THE STUDY OF BAND GAP ENGINEERING FOR PHONONIC CRYSTALS AND GAP STRUCTURES IN PHONONIC QUASICRYSTALS 2 by YUN LAI A Thesis Submitted to The Hong Kong University of Science and Technology in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Department of Physics August 2005, Hong Kong Copyright by Hong Kong University of ace & Technology and the author. DISTRIBUTION PROHIBITED.Authorization I hereby declare that I am the sole author of the thesis. Tauthorize the Hong Kong University of Science and Technology to lend this thesis to other institutions or individuals for the purpose of scholarly research, I further authorize the Hong Kong University of Science and Technology to repro- duce the thesis by photocopying or by other means, in total or in part, at the request of other institutions or individuals for the purpose of scholarly research, iiTHE STUDY OF BAND GAP ENGINEERING FOR PHONONIC CRYSTALS AND GAP STRUCTURES IN PHONONIC QUASICRYSTALS by YUN LAI This is to certify that I have examined the above PhD thesis and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the thesis examination committee have been made. Department of Physics August 29, 2005 iiiAcknowledgments Hereby, I would like to express iny gratitude to my supervisor, Prof. Z. Q. Zhang, for his full support, patient guidance, and invaluable advice on iy research. During the innumerous meetings with him in the past years, I have benefited a lot from his abundant knowledge and profound insight of physics. His aspiring lifestyle and persistent interests in physics have encouraged me to overcome difficulties and adhere to my goals, It is my pleasure to express my gratitude to Prof, C. H. Chan at the Department of Electrical Engineering in the City University of Hong Kong and Prof. L. Tsang at the Department of Electrical Engineering in the University of Washington Seattle, for their introduetion of the SMCG method and invaluable help in our collaborations. I would also like to thank Dr. P. W. Leung and Prof. C. 'T. Chan for generously providing us the parallel computing facilities which are crucial to my research. I would not forget the kind help and useful discussions from Prof, C.'T. Chan, X. D. Zhang, X. P,Q. J. Chu, Z. Y. Li, S. K. Cheung and Y. Wu [sincerely thank Prof. C. T. Hsu (ME HKUST), Prof. C. H. Chan (EE CityU), Prof. P. Sheng and Prof. C.'T. Chan for their kindness and readiness to serve in iny thesis examination, My special thanks are given to my parents, who have always loved me and sup- ported me during my PhD life, which is spent far away from their side and need. Last, I would thank all my friends in HKUST, who helped and accompanied ne through these years.TABLE OF CONTENTS Title Page Authorization Page Signature Page Acknowledgnents Table of Contents List of Figures Abstract 1 Introduetion 1.1 Periodic structures: 1.1.1 Photonic crystals. 1.1.2. Phononic crystals 1.2 Quasiperiodic structures L.2.1 Photonic quasicrystals 1.2.2 Phononic quasicrystals 13) Summary 2 Theory and numerical methods 2.1. Elastic and acoustic wave equations: ii iv xiii iL 4 u 16 162.2. Variational principle 2.3 Scattering of waves by a single cylinder . . . 24 The multiple seattering theory (MST) 24.1 Periodic case: the complex cell cas 24.2 Finite systems: the SMCG method 2.4.3. Transplant to elastic waves & Summary. eo odo 3 Band-gap engineering 3.1 Previous works 3.2 Enlarging acoustic band gaps 3.2.1. Steel cylinders in air host 3.2.2 Mercury cylinders in water host 3.2.3 Water cylinders in mereury host 3.3 Enlarging elastic band gaps 3.3.1 Aluninun cylinders in epoxy host 3.3.2 Rotational acoustical branch 3.3.3. Flat bands. . 3.34 Gap maps 3.3.5 Extension to phononic quasicrystals 34 Summary 4 Properties of gap structures and wave functions in phononie quasicrystals 4.1 Previous works 4.2 Gap structures due to short-range order 4.3 Gap structures due to long-range order vi 34 26 38 4 434.4 Properties of the wave functions... 6... 103 4.5 Summary cee cee ++. U8 5 Conclusion 119 Appendix: Lattice sums in the MST method 122 References 124 vii3. 3.2 33 3.4 3.6 38. 3.10 3.11 3.12 3.13 List of Figures Band structure for a square lattice of steel cylinders in air host Distributions of fields and their gradients for the band edge states LM and 2X in Fig. 3.1 ee ‘The estimated and calculated frequency shifts of the band edge states as functions of the inclusion radii R; for several acoustic systems with different insertion materials and positions, Band structure for a square lattice of water cylinders in mercury host Distributions of fields and their gradients for the band edge states 2M and 3M in Fig. 34 Band structure for a triangular lattice of alumninum cylinders in epoxy host Band structure for 9 triangular lattice of almninum cylinders in epoxy host with air inchisions Energy density distributions of the Bloch state LX in Fig. 3.6 Energy density distributions of the Bloch state 2X in Fig. 3.6 Energy density distributions of the Bloch state 30 in Fig. 3.6 Energy density distributions of the Bloch states 2X and 3P in Fig, 3.7 Band structure for a triangular lattice of aluminun cylinders in epoxy host with water inclusions The displacement fields dy and isp at a given time for the 2X and 30° states in Fig. 3.6 viii 52 693.14 3.16 3.17 3.18 3.19) 3.20 3.22 4.1 43. Displacement field maps for some resonances in an air eylinder with hard bomdary ... 0... ee Displacement field maps in some flat bands with wave vector f = 0 in Rigo Band structure for a triangular lattice of aluminum cylinders in epoxy host with smaller air inclusions ‘Transmission spectra after considering the viscosity in air cylinders for the s em in Fig. 3.16 Gap map for a triangular lattice of aluminum eylinders in epoxy host with air inclusions Gap map for a triangular lattice of steel cylinders in epoxy host with air inclusions Radiation power for the circular samples of a 12-fold symmetric phononic quasicrystal of alumuium cylinders in epoxy host, with and without air inclusions. Gap map for @ 12-fold symmetric phononic quasicrystal of aluninun cylinders in epoxy host with air inclusions Gap map for 4 12-fold symmetric phononie quasicrystal of steel eylin- ders in epoxy host with air inchusi pns Radiation power vs. band structure calculations for square and trian- gular lattices of rigid cylinders in air host Radiation power for a 12-fold symmetric quasiperiodic system consist- ing of rigid cylinders in air host... 2... Transmission for # 12-fold symmetric quasiperiodic system consisting of rigid eylinders in air host 72 74 78. 79 88 9044 45 46 48 4.10 all Gap maps due to the short-range order of the 12-fold symmetric quasiperi- odie system consisting of rigid cylinders in air host as a function of the filling fractions. The result is compared to the square and triangular lattices of rigid eylinders in air host Radiation power for small samples of the 12-fold symmetric phononic quasicrystal consisting of rigid cylinders in air host, the samples’ radii are Ry = 3a, Ga and 8a. 2.0.0... panos Radiation power for samples of the 12-fold symmetric phononie qua~ sicrystal consisting of rigid cylinders in air host, the samples’ radii are R, = 3a, 6a and 8a, 14a and 24a Bese eeEeUa Radiation power for samples of the 12-fold symmetric phononic qua- sicrystal consisting of rigid cylinders in air host, the samples’ radii are Ry =Mdaand 24a. 2... 2. Radiation power for large samples of the 12-fold symmetric phononic qnasicrystal consisting of rigid cylinders in air host, the samples’ radii are Ry = 24a, 300 and 34a. eee Ry Radiation power for sample 14a of the 12-fold symmetric phononie quasicrystal consisting of rigid cylinders in air host with perfect or partially destroyed long-range order in the structure. Radiation power in low-frequency gap areas for sample size Ry = Ma of the 12-fold symmetric phononic quasicrystal consisting of rigid eylin- ders in air host, with perfect or partially destroyed long-range order in the structure Radiation power vs. band structure calculations for a triangular lattice of rigid cylinders in air host 92 94 96 98 99 100 10L 1044.12 4.13 414 4.16 4.18 4.19 Average intensity for a frequency at a band edge in a phononic crystal sample. . Average intensity for a frequency inside a band in a phononie crystal sample. ee . Intensity distribution map for f = 0.670 near a band edge in a 12-fold. ymmetric phononic quasicrystal sample consisting of rigid cylinders in air with Rp = 34a... 0.2... 0.670 near a band edge in a 12-fold Average intensity I(r) for f symmetric phononic quasicrystal sample consisting of rigid cylinders in air with R, = 34a, ee Intensity distribution map for f = 0.42375 corresponding to a radiation power peak at a band edge ina 12-fold symmetric phononic quasicrystal sample consisting of rigid cylinders in air with , = 34a. I(r) for f = 0.42375 and f = 0.4239 corresponding Average intensiti to a radiation power peak and shifted a little away from the peak, re- spectively, which are near a band edge in a 12-fold syimmetrie phonontic = 34a, quasicrystal sample consisting of rigid cylinders in air with R, Intensity distribution map for f = 0.4239 near a radiation power peak at a band edge in a 12-fold symmetric phononie quasierystal sample consisting of rigid cylinders in air with R, = 1000 Average intensity I(r) for f = 0.4239 near a radiation power peak at a band edge in a 12-fold symmetric phononie quasierystal sample consisting of rigid cylinders in air with Ry = 1000 xi 106 106 108 109 110 ill 112 131.20 4.22 4.23 Average intensities I(r) at f = 0.42375 in a sample of Ry = 34a, ob- tained by using different normwise backward errors as the convergence criterion. tiger ET Intensity distribution map for f = 0.4687 corresponding to a radiation power peak in a small dip of the spectrum for a 12-fold symmetric phononic quasicrystal sample consisting of rigid cylinders in air with R, = 34a. ee Average intensity I(r) for f = 0.4687 corresponding to a radiation power peak in a small dip of the spectrum for a 12-fold symmetric phononic quasicrystal sample consisting of rigid cylinders in air with Ry = 34a. Average intensity I(r) for f = 0.92 inside a high-frequency band of the spectrum for a 12-fold symmetric phononic quasierystal sample consisting of rigid cylinders in air with R, = 34a: xii 4 L1G uyTHE STUDY OF BAND GAP ENGINEERING FOR PHONONIC CRYSTALS AND GAP STRUCTURES IN PHONONIC QUASICRYSTALS by Yun Lai Department of Physics The Hong Kong University of Science and Technology Abstract Phononic erystals and quasicrystals are solid-solid, fluid-solid or fluid-fluid composites with periodic or quasiperiodic long-range order, respectively. They are of interest not only because of the profound effects of their periodic or «nasiperiodic structure on wave propagation (e.g. the existence of elastic ot acoustic band gaps). but also because of potential applications like sound filters. transducer design and acoustic mirrors, etc. In this thesis, I will present some research progress in the study of both phononie crystals and quasierystals For phononie crystals, we have proposed a simple, systematic and efficient method to engineer clastic (acoustic) band gaps. A gap can be euhanced or cre- ated by altering the microstructure according to the field-energy distributions of the Bloch states at the band edges as well as their derivatives. Due to the structure of the acoustic wave equation, the engineering of acowstie band gaps is much more efficient than that of photonie band gaps. Por elastic waves, a large absolute band gap can be enhanced or created by inserting air inclusions int a two-component elastic phononic erystal or quasicrystal with sinall density con- trast and filling fraction. ‘The positions of the insertion are chosen to suppress the shear potential energy of the acoustic branches and lower their frequencies. ‘The validity of the proposed methods is supported by mmultiple-scattering calen- xiiiations in various two-dimensional systems, such as steel cylinders in air, water cylinders in mercury and aluminum cylinders in epoxy. Our methods make the acoustic band gap “designable” and the realization of a light and effective sonic insulator possible ‘The sonic band-gap structures of a 12-fold symmetric quasicrystal consist- ing of rigid cylinders in air are investigated by using the multiple scattering method. Large full gaps are found owing to its high symmetry. However, the gap structure evolves with the sample size. By using the SMCG method, we have observed a self-similar like behavior of states emerging inside original gaps and gaps emerging inside original bands when the sample size is increased. ‘The long-range order has been proved important for the formation of gaps at low frequencies. We have also studied the properties of wave functions in large samples (1p to 33919 cylinders). Various types of wave functions have been seen, including extended state, self-similar state with a power-law decay and confined state. xivChapter 1 INTRODUCTION Crystals and quasicrystals are two kinds of ordered materials in the world. Peri- odie structures, which were discovered in crystals nearly a century ago, stimulated the development of solid si ate physics [1]. Compared to crystals, quasicrystals are a kind of new material found just two decades ago [2]. Bearing quasiperiodicity instead of peri icity, quasierystals have shown very different physical properties and attracted wide interests in the past two decades. In recent years, people have created artificial materials in the macroworld, ac- cording to the imicrostructures of erystals or quasicrystals. For dielectric materials with electromagnetic waves, they are called “photonic crystals” or “photonic qua- sicrystals”, respectively. For elastic materials with elastic (acoustic) waves, they are called “phononic erystals” or “phononie quasicrystals”, respectively. ‘These artificial materials exhibit similar physical properties as in crystals or quasicrystals, which are neatly induced by the effects on the wave equations from the artificial structures. In the following, I give @ brief introduction to these new artificial materials 1.1 Periodic structures ‘The crystal structure was first discovered by Lane, Friedrich and Knipping in 1912 [3]. By doing x-ray diffraction studies, they found that crystals are composed2 of periodic array of atoms. During the following research on periodic structures, some important physical concepts and principles have been developed in solid state physics. One of the most important concepts is the band (dispersion) structure, which is derived from the Bloch theorem in a periodic system. ‘The band structure is composed of many bands. Each point in the bands denotes a Bloch state, which can be mathematically written in the form of a Bloch wave function. Band (dispersion) structure and Bloch states are induced by the interference effects of the periodically scattered waves in crystals. Sometimes, when the interference effects are strong in certain directions, waves will be forbidden to propagate in those directions. Then partial band gaps appear in the energy (frequency) spectrum of the band structure. If the interference effects are strong in all directions, the partial band gaps in differ ent directions will overlap aud combine into a complete band gap, where the wave propagations are forbidden in all directions. The band gaps in crystals are crucial for many applications, For example, in semiconductors there is a complete band gap between the valence and conductance bands. 