Thesis - Open Research Exeter (ORE)
Transcription
Thesis - Open Research Exeter (ORE)
An Analytical and Numerical Investigation of Auxeticity in Cubic Crystals and Frameworks T. P. Hughes Submitted by Thomas Peter Hughes, to the University of Exeter as a thesis for the degree of Doctor of Philosophy in engineering, August 2012. This thesis is available for Library use on the understanding that it is copyright material and that no quotation from the thesis may be published without proper acknowledgement. I certify that all material in this thesis which is not my own work has been identified and that no material has previously been submitted and approved for the award of a degree by this or any other University. (Signature).............................. 1 Abstract Negative Poisson’s ratio, or auxetic, materials present the possibility of designing structures and components with tailored or enhanced mechanical properties. This thesis explores the phenomenon of auxetic behaviour in cubic crystals using classical and quantum modelling techniques and assesses the validity of these techniques when predicting auxetic behaviour in cubic elemental metals. These techniques are then used to explore the mechanism of this behaviour. The findings of the atomistic modelling are then used as a template to create networks of bending beams with tailored Poisson’s ratio behaviour. 2 Contents 1 Background: Properties of materials 22 1.1 Negative Poisson’s ratio materials . . . . . . . . . . . . . . . . . . . . . . . 22 1.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 1.1.2 Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 1.1.2.1 Simple Crystals . . . . . . . . . . . . . . . . . . . . . . . . 24 1.1.2.2 Silicates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 1.1.2.3 Paratellurite . . . . . . . . . . . . . . . . . . . . . . . . . . 26 1.1.2.4 Metal organic frameworks . . . . . . . . . . . . . . . . . . 27 Fabricated structures . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 1.1.3.1 Honeycombs . . . . . . . . . . . . . . . . . . . . . . . . . . 27 1.1.3.2 Foams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 1.1.3.3 Molecular Auxetics . . . . . . . . . . . . . . . . . . . . . . 29 1.1.3.4 Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 1.1.3.5 Granular materials . . . . . . . . . . . . . . . . . . . . . . . 31 1.1.3.6 Applications of mono-crystalline structures . . . . . . . . 31 Investigative Methodologies . . . . . . . . . . . . . . . . . . . . . . 31 1.1.4.1 Crystallography . . . . . . . . . . . . . . . . . . . . . . . . 32 1.1.4.2 Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . 32 1.1.4.3 Molecular Modelling Techniques . . . . . . . . . . . . . . 33 1.1.4.4 Analytical Modelling . . . . . . . . . . . . . . . . . . . . . 34 1.1.3 1.1.4 1.1.4.4.1 1.2 Rotating Rigid Units . . . . . . . . . . . . . . . . 35 Other unusual material behaviour . . . . . . . . . . . . . . . . . . . . . . . 36 3 1.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 1.2.2 Negative Thermal Expansion . . . . . . . . . . . . . . . . . . . . . . 36 1.2.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 1.2.2.2 Crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 1.2.2.3 1.2.2.4 1.2.3 2 1.2.2.2.1 Cubic crystals . . . . . . . . . . . . . . . . . . . . 36 1.2.2.2.2 Metal Organic Frameworks . . . . . . . . . . . . 37 1.2.2.2.3 Other framework structures . . . . . . . . . . . . 37 Mechanisms of NTE . . . . . . . . . . . . . . . . . . . . . . 39 1.2.2.3.1 Increasing symmetry . . . . . . . . . . . . . . . . 39 1.2.2.3.2 Positive expansion of bonds . . . . . . . . . . . . 39 1.2.2.3.3 Electron valence transition/cation movement . . 40 1.2.2.3.4 Rotation of rigid units . . . . . . . . . . . . . . . 40 Methodologies . . . . . . . . . . . . . . . . . . . . . . . . . 40 1.2.2.4.1 Experimental . . . . . . . . . . . . . . . . . . . . . 40 1.2.2.4.2 Analytical Modelling - Rotating unit modes . . . 41 1.2.2.4.3 Molecular modelling . . . . . . . . . . . . . . . . 41 Negative Linear Compressibility . . . . . . . . . . . . . . . . . . . . 42 Background: Theoretical framework and computational modelling 43 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.2 General elasticity equations for stress and strain . . . . . . . . . . . . . . . 43 2.2.1 Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.2.1.1 Mandel notation . . . . . . . . . . . . . . . . . . . . . . . . 44 2.2.1.2 Voigt Notation . . . . . . . . . . . . . . . . . . . . . . . . . 44 2.2.2 The strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 2.2.3 The stress tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 2.2.4 Generalised Hooke’s Law . . . . . . . . . . . . . . . . . . . . . . . . 48 2.2.5 Conversion between stiffness and compliance . . . . . . . . . . . . 49 2.2.6 Rotations of constants . . . . . . . . . . . . . . . . . . . . . . . . . . 50 2.2.7 Relationship between elastic constants and engineering constants . 53 2.2.7.1 54 Young’s Modulus . . . . . . . . . . . . . . . . . . . . . . . 4 2.2.7.2 Poisson’s ratio . . . . . . . . . . . . . . . . . . . . . . . . . 54 2.2.7.3 Shear Modulus . . . . . . . . . . . . . . . . . . . . . . . . . 55 2.2.7.4 Bulk Modulus . . . . . . . . . . . . . . . . . . . . . . . . . 55 2.2.7.5 Zener ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 Application to auxetic media . . . . . . . . . . . . . . . . . . . . . . 56 Classical atomistic modelling techniques . . . . . . . . . . . . . . . . . . . 58 2.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 2.3.2 Two Body Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 2.3.2.1 The Morse Potential . . . . . . . . . . . . . . . . . . . . . . 59 2.3.2.2 The Lennard-Jones Potential . . . . . . . . . . . . . . . . . 60 2.3.2.3 The Cauchy relation . . . . . . . . . . . . . . . . . . . . . . 60 Many Body Potentials . . . . . . . . . . . . . . . . . . . . . . . . . . 61 2.3.3.1 ’Basic’ many-body potentials . . . . . . . . . . . . . . . . . 61 2.3.3.2 Embedded atom model for metals . . . . . . . . . . . . . . 62 2.3.3.3 The Finnis-Sinclair potential . . . . . . . . . . . . . . . . . 63 2.3.3.4 Sutton-Chen potential . . . . . . . . . . . . . . . . . . . . . 64 2.3.3.5 The Cleri-Rosato potential . . . . . . . . . . . . . . . . . . 64 Quantum Atomistic Modelling Techniques . . . . . . . . . . . . . . . . . . 66 2.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 2.4.2 Application to auxetic media . . . . . . . . . . . . . . . . . . . . . . 66 2.4.2.1 Selected applications of DFT to find elastic properties . . 66 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 The finite-element method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 2.2.8 2.3 2.3.3 2.4 2.4.3 2.5 3 Fundamental atomistic modelling of the elastic properties of cubic crystals 70 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 3.2 Modelling methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3.2.2 Derivation of elastic energy . . . . . . . . . . . . . . . . . . . . . . . 71 3.2.3 Strain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 3.2.3.1 73 Calculation of bulk modulus . . . . . . . . . . . . . . . . . 5 3.2.3.2 Calculation of tetragonal shear modulus . . . . . . . . . . 74 3.2.3.3 Calculation of rhombohedral shear modulus . . . . . . . . 74 3.2.3.4 Summary: Finding elastic properties from strains . . . . . 75 3.2.3.5 Poisson’s ratio . . . . . . . . . . . . . . . . . . . . . . . . . 75 Geometry considerations for cubic crystal systems . . . . . . . . . . 76 3.2.4.1 Face-centred cubic unit cell . . . . . . . . . . . . . . . . . 76 3.2.4.2 Body-centred cubic unit cell . . . . . . . . . . . . . . . . . 78 Elastic properties calculated from analytical modelling . . . . . . . . . . . 81 3.3.1 A simple model: volume dependence . . . . . . . . . . . . . . . . . 81 3.3.1.1 Bulk modulus . . . . . . . . . . . . . . . . . . . . . . . . . 81 3.3.1.2 Tetragonal shear . . . . . . . . . . . . . . . . . . . . . . . . 82 3.3.1.3 Rhombohedral shear . . . . . . . . . . . . . . . . . . . . . 82 3.3.1.4 Poisson’s ratios . . . . . . . . . . . . . . . . . . . . . . . . . 83 3.3.1.5 Material properties from volume potential . . . . . . . . . 83 A simple pair potential: The spring model . . . . . . . . . . . . . . 84 3.3.2.1 84 3.2.4 3.3 3.3.2 3.3.2.2 3.3.3 Application to a face-centred cubic unit cell . . . . . . . . 3.3.2.1.1 Bulk modulus: . . . . . . . . . . . . . . . . . . . . 84 3.3.2.1.2 Tetragonal shear: . . . . . . . . . . . . . . . . . . 84 3.3.2.1.3 Rhombohedral shear: . . . . . . . . . . . . . . . . 85 3.3.2.1.4 Poisson’s ratio: . . . . . . . . . . . . . . . . . . . . 85 Application to the body-centred cubic unit cell . . . . . . 86 3.3.2.2.1 Bulk modulus: . . . . . . . . . . . . . . . . . . . . 86 3.3.2.2.2 Tetragonal Shear modulus: . . . . . . . . . . . . 86 3.3.2.2.3 Rhombohedral shear modulus: . . . . . . . . . . 86 3.3.2.2.4 Poisson’s ratio: . . . . . . . . . . . . . . . . . . . . 87 3.3.2.3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 3.3.2.4 Combination of the spring and volumetric potentials . . . 87 The Lennard-Jones Potential . . . . . . . . . . . . . . . . . . . . . . 89 3.3.3.1 89 Face Centred Cubic cells . . . . . . . . . . . . . . . . . . . 3.3.3.1.1 First nearest neighbours . . . . . . . . . . . . . . 6 89 3.3.3.1.2 3.3.3.2 3.3.3.3 3.4 92 Body centred cubic cells . . . . . . . . . . . . . . . . . . . . 96 3.3.3.2.1 First nearest neighbours . . . . . . . . . . . . . . 96 3.3.3.2.2 Second nearest neighbours . . . . . . . . . . . . . 98 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 3.3.3.3.1 First nearest neighbours . . . . . . . . . . . . . . 101 3.3.3.3.2 Second nearest neighbours . . . . . . . . . . . . . 102 3.3.3.3.3 Influence of potential range . . . . . . . . . . . . 102 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 3.4.1 4 Second nearest neighbours . . . . . . . . . . . . . Summary of calculated properties . . . . . . . . . . . . . . . . . . . 105 Atomistic modelling of the elastic properties of cubic crystals 4.1 4.2 106 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 4.1.1 Elemental cubic metals . . . . . . . . . . . . . . . . . . . . . . . . . . 106 4.1.2 Covalently bonded zincblende compounds . . . . . . . . . . . . . . 107 Classical Potentials: Methodology . . . . . . . . . . . . . . . . . . . . . . . 109 4.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 4.2.2 Numerical modelling methodology . . . . . . . . . . . . . . . . . . 109 4.2.2.1 Calculation of Poisson’s ratio for elemental cubic metals . 110 4.2.2.1.1 Optimisation method . . . . . . . . . . . . . . . . 111 4.2.2.1.2 Direct measurement method . . . . . . . . . . . . 111 4.2.2.2 Parameter space exploration . . . . . . . . . . . . . . . . . 112 4.2.2.3 Modelling of a coupled three-body/harmonic system . . . 113 4.2.2.3.1 Fitting of potentials to covalently bonded cubic crystals . . . . . . . . . . . . . . . . . . . . . . . . 114 4.3 Elastic properties calculated from classical potentials . . . . . . . . . . . . 115 4.3.1 Experimental data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.3.2 Elastic and engineering constant data from potential models . . . . 117 4.3.2.1 Accuracy of calculated properties . . . . . . . . . . . . . . 117 4.3.2.2 Morse Potential . . . . . . . . . . . . . . . . . . . . . . . . 118 4.3.2.2.1 Analysis: Morse Potential . . . . . . . . . . . . . 119 7 4.3.2.3 Finnis-Sinclair potential . . . . . . . . . . . . . . . . . . . . 121 4.3.2.3.1 4.3.2.4 Sutton-Chen potential . . . . . . . . . . . . . . . . . . . . . 123 4.3.2.4.1 4.3.2.5 4.3.3 4.3.4 Analysis: Sutton-Chen Potential . . . . . . . . . . 124 Cleri-Rosato potential . . . . . . . . . . . . . . . . . . . . . 125 4.3.2.5.1 4.3.2.6 Analysis: Finnis-Sinclair . . . . . . . . . . . . . . 122 Analysis: Cleri-Rosato Potential . . . . . . . . . . 126 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 Deformed structures calculated from classical potentials . . . . . . 127 4.3.3.1 Body-centred cubic crystals . . . . . . . . . . . . . . . . . 128 4.3.3.2 Face-centred cubic crystals . . . . . . . . . . . . . . . . . . 128 Variation of predicted properties with phase space exploration . . 131 4.3.4.1 Morse potential . . . . . . . . . . . . . . . . . . . . . . . . 131 4.3.4.2 Finnis-Sinclair potential . . . . . . . . . . . . . . . . . . . . 133 4.3.4.3 Sutton-Chen Potential . . . . . . . . . . . . . . . . . . . . . 134 4.3.4.4 Analysis: Variation of predicted Poisson’s ratio with phasespace exploration . . . . . . . . . . . . . . . . . . . . . . . 135 4.3.4.5 4.3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . 136 Elastic constants calculated from coupled three-body/harmonic system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 4.3.5.1 Influence of the relationship of three-body and axial forces on predicted properties . . . . . . . . . . . . . . . . . . . . 139 4.4 5 Can classical potentials predict auxetic behaviour? . . . . . . . . . . . . . 140 Density Functional Theory modelling of elastic properties of cubic elemental metals 141 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141 5.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142 5.2.1 5.3 Convergence example . . . . . . . . . . . . . . . . . . . . . . . . . . 142 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 5.3.1 LDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 5.3.2 GGA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 8 5.4 6 5.3.3 PAW LDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 5.3.4 PAW PBE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156 5.3.5 Magnetic moment consideration . . . . . . . . . . . . . . . . . . . . 163 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 Finite Element Modelling of cubic frameworks 168 6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 6.2 From crystalline bonds to beam networks . . . . . . . . . . . . . . . . . . . 169 6.3 Analytical derivation of mechanical properties of cubic beam structures . 171 6.4 6.3.1 Assumptions used in analytical modelling . . . . . . . . . . . . . . 171 6.3.2 Type I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 6.3.3 Type II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174 6.3.4 Type III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 6.3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 Finite element modelling methodology . . . . . . . . . . . . . . . . . . . . 178 6.4.1 Element type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178 6.4.2 Mesh convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179 6.4.3 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 180 6.4.3.1 Representative volume element . . . . . . . . . . . . . . . 181 6.4.3.2 Panel model . . . . . . . . . . . . . . . . . . . . . . . . . . 181 6.4.3.2.1 6.4.4 Comparison of RVE and Panel Model . . . . . . 182 6.4.3.3 Material Properties . . . . . . . . . . . . . . . . . . . . . . 182 6.4.3.4 Loading conditions . . . . . . . . . . . . . . . . . . . . . . 182 High throughput techniques . . . . . . . . . . . . . . . . . . . . . . 183 6.4.4.1 Varying material properties . . . . . . . . . . . . . . . . . 183 6.4.4.2 Automatic generation of beam networks from crystallographic structure data . . . . . . . . . . . . . . . . . . . . . 183 6.4.4.3 6.4.5 6.5 Calculation of properties . . . . . . . . . . . . . . . . . . . 184 Validation of finite element modelling with analytical model . . . . 185 Results of finite element modelling . . . . . . . . . . . . . . . . . . . . . . . 186 6.5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 9 6.5.2 Comparison of panel model and RVE . . . . . . . . . . . . . . . . . 187 6.5.3 Elastic properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189 6.5.3.1 Poisson’s ratio . . . . . . . . . . . . . . . . . . . . . . . . . 191 6.5.3.2 Young’s modulus . . . . . . . . . . . . . . . . . . . . . . . 193 6.5.3.3 Shear modulus . . . . . . . . . . . . . . . . . . . . . . . . . 195 6.5.3.4 Comparison with [100] axis . . . . . . . . . . . . . . . . . 196 6.5.3.5 Summary of results . . . . . . . . . . . . . . . . . . . . . . 197 6.5.3.6 Mechanical performance of beam networks relative to conventional honeycomb . . . . . . . . . . . . . . . . . . . . . 198 6.5.3.7 7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . 200 Discussion 7.1 7.2 7.3 201 Synopsis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201 7.1.1 Classical Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202 7.1.2 Ab-initio modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 7.1.3 Mechanical modelling . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Further Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 7.2.1 Exploration of auxetic behaviour in crystals . . . . . . . . . . . . . . 206 7.2.2 Beam networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207 Key Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208 A Methodology of DFT 210 A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 A.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210 A.2.1 The Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . . . 210 A.2.2 Born Oppenheimer Approximation . . . . . . . . . . . . . . . . . . 211 A.2.3 Hohenberg-Kohn Theorem and Kohn-Sham Theorem . . . . . . . . 211 A.2.4 Self consistent loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212 A.2.5 Hellman-Feynman Theorem . . . . . . . . . . . . . . . . . . . . . . 214 A.2.6 Choice of Pseudpotentials . . . . . . . . . . . . . . . . . . . . . . . . 215 A.2.7 Periodicity, reciprocal space and k-point mesh . . . . . . . . . . . . 215 10 A.2.8 Basis, plane waves and energy cutoff . . . . . . . . . . . . . . . . . 216 A.2.8.1 Magnetic moment consideration . . . . . . . . . . . . . . . 216 B The Finite Element Method 219 B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 B.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 B.2.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 B.2.2 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 B.2.2.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220 B.2.2.2 Material properties . . . . . . . . . . . . . . . . . . . . . . 221 B.2.2.3 Analysis type, boundary conditions and loading . . . . . 221 B.2.3 Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221 B.2.4 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222 C Derivation of off-axis elastic constants from on-axis deformation using the finite element method 223 11 List of Figures 1.1 Schematic orthographic projection of non-auxetic, auxetic and completely auxetic material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.2 The auxetic mechanism of a re-entrant honeycomb . . . . . . . . . . . . . . 28 1.3 The auxetic mechanism of one cell of a tetrachiral honeycomb . . . . . . . 28 2.1 Schematic showing normal and shear stresses . . . . . . . . . . . . . . . . . 47 2.2 Angles used to describe rotated properties . . . . . . . . . . . . . . . . . . 51 2.3 Schematic representation of a classical potential energy function . . . . . . 59 2.4 Schematic showing a two-body (axial), and three-body (angular spring) potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 3.1 A face centred cubic unit cell . . . . . . . . . . . . . . . . . . . . . . . . . . 76 3.2 Strains applied for c’ shear in an FCC structure. . . . . . . . . . . . . . . . 77 3.3 Strains applied for c44 shear in an FCC structure. . . . . . . . . . . . . . . 78 3.4 A body centred cubic unit cell . . . . . . . . . . . . . . . . . . . . . . . . . . 78 3.5 Strain for C‘ shear for body-centred cubic, (110) plane . . . . . . . . . . . . 79 3.6 Shear displacement for c44 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 3.7 Variation of the (11̄0, 110) Poisson’s ratio with change in cutoff, for an arbitrary FCC system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 3.8 Variation of the lattice constant (normalised relative to the lattice constant for the first nearest neighbour value) with change in cutoff, for an arbitrary FCC system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103 3.9 Variation of the (11̄0, 110) Poisson’s ratio with change in cutoff, for an arbitrary BCC system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 12 3.10 Variation of the lattice constant (normalised relative to the lattice constant for the first nearest neighbour value) with change in cutoff, for an arbitrary BCC system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 4.1 Relative orientation of a body-centred cubic unit cell . . . . . . . . . . . . . 111 4.2 Schematic representation of the Keating model . . . . . . . . . . . . . . . . 113 4.3 Structure of covalently bonded zincblende type crystal structure . . . . . . 114 4.4 Deformation mechanism of body-centred cubic crystal . . . . . . . . . . . 128 4.5 A ’side’ view and a ’top’ view of the deformed unit cell . . . . . . . . . . . 129 4.6 The ’external’ deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 4.7 The hinging of the inter-layer bonds . . . . . . . . . . . . . . . . . . . . . . 130 4.8 The ’internal’ deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 4.9 ν with variation of Morse potential α . . . . . . . . . . . . . . . . . . . . . . 131 4.10 ν with variation of Morse potential r0 . . . . . . . . . . . . . . . . . . . . . . 132 4.11 ν with variation of Finnis-Sinclair potential density . . . . . . . . . . . . . 133 4.12 ν with variation of Finnis-Sinclair potential density . . . . . . . . . . . . . 133 4.13 ν with variation of Sutton-Chen potential density . . . . . . . . . . . . . . 134 4.14 ν with variation of Sutton-Chen potential repulsive component . . . . . . 134 4.15 Poisson’s ratio (ν110,11̄0 ) and relative spring stiffnesses calculated from the three-body/harmonic model for selected body-centred cubic elemental metals136 4.16 Poisson’s ratio (ν110,11̄0 ) and relative spring stiffnesses calculated from the three-body/harmonic model for selected covalently bonded cubic metal compounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 5.1 Convergence of elastic constants with change in energy cutoff . . . . . . . 143 5.2 Poisson’s ratio in the off-axis direction for cubic elemental metals, calculated from LDA pseudopotential . . . . . . . . . . . . . . . . . . . . . . . . 147 5.3 Poisson’s ratio in the off-axis direction for cubic elemental metals, calculated from GGA pseudopotential . . . . . . . . . . . . . . . . . . . . . . . . 151 5.4 Poisson’s ratio in the off-axis direction for BCC elemental metals, calculated from PAW-LDA pseudopotential . . . . . . . . . . . . . . . . . . . . . . . . 154 13 5.5 Poisson’s ratio in the off-axis direction for FCC elemental metals, calculated from PAW-LDA pseudopotential . . . . . . . . . . . . . . . . . . . . . . . . 154 5.6 Poisson’s ratio in the off-axis direction for BCC elemental metals, calculated from PAW-PBE pseudopotential . . . . . . . . . . . . . . . . . . . . . . . . . 162 5.7 Poisson’s ratio in the off-axis direction for FCC elemental metals, calculated from PAW-PBE pseudopotential . . . . . . . . . . . . . . . . . . . . . . . . . 162 6.1 Representative unit cell for the simple cubic, body-centred cubic and facecentred cubic cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170 6.2 Structures under consideration . . . . . . . . . . . . . . . . . . . . . . . . . 171 6.3 Type I structure, [110] strain . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 6.4 Free body diagram of simplified structure . . . . . . . . . . . . . . . . . . . 173 6.5 Type II structure, [100] strain . . . . . . . . . . . . . . . . . . . . . . . . . . 174 6.6 Load configuration for type II structure ([110] direction), side view . . . . 174 6.7 Load configuration for type II structure ([110] direction), top view 6.8 Type III cubic structure, [110] strain. . . . . . . . . . . . . . . . . . . . . . . 175 6.9 Type III cubic structure, [110] strain . . . . 175 . . . . . . . . . . . . . . . . . . . . . . 176 6.10 Schematic showing the deformed shape of a cantilevered beam . . . . . . 179 6.11 Evolution of Poisson’s ratio with mesh density for an arbitrary type II beam network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 6.12 Schematic showing the boundary conditions applied for axial and shear loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 6.13 Representative volume element . . . . . . . . . . . . . . . . . . . . . . . . . 181 6.14 Poisson’s ratio of structure for all considered cases, for strains in the [110] direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 6.15 In-plane deformed configuration of panel model . . . . . . . . . . . . . . . 188 6.16 Deformed shape of the RVE model . . . . . . . . . . . . . . . . . . . . . . . 189 6.17 Representative unit cell for the structure with two dominant types of beams 190 6.18 Poisson’s ratio of structure for cases where two types of beams are dominant191 6.19 Tensile modulus of structure for cases where two types of beams are dominant193 6.20 Shear modulus of structure for cases where two types of beams are dominant195 14 7.1 Possible network structures of MOF5 and α-cristobalite. . . . . . . . . . . 207 A.1 Flow diagram of DFT process . . . . . . . . . . . . . . . . . . . . . . . . . . 213 C.1 Poisson’s ratio in the (110, 11̄0) directions . . . . . . . . . . . . . . . . . . . 225 15 List of Tables 3.1 Elastic constants calculated for each of the potential models . . . . . . . . 105 3.2 Poisson’s ratio calculated for each of the potential models using the elastic constants shown in table 3.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 105 4.1 Poisson’s ratio maxima and minima for elemental cubic metals . . . . . . . 108 4.2 Varied potential parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 4.3 Elements used as basis for parameter space exploration. 4.4 Experimentally observed elastic properties of elemental metals . . . . . . 116 4.5 Elastic constants for elemental metals, calculated using Morse potential . 118 4.6 Engineering constants for elemental metals, calculated using Morse potential119 4.7 Elastic constants for elemental metals, calculated using Finnis-Sinclair po- . . . . . . . . . . 113 tential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 4.8 Engineering constants for elemental metals, calculated using Finnis-Sinclair potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 4.9 Elastic constants for elemental metals, calculated using Sutton-Chen potential123 4.10 Engineering constants for elemental metals, calculated using Sutton-Chen potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 4.11 Elastic properties for elemental metals, calculated using Cleri-Rosato potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4.12 Engineering constants for elemental metals, calculated using Cleri-Rosato potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 4.13 Elastic properties for body-centred cubic metals, calculated using Threebody/harmonic model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 16 4.14 Elastic constant data from Three-body/harmonic model for covalently bonded cubic crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138 5.1 Experimentally observed elastic constant data for elemental cubic metals. Data taken from Landolt-Bornstein (1985). Poisson’s ratio in the [110] direction calculated from these data. . . . . . . . . . . . . . . . . . . . . . . 144 5.2 Elastic properties of elemental metals calculated using LDA pseudopotentials. 5.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 Engineering constants of elemental metals calculated using LDA pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146 5.4 Elastic properties of elemental metals calculated using GGA pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 5.5 Engineering constants of elemental metals calculated using GGA pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 5.6 Elastic properties of elemental metals calculated using PAW-LDA pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152 5.7 Engineering constants of elemental metals calculated using LDA-PAW pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 5.8 Elastic properties of elemental metals calculated using PAW-PBE pseudopotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 5.9 Engineering constants of elemental metals calculated using PAW-PBE pseudopotentials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 5.10 Elastic properties of elemental metals calculated using four pseudpotentials 164 5.11 Engineering constants of elemental metals calculated using using four pseudpotentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 6.1 Displacements of nodal points for BCC structure loaded in [110] direction 177 6.2 Comparison of panel and RVE model . . . . . . . . . . . . . . . . . . . . . 182 6.3 Displacements of nodal points for Type I structure . . . . . . . . . . . . . . 185 6.4 Comparative (dimensionless) values for tensile and shear modulus, and Poisson’s ratio for maximum component beam stiffness. . . . . . . . . . . 197 17 6.5 Parameters used in the finite element analysis, and subsequently in determining normalised material properties. . . . . . . . . . . . . . . . . . . . . 199 6.6 Apparent and relative mechanical properties for structures where one, two and three beams are present within the structure, and the component beams have equal material properties (see table 6.5). 6.7 . . . . . . . . . . . . . . . . 199 Apparent and relative mechanical properties for conventional honeycomb structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199 7.1 Poisson’s ratio calculated for each of the potential models . . . . . . . . . . 202 7.2 Ability of each potential to predict Poisson’s ratio . . . . . . . . . . . . . . 204 C.1 Comparison of Poisson’s ratio calculated directly from strains in the (110) direction, and from rotating the elastic constants derived from strains in the (100) direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224 18 Acknowledgements Foremost, I would like to express my sincere gratitude to Dr Arnaud Marmier for his continuous support and guidance, for his patience, motivation and enthusiasm. I could not have imagined having a better advisor and mentor. I would like to thank Professor Ken Evans for his encouragement, insightful comments, and challenging questions. I thank my fellow researchers: Chris Taylor, Nunzio Palumbo, Marc Boucher, Rob Allen, Matt Johns, Ste Mellor and David Barnes. A special thanks goes to Dr. Wayne Miller who has supported me both academically and pastorally throughout my time at Exeter. Last but not the least, I would like to thank my family: my parents Paul and Wendy, for all the support and help that they have given me throughout the years. 19 Introduction This thesis details a thorough investigation of negative Poisson’s ratio, or auxetic, behaviour in cubic systems, from metallic and covalent crystals to macroscopic beam networks. Negative Poisson’s ratio materials are those which contract laterally when compressed and expand laterally when stretched. The main strength of the study is that many techniques, analytical and numerical, have been employed to understand the main drivers of auxeticity in cubic structures. Despite the range of the techniques used, and the variety of subject materials, the conclusions are strikingly similar, showing that the structure of the bonding networks dominates the details of the "interaction". While auxetic behaviour has been shown to exist in face centred cubic and body centred cubic crystals, the underlying causes and mechanisms are still little known. Understanding causes and mechanisms of auxeticity in crystals is important in itself for fundamental reasons, but the main attraction of a multi-scale, multi-techniques approach is that it opens the way for the design of macroscopic, very lightweight, three dimensional beam structures with very specific elastic properties. The idea is to take advantage of large databases of measured elastic moduli in crystals (or of cheap computational techniques) to provide already very good candidate structures for beam networks, where the beams are analogous to the bonds. The first two chapters review the current literature in the field, at first exploring materials known to exhibit negative Poisson’s ratio behaviour, then presenting the tools used to describe and calculate elastic properties. Purely analytical techniques are used in the third chapter to isolate the role of various effects of bonding; structure and range of interaction within cubic crystals. The fourth chapter demonstrates that empirical classical potential models can be used to access the elastic properties of cubic elemental metals, and highlights the mechanism of this behaviour, in the case of face-centred cubic crystals, for the first time. The limitations of such model are analysed at length. In the following chapter, quantum modelling provides a more robust methodology for finding the elastic properties of cubic crystals, mitigating the shortcomings of empirical models. The sixth chapter abandons crystals and details of interatomic bonding, as the insights obtained from the crystallographic modelling are applied to scale-independent beam networks. 20 In this section of the work, a combination of analytical and finite element modelling techniques results in findings which are very similar to those obtained at the atomistic scale. In the final chapter, the conclusions from diverse techniques at various scales are summarised and generic, scale independent, rules for Poison’s ratios in cubic structures are postulated. As the generic method of using bond structures from promising crystals to generate real three dimensional structures has been validated, I finally suggest possible ways to extend this work towards realistic applications. 21 Chapter 1 Background: Properties of materials 1.1 1.1.1 Negative Poisson’s ratio materials Introduction Poisson’s ratio (ν) is the ratio of lateral to axial strain, ν=− εlateral , εaxial (1.1) and for many common materials ranges from 0.25 to 0.3 (Callister, 2003). Negative Poisson’s ratio, or "auxetic" (Evans et al., 1991) materials exhibit the counterintuitive property that they expand laterally when stretched, or conversely, contract laterally when compressed, in one or more directions. Auxetic materials are worthy of investigation; as a by-product of this behaviour they often exhibit high shear modulus and indentation resistance (Evans and Alderson, 2000). Another potential application of auxetic materials is where sinclastic (doubly curved, or bowl shaped) curvature is required. The nature of auxetic materials means that in bending, the tensile/compressive behaviour occurs both parallel and perpendicular to the profile of the curvature, unlike in conventional materials which tend to form anticlastic shapes when bent in one direction. Applications of auxetic materials are wide ranging, including arterial prosthesis (Caddock and Evans, 1995), blast curtains for protection of building occupants from debris (Wright et al., 2010) and seat cushions (Lowe and Lakes, 2000). 22 In this context, materials may be classified as non-auxetic, where Poisson’s ratio is positive in all directions for all strains; auxetic, where they exhibit negative Poisson’s ratio behaviour in one direction when strained in a specific direction (accompanied by a corresponding positive orthogonal strain); averagely auxetic, where for most strains, negative Poisson’s ratio behaviour is observed in one or more directions or completely auxetic, where for a specific directional strain, the corresponding negative Poisson’s ratios are all negative (see figure 1.1). Figure 1.1: Schematic orthographic projection showing, left to right, deformed shape of non-auxetic, auxetic and completely auxetic material (unstrained green, strained cream) One of the earliest recorded instances of a material with a negative Poisson’s ratio is iron pyrite, calculated to be - 17 (Love, 1892). More recently, Gibson et al. (1982) showed both experimentally and analytically that it is possible to create a honeycomb structure with a Poisson’s ratio approaching -1. Work by Lakes (1987) showed that it was possible to modify a conventional open-celled foam structure to form a re-entrant structure with a Poisson’s ratio of -0.7. As well as the aforementioned Iron Pyrite, auxetic behaviour can also be seen in a variety of crystal structures. It has been shown that 69% of cubic metals exhibit a negative Poisson’s ratio in the (110) plane (Baughman et al., 1998). More complex crystal structures such as Zeolites and Cristobalites also exhibit a negative Poisson’s ratio due to their 23 framework structure; Lethbridge et al. (2010) conducted a systematic investigation into the Poisson’s ratio of 472 materials, both elements and compounds of many symmetries, and found 36% of them to be auxetic in one or more directions. 1.1.2 1.1.2.1 Crystals Simple Crystals Auxetic behaviour has been observed in simple crystal structures: face-centred and bodycentred cubic, and hexagonal close packed structures (FCC, BCC and HCP respectively). Milstein and Huang (1979) used experimental values of elastic moduli to determine the Poisson’s ratio for face-centred cubic crystals. Of the twelve elements investigated, only aluminium did not show negative Poisson’s ratio behaviour, the remaining 11 showed some degree of auxetic behaviour in the [11̄0] direction. It has been shown analytically (Milstein and Huang, 1979; Baughman et al., 1998), that BCC metals can exhibit negative Poisson’s ratio in the (110) plane. Experimentally derived elastic constants are used to compute the Poisson’s ratio for 32 metals with cubic phases. 69% are found to exhibit non-axial negative elastic behaviour. 20 hexagonal structures are studied and only Zn and Be found to be auxetic. Baughman suggests a mechanism for negative elastic behaviour due to hinging of the bonds in the crystal. By representing the inter-atomic forces in two adjacent body-centred cubic cells as rigid beams able to hinge at the atoms, it is shown that for a stretch in the [110] direction the corresponding atoms within the cell are displaced in a scissor-jack like mechanism and there is a resulting contraction in the vertical plane. Tokmakova (2005) used stereographic projection to investigate Poisson’s ratio in auxetic crystals. Tokmakova defines the Poisson’s ratio in the direction m for a stress in the direction q as a function of two Euler angles (to identify the direction of q) and the lateral strain in the cross section. The Poisson’s ratio for any possible stretch for cubic, monoclinic and hexagonal crystals is investigated using this reference system, and plotted. The minimum and maximum Poisson’s ratios are found for Copper, Silver and Iron alloys. All the alloys considered in this study were calculated to have a negative Poisson’s ratio in one or more directions. Tokmakova found labradorite to have a Poisson’s ratio of -0.102 24 for the stretch direction (0°, 65.4 °, 90°) and augite to have a Poisson’s ratio of -0.008 for the stretch direction (0°, 81.4°, 90°). Investigation into hexagonal structures gave a negative Poisson’s ratio for Zn, as reported by Lubarda and Meyers (1999), molybdenum sulphide, graphite polypropylene, and carbon. Common monoclinic crystals were not found to exhibit negative elastic behaviour. Auxetic behaviour is reported by Li (1976) in Cadmium, however no experimental data or elastic constants are provided, and this result is not found by other studies. 1.1.2.2 Silicates Silicates are a form of silicon dioxide based minerals. Common examples of silicates include sand and quartz. Generally the structure of silicates consists of one central Silicon (Si) atom surrounded by four Oxygen (O) atoms in a tetrahedral configuration. These tetrahedra are linked at the vertices, which then gives an open framework structure. It has been thought that this open framework structure could lead to negative Poisson’s ratio as there is the possibility for rotational displacements within the structure. Quartz is made up of a continuous framework of SiO4 silicon-oxygen tetrahedra, with each oxygen being shared between two tetrahedra. Negative Poisson’s ratio has been found in quartz, and the analytical model attributes this to both the rotation and dilation of tetrahedra (Alderson and Evans, 2009) . Cristobalite is a polymorph of silicon dioxide that has a high temperature (β) and a low temperature (α) modification (Yeganeh-Haeri et al., 1992). Using Brillouin spectroscopy, Yeganeh-Haeri showed that a single crystal of α-cristobalite has a Poisson’s ratio (ν) of between 0.08 to -0.5, but predominantly ν is negative. The elastic constants of α-cristobalite showed it is highly anisotropic, and the shear modulus was found to be around 2.4 times the bulk modulus, indicating that it is far more rigid than compressible. The transition from α to β phase was modelled using molecular dynamics by Kimizuka et al. (2000). The β phase was calculated to also be auxetic (νmin = −0.28 at 1800K). The mechanism for the auxetic behaviour of α-cristobalite is due to its tetrahedral framework structure. The unit cell can be described as a tetragon, comprised of four 25 smaller tetrahedra (Alderson and Evans, 2001). Rotation of the rigid tetrahedra gives rise to the negative Poisson’s ratio mechanism. An alternative rotating squares or rectangles mechanism is suggested (Grima et al., 2006, 2005). This theory is suggested after conducting molecular modelling investigations. Rectangular units can be observed in the structure but this model is only a two dimensional, projected, representation of the tetrahedral model. Other SiO2 framework structures that exhibit a negative Poisson’s ratio include the zeolites Natrolite, Thompsonite and Analcime (Grima et al., 2000). Natrolite and Analcime were investigated using Brillouin scattering (Sanchez-Valle et al., 2005). The Poisson’s ratios found by experiment were found to be all positive, in contrast with the computational study by Grima (2000). Experimental measurements of the bending, compression and indentation behaviour of Natrolite were conducted and compared to force-field calculations (Lethbridge et al., 2006). The experimental ultrasonic data were subject to large uncertainties due to the shape of the crystals examined. However despite this, all the simulated and experimental data were found to be in reasonable agreement. It was concluded that there was no evidence for on-axis auxetic behaviour. The predicted off-axis negative Poisson’s ratio behaviour is attributed to rotation of the tetrahedral units in an idealised structure. In the actual Natrolite structure, cations not present in the idealised structure prohibit this rotation, and the positive Poisson’s ratio originates from the deformation of the tetrahedral units. Contrary to this, negative Poisson’s behaviour was found in the (001) plane by Grima et al. (2007) from analysis of the elastic constants, and the mechanism is attributed to rotating squares. 