Minkowski dimension and nonmeasurability of lattice-type fractals - - PowerPoint PPT Presentation

minkowski dimension and nonmeasurability of lattice type
SMART_READER_LITE
LIVE PREVIEW

Minkowski dimension and nonmeasurability of lattice-type fractals - - PowerPoint PPT Presentation

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension and nonmeasurability of lattice-type fractals Erin P . J. Pearse epearse@calpoly.edu California Polytechnic State University Featuring joint work with Michel


slide-1
SLIDE 1

Minkowski dimension and nonmeasurability of lattice-type fractals

Minkowski dimension and nonmeasurability of lattice-type fractals

Erin P . J. Pearse

epearse@calpoly.edu

California Polytechnic State University

Featuring joint work with Michel Lapidus, Sabrina Kombrink, and Steffen Winter

  • Mar. 6, 2015

California Polytechnic State University San Luis Obispo, CA

slide-2
SLIDE 2

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Eigenvalues of the Dirichlet Laplacian

The Dirichlet problem for a bounded domain Ω ⊆ Rd:

  • ∆u = λu,
  • n Ω,

u = 0,

  • n ∂Ω,

where ∆ = −

  • ∂2

∂x2

1 + · · · + ∂2

∂x2

d

  • is the Laplace operator.
slide-3
SLIDE 3

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Eigenvalues of the Dirichlet Laplacian

The Dirichlet problem for a bounded domain Ω ⊆ Rd:

  • ∆u = λu,
  • n Ω,

u = 0,

  • n ∂Ω,

where ∆ = −

  • ∂2

∂x2

1 + · · · + ∂2

∂x2

d

  • is the Laplace operator.

The spectrum of this operator is 0 < λ1 ≤ λ2 ≤ λ3 ≤ . . .

slide-4
SLIDE 4

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Eigenvalues of the Dirichlet Laplacian

The Dirichlet problem for a bounded domain Ω ⊆ Rd:

  • ∆u = λu,
  • n Ω,

u = 0,

  • n ∂Ω,

where ∆ = −

  • ∂2

∂x2

1 + · · · + ∂2

∂x2

d

  • is the Laplace operator.

The spectrum of this operator is 0 < λ1 ≤ λ2 ≤ λ3 ≤ . . . The eigenvalue counting function is N(λ) = max{j .

. . λj ≤ λ}.

l1 l2 l3=l4 l5 l6 l N(l)

slide-5
SLIDE 5

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Weyl’s spectral asymptotic formula

Conjecture (Lorentz, 1910). One can hear the area of a drum: lim

λ→∞

2πN(λ) λ = A.

slide-6
SLIDE 6

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Weyl’s spectral asymptotic formula

Conjecture (Lorentz, 1910). One can hear the area of a drum: lim

λ→∞

2πN(λ) λ = A. Theorem (Weyl, 1911). The eigenvalue counting function N(λ) = max{j .

. . λj ≤ λ} satisfies

lim

λ→∞

N(λ) λd/2 = cd vold(Ω), cd = (2π)−dωd.

slide-7
SLIDE 7

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Weyl’s spectral asymptotic formula

Conjecture (Lorentz, 1910). One can hear the area of a drum: lim

λ→∞

2πN(λ) λ = A. Theorem (Weyl, 1911). The eigenvalue counting function N(λ) = max{j .

. . λj ≤ λ} satisfies

lim

λ→∞

N(λ) λd/2 = cd vold(Ω), cd = (2π)−dωd.

“My work always tried to unite the truth with the beautiful, but when I had to choose one or the other, I usually chose the beautiful.” — Hermann Weyl

slide-8
SLIDE 8

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Weyl’s spectral asymptotic formula

Conjecture (Lorentz, 1910). One can hear the area of a drum: lim

λ→∞

2πN(λ) λ = A. Theorem (Weyl, 1911). The eigenvalue counting function N(λ) = max{j .

. . λj ≤ λ} satisfies

lim

λ→∞

N(λ) λd/2 = cd vold(Ω), cd = (2π)−dωd.

“My work always tried to unite the truth with the beautiful, but when I had to choose one or the other, I usually chose the beautiful.” — Hermann Weyl “In these days the angel of topology and the devil of abstract algebra fight for the soul of each individual mathematical domain.” — Hermann Weyl

slide-9
SLIDE 9

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Weyl’s spectral asymptotic formula

Conjecture (Lorentz, 1910). One can hear the area of a drum: lim

λ→∞

2πN(λ) λ = A. Theorem (Weyl, 1911). Weyl’s asymptotic law: N(λ) = cdλd/2 vold(Ω) + o(λd/2), as λ → ∞.

“My work always tried to unite the truth with the beautiful, but when I had to choose one or the other, I usually chose the beautiful.” — Hermann Weyl “In these days the angel of topology and the devil of abstract algebra fight for the soul of each individual mathematical domain.” — Hermann Weyl

slide-10
SLIDE 10

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry Weyl asymptotics

Weyl’s spectral asymptotic formula

Conjecture (Lorentz, 1910). One can hear the area of a drum: lim

λ→∞

2πN(λ) λ = A. Theorem (Weyl, 1911). Weyl’s asymptotic law: N(λ) = cdλd/2 vold(Ω) + o(λd/2), as λ → ∞. Conjecture (Weyl, 1911). N(λ) = cdλd/2 vold(Ω) + cd−1λ(d−1)/2 vold-1(∂Ω) + o(λ(d−1)/2), as λ → ∞.

slide-11
SLIDE 11

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl Conjecture

Conjecture proved by Courant (1922) with bound O(λ(d−1)/2 log λ). Levitan (1952): O(λ(d−1)/2) if Ω is compact and closed. Duistermaat and Guillemin (1975): o(λ(d−1)/2) for nice Ω. M´ etivier (1977): extended to more general elliptic operators. Ivrii (1980): generalized to milder restrictions on Ω.

(set of periodic billiards has measure 0).

slide-12
SLIDE 12

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl Conjecture

Conjecture proved by Courant (1922) with bound O(λ(d−1)/2 log λ). Levitan (1952): O(λ(d−1)/2) if Ω is compact and closed. Duistermaat and Guillemin (1975): o(λ(d−1)/2) for nice Ω. M´ etivier (1977): extended to more general elliptic operators. Ivrii (1980): generalized to milder restrictions on Ω.

(set of periodic billiards has measure 0).

Mark Kac (1966): “Can one hear the shape of a drum?”

slide-13
SLIDE 13

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl Conjecture

Conjecture proved by Courant (1922) with bound O(λ(d−1)/2 log λ). Levitan (1952): O(λ(d−1)/2) if Ω is compact and closed. Duistermaat and Guillemin (1975): o(λ(d−1)/2) for nice Ω. M´ etivier (1977): extended to more general elliptic operators. Ivrii (1980): generalized to milder restrictions on Ω.

(set of periodic billiards has measure 0).

Mark Kac (1966): “Can one hear the shape of a drum?” Gordon, Webb, Wolport (1992): “No.”

