1. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Enclosure method for the p-Laplacian
Tommi Brander
Joint with Manas Kar and Mikko Salo
University of Jyv¨askyl¨a
9.12.2014
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
2. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Outline
1 Calder´on’s problem for the p-Laplace equation
2 Detecting inclusions
3 Main idea and outline
4 Details of the proof
5 Questions or comments?
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
3. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Calder´on’s problem for the p-Laplace equation
Suppose Ω is a bounded open set, and 0 < 1/C < σ < C in Ω.
Consider the weighted p-Laplace equation with 1 < p < ∞
∆σ
p (u) = div σ(x)| u|p−2 u = 0 in Ω,
u = f on ∂Ω.
Calder´on’s problem asks if we can determine the conductivity σ
from the knowledge of the DN map f → Λσ(f ) = σ| u|p−2 u · ν.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
4. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Calder´on’s problem for the p-Laplace equation
In the (isotropic) p = 2 case much is known: One can recover
Lipschitz conductivities when d ≥ 3 [Haberman-Tataru 2013,
Caro-Rogers 2014] and L∞ conductivites when d = 2
[Astala-P¨aiv¨arinta 2006].
The p = 2 case has been investigated less: One can recover
continuous conductivity on the boundary of C1 domain
[Salo-Zhong 2012] and, with stronger regularity assumptions,
σ on the boundary [Brander 2014].
d-Laplacian is useful in conformal geometry. 0- and
1-Laplacian are used in UMEIT and CDI, respectively.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
5. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Detecting inclusions
Suppose we have an inclusion D ⊆ Ω; that is, σ 1 or 1 in D
and σ = 1 elsewhere. What can we say about D based on
boundary measurements?
Theorem (B.-Kar-Salo 2014)
Suppose D has Lipschitz boundary, 0 < c < σ < C < 1 or
1 < c < σ < C < ∞ in D, and σ = 1 in Ω D. The we can
recover the convex hull of D from the Dirichlet to Neumann map.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
6. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Detecting inclusions
In the p = 2 case there are several well-studied methods.
[Isakov 1988] suggested using singular solutions to detect
obstacles in known background medium. The suggestion has
been implemented in sampling and probing methods, such as
linear sampling, factorisation, probe and singular source
methods.
[Ikehata 1999] introduced the enclosure method which avoids
some problems of the methods based on singular solutions.
We use Ikehata’s enclosure method.
The enclosure method has been applied to many linear
equations, including Hemlholtz and Maxwell equations.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
7. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Main idea and outline
Theorem (B.-Kar-Salo 2014)
If σ0, σ1 ∈ L∞
+ (Ω) and 1 < p < ∞, and if f ∈ W 1,p(Ω), then
(p − 1)
Ω
σ0
σ
1/(p−1)
1
σ
1
p−1
1 − σ
1
p−1
0 | u0|p
dx (1)
≤ ((Λσ1 − Λσ0 )f , f ) ≤
Ω
(σ1 − σ0) | u0|p
dx, (2)
where u0 ∈ W 1,p(Ω) solves div(σ0 | u0|p−2
u0) = 0 in Ω with
u0|∂Ω = f .
We will later define the indicator function to equal ((Λσ − Λ1)f , f )
for special solutions f .
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
8. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Main idea and outline
Define the p-harmonic function
f = u0(x, ρ, t, τ) = exp (τ (x · ρ − t)) a(τx · ρ⊥
),
where a is periodic and rapidly oscillating, ρ is direction with
ρ · ρ⊥ = 0, t is time and τ is large parameter.
When x · ρ < t, the function and its oscillations are small.
When x · ρ > t, the function and its oscillations are large.
For large τ almost all of the energy of the system will be on the
side x · ρ > t.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
9. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Main idea and outline
Define the indicator function
I(ρ, t, τ) = τn−p
((Λσ − Λ1)f , f ).
By selecting the direction ρ and the time t we focus energy on the
half-space x · ρ > t.
We can detect when inclusion intersects the half-space: by the
monotonicity inequality the indicator function becomes large
as τ grows.
We can also detect when the inclusion is completely outside
the half-space, since the energy will be very small (by
monotonicity inequality) as τ grows.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
10. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Main idea and outline
The identity
I(ρ, t, τ) = exp(2τ(hD(ρ) − t))I(ρ, hD(ρ), τ)
implies that we only need to investigate the time
t = hD(ρ) = sup
x∈D
x · ρ.
The wave of energy and the inclusion barely touch at this time.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
11. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Main idea and outline
Theorem (B.-Kar-Salo 2014)
There exist c, C > 0 such that
c < |I(ρ, hD(ρ), τ)| < Cτn
(3)
for τ 1.
With the previous identity this implies that the indicator function
grows exponentially fast when t < hD(ρ) and vanishes
exponentially when t > hD(ρ).
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
12. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Main idea and outline
The upper bound is easy to prove and only requires τ > 0. The
proof uses the definition of Wolff solutions, monotonicity
inequality, and the fact that σ > 1 in the inclusion D.
The lower bound is much more tricky, and also uses the Lipschitz
regularity of the inclusion.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
13. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Details of the proof
From now on, we consider the direction ρ to be fixed. We are
trying to show that the inequalities
c < |I(ρ, hD(ρ), τ)| < Cτn
(4)
hold for large τ. We will omit ρ as an argument to make the
expressions simpler.
We first prove the upper bound.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
14. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Details of the proof
By monotonicity inequality we have
|I(hD, τ)| ≤ τn−p
Ω
(σ − 1) | u0|p
dx. (5)
A property of the Wolff solutions u0 is that
| u0(x)|p
≈ τp
exp(pτ(x · ρ − hD)) ≤ Cτp
, (6)
since hD = supx∈D x · ρ. By assumption σ > 1 + ε on a set D of
positive measure.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
15. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Details of the proof
We now start the proof of the lower bound. First, monotonicity
inequality implies
|I(τ, hD)| ≥ τ(n−p)
Ω
(p − 1)
1
σ
1
p−1
(σ
1
p−1 − 1)| u0|p
dx (7)
≥ Cτ(n−p)
D
| u0|p
dx.
The estimate used for upper bound is not helpful, as it gives
exponentially decreasing lower bound.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
16. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Details of the proof
We cover the compact set ∂D ∩ {x ∈ Ω; x · ρ = hD} by a finite
collection of balls B(αj , δ) with αj in the set, and define
Dδ = D ∩ (∪j B(αj , δ)).
Then there is c > 0 with
DDδ
e−pτ(hD (ρ)−x·ρ)
dx = O(e−pcτ
) (8)
as τ → ∞.
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian
17. Calder´on’s problem for the p-Laplace equation
Detecting inclusions
Main idea and outline
Details of the proof
Questions or comments?
Questions or comments?
The article this presentation is based on can be found in the arXiv:
http://arxiv.org/abs/1410.4048
Tommi Brander Joint with Manas Kar and Mikko Salo Enclosure method for the p-Laplacian