Beta-lattice multiresolution of quasicrystalline Bragg peaks

Transcription

1 Beta-lattice multiresolution of quasicrystalline Bragg peaks A. Elkharrat, J.P. Gazeau, and F. Dénoyer CRC701-Workshop Aspects of Aperiodic Order 3-5 July 2008, University of Bielefeld Proceedings of SPIE Vol. 6701, Wavelets XII, (Optics & Photonics, San Diego, August ) APC, CNRS UMR 7164, Université Paris Diderot, Paris Cedex 13, France

2 Some words about quasicrystals Quasicrystallography dates back to 1982 with the discovery, by Schechtman, Blech, Gratias, and Cahn a, of an alloy, namely Al 0.86 Mn 0.14, showing i) diffraction pattern like a constellation of more-or-less bright spots longdistance order, ii) spatial organisation of those Bragg peaks obeying five- or ten-fold symmetries, at least locally icosahedral organisation in real space with five-fold symmetries, iii) spatial organisation of those Bragg peaks obeying specific scale invariance, more precisely invariance under dilations by a factor equal to some power of the golden mean τ = 1+ 5 = 2 cos 2π a D. Shechtman, I. Blech, D. Gratias, and J.W. Cahn, prl 53(1984), 1951

3 Tsai et al. First Prev Next Last Go Back Full Screen Close Quit

4 Lifshitz First Prev Next Last Go Back Full Screen Close Quit

5 Bragg peaks invariance π/5 quasi- periodic density

6 A Bragg peak in R n is a pair Bragg peak in R n p = (x, I x ), where x is a point in R n, called the (geometrical) support of p, and I x > 0 is a positive real number called the intensity of p. The support of all Bragg peaks in a given pure-point diffraction spectrum is a point set Z R n (a lattice for crystals) with Lebesgue measure 0 in R n and the range of their intensities is a bounded subset I of the positive real line (constant for periodic crystals). So the corresponding pp diffraction spectrum Π is the set of Bragg peaks which reads as : Π = {p = (x, I x ) x Z, I x I}. Moreover, there exists a maximal intensity, I max I (0, I max ], lowest upper bound of I, and all points supporting it are symmetry center for Z with respect to space inversion. One of them will be chosen as the origin of R n.

7 Cyclotomic Pisot-Vijayaraghavan Pure-Point Diffraction Spectrum A k-fold cyclotomic Pisot Vijayaraghavan (PV) pure-point diffraction spectrum is a 2d pure-point diffraction spectrum which is supported by the (generically dense) point set of cyclotomic integers with ζ = e i 2π k Z[ζ] = k 1 q=0 Zζ q = Z[β] + Z[β]ζ, β is a cyclotomic Pisot-Vijayaraghavan number a, i.e. [ Z[β] = Z 2 cos 2π ] k such that as a scaling invariance factor it is relevant to the pure-point diffractive properties a A Pisot-Vijayaraghavan, or PV, number β > 1 is an algebraic integer, dominant root of the polynomial P (X) = X n a n 1 X n 1 a 1 X a 0, with a 0,, a n 1 Z, such that the other roots of P (X), are strictly smaller than 1. It is said to be a unit if a 0 = ±1.

8 Examples of Cyclotomic Pisot-Vijayaraghavan numbers They appear in quasi-crystallographic structures in the plane : k = 5 β = τ = = cos 2π 2 5 (pentagonal), k = 10 β = 1 + τ = τ 2 = = cos π 2 5 (decagonal), k = 10 β = τ 2 = = cos π 2 5 (decagonal), k = 8 β = δ = = cos π 4 (octogonal), k = 12 β = θ = = cos π 6 (dodecagonal). Note that in the case k = 7, we have β = cos 2π 7 cubic equation X 3 2X 2 X + 1 = 0. which is solution to the

9 The set of β-integers Every one is familiar with the binary system used to express real numbers as series in powers of 2 : x = ±(x j 2 j + x j 1 2 j x l 2 l + ), j = j(x) N : highest power of 2 such that 2 j x < 2 j+1, x j = integral part of x /2 j x /2 j {0, 1}. The same algorithm, called greedy algorithm, can be employed to represent real numbers in a system based on an arbitrary real number β > 1 : x = ±(x j β j +x j 1 β j 1 + +x l β l + ) ±x j x j 1... x 1 x 0 x 1 x 2... where the letters assume their values in the alphabet {0, 1,..., β 1} if β is natural, and β if β is not }. But if β is not natural, all words are not allowed. For instance, the first positive τ-integers are given by 1 1 τ τ 1 {}}{{}}{{}}{{}}{{}}{{}}{ 0 1 τ τ 2 τ τ 3 τ (0) (1) (10) (100) (101) (1000) (1001).

