Discrete Tomography of Icosahedral Model Sets

The discrete tomography of B-type and F-type icosahedral model sets is investigated, with an emphasis on reconstruction and uniqueness problems. These are motivated by the request of materials science for the unique reconstruction of quasicrystalline…

Authors: Christian Huck

Discrete Tomography of Icosahedral Model Sets
DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS CHRISTIAN HUCK Abstra t. The disrete tomograph y of B-t yp e and F-t yp e iosahedral mo del sets is in v es- tigated, with an emphasis on reonstrution and uniqueness problems. These are motiv ated b y the request of materials siene for the unique reonstrution of quasirystalline stru- tures from a small n um b er of images pro dued b y quan titativ e high resolution transmission eletron mirosop y . 1. Intr odution Disr ete tomo gr aphy (the w ord tomograph y is deriv ed from the Greek τ oµoσ , meaning a slie) is onerned with the in v erse problem of retrieving information ab out some nite ob jet from (generally noisy) information ab out its slies. A t ypial example is the r e  onstrution of a nite p oin t set in Eulidean 3 -spae from its line sums in a small n um b er of diretions. More preisely , a ( disr ete p ar al lel ) X-r ay of a nite subset of Eulidean d -spae R d in diretion u giv es the n um b er of p oin ts of the set on ea h line in R d parallel to u . This onept should not b e onfused with X -ra ys in diration theory , whi h pro vide rather dieren t information on the underlying struture that is based on statistial pair orrelations; ompare [10℄, [12 ℄ and [19 ℄. In the lassial setting, motiv ated b y rystals, the p ositions to b e determined form a subset of a ommon translate of the ubi lattie Z 3 or, more generally , of an arbitrary lattie L in R 3 . In fat, man y of the problems in disrete tomograph y ha v e b een studied on Z 2 , the lassial planar setting of disrete tomograph y; see [ 21 ℄, [17 ℄ and [16 ℄. Bey ond the ase of p erfet rystals, one has to tak e in to aoun t wider lasses of sets, or at least signian t deviations from the lattie struture. As an in termediate step b et w een p erio di and random (or amorphous) Delone sets, w e onsider systems of ap erio di or der , more preisely , so-alled mo del sets (or mathemati al quasirystals ), whi h are ommonly regarded as go o d mathematial mo dels for quasirystalline strutures in nature [38 ℄. Our in terest in the disrete tomograph y of mo del sets is mainly motiv ated b y the task of struture determination of quasirystals, a new t yp e of solids diso v ered 25 y ears ago; see [ 33 ℄ for the pioneering pap er and [37 , 25 , 11℄ for ba kground and appliations. More preisely , w e address the problem of uniquely reonstruting three-dimensional quasirystals from their images under quan titativ e high r esolution tr ansmission ele tr on mir os opy (HR TEM) in a small n um b er of diretions. In fat, in [26℄ and [36℄ a te hnique is desrib ed, based on HR TEM, whi h an eetiv ely measure the n um b er of atoms lying on lines parallel to ertain diretions; it is alled QUANTITEM ( QU an titativ e AN alysis of T he I nformation from T ransmission E letron M irosop y). A t presen t, the measuremen t of the n um b er of atoms lying on a line an only b e appro ximately a hiev ed for some rystals; f. [26 , 36℄. Ho w ev er, it is reasonable to exp et that future dev elopmen ts in te hnology will impro v e this situation. 1 2 C. HUCK In this text, w e onsider b oth B-typ e and F-typ e i osahe dr al mo del sets Λ in 3 -spae whi h an b e desrib ed in algebrai terms b y using the i osian ring ; f. [8 ℄, [27℄ and [29 ℄. Note that the terminology originates from the fat that the underlying Z -mo dules (to b e explained in Setion 3) of B-t yp e and F-t yp e iosahedral mo del sets an b e obtained as pro jetions of b o dy-en tred and fae-en tred h yp erubi latties in 6 -spae, resp etiv ely . The F-t yp e iosa- hedral phase is the most ommon among the iosahedral quasirystals. Belo w, w e nev ertheless dev elop the theory for b oth the B-t yp e (also alled I-t yp e) and the F-t yp e phase. W ell kno wn examples of iosahedral quasirystals inlude the aluminium allo ys AlMn and AlCuF e; f. [22 ℄ for further examples. In pratie, only X -ra ys in Λ -diretions, i.e. , diretions parallel to non-zero elemen ts of the dierene set Λ − Λ of Λ ( i.e. , the set of in terp oin t v etors of Λ ) are reasonable. This is due to the fat that X -ra ys in non- Λ -diretions are meaningless sine the resolution oming from su h X -ra ys w ould not b e go o d enough to allo w a quan titativ e analysis  neigh b ouring lines are not suien tly separated. In fat, in order to obtain appliable results, one ev en has to nd Λ -diretions that guaran tee HR TEM images of high resolution, i.e. , yield dense lines in the orresp onding quasirystal Λ . An y lattie L in R d an b e slied in to latties of dimension d − 1 . More generally , mo del sets ha v e a dimensional hierar h y , i.e. , an