4 The essential norm and essential numerical range of the double-layer operator on Lipschitz domains (proofs of
4.2 Proof of Theorems 1.1 and 1.2 in the 2-d case
4.2.1 The norm and numerical range of the double-layer potential operator on a periodic Lipschitz graph
As the main step in the proofs of Theorems1.1and1.2, we calculate in this section lower bounds on the norm and numerical range (and their essential variants) for the double-layer potential operator D on a particular “sawtooth” periodic Lipschitz graph.
Definition 4.4 (The “sawtooth” Lipschitz graph ΓM) Given M > 0 (the Lipschitz constant) define fM : R → R by
fM(s) :=
M s − 2m, 2m ≤ M s ≤ 2m + 1,
2m + 2 − M s, 2m + 1 ≤ M s ≤ 2m + 2, (4.17) for m ∈ Z, so that fM is periodic with period 2/M and
|fM0 (s)| = M,
for almost all s ∈ R. Let ΓM := {(s, fM(s)) : s ∈ R} be the graph of fM, so that ΓM = [
m∈Z
Γm,
where Γmis the open line segment connecting ((1−m)/M, 0) and (−m/M, 1), for m odd, connecting ((1 − m)/M, 1) and (−m/M, 0), for m even. (See Figure4.2for ΓM when M = 1.)
Let DM : L2(ΓM) → L2(ΓM) denote the double-layer potential operator on ΓM, defined with the unit normal n on ΓM such that n2(x) > 0 for almost all x ∈ ΓM. Define V+M ⊂ VM ⊂ L2(ΓM)
by
The next two lemmas do most of the quantitative work for us in proving Theorems 1.1 and 1.2. In particular, Lemma 4.6 provides precise characterisations for kDVMkVM ≤ kDMkL2(ΓM)
and W (DVM) ⊂ W (DM) in terms of symbols of associated infinite Toeplitz matrices.
Lemma 4.5 For M > 0,
In the following lemma we use the notations of§2.3. In particular, wr(CN) denotes the numer-ical abscissa of a square matrix CN, and BN is the matrix defined by (2.25). We denote the norm on L∞(−π, π) by k · k∞.
Lemma 4.6 (The double-layer operator on ΓM) Given N ∈ N, define the orthonormal set {ψ1, ..., ψN} ⊂ V+M ⊂ L2(ΓM) by (4.7), but with Γm as in Definition4.4, and define the Galerkin matrix DN by (4.1), where D = DM is the double-layer potential operator on ΓM. Then:
(a) where E : `2(N) → `2(N) is (multiplication by) the infinite Toeplitz matrix defined by
(E)jm:= sign(m − j)d0m−j, j, m ∈ N,
where er∈ C(R) is 2π-periodic with er(0) = 0, so that
kDVMkVM = kEk`2(N)= kek∞≥ M
2 . (4.26)
(c) For N ≥ 3,
W (DN) ∩ R ⊃ [−a, a], where a :=√
2 d01, (4.27)
with equality when N = 3. Further, where DθN denotes the real part of eiθDN, wr(eiθDN) = wr(e−iθDN) = wr(−eiθDN) = kDNθk2, θ ∈ R, and
wr(eiθDN) = kDNθk2→ kHθk2= khθk∞, as N → ∞, (4.28) where Hθ: `2(N) → `2(N) is (multiplication by) the infinite Toeplitz matrix defined by
Hθ
jm:=1 − e−2iθ(−1)j−m
2 sign(m − j)d0m−j, j, m ∈ N, and hθ∈ L∞(−π, π) is its symbol, given by
hθ(t) = −i
∞
X
m=1
(1 − e−2iθ(−1)m)d0msin(mt) = −iMsign(t)(π − |t|) + e−2iθt
4π + hθr(t), (4.29) for −π < t < π, where hθr∈ C(R) is 2π-periodic with hθr(0) = 0, so that
lim
N →∞wr(eiθDN) = khθk∞≥M
4 , θ ∈ R. (4.30)
Moreover,
W (DVM) =
∞
[
N =1
W (DN) = \
0≤θ≤2π
λ ∈ C : <(eiθλ) ≤ khθk∞ ⊃ {λ ∈ C : |λ| ≤ M/4}.
(4.31) Proof. Part (a). Arguing as in the proof of Lemma4.2, (DN)jm is given by the right hand side of (4.8), but with djm replaced by d0jm, defined by (4.9) and (4.10) with Γm as in Definition4.4.
By symmetry of ΓM and since fM is periodic with period 2/M , d0jm = d0|j−m|,0 = d0|j−m|, for 1 ≤ j, m ≤ N . That d0−` = d0` = d0`,0 → 1/2 as M → ∞, for each ` ∈ N, so that (4.22) holds, follows exactly as in the proof of Lemma4.2, and the asymptotics (4.23) follow easily from the definitions (4.20) and (4.21).
