Large Strain Time-Dependent Behavior of Elastomeric Materials by Jorgen S. Bergstrom Submitted to the Department of Mechanical Engineering in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Mechanical Engineering at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 1999 © Massachusetts Institute of Technology 1999. All rights reserved. week of Mechanical Engineering May 14, 1999 Mary C. Boyce Professor Thesis Supervisor Accepted byLarge Strain Time-Dependent Behavior of Elastomeric Materials by Jorgen 8. Bergstrm Submitted to the Department of Mechanical Engineering on May 14, 1999, in partial fulfillment of the requirements for the degree of Doctor of Phiiosophy in Mechanical Engineering Abstract ‘The mechanical behavior of elastomeric materials is known to be rate-dependent and to exhibit hysteresis upon cyclic loading. Although these features of the rubbery constitutive response are well-recognized and important to its function, few models attempt to quantify these aspects of response. Based on a detailed experimental in- vestigation a new constitutive model for the time-dependence of unfilled elastomers has been developed. The foundation of the model is that the mechanical behavior can be decomposed into two parts: an equilibrium network corresponding to the state that is approached in long time stress relaxation tests; and a second network capturing the non-linear rate-dependent deviation from the equilibrium state. The time-dependence of the second network is further assumed to be governed by the rep- tational motion of molecules having the ability to significantly change conformation and thereby relaxing the overall stress state. To model the behavior of particle filled elastomers the newly developed constitutive framework is then extended to include filler interactions by amplification of the first strain invariant. In an effort to examine some of the assumptions that are common in the constitutive modeling of particle filled elastomers, a detailed series of micromechanical models were constructed using two- and three-dimensional finite element simulations. The results indicate that the effect of filler particles can be accurately predicted using stochastic three-dimensional simulations suggesting that successful modeling mainly requires a rigorous treatment, of the composite nature of the microstructure and not molecular level concepts such as alteration of mobility or effective crosslinking density in the elastomeric phase of the material. A diréct comparison between the new model and experimental data for a numbiog of different elastomerg suggest that the new framework successfully captures the observed behavior. = ~SS.5.. Thesis Supervisor: Mary C. Boyce Title: ProfessorAcknowledgments A number of individuals and organizations have made invaluable contributions to the completion of the research represented by this thesis. I first wish to express sincere appreciation to my thesis supervisor, Prof. Mary Boyce, whose guidance, advice and encouragement were a constant source of inspiration. I also wish to thank my other committee members, Professors Ali Argon, David Parks, and Edwin ‘Thomas, for their helpful suggestions and insights through the course of this work. ‘The research was funded by the U.S. National Science Foundation grant no. DMM-9157899 and CMS-9622526. I like to thank Alan Dickey and Teresa Cush- ing of the Caterpillar Corporation for supplying the materials tested in this work. I would also like to sincerely thank Yuanming and Benita for taking the time and helping me with the TEM investigation. Many friends and classmates have added much to my experience at MIT. I would like to thank Oscar and Nick for making my first two years at MIT more interesting, and Gu for greatly reducing the burden of the system administration tasks. I also would like to thank Yioula for all our discussions and for being such a good friend. ‘Thanks also to Kostas for getting me back in shape, Susan for all our weekend ad- ventures, and Melissa for being the ultimate role model. I would also like to thank Clarence, Ganti, Hong, Patricia, Tom, Ethan, Harish and Sezyei. Finally, I am forever grateful to my parents and family for all the support that they have given me throughout my studies. I could not have done it without you.Contents 1 Introduction 2 Experimental Results 2.1 Experimental Procedure ....... ey 2.2 Chloroprene Rubber with 7 vol% Carbon Black ........... 2.3 Chloroprene Rubber with 15 vol% Carbon Black. . . 24 Chloroprene rubber with 25 vol% Carbon Black ....... 20.0. 2.5 Comparison Between the Different Chloroprene Rubbers . . . . . 26 Unfilled Natural Rubber ...........0.0.05 2.7 Natural Rubber with 7 vol% Carbon Black... .... 2.0.0... 28 Natural Rubber with 9 vol% Carbon Black... 0.202.200. 2.9 Natural Rubber with 17 vol% Carbon Black .............. 2.10 Natural Rubber with 24 vol% Carbon Black .............5 2.11 Comparison Between the Different Natural Rubbers . . 2.12 Nitrile Butadiene Rubber (NBR)... <2... 0.000000. 0 05 2.13 Acrylate Butadiene Rubber (ABR) .. . . see enews 214 Other Elastomers... 2. eee 2.15 Conclusions ©. 6... eee 3 Equilibrium Behavior of Unfilled Elastomers 3.1 Introduction. 2... eee 32 Modeling... 0.0.0.0... cece eee eee 3.3. Determination of Material Constants... 2... 20.00.00003.4 Comparison with Experimental Data ............. .. 4 Equilibrium Behavior of Filled Elastomers 101 4.1 Introduction. 2.0.00... . 101 4.2 Experimental Data . . 104 4.2.1 Initial Modulus 105 4.3. Theoretical Models . . 106 4.3.1 Large Strains and Strain Amplification . . 113 4.4 Finite Element Simulations of Particle-Filled Elastomers. . 123 4.4.1 Two-dimensional Simulations 124 4.4.2. Three-Dimensional Simulations 130 4.5 Conclusions . 148 ‘Time-Dependence of Unfilled Elastomers 150 6.1 Introduction. 6.0.6... cece eee ene eee nes 150 5.2 Experimental Data... ...-.......- cece ees WL 5.3 A Mechanism for Time-Dependence in Elastomers... ........ 154 5.3.1 Constitutive Modeling ©... 22... e eee 157 5.3:2~ Verification of the Clausius-Duhem inequality... .. 2... 174 5.3.3 One Approach to Improve the Unloading Predictions ..... 176 54 Determination of Material Constants»... 2.2 ........-05 180 5.5 Conclusions»... 2.2... eee 184 ‘Time-Dependence of Filled Elastomers 186 6.1 Introduction.......... eee oneness 186 62 Modeling... 2.0.2... eevee eee eee eee . 187 63 Conclusions .. 22.22.20... 2c eee eee eee eee 199 Molecular Simulations 201 TA Introduction... 0... eee eee sees 201 7.2. Molecular Simulation Techniques... 0.2... 000-00 0005 203 7.21 Monte Carlo Algorithm... ....... vee cee 2077.22 Molecular Dynamics Algorithm ............-00-- 208 7.23 Wnitial Configuration .. 2.0.2.2... eee eee 210 7.3 Results... 2... eee 211 TA Conclusions 22.2.0... eee eee eee beeen eee 238 Conclusions 242 81 Summary of Results... 2.2.0.0... 00.00.0200 22 82 Future Work... 2... ee eee 246 List of Symbols 255 Correlation between Two-Dimensional and Three-Dimensional Sim- ulations 257 Simple Shear of a Cylindrical Specimen 259List of Figures 24 22 23 24 25 2-6 27 28 29 Schematic showing the experimental setup: (a) uniaxial compression, (b) plane strain compression, (c) simple shear. . TEM micrograph of a chloroprene rubber filled with 7 vol% carbon black N60. 2. eee Mullins effect in a chloroprene rubber filled with 7 vol% carbon black rr Stress-strain behavior when loaded to different final strains in a chloro- prene rubber filled with 7 vol% carbon black N600. ........ Effect of different strain rates on a chloroprene rubber filled with 7 vol% carbon black N600. Stress relaxation test on a chloroprene rubber filled with 7 vol% carbon Black. eee eee Str relaxation test showing the dependence on relaxation time for a chloroprene rubber filled with 7 vol% carbon black. ....... 0. Oscillatory test with frequency f = 0.25 Hz showing stress as a function of applied strain of a chloroprene rubber filled with 7 vol% carbon black. Oscillatory test with frequency f = 0.25 Hz showing stress as a function of time of a chloroprene rubber filled with 7 vol% carbon black. 2-10 Oscillatory test with frequency f = 0.02 Hz showing stress as a function of applied strain of a chloroprene rubber filled with 7 vol% carbon black. 2-11 Oscillatory test with frequency f = 0.02 Hz showing stress as a function of time of a chloroprene rubber filled with 7 vol% carbon black. 29 30 31 32 33 33 34 35 36 362-12 Uniaxial and plane strain response of a chloroprene rubber filled with 7 vol% carbon black... eee eee 2-13 Stress-strain behavior in uniaxial compression at different tempera- tures of a chloroprene rubber filled with 7 vol% carbon black. 2-14 TEM micrograph showing the filler morphology of a chloroprene rubber with 15 vol% N600 carbon black... 2.22.2... eee eee 2-15 Stress-strain behavior during the first three load cycles for a chloro- prene rubber filled with 15 vol% carbon black. ..... 2.0.0... 2.16 Stress-strain behavior when loaded to different final s:rains of a chloro- prene rubber filled with 15 vol% carbon black. ........... 0. 2-17 Effect of different strain rates on the stress-strain response of a chloro- prene rubber filled with 15 vol% carbon black... 0.20.02... 2-18 Stress relaxation tests showing stress as a function of strain for a chloro- prene rubber filled with 15 vol% carbon black. ......-.2.-.- 2-19 Stress relaxation test showing stress as a function of time for a chloro- prene rubber filled with 15 vol% carbon black. ........-.. 0. 2-20 Stress-strain response in a stress relaxation test on a chloroprene rubber filled with 15 vol% carbon black. 2-21 Stress relaxation as a function of relaxation time in a chloroprene rub- ber filled with 15 vol% carbon black. . 2... 0.00.0... 2-22 TEM micrograph of a chloroprene rubber filled with 25 vol% carbon N60. 2-24 Stress-strain behavior when loaded to different final strains of a chloro- prene rubber filled with 25 vol% carbon black... 2.2... 0. - 2-25 Effect of different strain rates on a chloroprene rubber filled with 25 vol% carbon black. . . . 