y mo del set in d dimensions an b e slied in to mo del sets of dimension d − 1 . In Prop osition 3.16, it is sho wn that generi (to b e explained in Setion 3) B-t yp e and F-t yp e iosahedral mo del sets an b e slied in to (planar) ylotomi mo del sets , whose disrete tomograph y w e ha v e studied earlier; f. [4 , 24 ℄ and [23℄. The latter observ ation will b e ruial, sine it enables us to use the results on the disrete tomograph y of ylotomi mo del sets, slie b y slie. Using the sliing of generi iosahedral mo del sets in to ylotomi mo del sets and the results from [4℄, it w as sho wn in [24℄ that the algorithmi problem of r e  onstruting nite subsets of a large lass of generi iosahedral mo del sets Λ ( i.e. , those with p olyhedral windo ws) giv en X -ra ys in two Λ -diretions an b e solv ed in p olynomial time in the real RAM-mo del of omputation (Theorem 4.3 ). Sine this r e  onstrution pr oblem an p ossess rather dieren t solutions, one is led to the in v estigation of the orresp onding uniqueness pr oblem , i.e. , the (unique) determination of nite subsets of a xed iosahedral mo del set Λ b y X -ra ys in a small n um b er of suitably presrib ed Λ -diretions. Here, a subset E of the set of all nite subsets of a xed iosahedral mo del set Λ is said to b e determine d b y the X -ra ys in a nite set U of diretions if dieren t sets F and F ′ in E annot ha v e the same X -ra ys in the diretions of U . Sine, as demonstrated in Prop osition 5.1, an y xed n um b er of X -ra ys in Λ -diretions is insuien t to determine the en tire lass of nite subsets of a xed iosahedral mo del set Λ , it is neessary to imp ose some restrition in order to obtain p ositiv e uniqueness results. In Prop osition 5.3 , it is sho wn that the nite subsets F of ardinalit y less than or equal to some k ∈ N of a xed iosahedral mo del set Λ are determined b y an y set of k + 1 X -ra ys in pairwise non-parallel Λ -diretions. Prop osition 5.6 then sho ws that, for ev ery R > 0 and an y xed iosahedral mo del set Λ , there are t w o non-parallel Λ -diretions su h that the set of b ounded subsets of Λ with diameter less than R is determined b y the X -ra ys in these diretions. F or our main result, w e restrit the set of nite subsets of a xed iosahedral mo del set Λ b y onsidering the lass of  onvex subsets of Λ . They are nite sets C ⊂ Λ whose on v ex h ulls on tain no new p oin ts of Λ , i.e. , nite sets C ⊂ Λ with C = con v ( C ) ∩ Λ . By using the DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 3 sliing of generi iosahedral mo del sets in to ylotomi mo del sets again, it is sho wn that there are four pairwise non-parallel Λ -diretions su h that the set of on v ex subsets of an y iosahedral mo del set Λ are determined b y their X -ra ys in these diretions (Theorem 5.12 ). In fat, it turns out that one an  ho ose four Λ -diretions whi h pro vide uniqueness and yield dense lines in iosahedral mo del sets, the latter making this result lo ok promising in view of real appliations (Example 5.14 and Remark 5.15). Finally , w e demonstrate that, in an appro ximativ e sense, this result holds in a far more general (and relev an t) situation, where one deals with a whole family of generi iosahedral mo del sets at the same time, rather than dealing with a single xed iosahedral mo del set. 2. Preliminaries and not a tion Natural n um b ers are alw a ys assumed to b e p ositiv e, i.e. , N = { 1 , 2 , 3 , . . . } . Throughout the text, w e use the on v en tion that the sym b ol ⊂ inludes equalit y . W e denote the norm in Eulidean d -spae R d b y k · k . The unit sphere in R d is denoted b y S d − 1 , i.e. , S d − 1 = { x ∈ R d | k x k = 1 } . Moreo v er, the elemen ts of S d − 1 are also alled dir e tions . Reall that a homothety h : R d → R d is giv en b y x 7→ λx + t , where λ ∈ R is p ositiv e and t ∈ R d . W e all a homothet y exp ansive if λ > 1 . If x ∈ R , then ⌊ x ⌋ denotes the greatest in teger less than or equal to x . F or r > 0 and x ∈ R d , B r ( x ) is the op en ball of radius r ab out x . F or a subset S ⊂ R d , k ∈ N and R > 0 , w e denote b y card( S ) , F ( S ) , F ≤ k ( S ) , D 0 su h that ev ery ball B r ( x ) with x ∈ R d on tains at most one p oin t of Λ . F urther, Λ is alled r elatively dense if there is a radius R > 0 su h that ev ery ball B R ( x ) with x ∈ R d on tains at least one p oin t of Λ . Remark 3.6. Let Λ b e an iosahedral mo del set with windo w W . Then, Λ is a Delone set in R 3 ( i.e. , Λ is b oth uniformly disrete and relativ ely dense) and is of nite lo  al  omplexity ( i.e. , Λ − Λ is losed and disrete). Note that Λ is of nite lo al omplexit y if and only if for ev ery r > 0 there are, up to translation, only nitely man y p oin t sets (alled p athes of diameter r ) of the form Λ ∩ B r ( x ) , where x ∈ R 3 ; f. [35, Prop osition 2.3℄. In fat, Λ is ev en a Meyer set , i.e. , Λ is a Delone set and Λ − Λ is uniformly disrete; ompare [ 27℄. F urther, Λ is an ap erio di mo del set, i.e. , Λ has no translational