1 1 2 2 3 3 1 1 2 2 3 3

slide-14
SLIDE 14

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl-Berry Conjecture

Michael Berry (1979): For non-smooth boundaries, change N(λ) = cdλd/2 vold(Ω) + cd−1λ(d−1)/2 vold-1(∂Ω) + o(λ(d−1)/2) to N(λ) = cdλd/2 vold(Ω) + cDλD/2HD(∂Ω) + o(λ(D−1)/2) where D is the Hausdorff dimension of the boundary.

Here, D is Hausdorff dimension (and need not be an integer). HD is the measure associated to dimension D.

slide-15
SLIDE 15

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl-Berry Conjecture

Michael Berry (1979): For non-smooth boundaries, change N(λ) = cdλd/2 vold(Ω) + cd−1λ(d−1)/2 vold-1(∂Ω) + o(λ(d−1)/2) to N(λ) = cdλd/2 vold(Ω) + cDλD/2HD(∂Ω) + o(λ(D−1)/2) where D is the Hausdorff dimension of the boundary. Brossard and Carmona (1986): “No. But maybe Minkowski dimension instead?”

slide-16
SLIDE 16

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl-Berry Conjecture

Michael Berry (1979): For non-smooth boundaries, change N(λ) = cdλd/2 vold(Ω) + cd−1λ(d−1)/2 vold-1(∂Ω) + o(λ(d−1)/2) to N(λ) = cdλd/2 vold(Ω) + cDλD/2HD(∂Ω) + o(λ(D−1)/2) where D is the Hausdorff dimension of the boundary. Brossard and Carmona (1986): “No. But maybe Minkowski dimension instead?” Lapidus and Pomerance (1993): “Yes, for subsets of R.” N(λ) = λ vol1(Ω) + cDMD(∂Ω)λD + o(λD), as λ → ∞, where D is Minkowski dimension and MD is Minkowski content.

slide-17
SLIDE 17

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

The Weyl-Berry Conjecture

Michael Berry (1979): For non-smooth boundaries, change N(λ) = cdλd/2 vold(Ω) + cd−1λ(d−1)/2 vold-1(∂Ω) + o(λ(d−1)/2) to N(λ) = cdλd/2 vold(Ω) + cDλD/2HD(∂Ω) + o(λ(D−1)/2) where D is the Hausdorff dimension of the boundary. Brossard and Carmona (1986): “No. But maybe Minkowski dimension instead?” Lapidus and Pomerance (1993): “Yes, for subsets of R.” N(λ) = λ vol1(Ω) + cDMD(∂Ω)λD + o(λD), as λ → ∞, where D is Minkowski dimension and MD is Minkowski content. Kigami and Lapidus (1993): similar results for ∆ on fractals.

slide-18
SLIDE 18

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

Can you hear the shape of a fractal drum?

Let I ⊆ R be a compact interval. Let Ω ⊆ I be an open set consisting of infinitely many open intervals.

slide-19
SLIDE 19

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

Can you hear the shape of a fractal drum?

Let I ⊆ R be a compact interval. Let Ω ⊆ I be an open set consisting of infinitely many open intervals. Can one hear D (the dimension of ∂Ω)? Lapidus and Pomerance (1993): “Yes.”

slide-20
SLIDE 20

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

Can you hear the shape of a fractal drum?

Let I ⊆ R be a compact interval. Let Ω ⊆ I be an open set consisting of infinitely many open intervals. Can one hear D (the dimension of ∂Ω)? Lapidus and Pomerance (1993): “Yes.” Can one hear if ∂Ω is Minkowski measurable? Lapidus and Maier (1995): “Iff ζ(s) has no zero on Re s = D.”

Here ζ(s) =

  • n=1

n−s is the Riemann zeta function.

slide-21
SLIDE 21

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

Can you hear the shape of a fractal drum?

Let A, B ⊆ Rd be a bounded open set. Let Ω ⊆ A consist of infinitely many images of B.

B A W

slide-22
SLIDE 22

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

Can you hear the shape of a fractal drum?

Let A, B ⊆ Rd be a bounded open set. Let Ω ⊆ A consist of infinitely many images of B. Can one hear D (the dimension of ∂Ω)? Lapidus and van Frankenhuijsen (2000): “Yes.” Can one hear if ∂Ω is Minkowski measurable? Now measurability can fail for more than one reason. Arithmetic properties of scaling ratios. Minkowski measurability of ∂B.

slide-23
SLIDE 23

Minkowski dimension and nonmeasurability of lattice-type fractals Motivation: inverse problems in spectral geometry The Weyl conjectures

Can you hear the shape of a fractal drum?

Let A, B ⊆ Rd be a bounded open set. Let Ω ⊆ A consist of infinitely many images of B. Can one hear D (the dimension of ∂Ω)? Lapidus and van Frankenhuijsen (2000): “Yes.” Can one hear if ∂Ω is Minkowski measurable? Now measurability can fail for more than one reason. Arithmetic properties of scaling ratios. Minkowski measurability of ∂B.

So there is some interest in having a robust collection of examples

  • f sets which are (or are not) Minkowski measurable.
slide-24
SLIDE 24

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Minkowski dimension is box-counting dimension

ε = 150km ε = 75km ε = 37.5km N(ε) = 22 N(ε) = 53 N(ε) = 130

N(ε) = #{boxes of size ε required to cover the set (coastline)}.

slide-25
SLIDE 25

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Minkowski dimension is box-counting dimension

ε = 150km ε = 75km ε = 37.5km N(ε) = 22 N(ε) = 53 N(ε) = 130

N(ε) = #{boxes of size ε required to cover the set (coastline)}. Idea: if C is the measure of the set, then (number of boxes) · (box size) = N(ε) · (ε)D ≈ C

slide-26
SLIDE 26

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Minkowski dimension is box-counting dimension

ε = 150km ε = 75km ε = 37.5km N(ε) = 22 N(ε) = 53 N(ε) = 130

N(ε) = #{boxes of size ε required to cover the set (coastline)}. Idea: if C is the measure of the set, then (number of boxes) · (box size) = N(ε) · (ε)D ≈ C = ⇒ N(ε) ≈ C 1

ε

D

slide-27
SLIDE 27

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

How to find D: box-counting

N(ε) ≈ C( 1

ε)D

slide-28
SLIDE 28

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

How to find D: box-counting

N(ε) ≈ C( 1

ε)D

= ⇒ log N(ε) ≈ log

  • C( 1

ε)D

log N(ε) ≈ log C + D log( 1

ε) log N(ε) log( 1 ε ) ≈ log C log( 1 ε ) + D

D = lim

ε→0

log N(ε) log( 1

ε)

slide-29
SLIDE 29

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

slide-30
SLIDE 30

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

Next(ε) = min{J .

. . A ⊆ J j=1 B(xj, ε), where (xj)j∈J ⊆ Rd}

Nint(ε) = min{J .

. . A ⊆ J j=1 B(xj, ε), where (xj)j∈J ⊆ A}

Npack(ε) = max{J .

. . B(xj, ε) ∩ B(xk, ε) = ∅, j = k, where (xj)j∈J ⊆ A}

Nnet(ε) = max{J .

. . |xj − xk| ≥ ε, where (xj)j∈J ⊆ A}

slide-31
SLIDE 31

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

Next(ε) = min{J .

. . A ⊆ J j=1 B(xj, ε), where (xj)j∈J ⊆ Rd}

Nint(ε) = min{J .