1.1.1 Photonic crystals In 1987, E Yablonoviteh [4] and S. John {5] proposed for the first time that pe- riodie structures of dielectric materials could be designed to have complete photonic are forbidden. This novel idea and its potential 1s in which optical modes applications in optical devices have attracted wide interest. In the following years, people have studied the photonic crystals on various aspects, including the fabrica tion of photonic band gaps, the effect of defects, wave guiding and bending, Anderson localization. negative refraction and super-lens effects, etc. Great progress has been3 made in both theory and applications. For example, at first, only partial gaps were observed for three-dimensional photonic crystals [6]. Ho, Chan and Soukoulis pre~ vinmetry dicted theoretically that a full band gap can be realized by lowering the s of the crystal [7], which was later experimentally verified. The defects in photonic crystals have been extensively studied with potential applications to the design of lasers, waveguides, and the integrated optical devices [8]. ‘The debate on negative refracion (9] has been brought. to close by experiments [10] At the same time, the study of phononic crystals began soon after the discovery of photonic crystals. Most of the phenomena found for electromagnetic waves in pho- tonic crystals have been also found for elastic (acoustic) waves in phononic crystals. such as complete band gaps, defect effects, wave guiding and bending, uegative refrac tion, ete, The physical mechanisms behind these phenomena are similar. Since this thesis is focused on elastic (acoustic) composites, we will not introduce the research tals in detail. However, it is important to point out that de- progress of photonic er spite the similarity between photonic crystals and phononic crystals, the differences between the physical natures of electromagnetic waves and elastic (acoustic) waves als, could bring new physics into phononie crys 1.1.2. Phononic crystals The first numerical study of phononic crystals was presented by M. Sigalas et al, [11] in 1992. By using the plane wave expansion method (PWE), they calculated the band structure and density of states (DOS) for elastic and acoustic waves in periodic structures consisting of identical spheres placed periodically within a host homogeneous material, In their work, analogous to photonic crystals, phononie band4 gaps were shown to exist in systems composed of properly chosen materials in the background. In the next. year, the band structures for phononic crystals in two- dimensions were also calculated by M. Sigalas et al. [12] and M. $. Kushwaha et al, [13]. As a comparison to photonic crystals, because of the additional physical parameters in the phononie crystals, as well as the coupling of transverse and lon verst gitudinal waves, phononic crystals are expected to generate new features over the physies in the photonic cry ‘The early works on phononic crystals were focused on the pursuit of phononic band extensive calculations of band structures gaps. By using the PWE and other methods for elastic (acoustic) waves in two-dimensional (rods in a host material) or three dimensional (scatterers in host material) two-component periodic composites have been performed. with arbitrary or realistic physical parameters used in computations [11-29]. The results indicate that the existence and the size of a phononic band gap are mainly affected by the following factors: topology [30]; lattice structure; filling fraction of the two components; and the parameters contrasts of the scatterers and the host material, The parameters contrasts include the density contrast, wave speed contrast of the scatterers and the host material, aud the longitudinal to transverse s in the scatterers and the host material, respectively. Among wave speeds cont those factors, one of the most important factors is the density contrast of the seatterers and the host material. For example, for elastic waves it is found that, the phononic band gap is easier to realize in the system of high-density inclusions in a low-density host; while for acoustic waves, the phononic band gap prefers to appear in the system of low-density inclusions in a high-density host {12}. The wave speed contrast (or the and the background also plays an important role clastic constant contrast) of the bi in the creation of phononic band gaps [15]. Similar to the density contrast, the higher5 wave speed contrast, the easier to realized phononie band gaps. The filling fraction is another critical factor in determining phononic band gaps, the effects of which 1 For the lattice structure, it is found that due to the higher symmetry, the triangular are often presented in the “gap ~ filling fraction” maps in literature [14-17,20 lattice often favors larger phononic band gaps than other two-dimensional lattices, While most efforts were devoted to realization of phononic band gaps, some efforts were made to understand the band structure results, e.g., by analogy with the LCAO approach [23, 24] Tn contrast to theoretical studies, there have been relatively fewer experimen- tal investigations on phononic cryst als [25,27-29,31-37]. ‘The first sonic band gap measurement was done with a sculpture exhibited at the Juan March Foundation in Madrid (31). In later experiments, people have investigated varions kinds of phononic crystals consisting of solid inclusions in fluid host, fluid inclusions in solid host, and solid inclusions in solid host, ete, By transmitting elastic (acoustic) waves through a phononie crystal, strong attenuations in some frequency ranges are observed, which indicate partial band gaps in that direction. Most of the experimental transinission data coincide well with the numerical transmission or band structure calculations. However, sometimes due to the anti-symmetry of the eigenmodes in some bands, these modes cannot be excited by incident plane waves, and those unexcited bands are called “deaf bands” [35,38]. Recently, hypersonic phononic erystals [39] were fabricated, which could possess both phononic and photonic band gaps In the previous studies of two-component simple cell phononie crystals, the ex- ity contrast and filling istence of a large band gap relies heavily on the large di fraction of the scatterers, which could lead to difficulties in fabrications and applica- tions. Thus people are trying other methods to enhance the properties of phononic6 band gaps. Caballero ef al. have demonstrated both theoretically and experimen- tally that sonic band gaps can be enhanced by reducing the structure symmetry of the phononic crystals [40,41]. Lin et al., through calculations by using the multi- nanifested that not only Bragg scattering, locally ple sattering method (MST) [42] resonant units could also produce phononic band gaps. This idea was realized in experiments [43] with complete band gaps opened at low frequencies in the audible range of human ears, which makes it a promising sound proofing material. Since then, the locally resonance band gap materials were extensively investigated [44-47] in both two and three dimensions. For phononic crystals with anisotropic inclusions, such as square cylinders, the effects of shapes and symmetries of scatterers have been stud- ied for both acoustic waves and elastic waves munerically [48-51]. Phononic erystals consisting of hollow cylinders have also been studied [52-54] Based on the variational principle, we have proposed a simple and general method to enhance or engineer both acoustic [55] or elastic [56] band gaps in phonoie crystal By inserting a third component at proper positions in the two-component systems, the phononic band gaps can be immediately opened or enlarged. The insertion positions were chosen according to the field distributions of the original band edge states Phe application to cermet topology (isolated inchisions) was manifested by MST calculations in two dimensions. In this thesis, we only consider cermet topology. While people gained more and more knowledge in the realization of phononic and gaps, they were also investigating some other interesting physical phenomena in phononic crystals, Yang et al, have studied the tunneling effects in a phononic band gap [57,58]. In their work, both experimental and theoretical results show that the while group velocity in a phononic band gap increases linearly with sample thickness, the phase velocity isa constant. Another hot topic is the unusual wave propagations in7 phononie crystals such as the negative refraction and wave focusing effects in phononic crystals [58-64]. The mechanism of those unusual refraction effects is explained by with Bloch wave matching the refraction boundary conditions with the Bloch wave vectors opposite to their energy poynting vectors. Ultrasonic wedge [65] for elastic wave bending and splitting without requiring a full band gap was also constructed Similar to photonic crystals, defects can be introduced into phononic crystals. Several types of defects have been studied [36, 66-73]. The point defects can act like microcavities with defect modes in the phononic band gaps strongly localized about the local defect. The line defect, often created by removing a line of scatterers in the phononie crystal, can act like a wave guide in the phononic band gap. However, due to complexity of elastic mixed modes, and the periodicity of the guide boundaries along the guide line, gaps may appear in the line defect wave guide otherwise allowed frequencies [71]. A lot of work has been done on investigating the coupling between, the defect modes, e.g., between several or a consecutive line of point defects, between two line defects (bending wave guides), and between line defect and point defects stublike waveguide). For example, the primary effect of the stub (point defect) beside a guide line was found to induce zeros in the transmission spectrum at its resonance frequencies [66]. Besides defects, surface waves due to phononic band gaps were also studied [36,74, 75]. In addition to the increasing knowledge on the characteristies of phononie crystals, some theoretieal and munerical methods were developed for phononie crystals. The theoretical work inchides Green's functions [76], local density of state (LDOS) [77] etc. The most popular numerical methods are: Plane wave expansion inethod (PWE) {11,14}, multiple scattering theory (MST) (42.56. 78-80], finite difference time domain (FDTD) [32.81], variational method [47], and lumped-mass method [82]. The PWE.8 method is the simplest and was widely used to obtain phononic band structures. Its principle is to expanse the system parameter functions (density, speeds) and wave functions by plane waves in the wave equation. The PWE method is applicable to any shapes of scatterers, But when the parameters contrasts are large, the PWE method converges slowly due to the need for a large amount of plane waves. For a mixed system composed of fluids and solids, it may lead to unphysical results [29]. On the contrary, the MST method is quite robust and applicable to large contrast systems including mixed systems of fluids and solids, The principle of the MST method is similar to the KKR (Korringa, Kohn and Rostoker) approach for electrons and is explained later in Sec. 2.4 in this thesis. Normally, the MST method was only used to solve systems of spherical or eylindrical scatters, which have simple forms of scattering matrices. The FDTD method is a real-time simulation method. The wave equation is discretized in both space and time domains, with proper boundary conditions. The FDTD method is mostly used to calculate transinissions. ‘The variational method and the Iumped-mass inethod were proposed recently and showed advantage in certain conditions, In conchision, various numerical methods have been developed to deal with different cases in research. In this thesis, we will use the MST method, since it is robust for cases of large contrasts and we only deal with cylindrical scatterers. 1.2 Quasiperiodic structures Since the discovery in 1984 of crystals with “forbidden” symmetry by Shechtman, Blech, Gratias and Cahn , materials that were dubbed quasicrystals, opened a new branch of crystallography and solid state physics. These structures possess long- range aperiodic order and crystallographically forbidden symmetries, and are thas9 alline fundamentally different from the two other known types of solid structure, ery and amorphous. The structure of quasicrystals is defined by the following conditions [83] 1, Quasiperiodic translational order: The fourier transform of a quasiperiodic structure is composed of discrete sharp peaks (diffraction peaks), which can be spanned by a minimal set of basis vectors. The mumber of basis vectors ex- ceeds the dimensions of the quasiperiodic structure. Thus the reciprocal space is densely filled by diffraction peaks and the fourier spectrum is singular con- timous. Minimal separation between atoms: This distinguishes the quasicrystal from a single set of sites obtained by superimposing two periodic lattices with periods whose ratio is irrational. 