symmetries. Moreo v er, if Λ is r e gular , Λ is pur e p oint dir ative , i.e. , the F ourier transform of the auto orrelation densit y that arises b y plaing a delta p eak (p oin t mass) on ea h p oin t of Λ lo oks purely p oin t-lik e; f. [35 ℄. If Λ is generi, Λ is r ep etitive , i.e. , giv en an y pat h of radius r , there is a radius R > 0 su h that an y ball of radius R on tains at least one translate of this pat h; f. [35 ℄. If Λ is regular, the frequeny of rep etition of nite pat hes is w ell dened, i.e. , for an y pat h of radius r , the n um b er of o urrenes of translates of this pat h p er unit v olume in the ball B r (0) of radius r > 0 ab out the origin 0 approa hes a non-negativ e limit as r → ∞ ; f. [34 ℄. Moreo v er, if Λ is b oth generi 8 C. HUCK Figure 1. A few slies of a pat h of the iosahedral mo del set Λ B ico (left) and their . ⋆ -images inside the iosahedral windo w in the in ternal spae (righ t), b oth seen from the p ositiv e x -axis. and regular, and, if a suitable translate of the windo w W has full iosahedral symmetry ( i.e. , if a suitable translate of the windo w W is in v arian t under the ation of the group Y ⋆ h of order 120 , where Y ⋆ h := Y ⋆ ∪ ( − Y ⋆ ) and Y ⋆ is the group of rotations of order 60 generated b y the t w o matries that arise from the t w o matries in ( 3 ) b y applying the onjugation . ′ to ea h en try), then Λ has full iosahedral symmetry Y h := Y ∪ ( − Y ) in the sense of symmetries of LI-lasses, meaning that a disrete struture has a ertain symmetry if the original and the transformed struture are lo ally indistinguishable (LI) ( i.e. , up to translation, ev ery nite pat h in Λ also app ears in an y of the other elemen ts of its LI-lass and vi e versa ); see [3℄ for details. T ypial examples are balls and suitably orien ted v ersions of the iosahedron, the do deahedron, the rhom bi triaon tahedron (the latter also kno wn as Kepler's b o dy) and its dual, the iosido deahedron. Example 3.7. F or a generi regular iosahedral mo del set with full iosahedral symmetry Y h , onsider Λ B ico := Λ B ico (0 , s + W ) , where s := 10 − 3 (1 , 1 , 1) t and W is the regular iosahedron with v ertex set Y ⋆ h ( τ ′ , 0 , 1) t ; see Figure 1 for an illustration. 3.2. Cylotomi mo del sets as planar setions of iosahedral mo del sets. In this setion, w e shall demonstrate that b oth B-t yp e and F-t yp e iosahedral mo del sets Λ an b e niely slied in to ylotomi mo del sets with underlying Z -mo dule Z [ ζ 5 ] , where the slies are in tersetions of Λ with translates of the h yp erplane H ( τ , 0 , 1) in R 3 orthogonal to ( τ , 0 , 1) t . DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 9 F rom no w on, w e alw a ys let ζ 5 := e 2 π i/ 5 , as a sp ei  hoie of a primitiv e 5 th ro ot of unit y in C . Oasionally , w e iden tify C with R 2 . Remark 3.8. It is w ell kno wn that the 5 th ylotomi eld Q ( ζ 5 ) is an algebrai n um b er eld of degree 4 o v er Q . Moreo v er, the eld extension Q ( ζ 5 ) / Q is a Galois extension with Ab elian Galois group G ( Q ( ζ 5 ) / Q ) ≃ ( Z / 5 Z ) × , where a ( mo d 5) orresp onds to the automorphism giv en b y ζ 5 7→ ζ a 5 ; f. [39, Theorem 2.5℄. Note that, restrited to the quadrati eld Q ( τ ) , b oth the Galois automorphism of Q ( ζ 5 ) / Q that is giv en b y ζ 5 7→ ζ 3 5 and its omplex onjugate automorphism ( i.e. , the automorphism giv en b y ζ 5 7→ ζ 2 5 ) indue the unique non-trivial Galois automorphism . ′ of Q ( τ ) / Q (determined b y τ 7→ 1 − τ ). F urther, Z [ ζ 5 ] is the ring of in tegers in Q ( ζ 5 ) ; f. [39, Theorem 2.6℄. The ring Z [ ζ 5 ] also is a Z [ τ ] -mo dule of rank t w o. More preisely , one has the equalit y Z [ ζ 5 ] = Z [ τ ] ⊕ Z [ τ ] ζ 5 ; f. [4 , Lemma 1(a)℄. Sine ζ 3 5 is also a primitiv e 5 th ro ot of unit y in C , one further has the equalit y Z [ ζ 5 ] = Z [ ζ 3 5 ] = Z [ τ ] ⊕ Z [ τ ] ζ 3 5 . Denition 3.9. Cylotomi mo del sets with underlying Z -mo dule Z [ ζ 5 ] Λ cyc ( t, W ) arise from the ut and pro jet s heme (4 ) b y setting d := m := 2 , L := Z [ ζ 5 ] and letting the star map . ⋆ 5 : L → R 2 b e either giv en b y the non-trivial Galois automorphism of Q ( ζ 5 ) / Q , dened b y ζ 5 7→ ζ 3 5 , or its omplex onjugate automorphism. Remark 3.10. The star map . ⋆ 5 as dened in Denition 3.9 is a monomorphism of Ab elian groups. F urther, the image of the map ˜ . 5 : L → R 2 × R 2 , dened b y α 7→ ( α, α ⋆ 5 ) , is indeed a lattie in R 2 × R 2 . Finally , one an v erify that the image L ⋆ 5 is indeed a dense subset of R 2 . F or the general setting, w e refer the reader to [4, 24 , 23 ℄. By [24 , Lemma 1.84(a)℄ (see also [23 , Lemma 25(a)℄), for all ylotomi mo del sets Λ with underlying Z -mo dule Z [ ζ 5 ] , the set of Λ -diretions is preisely the set of Z [ ζ 5 ] -diretions. Example 3.11. F or illustrations of ylotomi mo del sets with underlying Z -mo dule Z [ ζ 5 ] , see Figure 2 on the left and Figure 3; f. Prop osition 3.16 and Example 3.17 b elo w. Lemma 3.12. F or L ∈ { Im( I ) , I 0 } , the fol lowing e quations hold: (a) L ∩ H ( τ , 0 , 1) = Z [ τ ](0 , 1 , 0) t ⊕ Z [ τ ] 1 2 ( − 1 , − τ ′ , τ ) t . (b) ( L ∩ H ( τ , 0 , 1) ) ⋆ = L ⋆ ∩ H ( τ ′ , 0 , 1) . Pr o of. P art (a) follo ws from Equations (1) and (2) together with the relations Im( I ) = 1 2 M B and I 0 = 1 2 M F . P art (b) follo ws from the iden tit y (( τ , 0 , 1) t ) ⋆ = ( τ ′ , 0 , 1) t .  