (∂1u, . . . , ∂nu)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-12
SLIDE 12

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

T p

θ : weighted tent space (definition on next slide)

˙ Hp

θ : homogeneous Hardy–Sobolev space of order θ

tangential gradient: ∇u = (∂1u, . . . , ∂nu) A-conormal derivative: ∂νAu = e0 · A∇u (e0: unit vector in the t-direction).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-13
SLIDE 13

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

T p

θ : weighted tent space (definition on next slide)

˙ Hp

θ : homogeneous Hardy–Sobolev space of order θ

tangential gradient: ∇u = (∂1u, . . . , ∂nu) A-conormal derivative: ∂νAu = e0 · A∇u (e0: unit vector in the t-direction). Say (N)p

θ,A is well-posed if for every boundary data ∂νAf there exists a unique

solution u satisfying the given conditions.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-14
SLIDE 14

Boundary value problems

For θ ∈ [−1, 0) and p > 1, formulate the Neumann problem (N)p

θ,A :

       div A∇u = 0 in R1+n

+

, ||∇u||T p

θ ||∂νAf|| ˙

Hp

θ ,

limt→∞ ∇u(t, ·) = 0 in (S′/P)(Rn : Cn), limt→0 ∂νAu(t, ·) = ∂νAf ∈ ˙ Hp

θ (Rn : C).

T p

θ : weighted tent space (definition on next slide)

˙ Hp

θ : homogeneous Hardy–Sobolev space of order θ

tangential gradient: ∇u = (∂1u, . . . , ∂nu) A-conormal derivative: ∂νAu = e0 · A∇u (e0: unit vector in the t-direction). Say (N)p

θ,A is well-posed if for every boundary data ∂νAf there exists a unique

solution u satisfying the given conditions. Goal: find a useful characterisation of well-posedness of (N)p

θ,A.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 3 / 17

slide-15
SLIDE 15

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p .

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-16
SLIDE 16

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p . a solution u to (N)p

θ,A with boundary data ∂νAf must satisfy

||∇u||T p

θ ||∂νAf|| ˙

Hp

θ . Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-17
SLIDE 17

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p . a solution u to (N)p

θ,A with boundary data ∂νAf must satisfy

||∇u||T p

θ ||∂νAf|| ˙

Hp

θ .

History of tent spaces:

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-18
SLIDE 18

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p . a solution u to (N)p

θ,A with boundary data ∂νAf must satisfy

||∇u||T p

θ ||∂νAf|| ˙

Hp

θ .

History of tent spaces: Unweighted tent spaces (θ = 0): Coifman–Meyer–Stein 1985.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-19
SLIDE 19

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p . a solution u to (N)p

θ,A with boundary data ∂νAf must satisfy

||∇u||T p

θ ||∂νAf|| ˙

Hp

θ .

History of tent spaces: Unweighted tent spaces (θ = 0): Coifman–Meyer–Stein 1985. First definition with θ = 0: Hofmann–Mayboroda–McIntosh 2011.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-20
SLIDE 20

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p . a solution u to (N)p

θ,A with boundary data ∂νAf must satisfy

||∇u||T p

θ ||∂νAf|| ˙

Hp

θ .

History of tent spaces: Unweighted tent spaces (θ = 0): Coifman–Meyer–Stein 1985. First definition with θ = 0: Hofmann–Mayboroda–McIntosh 2011. ‘General theory’: Huang 2016 (complex interpolation and factorisation), A. 2017 (real interpolation and embeddings).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-21
SLIDE 21

Weighted tent spaces

||F||T p

θ :=

Rn

  • B(x,t)

|t−θF(t, y)|2 dy dt tn+1 p/2 dx 1/p . a solution u to (N)p

θ,A with boundary data ∂νAf must satisfy

||∇u||T p

θ ||∂νAf|| ˙

Hp

θ .