3. Orientational order: The quasicrystals possesses crystallographically disallowed orientational symmetry. This condition separates it with the incommensurate crystal which has erystallographically allowed orientational symmetry. However this requirement has been seriously questioned recently [84] Besides the previous conditions, another important characteristic of the quasiperiodic structure is self-similarity, Inflation (deflation) rules are often used to construct qua- sicrystal structures, as well as other methods such as the generalized dual or multigrid method [83]. The quasiperiodic structure can also be generated by projection or eut from higher dimensional space [83]. Another important property for the quasiperiodic structure is known as Conway's theorem [85], in the case of Penrose lattice. Conway's theorem states that any finite-size section of the lattice repeats its figure an infinite number of times quasiperiodically (86)10 Due to the lack of periodicity, some of the concepts for crystals such as the en- ergy (dispersion) bands can not be transplanted to quasicrystals. Most calculations are based on the tight-binding method. For one-dimensional systems (87-96), the quasiperiodicity can be described in two models [87,96], i.e., the on-site model where the site energies are arranged quasiperiodically, and the transfer model where the hopping integrals are arranged quasiperiodically. For two- and higher-dimensional systems [86,97-109], the setup of the model can be more complicated. For a two- dimensional system arranged according to the Penrose lattice, the models can be categorized into two types [109]. One is the vertex model [101], where atomic sites are placed on vertices of rhombuses, The other is the center model [86], where atomic sites are placed at the centers of rhombuses. Different models may bring different results, But, in general, people have found the spectra of quasicrystals to exhibit some singular continuous behaviors, It is also generally believed that wave functions: with critical behaviors exist in quasicrystals [89,95, 98, 105]. They are denoted as the critical state, which is intermediate between the extended state in erystals and the localized state in amorphous materials. The critical state decays weaker than exponential decay, inost likely by a power law, and has rich self-similar structures. tal. ‘The one-dimensional Fibonacci s stem inay be the most studied quasic ‘The Fibonacci system is generated by the rule Sixa = {$)-1Si}, with Sp = {B} and S$) {A} [89,95]. By using the transfer matrix and dynamical mapping mneth- ods [89,95], many rigorons results were obtained. For instance, the energy spectrum of a Fibonacci system consists of a self-similar Cantor set with zero Lebesgue inea- sure [95]. Most of the states in a Fibonacci system are believed to be critical states. ‘The sell ilar behavior of band splitting and the spatial distributions of critical states (including chaotic states) can be explained in a simple physical picture origi-cor nated from the renormalization group ideas [87]. Extended and localized states were also found munerically by Lin et al. [96] for the on-site model. Two-dimensional and three-dimensional quasicrystals, however, apart from a few particular examples, have mainly been investigated by numerical calculations for finite systems or periodic ap- proximants. Penrose lattices were studied for both vertex model and center model. ‘The DOS is unsmooth and spiky for large systems [86] and all extended, critical and localized states have been found numerically [109]. Mathematically, strictly localized states without tails were found due to the local connectivity of the Penrose lattice and were named “confined states” by Fujiwara et al. [97]. They have also found string states”, which appear on some strings of rhombuses. Some multifractal states were found theoretically for some particular Hamiltonian and configuration of the systems (98, 105] 1.2.1. Photonic quasicrystals The optical analogue of one-dimensional quasierystal is realized by stacking two different dielectric layers according to a simple generation rule [110-120]. Siti to the electronic structures in one-ditnensional quasicrystals, the problems in one- dimensional photonic quasicrystals can be rigorously solved by the transfer matrix and the dynamical mapping method [113]. The typical DOS structure (or gap structure) in a Fibonacci system has a self-similar nature and consists of narrow resonances separated by numerous gaps, where the DOS tends to zero. The self-similarity is also exhibited in the transmission spectra and the field distributions of the critical states [111,113]. The pulse propagation in Fibonacci quasicrystals has been observed experimentally. Large pulse delays and pulse stretching are observed when excitingthe band edge states of the Fibonacei structure [111]. A two-dimensional photonic quasicrystal consists of cylinders placed perpendicular to the basal plane on vertices of tiles composing a quasiperiodic pattern. The first study of two-dimensional photonic quasicrystal was done by Chan et al. [121] in 1998. ‘Their main purpose wa to investigate whether a structure without periodic order could have complete gaps or not. After investigating the transmission spectra of the quasicrystal samples by using a FDTD method [81] with periodic boundary condition, they have found complete gaps in eightfold symmetric photonic quasicrystals. The next question is what mechanism has produced the gaps in the absence of periodicity. Since the transmission spectral gaps could be found in rather small subsections of smi a structure, their conclusion was that local scattering is governing the formation of gaps rather than global scattering and, therefore, long-range periodic order is not ‘a prerequisite for the existence of a gap, Another interesting phenomenon is the existence of localized states in small samples, which is induced by quasiperiodicity and has nothing to do with defect or disorder. However, due to the incapability to calculate large-size samples, the localization properties were not characterized, In the next year, Jin et al. [122], measured the transmittance in the microwave re- gion for the TM wave in a two-dimensional octagonal photonic quasicrystal composed of alumina cylinders. Isotropic gaps which do not depend on the incident direction were observed, even in a small piece of the quasicrystal. Similar properties have been found in two-dimensional dodecagonal and decagonal photonic quasicrystals [123]. q \¢ local density of states (LDOS) and DOS were also investigated [124,12 saine group found photonic gaps in amorphons photonic materials {126.127}. which is another manifestation that short-range order can produce gaps. Zoorob ¢t al., have armounced experimental observation of complete photonic gaps13, in a 12-fold symmetric photonic quasicrystal of sinall dielectric contrast (128. 129]. ‘lattice is such as air holes in silicon nitride or glass. The 12-fold symmetric qua it may lead to complete gaps easier more isotropic than any periodic lattices; th However, in a following paper, Zhang et al. found no absolute gaps in the above stem with such low dielectric constast [130], although the 12-fold symmetric square- triangle tiling quasilattice is indeed very good for the realization of photonic gaps at higher dielectric contrasts. Hase et al. have shown experimental results of both isotropic and anisotropic gap structures in transmission spectra of two-dimensional fivefold and eightfold photonic quasicrystals [131,132] Besides the study of gap structure and localized states, the study of defects in photonic quasicrystals (121, 133-135] has started since the seminal work of Chan et ai. in 1998. In a photonic quasicrystal, each cylinder (site) is located in # different euviroument, therefore, by removing one cylinder from a different location, defect states with different frequencies and mode patterns ean be produced. Thus the defects states are more complicated and interesting than those in photonic crystals, Besides point defects, line defects (wave guides) have also been investigated [122, 134, 135} where similar wave guiding effects exist Notomi et al, [136] have observed coherent lasing actions due to the optical feed- back from quasiperiodicity, exhibiting a variety of 10-fold-syinmetric lasing spot pat- terns. Those spot patterns indicate the result of well-defined extended modes that are coherently spread throughout the sample, Due to sinall contrasts in their system, their results were explained by considering the reciprocal lattice. Recently, Della Villa et al, {137] have studied the behaviors of gaps in. photonic quasicrystals. They found that some frequency gaps exhibit a behavior similar to that typical of photonic crystal, while others do not. And they conclude that long-rangeld interactions, as well as multiple scattering, are involved in the formation of gaps. However, due to small size limitation in calculation, direct proof is still lacking. 1.2.2. Phononic quasicrystals Yompared to photonic quasicrystals, there have been fewer studies of phononic quasicrystals. One-dimensional phononic quasicrystals was investigated by Velasco et al, [138], including Fibonacci, Trne-Morse, and Rudin-Shapiro sequences. In 2002, we published a paper on gap structures for small samples in two-dimensional phononie quasierystals 139]. which consisted of a 12-fold symmetric quasilattice (square-triangle tiling) of rigid cylinders in air. The existence of mnulti-gap struc- tures in sonic quasicrystals was proved. Sutter ef al. [140] have done experiments involving the two-dimensional phononie quasicrystals of steel rods in water. The ar- rangements inchide penrose tilings with different decorations. However, characteristic broad dips instead of gaps were found. Some dips are isotropic and some are not. The two-dimensional clastic phononic quasicrystal was investigated by us (141. The inethod of inserting air cylinders was also applied to create complete gaps in phononic quasicrystals, just as in the case of phononie crystals. Various specific effects in elas: tie medium, such as the effects of the coupling of longitudinal wave and transverse waves have not been investigated in phononic quasicrystals yet 1.3. Summary In this chapter, we have introduced the recent progress in the study of phononic crystals and quasierystals. Due to the lack of works in phononic quasicrystals, theprogress in photonic quasierystals is introduced, As mentioned before, periodie struc . while their proper- tures and quasiperiodic structures both exhibit long-range orde als is focused on the ties are totally different. The study of phononie (photonic) en various properties and potential applications of this artificial material, including the enhancement of complete gaps, the negative refraction and wave focusing effects, the defects, Anderson localization, ete. On the contrary, the properties of quasiperiodic systems is not fully explored, The study of phononic (photonic) quasicrystals may help understand the physics in quasiperiodic systems, as well lead to potential appli- rystals. In this thesis, I will firstly cations like the lasing actions in photonic qua: give a detailed investigation on the band gap engineering for acoustic phononic er; tals in See. 3.2 and elastic phononic erystals in See. 3.3, as well as extension to elastic phononic quasicrystals. Then [ will present the research progress on the properties of gap structures due to short-range order in Sec, 4.2, gap structures due to long-range order in Sec. 4.3 and wave functions in See, 44, for a 12-fold symmetric acoustic sicrystal cons phononic que ting of rigid cylinders in airChapter 2 THEORY AND NUMERICAL METHODS 2.1 Elastic and acoustic wave equations If it returns A solid body becomes deformed when subjected to external forces to its original form when the forces are removed, it is said to be clastic. As long as the wavelength is large compared with the interatomic distances, any solid can be regarded as a material continuum. Its mechanical state is described by two tensors of rank two, the stress tensor @ and the strain tensor S, or the time- and space-variant clastic displacement field @ which is governed by the elastie wave equation [142, 143} When a solid is deformed, the deformation information is contained in the gradi ents of displacements Ou,/dx;. However, the anti-symmetric parts of the gradients, 1 (Ou; Ou, ( 3H), 1) should be excluded since they describe the simple overall rotations without deforma- tion of the solid, Only the symmetric parts, which are defined as the strain tensor 8, describe the deformation of the solid, (2.2) Ina two-dimensional solid, the strain tensor can be expressed in cartesian coordinates 16os Ou, 1 ( Ou, Ou, Or 2 (a oe) 1 (Ou Ou, du, 2 (% + aus) wy The diagonal terms indicate simple extensions or compressions. ‘The off-diagonal (2.3) terms indicate shear distortions. Elastic restoring forces develop inside a solid when it is deformed. In order to describe the forces, the stress tensor oi; is defined as the force applied in the # direction to a mit area of a plane whose normal lies in the j direction, by the material on the other side (142, 143]. The total force in the i direction applied on the overall surface of a sinall part of the solid by the surrounding inaterial ean be expressed as = [ auhyds [ Soay (2.4) 0 Oty Here [, denotes the direction cosine of the surface normal vector and the unit j vector According to Newton's Law, the dynamic equation is written as doy _ us On; OF or in the vector form (2.6) Here pis the density Experimental evidence has shown that the clastic behavior of inost substances is well described, assuming small deformations, by the first-order term in the following18, ‘Taylor expansion for the stress: doi, : (Su) = 05(0) + (4) Sut This gives, since 0;;(0) Oy = Ciynt Si (2.8) where cia = (33) denotes the fourth rank stiffness tensor, which describes Sumo the most general linear relation between the stress tensor oj; and strain tensor Sx. ‘The linear law between stress and strain was first stated by Hooke in the 17th century. If the solid is isotropic, after symmetry considerations, it is found that only two independent parameters exist in the stiffness tensor, such as the Young’s modulus and the Poisson's ratio v, or the Lamé constants \ and s. The tensor components can be written in terms of the Lamé constants as: cxjat = Myer + 1(5.ad;t + 5:95u) and the stress-strain relation is simplified as 04; = ASexdiy + 2WSi; (2.10) where Ske = Dy Sex V.- i, following the Einstein summation notation. The tilded notation Sky denotes the diagonal strain component which does not follow the Einstein sumnation notation in expressions, to be distinguished with the summation Sx, By substuting Eq. (2.2) and (2.10) into Eq. (2.5), the general elastic wave equation.19 which describes the displacement field i, is obtained as [L4, 143} “ =V-(uvu) +¥: (nS2) 2 yi eu) + ni) On; Here we can sce that all components of displacement i are coupled together in a general inhomogeneous solid. However, the displacement field i can be written as the sumn of the transverse field 7, and the longitudinal field i as a= +t, (2.12) with conditions V+ i =0 and V x i = 0, respectively. If the solid is homogeneous, ie. p, Nand scare all constants, then Eq. 2.11 is also decoupled into two independent equations: Pit, = Vt, (2.13) +2) Va, (2.14) for the transverse field 7, and longitudinal field dj, respectively. The transverse wave speed ¢; and longitudinal wave speed ¢; are inferred from Eq. 2.13 and Eq, 2.14 as = im [i [ (A+ Qe oS (2.15) Ve ? respectively. For a normal elastie solid, both \ and je are positive real numbers, thus we have ¢ > V2c Ifthe clastic solid has translational symmetry in one ditection, e.g. 2, then p, Aand p only depend on coordinates x and y. In this case, we limit the waves propagating20 to the ry plane, which is the most studied case in similar systems such as two- dimensional photonic crystals. In this case, the displacement field i also depends on -r, y coordinates only and its x and y components are still coupled together as Fu, Om; a o a PEt = vp: (wWru) + Ve (AGE) + GOVE tr). (216) an, ) * ae, Here ty = tu; + Guy and Vp = 22 + ag. However, the z component of the displacement field u, is decoupled from other components in Eq. 2.11 and is governed by the following scalar wave equation. (2.17) ional Thus the system reduces into two independent subsystems [Id]: a two-dimer vector subsystem which governs the vibrations in the ry plane, and a two-dimensional sealar subsystem which governs the vibrations parallel to the translational axis 2. The two subsystems are described by Eq. 2.16 and 2.17, respectively. In Eq. 2.16, both and ) appear, hence both the longitudinal and transverse waves exist and remain coupled if the subsystem is inhomogencous, ‘Thus the vibration mode in the vector subsystem is also referred as “mixed polarization mode”. In Eq. 2.17, only appears, thus only transverse waves exist in the scalar subsystem. Thus vibration mode in the scalar subsystem is also referred as “transverse polarization mode”. In the following studies, we will study physical cases with translation symmetry in one direction, stich as a system composed of nany parallel cylinders with infinite lengths. And we will focus on the two-dimensional vector subsystem since the mixed polarization mode is, nore interesting in physics and is absent from other physical systems sueh as photonic crystals. The subscript T in Vp and