Denition 3.13. W e denote b y Φ the R -linear isomorphism Φ : H ( τ , 0 , 1) → C , determined b y (0 , 1 , 0) t 7→ 1 and 1 2 ( − 1 , − τ ′ , τ ) t 7→ ζ 5 . F urther, Φ ⋆ will denote the R -linear isomorphism Φ ⋆ : H ( τ ′ , 0 , 1) → C , determined b y (0 , 1 , 0) t 7→ 1 and 1 2 ( − 1 , − τ , τ ′ ) t 7→ ζ 3 5 . Lemma 3.14. The maps Φ and Φ ⋆ ar e isometries of Eulide an ve tor sp a es, wher e H ( τ , 0 , 1) , H ( τ ′ , 0 , 1) and C ar e r e gar de d as two-dimensional Eulide an ve tor sp a es in the  anoni al way. Mor e over, identifying C with the xy -plane in R 3 , Φ and Φ ⋆ extend uniquely to dir e t rigid motions of R 3 , i.e., elements of the gr oup SO(3 , R ) . Pr o of. The rst assertion follo ws from the follo wing iden tities: w w r (0 , 1 , 0) t + s 1 2 ( − 1 , − τ ′ , τ ) t w w = | r + s ζ 5 | = p r 2 + s 2 − r sτ ′ , w w r (0 , 1 , 0) t + s 1 2 ( − 1 , − τ , τ ′ ) t w w = | r + s ζ 3 5 | = p r 2 + s 2 − r sτ . 10 C. HUCK Figure 2. The en tral slie of the pat h of Λ B ico from Figure 1 (left) and its . ⋆ -image inside the (mark ed) deagon ( s + W ) ∩ H ( τ ′ , 0 , 1) (righ t), b oth seen from p erp endiular viewp oin ts. The additional statemen t is immediate.  Lemma 3.15. L et L ∈ { Im( I ) , I 0 } . Via r estrition, the maps Φ and Φ ⋆ indu e isomorphisms of r ank two Z [ τ ] -mo dules: L ∩ H ( τ , 0 , 1) Φ − → Z [ ζ 5 ] , L ⋆ ∩ H ( τ ′ , 0 , 1) Φ ⋆ − → Z [ ζ 5 ] . Pr o of. This follo ws immediately from the denition of Φ and Φ ⋆ together with Lemma 3.12 and Remark 3.8.  Prop osition 3.16. L et Λ b e a generi i osahe dr al mo del set with underlying Z -mo dule L , say Λ = Λ ico ( t, W ) . Then, for every λ ∈ Λ , one has the identity Φ  ( Λ ∩ ( λ + H ( τ , 0 , 1) )) − λ  =  z ∈ Z [ ζ 5 ]   z ⋆ 5 ∈ W λ  , wher e . ⋆ 5 is the Galois automorphism of Q ( ζ 5 ) / Q , dene d by ζ 5 7→ ζ 3 5 and W λ := Φ ⋆  ( W ∩ (( λ − t ) ⋆ + H ( τ ′ , 0 , 1) )) − ( λ − t ) ⋆  . Thus, the sets of the form Φ  ( Λ ∩ ( λ + H ( τ , 0 , 1) )) − λ  , (5) wher e λ ∈ Λ , ar e ylotomi mo del sets with underlying Z -mo dule Z [ ζ 5 ] . Pr o of. First, onsider Φ( µ ) , where µ ∈ ( Λ ∩ ( λ + H ( τ , 0 , 1) )) − λ . It follo ws that µ ∈ L ∩ H ( τ , 0 , 1) and ( µ + ( λ − t )) ⋆ = µ ⋆ + ( λ − t ) ⋆ ∈ W . Lemma 3.15 implies that Φ( µ ) ∈ Z [ ζ 5 ] , sa y Φ( µ ) = α + β ζ 5 for suitable α, β ∈ Z [ τ ] . One has Φ( µ ) ⋆ 5 = α ′ + β ′ ζ 3 5 = Φ ⋆ ( µ ⋆ ) ∈ W λ . DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 11 Figure 3. Another t w o slies of the pat h of Λ B ico from Figure 1. Con v ersely , supp ose that z ∈ Z [ ζ 5 ] satises z ⋆ 5 ∈ W λ . Then, there are suitable α, β ∈ Z [ τ ] su h that z = α + β ζ 5 and, onsequen tly , z ⋆ 5 = α ′ + β ′ ζ 3 5 ∈ W λ . By denition of W λ , one has z ⋆ 5 = Φ ⋆ ( µ ) , where µ ∈ H ( τ ′ , 0 , 1) satises µ + ( λ − t ) ⋆ ∈ W . Clearly , there exist r , s ∈ R su h that µ = r (0 , 1 , 0) t + s 1 2 ( − 1 , − τ , τ ′ ) t , whene Φ ⋆ ( µ ) = r + sζ 3 5 . The linear indep endene of 1 and ζ 3 5 o v er R no w implies that r = α and s = β , so that µ ∈ L ⋆ . Moreo v er, one an v erify that one has µ − ⋆ ∈ ( Λ ∩ ( λ + H ( τ , 0 , 1) )) − λ and Φ( µ − ⋆ ) = α + β ζ 5 = z . This pro v es the laimed iden tit y . The assertion is no w immediate.  Example 3.17. F or an illustration of the on ten t of Prop osition 3.16 in ase of the iosahedral mo del set Λ B ico from Example 3.7 , see Figures 2 and 3. 3.3. The translation mo dule of iosahedral mo del sets. In order to shed some ligh t on the set of Λ -diretions of an iosahedral mo del set Λ with underlying Z -mo dule L , w e rst ha v e to establish a relation b et w een iosahedral mo del sets and their underlying Z -mo dules. W e denote b y m τ the Z [ τ ] -mo dule endomorphism of Q ( τ ) 3 , giv en b y m ultipliation b y τ , i.e. , α 7→ τ α . F urthermore, w e denote b y m τ ⋆ the Z [ τ ] -mo dule endomorphism of ( Q ( τ ) 3 ) ⋆ , giv en b y α ⋆ 7→ ( τ α ) ⋆ . Lemma 3.18. The map m τ ⋆ is  ontr ative with  ontr ation  onstant 1 /τ ∈ (0 , 1) , i.e. , the e quality k m τ ⋆ ( α ⋆ ) k = (1 /τ ) k α ⋆ k holds for al l α ∈ Q ( τ ) 3 . Pr o of. F or α ∈ Q ( τ ) 3 , observ e that k m τ ⋆ ( α ⋆ ) k = k ( τ α ) ⋆ k = k τ ′ α ⋆ k = (1 /τ ) k α ⋆ k .  Lemma 3.19. L et Λ b e an i osahe dr al mo del set with underlying Z -mo dule L , say Λ = Λ ico ( t, W ) . Then, for any F ∈ F ( L ) , ther e is an exp ansive homothety h : R 3 → R 3 suh that h ( F ) ⊂ Λ . Pr o of. F rom in t( W ) 6 = ∅ and the denseness of L ⋆ in R 3 , one gets the existene of a suitable α 0 ∈ L with α 0 ⋆ ∈ in t ( W ) . Consider the op en neigh b ourho o d V := in t( W ) − α 0 ⋆ of 0 in 12 C. HUCK R 3 . Sine the map m τ ⋆ is on trativ e b y Lemma 3.18 (in the sense whi h w as made preise in that lemma), the existene of a suitable k ∈ N is implied su h that ( m τ ⋆ ) k ( F ⋆ ) ⊂ V . Hene, one has { ( τ k α + α 0 ) ⋆ | α ∈ F } ⊂ in t( W ) ⊂ W and, further, h ( F ) ⊂ Λ , where h : R 3 → R 3 is the expansiv e homothet y giv en b y x 7→ τ k x + ( α 0 + t ) .  