Recent Progress in Mathematical Diffraction · Recent Progress in Mathematical Diffraction Uwe...
Transcript of Recent Progress in Mathematical Diffraction · Recent Progress in Mathematical Diffraction Uwe...
Recent Progress inMathematical Diffraction
Uwe Grimm
Department of Mathematics & StatisticsThe Open UniversityMilton Keynes, UK
(joint work with Michael Baake & others)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.1
MenueDiffraction theoryPP spectra
Poisson formulaModel sets
SC spectraThue–MorseCantorPlanar example
AC spectraBernoulliRudin–ShapiroBernoullisationRandom tilings
Dynamical versusdiffraction spectra
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.2
DiffractionOptical diffraction
L A S E R
Diffraction pattern
interference of scattered waves
structure analysis
X-ray, electron or neutron diffraction
information on order and symmetry
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.3
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 3
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 3
120o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 3
120o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 3
120o 120o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 4
90o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 4
90o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 4
90o 90o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 5
72o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 5
72o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 5
72o 72o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 5
72o 72o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 6
60o 60o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 8
45o 45o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Why is that true?
Consider two points of minimal distance in a 2d lattice
Order 8
45o 45o
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
Crystallographic restrictionA lattice in 2 or 3 dimensions can only have non-trivialrotational symmetry axes of order 2, 3, 4, or 6.
Same result holds for three dimensions, but moresymmetries can be realised in lattices of dimension ≥ 4.
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.4
DiffractionQuasicrystals 5-fold
3-fold
2-fold
from: D. Shechtman, I. Blech, D. Gratias and J.W. Cahn (1984), Metallic phase with long-range orientational order and no translational symmetry, Phys. Rev. Lett. 53, 1951–1953
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.5
Diffraction theory
Wiener diagram
g∗−−−→ g ∗ g
Fy
yF
g|.|2−−−→ |g|2
commutative for integrable function g (with g(x) := g(−x) )
Kinematic diffraction:diagonal map g 7→ |g|2
Mathematical diffraction theory:use path via autocorrelationfor translation bounded measures
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.6
Why measures?Measures
are natural mathematical objects to describedistributions (of scatterers or radiation) is space
provide a unified generalisation of continuous (density)and discrete (tiling) approaches
ensure that quantities are mathematically well-defined
Absolutely continuous measure µ (with density ) on Rd
µ(f) =
∫
Rd
f(x) dµ(x) =
∫
Rd
f(x) (x) dx
Pure point measure µ on Rd
µ(f) =(∑
i∈Iwi δxi
)(f) =
∑
i∈Iwi f(xi)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.7
Why measures?
Singular continuous measure
gives no weight to any single point
support is uncountable set of zero Lebesgue measure
‘trivial’ cases (e.g. measures on lines in the plane)
probability measureµ for middle-thirdsCantor set withcontinuousdistribution functionF (x) = µ([0, x])
0 11/3 2/3
1
1/2
1/4
3/4
x
F(x)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.8
Diffraction theory
Structure: translation bounded measure ω
assumed ‘amenable’
Autocorrelation: γ = γω = ω ⊛ ω := limR→∞
ω|R ∗ ω|Rvol(BR)
Diffraction: γ = γpp + γsc + γac (relative to Lebesgue measure)
pp: Bragg peaks
ac: diffuse scattering with density
sc: whatever remains ...
General setting: ω y γ = ω ⊛ ω y γ 6y ω
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.9
Pure point spectra: Periodic crystals
Point measures: δx , δS :=∑
x∈S δx
Poisson summation formula: δΓ = dens(Γ ) δΓ ∗
for lattice Γ , dual lattice Γ ∗
Periodic crystals: ω = µ ∗ δΓ (µ finite)
⊲ γ = dens(Γ ) (µ ∗ µ) ∗ δΓ⊲ γ =
(dens(Γ )
)2 ∣∣µ∣∣2 δΓ ∗
(pure point diffraction: Bragg peaks on dual lattice)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.10
Pure point spectra: Periodic example I
Periodic Dirac comb
ω = δZ + 12 δZ+ 1
2
+ 14 δZ+{ 1
4, 34}
This can be written as
ω = µ ∗ δZ with µ = δ0 + 14δ 1
4
+ 12δ 1
2
+ 14δ 3
4
We have
ω =(δ0 + 1
4 δ− 1
4
+ 12 δ− 1
2
+ 14 δ− 3
4
)∗ δZ = ω
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.11
Pure point spectra: Periodic example I
Using δx ∗ δy = δx+y, one obtains the autocorrelation
γ = ω ∗ ω =(
118 δ0 + 3
4 δ 1
4
+ 98 δ 1
2
+ 34 δ 3
4
)∗ δZ
The diffraction measure is
γ =(
118 + 9
8(−1)k + 32 cos(πk2 )
)∗ δZ
= 4 δ4Z + δ4Z+2 + 14 δ4Z+{1,3} .