iy will be omitted for simplicity.2 For systems composed of fluids instead of solids, such as water in mercury, or air bubble in water, most of the above physical quantities and equations are still applicable. Under most circumstances, the viscosity in fluids can be omitted and the condition jr = c = 0 is assumed. As a consequence, both the shear stress o4j45 and the transverse field i, are nonexistent in inviscid fluids. The stress tensor o is reduced into a scalar quantity dy = AV i, which is also known as the fluid pressure p. Taking the divergence of both sides in Eq. 2.11, the acoustic wave eqnation can be expressed for pressure p as 1 # Lay (iv) (2.18) p If the system is homogeneous, then p and \ are constants. This leads to the standard scalar wave equation with wave speed cy = y/4 When viscosity is not neglectable, the wave equation in a homogeneous viscous fluid can be described by the following equation, which comes from dropping, the convective term in the Navier-Stokes equation for acousties [144] ve or ow at : : ras py (g. o8\ — at 7 y, Ot) 2.19 p AV (Vi) + (4y'/3-+ )V (v a) nvx (vx 7) (2.19) where 1’ and ji, ave, respectively, the dynamic coefficients of the shear and bulk vis cosities, The solution of this equation can also be written as the linear superposition of a Iongitudinal wave and a transverse wave, as the elastic wave equation in a ho- mogeneous solid. However, due to the first-order derivatives with respect to time # in the wave equation, the solutions will decay exponentially in viscous fluid, as will be shown below The wave frequency w is introduced into the above systems by writing the time dependence of physical quantities as fourier transforms of their spectra components.For example. displacement field @ or pressure field p are written as a= [ite Aexp(—iut)du; p(t?) = f olen exo(ietite (2.20) For the spectra component of frequency w, such as tw, 7) and p(w, 7), the above wave . equations are still valid, except that the operators & and § are replaced by ~w* and —iw, respectively, For example, Eq. 2.19 is now written as pit = AV (Vit) i (4/34 uh wV (Vi) +i- Wl x (Vx it). (2.21) By separating the transverse wave and the longitudinal wave into their independent wave equations, for the case of plane wave propagation, the dispersion relations have the forms ? —— _ for longi al waves, 2.22 = reaps fot longitudinal waves (2.22) 2 = the for transverse waves (2.23) i in which the longitudinal wave vector ky and transverse wave vector k; should choose the square root with positive imaginary parts, indicating the decaying of wave and energy along propagation path due to the viscosity in fluid. Eqs. 2.22 and 2.23 lead to the following energy decay lengths: 4/3y0 + ui) pa? 2+ (4/340 + 1) APB + 1) an x > Vi for transverse waves. (2.25) for longitudinal waves. (2.24) ps23 where the energy decays as exp(—ayr) and exp(—a,7) for longitudinal and transverse waves, respectively. Einstein summation notation is used in the above equations. And tilded notations are used to denotes vector or tensor elements which disobeys Einstein summation rule in expressions. 2.2 Variational principle As derived in Sec. 2.1, the time-independent elastic wave equation in an inhomo- gencous medium takes the following form: — peru; = 9 (uVay) +0 (52) sar er ay (2.26) The RHS of the above equation can be viewed as a self-adjoint operator operating on the displacement j. On the LHS, .? is the eigenvalue with the weighting function ~p. Thus the Sturm-Liouville theory and other physical principles in quantum physics are applicable in systems described by Eq. 2.26, including the variational principle. The variational principle can be described in the following way. Firstly, by mul- tiplying both sides of Eq. 2.26 by uf, suming, over index é and integrating over the real space, we obtain - [Pouiuate = fuv (uVuj) dr + [uv (ze 2 a . + fuse (AV adr (2.27) In most physical cases, the integration region on the RHS has either a periodic bound-4 ary condition, which is commonly seen in periodic physical systems, or a vanishing ‘Then field boundary condition, which is common at the infinity or hard boundarie the integrations on the RHS of Eq. 2.27 can be integrated by parts, which leads to foi dr = [ove Vu t Vu SE de + fav a? dr. 28) ‘Thus, the eigenvalue w® can be written as a functional of the displacement field @, fu (var Gu + UH =) are faiveatar [ovatar Similar to the variational principle in quantuin mechanics, for an eigenstate of a fixed ive. wo [a] (2.29) physical system, i.e. p, \ and jr are fixed functions, the eigenvalue w* should take an extremum value as a variation of the displacement finction a(?), [vs (ve vm + Vuy =) dr+ [x (IVa?) dr Ei [oat dr (2.30) This equation is equivalent to the elastic wave equation in principle and should be satisfied for all eigenstates, At a given eigenstate, we are interested in the change of the eigenvalue A(w2) when the physical parameters p(7), (7) and (7) are varied Wohen the variations of functions p(7), A(7) and i(7?) are small enough to be consideredas perturbations, the change in eigenvalue can be now written as pow (eure wae val 5) dr f aly adr fovatar At’) wo i] Aplii? dr i pli? dr for the perturbations Ap, AA, Aj and variation di. The terms including di are (2.31) cancelled in Eq. 2.31 due to Eq. 2.30. Thus Eq. 2.31 gives a simple formula to estimate the eigenfrequency shift as a consequence of the modification of system parameters and jr. This formula could also be obtained by a perturbative approach as shown in the paper [145] Eqs, 2.30 and 2.31 will provide the basic analytical tools to engineer the elastic band gaps in later discussions. Similar equations for acoustic waves are also deduced as (2.32) and (2.33) The physical meaning of Bq. 2.30 can be described in the following. Firstly, we write down the energy expressions of the elastic medium which is studied in elasticity theory. The kinetic energy density in clastic medium is26 The potential energy density in elastic medium is, 1 Ey = yewneSiy Sut = Asi) + nS) iw a+ vu Vurt area ee 2.35) aa (2.35) Then for a harmonic wave at frequency w, we find that the time-averaged kinetic and potential energy densities are just the integrands on the LHS and RHS of Eq. 2.28, respectively. Ey = Sela? (2.36) a H oe a E = giv +5 (Vu Vay + in V4) (2.37) ‘Thus Eq. 2.30 simply means that the time-averaged total kinetic energy equals the total potential energy in the physics sense, which is a basic statement for harmonic oscillations. For the convenience in energy densities discussions later in this thesis, here we s of de- decompose the elastic potential energy density E, according to different way composition, For one way, Ep is separated into the longitudinal pote jal energy density F1, which is induced by pure longitudinal plane waves, the transverse po- tential energy density £,, which is induced by pure transverse plane waves, and the interference potential energy density Ej, which is induced by the interference effects of longitudinal and transverse plane waves. This decomposition can be expressed in 20,598) the following. Since (V+ i)? = (Su)? = (Lj Su)? and (V x a)? = 203,thus 2, can be further written as 5 - (3 +n) LS HALA HM 388, +n 8 a a _(h 7 ai dit, Gai a+ geen ond (Fe TF In the above equation, tilded terms are elements that disobey the Einstein summa- tion notations. On the RHS of Eq. 2.38, the first, second and third terms represent longitudinal potential energy density E}, transverse potential energy density E, and interference potential energy density E), respectively. After time average over a pe- riod, we have 5) + By + B; (146), with d ee (p+) . (2.39) Bim BV xi (2.40) ai, BU, Di Bu; Oi; ii; A | ee ( Ox; Ox; x; (2.41) Sometimes the interference term £; could be very large in magnitude Another way of decomposition of E, is proposed by Landau [142]. Instead of longitudinal and transverse waves, he considers E,, as composed of the pure shear potential energy density E,, which is induced hy the shape deformation of the solid with its volume unchanged, and the hydrostatic compressional potential energy den- sity E.. which is induced by the volume change of the solid with its shape fixed. ‘The advantage of this decomposition is that no interference term exists for Ey and To derive the formula of E, and E,. we first introduce the zero trace stresses:28 4 — Su/d, such that Sf, = 0. Here d is the dimension of the system. By using this notation, E, can be expressed as Ey = Si)? + Q $8) ig = G+ ES HHL H+ LHD i ee Aen ae 1 = (5+ GVA + | 5(Vur- Vat ou 1 2 a VG | (2.42) The first and second terms on the RHS represent hydrostatic compressional potential energy density E. and pure shear potential energy density E., respectively. The corresponding time-averaged potential energy densities are (2.43) Vu) — 41V-a (2.44) Since this decomposition of E) into E, and E, is more simple (only two terms), we will use this decomposition for analysis of energy densities in our studies. 2.3. Scattering of waves by a single cylinder In Sec, 2.1, we have shown that a displacement field ican always be separated into the longitudinal and transverse components, ie., i = a + i, satisfying conditions V x a =0 and V- ii; =0. respectively. Thus. we can define two scalar potentials 629 and o as a = Vu (2.45) = Vxd=V x (2m). (2.46) Substituting Eqs. 2.45 and 2.46 into Eqs. 2.14 and 2.13, respectively, we get the wave equations for 1 and 1 in standard form Puy A+ Bory, au (2.47) Pu Hoey, ¥ 7 wwe (2.48) Now we consider wave scattering by a cylindrical scatterer located at the origin, For the spectra components of frequency w, the general solutions to Eqs. 2.47 aud 2.48 in two-dimensions for waves outside the scatterer can be expressed by Bessel and Hankel functions, i.e., Gy = Yo (anrdalherde” + GFE (kere) (2.49) YS (neta kere’ + Sas? (ker e™?) (2.50) ‘The coeflici its Qt Bats Outs Bur are the wave whee b= Tegan Tule coefficients corresponding to angular momentum quantum number 1 for longitudinal waves and transverse waves, respectively. In the physical sense, the part of Bessel functions denote waves incident to the scatterer, while the part of Hankel functions denote waves scattered away from the scatterer. Given the incident wave, the scat- tered wave is always determined by the boundary conditions of the scatterer, which is to be discussed in detail in this section,30 ‘The boundary condition s of a scatterer are influenced by the geometric shape and physical constitution of the scatterer and the background. In our study, we only consider the simplest geometric shape in two dimensions, ic., cylindrical seatterer For an elastic cylinder with parameters: density p, and Lamé constants A, and j1y inside an elastic background with different physical parameters pj, 5 and jy, the boundary conditions require that [78], urlin = Urlout olin = Ualout oorlin Orelouts Orlin Oraleuts $ Sy- And oy are both continuous which means the displacement @ and the stre across the boundary interface between the cylinder and the medium. Here subscripts in and out denote positions inside and outside the cylinder, respectively. Similarly, according to the physical considerations, the boundary conditions for an inviscid fluid cylinder in an elastic background are expressed in the following, trlin = Urlouts (2.55) Orelin = Frrlow (2.56) Salou = 0. (2.57) Here equations numbers are reduced to three, since for an inviscid fluid, no shear stress exists, thus the continuity of up across the boundary interface is broken. From the point of view of solving equations, the reduetion of equation number is also due to the reduction of the unknowns in the boundary equations. which correspond to31 transverse waves in the eylinder. Similarly, for the opposite case of an elastic cylinder in an inviscid fluid background, the boundary conditions are Urlin = Urloue (2.58) Orrin Orrlouts (2.59) orlin = 0 (2.60) And for the pure acoustic case of an inviscid fluid cylinder in an inviscid fluid back- ground, the boundary conditions are redneed again into Urlin Urlour (2.61) Frrlouts Orlin in the Huid pressure notation, due to another reduction of unknowns in the equa- tions, corresponding to the transverse waves in the background. In conclusion, in the id fluid, the displacement field up, previous three boundary conditions involving inv which is parallel to the boundary surface, does not need to be continuous, thus the total displacement field i also does not to be continuous across the boundary surface. ~ Ina These kind of boundary conditions are also called “slip boundary condition: real physical case, the fluid viscosity will always prevent from the slip boundary con- tic background. the bounda ditions. For example, for a viscous fluid cylinder in ela32 conditions turns into trlin = Urlout (2.65) Uolin Uolout (2.66) Orrlin + Trrlin = Frrlout (2.67) Tralin Srolout (2.68) Here 7;, and 7,9 are the viscosity stress tensors which are induced by the viscosity. In this boundary condition, equations number is restored to four and the displacement field @ is continuous across the boundary, since both shear stress and transverse ases, the boundary ide a viscous fluid cylinder. ‘Thus in real physical ip boundary conditious”. However, if the viscosity conditions are always “non small enough, like in some common fluids like air and water, it is known that only a thin layer of fluid will be ittached” to the boundary interface which support transverse components and make i continuous [147]. The major part of the fluid behaves as an inviscid fluid since the viscosity is very stnall. In this ease, the inviscid fluid and the slip boundary conditions are good approximations and are adequate for most purposes. By using proper boundary conditions, the scattered waves is related to the incident waves by the scattering matrix T of the scatterer. If we write the scattering coefficients33 Ont Ants Spt Snr into vector forms: on ann Bat =] a |. &=| on | %=] Bo Gm [+ (2.69) an oy Bu Bre then the scattered waves are related to incident waves by the following formula: & Tay a Sep ee eater (2.70) By a Ta Tu a For cylindrical boundary conditions, conservation of angular momentum requires that the sub-inatrices Ty, Ty. Ti and Ty are all diagonal matrices which indicate four kinds of scatterings events: the scattering of longitudinal waves into longitudinal waves, the seattering of longitudinal waves into transverse waves, the scattering of transverse waves into longitudinal waves and the scattering of transverse waves into transverse waves, respectively. It is obvious that through scattering of the scatterer, the longitudinal and transverse waves inside the background are coupled together If the bac ground is an inviscid fluid, then the above formula reduces into Tu (2.71) since the other quantities related to transverse waves, i.c., 5, &. Tis Ta and Ty nonexistent in the inviscid fluid background. In this case, the unknowns of the system (background) are halved from the case of elastic background, and thus the system is34 2.4 The multiple scattering theory (MST) ‘The multiple scattering theory (MST) [42,56,78-80] is a general numerical method that can be applied to calculate waves propagating and scattering in a system of many scatterers. Here we will firstly deduce the formulas for scalar waves and then consider the case of elastic waves in Sec. 2.4.3. The multiple scattering formalism is based on the physical picture that the incident waves to one scatterer are composed of the em and the seattered waves from other external incident wave from outside the sy scatterers in the system. For scalar waves, this idea can be expressed mathematically in the following: wite(F) = 07) + Do vy ow (7). 