2.26 Stress relaxation tests showing the stress-strain cesponse of a chloro- prene rubber filled with 25 vol% carbon black...........-05 37 38 43 45 46 462-27 Uniaxial and plane strain response of a chloroprene rubber filled with 25 vol% carbon black. 2-28 Effect of different temperatures on the stress-strain response of a chloro- prene rubber filled with 25 vol% carbon black. 2-29 Comparison between the total stress-strain response of the three chloro- prene rubbers tested. 2-30 Comparison between the equilibrium loci for three different chloroprene rubbers. 2-31 Comparison between the time-dependent stress for three different chloro- prene rubbers. 2-32 Time-dependent stress for a chloroprene rubber filled with 7 vol% car- bon black tested at three different strain rates. 2-33 Time-dependent stress for a chloroprene rubber filled with 15 vol% carbon black tested at four different strain rates. 2-34 Effect of different strain rates on the stress-strain response for an un- filled natural rubber. 2-35 Large strain response of an unfilled natural rubber. .......... 2-36 Stress-strain behavior of an unfilled natural rubber in simple shear. 2-37 Mullins effect in a natural rubber filled with 7 vol% carbon black N351. 2-38 Effect of different final strains and strain rates on the stress-strain response of a natural rubber filled with 7 vol% carbon black... . . . 2.39 Stress relaxation behavior at different strain rates of a natural rubber filled with 7 vol% carbon black... 2.22000 2-40 Stress-time response in a stress relaxation experiment on a natural rubber filled with 7 vol% carbon black... 2.6... ee eee ee 2.41 Stress relaxation test showing the stress-strain response of a natural rubber filled with 7 vol% carbon black. 2-42 TEM micrograph of a natural rubber filled with 9 vol% carbon black NBL eee 2-43 Mullins effect in a natural rubber filled with 9 vol% carbon black N351. 47 48 49 50 51 51 52 53 54 54 55 36 56 87 57 58 592-44 Effect of different strain rates on a natural rubber filled with 9 vol% (carbon) DIAC: teen er a 60 2-45 Stress relaxation behavior of a natural rubber filled with 9 vol% carbon black. 60 2-46 TEM micrograph of a natural rubber filled with 17 vol% carbon black 351. Note rather uniform dispersion of particles throughout the rub- ber matrix, ©. eee eee eee 2 2-47 Effect of different strain rates and stress relaxation behavior on a nat- ural rubber filled with 17 vol% carbon black N351. .......... 63 2.48 Effect of different strain rates on the stress-strain response of a natural rubber filled with 17 vol% carbon black N351. . . 63 2.49 Stress relaxation behavior during a high strain rate compression test on a natural rubber filled with 17 vol% carbon black N351..... . « 64 2-50 Stress-time response in a high strain rate relaxation experiment on a natural rubber filled with 17 vol% carbon black N35L.... 2... « 64 2-51 Stress relaxation behavior during a low strain rate compression test. on a natural rubber filled with 17 vol% carbon black N351..... 2... 65 2-52 Stress-time response in a low strain rate relaxation experiment on a natural rubber filled with 17 vol% carbon black N35L.......... 65 2-53 Stress-strain response in simple shear of a natural rubber filled with 17 vol% carbon black N35... 6... eee eee eee 66 2.54 Effect of different strain rates on the stress-strain response of a natural rubber filled with 17 vol% carbon black N231...........- 66 2-55 Stress relaxation behavior of a natural rubber filled with 17 vol% car- bon black N231, applied strain rate é=—0.01/s............. 67 2.56 Stress relaxation behavior of a natural rubber filled with 17 vol% car- bon black N231, applied strain rate ¢ = o7 2-57 Stress relaxation behavior of a natural rubber filled with 17 vol% car- bon black N231, applied strain rate é = ODN Aenee . 68 2-58 Influence of different carbon black particles on the stress-strain response. 68 102-59 TEM micrograph of a natural rubber filled with 24 vol% carbon black N3BL. 0 ee 2-60 Effect of different strain rates on the stress-strain response of a natural rubber filled with 24 vol% carbon black N351. . 6.0... -.0--- 261 Mullins effect in simple shear of a natural rubber filled with 24 vol% carbon black N351. 6 ee 2-62 Simple shear behavior at different shear rates of a natural rubber filled with 24 vol% carbon black N351. 2-63 Comparison between the total stress-strain response of the different natural rubbers tested. 2-64 Comparison between the total stress-strain response of the different natural rubbers tested. 2-65 Comparison between the equilibrium loci for the four different natural rubbers tested. 2-66 Comparison between the time-dependent stress for the four different natural rubbers tested. . 2-67 Stress relaxation behavior of NBR. 2-68 Stress relaxation test showing the dependence on relaxation time of NBR. 2-69 Effect of different strain rates on the stress-strain response of ABR. 2-70 Results from an oscillatory loading experiment showing the stress- strain response, frequency f = 1 Hz. 2-71 Results from an oscillatory loading experiment showing the stress as a function of time, frequency f = 1 Hz 2-72 Results from an oscillatory loading experiment showing peak stress as a function of time, frequency f = 1 Hz. . 2-73 Results from an oscillatory loading experiment showing double stress amplitude as a function of time, frequency f = 1 Hz. 2-74 Results from an oscillatory loading experiment. showing the stress- strain response, frequency = 5 Hz.........----+45 ul 69 7 70 7 2 73 74 74 icy 76 7B 78 79 79 802-75 Results from an oscillatory loading experiment showing the stress as a function of time, frequency = 5 Hz... . 2-76 Results from an oscillatory loading experiment showing the peak stress, as a function of time, frequency = 5 Hz. 2-77 Results from an oscillatory loading experiment showing the double stress amplitude as a function of time, frequency = 5 Hz. 2-78 Effect. of different strain rates on the stress-strain behavior of EPDM. 2-79 Effect of different strain rates on the stress-strain behavior of FKM. . 2-80 Effect: of different strain rates on the stress-strain behavior of Viton. . 2-81 Effect of different strain rates on the stress-strain behavior of VMQ. . 3-1 Representative volume element used in the 8-chain model. 3-2 Langevin and inverse Langevin functions. . 3-3 Estimated equilibrium locus of a chloroprene rubber filled with 7 vol% carbon black, ¢¢ = —0.4, o¢ = —0.78 MPa giving y? = 0.75 MPa The limiting chain stretch is obtained from the stress-strain behavior at large strains, €maz = —2.24 giving N° = 2.5. . 3-4 Comparison between experimental data and eight-chain predicted be- havior of a chloroprene rubber filled with 7 vol% carbon black tested in uniaxial compression. 3-5 Comparison between experimental data and eight-chain predicted be- havior of a chloroprene rubber filled with 7 vol% carbon black tested in plane strain compression... . . . 3.6 Comparison between experimental data and eight-chain predicted be- havior of a chloroprene rbber filled with 15 vol% carbon black. . 3-7 Comparison between experimental data and eight-chain predicted be- havior of a chloroprene rubber filled with 25 vol% carbon black... . 3-8 Comparison between experimental data and eight-chain predicted be- havior of an unfilled natural rubber tested in uniaxial compression. 12 80 81 81 82 83 83 84 87 94 96 96 97 a73-9 Comparison between experimental data and eight-chain predicted be- havior of an unfilled natural rubber tested in simple shear... . . . « 3-10 Comparison between experimental data and cight-chain predicted be- havior of a natural rubber filled with 7 vol% carbon black... . . . « 3-11 Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 9 vol% carbon black. . .. . . . 3-12 Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 17 vol% carbon black... . . 3-13 Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 24 vol% carbon black tested in uniaxial compression. 3-14 Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 24 vol% carbon black tested in simple shear. 4-1 Stress-strain behavior of a chloroprene rubber filled with different levels of carbon black. . 4-2 Comparison between the equilibrium loci for three different chloroprene rubbers. 4-3 Compi n between the equilibrium loci for the four different natural rubbers tested. 4-4 Experimentally determined normalized Young's modulus E/Em as a function of volume fraction particles. . . . 4-5 Theoretical predictions of the initial Young’s modulus, E;/Em = 100. 4-6 Comparison between theoretical predictions and experimental data for a chloroprene rubber with 7 vol% carbon black. ...........- 4-7 Comparison between theoretical predictions and experimental data for achloroprene rubber with 15 vol% carbon black. - 4-8 Comparison between theoretical predictions and experimental data for a chloroprene rubber with 25 vol% carbon black. . . . . 13 98 98 99 99 100 100 104 105 106 107 110 116 uz4-9 Comparison between theoretical predictions and experimental data for three different chloroprene rubber filled with carbon black N600. 19 4-10 Comparison between theoretical predictions and experimental data for a natural rubber with 7 vol% carbon black... . . 120 4-11 Comparison between theoretical predictions and experimental data for a natural rubber with 17 vol% carbon black. . . . 120 4-12 Comparison between theoretical predictions and experimental data for a natural rubber with 24 vol% carbon black 121 4-13 Comparison between theoretical predictions and experimental data for three different natural rubber filled with carbon black N351. 1a. 4-14 Model prediction vs. experimental data for the stress-strain behavior in both uniaxial and plane-strain loading situations. 122 4-15 Mosel prediction vs. experimental data for the str in simple shear... . . . 122 4-16 Influence of different particle shapes on two dimensional simulations of a RVE containing one particle. Plane stress loading, area fraction filler of =0.15. ..... 125 4-17 Influence of different particle shapes for a more refined mesh. Plane stress loading 126 4-18 Influence of particle stiffness on the stress-strain behavior. Plane stress loading, shape parameter C = 0.8, area fraction filler ay = 0.15... . 