History of tent spaces: Unweighted tent spaces (θ = 0): Coifman–Meyer–Stein 1985. First definition with θ = 0: Hofmann–Mayboroda–McIntosh 2011. ‘General theory’: Huang 2016 (complex interpolation and factorisation), A. 2017 (real interpolation and embeddings). Weighted tent spaces satisfy Hardy–Littlewood–Sobolev-type embeddings: T p0

θ0 ֒

→ T p1

θ1

(p1 − p0) = 1 n(θ1 − θ0), p1 ≥ p0.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 4 / 17

slide-22
SLIDE 22

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 5 / 17

slide-23
SLIDE 23

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach:

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 5 / 17

slide-24
SLIDE 24

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 5 / 17

slide-25
SLIDE 25

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’) Classify solutions to Cauchy–Riemann systems in various function spaces (eg. tent spaces) via functional calculus/semigroups

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 5 / 17

slide-26
SLIDE 26

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’) Classify solutions to Cauchy–Riemann systems in various function spaces (eg. tent spaces) via functional calculus/semigroups See boundary data for (N)p

θ,A as a projection NA,p,θ of the initial value of a

Cauchy–Riemann system.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 5 / 17

slide-27
SLIDE 27

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’) Classify solutions to Cauchy–Riemann systems in various function spaces (eg. tent spaces) via functional calculus/semigroups See boundary data for (N)p

θ,A as a projection NA,p,θ of the initial value of a

Cauchy–Riemann system. ‘Theorem’: For a range of parameters (p, θ) depending on A, (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 5 / 17

slide-28
SLIDE 28

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’) Classify solutions to Cauchy–Riemann systems in various function spaces (eg. tent spaces) via functional calculus/semigroups See boundary data for (N)p

θ,A as a projection NA,p,θ of the initial value of a

Cauchy–Riemann system. ‘Theorem’: For a range of parameters (p, θ) depending on A, (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 6 / 17

slide-29
SLIDE 29

Dirac operators and Cauchy–Riemann systems

the Dirac operator D acts on distributions F : R1+n

+

→ C1+n:

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 7 / 17

slide-30
SLIDE 30

Dirac operators and Cauchy–Riemann systems

the Dirac operator D acts on distributions F : R1+n

+

→ C1+n: DF =

  • div

−∇ F⊥ F

  • =
  • div F

−∇F⊥

  • .

(here F⊥ : R1+n

+

→ C and F : R1+n

+

→ Cn)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 7 / 17

slide-31
SLIDE 31

Dirac operators and Cauchy–Riemann systems

the Dirac operator D acts on distributions F : R1+n

+

→ C1+n: DF =

  • div

−∇ F⊥ F

  • =
  • div F

−∇F⊥

  • .

(here F⊥ : R1+n

+

→ C and F : R1+n

+

→ Cn) Let B ∈ L∞(Rn : L(C1+n)) satisfy the same assumptions as A (elliptic, t-independent), and define the perturbed Dirac operator DB as an unbounded

  • perator on L2(Rn : C1+n) with natural domain.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 7 / 17

slide-32
SLIDE 32

Dirac operators and Cauchy–Riemann systems

the Dirac operator D acts on distributions F : R1+n

+

→ C1+n: DF =

  • div

−∇ F⊥ F

  • =
  • div F

−∇F⊥

  • .

(here F⊥ : R1+n

+

→ C and F : R1+n

+

→ Cn) Let B ∈ L∞(Rn : L(C1+n)) satisfy the same assumptions as A (elliptic, t-independent), and define the perturbed Dirac operator DB as an unbounded

  • perator on L2(Rn : C1+n) with natural domain.

The Cauchy–Riemann system for DB is (CR)DB : ∂tF + DBF = 0 in R1+n

+

, F ∈ R(D), with solutions considered in the usual weak sense.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 7 / 17

slide-33
SLIDE 33

CR-systems vs. elliptic equations

A-Conormal gradient: ∇Au := ∂νAu ∇u

  • Alex Amenta (TU Delft)

BVP / L∞ coefficients / fractional regularity data August 18, 2017 8 / 17

slide-34
SLIDE 34

CR-systems vs. elliptic equations

A-Conormal gradient: ∇Au := ∂νAu ∇u

  • This transforms solutions of div A∇u = 0 to solutions of (CR)D ˆ

A, where ˆ

A is a transformed coefficient matrix, and conversely:

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 8 / 17

slide-35
SLIDE 35

CR-systems vs. elliptic equations

A-Conormal gradient: ∇Au := ∂νAu ∇u

  • This transforms solutions of div A∇u = 0 to solutions of (CR)D ˆ

A, where ˆ

A is a transformed coefficient matrix, and conversely: Theorem (Auscher–Axelsson–McIntosh 2010) F solves (CR)D ˆ

A ⇔ F = ∇Au for a (unique) u such that div A∇u = 0.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 8 / 17

slide-36
SLIDE 36

CR-systems vs. elliptic equations

A-Conormal gradient: ∇Au := ∂νAu ∇u

  • This transforms solutions of div A∇u = 0 to solutions of (CR)D ˆ

A, where ˆ

A is a transformed coefficient matrix, and conversely: Theorem (Auscher–Axelsson–McIntosh 2010) F solves (CR)D ˆ

A ⇔ F = ∇Au for a (unique) u such that div A∇u = 0.

In the splitting C1+n = C ⊕ Cn, ˆ A is defined by A =: A⊥⊥ A⊥ A⊥ A

  • ,

ˆ A := I A⊥ A A⊥⊥ A⊥ I −1 .

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 8 / 17

slide-37
SLIDE 37

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’) Classify solutions to Cauchy–Riemann systems in various function spaces (eg. tent spaces) via functional calculus/semigroups See boundary data for (N)p

θ,A as a projection NA,p,θ of the initial value of a

Cauchy–Riemann system. ‘Theorem’: For a range of parameters (p, θ) depending on A, (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 9 / 17

slide-38
SLIDE 38

DB-adapted Hardy–Sobolev spaces

DB is bisectorial, with bounded H∞ functional calculus on R(DB) ⊂ L2(Rn) (Axelsson–Keith–McIntosh 2006, Auscher–Axelsson–McIntosh 2010)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 10 / 17

slide-39
SLIDE 39

DB-adapted Hardy–Sobolev spaces

DB is bisectorial, with bounded H∞ functional calculus on R(DB) ⊂ L2(Rn) (Axelsson–Keith–McIntosh 2006, Auscher–Axelsson–McIntosh 2010) For every ϕ ∈ H∞(Sµ) we can define ϕ(DB) ∈ B(R(DB)), and an ‘extension

  • perator’

(Qϕ,DBf)(t, x) = (ϕ(tDB)f)(x) (t > 0, f ∈ R(DB)).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 10 / 17

slide-40
SLIDE 40

DB-adapted Hardy–Sobolev spaces

DB is bisectorial, with bounded H∞ functional calculus on R(DB) ⊂ L2(Rn) (Axelsson–Keith–McIntosh 2006, Auscher–Axelsson–McIntosh 2010) For every ϕ ∈ H∞(Sµ) we can define ϕ(DB) ∈ B(R(DB)), and an ‘extension

  • perator’

(Qϕ,DBf)(t, x) = (ϕ(tDB)f)(x) (t > 0, f ∈ R(DB)). DB-adapted Hardy–Sobolev spaces Hp

θ,DB are formally defined by

||f||Hp

θ,DB := ||Qϕ,DBf||T p θ .

The norm is (almost) independent of ϕ.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 10 / 17

slide-41
SLIDE 41

DB-adapted Hardy–Sobolev spaces

DB is bisectorial, with bounded H∞ functional calculus on R(DB) ⊂ L2(Rn) (Axelsson–Keith–McIntosh 2006, Auscher–Axelsson–McIntosh 2010) For every ϕ ∈ H∞(Sµ) we can define ϕ(DB) ∈ B(R(DB)), and an ‘extension

  • perator’

(Qϕ,DBf)(t, x) = (ϕ(tDB)f)(x) (t > 0, f ∈ R(DB)). DB-adapted Hardy–Sobolev spaces Hp

θ,DB are formally defined by

||f||Hp

θ,DB := ||Qϕ,DBf||T p θ .