10 Beta-lattices Z β : set of beta-integers i.e. set of real numbers such that their beta-expansion is polynomial, Z β = {x R x = x k x 0 } Z β is relatively dense, self-similar and symmetrical with respect to the origin, If β is a Pisot-Vijayaraghavan number, then the set Z β is uniformly discrete and hence is Delaunay. Let (e i ) be a base of R n and β a a Pisot-Vijayaraghavan (and more generally a Parry number). The corresponding beta-lattice is the Delaunay set in R n : Γ = d Z β e i. i=1 This set is obviously self-similar and symmetrical with respect to the origin : βγ Γ, Γ = Γ.

11 A Tau-lattice in R 2 ζ = e iπ 5 Γ(τ) = Z τ + Z τ ζ z 2,3 = b 2 + b 3 ζ (2, 3) z m,n = b m + b n ζ

12 2D QC embedded in a Tau-lattice or Tau-lattice as a counting background for QC First Prev Next Last Go Back Full Screen Close Quit

13 Monochromatic x-ray diffraction pattern (Al63 Cu17.5 Co17.5 Si2, synchrotron LURE, Orsay) inplanted in a τ -lattice at a suitable scale. First Prev Next Last Go Back Full Screen Close Quit

14 On the use of beta-lattices as nested sequences - Proposition I Let β be a Pisot-Vijayaraghavan unit of degree d and Z β its corresponding set of β-integers. Let Γ = n Z i=1 βe i be the corresponding beta-lattice in R n, and Γ = n Z[β]e i=1 i the Z-module built from the basis e i and the algebraic β. Suppose that all integers have finite β-expansion. (ι) Suppose that there exists a finite l such that Z β + Z β Z β /β l. Let Γ j = Γ/β j be the scaled beta-lattices issued from Γ Γ 0. Then we have the infinite nested sequence of Delaunay sets and the inductive limit : Γ j 1 Γ j Γ j+1 and Γ = ind lim j Γ j. (ιι) Let F d be a finite set in Z[β] such that Z β + Z β Z β + F d. }{{} d times ) be the decorated (resp. scaled&decorated) beta-lattice defined by Γ F d := Γ + n i=1 F de i (resp. Γ F d kj = ΓF d /β kj ). Then we have the infinite nested sequence of Delaunay sets and the inductive limit : Suppose that there exists a finite k such that β k F d Z β. Let Γ F d (resp. ΓF d kj Γ F d k(j 1) ΓF d jk ΓF d k(j+1) and Γ = ind lim j Γ F d j.

15 On the use of beta-lattices as nested sequences - Proposition II Let β be a Pisot-Vijayaraghavan unit of degree d and Z β its corresponding set of β-integers. Let Γ = n Z i=1 βe i be the corresponding beta-lattice in R n, and Γ = n Z[β]e i=1 i the Z-module built from the basis e i and the algebraic β. Suppose that there exist integers which do not have finite β-expansion. There exists a finite set F N Z[β] such that all integers have finite β-expansion mod F N. Define the finite set G d = F N + F N β + + F N β d 1 Z[β] and the corresponding finite point set in R n, G d = n G i=1 de i. Let F d be a finite set in Z[β] such that Z β + Z }{{} β Z β + F d. Suppose there exists a finite k such d times that β k F d Z β. Let Γ F d and Γ F d jk be like in Prop. I. Then we have the infinite nested sequence of Delaunay sets and the inductive limit : Γ F d k(j 1) + G d Γ F d kj + G d Γ F d k(j+1) + G d and Γ = ind lim j Γ F d kj + G d.

16 Multiresolution analysis of Bragg spectra Here is a method of analysing and ) classifying Bragg spectra in R n given some (infinite) nested sequence ( Γj having the support Z of Bragg peaks as inductive limit : j Z Γ j 1 Γ j Γ j+1 and Z = ind lim j Γj. The supports of most quasi-crystalline Bragg peaks with k-fold symmetry are precisely the dense rings of cyclotomic integers built from ζ = e i 2π k (up to an irrelevant scale factor). Precisely, for β-lattices Γ = n i=1 Z βe i or some decorated versions Γ of them, where β is Pisot unit, we have Γ j 1 Γ j def = Γ/β j Γ j+1 and Γ = ind lim j Γj. with Γ = n i=1 Z[β]e i.

17 Multiresolution of Bragg Peaks - I Introduce the diffraction spectrum with cut-off X > 0 : The subset of Bragg peaks Π(X) Π is defined by : Π(X) = {p = (x, I x ) Π I x X}. Clearly the set-valued function X Π(X) is decreasing till reaching the empty set value : X Y Π(X) Π(Y ), and X > I max, Π(X) =.