2) Here yi" denotes the total incident wave to the scatterer p: w® denotes the incident wave from outside the system and us* denotes the scattered wave from the scatterer q other than p. In two dimensions, it is known that the incident waves and scattered waves in the background near the scatterer p ean be expressed as wee) = Lan(p)nlkar(p)jene”. (2.73) YS Balp) Hl (or(p) ye". (2.74) we) Here r(p) and (p) are circular cylindrical coordinates of the position veetor #— fp. we speed A, denotes the position of the scatterer p. ky = w/cy, where op is the35 in the background medium. a,(p) and 3,(p) are the incident and scattered wave coefficients with angular momentwn quantum number n for the scatterer p, which are related by the scattering matrix T(p), Ap) = T(p)- a), ce s a as discussed in previous section. By using Graf's addition theorem, the scattered wave from scatterer q with angular momentum quantum number m to the scatterer p can be expressed as HD (hora = SO dnm (Pa) In(Ror(p) er", (2.76) where Gum (PQ) = Hn-m(Kergy oP, (2.77) are the vector structure constants, Here rg, denotes the distance between the scatterer pand q. 6,(p) denotes the angle between the x axis and the vector ,— Fp. In Eq. 2.76 the scattered wave from scatterer g with angular momentum quantum number am is rowritten into the sum of the incident waves to seatterer p with different angular momentuin quantum number n and coefficents gan(pq). Then the total incident waves to seatterer p from external and other scatterers can be written as Fe = Lad(psnlher(p) yet + 523° Bla) Hm (korg) = Lakrpsuthor(oe + 2 YS im lagnm(ea) Inf arp)” wm=e (1 EES loin) In(yr(p)yere? (2.78) pm Combining Eq. 2.78 and Eq, 2.75, a set of self-consistent equations are obtained terers for the scattered wave coefficients JF for all as the following equation. tp) = Tn) (+ Dem aa), P (2.79) oe Here G(pq) is the tensor composed by elements gym(pq) with m as the row index and mas the column index. This set of self-consistent equations can be written into the standard form of linear equations Ar = B. Thus the multiple scattering problem can be numerically solved, 2.4.1. Periodic case: the complex cell case For an infinite periodic lattice of scatterers in 2D, there is no incident wave thus we let 4 = 0. The self-consistent equations turns into Bip) = T(p) 2 G(r) 3a), p= 1-90. (2.80) oe First, we condier the simple cell case, If the basis of the lattice contains only one seatterer, then T(p) = T. p = 1.....90. Moreover, from the Bloch theorem, and assuming Ry = 0, we have 5g) = Fje*, (2.81) where f is the wave vector and R, is the position vector of the scatterer q. which also denotes lattice vector of the qth cell containing the scatterer q. Thus the self-ent equations Eqs. 2.80 reduce into (2.82) (r SL Gage** ) -H() va where J is the identity matrix. For a certain wave vector k, the condition that an eigenmode exists ean be written as (2.83) det [F-Seounet® -1| a This condition is only satisfied for certain eigenfrequencies w, gn = 1,2,... with wave vector f, Here n denotes the band number. The lattice sum ,,, G(1ge™™ can be obtained by using the method described in the Appendix, If the system basis contains more than one scatterers, say M scatterers per unit, cell positioned at 74,.... yy relative to the position vector of the cell, e.g. Ff, then the number of self-consistent equations is increased to A times the number for the simple cell case (basis containing only one scatterer). Here, the notations i and j denote the lattice cells instead of scatterers. If we denote the sth scatterer (1 < s Grs(U)e()e* s MM. (2.93) cA Tals) SO (Secawne ') - Hey) te NG43 4+Tuls) Do Gulsyeye" -H(s() Tul oe eae (yes pet) aut(.)) +Tuls) So G(s(1yt)e*™ - He(s(1))s M. (2.94) wa” And the SMCG method can be now written as ae Ap) = Tir) (stn SS Culpa) Hla) + Hilvi) > Gulia) - YG) xa) aehFein jay 1 ac jSein jay 9 ae +Ti(P) (a0 SS Gilpa) Bulg) + Helpe) S29) - Sia) (a), Ny (2.95) pal. . an : Hp) = Talr)- {P+ Y Gilva) - Gila) + Hal’) YO Gig) 7 Gia) - Hla) acnaein 7 o ae +Tu(P) (oo SS Gila) - Ba) + Hilpi) 2 Gt) SO Hea) ww). sien ier «i pel. (2.96) Here Gy, Hy and G;, H; corrpesond to the G and H for longitudinal and transverse waves, respectively 2.5 Summary In this chapter we have introduced the basic theory of elasticity, as well as some simple expressions for fluid with or without viscosity. ‘The elastic (acoustic) wave equation is empirically set up in the classic picture of Newtonian Mechanies. Com44 pared with the electromagnetic waves governed by the Maxwell equations, the elastic (acoustic) wave equation shows different structures and behaviors. By using the vari- ational principle, the engineering formula for the elastic (acoustic) waves is deduced, which gives instructions on how to modify the system parameters in order to shift the eigenfrequency of an eigenmode, The direct application of this engineering for- mula is to engineer the phononic band gaps in phononic crystals shown in the next chapter. The boundary conditions are discussed for various systems including solid (fluid) cylinders in solid (fluid) background, with or without viscosity in fluids. Th e boundary conditions are represented by the scattering matrices (T' matrices), which is used in the numerical method called the multiple scattering theory (MST). The MST inethod solves the problem of waves scattering between many scatterers. It is the basis for our unmerical calculations. For calculations of large-size finite samples, the sparse matrix canonical grid (SMCG) method is quite helpful in reducing compu- tational cost in both time and memory, thus much larger samples can be calculated. We will apply the SMCG method in the study of phononic quasierystals.Chapter 3 BAND-GAP ENGINEERING 3.1. Previous works In many periodic structures, the existence of a wide and good quality gap is the basis of all applications. However, such a gap can only be found under certain conditions. It is important to establish a systematic way to enlarge or open a gap for a given microstructure, Since the proposal of phononic crystals, people have been searching for the optimal configuration and engincering methods to open the ideal phononic band gap for potential applications and researches [11-29]. A brief review on these efforts is presented in Chap. 1, Due to the additional system parameters and the coupling of the longitudinal and transverse waves, the conditions t produce phononic band gaps are more complicated. The first and most studied system is the two-component simple cell phononic crystal. Complete band gaps have been observed in phononic crystals constructed with various materials. ‘The optimal configuration for acoustic band gaps were found to be low density, low velocity inclusions embedded in high density, high velocity matrix, like air bubbles in water. On the other hand, for clastic band gaps, the situation is more complicated. M. Siglas et al. found [12] that the optimal condition was obtained when high density. low velocity inclusions were embedded in low density, high velocity matrix. However. elastic band gap was found in a phononic crystal of carbon evlinders in epoxy matrix instead of the opposite [15]46 ‘Therefore, there isn’t a general rule about the best velocity configuration for an elastic band gap; rather, the contrast of elastic constants as well as the mass densities should be taken into account together. Recently there are some other ideas about the role of impedance instead of velocity [151]. In general, the existence of a large band gap relies heavily on the parameter contrasts and filling fraction of the inclusions. After the investigation on two-component simple cell phononic crystals. ‘There were some examinations on band structure properties of more complex phononic crystals. Caballero et al, have demonstrated both theoretically and experimentally that, acoustic band gaps can be increased by reducing the rotational symmetry [40]. For example, in a square lattice of rigid cylinders in air, they added a smaller rigid cylinder at the corner of each unit cell (the center of unit cell is positioned on the original rigid cylinders) to destroy the rotational symmetry. When the inserted cylin- der radius is increased to certain extent, they found the original gap in the original systein disappears while new gaps appear at higher frequencies. ‘This example will be referred later in our study, Complex cell structures, such as honeycomb struc ture [10], Suzuki phase [41], have all shown some advantages in opening ot enhancing band gaps. Liu et al. have taken au innovative step to propose the locally resonant materials: [43,44], which produce gaps based on local resonances instead of Bragg, scatterings. In their theoretical calculations and experiments, they used a composite consisting of a solid core material (lead) with relatively high density and a coating of elastically soft material (silicon rubber), and suecessfully opened up locally resonant gaps at low frequencies. In the sense of effective medium, close to resonance frequencies, the effective modulus turns negative. Thus “the material behaves like a total reflector been developed in two dimensions [45-47] {43}. Similar works hia Recently, Zhang ct al. have proposed a simple and efficient method [145] to engineer a photonic gap by adding a small fraction of a new component into each unit cell. The material of the new component as well as its insertion position are chosen according to the field-energy distributions of the Bloch states at the band edges. In the following, we will develop a similar method for engineering phononic band gaps in the phononie crystals. Our method also enhances band gaps by introducing a third component into the binary system, like the locally resonant materials. But the physical nature of our method is based on Bragg scatterings instead of local resonances. For comparison between photonic and phononic crystals later, we give the en- gineering formulas for two-dimensional photonic erystals here, which were derived through the perturbative approach, as )_ fac |B) 7 [« Bal a for p wave (3) and ar oe [|B eee m faa Here (7) is the periodic dicleetrie constant. E,; and D,z are the electrie and dis- for s wave. (3.2) ar placement eigenfield with Bloch wave vector F and band number n, respectively For a microstructure change denoted by A(e!) or Ac in the photonic crystal, e.g., s of another materials, the eigenfrequency shift can be estimated as —w,g from Eq. 3.1 or Eq. 3.2, for p waves or s waves, respectively Thus the engincering formulas Eq. 3.1 and Eq. 3.2 can be used as guidances for en- gineering photonic band gaps. This method is very general and its validity has beendemonstrated by microwave experiments, 3.2 Enlarging acoustic band gaps Unlike photonic crystals where the existence of gaps for electromagnetic waves depends only on the periodic arrangement of the dielectric constant (or light speed), the formation of gaps for clastic/acoustic waves in phononic erystals depends on the periodic arrangement of both sound speeds and mass density in a composite. Thus, the engineering of phononic band gaps is not only richer in physies but it is also more officient The acoustic wave equation for the propagation of pressure waves p,g(F) at fre- ty function p(F) and Lamé quency w,; ina phononie crystal with periodic mass den: constant function A(7) can be written as 2 pe a nee ye (vn) (33) Here, as in the photonic case, F is the Bloch wave vector and 1 is the band num- ber. The corresponding formula to estimate the effects of microstructure change for acoustie waves in phononic crystals can be obtained from the perturbative approach as shown in Eq. 2.33, Ifthe microstructure change is denoted by the change in paraineter functions as Ap(F)49 and AA(7}, then the effect of microstructure change on w,g can be directly estimated by A(1/p) = 1/9 - 1p = —Ap/(ap) and A(I/A) = 1/) - 1/4 = —AN/ (AY). as shown in Eq. 34. Here p = p+ Ap and A= A+ A) denote the new density function and Lamé constant function in the unit cell, respectively. Here we note that contrary to the equation for photonic crystals, Eq. 34 contains two terms. The first term comes from the factor 1/, which plays the role of 1/¢ in the p-polarized waves of the two-dimensional photonic erystals [145]; the second term comes from the factor 1/p in Eq. 3.4, which plays the role of ¢ in the s-polarized waves of the two-dimensional photonic crystals [145]. The presence of both terms with opposite signs makes the engineering of the band structures more efficient in an acoustic phononic crystal than in a photonic erystal, Now, suppose a sinall band gap exists inside the band structure of an acoustic phononic crystal. Let war and waar be the eigenfrequencies at the lower and upper band edges of the band gap, respectively, where L and U denote the corresponding mmetry points in the band structure. Due to the orthogonality condition of eigen- field functions, the corresponding eigenfield intensities |ppz|? and |pyr|?. as well as the gradient intensities [Vpyz|? and |Vpmc|?, will distribute mutually exclusively in the unit cell. Thus the engineering of band gaps by shifting the upper and lower band edges in opposite directions is feasible. Here is the simplest engineering case, which is similar to the ease of photonic crystals {145}. From the engineering formnla Eq. 3.4, by using proper inclusions with Ap = 0, AX #0 or inclusions with AA = 0. Ap 4 0 at the right places in the unit cell, we ean almost always shift the eigenfrequency of the one band edge in the preferred direction, while keeping the eigenfrequency of the other band edge nearly unperturbed, For example, with the inclusion of Ap > 0 and AX = 0, we can decrease50. wz While keeping wyc unperturbed by inserting the inchision in the unit cell where [as] is large and |ppe| is small. If we want to inerease wy, instead, we simply need to change the inclusion material to another type with Ap < 0. The feasibility of engineering is guaranteed by the exclusive distribution of |pmu| and |pyz| in the unit cell. In this case, only one term in Eq. 3.4 has taken effect. ‘The advantage of two terms in Eq. 3.4 is shown in the following case, in which we can shift the two band edges in opposite directions by a single type of insertion at the right places in the unit cell. For example, if we insert the inclusion of both positive Ap and positive Ad at the places where [pps] is large and |Vpac| is small, then the lower band edge wp, will be shifted downwards according to Eq. 34. At the same time, if the |pqu| is small and |Vp.v| is large at the insertion position, which is often the case due to exclusive distributions, the upper band edge w»; will be shifted upwards by Awyny according to Eq. 3.4. Thus the two band edges will be shifted in opposite directions, which makes the engineering of band gaps more efficient. If we denote the shift of wa, by Awnr and the shift of wn hy Awnu:, the change of the size of the band gap can be obtained as Aw, = Aww — Aur. The enhancing of a band gap by shifting both band edges with a single insertion is impossible to realize in the case of