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.11
Pure point spectra: Periodic example IITwo-dimensional crystal with lattice of periods Z
2 and twoscatterers at positions (0, 0) and (a, b), so
Λ = Z2 ∪(Z
2 + (a, b))
with Dirac comb
ω = δΛ = (δ(0,0) + δ(a,b)) ∗ δZ2
Autocorrelationγ = (δ(0,0) + δ(a,b)) ∗ (δ(0,0) + δ−(a,b)) ∗ δZ2
=(2 δ(0,0) + δ(a,b) + δ−(a,b)
)∗ δ
Z2
Diffractionγ =
(2 + 2 cos
((2π(ka+ ℓb)
))δZ2
=(2 cos
(π(ka+ ℓb)
))2δZ2 for k, ℓ ∈ Z
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.12
Pure point spectra: Periodic example II
Diffraction
γ =(2 + 2 cos
((2π(ka+ ℓb)
))δZ2
=(2 cos
(π(ka+ ℓb)
))2δZ2 for k, ℓ ∈ Z
Example with (a, b) = (13 ,
1√3)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.12
Pure point spectra: Aperiodic example
Aperiodic structure as a limit of periodic structures withgrowing periods:
Coloured point set Pi+1 obtained from Pi by changing colourat all positions n ≡ (2 · 4i − 1) mod 4i+1, with P0 = Z.
Consider Z = Pred ∪ Pblue and
ω = δPred=
∑
n∈Pred
δn
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.13
Pure point spectra: Aperiodic example
Autocorrelation γ =∑
m∈Za(m)δm with
a(m) =2
3
(1 − 1
2r+1
)
for m = (2ℓ+ 1)2r with r ≥ 0 and ℓ ∈ Z.
Diffraction γ =∑
k∈Z[ 12] I(k)δk with
I(k) =
{49 k ∈ Z
19·4r−1 k = (2ℓ+ 1)/2r with ℓ ∈ Z and r ≥ 1
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.13
Pure point spectra: Aperiodic example
Period doubling substitution
:r 7→ rb
b 7→ rr
Fixed point sequences under 2:
r|r 7−→ rb|rb 7−→ rbrr|rbrr 7−→ rbrrrbrb|rbrrrbrb7−→ rbrrrbrbrbrrrbrr|rbrrrbrbrbrrrbrr 7−→ . . .