127 4-19 Influence of mesh density on the normalized Young’s modulus. Plane stress loading, quadrilateral particles, area fraction particles ay = 0.15. 127 4-20 Influence of area fraction of particles on the normalized Young's mod- ulus. Plane stress loading. 129 421 Example contours of the first strain invariant J,. Plane stress loading, octalateral particles, shape parameter Cy = 0.8, area fraction filler ay = 0.24, applied strain ¢ = 129 44-22 Axisymmetric three-dimensional simulation with one particle illustrat- ing the influence of volume fraction of particles on the normalized Young's modulus. The parameter Cy is defined in Figure 4-17... .. 130 4-23 Comparison between predicted Young’s modulus for two-dimensional and three-dimensional simulations... ... . becee eee . 131 4-24 Predicted influence of volume fraction particles on the normalized Young's modulus from stochastic three-dimensional simulations. Each data point is the stochastic average of 10 simulations............- 132 4-25 Mesh refinement study of a RVE containing cubocahedral particles, wp SOL cece eee eee 133 4-26 Comparison between FE simulations and the experimental data for filled elastomers presented in Figure 4-4. Each data point is the stochas- tic average of 10 simulations... 6... 0.0... eee eee 134 4-27 I,-amplification in the matrix as a function applied strain. Uniaxial compression, vy = 0.15... 00.00 eee eee eee 135 4-28 I,-amplification in the matrix as a function applied strain. Plane strain compression, vy = 0.15. . 136 4.29 -amplification in the matrix as a function applied strain. Simple shear, vy = 0.15. 136 4-30 Amplified principal stretches in the matrix, uniaxial loading, vy = 0.15. Five simulations with dodecahedral particles (mesh density 96)... . 137 4-31 Amplified principal stretches in the matrix, plane-strain loading, vy 0.15. Three simulations with dodecahedral particles (mesh density 96). 138 4-32 Amplified principal stretches in the matrix, simple shear loading, vy = 0.15. Two simulations with dodecahedral particles (mesh density 96). 138 4-33 Uniaxial deformation of a RVE containing 15 vol% filler. The matrix is modeled with the eight-chain model. (a) Mesh used in the simulation, (b) filler particle distribution. 139 14-34 Deformed particle distribution of a RVE containing 15 vol% filler par- ticles tested in uniaxial compression. The matrix is modeled with the eight-chain model... 02... 2 eee 140 4-35 Components of the tensorial difference between the average rotation tensor of the filler particles (R), and the applied rotation tensor R. Uniaxial compression of a RVE containing 15 vol% filler particles. . . 140 4-36 Tensor norms of the difference between the average rotation tensor of the filler particles (R), and the applied rotation tensor R. Uniaxial compression of a RVE containing 15 vol% filler particles. .... . . . ML 4-37 Distribution in (Rj2), for a RVE containing 15 vol% fillers tested in uniaxial compression to a final strain of 0.0. ..........0-5 M1 4-38 FEM mesh and particle distribution used to study the simple shear response of a RVE containing 15 vol% filler particles, (a) perspective view showing the inesh, (b) perspective view showing the particle dis- tribution, (c) side view showing the particle distribution... ... . . 143 4.39 Perspective view of the deformed particle distributions. ....... 144 4-40 Stress-strain response in simple shear of a RVE continuing 15 vol% filler particles, = 0.67 MPa and AK =2........~. wee 144 4-41 Components of the tensorial difference between the average rotation tensor of the filler particles (R)y and the applied rotation tensor R. Simple shear of a RVE containing 15 vol% filler particles. .... . . . M5 4-42 Tensor norms of the difference between the average rctation tensor of the filler particles (R), and the applied rotation tensor R. Simple shear of a RVE containing 15 vol% filler particles. ... . . . 145 4-43 Distribution in (Ri2)y for a RVE containing 15 vol% fillers tested in simple shear, Riz = 0.20. 146 4-44 Example stress-strain response for a RVE with rhombic dodecahedral particles. Each data point is the stochastic average of 8 simulations, = 0.57 MPa, AP* = 25. 0.0.0.2 ee cee ee es 17 165-1 (a) Stress-strain response, ¢ = —0.1/s, relaxation time=20 s. (b) Stress-time response. 5-2 (a) Estimated equilibrium stress as a function of strain. (b) Time- dependent stress as a function of strain... ...........005 153 5-3. One free chain in anetwork. .. 2.0.00 eee eee 155 5-4 A dangling chain end in a network. 5-5 One dimensional rheological representation of the constitutive model. 159 5-6 Multiplicative decomposition of deformation... ....... . +++ 160 5-7 Uniaxial compression to different final strains, chloroprene rubber filled with 7 vol% carbon black N600. .. 6.6... ee ee eee 162 5-8 Stress relaxation test on a chloroprene rubber filled with 7 vol% carbon black, oe ee eee ee 163 5-9 Stress relaxation test showing the dependence on relaxation time for a chloroprene rubber filled with 7 vol% carbon black. .........- 163 5-10 Stress relaxation test showing the stress in the time-dependent network for a chloroprene rubber filled with 7 vol% carbon black. . . . . . 164 5-11 Uniaxial compression with different strain rates, chloroprene rubber filled with 7 vol% carbon black N600. . . . 5-12 Oscillatory test with frequency f = 0.02 Hz showing stress as a function of applied strain. ©... 0... eee eee \ PEEEEee 165 5-13 Oscillatory test with frequency f = 0.02 Hz showing stress as’ function 164 oftime. 2... ee ... 166 5-14 Oscillatory test with frequency f = 0.25 Hz showing stress as a funtion of applied strain. oo. eee eee eee eee he 167 5-15 Oscillatory test with frequency f = 0.25 Hz showing stress as a function oftime. ©... ee 168 5-16 Plane strain response of a chloroprene rubber filled with 7 vol% carbon black eee eee ee cece ees 168 5-17 Effect of different strain rates on the stress-strain response of a chloro- prene rubber filled with 15 vol% carbon black... ... 0... 6055 169 V75-18 Stress relaxation test showing stress as a function of time for a chloro- prene rubber filled with 15 vol% carbon black. ............- 5-19 Plane strain response of a chloroprene rubber filled with 25 vol% carbon black. 5-20 Large strain response of an unfilled natural rubber. ......... - 5-21 Effect of different strain rates on the stress-strain response for an un- filled natural rubber. 2.02... eee 5-22 Comparison between simple shear predictions and experimental data, unfilled natural rubber. 5-23 Comparison between simple shear predictions and experimental data, natural rubber filled with 17 vol% carbon black. . . 5-24 Effect of different strain rates on the stress-strain response of a natural rubber filled with 24 vol% carbon black N351............ 5-25 Simple shear behavior of a natural rubber filled with 24 vol% N351. . 5-26 Uniaxial compression to different final strains, chloroprene rubber filled with 7 vol% carbon black N60... 0.0.0.0. eee eee ee 5-27 Uniaxial compression with different strain rates, chloroprene rubber filled with 7 vol% carbon black N60... 2.2.2.0... eee black, 6. eee 5-29 Stress relaxation test showing the dependence on relaxation time for a chloroprene rubber filled with 7 vol% carbon black. .......... 5-30 Oscillatory test with frequency f = 0.02 Hz showing stress as a function oftime, 66. ee eee bee 5-32 Stress in network B as a function of applied strain. ... 20... . 5-33 Stress in the time-dependent network as a function of time... . . 5-34 Stress in the time-dependent network as a function of time. . 5-35 Inelastic strain in network B as a function of time. .......... 18 173 178 1795-36 Inelastic strain in network B as a function of time. ......... + 6-1 Effective shear stress 7 as a function of the applied strain, uniaxial compression, 7 vol% fillers... 2. ove eee eee 62 Elastic chain stretch of network B as a function of the applied strain, uniaxial compression, 7 vol% fillers... 2... eee ee eee 6-3 Inelastic chain stretch in network B as a function of the applied strain, uniaxial compression, 7 vol% fillers. 6-4 Chain stretch )* as a function of the applied strain, uniaxial compres- sion, 7 vol% fillers. 6-5 Effective stress r as a function of applied strain, uniaxial compression, Tvol% fillers... 2.0.0... cove eee beeen 6-6 Effective stress 7 as a function of applied strain, uniaxial compression, 17 vol% fillers... 2. eee 6-7. Effective stress 7 as a function of applied strain, simple shear, 17 vol% fillers. 6-8 Strain energy density as a function of applied strain, uniaxial compres- sion, 7 vol% fillers. . 6-9 Volumetric deformation J as a function of applied strain, uniaxial com- pression, 7 vol% fillers. 6-10 Pressure P as a function of applied strain, uniaxial compression, 7 vol% filers. cece teens 6-11 Stress Ty as a function of applied strain, uniaxial compression, 7 vol% fillers. ©... 00. eee 6-12 Deviatoric stress Tj, as a function of applied strain, uniaxial compres- sion, 7 vol% fillers... 2.0... . eee eee errr 6-13 Effective stress 7 normalized with the pressure P, uniaxial compression, Tvol% fillers. eee 6-14 Direct /,-amplification of the eight-chain networks, chloroprene rubber filled with carbon black N60... 2... 00.0 eee e eee eee ee 196-15 J,-amplification of networks only, natural rubber filled with carbon black N351. 2 eee 199 7-1 Pressure-density response of a Lennard-Jones fluid at 6 = 0.9, N = 400. 211 7-2. Distribution of atoms in the RVE in the undeformed state, N = 20000. 212 7-3 Bond length distribution in the undeformed state....... 2... 213 7-4 Radial distribution function of the simulated system in the undeformed State cece teen eee e eee es 24 7-5 Bond angle pole figure of the initial state of the simulated system... 214 7-6 Applied strain history. 2.0... 215 7-7 Distribution of atoms in the RVE: (a) undeformed state, (b) deformed state (¢ = 0.7), N= 20000............ cove cece eee 216 7-8 Energies of the RVE as a function of simulation time, MD simulation (NVO).. 