. . A ⊆ J j=1 B(xj, ε), where (xj)j∈J ⊆ A}

Npack(ε) = max{J .

. . B(xj, ε) ∩ B(xk, ε) = ∅, j = k, where (xj)j∈J ⊆ A}

Nnet(ε) = max{J .

. . |xj − xk| ≥ ε, where (xj)j∈J ⊆ A}

Nnet(2ε) = Npack(ε) ≤ 2nNext(ε) and Next(ε) ≤ Nint(ε) ≤ Nnet(ε).

slide-32
SLIDE 32

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

Minkowski’s formulation: D = n − lim

ε→0

log VA(ε) log ε . Here, VA(ε) = vold(A−ε) is an inner tube formula, and A−ε is the inner ε-parallel set of A: A−ε = {x ∈ A .

. . d(x, A) < ε}.

A A

slide-33
SLIDE 33

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

Minkowski’s formulation: D = n − lim

ε→0

log VA(ε) log ε . Here, VA(ε) = vold(A−ε) is an inner tube formula, and A−ε is the inner ε-parallel set of A: A−ε = {x ∈ A .

. . d(x, A) < ε}.

A-e

slide-34
SLIDE 34

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

Minkowski’s formulation: D = n − lim

ε→0

log VA(ε) log ε . Here, VA(ε) = vold(A−ε) is an inner tube formula, and A−ε is the inner ε-parallel set of A: A−ε = {x ∈ A .

. . d(x, A) < ε}.

A-e So you can find D if you can find VA(ε)

slide-35
SLIDE 35

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Box-counting

Alternative formulations

Box-counting formulation: D = lim

ε→0

log N(ε) log( 1

ε) .

Minkowski’s formulation: D = n − lim

ε→0

log VA(ε) log ε . Here, VA(ε) = vold(A−ε) is an inner tube formula, and A−ε is the inner ε-parallel set of A: A−ε = {x ∈ A .

. . d(x, A) < ε}.

PROBLEM: The limits may not exist.

slide-36
SLIDE 36

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski measurability

Definition. The set A ⊆ Rd is Minkowski measurable in dimension α iff Mα(A) = lim

ε→0+ VA(ε)ε−(d−α)

exists with 0 < Mα(A) < ∞. In this case, Mα(A) is the α-dimensional Minkowski content of A.

slide-37
SLIDE 37

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski measurability

Definition. The set A ⊆ Rd is Minkowski measurable in dimension α iff Mα(A) = lim

ε→0+ VA(ε)ε−(d−α)

exists with 0 < Mα(A) < ∞. In this case, Mα(A) is the α-dimensional Minkowski content of A. One also has the average Minkowski content Mα(A) := lim

δ→0+

1 | ln δ| 1

δ

εα−dVA(ε)dε ε which always exists for self-similar sets [Gatzouras].

slide-38
SLIDE 38

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski measurability

Definition. The set A ⊆ Rd is Minkowski measurable in dimension α iff Mα(A) = lim

ε→0+ VA(ε)ε−(d−α)

exists with 0 < Mα(A) < ∞. In this case, Mα(A) is the α-dimensional Minkowski content of A. Since Mα(A) =      ∞, α < D, MD(A), α = D, 0, D < α, let M(A) = MD(A).

slide-39
SLIDE 39

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Measuring with respect to different dimensions

a-dimensional measure dimension a

a=D ¥

a(A)

D(A)

slide-40
SLIDE 40

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski vs. Hausdorff

Hausdorff: Hα,ε(A) = inf{

  • j∈J

|Uj|α .

. . A ⊆

  • j∈J

Uj, |Uj| ≤ ε}

ε→0+

− − − − − → Hα(A)

slide-41
SLIDE 41

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski vs. Hausdorff

Hausdorff: Hα,ε(A) = inf{

  • j∈J

|Uj|α .

. . A ⊆

  • j∈J

Uj, |Uj| ≤ ε}

ε→0+

− − − − − → Hα(A) Minkowski: N(ε)εα = inf{

  • j∈J

|Uj|α .

. . A ⊆

  • j∈J

Uj, |Uj| = ε}

ε→0+

− − − − − →? Mα(A) Hausdorff allows countable covers by arbitrarily sized sets. Minkowski uses finite covers by uniformly sized sets.

(Compare to partitions used in Lebesgue vs Riemann integration.)

slide-42
SLIDE 42

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski vs. Hausdorff

Why Minkowski dimension instead of Hausdorff?

slide-43
SLIDE 43

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski vs. Hausdorff

Why Minkowski dimension instead of Hausdorff?

It is not as refined a notion as Hausdorff dimension!

slide-44
SLIDE 44

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski vs. Hausdorff

Hausdorff measure: σ-additive. Hausdorff dimension is sensitive to translations of connected components.

dim(¶W) = log32 dim(¶W) = 0

slide-45
SLIDE 45

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Minkowski vs. Hausdorff

Hausdorff measure: σ-additive. Hausdorff dimension is sensitive to translations of connected components.

dim(¶W) = log32 dim(¶W) = 0 dim(¶W) = log32 dim(¶W) = log32

Minkowski content: only finitely additive. Minkowski dimension is invariant under translations of connected components. = ⇒ dimM makes more sense for drums

slide-46
SLIDE 46

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

What VA(ε) looks like, for open subsets of R

  • Definition. A fractal string L is the complement of a fractal subset
  • f R. L is a bounded open subset of R, so

L = {Ln}∞

n=1,

where each Ln is an open interval with length denoted ℓn.

slide-47
SLIDE 47

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

What VA(ε) looks like, for open subsets of R

  • Definition. A fractal string L is the complement of a fractal subset
  • f R. L is a bounded open subset of R, so

L = {Ln}∞

n=1,

where each Ln is an open interval with length denoted ℓn. The Cantor string: CS = {1

3, 1 9, 1 9, 1 27, 1 27, 1 27, 1 27, 1 81, . . . } L1 L2 L3 L4 L5 L6 L7 L8

slide-48
SLIDE 48

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

What VA(ε) looks like, for open subsets of R

  • Definition. A fractal string L is the complement of a fractal subset
  • f R. L is a bounded open subset of R, so

L = {Ln}∞

n=1,

where each Ln is an open interval with length denoted ℓn. The Cantor string: CS = {1

3, 1 9, 1 9, 1 27, 1 27, 1 27, 1 27, 1 81, . . . }

For fractal strings: VL(ε) =

  • ℓn>2ε

2ε +

  • ℓn≤2ε

ℓn.

slide-49
SLIDE 49

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Using VL(ε) to see Minkowski measurability

1 1 1

V

(e) 6 18

6 18 54

VL(ε) =

  • ℓn>2ε

2ε +

  • ℓn≤2ε

ℓn

ε 2ε + 2ε + 2ε + l4 + l5 + l6 + l7 + l8 + l9 + ...

slide-50
SLIDE 50

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Using VL(ε) to see Minkowski measurability

1 1 1

V

(e) 6 18

6 18 54

VL(ε) =

  • ℓn>2ε

2ε +

  • ℓn≤2ε

ℓn

ε 2ε + 2ε + 2ε + 2ε + l5 + l6 + l7 + l8 + l9 + ...