The norm is (almost) independent of ϕ. D-adapted spaces are ‘classical’: Hp

θ,D = ˙

Hp

θ (Rn) ⊕ ( ˙

Hp

θ (Rn : Cn) ∩ N(curl))

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 10 / 17

slide-42
SLIDE 42

Spectral subspaces and Cauchy extensions

Bounded H∞ calculus of DB on R(DB) extends to adapted spaces Hp

θ,DB.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 11 / 17

slide-43
SLIDE 43

Spectral subspaces and Cauchy extensions

Bounded H∞ calculus of DB on R(DB) extends to adapted spaces Hp

θ,DB.

Useful operators can be constructed:

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 11 / 17

slide-44
SLIDE 44

Spectral subspaces and Cauchy extensions

Bounded H∞ calculus of DB on R(DB) extends to adapted spaces Hp

θ,DB.

Useful operators can be constructed: The spectral projections χ+(DB) and χ−(DB) defined via χ+(z) := 1z:Re(z)>0, χ−(z) := 1z:Re(z)<0,

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 11 / 17

slide-45
SLIDE 45

Spectral subspaces and Cauchy extensions

Bounded H∞ calculus of DB on R(DB) extends to adapted spaces Hp

θ,DB.

Useful operators can be constructed: The spectral projections χ+(DB) and χ−(DB) defined via χ+(z) := 1z:Re(z)>0, χ−(z) := 1z:Re(z)<0, which induce a decomposition Hp

θ,DB = Hp,+ θ,DB ⊕ Hp,− θ,DB.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 11 / 17

slide-46
SLIDE 46

Spectral subspaces and Cauchy extensions

Bounded H∞ calculus of DB on R(DB) extends to adapted spaces Hp

θ,DB.

Useful operators can be constructed: The spectral projections χ+(DB) and χ−(DB) defined via χ+(z) := 1z:Re(z)>0, χ−(z) := 1z:Re(z)<0, which induce a decomposition Hp

θ,DB = Hp,+ θ,DB ⊕ Hp,− θ,DB.

The Cauchy extension CDBf(t) := e−tDBχ+(DB)f (t > 0) which acts as a strongly continuous semigroup on Hp,+

θ,DB.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 11 / 17

slide-47
SLIDE 47

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-48
SLIDE 48

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

(This identification has a precise, technical interpretation.)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-49
SLIDE 49

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

(This identification has a precise, technical interpretation.) Then F solves (CR)DB, F ∈ T p

−1 and limt→∞ F(t) = 0 in S′/P

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-50
SLIDE 50

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

(This identification has a precise, technical interpretation.) Then F solves (CR)DB, F ∈ T p

−1 and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

−1,DB ⊂ Hp −1,D.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-51
SLIDE 51

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

(This identification has a precise, technical interpretation.) Then F solves (CR)DB, F ∈ T p

−1 and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

−1,DB ⊂ Hp −1,D.

In this correspondence, f ˙

Hp

−1 ≃ FT p −1. Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-52
SLIDE 52

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

(This identification has a precise, technical interpretation.) Then F solves (CR)DB, F ∈ T p

−1 and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

−1,DB ⊂ Hp −1,D.

In this correspondence, f ˙

Hp

−1 ≃ FT p −1.

θ = 0 case: replace T p

−1 with a certain nontangential maximal function norm.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-53
SLIDE 53

Classification of solutions to (CR)DB: endpoint cases

Theorem (Auscher–Mourgoglou 2015, Auscher–Stahlhut 2016) Fix p ∈ (1, ∞) such that Hp

−1,DB ≃ Hp −1,D.

(This identification has a precise, technical interpretation.) Then F solves (CR)DB, F ∈ T p

−1 and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

−1,DB ⊂ Hp −1,D.

In this correspondence, f ˙

Hp

−1 ≃ FT p −1.

θ = 0 case: replace T p

−1 with a certain nontangential maximal function norm.

  • lder space results (‘p ≥ ∞’) are also available.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 12 / 17

slide-54
SLIDE 54

Classification of solutions to (CR)DB: endpoint cases

Theorem (A.–Auscher 2017) Let θ ∈ (−1, 0) and p ∈ (1, ∞) be such that Hp

θ,DB ≃ Hp θ,D.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 13 / 17

slide-55
SLIDE 55

Classification of solutions to (CR)DB: endpoint cases

Theorem (A.–Auscher 2017) Let θ ∈ (−1, 0) and p ∈ (1, ∞) be such that Hp

θ,DB ≃ Hp θ,D.