02 eee ee eve eee cece eee 27 7-9 Average bond angle of all bonds as a function of simulation time, MD simulation (NV0). 0.0.0 eee eee eee 27 7-10 Bond angle pole figures, NV@, MD simulation...... 2... . . 219 7-11 Maximum, average and minimum bond lengths as a function of simu- lation time, MD simulation (NV@)........ ccc es 220 7-12 Lateral pressure in the RVE as a function of simulation time, MD simulation (NV8). 2.0.00... eee ee ne 220 7-13 Stress-strain response of the RVE, MD simulation (NVQ). ...... 221 7-14 Stress-time response of the RVE, MD simulation (NV@). ...... « 221 7-15 Decomposition of the stress in to bonded and non-bonded contribu- tions, MD simulation (V8). ..........- seventeen 222 7-16 Decomposition of the stress in to bonded and non-bonded contribu- tions, MD simulation (NV8). 2.2... 6... : vee 222 7-17 Average atom displacement. as a function of relaxation time, MD sim- ulations (NV8).. 0. eee ee 223 207-18 Energies of the RVE as a function of simulation time, MC simulation 224 7-19 Average bond angle of all bonds as a function of simulation time, MC simulation (NP8). 225 7-20 Bond angle pole figures, MC simulation (NP8). .......... 226 7-21 Maximum, average and minimum bond lengths as a function of simu- lation time, MC simulation (NP8). ....... 227 7-22 Lateral pressure in the RVE as a function of simulation time, MC simulation (NP8@). 2.02.00 ee eee 207 7-23 Volume of the RVE as a function of simulation time, MC simulation CO 2 Sete e tees 228 7-24 Stress-strain response of the RVE, MC simulation (NP8@). ..... . 228 7-25 Stress-time response of the RVE, MC simulation (NP). ....... 229 7-26 Decomposition of the stress in to bonded and non-bonded contribu- tions, MC simulation (NPQ). 20.02.2000 eee eee eee 229 7-27 Decomposition of the stress in to bonded and non-bonded contribu- tions, MC simulation (NP@). . 0.0.0.0 2020 2. 230 7-28 Crosslinking of the RVE as a function of simulation time (MD simula- C5 231 7-29 Bond length distribution in the undeformed state, MD simulation (NV0)232 7-30 Radial distribution function of the simulated system in the undeformed state, MD simulation (NV8). 0.0.00 eee ee eee . 233 7-31 Energies of the RVE as a function of simulation time, MD simulation OO) 233 7-32 Average bond angle of all bonds as a function of simulation time, MD simulation (NV@). ..... « bec eee eevee eee 1. 234 7-33 Bond angle pole figures, MD simulation (NV8)..........-. 5 235, 7-34 Lateral pressure in the RVE as a function of simulation time, MD simulation (NV0). 0.0.0 c cece eee eee eee 236 7-35 Stress-strain response of the RVE, MD simulation (NV8). .... . . 236 217-36 Stress-time response of the RVE, MD simulation (NV8). . . . 237 7-37 Decomposition of the stress in to bonded and non-bonded contribu- tions, MD simulation (NV@). 237 7-38 Average end-to-end distance for all chains, MD simulation (NV)... 238 7-39 Average chain stretch for all chains, MD simulation (NV@). .... . 239 7-40 Average chain stretch for all chains, MD simulation (NV8). 239 7-41 Average atom displacement as a function of relaxation time, MD sim- ulations (NVO). 00. ete eee 240 B-1 Particle radius for a circular/spherical particle as a function of area/volume fraction particles. ....... See eee me) C-1 Simple shear of a cylindrical specimen (diameter d = 28 mm, height h=13mm).... 259 22List of Tables 7.1 Input parameters used in the molecular simulations............ 213 7.2 Input parameters used in the multi-chain MD molecular simulation. . 231 23Chapter 1 Introduction ‘The ability to accurately predict the mechanical behavior of elastomeric materials is an important technological problem that is still far from being completely understood. In many applications elastomeric components are subjected to cyclic deformation at a certain frequency or over a range of frequencies. Typical examples of this include tires and engine mounts. In this type of applications the mechanical properties are often strongly dependent on the loading conditions such as temperature, frequency, deformation state, and the environment. To properly design new components of elastomeric material it is therefore of importance to be able to model the material behavior under different loading conditicns. Although a number of different, constitu- tive models addressing different characteristics of the elastomeric response have been proposed, the complex connection between deformation, microstructure, time and temperature is still not understood or well modeled. In this work both mechanical and computer experiments have been used in an effort to gain better insight into the mechanisms causing the observed behavior and to facilitate the development of a new constitutive framework. ‘The mechanical response of elastomers have a number of interesting characteris- tic features. For example, it is well known that an elastomer undergoes significant softening during the first couple of load cycles and that: the material response after that becomes repeatable (Mullins and Tobin 1965; Harwood et al. 1965; Harwood and Payne 1966a; Harwood and Payne 1966b). This phenomenon is often referred to 24as the Mullins effect. It is also well known that the stress in an elastomeric specimen after being subjected to a step in the applied strain will relax towards an equilibrium state (e.g. Ferry 1980). This observation is often interpreted as an indication of the existence of a hyperelastic strain energy function for sufficiently slow loading rates. A number of more intricate experiments have further shown that elastomers subjected to dynamic loading histories exhibit complicated time and temperature effects (e.g. Lion 1996; Hausler and Sayir 1995; Bergstrém and Boyce 1998). Throughout the years, a number of constitutive models have been developed ad- dressing different aspects of these observations. Most proposed models, however, are limited in scope in that they address only a subset of the known behavioral char- acteristics of the rubbery constitutive response. The reason for this is that rubbery materials exhibit very complicated non-linear time, temperature and history depen- dent behavior and that the amorphous character of the microstructure makes it dif ficult to develop powerful continuum models. The proposed models can be broadly divided into two categories: equilibrium models and time- and temperature- depen- dent models. Both of these categories can be further divided into models for unfilled elastomers and models for filled elastomers. Most of the early work (James and Guth 1943; Treloar 1975; Flory 1977; Wall and Flory 1951) was devoted to the prediction of the equilibrium response of unfilled elastomers, but there has also been a continuous effort in this field. One of the more successful models for this loading situation is the eight-chain network representation of Arruda and Boyce (1993). This model has been shown to successfully capture the equilibrium response under different loading conditions for many elastomers. The modeling of the equilibrium response of unfilled elastomers can at this point be considered to be a fairly well developed field, as evi- denced by a namber of detailed reviews (Treloar 1975; Mark and Erman 1994; Arruda 1992). It has been known for many years that adding small amounts of filler particles, such as carbon black, to an elastomer can significantly improve both the stiffness and the strength of the material. A significant amount of work has been done trying to understand and model how the filler particles influence the stiffness of the material 25(Smallwood 1944; Guth and Simha 1936; Vand 1948; Mooney 1951), Not nearly as much attention has been given to predictions of how filler particles change the large strain equilibrium behavior. Perhaps the most common approach for this particular problem is the Mullins and Tobin (1965) strain amplification concept, although there has also been some recent developments in this field by Govindjee (1997) and Ponte Castafieda (1989) Predictions of the large strain time-dependent behavior of unfilled elastomers sub- jected to general strain histories is another area in need of elucidation. Recently, there has been a renewed effort to understand and model these effects (e.g. Lion 1996; Hausler and Sayir 1995; Johnson et al. 1995; Dafalias 1991; Zdunek 1993). Most of the proposed models, however, capture only a subset of the experimentally observed phenomena and are mainly phenomenologically based. One further compli- cation of this general class of models is that they normally contain material dependent functionals that can be hard to experimentally determine for a new material One final problem that will be studied in this thesis is the large strain time- dependence of filled elastomers. Very few models that explicitly take the filler particles. into consideration have been proposed in the literature, in fact, only one model (Chen and Cheng 1997) specializing on the creep behavior was found. The goal of this work was to gain a better understanding and to develop a useful constitutive model that allows for accurate predictions of the observed large strain time-dependent behavior of particle filled elastomers. The first step towards this goal, described in Chapter 2, involves a detailed experimental investigation probing the mechanical behavior of a number of different elastomeric materials containing a range of different carbon black filler concentrations under different loading conditions. A new slightly different version of the eight-chain model of Arruda and Boyce (1993) is then presented in Chapter 3. The influence of filler particles on the equilibrium behavior of elastomers is the topic of Chapter 4. In this chapter is presented a new model (Bergstrém and Boyce 1999) based on amplification of the first strain invari- ant 1). To directly assess the influence of filler particles on the mechanical behavior, a set of finite element (FE) based micromechanical models are also studied. The 267Chapter 2 Experimental Results 2.1 Experimental Procedure Elastomeric