slide-51
SLIDE 51

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Using VL(ε) to see Minkowski measurability

1 1 1

V

(e)

e1-D

6 18

6 18 54

M = lim

ε→0+ VL(ε)ε−(1−D)

slide-52
SLIDE 52

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Using VL(ε) to see Minkowski measurability

1 1 1

V

(e)

e1-D

6 18

6 18 54

M = lim

ε→0+ VL(ε)ε−(1−D)

1 1 1 6 18 54

* *

e

e1-D V

(e)

slide-53
SLIDE 53

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Using VL(ε) to see Minkowski measurability

1 1 1

V

(e)

e1-D

6 18

6 18 54

M = lim

ε→0+ VL(ε)ε−(1−D) * * 1

54 1 1 1 2 6 18

e1-D V

(e)

loge

slide-54
SLIDE 54

Minkowski dimension and nonmeasurability of lattice-type fractals Minkowski dimension Minkowski measurability

Using VL(ε) to see Minkowski measurability

1 1 1

V

(e)

e1-D

6 18

6 18 54

There are “geometric oscillations” in VL(ε) of order D. This means limε→0+ ε−(1−D)VL(ε) cannot exist. ε−(1−D)VL(ε) contains terms of the form cωεIm(ω)✐ = cωeIm(ω)✐ log ε.

slide-55
SLIDE 55

Minkowski dimension and nonmeasurability of lattice-type fractals Tube formulas of self-similar sets

How to compute the tube formula for a self-similar set?

Use inner tube formula for A∁ to find outer tube formula for A.

= ∪

slide-56
SLIDE 56

Minkowski dimension and nonmeasurability of lattice-type fractals Tube formulas of self-similar sets

How to compute the tube formula for a self-similar set?

Use inner tube formula for A∁ to find outer tube formula for A.

= ∪

(1) Obtain the components of A∁ from the IFS. (2) Determine compatibility conditions. (3) Compute V(ε) = vold((A∁)−ε) using complex dimensions.

slide-57
SLIDE 57

Minkowski dimension and nonmeasurability of lattice-type fractals Tube formulas of self-similar sets Self-similar sets

Self-similarity sets and IFSs

SG is invariant under Φj(x) = x

2 + qj, j = 1, 2, 3

where {q1, q2, q3} are the vertices of a triangle.

q1 F2(q2) = q2 q2 q3

slide-58
SLIDE 58

Minkowski dimension and nonmeasurability of lattice-type fractals Tube formulas of self-similar sets Self-similar sets

Self-similarity sets and IFSs

SG is invariant under Φj(x) = x

2 + qj, j = 1, 2, 3

where {q1, q2, q3} are the vertices of a triangle.

q1 F2(q2) = q2 q2 q3

  • Definition. A self-similar set F ⊆ C(X) is a fixed point

F = Φ(F) := J

j=1 Φj(F),

F = ∅. Each Φj should be contractive: |Φj(x) − Φj(y)| < |x − y|.

slide-59
SLIDE 59

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling

Initiating the tiling construction

For the construction to be possible, assume F satisfies the open set condition If O is a feasible open set for F, this means:

1 Φj(O) ∩ Φk(O) = ∅ for j = k, 2 Φj(O) ⊆ O for each j, 3 F ⊆ O,

slide-60
SLIDE 60

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling

Initiating the tiling construction

For the construction to be possible, assume F satisfies the open set condition , and int F = ∅. If O is a feasible open set for F, this means:

1 Φj(O) ∩ Φk(O) = ∅ for j = k, 2 Φj(O) ⊆ O for each j, 3 F ⊆ O, and 4 O Φ(O).

For a feasible open set O, construct a tiling of K := O.

slide-61
SLIDE 61

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-62
SLIDE 62

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-63
SLIDE 63

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-64
SLIDE 64

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-65
SLIDE 65

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-66
SLIDE 66

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-67
SLIDE 67

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-68
SLIDE 68

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-69
SLIDE 69

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-70
SLIDE 70

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Two tiling examples: Koch curve and Sierpinski gasket

The Koch tiling and the Sierpinski gasket tiling

Φ1(z) = ξz, Φ2(z) = (1 − ξ)(z − 1) + 1, for ξ = 1

2 + ✐ 2 √ 3 ∈ C.

Φ1(x) = x

2 + p1, Φ2(x) = x 2 + p2, Φ3(x) = x 2 + p3.

slide-71
SLIDE 71

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Definition of the tiling

Each tile is the image of a generator

  • Definition. The generators {Gq}Q

q=1 are the connected components

  • f O \ Φ(K).

Some examples may have multiple generators.

slide-72
SLIDE 72

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling Definition of the tiling

Each tile is the image of a generator

  • Definition. The generators {Gq}Q

q=1 are the connected components

  • f O \ Φ(K).
  • Definition. The self-similar tiling associated with Φ and O is

T = T (O) = {Φw(Gq) .

. . w ∈ W, q = 1, . . . , Q},

where W := ∞

k=0{1, . . . , N}k is all finite strings on {1, . . . , N}, and

Φw := Φw1 ◦Φw2 ◦. . .◦Φwn. Let T =

R∈T R denote the union of the tiles.

slide-73
SLIDE 73

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling The compatibility condition for generators

Choosing a good O (or K)

Theorem [Compatibility Theorem]: Let int F = ∅ satisfy OSC with feasible set O and associated tiling T (O). Then TFAE:

1 bd T = F. 2 bd Gq ⊆ F, for all q ∈ Q. 3 Fε ∩ K = T−ε for all ε ≥ 0. 4 Fε ∩ K∁ = Kε ∩ K∁ for all ε ≥ 0.

So for a given Φ and F, check that one of 1–4 is satisfied. Then 5–6 ensure the inner/outer decomposition:

slide-74
SLIDE 74

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling The compatibility condition for generators

Choosing a good O (or K)

Specific possibilities: (1) Choose K = [F] and O = int K. Feasible iff int Φj(K) ∩ Φk(K) = ∅ for j = k. (Tileset condition) In this case, int F = ∅ iff F is convex. (Nontriviality condition)

slide-75
SLIDE 75

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling The compatibility condition for generators

Choosing a good O (or K)

Specific possibilities: (1) Choose K = [F] and O = int K. Feasible iff int Φj(K) ∩ Φk(K) = ∅ for j = k. (Tileset condition) In this case, int F = ∅ iff F is convex. (Nontriviality condition) (2) Let U be the unbounded component of F∁. Choose K = U∁ (the envelope of F) and O = int K. For the envelope, one always has bd K ⊆ F ⊆ K ⊆ [F], and K is convex iff K = [F].

slide-76
SLIDE 76

Minkowski dimension and nonmeasurability of lattice-type fractals The canonical self-affine tiling The compatibility condition for generators

Choosing a good O (or K)

Specific possibilities: (1) Choose K = [F] and O = int K. Feasible iff int Φj(K) ∩ Φk(K) = ∅ for j = k. (Tileset condition) In this case, int F = ∅ iff F is convex. (Nontriviality condition) (2) Let U be the unbounded component of F∁. Choose K = U∁ (the envelope of F) and O = int K. For the envelope, one always has bd K ⊆ F ⊆ K ⊆ [F], and K is convex iff K = [F].

slide-77
SLIDE 77

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula

Encoding the data into a formula

We have seen A decomposition of the set to be measured (via tiling). Conds on {Φ1, . . . , ΦJ} that ensure tiling is possible. Conds on O (or K = O) that allow vold(Fε) = VT (ε) + vold(Kε).

slide-78
SLIDE 78

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula

Encoding the data into a formula

We have seen A decomposition of the set to be measured (via tiling). Conds on {Φ1, . . . , ΦJ} that ensure tiling is possible. Conds on O (or K = O) that allow vold(Fε) = VT (ε) + vold(Kε). Next: how to encode the resulting data from The scaling factors rw, where rw = rw1 . . . rwn is scaling ratio of Φw := Φw1 ◦. . .◦Φwn. The generator(s) G = O \ Φ(K).

slide-79
SLIDE 79

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

The scaling data of the tiling

  • Definition. Let rw = rw1rw2 . . . rwk be the scaling ratio of Φw.