Point setsPred =
⋃
i≥0
(2 · 4iZ + (4i − 1)
)∪ {−1}
Pblue =⋃
i≥0
(4iZ + (2 · 4i−1 − 1)
)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.13
Incommensurate composite structureFor irrational α > 0 and u ∈ R
2 consider
ω = δZ2 + δu+Γ = δ
Z2 + δu ∗ δΓ ,
where Γ = αZ × Z ⊂ R2
Example withu = (1
3 ,12) and
α = τ = 12(1 +
√5)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.14
Incommensurate composite structureAutocorrelation
γ = δZ2 +
1
αδΓ +
1
α(δu + δ−u) ∗ (λ⊗ δZ)
Diffraction
γ = δZ2 +
1
α2δΓ ∗ +
2
αcos(2πk2u2) (δ0 ⊗ δZ)
with the dual lattice Γ ∗ = ( 1αZ) × Z
Total intensity of Bragg peak at position (0, n) (n ∈ Z)
γ({(0, n)}
)= 1 +
1
α2+
2
αcos(2πnu2) ≥
(1 − 1
α
)2 ≥ 0
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.14
Incommensurate composite structureDiffraction for u = (1
3 ,12) and α = τ
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.14
The Fibonacci sequence
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.15
The Fibonacci sequence
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.15
The Fibonacci sequence
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.15
The Fibonacci sequence
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.15
The Fibonacci sequence
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.15
The Fibonacci sequence
Liber Abaci (1202) by Leonardo of Pisa (Fibonacci)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.15
The Fibonacci sequenceSubstitution rule and matrix
:ℓ 7→ ℓs
s 7→ ℓM =
(1 1
1 0
)
One-sided fixed point w = (w) by iteration of on w(0) = ℓ:
ℓ 7→ ℓs 7→ ℓsℓ 7→ ℓsℓℓs 7→ ℓsℓℓsℓsℓ 7→ . . . 7→ w(n) n→∞−−−−→ w
Fibonacci numbers:
|w(n)| = fn+2 with cardℓ(w(n)) = fn+1 and cards(w
(n)) = fn
where f0 = 0, f1 = 1 and fn+1 = fn + fn−1
Golden ratio:
limi→±∞
fi+1
fi=
1 ±√
5
2=
{τ
τ ′
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.16
The Fibonacci sequence
Recursion: w(n+1) = w(n)w(n−1)
Two-sided Fibonacci sequence
ℓ|ℓ 7−→ ℓs|ℓs 7−→ ℓsℓ|ℓsℓ7−→ ℓsℓℓs|ℓsℓℓs 7−→ ℓsℓℓsℓsℓ|ℓsℓℓsℓsℓ7−→ ℓsℓℓsℓsℓℓsℓℓs|ℓsℓℓsℓsℓℓsℓℓs 7−→ · · ·
limiting 2-cycle ⊲ two fixed points under 2
Geometric realisation:
s ℓ
ℓ ℓ s
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.17
Pure point spectra: Model sets
Fibonacci substitution: ℓ 7→ ℓs, s 7→ ℓ (λPF = τ = 1+√
52 )
Fibonacci point set: Λ ={x ∈ Z[τ ] | x′ ∈ (−1, τ − 1]
}
Pairs (x, x′) form lattice:
s ℓ s ℓ ℓ s ℓ ℓ s ℓ s ℓ
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.18
Pure point spectra: Model sets
Silver mean substitution: a 7→ aba, b 7→ a (λPF = 1 +√
2 )
Silver mean point set: Λ ={x ∈ Z[
√2 ] | x′ ∈ [−
√2
2 ,√
22 ]}
Pairs (x, x′) form lattice:
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.19
Pure point spectra: Model sets
CPS:
Rd π←−−− R
d × Rm
πint−−−−→ R
m
∪ ∪ ∪ dense
π(L)1−1←−−−− L −−−−→ π
int(L)
‖ ‖L
⋆−−−−−−−−−−−−−−−−−−−−→ L⋆
Model set: Λ = {x ∈ L | x⋆ ∈ W } (assumed regular)
with W ⊂ Rm compact, λ(∂W ) = 0
Diffraction: γ =∑
k∈L⊛ |A(k)|2 δk
with L⊛ = π(L∗) (Fourier module of Λ)
and amplitude A(k) = dens(Λ)vol(W ) 1W (−k⋆)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.20
Fibonacci diffraction
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.21
Example: Ammann–Beenker
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.22