materials exhibit a mechanical response which strongly depends on time, temperature and loading history as will be illustrated in this chapter via a series of experimental results for a number of different elastomers. Most experiments were per- formed in uniaxial compression but in order to test more general deformation states a few experiments were also performed in plane strain compression and simple shear, see schematic of the experimental setups shown in Figure 2-1. The uniaxial vompression experiments were performed on ASTM sized specimens (height and diameter were 13 mm and 28 mm, respectively.) The specimens were compressed between hardened steel compression platens which contained a spherical seat for improved alignment. Barrelling was prevented by inserting thin Teflon sheets between the specimen sur- faces and the compression platens. ‘The plane strain compression experiments were conducted using a channel die in which teflon sheets were inserted to reduce the inter- face friction, the geometry of the specimens used in these experiments was 10.2 mm 12.6 mm x14.1 mm. The specimen geometry used in the simple shear experiments was the same as in the uniaxial compression experiments. All tests were conducted using a computer controlled Instron servohydraulic uni- axial testing machine operated in strain control mode. All specimens were conditioned by five to six load cycles with increasing amplitude to a final true strain of about 1 28a a (b) (©) Figure 2-1: Schematic showing the experimental setup: (a) uniaxial compression, (b) plane strain compression, (c) simple shear. prior to testing to remove the influence of the Mullins effect and to insure repeata- bility in the tests. ‘The results from the experimental investigation of the different materials tested are summarized in the next few sections. 2.2 Chloroprene Rubber with 7 vol% Carbon Black The mechanical response of a chloroprene rubber filled with 7 vol% carbon black N600 is presented in this section. This particular type (N600) of carbon black particles have a mean diameter of approximately 60 nm and an example of the microstructure of the material is given in Figure 2-2. The micrograph shows that the particles are approximately spherical in shape and have a strong tendency to cluster into irregularly shaped aggregates. This clustering into aggregates is important for the mechanical response and will be discussed fuither in Section 4.2.1. A first example of the mechanical response of this material is shown in Figure 2.3 illustrating the magnitude of the strain-history induced softening during the first. few load cycles of a previously undeformed specimen, a phenomenon that has been 29Figure 2-2: TEM micrograph of a chloroprene rubber filled with 7 vol% carbon black N600. 30termed the Mullins effect in the literature (Mullins and Tobin 1965). The figure shows the stress response during the first four constant strain-rate load cycles. After this initial test series the stress-strain response becomes repeatable and it is this repeatable behavior that has been the focus of the current work. The magnitude of the Mullins effect for this material is shown to be rather small, the relative difference between the first load cycle and the repeatable behavior only being about 12%. load eycle # (sti: bI-O1) toad eyele #2 (esti: BIO2) load eye #3 (esti: B103) Head eye a (esti: BIO8) Figure 2-3: Mullins effect in a chloroprene rubber filled with 7 vol% carbon black N600. Figure 2-4 shows the stress-strain response when loaded to different final strains. As shown in the figure, there is significant hysteresis at all levels of final strain and the amount of hysteresis is roughly proportional to the maximum stress level. All solid lines in the figure were obtained from the same test specimen illustrating the repeatability between different tests on the same specimen, whereas the dashed curve was obtained from a different smaller specimen at a later time. Also note in the figure that there is virtually no residual strain after one complete deformation cycle, and in fact, the small amount that is seen is recovered with time after the test is completed. A first example of the time-dependent behavior of the material is shown in Figure 2-5 illustrating the influence of different applied strain rates on the stress 31fina straine=1.3 (test id: b3-06) Sess (MPa) chloroprene rubber 77 vol% N6O0 carbon Black ‘uniaxial compression ‘tran ate = 0.01. “2046S = 12 True Sain Figure 2-4: Stress-strain behavior when loaded to different final strains in a chloro- prene rubber filled with 7 vol% carbon black N600. response. From the figure it is clear that the amount of hysteresis increases with applied strain rate, and that the behavior during uploading is more rate-dependent, than the behavior during unloading. The amount of residual strain is also shown to be only weakly influenced by the applied strain rate. Another example of the time-dependent behavior is shown in Figure 2-6. Here the strain history consisted both of constant strain rate loading segments and segments during which the applied strain was held constant allowing the stress to relax. The details of the strain history are shown in the inset in Figure 2-6. The solid curve in the figure illustrates that, during the relaxation segments in the uploading the stress starts to relax to a lower level, but during the unloading the stress actually increases during the relaxation segments. For comparison, the figure also shows experimental data from a constant strain rate experiment and an estimated equilibrium curve representing the behavior for the case when the applied strain rate goes to zero. The experimental data from the same relaxation test is also shown in Figure 2-7, but here the stress is plotted as. a function of time illustrating the rate at which the stress relaxes. A third example of time dependence is shown in Figure 2-8 where the stress 320 01. 02-0304 05-06-0708 ‘True Stain Figure 2-5: Effect of different strain rates on a chloroprene rubber filled with 7 vol% carbon black N600. final strnin=-0.65, eax tieen20 5 (esti 9-07) final strain=-0. (jest: 24-04) estimated equlibeum a chloropene nubber 7 vol8 N600 eubon black Figure 2-6: Stress relaxation test on a chloroprene rubber filled with 7 voi% carbon black. In the experimental data represented by the solid curve the applied strain was held constant for 20 seconds at strain levels of {-0.07, -0.17, -0.27, -0.37, -0.47, -0.57}. 33_hloroprene rubber —cestid: ot-07) 7 vol% N600 carbon black ‘uniaxial compression strain rate = -0.01/ relaxation tine = 20 8 Time(s) ure 2-7: Stress relaxation test showing the dependence on relaxation time for a chloroprene rubber filled with 7 vol% carbon black. response when subjected to an oscillatory loading under a finite pre-stzain is depicted. ‘The applied strain history, as shown in the inset in the figure, consisted of a loading segment with a constant strain rate of é = —0.01/s to a final strain of ¢ = —0.4 after which five cycles of oscillatory strain with a strain amplitude of 0.1 and frequency of f = 0.25 Hz was applied. After this, the applied strain was further increased to a strain of ¢ = —0.6 and then immediately decreased back to ¢ = —0.4 and then the same oscillatory loading was applied once more. Finally the applied strain was taken back to zero. The experimentally observed stress response from this strain history is shown in Figures 2-8 and 2-9 as a function of the applied strain and time, respectively. The data shows that the stress-strain response during the superposed oscillatory loading depends on rate where much stiffer response is exhibited at the higher rates. ‘The figures also show that during the oscillatory segments the mean stress starts to relax as was observed in the relaxation experiments presented in Figures 2-6 and 2-7. The same type of experiment illustrating the same type of qualitative behavior is shown in Figure 2-10 and 2-11 for the case when the frequency of the applied strain was 0.02 Hz. 34— ent id: ot Figure 2-8: Oscillatory test with frequency f = 0.25 Hz showing stress as a function of applied strain of a chloroprene rubber filled with 7 vol% carbon black. a E-os| Bo oa chloroprene rubber "7 vol% N600 carbon black uniaxial compression strain rate= 01/5, 02 {frequency =025 Hz strain amplitude = 0.1 ‘number of cycles = 5 ar a ae) 1002040 0 Time (s) — testis: 17) Figure 2-9: Oscillatory test with frequency f = 0.25 Hz showing stress as a function of time of a chloroprene rubber filled with 7 vol% carbon black. 35,testi ot-18) Figure 2-10: Oscillatory test with frequency f = 0.02 Hz showing stress as a function of applied strain of a chloroprene rubber filled with 7 vol% carbon black. cloroprene rubber "7 vol% NEOO carbon black ° 100 200 00 00 500 00 Figure 2-11: Oscillatory test with frequency J = 0.02 Hz showing stress as a function of time of a chloroprene rubber filled with 7 vol% carbon black. 36The stress response when loaded in plane strain compression using the setup discussed in Section 2.1 is shown in Figure 2-12, The material exhibits a stiffer response in plane strain compression than in uniaxial compression and the plane strain stiffness increases with strain to a greater extent than the observed uniaxial compression. Figure 2-13 shows the effect of two different testing temperatures. The 3 TT= MRLenseeeisa eaetet 5 chloroprene rubber 7 voi N6OO carbon black sinin rate = 0.018 2s ‘True Stress (MPa) 05} 0 Ol 0203 04 0S 06 7 a8 ‘Tre Stain Figure 2-12: Uniaxial and plane strain response of a chloroprene rubber filled with 7 vol% carbon black. figure illustrates that increasing the temperature from 24°C to 80°C does not change the equilibrium stress in any significant way but the amount of hysteresis appears to slightly decrease with increasing temperature in agreement with experimental data from Lion (1997a). 37hloropene rubber 7 vol N6OO exon black ‘uniaxial compression strain ate= 001s 0 “a “02 “93 “0 “5 “06 ‘Tre Stain Figure 2-13: Stress-strain behavior in uniaxial compression at different temperatures of a chloroprene rubber filled with 7 vol% carbon black. 