engineering photonic band gaps. In the following, we will demonstrate explicitly the feasibility of this engineering method through some concrete examples, 3.2.1 Steel cylinders in air host A periodic lattice of steel cylinders in the air background, is one of the most studied phononic crystals. For this system, it has been shown [78] that because the density5h contrast of steel and ait, ps/pe, i8 very large, the shear stress and trausverse waves inside the steel cylinders will not make a noticeable contribution to the scattering of the acoustic waves in the air background, and the scalar acoustic wave equation Eq. 3.4 is adequate to describe the scattering events of pressure waves at the steel cylinders. ‘Therefore, the variational principle and the engineering formula for acoustic waves are also adequate for this system. The system parameters are p, = 7.8 x 10kg/m°. Cis = 5960m/s for steel, and py = 1.2kg/m* and ci = 331m/s for air. And we choose a square lattice, 1.2 Reduced Frequency(«a/2nc,) ° ® x Figure 3.1: Calculated aconstic band structure for a square lattice of steel cylinders in an air background. The frequency is scaled by ¢9/a, in which ¢, is the sound speed in air, and a is the lattice constant The MST method for acoustic waves [55] is used to calculate the band structure w,j- gap positions and cigenfields uumerically. Fig, 3.1 shows the band structure52, with a filling fraction of 0.4 (2 = 0.3568a, where @ denotes the lattice constant). This result is similar to the band structure obtained by using plane wave expansion with R = 0.4180. In terms of the dimensionless frequeney, » = w/2mci, the first gap appears from 04,17 = 0.5528 to v2, = 0.6168. In Fig. 3.2, we plot the pressure field distributions |p(7)] and the gradients of pressure field distributions |Vp(7}] for Bloch states LM and 2X in a unit cell. The center of the unit cell corresponds to the center of the steel cylinder Figure 3.2; Pressure field distribution (left) |p|, and the gradient of the pressure field distribution (right) |Vp| in a cubic cell, of the band edge states LM and 2X in the steelair system whose band structure is shown in Fig, 3.1. The |p| fields are normalized to have a maximum value of 1. The center of the cell corresponds to the center of the steel eylinder We first consider the insertion of steel cylinders of radius Rj at the corner of each53 unit cell, as suggested in Caballero et als paper [40]. Thus, both Ap and AX are positive at the insertion position. However. from Fig. 3.2, we can see that |p(7)| is rs for the state at 1M, sinall for the state at 2X, and i large at the cell co just opposite for |Vp(7)|. ‘Thus the eigenfrequeney shift Aw estimated from Eq. 3.4 is positive for the LA state and negative for the 2X state. It should be pointed out that the eigenfrequency shift of the 2. state is verified to follow the trend of the 2X state and is always higher. This implies that there is a reduction of the gap ze, which is in qualitative agreement with the band structure calculation in Ref. [40]. In order to enlarge the gap, according to Eq. 3.4, we choose the insertion positions at the middle points of the four border lines of the unit cell. The estimated gap position from Eq. 3-4 is plotted in Fig. 3.3 as a function of R,/a as dashed lines. We have also performed the band structure calculations and plot the band edge positions as solid lines in Fig, 3.3. Comparing the perturbation results with the exact munerical calculation, we find that Eq. 3.4 can accurately reproduce the result of the band structure calculation at the 2X state, However, it is only qualitatively correct for the 1M state when R,/a > 0.02. At Ri/a > 0.1, the gap size is increased by a factor of 4a. 3.2.2. Mercury cylinders in water host We now consider another system, ic. a square lattice of mercury cylinders of a filling fraction 0.35 (radius @ = 0.3338a) in a water background (17]. Here both the density ratio pyx/pye = 13.171 and the speed ratio ¢iyn/ Ce = 0.9471 are much smaller than the ratios of the steel-air system. The band structure of the mereury-water system is similar to that of the steel-air system, as are the pressure field distributions54 0 01027 0.047 0106 00s 011 Ria Figure 3.3: The solid lines are calculated by the multiple-scattering method. The dashed lines are calculated using Eq. 3.4. (a) Gap structure after the insertion of steel cylinders with radins Rj. The original system is composed of a square lattice of steel cylinders with radius 0.35684 embedded in a background of air. ‘The insertions are at the middle points of the border lines of a unit cell as in Fig, 3.2. (b) Gap structure after insertion of mercury cylinders with radius R, . The original system is composed of a square lattice of mercury cylinders with radins 0.33384 embedded in a background of water. The insertions are at the middle points of the border lines of a unit cell as in Fig, 3.2. (c) Gap structure after the insertion of water eylinders with radius R, . The original system is composed of a square lattice of water cylinders with radius 0.292 embedded in a background of mercury. The insertions are at the corners of a unit cell as in Fig. 3.2.55 and their gradients at the band edge states (still LA/, 2X). The first gap appears from viyr(= wa/2mey) = 0.5929 to vex = 0.5970. In order to enlarge the gap, we insert mercury cylinders of radius R; at the middle points of the fonr border lines of the unit cell. In Fig, 3.3 we plot both the estimated results from Eq. 34 and the exact results from the band structure calculation as dashed and solid lines, respectively. Again, Eq. 3.4 correctly predicts the trends of the two band edges. Here, significant enlargement is found in this system. At R, = 0-La, the gap size is enlarged 33.2 times. 3.2.3 Water cylinders in mercury host ° © o ® Reduced Frequency(wa/2zc,,) 2 R o iy 0 a a Tr x Figure 3.4: Calculated acoustic band structure for a square lattice of water cylind ina mercury background. The frequency is scaled by c/a, in which ¢,, is the acoustic wave speed in mercury and a is lattice coustant.56, The two cases above are both systems of high-density scatterers in low-density backgrounds and thus exhibit similar properties. We also consider the case of a square lattice of water cylinders in a mercury background {17], with radius P = 0.29a Pol Pm = 0.076, and C,w/¢rm = 1.056. The band structure is shown in Fig. 3.4. The second band gap appears from v2,/(= wa/2r¢4m) = 0.8047 to vsx = 0.9278. The cigenfield distributions of 2M state and 3M state are plotted in Fig. 3.5, in which we find that [p(7] is lange at the cell sides for the state at 2, small for the state at 3M, for IVP(| and it is just opp are field distribution (left) |p, and the gradient of the pressure field ight) [Vp] in a cubic cell, of the band edge states 27 and 3A/ in the stem whose band structure is shown in Fig. 3.4. The |p| fields are ue of 1. The center of the cell corresponds to the Figure 3.5: Pt distribution ( water-inercur normalized to have a maxi center of the water cylinde ‘Thus, in order to enlarge this gap, we insert water cylinders of radius Rj at the57 four corners of the unit cell. The estimated eigenfrequency shifts of the 2M and 3 Bloch states (dashed lines), as well as the calculated gap edge positions (solid lines) are shown in Fig. 3.3. Again we find good agreement between Eq. 3.4 and band structure calculation. When R, > 0.1a, the gap size is inereased by 3.7 times. 3.3. Enlarging elastic band gaps In previous section we have successfully enlarged acoustic band gaps by using thid-component inclusions at proper places in the unit cell. Now we consider the clastic case. Similar to the case of acoustic phononie crystals, the engineering formula for elastic waves can also be deduced from the variational principle, or by using the perturbative approach. ‘The wave equation for Bloch waves of frequency w,;. with wave vector F in the nth band in clastic phononie crystals is, Ott = natty yg = V- (WPuyye) +0 C a aa QV: a,¢) (3.5) And the variational principle can be expressed in the physical way as, / where Eg and E,,,g are, respectively, the kinetic and potential energy densities. enh T= [Foca (3.6) which have the following expressions: (3.7) na? [a= Bengt Bank Ak te = Geb wea, oa Be, Vink (3.8) 1 HS (Vui ng: Vey ink 1 Be = AV he where E,,,¢ and E,,; are the hydrostatic compressional potential energy density and pure shear potential energy density, respectively. d here denotes the dimension of the system, And using Eqs. 3.7 and 3.8, Eq. 231 can be rewritten as [die (Sugg: Veaet Vege el = FIV tal) A ond f(a : ian) |V a)? ar [aolmlar [obival [olial? ar ‘Thus, we can engineer the band structure through similar process as the engineering dr for acoustic band gaps. In the following, we give an effective engineering example by using air inclusions. 3.3.1 Alum num cylinders in epoxy host Firstly, we demonstrate in the following example the creation of a large elas- tic band gap by using air inclusions. Consider a 2D phononic crystal consisting of a triangular lattice of aluminum cylinders in an epoxy host. The parameter con- trast in the system are found to be par/Pepory = 24333. Chepory/Ceepory = 24396, c1,at/Ctepary = 5.5603 and ¢4,11/Crepory = 2-758. When the radius of the aluminum cylinders yy = 0.2a, where a is the lattice constant. the 2-component system pos-59 x Tv J x Figure 3.6: Band structure of a phononic crystal consisting of a triangular lattice of aluninum cylinders with radius ra; = 0.2a, where a is lattic constant, in an epoxy host. The frequency is sealed by ¢ epory/d. Where Cepory iS the transverse wave speed in epoxy. The inset demonstrates the phononic exystal configuration.60 sesses no gaps, The band structure calculated by the multiple-scattering method (56) is shown in Fig. 3.6, where the dimensionless frequency, fa/crepory: is used. The inset shows the phononie crystal configuration. Then we insert air cylinders of radii 0.20 in epoxy host just between each pair of nearest-neighbor aluminum cylinders. Considering the air as inviscid fluid and using the multiple scattering method, the band structure for the 3-component system is calculated and shown in Fig. 3.7. The inset demonstrates the phononic crystal configuration after insertion of air cylinders. Comparing with Fig. 3.6, the first three bands are enormously shifted downwards. The higher bands are also greatly altered, but not shifted downwards as the first three bands. As a result, a large absolute elastic band gap with gap/midgap ratio Aw/w, = 0.72 is opened. In Fig. 3.7, it is also seen that some flat bands emerge inside the gap. They are plotted as dashed lines in the band structure. These bands are induced by local resonances of air cylinders, and most of which will be dimished considering the viscosity in air, Thus, these flat bauds will have insignificant effects on the quality of the gap created in Fig, 3.7. The detailed proof will be shown later. In order to understand the physical reason behind the creation of large band gaps with air inclusions, i.e. the suppression of the first three branches, we plot, respectively, in Fig. 3.8, Fig. 3.9 and Fig, 3.10 the energy density distributions of the eigenstates LX, 2X and 3P in the original ALepoxy system. The energy densities are renormalized to the maximuin kinetic energy density in the primitive cell. The chosen eigenstates are marked by arvows in the band structure of Fig. 3.6. It is interesting to see from Figs. 3.8(a), 3.9(a) and Fig. 3.10(a) that the kinetic energy densities of LX, 2X and I" states concentrate in the aluminum cylinders, whereas, the corresponding potential energy densities are concentrated in the epoxy host between nearest-neighbor aluminum cylinders as can be seen from Figs. 3.8(b), 3.9(b) and61 Figure 3.7: Band structure after insertions of air cylinders with radius rai = 0.20 into the phononic crystal described in Fig. 3.6. The band gap is marked in grey. The dashed lines are the flatbands induced by locally resonant states in air cylinders. The inset demonstrates the phononic crystal configuration with air inclusions.62 Fig. 3.10(b). In order to shift these three eigenfrequencies downwards to create a gap. we can, from Eq. 3.6 and Eq. 3.9, either increase the mass density, p, inside or around moduli, \ the aluminum cylinders where the field, |@2, is high, or decrease the elas and j1, in the epoxy background where the potential energy density, Eis high. The former way is generally achieved by increasing the mass density and filling fraction of the scatterers. Since the elastic modulus of air is about 5 orders of magnitude sinaller than those of epoxy, by inserting air cylinders between the aluminum cylinders, we have effectively lowered the elastic moduli in the epoxy background and shifted the cigenfrequencies downwards, as shown in Fig. 3.7. As an explicit verification of the above analysis, the kinetic and potential energy density distributions for the 2X and BP states after air insertions are plotted in Fig. 3.11, The potential energy densities ax and E,,ar are obviously squeezed out from the imiddles between pairs of nearest neighbor aluminum cylinders. Another interesting phenomenon in the energy density distributions shown in Figs 3.8, 3.9 and 3.10 is that for all three eigenstates LY, 2X and 3D, the hydro- static compressional potential energy, E,. shown in Figs. 3.8(c), 3.9(c), and 3.10(c) is inch smaller than the pure shear potential energy, E,, shown in Figs. 3.8(d) 3.9(d), and 3.10(d), For the first and third branches, whieh correspond to transverse acoustical branch and the rotational acoustical branch (To be discussed in Sec, 3.3.2), respectively, E, always dominates in E,, For the second branch, i.e. the longitudinal acoustical branch, E, dominates for small area, while E, catches up and exceeds E, as k is increased, and ends up more than 10 times larger than £, at the Brillouin ‘one boundaries, This indicates that near the band edges of the three acoustical branches, the pure shear potential energy E, dominates in E,. Thus we should be able to shift the cigenfrequencies downwards with any fluid inclusions besides air, as63 os 0.13 a os os 00 0.0 063 04 078 oa “ 109 08°54 00 04 08 (a) oe. 04. a 00. 2 04 08 08 04 00 (©) Figure 3.8: (a), (b), (c) and (d) show respectively, the kinetic, the potential, the hydrostatic compressional potential and the pure shear potential energy density dis- tributions Exax, Epax. Eoax and E,x for the 1X state in Fig, 3.6. The energy densities are renorm The graphs cover a rectangular area including a whole alwninum cylinder at (0,0), and halves of six aluminum. cylinders at (-1,0), (0,1), (-1/2, V3/2), (-1/2,-v3/2). (1/2, v3/2) (1/2, ~v3/2) in units of « lized to the maximum kinetic energy dei64 08. 0 0.13 4 0.25 038 00} oso 068 oA. 078 ose 1.00 08. 08 04 00 04 08 08 04 00 04 (b) 08 04 00 04 08 08 04 00 04 08 () (a) Figure 3.9: (a), (b), (c) and (a) show respeetively, the kinetic, the potential, the hydrostatic compressional potential and the pure shear potential energy density dis- tributions Eyax. E; and E,2x for the 2X state in Fig. 3.6, x08 04 00 04 (a) (©) Figure 3.10: (a), (b), (c) and (d) show respectively, the kinetic, the potential, the hydrostatic compressional potential and the pure shear potential energy density dis- tributions Exar. Byars Boar and Bar for the 3° 3.6.Figure density distributions Ey2y and Ey,2x for the 2X state in F 08. o 0.13 a 0.25 0.38 00. 