The scaling zeta function is given by the scaling ratios of Φ via ζs(s) =

  • w∈W

rs

w =

1 1 − J

j=1 rs j

, for s ∈ C. ζs records the sizes (and multiplicities) of the tiles T = {Φw(G)}.

W :=

  • k=0

{1, . . . , J}k is all finite seqs of indices, and Φw := Φw1◦Φw2◦. . .◦Φwk.

slide-80
SLIDE 80

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

The scaling data of the tiling

  • Definition. Let rw = rw1rw2 . . . rwk be the scaling ratio of Φw.

The scaling zeta function is given by the scaling ratios of Φ via ζs(s) =

  • w∈W

rs

w =

1 1 − J

j=1 rs j

, for s ∈ C. ζs records the sizes (and multiplicities) of the tiles T = {Φw(G)}. Poles of ζs encode scaling information: Ds = {ω .

. . ω is a pole of ζs}.

slide-81
SLIDE 81

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

The scaling data of the tiling

  • Definition. Let rw = rw1rw2 . . . rwk be the scaling ratio of Φw.

The scaling zeta function is given by the scaling ratios of Φ via ζs(s) =

  • w∈W

rs

w =

1 1 − J

j=1 rs j

, for s ∈ C. ζs records the sizes (and multiplicities) of the tiles T = {Φw(G)}. Poles of ζs encode scaling information: Ds = {ω .

. . ω is a pole of ζs}.

The complex dimensions Ds generalize the Minkowski dimension. Theorem [Lapidus et al.] D ∈ Ds and D = inf{σ ∈ R .

. . ζs(σ) < ∞}.

slide-82
SLIDE 82

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Consider a self-similar set with mappings Φ1, Φ2 having scaling ratios r1 = 2−1 and r2 = 2−φ. D is the set of solutions of the transcendental equation 2−s + 2−φs = 1 (s ∈ D),

slide-83
SLIDE 83

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Consider a self-similar set with mappings Φ1, Φ2 having scaling ratios r1 = 2−1 and r2 = 2−φ. D is the set of solutions of the transcendental equation 2−s + 2−φs = 1 (s ∈ D), The fractal can be approximated by a sequence with scaling ratios 2 1, 3 2, 5 3, 8 5, 13 8 , 21 13, 34 21, 55 34, . . . − → φ. For any such approximation, r1 and r2 are integer powers of some common base r (F is lattice type) r1 = rk1, r2 = rk2, k1, k2 ∈ N

slide-84
SLIDE 84

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

1 50 p −1

f » 2/1

50 p −1

f » 3/2

1 50 p −1

f » 5/3

1 p −1

f » 8/5

1 p −1

f » 13/8

1 50 p −1

f » 21/13

1 100 p −1

f » 34/21

1 100 p −1

f » 55/34

1 100

D D D D D D D D 100 700 1 D

slide-85
SLIDE 85

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Lattice vs. Nonlattice

For self-similar strings, a dichotomy exists.

The lattice case: {log r1, . . . , log rJ} are rationally dependent. #{Re ω} is finite. ∃ a row of dimns on Re s = D. Re ω = D for infinitely many ω ∈ Ds. ∂L is not Minkowski measurable. Example: Cantor String CS.

slide-86
SLIDE 86

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Lattice vs. Nonlattice

For self-similar strings, a dichotomy exists.

The lattice case: {log r1, . . . , log rJ} are rationally dependent. #{Re ω} is finite. ∃ a row of dimns on Re s = D. Re ω = D for infinitely many ω ∈ Ds. ∂L is not Minkowski measurable. Example: Cantor String CS. The nonlattice case: Some log rj are rationally independent. #{Re ω} is infinite. Re ω dense in parts of [σl, D]. Re ω = D = ⇒ ω = D. ∂L is Minkowski measurable. Example: The golden string GS.

slide-87
SLIDE 87

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Conjecture: Lattice/Nonlattice Dichotomy

  • Theorem. A self-similar subset F ⊆ R is Minkowski measurable

⇐ ⇒ F is nonlattice. (Proved independently by Falconer and Lapidus.)

slide-88
SLIDE 88

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Conjecture: Lattice/Nonlattice Dichotomy

  • Theorem. A self-similar subset F ⊆ R is Minkowski measurable

⇐ ⇒ F is nonlattice. (Proved independently by Falconer and Lapidus.) Lapidus et. al conjectured that (under suitable hypotheses):

1 Lattice-type fractals are not Minkowski measurable. 2 Nonlattice-type fractals are Minkowski measurable.

slide-89
SLIDE 89

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Conjecture: Lattice/Nonlattice Dichotomy

  • Theorem. A self-similar subset F ⊆ R is Minkowski measurable

⇐ ⇒ F is nonlattice. (Proved independently by Falconer and Lapidus.) Lapidus et. al conjectured that (under suitable hypotheses):

1 Lattice-type fractals are not Minkowski measurable. 2 Nonlattice-type fractals are Minkowski measurable.

  • Theorem. A self-similar subset F ⊆ R is Minkowski measurable if

F is nonlattice. (Proved by Gatzouras.) The converse has kept us busy.

slide-90
SLIDE 90

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Converting “scales” to “sizes”

  • Definition. The inradius of A ⊆ Rd is the radius of the largest

metric ball contained in A. Equivalently, ρ(A) := inf{ε > 0 .

. . A−ε = A}.

For A = G, write g := ρ(G). The tile Φw(G) has inradius rwg.

slide-91
SLIDE 91

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The scaling data of the tiling

Converting “scales” to “sizes”

  • Definition. The inradius of A ⊆ Rd is the radius of the largest

metric ball contained in A. Equivalently, ρ(A) := inf{ε > 0 .

. . A−ε = A}.

For A = G, write g := ρ(G). The tile Φw(G) has inradius rwg. We now have all the scaling data we need. − → Next, we look at the geometry of the generator.

slide-92
SLIDE 92

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The geometric data of the generators

Monophase and pluriphase generators

  • Definition. G is monophase iff VG(ε) is polynomial on [0, g]:

VG(ε) =

d−1

  • k=0

κkεd−k, 0 ≤ ε ≤ g.

Motivation: Steiner’s Theorem. If A is convex, then vold(Aε) =

d−1

  • k=0

κkεd−k for ε ≥ 0. Here, κk = µk(A) vold-k(Bd−k), and µk are the intrinsic volumes.

slide-93
SLIDE 93

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The geometric data of the generators

Monophase and pluriphase generators

  • Definition. G is monophase iff VG(ε) is polynomial on [0, g]:

VG(ε) =

d−1

  • k=0

κkεd−k, 0 ≤ ε ≤ g.

  • Definition. G is pluriphase iff VG(ε) is polynomial on each

subinterval of [0 = a0, a1, a1, . . . , aM = g].

slide-94
SLIDE 94

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The geometric data of the generators

How to compute the tube formula

Suppose you have Φ satisfying OSC, with int F = ∅, and a feasible open set O satisfying the compatibility theorem.

slide-95
SLIDE 95

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The geometric data of the generators

How to compute the tube formula

Suppose you have Φ satisfying OSC, with int F = ∅, and a feasible open set O satisfying the compatibility theorem. Scaling data: ζs(s) =

  • w∈W

rs

w.

Generator geometry: VG(ε) =

d−1

  • k=0

κkεd−k. We compute VT (ε) = vold(T−ε), the inner tube formula for the tiling, in terms of these ingredients.

slide-96
SLIDE 96

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula The fractal tube formula

Fractal tube formula

Theorem [Pearse 2006], [Lapidus, Pearse, Winter 2011] Suppose T has a single monophase generator and ζs has only simple poles. Then for 0 ≤ ε ≤ g, VT (ε) =

  • ω∈Ds∪{0,1,...,d−1}

cωεd−ω, where cω =

  • res (ζs; ω) d

k=0 gω−k ω−k κk,

ω ∈ Ds, κkζs(k), ω = k ∈ {0, 1, . . . , d}. Recall Ds := {poles of ζs}.

slide-97
SLIDE 97

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Lattice type fractals are not Minkowski measurable

Theorem [Lapidus, Pearse, Winter 2013]. Suppose F is a fractal with one monophase generator, d − 1 < D < d, and satisfying some technical conditions. If F is lattice, then F is not Minkowski measurable.

slide-98
SLIDE 98

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Lattice type fractals are not Minkowski measurable

Theorem [Lapidus, Pearse, Winter 2013]. Suppose F is a fractal with one monophase generator, d − 1 < D < d, and satisfying some technical conditions. If F is lattice, then F is not Minkowski measurable. Proof uses the pointwise tube formula, with an estimate based on careful choice of a window (contour).

D Wn+1 Wn

slide-99
SLIDE 99

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

The screen and the window

Choose a vertical contour S = {S(t) + ✐t .

. . t ∈ R}

where S : R → R is Lipschitz. S is the screen. Define the window W = {z ∈ C .

. . Re z ≥ S(Im z)}.

 D W 

slide-100
SLIDE 100

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Fractal tube formula with error term

Theorem [Lapidus, Pearse, Winter 2013] Suppose T has a single monophase generator and ζs has only simple poles. Then for 0 ≤ ε ≤ g, VT (ε) =

  • ω∈Ds∩W

cωεd−ω +

  • k∈{0,1,...,d−1}∩W

ckεd−k + R(ε), where Ds := {poles of ζs} and cω = res (ζs; ω)

d

  • k=0

gω−k ω − kκk, ck = κkζs(k), as before,

slide-101
SLIDE 101

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Fractal tube formula with error term

Theorem [Lapidus, Pearse, Winter 2013] Suppose T has a single monophase generator and ζs has only simple poles. Then for 0 ≤ ε ≤ g, VT (ε) =

  • ω∈Ds∩W

cωεd−ω +

  • k∈{0,1,...,d−1}∩W

ckεd−k + R(ε), where Ds := {poles of ζs} and cω = res (ζs; ω)

d

  • k=0

gω−k ω − kκk, ck = κkζs(k), as before, but now also R(ε) = 1 2π✐

  • S

εd−sζs(s) d − s d−1

  • k=0

gs−k s − k(d − k)κk(G)

  • ds,

and R(ε) = O(εd−sup S), as ε → 0+.

slide-102
SLIDE 102

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Recent results

Theorem [Kombrink, Pearse, Winter 2014]. Suppose F is a fractal with pluriphase generators. If F is lattice, then F is not Minkowski measurable.

Pluriphase means generator has a piecewise polynomial inner tube formula.

slide-103
SLIDE 103

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Recent results

Theorem [Kombrink, Pearse, Winter 2014]. Suppose F is a fractal with pluriphase generators. If F is lattice, then F is not Minkowski measurable.

Pluriphase means generator has a piecewise polynomial inner tube formula.

Improvements on prior result: Compatibility (bd O ⊆ F) no longer required. Works for countably many generators. No longer require d − 1 < D < d. More elementary tools (“simpler” proof).

slide-104
SLIDE 104

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

More precisely ...

  • Theorem. F is the attractor of a lattice IFS satisfying OSC, and

D = dimM F. Also, O is a strongly feasible open set satisfying a projection condition, for which the generator is pluriphase. Strongly feasible means F ∩ O = ∅ (SOSC). Projection condition means SjO ⊆ π−1

F (SjF).

slide-105
SLIDE 105

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

More precisely ...

  • Theorem. F is the attractor of a lattice IFS satisfying OSC, and

D = dimM F. Also, O is a strongly feasible open set satisfying a projection condition, for which the generator is pluriphase.

1 If D = dim aff F, then F is Minkowski measurable.

In particular, this is true for D = d.

slide-106
SLIDE 106

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

More precisely ...

  • Theorem. F is the attractor of a lattice IFS satisfying OSC, and

D = dimM F. Also, O is a strongly feasible open set satisfying a projection condition, for which the generator is pluriphase.

1 If D = dim aff F, then F is Minkowski measurable.

In particular, this is true for D = d.

2 If D /

∈ N, then F is not Minkowski measurable.

slide-107
SLIDE 107

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

More precisely ...

  • Theorem. F is the attractor of a lattice IFS satisfying OSC, and

D = dimM F. Also, O is a strongly feasible open set satisfying a projection condition, for which the generator is pluriphase.

1 If D = dim aff F, then F is Minkowski measurable.

In particular, this is true for D = d.

2 If D /

∈ N, then F is not Minkowski measurable.

3 If D ∈ N, then F is not Minkowski measurable ⇐

⇒ certain algebraic relations involving the data of the pluriphase representation are satisfied. Lm(ε) =

  • logr

am ε

  • 1

28 1 14 1 7 2 7 4 7 4 7 3 28 3 14 3 7 6 7

  • 5
  • 4
  • 3
  • 2
  • 1

1 2

L2(e) L1(e) = a2 g = a1

1 2

slide-108
SLIDE 108

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

More precisely ...

  • Theorem. F is the attractor of a lattice IFS satisfying OSC, and

D = dimM F. Also, O is a strongly feasible open set satisfying a projection condition, for which the generator is pluriphase.

1 If D = dim aff F, then F is Minkowski measurable.

In particular, this is true for D = d.

2 If D /

∈ N, then F is not Minkowski measurable.