382.3 Chloroprene Rubber with 15 vol% Carbon Black This section presents experimental data for a chloroprene rubber filled with 15 vol% carbon black N600. The microstructure of this material is shown in Figure 2-14. The type N600 carbon black particles appear similar to the 7 vol% sample but due to the increased volume fraction of particies, clusters are more numerous. Figure 2-15 shows Figure 2-14: TEM micrograph showing the filler morphology of a chloroprene rubber with 15 vol% N600 carbon black. the response during the first three load cycles illustrating that the Mullins effect is not very strong for this higher filler content material. The behavior when loaded to different final strains is shown in Figure 2-16 demonstrating that the characteristic 390 or 2 OAS ‘Troe Susin 6 07-08 Figure 2-15: Stress-strain behavior during the first three load cycles for a chloroprene rubber filled with 15 vol% carbon black. rapid increase in stress with applied strain occurs a: a lower value of the applied strain for this more highly filled material. The figure also demonstrates that the unloading slope at strain reversals is strain dependent and becomes steeper at larger strains. ‘The influence of different applied strain rates is shown in Figure 2-17 for four different rates, It is clear that the material is rate-depencent and that the rate-dependence is higher during loading than during unloading. Two different stress relaxation tests are shown in Figure 2-18, in the small strain test the strain was held constant for 10 seconds at strain levels of {—0.05, ~0.10, 0.15} and in the large strain test the strain was held constant for 40 seconds at the strains {-0.18, —0.39}. In the figure is also shown the estimated equilibrium locus for the material. The variation in stress with time for the large strain relaxation testis further shown in Figure 2-19 illustrating that the stress relaxes rather fast during the first 10 to 20 seconds and that the relaxation rate gradually decreases with time. Another more dramatic example of this behavior is shown in Figures 2-20 and 2-21 illustrating the same qualitative behavior.final strain = -0.2 (est id: 41-11) final strain = 04 (test id: 1-10) final strain = 0 (test: 1-09) final strain = 038 (test: 1-07) final strain = ~L0 (test i: 1-08) chloroprene mabber 15 vol carbor Black N60O ‘nianiel compression stain re = 00/5 “01 02 03 04 05 06 07 08 09 -1 ‘Troe Stain Figure 2-16: Stress-strain behavior when loaded to different final strains of a chloro- prene rubber filled with 15 vol% carbon black. strain rte = -0.Us (testi: 1-13) Stain rate = -O.08/ (esti d-12) Strain rte = -O.OUs (esti: 1-07) Strain rte = ~O.002/ (est: d1=14) Figure 2-17: Effect of different strain rates on the stress-strain response of a chloro- prene rubber filled with 15 vol% carbon black. 41relaxation time=10s (test i: 3-04) Flaxation timed s (testi: d3-06) ‘sumated equiibaum “i -02—~SS CSCC Te Sain Figure 2-18: Stress relaxation tests showing stress as a function of strain for a chloro- prene rubber filled with 15 vol% carbon black. = — esis: 43-06) hloroprene rubber 15 vol eabon black N6OO -25} ‘uniaxial compression ‘elaration strain = (-0.18;-0.39:-029;-0.18), 0 30 100 150 20 250 300 Time(s) Figure 2-19: Stress relaxation test showing stress as a function of time for a chloro- prene rubber filled with 15 vol% carbon black. 42—Cestid:nt-21) hloroprene rubber 1S vol carbon black NSOO uniaxial compression strain aie =-0.01/s relaxation time = 605 “True Suess (MPa) 0-01 -02~—03 04-0506 a7 ~—~OR ‘Tee Strain Figure 2-20: Stress-strain response in a stress relaxation test on a chloroprene rubber filled with 15 vol% carbon black. ehloroprese rubber 15 vol carbon black N6OO ‘iaxial compression train ate =-0.0/¢ relaxation tine = 60 sinin =-0.59 i. _ i , 7 10" 10 10° 10 Relation Time (6) Figure 2-21: Stress relaxation as a function of relaxation time in a chloroprene rubber filled with 15 vol% carbon black. 432.4 Chloroprene rubber with 25 vol% Carbon Black This section shows results for one last type of chloroprene rubber filled with 25 vol% carbon black N600. The microstructure morphology of the material is shown in Figure 2-22. The TEM micrograph illustrates that when the material is this highly filled the spherical filler particles clearly cluster into large aggregate structures and that there are fewer isolated particles in this case compared to the 7 vol% and 15 vol% filled chloroprene rubbers presented in the last two sections, Figure 2-22: TEM micrograph of a chloroprene rubber filled with 25 vol% carbon black N600. The influence of Mullins effect during the first five load cycles is shown in Fig- 44ure 2-23. In this case each load cycle took the material to a larger strain than the previous one. From the figure it is clear that the damage induced in the material is controlled by the maximum value of the applied strain in the prior strain history. Figure 2-24 shows the stress-strain response when loaded to different final strains. load eycle #1: Gina srnine-0.2 (est load epele #2: final strin=—0.4 (eat = toad eget #3: final strane -0.6 est load eyete #4: final strain=-0.8 (est id = foadepele #5: final strain=—1.0 (est id 1-05) chloroprene rubber 25 vol carbon Black NGO ‘uniaxial compression san rate =-0.01/s 01 02 03 04 05 06 07 8 -09 -1 ‘Tree Suain Figure 2-23: Mullins effect in a chloroprene rubber filled with 25 vol% carbon black N600. The characteristic strain stiffening that occurs at large applied strains is shown very clearly. The influence of different applied strain rates is shown in Figure 2-25 having the same characteristics as the less filled chloroprene rubber presented in the last two sections. Figure 2-26 shows the stress relaxation behavior for both small and large strains, and based on this data an estimated equilibrium locus has been determined. ‘The response in plane-strain compression is presented in Figure 2-27 showing that the stress-strain response is significantly higher in plane-strain than in uniaxial load- ing. One example of the high temperature behavior is shown in Figure 2-28. When comparing the response at @ = 24°C with the response at @ = §3°C it is clear that the equilibrium response is not influenced very strongly but that the hysteresis decreases with increasing temperature. 45Sina strain=—0:2 (testi: f1-12) final sraine=O 8 (testi: 1-08) final stain=—1.0 (testi: £1-09) chiloroprene rubber 25 vol carbon black N60 ‘uniaxial compression sain ate =-001/5, 0-01-02 03 04 05 06 07 08 ~09 Tre Stain Figure 2-24: Stress-strain behavior when loaded to different final strains of a chloro- prene rubber filled with 25 vol% carbon black. strain te=-O Us test Figure 2-25: Effect of different strain rates on a chloroprene rubber filled with 25 vol% carbon black. 46= asmion mento ent it 3 Tz IN nesta en: 3-8 2. titel ots _hloroprene rubber “1 25 vol carbon black N6OO ‘Uniaxial compression os ‘sain aie =-0.01/s 0 =a 02 “03 “oa ~as 06 ‘Troe Stain Figure 2-26: Stress relaxation tests showing the stress-strain response of a chloroprene rubber filled with 25 vol% carbon black. “ —— pane rine gato Sa 85 4 z é Bb : 4 “ E clgcpae iter 2s vin enone sie ate p = a3 a a as Tia ein Figure 2-27: Uniaxial and plane strain response of a chloroprene rubber filled with 25 vol% carbon black. 47chloropene rubber 25 vol carbon back N60O, 02 “4 Tre Stain Figure 2-28: Effect of different temperatures on the stress-strain response of a chloro- prene rubber filled with 25 vol% carbon black. 482.5 Comparison Between the Different Chloroprene Rubbers In this section the results from the three different chloroprene rubbers will be directly compared. A first example of the different behavior of the three materials is shown in Figure 2-29 illustrating the total stress as a function of applied strain. The figure shows that the separation between the uploading and unloading curves at fixed strain increases with filler concentration resulting in larger amounts of hysteresis with filler concentration when subjected to the same strain history. The estimated equilibrium chloroprene rubber carbon black: N6OO uniaxial compression strinrate =-00I/¢ OI 02-03 0405-06 ~—-07~—O8 ‘Tre Stain Figure 2-29: Comparison between the total stress-strain response of the three chloro- prene rubbers tested. loci of the materials, i.e. the response that is approached in the limit when the ap- plied strain rate goes to zero, are shown in Figure 2-30. This figure shows that not only does the initial stiffness of the material increase with filler concentration but the strain induced stiffening also becomes more pronounced. ‘The figure also shows that the character of the curves changes with filler concentration: at low filler concentra- tions the stress-strain curve is convex but at high filler concentrations the curve has a sigmoidal shape. Figure 2-31 shows the time-dependent stress as a function of the ap- 49hloroprene rubber carbon black: N6OO nian] compression ry a Cy a Tree Stain Figure 2-30: Comparison between the equilibrium loci for three different chloroprene rubbers. plied strain, Here the time-dependent stress is defined as the experimentally observed stress minus the equilibrium stress. The figure clearly shows that the hysteresis for a given strain cycle increases with an increase in filler concentration. Finally is shown in Figures 2-32 and 2-33 examples of how the time-dependent, stress depends on applied strain for three different strain rates of two different chloro- prene rubber. The figures show that the time-dependent stress at different strain rates has similar characteristics as shown in Figure 2-31 for different filler concentrations. 50relax time=40s (esti 3-04) iments (esti: 3-06) i't-04) os! Figure 2-31: Comparison between the time-dependent stress for three different chloro- prene rubbers. Figure 2-32: Time-dependent stress for a chloroprene rubber filled with 7 vol% carbon black tested at three different strain rates. 