1.1.2.3 Paratellurite Paratellurite (α−TeO2 ) is a crystal that is widely used in a variety of applications due to its piezoelectric properties. It also has remarkable elastic anisotropy and negative Poisson’s ratio behaviour (Ogi et al., 2004). This behaviour is attributed to a star shaped truss structure that when subjected to a compression in the [110] direction, creates contraction in the perpendicular direction. Further investigation revealed that ν is negative for 26 displacements around the b and c axis (Ogi et al., 2006). Ab-initio calculations support the experimental findings (Ceriotti et al., 2006). 1.1.2.4 Metal organic frameworks Metal organic framework structures (MOFs) are single-phase crystalline hybrid organicinorganic materials (Tan and Cheetham, 2011). These materials can demonstrate the functionality of organic materials whilst exhibiting the thermal stability of inorganic materials. MOFs are an open framework structure, and it is this that makes them both useful for molecular separation, gas storage and catalysis applications, as well as presenting the possibility of negative Poisson’s ratio behaviour. Synthesis of MOFs is still in its infancy and very much a developing field, there are to date, around 2000 MOF structures (Yaghi, 2012), and the nomenclature is yet to be fully developed. Most mechanical characterisation of MOFs has thus far employed nano indentation techniques to find the elastic constants however the accuracy of this technique, particularly given the dependence of elastic properties on material orientation in such highly anisotropic structures, combined with the size of the crystals being investigated means that these results may benefit for further validation and often a range of values is to be found in the literature for a specific crystal. 1.1.3 1.1.3.1 Fabricated structures Honeycombs Several auxetic honeycomb structures have been devised, these can mainly be classified as either re-entrant, chiral, or anti-chiral. The first auxetic honeycomb structures to be developed were re-entrant honeycomb structures, a two dimensional structure with a Poisson’s ratio of -1 (Gibson et al., 1982; Almgren, 1985) and variations thereof (Evans, 1989). The deformation mechanism is such that inplane, on axis compression results in hinging of the ribs of the structure which is forced to occur in the orthogonal axis and thus the structure has a Poisson’s ratio of -1 (see fig. 1.2). A unit cell for this mechanism in three dimensions is suggested by Lakes (1987). The effect on the Poisson’s ratio of varying the geometry of re-entrant honeycombs was considered by Wan et al. (2004) who 27 showed that varying the aspect ratio of the cells could give a Poisson’s ratio as low as -4.211. Figure 1.2: The auxetic mechanism of a re-entrant honeycomb. (Taken from http://silver.neep.wisc.edu/ lakes/Poisson.html) Chiral honeycombs first suggested by Lakes (1991) are composed of nodes of equal radius joined by ribs of equal length (Prall and Lakes, 1997), and may have three, four or six ribs at each node. The deformation mechanism is such that an on-axis compression forces the ribs of the structure to bend and the nodes to rotate (see fig. 1.3), drawing in the ribs that are off-axis and thus giving the structure auxetic behaviour with a Poisson’s ratio of -1. Anti-chiral honeycombs utilise the same mechanism but have the ribs arranged in such a fashion that they exhibit symmetry. Figure 1.3: The auxetic mechanism of one cell of a tetrachiral honeycomb (left undeformed, right - deformed) 1.1.3.2 Foams Conventional thermoplastic and polymeric foams can be modified to create an auxetic structure by triaxial compression followed by heat treatment; the resulting structure has curved ribs that create a re-entrant structure under compression. Foams generally have a degree of anisotropy owing to the manufacturing process (Weaire and Hutzler, 2005). 28 A negative Poisson’s ratio of up to -0.25 in one plane, and -0.15 orthogonal to this was achieved in preliminary investigations into this technique using a polyester urethane foam (Friis et al., 1988). Polyethylene foams can be processed to give a Poisson’s ratio of around -0.7 for small strains (Brandel and Lakes, 2001). More recently, foams with a Poisson’s ratio as low as -1.4 in the ν yx direction have been manufactured Bianchi et al. (2011). 1.1.3.3 Molecular Auxetics Polymers synthesised to create materials with a negative Poisson’s ratio at the molecular scale have been developed. These employ a variety of mechanisms. At a molecular level the structure ’(n,m)-reflexyne’ was proposed (Evans et al., 1991). This gives a two dimensional re-entrant honeycomb structure with a theoretical Poison’s ratio, calculated using molecular dynamics, of -0.96. Polytetraflouroethylene (PTFE) was found to be highly anisotropic with a Poisson’s ratio as extreme as -11 due to the complex microstructures from which it is formed (Caddock and Evans, 1989). This can be approximated as a particle-string model to give an analytical model in close agreement to experimental data (Evans and Caddock, 1989). Ultra high molecular weight polyethylene (UHMWPE) can be processed to give it a similar structure to PTFE (Neale et al., 1993) and thus a Poisson’s ratio of approximately -7 (Alderson and Evans, 1993). A similar network of hinging interconnecting rods was proposed by He et al. (2005) who used para-quaterphenyl rods laterally attached to a polymer backbone to give a possible mechanism for auxetic behaviour at the molecular scale. Polyphenylacetylene networks comprise triangular units of benzene rings ’nodes’ joined by acetylene chains to make a structure of triangular repeating units. Molecular dynamics calculations show that by a ’rotating triangles’ mechanism (similar to the rotating squares mechanism of zeolites) this configuration can produce a Poisson’s ratio of -0.97 (Grima and Evans, 2000). 29 1.1.3.4 Composites Composite materials can be designed such that their properties are optimised for their intended use and thus are generally high strength, low weight materials. The key parameter in making a negative Poisson’s ratio composite from auxetic inclusions embedded in a conventional matrix is the ratio of auxetic to non-auxetic components (Wei and Edwards, 1998). By alternating plates of auxetic and conventional materials, homogenization theory can be used to estimate the global modulus of the structure, and it is found that inclusion of auxetic layers results in a stiffer structure (Donescu et al., 2008; Lim, 2007). Composite inclusions of various shapes can be used to vary the properties of the composite, and the exact response of the material is dependent not only on the ratio of auxetic to conventional materials but also the shape of the auxetic inclusions. A composite containing star shaped rigid inclusions set in a compressible matrix can be found to give auxetic behaviour by either dilation of the triangles within the structure (Milton, 1992) and by a re-entrant like structure (Theocaris et al., 1997). Auxetic fibre reinforced composites may be made entirely from conventional materials by choosing an optimal stacking configuration (Alderson et al., 2005). A material manufactured in this way from standard carbon epoxy prepreg was shown to have a Poisson’s ratio of -0.156, and demonstrated greater resistance to indentation than a similar conventional composite material. A high/negative Poisson’s ratio elastomer-matrix composite has the potential to produce a structure with extraordinary Poisson’s ratio: from 100 to -60 dependent on composition and angle of lay-up (Peel, 2007). Double-helix yarn fibres comprises a low-modulus central strand, with a thinner high modulus strand wound around it. Under tension, the higher modulus strand forces the lower modulus strand to adopt a helical shape, and drives a perpendicular expansion. The magnitude of this effect is dependent on the parameters of the system; comparative radius of the yarns and angle of the wrap, but Poisson’s ratios of -2.1 have been shown Miller et al. (2012a). Using this system, it has been possible to construct a negative Poisson’s ratio Carbon fibre composite (Miller et al., 2012b). 30 1.1.3.5 Granular materials It has been shown analytically that, in both two and three dimensional systems the Poisson’s ratio of a granular solid is dependent on the ratio of the axial to tangential stiffness, λ (Bathurst and Rothenburg, 1988). For λ greater than 1, Poisson’s ratio is negative. Numerical simulations of bonded disc assemblies show that in two dimensions, an assembly of 1000 discs with a λ value of 3 gives a Poisson’s ratio of -0.301. The link between granular materials and a network of bending beams was made by Koenders (2008), modelling the interactions between particles as beams and the particles themselves as nodes. It was shown than auxetic behaviour could result from anisotropy in the packing arrangement of the beams and that this problem was analogous to that of a granular system. 1.1.3.6 Applications of mono-crystalline structures Micro electro-mechanical systems (MEMS) are being widely used for a variety of both mechanical and electronic applications. A recent application for a micro-mechanical system is in the control of lightweight mirrors for space telescopes (see, for example Hishinuma et al. (2005)). Current developments in this field are based around silicon based structures, however, by incorporating a highly anisotropic, monocrystalline structural element into the actuator at a specific orientation to maximise the Poisson effect, there is the possibility that the small strain of the silicon component could be magnified by the monocrystalline element, and thus a far greater range of movement made available. Accurate modelling of these off-axis properties is critical to implementation of these material in this, and other developing technologies. 1.1.4 Investigative Methodologies Investigation of auxetic properties has taken several routes: experimental, mathematical and more recently computational modelling have been used to investigate auxetic properties. For a more comprehensive review of the techniques employed for this thesis refer to chapter 2. 31 1.1.4.1 Crystallography Various crystallographic techniques have been applied to the study of elasticity, however generally all involve interaction of a wave with lattice phonons and extracting the elastic constants from dispersion curves. Neutron scattering methods are capable of probing lattice dynamics (Dove, 2005). A neutron beam is directed at the sample being considered and is diffracted as it passes through the sample. The phonon spectrum of the material can be observed from the resulting change in wave energy and wave vector. Neutron diffraction can be used for both single and poly-crystalline samples. Neutron diffraction allows the phonons within the crystal to be observed, and from these the elastic properties can be calculated. Crystalline materials can be investigated using x-ray scattering techniques (for example (Mendham et al., 2000)) in a process similar to neutron scattering however at a far greater energy. X-ray scattering is said to be in its infancy compared to neutron scattering techniques, however, with the development of synchrotron x-ray sources, dispersion curves comparable to that produced by neutron scattering can be produced (Dove, 2005). The theory of both X-ray and neutron scattering is discussed in depth by Ashcroft and Mermin (1976) and by Burkel (2001) amongst others. 1.1.4.2 Spectroscopy Raman scattering uses light to probe the lattice phonons of the sample. The technique has a high resolution and is suitable for both single and polycrystalline samples, however is particularly useful for investigating the spectra of single crystals at differing alignments to investigate the symmetries of the vibrations (Palmer et al., 1994). Brillouin spectroscopy has been widely used as a method for determining crystal structures, for example Yeganeh-Haeri et al. (1992). Brillouin scattering uses acoustic modes to determine the elastic constants and Poisson’s ratio of a single crystal (SanchezValle et al., 2005). The energy of the scattered light may be increased or decreased; usually referred to as Brillouin shift. This is equal to the energy of the interacting phonon and magnon and thus Brillouin scattering can be used to measure phonon and magnon energies, and to determine elastic properties. 32 1.1.4.3 Molecular Modelling Techniques Molecular mechanics provides a ’virtual experiment’ to model the interactions of a system of atoms. The potential of the system can be calculated using either classical, or quantum mechanics. The use of molecular modelling techniques to model crystalline structures and their elastic properties is widespread. For the modelling of Silicates, a specific force field model has been developed from experimental data, the Burchart force field. It assumes that the frameworks are largely covalent however, one shortcoming of this method is that it does not accurately reproduce the on-axis Poisson’s ratio found experimentally (Alderson et al., 2005). Grima and Evans (2000) use Burchart, BKS, Universal and Constant Valence forcefield modelling (using Cerius2 ) to predict the negative Poisson’s ratio in idealised zeolite cage structures. Similarly, the CVFF force-field model was applied by Grima et al. (2006) to model the deformation mechanism in alpha-cristobalite. Cerius2 was also used by Grima et al. (2005) to model calix-[4]-arene network structures using four force-fields. The ’egg-rack’ microstructure was found to exhibit negative Poisson’s ratio in two planes and through variation of the structure, the Poisson’s ratio could be altered from -0.51 to -0.95. The use of ’dummy-atoms’ to investigate the properties of a re-entrant honeycomb at nano-scale showed that this scale structure can be created with auxetic behaviour (Grima et al., 2005). The EMUDA (Empirical modelling using dummy atoms) approach can be used as a tool for modelling new structures where analytical methods are yet to be developed, and is a cheaper alternative to experiment. Density functional theory was used to investigate the lattice dynamics of solid benzene (Kearley et al., 2006). It was found that elongation of the crystal along the b axis leads to an increase in the a and c directions. Monte-Carlo simulations of theoretical materials (Tretiakov and Wojciechowski, 2005) show that, as found in the analytical model (Wojciechowski, 2004), auxetic behaviour can occur. Simulations of an auxetic ferrogel shows that magnetic effects can be combined with auxetic behaviour (Dudek et al., 2007). 33 1.1.4.4 Analytical Modelling Modelling of auxetic microporous polymers was investigated using a ’node/fibril’ model to try to reproduce the mechanisms seen with PTFE and UHMWPE (Alderson and Evans, 1995). The fibrils are modelled as free to rotate, and connected by hinged inextensible rods. The Poisson’s ratio and Young’s modulus can be found in this way. Comparing the models to experimental data it is found that the model is valid for up to around 10% strain, at which point the Poisson’s ratio begins to become less negative as the deformation mechanism becomes extension of the fibrils. A model for negative Poisson’s ratio behaviour in foams was developed by Masters and Evans (1996). It was noticed that samples of both auxetic and conventional foams had broken ribs within their structure as a result of the manufacturing process, this was used as the basis of the ’missing rib’ model (Smith et al., 2000). When modelling a structure such as a polymeric foam, the behaviour of the foam at the macro scale is far different from the behaviour of the foam at the micro scale. A large sample of foam would behave as though a solid homogeneous medium whereas at the microscale, it is composed of a complex network of ribs. It is not possible to model this microscale using continuous elasticity and thus homogenization must be employed (Gaspar et al., 2008). Using homogenisation, Gaspar showed that for an isotropic small scale material where the modulus fluctuates it is possible to produce an isotropic auxetic material. For anisotropic small-scale media, to produce auxetic behaviour at the macro scale the fluctuations in modulus allowed are reduced. Modelling of hard spheres as a framework for auxetic behaviour has been conducted using molecular dynamics (MD) calculations - the MD framework being applied to a purely theoretical structure. Modelling hard cyclic multimers can be shown to exhibit negative Poisson’s ratio under certain conditions (Wojciechowski, 2004) where the arrangement is such that the structure forms a triangular lattice at close packing. Variation of the ratio of diameters of this model and introduction of disorder into the system gives a model that shows the effect of molecular shape asymmetry on elastic properties (Narojczyk and Wojciechowski, 2008). Classical elasticity provides an ideal framework for the modelling of negative Pois34 son’s ratio materials, and is discussed in detail in section 2.2. 1.1.4.4.1 Rotating Rigid Units Rotating polygons (and in 3 dimensions) polyhedra provide an accurate and convenient general model for negative behaviour in a variety of structures, and are comparable with the Rigid Unit Modes models suggested by Dove (1997). The rotating rectangles method (see, for example (Grima et al., 2005b) and similar), provides a 2D model for planar structures exhibiting auxetic behaviour. The rotating rectangles model assumes that the structure being considered, generally a framework structure such as a zeolite, can be discretised into a system of rigid rectangular elements linked by bonds that allowed the rigid units to hinge. Deformation of the structure results in hinging of the rigid units allowing a Poisson’s ratio of -1. Using this model a compliance matrix for the structure can be derived. This model can be applied to zeolites and silicates and is also proposed as an alternative mechanism for auxeticity in foam structures (Grima et al., 2005a). A similar model was proposed by Ishibashi and Iwata (2000) and reviewed in detail by Dmitriev (2007), whereby a system of squares linked by springs is employed to model the structure. A variation on this analytical model was developed by Williams et al. (2007) for auxetic behaviour in Zeolites. Rotating linked parallelogram structures are used to give a 2D model. This model gives a theoretical Poisson’s ratio of -1. Including a term for the distortion of the parallelograms gives the possibility of a lower ν. Stretching, rather than rotation of squares is a further mechanism for auxetic behaviour (Grima et al., 2008). This mechanism has potential to be adjustable to give the required properties in the structure being designed. As a model it is valid but seemingly does not describe any pre-existing behaviour, and is conditional on the unit cells having the described ’piston like’ structure thus seems unlikely to be applicable to any molecular or honeycomb like structure as well as presenting various manufacturing difficulties. 35 1.2 1.2.1 Other unusual material behaviour Introduction Whilst this work is concerned with the discovery and exploitation of negative Poisson’s ratio (NPR) behaviour, considering other ’negative’ elastic properties (negative thermal expansion (NTE) and negative linear compressibility (NLC)) could provide an insight into the mechanism of NPR behaviour. The discovery, prevalence and mechanism of NTE and NLC materials are briefly reviewed in the next section. 1.2.2 Negative Thermal Expansion 1.2.2.1 Background Negative thermal expansion (NTE), the phenomenon whereby a material shrinks on heating or expands on cooling can be observed in many natural structures, the most commonly known example is the expansion of water during freezing (although this is brought about by a phase change) but NTE is also seen in more complex crystals and in fabricated structures (Oruganti et al., 2004). There are several mechanisms by which materials exhibit negative thermal behaviour, dependent on the structure of the material. Investigation of negative thermal expansion materials provides a useful background from which novel composites, with potential for zero thermal expansion, can be developed. 1.2.2.2 1.2.2.2.1 Crystals Cubic crystals Whilst NTE behaviour is generally observed in complex framework structures such as those discussed below, it is also found in Caesium. There are six known phases of caesium at room temperature. Ab-initio calculations by Christensen et al. (2000) revealed the coefficient of thermal expansion for Caesium in its II (face-centred cubic) phase to be negative at pressures above 3 GPa for all temperatures. Whilst there is no experimental evidence offered to support this for FCC, the results for BCC caesium are compared to experimental data and found to be in agreement. 36 1.2.2.2.2 Metal Organic Frameworks Metal organic framework (MOF) structures comprise atoms or clusters of metals with organic linkers (Zhou et al., 2008). MOF-5 (or Zn4 O13 − (C8 H4 )) has been shown in molecular dynamics simulations to have a thermal expansion coefficent as low as -20E-6K−1 (Dubbeldam et al., 2007). MOF-5 comprises ZnO4 clusters linked by a 1,4-benzenedicardoxylate FCC crystal (Li et al., 1999). The mechanism for NTE suggested by Zhou et. al. is that the ZnO4 tetragonal units are bridged to the benzene rings in the structure by carboxyl groups. Large amplitude vibrations of these carboxyl groups leads to negative thermal expansion. Experiment found that the linear thermal expansion coefficient of MOF-5 was -16E-6K−1 . The lattice phonon modes were investigated and it was found that low frequency lattice modes were responsible for the NTE in the structure. It is suggested that MOFs that do not have a flexible bridging group such as the carboxyl group in MOF-5 would not exhibit NTE. Wu et al. (2008) investigated Cu3 (1, 3, 5 − benzenetricarboxylate)2 experimentally, it was found to have a negative thermal coefficient owing to translation of the aromatic ring in a similar fashion to that of MOF-5. 1.2.2.2.3 Other framework structures As with auxetic behaviour, framework structures such as zeolites have potential to behave as negative thermal expansion materials. One such structure is silver hexacyanocobalate. The structure is that of a ’Kagome’ lattice, hexagonal units formed by Ag atoms with cyanocobalate ions positioned above the Kagome holes (Goodwin et al., 2008). Experimental investigation using x-ray diffraction by Goodwin et al. revealed the coefficient of thermal expansion to be in the order of -120E-6 to -130E-6 K−1 , an order of magnitude larger than is generally found with framework structures. DFT calculations reveal that this behaviour was due to the structure behaving like a garden trellis; the bond lengths undergo very small changes under heating but the bond angles change considerably (Calleja et al., 2008). Zirconium tungstate exhibits negative thermal expansion over a very wide temperature range (Evans et al., 1997). Despite that there is a structural phase transformation within this temperature range, it exhibits negative thermal expansion in both phases and 37 thus this leads to isotropic thermal expansion (Pryde et al., 1996). A similar effect is seen in the compound ZrV2 O7 . The negative thermal expansion of Zirconium Tungstate is caused by transverse vibrations of the Oxygen molecule in Zr-O-W bonds which lead to shortening of the Zr-W distance and thus a negative thermal expansion (Mary et al., 1996). The structure of zirconium tungstate is such that it comprises WO4 tetrahedra linking ZrO6 octahedra, the combination of linked octahedra and tetrahedra forming a framework structure. Decreasing the distance between the adjacent Zirconium and Tungsten leads to opposing rotations of the polyhedra. The mechanism of NTE by rotation of these polyhedra without distorting is known as a rotating unit mode (RUM) (Giddy et al., 1993). The shape of a typical metal-oxide potential implies that longitudinal oxygen vibrations creates an increase in M-O and M-M distances, however a transverse vibration in which M-O bond distances are unchanged can result in a reduction in the M-M distance (Evans et al., 1997). Tao and Sleight (2003) investigated NTE in five oxide structures with varying ratios of octahedral and tetrahedral cations. Negative thermal expansion is found in Sc2 (WO4 )3 from 10 to 1073K (Evans et al., 1998). The structure of Sc2 (WO4 )3 comprises ScO6 octahedra joined at the corners to six WO4 tetrahedra, each tetrahedron being joined to four octahedra. It has been observed that an increase in temperature results in minimal change in bond angles in the polyhedra and thus the rigid unit modes model is applicable to this structure. Where the hinging of the rigid units is enabled due to M-O-M oxide based frameworks, this can also be seen in cyanide bridged frameworks where the bridging unit becomes M-CN-M (Goodwin and Kepert, 2005). Unlike in oxide frameworks where the linking element is formed of just one molecule, in CN bonded structures, the diatomic linking element gives rise to a larger range of RUMs. For the perovskite structure, investigated by Goodwin, it was found that whilst with the M-O-M linkages the was a single RUM at one wave vector, using a diatomic linkage results in RUMs at all wave vectors across the Brillouin zone. The framework structure of zeolites make them a likely class of solids that exhibits negative thermal expansion. Computational investigations into the effect of tempera- 38 ture on siliceous faujasite conducted by Couves et al. (1993) predicted a reduction in lattice parameters under heating. These data were verified by high-resolution powder diffraction studies. Lattice dynamics calculations for 18 zeolite structures found NTE behaviour which was attributed to the structure expanding into the pores within the network (Tschaufeser and Parker, 1995). Strong NTE was observed by Woodcock et al. (1999). It was thought that water trapped within the pores of the structure may be evacuated at high temperatures, and this may be the cause of this behaviour however further investigation did not confirm this. Reisner et al. (2000) investigated the zeolite RHO experimentally and found there to be two stages of NTE, rapid contraction from 25-75°Cand slower contraction up to 500°C. Cation relocation caused by dehydration of the structure under heating means that the structure changes shape. In dehydrated zeolite RHO this was not observed. 1.2.2.3 1.2.2.3.1 Mechanisms of NTE Increasing symmetry Increasing symmetry thermal expansion is seen in PbTiO3 and similar AMO3 compounds that have a perovskite structure (Sleight, 1998a). Such compounds contain MO6 and AO12 polyhedra that become more regular with increasing temperature. As the polyhedra become more regular, the M-O distances decrease. This leads to a thermal expansion in two planes, and thermal contraction in the third plane. 1.2.2.3.2 Positive expansion of bonds Positive thermal expansion of some M-O bonds in hexagonal structures can contribute to negative thermal expansion (Sleight, 1998b). The mechanism for this is described in detail for Cordierite, Mg2 Al2 Si5 O18 by Sleight (1998a). Assuming that the thermal expansion of the AL-O, Si-O, P-O and Zr-O bonds is very small, then the mechanism is dependent on the thermal expansion of the Mg-O, Li-O and Na-O bonds. Because in this class of oxides the structure is comprised of face/edge sharing polyhedra, it is possible for the polyhedron to expand in only one or two directions (expansion is not possible in all three planes due to cation-cation repulsion across shared edges) and some negative expansion 39 is seen. 1.2.2.3.3 Electron valence transition/cation movement The movement of cations is used by Sleight (1998a) to explain the negative thermal behaviour in β-eucryptite. The Li+ cations migrate from tetrahedral to octahedral sites to contribute to negative volume expansion. Electronic valence transition is proposed as a mechanism for zero volumetric thermal expansion in YbGaGe by Salavdor et al. (2003). As the material is cooled, electrons depopulate the conduction band and increase the fraction of Yb2+ ions. As Yb2+ ions are larger than Yb3+ ions, there is an expansion in the a-b plane. This mechanism, and the existence of zero thermal expansion in YbGaGe is disputed by Bobev et al. (2004) and Janssen et al. (2005) as in similar studies this effect was not observed, and it seems that the stoichiometry of the sample may define its behaviour. Janssen found the thermal expansion to be comparable to that of copper. Delta plutonium is suggested as a possible material that does exhibit electronic valence transition to give rise to NTE behaviour. 1.2.2.3.4 Rotation of rigid units In structures composed of polyhedra, such as tetragon or hexagons, linked by shared atoms the corners of the structure, rotation of these units can lead to a negative thermal expansion. Co-operative rotation of the polyhedra leads to a reduction of the spaces between the polyhedra. 1.2.2.4 1.2.2.4.1 Methodologies Experimental Experimental methodolgies used in the literature are generally the same as those employed for observation of negative Poisson’s behaviour; neutron diffraction (Evans et al., 1998, 1996), x-ray diffraction (Lightfoot et al., 2001; Woodcock et al., 1999; Wu et al., 2008) and raman spectroscopy. Infra-red spectroscopy is used by Evans et al. (1996). Different vibration modes give differing absorption spectra, thus the behaviour of the sample over a range of temperatures can be analysed. 40 1.2.2.4.2 Analytical Modelling - Rotating unit modes Rigid unit modes (RUMs) has been used as a model for negative thermal expansion in framework structures (for example: (Pryde et al., 1996; Heine et al., 1999; Tao and Sleight, 2003)). In a RUMs model the framework structure, such as zirconium tungstate, is modelled as rigid polyhedral units joined at corners. These units are assigned classical properties such as mass, moment of inertia and three vibrational and translational degrees of freedom (Tao and Sleight, 2003). The only connectivity in the model is that between the units. The units share oxygen atoms, which serve to link them together. Investigations of compounds with these structures has been conducted using the program CRUSH (Giddy et al., 1993), where a split atom model has been implemented. The split atom model treats the oxygen molecules linking the units as two atoms with a separation of zero, and the constraint that the two split atoms are bound by an invented potential. From this model it is possible to predict if RUMs occur in a given structure, and thus explain the mechanism of NPR in the compound under consideration. The degree to which an expansion is seen is dependent on the degree of rotation, which is a function of the Boltzmann constant, temperature and the frequency of the rotational phonon (Pryde et al., 1996). Summing these over all the units in the lattice gives a net volume change in the structure. Where the modes result in a small distortion of the polyhedra, the RUMs become known as quasi-RUMS. 1.2.2.4.3 Molecular modelling As well as CRUSH, developed for determining the Rigid Unit Modes in structures, several other computational methods have been employed to investigate NTE behaviour. A dynamic version of the split atom method was developed to investigate RUMs in cristobalite and quartz (Gambhir et al., 1999). Free energy minimisation conducted with the program PARAPOC was used to in some of the earliest work on zeolites to model the cell dimension at a range of temperatures (Tschaufeser and Parker, 1995). Interatomic potentials were used by Pryde et al. (1996) to determine the behaviour of zirconium tungstate. 41 1.2.3 Negative Linear Compressibility Negative linear compressibility is a phenomenon whereby a material expands in one direction when subjected to hydrostatic loading. Negative linear compressibility has been found in several compounds by Skelton et al. (1976) in paratellurite (TeO2 ), and by Mariathasan et al. (1985) in lathanum niobate. Baughman et al. (1998) investigated around 500 noncubic crystal compounds, using experimental elastic constants. He found 13 compounds that exhibit this behaviour. Baughman suggests a ’wine-rack like’ deformation structure. The possible mechanisms for this behaviour are suggested to be dependent on the structure, but are all based around interconnected helices. Two helices, opposingly wound would result in a hinged structure that decreases in volume when subjected to an axial tensile load, or if subjected to a pressure from within the helices, the structure can be seen to expanded axially. Negative compressibility was also considered by Grima et al. (2008) who proposes a truss structure formed of triangular hinged units. The model suggested is a triangle whereby one side is comprised of a material with lower compressibility than the other two sides. This gives structure that exhibits negative linear compressibility in the plane perpendicular to the side with lowest compressibility. A further structure that exhibits negative compressibility is suggested by Gatt and Grima (2008), comprising of a structure similar to the chiral structures of Prall and Lakes (1997) however the round central nodes are replaced by square rectangular blocks and the interconnecting ribs are comprised a bi-metallic strips. Applications for negative linear compressibility structures are yet to be realised. 42 Chapter 2 Background: Theoretical framework and computational modelling 2.1 Introduction Negative Poisson’s ratio, and other unusual material properties have been observed in variety of crystal and fabricated structures. This chapter describes techniques used to investigate these properties both analytically and numerically. 2.2 General elasticity equations for stress and strain The linear elastic behaviour of an anisotropic material or structure can be fully described by a set of 21 elastic constants, relating the stress and strain in any given direction (Nye, 1985). This section details the relationship between stress and strain, the conversion from elastic constants to so called "engineering" constants, the relationship between stiffnesses and compliances, and the derivation of elastic properties in any direction from the elastic constants for the standard orientation. 2.2.1 Notation When discussing properties of elastic media, it is usual to talk in terms of stresses and strains within a reference framework. Axial directions are generally referred to as 1, 2, 3, 43 and thus stresses and strains can be expressed as σii , or σi j dependent on whether they are in axial or shear directions. The full components of the stiffness (and compliance) tensor that relate stress and strain and described the elastic properties of the material can be expressed as si jkl . The most commonly employed notation schemes are the Mandel and Voigt notation. 2.2.1.1 Mandel notation The Mandel notation allows a symmetric second rank tensor to be expressed as a vector (albeit one of higher dimension). This is useful as it also allows a rank four symmetric tensor (such as that described in section 2.2.4) to relate stresses and strains in an elastic solid) to be reduced from 81 components to a six by six matrix. The symmetry dictates that Di jkl = D jikl and Di jkl = Di jlk and thus the tensor is reduced to: DM D 1111 D2211 D3311 = √ 2D 1211 √ 2D2311 √ 2D1311 √ √ D1122 D1133 2D1112 √ D2222 D2233 2D2212 D3333 √ 2D1233 √ 2D2333 √ 2D1333 2D2223 √ √ D3322 √ 2D1222 √ 2D2322 √ 2D1322 2D1123 √ 2D3312 2D3323 2D1212 2D1223 2D2312 2D2323 2D1312 2D1323 √ 2D1113 √ 2D2213 √ 2D3313 2D1213 2D2313 2D1313 (2.1) This transformation is generic and symmetric, but rarely used. 2.2.1.2 Voigt Notation For historical reasons, the notation of Voigt is more commonly used (Nye, 1985). This also transforms the three-dimensional fourth-order tensor to a 6 x 6 matrix. The notation follows the rules that a double index e.g. xx becomesx, and for indices of the type xy, the Voigt index becomes 9 − (x + y). The Voigt notation is similar to the Mandel notation, but does describe stress and stiffness, for which the transformation is more convenient. As a consequence, the strain and compliance transformation must account for double or quadruple counting. For 44 convenience when performing such operations as rotations, it is possible to modify the matrix accordingly; when transforming from the true tensor notation to Voigt (matrix) notation, numerical factors must be introduced to identify x11 with x1 and similarly x13 with 12 x5 etc. (Hearmon, 1957). These factors are: sxxyy = sqr (When q and r = 1, 2 or 3) 1 sxxyz = sqr (When q = 1, 2 or 3, and r = 4, 5, or 6) 2 1 sxyxy = sqr (When q and r = 4, 5 or 6) 4 (2.2a) (2.2b) (2.2c) An example of this is given in equation 2.15. 2.2.2 The strain tensor The strain tensor can be used to describe linear elastic strains; for a solid cuboid, of side lengths l1 , l2 and l3 , with a uniform normal pressure applied to one face in the 1 direction (where 1,2 and 3 are equivalent to x, y and z), the face is displaced by du1 . The axial pressure (for almost all materials) results in a corresponding contraction or expansion of the material in the lateral directions. The axial strain in the load direction can be described as: ε11 = du1 l1 (2.3) Similarly, the corresponding orthogonal strains (in the 2 and 3 direction) can be written as: ε22 = du2 l2 (2.4a) ε33 = du3 l3 (2.4b) Applying a displacement parallel to the faces of cuboid results in corresponding shear deformation. In pure shear the deformation is characterised by deformations in the x and y directions. The deformation can be expressed in terms of the angle. 45 θ1 = du1 l2 θ2 = (2.5a) du2 l1 (2.5b) The total shear strain can be written as ε12 1 du1 du2 + = 2 l2 l1 ! (2.6) This can be generalised to the form: 1 dui du j εi j = + 2 lj li ! (2.7) Combining the equations for the axial and shear strains, the total strain tensor can be written as: ε 11 ε12 ε13 ε21 ε22 ε23 ε31 ε32 ε33 = du1 l1 1 2 du 1 2 du 2 l1 l1 3 + + 1 2 du1 l2 du1 l3 du 1 l2 + du2 l1 du2 l3 du2 l2 1 2 du l2 3 + l3 du du3 1 2 + 2 l3 l2 du3 1 2 du 1 + du3 l1 (2.8) l3 The symmetry of the tensor (ε12 = ε21 etc.) can be used to replace the strain tensor to a six dimensional vector: ε 11 ε22 ε33 ε = 1 ε 2 23 1 ε31 2 1 2 ε12 46 (2.9) 2.2.3 The stress tensor Figure 2.1: Schematic showing normal (S1,2,3 ) and shear stresses (S12,13,23 ) for am arbitrary cuboidal solid. The stress tensor can also be constructed in a similar way to the strain tensor. Considering the cuboidal example above, a displacement has a force associated with it. A force over a given area results in a stress in the specified direction. Components of the stress tensor take the form of σi j where i represents the normal to the surface on which the stress acts, and j is the direction of the stress component. The stress tensor can be represented as: σ 11 σ12 σ13 σ = σ21 σ22 σ23 σ31 σ32 σ33 (2.10) Where components of the form σii represent the axial stresses, and components of the from σi j represent the shear stresses. As with the strain tensor, the stress tensor can be represented as a six dimensional vector. 47 σ = 2.2.4 σ11 σ22 σ33 σ23 σ31 σ12 (2.11) Generalised Hooke’s Law Hooke’s law states that: F = kx (2.12) where F is the applied force, k the stiffness of the material and x the displacement. Extending this to stress and strain for isotropic materials, the stress is related to the strain by the Young’s modulus: ε = Eσ (2.13) For anisotropic materials the strain can be related to the stress by the elastic compliance matrix, thus: εi j = si jkl σkl (2.14) In the matrix form: ε11 ε22 ε33 ε23 ε31 ε12 s 1111 s1122 s1133 s2211 s2222 s2233 s3311 s3322 s3333 = 2s 2311 2s2322 2s2333 2s3111 2s3122 2s3133 2s1211 2s1222 2s1233 2s1123 2s1131 2s2223 2s2231 2s3323 2s3331 4s2323 4s2331 4s3123 4s3131 4s1223 4s1231 48 2s1112 2s2212 2s3312 4s2312 4s3112 4s1212 σ11 σ22 σ33 σ23 σ31 σ12 (2.15) Similarly, the elastic stiffness matrix relates the stress to the strain: σi j = ci jkl εi jkl (2.16) Or, in full: σ11 σ22 σ33 σ23 σ31 σ12 = c1111 c1122 c1133 c1144 c1155 c1166 c2211 c2222 c2233 c2244 c2255 c2266 c3311 c3322 c3333 c3344 c3355 c3366 c4411 c4422 c4433 c4444 c4455 c4466 c5511 c5522 c5533 c5544 c5555 c5566 c6611 c6622 c6633 c6644 c6655 c6666 ε11 ε22 ε33 ε23 ε31 ε12 (2.17) Each index (i, j, k, l) may be either 1, 2 or 3, signifying its direction in three dimensional space, hence the 81 constants. Due to the symmetry of the tensor, this can be reduced to 21 unique elastic constants. 2.2.5 Conversion between stiffness and compliance Compliances are the inverse of the stiffnesses, thus by inverting the stiffness matrix the compliance matrix can be found and vice versa. For cubic media where symmetry dictates the elastic constants as c11 , c12 and c44 , the inversion is straightforward: −c11 − c12 + 2c212 − c11 c12 −c21 = 2 −c11 + 2c212 − c11 c12 1 s44 = s55 = s66 = c44 s11 = s22 = s33 = s12 = s13 = s23 −c211 (2.18) (2.19) (2.20) At lower symmetries, the number of unique elastic constants increases the number of terms in the tensor increases. For tetragonal media, the symmetry dictates that seven 49 constants are required, thus the compliance matrix is: s 0 s16 11 s12 s13 0 s12 s11 s13 0 0 −s 16 s13 s13 s33 0 0 0 s = 0 0 0 s 0 0 44 0 0 0 0 s 0 44 s16 −s16 0 0 0 s66 (2.21) The inverse of this matrix describes the stiffnesses: c11 = s11 s33 − s213 s33 s211 − 2s213 s11 + 2s213 s12 (2.22a) c33 = s11 + s12 s11 s33 − 2s213 + s12 s33 (2.22b) c13 = s13 s11 s33 − 2s213 + s12 s33 (2.22c) −s16 s11 − s12 s66 1 = s44 1 = s66 c16 = (2.22d) c44 (2.22e) c66 (2.22f) (2.22g) 2.2.6 Rotations of constants A full set of elastic constants for a given direction can be used to determine the elastic constants for any arbitrary direction. Consider the original reference axes x, y, z , these can be rotated through angles θ and φ into the position x0 , y0 and z0 (see figure 2.2). Methodolgies for this are discussed at length in the literature (see (Nye, 1985),(Hearmon, 1957), (Lieberman and Zirinsky, 1956)). 