Then F solves (CR)DB, F ∈ T p

θ and limt→∞ F(t) = 0 in S′/P

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 13 / 17

slide-56
SLIDE 56

Classification of solutions to (CR)DB: endpoint cases

Theorem (A.–Auscher 2017) Let θ ∈ (−1, 0) and p ∈ (1, ∞) be such that Hp

θ,DB ≃ Hp θ,D.

Then F solves (CR)DB, F ∈ T p

θ and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

θ,DB ⊂ Hp θ,D.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 13 / 17

slide-57
SLIDE 57

Classification of solutions to (CR)DB: endpoint cases

Theorem (A.–Auscher 2017) Let θ ∈ (−1, 0) and p ∈ (1, ∞) be such that Hp

θ,DB ≃ Hp θ,D.

Then F solves (CR)DB, F ∈ T p

θ and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

θ,DB ⊂ Hp θ,D.

In this correspondence, f ˙

Hp

θ ≃ FT p θ . Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 13 / 17

slide-58
SLIDE 58

Classification of solutions to (CR)DB: endpoint cases

Theorem (A.–Auscher 2017) Let θ ∈ (−1, 0) and p ∈ (1, ∞) be such that Hp

θ,DB ≃ Hp θ,D.

Then F solves (CR)DB, F ∈ T p

θ and limt→∞ F(t) = 0 in S′/P

  • F = CDBf for some (unique) f ∈ Hp,+

θ,DB ⊂ Hp θ,D.

In this correspondence, f ˙

Hp

θ ≃ FT p θ .

This does not follow from the previous theorem by interpolation! Our proof only works for θ ∈ (−1, 0).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 13 / 17

slide-59
SLIDE 59

The first-order approach to elliptic BVPs

Approach initiated by Auscher–Axelsson–McIntosh 2010. Key steps of the approach: Identify the second-order equation div A∇u = 0 with a first-order evolution equation (a ‘Cauchy–Riemann system’) Classify solutions to Cauchy–Riemann systems in various function spaces (eg. tent spaces) via functional calculus/semigroups See boundary data for (N)p

θ,A as a projection NA,p,θ of the initial value

  • f a Cauchy–Riemann system.

‘Theorem’: For a range of parameters (p, θ) depending on A, (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 14 / 17

slide-60
SLIDE 60

Classification of well-posedness

Suppose θ ∈ [−1, 0], p ∈ (1, ∞), and Hp

θ,D ˆ A = Hp θ,D.

(required to classify solutions to (CR)D ˆ A in T p θ ) Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 15 / 17

slide-61
SLIDE 61

Classification of well-posedness

Suppose θ ∈ [−1, 0], p ∈ (1, ∞), and Hp

θ,D ˆ A = Hp θ,D.

(required to classify solutions to (CR)D ˆ A in T p θ )

N⊥ : Hp

θ,D = ˙

Hp

θ (Rn) ⊕ ( ˙

Hp

θ (Rn : Cn) ∩ N(curl))) → ˙

Hp

θ (Rn).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 15 / 17

slide-62
SLIDE 62

Classification of well-posedness

Suppose θ ∈ [−1, 0], p ∈ (1, ∞), and Hp

θ,D ˆ A = Hp θ,D.

(required to classify solutions to (CR)D ˆ A in T p θ )

N⊥ : Hp

θ,D = ˙

Hp

θ (Rn) ⊕ ( ˙

Hp

θ (Rn : Cn) ∩ N(curl))) → ˙

Hp

θ (Rn).

Identify Hp,+

θ,D ˆ A ⊂ Hp θ,D and restrict the projection N⊥ to define

NA,p,θ : Hp,+

θ,D ˆ A → ˙

Hp

θ (Rn).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 15 / 17

slide-63
SLIDE 63

Classification of well-posedness

Suppose θ ∈ [−1, 0], p ∈ (1, ∞), and Hp

θ,D ˆ A = Hp θ,D.