As an easy appliation of Lemma 3.19 , one obtains the follo wing result on the set of Λ - diretions for iosahedral mo del sets Λ . Prop osition 3.20. L et Λ b e an i osahe dr al mo del set with underlying Z -mo dule L . Then, the set of Λ -dir e tions is pr e isely the set of L -dir e tions. Pr o of. Sine one has Λ − Λ ⊂ L , ev ery Λ -diretion is an L -diretion. F or the on v erse, let u ∈ S 2 b e an L -diretion, sa y parallel to α ∈ L \ { 0 } . By Lemma 3.19, there is a homothet y h : R 3 → R 3 su h that h ( { 0 , α } ) ⊂ Λ . It follo ws that h ( α ) − h (0) ∈ ( Λ − Λ ) \ { 0 } . Sine h ( α ) − h (0) is parallel to α , the assertion follo ws.  4. Complexity In the pratie of quan titativ e HR TEM, the determination of the rotational orien tation of a quasirystalline prob e in an eletron mirosop e an rather easily b e a hiev ed in the diration mo de. This is due to the iosahedral symmetry of gen uine iosahedral quasirystals. Ho w ev er, the X -ra y images tak en in the high-resolution mo de do not allo w us to lo ate the examined sets. Therefore, as already p oin ted out in [4℄, in order to pro v e pratially relev an t and rigorous results, one has to deal with the non-anhor e d  ase of the whole lo  al indistinguishability lass (or LI-lass, for short) LI( Λ ) of a regular, generi iosahedral mo del set Λ , rather than dealing with the anhor e d  ase of a single xed iosahedral mo del set Λ ; reall Remark 3.6 for the equiv alene relation giv en b y lo al indistinguishabilit y and ompare also [18℄. Remark 4.1. In the rystallographi ase of a lattie L in R 3 , the LI-lass of L onsists of all translates of L in R 3 , i.e. , one has LI( L ) = { t + L | t ∈ R 3 } . In partiular, LI( L ) simply onsists of one translation lass. The en tire LI-lass LI( Λ ico ( t, W )) of a regular, generi iosahedral mo del set Λ ico ( t, W ) an b e sho wn to onsist of all generi iosahedral mo del sets of the form Λ ico ( t, s + W ) and all patterns obtained as limits of sequenes of generi iosahedral mo del sets of the form Λ ico ( t, s + W ) in the lo al top ology (L T). Here, t w o patterns are ε -lose if, after a translation b y a distane of at most ε , they agree on a ball of radius 1 /ε around the origin; see [3, 35 ℄. Ea h su h limit is then a subset of some Λ ico ( t, s + W ) , but s migh t not b e in a generi p osition. Note that the LI-lass LI( Λ ) of an iosahedral mo del set Λ on tains un ountably many (more preisely , 2 ℵ 0 ) translation lasses; f. [3℄ and referenes therein. In view of the ompliation desrib ed ab o v e, w e m ust mak e sure that w e deal with nite subsets of generi iosahedral mo del sets of the form Λ ico ( t, s + W ) , i.e. , subsets whose . ⋆ -image lies in the interior of the windo w. This restrition to the generi ase is the prop er analogue of the restrition to p erfe t latties and their translates in the rystallographi ase. Analogous to the lattie ase [15 , 16℄ and the ase of ylotomi mo del sets [ 4℄, the main algorithmi problems of the disrete tomograph y of iosahedral mo del sets lo ok as follo ws. DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 13 Denition 4.2 (Consisteny , Reonstrution, and Uniqueness Problem) . Let L = Im( I ) (resp., L = I 0 ), let W ⊂ R 3 b e a windo w and let u 1 , . . . , u m ∈ S 2 b e m ≥ 2 pairwise non- parallel L -diretions. The orresp onding onsisteny , reonstrution and uniqueness problems are dened as follo ws. Consisteny . Giv en funtions p u j : L 3 u j → N 0 , j ∈ { 1 , . . . , m } , whose supp orts are nite and satisfy supp( p u j ) ⊂ L L u j , deide whether there is a nite set F whi h is on tained in an elemen t of I B g ( W ) (resp., I F g ( W ) ) and satises X u j F = p u j , j ∈ { 1 , . . . , m } . Reonstr ution . Giv en funtions p u j : L 3 u j → N 0 , j ∈ { 1 , . . . , m } , whose supp orts are nite and satisfy supp( p u j ) ⊂ L L u j , deide whether there exists a nite subset F of an elemen t of I B g ( W ) (resp., I F g ( W ) ) that satises X u j F = p u j , j ∈ { 1 , . . . , m } , and, if so, onstrut one su h F . Uniqueness . Giv en a nite subset F of an elemen t of I B g ( W ) (resp., I F g ( W ) ), deide whether there is a dieren t nite set F ′ that is also a subset of an elemen t of I B g ( W ) (resp., I F g ( W ) ) and satises X u j F = X u j F ′ , j ∈ { 1 , . . . , m } . One has the follo wing tratabilit y result, whi h w as pro v ed for the ase of B-t yp e iosahedral mo del sets b y om bining the results from Setion 3.2 with those presen ted in [4℄; f. [24 , Theorem 3.33℄ for the details. The pro of for the F-t yp e ase is similar and w e prefer to omit the straigh tforw ard details here. Belo w, for L ∈ { Im( I ) , I 0 } , the L -diretions in S 2 ∩ H ( τ , 0 , 1) will b e alled L ( τ , 0 , 1) -dir e tions . By Lemma 3.12 (a), the set of Im( I ) ( τ , 0 , 1) -diretions and the set of I ( τ , 0 , 1) 0 -diretions oinide. Theorem 4.3. L et L ∈ { Im( I ) , I 0 } . When r estrite d to two L ( τ , 0 , 1) -dir e tions and p olyhe- dr al windows, the pr oblems Consisteny , Reonstr ution and Uniqueness as dene d in Denition 4.2  an b e solve d in p olynomial time in the r e al RAM-mo del of  omputation. Remark 4.4. F or a detailed analysis of the omplexities of the ab o v e algorithmi problems in the B-t yp e ase, w e refer the reader to [ 24 , Chapter 3℄. Note that ev en in the an hored planar lattie ase Z 2 the orresp onding problems Consisteny , Reonstr ution and Uniqueness are NP -hard for three or more Z 2 -diretions; f. [15 , 16℄. 5. Uniqueness 5.1. Simple results on determination of nite subsets of iosahedral mo del sets. In this setion, w e presen t some uniqueness results whi h only deal with the anhor e d  ase of determining nite subsets of a xed iosahedral mo del set Λ b y X -ra ys in arbitr ary Λ - diretions; f. Prop osition 3.20. As already explained in Setion 1, X -ra ys in non- Λ -diretions are meaningless in pratie. Without the restrition to Λ -diretions, the nite subsets of a xed iosahedral mo del set Λ an b e determined b y one X -ra y . In fat, an y X -ra y in a non- Λ -diretion is suitable for this purp ose, sine an y line in 3 -spae in a non- Λ -diretion passes through at most one p oin t of Λ . The next result represen ts a fundamen tal soure of diulties 14 C. HUCK in disrete tomograph y . There exist sev eral v ersions; ompare [21, Theorem 4.3.1℄, [13, Lemma 2.3.2℄, [5 , Prop osition 4.3℄, [24 , Prop osition 2.3 and Remark 2.4℄ and [23 , Prop osition 8℄. Prop osition 5.1. L et Λ b e an i osahe dr al mo del set with underlying Z -mo dule L , say Λ = Λ ico ( t, W ) . F urther, let U ⊂ S 2 b e an arbitr ary, but xe d nite set of p airwise non-p ar al lel L -dir e tions. Then, F ( Λ ) is not determine d by the X -r ays in the dir e tions of U . Pr o of. W e argue b y indution on card( U ) . The ase card( U ) = 0 means U = ∅ and is ob vious. Supp ose the assertion to b e true whenev er card( U ) = k ∈ N 0 and let card( U ) = k + 1 . By indution h yp othesis, there are dieren t elemen ts F and F ′ of F ( Λ ) with the same X -ra ys in the diretions of U ′ , where U ′ ⊂ U satises card( U ′ ) = k . Let u b e the remaining diretion of U . Cho ose a non-zero elemen t α ∈ L parallel to u su h that α + ( F ∪ F ′ ) and F ∪ F ′ are disjoin t. Then, F ′′ := ( F ∪ ( α + F ′ )) − t and F ′′′ := ( F ′ ∪ ( α + F )) − t are dieren t elemen ts of F ( L ) with the same X -ra ys in the diretions of U . By Lemma 3.19 , there is a homothet y h : R 3 → R 3 su h that h ( F ′′ ∪ F ′′′ ) = h ( F ′′ ) ∪ h ( F ′′′ ) ⊂ Λ . It follo ws that h ( F ′′ ) and h ( F ′′′ ) are dieren t elemen ts of F ( Λ ) with the same X -ra ys in the diretions of U ; f. Lemma 2.6 .  Remark 5.2. An analysis of the pro of of Prop osition 5.1 sho ws that, for an y nite set U ⊂ S 2 of k pairwise non-parallel L -diretions, there are disjoin t elemen ts F and F ′ of F ( Λ ) with card( F ) = card( F ′ ) = 2 ( k − 1) and with the same X -ra ys in the diretions of U . Consider an y on v ex subset C of R 3 whi h on tains F and F ′ from ab o v e. Then, the subsets F 1 := ( C ∩ Λ ) \ F and F 2 := ( C ∩ Λ ) \ F ′ of F ( Λ ) also ha v e the same X -ra ys in the diretions of U . Whereas the p oin ts in F and F ′ are widely disp ersed o v er a region, those in F 1 and F 2 are on tiguous in a w a y similar to atoms in a quasirystal; ompare [ 15 , Remark 4.3.2℄ and [23 , Remark 2.4 and Figure 2.1℄ (see also [23 , Remark 32 and Figure 5℄). Originally , the pro of of the follo wing result is due to Rén yi; f. [32 ℄ and ompare [21 , Theorem 4.3.3℄. Prop osition 5.3. L et Λ b e an i osahe dr al mo del set with underlying Z -mo dule L . F urther, let U ⊂ S 2 b e any set of k + 1 p airwise non-p ar al lel L -dir e tions, wher e k ∈ N 0 . Then, F ≤ k ( Λ ) is determine d by the X -r ays in the dir e tions of U . Mor e over, for al l F ∈ F ≤ k ( Λ ) , one has G F U = F . Pr o of. Let F , F ′ ∈ F ≤ k ( Λ ) ha v e the same X -ra ys in the diretions of U . Then, one has card( F ) = card( F ′ ) b y Lemma 2.2 (a) and F , F ′ ⊂ G U F b y Lemma 2.4. But w e ha v e G U F = F sine the existene of a p oin t in G U F \ F implies the existene of at least card( U ) ≥ k + 1 p oin ts in F , a on tradition. It follo ws that F = F ′ .  