5104h ehloropree rubber 1 voi N6OD carbon black tesa compression 0 Ol 02 03 04 05S 06 07 08 ‘Tre Strain Figure 2-33: Time-dependent stress for a chloroprene rubber filled with 15 vol% carbon black tested at four different strain rates. 522.6 Unfilled Natural Rubber In this section results from a set of tests performed on unfilled natural rubber are summarized. The experimental setup and techniques used are the same as those used for the chloroprene rubber presented in the last few sections. As a first example, Figure 2-34 shows the influence of two different strain rates. As before, the material appears to have slightly higher rate-dependence during loading than during unloading. Figure 2-35 shows the behavior when tested to two different final strains, and also the — stein rate=0.1/s (est id: 900—4-me-06) = = strain ate=-0.015 (testi: rs00-4-me-07) ~~ sain rate=-0.01/ (esti ne00-4) ‘estimated eqilibsium é Su fad od 4 nt er td apa i a Figure 2-34: Effect of different strain rates on the stress-strain response for an unfilled natural rubber. estimated equilibrium behavior. In Figure 2-36 the simple shear response is plotted, illustrating a weak rate-dependence and very little hysteresis. In the simple shear experiments, the shear stress was calculated by dividing the applied force with the cross-sectional area. 53final suain== 8 (testi ne0-1-me~10) final strain=—1 49 (est 100-3-me—03) estimated equilibria vnfed natura rubber ~s ‘nial compression ‘ain te= 0.015 ‘Tre Stain Figure 2-35: Large strain response of an unfilled natural rubber. 03 025} ous} “Tre Stress ,, (MPa) 00s} ‘unfilled ura exer simple shear or 02 04 05 06 03 Enginering Stain, Figure 2-36: Stress-strain behavior of an unfilled natural rubber in simple shear. 542.7 Natural Rubber with 7 vol% Carbon Black ‘The behavior of natural rubbe: ‘illed with 7 vol% carbon black N351! will be presented in this section. A first example shown in Figure 2-37 demonstrate the observed magnitude of the Mullins effect. Figure 2-38 shows the influence of different strain toad cycle (esti Wai hfe Head eye 8 (estar ;, a aun robber vol carbon black W331 ‘nial eonpresion ‘eninraie 0018 3203 “5 06-07-08 “4 ‘Troe Stain Figure 2-37: Mullins effect in a natural rubber filled with 7 vol% carbon black N351. rates and different final strains. Quantitatively very similar behavior is observed for this material as for the unfilled natural rubber. The influence of different strain rates is also illustrated in the stress relaxation tests shown in Figure 2-39. The experimental data for the slower test in Figure 2-39 is also plotted as a function of time in Figure 2-40. From this figure it is clear that the magnitude and rate of stress relaxation increases with applied strain. One last cxample of this type, at lower strain rate, is shown in Figure 2-41 in which the estimated equilibrium curve has been added. "This type of carbon black has a number average particle diameter of 31 nm (ie. see Figure 2-42). 55ol MRIESB atest asia) , ly sacl uber Za vo caon lack N33 anal compresion “22 “2a 05 “as True Surin Figure 2-38: Effect. of different final strains and strain rates on the stress-strain re- sponse of a natural rubber filled with 7 vol% carbon black. 22 fae tme=s min, stan te=0 5 (etd: ne} S-1 -ch-09) caf ===" felts tme=305, Simin rates-0.s Gest id: 9-1-2) is -13 " -16 ral rubber 7 val cathon Mack N¥ST uniaxial compression 0 Or O2~OS OSCR ‘Troe Stain Figure 2-39: Stress relaxation behavior at different strain rates of a natural rubber filled with 7 vol% carbon black. 56— (test net 5-1-ch-02) -15| “Trve Stress (MPa) os natural rubber 7 vol carbon black N3S1 ‘uiaxal compression ‘relaxation Ume= 308 ‘san rate =—O.15 a a er) ime (5) Figure 2-40: Stress-time response in a stress relaxation experiment on a na’ ural rubber filled with 7 vol% carbon black. natural rubber 7 vol carbon black N351 ‘uniaxial compression relaxation time = § min ‘stain ate= D008, “Ol -02—~03— OA SOC SCO “Tree Stain Figure 2-41: Stress relaxation test showing the stress-strain response of a natural rubber filled with 7 vol% carbon black. 872.8 Natural Rubber with 9 vol% Carbon Black In this section the results for natural rubber filled with 9 vol% carbon black N351 will be presented. One example of the microstructure is shown in Figure 2-42 illustrating that the average particle size if about 30 nm and that the particles tend to cluster into larger aggregates. The influence of the Mullins effect, as illustrated in Figure Figure 2-4: N351 ‘TEM micrograph of a natural rubber filled with 9 vol% carbon black 2-43, is qualitatively very similar to what was observed in the last section. In Figure 2-44 is shown the response when subjected to two different strain rates and in Figure 2-45 is shown the stress relaxation response and the estimated equilibrium behavior. 58fond cycle #1 (esi n20-1-me-1) load eyele #2 (est id n20-I-me-2) load eyete #3 (testi: ne20-I-me-3) t Toad eyele #4 (est id 20-1-me—4) stun rubber 9 volt carbon black NST ‘uniaxial compression ‘stn utes 0.1 ~02 030s 05-06 07-08 ‘Tre Stain Figure 2-43: Mullins effect in a natural rubber filled with 9 vol% carbon black N351. 59TT titin tO tid 29 to me-6 "= imi ate=-0.010 eats 1220-Tomo » anual rubber 9 voll carbon blnck N3SI ‘uniaxial compression “1 02-03. 0s 05806078 ‘Tre Surin Figure 2-44: Effect of different strain rates on a natural rubber filled with 9 vol% carbon black. strain rates, relation time=60 5 (esti: re20-1-ch-1 stmated equim ‘ 7 “True Suess (MPa) ey atu rubber 9 vols carbon Blick N3SI ‘niaial compression “2 03-0405 6ST ‘Troe Stain Figure 2-45: Stress relaxation behavior of a natural rubber filled with 9 vol% carbon black. 602.9 Natural Rubber with 17 vol% Carbon Black This section presents experimental results for a natural rubber containing 17 vol% of two different types of carbon black particles. The first type, N351, is the same that was used in the last two sections and the second type, N231, has a smaller mean particle diameter (d + 20 nm). By comparing the experimental data for these two materials it is possible to also directly investigate the influence of different filler systems in addition to filler concentrations. ‘The results for the N351 carbon black filled elastomer (see the micrograph in Figure 2-46) are presented first. Figure 2-47 summarizes experimental results for different strain rates and stress relaxation behavior both at large and small strains. Figure 2-48 shows the dependence on different strain rates. Figure 2-49 shows the stress relaxation behavior as a function of strain and Figure 2-50 as a function of time. Figure 2-51 shows stress relaxation behavior at a slower applied strain rate and the estimated equilibrium behavior. In Figure 2-52 is shown the same experimental data as in 2-51 but the stress is now plotted as a function of time. As a last example, Figure 2-53 shows the simple shear response at two different strain rates. The figure demonstrates both rather weak strain rate dependence and relatively small amount of hysteresis. Figure 2-54 shows the strain rate dependence for a natural rubber filled with 17 vol% carbon black N231 and Figure 2-55 illustrates the stress relaxation behavior and the estimated equilibrium curve. Figures 2-56 and 2-57 show the stress relaxation behavior at two different strain rates. A direct comparison between the two types of carbon black particles is shown in Figure 2-58 illustrating that the equilibrium response appears to be slightly higher for the N351-filled rubber but that the hysteresis is higher for the N231-filled natural rubber. 61Figure 2-46: TEM micrograph of a natural rubber filled with 17 vol% carbon black N351. Note rather uniform dispersion of particles throughout the rubber matrix. 62strain rte=-O 1 (est id: 0-1-ch-O1) Sra mie=0. 1s (et: resdt—et-01) [== Stain ate=-0.01/5 (estid nedO-I=me=11) wh 25) a z yee $2 ov i! 15 ut” + sata rubber os) 17 vol% carbon lack N351 anal compression relaxation ime = 60 0 Or 2 03 04 0S 06 07-08 ‘Tre Stain Figure 2-47: Effect of different strain rates and stress relaxation behavior on a natural rubber filled with 17 vol% carbon black N351. a4 4 a i 2 a by “ 5 17a cue te empesion a CT) ‘Tne Stain Figure 2-48: Effect of different strain rates on the stress-strain response of a natural rubber filled with 17 vol% carbon black N351.3st —~ (estid:e40-2- R° a macroscopic motion of Bp in a time interval t € (0, t,] C IR. which maps any material point p to the point x = x,(p) € R® 86in the deformed configuration B, = x;(Bo) C R*. In the following the reference configuration is taken to be stress-free with homogeneous reference temperature 4) € R,. The total deformation gradient defined by F := dp x(p,t) can be decomposed into distortional and dilatational parts: F = (J)!/°F* where J = det(F). To get a physical picture of the deformation process of an elastomer assume that the chain molecules, in a statistical sense, can be considered to be located along the diagonals of a unit cell located in principal stretch space as illustrated in Figure 3-1 (this is the eight-chain assumption of Arruda and Boyce 1993.) Denote the side lengths of 42.0 x 1 Figure 3-1: Representative volume element used in the 8-chain model. the unit cell in the reference state by ao and the undeformed chain length by ro (it then directly follows that ro = apV3.) Further take the macromolecules to be freely jointed with n rigid links each of length !. For this chain model the average end-to- end distance in the absence of an external force field is V/A (e.g., Doi and Edwards 1986). By defining {A} |i = 1, 2,3} to be the applied principal distortional stretches the effective distortional chain length becomes r* = aa[(A{)? + (A3)? + (A3)?]!?, giving trCt V3 the effective distortional chain stretch OGY? + 9)? + 09)? 87where B* = (J)-?