3 If D ∈ N, then F is not Minkowski measurable ⇐

⇒ certain algebraic relations involving the data of the pluriphase representation are satisfied. Proof uses renewal theory.

slide-109
SLIDE 109

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Renewal Theorem. Suppose p1, . . . , pJ > 0 with J

j=1 pj = 1.

Let z : R → R have a discrete set of discontinuities and satisfy |z(t)| ≤ c1e−c2|t|, for some 0 < c1, c2 < ∞.

slide-110
SLIDE 110

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Renewal Theorem. Suppose p1, . . . , pJ > 0 with J

j=1 pj = 1.

Let z : R → R have a discrete set of discontinuities and satisfy |z(t)| ≤ c1e−c2|t|, for some 0 < c1, c2 < ∞. Let y1, . . . , yn > 0, and let Z be the unique solution of Z(t) = z(t) + J

j=1 pjZ(t − yj).

← (Renewal Eqn)

slide-111
SLIDE 111

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Renewal Theorem. Suppose p1, . . . , pJ > 0 with J

j=1 pj = 1.

Let z : R → R have a discrete set of discontinuities and satisfy |z(t)| ≤ c1e−c2|t|, for some 0 < c1, c2 < ∞. Let y1, . . . , yn > 0, and let Z be the unique solution of Z(t) = z(t) + J

j=1 pjZ(t − yj).

← (Renewal Eqn) Then Z(t) has one of two asymptotic behaviors: (1) If {y1, . . . , yn} ⊆ hZ, then Z(t) ∼ h η

  • k∈Z

z(t − kh), t → ∞. (2) If there is no such h, then lim

t→∞ Z(t) = 1

η ∞

−∞

z(τ) dτ. Here, η = J

j=1 pjyj.

slide-112
SLIDE 112

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Tube formula as renewal process

Since s = D is the solution of rs

1 + · · · + rs n = 1, let pj = rD j .

V

(e)

rdV

(r-1e)

v(e)

VT (ε) =

J

  • j=1

rd

j VT (r−1 j

ε) + v(ε)

slide-113
SLIDE 113

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Tube formula as renewal process

Since s = D is the solution of rs

1 + · · · + rs n = 1, let pj = rD j .

V

(e)

rdV

(r-1e)

v(e)

εD−dVT (ε) =

J

  • j=1

rD

j VT (r−1 j

ε)

  • r−1

j

ε D−d + v(ε)εD−d

slide-114
SLIDE 114

Minkowski dimension and nonmeasurability of lattice-type fractals Computation of the tube formula Minkowski (non)measurability results

Tube formula as renewal process

Since s = D is the solution of rs

1 + · · · + rs n = 1, let pj = rD j .

V

(e)

rdV

(r-1e)

v(e)

εD−dVT (ε) =

J

  • j=1

rD

j VT (r−1 j

ε)

  • r−1

j

ε D−d + v(ε)εD−d Z(ε) =

J

  • j=1

pjZ(r−1

j

ε) + z(ε)

slide-115
SLIDE 115

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

The distance and tube zeta functions

For a bounded set A ⊆ Rd, define the distance zeta function ζA(ε, s) :=

dist(x, A)s−d dx, where Aε := {x ∈ Rd .

. . dist(x, A) ≤ ε} is the outer ε-neighbourhood

and dx is d-dim Lebesgue measure.

slide-116
SLIDE 116

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

The distance and tube zeta functions

For a bounded set A ⊆ Rd, define the distance zeta function ζA(ε, s) :=

dist(x, A)s−d dx, where Aε := {x ∈ Rd .

. . dist(x, A) ≤ ε} is the outer ε-neighbourhood

and dx is d-dim Lebesgue measure. The tube zeta function is the Mellin transform of ε → vold(Aε):

  • ζA(ε, s) :=

ε ts−d−1 vold(At) dt,

slide-117
SLIDE 117

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

The distance and tube zeta functions

For a bounded set A ⊆ Rd, define the distance zeta function ζA(ε, s) :=

dist(x, A)s−d dx, where Aε := {x ∈ Rd .

. . dist(x, A) ≤ ε} is the outer ε-neighbourhood

and dx is d-dim Lebesgue measure. The tube zeta function is the Mellin transform of ε → vold(Aε):

  • ζA(ε, s) :=

ε ts−d−1 vold(At) dt, Easy theorem: ζA(ε, s) = εs−d vold(Aε) + (d − s) ζA(ε, s)

slide-118
SLIDE 118

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

Fractal zeta functions: a different approach

ζA(ε, s) =

  • Aε dist(x, A)s−d dx does not coincide with ζT for any A.

(e.g., for A = T = Φw(Gq)). However, it appears to contain comparable information. Theorem (Lapidus, Radunovic, Zubrinic 2011). dimM A = absc(ζA).

slide-119
SLIDE 119

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

Fractal zeta functions: a different approach

ζA(ε, s) =

  • Aε dist(x, A)s−d dx does not coincide with ζT for any A.

(e.g., for A = T = Φw(Gq)). However, it appears to contain comparable information. Theorem (Lapidus, Radunovic, Zubrinic 2011). dimM A = absc(ζA). Theorem (Kombrink, Pearse, Rock (last weekend)). For self-similar fractals F ⊆ Rd, the structure theorem for the complex dimensions of ζF is the same as for the scaling zeta function.

slide-120
SLIDE 120

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

A little more detail ...

Let O ⊆ Rd be open with λd(O) < ∞ and let A ⊆ Rd (poss. unbdd). The relative distance zeta function of A with respect to O is ζA(ε, s; O) =

  • O

dist(x, A)s−d dx.

slide-121
SLIDE 121

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

A little more detail ...

Let O ⊆ Rd be open with λd(O) < ∞ and let A ⊆ Rd (poss. unbdd). The relative distance zeta function of A with respect to O is ζA(ε, s; O) =

  • O

dist(x, A)s−d dx. The relative tube zeta function of A with respect to O is

  • ζA(ε, s; O) =

ε ts−d−1 vold(At ∩ O) dt.

slide-122
SLIDE 122

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

A little more detail ...

  • Theorem. Let F ⊆ Rd be a self-similar set satisfying OSC and let

O be a strongly feasible open set satisfying the projection

  • condition. Then
  • ζF(ε, s; O) =
  • ζF(ε, s; Γ) + J

j=1 rs j

r−1

j

ε ε

ts−d−1 vold(Ft ∩ O) dt 1 − J

j=1 rs j

slide-123
SLIDE 123

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

A little more detail ...

  • Theorem. Let F ⊆ Rd be a self-similar set satisfying OSC and let

O be a strongly feasible open set satisfying the projection

  • condition. Then
  • ζF(ε, s; O) =
  • ζF(ε, s; Γ) + J

j=1 rs j

r−1

j

ε ε

ts−d−1 vold(Ft ∩ O) dt 1 − J

j=1 rs j

Furthermore,

  • ζF(ε, s) =

ζF(ε, s; O) + ζF(ε, s; Oc).

slide-124
SLIDE 124

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

A little more detail ...