Part (b). kDVMkVM = kDVM
+ kVM by Lemma 4.5, and kDNk2 → kDVM
+ kVM as N → ∞ by (4.4), since H1⊂ H2⊂ ..., where HN is the space spanned by {ψ1, ..., ψN}, and V+M = ∪∞N =1HN. Moreover, it is easy to see that, for each N ∈ N, kDNk2= kENk2, where EN is the order N finite section of E. That kENk2 → kEk2 = kek∞, with e given by the first of the equalities in (4.25), are standard properties of infinite Toeplitz matrices with bounded symbols [10]. The last equation in (4.25), with
er(t) := −2i
∞
X
m=1
(d0m− M/(2πm)) sin(mt), t ∈ R, (4.32) follows since
sign(t)(π − |t|) = 2
∞
X
m=1
sin(mt)
m , −π ≤ t ≤ π, (4.33)
and that er is continuous and er(0) = 0 follows since the series (4.32) is absolutely and uniformly convergent by (4.23). The bound (4.26) follows since kek∞≥ limt→0|e(t)| = M/2.
Part (c). Where CN is the matrix in Lemma 2.8, DN = CN if we set ejm = d0|j−m|, for 1 ≤ j, m ≤ N . Thus (4.27) follows from Lemma2.8(iii), and that wr(eiθDN) = wr(e−iθDN) = wr(−eiθDN) = kDθNk2, for θ ∈ R, from Lemma 2.8(ii) and (iv). Using (2.27), we see also that
Figure 4.3: Graphs of 2|e(t)|/M and 4 max0≤θ≤2π|hθ(t)|/M against t/π, where e and hθ are the symbols of the infinite Toeplitz matrices E and Hθ, given by (4.25) and (4.29), respectively.
Hθ
jm= e−iθ(−1)m+1(DNθ)jm, 1 ≤ j, m ≤ N , so that kDNθk2 = kHNθk2, where HNθ is the order N finite section of Hθ. The remaining results up to and including (4.30) follow in the same way as we proved (4.24)-(4.26), using (4.33) and that
−t = 2
∞
X
m=1
(−1)msin(mt)
m , −π < t < π.
The final statement (4.31) follows from (4.4), (2.26), and (4.30).
Combining Lemma4.5and4.6we see that, for M > 0, kDMkL2(ΓM) ≥ lim
N →∞kDNk2= kDVMkVM ≥M
2 and (4.34)
W (DM) ⊃
∞
[
N =1
W (DN) = W (DVM) ⊃ {λ ∈ C : |λ| ≤ M/4} , (4.35)
so that w(DM) ≥ M/4. On the other hand, by (2.4) and Theorem3.3, for some C, µ > 0, 1
2kDMkL2(ΓM)≤ w(DM) ≤ kDMkL2(ΓM)≤ CM (1 + M )µ, M > 0.
For sufficiently large M > 0 it appears that the last of the bounds in (4.34) and the last inclusion in (4.35) are in fact equalities. Indeed, it follows from (4.26) and (4.30) that (4.34) holds with the last “≥” replaced by “=” if kek∞= M/2, and (4.35) holds with the last “⊃” replaced by “=” if kh∞k∞= M/4, for θ ∈ R, and the plots in Figure4.3suggest the conjecture that
kek∞= M/2, at least for M ≥ 1, and khθk∞= M/4, for θ ∈ R, at least for M ≥ 2.
4.2.2 The proof of Theorems 1.1and 1.2
We now use the above results to prove Theorems1.1and1.2in the 2-d case, as Theorem4.9below.
Let
ΓM,m:=
2m
[
j=1
Γj, m ∈ N, (4.36)
where Γj is as defined in Definition4.4(and see Figure4.2).
Definition 4.7 (ΓMd and ΩMd for d = 2) Choose β ∈ (0, 1) and, given M > 0, define (γj)∞j=1 ⊂ (0, ∞) by
γj :=M (βj−1− βj)
2j , j ∈ N.
0 ... β4 β3 β2 β 1
Figure 4.4: The curve ΓM2 , as specified in Definition4.7, in the case M = 1 and β = 0.6. The labels are the x1-coordinates of the point 0 = (0, 0) and of the first 5 of the points (βj, 0), j = 0, 1, ....
All these points lie on ΓM2 .