Remark 5.4. In partiular, the additional statemen t of Prop osition 5.3 demonstrates that, for a xed iosahedral mo del set Λ with underlying Z -mo dule L , the unique reonstrution of sets F ∈ F ≤ k ( Λ ) from their X -ra ys in arbitrary sets of k + 1 pairwise non-parallel L -diretions U ⊂ S 2 merely amoun ts to ompute the grids G U F . Let Λ b e an iosahedral mo del set with underlying Z -mo dule L . Remark 5.2 and Prop osition 5.3 sho w that F ≤ k ( Λ ) an b e determined b y the X -ra ys in an y set of k + 1 pairwise non-parallel L -diretions but not b y 1 + ⌊ log 2 k ⌋ pairwise non-parallel X -ra ys in L -diretions. Ho w ev er, in pratie, one is in terested in the determination of nite sets b y X -ra ys in a small n um b er of diretions sine after ab out 3 to 5 images tak en b y HR TEM, the ob jet ma y b e damaged or ev en destro y ed b y the radiation DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 15 energy . Observing that the t ypial atomi strutures to b e determined omprise ab out 10 6 to 10 9 atoms, one realizes that the last result is not pratial at all. The follo wing result w as pro v ed in [ 24, Theorem 2.8(a)℄; see also [ 23 , Theorem 13(a)℄. Prop osition 5.5. L et d ≥ 2 , let R > 0 , and let Λ ⊂ R d b e a Delone set of nite lo  al  omplexity. Then, the set D 0 . Then, the set D 0 (whi h is rather natural in pratie), then Theo- rem 5.16 allo ws us to write DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 19 1 v ol( W ) Z s + W y d λ ( y ) ≈ 1 card ( F − t ) X x ∈ F − t x ⋆ = 1 card ( F ′ − t ) X x ∈ F ′ − t x ⋆ ≈ 1 v ol( W ) Z ( s ′ +( t ′ − t ) ⋆ )+ W y d λ ( y ) . Consequen tly , s + Z W y d λ ( y ) ≈ ( s ′ + ( t ′ − t ) ⋆ ) + Z W y d λ ( y ) , and hene s ≈ s ′ + ( t ′ − t ) ⋆ . The latter means that, appro ximately , b oth F − t and F ′ − t are elemen ts of the set C ( Λ B ico (0 , s + W )) . No w, it follo ws in this appro ximativ e sense from prop ert y (C) and Theorem 5.12 that F − t ≈ F ′ − t , and, nally , F ≈ F ′ . Remark 5.19. The ab o v e analysis suggests that, for all xed windo ws W ⊂ R 3 whose b oundary b d( W ) has Leb esgue measure 0 in R 3 , the sets of the form ∪ Λ ∈I B g ( W ) C ( Λ ) (resp., ∪ Λ ∈I F g ( W ) C ( Λ ) ) are appro ximately determined b y the X -ra ys in the four presrib ed L ( τ , 0 , 1) - diretions of U ico , where L = Im( I ) (resp., L = I 0 ); f. Examples 6.12 and 6.15. A dditionally , in the pratie of quan titativ e HR TEM, the resolution oming from the diretions of U ico is lik ely to b e rather high, whi h mak es this appro ximativ e result lo ok ev en more promising in view of real appliations; f. Remark 5.15 . 6. Outlook F or a more extensiv e aoun t of b oth uniqueness and omputational omplexit y results in the disrete tomograph y of Delone sets with long-range order, w e refer the reader to [ 24 ℄. This referene also on tains results on the in terativ e onept of su  essive determination of nite sets b y X -ra ys and further extensions of settings and results that are b ey ond our sop e here; ompare also [23℄. Although the results of this text and of [24℄ giv e satisfying answ ers to the basi problems of disrete tomograph y of iosahedral mo del sets, there is still a lot to do to reate a to ol that is as satisfatory for the appliation in materials siene as is omputerized tomograph y in its medial or other appliations. First, w e b eliev e that it is an in teresting problem to  haraterize the sets of Λ -diretions in gener al p osition ha ving the prop ert y that, for all iosahedral mo del sets Λ , the set of on v ex subsets of Λ is determined b y the X -ra ys in these diretions; ompare [13, Problems 2.1 and 2.3℄. Seondly , it w ould b e in teresting to ha v e exp erimen tal tests in order to see ho w w ell the ab o v e results w ork in pratie. Sine there is alw a ys some noise in v olv ed when ph ysial measuremen ts are tak en, the latter also requires the abilit y to w ork with impreise data. F or this, it is neessary to study stabilit y and instabilit y results in the disrete tomograph y of iosahedral mo del sets in the future; f. [ 1℄. A kno wledgements It is m y pleasure to thank Mi hael Baak e, Uw e Grimm, P eter Gritzmann, Barbara Langfeld and Reinhard Lü k for v aluable disussions and suggestions. This w ork w as supp orted b y 20 C. HUCK the German Resear h Counil (DF G), within the CR C 701, and b y EPSR C, via Gran t EP/D058465/1. Referenes [1℄ Alp ers A and Gritzmann P 2006 On stabilit y , error orretion, and noise omp ensation in disrete tomog- raph y SIAM J. Disr ete Math. 20 227239 [2℄ Baak e M 1997 Solution of the oinidene problem in dimensions d ≤ 4 . In: The Mathematis of L ong- R ange Ap erio di Or der (Ed. Mo o dy R V) pp. 944. (NA TO-ASI Series C 489 , Klu w er, Dordre h t) Revised v ersion arXiv:math/0605222v1 [math.MG℄ [3℄ Baak e M 2002 A guide to mathematial quasirystals. In: Quasirystals. A n Intr o dution to Strutur e, Physi al Pr op erties, and Appli ations (Eds. Su k J-B, S hreib er M and Häussler P) pp. 1748. (Springer, Berlin) arXiv:math-ph/9901014v1 [4℄ Baak e M, Gritzmann P , Hu k C, Langfeld B and Lord K 2006 Disrete tomograph y of planar mo del sets A 62 419433 arxiv:math.MG/0609393 [5℄ Baak e M and Hu k C 2007 Disrete tomograph y of P enrose mo del sets Philos. Mag. 87 28392846 arXiv:math-ph/0610056v1 [6℄ Baak e M and Mo o dy R V (Eds.) 2000 Dir e tions in Mathemati al Quasirystals (CRM Monograph Series, v ol. 13 , AMS, Pro videne, RI) [7℄ Baak e M, Pleasan ts P A B and Rehmann U 2006 Coinidene site mo dules in 3 -spae