On 2D Inverse Problems/The case of the unit disc

From testwiki
Revision as of 00:43, 22 January 2016 by imported>DVD206
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

The operator equation

The continuous Dirichlet-to-Neumann operator can be calculated explicitly for certain domains, such as a half-space, a ball and a cylinder and a shell with uniform conductivity. For example, for a unit ball in N-dimensions, writing the Laplace equation in spherical coordinates:
Δf=r1Nr(rN1fr)+r2ΔSN1f,

and, therefore, the Dirichlet-to-Neumann operator satisfies the following equation:

Λ(Λ(N2)Id)+ΔSN1=0.

In two-dimensions the equation takes a particularly simple form:

Λ2=ΔS1.

The study of material of this chapter is largely motivated by the question of Professor of Mathematics at the University of Washington Gunther Uhlmann: "Is there a discrete analog of the equation?"

The network setting

To match the functional equation for the Dirichlet-to-Neumann operator of the unit disc with uniform conductivity, is to find the self-dual layered planar network with rotational symmetry. The Dirichlet-to-Neumann operator for such graph G is equal to:

ΛG2=L,

where -L is equal to the Laplacian on the circle:

L=(2101121001211012).
Exercise(*). Prove that the entries of the cofactor matrix of ΛG are ±1 w/the chessboard pattern.
The problem then reduces to calculating a Stieltjes continued fraction equalled to 1 at the non-zero eigenvalues of L. For the (2n+1)-case, where n is a natural number, the eigenvalues are 0 with the multiplicity one and
2sin(kπ2n+1),k=1,2,n

w/multiplicity two. The existence and uniqueness of such fraction with n levels follow from our results on layered networks, see [BIMS].

Exercise (***). Prove that the continued fraction is given by the following formula:
β(z)=cot(nπ2n+1)z+1cot((n1)π2n+1)z+1+1cot(π2n+1)z.
Exercise 2 (*). Use the previous exercise to prove the trigonometric formula:
tan(nπ2n+1)=2ksin(kπ2n+1).
Exercise 3(**). Find the right signs in the following trigonometric formula
tan(lπ2n+1)=2k(±)sin(kπ2n+1),l=1,2,n.

Example: the following picture provides the solution for n=8 w/white and black squares representing 1s and -1s.

Correct signs

Template:BookCat