Example: Ammann–BeenkerL = Z[ξ] L ∼ Z
4 ⊂ R2 × R
2 O: octagon
ξ = exp(2πi/8) φ(8) = 4 ⋆-map: ξ 7→ ξ3
ΛAB ={x ∈ Z1 + Zξ + Zξ2 + Zξ3 | x⋆ ∈ O
}
1
ξ
ξ2
ξ3
1⋆
ξ⋆
ξ2⋆
ξ3⋆
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.22
Example: Ammann–Beenker
physical space internal space
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.22
Example: Ammann–BeenkerDiffraction intensity:
Bragg peaks at positions k1 + k2ξ2 ∈ 1
2Z[ξ] ⊂ C
(where ξ = exp(2πi/8)) with intensities
I((k1, k2)
)=
1(4π2(k′2+k′1)(k
′2−k′1)
)2(
cos(k′2π) cos(λk′1π)
− cos(k′1π) cos(λk′2π) − k′1k′2
sin(k′2π) sin(λk′1π)
+k′2k′1
sin(k′1π) sin(λk′2π)
)2
with λ = 1 +√
2 and algebraic conjugation ′ :√
2 7→ −√
2
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.22
Example: Ammann–Beenker
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.22
Interlude: Homometry
Problem: distinct structures with identical autocorrelation
Example 1: δ6Z∗
5∑
j=0
cj δjj 0 1 2 3 4 5
cj 11 25 42 45 31 14
cj 10 21 39 46 35 17
same correlations up to order 5 (Grünbaum & Moore)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.23
Interlude: Homometry
Problem: distinct structures with identical autocorrelation
Example 1: δ6Z∗
5∑
j=0
cj δjj 0 1 2 3 4 5
cj 11 25 42 45 31 14
cj 10 21 39 46 35 17
same correlations up to order 5 (Grünbaum & Moore)
Example 2: homometric models sets with distinct windows
windows covariogram
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.23
Interlude: Homometry
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.23
Interlude: Homometry
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.23
Interlude: Homometry
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.23
Dirac combs on integer lattices
Dirac comb (on Zd): ω =
∑
n∈Zd
w(n) δn
Autocorrelation: γ =∑
m∈Zd
η(m) δm
Autocorrelation coefficients:
η(m) = limN→∞
1
(2N+1)d
∑
n∈[−N,N ]d
w(n)w(n−m)
Diffraction: γ = γpp + γsc + γac
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.24
Thue–Morse diffraction
Substitution: :1 7→ 11
1 7→ 11( 1 = −1 )
Iteration and fixed point:
1 7→ 11 7→ 1111 7→ 11111111 7→ . . . −→ v = (v) = v0v1v2v3 . . .
v2i = vi and v2i+1 = vi
⊲ recursion for autocorrelation coefficients:
η(2m) = η(m) and η(2m+1) = −12
(η(m) + η(m+1)
)
for all m ∈ Z, with η(0) = 1
⊲ diffraction is purely singular continuous
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.25
Thue–Morse diffractionAbsence of pure point part:
γ = µ ∗ δZ
with µ = γ∣∣[0,1)
and η(m) =
∫ 1
0e2πimy dµ(y)
(Herglotz-Bochner)
Wiener’s criterion: µpp = 0 ⇐⇒ Σ(N) = o(N)
where Σ(N) =N∑
m=−N
(η(m)
)2
Argument: Σ(4N) ≤ 32Σ(2N) (by recursion for η)
⊲ µ = µcont = µsc + µac
Define: F (x) = µ([0, x]
)for x ∈ [0, 1], where F = Fac + Fsc
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.25
Thue–Morse diffractionAbsence of absolutely continuous part:
Functional relation:dF(x2
)+dF
(x+1
2
)= dF (x)
dF(x2
)−dF
(x+1
2
)= − cos(πx) dF (x)
valid for Fac and Fsc separately (µac ⊥ µsc)
Define: ηac(m) =∫ 10 e2πimx dFac(x)
y same recursion as for η(m), but ηac(0) free
Riemann-Lebesgue lemma: limm→±∞ ηac(m) = 0⊲ ηac(0) = 0 ⊲ ηac(m) ≡ 0 ⊲ Fac = 0
(Fourier uniqueness thm)
Theorem: µ = µsc and γ is purely sc.