Disr. Comput. Ge om. 38 111138 arXiv:math/0609793v1 [math.MG℄ [8℄ Chen L, Mo o dy R V and P atera J (1998) Non-rystallographi ro ot systems. In: Quasirystals and Disr ete Ge ometry (Ed. P atera J) pp. 135178. (Fields Institute Monographs, v ol. 10, AMS, Pro videne, RI) [9℄ Con w a y J H and Sloane N J A 1999 Spher e Pakings, L atti es and Gr oups (3rd ed., Springer, New Y ork) [10℄ Co wley J M 1995 Dir ation Physis (3rd rev. ed., North-Holland, Amsterdam) [11℄ Dub ois J-M 2005 Useful Quasirystals (W orld Sien ti, Singap ore) [12℄ F ewster P F 2003 X-r ay S attering fr om Semi ondutors (2nd ed., Imp erial College Press, London) [13℄ Gardner R J 2006 Ge ometri T omo gr aphy (2nd ed., Cam bridge Univ ersit y Press, New Y ork) [14℄ Gardner R J and Gritzmann P 1997 Disrete tomograph y: determination of nite sets b y X-ra ys T r ans. A mer. Math. So . 349 22712295 [15℄ Gardner R J and Gritzmann P 1999 Uniqueness and omplexit y in disrete tomograph y . In: [21 ℄ pp. 85114 [16℄ Gardner R J, Gritzmann P and Prangen b erg D 1999 On the omputational omplexit y of reonstruting lattie sets from their X-ra ys Disr ete Math. 202 4571 [17℄ Gritzmann P 1997 On the reonstrution of nite lattie sets from their X-ra ys. In: L e tur e Notes on Computer Sien e (Eds.: Ahrono vitz E and Fiorio C) pp. 1932 (Springer, London) [18℄ Gritzmann P and Langfeld B On the index of Siegel grids and its appliation to the tomograph y of quasirystals Eur. J. Comb. to app ear [19℄ Guinier A 1994 X-r ay Dir ation in Crystals, Imp erfe t Crystals, and A morphous Bo dies (Do v er Publi- ations, New Y ork) [20℄ Hardy G H and W righ t E M 1979 A n Intr o dution to the The ory of Numb ers (5th ed., Clarendon Press, Oxford) [21℄ Herman G T and Kuba A (Eds.) 1999 Disr ete T omo gr aphy: F oundations, A lgorithms, and Appli ations (Birkhäuser, Boston) [22℄ de Boissieu M, Guy ot P and Audier M 1994 Quasirystals: quasirystalline order, atomi struture and phase transitions. In: L e tur es on Quasirystals (Eds. Hipp ert F and Gratias D) pp. 1152 (EDP Sienes, Les Ulis) [23℄ Hu k C 2007 Uniqueness in disrete tomograph y of planar mo del sets. Notes arXiv:math/0701141v2 [math.MG℄ [24℄ Hu k C 2007 Disr ete T omo gr aphy of Delone Sets with L ong-R ange Or der (Ph.D. thesis (Univ ersität Bielefeld), Logos V erlag, Berlin) [25℄ Janot C 1994 Quasirystals: A Primer (2nd ed., Clarendon Press, Oxford) DISCRETE TOMOGRAPHY OF ICOSAHEDRAL MODEL SETS 21 [26℄ Kisielo wski C, S h w ander P , Baumann F H, Seibt M, Kim Y and Ourmazd A 1995 An approa h to quan titativ e high-resolution transmission eletron mirosop y of rystalline materials Ultr amir os opy 58 131155 [27℄ Mo o dy R V 2000 Mo del sets: a surv ey . In: F r om Quasirystals to Mor e Complex Systems (Eds. Axel F, Déno y er F and Gazeau J-P) pp. 145166 (EDP Sienes, Les Ulis, and Springer, Berlin) arXiv:math/0002020v1 [math.MG℄ [28℄ Mo o dy R V 2002 Uniform distribution in mo del sets Canad. Math. Bul l. 45 123130 [29℄ Mo o dy R V and P atera J 1993 Quasirystals and iosians J. Phys. A 26 28292853 [30℄ Pleasan ts P A B 2000 Designer quasirystals: ut-and-pro jet sets with pre-assigned prop erties. In: [6 ℄ pp. 95141 [31℄ Pleasan ts P A B 2003 Lines and planes in 2 - and 3 -dimensional quasirystals. In: Coverings of Disr ete Quasip erio di Sets (Eds. Kramer P and P apadop olos Z) pp. 185225 (Springer T rats in Mo dern Ph ysis, v ol. 180 , Springer, Berlin) [32℄ Rén yi A 1952 On pro jetions of probabilit y distributions A ta Math. A  ad. Si. Hung. 3 131142 [33℄ She h tman D, Ble h I, Gratias D and Cahn J W 1984 Metalli phase with long-range orien tational order and no translational symmetry Phys. R ev. L ett. 53 19511953 [34℄ S hlottmann M 1998 Cut-and-pro jet sets in lo ally ompat Ab elian groups. In: Quasirystals and Disr ete Ge ometry pp. 247264 (Fields Institute Monographs, v ol. 10 , AMS, Pro videne, RI) [35℄ S hlottmann M 2000 Generalized mo del sets and dynamial systems. In: [6 ℄ pp. 143159 [36℄ S h w ander P , Kisielo wski C, Seibt M, Baumann F H, Kim Y and Ourmazd A 1993 Mapping pro jeted p oten tial, in terfaial roughness, and omp osition in general rystalline solids b y quan titativ e transmission eletron mirosop y Phys. R ev. L ett. 71 41504153 [37℄ Steinhardt P J and Ostlund S (Eds.) 1987 The Physis of Quasirystals (W orld Sien ti, Singap ore) [38℄ Steurer W 2004 T w en t y y ears of struture resear h on quasirystals. P art I. P en tagonal, o tagonal, deago- nal and do deagonal quasirystals Z. Kristal lo gr. 219 391446 [39℄ W ashington L C 1997 Intr o dution to Cylotomi Fields (2nd ed., Springer, New Y ork) [40℄ W eyl H 1916 Üb er die Glei h v erteilung v on Zahlen mo d. Eins Math. A nn. 77 313352 Dep ar tment of Ma thema tis and St a tistis, The Open University, W al ton Hall, Mil ton Keynes, MK7 6AA, United Kingdom E-mail addr ess : .hukopen.a.uk URL : http://www.mathematis.open.a.u k/Peo ple/ .hu k

Original Paper

Loading high-quality paper...

Comments & Academic Discussion

Loading comments...

Leave a Comment