(required to classify solutions to (CR)D ˆ A in T p θ )

N⊥ : Hp

θ,D = ˙

Hp

θ (Rn) ⊕ ( ˙

Hp

θ (Rn : Cn) ∩ N(curl))) → ˙

Hp

θ (Rn).

Identify Hp,+

θ,D ˆ A ⊂ Hp θ,D and restrict the projection N⊥ to define

NA,p,θ : Hp,+

θ,D ˆ A → ˙

Hp

θ (Rn).

Theorem (Auscher–Mourgoglou 2014, A.–Auscher 2017) (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 15 / 17

slide-64
SLIDE 64

Classification of well-posedness

Suppose θ ∈ [−1, 0], p ∈ (1, ∞), and Hp

θ,D ˆ A = Hp θ,D.

(required to classify solutions to (CR)D ˆ A in T p θ )

N⊥ : Hp

θ,D = ˙

Hp

θ (Rn) ⊕ ( ˙

Hp

θ (Rn : Cn) ∩ N(curl))) → ˙

Hp

θ (Rn).

Identify Hp,+

θ,D ˆ A ⊂ Hp θ,D and restrict the projection N⊥ to define

NA,p,θ : Hp,+

θ,D ˆ A → ˙

Hp

θ (Rn).

Theorem (Auscher–Mourgoglou 2014, A.–Auscher 2017) (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Idea: every F0 ∈ Hp,+

θ,D ˆ A is the initial value of a solution F = ∇Au of (CR)D ˆ A,

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 15 / 17

slide-65
SLIDE 65

Classification of well-posedness

Suppose θ ∈ [−1, 0], p ∈ (1, ∞), and Hp

θ,D ˆ A = Hp θ,D.

(required to classify solutions to (CR)D ˆ A in T p θ )

N⊥ : Hp

θ,D = ˙

Hp

θ (Rn) ⊕ ( ˙

Hp

θ (Rn : Cn) ∩ N(curl))) → ˙

Hp

θ (Rn).

Identify Hp,+

θ,D ˆ A ⊂ Hp θ,D and restrict the projection N⊥ to define

NA,p,θ : Hp,+

θ,D ˆ A → ˙

Hp

θ (Rn).

Theorem (Auscher–Mourgoglou 2014, A.–Auscher 2017) (N)p

θ,A is well-posed ⇔ NA,p,θ is an isomorphism.

Idea: every F0 ∈ Hp,+

θ,D ˆ A is the initial value of a solution F = ∇Au of (CR)D ˆ A,

and NA,p,θF0 = (∇Au|t=0)⊥ = ∂νAu|t=0.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 15 / 17

slide-66
SLIDE 66

Consequences

Duality: well-posedness of (N)p

θ,A implies well-posedness of (N)p′ −1−θ,A∗

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 16 / 17

slide-67
SLIDE 67

Consequences

Duality: well-posedness of (N)p

θ,A implies well-posedness of (N)p′ −1−θ,A∗

Interpolation of (compatible) well-posedness (takes too much effort to write rigorously)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 16 / 17

slide-68
SLIDE 68

Consequences

Duality: well-posedness of (N)p

θ,A implies well-posedness of (N)p′ −1−θ,A∗

Interpolation of (compatible) well-posedness (takes too much effort to write rigorously) Extrapolation: well-posedness of (N)p

θ,A implies well-posedness of (N)˜ p ˜ θ,A for

(˜ p, ˜ θ) near (p, θ)

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 16 / 17

slide-69
SLIDE 69

Consequences

Duality: well-posedness of (N)p

θ,A implies well-posedness of (N)p′ −1−θ,A∗

Interpolation of (compatible) well-posedness (takes too much effort to write rigorously) Extrapolation: well-posedness of (N)p

θ,A implies well-posedness of (N)˜ p ˜ θ,A for

(˜ p, ˜ θ) near (p, θ) Some stability in coefficients: w-p of (N)p

θ,A implies w-p of (N)p θ, ˜ A for

˜ A − A∞ sufficiently small (with some restrictions on (p, θ)).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 16 / 17

slide-70
SLIDE 70

What about Besov spaces?