50 Figure 2.2: Angles θ and φ used to describe rotated properties. χ shows the plane orthogonal to the vector described by these angles (a), b is the vector orthogonal to this. As a simple example, consider the rotation of the elastic constants of a cubic crystal 0 from the [100] to [110] direction. The constants Ti jkl is rotated to the new orientation Ti jkl according to: Ti0jkl = aim a jn ako alp Ti jkl (2.23) The rotation can be defined by the matrix, A, where a 11 a12 a13 A = a21 a22 a23 a31 a32 a33 (2.24) The tensor can be rotated about the z axis in a revised configuration using the formula: cosθ sinθ 0 A = −sinθ cosθ 0 0 0 1 51 (2.25) Thus: a11 = a22 = cosθ (2.26a) a33 = 1 (2.26b) a12 = sinθ (2.26c) a21 = −sinθ (2.26d) For a cubic crystal, the symmetry of the crystal means that there are only three unique elastic constants (in Voigt notation), for the principal axes (s11 , s22 , s33 ), the shear component (s44 , s55 , s66 ) and the perpendicular components (s12 , s32 , s23 ). This leads to any terms where the index is not one of these, or an equivalent index, being disregarded. Furthermore, as the rotation is about the z axis, any index (in tensor notation) where the 3 term is present can be ignored as this does not change. Thus: s01111 = s1111 − − − − − s1112 s1121 s1211 s2111 s1122 s1212 s1221 s2121 s2211 s2112 s2221 s1222 s2122 s2112 − s2222 − − − − − (2.27) This can be simplified using the symmetry of the crystal to become: s01111 = s1111 − − − − − s1122 s2211 s1212 s1221 s2121 s2112 s2112 − − − − s2222 − − − − − (2.28) Thus, converting the tensor notation to matrix notation, and including the above 52 0 0 factors, and applying the same methodology for s12 and s44 : 0 s11 = R · 2s11 + 2s12 + s44 0 s12 = R · 2s11 + 2s12 − s44 (2.29a) (2.29b) 0 s44 = R · 8(s11 − s12 ) (2.29c) For the example of mapping to the constants from [100] to [110]: cos π 4 A = −sin π4 0 sin π4 cos π4 0 0 0 1 (2.30) Thus: 0 1 s11 = (2s11 + 2s12 + s44 ) 4 0 1 s12 = (2s11 + 2s12 − s44 ) 4 0 1 s44 = (8(s11 − s12 )) 4 2.2.7 (2.31a) (2.31b) (2.31c) Relationship between elastic constants and engineering constants Traditionally, materials are characterised by the engineering constants, as these can be easily measured experimentally, and provide a useful indicator of a materials suitability for a specific application. These constants can be summarised as the Young’s (elastic), shear and bulk moduli and the Poisson’s ratios. The elastic constants can be used to derive the engineering constants for isotropic media (Ashby and Gibson, 1997): 53 εxx ε yy εzz = γ yz γzx γxy 1 E − Eν − Eν 0 0 − Eν 1 E − Eν 0 0 − Eν − Eν 1 E 0 0 0 0 0 0 1 G 0 0 0 0 1 G 0 0 0 0 0 0 0 0 0 0 1 G σxx σ yy σzz σ yz σzx σxy (2.32) Where ε signifies extensional strain, γ represents shear strain, ν is Poisson’s ratio, E is Young’s modulus, G is shear modulus and σ stress in a given direction. Engineering constants in terms of the elastic constants are as follows: 2.2.7.1 Young’s Modulus Young’s modulus relates the extensional or compressive strain to the stress in the direction of loading. For a given sample there are up to three Young’s (or elastic) moduli, corresponding to the principal axes. For n = 1,2,3: En = 2.2.7.2 1 snn (2.33) Poisson’s ratio An isotropic material has a Poisson’s ratio bounded by thermodynamics; the Young’s modulus, bulk modulus and Poisson’s ratios must have positive values, thus Poisson’s ratio must fall between -1 and 0.5 (see eqn. 2.2.7.4). Anisotropic materials are not restricted by these bounds (Ting and Chen, 2005). For an anisotropic material, an axial strain results in a major and a minor Poisson’s ratio corresponding to the ratios of the axial and lateral strains in each perpendicular direction. s21 s31 s32 , ν13 = − , ν23 = − s11 s11 s22 s12 s13 s23 = − , ν31 = − , ν32 = − . s22 s33 s33 ν12 = − ν21 54 (2.34) 2.2.7.3 Shear Modulus Shear modulus relates the shear stress to the shear strain. Shear modulus can be be known as modulus of rigidity as it describes the materials resistance to lateral rather than compressive loads. Again, there are three shear moduli corresponding to the faces normal to each principal axis. G23 = 2.2.7.4 1 , s44 G31 = 1 , s55 G12 = 1 . s66 (2.35) Bulk Modulus Bulk modulus is a measure of a materials resistance to uniform volumetric compression. For anisotropic media, this is k= 1 s11 + s22 + s33 + 2s12 + 2s13 + 2s23 (2.36) For cubic media, this factorises to: K= 1 , 3(s11 + 2s12 ) or for isotropic media: K= E . 3(1 − 2ν) (2.37) This form shows how the restricton that E, ν and K must be greater than zero which limits Poisson’s ratio for isotropic media, as the inequality (1 − 2ν) > 0 must be satisfied. 2.2.7.5 Zener ratio The Zener ratio is a measure of a materials anisotropy, based on the stiffnesses, for cubic materials: Z= 2C44 . c11 − c12 55 (2.38) A Zener ratio of 1 indicates isotropic behaviour. It is possible to characterise materials of lower symmetry by comparing the ratio of the maximum and minimum shear wave velocities in the material (Ledbetter and Migliori, 2006). 2.2.8 Application to auxetic media Investigation of Poisson’s ratio using classical elasticity theory for auxetic media has been developed for well over a hundred years and has provided an accurate and robust method for determining the elastic behaviour of materials but seldom reveals the mechanisms of this behaviour. In most materials, the Young’s modulus, shear modulus and Poisson’s ratio are directional. In an isotropic elastic medium, the stresses and strains are linked by Hooke’s law, thus = Eσ . For isotropic cubic media, Poisson’s ratio must be between -1 and 0.5 for the material to be stable. In an anisotropic material this is not the case. Ting and Chen (2005) show that Poisson’s ratio for anisotropic media can be unbounded. This is then developed by Ting (2004) [n.b.(Ting and Chen, 2005) was written and submitted prior to (Ting, 2004) but publication appears to have been delayed] who shows that the large transverse strain in this model does not comply with linear elasticity theory, and that transverse strain must be bounded to comply with these laws. Some symmetries have been investigated in detail; the stress-strain relationships for cubic crystals are shown by Thomas (1966), Hooke’s law can be generalised for an anisotropic cubic solid to take into account the direction of stresses and resulting strains and in a cubic system, the resulting compliance tensor can then be transformed depending on the orientation of the stress. Symmetry can be used to reduce the number of terms from 21 to 3 for a cubic anisotropic homogeneous material (Turley and Sines, 1971b) (Turley and Sines, 1971a). The resulting equations allow the elastic properties of the crystal under consideration to be plotted for different planes, rotating about a specified axis. Negative Poisson’s ratio in anisotropic media is considered by Ting and Barnett (2005). Using the equation of Hayes and Shuvalov (1998), it is shown that for a cubic crystal to be completely auxetic the compliance S12 must be greater than 0, however searches have yet to reveal such a crystal. The conditions for a non cubic/hexagonal crystal to be completely 56 auxetic are derived in terms of the compliances of the crystal, giving parameters from which auxetic behaviour can be found from elastic constants published in the literature. The relationship between Poisson’s ratio and elastic properties of an anisotropic crystal is further investigated by Guo and Wheeler (2006) who define the criteria for ν to be negative in terms of directional compliances and compared the analytical finding to the experimental work of Yeganeh-Haeri et al. (1992). The directions of auxeticity for monoclinic crystals, which have one plane of elastic symmetry, are considered by Rovati (2004). The number of elastic constants can be reduced to 13. Using these constants to find an expression for ν, it is shown that when the material is stretched in a plane orthogonal to the symmetry plane, the Poisson’s ratio in the same plane is always negative. Brańka et al. (2009) show analytically that it is possible for some crystals to exhibit a Poisson’s ratio as low as -5 in what they term a V3 direction, that is a direction very close to the [111] direction but not defined by Miller indices. Branka et al. discuss the condition by which this may be seen, based on the ratios of bulk, tetragonal shear and simple shear moduli. 57 2.3 2.3.1 Classical atomistic modelling techniques Introduction Atomic forces and structure can be modelled as mathematical functions which describe the potential energy of the system as a function of the interatomic separation (Finnis, 2003). These functions, known as classical potential models, can be used to model the elastic constants and lattice parameters of materials in order to determine their behaviour under pressure, temperature, strain etc. at an atomic level. Classical potentials are based on empirical data; the calculations are generally not computationally expensive but are reliant on the accuracy and availability of the experimental (or sometimes quantum) data on which they are based, for fitting parameters to be determined. Use of classical potential models can provide the elastic properties of the crystals under consideration, as well as insight into the mechanism of negative Poisson’s ratio behaviour. 2.3.2 Two Body Potentials There are many ways to classify chemical potentials, but for this study it is convenient to group them into two-body and many-body potentials. The two-body approximation assumes the interaction of a set of n atoms can be replaced by the sum of n(n−1) pointwise interactions. This is often reduced to those points where the separation is less than a specified distance, known as a cutoff; the presence of surrounding atoms does not affect the interaction of a pair of atoms. Two-body potentials accurately capture the directional nature of ionic or Van der Waals bonding, but are less well suited for metallic or covalent materials. Despite this, they can be a good first approximation, and reproduce both the structure and energy of many materials. Several such potential models have been developed over the last century to describe the inter-atomic potential energy of atoms. The first of these was by Morse (1929) who derived a two-body potential, based on the forces between two neighbours, many-body potentials followed to describe include near-neighbour effects. 58 2.3.2.1 The Morse Potential Figure 2.3: Schematic representation of a classical potential energy function, showing the variation of energy with interatomic spacing (arbitrary units). Morse proposed a solution to Schrödingers’ equation based on a simplified form of the energy function. Previous work has been based on series forms however these only work over a small range, and contain further approximations. Morse determined the energy function must fulfil several criteria; it must tend to a finite value as the inter-atomic separation tends to infinity, it must tend to infinity when r = 0 and it must have a minimum at r = r0 . The proposed potential takes the form: U(r) = De (1 − e−2α(rij −r0 ) − 2e−α(rij −r0 ) ) (2.39) D describes the energy, or the depth of the potential well, r0 the separation at the minimum energy and α the width of the potential well. The Morse potential was applied to cubic metals by Girifalco and Weizer (1959) to find the elastic constants, who concluded that the Morse potential could be used to study lattice deformations in cubic metals where the crystals were perfect however if defects were present, the electron distribution of the crystal would become distorted and the derived constants would no longer be applicable. Figure 2.3 shows an example schematic of the change in energy relative to interatomic spacing. 59 Figure 2.4: Schematic showing (left) a two-body, and (right) three-body angular spring potential 2.3.2.2 The Lennard-Jones Potential Lennard-Jones (1931) derived the two-body potential based on the Van der Waals forces in the system. A repulsive and a cohesive term are included in the model. The shape of the potential is such that for separations of r < r0 , the potential is very steep, indicating the strength of the repulsive interaction. The Leonard-Jones potential has been widely used to study the properties of liquids (Broughton et al., 1982) and is a good model for the solid states of the noble gases (Ashcroft and Mermin, 1976). " 12 6 # σ σ − U(r) = 4 r r (2.40) The ( 1r )12 term describes the repulsive element of the potential, the ( 1r )6 the cohesive term. and σ are the depth of the potential well and distance at which the energy becomes zero. The Leonard-Jones potential was used by Barron and Domb (1955) to determine the stability of hexagonally close packed structures. 2.3.2.3 The Cauchy relation Elastic constants calculated from a two-body potential are governed by the Cauchy relation. A two-body potential is only able to describe the change in energy resulting from a change in the interatomic spacing of two atoms, there is no rotational component of this energy. The potential is unable to account for the difference in energy change for a strain in the shear (ε1212 ) or lateral (ε1122 ) directions and subsequently, elastic constants predicted using a pair potential will always show s44 = s12 . 60 2.3.3 Many Body Potentials Two-body potentials have the advantage of being comparatively easy to solve, however they are obviously limited by only describing the forces on two adjacent atoms. To consider the effect of clusters of atoms, many-body potentials based on near-neighbour interactions have been developed. A comparison of pairwise and many-body forces is conducted by Holian et al. (1991). Two dimensional simulations are also conducted for high velocity impacts where the projectile penetrates the wall. It is found that the formation of defects in the structure is better represented in the many-body model. 2.3.3.1 ’Basic’ many-body potentials The most simple three-body potentials depend on the change of angle between three atoms (see figure 2.4). These can be extended to include a fourth atoms, and to include non linear effects. Examples of these types of potential include those based solely on change of angle such as the Linear-threebody and Three-body harmonic; those dependent solely on the separation, such as the Urey-Bradley potentials; and those dependent on both angle and separation such as the Axilrod-Teller potential which is particularly suited to modelling Van der Walls forces. 1 1 1 U = k2 (θ − θ0 )2 + k3 (θ − θ0 )3 + k4 (θ − θ0 )4 2 6 24 U= 1 k4 (θ − θ0 )4 24 1 U = − (r23 − r023 )2 2 1 + 3 cos γ cos γ cos γ i j k U = E0 3 ri j r jk rik (2.41a) (2.41b) (2.41c) (2.41d) Equation 2.41a shows the three-body linear potential, 2.41b the three-body harmonic, 2.41c the Urey-Bradley and 2.41d the Axilrod-Teller. θ represents a change in angle, r the interatomic separation, k the spring constant, γ the angle between the vectors ri j and E0 is a function of the energies of the p orbitals. 61 2.3.3.2 Embedded atom model for metals Daw and Baskes (1983) devised a semi-empirical model based on a generalisation of quasiatom theory, which they called the embedded atom method; each atom is viewed as being an impurity embedded in a host lattice comprising all other atoms in the model. Stott and Zaremba (1980) showed that the energy of an impurity in a host is a functional of the electron density of the host, without the impurity. It can be assumed that the embedding energy depends only on the environment around the impurity or the impurity has a uniform local electron density. This equation can be used to calculate the ground state properties of solids, however Daw and Baskes used experimental data to fit their model to enable it to calculate the elastic properties correctly. The equation is tested using a model of a slab of Nickel, a strain is applied to the model and the result compared to that of existing elastic theory and the data that were used to fit the model. The model is then modified to simulate vacancies in the slab, and then include a Hydrogen atom in the slab to model Hydrogen embrittlement. Daw and Baskes (1984) then develop the model to include a core-core repulsion term (modelled as a short pair-wise repulsion between cores). The assumptions that the host density is close to the sum of the atomic densities of the constituents means that the energy becomes a function of the position of the atoms. The lattice vectors can be calculated as a function of the position vectors to the near neighbours. Foiles et al. (1986) derive a set of embedding energies, pair interactions and atomic densities to be used with the embedded atom method for 6 transition metals. These were fitted using the most simple parameterised forms possible with the least parameters in order to maximise the number of alloys that could be modelled using these data. The functions were then validated by applying them to both the pure metals and alloys for bulk and surface calculations . Comparison with experimental data shows that generally agreement is found However in some cases, for instance calculation of surface energies, the experimental value is around 50% higher than the calculated value. Avinc and Dimitrov (1999) derive a generalised long-ranged potential using the Leonard-Jones potential, with an EAM element. The advantage of this potential over 62 the Sutton-Chen potential is that it provides a better estimation of the elastic constants, and an exact fit for the lattice parameters. This modified Leonard-Jones potential is most applicable to FCC transition metals with computed values being within around 10% of experiment data. When applied to BCC metals (Fe, Mo, Nb, W and V), the calculated values around generally far lower than the experimental data, around 50% in some cases. 2.3.3.3 The Finnis-Sinclair potential Finnis and Sinclair (1984) highlight the shortcomings of the Morse, and similar two-body potentials. The standard way of dealing with this is to include a term dependent on the macroscopic volume of the crystal. This leads to the problem whereby the bulk modulus calculated from constant volume differs from the bulk modulus calculated from a homogeneous deformation, a problem that is more apparent when internal cavities or cracks are present in the simulation. To combat this, Finnis and Sinclair propose a model that includes the band character of metallic cohesion. The cohesive energy is modelled √ as z where z is atomic co-ordination depending on the structure being considered, 1 for a diatomic molecule, 8 for a BCC crystal. The total energy is the sum of the many-body (UN ) and pair potential (UP ) terms: UFS = UN + UP X UN = −A f (ρi ) (2.42) (2.43) i UP = 1X V(Ri j ) 2 (2.44) ij ρi = X φ(Ri j ) (2.45) j Ri j = |Ri j | = |R j − Ri | (2.46) where R represents the interatomic separation, U the total energy, ρi the local charge density, σ the atomic charge densities and V the volume of the system. Unlike the twobody potentials, it is not possible to plot the change in energy relative to interatomic separation because of the potentials dependence on the local density. 63 The Finnis-Sinclair potential was extended by Dai et al. (2006) by extending the repulsive component. The result of this modification is that for BCC metals the pressure-volume relationship can be more accurately modelled, and also, some of the properties of FCC metals can be modelled, something not possible with the original formulation. 2.3.3.4 Sutton-Chen potential The potential suggested by Sutton and Chen (1990) is a long range version of the FinnisSinclair potential, adding a long-range Van der Waals term. The potential always favours the FCC structure over the BCC structure, and the cohesive energy of the BCC structure is always predicted to be less than the the energy of the FCC structure. For this reason, fitting parameters are only given for a range of FCC metals. The potential comprises an EAM functional term, and EAM density term and a pair potential. Fi (ρ̄i ) = − X p Ai ρ̄i (2.47) i ρ̄i = X Cr−6 ij (2.48) A r7 (2.49) i φi j (ri j ) = For the EAM functional term (equation 2.47), A is a fitting parameter, expressed in eV. For the EAM density term, C is a fitting parameter expressed in Å6 . For the pair potential component, A is expressed in eV Å7 . 2.3.3.5 The Cleri-Rosato potential Like the free-electron-gas model, the tight-binding model assumes electrons are independent of the atomic structure but contrary to the free-electron picture, the tight-binding model describes the electronic states starting from the limit of isolated-atom orbitals. This simple model gives good quantitative results for bands derived from strongly localised atomic orbitals (Cannini, 2008). 64 The Cleri-Rosato potential uses the second-moment approximation of the tight-binding model (Cleri and Rosato, 1993) to model the metallic nature of cohesive bonds up to the fifth neighbour. Applied to FCC metals it is found that the elastic constants, atomic volume and cohesive energies calculated are comparable with experimental data. The Cleri-Rosato potential calculates the energy of the system (Ec )as the sum of the band energy (EiB ) and the ion-ion repulsions (EiR ). Ec = X EiB + EiR (2.50) i 21 X EiB = − ξ−2 αβ −2qαβ rij /r −1 0 e j EiR = X j (2.51) αβ A αβ −Pαβ(rij /r −1) 0 αβe (2.52) αβ r0 is the first nearest neighbour bonding distance, ξ is an effective hopping integral, and q describes the dependence on the interatomic spacing. P describes the compressibility of the bulk metal. 65 2.4 2.4.1 Quantum Atomistic Modelling Techniques Introduction The empirical, potential based methods described in the previous section have been very successful in calculating elastic properties of crystals, but in some cases, the explicit, electronic nature of the interaction must be considered. This is especially true in the case of metals, where electrons are delocalised. The electrons must be treated quantum mechanically, and many theoretical methods have been developed to this effect. These are often referred to as ab initio, first principles, or quantum techniques. One of the most successful ab initio methods is Density Functional Theory (DFT). DFT has been chosen for this study as it is well adapted to periodic crystals and is implemented in mature, convenient to use software which allows complex analyses such as elastic constant calculations to be carried out quite easily. As is the case for most quantum based methods, DFT is complex and very subtle; its formalism has been described in many publications (see for example: (Parr, 1983; Koch et al., 2001; Sholl and Steckel, 2009)). A brief description of the methodology employed by DFT is given in Appendix A, this section provides an overview of applications of DFT to determine the elastic properties of materials. 2.4.2 Application to auxetic media Density functional theory has been used to investigate the structural and elastic properties of a range of crystals. Studies employing ab intio methods to investigate auxetic behaviour have primarily been confined to silicates and other similar framework molecules, as discussed in section 1.1.2. The use of density functional theory to investigate the elastic properties of cubic metals has been conducted, however these studies have generally focussed on the structural, and on-axis properties, rather than auxetic behaviour. 2.4.2.1 Selected applications of DFT to find elastic properties Pour et al. (2009) used DFT to investigate the properties of biprismanes. These are polycyclic hydrocarbons, made up of an internal and external ring structure, that exhibit negative Poisson’s ratio. It was initially suggested that this behaviour was driven by the 66 bow-tie shape of the external rings of the molecule. This would mirror the mechanism of auxetic re-entrant honeycombs (see section 1) however due to the way that the external rings shrink under stretching, this was not found to be the case. It was found from DFT calculations that whilst the external rings of the structure undergo a positive Poisson’s ratio deformation, it is the internal rings of the structure that drive the auxetic behaviour through a change in internal angles. Hammerschmidt et al. (2007) used both analytical modelling and DFT to determine the elastic behaviour of InAs and GaAs crystals under a biaxial strain. Negative Poisson’s ratio as low as -3.21 is found in the InAs for very high, non-linear strains. Cubic metals are explored in this thesis, using both classical and quantum methods. The partially filled d shells of the cubic transition metals complicates prediction of their elastic properties. Wills et al. (1992) investigate the properties of the transition metals using DFT to try to find trends in the total energies and the values for c’ c0 = c11 − c12 . 2 (2.53) It is found that for crystals where there is a very large difference between the energies of the BCC and FCC configurations, the value of c’ is very high. Iridium and its Ir3 X compounds have a face-centred cubic structure. Chen et al. (2003) use a Generalised Gradient Approximation (see Appendix A) approach to find the elastic constants of these structures by distorting the lattice and computing the resulting energies. The Poisson’s ratio of the crystals is computed, and this is correlated to the ratio of shear modulus to bulk modulus to find a measure of a material’s brittleness in terms of Poisson’s ratio. Sin’ko and Smirnov (2002) used ab-initio methods to model Aluminium in its BCC, HCP and FCC phase to determine its structural and elastic properties under a range of pressures and temperatures. Keskar and Chelikowsky (1992) explicitly explore auxetic behaviour in α-cristobalite, comparing the results gleaned from DFT to those from a pair potential. The Poisson’s ratios found are -0.2 and -0.17, and from these calculations the mechanism for NPR in α-cristobalite is derived. It is shown that the mechanism is dependent on the rigidity of the tetrahedra. This was determined by varying the rigidity 67 of the tetrahedra, straining the structure and finding that Poisson’s ratio becomes more negative when the rigidity of the tetrahedra is increased. 2.4.3 Conclusions Whilst no investigations have explicitly considered negative Poisson’s ratio in cubic elemental metals using DFT, studies have been conducted where auxetic behaviour of crystals has been modelled, and where the elastic properties of cubic metals have been calculated when modelling on-axis properties. 68 2.5 The finite-element method The finite element method has been used to explore the properties of a variety of structures that exhibit negative Poisson’s ratio, such as those described in section 1.1.3. A brief explanation of the methodology of the finite-element method is given in Appendix B. Most of the studies where the finite element method has been used to model auxetic media have characterised the auxetic behaviour in honeycombs. The re-entrant honeycomb, first reported by Lakes (1987) has been investigated extensively. Evans et al. (1994) modelled re-entrant cells using both analytical and finite element methods and showed that the Poisson’s ratio of the structure could be varied from 0.6 to -1.2 dependent on the geometry of the structure being considered. Whitty et al. (2003) used the finite element method to investigate the effect of geometry variation on the failure method for both conventional and auxetic honeycombs. Bezazi et al. (2008) also used the finite element method to investigate the properties of honeycombs with differing geometries. Scarpa et al. (2008) extended the use of this technique to auxetic tubes. In addition to re-entrant structures, studies have investigated the mechanical properties of chiral honeycombs. The structure of chiral honeycombs, being made up of a central cylinder and tangential ribs gives the possibility of designing a structure with a Poisson’s ratio of -1 whilst tailoring the elastic properties. The original structure was proposed by Prall and Lakes (1997), but more recently these structures have been investigated using both experimental testing and the finite-element method. The finite element has been used to investigate the buckling behaviour (Spadoni et al., 2005), the wave propagation (Tee et al., 2010), the out of plane properties (Miller et al., 2010; Lorato et al., 2010) and in-plane properties (Alderson et al., 2010). In summary, from the literature surveyed it is apparent that the use of the finite element method provides a method for investigating the structural properties of auxetic properties in an efficient and robust manner, particularly where a range of geometries are to be considered. 69 Chapter 3 Fundamental atomistic modelling of the elastic properties of cubic crystals 3.1 Introduction In this chapter, analytical expressions of the elastic constants and Poisson’s ratio of BCC and FCC crystals are derived from simple models of interatomic interactions. In principle, this is very similar to the discussions of Young’s modulus for simple cubic crystals found in basic materials science textbooks (for example Ashby (2011)), but these analyses determine the influence of the form of the potential model and the number of nearest neighbour interactions on the bounds of Poisson’s ratio. For more realistic systems, albeit still rather simple, the mathematical effort is already considerable and limits the analysis to second nearest neighbours and simple potential forms. The analytical nature of the models allow the unambiguous discovery of trends and relationships. This section aims to explore the hypotheses that the BCC structure is inherently more auxetic than the FCC one, the inclusion of second neighbour interactions also increases the auxeticity, while a volume based many-body contribution decreases it. This information is very useful when interpreting the results obtained from the sophisticated models of interaction used in the following chapters. 70 3.2 3.2.1 Modelling methodology Introduction Whilst numerical modelling of crystal structures allows prediction and visualisation of the elastic behaviour, analytical modelling allows a deeper understanding of how the potential form can affect the predicted properties and explain the results that are observed. This section details the foundation of the classical potentials, and the strains imposed in order to calculate the elastic properties. 3.2.2 Derivation of elastic energy The elastic energy per unit volume is the integral of stress with respect to strain: Eelas = 1X 1X σi · ε j = ci j · εi · ε j . 2 2 i,j (3.1) i, j For a cubic crystal, the elastic stiffness matrix is greatly simplified, with only three independent components, owing to the symmetry of the crystal. ci j = c11 c12 c13 0 0 0 c21 c22 c23 0 0 0 c31 c32 c33 0 0 0 0 0 0 c44 0 0 0 0 0 0 c55 0 0 0 0 0 0 c66 . (3.2) Therefore, the elastic energy (equation 3.1) simplifies to: Eelas = 1 2 c11 ε1 + ε22 + ε23 + c12 (ε1 ε2 + ε2 ε3 + ε3 ε1 ) + c44 ε24 + ε25 + ε26 . 2 (3.3) It can be seen that application of specific strains makes it possible to vary the dependence of the elastic energy, and thus to determine the elastic constants. In the cubic case, three strains are sufficient (axial compression, tetragonal shear and rhombohedral shear), but note that c12 can never be obtained directly by itself (it requires two strain components 71 to be non-zero, and this in turn requires c11 ). 3.2.3 Strain To apply strain to lattice cells, it is more convenient to use the tensor notation and to separate the direction of strain from its amplitude by using a scalar, γ, and a symmetric matrix (order 2 tensor), T: εαβ = γ · Tαβ . (3.4) The new lattice vectors coordinates, a0iα are then obtained from the original lattice vector coordinates, aiα , by: a0iα = aiα + γ X Tαβ aiβ . (3.5) β The strained volume is the determinant of the matrix a0iα . The elastic energy per volume can be rewritten in terms of the strain (transformation) matrix (Finnis, 2003), which in turn can be shown to be the sum of the energy contributions from the axial (K), tetragonal shear (c0 ) and rhombohedral shear (c44 ) deformations: Eelas K = elas elas Eelas = Eelas K + Ec0 + Ec44 (3.6) 1 (c11 + 2c12 )(Txx + T yy + Tzz )2 γ2 6 (3.7) 1 2 2 2 2 Eelas c0 = (c11 − c12 )(Txx + T yy + Tzz − Txx T yy − T yy Tzz − Tzz Txx )γ 3 (3.8) 2 2 2 2 Eelas c44 = 2c44 (Txy + T yx + Tzx )γ . (3.9) The energy can be rearranged to determine the bulk modulus, 1 K = (c11 + 2c12 ), 3 and a cubic shear modulus, 72 (3.10) c‘ = (c11 − c12 ). (3.11) To determine the elastic constants for a cubic medium the bulk, tetragonal and rhombohedral shear strains can be applied, the change in energy calculated and the elastic constants determined. 3.2.3.1 Calculation of bulk modulus The obvious strain to obtain the bulk modulus (K) is a triaxial compression (equal compression in x, y and z). This gives the change in energy, from which the axial elastic constants can be determined: 1 0 0 T = 0 1 0 0 0 1 (3.12) thus the scaled lattice parameters become: a(1 + γ) 0 0 a0 = 0 a(1 + γ) 0 0 0 a(1 + γ) (3.13) The determinant of this matrix is a3 (γ + 1)3 and so for this simple isotropic strain, the energy can then be expressed in terms of the applied strain (γ) and the elastic constants c11 and c12 : Eelas = K 3 (c11 + 2c12 )γ2 2 (3.14) or, the change in energy with respect to the change in strain: 2 elas 1 d (EK ) (c11 + 2c12 ) = 3 dγ2 73 (3.15) 3.2.3.2 Calculation of tetragonal shear modulus In order to obtain the first, or tetragonal, shear modulus c’ , the strain 1 0 0 T = 0 −1 0 0 0 0 (3.16) is applied (for example, in the case of an FCC system, figure 3.2). From this, the strained lattice parameters are: a(1 + γ) 0 0 a0 = 0 a(1 − γ) 0 . 0 0 a (3.17) The energy can then be calculated as: Eelas = (c11 − c12 )γ2 , c0 (3.18) and the lattice parameters in terms of the energy expressed as: (c11 − c12 ) = 3.2.3.3 1 d2 Ec‘ 2 dγ2 (3.19) Calculation of rhombohedral shear modulus To find the three elastic constants that describe the elastic properties of cubic media, it is necessary to find the rhombohedral shear modulus, C44 . As with the first two strains, the strain corresponding to the rhombohedral shear deformation is applied: a aγ 0 a0 = aγ a 0 0 0 a This results in an energy of 74 (3.20) 2 Eelas c44 = 2c44 γ (3.21) and as with the previous strains, the corresponding elastic constant calculated: c44 = 3.2.3.4 1 d2 Ec44 . 4 dγ2 (3.22) Summary: Finding elastic properties from strains The methodology to obtain the elastic constants from a model of the interaction energy can be summarised as four stages that are applicable to all the potentials considered: • Apply the relevant strains described in section 3.2.3 • Express the geometric changes (distances, volume) as functions of γ • Find the energy. The second derivatives of the energy with respect to γ then lead to combinations of elastic constants. • Solve these linear equations to find the elastic constants This methodology can then be applied to a range of simple potential models to determine the predicted elastic properties of the cubic crystals. 3.2.3.5 Poisson’s ratio Poisson’s ratio, in terms of elastic compliances, is equal to tions, −s012 s011 −s12 s11 , or for rotated configura- . As the compliance matrix is the inverse of the stiffness matrix, the compliance ratios can also be expressed as functions of the stiffnesses. Most elemental cubic metals exhibit negative Poisson’s ratios, when they occur, in the [110] (and perpendicular) directions. As an aside, it should be noted that whilst experimental data usually shows auxetic behaviour in the [110] direction, theoretically Poisson’s ratios as low as -5 can be seen for strains in specific directions close to the [111] direction (Ting and Chen, 2005; Ting, 2004; Brańka et al., 2009). 75 The following formulae (found using the methodology described in section 2.2.6) enables the determination of in-plane Poisson’s ratio for an axial strain and corresponding directions of interest. −s12 c12 = s11 c11 + c12 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) = 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 = 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) ν[010][100] = ν[11̄0][110] ν[001][110] 3.2.4 (3.23) (3.24) (3.25) Geometry considerations for cubic crystal systems Potential energy models usually depend on interatomic distances or angular change and it is therefore useful to compute these discretely in order that they can be applied in the potential functions under consideration. This section calculates the geometry change for face-centred and body-centred cubic crystals, and these will later be applied to section 3.3. 3.2.4.1 Face-centred cubic unit cell Figure 3.1: A face centred cubic unit cell, showing the 4 atoms (large, numbered), and some periodic images (small). The 12 nearest neighbour bonds of atom 4 are shown in red, using the standard chemical convention for perspective. 76 There are 4 atoms in a FCC unit cell, and each atom has 12 first nearest neighbours (4 in each basal plane (see Fig. 3.1)). The first nearest neighbour distance is r, and for the original cell: √ r = r0 = Tri-axial compression: 2 a 2 (3.26) Tri-axial compression, for calculation of the bulk modulus is simply a scaling of the lattice parameters: √ r = (1 + γ)r0 = 2 (1 + γ)a 2 (3.27) Tetragonal shear: For the first (tetragonal) shear, c0 , each basal plane must be treated independently, thus resulting in equations 3.28-3.30 to describe the xy, xz and yz planes, as shown in figure 3.2: q q a 2 2 rxy = (1 + γ) + (1 − γ) = r0 1 + γ2 2 r q a 1 rxz = (1 + γ)2 + 1 = r0 1 + γ + γ2 2 2 r q a 1 r yz = (1 − γ)2 + 1 = r0 1 − γ + γ2 2 2 Figure 3.2: Strains applied for c’ shear in an FCC structure. 77 (3.28) (3.29) (3.30) Rhombohedral shear: Calculation of the second shear, c44 for the face-centred cubic configuration has three components, for strains in the xy, xz and yz planes (see figure 3.3). q a (1 + γ)2 + (1 + γ)2 = r0 (1 + γ2 ) = 2 q a r−xz = (1 + γ)2 + (1 − γ)2 = r0 (1 − γ) 2 r q γ2 a 1 + (1 − γ)2 = r0 1 − r yz = 2 2 r+xy (3.31) (3.32) (3.33) Figure 3.3: Strains applied for c44 shear in an FCC structure. 3.2.4.2 Body-centred cubic unit cell Figure 3.4: A body centred cubic (BCC) unit cell, showing the 2 atoms (labelled 1 and 2), and some periodic images (small). The 8 nearest neighbour bonds of atom 2 are shown in red, using the standard chemical convention for perspective. Using the methodology outlined above for the face centred unit cell, the body-centred cubic configuration can be modelled in the same way, accounting for the change in the primitive unit cell. 78 The body centred cubic crystal has a primitive unit cell of two atoms, which can be visualised as being located at a vertex and centre point of a cube (labelled atoms 1 & 2 in figure 3.4). Each atom within the structure has eight nearest neighbours, and each plane adjacent to the central atom comprises four atoms. The first nearest neighbour distance is therefore √ 3 r = r0 = a. 2 Bulk modulus: (3.34) Applying a tri-axial compression to the body-centred cubic unit cell gives a simple scaling of the lattice vector, to derive the bulk modulus: √ 3 a r = (1 + γ)r0 = (1 + γ) 2 (3.35) Figure 3.5: Strain for C‘ shear for body-centred cubic, (110) plane Tetragonal shear: The first shear modulus, c‘, is described in terms of the (110) and (101) planes, as shown in figure 3.5: r110 = r101 a = 2 r q 2 2 + 2γ2 + 1 = r0 1 + γ2 3 (3.36) Figure 3.6: Shear displacement for c44 , showing a) (110), b) (11̄0), c) (101) planes 79 Rhombohedral shear modulus: To find the second shear modulus, c44 , the bonds on the sheared (110), (11̄0)& (101) planes (see figure 3.6) must be considered separately: r q 2 4 r11̄0 2(1 + γ)2 + 1 = r0 1 + γ + γ2 3 3 r q a 2 4 r110 = 2(1 − γ)2 + 1 = r0 1 − γ + γ2 , 2 3 3 r q a 2 r101 = 3 + 2γ2 = r0 1 + γ2 . 2 3 a = 2 (3.37) (3.38) (3.39) Having described the strains required to calculate the elastic properties from the fundamental potentials, these can be applied to a volumetric, spring and Lennard-Jones potentials and the results are considered in section 3.3. 80 3.3 Elastic properties calculated from analytical modelling This section describes the elastic properties calculated from the simple potential models, starting with a very simple volume dependent model, increasing in complexity to a simple spring model and then a Lennard-Jones potential. 3.3.1 A simple model: volume dependence To model the relationship between the energy and elastic properties of solids, a fundamental volumetric model is used: Eelas V = 1 kv · (V − V0 )2 2 V0 (3.40) where kv is an equivalent spring constant (the parameter used to describe the linear bond stiffness) and V and, V0 , are the strained and original cell volume. For a cubic cell of length a, the original volume V 0 is a3 . This simple model neglects the Bravais lattice structure of the crystal (whether BCC or FCC) and as there is only one fitting parameter, it is simple to examine the characteristics of the model. 3.3.1.1 Bulk modulus Under tri-axial compression, the new volume is simply obtained from equation 3.13, and thus the energy can be found by substituting this into equation 3.40: Eelas K = V1 = a3 (1 + γ)3 (3.41) kv · (a3 (1 + γ)3 − a3 )2 2a3 (3.42) A Taylor expansion yields: 9 3 2 a kv γ + . . . 2 Equating this with equation 3.14 gives 81 (3.43) (c11 + 2c12 ) = 3a3 kv . 3.3.1.2 (3.44) Tetragonal shear The same process is repeated for the first shear modulus with the associated tetragonal strain: V0 = a3 (1 + γ)(1 − γ). (3.45) The energy can be found from equation 3.18 Eelas c0 = kv · (a3 (1 + γ)(1 − γ) − a3 )2 2a3 −1 3 4 Eelas a kv γ , c0 = 2 (3.46) (3.47) Recalling equation 3.21, it is apparent that there is no term in γ2 , and thus no contribution from this potential energy form to the c‘ shear: 3.3.1.3 (c11 − c12 )γ2 = 0. (3.48) V0 = a3 (1 − γ2 ) (3.49) Rhombohedral shear Eelas c44 = kv · (a3 (1 − γ2 ) − a3 )2 2a3 1 3 4 Eelas c44 = a kv γ . 2 (3.50) (3.51) As with the first shear derivation, expanding equation 3.50 it is apparent that there is no term in γ2 , so again there is no contribution from this potential energy form to the shear, and c44 = 0. 82 3.3.1.4 Poisson’s ratios Equations 3.14, 3.47 and 3.50 can be solved to obtain: c11 = c12 = a3 kv (3.52) c44 = 0 (3.53) Using these values the Poisson’s ratios for specific strains can be found : s12 c12 1 = = s11 c11 + c12 2 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) = −1 ν[11̄0][110] = 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 ν[001][110] = =0 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) ν[010][100] = 3.3.1.5 (3.54) (3.55) (3.56) Material properties from volume potential The volume potential gives only a primitive representation of material properties but is still effectively a many-body potential, being a function of the homogenised material rather than describing specific inter-atomic interactions. The elastic constants generated from this model are not governed by the Cauchy relation, however there is no shear rigidity in this model as it only describes the energy under a volume change. A pure shear conserves volume, hence c44 = 0. The axial (c11 ) and lateral (c12 ) components are equal, as the model is unable to account for the mode of the volume change. The negative Poisson’s ratio for the ν(11̄0,110) direction shows that the ’default’ behaviour for the simple potential is to predict auxetic behaviour even when the relationships between the axial and tangential stiffnesses are not fully defined. 83 3.3.2 A simple pair potential: The spring model This potential assumes that the bonds can be represented by linear springs. It is the simplest way to represent pairwise bonded interactions, and the energy (per volume) of a bond is given by Hooke’s Law: Eelas = r 3.3.2.1 1 kr · (r − r0 )2 2 V0 (3.57) Application to a face-centred cubic unit cell As described in section 3.2.4.1, there are 4 atoms in a one primitive face-centred cubic unit cell, and each has 12 first neighbours. Each bond is shared between two atoms, thus there are 24 bonds, 8 in each basal plane. 