  • Theorem. Let F ⊆ Rd be a self-similar set satisfying OSC and let

O be a strongly feasible open set satisfying the projection

  • condition. Then
  • ζF(ε, s; O) =
  • ζF(ε, s; Γ) + J

j=1 rs j

r−1

j

ε ε

ts−d−1 vold(Ft ∩ O) dt 1 − J

j=1 rs j

Furthermore,

  • ζF(ε, s) =

ζF(ε, s; O) + ζF(ε, s; Oc).

  • Corollary. The structure theorem for the complex dimensions of

the tube and distance zeta functions is the same as for the scaling zeta function.

slide-125
SLIDE 125

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions A new hope; a new beginning

A little more detail ...

  • Theorem. Let F ⊆ Rd be a self-similar set satisfying OSC and let

O be a strongly feasible open set satisfying the projection

  • condition. Then
  • ζF(ε, s; O) =
  • ζF(ε, s; Γ) + J

j=1 rs j

r−1

j

ε ε

ts−d−1 vold(Ft ∩ O) dt 1 − J

j=1 rs j

Furthermore,

  • ζF(ε, s) =

ζF(ε, s; O) + ζF(ε, s; Oc).

  • Corollary. The structure theorem for the complex dimensions of

the tube and distance zeta functions is the same as for the scaling zeta function.

  • Corollary. For F ⊆ R, the lattice/nonlattice dichotomy extends to

the distance and tube zeta functions.

slide-126
SLIDE 126

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The Steiner formula

The Steiner formula: a tube formula for convex bodies

Let A ∈ Kd be a convex body (nonempty, convex, compact set). vold(A + εBd) =

slide-127
SLIDE 127

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The Steiner formula

The Steiner formula: a tube formula for convex bodies

Let A ∈ Kd be a convex body (nonempty, convex, compact set). vold(A + εBd) = c0ε2 + . . .

slide-128
SLIDE 128

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The Steiner formula

The Steiner formula: a tube formula for convex bodies

Let A ∈ Kd be a convex body (nonempty, convex, compact set). vold(A + εBd) = c0ε2 + c1ε1 + . . .

slide-129
SLIDE 129

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The Steiner formula

The Steiner formula: a tube formula for convex bodies

Let A ∈ Kd be a convex body (nonempty, convex, compact set). vold(A + εBd) = c0ε2 + c1ε1 + c2ε0 = d

i=0 ci εd−i.

slide-130
SLIDE 130

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The Steiner formula

The coefficients of the Steiner formula describe curvature

The coefficients ci = d

i

  • Ci(A) are interesting:

Ci is the ith total curvature of A. C0 is the Euler characteristic C1 is mean width Cd−1 is surface area Cd is volume Ci(A) is defined via integral geometry in terms of average volumes

  • f projections to lower-dimensional subspaces.
slide-131
SLIDE 131

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The Steiner formula

The coefficients of the Steiner formula describe curvature

The coefficients ci = d

i

  • Ci(A) are interesting:

Ci is the ith total curvature of A. C0 is the Euler characteristic C1 is mean width Cd−1 is surface area Cd is volume Also, ci = ωd−iµi(A), where µi is the ith intrinsic volume of A and ωi = voli(Bi). Ci is homogeneous of degree i: Ci(λA) = λiCi(A), λ > 0.

slide-132
SLIDE 132

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

The coefficients Ci can be localized as measures

Ci is total i-dimensional curvature measure. Localization: Ci(A, ·) is a measure defined on B(Rd). For B ⊆ Rd, Ci(A, B) describes curvature of B ∩ ∂A.

slide-133
SLIDE 133

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

The coefficients Ci can be localized as measures

Ci is total i-dimensional curvature measure. Localization: Ci(A, ·) is a measure defined on B(Rd). For B ⊆ Rd, Ci(A, B) describes curvature of B ∩ ∂A. R = {x ∈ Aε .

. . p(A, x) ∈ B}. VA(ε, B) = d−1 i=0 Ci(A, B)εd−i = vol2(R).

slide-134
SLIDE 134

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

The coefficients Ci can be localized as measures

Ci is total i-dimensional curvature measure. Localization: Ci(A, ·) is a measure defined on B(Rd). For B ⊆ Rd, Ci(A, B) describes curvature of B ∩ ∂A. R = {x ∈ Aε .

. . p(A, x) ∈ B}. VA(ε, B) = d−1 i=0 Ci(A, B)εd−i = vol2(R).

C1(A, B) > 0, but C0(A, B) = 0. (VA(ε, B) grows linearly in ε)

slide-135
SLIDE 135

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

The coefficients Ci can be localized as measures

Ci is total i-dimensional curvature measure. Localization: Ci(A, ·) is a measure defined on B(Rd). For B ⊆ Rd, Ci(A, B) describes curvature of B ∩ ∂A.

R′ = {x ∈ Aε .

. . p(A, x) ∈ B′}. VA(ε, B′) = d−1 i=0 Ci(A, B′)εd−i = vol2(R′).

Now C1(A, B′) > 0 and C0(A, B′) > 0.

slide-136
SLIDE 136

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

Additive extension of the Steiner formula

A is polyconvex iff A is a finite union of convex sets. A ∈ U(Kd) iff A = N

n=1[An],

slide-137
SLIDE 137

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

Additive extension of the Steiner formula

A is polyconvex iff A is a finite union of convex sets. A ∈ U(Kd) iff A = N

n=1[An],

The curvature measures are valuations. If A1, A2, A1 ∪ A2 ∈ Kd, then Ci(A1 ∪ A2, ·) + Ci(A1 ∩ A2, ·) = Ci(A1, ·) + Ci(A2, ·). Inclusion-exclusion principle can be used to define Ci(A, ·) for A ∈ U(Kd).

slide-138
SLIDE 138

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

Additive extension of the Steiner formula

Summary: If A ∈ U(Rd) is a polyconvex set, then∗ VA(ε, B) = d−1

i=0 Ci(A, B)εd−i.

More conceptually: VA(ε) =

  • i∈{0,...,d−1} ciεd−i,

where i = 0, . . . , d − 1 are “integer dimensions” of A, the coefficients ci are related to curvature.

slide-139
SLIDE 139

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

ζA may lend itself more easily to localization

For Σ := Rd × Sd−1, consider a Borel subset η ∈ B(Σ). For A convex, let p(A, ·) be metric projection to A. Let u(A, x) denote the unit vector pointing from p(A, x) to x. A S1

x p(A,x) u(A,x)

The normal bundle of A is Nor A := {(p(A, x), u(A, x)) .

. . x ∈ Rd}.

slide-140
SLIDE 140

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

A local parallel set is MA(ε, η) := {x ∈ Rd .

. . 0 < dist(x, A) ≤ ε and (p(A, x), u(A, x)) ∈ η}.

A A β η=β×σ MA(ε,η) σ

Taking η = β × Sd−1 recovers the earlier case.

slide-141
SLIDE 141

Minkowski dimension and nonmeasurability of lattice-type fractals Fractal zeta functions The curvature measures

A local parallel set is MA(ε, η) := {x ∈ Rd .

. . 0 < dist(x, A) ≤ ε and (p(A, x), u(A, x)) ∈ η}.

For a bounded set A ⊆ Rd, define ζA(ε, s; η) :=

  • MA(ε,η)

dist(x, A)s−d dx. This approach builds in localization from the beginning.