0.50 083 a 075 0.88 1.00 08 08 04 00 08 08 08 04 00 04 08 (c) 3.11: (a) and (b) show respectively 08 o oa loss 131 00. 175 219 04 263 3.08 350 48 08 04 09 04 08 (b) 08 lo 11.00 oe 200 3.00 00. Pi s00 B50 04. 609 7.00 08 8.09 08-04 00 04 08 (d) the kinetic and the 66 potential energy 3.7; (c) and (d) show respectively, the kinetic and the potential energy density distributions Exar aud Ep st for the 3P state in Fig. 3.7.67 1 =0 for any fluid. Below we demonstrate the insertion of water cylinders instead of air cylinders in the original ALepoxy system. The band structure calculated using the multiple scattering method is shown in Fig. 3.12. As can be seen, the acoustical branches are indeed shifted downwards. However the higher optical branches are also shifted downwards, thus resulting in a smaller band gap. Obviously, air is a better choice than other fluids, which have much larger mass density, for the creation of large band gaps. x . J x Figure 3.12: Band structure after insertions of water cylinders with radius ruater = 0.2a into the phononic crystal described in Fig, 3.6. The insertion positions are the same as the air cylinders in Fig. 3.7. The band gap is marked in grey. 3.3.2 Rotational acoustical branch68 From the physics point of view, the first two bands corresponds to the acous- tical branches in the solid state physics. The first band corresponds to the trans- ical verse acoustical branch. The second band corresponds to the longitudinal acous al lattice branch. While the third band does not have a correspondence in the cr vibration modes, As is seen in Fig. 3.10(a) and 3.10(b), in the Al-epoxy system, the kinetic energy density Exar concentrates inside the aluminum cylinders and behaves as a quadratical function of r reaching the maximum at the Al-epoxy interface, here ris the distance to the cylinder center; the potential energy density E,,sr, however, concentrates inside the epoxy host precisely between each pair of nearest-neighbor aluminum cylinders. Moreover, from Fig, 3.10(c) and 3.10(d), we can find that the potential energy density is almost pure shear, with Br 4 orders of magnitude smaller than E,ar. After combining and analyzing the above phenomena, we come to the conclusion that the aluminum cylinders do in phase oscillating rotations. ‘The epoxy inedium between any pair of aluminum cylinders is simply shear-deformed and thus producing restoring forces for the alunimun cylinders. For a clear demonstration, the displacement fields of 2X and 3D states in the ALepoxy system at a given time are plotted in Fig. 3.13(a) and 3.13(b), respectively. Obviously, for the 3D state, each aluminum cylinder and its near by epoxy in the Wigner-Seitz cell rotates as a whole. ‘Thus the 3P state can be regarded as a rotational acoustical state and the corre- sponding third band can be regarded as a generalized rotational acoustical branch. Different from the transverse and longitudinal acoustic branches, the frequency of ro- tational acoustic branch does not go to zero when the wave vector goes to zero. The existence of three acoustical branches is a special phenomenon in clastic phononic crystals,69 Figure 3.13: (a) and (b) show respectively, the displacement fields ix and figr at a given time for the 2X and 3P states in Fig. 3.6. The graphs cover a rectangular area including a whole aluminum cylinder at (0,0), and halves of six aluminum cylinders t (-1,0), (0,1), (—1/2, V3/2), (-1/2, -V3/2), (1/2. V3/2). (1/2, -V3/2) in units ofa. 3.3.3 Flat bands In the band structure, the flat bands are actually induced by local resonances of air cylinders, This can be verified in the following way. Considering the boundary condition of air inclusions as hard boundary conditions.i.e. u,(r = Rj) = 0, then the cigenfrequency for a local resonance can be obtained from the following condition: dn kair?) =0. 3.10) dr 2 i ) r= Ry where kgir = 2/¢aire From the Eq, 3.10, dimensionless eigentrequencies wn m/(21Cepary) which denotes the mth solution of angular momentum quantum nunber n can be ob- tained. We can easily verify that, the first eigenvalues for n = 0, -£1, £2, +3. ie, w/(2nCrepary) = 0.8937, 0.4294, 0.7123, 0.9798. are exactly the frequencies of the four sets of flat-bands in the band structure, respectively. As a further examination:70 the cigenfield maps |a|? are also calculated and plotted in Fig. 3.14 and Fig. 3.15, respectively. In Figs. 3.14(a), 3.14(b), 3.14(¢) and 3.14(d), we plot the displacement field map ||? corresponding to the first eigenvalue with angular momentum quantum number m = 0, 1, 2 and 3 respectively 04 04 02 02 00 00 02 02 “04 04 04-02 00 02 04 04 0.2 00 02 04 (a) (b) 04 a i 02 02 00 0.0. 02 02 0.4 04: 0.4 0.2 00 02 04 04 -0.2 00 02 04 (c) (d) ve 3.14: (a),(b),(c) and (d) denote respectively, the displacement field naps for the first cigenvalues of an air cylinder with hard bounday, at angular momentum The dotted white lines denote the hard quantum number n = 0, £1, +2, poundaries In Figs. 3.15(b), 3.15(c), and 3.15, we plot the displacement field map [@/?, with wave vector k = 0 at frequencies w/ (2c; epory) = 0.8937, 0.4239 and 0.7123, respec- tively. Fig. 3.15(a) is the corresponding configuration map of the aluminum and air cylinders for Figs. 3.15(b), 3.15(c), and 3.15. The correspondence between the states 3.14 and Fig. 3.15 is evidentoA On%, ed D 00 05 10-05 00 (c) Figure 3.15: (a) denotes the configuration of the field maps. (b),(c) and (d) denote respectively, the displacement field naps for the flat bands with wave veetor F = 0 in Fig, 3.7, at frequencies w/(21¢;epary) = 0.8937, 0.4239 and 0.7123, which corresponds ton =0 £1.42, respectively72 ‘The presence of these flat bands will produce very sharp peaks in the transmission spectrum. However, due to the viscosity in air, it can been shown that these sharp peaks can be smeared out completely, as long as the dwell time of the system at, resonance is much larger than the absorption time due to air viscosity. In Fig. 3.16 we show the band structure of the same structure as in Fig. 3.7, but with a smaller air cylinder radii rai, = 0.17a. The corresponding transmission spectra calculated by the multiple-scattering method for a sample of size 10a x 33a with rai. = 0.17a are shown in Figs. 3.17(a) and 3.17(b) for, respectively, the 'J and PX directior 1.2 0.6 $l, epory 0.4 0.2 © 0 x r J x Figure 3.16: Band structure after insertions of air cylinders with radius reje = 0.174 into the phononic crystal described in Fig. 3.6. The band gap is marked in grey. The dashed lines are the flatbands induced by locally resonant states in air cylinders. The inset demonstrates the phononic crystal configuration with air inclusions. ‘The existence of a sharp resonant transmission at fa/cyepory = 0.838 is due to the flatband inside the band gap, which corresponds to the eigenmode of the air cylinder3 5 vy rx 3 B ic IND B10? 5 | S10 F | @ MEF &) ee & ie 3 Ee 2 | 2 | & | Fao L ee eves Oe 0 02 04 06 08 1 Fic eciame ec erst eer eae fear. fA/C, epoxy (A=16M) FAC, ssoxy (@=10M) Figure 3.17: (a) and (b) show the transmission spectra of a 10a x 3a sample of the phononic crystal described in Fig. 3.16 in, respectively, the PJ and the CX directions. (c) and (d) are the transmissions when viscosity is included in the air inclusions and ais chosen as lem with angular momentum quantum number m = 2. The width of the transmission resonance is about Afa/cyepory © 10~, which corresponds to a dwell time of At © 10'a/crepory- Such a sharp resonance will not survive when viscosity is inchided in the air cylinder, When viscosity is considered in the air cylinder, the wave is governed by Bg, 2.19 which was introduced in Sec. 2.1 [144,147]. The slip boundary conditions are now changed into the non-slip bomdary conditions, whieh was also discussed in Sec. 2.3. The change of boundary conditions will introduce a transverse component into the eigenmodes concentrated in a thin layer near the surface of air cylinder [147] except for the eigenmodes with m = 0 state, e.g. the one at fa/cyepory = 1-051 The absorption time. At,, arising from the transverse component can be estimated from the dispersion relation Eq, 2.23. Using the value of f = 0.838¢,epary/a. we find the dwell time is about four orders of magnitude larger than the absorption time74 and the resonant transmission will he smeared out, completely by absorption. To check this, we have recalculated the transmission spectra with the shear viscosity of the air component included in the calculations. In our calculations, we take pi, = 0 and p/p = 0.15em?s~! [147]. When a = Lem, the transmission results are shown in Figs. 3.17(c) and 3.17(d). Indeed, the resonant transmission inside the gap disappears entirely, leading to a large transmission gap. 3.3.4 Gap maps Figure 3.18: Gap map for different filling fractions of aluminum and air cylinders in the phononie crystal consisting of a triangular lattice of aluminum cylinders in epoxy host with air inchisions. The lines with triangles down and up indicate upper and lower band gap edges, respectively.co) 0.05 0.1 0.15 02 la Figure 3.19: Gap map for different filling fractions of steel and air cylinders in the phononic crystal consisting of a triangular lattice of steel cylinders in epoxy host with air inclusions. The lines with triangles down and up indicate upper and lower band gap edges, respectively76 Above we have discussed the creation of elastic gap by fluid inclusions. Due to the small mass density and elastic modulus, air is the optimal choice to open up the fi st band gap. Below, in Fig, 3.18, we draw a gap map for different filling fractions of air and aluninum cylinders in the 3-component Al-epoxy-air phononic cryst The radii of air rair and aluminum ray are limited by ryir < 0.20, rat-+ Pair S O-4a to avoid close touch of cylinders As we can see, for a fixed rar, the band gap increases With Tair. Even for the case of rar = 0, a band gap can still be created when reir is sufficiently large. In Fig. 3.19, we plotted the gap map for the system of steel cylinders in epoxy in a triangular lattice with air inclusions. Much larger gaps are found in the Fé system than in the Al-epoxy-air system. In fact, in the poxy-air configuration of rre = 0.2a aud rir = 02a, we are able to obtain a band gap with gap/midgap ratio Aw/w, = 0.93, while our system is umeh lighter than the normal 2-component. phononic crystals. 3.3.5 Extension to phononic quasicrystals Since the principle of suppressing, the acoustical branches by air inclusions is very general and successfull for elastic phononic crystals, we try to apply it to the case of quasierystals {83, 152]. In order to identify a gap in a quasicrystal, we calculate the radiation power of a circular sample quasicrystal by using the multiple seatter- ing method, In Fig, 3.20, we present the radiation power of a 12-fold symmetric quasicrystal consisting of aluminum cylinders in epoxy host, with or without air in- clusions, ‘The 12-fold symmetric quasiperiodic lattice is generated by the Stampfii inflation rule [153]. Here. again, an air cylinder is inserted between each pair of nearest-neighbor aluminum cylinders. The inset shows the quasicrystal sample with10° 10 * epoxy Figure 3.20; Radiation power for the circular samples of phononie quasicrystals con- sisting of a 12-fold symmetric quascrystal of alunniwn cylinders in epoxy host, with or without air inclusions. ‘The radii of akuminnm cylinders 741 = 0.2a, where a is nearest-neighbor distance of the quascrystal. The sample radius R controls the size of the quasicrystal. ‘The inset demonstrates the quasicrystal configuration with air inclusions.07 [ae 0208 Ar ec 170.288 1. 0.5 | = 0.05 0.1 0.15 02 rad Figure 3.21; Gap map of a 12-fold symmetric phononic quasicrystal consisting of aluminun cylinders in epoxy host with air inclusions at different filling fractions of aluininum, The lines with triangles down and up indicate upper and lower band gap edges, respectively79 03 0 0.05, o4 0.15) 02 Figure 3.22; Gap map of a 12-fold symmetric phononic quasicrystal consisting of steel cylinders in epoxy host with air inclusions at different filling fractions of steel The lines with triangles down and up indicate upper and lower band gap edges, respectively80 air inclusions, The sample radius is R = 6a, where a is the nearest-neighbor distance of the 12-fold symmetric quasicrystals. The cylinder radii of aluminum and air are rar = 0.20 and rir = 0.2a, respectively. It can be seen that the original ALepoxy quasierystal possesses no band gap in the frequency region of 0 < fa/crepary < 0.9. After air cylinders are inserted, a gap appears from fa/crepory = 0.6 to 0.8. More- over, when the sample size is increased from radius R = Ga (121 Al cylinders and 300 air cylinders) to R = 8a (217 Al cylinders and 564 air cylinders), the radiation power inside the gap decreases exponentially, thus verifying the existence of this gap in the quasicsystal. Below in Fig. 3.21 we plot the gap map for different filling frac- tions of aluminum and air cylinders in Alepoxy-air phononic quasicrystal. The radii of air cylinders, raie, and aluminum cylinders, rai, are again limited by raix < 0.24 Tart Tair $ Oda to avoid close touch of eylinders. The results show that insertion of ps. Com air cylinders in phononie quasicrystals is also capable of creating large pared with the gap map for the phononic crystal of triangular lattice in Fig. 3.18, we find the upper edges of the band gap also shift downwards as the lower edges do, when the radii of air cylinders increase. Thus these gaps are smaller than the previously created gaps in phononic crystals, When the radii of the aluninmn cylinders become 0.2a. For comparison, in smnaller than 0.12a, there is no gaps created, even with rn, Fig. 3.22, we also plotted the gap map for the Fe-epoxy-air phononie quasicrystals at different filling fractions of the steel. Compared to Fig. 3.21, larger gaps are obtained as expected, 3.4. Summary In this chapter, we have implemented the method of gap engineering discussed81 in Chap. 2 for both acoustic and elastic band gaps. We have also demonstrated that the methods are successful and efficient in enlarging an existing band gap or creating a new band gap for phononic crystals. In the case of elastic band gap. an interesting phenomenon of rotational acoustic branch is discussed, which is a characteristic behavior of phononic crystals. Gap-maps are calculated and we find even in an homogeneous medium, gaps can still be created by inserting air inclusions. Finally, we have shown that the method of inserting air inclusions is also effective in the opening of an elastic gap for phononic quasicrystals. The existence of a gap in quasicrystals is known to depend on the short-range order. However, when the sample of a quasicrystal is increased, long-range order sets in. Such long-range order can alter the gap structure found in small samples. How the gap structures evolve with ple size is an interesting topic which has not been explored before. This will the say be the focus of our discussions in the next chapter.Chapter 4 PROPERTIES OF GAP STRUCTURES AND WAVE FUNCTIONS IN PHONONIC QUASICRYSTALS 4.1 Previous works Although quasicrystal was found only two decades ago [2], its essential differences, with both periodic or random structures have attracted wide interests. Photonic or phononic quasicrystals