3.3.2.1.1 Bulk modulus: By replacing r with a triaxial strain (see section 3.2.4.1), the energy for this strain, and thus the bulk modulus are derived: Eelas r 3.3.2.1.2 √ !2 1 kr kr 2 kr 2 = 24 · · (r − r0 ) = 12 3 a (1 + γ − 1)2 = 6 γ2 2 V0 2 a a kr 3 (c11 + c12 )γ2 = 6 γ2 2 a kr (c11 + c12 ) = 4 a (3.58) (3.59) (3.60) Tetragonal shear: The shear strain, as described in section 3.2.4.1, has components to describe the strains in the xy, xz and yz directions. Eelas r r 2 r 2 q 2 1 1 1 kr = 8 r0 1 + γ2 − r0 + 8 r0 1 + γ + γ2 − r0 + 8 r0 1 − γ + γ2 − r0 2 V0 2 2 (3.61) 84 Expanding this Taylor series gives: (c11 − c12 ) = 3.3.2.1.3 kr a (3.62) Rhombohedral shear: For the strains described in section 3.2.4.1, there are three strains that are applied. These are then summed to give the total energy: Eelas r 1 kr = 2 V0 2 r 1 2 2 4 r0 (1 + γ) − r0 + 4 r0 (1 − γ) − r0 + 16 r0 1 + γ2 − r0 2 Eelas = r 3.3.2.1.4 1 (4γ2 + 4γ2 ) 4a kr c44 = a (3.63) (3.64) (3.65) Poisson’s ratio: Having determined the relationship between the bulk modulus and shear modulus, the elastic constants can be found from equations 3.60, 3.62 & 3.65: c11 = 2 kr a c12 = c44 = (3.66) kr a (3.67) The simple spring model is a two-body potential and thus has no rotational component, resulting in the Cauchy relation being obeyed. From the elastic constants, the analytical values of Poisson’s ratio for the directions shown in section 3.3.1.4 are c12 1 = c11 + c12 3 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) =0 ν[11̄0][110] = − 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 1 ν[001][110] = = . 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 2 ν[010][100] = 85 (3.68) (3.69) (3.70) 3.3.2.2 Application to the body-centred cubic unit cell The same paradigm can be applied to the body-centred cubic unit cell, to give the elastic properties in terms of the pair potential: 3.3.2.2.1 Bulk modulus: Each atom within the body-centred cubic unit cell has eight nearest neighbours, at a √ spacing of 3 2 a, so for a triaxial compression √ Eelas r 3.3.2.2.2 !2 3 a (1 + γ − 1)2 2 kr Eelas = 3 γ2 r a 3 kr (c11 + 2c12 )γ2 = 3 γ2 2 a kr (c11 + 2c12 ) = 2 a kr =8 2 2a (3.71) (3.72) (3.73) (3.74) Tetragonal Shear modulus: Eelas r kr =4 3 a r 2 2 1 + γ2 − 1 3 (3.75) Expanding this as a Taylor series, there is no term in γ2 , E= 4 k γ4 + ... 9 a3 (3.76) and thus c11 − c12 = 0. 3.3.2.2.3 Eelas r Rhombohedral shear modulus: r 2 r 2 r 1 4 2 4 2 2 = 2 r0 1 + γ + γ2 − r0 + 2 r0 1 − γ + γ2 + 4 r0 1 + γ2 (3.77) 2 3 3 3 3 3 2c44 γ2 = c44 = c12 86 4 kr 2 γ 3a 2 kr = 3a (3.78) (3.79) 3.3.2.2.4 Poisson’s ratio: From equations 3.74, 3.76 & 3.79, we find the elastic constants are equal for this model, and thus the Poisson’s ratios can be calculated: c11 = c12 = c44 c12 1 = c11 + c12 2 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) = −1 ν[11̄0][110] = − 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 ν[001][110] = = 2. 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) ν[010][100] = 3.3.2.3 (3.80) (3.81) (3.82) (3.83) Analysis The spring model is a more sophisticated model than the volume model, as it includes terms to describe the deformations of specific interatomic attractions in the crystal, hence having a different form for the body-centred and face-centred cubic lattices. The face-centred cubic crystal model predicts elastic constants which obey the Cauchy relation; the body-centred cubic form predicts that c11 = c44 = c12 . The predicted Poisson’s ratios calculated from equations 3.81, 3.82 and 3.83 show that for the face-centred cubic unit cell, the predicted Poisson’s ratio in the (110,11̄0) direction is zero, not negative, despite the Cauchy relation being obeyed. This is consistent with the theory that the negative Poisson’s ratio behaviour in the face centred cubic crystal is not directly governed by first-nearest neighbour axial interactions, but by the rotational rigidity of the bonds, something which is not accounted for in this model (This effect is explored in depth using numerical modelling techniques, and described in chapter 4). The magnitude of the stiffness of the interactions is also greater for the face-centred cubic cell than the body-centred cubic cell, presumably owing to the greater number of interactions. 3.3.2.4 Combination of the spring and volumetric potentials The volumetric potential alone is not able to successfully predict Poisson’s ratios, however, by combining the spring and volume potentials the many-body character of the volume 87 potential can offset the Cauchy relation characteristics of the spring potential. Using the calculated values for the volume and spring potentials for each lattice, the ratios of the elastic constants can be simplified. If the potential parameters are described by the term ε, the elastic constants for the face-centred cubic become c11 = a3 kv + 2 kr = 1 + 2ε a kr =ε a kr = a3 kv + = 1 + ε a 1 ν(11̄0,110) = 4 c44 = c12 (3.84) (3.85) (3.86) (3.87) This indicates that by adding a many-body component to the potential form, the predicted Poisson’s ratio changes from negative to positive. 88 3.3.3 The Lennard-Jones Potential The Lennard-Jones (1931) potential is a non-linear, two-body potential which describes the interatomic potential energy as: Er = B −A + 12 r6 r (3.88) Where A and B are fitting parameters, and r is the interatomic separation. The Lennard-Jones potential is an anharmonic potential that accounts for the fact that there is both an attractive and repulsive component in interatomic bonding. Given that only two parameters govern the potential, elastic constants can be derived analytically. 3.3.3.1 Face Centred Cubic cells 3.3.3.1.1 First nearest neighbours Initially, the potential was truncated such that only first nearest neighbours interact. In this case, the total energy (for a cell containing 4 atoms) is −A B Er = 24 6 + 12 r0 r0 B −A Er = 24 √ 6 + √ 12 2 2 2 a 2 a (3.89) (3.90) where a is the lattice parameter and r0 is the first nearest neighbour bond distance (for √ FCC, r0 = 2 2 a). To find the equilibrium lattice parameter a1nn , the derivative of the energy is set to zero. The first derivative of the interatomic energy with respect to strain is the inter atomic force. At equilibrium the force is zero thus: dE 6 · A 12 · B = − 13 dr0 r70 r0 and 89 (3.91) a1nn √ = 2 r 6 2B A (3.92) The strains applied for the previous potentials under consideration can be applied in the same manner, using the lattice parameter calculated in equation 3.92. Bulk Modulus strains: r B =4 A r 6 2B r1 = γ + 1 A 2 A2 A E1 = 24 6 12 − 4 γ+1 B 2 γ+1 B V1nn = E1 = − a31nn 6 A2 216 A2 γ2 1512 A2 γ3 + − B B B 216A2 3 (c11 + 2c12 ) = 2 B 144A2 (c11 + 2c12 ) = B (3.93) (3.94) (3.95) (3.96) (3.97) (3.98) Tetragonal shear: r q q 6 2B 1 + γ2 r1xy = r0 1 + γ2 = A r r r γ2 γ2 6 2B r1xz = r0 1 + γ + = +1 γ+ 2 A 2 r r r γ2 γ2 6 2B r1yz = r0 1 − γ + = −γ + +1 2 A 2 90 (3.99) (3.100) (3.101) −A B −A B −A B E1 = 8 6 + 12 + 6 + 12 + 6 + 12 r1xy r1xy r1xz r1xz r1yz r1yz A2 A2 A2 + − E1 = 8 − 3 . . . 2 2 γ2 + 13 B 4 γ2 + 16 B γ +γ+1 B 2 (3.102) (3.103) 2 E1 = − 6 A2 36 A2 γ2 195 A2 γ4 + + + ... B B B A2 c11 − c12 = 36 . B (3.104) (3.105) Rhombohedral shear: r1xyp = r0 (1 + γ) r1xym = r0 (1 − γ) r γ2 r1x = r0 1 + 2 −A −A −A B B B E1 = 4 6 + 12 + 6 + 12 + 6 + 12 r1xyp r1xyp r1xym r1xym r1x r1x 2 2 A2 A A2 A + 4 E1 = 16 − − 6 3 6 12 γ2 γ2 2 γ + 1 B 4 γ + 1 B 4 2 + 1 B + 1 B 2 2 2 2 A A +4 6 12 − 4 1−γ B 2 1−γ B E1 = − 6 A2 72 A2 γ2 + + ... B B A2 c44 = 36 B 91 (3.106) (3.107) (3.108) (3.109) (3.110) (3.111) (3.112) Poisson’s ratio: c11 = 2A2 q B AB A2 q c12 = c44 = B (3.113) (3.114) B A 1 c12 = c11 + c12 3 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) ν[11̄0][110] = − =0 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 1 4c12 c44 = . ν[001][110] = 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 2 ν[010][100] = 3.3.3.1.2 (3.115) (3.116) (3.117) Second nearest neighbours The interatomic distance varies if the Lennard-Jones is also allowed to act between the second nearest neighbours. Therefore, the equilibrium lattice parameters must first be calculated. Recalling the structure of the face-centred cubic unit cell, there are 24 first √ and 12 second nearest neighbour bonds to be considered at a separation of 2 2 a and a respectively. The strains applied to calculate the properties for the first nearest neighbour interactions can be applied to find the second nearest neighbour interactions, and the geometry change of these bonds summed with those of the first-nearest neighbour bonds to find the total energy. E2nn A A B B = 24 − √ 6 + √ 12 + 12 − 6 + 12 a2nn a2nn 2 2 2 a2nn 2 a2nn (2 · 8 + 1)A (2 · 64 + 1)B + E2nn = 12 − a62nn a12 2nn (3.118) (3.119) From the first derivative of this, the interatomic spacing is: r a2nn = 6 92 258B 17A (3.120) Comparing this to equation 3.92 it is evident adding nearest neighbours contracts the cell. This revised lattice parameter can be used to calculate the energies of the strained crystal so that the effect off adding a greater number of nearest neighbour bonds in the model can be observed. Bulk modulus: Applying the relevant strains, the energy component for the bulk modulus, from both the first and second nearest neighbour can be found. The distances for the first, r1 , and second nearest neighbour, r2 , under a strain γ can be found. q 6 r1 = q 6 r2 = 258 AB (γ + 1) √ √ 6 2 17 (3.121) 258 AB (γ + 1) √ 6 17 (3.122) The total energy is is the sum of the energy of the first and second nearest neighbour energies: Etot = 12 Enn1 = −A B 4624 A2 68 A2 + = − 6 r61 r12 16641 γ + 1 12 B 129 γ + 1 B 1 (3.123) Enn2 = −A B 289 A2 17 A2 + = − 6 r62 r12 66564 γ + 1 12 B 258 γ + 1 B 2 (3.124) Etot = Enn1 + Enn2 (3.125) 2 2 289 A2 17 A2 4624 A 68 A 6 + 24 6 12 − 12 − 66564 γ + 1 B 258 γ + 1 B 16641 γ + 1 B 129 γ + 1 B (3.126) Expanding this as a Taylor series yields 93 Etot 289 A2 10404 A2 γ2 72828 A2 γ3 =− + − + ... 43 B 43 B 43 B 3 10404 A2 (c11 + 2c12 ) = 2 43B (3.127) (3.128) Tetragonal shear modulus: The shear strains (rnm ) can then be applied to each of the shear planes, to find the resulting energies (Enm ), which can then be summed to find the total energy. q γ2 + γ + 1 rxy = √ √ 6 2 17 q q γ2 6 258B A 2 +γ+1 rxz = √ √ 6 2 17 q q γ2 6 258B A 2 −γ+1 r yz = √ √ 6 2 17 q q q 6 258B 6 258B 6 258B (1 + γ) (1 − γ) A A A rx = ry = rz = √ √ √ 6 6 6 17 17 17 2 2 4624 A 68 A Exy = 8 − 6 3 16641 γ2 + 1 B 129 γ2 + 1 B 4624 A2 68 A2 Exz = 8 − 2 6 2 3 16641 γ + γ + 1 B 129 γ + γ + 1 B 2 2 6 258B A p 94 (3.129) (3.130) (3.131) (3.132) (3.133) (3.134) E yz 68 A2 4624 A2 − = 8 2 6 2 3 16641 γ − γ + 1 B 129 γ − γ + 1 B 2 2 17 A2 289 A2 Ex = 4 6 12 − 66564 γ + 1 B 258 γ + 1 B 17 A2 289 A2 E y = 4 6 12 − 66564 1 − γ B 258 1 − γ B Ez = − 4097A2 16641B Etot = Ex + E y + Ez + Exy + Exz + E yz (3.135) (3.136) (3.137) (3.138) (3.139) A Taylor expansions reduces this to Etot = − 289 A2 61064 A2 γ2 369410 A2 γ4 + + + ... 43 B 1849 B 1849 B (3.140) From this, (c11 − c12 ) = 61064 A2 1849 B (3.141) Rhombohedral shear modulus q 6 258 AB (1 + γ) 2 rxyp = ann2 (1 + γ) = √ √ 6 2 2 17 q 6 √ 258 AB (1 − γ) 2 rxym = ann2 (1 − γ) = √ √ 6 2 2 17 q q r γ2 6 √ B 258 A 2 + 1 2 1 2 r yz = ann2 1 + γ = √ √ 6 2 2 2 17 q 6 258 AB rz = ann2 = √ 6 17 q p 6 q 258 AB γ2 + 1 rx = ann2 1 + γ2 = √ 6 17 √ 95 (3.142) (3.143) (3.144) (3.145) (3.146) The energy for each strain can then be found, these summed and expanded as a Taylor series: 289 A2 474368 A2 γ2 13929766 A2 γ4 + + + ... − 43 B 5547 B 5547 B 237184A2 c44 = 5547 B (3.147) (3.148) Poisson’s ratio: 420310A2 5547B 237184A2 = c44 = 5547B c11 = c12 (3.149) (3.150) c11 ' 1.7 · c44 (3.151) c12 = 0.36 c11 + c12 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) ν[11̄0][110] = − = −0.09 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 = 0.6. ν[001][110] = 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) ν[010][100] = 3.3.3.2 3.3.3.2.1 (3.152) (3.153) (3.154) Body centred cubic cells First nearest neighbours √ For the body-centred cubic cell the first nearest neighbour bond distance is 3 2 a. Each atom has eight nearest neighbour bonds, and 6 second nearest neighbour bonds. There are two atoms in the primitive unit cell. The energy for the first nearest neighbour interaction is: Enn1 = 8 −A B + 12 6 a a From the first derivative of the energy: 96 (3.155) r r0 = 6 2B A (3.156) The strains for bulk, tetragonal and rhombohedral shear can then be applied, as with the face centred unit cell. Bulk modulus strains: r 6 2B r1 = r0 (1 + γ) = γ + 1 A 2 −A B A2 A E1 = 8 6 + 12 = 8 6 12 − r1 r1 4 γ+1 B 2 γ+1 B E1 = − 2 A2 72 A2 γ2 504 A2 γ3 + − + ... B B B 72A2 γ2 3 (c11 + 2c12 ) = 2 B 48A2 γ2 (c11 − c12 ) = B (3.157) (3.158) (3.159) (3.160) (3.161) Tetragonal shear modulus: r1p = a1nn p E1 = 8 q 2 γ2 + 3 = 2 γ2 + 3 2 A B E1 = 8 6 + 12 r1p r1p 729 A2 r B √ 6 2 3 A 27 A2 6 − 3 4 2 γ2 + 3 B 2 2 γ2 + 3 B E1 = − (3.162) (3.163) 2 A2 8 A2 γ4 + + ... B B (3.164) (3.165) There is no γ2 term in the energy for the tetragonal shear c11 − c12 = 0. Rhombohedral shear modulus: 97 (3.166) q 2 + 4 γ + 3 6 2B 2 γ A a1nn r1p = √ 2 3 q p 6 2 q 2 γ − 4 γ + 3 2B A a1nn r1m = 3 − 4γ + 2γ2 = √ 2 3 −A −B −A −B E1 = 4 6 + 12 + 4 6 + 12 r1p r1m r1m r1p 2 2 27 A 729 A E1 = 4 6 − 3 2 2 4 2γ + 4γ + 3 B 2 2γ + 4γ + 3 B 2 2 27 A 729 A − +4 6 3 4 2 γ2 − 4 γ + 3 B 2 2 γ2 − 4 γ + 3 B p q 3 + 4γ + 2γ2 = E1 = − 2 A2 32 A2 γ2 + + ... B B A2 c44 = 16 B (3.167) (3.168) (3.169) (3.170) (3.171) (3.172) Poisson’s ratio: From equations 3.161, 3.166 & 3.172, the elastic constants can be found, and Poisson’s ratios derived. c11 = c12 = c44 1 c12 = c11 + c12 2 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) = −1 ν[11̄0][110] = − 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 = 2. ν[001][110] = 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) ν[010][100] = 3.3.3.2.2 (3.173) (3.174) (3.175) (3.176) Second nearest neighbours As with the first nearest neighbours, to find the lattice parameter for the the crystal including the second nearest neighbour interactions it is necessary to find both the first and second nearest neighbour separation: 98 −A = 6 2 6 + √ r 0 −A B B Enn2 + 8 + 2 12 √ r60 r12 r0 0 ( 3 ( 3 ! ! 6 A 12 B 81 A 2187 B dE +8 =6 − − 13 da 32 r0 7 1024 r0 13 r0 7 r0 r 6 (55713 · 16 B) r0 = (1011 · 512 A) q √ 3 6 379B 2 7 A 2 a2nn = √ r0 = √ √ 6 3 3 10784 (3.177) (3.178) (3.179) (3.180) Bulk modulus: The strain for bulk modulus can then be applied using the same methodology as for the face-centred cubic unit cell, with the revised separation from equation 3.179: √ 3 q 7(γ + 1) 379B A r1 = r0 (1 + γ) = √ 6 10784 q √ 3 6 27(γ + 1) 379B A r2 = a2nn (1 + γ) = √ √ 6 3 10784 −A −A B B E2 = 8 6 + 12 + 6 6 + 12 r1 r2 r1 r2 2 2 82791801 A 9099 A E2 = 6 6 12 − 1379528164 γ + 1 B 37142 γ + 1 B 2 2 116294656 A 10784 A +8 6 12 − 344882041 γ + 1 B 18571 γ + 1 B E2 = − 6 113569 A2 2044242 A2 γ2 2044242 A2 γ3 + − + ... 37142 B 18571 B 2653 B 3 2044242A2 (c11 + 2 c12 ) = 2 18571B 99 (3.181) (3.182) (3.183) (3.184) (3.185) (3.186) Tetragonal shear modulus: q p √ 3 2+3 6 B 7 2γ A a2nn r1p = √ √ 6 2 3 10784 q √ 3 6 7(γ + 1) AB rx = a2nn (1 + γ) = √ √ 6 3 10784 q √ 3 6 7(1 − γ) AB r y = a2nn (1 − γ) = √ √ 6 3 10784 q √ 3 6 7 AB rz = a2nn = √ √ 6 3 10784 ! ! −A −A B −A −A B B B E2 = 8 6 + 12 + 2 6 + 12 + 2 6 12 + 2 6 + 12 r1p r1p rx ry ry rz rx rz q 3 + 2 + γ2 = E2 = − 113569 A2 165201444 A2 γ2 − + ... 37142 B 49268863 B −165201444A2 c11 − c12 = 49268863B (3.187) (3.188) (3.189) (3.190) (3.191) (3.192) (3.193) Rhombohedral shear modulus: q p √ 3 6 2 7 2γ + 4γ + 3 379B A a2nn r1p = 3 + 4γ + 2γ2 = √ √ 6 2 3 10784 q p √ 3 2 − 4γ + 3 6 379B q 7 2γ A a2nn r1m = 3 − 4γ + 2γ2 = √ √ 6 2 3 10784 q p √ 3 6 2 q 7 2γ + 1 379B A r2xy = a2nn 1 + γ2 = √ √ 6 3 10784 q 100 (3.194) (3.195) (3.196) q √ 3 6 7 379B A rz = a2nn = √ √ 6 3 10784 −A −A −A −A B B B B E2 = 8 6 + 12 + 4 6 + 12 + 4 6 + 12 + 2 6 + 12 r1p r1p r1m r1m r1xy r1xy r1z r1z E2 = − (3.197) (3.198) 2780889963 A2 517114368 A2 γ 26206931712 A2 γ2 14544769024 A2 γ3 − + − + ... 689764082 B 344882041 B 344882041 B 147806589 B (3.199) 2c44 = 26206931712 A2 γ2 344882041 B a32nn (3.200) Poisson’s ratio: c12 ≈ 0.53 c11 + c12 2c11 c44 − (c11 − c12 )(c11 + 2c12 ) ν[11̄0][110] = − ≈ −1.55 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) 4c12 c44 ν[001][110] = ≈ 2.93 2c11 c44 + (c11 − c12 )(c11 + 2c12 ) ν[010][100] = 3.3.3.3 3.3.3.3.1 (3.201) (3.202) (3.203) Discussion First nearest neighbours The Lennard-Jones potential includes terms for both the attractive and repulsive atomic interactions. Considering only the first nearest neighbour bonds, the properties calculated show that for face-centred cubic cells, negative Poisson’s ratio is not predicted in the (11̄0, 110) direction, but the predicted value is zero. The predicted elastic constants for the body-centred cubic cell are all equal. This mimics the result of the simple spring potential, and it can be seen that when considering only the first nearest neighbour interactions the predicted Poisson’s ratio in the off-axis direction is −1 2 . These results again show that the first nearest neighbour bonding in the face-centred cubic cell does not give rise to auxetic behaviour. For the body-centred cubic cell, negative Poisson’s ratio is predicted when considering the only the first nearest neighbour bonding. 101 3.3.3.3.2 Second nearest neighbours Considering the second nearest neighbour interactions, the calculated properties differ from those found when only the first nearest neighbours are considered. The predicted values for the face centred cubic unit cell show c11 > c44 and c44 = c12 . The Cauchy relation is respected, and the crystal shows greater axial stiffness than shear stiffness. The predicted Poisson’s ratio for the (100,010) directions is positive, however, the Poisson’s ratio for the (11̄0, 110) direction is negative (-0.09). This supports the theory that the negative Poisson’s ratio behaviour in the face-centred cubic metals is contingent not on the first nearest neighbour bonding, but on the second and third nearest neighbour bonds. The calculated values for the body-centred cubic unit cell show c11 < c12 . This leads to a Poisson’s ratio of -1.55 for the (11̄0, 110) direction, with an associated very high (2.9) corresponding orthogonal positive strain. This result is perhaps counter intuitive; it is expected that negative Poisson’s ratio behaviour in the body-centred cubic cell can be attributed to the first nearest neighbour bonding, however this result shows how adding additional nearest neighbours accentuates this. 3.3.3.3.3 Influence of potential range It is implied from the analysis of the Lennard-Jones potential that the range of the potential has a greater effect on the calculated properties than any other parameter. To explore this numerical analysis was conducted computationally (further explanation of this technique is presented in chapter 4). It can be seen from section 3.3.3.2.2 that using just two nearest neighbours results in a lengthy calculation, adding further nearest neighbours becomes increasingly so, even with the aid of a symbolic mathematics program. An arbitrary face-centred cubic unit cell is used as a starting point for this analysis. The cutoff of the potential is varied so that the interactions up to the ninth nearest neighbour are modelled. The calculated (11̄0, 110) Poisson’s ratio, calculated from the numerical methods, is shown in figure 3.7. The change in interatomic separation is shown in figure 3.8. It can be seen that adding more neighbours contracts the cell, as has been shown with the analytical modelling. The variation in Poisson’s ratio mimics the change when neighbours are added, and the Poisson’s ratio converges to less than 1% variation with 5 102 nearest neighbours, and less than 10% variation with two nearest neighbours. Repeating this analysis (figures 3.9 and 3.9) shows the BCC crystal system mirrors this. Figure 3.7: Variation of the (11̄0, 110) Poisson’s ratio with change in cutoff, for an arbitrary FCC system. Figure 3.8: Variation of the lattice constant (normalised relative to the lattice constant for the first nearest neighbour value) with change in cutoff, for an arbitrary FCC system. 103 Figure 3.9: Variation of the (11̄0, 110) Poisson’s ratio with change in cutoff, for an arbitrary BCC system. Figure 3.10: Variation of the lattice constant (normalised relative to the lattice constant for the first nearest neighbour value) with change in cutoff, for an arbitrary BCC system. 104 3.4 3.4.1 Conclusions Summary of calculated properties Potential c11 c44 c12 Volume a3 kv 0 a3 kv Spring - FCC 2 kar kr a kr a Spring - BCC 2 kr 3a 2 kr 3a 2 kr 3a Lennard-Jones - FCCnn1 2 2A √ B AB 2 A √ Lennard-Jones - BCCnn1 16 AB Lennard-Jones - FCCnn2 2 75 AB Lennard-Jones - BCCnn2 22.2 AB B 2 16 2 2 A √ B A A2 B 16 B B A A2 B 2 42.7 AB 42.7 AB 2 25.7 AB 25.7 AB 2 2 Table 3.1: Elastic constants calculated for each of the potential models ν(010,100) ν(11̄0),(110) ν(001,110) Volume 1 2 -1 0 Spring - FCC 1 3 0 1 2 Spring - BCC 1 2 -1 2 Lennard-Jones - FCCnn1 1 3 0 1 2 Lennard-Jones - BCCnn1 1 2 -1 2 Lennard-Jones - FCCnn2 0.36 -0.09 0.62 Lennard-Jones - BCCnn2 0.53 -1.55 2.93 Potential Table 3.2: Poisson’s ratio calculated for each of the potential models using the elastic constants shown in table 3.1 It has been demonstrated that simple two-body potentials exhibit negative Poisson’s ratio for strains in the (11̄0, 110) direction. The predicted values are sensitive to number of nearest neighbours included in the analysis. The Bravais lattice of the crystal has an effect on both the interatomic separation and the number of nearest bonds in the cell, and this in turn changes the way in which the potential predicts the elastic properties. 105 Chapter 4 Atomistic modelling of the elastic properties of cubic crystals 4.1 Introduction This chapter describes the numerical investigation of the elastic properties of crystalline materials. The primary aim of this study is to evaluate whether classical atomistic modelling techniques are able to predict negative Poisson’s ratio behaviour in cubic elemental metals. Both classical potentials and quantum mechanical modelling are considered in this thesis. In this chapter the numerical results of the classical studies are used to investigate the underlying cause of auxetic behaviour, and to provide a template for the finite element modelling investigation of scale independent beam networks (see chapter 6). 4.1.1 Elemental cubic metals Elemental cubic metals are the focus of this chapter as the elastic constant data for these materials is readily available and the existence of negative Poisson’s ratio behaviour is well documented, allowing the techniques used to be validated. What is not explicitly known is why some crystals exhibit this behaviour whilst those that are extremely similar in structure do not. The elastic constant data from Landolt-Bornstein (1985), a compilation of experimentally derived elastic constants, is used to calculate the experimentally derived Poisson’s 106 ratios for elemental cubic metals. The experimental data is published for only the principal crystallographic orientations, however as discussed in section 1.1.2.1, negative Poisson’s ratio is usually observed in off-axis directions. It is possible to use the on-axis data to compute the Poisson’s ratio for all directions as shown in section 2.2.6, however to do this manually is very laborious, as all angles must be considered, and even with spreadsheets, this is very time consuming. For convenience, the program ElAM (Marmier et al., 2010) is used to perform these calculations . ElAM provides an automated methodology to calculate elastic properties in all directions (for a specified resolution) and report the maximum and minimum values, and there associated directions. The calculated maximum and minimum Poisson’s ratios are shown in table 4.1. 4.1.2 Covalently bonded zincblende compounds In addition to the elemental cubic metals that are the main focus of this work, the properties of covalently bonded compounds are investigated. The focus of the analysis is to determine whether there is a link between the axial and rotational bonding within the crystal and the resulting Poisson’s ratio. Covalent bonding is far more directional in character than the metallic bonding of the elemental metals considered. Whilst it would have been logical to use covalently bonded BCC and FCC structures, the analysis depends on having experimentally measured elastic properties with which to validate the modelling and thus the zincblende structures of these compounds were used in the analysis as this data. 107 Element Symmetry Group νmin νmax Cs bcc 1 -0.46 1.22 Na bcc 1 -0.44 1.20 K bcc 1 -0.43 1.21 Rb bcc 1 -0.40 1.15 Ba bcc 2 -0.29 0.82 Ca fcc 2 -0.26 0.89 Sr fcc 2 -0.26 0.85 Ta bcc 5 0.17 0.50 Nb bcc 5 0.21 0.61 V bcc 5 0.29 0.45 Cr bcc 6 0.13 0.32 Mo bcc 6 0.21 0.40 W bcc 6 0.28 0.28 Fe bcc 8 -0.06 0.62 Rh fcc 9 0.07 0.44 Ir fcc 9 0.10 0.38 Ni fcc 10 -0.07 0.67 Pd fcc 10 -0.05 0.81 Pt fcc 10 0.21 0.57 Cu fcc 11 -0.13 0.82 Ag fcc 11 -0.09 0.83 Au fcc 11 -0.04 0.88 Al fcc 13 0.27 0.42 Pb fcc 14 -0.19 1.01 Table 4.1: Experimentally determined Poisson’s ratio maxima and minima for elemental cubic metals (listed according to the 1990 IUPAC group system (Fluck, 1988)). Data taken from (Landolt-Bornstein, 1985). 108 4.2 4.2.1 Classical Potentials: Methodology Introduction Classical potentials are used to find the energy of the crystals under consideration and from this calculate the elastic properties of a range of elemental cubic metals. Simple potentials have been used in chapter 3 to investigate how the Bravais lattice and the number of nearest neighbours in the model affect the predicted properties of the crystal. By using more sophisticated potentials, the properties of the elemental cubic metals can be calculated, and the understanding gained in the previous chapter can be used to analyse how well the potentials predict negative Poisson’s ratio behaviour. The total internal energy of a crystal structure is the sum of the energies resulting from interactions of atoms within the crystal (nearest neighbour, second nearest neighbour etc.). The accuracy of the description of the energy is to some extent conditional on the number of interactions considered but it is also true that, as the order of the interaction increases, its contribution to the total energy diminishes; second order interactions have less effect on calculated properties than first. The elastic properties of the crystals are calculated from the energy, and thus has a fundamental effect on the accuracy of the predicted properties. This section first details how the elastic properties are calculated using potential forms and parameters found in the literature. The parameters of these potentials are then varied to determine how these affect Poisson’s ratio. A fundamental potential model is then applied to selected covalent compounds to investigate how Poisson’s ratio varies with comparative bond energies. 4.2.2 Numerical modelling methodology The numerical modelling is conducted using well established potentials (outlined in section 2.3), the two-body Morse potential, and the many-body Finnis-Sinclair, SuttonChen and Cleri-Rosato potentials. The General Utility Lattice Program (Gulp) (Gale, 2007) is used to perform the analysis of the crystal structures. Fitting parameters for the Morse potential are taken from Girifalco and Weizer (1959), and for the many-body potentials 109 from the library within Gulp (these are from respective original publications; (Finnis and Sinclair, 1984; Sutton and Chen, 1990; Cleri and Rosato, 1993)). Structural information is obtained from the literature (Ashcroft and Mermin, 1976; Landolt-Bornstein, 1985) to determine the lattice constants, elastic constants and Poisson’s ratios in both the [100] and [110] directions for a range of cubic metals. These input parameters are used to construct the Gulp input files. The potential modelling methodology can be described in three stages: an optimisation; followed by a relaxation; followed by calculation of the bulk properties. The structural optimisation is conducted by finding the local energy minimum, closest to the starting configuration. The energy can then be expanded as a Taylor series: U(x + δx) = U(x) + 1 ∂2 U ∂U (δx)2 δx + 2! ∂x2 ∂x (4.1) The first derivative of this Taylor series is the gradient vector, and the second derivative is the Hessian matrix. Gulp uses a second order truncated form of this, and a NewtonRaphson iterative scheme to find the local minimum (Gale, 2007). Once the local energy minimum has been found and the resulting structure calculated, it is possible to derive the physical properties of the structure from the gradient of the energy surface. The elastic constants are the second derivative of the energy density (with respect to strain): 1 ∂2 U ci j = V ∂εi ∂ε j ! (4.2) Having determined the elastic constant matrix it is then possible to further derive the elastic properties using the procedures outlined in section 2.2.6. 4.2.2.1 Calculation of Poisson’s ratio for elemental cubic metals To investigate the Poisson’s ratio (measured in the [110], [11̄0] direction) for the cubic crystals, two methodologies are employed; calculation of the elastic constants in the standard orientation and computation of the off-axis properties from these data, and direct measurement from a rotated structure. 110 4.2.2.1.1 Optimisation method For each of the potential functions, the experimentally calculated structure was used as a starting point. The structure is optimised, and allowed to fully relax. The relaxed structure and elastic stiffnesses are calculated, and Poisson’s ratio in the [110, 11̄0] directions (the direction for which a minimum Poisson’s ratio has been observed from experimental data) calculated from the stiffness tensor. The accuracy of the potential was then appraised relative to the published experimental data. 4.2.2.1.2 Direct measurement method The direct measurement method is employed to verify that the off-axis strains calculated by the optimisation method are replicated when measured directly. The unit cell is rotated through 45◦ about the [001] axis. The atomic positions in the 1’ (revised [100]) direction are constrained at a value of 1.005 times the rotated lattice constant, to generate a 0.5% strain in the original [110] direction, and the position of the (unconstrained) orthogonal atoms calculated after optimisation (see figure 4.1). The resulting strains are calculated from the lattice constants resulting from the relaxation and the Poisson’s ratio verified against that produced from the optimised elastic constants. The main advantage of the direct measurement method is that the deformed structure can be visualised in order to determine the mechanism of the deformations. Figure 4.1: Relative orientation of a body-centred cubic unit cell for optimisation (left) and direct measurement methods (right) (n.b. axes, bottom left) 111 4.2.2.2 Parameter space exploration Each of the potentials used is controlled by fitting parameters that govern the shape of the potential function, for example the width and depth of the potential well and the interatomic separation at equilibrium. Given that the elastic constants are calculated from the second derivative of the energy, changing these parameters and thus the energy function affects the calculated constants and resulting Poisson’s ratio. The parameters in the model are fitted to the experimental data, usually the lattice parameters and c11 , c12 . For many-body potentials, which are not bound by the Cauchy relation, c44 is also included. Each parameter models a different physical property in the crystal (electron density etc.) which has a different effect on the calculated energy and Poisson’s ratio. To understand which potential parameters regulate the negative Poisson’s ratio behaviour crystals and align this with physical properties of the crystals, a parameter space exploration is conducted. Fitting parameters for the potentials are varied to investigate the controlling factors of the calculation of the elastic constants and subsequent Poisson’s ratio. The parameters varied for each potential are shown in table 4.2. Values from the literature were used as a starting point and high throughput techniques help to conduct the analyses. Potential Varied parameters Morse Separation at minimum energy (r0 ) Well width (α) Finnis-Sinclair EAM functional (A) EAM Density (φ) Sutton-Chen Lennard-Jones (repulsive) component (ρ) EAM Density (φ) Table 4.2: Varied potential parameters. (n.b. Potentials detailed in section 2.3) The geometry of an elemental cubic metal, and associated fitting parameters are used as the basis for the analysis. High throughput computing techniques are then employed 112 to vary each of the parameters of the potential. The respective element used in each analysis is shown in table 4.3. Potential Element Morse Al Finnis-Sinclair Fe Sutton-Chen Ni Table 4.3: Elements used as basis for parameter space exploration. 4.2.2.3 Modelling of a coupled three-body/harmonic system Figure 4.2: Schematic representation of the Keating model A simple model of the bonding in crystals is the so called "Keating model" where the bonding is modelled by the combination of a harmonic and a three-body potential. The original model, published by Keating (1966a) and subsequently extended (Keating, 1966b), models the interaction of diamond like crystals to find the elastic properties. This fundamental model allows the relationship between the ratio of axial to rotational forces and Poisson’s ratio to be analysed. The potentials used for this analysis are the harmonic and simple three-body potentials; 1 Eharm = Kharmonic (r − r0 )2 2 1 Ethree = Kthree (θ − θ0 )2 , 2 (4.3) (4.4) where En is the respective energy, Kn the associated spring constant, (r − r0 ) the axial deformation, and (θ − θ0 ) the angular deformation. 113 The contribution to the overall energy from each potential can be used to determine the influence each has on the overall properties and the ratio of lateral to axial stiffness can be found. 4.2.2.3.1 Fitting of potentials to covalently bonded cubic crystals The bonding within the elemental cubic metals is, implicitly, metallic bonding. The nature of metallic bonding is such that it is non-directional but can be approximated as such. In contrast, covalent bonding is highly directional and thus is more suited to being described by directional potentials. The covalently bonded crystals considered all have a zincblende structure (see figure 4.3). The fitting parameters for covalent cubic crystals are not readily available and thus it is necessary to use the fitting capabilities of Gulp to conduct this analysis. As the focus of this thesis is the elastic properties of BCC and FCC crystals, analysis of BCC and FCC covalent compounds would have been more directly relevant, however the analysis depends on accurate experimentally measured elastic data and thus the zincblende structures were modelled as these data were available in the literature. The elastic constants of the crystals are obtained from the experimental data (Vukcevich, 1972). The structure is constrained to that observed experimentally, the elastic constants used as observables, and the spring constants of the potential allowed to relax. Having found the fitting parameters, we can then perform an optimisation to find the energy of the system. Figure 4.3: Structure of covalently bonded zincblende type crystal structure - metal grey, non-metal yellow. Unconnected atoms are periodic images. 114 4.3 Elastic properties calculated from classical potentials This section presents the results of the potential modelling. It is divided into three sections. Initially, the elastic properties calculated from the classical potentials is shown. These models are then used as a basis for the exploration of how the parameters of the model vary the calculated Poisson’s ratio. Finally, the results of the fundamental three-body modelling is presented. 4.3.1 Experimental data N.b. For convenience, the experimental data from the literature is provided below in table 4.4 to enable comparison with the predicted elastic properties. 115 Element Symmetry Group Period Lattice const. c11 c44 c12 ν110 ν12 K G E Na b 1 3 4.2 7.6 4.3 6.3 -0.4 0.5 6.8 4.3 1.8 K b 1 4 5.2 3.7 1.9 3.1 -0.4 0.5 3.3 1.9 0.8 Rb b 1 5 5.6 3.0 1.6 2.4 -0.4 0.5 2.6 1.6 0.8 Cs b 1 6 6.0 2.5 1.5 2.1 -0.5 0.5 2.2 1.5 0.6 Ca f 2 4 5.6 22.8 14.0 16.0 -0.3 0.4 18.3 14.0 9.6 Sr f 2 5 6.1 15.3 9.9 10.3 -0.3 0.4 12.0 9.9 7.0 Ba b 2 6 5.0 12.6 9.5 8.0 -0.3 0.4 9.5 9.5 6.4 V b 5 4 3.0 230.0 43.1 120.0 0.5 0.3 156.7 43.1 147.7 Nb b 5 5 3.3 245.0 28.4 132.0 0.6 0.3 169.7 28.4 152.6 Ta b 5 6 3.3 264.0 82.6 158.0 0.2 0.4 193.3 82.6 145.7 Cr b 6 4 2.9 348.0 100.0 67.0 0.3 0.2 160.7 100.0 326.4 Mo b 6 5 3.1 465.0 109.0 163.0 0.4 0.3 263.7 109.0 380.4 W b 6 6 3.2 523.0 160.0 203.0 0.3 0.3 309.7 160.0 409.5 Fe b 8 4 2.9 230.0 117.0 135.0 -0.1 0.4 166.7 117.0 130.1 Rh f 9 5 3.8 413.0 184.0 194.0 0.1 0.3 267.0 184.0 289.0 Ir f 9 6 3.8 580.0 256.0 242.0 0.1 0.3 354.7 256.0 437.5 Ni f 10 4 3.5 247.0 122.0 153.0 -0.1 0.4 184.3 122.0 130.0 Pd f 10 5 3.9 221.0 70.8 171.0 -0.1 0.4 187.7 70.8 71.8 Pt f 10 6 3.9 347.0 76.5 251.0 0.2 0.4 283.0 76.5 136.3 Cu f 11 4 3.6 169.0 75.3 122.0 -0.1 0.4 137.7 75.3 66.7 Ag f 11 5 4.1 122.0 45.5 92.0 -0.1 0.4 102.0 45.5 42.9 Au f 11 6 4.1 191.0 42.2 162.0 0.0 0.5 171.7 42.2 42.3 Al f 13 3 4.0 108.0 28.3 62.0 0.3 0.4 77.3 28.3 62.8 Pb f 14 6 5.0 48.8 14.8 41.4 -0.2 0.5 43.9 14.8 10.8 Table 4.4: Experimentally observed elastic properties of elemental metals (Ashcroft and Mermin, 1976; Landolt-Bornstein, 1985). Elastic stiffnesses in GPa. 116 4.3.2 4.3.2.1 Elastic and engineering constant data from potential models Accuracy of calculated properties The numerical modelling is conducted on both body-centred and face-centred cubic metals. The elastic constants for each metal are computed, and the engineering constants calculated from these. The accuracy of the properties calculated using the classical potentials could then be compared to experimental data. This section discusses the accuracy of the values calculated from the numerical modelling, and considers the shortcomings of these techniques. 117 c11 Expt c44 Expt c12 Expt c11 Potential c44 Potential c12 Potential Error c11 % Error c44 % Error c12 % Morse Potential Element 4.3.2.2 Cs 2.5 1.5 2.1 2.4 1.7 1.7 -4.6 13.7 -18.3 Rb 3.0 1.6 2.4 3.2 2.3 2.3 7.2 44.9 -5.0 K 3.7 1.9 3.1 3.7 2.8 2.8 -1.2 49.2 -10.9 Na 7.6 4.3 6.3 7.3 5.8 5.8 -4.3 35.1 -8.2 Ba 12.6 9.5 8.0 10.6 9.2 9.2 -15.7 -3.5 14.6 Sr 15.3 9.9 10.3 14.9 10.8 10.8 -2.4 9.3 5.1 Ca 22.8 14.0 16.0 20.8 15.1 15.1 -8.7 7.9 -5.6 Cr 348.0 100.0 67.0 202.1 197.1 197.1 -41.9 97.1 194.1 Mo 465.0 109.0 163.0 290.5 289.3 289.3 -37.5 165.4 77.5 W 523.0 160.0 203.0 340.7 330.0 330.0 -34.9 106.2 62.5 Fe 230.0 117.0 135.0 180.6 167.0 167.0 -21.5 42.8 23.7 Ni 247.0 122.0 153.0 233.0 163.4 163.4 -5.7 34.0 6.8 Ag 122.0 45.5 92.0 133.2 90.1 90.1 9.2 98.1 -2.0 Cu 169.0 75.3 122.0 173.6 122.6 122.6 2.8 62.8 0.5 Al 108.0 28.3 62.0 94.4 67.5 67.5 -12.6 138.7 8.9 Pb 221.0 70.8 171.0 55.8 37.2 37.2 -74.8 -47.5 -78.3 Table 4.5: Elastic constants for elemental metals, calculated using Morse potential. Elastic stiffnesses in GPa. 118 ν(11̄0,110) ν(11̄0,110) ν(010,100) K G E Error ν(11̄0,110) % Error ν(010,100) % Error K % Error G % Error E % -0.28 -0.46 0.45 2.20 1.48 0.60 24.69 8.35 13.16 -13.74 -59.72 Na -0.35 -0.40 0.45 6.75 4.30 1.83 -16.30 2.27 6.76 -35.10 -14.48 K -0.65 -0.43 0.46 3.34 1.88 0.82 -3.42 5.57 7.32 -49.25 -50.79 Rb -0.37 -0.44 0.45 2.61 1.60 0.75 5.68 6.58 0.39 -44.88 -61.08 Ba -0.93 -0.29 0.39 9.53 9.50 6.39 -121.98 -19.32 -1.24 3.50 66.78 Ca -0.34 -0.26 0.41 18.27 14.00 9.60 -39.41 -1.96 6.87 -7.90 15.47 Sr -0.33 -0.26 0.40 11.97 9.90 7.01 -42.79 -4.45 -1.89 -9.33 16.80 W -0.79 0.32 0.28 309.67 160.00 409.48 425.18 -75.97 -7.71 -106.24 96.11 Cr -0.44 0.40 0.16 160.67 100.00 326.37 388.90 -205.81 -23.70 -97.07 97.71 Mo -0.99 0.28 0.26 263.67 109.00 380.39 344.51 -92.26 -9.88 -165.44 99.55 Fe -0.51 -0.06 0.37 166.67 117.00 130.14 -1170.70 -29.91 -2.92 -42.75 84.57 Ni -0.32 -0.07 0.38 184.33 122.00 129.96 -338.41 -7.80 -1.24 -33.98 24.46 Cu -0.26 -0.09 0.42 137.67 75.30 66.70 -146.34 1.28 -1.42 -62.84 -8.14 Ag -0.38 -0.13 0.43 102.00 45.50 42.90 -195.71 6.12 -2.43 -98.07 -40.89 Al -0.36 0.27 0.36 77.33 28.3 62.78 228.33 -14.37 1.09 -138.67 39.41 Pb -0.91 -0.05 0.44 187.67 70.8 71.81 -394.87 8.37 76.89 47.52 63.65 Expt Element Cs Table 4.6: Engineering constants for elemental metals, calculated using Morse potential. Experimental off-axis PR shown for comparison. K,G & E in GPa. 4.3.2.2.1 Analysis: Morse Potential The Morse potential is used to investigate 16 metals. The calculated elastic constants are shown in table 4.5, the derived elastic properties in table 4.6. The most apparent feature of these data is that the Cauchy relation, c44 = c12 , is obeyed for all of the metals considered. The computed values of c11 show good agreement for group one and two elements, with less than 10% error. The computed values for c44 and c12 are generally an average of the true values of c44 and c12 . For the group one and two elements, these are still less than 119 50% error. For the transition metals and poor metals, the magnitude of error for c11 is greater than 50%. Engineering constants are often used as a measure of accuracy when analysing potential data. The computed values for ν(010,100) show an error of less than 20% for all except four of the transition metals. Similarly, the predicted values for bulk modulus show good agreement with the experimentally calculated values. It is only when considering the values that have off-axis components (ν(11̄0,110) and shear modulus, G) that the erroneous prediction of the elastic constants is evident. The values for Young’s modulus are calculated from the compliances, which are in turn calculated from the inverted stiffness matrix. Because of this, the Young’s modulus value, whilst only being a function of axial compliances, is not accurately predicted due to the errors in the stiffness values. Experimental elastic constants for the body-centred cubic metals show a negative Poisson’s ratio for the group one and two metals, and a positive Poisson’s ratio for the transition metals, with the exception of body-centred cubic Iron. Comparison of the computed Poisson’s ratio of the group one elements to experimental results shows that for these metals the Morse potential is able to accurately predict the Poisson’s ratio, both in magnitude and sign. The result for Barium, group two, predicts auxetic behaviour however, the calculated value is double that of the experimental value. The face-centred cubic transition metals have a negative Poisson’s ratio when calculated from experimental data. The Morse potential correctly predicts this auxetic behaviour however generally over predicts the value of ν by a factor of around 3. The shear modulus calculated from the Morse potential shows good agreement for the group two metals but generally has greater than 50% error for the remaining metals. 