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 17 / 17

slide-71
SLIDE 71

What about Besov spaces?

Replace Hardy–Sobolev spaces ˙ Hp

θ with Besov spaces ˙

Bp,p

θ , and tent spaces T p θ

with Z-spaces Zp

θ ,

||F||Zp

θ :=

  • R1+n

+

2t

t/2

  • B(x,t)

|τ −θF(τ, ξ)|2 dξ dτ p/2 dx dt t 1/p .

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 17 / 17

slide-72
SLIDE 72

What about Besov spaces?

Replace Hardy–Sobolev spaces ˙ Hp

θ with Besov spaces ˙

Bp,p

θ , and tent spaces T p θ

with Z-spaces Zp

θ ,

||F||Zp

θ :=

  • R1+n

+

2t

t/2

  • B(x,t)

|τ −θF(τ, ξ)|2 dξ dτ p/2 dx dt t 1/p . Note that ( ˙ Hp0

θ0 , ˙

Hp1

θ1 )α,p ≃ ˙

Bp,p

θ

(classical) (T p0

θ0 , T p1 θ1 )α,p ≃ Zp θ

(A. 2017).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 17 / 17

slide-73
SLIDE 73

What about Besov spaces?

Replace Hardy–Sobolev spaces ˙ Hp

θ with Besov spaces ˙

Bp,p

θ , and tent spaces T p θ

with Z-spaces Zp

θ ,

||F||Zp

θ :=

  • R1+n

+

2t

t/2

  • B(x,t)

|τ −θF(τ, ξ)|2 dξ dτ p/2 dx dt t 1/p . Note that ( ˙ Hp0

θ0 , ˙

Hp1

θ1 )α,p ≃ ˙

Bp,p

θ

(classical) (T p0

θ0 , T p1 θ1 )α,p ≃ Zp θ

(A. 2017). Our whole theory works identically for θ ∈ (−1, 0).

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 17 / 17

slide-74
SLIDE 74

What about Besov spaces?

Replace Hardy–Sobolev spaces ˙ Hp

θ with Besov spaces ˙

Bp,p

θ , and tent spaces T p θ

with Z-spaces Zp

θ ,

||F||Zp

θ :=

  • R1+n

+

2t

t/2

  • B(x,t)

|τ −θF(τ, ξ)|2 dξ dτ p/2 dx dt t 1/p . Note that ( ˙ Hp0

θ0 , ˙

Hp1

θ1 )α,p ≃ ˙

Bp,p

θ

(classical) (T p0

θ0 , T p1 θ1 )α,p ≃ Zp θ

(A. 2017). Our whole theory works identically for θ ∈ (−1, 0). By interpolation, w-p of (N)p0

0,A and (N)p1 −1,A implies w-p of corresponding

Neumann problems with boundary data in ˙ Bp,p

θ

and gradient in Zp,p

θ

.

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 17 / 17

slide-75
SLIDE 75

What about Besov spaces?

Replace Hardy–Sobolev spaces ˙ Hp

θ with Besov spaces ˙

Bp,p

θ , and tent spaces T p θ

with Z-spaces Zp

θ ,

||F||Zp

θ :=

  • R1+n

+

2t

t/2

  • B(x,t)

|τ −θF(τ, ξ)|2 dξ dτ p/2 dx dt t 1/p . Note that ( ˙ Hp0

θ0 , ˙

Hp1

θ1 )α,p ≃ ˙

Bp,p

θ

(classical) (T p0

θ0 , T p1 θ1 )α,p ≃ Zp θ

(A. 2017). Our whole theory works identically for θ ∈ (−1, 0). By interpolation, w-p of (N)p0

0,A and (N)p1 −1,A implies w-p of corresponding

Neumann problems with boundary data in ˙ Bp,p

θ

and gradient in Zp,p

θ

. Thanks for your attention!

Alex Amenta (TU Delft) BVP / L∞ coefficients / fractional regularity data August 18, 2017 17 / 17