were proposed just a few years ago and many interesting results have been found. A brief review on the previous works of photonic and phononic qnasicrystals has been given in Sec. 1.2, especially for two-dimensional quasicrystals, However, due to difficulties in calculations, the gap structures and wave functions in photonic and phononic quasicrystals have only been studied in small-size systems, while it is generally known that many interesting phenomena of quasicrystals appear only in large or infinite systems. With the help of the SMCG method [148-150], we are able to calculate larger samples of phononic (photonic) quasicrystals numerically This advantage stimulates us to make a deeper investigation on the gap structures and wave functions of the phononic quasieryst As mentioned in Sec, 1.2, the one-dimensional quasicrystals have been investi- gated both mathematically and numerically. The existence of singular continuous self-similar spectra and critical states, which are the essential differences with the periodic or random systems. have been manifested by the transfer matrix and dy- 8283 namical mapping methods, for both the electronic tight-binding model and photonic (phononic) quasierystals. It is believed that most eigenstates in one-dimensional quasicrystals are critical, i.e. neither extended all over the lattice as are the Bloch states of Anderson states, nor exponentially localized as are the strongly localized localization (95,96, 103]. The typical spatial distribution of a critical state is decay- ing practically to zero and recovering to large values then decaying again alternately: Some of the critical states are proved to be localized in a power law with a self-similar structure. Others are chaotic or random. However, exceptions of extended states have been found at certain energies rigorously and all extended, critical and localized states have been observed in numerical results [96]. Here we emphasize an interesting fact that the singular continuous spectra and the critical behaviors of the wave functions do not vanish no matter how small the quasiperiodic potential is, but the potential strength affects only the value of exponents. All of the above phenomena have been verified in numerical calculations and the formation of self-similar spectra and critical states can be understood by the renormalization group idea [87] ‘The case is different in two or higher dimensions, From the Anderson localiza- tion problem, we know that the one-dimensional system is quite special. Some of the results obtained in one dimension can uot be applied to systems in high dimen- sions. Especially, one-dimensional quasicrystals have lost the “topological” nonperi- odicity [86], which is a characteristic feature of high-dimensional quasicrystals. This topological nonperiodicity is believed to induce a remarkable effect on the electronic proper frustration of the wave functions, In comparison with the one-dimensional quasicrystals, the two- and higher-dimensional quasierystals have been studied by fewer researchers, Despite the many rigorous r sults obtained by the transfer matrix and dynamical mapping methods in general one-84 dimensional quasiperiodic systems [95, 103], mathematically rigorous results in two dimensions have only been found for some particular eigenstates located at particular energies. Firstly, some multifractal eigenstates [98, 105] are obtained by constructing a tight-binding Hamiltonian and setting the transfer energies according to a complex, rule, However, this procedure is quite artificial, it is not sure whether there are mul- tifractal eigenfunctions in general simple models, such as nearest-neighbor hopping models. ondly, it is known that the nearest-neighbor tight-binding models may lead to some degenerate states, called confined states and string states [97,99]. Con- fined states are strictly localized in some special regions and the degeneracy is due to Conway's theorem. String states exist on certain strings of rhombuses of the Penrose lattice. Besides these rigorous results, there are also many numerical works on finite structures or quasicrystal approximants [86, 100, 101, 103, 106, 109). All three kinds of wave function behavior, i.e. extended, critical (intermediate) and localized states. have been numerically observed in a two-dimensional Penrose lattice [109] Tsrmetsugu ef al. have made some intuitive arguments [86] on quasicrystal They expect that quasierystals have an unsmooth energy spectrum and localized wave functions, but the total band-width will be generally nonzero in more than one dimension, Here nnsmooth energy spectrum means that the level spacings have an anomalous system-size dependence AE x 1/N? (7 # 1). Based on # thought experiment of the real-space renormalization-group theory, they argued that wave |, but would functions cannot be localized exponentially in qasierystals in genera show power-law decays instead. ‘The power-law decays are caused by competition of the two effects in quasicrystals — nonperiodicity and self-similarity (Conway's theorem). The phononie (photonic) structures bear many differences with the electronic85 tight-binding model. For example, in the tight-binding method, only the couplings between several neighbors are considered; while in calculations of phononic (photonic) structures, eg, by the MST method, we have to include all the couplings between two scatterers no matter how far they are. A finite system in the tight-binding model is a closed system, while a finite sample of the phononic (photonic) structure is an open system with outgoing flux. Kafesaki et al. tried to interpret the band structure results for elastic and acoustic waves by analogy with the LCAO approach [23]- They have found two kinds of propagation modes. For the first kind, the wave propagates mainly through the host material; for the second kind, the wave hops coherently from a scatterer to its neighbors at the resonances in the single scattering cross section, The second mode can be viewed as the analog of the linear combination of atomic orbitals (LCAO) in electronic propagation, while the first mode cannot be explained by analog with the LCAO approach. Thus the phononic (photonic) structures exhibit richer features than in the tight-binding model. In fact, the origin of some gaps are nnuch better explained by the Bragy scatterings and multiple scatterings. On the other hand, the idea of phononie (photonic) erystals has been successfully implemented and the familiar phenomena in crystals such as band gaps, Bloch waves, etc,, have been observed in phononic (photonic) crystals. ‘Thus we also expect to see sone interesting phenomena in phononic (photonic) quasicrystals, which were firstly observed in the tight-binding model for electrons. To the best of our knowledge, the self-similar spectra as well as the critical behavior of wave functions have not been observed or studied in two-dimensional phononic (photonic) quasicrystals yet. In the following, with the help of the SMCG method, we will investigate two main yinmnetric features in a two-dimensional phononic quasicrystal consisting of a 12-fold s quasiperiodic lattice of aluminum cylinders in air background. One feature is about86 the gap structures in phononic crystals; the other feature is about the properties of wave functions. In our system, the sound wave energies are expelled out of the aluminum eylinders because of the high density contrast between aluminum and air Thi jewed as rigid cylinders and the transverse waves s the aluminum cylinders can be inside can be ignored without inducing any noticable changes in the background fields, ‘The main cause of gap structures is due to Bragg-like scatterings. In this respect, our system is fundamentally different from the tight-binding model studied for electrons. The 12-fold symmetric quasiperiodic lattice is firstly proposed by Oxborrow ef al, in 1993 [153]. The seed (parent) structure of this quasiperiodic lattice is a dodecagon consisting of square-triangle tilings. And the quasiperiodic lattice can be generated through an inflation process. For each generation in the inflation process, the of the system is enlarged (2 + V3) © 3.7 times in length. This 12-fold symmetric quasiperiodic system has been found good for the realization of gaps in spectrum (128, 130). 4.2 Gap structures due to short-range order Consider a circular sample of radius R,. Inside this circular domain, the eylinders are arranged in a quasiperiodic order. ‘The system is excited by a line source at a fixed frequency located near the center of the sample. Here the source is put a little away from the exact system center in order to stimulate asymmetric states inside the sample. The far-field radiated cnergy flux at any angle @ , s(@), can be calculated by the multiple scattering method [56]. from which the total radiation power P, can be obtained. Since P, includes the transmnission from all directions, the spectrum of P, can be used to determine the position of a complete gap. For frequencies insideReduced Frequency p contrast 0 vos ve 2 10" 10 10 7 a ii Figure 4.1: (a) Acoustic band structure for a square lattice of rigid cylinders in air (b) Corresponding radiation power background, with cylinder radius R = 0.380. Solid line and dotted line correspond to the square lattice in (a) with sample sizes R, = 8a and R, = 12a respectively. Dashed line corresponds to the triangle lattice in (c) with sample size Ry = 8a. (c) Acoustic band structure for a triangle lattice of rigid cylinders in air background, with cylinder radius R = 0.380. The reduced frequency is scaled by c/a, in which ¢, is the sonic wave speed in air, and a the lattice constant.88 ‘a complete gap, the density of states is zero, which in tum gives a divergent input impedance and vanishing P,. Therefore, a dip in the spectrum of P, represents a complete gap of the system. ee ih ho & 10 | fi i I 5 Ki FEeeeale S10" }) Weel ace 1 \ | i! iH 10? tt tt | i ht o . iH ' 10" eet sop | | uae | \ y ee ae ae Ose rose aos taster A ior fees 2 Reduced Frequency Figure 4.2: Radiation power (Ry = 8a for solid line, R, = 12a for dashed line, and R, = 16a for dotted line) for 12-fold symmetric quasicrystal of rigid cylinders (R = 0.380) in air background, The quasierystal structure is shown in the inset graph. Four major gaps is marked by arrow In order to show the validity of using radiation power to locate the position of a complete gap, we first apply the method to a square lattice with cylinders of radius R = 0.384, where a is the lattice constant. The band structures calculated by the multiple scattering method are shown in Fig. 4.1 (a). The first complete gap is located between symmetry points LAY and 2X at frequencies f = 0.5263 and f = 0.6299, respectively, Here we use a reduced frequency, f = wa/2ney . where cy is the speed89 of sound in air. The second gap appears from f = 1.8725 to f 2.0079. Figure 4.1 (b) plots, in solid and dotted curves, the frnetion P,(f) calculated from two circular samples of radii R, = 8a and 12a, respectively. It is clearly that the positions of the dips in P,(f) coincide excellently with the gaps in the band strnetures. Furthermore, the depth of a dip increases with the sample size. We have repeated the calculations for the case of a triangular lattice. The band strnetures and the function P,(f) are plotted in Figs. 4.1 (c) and 4.1 (b) (dashed curve), respectively. Again their gap positions coincide excellently. Next we study the gap structures of a 12-fold symmetric quasiperiodic system [128, 130, 153] with cylinder radius 0.38a (here a represents the distance between nearest neighbors). Such a system exhibits a square-triangle tiling as shown in the inset of Fig, 4.2. The function P,(f) calculated from two samples of sizes R, = 8a and RR, = 12a are shown as solid and dashed curves in Fig. 4.2, respectively. The fimnetion Pf) obtained from a 12-fold symmetric quasierystal shows many more structures than that obtained from the periodic lattices shown in Fig. 4.1 (b). ‘This suggests that a 12-fold symmetrie quasicrystal is capable of supporting more gaps than a square or triangular lattice does, Here we concentrate on the four major dips marked by arrows int Fig, 4.2. The depth of each dip increases with the sample size, whereas its position remains unchanged. Thus, these results strongly indicate that there exist at Jeast four gaps in the system for frequencies below f = 2 (f = wa/2req). We have further increased the sample size to R, = 16a and recalculated the function P,(f) around these four gaps. The result is shown with dotted lines in Fig. 4.2. It confirms the robustness of these gaps. In addition to the total radiation power, we have also calculated the transmission coefficient T() of the same system by using the multiple scattering methods, To cal-90 Transmission coefficient 3 10 (a) 10° aa 7 UTUAY MM ae atel wim 4 i ete $ Fa 2 Vy 3 iM 8 Mt 510% | yt 4 : B19? E10 (b) 10" - 02 07 12 17 22 Reduced Frequency Figure 4.3: Transmission coefficients (W = 20a and d = 4a for solid lines, Ww 0a and d= 10a for dashed lines) for the 12-fold symmetric quasicrystal of rigid cylinders (R= 0.382) in air background, The dips represent partial gaps in the corresponding direction. (a) — J direction (b) T — X direction, Four major gaps is marked by arrows.91 culate the transmission coefficient, we cut a rectangular sample from the quasicrystal instead of a circular sample. ‘The source is prepared by passing a plane wave at a fixed frequency f through a slit, which is parallel to the wider side of the rectangular sample. ‘The transmission coefficient is defined as the ratio of the total flux of the transmnitted wave F, to that of the incident wave Fj. The flux F; can be obtained analytically and is proportional to the width of the slit, whereas the flux F, can be obtained by integrating the near-field energy poynting vector on the output surface lated from of the rectangular sample. Figure 4.3, plots the transmission spectra cal two samples with the same width 1 but different thickness d (W’ = 20a and d= 4a and 10a) in both the PJ (Fig. 4.3 (a) and PX (Fig. 4.3 (b)) directions. Similar to Fig. 4.2, Fig, 4.3 shows many transmission dips, The positions of the four major gaps found in Fig, 4.2 are also marked in Fig. 4.3 by arrows. Both Figs. 4.3 (a) and 1.3(b) show the existence of transmnission dips at these positions, reconfirming the existence of the complete gaps predicted by the radiation power. In addition to these four major gaps, there could exist other sinaller gaps in this system. ‘The existence of such a multigap structure is an inherent property of quasicrystals (130, 134). For 2-dimensional quasicrystals, structures possess self-similar behavior under the change of length scales in addition to the rotational syinmetry. Due to such scaling property, the spectra of such systems can also exhibit self-similar behavior. Such fractal-type spectra have been widely scen in 1D quasierystals. The multigap structure we have seen here in sinall samples is a result of the scaling property in quasicrystals. ‘The presence of multigap structure shown in Figs. 4.2 and 4.3 could make 12-fold sym- mnetrie quasicrystals useful as multiband filters. Finally, we study the gap distribution of the 12-fold symmetric quasiperiodic sys- tem as a function of filling fraction. The resulting gap map associated with the four