120 c11 Expt c44 Expt c12 Expt c11 Potential c44 Potential c12 Potential Error c11 % Error c44 % Error c12 % Finnis-Sinclair potential Element 4.3.2.3 V 230.0 43.1 120.0 227.9 42.6 118.7 -0.9 -1.2 -1.1 Nb 245.0 28.4 132.0 246.6 28.1 133.2 0.7 -1.1 0.9 Ta 264.0 82.6 158.0 266.0 82.4 161.2 0.8 -0.2 2.0 Cr 348.0 100.0 67.0 387.1 100.8 103.5 11.2 0.8 54.5 Mo 465.0 109.0 163.0 464.7 108.9 161.5 -0.1 -0.1 -0.9 W 523.0 160.0 203.0 522.4 160.6 204.4 -0.1 0.4 0.7 Fe 230.0 117.0 135.0 243.1 121.1 138.1 5.7 3.5 2.3 ν(11̄0,110) ν(11̄0,110) ν(010,100) K G E Error ν(11̄0,110) % Error ν(010,100) % Error K % Error G % Error E % V 0.45 0.45 0.34 155.10 42.60 146.60 -0.31 0.11 1.00 1.16 0.76 Nb 0.62 0.61 0.35 171.01 28.10 153.18 -0.79 -0.17 -0.79 1.05 -0.41 Ta 0.17 0.17 0.38 196.15 82.41 144.36 0.61 -0.78 -1.46 0.23 0.91 Cr 0.37 0.32 0.21 198.05 100.81 343.44 -14.25 -30.68 -23.27 -0.81 -5.23 Mo 0.40 0.40 0.26 262.57 108.90 381.40 -0.14 0.64 0.42 0.09 -0.27 W 0.28 0.28 0.28 310.43 160.62 407.46 1.08 -0.58 -0.25 -0.38 0.49 Fe -0.04 -0.06 0.36 173.11 121.09 143.04 38.55 2.05 -3.87 -3.50 -9.91 Expt Element Table 4.7: Elastic constants for elemental metals, calculated using Finnis-Sinclair potential. Elastic stiffnesses in GPa. Table 4.8: Engineering constants for elemental metals, calculated using Finnis-Sinclair potential. Experimental off-axis PR shown for comparison. K,G & E in GPa. 121 4.3.2.3.1 Analysis: Finnis-Sinclair The Finnis-Sinclair potential is used to predict the behaviour of seven body-centred cubic transition metals, shown in tables 4.7 and 4.8. The calculated elastic constants are in good agreement for all of the metals considered. The engineering constants calculated from these values are thus in good agreement with the experimental results. For V, Nb, Ta, Cr, W and Mo the Poisson’s ratio is found to be positive experimentally and this is replicated in the simulations. For Fe, the experimental Poisson’s ratio is negative, and which is also found with the computed value. Not only does this model correctly predict whether the Poisson’s ratio is positive or negative but also the value of the result is essentially correct. The main conclusions that can be drawn are that the Finnis-Sinclair is not only able to accurately predict whether Poisson’s ratio will be positive or negative, but also can accurately determine the magnitude of ν. The Morse potential, being a two-body potential, is not able to predict when an element has a positive Poisson’s ratio, due to the Cauchy relation as discussed in section 2.3. 122 c11 Expt c44 Expt c12 Expt c11 Potential c44 Potential c12 Potential Error c11 % Error c44 % Error c12 % Sutton-Chen potential Element 4.3.2.4 Rh 413 184 194 339.0 142.7 231.9 -17.9 -22.4 19.5 Ir 580 256 242 475.2 210.0 309.5 -18.1 -17.9 27.9 Pd 221 70.8 171 248.5 93.4 175.8 12.5 32.0 2.8 Ni 247 122 153 230.8 79.6 177.1 -6.5 -34.7 15.8 Pt 347 76.5 251 314.5 73.7 258.4 -9.4 -3.7 3.0 Ag 122 45.5 92 139.9 58.8 95.6 14.6 29.3 4.0 Cu 169 75.3 122 168.7 58.2 129.4 -0.2 -22.8 6.1 Au 191 42.2 162 179.9 42.2 147.8 -5.8 -0.1 -8.7 Al 108 28.3 62 81.2 15.5 71.6 -24.8 -45.3 15.4 Pb 48.8 14.8 41.4 49.4 15.5 38.0 1.3 4.7 -8.2 Table 4.9: Elastic constants for elemental metals, calculated using Sutton-Chen potential. Elastic stiffnesses in GPa. 123 ν(11̄0,110) ν(11̄0,110) ν(010,100) K G E Error ν(11̄0,110) % Error ν(010,100) % Error K % Error G % Error E % -0.06 0.07 0.41 267.60 142.69 150.63 182.20 -27.09 -0.23 22.45 47.88 Ir -0.05 0.10 0.39 364.75 210.04 231.06 150.34 -33.97 -2.84 17.95 47.19 Pd -0.03 -0.05 0.41 200.05 93.43 102.87 41.38 5.03 -6.60 -31.97 -43.24 Ni -0.08 -0.07 0.43 195.03 79.62 77.01 -6.08 -13.51 -5.81 34.74 40.74 Pt 0.00 0.21 0.45 277.12 73.69 81.41 98.54 -7.46 2.08 3.67 40.27 Ag -0.06 -0.09 0.41 110.38 58.85 62.18 38.30 5.53 -8.22 -29.34 -44.94 Cu -0.08 -0.13 0.43 142.52 58.17 56.32 42.12 -3.55 -3.53 22.75 15.58 Au 0.00 -0.04 0.45 158.54 42.16 46.57 108.09 1.72 7.65 0.10 -10.08 Al -0.08 0.27 0.47 74.78 15.49 14.12 128.28 -28.47 3.30 45.26 77.51 Pb -0.03 -0.19 0.43 41.81 15.50 16.36 82.39 5.27 4.68 -4.71 -51.51 Expt Element Rh Table 4.10: Engineering constants for elemental metals, calculated using Sutton-Chen potential. Experimental off-axis PR shown for comparison. K, G & E in GPa 4.3.2.4.1 Analysis: Sutton-Chen Potential The data from the Sutton-Chen potential is shown in tables 4.9 and 4.10. Where Poisson’s ratio is experimentally predicted to be negative, the potential predicts a negative Poisson’s ratio to a reasonable degree of accuracy. Where Poisson’s ratio is predicted to be positive, the potential predicts a Poisson’s ratio of zero. Predictions of bulk modulus are generally within 10% accuracy, however the potential over-estimates the shear and Young’s moduli in the majority of cases. 124 c11 Expt c44 Expt c12 Expt c11 Potential c44 Potential c12 Potential Error c11 % Error c44 % Error c12 % Cleri-Rosato potential Element 4.3.2.5 Rh 413.0 184.0 194.0 399.0 201.9 242.3 -3.4 9.8 24.9 Ir 580.0 256.0 242.0 554.5 261.4 345.8 -4.4 2.1 42.9 Pd 221.0 70.8 171.0 231.9 72.6 178.5 5.0 2.5 4.4 Ni 247.0 122.0 153.0 298.2 157.2 184.1 20.7 28.8 20.3 Pt 347.0 76.5 251.0 341.1 90.6 273.4 -1.7 18.4 8.9 Ag 122.0 45.5 92.0 131.7 50.6 96.7 7.9 11.2 5.0 Cu 169.0 75.3 122.0 176.7 82.2 125.2 4.6 9.2 2.6 Au 191.0 42.2 162.0 187.4 44.7 154.4 -1.9 6.0 -4.7 Al 108.0 28.3 62.0 95.0 37.0 74.5 -12.1 30.8 20.2 Pb 221.0 70.8 171.0 48.4 12.8 39.8 -78.1 -82.0 -76.8 Table 4.11: Elastic properties for elemental metals, calculated using Cleri-Rosato potential. Elastic stiffnesses in GPa. 125 ν(11̄0,110) ν(11̄0,110) ν(010,100) K G E Error ν(11̄0,110) % Error ν(010,100) % Error K % Error G % Error E % -0.08 0.07 0.38 294.53 201.95 215.86 -3.40 -18.22 -10.31 -9.76 25.31 Ir -0.05 0.10 0.38 415.36 261.39 288.93 -4.39 -30.45 -17.11 -2.11 33.96 Pd -0.03 -0.05 0.43 196.32 72.61 76.68 4.95 0.30 -4.61 -2.55 -6.79 Ni -0.10 -0.07 0.38 222.14 157.17 157.68 20.74 0.21 -20.51 -28.83 -21.34 Pt -0.01 0.21 0.44 295.99 90.61 97.83 -1.69 -6.00 -4.59 -18.44 28.22 Ag -0.08 -0.09 0.42 108.32 50.62 49.84 7.92 1.53 -6.20 -11.25 -16.19 Cu -0.14 -0.13 0.41 142.40 82.23 72.84 4.57 1.07 -3.44 -9.20 -9.19 Au -0.01 -0.04 0.45 165.40 44.72 47.89 -1.89 1.56 3.65 -5.96 -13.19 Al -0.17 0.27 0.44 81.34 37.00 29.45 -12.06 -20.55 -5.19 -30.76 53.09 Pb -0.05 -0.05 0.45 42.64 12.75 12.61 -78.08 -3.32 77.28 81.99 82.44 Expt Element Rh Table 4.12: Engineering constants for elemental metals, calculated using Cleri-Rosato potential. Experimental off-axis PR shown for comparison. K, G & E in GPa. 4.3.2.5.1 Analysis: Cleri-Rosato Potential The results of modelling using the Cleri-Rosato potential are shown in tables 4.11 and 4.12. A range of transition metals were analysed. The potential is not able to accurately recreate the elastic constants for Lead (group 14). For the remaining metals (groups 9 13) the potential underestimates the value of c11 and under estimates the values of c12 and c44 . The most accurate results are for the group 10 and 11 metals. Comparing the experimental and predicted properties for the off-axis Poisson’s ratio, the predicted value is always negative, even when the experimental value is positive. The computed values for the bulk modulus are generally within 10% error, and the values for shear modulus and Young’s modulus generally within a 30% error. 126 4.3.2.6 Summary In general, predictions for group one and group two metals are more accurate than those for transition metals, with the exception of those calculated using the Finnis-Sinclair potential. 4.3.3 Deformed structures calculated from classical potentials The classical potential method allows the deformed structure of the crystal to be easily calculated from the direct measurement method: the deformed lattice parameters are automatically generated by Gulp. There are four possible deformed structures that can be observed: • Body-centred cubic with positive Poisson’s ratio • Body-centred cubic with negative Poisson’s ratio • Face-centred cubic with positive Poisson’s ratio • Face-centred cubic with negative Poisson’s ratio The crystals are subjected to a 0.05% strain in the [110] direction, and the magnitude of the atomic displacements is recorded. These displacements are replicated and translated so that the deformation of the unit cell can be observed. 127 4.3.3.1 Body-centred cubic crystals (a) Positive ν (b) Negative ν Figure 4.4: Schematic of the deformation mechanism of body-centred cubic crystals. Left: Negative Poisson’s ratio, with corresponding perpendicular positive Poisson’s ratio deformation. right: positive Poisson’s ratio in both planes. The mechanism for auxetic behaviour in body-centred cubic crystals is suggested by Baughman et al. (1998). Baughman suggested a scissor type, hinging arrangement where the perpendicular out of plane behaviour is a high positive Poisson’s ratio behaviour and the in-plane behaviour is negative. This gives a theoretical minimum in-plane Poisson’s ratio of -1, and the bond lengths remain unchanged. By computing the deformed structure of the body-centred cubic crystals that are shown to have a negative Poisson’s ratio, it can be seen that the mechanism suggested by Baughman is replicated in the Gulp models (see figure 4.4). The positive Poisson’s ratio mechanism is perhaps less intuitive when considering the atomistic picture, it corresponds to a contraction of the planes perpendicular to the strained direction. The varying interatomic distances make this a more complex deformation that the auxetic behaviour. 4.3.3.2 Face-centred cubic crystals The mechanism for auxetic behaviour in face-centred cubic crystals is far less obvious than that for body-centred cubic. The unit cell of the face-centred cubic cell consists of four atoms, however, in the rotated configuration the unit cell assumes tetragonal symmetry (the x and y axes are rotated, z is not). This results in an eight atom unit cell. 128 Computing the displacements of this cell alone does not provide a clear understanding of the deformation mechanism, and thus it is necessary to further manipulate the resulting configuration, taking care to preserve the symmetry of the crystal. Initially, a macro-cell is generated, and from this it is possible to observe the auxetic behaviour. As with the body-centred cubic unit cell there is a large positive orthogonal strain associated with the negative in-plane Poisson’s deformation. Visualising the deformed unit as the cubic cell most readily associated with a facecentred cubic crystal, the strains can be more readily observed. A possible mechanism is derived from the Cleri-Rosato potential modelling of Iridium: The deformed unit cell appears to be subject to two deformation mechanisms. The first is analogous to the deformation mechanism of the body-centred unit cell described above, the second is a combination of stretching and hinging within the cell. Figure 4.5: A ’side’ view (left) and a ’top’ view (right) of the deformed unit cell. Deformed atomic positions shown as dashed circles, undeformed as solid. Not to scale. Figure 4.5 shows a view down the (11̄0) direction, a side view, and a perpendicular view down the (001) axis, a top view. Considering the cubic cell as being constructed of 3 layers, the upper and lower layers deform in a manner very similar to the body-centred cubic unit cell, the ’in-plane’ bonds undergoing a stretching, the ’out of plane’ bonds being shortened, as shown as dashed lines in figure 4.6. It can be seen that the bonds connecting the central layer to the upper and lower layers undergo hinging, as shown in figure 4.7. The bonds within the central layer undergo stretching, as indicated by dashed lines in figure 4.8. The magnitude of the displacements of the ’mid-layer’ atoms is less than that of the top layer atoms. 129 Figure 4.6: The ’external’ deformation. Deformed atomic positions shown as dashed circles, undeformed as solid. Not to scale. Figure 4.7: The hinging of the inter-layer bonds. Initial bonds shown as solid lines, deformed bonds shown as dashed. Deformed atomic positions shown as dashed circles, undeformed as solid. Only half of the unit is shown for clarity. Figure 4.8: The ’internal’ deformation. Deformed atomic positions shown as dashed circles, undeformed as solid. Not to scale. This mechanism suggests that the auxetic behaviour of the face-centred cubic cells is conditional on a combination of both rotational and axial forces, the stretching of the upper and lower layers and the rotation of the inter-layer bonding. This mechanism gives a lower magnitude of Poisson’s ratio than the body-centred cubic deformation, in agreement with the result shown by the simple spring models in section 3.3. 130 4.3.4 Variation of predicted properties with phase space exploration Figures 4.9 to 4.14 show the change in calculated Poisson’s ratio as the respective potential parameter is varied. The calculated ’virtual’ value is shown by a blue rhombus. The value calculated from the fitted value is shown as a pink square. An explanation of the potential forms, and their component terms is given in section 2.3.3.2. 4.3.4.1 Morse potential Figure 4.9: Poisson’s ratio, calculated from the Morse potential, showing the change in ν11̄0,110 with variation of the well width (α) component of the Morse potential. Pink square shows the fitted values for Aluminium, for reference. 131 Figure 4.10: Poisson’s ratio, calculated from the Morse potential, showing the change in ν110 with variation of the interatomic separation (r0 ) component component of the Morse potential. Pink square shows the fitted values for Aluminium, for reference. 132 4.3.4.2 Finnis-Sinclair potential Figure 4.11: Poisson’s ratio, calculated from the Finnis-Sinclair potential, showing the change in ν110 with variation of the density component fitting parameter. Pink square shows the fitted values for Iron, for reference. Figure 4.12: Poisson’s ratio, calculated from the Finnis-Sinclair potential, showing the change in ν110 with variation of the EAM functional component fitting parameter. Pink square shows the fitted values for Iron, for reference. 133 4.3.4.3 Sutton-Chen Potential Figure 4.13: Poisson’s ratio, calculated from the Sutton-Chen potential, showing the change in ν110 with variation of the density component fitting parameter. Pink square shows the fitted values for Copper, for reference. Figure 4.14: Poisson’s ratio, calculated from the Sutton-Chen potential, showing the change in ν110 with variation of the pairwise repulsive (Lennard-Jones) component. Pink square shows the fitted values for Copper, for reference. 134 4.3.4.4 Analysis: Variation of predicted Poisson’s ratio with phase-space exploration The results of the analysis are shown in figures 4.9 to 4.14. The analysis is conducted to investigate the sensitivity of the calculated Poisson’s ratio to changes in the potential parameters. The components of each model were varied and the resulting ν(11̄0,110) Poisson’s ratio computed. Where there appears to be little, or significant variation the sensitivity of the analysis was adjusted accordingly to capture this effect. The Morse potential uses three terms, to describe the interatomic separation and both the depth, and width of the potential well. Variation of the depth of the potential well has no impact on the calculated Poisson’s ratio once a minimum value has been found. Variation with the width of the potential well, α, is shown in figure 4.9. The Poisson’s ratio is very sensitive to variation in α, until a maximum value of around zero is found, but as the Morse is a two-body potential, the Poisson’s ratio never becomes positive. Varying the atomic separation at minimum energy results in a variation of the Poisson’s ratio from -0.6 to almost zero. The Finnis-Sinclair potential has two main parameters that affect the calculated properties. As discussed in section 2, the Finnis-Sinclair is a form of embedded-atom model (EAM). Within the potential are terms for the density and the EAM functional. Varying the density component (shown in figure 4.11) has the effect of varying the Poisson’s ratio from -0.3 to -0.2. Varying the EAM functional term gives the possibility of varying the Poisson’s ratio from -1 to around 2. The Sutton-Chen and Cleri-Rosato component are very closely related in form, and thus only the Sutton-Chen potential is considered. The Sutton-Chen potential is controlled by the an embedded atom model and by a Lennard-Jones repulsive component. The Sutton-Chen potential is far less sensitive to variation in potential parameters than the Finnis-Sinclair and Morse potentials, however it can be seen that the computed results deviate from the trend for some cases. Investigation of these shows that this is an artefact of the modelling; as the density or repulsive component changes, the separation at minimum energy also changes. The cutoff in the simulation is fixed, and thus this small change in cutoff determines the number of nearest neighbours in the analysis which in turn leads to a change in the overall energy, and thus the elastic constants derived from 135 this. 4.3.4.5 Conclusions The sensitivity of the potentials to input parameters depends greatly on the form used. The Poisson’s ratio of the Finnis-Sinclair and Morse potential is very sensitive to small changes in input parameters, but the Sutton-Chen potential less so. From this analysis, it seems impossible to generalise. Fortunately it is computationally inexpensive to conduct such sensitivity analyses on a case-by-case basis. 4.3.5 Elastic constants calculated from coupled three-body/harmonic system This model is described in section 4.2.2.3. The values presented here are those derived from the fitted data. Figure 4.15: Poisson’s ratio (ν110,11̄0 ) and relative spring stiffnesses calculated from the three-body/harmonic model for selected body-centred cubic elemental metals. Violet Calculated Poisson’s ratio. Burgundy - Poisson’s ratio as reported in the literature. Cream - ratio of three-body and harmonic spring constants. 136 Crystal c11 c44 c12 Error c11 % Error c44 % Error c12 % Φharmonic Φthree kharm kthree Normalised kthree Figure 4.16: Poisson’s ratio (ν110,11̄0 ) and relative spring stiffnesses calculated from the three-body/harmonic model for selected covalently bonded cubic metal compounds. Green - Calculated Poisson’s ratio, red - Poisson’s ratio as reported in the literature, blue - ratio of three-body and harmonic spring constants. Ba 12.44 9.87 7.80 -1.30 3.85 -2.54 3.50E-07 2.34E-04 0.44 0.11 0.58 Cr 320.64 161.30 33.09 -7.86 61.30 -50.62 6.20E-07 2.75E-03 3.48 1.34 3.87 Cs 2.24 1.99 1.78 -9.21 34.31 -13.56 9.10E-07 4.63E-04 0.11 0.02 0.14 Fe 212.17 157.16 112.88 -7.75 34.32 -16.38 8.20E-07 9.35E-04 3.92 0.46 1.31 K 3.28 2.89 2.58 -11.58 53.96 -18.08 4.77E-05 3.81E-05 0.14 0.02 0.10 Li 12.62 11.35 10.33 -5.83 18.24 -8.59 2.70E-07 3.88E-05 0.36 0.02 0.07 Mo 409.61 234.55 93.34 -11.91 115.19 -42.74 1.49E-03 3.91E-03 5.84 1.91 6.03 Na 6.84 6.03 5.37 -9.87 40.14 -15.21 1.00E-05 4.37E-05 0.23 0.02 0.09 Nb 199.40 130.66 75.34 -18.61 360.07 -42.93 1.00E-07 1.78E-03 3.61 0.87 2.88 Rb 2.64 2.31 2.05 -10.72 44.49 -16.16 0.00E+00 4.17E-05 0.12 0.02 0.11 Ta 227.75 164.08 112.83 -13.73 98.64 -28.59 5.54E-06 1.66E-03 4.68 0.81 2.70 V 193.15 126.89 73.43 -16.02 194.41 -38.81 3.73E-04 1.30E-03 3.20 0.64 1.93 W 468.78 283.63 134.29 -10.37 77.27 -33.85 1.28E-03 4.18E-03 7.25 2.04 6.46 Table 4.13: Elastic properties for body-centred cubic metals, calculated using Three-body/harmonic model. Elastic stiffnesses in GPa 137 Element c11 c44 c12 ν(11̄0,110) Error c11 % Error c44 % Error c12 % Φharmonic Φthree kharm kthree kharm k3 kthree a Φthree Φharm kthree norm kharm 138 ZnS 99.1 63.0 44.3 -0.10 15.4 24.6 9.1 0.015 18.96 6.29 2.06 3.05 11.68 2.56 1.86 ZnSe 87.2 51.2 39.9 -0.06 -11.1 -18.3 -10.8 0.028 11.73 5.85 1.90 3.08 10.29 0.85 1.76 ZnTe 72.1 41.1 30.8 -0.04 1.2 1.0 -1.4 0.021 8.86 5.05 2.04 2.48 12.45 0.85 2.47 InP 103.4 58.2 45.4 -0.03 1.2 1.1 -1.3 0.026 10.68 7.05 2.54 2.78 14.88 0.83 2.11 InAs 84.2 45.8 39.1 -0.03 1.1 1.1 -1.1 0.026 13.10 7.16 3.49 2.05 24.85 0.99 3.47 InSb 66.7 36.5 30.2 -0.02 0.0 0.0 0.0 0.026 7.58 5.10 2.19 2.33 14.19 0.59 2.78 AlSb 90.3 45.0 41.1 0.02 1.0 1.5 -1.1 0.018 3.77 6.53 2.28 2.86 13.93 0.43 2.13 GaSb 89.4 41.1 42.7 0.05 1.1 2.0 -1.1 0.018 1.81 6.61 2.10 3.15 12.81 0.20 1.94 GaAs 120.2 55.0 58.1 0.05 1.8 3.3 -1.8 0.046 1.89 8.25 2.15 3.84 12.16 0.08 1.47 GaP 142.4 63.5 69.9 0.06 0.9 1.7 -0.9 0.011 1.61 9.55 2.34 4.08 12.74 0.28 1.33 Table 4.14: Elastic constant data from Three-body/harmonic model for covalently bonded cubic crystals 4.3.5.1 Influence of the relationship of three-body and axial forces on predicted properties The original three-body/harmonic or Keating model was devised to model the interatomic interactions of Diamond (Keating, 1966a). In the previous section this technique was adapted to investigate the bonding in both elemental body-centred cubic metals and some covalently bonded compounds which have a zinc-blende structure. By analysing whether the Poisson’s ratio (in the (110) direction) predicted by the potential is negative, and how the two components of the potential relate, a link between the bond order and auxetic behaviour may be seen. Body-centred cubic metals, were the first to be examined followed by covalent crystals with a zinc-blende structure. Results are shown in table 4.13. The Keating model is not able to perfectly reproduce the elastic properties of any of the elemental metals, however, it is able to reproduce c11 to within a 10% error for most of the crystals and c12 to within a 30% error. c44 is calculated to within a 50% error for 5 of the crystals, but for the remaining crystals, a good fit could not be found. Results where the experimental and calculated Poisson’s ratio are in close agreement are shown in figure 4.15. The analysis of these data encompasses comparison of the energies and spring constants for the three-body and harmonic components of the potential form. No clear trends can be observed, but comparing the ratio of the three-body and harmonic spring constants for the metals where a satisfactory fit can be found, there is some correlation between the spring stiffness ratio and the Poisson’s ratio. Comparing the energy values shows a several orders of magnitude of variation, as the fitted values are not able to accurately reflect the bonding in the crystal. The covalent cubic crystals are modelled far more accurately, with less than 30% error for all the elastic constants, which given the simplicity of the model seems a reasonable level of accuracy. Both the energy and the stiffness value, for each component of the potential, can be compared. The ratio of the harmonic and three-body contributions to the energy are shown in figure 4.16, and the numerical values for both the spring constants and energy values can be seen in table 4.14. There is a clear trend that shows that as the harmonic/three-body energy ratio increases Poisson’s ratio becomes more positive, again 139 showing how the Poisson’s ratio is sensitive to on the ratio of axial to lateral stiffness in the bonding. In the metallic crystals, the bonding is non-directional, and thus the lateral stiffness of the bonds is provided by the higher order bonding, however, for the covalent bonding, the bonds are highly directional, and thus the lateral contribution from the first neighbour bonding dominates. 4.4 Can classical potentials predict auxetic behaviour? Classical potentials are used to investigate the elastic properties of elemental cubic metals. Two-body potentials, whilst able to accurately predict the structure of the crystals, are unable to predict when positive Poisson’s ratio behaviour occurs, as they are governed by the Cauchy relation. Many-body potentials are more successful, the Finnis-Sinclair potential in particular is able to predict elastic properties for a range of transition metals. Of course, this can be attributed to how well the model is fitted to experimental data. Investigation of the deformation mechanism based on the potential modelling shows that the scissor type mechanism is responsible for the negative Poisson’s ratio deformation in body-centred cubic metals. For the face-centred cubic metals, the mechanism is more complex, but appears to be a combined stretching and rotation mechanism, similar to that of the body-centred cubic crystals. Analysis of the relationship of the three-body and harmonic spring stiffnesses on Poisson’s ratio shows that there is no obvious link between Poisson’s ratio and the ratio of the stiffnesses for metals, but by considering the data as a whole, particularly the covalently bonded crystals, a weak trend emerges. In other words, the degree of auxeticity of cubic metals can be accounted for by some specific potential models but only if the experimental elastic constant data was intrinsic to the derivation of the model. This is clearly not the case for such potential as the Cleri-Rosato. The conclusion should be heeded by those wishing to use potential energy methods for modelling nano-scale structures. 140 Chapter 5 Density Functional Theory modelling of elastic properties of cubic elemental metals 5.1 Introduction The classical potentials described and used in chapter 4 show that in many cases it is possible to accurately model the Poisson’s ratio, both negative and positive, for elemental cubic metals and alloys. This study also shows that predicting the elastic properties of these materials using this technique, whilst being computationally inexpensive, is conditional on the validity of the potential chosen, and the accuracy of the parameters. This dependence on empirical data makes this analysis impossible when using these techniques to model systems where experimental data is not available. In order to provide an additional source of numerical data from which the Poisson’s ratios of elemental metals can be calculated, a first principal quantum based technique is required: Density Functional Theory (DFT). DFT techniques are not dependent on fitting to experimental data to derive model parameters and provide more generic solutions. They allow calculation of Poisson’s ratio for a range of crystals without dependence on the accuracy of the fitting parameters. 141 5.2 Methodology The purpose of this investigation is to appraise whether negative Poisson’s ratio can be accurately predicted by DFT techniques, using a set of crystals for which the behaviour is well characterised experimentally. An outline summary of the principles behind DFT is given in Appendix A. The elastic properties for the elemental cubic metals are calculated using Vienna Abinitio Software Project (VASP) (Kresse and Hafner, 1993, 1994a; Kresse and Furthmüller, 1996a,b). The input files are generated from established lattice constant data (Ashcroft and Mermin, 1976). The elastic constants are calculated during the analysis using the built-in functionality of the software package, negating the need for any manual post processing. The pseudopotentials chosen are ultrasoft Vanderbilt LDA and GGA pseudopotentials. Strictly speaking, LDA and GGA are approximations of the exchange correlation functional (Exc ), but because the pseudopotentials are generated from all-electron solutions, it is normal to refer to them as pseudopotentials. Additionally, LDA-PAW and PBE projector augmented wave methodologies were also used (Kresse and Hafner, 1994b; Blöchl, 1994; Perdew et al., 1996, 1997; Kresse and Joubert, 1999). Where a range of methodologies were available, all were used for comparison as differing pseudopotentials formulations are better suited to specific crystal types. These are noted in the results as follows: • h - a harder potential than standard • s - a softer potential than standard • pv, sv - the p or s semi-core states are treated as valence states. • new - a revised version of the original potential 5.2.1 Convergence example A convergence for the k-point mesh and cut off was conducted, and suitable values for mesh density and energy cut-off decided upon based on these simulations. To aid this process, scripting was used to generate the input files, conduct the analysis and extract 142 the elastic constants. Periodic (Bloch) boundary conditions are used in the analysis (see section A.2.7). For each crystal/pseudopotential combination, an optimisation was conducted to find the calculated elastic constants. Once this had been determined, a final simulation could be conducted to find the elastic properties of the crystals. Figure 5.1: Example convergence of elastic constants with change in energy cutoff, using GGA pseudpotential . c11 - blue, c12 - green, c44 - red. Figure 5.1 shows an example convergence test for the GGA pseudopotential methodology. These tests were conducted to determine the optimum energy cutoff and k-point mesh density for the analysis. In this case, it is sufficient to use a plane wave cutoff of 350 eV. 5.3 Results The results of the analyses are shown in the following tables. Each pseudopotential yields a set of elastic constants. These elastic constants are used to calculate the engineering constants for each metal, which are presented in a separate table. First, the results of the LDA pseudopotential modelling is presented. This is followed by the results of the GGA, PAW-LDA and PAW-PBE models. The accuracy of the elastic 143 properties calculated from the DFT modelling is appraised relative to that found in the literature. For convenience, this is shown in table 5.1. Element Group Period c11 c44 c12 ν110 Na 1 3 7.59 4.3 6.33 -0.44 K 1 4 3.71 1.88 3.15 -0.43 Rb 1 5 2.96 1.6 2.44 -0.4 Cs 1 6 2.47 1.48 2.06 -0.46 Ca 2 4 22.8 14 16 -0.26 Sr 2 5 15.3 9.9 10.3 -0.26 Ba 2 6 12.6 9.5 8 -0.29 V 5 4 230 43.1 120 0.45 Nb 5 5 245 28.4 132 0.61 Ta 5 6 264 82.6 158 0.17 Cr 6 4 348 100 67 0.32 Mo 6 5 465 109 163 0.4 W 6 6 523 160 203 0.28 Fe 8 4 230 117 135 -0.06 Rh 9 5 413 184 194 0.07 Ir 9 6 580 256 242 0.1 Ni 10 4 247 122 153 -0.07 Pd 10 5 221 70.8 171 -0.05 Pt 10 6 347 76.5 251 0.21 Cu 11 4 169 75.3 122 -0.13 Ag 11 5 122 45.5 92 -0.09 Au 11 6 191 42.2 162 -0.04 Al 13 3 108 28.3 62 0.27 Pb 14 6 48.8 14.8 41.4 -0.19 Table 5.1: Experimentally observed elastic constant data for elemental cubic metals. Data taken from Landolt-Bornstein (1985). Poisson’s ratio in the [110] direction calculated from these data. 144 Symmetry Group Lattice const. Lattice error %. c11 c44 c12 Error c11 % Error c44 % Error c12 % LDA Element 5.3.1 Na b 1 1.71 4.26 8.26 6.22 7.19 -8.83 -44.65 -13.59 K b 1 1.84 4.02 3.98 2.65 3.33 -7.28 -40.96 -5.71 Rb b 1 1.59 3.76 3.03 1.93 2.62 -2.36 -20.62 -7.38 Cs b 1 1.96 -1.49 2.73 1.69 2.18 -10.53 -14.19 -5.83 Ca f 2 2.75 1.47 22.30 14.30 15.10 2.19 -2.14 5.63 Sr f 2 3.02 0.76 12.20 8.40 7.30 20.26 15.15 29.13 Ba b 2 2.08 4.24 14.20 11.90 8.87 -12.70 -25.26 -10.87 V b 5 1.46 3.05 312.90 58.15 146.30 -36.04 -34.92 -21.92 Nb b 5 1.65 -0.01 280.50 -86.17 145.10 -14.49 403.41 -9.92 Cr b 6 2.02 -1.04 561.80 89.10 150.00 -61.44 10.90 -123.88 Mo b 6 1.77 1.27 523.70 96.00 179.10 -12.62 11.93 -9.88 W b 6 1.40 1.27 578.00 143.00 210.00 -10.52 10.62 -3.45 Fe b 8 1.90 2.70 110.00 218.30 472.60 52.17 -86.58 -250.07 Rh f 9 1.88 0.82 482.80 208.50 214.42 -16.90 -13.31 -10.53 Ir f 9 1.94 -1.20 563.10 234.00 216.90 2.91 8.59 10.37 Ni f 10 1.76 -0.02 256.70 112.10 177.50 -3.93 8.11 -16.01 Pd f 10 1.98 -1.87 192.70 60.00 148.90 12.81 15.25 12.92 Pt f 10 2.00 -1.85 289.30 54.98 209.80 16.63 28.13 16.41 Cu f 11 1.82 -0.94 171.00 62.70 126.50 -1.18 16.73 -3.69 Ag f 11 2.08 -1.93 98.04 24.90 83.50 19.64 45.27 9.24 Au f 11 2.09 -2.54 131.00 6.24 132.50 31.41 85.21 18.21 Al f 13 2.02 0.13 113.60 45.20 53.60 -5.19 -59.72 13.55 Pb f 14 2.52 -1.64 56.50 20.40 33.50 -15.78 -37.84 19.08 Table 5.2: Elastic properties of elemental metals calculated using LDA pseudopotentials. 145 ν11̄0,110 ν11̄0,110 ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % -0.62 -0.44 0.47 7.55 6.22 1.57 29.19 -2.34 -11.8 -44.65 14.46 K -0.51 -0.43 0.46 3.55 2.65 0.95 15.7 0.79 -6.29 -40.96 -15.78 Rb -0.55 -0.4 0.46 2.76 1.93 0.6 27.66 -2.63 -5.48 -20.62 20.51 Cs -0.41 -0.46 0.44 2.36 1.69 0.79 -13.42 2.36 -7.59 -14.19 -33.15 Ca -0.26 -0.26 0.4 17.5 14.3 10.11 -2.78 2.09 4.2 -2.14 -5.24 Sr -0.22 -0.26 0.37 8.93 8.4 6.73 -16.84 6.96 25.35 15.15 3.96 Ba -0.33 -0.29 0.38 10.65 11.9 7.38 11.91 1 -11.68 -25.26 -15.55 V 0.47 0.45 0.32 201.83 58.15 219.68 5.14 7.08 -28.83 -34.92 -48.72 Nb 4.34 0.61 0.34 190.23 -86.17 181.56 85.94 2.63 -12.12 403.41 -19.01 Cr 0.56 0.32 0.21 287.27 89.1 498.58 42.65 -30.53 -78.8 10.9 -52.77 Mo 0.5 0.4 0.25 293.97 96 432.42 19.62 1.82 -11.49 11.93 -13.68 W 0.38 0.28 0.27 332.67 143 466.07 26.26 4.69 -7.43 10.62 -13.82 Fe 1.29 -0.06 0.81 351.73 218.3 -656.74 104.85 -119.32 -111.04 -86.58 604.65 Rh 0.1 0.07 0.31 303.88 208.5 350.91 26.34 3.78 -13.81 -13.31 -21.43 Ir 0.13 0.1 0.28 332.3 234 442.47 28.8 5.55 6.31 8.59 -1.13 Ni -0.09 -0.07 0.41 203.9 112.1 111.58 14.07 -6.88 -10.61 8.11 14.14 Pd -0.04 -0.05 0.44 163.5 60 62.89 -43.83 0.08 12.88 15.25 12.42 Pt 0.28 0.21 0.42 236.3 54.98 112.92 24.17 -0.15 16.5 28.13 17.15 Cu -0.06 -0.13 0.43 141.33 62.7 63.42 -110.75 -1.42 -2.66 16.73 4.92 Ag -0.12 -0.09 0.46 88.35 24.9 21.23 19.52 -6.99 13.39 45.27 50.51 Au -2.14 -0.04 0.5 132 6.24 -2.25 98.22 -9.57 23.11 85.21 105.33 Al 0.13 0.27 0.32 73.6 45.2 79.23 -114.46 12.1 4.83 -59.72 -26.22 Pb 0.1 -0.19 0.37 41.17 20.4 31.56 287.08 18.9 6.16 -37.84 -192.33 Expt Element Na Table 5.3: Engineering constants of elemental metals calculated using LDA pseudopotentials 146 Figure 5.2: Poisson’s ratio in the off-axis direction from cubic elemental metals, calculated from LDA pseudopotential. (ν12 DFT - Filled squares, ν12 Expt. - Hollow squares, ν110 DFT - Filled triangles, ν110 Expt. - Hollow triangles) The elastic properties calculated using LDA pseudopotentials are shown in tables 5.2 and 5.3, Poisson’s ratio is shown in figure 5.2. The predicted lattice constants are all within 5% of the experimentally measured value with the exception of Niobium. The predicted elastic constants vary in accuracy. General agreement with c11 is good, and similar to those calculated by the GGA pseudopotential. Group one and two elements show a good agreement with the exception of Strontium. The results for the transition metals show less agreement, particularly for Cr and Fe. The predicted values for c11 show better agreement than for c12 and c44 . The Poisson’s ratios for the elemental metals, calculated using the LDA pseudopotentials generally shows agreement, where the Poisson’s ratio is observed to be negative experimentally, this is shown in the pseudopotential calculations, similarly this is seen where Poisson’s ratio is positive. Iron and Lead are exceptions to this, where the DFT calculations are unable to predict the Poisson’s ratio. For the engineering constants, bulk modulus is calculated to within a 25% error, aside from for Iron and Chromium. Shear modulus shows a lesser agreement, even for the group one and two metals, but is still within a 50% error for most of the elements. Finally, the Young’s modulus shows the greatest agreement with only Iron, Lead, Silver and Gold not being able to be predicted to within a reasonable degree of accuracy. 147 Symmetry Group Lattice const. Lattice error %. c11 c44 c12 Error c11 % Error c44 % Error c12 % GGA Element 5.3.2 Na b 1 4.19 0.96 8.26 6.22 7.19 8.11 30.87 11.96 K b 1 5.29 -1.09 3.98 2.66 3.33 6.78 29.24 5.41 Rb b 1 5.67 -1.45 3.03 1.93 2.62 2.31 17.10 6.87 Cs b 1 6.16 -1.74 2.11 1.35 1.82 -17.06 -9.63 -13.19 Ca f 2 2.75 1.47 22.30 14.30 15.10 -2.24 2.10 -5.96 Sr f 2 3.02 0.76 9.39 5.40 5.20 -62.94 -83.33 -98.09 Ba b 2 5.01 0.12 12.20 10.50 7.73 -3.28 9.52 -3.49 V b 5 2.99 1.04 265.80 4.21 132.60 13.47 -923.75 9.50 Nb b 5 4.19 -26.95 249.20 -27.50 131.40 1.69 203.27 -0.45 Ta b 5 3.30 0.43 275.60 63.00 157.30 4.21 -31.11 -0.44 Cr b 6 2.79 3.27 482.80 83.79 130.50 27.92 -19.35 48.6 Mo b 6 3.15 -0.10 462.00 86.70 156.00 -0.65 -25.72 -4.48 W b 6 3.17 -0.38 516.80 130.10 189.90 -1.20 -22.98 -6.89 Fe b 8 2.77 3.56 66.10 166.50 35.83 -247.96 29.73 -276.78 Rh f 9 1.88 0.82 393.60 170.10 174.00 -4.93 -8.17 -11.49 Ir f 9 1.94 -1.20 563.10 234.00 216.90 -3.00 -9.40 -11.57 Ni f 10 1.76 -0.02 256.30 112.10 177.30 3.63 -8.83 13.705 Pd f 10 1.98 -1.87 192.70 60.00 148.90 -14.69 -18.00 -14.84 Pt f 10 2.00 -1.85 289.30 54.90 209.80 -19.94 -39.34 -19.63 Cu f 11 1.82 -0.94 171.04 62.84 126.50 1.19 -19.83 3.55 Ag f 11 2.08 -1.93 131.70 6.24 132.60 7.37 -629.17 30.61 Au f 11 2.09 -2.54 131.60 5.72 132.50 -45.14 -637.76 -22.26 Al f 13 2.02 0.13 113.60 45.20 53.60 4.93 37.39 -15.67 Pb f 14 2.52 -1.64 56.50 20.40 33.50 13.63 27.45 -23.58 Table 5.4: Elastic properties of elemental metals calculated using GGA pseudopotentials 148 ν11̄0,110 ν11̄0,110 ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % -0.62 -0.44 0.47 7.55 6.22 1.57 29.19 -2.34 -11.8 -44.65 14.46 K -0.51 -0.43 0.46 3.55 2.66 0.95 15.87 0.79 -6.29 -41.33 -15.78 Rb -0.55 -0.4 0.46 2.76 1.93 0.6 27.66 -2.63 -5.48 -20.62 20.51 Cs -0.55 -0.46 0.46 1.92 1.35 0.42 15.87 -1.84 12.75 8.78 28.86 Ca -0.26 -0.26 0.4 17.5 14.3 10.11 -2.78 2.09 4.2 -2.14 -5.24 Sr -0.1 -0.26 0.36 6.6 5.4 5.68 -155.05 11.42 44.87 45.45 18.94 Ba -0.35 -0.29 0.39 9.22 10.5 6.2 16.7 0.13 3.29 -10.53 2.86 V 0.94 0.45 0.33 177 4.21 177.53 52.52 2.92 -12.98 90.23 -20.19 Nb 1.59 0.61 0.35 170.67 -27.5 158.47 61.57 1.4 -0.59 196.83 -3.87 Ta 0.34 0.17 0.36 196.73 63 161.29 49.35 2.95 -1.76 23.73 -10.71 Cr 1.03 0.32 0.98 87.93 83.79 -252.76 68.77 -506.46 45.27 16.21 177.45 Mo 0.49 0.4 0.25 258 86.7 383.24 18.27 2.75 2.15 20.46 -0.75 W 0.37 0.28 0.27 298.87 130.1 414.74 24.63 3.9 3.49 18.69 -1.29 Fe -0.68 -0.06 0.35 45.92 166.5 40.91 90.85 4.96 72.45 -42.31 68.56 Rh 0.1 0.07 0.31 247.2 170.1 286.92 26.63 4.08 7.42 7.55 0.72 Ir 0.13 0.1 0.28 332.3 234 442.47 28.8 5.55 6.31 8.59 -1.13 Ni -0.09 -0.07 0.41 203.63 112.1 111.3 15.18 -6.9 -10.47 8.11 14.35 Pd -0.04 -0.05 0.44 163.5 60 62.89 -43.83 0.08 12.88 15.25 12.42 Pt 0.28 0.21 0.42 236.3 54.9 112.92 24.36 -0.15 16.5 28.24 17.15 Cu -0.06 -0.13 0.43 141.35 62.84 63.48 -108.36 -1.41 -2.67 16.55 4.84 Ag -1.56 -0.09 0.5 132.3 6.24 -1.35 93.91 -16.7 -29.71 86.29 103.15 Au -1.62 -0.04 0.5 132.2 5.72 -1.35 97.65 -9.32 22.99 86.45 103.19 Al 0.13 0.27 0.32 73.6 45.2 79.23 -114.46 12.1 4.83 -59.72 -26.22 Pb 0.1 -0.19 0.37 41.17 20.4 31.56 287.08 18.9 6.16 -37.84 -192.33 Expt Element Na Table 5.5: Engineering constants of elemental metals calculated using GGA pseudopotentials 149 Table 5.4 displays elastic properties calculated using the GGA pseudopotentials, table 5.5 shows the engineering constants derived from these data. In addition the Poisson’s ratio, in the (11̄0,110) direction, is shown in figure 5.3. The calculated lattice parameters for the cubic metals are all within 5% of the experimentally predicted value with the exception of Niobium. The failure of DFT to successfully predict the elastic properties of Niobium was encountered by Louail et al. (2004) who investigated the elastic constants of transition metals and found the calculated values for Niobium to be erroneous. No explanation of this is offered by Louail, however it is thought that the electronic configuration of transition metals being such that the 3d and 4s orbitals are both partially occupied (Hayward, 2002) results in the DFT approximation not being able to fully model the behaviour of these elements without a specific non-general approach. Both the bulk modulus and Young’s modulus are accurately predicted by the GGA methodology to within 10% error in most cases. Poisson’s ratio in the off-axis direction shows good agreement with the experimental result. The potential correctly predicts whether Poisson’s ratio is negative or positive for all of the elements except Lead. For most elements, the error is less than 30%, and generally it can be seen that the smaller group one and two body-centred cubic elements show greater agreement with experimental data than the face-centred cubic group 11-14 elements. 150 Figure 5.3: Poisson’s ratio calculated from GGA pseudopotential. (ν12 DFT - Filled squares, ν12 Expt. - Hollow squares, ν110 DFT - Filled triangles, ν110 Expt. - Hollow triangles) 151 Symmetry Group Lattice const. Lattice error %. c11 c44 c12 Error c11 % Error c44 % Error c12 % PAW LDA Element 5.3.3 Na b 1 4.06 4.13 10.01 7.13 8.89 -31.9 -65.87 -40.38 Rb b 1 5.37 3.86 3.91 2.29 3.5 -32.08 -43.41 -43.43 K b 1 5.73 3.73 0.52 4.41 2.08 85.9 -134.73 34.06 Cs b 1 19.53 5.22 0.57 2.17 2.58 76.78 -46.35 -25.3 Ba b 2 4.77 4.93 14.08 11.89 8.84 -11.74 -25.21 -10.48 Ca f 2 5.58 0 19.82 11.92 15.1 -15.04 -17.45 -5.96 Sr f 2 5.8 4.68 16.62 14.01 5.2 7.92 29.33 -98.08 V b 5 2.91 3.6 325.68 -5.1 159.71 -41.6 111.84 -33.09 Ta b 5 3.25 1.88 313.87 64.9 177.67 -18.89 21.42 -12.45 Nb b 5 3.26 1.11 277.5 10.96 154.58 -13.27 61.4 -17.11 Cr b 6 2.78 3.52 586.43 93.99 166.9 -68.51 6.01 -149.1 W b 6 2.91 7.87 1003.43 250.33 460.54 -91.86 -56.46 -126.87 Mo b 6 3.11 1.42 530.48 98.47 190.53 -14.08 9.66 -16.89 Fe b 8 2.69 6.11 97.23 207.14 443.57 57.72 -77.05 -228.57 Rh f 9 3.77 0.9 500.55 219.4 174 17.49 16.13 -11.49 Ir f 9 3.82 0.56 675.19 282.18 216.9 14.1 9.28 -11.57 Ni f 10 3.42 2.79 319.11 141.08 177.3 22.6 13.53 13.71 Pd f 10 3.85 0.91 266.38 83.4 148.9 17.04 15.11 -14.84 Pt f 10 3.91 0.36 378.78 80.52 209.8 8.39 5 -19.64 Cu f 11 3.52 2.4 221.26 84.54 126.5 23.62 10.93 3.56 Ag f 11 4.01 1.84 155.14 39.28 128.27 21.36 -15.83 28.28 Au f 11 4.06 0.45 199.42 26.63 132.5 4.22 -58.48 -22.26 Al f 13 3.99 1.58 127.58 51.8 53.6 15.35 45.36 -15.67 Pb f 14 4.91 0.8 67.16 23.68 33.5 27.33 37.51 -23.58 Table 5.6: Elastic properties of elemental metals calculated using PAW-LDA pseudopotentials 152 ν11̄0,110 ν11̄0,110 ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % -0.64 -0.44 0.47 9.26 7.13 1.65 31.66 -3.41 -37.2 -65.87 9.76 K 4.48 -0.43 0.8 1.56 4.41 -2.8 109.52 -73.97 53.28 -134.73 442.1 Rb -0.6 -0.4 0.47 3.64 2.29 0.6 33.73 -4.53 -39.14 -43.41 20.04 Cs 1.55 -0.46 0.82 1.91 2.17 -3.65 129.7 -79.93 12.97 -46.35 712.02 Ca -0.33 -0.26 0.43 16.67 11.92 6.76 21.21 -4.86 8.72 14.86 29.6 Sr -0.2 -0.26 0.24 9.01 14.01 14.14 -26.04 40.76 24.74 -41.51 -101.63 Ba -0.34 -0.29 0.39 10.59 11.89 7.26 13.5 0.69 -11.04 -25.21 -13.72 V 1.06 0.45 0.33 215.03 -5.1 220.59 58.12 4.03 -37.25 111.84 -49.33 Nb 0.84 0.61 0.36 195.56 10.96 166.89 27.71 -2.18 -15.26 61.4 -9.39 Ta 0.38 0.17 0.36 223.07 64.9 185.42 55.51 3.46 -15.38 21.42 -27.28 Cr 1.02 0.32 0.98 112.19 93.99 -325.55 68.49 -509.26 30.17 6.01 199.75 Mo 0.5 0.4 0.26 303.85 98.47 429.78 18.48 -1.81 -15.24 9.66 -12.99 W 0.35 0.28 0.31 641.5 250.33 713.67 20.24 -12.51 -107.16 -56.46 -74.29 Fe 1.27 -0.06 0.82 328.13 207.14 -630.41 104.92 -121.76 -96.88 -77.05 584.42 Rh 0.12 0.07 0.26 282.85 219.4 410.78 38.12 19.29 -5.94 -19.24 -42.14 Ir 0.14 0.1 0.24 369.66 282.18 569.71 33.26 17.41 -4.23 -10.23 -30.22 Ni 0.03 -0.07 0.36 224.57 141.08 192.45 349.36 6.62 -21.83 -15.64 -48.09 Pd 0.2 -0.05 0.36 188.06 83.4 159.6 126.8 17.81 -0.21 -17.8 -122.26 Pt 0.38 0.21 0.36 266.13 80.52 229.22 44.04 15.08 5.96 -5.26 -68.18 Cu 0.09 -0.13 0.36 158.09 84.54 129.23 247.26 13.24 -14.83 -12.28 -93.74 Ag -0.05 -0.09 0.45 137.23 39.28 39.02 -95.22 -5.28 -34.54 13.67 9.03 Au 0.49 -0.04 0.4 154.81 26.63 93.63 107.78 13.02 9.82 36.9 -121.31 Al 0.14 0.27 0.3 78.26 51.8 95.87 -100.09 18.88 -1.2 -83.03 -52.71 Pb 0.17 -0.19 0.33 44.72 23.68 44.86 212.25 27.49 -1.94 -60.02 -315.5 Expt Element Na Table 5.7: Engineering constants of elemental metals calculated using LDA-PAW pseudopotentials 153 Figure 5.4: Poisson’s ratio: Calculated using PAW-LDA, for BCC metals (ν12 DFT - Filled squares, ν12 Expt. - Hollow squares, ν110 DFT - Filled triangles, ν110 Expt. - Hollow triangles) Figure 5.5: Poisson’s ratio: Calculated using PAW-LDA, for FCC metals (ν12 DFT - Filled squares, ν12 Expt. - Hollow squares, ν110 DFT - Filled triangles, ν110 Expt. - Hollow triangles) The elastic properties calculated using the LDA-PAW methodology are shown in tables 5.6 and 5.7. Poisson’s ratio is shown in figure 5.4. With the exception of Tungsten and Iron, the lattice constant is calculated to less than a 5% error. The elastic constants for these metals are also in poor agreement with the experimental data. For group one metals, the elastic constant data shows c11 and c12 are 154 in closer agreement with the experimental data than c44 but the general agreement is poor. The results for the transition metals generally shows agreement with less than 50% error for the majority of the elements, in contrast to the GGA pseudopotentials. The Poisson’s ratio does not reflect the sign for eight of the 24 elements. Poisson’s ratio is however predicted within 30% error for the group one and group two elements indicating that the methodology can determine the relative, if not absolute values of the constants. Bulk modulus shows less than 30% error for all except Chromium and Iron, shear modulus and Young’s modulus are also in good agreement for the group one and two metals, but less so for elements in group 10 and above. 