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.25
Thue–Morse diffraction
Exponential sum: gn(k) =2n−1∑ℓ=0
vℓ e−2πikℓ
Recursion: gn+1(k) =(1 − e−2πik2n)
gn(k)
with g0(k) = 1
(follows from v(n+1) = v(n)v(n))
Riesz product: γ =∏
n≥0
(1 − cos(2n+1πx)
)
(vague convergence)
⊲ approach can be generalised to other sequencesand higher-dimensional block substitutions
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.25
Thue–Morse measure
0 0.5 1
0
0.5
1
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.26
Comparison: Cantor measure
0 11/3 2/3
1
1/2
1/4
3/4
x
F(x)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.27
Thue–Morse measure
0 0.5 1
0
10
20
0 0.5 1
0
10
20
0 0.5 1
0
10
20
0 0.5 1
0
0.5
1
0 0.5 1
0
0.5
1
0 0.5 1
0
0.5
1
(k = 4) (k = 5) (k = 6)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.28
Thue–Morse scalingGrowth rate β(k) = limn→∞
log(fn(k))n log(2) (if it exists)
extinctions for all k = m2r with m ∈ Z and r ≥ 0
Lebesgue-almost all k ∈ R do not contribute
growth rate β(k) = log(3/2)log(3) ≈ 0.585 for k = m
3·2r for m ∈ Z
and not divisible by 3 and r ≥ 0
growth rate β(k) = log(5/4)2 log(2) ≈ 0.161 for k = m
5·2r for m ∈ Z
and not divisible by 5 and r ≥ 0
explicit results for k = pq·2r with q ≥ 3 odd, gcd(p, q) = 1
and r ≥ 0
there are values of k where β(k) does not exist
uncountably many values of k in the supporting setWorkshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.29
Period doubling
Block map: ψ : 11, 11 7→ a, 11, 11 7→ b
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.30
Period doubling
Block map: ψ : 11, 11 7→ a, 11, 11 7→ b
1111111111111111 1111111111111111
abaaabababaaabab abaaabababaaabaa
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.30
Period doubling
Block map: ψ : 11, 11 7→ a, 11, 11 7→ b
1111111111111111 1111111111111111
abaaabababaaabab abaaabababaaabaa
y period doubling: ′ :a 7→ ab
b 7→ aaXTM−−−→ XTM
ψ
yyψ (2:1)
Xpd′−−−→ Xpd
π
yyπ (a.e. 1:1)
Sol × 2−−−→ Sol
↑coincidence
⊲ model set(Dekking)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.30
Planar example
Block inflation:
Substitution on v ∈ {1, 1}Z2
(with m,n≥0, 0≤r,s≤2)
(v)3m+r,3n+s =
{vm,n, r ≡ s ≡ 0 mod 2,
vm,n, otherwise,
Autocorrelation coefficients η(m,n) satisfy
η(3m+ r, 3n+ s) =
min(1,r)∑
k=0
min(1,s)∑
ℓ=0
α(r,s)k,ℓ η(m+ k, n+ ℓ)
with α(r,s)k,ℓ ∈ {−2
9 , 0,19 ,
13 , 1}. In particular, η(3m, 3n) = η(m,n).
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.31
Planar exampleη(3m, 3n) = η(m,n)
η(3m, 3n+1) = −29η(m,n) + 1
3η(m,n+1)
η(3m, 3n+2) = 13η(m,n) − 2
9η(m,n+1)
η(3m+1, 3n) = −29η(m,n) + 1
3η(m+1, n)
η(3m+1, 3n+1) = −29
(η(m+1, n) + η(m,n+1)
)+ 1
9η(m+1, n+1)
η(3m+1, 3n+2) = −29
(η(m,n) + η(m+1, n+1)
)+ 1
9η(m+1, n)
η(3m+2, 3n) = 13η(m,n) − 2
9η(m+1, n)
η(3m+2, 3n+1) = −29
(η(m,n) + η(m+1, n+1)
)+ 1
9η(m,n+1)
η(3m+2, 3n+2) = 19η(m,n) − 2
9
(η(m+1, n) + η(m,n+1)
)
⊲ diffraction purely singular continuous
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.31
Planar exampleDistribution function and density
F (x, y) = xy∑
m,n∈Z
η(m,n) sinc(2πmx) sinc(2πny)
f (N)(x, y) =∂2F (N)
∂x ∂y(x, y) =
N−1∏
ℓ=0
ϑ(3ℓx, 3ℓy
)
with ϑ(x,y) = 1
9
(1+2 cos(2πx)+2 cos(2πy)−2 cos(2π(x+y))−2 cos(2π(x−y))
)2
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
0
0.2
0.4
0.6
0.8
0
0.2
0.4
0.6
0.8
1
0
0.2
0.4
0.6
0.8
1
0
50
100
150
0
0.2
0.4
0.6
0.8
0
0.2
0.4
0.6
0.8
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.31
Planar exampleA pure point diffractive factor:
Continuous mapping ψ : {1, 1}Z2 −→ {1, 1}Z
2
defined by
(ψw)m,n = wm,nwm+1,nwm,n+1wm+1,n+1
Induced inflation rule for factor (with a ∈ {1, 1})
1 1 a
1 1 1
a 7−→ 1 1 1
Fixed points are model sets (with 3-adic internal space)⊲ pure point diffractive⊲ Fourier module Z[13 ] × Z[13 ]
⊲ recovers point part of dynamical spectrum
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.31
Interlude: Pinwheel tiling
2
1
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.32
Interlude: Pinwheel tiling
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.32
Interlude: Pinwheel tiling
Autocorrelation is circularly symmetric,
γΛ = δ0 +∑
r∈D\{0}η(r)µr =
∑
r∈Dη(r)µr,
with µr the normalised uniform distribution on rS1 and µ0 = δ0
R.V. Moody, D. Postnikoff and N. Strungaru, Circular symmetry of pinwheel diffraction,Ann. H. Poincaré 7 (2006) 711–730
⊲(γΛ)pp
=(dens(Λ)
)2δ0 = δ0
⊲ diffraction intensity on rings (singular component)
⊲ also absolutely continuous component?