Define fβM : R → [0, ∞) by
fβM(s) :=
0, if s ≤ 0 or s ≥ 1,
γjfM((s − βj)/γj), if βj ≤ s < βj−1, (4.37) for j ∈ N, where fM is defined by (4.17), and note that the definition we make for γj ensures that fβM ∈ C0,1(R), with |(fβM)0(s)| = M for almost all s ∈ (0, 1), and that
{fβM(s) : βj≤ s ≤ βj−1} = (βj−1, 0) + γjΓM,j, j ∈ N, (4.38) where ΓM,j is defined by (4.36). Let ε := 0.1 and let
ΓM2 :=
s, fβM(s) : −ε ≤ s ≤ 1 + ε
= {(s, 0) : −ε ≤ s ≤ 0 or 1 ≤ s ≤ 1 + ε} ∪ [
j∈N
(βj−1, 0) + γjΓM,j ; (4.39)
see Figure4.4for Γ12 when β = 0.6. Set x0:= (−ε, −ε) and x00:= (1 + ε, −ε), and let ΩM2 :=(x1, x2) : −ε < x1< 1 + ε and − 2ε < x2< fβM(x1) ∪ Bε(x0) ∪ Bε(x00);
see Figure 1.1for Ω12 when β = 0.6. Note that ΩM2 ⊂ R2 is a simply-connected Lipschitz domain with Lipschitz constant M , and the boundary Γ of ΩM2 contains ΓM2 and is C1except at a countable set of points on
s, fβM(s) : 0 ≤ s ≤ 1 ⊂ ΓM2 .
The proofs of Theorems1.1and1.2, in both the 2-d and 3-d cases, depend on (4.34) and (4.35), and on the localisation result Theorem3.2. They also depend on the simple observation that the norm and numerical range of the double-layer potential on a curve or surface Γ0 are the same as those of the double-layer potential operator on its translate, Γ0+ x, for x ∈ Rd. These quantities are also invariant under scaling in the sense of the following lemma.
Lemma 4.8 Suppose that Γ0 := {(y0, f (y0)) : y0 = (y1, ..., yd−1) ∈ N }, for some open N ⊂ Rd−1, d = 2 or 3, and some f ∈ C0,1(Rd−1), and let Dκ denote the double-layer potential operator on κΓ0 for κ > 0. Then
kDκkL2(κΓ0)= kD1kL2(Γ0) and W (Dκ) = W (D1), κ > 0.
Proof. Define V : L2(κΓ0) → L2(Γ0) by V φ(y) = κ(d−1)/2φ(κy), for φ ∈ L2(Γ0), y ∈ Γ0. Then, arguing exactly as in the proofs of Lemmas3.12and3.13, we see that V is an isometric isomorphism and D1V = V Dκ, so that D1 and Dκ are unitarily equivalent, and the result follows.
Theorem 4.9 (Theorems 1.1 and 1.2 in the 2-d case) Suppose that Γ is the boundary of ΩM2 , defined as in Definition4.7, for some M > 0 and β ∈ (0, 1). Then
kDkL2(Γ),ess≥ M
2 and Wess(D) ⊃ {λ ∈ C : |λ| ≤ M/4}.
Proof. Let x∗:= 0 ∈ ΓM2 ⊂ Γ. By Theorem3.2, kDkL2(Γ),ess≥ lim
δ→0kDx∗,δkL2(Γ) and Wess(D) ⊃ \
δ>0
W (Dx∗,δ).
Thus the result follows if we show that kDx∗,δkL2(Γ) ≥ M/2 and W (Dx∗,δ) ⊃ {λ ∈ C : |λ| ≤ M/4}. By (4.34) and (4.35), this in turn follows if we can show that kDx∗,δkL2(Γ)≥ kDNk2 and W (Dx∗,δ) ⊃ W (DN), for every N ∈ N and δ > 0, where DN is as defined in Lemma4.6. Given m ∈ N, set eΓ := ΓM,m, defined by (4.36). Then, by (4.3) and (4.2), kDNk2 ≤ k eDkL2(eΓ) and W (DN) ⊂ W ( eD), for N = 1, ..., 2m, where eD denotes the double-layer potential operator on eΓ.
But, by construction of ΓM2 in Definition4.7 (see (4.38)), for every δ > 0 there exists κ > 0 and s ∈ R such that bΓ := (s, 0) + κeΓ ⊂ Bδ(x∗) ∩ Γ, so that
k bDkL2(bΓ)≤ kDx∗,δkL2(Γ) and W ( bD) ⊂ W (Dx∗,δ),
by (2.5), where bD denotes the double-layer potential operator on bΓ. Furthermore, by Lemma4.8, k bDkL2(bΓ)= k eDkL2(eΓ) and W ( bD) = W ( eD),
so that
kDx∗,δkL2(Γ)≥ kDNk2 and W (Dx∗,δ) ⊃ W (DN),
for N = 1, ..., 2m. Since this holds for every m ∈ N and δ > 0, the proof is complete.