155 Symmetry Group Lattice const. Lattice error %. c11 c44 c12 Error c11 % Error c44 % Error c12 % PAW PBE Element 5.3.4 Na b 1 4.19 0.95 8.59 6.27 7.55 11.61 31.4 16.12 Napv b 1 4.19 0.95 0.57 10.41 11.67 -1223.6 58.71 45.75 Nasv b 1 4.19 0.95 9.42 6.45 7.94 19.42 33.37 20.25 Nbpv b 1 3.29 0.30 258 10.04 142.1 98.85 84.07 98.28 Nbsv b 1 3.29 0.30 258 10.04 142.1 98.85 84.07 98.28 Nbsv new b 1 3.29 0.30 258 10.04 142.1 98.85 84.07 98.28 Cs b 1 5.73 5.29 -2.1 -2.02 -2.19 217.54 173.21 194.1 Cssv b 1 5.73 5.29 2.08 2.54 3.22 -18.53 41.69 35.97 Ca f 2 5.56 0.36 22.78 14.74 15.57 -0.1 5.04 -2.78 Capv f 2 5.58 0.00 21.17 14.02 13.83 -7.72 0.16 -15.7 Casv f 2 5.58 0.00 20.83 13.84 13.64 -9.46 -1.16 -17.26 Srsv f 2 6.08 0.00 12.9 8.82 7.86 -18.63 -12.23 -30.98 Ba b 2 4.77 4.98 16.02 12.45 9.63 21.36 23.72 16.96 V b 5 2.91 3.64 337.6 8.87 171.4 31.87 -386.11 30 Vpv b 5 2.91 3.64 345.7 24.81 177.3 33.46 -73.74 32.32 Vsv b 5 2.99 0.99 359.5 52.63 172.2 36.02 18.11 30.33 Vsvh b 5 2.99 0.99 259.3 -1.37 127.3 11.31 3257.3 5.72 Vsv new b 5 2.99 0.99 279.6 26.77 143.7 17.74 -60.99 16.52 Nb b 5 3.29 0.30 258 10.04 142.1 5.02 -182.77 7.13 Ta b 5 3.3 0.30 276.7 58.59 158.7 4.59 -40.98 0.46 Tapv b 5 3.3 0.30 288 81.9 170.4 8.34 -0.85 7.28 Cr b 6 2.78 3.47 600.2 109.4 176.2 42.02 8.63 61.98 Crpv b 6 2.78 3.47 615.9 128.1 187.5 43.49 21.93 64.26 Crpv new b 6 2.78 3.47 611.3 127.2 186.2 43.07 21.39 64.02 156 Element Symmetry Group Lattice const. Lattice error %. c11 c44 c12 Error c11 % Error c44 % Error c12 % Mo b 6 3.11 1.27 538.8 107.8 196.6 13.7 -1.11 17.08 Mopv b 6 3.15 0.00 478.6 106.3 175.5 2.83 -2.5 7.13 Mopv new b 6 3.15 0.00 59.73 25 59.93 -678.45 -336.08 -172 Mosv b 6 3.15 0.00 477.4 107.2 168.5 2.59 -1.67 3.26 W b 6 3.13 0.95 603 157.7 235.6 13.27 -1.46 13.84 Wpv b 6 3.17 -0.32 542.3 149.1 211.4 3.55 -7.32 3.97 Wpv new b 6 2.91 7.91 1030 310.6 500.5 49.24 48.49 59.44 Fe b 8 2.69 6.27 123.9 224.9 454.3 -85.61 47.98 70.28 Feh b 8 2.69 6.27 1535 647 605.3 85.02 81.92 77.7 Fepv b 8 2.69 6.27 122.4 234.9 461 -87.86 50.19 70.72 Fepv new b 8 2.69 6.27 116.5 230 452 -97.41 49.14 70.13 Fesv b 8 2.69 6.27 391.4 324.9 552.4 41.24 63.99 75.56 Rh f 9 3.8 0.00 452.2 196.9 195.7 8.67 6.57 0.86 Rhnew f 9 3.8 0.00 452.2 196.9 195.7 8.67 6.57 0.86 Rhpv f 9 3.8 0.00 452.2 196.9 195.7 8.67 6.57 0.86 Rhpv new f 9 3.8 0.00 452.2 196.9 195.7 8.67 6.57 0.86 Ir f 9 3.83 0.26 663.5 285.1 270.5 12.59 10.19 10.54 Ni f 10 3.52 0.00 234 101.7 157.7 -5.57 -20.01 2.98 Ninew f 10 3.52 0.00 234 101.7 157.7 -5.57 -20.01 2.98 Nipv f 10 3.52 0.00 234 101.7 157.7 -5.57 -20.01 2.98 Pd f 10 3.89 0.00 249.5 83.64 188.3 11.42 15.35 9.17 Pdnew f 10 3.89 0.00 238.4 78.99 180 7.28 10.36 5.02 Pdpv f 10 3.89 0.00 250.3 84.17 190.1 11.69 15.89 10.03 Pdpv new f 10 3.89 0.00 237.6 72.9 176.6 6.99 2.88 3.17 Pdvnew f 10 3.89 0.00 238.4 78.99 180 7.28 10.36 5.02 157 Element Symmetry Group Lattice const. Lattice error %. c11 c44 c12 Error c11 % Error c44 % Error c12 % Pt f 10 3.92 0.00 372.23 84 265.13 6.78 8.93 5.33 Ptnew f 10 3.92 0.00 362.31 80.67 256.83 4.23 5.17 2.27 Ptpv f 10 3.92 0.00 365.4 80.37 258.15 5.03 4.81 2.77 Ptpv ZORA f 10 3.92 0.00 365.77 80.49 258.58 5.13 4.96 2.93 PtZORA f 10 3.92 0.00 361.73 74.59 256.3 4.07 -2.56 2.07 Cu f 11 3.61 0.00 167.24 57.87 123.41 -1.05 -30.12 1.14 Cu f f 11 3.61 0.00 167.24 57.87 123.41 -1.05 -30.12 1.14 Cunew f 11 3.61 0.00 167.24 57.87 123.41 -1.05 -30.12 1.14 Cupv f 11 3.61 0.00 167.24 57.87 123.41 -1.05 -30.12 1.14 Cupv f f 11 3.61 0.00 188.19 74.36 136.38 10.2 -1.26 10.54 Ag f 11 4.08 0.24 132.67 37.02 109.45 8.05 -22.9 15.94 Agnew f 11 4.08 0.24 126.36 34.29 104.62 3.45 -32.7 12.07 Agpv f 11 4.08 0.24 129.21 34.85 106.75 5.58 -30.55 13.82 Au f 11 4.07 0.25 205.08 35.19 190.5 6.86 -19.92 14.96 Aunew f 11 4.07 0.25 196.27 29.82 184.43 2.68 -41.52 12.16 Al f 13 4.04 0.25 118.56 48.89 57.39 8.91 42.11 -8.03 Pb f 14 4.95 0.00 61.96 21.47 37.16 21.24 31.07 -11.41 Table 5.8: Elastic properties of elemental metals calculated using PAW-PBE pseudopotentials 158 ν11̄0,110 ν11̄0,110 ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % -1.14 -0.46 0.51 -2.16 -2.02 0.13 59.7 -12.19 198.32 236.60 77.9 Cssv -21.87 -0.46 0.61 2.84 2.54 -1.8 97.89 -33.46 -29.26 -71.50 405.3 Na -0.63 -0.44 0.47 7.89 6.27 1.53 30.2 -2.86 -16.93 -45.77 16.7 Napv 1.09 -0.44 0.95 7.97 10.41 -21.6 140.0 -109.61 -18.08 -142.19 1282.3 Nasv -0.53 -0.44 0.46 8.43 6.45 2.1 17.2 -0.57 -24.91 -50.09 -17.8 Ba -0.28 -0.29 0.38 11.76 12.45 8.79 -4.57 3.3 -23.39 -31.09 -37.6 Ta 0.37 0.17 0.36 198.05 58.59 160.97 53.73 2.64 -2.44 29.07 -10.5 Tapv 0.22 0.17 0.37 209.61 81.90 161.33 23.09 0.72 -8.42 0.84 -10.7 V 0.90 0.45 0.34 226.81 8.87 222.14 50.45 1.78 -44.77 79.43 -50.4 Vpv 0.75 0.45 0.34 233.42 24.81 225.47 40.27 1.1 -48.99 42.44 -52.6 Vsv 0.55 0.45 0.32 234.65 52.63 247.89 19.54 5.52 -49.77 -22.12 -67.8 Vsvh 1.02 0.45 0.33 171.29 -1.37 175.51 56.36 3.98 -9.33 103.17 -18.8 Vsvn 0.67 0.45 0.34 189.02 26.77 181.97 33.94 0.96 -20.65 37.88 -23.2 Nbpv 0.85 -0.40 0.36 180.74 10.04 156.98 146.9 21.4 -6816.07 -527.72 -20692.4 Nbsv 0.85 -0.40 0.36 180.74 10.04 156.98 146.9 21.4 -6816.07 -527.72 -20692.4 Nbsvn 0.85 -0.40 0.36 180.74 10.04 156.98 146.9 21.4 -6816.07 -527.72 -20692.4 Nb 0.85 0.61 0.36 180.74 10.04 156.98 27.99 -1.46 -6.53 64.64 -2.9 Cr 0.51 0.32 0.23 317.54 109.44 520.19 36.93 -40.6 -97.64 -9.44 -59.4 Crpv 0.46 0.32 0.23 330.28 128.09 528.35 29.89 -44.6 -105.57 -28.09 -61.9 Crpvn 0.46 0.32 0.23 327.90 127.21 524.31 29.85 -44.6 -104.09 -27.21 -60.65 Mo 0.47 0.40 0.27 310.65 107.81 433.73 13.28 -2.99 -17.82 1.09 -14.0 Mopv 0.42 0.40 0.27 276.52 106.34 384.35 4.60 -3.4 -4.88 2.44 -1.0 Mopvn -1.02 0.40 0.50 59.86 25.00 -0.29 139.5 -92.95 77.30 77.07 100.0 Mosv 0.42 0.40 0.26 271.45 107.21 389.45 4.12 -0.5 -2.95 1.64 -2.4 Expt Element Cs 159 Element ν11̄0,110 ν11̄0,110 Expt ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % W 0.35 0.28 0.28 358.07 157.70 470.62 20.02 -0.48 -15.63 1.44 -14.9 Wpv 0.33 0.28 0.28 321.68 149.09 423.68 14.67 -0.31 -3.88 6.82 -3.5 Wpvn 0.25 0.28 0.33 677.10 310.60 703.02 -10.02 -16.93 -118.65 -94.13 -71.7 Fe 1.39 -0.06 0.79 344.17 224.91 -589.96 104.49 -112.43 -106.50 -92.23 553.3 Feh 0.12 -0.06 0.28 915.32 647.01 1193.04 149.94 23.55 -449.19 -453.0 -816.8 Fepv 1.39 -0.06 0.79 348.15 234.90 -606.10 104.49 -113.64 -108.89 -100.77 565.7 Fepvn 1.37 -0.06 0.80 340.15 230.02 -602.19 104.55 -114.96 -104.09 -96.60 562.7 Fesv -36.63 -0.06 0.59 498.71 324.91 -255.17 99.83 -58.24 -199.22 -177.70 296.1 Ca -0.27 -0.26 0.41 17.97 14.74 10.14 1.46 -1.58 -1.65 5.04 5.3 Capv -0.25 -0.26 0.40 16.27 14.02 10.24 -6.30 -4.35 -12.24 0.16 6.2 Casv -0.25 -0.26 0.40 16.04 13.84 10.03 -5.02 -4.19 -13.89 -1.16 4.2 Srsv -0.22 -0.26 0.38 9.54 8.82 6.94 -13.91 -6.22 -25.42 -12.23 -1.0 Rh 0.10 0.07 0.30 281.19 196.93 334.01 26.26 -5.82 5.05 6.57 13.5 Rhnew 0.10 0.07 0.30 281.19 196.93 334.01 26.26 -5.82 5.05 6.57 13.5 Rhpv 0.10 0.07 0.30 281.19 196.93 334.01 26.26 -5.82 5.05 6.57 13.5 Rhpvn 0.10 0.07 0.30 281.19 196.93 334.01 26.26 -5.82 5.05 6.57 13.5 Ir 0.11 0.10 0.29 401.52 285.06 506.85 14.54 -1.66 11.67 10.19 13.7 Ni -0.06 -0.07 0.40 183.12 101.66 106.96 -16.34 5.00 -0.66 -20.01 -21.5 Ninew -0.06 -0.07 0.40 183.12 101.66 106.96 -16.34 5.00 -0.66 -20.01 -21.5 Nipv -0.06 -0.07 0.40 183.12 101.66 106.96 -16.34 5.00 -0.66 -20.01 -21.5 Pd -0.04 -0.05 0.43 208.67 83.64 87.57 -24.46 -1.43 10.06 15.35 17.9 Pdnew -0.04 -0.05 0.43 199.48 78.99 83.42 -39.56 -1.38 5.92 10.36 13.9 Pdpv -0.05 -0.05 0.43 210.13 84.17 86.18 -1.24 -1.06 10.69 15.89 16.7 Pdpvn 0.02 -0.05 0.43 196.94 72.90 87.04 365.56 -2.32 4.71 2.88 17.459 Pdvnew -0.04 -0.05 0.43 199.48 78.99 83.42 -39.56 -1.38 5.92 10.36 13.9 160 Element ν11̄0,110 ν11̄0,110 Expt ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % Pt 0.21 0.21 0.42 300.83 84.00 151.66 1.52 -0.90 5.93 8.93 10.13 Ptnew 0.23 0.21 0.41 291.99 80.67 149.23 6.18 -1.18 3.08 5.17 8.67 Ptpv 0.23 0.21 0.41 293.90 80.37 151.64 9.68 -1.38 3.71 4.81 10.12 Ptpvzora 0.23 0.21 0.41 294.31 80.49 151.59 9.38 -1.35 3.84 4.96 10.09 Ptzora 0.26 0.21 0.41 291.44 74.59 149.16 19.28 -1.21 2.90 -2.56 8.63 Cu -0.03 -0.13 0.42 138.02 57.87 62.44 -318.17 1.26 0.26 -30.12 -6.83 Cu f -0.03 -0.13 0.42 138.02 57.87 62.44 -318.17 1.26 0.26 -30.12 -6.83 Cunew -0.03 -0.13 0.42 138.02 57.87 62.44 -318.17 1.26 0.26 -30.12 -6.83 Cupv -0.03 -0.13 0.42 138.02 57.87 62.44 -318.17 1.26 0.26 -30.12 -6.83 Cupv f -0.08 -0.13 0.42 153.65 74.36 73.59 -70.16 0.22 10.40 -1.26 9.35 Ag -0.09 -0.09 0.45 117.19 37.02 33.73 -2.82 4.90 12.96 -22.90 -27.19 Agnew -0.09 -0.09 0.45 111.87 34.29 31.58 -10.38 5.09 8.82 -32.70 -35.82 Agpv -0.08 -0.09 0.45 114.24 34.85 32.63 -20.98 4.97 10.71 -30.55 -31.48 Au -0.26 -0.04 0.48 195.36 35.19 21.59 85.12 4.70 12.13 -19.92 -95.95 Aunew -0.27 -0.04 0.48 188.38 29.82 17.56 86.01 5.27 8.87 -41.52 -140.89 Al 0.10 0.27 0.33 77.78 48.89 81.12 -162.10 -11.81 0.58 42.11 22.61 Pb 0.12 -0.19 0.37 45.43 21.47 34.09 263.49 -22.42 3.43 31.07 68.33 Table 5.9: Engineering constants of elemental metals calculated using PAW-PBE pseudopotentials. 161 Figure 5.6: Poisson’s ratio: Calculated from PAW-PBE, for BCC metals (ν12 DFT - Filled squares, ν12 Expt. - Hollow squares, ν110 DFT - Filled triangles, ν110 Expt. - Hollow triangles) Figure 5.7: Poisson’s ratio: Calculated from PAW-PBE, for FCC metals (ν12 DFT - Filled squares, ν12 Expt. - Hollow squares, ν110 DFT - Filled triangles, ν110 Expt. - Hollow triangles) The PAW-PBE parameterisation offers a range of pseudopotentials, each with a differing form to reflect either a differing model for the valence electrons of the element, or a differing assumption. The results for each available pseudopotential form are shown in tables 5.8, 5.9 and figures 5.6 and 5.7. The PAW-PBE data show that the from of the pseudopotential can have a large effect on the calculated properties of the metals. In general, the calculated lattice constant is the same for each of the pseudopotentials, only the elastic constants change. It is evident from the data that for many of the pseudopotential variations that there is a trade-off between c11 , c44 and c12 . An example of this is Vanadium, where the sv form predicts values with the lowest average errors, whereas the updated form gives a greater accuracy for c11 and c12 at the expense of the c44 prediction. Using the form of the pseudopotential which gives the most accurate constants relative to the experimental, it is possible to appraise the potential in the same fashion as the 162 alternate methodologies. The lattice constant is within 5% for each of the crystals. For the group one and two metals, agreement within 30% error for all constants is achieved, with the exception of Niobium. The variation in potential form also allows this level of accuracy to be determined for the majority of the transition metals, and group 10 and 11 elements. As with the other methodologies, the value for c44 shows a far lower degree of accuracy that that for the other two constants. The sign (whether negative or positive) for Poisson’s ratio is correctly predicted. Notable exceptions to this are Iron and Lead, where Poisson’s ratio is not successfully predicted by any of the potential forms. The magnitude of Poisson’s ratio is also generally predicted to within 30% error for the majority of the metals. The engineering constants are not successfully predicted for Niobium, Iron and Caesium. 5.3.5 Magnetic moment consideration For Chromium, Iron and Nickel, the DFT analysis is unable to correctly model the elastic properties of the material to an acceptable degree of accuracy. These are ferromagnetic materials. The analysis conducted has not considered the magnetic moment (spin) of these materials, which could explain the anomalous results. These elements do not obey the Madelung rule. The analyses for these elements is repeated, considering the magnetic moment of the atoms and the results shown in tables 5.10 and 5.11. These analyses were conducted using the built in functionality of VASP, however for completeness, an overview of the techniques employed is given in appendix A. 163 Element PSP Symmetry Lattice const. c11 c44 c12 Error c11 % Error c44 % Error c12 % Ni LDA FCC 3.52 359.5 215.8 160.3 31.3 43.5 4.6 Ni GGA FCC 3.52 380.7 231.8 174.4 35.1 47.4 12.3 Ni PAW-LDA FCC 3.52 275.8 159.6 129.6 10.4 23.6 -18.1 Ni PAW-PBE FCC 3.52 217.0 118.5 54.0 5.29 16.72 -36.5 Fe LDA BCC 2.87 215.4 114.2 59.3 -6.8 -2.5 -127.8 Fe GGA BCC 2.87 156 105.3 70.1 -47.5 -11.1 -92.7 Fe PAW-LDA BCC 2.87 218.2 108.9 68.7 -5.4 -7.4 -96.6 Fe PAW-PBE BCC 2.87 225.6 121.1 63.5 -1.9 3.4 -112.7 Cr LDA BCC 2.85 467.6 120 74.9 25.6 16.7 10.5 Cr GGA BCC 2.85 488.7 129.5 85.5 28.8 22.8 21.6 Cr PAW-LDA BCC 2.85 487.2 132 88.2 28.6 24.2 24 Cr PAW-PBE BCC 2.85 467.6 125.1 74.7 25.6 20.1 10.3 Table 5.10: Elastic properties of elemental metals calculated using four pseudpotentials 164 PSP ν11̄0,110 ν11̄0,110 ν010,100 K G E Error ν11̄0,110 % Error ν010,100 % Error K % Error G % Error E % GGA -0.08 -0.07 0.31 243.2 231.8 271.1 7.2 -21.7 24.2 47.4 52.1 Ni LDA -0.07 -0.07 0.31 226.7 215.8 260.7 -9.2 -24 18.7 43.5 50.1 Ni PAWLDA -0.06 -0.07 0.32 178.3 159.6 192.9 -25 -19.6 -3.4 23.6 32.6 Ni PAWPBE -0.03 0.07 0.30 161.5 146.5 193.61 -157.5 -27.4 -14.11 16.72 32.88 Fe GGA -0.13 -0.06 0.31 98.7 105.3 112.5 51 -19.3 -68.9 -11.1 -15.7 Fe LDA 0.62 -0.06 2.04 80.5 84.1 -743.6 110.1 81.9 -107 -39.1 117.5 Fe PAWLDA 0.06 -0.06 0.24 185.3 211.8 -54.5 -40.6 -7.4 29.8 -7.4 29.8 Fe PAWPBE 0.02 -0.06 0.22 117.5 121.1 197.8 375.9 -68.5 -41.8 3.4 34.2 Cr GGA 0.36 0.32 0.15 219.9 129.5 463.3 9.6 -8.4 26.9 22.8 29.6 Cr LDA 0.37 0.32 0.14 205.8 120 447 12.5 -17 21.9 16.7 27 Cr PAWLDA 0.35 0.32 0.15 98.7 105.3 112.5 7.2 -5.3 27.4 24.2 29.1 Cr PAWPBE 0.35 0.32 0.14 205.7 125.1 447 8 -17.2 21.9 20.1 27 Expt Element Ni Table 5.11: Engineering constants of elemental metals calculated using using four pseudpotentials For these methodologies, Poisson’s ratio is calculated for both Nickel and Chromium; negative Poisson’s ratios are correctly modelled. For Iron, this is not the case, but the off axis Poisson’s ratio value is of a very low magnitude, and this is captured by the analysis. The inclusion of the magnetic moment in the analysis yields an improvement of the calculated properties of the ferromagnetic materials for each of the potential type. The results for Chromium give elastic constants within 30% for c11 and c12 . For LDA, all three constants show improved values, for GGA, the value of c12 is much improved especially. For the PAW analysis results values for c11 and c12 are improved with no detrimental effect on the accuracy of the value of c44 . This gives far greater accuracy, with less than 30% error for both the elastic constant and derived engineering constants. 165 For Iron, the results show less agreement with the experimental properties, but still an improvement over then analysis with no magnetic moment consideration. For LDA, error in the analysis is reduced by a half when compared to the previous calculations. Similarly, the results for GGA show an error of less than 50% for c11 and c44 , and less than 100% error for c12 , from the previous 276%. The PAW results show a similar trend with c11 and c44 being predicted to less than 10% error, and c12 not being accurately modelled. The result of this is that the derived engineering properties do not show good agreement. The results for Nickel show a better agreement for the PAWs pseudopotentials, with less than 30% error for c11 and c44 , and 35% error for c12 . The LDA and GGA pseudopotentials give a result with a greater degree or error. These erroneous results mirror those of Louail et al. (2004), who found the elastic constants of Nickel could not be well predicted by the DFT calculations employed. 5.4 Conclusion Density functional theory techniques have been employed to predict the elastic properties of elemental cubic metals. Unlike most classical potential modelling, density functional theory is able to predict whether Poisson’s ratio is negative or positive. Differing pseudopotential methods are better suited to differing elements and in particular, transition metals cannot be easily modelled. The usage of projector augmented waves provides a methodology where the electronic structure can be accounted for enables a more accurate prediction of the properties. LDA and GGA pseudopotentials are able to predict the behaviour, however the PAW methodologies provide far more accurate elastic constant data when compared to the experimentally calculated properties. The metals considered are a range of group one, group two and transition metals. Generally, the small group one and two metals show a greater level of accuracy than the transition metals. Additionally, for the ferromagnetic elements it is necessary to consider the electron spin polarization. It should be noted that, as would be expected, the calculated Poisson’s ratio and engineering constants are entirely dependent on the accuracy of the elastic constant data. As with the classical potentials, C11 is generally well 166 replicated, but C12 and c44 show worse agreement. The computational cost of the DFT simulations is considerably greater than that of the simple potential models (hours rather than seconds), but the accuracy of the data combined with the lack of dependence on empirical data gives a far better modelling methodology. 167 Chapter 6 Finite Element Modelling of cubic frameworks 6.1 Introduction It has been shown in chapter 1 that crystal structures can exhibit negative Poisson’s ratio behaviour. This chapter investigates whether by keeping the crystalline structures and replacing the atomic bonding by linear-elastic beams it is possible to recreate the auxetic behaviour at a scale independent level. Such beam networks could potentially find a variety of applications in situations where truss or beam networks are currently employed, but where negative Poisson’s ratio, or tailored directional mechanical properties could be advantageous. This section outlines the methodology used in this analysis before presenting the structures considered and the corresponding results of these studies. It is worth noting that the purpose of this study is not to expressly model or exactly recreate the auxetic behaviour observed in the beam networks using the finite element method, but to simply use these structures as a starting point to inspire scale independent network structures. If successful, this approach could be applied to other complex atomic networks with unusual mechanical properties in order to create similar scale independent structures. The basis for this approach is to replace the nearest neighbour bonding in crystalline structures with linear elastic beams in order to replicate the auxetic behaviour found in 168 crystal structures. A model that would allow bending of ribs rather that hinging at nodes is used; the beams can both bend and axially stretch. This is justified because this type of structure can be manufactured using both conventional and additive manufacturing techniques, whereas a hinging design would be more costly and complicated to produce. The following section describes how the beam networks are derived from the bonding in body-centred, face- centred and simple cubic crystals. This is followed by an analytical derivation of the deformation of the structures. Finally, the finite element method is used to analyse the elastic properties of combinations of the different beam structures. 6.2 From crystalline bonds to beam networks The crystallographic structures are obtained from experimental data published in the literature. For convenience, the Crystallographic Database Service (CDS) (Fletcher et al., 1996) is used to find the structural information for the crystals under consideration. As well as providing information on the space group and lattice parameters of the crystals, it is also possible to obtain structural information files that can be used both to graphically display the structure and formulate the finite element model. Having obtained the basic structure of the crystal, this data is used to construct the network of beams in a finite element model. The atomic co-ordinates are used as nodal points and the vectors corresponding to the nearest neighbour bonds provide the geometry and location of the connecting beams. There are several crystal space groups that exhibit negative Poisson’s ratio behaviour however it is not possible to study all of these. This section will explore the possibility of creating scale independent beam networks from cubic crystal inspired frameworks. Nomenclature The elastic properties of simple cubic crystals are discussed at length in previous chapters. It has been shown both in the literature and also by classical and quantum modelling techniques that negative Poisson’s ratio behaviour has been observed in elemental cubic metal crystals. Cubic crystals have the highest symmetry and the least number of elastic 169 constants and thus form the logical starting point when trying to reproduce the negative Poisson’s ratio behaviour observed in crystal as a scale independent beam network. The cubic symmetry class contains the simple cubic, body-centred cubic and face-centred cubic Bravais lattices. (a) (b) (c) Figure 6.1: Representative unit cell for the simple cubic (a), body-centred cubic (b) and face-centred cubic (c) cells. Figure 6.1 shows that the nearest neighbour bonding in simple cubic crystal is analogous to the second nearest neighbour bonding in body-centred and face-centred cubic crystals. Auxetic behaviour is not seen in simple cubic crystals, but is in face-centred and body-centred cubic crystals. Using the mechanism for body-centred cubic as a starting point, the nearest first, second and third nearest neighbour bonds are modelled as beams (see figure 6.2). For simplicity, simple cubic analogous beams are labelled as type I, facecentred cubic analogous as type II and body-centred cubic analogous as type III. In order to tailor the Poisson’s ratio of the structure, the relative stiffness of each type of beam is varied and the mechanical response of the structure calculated. 170 Figure 6.2: Structures under consideration, that form the macro scale beam network (left to right) Type I - simple cubic analogous, Type II - face-centred cubic analogous, Type III - body-centred cubic analogous. 6.3 Analytical derivation of mechanical properties of cubic beam structures In order to examine the deformation of the structures, each of the three beam configurations considered is modelled using analytical beam equations. In all cases, the in-plane Poisson’s ratio of the structure is calculated for a strain in the [110] direction. 6.3.1 Assumptions used in analytical modelling It is assumed that the structure is made up of linear elastic beams of solid, homogeneous, isotropic circular cross section. These simple analytical models echo the modelling conducted in chapter 3, however in this case the linear springs are replaced with bending beams. In modelling the beams, assumptions must be made. Elastic deformations are assumed to be quasi-static; any inertial effects in the loading can be ignored, and only the steady state solution is considered. Analytical beam modelling can follow two main formalisms: either Euler-Bernoulli or Timoshenko. The Timoshenko approach improves on Euler-Bernoulli by including terms for the shear deformation of the beams. However as the modelling conducted in this work is limited to the static case for slender beams where the deformation arises from bending and only for for small strains, this level of sophistication is not necessary (Ashby and Gibson, 1997). Euler-Bernoulli beam theory 171 assumes that the cross-sectional area remains unchanged, and that the maximum stress is less than the yield stress. This approximation provides a convenient but sufficiently robust methodology for calculating the deformed structure and the resulting Poisson’s ratio. 6.3.2 Type I Figure 6.3: Type I structure, [110] strain The strain in the [110] direction leads to an elongation of the top and bottom squares of the frame, with an in-plane perpendicular contraction in each. For an applied strain, rather than a uniformly distributed load, the axial displacement will be driven by this deformation and thus for a displacement, δx , the Poisson’s ratio can be easily calculated. From the symmetry of the structure, it can be seen that the displacement applied to each beam is equal, and thus the total deformation can be found by considering the deformation of one beam and using the symmetry to find the total deformation. Under symmetrical loading, the beams can be considered as guided end cantilever beams, and beam theory used to compute the displacement. For a square configuration, the load is at an angle of 45◦ to the load direction (for [110] loading). The load at each nodal point can be considered by summing the deflection of each beam in turn. Considering the top (or bottom) of the type I structure independently of the vertical beams, we can compute the in-plane Poisson’s ratio for the structure. Considering the top or bottom of the structure as a square framework (figure 6.4), made up of beams A,B,C,D, connected at points 1,2,3,4, subjected to a load, P (the resultant of the force applied to the 172 vertical beams in the [110] direction). The symmetry dictates that this can be simplified to a half and a quarter of the unit, see figure 6.4. (a) Top/bottom frame (b) Symmetric con- (c) Single beam ditions Figure 6.4: Free body diagram of simplified structure (a): Top frame, (b):Symmetric condition and (c): Single beam Considering beam B, it is subjected to axial (Pax ) and transverse load (Plat ) (the components of force P, acting at an angle to the beam). Given that the of the load angle, θ for a square configuration is 45◦ , Plat = Pax . The axial force is then exerted as the transverse load on beam A, and an equivalent lateral displacement is seen. This loading is symmetrical about both sides, and thus the displacement at point 3 is equal to the sum of the y displacements caused by deflection of beams A and B, and their symmetric equivalents. The transverse load causes a bending deflection on the beam, the deflection of a slender beam is given by Plat = p 2 sin θ L3 12EI , (6.1) where P is the load, L is the length of the beam, E is the Young’s modulus, I the second moment of inertia and θ the angle of the load relative to the beam. This is a standard solution as found in the literature (Roark and Young, 1982). The Poisson’s ratio of the structure can be found from: 173 2δx εlateral = εaxial √point2 2 2lx δy point 3 √ . (6.2) 2l2x The Poisson’s ratio for the type I structure for a [110] displacement is +1. 6.3.3 Type II Figure 6.5: Type II structure, [100] strain The deflection of the type II structure in the [110] direction gives rise to purely axial loading on beams A & D (see figure 6.6). This axial stretching dominates and gives an in-plane Poisson’s ratio equivalent to that of the beams, thus almost zero. This loading causes deflection of points in the z direction owing to bending of beams E & F, and K & L, the reaction moment exerted by the junction of beams A, E & F, causing inwards (z plane) curvature of beam A. Figure 6.6: Load configuration for type II structure ([110] direction), side view 174 Figure 6.7: Load configuration for type II structure ([110] direction), top view For a load, P, applied to beam A in the x direction the resulting strain ∆Ax can be found: ∆Ax = PAx πr2 · LAx E (6.3) The stretching of beam A causes no perpendicular in-plane deformation of beam B, resulting in an in-plane deformation of zero, and thus an in-plane Poisson’s ratio of zero. 6.3.4 Type III Figure 6.8: Type III cubic structure, [110] strain. By straining the body-centred cubic analogous structure (figure 6.8), a stretch in the [110] direction (points 2 and 4), results in a contraction in one perpendicular direction (points 1 and 3) which will in turn force an expansion in the corresponding perpendicular direction 175 (points 5 and 6) (the unit cell has been translated so that points 1 (or 3) correspond to the centre of the cubic unit shown in figure 6.8). The plane on which the strain/contraction occurs can be seen to be the same as that in figure 6.4(a). Figure 6.9: Type III cubic structure, [110] strain, showing unit cell - points 1 and 3 are central points of unit cell shown in figure 6.8. The perpendicular plane undergoes the same deformation, rotated through 90 degrees, the only variation in this case is that the deformation occurs in perpendicular planes and thus the resulting load on beams A,B,C,D forces compression of beams EFGH at points 1 and 3, causing expansion at points 5 and 6. The symmetry of the structure dictates that the deflection of all the component beams must be equal, and that thus the deflection of the orthogonal points must also be equal. Using the methodology outlined in section 6.3.2, the deflection, γ at each point, and the resulting Poisson’s ratio can be found γ= P 4 sinθ 12 E I 176 l (6.4) Point δx δy δz 1 0 0 γ 2 −γ −γ 0 3 0 0 −γ 4 γ −γ 0 5 0 γ 0 6 0 −γ 0 Table 6.1: Displacements of nodal points for BCC structure loaded in [110] direction The strains are derived, and Poisson’s ratio computed: 2γ −2γ 2γ εy = εz = 2l sinθ 2l θ 2l sinθ εy εz νxy = − = 1 νxz = − = −1 εx εx εx = 6.3.5 (6.5) (6.6) Conclusion The analytical modelling has shown the in-plane Poisson’s ratio behaviour for each of the three beam types. Type I beams give a Poisson’s ratio of 1, type II zero and type III -1. The analytical modelling gives an exact solution, however equations must be derived for each case. Additional beams add to the complexity of the analysis. In order to investigate the elastic properties of combinations of beam structures, the next section shows how the finite element method can provide an automated, robust analysis tool. 177 6.4 Finite element modelling methodology This section presents the finite element modelling of combinations of the beam networks modelled in section 6.3. The parameters of the analysis are discussed, and the validity of the results appraised. High throughput techniques are used to conduct analysis of beam networks and the resulting elastic properties calculated. Finally the conclusions of this study are presented. 6.4.1 Element type When using the finite element method to perform calculations of beam networks, several assumptions are made about the type of beams being used. The purpose of the study is to investigate the feasibility of generating scale independent networks with negative, or tailored Poisson’s ratio. The manufacturing techniques and constituent materials used to assemble the networks are dependent on the scale of the structure being generated and thus in this study, relative material properties are used. There are two possible methods for varying the relative stiffness of the beams (EI); varying the Young’s modulus of the constituent material (E), or varying the second moment of area, I by varying the thickness (for circular beams, radius) of the beams. One dimensional or three dimensional elements can be employed. Using one dimensional elements enables the linear-elastic deformation of the beams to be captured at a low computational cost. 178 Figure 6.10: Schematic showing the deformed shape of a cantilevered beam. Top to bottom - Blue - 1D beam elements, 3 nodes; Green - 1D beams elements, 15 nodes; Orange - 3D tetrahedral elements, 1200 nodes; Red - 3D tetrahedral elements, 96000 nodes; Black - Analytical (Euler-Bernoulli) model. 6.4.2 Mesh convergence To determine the validity of the finite element simulations, it is standard practice to conduct a mesh convergence. It is possible to capture the initial geometry of the beams with one element however this does not fully replicate the deformed shape of the beams. The mesh convergence determines the required mesh density in order to accurately simulate the structure. A fixed geometry and loading condition (applied displacement in the (110) direction) is issued the for analysis, and the Poisson’s ratio of the structure calculated for increasing mesh densities, shown in figure 6.11. As the mesh density increases, the Poisson’s ratio of the structure increases, as the finer mesh is able to replicate the lateral and axial displacement of the component beams with greater accuracy (see figure 6.10).. 179 Figure 6.11: Evolution of Poisson’s ratio with mesh density for an arbitrary type II beam network 6.4.3 Boundary Conditions Boundary conditions within the finite element model determine the translational and rotational constraints placed on the nodes within the model. When modelling crystalline structures, it is normal to use periodic or Bloch boundary conditions as the number of atoms in a crystal may run to several billion. The structures that are modelled in the finite element study have geometries based on the crystalline structures, but all are of a comparatively small number of unit cells. For this reason, true periodic boundary conditions are not imposed. A schematic representation of the applied boundary conditions is shown in figure 6.12. Figure 6.12: Schematic showing the boundary conditions applied for axial (left) and shear (right) loading. 180 6.4.3.1 Representative volume element For this study a Representative Volume Element (RVE) is chosen as the unit cell, as shown in figure 6.13. Whilst not the primitive cell of the reciprocal lattice, this structure was chosen as it is the smallest unit cell that encompasses the unit cell shown by Baughman et al. (1998). This RVE cell is subsequently verified against a larger lattice structure. The RVE model enables rapid generation and analysis of a large number of models at a comparatively low computational expense, as well as a simple way of visualising the deformation mechanism of the structures. Figure 6.13: Representative volume element, showing the Type I (red, dotted), Type II (green, solid) and Type III (blue, dashed) component parts. 6.4.3.2 Panel model An array of 10x10x2 RVE unit cells was constructed to represent a sandwich panel core style structure. Not only does this serve as a verification of the RVE model, but also shows how the RVE could be used in a sandwich panel or space-frame structure. Having a large array also simplified the boundary conditions required for shear deformation, as periodicity and edge effects can be disregarded. 181 6.4.3.2.1 Comparison of RVE and Panel Model The panel model and RVE were modelled for the structures where the Type I, II and III structures are dominant. The results are shown in table 6.2. Type I Type II Type III RVE Panel RVE Panel RVE Panel ν110 0.9 0.9 0 0 0.95 0.95 ν100 0 0 0.26 0.26 0.33 0.33 Table 6.2: Comparison of panel and RVE model 6.4.3.3 Material Properties The combined structure was modelled in the finite element package. Each type of beam was assigned discrete material properties, but the radius kept constant. The stiffnesses were varied from 1 to 9000 arbitrary units. The relative stiffnesses of the beams control the deformation of the structure, and govern the resulting Poisson’s ratio, shear modulus and compressive modulus. The structure is scale independent and thus the units of force, modulus and length are specified as relative, rather than absolute values. The dimensions of the type one beams were specified as a length of 10, and a circular cross-section of 0.3, kept constant for all three beam types. Abaqus 6.8 (Dassault, 2008) was used to perform the finite-element analysis. Beams were modelled as B31 type linear beam elements. Linear elastic stiffnesses were assigned ranging across several orders of magnitude. Two geometries were used in the analysis in order to determine the accuracy of the boundary conditions employed, a representative volume element and a ’panel’ model. 6.4.3.4 Loading conditions The structures under consideration were subjected to a known displacement in the [110] direction and the deformations in plane were measured such that Poisson’s ratio of the structure could be found. For the panel models, the structures are displaced at the nodes along the flat faces, and the Poisson’s ratio calculated in plane. For comparison the 182 structures were also loaded in the [100] direction. Shear deformations were applied to the flat faces of the panel model in both cases.For the RVE, load conditions were those used in the analytic modelling, section 6.3. 6.4.4 6.4.4.1 High throughput techniques Varying material properties To conduct the high throughput analysis, the model is defined such that the each type of beam can be assigned separate material properties. The input file from the initial model is used as a template for generation of the large number of input files required to conduct the study, and these can then be batch processed for analysis and post-processing. This technique is computationally very efficient; the high level scripting languages Python or Perl are used to edit the relevant text files. 6.4.4.2 Automatic generation of beam networks from crystallographic structure data As outlined above, the commercial FEA code employed in this study, Abaqus, has the facility for use of the Python scripting interface to generate and analyse models using the finite element method. Having demonstrated the possibility of crystal structures to provide a geometric framework for a negative Poisson’s ratio structure, an automated methodology is used to construct these networks. The crystalline structure of materials is well known from experimental data, and repositories of such data are readily available from such sources as the Crystallographic Database Service (Fletcher et al., 1996). This data is often available in the format of rectangular coordinates which can then be used as a starting point for the geometry of the finite element model. The script is designed to be run from within Abaqus with minimal input, or knowledge of the scripting interface from the user, and is designed to directly report stresses and strains from compressive and shear tests in order to fully characterise the material. The function of the code can be summarised as: • The coordinate points are used to calculate the vectors that connect these points, 183 duplicate points are removed. • The cutoff value is defined by the user, either as a nearest neighbour or a longest length. This value is used to determine the lengths of the beams to be constructed. Vectors longer than the cutoff value are discarded. Each vector is only constructed once, and replicated in the part as necessary. • The beams are drawn within the software. • Each beam is translated and rotated to the correct position. • The material properties for each length of beam can be individually assigned, or assigned as a uniform value. In order to maintain slender beam assumptions, the default diameter of the beams used in the calculations is 1 15 of the length value of the shortest beam within the structure. • The material properties are assigned to each beam, and the beams merged into one uniform part such that the deformation modes are stretching and bending, not hinging. Beam, not truss, elements are used so shear effects are included in the analysis. • Displacements are assigned for compressive strains in the 1,2,3 and shear in the 12, 23, 13 directions. Boundary conditions are such that the part is restrained in space, but is free to expand/contract. • After analysis, the stresses, strains and displacements (for Poisson’s ratio) are reported, from which elastic constants can be calculated. 6.4.4.3 Calculation of properties Poisson’s ratios were calculated for deformations in both the [100] and [110] direction to give ν(010,100) and ν(11̄0,110) . The behaviour was assumed to be linear elastic, and engineering strains were calculated. Data were taken from both the RVE and panel model for comparison. When calculating the Poisson’s ratio in the (100) plane, the computed strains were those in the [100] (load) direction and the [010] perpendicular in plane direction. 184 For loading in the [110] orientation, strains used were those in the [110] direction and the in-plane perpendicular direction. Loads were applied to edge nodes as a fixed displacement. Nodes were free to translate in the corresponding orthogonal directions to avoid over constraining the structures. The perpendicular displacements were recorded at the edge nodes of the structure. The tensile and shear moduli were calculated from the panel model (rather than the RVE) to avoid having to include periodic boundary conditions in the analysis periodicity into the analysis. When calculating the moduli for cases where one, two, or three types of beam are dominant the stiffness value of the dominant beams is the maximum value of 9000, and the resulting moduli of the structure are in the same arbitrary units. Whilst it may seem more sensible to normalise the values with respect to this maximum, this would not allow comparison for values where the dominant class of beam does not have the maximum stiffness value. 