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.32
Interlude: Pinwheel tilingpinwheel radial intensity (numerical)
k
I(k)
0 1 2 3 4
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.32
Interlude: Pinwheel tilingpinwheel radial intensity (numerical) square lattice powder diffraction
k
I(k)
0 1 2 3 4(central intensity suppressed; relative scale chosen such peaks at k = 1 match)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.32
Interlude: Pinwheel tiling
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.32
AC spectra: Coin tossing sequenceSequence: i.i.d. random variables Wn ∈ {±1}
with probabilities p and 1−p
Entropy: H(p) = −p log(p) − (1−p) log(1−p)
Autocorrelation: γB =∑
m∈Z
ηB(m)δm with
ηB(m) := limN→∞
1
2N+1
N∑
n=−NWnWn+m
(a.s.)=
{1, m = 0
(2p−1)2, m 6= 0
(strong law of large numbers)
Diffraction: γB
(a.s.)= (2p− 1)2δ
Z+ 4p(1 − p)λ
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.33
Rudin–Shapiro sequence
Recursive definition:
w(−1) = −1, w(0) = 1, and
w(4n+ ℓ) =
{w(n), for ℓ ∈ {0, 1}(−1)n+ℓw(n), for ℓ ∈ {2, 3}
⊲ autocorrelation γRS = δ0⊲ diffraction γRS = λ
⊲ homometric with coin tossing for p = 12
but zero entropy!
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.34
BernoullisationSequence: S ∈ {±1}Z (assumed ergodic)
with Dirac comb ωS =∑
n∈ZSn δn
and autocorrelation γS
Bernoullisation: ω :=∑
n∈ZSnWn δn
(Wn ∈ {±1}
)
Autocorrelation: γ(a.s.)= (2p− 1)2 γS + 4p(1 − p) δ0
(strong law of large numbers)
Application: Rudin–Shapiro, with γS = γRS = δ0⊲ γ = δ0 independently of p
⊲ diffraction γ ≡ λ
⊲ homometric, irrespective of entropy
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.35
Diffraction of random tilingsPatch of a square rhombus random tiling,and diffraction from this finite patch:
Expectation: mixed spectrum with trivial Bragg peak at 0,and a mixture of singular and absolutely continuous parts
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.36
Dynamical spectrum(X,Z, µ) measure-theoretic dynamical systemHilbert space H = L2(X, µ) with product
〈g | h〉 =
∫
Hg(x)h(x) dµ(x)
Shift map S induces unitary operator U on HUf := f ◦ S (Uf)(x) = f(Sx) for x ∈ X
For f ∈ H, the map n 7→ 〈f | Unf〉 defines a positive definitefunction on Z. Spectral measure of f
〈f | Unf〉 =
∫ 1
0e2πint dσf (t) = σf (n)
Dynamical spectrum (as a set)⋃f∈L2(X,µ) supp(σf )
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.37
Dynamical versus diffraction spectrumThe crucial link (Baake, Lenz, van Enter 2013):Consider ξ : X −→ R with w 7→ w0.Then σξ = σ
Xwhich is the fundamental diffraction, which
means that the diffraction measure is γ = σX∗ δ
Z.