18 Multiresolution of Bragg Peaks - II : Model Set Assumption In order to put in relation the nested ) continuously indexed family (Π(X)) X I and the nested discretely indexed sequence ( Γj we need the : j Z Assumption 1 For any X > 0 there exists j Z such that Π(X) is supported by Γ j Supp(Π(X)) Γ j. One next proceed to a discretization of the intensity interval I based on Assumption 2 To any j Z there corresponds the positive real X j defined by X j = min {X Supp(Π(X)) Γ } j. Concretely, if X X j, then Π(X) is supported by Γ j, whereas if X < X j, there exists p Π(X) such that x(p) / Γ j.

19 Multiresolution of Bragg Peaks - III : Multiresolving Filter and Partition Definition 1 Given the nested sequence ( Γj ) j Z such that Assumptions 1 and 2 hold true, two nested sequences of subsets of the spectrum are relevant : P i = Π(X i ), (1) G j = {p = (x, I x ) Π x Γ j } (2) Note that (2) depends more on the geometrical disposal of the spectrum whereas (1) is more centered on intensity features. Clearly, both increasing sequences have the pp diffraction spectrum Π as inductive limit : P i 1 P i P i+1 Π, G j 1 G j G j+1 Π.

20 Multiresolution of Bragg Peaks - IV : Multiresolving Filter and Partition Of course, if we make explicit one of the approximation sets in each sequence, say P i (resp. G j ), and if we wish to be more precise by dealing with the next approximation, P i+1 (resp. G j+1 ), it is not needed to reboot all the analysis : it is enough to add detail set Di P (resp. Dj G ) along the following partitions : P i+1 = P i D P i+1, D P i+1 := { p = (x, I x ) Π X i+1 I x < X i }, G j+1 = G j D G j+1, D G j+1 := { p = (x, I x ) Π x Γ j+1 \ Γ j }. In the language of wavelets and multiresolution analysis, the sets P i and G j are referred to as the tendency at scale i and j, respectively, and the sets Di+1 P and DG j+1 as the sets of details.

21 Multiresolution of Bragg Peaks - IV : The Fingerprint Hence we get the two partitions of the pp diffraction spectrum : Π = i= D P i, and Π = j= D G j. These two partitions are not identical and each one presents its own advantages in carrying out a classification of the Bragg peaks based on multiresolution. Obvious refinement is based on the intersection of the two partitions.

22 Multiresolution of Bragg Peaks - V : The Fingerprint Proposition 1 With the above definitions and supposing that Assumptions 1 and 2 are fulfilled, the most refined partition of the pp-diffraction spectrum Π reads as : Π = R i,j, where i,j= R i,j = D P i D G j = { p = (x(p), I x (p)) Π x(p) Γ j \ Γ j 1 and X i I x (p) < X i 1 } The table of values R i,j is the fingerprint.

23 (Academic) example : Fingerprint of Tau-Lattice Diffraction Pattern - I

24 (Academic) example : Fingerprint of Tau-Lattice Diffraction Pattern - II ( R ij ) j\i X 0 = 1 1 X 1 = X 2 = X 3 = X 4 = X 5 = X 6 = X 7 = First Prev Next Last Go Back Full Screen Close Quit

25 (Academic) example : Fingerprint of Penrose Diffraction Pattern - I

26 (Academic) example : Fingerprint of Penrose Diffraction Pattern - II ( R ij ) j\i X 0 = X 5 = X 6 = X 7 = X 8 = X 9 =

27 Steurer Kao model Diffraction pattern of the Steurer and Kuo model for the planel = 0. On the left-hand side : in the physical space together with the grid Γ 0 ; on the right-hand side : in the internal space together with the envelope τ 8 R τ.

28 Steurer Kao model plane l = 1

29 Steurer Kao model plane l = 2. -3

30 Steurer Kao model plane l = 3

31 0.1. plane l = plane l = 1 X i j X 6 = X 7 = X 9 = X i j X 5 = X 6 = X 7 = X 8 = X 9 =

32 0.3. plane l = plane l = 3 X i j X 6 = X 7 = X 8 = X 9 = X i j X 5 = X 6 = X 7 = X 8 = X 9 =

33 Practical use of this approach Equipped with these nested sequences of inflated/deflated versions of betaintegers and beta-lattices, we become able to build catalogues of diffraction patterns due to standard or manageable well-identified site occupation distributions, and then implement our multiresolution analysis of each of them. The issue will be a catalogue of diffraction patterns. Having this catalogue at our disposal, one can then examine experimental diffraction patterns, those ones displayed by real quasicrystals or other more or less exotic structures. The next step will be a (patient!) labor of (statistical) comparison, classification, and identification.