6.4.5 Validation of finite element modelling with analytical model To validate the computational model, a direct comparison is made with the analytically derived solution of the Poisson’s ratio. The result of this, compared to the finite element analysis of the same structure is tabulated in table 6.3. Nodal displacements are shown; for identical geometries, the displacements are equal (to three significant figures). This is replicated in the other two structures. Point Analytical δx Analytical δy FE δx FE δy 1 0 0 0 0 2 2.65 2.65 2.65 2.65 3 0 5.3 0 5.3 4 -2.65 2.65 -2.65 2.65 Table 6.3: Displacements of nodal points for Type I structure loaded in [110] direction. Example geometry = length 100mm, diameter 1mm, Young’s modulus 1600MPa 185 6.5 6.5.1 Results of finite element modelling Overview Figure 6.14: Poisson’s ratio of structure for all considered cases, for strains in the [110] direction. Horizontal axes show normalised comparative beam stiffness. Red region shows structures where Type I beams dominate, green where Type II dominate, and blue where III dominate. This section provides a general overview of the Poisson’s ratio trends exhibited by the structure when loaded in the [110] direction. The effects of varying two specific beam types within the structure is analysed in order to characterise the specific role of each beam type on the Poisson’s ratio, tensile and shear modulus. Comparisons are then made between the behaviour of the structure in the auxetic and non-auxetic [100] load directions. Data for the Poisson’s ratio was taken from the RVE model, and subsequently verified against data from the panel model. Data for the elastic moduli were taken from the panel model. Figure 6.14 shows values for Poisson’s ratio for all combinations of beam stiffnesses where the stiffness of the type I, II and III are varied by a factor of 20 (presented in dimensionless units of beam stiffnesses normalised with respect to the stiffness of the type three component). As shown in section 6.3, each type of beam structure can be seen to have a differing deformation mechanism resulting in a differing Poisson’s ratio. The influence of type I beams on the deformation of the structure is to give a Poisson’s ratio of +1, the upper 186 bound for in-plane strains for a type I structure. Deformation of type II structures gives a Poisson’s ratio of around zero when strained in the [110] direction. The deformation mechanism of the structures where type III beams are dominant gives a Poisson’s ratio of -1, as shown to be the lower limit by Baughman et al. (1998) (for an in-depth discussion of Poisson’s ratio bounds in the cubic crystal system see Ting and Chen (2005)). The in-plane negative Poisson’s ratio has an associated out of plane positive Poisson’s ratio deformation. Increasing the stiffness of type I and II beams will reduce this effect as the structure becomes stiffer. 6.5.2 Comparison of panel model and RVE Panel model The resulting structures for a positive, zero and negative Poisson’s ratio deformation is shown in figure 6.15(a), 6.15(b) and 6.15(c). The panel model shows how the representative volume element can be positioned, as would be employed in a sandwich panel or generic truss structure. The Poisson’s ratio, relative shear modulus and relative Young’s modulus of the structure are shown in figures 6.18, 6.19, 6.20. Absolute values are not given as these are conditional on the parent material of the structure and manufacturing techniques employed. 187 (a) ν = +1 (b) ν = 0 (c) ν = -1 Figure 6.15: In-plane deformed configuration of panel model showing (a) Positive Poisson’s ratio, (b) Zero Poisson’s ratio and (c) Negative Poisson’s ratio (Deformed green, Undeformed black) RVE Analysis of the RVE model allows the deformation mechanism to be observed in detail. The deformation behaviour in these structures results from flexure of the beams in the structure (see Fig. 2). The in-plane Poisson’s ratio for structures with all considered relative component beam stiffnesses is shown in figure 6.16. 188 Figure 6.16: Deformed shape of the RVE model showing (left to right) Poisson’s ratio of 1, 0 and -1 for strains in the [110] direction (where load direction is perpendicular to page. Deformed black, undeformed grey). 6.5.3 Elastic properties Each substructure has a differing deformation mechanism under tensile load, thus by varying the relationship of the component beam stiffnesses it is possible to vary the Poisson’s ratio of the whole structure. If it is assumed that once the stiffness of one type of beams is considerably lower than that of the other components its influence on the deformation mechanism can be ignored then seven discrete types of substructure can be identified: • Structures comprising only type I, type II, or type III beams. • Structures comprising type I and II, I and III, or II and III beams (see figure 6.17). • Structures where all types of beams have an influence on the deformation mechanism. 189 (a) (b) (c) Figure 6.17: Representative unit cell for (a) Type I and Type II; (b) Type I and Type III; and (c) Type II and Type III Combined structures. By considering each of the substructures independently it is then possible to analyse the effect that each type of beams has not only the deformation mechanism and resulting Poisson’s ratio of the structure but also on the tensile strength and shear moduli of the combined structure. It should be possible to calculate the required material properties to manufacture a structure with a desired Poisson’s ratio, shear and tensile modulus. Considering structures where two types of beam are present enables the interaction of the different sub structures to be analysed. 190 6.5.3.1 Poisson’s ratio Figure 6.18: Poisson’s ratio of structure for cases where two types of beams are dominant (strains are in the [110] direction). (a) Ratio: Type I/Type II structure. (b) Ratio: Type I/Type III structure. (c) Ratio: Type II/Type III structure. 191 Considering the structures where two types of beam are dominant, trends can be observed corresponding to the relative (thus dimensionless) stiffnesses of the component beams, as one beam type increases in stiffness relative to the other. The Poisson’s ratio of these substructures is shown in figure 6.18. The trend echoes that of the three component beam structure; a type III dominated structure exhibits a negative Poisson’s ratio, type II a Poisson’s ratio of around zero and type I a positive Poisson’s ratio. 192 6.5.3.2 Young’s modulus Figure 6.19: Tensile modulus of structure for cases where two types of beams are dominant. (a) Type I/Type II structure. (b) Type I/Type III structure. (c) Type II/Type III structure. 193 Comparison of the maximum tensile moduli of the structures where two types of beams govern the deformation shows that the maximum value for tensile modulus is found in the type I/type II structure with a comparative modulus of 610, the type I/type III has a comparative modulus of 395 and the type II/type III structure a comparative modulus of 561. This shows that the inclusion of type II beams in the structure leads to an increase in the relative stiffness of the structure at the expense of auxeticity, as these beams link the vertices displaced by the auxetic deformation. Considering each substructure in detail, the influence of each individual type of beam can be observed. Figure 6.19 shows the affect on tensile modulus of varying the component beam stiffness for a type I and II; type I and III; and type II and III structure, it can be seen that for a given ratio, the absolute magnitude of the stiffness increases as the magnitude of the component beams increases, but each configuration shows the same trend. For a fixed value of type II beams, increasing the stiffness of the type I beams by a factor of nine has the effect of approximately doubling the modulus of the structure. For a fixed value of type I beams, the effect of varying the type II beams by a factor of two is to approximately double the stiffness of the structure, as the deformation mode is axial stretching of these beams. As the initial value of the type I beam increases, this effect is reduced slightly. A type I/type III structure shows a similar trend. The lower modulus for values where the only one type of beam is present should be noted, as the solely type III beam structure is far less stiff than one comprising type II beams alone. Also, for the type I/type III structure, overall stiffness of the structure is far lower, highlighting the compromise that must be reached when a negative Poisson’s ratio structure is sought. The type II/type III structure shows that for a low value of type II beams, increasing the stiffness of the type III beams has little effect on the overall modulus of the structure. 194 6.5.3.3 Shear modulus Figure 6.20: Shear modulus of structure for cases where two types of beams are dominant. (a) Type I/Type II structure. (b) Type I/Type III structure. (c) Type II/Type III structure. As with the tensile modulus, the maximum shear modulus of 383 is found in the type I/type II structure, for a type I/type III structure Gmax =276 and for a type II/type III structure Gmax =175. Considering each substructure independently, we see that for a type I/type II structure the trend for shear is very similar to that seen in tension, at low values of 195 type II beams, increasing the stiffness of type II beams by a factor of nine increases the global stiffness by a factor of around 2. For a type I/type III structure, and type II/type III structures it should be again noted that the influence of the type II beams on the global modulus is much higher than that of the type I beams. 6.5.3.4 Comparison with [100] axis All previous data have been reported for the structure strained in the [110] direction. For comparison, the structure is also loaded in the [100] direction. Comparisons can be drawn between the structure when loaded in the extremal Poisson’s ratio, and the ’flat’ directions. The Poisson’s ratio for seven asymptotic structures is shown in Table 1. It can be seen that the Poisson’s ratio varies from around zero to around 0.33, but is always positive in all cases. The tensile moduli of the structures in the [100] orientation shows the same general trend as in the [110] direction, however it should be noted that by rotating the unit cell, the orientation of type I beams in the [100] direction means they are now loaded in the same axis as type II beams were in the [110] direction (and conversely the orientation of type II beams means that they now are loaded in the manner that type I were in the previous orientation). It should also be noted that the flexural behaviour differs slightly due to the change in relative beam lengths. 196 6.5.3.5 Summary of results Structure Tensile modulus Shear modulus Poisson’s ratio Type I 583 1.8 0 Type II 397 192 0.26 Type III 3.25 5 0.33 Type I/II 1078 310 0.14 Type I/III 711 224 0.12 Type II/III 413 263 0.29 Type I/II/III 1135 673 0.17 Type I 2.5 42 0.9 Type II 415 152 0.01 Type III 177 5 -0.95 Type I/II 610 383 0.4 Type I/III 395 276 0.5 Type II/III 561 175 -0.2 Type I/II/III 905 453 0.12 [100] direction [110] direction Table 6.4: Comparative (dimensionless) values for tensile and shear modulus, and Poisson’s ratio for maximum component beam stiffness. Table 6.4 summarises the results. The first thing to notice in Table 1 is the increase in global stiffness of the substructure by combination of two types of beams. The second aspect is that while strains for [100] direction never result in a negative Poisson’s ratio behaviour, deformations in the [110] direction give negative Poisson’s ratio in cases where type III beams dominate. The third point is that compromises can be reached where the structure is able to demonstrate negative Poisson’s ratio by a combination of type II and type III beams whilst maintaining a comparatively high tensile and shear modulus. It is generally accepted that auxetic materials have a high shear modulus (see 1). For 197 these structures, the negative Poisson’s ratio behaviour is at the expense of the shear rigidity as the mechanism depends on there being no cross linking within the structure. 6.5.3.6 Mechanical performance of beam networks relative to conventional honeycomb Conventional (hexagonal) honeycombs are widely employed in engineering applications due to their high stiffness and comparative low weight. Whilst their properties are well known, and can be easily modelled (see, for example, Ashby and Gibson (1997)), the Poisson’s ratio of such structures is fixed at 1 for regular hexagons, and always positive for variations thereof. Re-entrant honeycombs (as discussed in section 1.1.3) can be constructed with an in-Plane Poisson’s ratio of -1. As with the structures presented in this work, the mechanical properties of honeycombs is dependent on the geometry and parent material of the structure. Engineering design usually requires a low weight, high stiffness structure and thus to enable the comparison of the properties of differing honeycomb (or other substructures), performance is usually expressed in terms of the properties of the structure relative to that of its parent material. To enable comparison of the structures presented in this work with those of commercially available honeycombs, relative density (ρR ), relative shear modulus (GR ) and relative Young’s modulus (ER ) will be used to describe the relationship between the apparent properties (P∗ ) and the property of the parent material (Ps ). In order to normalise each of the structures, the parameters shown in table 6.5 are used. Whilst the finite element analysis uses consistent units, units are shown for clarity. The relative material properties can be calculated: E∗ Es G∗ GR = Gs ρ∗ ρR = . ρs ER = 198 (6.7) (6.8) (6.9) Property Value Units 1000 mm3 0.1 mm Es 1000 MPa Gs 384 MPa ρs 1 g mm−3 Apparent volume of unit cell Radius of circular beams Table 6.5: Parameters used in the finite element analysis, and subsequently in determining normalised material properties. ν110,11̄0 M (g) ρ∗ (gmm−3 ) ρR E∗ (MPa) ER G∗ (MPa) GR ER ρR GR ρR Type I 0.9 3.77 3.77E-03 3.77E-03 2.5 0 42 0.11 0.66 29.01 Type II 0.01 6.53 6.53E-03 6.53E-03 415 0.41 152 0.4 63.56 60.62 Type III -0.95 1.78 1.78E-03 1.78E-03 177 0.18 5 0.01 99.6 7.33 Type I/II 0.4 10.3 10.3E-03 10.3E-03 610 0.61 383 1 59.23 96.84 Type II/III 0.5 8.31 8.31E-03 8.31E-03 395 0.4 276 0.72 47.55 86.53 Type I/III -0.2 5.55 5.55E-03 5.55E-03 561 0.56 175 0.46 101.13 82.16 Type I/II/III 0.12 12.08 12.1E-03 12.1E-03 905 0.91 453 1.18 74.94 97.68 Structure Table 6.6: Apparent and relative mechanical properties for structures where one, two and three beams are present within the structure, and the component beams have equal material properties (see table 6.5). rho∗ (g/cm3) ρR E∗ (Mpa) ER G∗ (Mpa) GR ER ρR GR ρR 0.18 mm 0.104 3.85E-02 1896.06 2.75E-02 275.79 1.08E-02 0.713 0.281 0.03 mm 0.147 5.46E-02 2895.80 4.20E-02 365.42 1.43E-02 0.769 0.263 0.04 mm 0.199 7.35E-02 4481.59 6.50E-02 448.16 1.76E-02 0.883 0.239 Gauge Table 6.7: Apparent and relative mechanical properties for three examples of conventional honeycomb structures. Data taken from Hexcel (2012). Table 6.6 shows the normalised material properties of the structures. Table 6.7 shows comparable data, taken from experimental measurements (Hexcel, 2012). The compressive and shear stiffness, as a function of the relative density of the structure is shown for 199 both the beam networks and the classical honeycombs. It can be seen that the relative in-plane performance of the beam networks, when density is considered, is greater than that for the conventional honeycombs, however the absolute mechanical properties of the honeycombs are far superior, owing to the greater cross sectional area of the structures. It should also be noted that many applications of cellular solids utilise the out of plane properties of the structure (see for example Zhang and Ashby (1992)). 6.5.3.7 Conclusion A structure consisting of a network of bending beams can exhibit a negative Poisson’s ratio. The negative Poisson’s ratio behaviour is driven by the (bcc analogous) type III beams, the type II (fcc like) beams result in a structure with a Poisson’s ratio of around zero and type I (simple cubic configuration) beams result in a Poisson’s ratio of around +1. The tensile and shear moduli of the type III beams can be augmented by addition of type II and type III beams. By tailoring the relative stiffness of the component beams within the structure it is possible to design an auxetic truss structure with a specific Poisson’s ratio, shear modulus and tensile modulus. 200 Chapter 7 Discussion 7.1 Synopsis This work has explored negative Poisson’s ratio in cubic structures using a variety of analytic and modelling techniques and at different scales, from the atomic to the macroscopic. The evidence for negative Poisson’s ratio in a variety of materials, both naturally occurring crystals and man made structures, is reviewed through a thorough analysis of the literature. From this, it is apparent that while negative Poisson’s ratio in elemental cubic metals has been considered by previous studies, no work has systematically explored this phenomenon to the same extent, or used the same breadth of techniques as this thesis. Previous studies have focussed on either experimental data, simple modelling of isolated cases, or abstract mathematical consideration of the elastic tensors. A mechanism for auxeticity in body-centred cubic materials was proposed as early as 1998, but no mechanism for auxeticity in face-centred cubic metals has since been determined, despite this being the more common behaviour. Away from the cubic system, crystal structures of lower symmetries exhibit negative Poisson’s ratios; this has in rare cases been experimentally observed directly, but in the majority of cases, the presence of auxeticity in off-axis directions is calculated from either the experimentally observed axial data, or data obtained from nano-indentation, as a result little is known about the underlying mechanism of such behaviour. 201 7.1.1 Classical Modelling Models of differing levels of complexity can be used to explore elastic properties in general, and auxeticity in particular, using the formalism of classical elasticity. The methodology is conceptually simple as elastic stiffnesses can be obtained from the energy changes of a few specific strains (three in the case of materials with cubic symmetry). The explicit derivation of this approach, and its application to the study of auxetic media is shown analytically for the simpler models of interaction. Table 7.1 displays the results from this analytic work. The first model has a very simple harmonic volumetric dependence and it shows that for an arbitrary cubic system, the Poisson’s ratio is positive in the axial directions ((ν(100,010) ), and negative in the off axis direction ((ν(110,11̄0) ). Modelling the first nearest neighbour bonding with linear springs results a more sophisticated model which includes a description of the Bravais lattice of the structure. For the body-centred cubic unit cell, Poisson’s ratio in the (110, 11̄0) direction is predicted to be negative while in the face-centred cubic crystal, this is not the case and the Poisson’s ratio is found to be positive, albeit possibly zero. This supports the hypothesis that the negative Poisson’s ratio behaviour in BCC crystals is driven by the first nearest neighbour bonds, but in FCC crystals additional bonding must be taken into account to allow auxeticity. ν(010,100) ν(11̄0),(110) ν(001,110) Volume 1 2 -1 0 Spring - FCC 1 3 0 1 2 Spring - BCC 1 2 -1 2 Lennard-Jones - FCCnn1 1 3 0 1 2 Lennard-Jones - BCCnn1 1 2 -1 2 Lennard-Jones - FCCnn2 0.36 -0.09 0.62 Lennard-Jones - BCCnn2 0.53 -1.55 2.93 Potential Table 7.1: Poisson’s ratio calculated for each of the potential models Replicating the methodology applied to the simple linear springs, first nearest neighbour bonding is represented by the Lennard-Jones potential; a two-body potential that 202 accounts for the repulsive component within the atomic bonding. For the face-centred cubic cells, negative Poisson’s ratio is not allowed, and for the body centred cubic unit cells, the predicted value was -1. Adding second nearest neighbour bonding to the model in this analytical model yields more realistic results with auxeticity being more prevalent. For the body-centred cubic crystal, the anisotropic character of the deformation is captured, and the Poisson’s ratio is calculated to be -1.5 with a corresponding perpendicular positive Poisson’s ratio at 2.9. For the face-centred cubic cell the in plane Poisson’s ratio is finally negative, at -0.09. The Lennard-Jones model not only gives a more accurate representation of the bonding than the simple spring models as it includes a term for the modelling repulsive component but by permitting second nearest neighbour bonding, it is possible to further improve the accuracy of the model. Adding further neighbours increases the complexity of the model and thus analytical derivations are abandoned in favour of computational methods to find numerical solutions to the analyses. The potentials considered are the two-body Morse potential (fitted to both BCC and FCC crystals) and the many-body Finnis-Sinclair (fitted to BCC metals), Cleri-Rosato and Sutton Chen potentials (both fitted to FCC metals). The fitted values for the elemental cubic metals are obtained from the literature and these are used to investigate the off-axis Poisson’s ratio behaviour of the elements. Preliminary cutoff convergence studies are carefully conducted to determine the optimal range for the potentials. The Morse data shows that the two-body potential gives values where the c12 and c44 are equal, and thus always predicts negative Poisson’s ratio behaviour. The manybody potentials yield far more interesting results. It is possible to successfully model the experimentally observed negative Poisson’s ratio with the many-body potentials, but unlike the two-body potentials, these are also capable of modelling positive Poisson’s ratio behaviour where it exists. From this classical potential modelling, possible mechanisms for the negative Poisson’s ratio behaviour are derived. As previously mentioned, the mechanism for the auxetic behaviour in BCC metals has been discussed in the literature but no mechanism for the auxetic behaviour in face-centred cubic crystals has been suggested to date. By 203 using the deformation computed for negative Poisson’s ratio in the FCC crystals, a possible mechanism is found and detailed. The ability of the potentials to model Poisson’s ratio is summarised in table 7.2. BCC FCC ν+ ν- ν+ ν- Morse 7 3 7 3 Finnis-Sinclair 3 7 - - Sutton-Chen - - 3 3 Cleri-Rosato - - 3 3 Potential Table 7.2: Ability of each potential to predict Poisson’s ratio To understand the underlying cause of the auxetic behaviour, the numerical approach was applied to the simple spring methodology used earlier. Cubic, covalently bonded compounds are selected for the analysis as the bonding in these crystals is more directional in character and thus better suited to being modelled in this way. A simple model, based on the methodology of Keating, and originally used to investigate the properties of Diamond is employed. The parameters of the potentials are fitted to the elastic constant data available in the literature, and the calculated spring constants normalised to give a comparative stiffness value for the axial and tangential bonding stiffnesses. This analysis shows that, for the systems considered, the fitted constants are within an acceptable margin for error and able to reproduce the elastic properties. As the ratio of the axial to lateral stiffness energies increased, the Poisson’s ratio also increased. The classical potential modelling has been conducted to determine whether these relatively simple and computationally inexpensive models are able to accurately predict negative Poisson’s ratio in cubic elemental metals, and whether any insight could be gleaned as to what governs whether or not a crystal exhibiting auxetic behaviour. It is apparent that for most elements a correctly fitted many-body potential can predict Poisson’s ratio, but the accuracy is wholly dependent on the accuracy of the fitting parameters and the suitability of the potential form. 204 7.1.2 Ab-initio modelling To model the Poisson’s ratio of the elemental cubic metals without such a dependence on the experimental data, density functional theory techniques have been employed. These first principal techniques have been used to determine the elastic constants for a range of materials. As with the classical potentials the exact form of the modelling, specifically the pseudopotential used, has an effect on the accuracy of the solution. For ferromagnetic elements it is also necessary to include the spin polarisation to better replicate the elastic properties. Inclusion of the Projector augmented wave forms also yields an improved solution, and in the majority of cases, can not only accurately predict the sign of Poisson’s ratio but also its magnitude. The density functional theory approach, whilst incurring a greater computational expense gives a more robust modelling methodology when predicting the elastic constants of the elemental metals. These techniques are able to predict both positive and negative Poisson’s ratio behaviour, but the success of the analysis is dependent on the accuracy of the c44 term. 7.1.3 Mechanical modelling Finally, the information derived from the modelling of the interatomic interactions in the crystals has been used to inspire the scale independent beam networks that are explored using the finite element method. The beam networks are directly based on the first, second and third nearest neighbour interactions in cubic crystals. By replicating each of the component bonds as linear elastic bending beams of varying elastic (Young’s modulus), structures with tailored elastic properties have been generated. The same principles of how the bonding in the physical crystals can drive the auxetic behaviour seem to hold. Simple cubic analogous (type I) beams give a positive Poisson’s ratio, face-centred cubic analogous (type II) beams give Poisson’s ratio of zero and bodycentred cubic analogous (type III) beams lead to a negative Poisson’s ratio. Elastic and shear stiffnesses of the structure are also governed by the relative component beams. By generating structures where each component beam type has a differing elastic stiffness it is possible to construct an array of beam with tailored mechanical properties: Poisson’sratio, shear stiffness and axial stiffness. 205 7.2 Further Work Extensions of the work conducted within this thesis should follow two streams. • Further analysis of the beam networks modelled in this work • Extension of the modelling techniques employed to investigate auxetic behaviour in crystals of lower symmetry 7.2.1 Exploration of auxetic behaviour in crystals This work has provided an analysis of the suitability of modelling techniques to the observation of negative Poisson’s ratio. It shows that techniques that are able to model the structure of a material to a high degree of accuracy are in some cases unable to successfully predict one or more of the elastic constants, in particular c44 . As a result, the calculation of Poisson’s ratio is not possible with any degree of certainty for some combinations of material and methodology. As the symmetry of the crystal under consideration increases, the number of elastic constants and thus the potential for inaccuracies in the analysis increases. In extending this methodology to other symmetries of crystal, it would be prudent to start with structures where the symmetry of the structure provides a unit cell that can be replicated and give a panel geometry that has planar edges - some of the lower symmetry crystals will not provide such unit cells. In doing so, it will subsequently be possible to characterise the structures using both finite element and experimental methods. Consideration should be given to both the magnitude and orientation of the Poisson’s ratio that has been observed experimentally, examples of such materials are discussed at length in chapter 1. Structures such as MOF-5 and alpha cristobalite lend themselves well to being the starting framework. Modelling of these materials could in the future yield interesting structures with novel mechanical properties. 206 (b) α-cristobalite (a) MOF 5 Figure 7.1: Possible network structures of MOF5 and α-cristobalite. Classical potential modelling, whilst simple and not infallible has proven useful in understanding the underlying mechanism of these behaviours. Density functional theory techniques have proved more reliable in determining the magnitude and direction of negative Poisson’s ratio behaviour and would provide a good starting point when considering materials where the existence of auxetic behaviour is based on nano-indentation or other techniques subject to a high degree of experimental uncertainty. 7.2.2 Beam networks The beam networks investigated using the finite element method can be produced from a variety of manufacturing methods, dependent on the scale required. For manufacture where cell sizes are at the centimetre scale, additive manufacture would be most appropriate, even if materials anisotropy is present. Comparative beam stiffnesses could be achieved either by varying the radius of the beams or by using a system that has the ability to use different component materials to give a parent material with variable mechanical properties. Experimental validation of the finite-element modelling at a variety of scale could provide a starting point for the application of these structures. Using the methodology developed in this work to explore beam networks based on crystals of lower symmetries could yield beam networks with further enhanced mechanical properties; lower Poisson’s ratio structures could be found, based on crystals with a high degree of anisotropy, or networks that are completely auxetic may be possible, based on some silicate style frameworks. Further analysis could be conducted to investigate the out-of-plane properties of the structures. 207 7.3 Key Findings This work has explored negative Poisson’s ratio behaviour in cubic elemental metals using a full range of numerical methods not employed in previous studies. The success of these methods in predicting this behaviour has been appraised, and inferences drawn from this analysis. Each of the methodologies employed have shown negative Poisson’s ratio behaviour exists in cubic elemental metals. Additionally, it was shown using classical potential modelling that auxetic behaviour in covalent cubic crystals can be successfully modelled. Fundamental analysis, using linear springs, showed the ’default’ behaviour of the structure within the crystals using first nearest neighbour bonds; body-centred cubic crystals exhibit negative Poisson’s ratio behaviour, face-centred cubic have a Poisson’s ratio of around zero. Classical potential modelling methodologies have demonstrated that two-body potentials, whilst perfectly adequate at producing structural and on-axis properties, are highly unsuitable for calculating off-axis properties due to them being governed by the Cauchy relation. Additionally, it was shown that increasing the ’many-body’ character of the potential, increases the accuracy. The experimental data found in the literature shows that a greater proportion of facecentred cubic crystals exhibit negative Poisson’s ratio than body-centred cubic, further investigation into this reveals that as a general trend, the magnitude of the negative Poisson’s ratio in the face-centred cubic is far lower than that of the body-centred cubic. The negative Poisson’s ratio behaviour has been shown to derive from the body-centred cubic bonding in the crystals. The interaction of the differing nearest neighbour bonding types governs the magnitude of the Poisson’s ratio. Comparison of many-body and two-body classical potential techniques supports this hypothesis. Even for a face-centred cubic system, increasing the potential range gives a lower value of Poisson’s ratio. For the body-centred cubic system, this is also true. Investigation of the deformation mechanism of the face-centred cubic structures using simple analytical springs, classical potential modelling and beam networks all shows that the Poisson’s ratio of these bonds alone is around zero. Similarly, analysis of the 208 body-centred cubic frameworks shows a Poisson’s ratio of -1. The contribution of the additional bonding within the structure is what appears to govern the negative Poisson’s ratio behaviour in the face centred cubic crystals, but the long range of these bonds results in a lower magnitude. The finite element modelling showed that is possible to transfer the crystallographic mechanism into a beam network, despite the difference in nature between the atomic bonds in an elemental metal (or springs in simple models) and the linear elastic beams. Not only does this give confidence in the findings of the atomistic modelling, but additionally, shows that for unusual elastic properties atomic scale principles can be applied at the macro scale. 209 Appendix A Methodology of DFT A.1 Introduction The following section of this thesis is limited to a pragmatic explanation of the background, methodology and factors affecting the efficacy of the technique. A.2 A.2.1 Background The Schrödinger Equation DFT is a technique for calculating the electron density and subsequent material properties by solving an approximation to the Schrödinger equation. Solving the Schrödinger equation for a realistic, many ion, many electron system is a great challenge due to the large number of degrees of freedom and can only be achieved using approximations and numerical methods. The time-independent Schrödinger equation takes the form: − h2 2 ∇ Ψ + VΨ = EΨ 2m (A.1) ∂2 ∂2 ∂2 + + . ∂x2 ∂v2 ∂z2 (A.2) where: ∇= The potential operator, V, and the wavefunction, Ψ, of the system are functions of the 210 particles positions, thus equation A.1 can be re-arranged to give: ! h2 2 − ∇ + V Ψ = EΨ 2m (A.3) 2 h where − 2m + V is the Hamiltonian operator, Ĥ. The Hamiltonian operator represents the total energy of the system expressed as a function of the coordinates of the particles and their conjugate moments and can be described as the sum of the kinetic and potential energies. A.2.2 Born Oppenheimer Approximation The Born-Oppenheimer approximation states that as the electrons are much lighter than nuclei and move at a greater velocity, their motion can be de-coupled from that of the nuclei. This allows the wavefunctions to be split into the wavefunction of the nuclei and the wavefunction of the electrons. The wavefunctions for the nuclei can then be treated classically, and be a part of the hamiltonian for the electronic wave function. A.2.3 Hohenberg-Kohn Theorem and Kohn-Sham Theorem The electronic wavefunction is still a complex many-dimensional mathematical object. Further steps must be taken to simplify the problem. The next step is to note that the ground state of a quantum system can be calculated from electronic density alone (Hohenberg et al., 1964). This dependence on the electronic density is referred to as the Hohenberg-Kohn theorem. This is powerful, as the complex diagonalisation procedure to obtain the eigenvalues of the the Hamiltonian associated with the direct solution of Schrödinger’s equation is replaced by a one to one mapping from electronic density to ground-state energy. In other words, the ground state energy is a functional of the density, (a functional being a function which takes a function as its input argument). Still, the exact form of this energy functional is not known and is very difficult to approximate for systems of interacting electrons. The last, crucial approximation consists in replacing the system of interacting electrons 211 by a system of non-interacting electrons of the same density. As they are independent, each electron has its own Shrödinger equation, where the effective potential operator Ve f f must take into account the interaction with the other electrons (Kohn et al., 1965). Using these assumptions, the energy functional of the DFT system reduces to: Z E[ρ] = TNI [ρ] + ρ(r)ν(r)dr + 1 2 Z Z ρ(r’) drdr’ + Exc [ρ] |r − r’| (A.4) This then gives the energy for a system as a sum of the non-interacting particle (Kohn Sham) kinetic energy (TNI ), the external potential, the classically treated coulombic interacting energies and the exchange correlation functional. The exchange correlation functional captures everything that has been neglected, chiefly the quantum only exchange energy. However this is still a functional at this stage, and therefore still a complex mathematical object. A lot of the success of DFT can be attributed to the fact that this exchange-correlation functional can be approximated quite easily. The two most common schemes are the Local Density Approximation (LDA), and General Gradient Approximation (GGA), but more complex ’hybrid’ or schemes exist. The local density approximation assumes that the density can be treated as that of a homogeneous electron gas, and the exchange-correlation potential becomes the integral of the product of the electronic density by an exchange correlation function (derived from accurate homogeneous electron gas calculations). Z ELDA xc = [ρ] = ρ(r)xc (ρ)dr (A.5) There are many GGA approaches, but in general, they include a contribution due to the gradient of the electron density, and can be more successful in situations where the electron density varies significantly through the system. A.2.4 Self consistent loop There is an apparent shortcoming in the algorithm outlined previously. It possible to access the electronic density from summing the squares of the the wave functions obtained 212 Figure A.1: Flow diagram of DFT process from the Kohn-Sham equations, but the exchange correlation terms in these equations depends on the electronic density. A well known solution to such correlated problems is to use a so called ’self consistent loop’. The usual idea consists of starting from a reasonable electronic distribution, based, for instance, on independent atoms, derive the Kohn-sham equations, solve them and compare the resulting density with the input density. If the difference is lower than a user chosen threshold, the loop is terminated, and the observables (usually energy) can be calculated. If not, a new density is formed from mixing the input and output densities (different mixing schemes exist), and the process repeated, as shown in figure A.1. 213 A.2.5 Hellman-Feynman Theorem The Hellman-Feynman (Feynman, 1939; Hellmann, 1941) theorem relates the derivative of the total energy to the derivative of the Hamiltonian and allows the forces in the system to be obtained. Once the electronic distribution has been found, classical electrostatics can be used to find the forces in the system. In classical mechanics, force at a position R can be expressed in terms of the energy: F = −∇R U(R) (A.6) In quantum mechanics, this can be expressed as: F = −∇R hEi (A.7) where: hEi = hΨHΨi hΨΨi (A.8) To find hEi for the wave function Ψ, it can be expanded in terms of a set of fixed basis functions (equation A.9), and then the energy can be minimised with respect to ci . |Ψi = X ci |ϕi i (A.9) i Using plane waves as the basis functions the general expression for the forces becomes: * + X * + X ∂hEi ∂ X ∗ ∗ ∂H ∗ ∗ = ci ϕi (r)|H| c j ϕ j (r) = ci c j ϕi (r)| |ϕ j (r) ∂R ∂R ∂R i j (A.10) i,j The forces are then calculated from the from the expansion coefficients used to minimise the energy, reducing the computational cost. The choice of functional can determine the accuracy of the result; the LDA has a tendency to underestimate the ground state energy, the GGA to overestimate it. As a consequence, care must be taken when selecting the methodology to be employed. 214 A.2.6 Choice of Pseudpotentials Although it is by no means necessary, most modern implementations of DFT use an additional approximation, which consists of incorporating the non valence electrons into a pseudo-core, which replaces the original atomic nucleus. The benefits are obvious; the number of electrons is reduced to just the valence electrons. The cost is that the simple external, coulombic electron-nucleus potential is replaced by a quite complex electron-core pseudopotential. These pseudopotential are generated by conducting full electrons DFT calculations on atomic systems. Pseudopotentials are intimately linked to the functional that they have been generated from: It does not make sense to use a pseudopotential generated by full electron atomic LDA-DFT to simulate a crystal using GGA-DFT for instance. Many schemes and implementations exist, but it goes beyond the scope of this thesis to discuss them. In this work, psedupotentials of the LDA and GGA form have been used. The choice of pseudopotential can determine the accuracy of the result; the LDA has a tendency to underestimate the ground state energy, the GGA to overestimate. An alternative recent refinement to the pseudopotential methodology is to use so called Projector Augmented Waves (PAW). PAW methodology uses the total energy functional, rather than a potential, and the forces derived are from the total energy. The auxiliary wave functions are transformed by a transformation operator such that the auxiliary wave functions can be mapped onto the true wave functions and the total energy can be expressed in terms of the auxiliary functions. Use of the PAW methodology can give more accurate results with no increase in computational expense. A.2.7 Periodicity, reciprocal space and k-point mesh As customary in solid state physics (Ashcroft and Mermin, 1976), the crystal is assumed to be an infinite periodic structure, as even a small crystal may span several billion atoms. The periodic nature of crystals means that it is possible to simulate the large scale by considering the reciprocal (spatial frequency) space where small wavevectors correspond to large periods. In fact only one such reciprocal unit cell has to be considered, the first Brillouin zone to be exact. The integrals over the infinite system are replaced by the 215 integrals over the Brillouin zone in reciprocal space (Bloch Theorem) (Bloch, 1929). At a finite number of points within the Brillouin zone, the function can be evaluated. These points are known as the k-point mesh. The mesh density and configuration of the k-point mesh will determine the validity of the result and thus a convergence test is required in order to determine the density at which results can be considered to be valid. A.2.8 Basis, plane waves and energy cutoff The actual calculations on wave functions (solutions of Kohn-Sham differential equations) are carried out using linear algebra, by projecting the wave function onto a basis set, and dealing with the resulting coordinates. Depending on the nature of the system, different methods exist. For instance, for molecules where the electrons are well constrained, it is convenient to use localised functions, usually centred on the atoms. Gaussian functions are very common. The more such functions are included, the better the wave functions are described. In the case of periodic systems, a natural basis set is provided by simple plane waves (spatial sine). Here again, the quality of the basis set is controlled by the number of functions: how small their period is allowed to be, or how high their frequency. Traditionally, this maximum frequency is converted to energy, and this maximum energy is referred to as the cutoff. As in the case of the k-point mesh, the energy cutoff must be optimised by a preliminary convergence study. A.2.8.1 Magnetic moment consideration The magnetic moment is modelled using the inbuilt functionality of VASP. Ferromagnetism can be described by a simple Stoner model (Stoner, 1938) to relate the energies of the spin up and spin down electron bands to the strength of the exchange correlation. Is n↑ N Is n↓ E↓ (k) = E(k) − N E↑ (k) = E(k) − 216 (A.11) (A.12) where Is is the Stoner parameter that describes the energy reduction due to electronic spin correlation and n↑ and n↓ are the density of up and down spins. The spin excess, R, can be found from R= n↑ − n↓ N (A.13) The energy can then be expressed as E↑,↓ (k) = Ẽ(k) − Is R 2 (A.14) where Ẽ = E(k) − Is (n↑ + n↓ ) . 