Proof:
σξ(n) = 〈ξ | Unξ〉 =
∫
X
ξ(x) ξ(Snx) dµ(x)
= limN→∞
1
2N + 1
N∑
m=−Nξ(Smw)ξ(Sm+nw) (unique ergodicity)
= limN→∞
1
2N + 1
N∑
m=−Nwmwm+n = η(n) = σ
X(n)
generalises to subshift factors encoded by continuousfunctions
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.38
Summary & outlookDiffraction as useful toolContinuous spectraaccessibleHomometry more difficultInsensitivity to entropyExtension tohigher dimensionsGeneralisation beyondlattice systemsRandomness withinteractionScalingRelation todynamicalspectrum
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.39
Recent review articlesM Baake & U Grimm, Mathematicaldiffraction of aperiodic structures,Chem. Soc. Rev. 41 (2012) 6821,arXiv:1205.3633
M Baake &U Grimm,
Kinematic diffractionfrom a mathematicalviewpoint,
Z. Krist. 226 (2011) 711,arXiv:1105.0095
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.40
For yet more see...
MB & UG
Aperiodic OrderVol. 1: A Mathematical Invitation
Cambridge University Press (2013)
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.41
ReferencesN. Wiener, The spectrum of an array and its application to the study of the translationproperties of a simple class of arithmetical functions. Part I: The spectrum of an array,J. Math. Massachusetts 6 (1927) 145–157
K. Mahler, — dto. — Part II: On the translation properties of a simple class ofarithmetical functions, J. Math. Massachusetts 6 (1927) 158–163
M. Keane, Generalized Morse sequences, Z. Wahrscheinlichkeitsth. Verw. Gebiete 10(1968) 335–353
S. Kakutani, Strictly ergodic symbolic dynamical systems, in: Proc. 6th BerkeleySymposium on Math. Statistics and Probability, eds. L.M. LeCam, J. Neyman andE.L. Scott (eds.), Univ. of California Press, Berkeley (1972), pp. 319–326
M. Queffélec, Substitution Dynamical Systems — Spectral Analysis, LNM 1294(Springer, Berlin, 1987)
M. Baake & U. Grimm, The singular continuous diffraction measure of the Thue–Morsechain, J. Phys. A.: Math. Theor. 41 (2008) 422001; arXiv:0809.0580 (math.DS)
H.S. Shapiro, Extremal Problems for Polynomials and Power Series, Masters Thesis(MIT, Boston, 1951)
W. Rudin, Some theorems on Fourier coefficients, Proc. AMS 10 (1959) 855–859
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.42
ReferencesF.M. Dekking, The spectrum of dynamical systems arising from substitutions ofconstant length, Z. Wahrscheinlichkeitsth. verw. Geb. 41 (1978) 221–239
F.A. Grünbaum & C.C. Moore, The use of higher-order invariants in the determinationof generalized Patterson cyclotomic sets, Acta Cryst. A 51 (1995) 310–323
M. Höffe & M. Baake, Surprises in diffuse scattering, Z. Krist. 215 (2000) 441–444;arXiv:math-ph/0004022
M. Baake & U. Grimm, Homometric model sets and window covariograms, Z. Krist.222 (2007) 54–58; arXiv:math.MG/0610411
U. Grimm & M. Baake, Homometric point sets and inverse problems, Z. Krist. 223(2008) 777–781; arXiv:0808.0094
M. Baake & U. Grimm, Kinematic diffraction is insufficient to distinguish order fromdisorder, Phys. Rev. B 79 (2009) 020203(R); arXiv:0810.5750
X. Deng & R.V. Moody, How model sets can be determined by their two-point andthree-point correlations, J. Stat. Phys. 135 (2009) 621–637; arXiv:0901.4381
M. Baake, M. Birkner & R.V. Moody, Diffraction of stochastic point sets: Explicitlycomputable examples, Commun. Math. Phys. 293 (2010) 611–660; arXiv:0803.1266
M. Baake & T. Ward, Planar dynamical systems with pure Lebesgue diffractionspectrum, J. Stat. Phys. 140 (2010) 90–102; arXiv:1003.1536
Workshop on time-frequency analysis and aperiodic order, Trondheim, June 2015 – p.43