2N (A.15) The spin excess, R, can then be found from Fermi statistics R= 1 X f↑ − f↓ (k) N (A.16) k f↑,↓ = [exp(Ẽ(k) ± Is R EF −1 − )] . 2 kT (A.17) From a Taylor expansion: 1 X δ3 f (k) 1 X δ f (k) (Is R) − (I R)3 + ... 3 s N 24N δ Ẽ(k) δ Ẽ(k) k k Z X δf ! V −V = δk(−δ(Ẽ − EF )) = DEF 3 δE 2 (2π) N R= (A.18) (A.19) k where D is the density of states at the Fermi level. The Stoner condition for ferromagentism is dependent on the densty of states per atom per spin (D(EF )) and the Stoner parameter. If 217 D̃(EF ) = V D(EF ) 2N (A.20) then R = D̃(EF )Is R − O(3) (A.21) where O describes the third order terms in the analysis. −O(3) = R(1 − D̃(EF )Is ) (A.22) for the spin excess, R, to be positve 1 − D̃(EF )Is must be negative and thus for a material to be ferromagnetic: D(EF )Is > 1. (A.23) This is valid for Fe, Co and Ni. This band theory approach enables the magnetisation owing to the mobile d band electrons to be described. 218 Appendix B The Finite Element Method B.1 Introduction In this study the finite element method is employed to evaluate the mechanical response of the structures that are derived from the crystalline structures. The finite element method solves large geometric problems, often involving non-linear partial differential equations which can be reduced to a linear algebraic problem. The finite element method can be used to solve a variety of field problems (structural, thermal, magnetic) but this work concentrates solely on structural analysis problems. B.2 Methodology B.2.1 Outline The methodology of the finite element method may be described as three discrete stages; pre-processing, analysis and post-processing. A general outline of each of these given below, and a more specific description of the techniques employed in this thesis is given in chapter 6. Many commercial codes are available for conducting finite element analysis, and it is not uncommon for users to write their own code tailored to specific applications. The work conducted in this thesis uses the commercial finite element code Abaqus and thus this introduction is tailored to the nuances of the particular package. 219 B.2.2 Pre-processing The pre-processing stage compiles an input file based on information specified by the user. The requires specification of the geometry, materials, loading and boundary conditions and analysis type. B.2.2.1 Geometry Accurate description of the geometry under consideration is vital for a correction solution to the finite element analysis problem. Geometry may be specified using either the graphical user interface (GUI), or by calculating the vertices and connecting units within the structure and using the Python scripting interface to assign the geometry to the model. The geometry is then used to specify the mesh, the array of nodal points and connecting elements that will be analysed. The geometric problem is discretized into a number of finite elements (the mesh), each bounded by nodes (meshless techniques are also available for specific analyses). For models containing multiple interacting parts, contact elements can be used to describe the interface between surfaces. The elements that make up the mesh may be: one dimensional where they describe a beam, truss or wire; two dimensional where they describe a plate; or three dimensional where they describe a continuum. Each element is bounded by nodes at the vertices (or ends for 1d elements) and in higher order elements, may have central or mid side nodes. Formulation errors are caused by incorrectly modelling the problem, and thus care should be taken to to ensure that appropriate elements and constraints are used (e.g. slender beam assumptions taken into account, and correct elements selected accordingly). Discretization errors are the main source of error, where the number of elements is insufficient to describe the geometry under consideration. To minimise this error a mesh convergence is performed where solution is computed for a range of mesh densities to verify that the solution converges. Further validation can be achieved by comparing the solution to experimental or analytical data. 220 B.2.2.2 Material properties The material property specification required is dependent on the analysis type, for a simple linear elastic analysis of isotropic homogeneous materials only the Young’s modulus and Poisson’s ratio need be specified. For orthotropic or anisotropic materials a frame of reference must be assigned to define material orientation relative to the model, and materials properties may be specified either as engineering constants (Exx , Gxy , νxy ), or as an elastic stiffness matrix. Non-linear, for example elasto-plastic material properties can be entered, and if necessary, user subroutines can be written to model fracture, or other time dependent plastic effects. B.2.2.3 Analysis type, boundary conditions and loading The type of analysis used is entirely dependent on the physics of the situation being considered. For structural simulations where no time, thermal or other external effects are taken into account either a linear elastic, for small strains, or geometrically non-linear analysis can be used. Each node within the model has both rotational and translational degrees of freedom. Constraints or boundary conditions can be placed on the model to replicate the physical loading conditions by prescribing specific displacements or rotations to specified nodes. Application of boundary conditions also has the effect of reducing the degrees of freedom, thus reducing the problem size. B.2.3 Solution Once the geometry, material properties, boundary condition and analysis type has been defined, this information is used by— the solver to compute the solution. The solution of the finite element problem is found by relating the stress σab , strain εab and constitutive (material properties) matrices dab . As an example, a simple two dimensional example of a body in plain stress is thus described by σ = Eε : 221 σx σy σxy d 11 d12 0 = d21 d22 0 0 0 d33 εx εy 2εxy (B.1) For the structure to be in equilibrium, the body forces, fa , must equal the stresses (σab , τab ) in the structure: ∂σx ∂σ y + + fx = 0 ∂x ∂y ∂τxy ∂σ y + + fy = 0 ∂x ∂y (B.2) (B.3) The matrix of nodal strain displacements (B), and nodal displacement vectors (d) is related to the global strains by σ = Bd. For each element, the strain energy is: Z 1 (σx εx + σ y ε y + τxy γxy dV U= σ εdV = 2 v v Z Z 1 1 (Eε)T εdV = = εT EεdV 2 v 2 v Z 1 1 T = d Bt BEdVd = dT kd 2 2 v Z T (B.4) (B.5) (B.6) By minimising the energy function it is then possible to compute the nodal displacements and potential energy for the whole model. B.2.4 Post-processing From the displacement/energy solution, the forces and stresses within the model can be derived. Stresses are computed at integration points within each element. It is also possible to output the reaction forces at each node and from these compute the global elastic properties of the structure. Data maybe visualized as either contour plots showing loads, stresses and the deformed structure, or as numerical data for further post processing and interpretation. 222 Appendix C Derivation of off-axis elastic constants from on-axis deformation using the finite element method When calculating the elastic properties of a continuum, given the full elastic constant tensor for a given structure in a specific direction it is possible to calculate the elastic tensor for the structure in any orientation. This methodology (described in section 2.2) is well proven for continuum structures and thus it was assumed that this could be extended to the beam networks modelled in this study. The cubic based beam networks were constructed in two orientations; one such that it could be loaded in the [110] orientation, and one such that it could be loaded in the [100] orientation. Both structures were analysed using the finite element method and the directly measured Poisson’s ratio, and elastic constants, for an axial stretch computed. From strains in the (110) direction, the elastic constants are derived. These constants are then rotated to derive the elastic constants in the (110) (or any other) direction. 223 ν110 ν110 rotated Type I Type II Type III 1 1 1 0.07 0.11 10000 1 1 0.01 0.05 1 10000 1 0.98 0.99 10000 10000 1 0.29 0.35 1 1 10000 -0.92 -0.92 10000 1 10000 -0.24 -0.21 1 10000 10000 0.00 0.17 10000 10000 10000 0.07 0.11 f ea Table C.1: Comparison of Poisson’s ratio calculated directly from strains in the (110) direction, and from rotating the elastic constants derived from strains in the (100) direction Conducting these analyses for all possible configurations yields a large volume of data. Table C.1 shows the Poisson’s ratio computed using both the direct measurement and rotated constants methodologies for comparison. These data show that for structures where type III beams are dominant the agreement between the rotated Poisson’s ratio and measured result is good. For structures where type I is dominant, there is less agreement. To determine the underlying reason for this poor agreement, it necessary to consider the deformation of type I structures in the (100) direction as shown in section 6.3. It can be seen that an axial applied strain results in a lateral strain of zero due to the beam stretching mechanism and the lack of cross-linking in the structure. This gives a Poisson’s ratio of zero, and a very high value of s12 . When these constant values are used to predict rotated elastic properties, these values lead to skewed results. In contrast to this, the deformation of the type III structure is primarily bending and there is a high degree of cross linking in the structure. For structures where there are type III beams present there is sufficient cross linking in the structure such that deformations in both the (100) and (110) gives high axial deformations and thus the constants can be rotated. 224 Figure C.1: Poisson’s ratio in the (110, 11̄0) directions - calculated using (top) ElAM and (bottom) direct measurement from FE This analysis shows that when predicting elastic properties, applying continuum methods to anisotropic beam networks is dependent on the nature of the beam networks deformation mode. Networks where the deformation is accompanied by very small orthogonal strains give erroneous values of rotated constants thus extreme caution should be applied when using this methodology to predict off-axis material properties. 225 Bibliography Alderson, A., K. Alderson, D. Attard, K. Evans, R. Gatt, J. Grima, W. Miller, N. Ravirala, C. Smith, and K. Zied (2010). Elastic constants of 3-, 4-and 6-connected chiral and anti-chiral honeycombs subject to uniaxial in-plane loading. Composites Science and Technology 70(7), 1042–1048. Alderson, A., K. L. Alderson, K. E. Evans, J. N. Grima, M. R. Williams, and P. J. Davies (2005). Modelling the deformation mechanisms, structure-property relationships and applications of auxetic nanomaterials. Physica Status Solidi B-Basic Solid State Physics 242(3), 499–508. Alderson, A. and K. E. Evans (1995). Microstructural modelling of auxetic microporous polymers. Journal of Materials Science 30, 3319–3332. Alderson, A. and K. E. Evans (2001, November). Rotation and dilation deformation mechanisms for auxetic behaviour in the alpha-cristobalite tetrahedral framework structure. Physics and Chemistry of Minerals 28(10), 711–718. Alderson, A. and K. E. Evans (2009). Deformation mechanisms leading to auxetic behaviour in the alpha-cristobalite and alpha-quartz structures of both silica and germania. Journal of Physics: Condensed Matter 21(2), 025401. Alderson, K. L. and K. E. Evans (1993, January). Strain-dependent behaviour of microporous polyethylene with a negative Poisson’s ratio. Journal of Materials Science 28(15), 4092–4098. Alderson, K. L., V. R. Simkins, V. L. Coenen, P. J. Davies, A. Alderson, and K. E. Evans 226 (2005). How to make auxetic fibre reinforced composites. physica status solidi (b) 242(3), 509–518. Almgren, R. F. (1985). An isotropic three-dimensional structure with Poisson’s ratio =-1. Journal of Elasticity 15, 427–430. Ashby, M.F & Jones, D. (2011). Engineering Materials 1: An Introduction to Properties, Applications and Design. Elsevier Science. Ashby, M. and L. J. Gibson (1997). Cellular Solids: Structure and Properties (Second ed.). Cambridge Univeristy Press. Ashcroft and Mermin (1976). Solid State Physics. Saunders College. Avinc, A. and V. I. Dimitrov (1999, January). Effective lennard-jones potential for cubic metals in the frame of embedded atom model. Computational Materials Science 13(4), 211–217. Barron, T. H. K. and C. Domb (1955, February). On the cubic and hexagonal close-packed lattices. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences (1934-1990) 227(1171), 447–465. Bathurst, R. J. and L. Rothenburg (1988). Note on a random isotropic granular material with negative Poisson’s ratio. International Journal Of Engineering Science 26(4), 373–383. Baughman, R. H., J. M. Shacklette, A. A. Zakhidov, and S. Stafstrom (1998, March). Negative Poisson’s ratios as a common feature of cubic metals. Nature 392(6674), 362– 365. Baughman, R. H., S. Stafstrom, C. X. Cui, and S. O. Dantas (1998, March). Materials with negative compressibilities in one or more dimensions. Science 279(5356), 1522–1524. Bezazi, A., C. Remillat, P. Innocenti, and F. Scarpa (2008). In-plane mechanical and thermal conductivity properties of a rectangular-hexagonal honeycomb structure. Composite Structures 84(3), 248 – 255. Bianchi, M., F. Scarpa, M. Banse, and C. Smith (2011). Novel generation of auxetic open cell foams for curved and arbitrary shapes. Acta Materialia 59(2), 686 – 691. 227 Bloch, F. (1929). Uber die quantenmechanik der elektronen in kristallgittern. Z. Physik 52, 555–600. Blöchl, P. E. (1994, Dec). Projector augmented-wave method. Phys. Rev. B 50(24), 17953– 17979. Bobev, S., D. J. Williams, J. D. Thompson, and J. L. Sarrao (2004). Thermal expansion in YbGaGe. Solid State Communications 131(7), 431 – 433. Brandel, B. and R. S. Lakes (2001, December). Negative Poisson’s ratio polyethylene foams. Journal of Materials Science 36(24), 5885–5893. Brańka, A., D. Heyes, and K. Wojciechowski (2009). Auxeticity of cubic materials. physica status solidi (b) 246(9), 2063–2071. Broughton, J. Q., G. H. Gilmer, and K. A. Jackson (1982, Nov). Crystallization rates of a lennard-jones liquid. Phys. Rev. Lett. 49(20), 1496–1500. Burkel, E. (2001, August). Introduction to x-ray scattering. Journal of Physics Condensed Matter 13, 7477–7498. Caddock, B. D. and K. E. Evans (1989). Microporous materials with negative Poisson’s ratios .1. microstructure and mechanical-properties. Journal Of Physics D-Applied Physics 22(12), 1877–1882. Caddock, B. D. and K. E. Evans (1995). Negative Poisson’s ratios and strain-dependent mechanical-properties in arterial prostheses. Biomaterials 16(14), 1109–1115. Calleja, M., A. L. Goodwin, and M. T. Dove (2008, June). Origin of the colossal positive and negative thermal expansion in Ag3 [Co(CN)6 ]: an ab initio density functional theory study. Journal of Physics Condensed Matter 20(25), 255226. Callister, W. D. (2003). Materials Science and Enginnering an Introduction. Wiley and Sons, Inc. Cannini, N. (2008). The tight-binding model for electronic band structure. 228 Ceriotti, M., F. Pietrucci, and M. Bernasconi (2006). Ab initio study of the vibrational properties of crystalline TeO2 . Physical Review B (Condensed Matter and Materials Physics) 73(10), 104304. Chen, K., L. Zhao, and J. Tse (2003). Ab initio study of elastic properties of ir and ir3x compounds. Journal of applied physics 93(5), 2414–2417. Christensen, N. E., D. J. Boers, J. L. van Velsen, and D. L. Novikov (2000). Ab initio thermodynamics of body-centred cubic and face-centred cubic cs. Journal Of PhysicsCondensed Matter 12(14), 3293–3312. Cleri, F. and V. Rosato (1993). Tight-binding potentials for transition-metals and alloys. Physical Review B 48(1), 22–33. Couves, J. W., R. H. Jones, S. C. Parker, P. Tschaufeser, and C. R. A. Catlow (1993). Experimental-verification of a predicted negative thermal expansivity of crystalline zeolites. Journal Of Physics-Condensed Matter 5(27), L329–L332. Dai, X. D., Y. Kong, J. H. Li, and B. X. Liu (2006, May). Extended Finnis Sinclair potential for bcc and fcc metals and alloys. Journal of Physics Condensed Matter 18, 4527–4542. Dassault (2008). Abaqus 6.8. Daw, M. S. and M. I. Baskes (1983, Apr). Semiempirical, quantum mechanical calculation of hydrogen embrittlement in metals. Phys. Rev. Lett. 50(17), 1285–1288. Daw, M. S. and M. I. Baskes (1984, Jun). Embedded-atom method: Derivation and application to impurities, surfaces, and other defects in metals. Phys. Rev. B 29(12), 6443–6453. Dmitriev, S. V. (2007). Auxetic behaviour of crystals from rotational defrees of freedom. Ferroelectrics 349, 33–44. Donescu, S., V. Chiroiu, and L. Munteanu (2008). On the Young’s modulus of a auxetic composite structure. Mechanics Research Communications In Press, Uncorrected Proof, –. Dove, M. T. (1997, March). Theory of displacive phase transitions in minerals. American Mineralogist 82(3-4), 213–244. 229 Dove, M. T. (2005). Structure and Dynamics An Atomic View of Materials. Oxford University Press. Dubbeldam, D., K. S. Walton, D. E. Ellis, and R. Q. Snurr (2007). Exceptional negative thermal expansion in isoreticular metal-organic frameworks. Angewandte ChemieInternational Edition 46(24), 4496–4499. Dudek, M. R., B. Grabiec, and K. W. Wojciechowski (2007). Molecular dynamics simulations of auxetic ferrogel. Reviews On Advanced Materials Science 14(2), 167–173. Evans, J. S. O., T. A. Mary, and A. W. Sleight (1997). Negative thermal expansion materials. Physica B 241, 311–316. Evans, J. S. O., T. A. Mary, and A. W. Sleight (1998). Negative thermal expansion in sc-2(wo4)(3). Journal Of Solid State Chemistry 137(1), 148–160. Evans, J. S. O., T. A. Mary, T. Vogt, M. A. Subramanian, and A. W. Sleight (1996). Negative thermal expansion in zrw2o8 and hfw2o8. Chemistry Of Materials 8(12), 2809–2823. Evans, K., M. Nkansah, and I. Hutchinson (1994). Auxetic foams: modelling negative Poisson’s ratios. Acta metallurgica et materialia 42(4), 1289–1294. Evans, K. E. (1989). Tensile network microstructures exhibiting negative Poisson’s ratios. J. Phys. D: Appl. Phys. 22, 1870–1876. Evans, K. E. and A. Alderson (2000). Auxetic materials: Functional materials and structures from lateral thinking! Advanced Materials 12(9), 617–628. Evans, K. E. and B. D. Caddock (1989). Microporous materials with negative poisson ratios .2. mechanisms and interpretation. Journal Of Physics D-Applied Physics 22(12), 1883–1887. Evans, K. E., M. A. Nkansah, I. J. Hutchinson, and S. C. Rogers (1991). Molecular network design. Nature 353(6340), 124–124. Feynman, R. (1939, August). Forces in molecules. Physical Review 56, 340–343. Finnis, M. (2003). Interatomic Forces in Condensed Matter. Oxford. 230 Finnis, M. W. and J. E. Sinclair (1984). A simple empirical n-body potential for transitionmetals. Philosophical Magazine A-Physics Of Condensed Matter Structure Defects And Mechanical Properties 50(1), 45–55. Fletcher, D., R. McMeeking, and D. Parkin (1996). The united kingdom chemical database service. Journal of chemical information and computer sciences 36(4), 746–749. Fluck, E. (1988). New notation in the periodic table. Pure & Appl. Chem. 60(3), 431–436. Foiles, S. M., M. I. Baskes, and M. S. Daw (1986, Jun). Embedded-atom-method functions for the fcc metals cu, ag, au, ni, pd, pt, and their alloys. Phys. Rev. B 33(12), 7983–7991. Friis, E. A., R. S. Lakes, and J. B. Park (1988). Negative Poisson’s ratio polymeric and metallic foams. Journal of Materials Science 23, 4406–4414. Gale, J. (2007). General Utility Latice Program, Users Manual (3.1 ed.). Curtin University of Technology. Gambhir, M., M. T. Dove, and V. Heine (1999). Rigid unit modes and dynamic disorder: SiO2 cristobalite and quartz. Physics and Chemistry of Minerals 26, 484–495. Gaspar, N., C. Smith, and K. Evans (2008). Auxetic behaviour and anisotropic heterogeneity. Acta Materialia In Press, Corrected Proof, –. Gatt, R. and J. Grima (2008). Negative compressibility. Physica Status Solidi (RLL) 5, 236–238. Gibson, L. J., M. F. Ashby, G. S. Schajer, and C. I. Robertson (1982, July). The mechanics of two-dimensional cellular materials. Proceedings of the Royal Society of London. Series A, Mathematical and Physical Sciences (1934-1990) 382(1782), 25–42. Giddy, A. P., M. T. Dove, G. S. Pawley, and V. Heine (1993, September). The determination of rigid-unit modes as potential soft modes for displacive phase-transitions in framework crystal-structures. Acta Crystallographica Section A 49, 697–703. Girifalco, L. A. and V. G. Weizer (1959). Application of the morse potential function to cubic metals. Physical Review 114(3), 687–690. 231 Goodwin, A., M. Calleja, M. Conterio, M. Dove, J. Evans, D. Keen, L. Peters, and M. Tucker (2008, February). Colossal positive and negative thermal expansion in the framework material ag3 [co(cn)6 ]. Science 319, 794–. Goodwin, A. and C. Kepert (2005). Negative thermal expansion and low-frequency modes in cyanide-bridged framework materials. Physical Review B 71. Grima, J., D. Attard, and R. Gatt (2008). Truss-type systems exhibiting negative compressibility. Physica Status Solidi (b) 245, 2405–2414. Grima, J., P. Farrugia, C. Caruana, R. Gatt, and D. Attard (2008, September). Auxetic behaviour from stretching connected squares. Journal of Materials Science 43(17), 5962– 5971. Grima, J. N., A. Alderson, and K. E. Evans (2005a). An alternative explanation for the negative Poisson’s ratios in auxetic foams. Journal Of The Physical Society Of Japan 74(4), 1341–1342. Grima, J. N., A. Alderson, and K. E. Evans (2005b). auxetic behaviour from rotating rigid units. Physica Status Solidi (b) 242, 561–575. Grima, J. N. and K. Evans (2000). Self expanding molecular networks. Chemical Communications 16, 1531–1532. Grima, J. N., R. Gatt, A. Alderson, and K. E. Evans (2005). On the origin of auxetic behaviour in the silicate alpha-cristobalite. Journal Of Materials Chemistry 15(37), 4003– 4005. Grima, J. N., R. Gatt, A. Alderson, and K. E. Evans (2006). An alternative explanation for the negative Poisson’s ratios in alpha-cristobalite. Materials Science And Engineering A-Structural Materials Properties Microstructure And Processing 423(1-2), 219–224. Grima, J. N., R. Gatt, T. G. C. Bray, A. Alderson, and K. E. Evans (2005). Empirical modelling using dummy atoms (emuda): an alternative approach for studying "auxetic" structures. Molecular Simulation 31(13), 915–924. 232 Grima, J. N., R. Gatt, V. Zammit, J. J. Williams, K. E. Evans, A. Alderson, and R. I. Walton (2007). Natrolite: A zeolite with negative Poisson’s ratios. Journal Of Applied Physics 101(8). Grima, J. N., J. J. Williams, R. Gatt, and K. E. Evans (2005). Modelling of auxetic networked polymers built from calix[4] arene building blocks. Molecular Simulation 31(13), 907–913. Grima, J.N. Jackson, R., A. Alderson, and K. Evans (2000). Do zeolites have negative Poisson’s ratio. Advanced Materials 12, 1912–1918. Guo, C. Y. and L. Wheeler (2006). Extreme Poisson’s ratios and related elastic crystal properties. Journal Of The Mechanics And Physics Of Solids 54(4), 690–707. Hammerschmidt, T., P. Kratzer, and M. Scheffler (2007). Elastic response of cubic crystals to biaxial strain: Analytic results and comparison to density functional theory for inas. Physical Review B 75(23), 235328. Hayes, M. and A. Shuvalov (1998). On the extreme values of Young’s modulus, the shear modulus and Poisson’s ratio for cubic materials. Transactions of the ASME 65, 786. Hayward, D. (2002). Quantum Mechanics for Chemists. RSC. He, C., P. Liu, P. J. McMullan, and G. A. C. (2005). Toward molecular auxetics: Main chain liquid crystalline polymers consisting of laterally attached para-quaterphenyls. Physica Status Solidi (b) 242, 576–585. Hearmon, R. F. S. (1957, Feb). Equations for transforming elastic and piezoelectric constants of crystals. Acta Crystallographica 10(2), 121–124. Heine, V., P. R. L. Welche, and M. T. Dove (1999). Geometrical origin and theory of negative thermal expansion in framework structures. Journal Of The American Ceramic Society 82(7), 1793–1802. Hellmann, V. H. (1941). Einfuhrung in die quantenchemie. Angewandte Chemie 54(11-12), 156–156. Hexcel (2012). Hexweb honeycomb attributes and properties. 233 Hishinuma, Y., E. Yang, B. Levine, and E. Bloemhof (2005, January). Piezoelectric unimorph mems deformable mirror for ultra-large telescopes. 5717, 21–29. Hohenberg, P., W. Kohn, et al. (1964). Inhomogeneous electron gas. Physical Review 136(3B), B864–B871. Holian, B. L., A. F. Voter, N. J. Wagner, R. J. Ravelo, S. P. Chen, W. G. Hoover, C. G. Hoover, J. E. Hammerberg, and T. D. Dontje (1991, Mar). Effects of pairwise versus many-body forces on high-stress plastic deformation. Phys. Rev. A 43(6), 2655–2661. Ishibashi, Y. and M. Iwata (2000, August). A Microscopic Model of a Negative Poisson’s Ratio in Some Crystals. Journal of the Physical Society of Japan 69, 2702–+. Janssen, Y., S. Chang, B. Cho, A. Llobet, K. Dennis, R. McCallum, R. McQueeney, and P. Canfield (2005). Ybgage: normal thermal expansion. Journal of Alloys and Compounds 389(1-2), 10 – 13. Kearley, G. J., M. R. Johnson, and J. Tomkinson (2006). Intermolecular interactions in solid benzene. The Journal of Chemical Physics 124(4), 044514. Keating, P. (1966a). Effect of invariance requirements on the elastic strain energy of crystals with application to the diamond structure. Physical Review 145(2), 637. Keating, P. N. (1966b, Sep). Theory of the third-order elastic constants of diamond-like crystals. Phys. Rev. 149(2), 674–678. Keskar, N. R. and J. R. Chelikowsky (1992, July). Negative poisson ratios in crystalline sio2 from 1st-principles calculations. Nature 358(6383), 222–224. Kimizuka, H., H. Kaburaki, and Y. Kogure (2000, Jun). Mechanism for negative poisson ratios over the alpha - beta transition of cristobalite, sio2 : A molecular-dynamics study. Phys. Rev. Lett. 84(24), 5548–5551. Koch, W., M. Holthausen, and M. Holthausen (2001). A chemist’s guide to density functional theory, Volume 293. Wiley Online Library. Koenders, M. A. Gaspar, N. (2008). The auxetic properties of a network of bending beams. PHYSICA STATUS SOLIDI B BASIC RESEARCH 245, 539–544. 234 Kohn, W., L. Sham, et al. (1965). Self-consistent equations including exchange and correlation effects. Physical Review 140(4A), A1133–A1138. Kresse, G. and J. Furthmüller (1996a). Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Computational Materials Science 6(1), 15 – 50. Kresse, G. and J. Furthmüller (1996b, Oct). Efficient iterative schemes for ab initio totalenergy calculations using a plane-wave basis set. Phys. Rev. B 54(16), 11169–11186. Kresse, G. and J. Hafner (1993, Nov). Ab initio molecular dynamics for open-shell transition metals. Phys. Rev. B 48(17), 13115–13118. Kresse, G. and J. Hafner (1994a, May). Ab initio molecular-dynamics simulation of the liquid-metal–amorphous-semiconductor transition in germanium. Phys. Rev. B 49(20), 14251–14269. Kresse, G. and J. Hafner (1994b). Norm-conserving and ultrasoft pseudopotentials for first-row and transition-elements. J. Phys.: Condens. Matter 6, 8245. Kresse, G. and D. Joubert (1999). From ultrasoft pseudopotentials to the projector augmented-wave method. Physical Review B 59(3), 1758. Lakes, R. (1987). Foam structures with a negative Poisson’s ratio. Science 235(4792), 1038–1040. Lakes, R. (1991, May). Deformation mechanisms in negative Poisson’s ratio materials: structural aspects. Journal of Materials Science 26(9), 2287–2292. Landolt-Bornstein (1985). Landolt-Bornstein Numerical data and functional relationships in science and technology. Springer-Verlag. Ledbetter, H. and A. Migliori (2006). A general elastic-anisotropy measure. Journal of applied physics 100, 063516. Lennard-Jones, J. E. (1931). Cohesion. Proceedings of the Physical Society 43(5), 461–482. 235 Lethbridge, Z., R. Walton, A. Marmier, C. Smith, and K. Evans (2010). Elastic anisotropy and extreme Poisson’s ratios in single crystals. Acta Materialia 58(19), 6444–6451. Lethbridge, Z. A., J. J. Williams, R. I. Walton, C. W. Smith, R. M. Hooper, and K. E. Evans (2006). Direct, static measurement of single-crystal Young’s moduli of the zeolite natrolite: Comparison with dynamic studies and simulations. Acta Materialia 54(9), 2533 – 2545. Li, H., M. Eddaoudi, M. O’Keeffe, and O. M. Yaghi (1999, November). Design and synthesis of an exceptionally stable and highly porous metal-organic framework. Nature 402(6759), 276–279. Li, Y. (1976). The anisotropic behavior of Poisson’s ratio, Young’s modulus, and shear modulus in hexagonal materials. Physica Status Solidi (a) 38(1), 171–175. Lieberman, D. S. and S. Zirinsky (1956, May). A simplified calculation for the elatic constants of arbitrarily oriented single crystals. Acta Crystallographica 9(5), 431–436. Lightfoot, P., D. A. Woodcock, M. J. Maple, L. A. Villaescusa, and P. A. Wright (2001). The widespread occurrence of negative thermal expansion in zeolites. Journal Of Materials Chemistry 11(1), 212–216. Lim, T. C. (2007, March). On simultaneous positive and negative Poisson’s ratio laminates. Physica Status Solidi B-Basic Solid State Physics 244(3), 910–918. Lorato, A., P. Innocenti, F. Scarpa, A. Alderson, K. Alderson, K. Zied, N. Ravirala, W. Miller, C. Smith, and K. Evans (2010). The transverse elastic properties of chiral honeycombs. Composites Science and Technology 70(7), 1057–1063. Louail, L., D. Maouche, A. Roumili, and F. A. Sahraoui (2004). Calculation of elastic constants of 4d transition metals. Materials Letters 58(24), 2975 – 2978. Love, A. (1892). A Treatise On the Mathematical Theory of Elasticity, Vol. I. Cambridge University Press. Lowe, A. and R. S. Lakes (2000). Negative Poisson’s ratio foam as seat cushion material. Cellular Polymers 19(3), 157–167. 236 Lubarda, V. A. and M. A. Meyers (1999). On the negative poisson ratio in monocrystalline zinc. Scripta Materialia 40(8), 975–977. Mariathasan, J. W. E., L. W. Finger, and R. M. Hazen (1985, Jun). High-pressure behavior of LaNbO4 . Acta Crystallographica Section B 41(3), 179–184. Marmier, A., Z. Lethbridge, R. Walton, C. Smith, S. Parker, and K. Evans (2010). ElAM: A computer program for the analysis and representation of anisotropic elastic properties. Computer Physics Communications. Mary, T. A., J. S. O. Evans, T. Vogt, and A. W. Sleight (1996). Negative thermal expansion from 0.3 to 1050 kelvin in zrw2o8. Science 272(5258), 90–92. Masters, I. G. and K. E. Evans (1996). Models for the elastic deformation of honeycombs. Composite Structures 35(4), 403 – 422. Mendham, J., R. C. Denney, B. J. D., and M. J. K. Thomas (2000). Vogel’s Textbook of Quantitative Chemical Analysis, 6th Ed. Prentice Hall. Miller, W., Z. Ren, C. Smith, and K. Evans (2012a). A negative Poisson’s ratio carbon fibre composite using a negative Poisson’s ratio yarn reinforcement. Composites Science and Technology 72(7), 761 – 766. Miller, W., Z. Ren, C. Smith, and K. Evans (2012b). A negative Poisson’s ratio carbon fibre composite using a negative Poisson’s ratio yarn reinforcement. Composites Science and Technology 72(7), 761 – 766. Miller, W., C. Smith, F. Scarpa, and K. Evans (2010). Flatwise buckling optimization of hexachiral and tetrachiral honeycombs. Composites Science and Technology 70(7), 1049– 1056. Milstein, F. and K. Huang (1979). Existence of a negative poisson ratio in fcc crystals. Physical Review B 19(4), 2030–2033. Milton, G. W. (1992). Composite materials with Poisson’s ratios close to – 1. Journal of the Mechanics and Physics of Solids 40(5), 1105 – 1137. 237 Morse, P. M. (1929, July). Diatomic molecules according to the wave mechanics. ii. vibrational levels. Physical Review 34(1), 57–64. Narojczyk, J. W. and K. W. Wojciechowski (2008). Elastic properties of the fcc crystals of soft spheres with size dispersion at zero temperature. Physica Status Solidi B-Basic Solid State Physics 245(3), 606–613. Neale, P. J., K. L. Alderson, A. P. Pickles, and K. E. Evans (1993, January). Negative Poisson’s ratio of microporous polyethylene in compression. Journal of Materials Science Letters 12(19), 1529–1532. Nye, J. (1985). Physical Properties of Crystals: Their Representation by Tensors and Matrices. Oxford University Press. Ogi, H., M. Fukunaga, M. Hirao, and H. Ledbetter (2004, January). Elastic constants, internal friction, and piezoelectric coefficient of alpha-teo2. Physical Review B 69(2), 024104–1. Ogi, H., T. Ohmori, N. Nakamura, and M. Hirao (2006). Elastic, anelastic, and piezoelectric coefficients of alpha-quartz determined by resonance ultrasound spectroscopy. Journal of Applied Physics 100(5), 053511. Oruganti, R. K., A. K. Ghosh, and J. Mazumder (2004). Thermal expansion behavior in fabricated cellular structures. Materials Science And Engineering A-Structural Materials Properties Microstructure And Processing 371(1-2), 24–34. Palmer, D. C., R. J. Hemley, and C. T. Prewitt (1994, December). Raman spectroscopic study of high-pressure phase transitions in cristobalite. Physics and Chemistry of Minerals 21(8), 481–488. Parr, R. (1983). Density functional theory. Annual Review of Physical Chemistry 34(1), 631–656. Peel, L. D. (2007, March). Exploration of high and negative Poisson’s ratio elastomermatrix laminates. Physica Status Solidi B-Basic Solid State Physics 244(3), 988–1003. 238 Perdew, J. P., K. Burke, and M. Ernzerhof (1996). Generalized gradient approximation made simple. Phys. Rev. Lett. 77, 3865. Perdew, J. P., K. Burke, and M. Ernzerhof (1997). Erratum: Generalized gradient approximation made simple. Phys. Rev. Lett. 78, 1396. Pour, N., E. Altus, H. Basch, and S. Hoz (2009). The origin of the auxetic effect in prismanes: Bowtie structure and the mechanical properties of biprismanes. The Journal of Physical Chemistry C 113(9), 3467–3470. Prall, D. and R. S. Lakes (1997). Properties of a chiral honeycomb with a Poisson’s ratio of -1. International Journal Of Mechanical Sciences 39(3), 305. Pryde, A. K. A., K. D. Hammonds, M. T. Dove, V. Heine, J. D. Gale, and M. C. Warren (1996). Origin of the negative thermal expansion in zrw2o8 and zrv2o7. J. Phys.: Condens. Matter 8, 10973–10982. Reisner, B. A., B. H. Toby, Y. Lee, J. B. Parise, J. Z. Larese, V. Kahlenberg, J. C. Hanson, A. Freitag, G. A. Jones, and D. R. Corbin (2000). Understanding negative thermal expansion and trap door cation relocations in zeolite rho. Chem. Commun., 2221 – 2222,. Roark, R. and W. Young (1982). Formulas for stress and strain. Pearson McGraw-Hill Book Company. Rovati, M. (2004). Directions of auxeticity for monoclinic crystals. Scripta Materialia 51(11), 1087–1091. Salavdor, J. R., F. Guo, T. Hogan, and M. Kanatzdis (2003). Zero thermal expansion in ybgage due to an electronic valence transition. Letters to Nature 425, 702–705. Sanchez-Valle, C., S. V. Sinogeikin, Z. A. D. Lethbridge, R. I. Walton, C. W. Smith, K. E. Evans, and J. D. Bass (2005). Brillouin scattering study on the single-crystal elastic properties of natrolite and analcime zeolites. Journal Of Applied Physics 98(5). Scarpa, F., C. Smith, M. Ruzzene, and M. Wadee (2008). Mechanical properties of auxetic tubular truss-like structures. physica status solidi (b) 245(3), 584–590. 239 Sholl, D. and J. Steckel (2009). Density functional theory. Wiley Online Library. Sin’ko, G. and N. Smirnov (2002). Ab initio calculations of elastic constants and thermodynamic properties of bcc, fcc, and hcp al crystals under pressure. Journal of Physics: Condensed Matter 14, 6989. Skelton, E. F., J. L. Feldman, C. Y. Liu, and I. L. Spain (1976, Mar). Study of the pressureinduced phase transition in paratellurite (teo2). Phys. Rev. B 13(6), 2605–2613. Sleight, A. W. (1998a). Compounds that contract on heating. Inorganic Chemistry 37(12), 2854–2860. Sleight, A. W. (1998b). Negative thermal expansion materials. Current Opinion in Solid State & Materials science 3, 128–131. Smith, C. W., J. N. Grima, and K. E. Evans (2000). A novel mechanism for generating auxetic behaviour in reticulated foams: missing rib foam model. Acta Materialia 48(17), 4349 – 4356. Spadoni, A., M. Ruzzene, and F. Scarpa (2005). Global and local linear buckling behavior of a chiral cellular structure. Physica Status Solidi-B-Basic Research 242(3), 695–709. Stoner, E. (1938, April). Collective electron ferromagnetism. Royal Society of London Proceedings Series A 165, 372–414. Stott, M. J. and E. Zaremba (1980, Aug). Quasiatoms: An approach to atoms in nonuniform electronic systems. Phys. Rev. B 22(4), 1564–1583. Sutton, A. P. and J. Chen (1990). Long-range finnis-sinclair potentials. Philosophical Magazine Letters 61(3), 139–146. Tan, J. C. and A. K. Cheetham (2011). Mechanical properties of hybrid inorganic- organic framework materials: establishing fundamental structure-property relationships. Chem. Soc. Rev 40, 1059–1080. Tao, J. Z. and A. W. Sleight (2003). The role of rigid unit modes in negative thermal expansion. Journal Of Solid State Chemistry 173(2), 442–448. 240 Tee, K., A. Spadoni, F. Scarpa, and M. Ruzzene (2010). Wave propagation in auxetic tetrachiral honeycombs. Journal of Vibration and Acoustics 132, 031007. Theocaris, P. S., G. E. Stavroulakis, and P. D. Panagiotopoulos (1997). Negative Poisson’s ratios in composites with star-shaped inclusions: a numerical homogenization approach. Archive of Applied Mechanics (Ingenieur Archiv) 67, 274 – 286. Thomas, T. Y. (1966). On the stress strain rleations for cubic crystals. Proc. N. A. S. 55, 235–239. Ting, T. C. T. (2004, November). Very large Poisson’s ratio with a bounded transverse strain in anisotropic elastic materials. Journal of Elasticity 77(2), 163–176. Ting, T. C. T. and D. M. Barnett (2005). Negative Poisson’s ratios in anisotropic linear elastic media. Journal Of Applied Mechanics-Transactions Of The Asme 72(6), 929–931. Ting, T. C. T. and T. Chen (2005). Poisson’s ratio for anisotropic elastic materials can have no bounds. Q J Mechanics Appl Math 58(1), 73–82. Tokmakova, S. (2005). Stereograohic projections of Poisson’s ratio in auxetic crystals. Physica Status Solidi (b) 242, 721–729. Tretiakov, K. V. and K. W. Wojciechowski (2005). Monte carlo simulation of two- dimensional hard body systems with extreme values of the Poisson’s ratio. Physica Status Solidi B-Basic Solid State Physics 242(3), 730–741. Tschaufeser, P. and S. C. Parker (1995). Thermal expansion behavior of zeolites and AlPO4 s. The Journal of Physical Chemistry 99(26), 10609–10615. Turley, J. and G. Sines (1971a). Anisotropic behaviour of compliance and stiffness coefficients for cubic materials. Journal Of Physics D-Applied Physics 4(11), 1731–&. Turley, J. and G. Sines (1971b). Anisotropy of Young’s modulus, shear modulus and Poisson’s ratio in cubic materials. Journal Of Physics D-Applied Physics 4(2), 264–&. Vukcevich, M. R. (1972). The elastic constants of cubic crystals with covalent and partially covalent bonds. Physica Status Solidi (b) 50(2), 545–552. 241 Wan, H., H. Ohtaki, S. Kotosaka, and G. Hu (2004). A study of negative Poisson’s ratios in auxetic honeycombs based on a large deflection model. European Journal of Mechanics - A/Solids 23(1), 95 – 106. Weaire, D. and S. Hutzler (2005). The Physics of Foams. Clarendon Press, Oxford. Wei, G. and S. F. Edwards (1998). Auxeticity windows for composites. Physica A: Statistical Mechanics and its Applications 258(1-2), 5 – 10. Whitty, J., A. Alderson, P. Myler, and B. Kandola (2003). Towards the design of sandwich panel composites with enhanced mechanical and thermal properties by variation of the in-plane Poisson’s ratios. Composites Part A: Applied Science and Manufacturing 34(6), 525–534. Williams, J. J., C. W. Smith, K. E. Evans, Z. A. D. Lethbridge, and R. I. Walton (2007). An analytical model for producing negative Poisson’s ratios and its application in explaining off axis elastic properties of the nat-type zeolites. Acta Materialia 55, 5697– 5707. Wills, J., O. Eriksson, P. Söderlind, and A. Boring (1992). Trends of the elastic constants of cubic transition metals. Physical review letters 68(18), 2802–2805. Wojciechowski, K. W. (2004). Simple models of unusual elastic properties. Slow Dynamics In Complex Systems 708, 735–738. Woodcock, D. A., P. Lightfoot, P. A. Wright, L. A. Villaescusa, and M. A. Camblor (1999). Strong negative thermal expansion in the siliceous zeolites itq-1, itq-3 and ssz-23. J. Mater. Chem. 9, 349–351. Wright, J., M. Sloan, and K. Evans (2010). Tensile properties of helical auxetic structures: A numerical study. Journal of Applied Physics 108(4), 044905–044905. Wu, Y., A. Kobayashi, G. Halder, V. K. Peterson, K. W. Chapman, N. Lock, P. D. Southon, and C. J. Kepert (2008). Negative thermal expansion in the metal-organic framework material Cu(1,3,5-benzenetricarboxylate). Angewandte Chemie 120(46), 9061–9064. Yaghi, O. M. (2012, May). 242 Yeganeh-Haeri, A., D. J. Weidner, and J. B. Parise (1992). Elasticity of alpha-cristobalite: A silicon dioxide with a negative Poisson’s ratio. Science 257(5070), 650–652. Zhang, J. and M. Ashby (1992). The out-of-plane properties of honeycombs. International Journal of Mechanical Sciences 34(6), 475 – 489. Zhou, W., H. Wu, T. Yildirim, J. R. Simpson, and A. R. H. Walker (2008). Origin of the exceptional negative thermal expansion in metal-organic framework-5 zn4 o(1,4benzenedicarboxylate)3. Physical Review B (Condensed Matter and Materials Physics) 78(5), 054114. 243