/*B. Based on this physical model, an eight-chain material is defined as an isotropic thermoelastic material whose strain energy density W only depends on the two invariants 1*(B*), J(F), and the temperature @. By ni that the effective chain stretch is related to the first invariant of B* through X* [1,(B*)/3]/? it follows that the strain energy density can be written W(A*, J,0), or alternatively W(I}, J,0) where J} = [,(B*) = tr(B*). ‘To determine the strain energy density a similar approach as used by Anand (1996) is here implemented. First use the experimental observation that the internal energy is not a function of the applied distortional stretch (Treloar 1975), ie. e(J,6). The specific heat at constant volume, defined by cy = de(J,6)/00, can then be written On(°, J,8) 0) + ORO (3.1) where 7 is the specific entropy and the definition of the Helmholtz free energy p(X", J, 8) = e(J,9) — 8n(X*, J,8) has been used. Since the first two terms of the right hand side of (3.1) have io cancel to satisfy the Clausius-Duhem inequality the specific heat becomes: (3.2) Further assume that the material is almost incompressible and that the temperature does not vary too much such that ¢,, to a first order approximation, can be taken to be a constant. Equation (3.2) then give e(J,0) = c+ e(J) and 9(X*, J,0) = cyln@ + m(A*, J) giving the Helmholtz free energy yd", J,8) = cy0(1 — In) + ex (J) — Om (2*, J). (3.3) Assuming small volumetric deformations, the linearized relationship between the Cauchy stress T and the volumetric deformation J is taken as aw as =Tom(J—1)1 (3.4) 88where « is the bulk modulus, giving the convex relationship e,(J) = «(J — 1)?/(2p0) which when inserted into (3.3) gives W(, J,) = pocyA[1 — Ind] + =| gl - ar - 8pon(2"). (3.5) Note that the dependence on J in n(A*,J) has been neglected due to the the as- sumption of small volume change. It now only remains to determine how the entropy depends on the effective chain stretch. By assuming the molecules to be freely jointed with a fixed bond length the entropy can be determined from the statistical mechanics relation [pon] « Nkg In Q(r) where 9(r) is the probability distribution of the end-to- end distance of the molecular chain and N is the number of chains per unit reference volume. Flory (1988) showed that the probability distribution under these conditions can be written a) = se [asin [82 te (3.6) One approximation of (3.6) good for n >> 1 and r ~ n is the Langevin expression attributed to Kuhn and Griihn (1942): a [se By" en [ where 8 = #~\(r/(nl)) and #(8) = coth(8) — 1/6 is the Langevin function, see Figure 3-2. For small z the inverse Langevin function can be Taylor expanded into Qr) 7 ] (3.7) “Na = 324 20° + Bas + 02"), (38) L and for large |x| the asymptotic form of the Langevin function is 2 (8) * sign() ~ z ita] >1, (9) 89B=L"enay) Figure 3-2: Langevin and inverse Langevin functions. The three term Taylor expan- sion shown in the figure is taken from Equation (3.8) and the curve fit from Equation (3.11). giving 1 * Jena) —2' @e) if 2] = 1—<, where e <1. (3.10) A computationally efficient curve fit of the inverse Langevin function valid’ for all x is 1.31446 tan(1.58986z) + 0.912092, if |x| < 0.84136, 2 (z)= (3.11) 1/(sign(x) — 2), if 0.84136 < |x| < 1. In the limit n + 00, Equation (3.7) becomes a Gaussian distribution a(n) = lal men [| . (12) By inserting (3.7) into (3.5) and using r/l = "A the strain energy density can be ‘a relative error that is less than 6.4-10-* for all x € (—1,1). "The curve fit 90written WO, J,0) = NkpOBN*X* — Niko (1)? In (34) + 3 -ap + pocy9[1 — Ind] + Wo (3.13) where ) yok = a, (3.14) To obtain the Cauchy stress for an eight-chain material start with the continuum mechanics expression for an objective thermoelastic material”: 2 paMC.o) 9) pr T(C,0) = (3.15) By imposing isotropy and decomposing the deformation gradient into dilatational and distortional components a direct application of the chain rule to (3.15) gives 2 ow — 41; 0W Oa ~ Bi aR ~ 37 aR |» 19 which in this case with no dependence on J} can be simplified to =5 oT W dev + a (3.17) or when expressed in terms of the effective chain stretch )* (3.18) 2A thermoelastic material is here defined as a material in which the field variables ¢(-), T(-), (+), a(-) only depend on the state variables F, @, and g. 91Inserting (3.13) into (3.18) gives a . T re re (a) dev[B*] + «[J — 1]1. (3.19) For the special case of simple shear, defined by F = 1+ 7¢; @e», the shear stress is given by ' ( =) 7 (3.20) where ¥* = 1+ 7/3. The initial shear modulus of the material is given by p° = FL =0 yielding Mate yea! ( a) (3.21) which when inserted in (3.19) gives the Cauchy stress as T gal * "(a FR eB + aly - a (3.22) >) For the special case of incompressible uniaxial deformation the corresponding equation 2 Cy becomes vA 4 , (3.23) where }* = [(\? + 2/A)/3]"/. For the special case of incompressible plane strain 92loading the Cauchy stress is given by DE 2(5) (2-1) 00 80 1 Ty= 0 o 90 5 2g) 0 0 (1-1/2?) (3.24) where A* = [(A? + 1/A?-+1)/3}"”2, And for simple shear the Cauchy stress is given by : (3.25) where F = /T+ 7773. The constitutive relationship for the case of Gaussian chains is directly obtained by replacing the inverse Langevin function by the first term in its series expansion &-\(x) = 3z, giving: ° z 5 dev[B*] + x[J — 1]2. (3.26) 3.3 Determination of Material Constants In this section, a systematic procedure for determining the material constants used in the version of the eight-chain model given by Equation (3.22) will be presented. ‘The procedure will be demonstrated for the chloroprene rubber filled with 7 vol% carbon black that was discussed in Section 2.2. Start with the estimated equilibrium locus shown in Figure 2-18 and pick one point (c¢,¢¢) on that locus at a relatively moderate strain level, as shown in Figure 3-3. The initial shear modulus :° can then be approximated by (3.27) 93“© predicted equilibria ¢hloroprene rubber 1 ole carbon back N6OO uniaxial compression “True Suess (MPa) Figure 3-3: Estimated equilibrium locus of a chloroprene rubber filled with 7 vol% carbon black, €¢ = —0.4, 6, = —0.78 MPa giving y° = 0.75 MPa. The limiting chain stretch is obtained from the stress-strain behavior at large strains, émar = —2.24 giving Nov = 2.5. One way to determine the limiting chain stretch A! is to estimate the strain at which the stress starts to increase without limit, see Figure 3-3. The limiting chain stretch \'% can then be calculated from dreck (3.28) ‘The material constant « is the bulk modulus and cannot be obtained from a simple uniaxial test. Instead, it can be looked up in a handbook or it can simply be taken sufficiently large, say O(x/u°) = 10°. 3.4 Comparison with Experimental Data ‘The model presented in Section 3.2 will be compared to a selection of different ma- terials tested under different loading conditions. The results from the comparisons, shown in Figures 3-4 to 3-14, illustrate that the eight-chain model can be used to 94accurately predict the equilibrium response of many different elastomers tested in different loading modes. The predicted simple shear response shown in Figures 3-9 and 3-14 was calculated using the procedure discussed in Appendix C. The figures demonstrate that if the material constants are chosen appropriately the equilibrium response also of filled elastomers can be predicted, even though the eight-chain model as presented in this chapter was developed for unfilled elastomers. To explicitly in- corporate filler interactions in the modeling requires a more systematic study of the influence of filler particles on the equilibrium response and is the topic of the next chapter. strain rate=-0 01, relax time=20 5 (testi: of-07) ‘ight-chain model hloroprene rubber 77 vole carbon black N&O ‘uniaxial compression w°<0.75 MPa aM=2.5 a a a a “Tre Stain Figure 3-4: Comparison between experimental data and eight-chain predicted be- havior of a chloroprene rubber filled with 7 vol% carbon black tested in uniaxial compression. 95lane strain cc ion (test / sgesrancmgeen oon Ecce aT on eaomnacns Seat chloroprene rubber “Tole earbon black N6OO smn ate = -001/6 y xy} W°=0.75 MPa, X25 2 04 ~~ —06_~—08 “Tie Stain Figure 3-5: Comparison between experimental data and eight-chain predicted behav- ior of a chloroprene rubber filled with 7 vol% carbon black tested in plane strain compression, age gge: 204 lag iment set i: 3-00 hloroprene rubber 15 volt eabon back N6OO ‘niaxal compression p°=1.2 MPa i219 0-01 -02—~=C OSC “True Surin Figure 3-6: Comparison between experimental data and eight-chain predicted behav- ior of a chloroprene rubber filled with 15 vol% carbon black. 96an a et i 25902) ate can atl asm) sealer Cth od antxat) ~ “Chloroprene rubber 7°77” 25 volt carbon Back N6OO 7 steam ate= -0.01/s wet 9s MPa, uSt 45 0 “or “02 “3 ~o4 “05 706 ‘Tre Suain Figure 3-7: Comparison between experimental data and eight-chain predicted behav- ior of a chloroprene rubber filled with 25 vol% carbon black. ies-0 01 (esti: ns00-3-me-05) inmosel ‘unfilled astral ra ‘uniaxial compression Figure 3-8: Comparison between experimental data and eight-chain predicted hehav- ior of an unfilled natural rubber tested in uniaxial compression. 97025; TIT epee mesg ons este: ars00) oa] - é = ous fe ¢ a 2 05 untied nara rbber ple shear v0 MPa, abe 00501 01S 0202803035 04 Engincering Stain, Ey, Figure 3-9: Comparison between experimental data and eight-chain predicted behav- ior of an unfilled natural rubber tested in simple shear. * pepe 20 mtn nes min eta «6 =1.4t “ z Eng 2: ° Bod C od : rane tes od A 1 va eaten bet NST teil congesion 0.2} g ¥°=0.7 MPa, i dl cat 02)a 0s ‘Tre Strain Figure 3-10: Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 7 vol% carbon black. 98SE eget lg rtnation times (et 20-11) : gne-chain mode aural ubber 9 vole carbonblack N3SI uniaxial compression 03} 1°-0.75 MPa, 232.0 % “or 02 03 0a 05-06-0708 ‘Troe Sin Figure 3-11: Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 9 vol% carbon black. SRR NOON lato mens min esi 2-eh-01) aural rubber 17 vol carbon Black N331 ‘niaxal compression w°=L.0 MPa, A219 “or 02~—-03.~—Oa— SOC ‘Tre Stain Figure 3-12: Comparison between experimental data and cight-chain predicted be- havior of a natural rubber filled with 17 vol% carbon black.strain rate=-0.01/s (test i: nr60-1-me-11) ‘ight-chain model satura ruber 24 vol carbon Back N3S1 ‘uniaxial compression e122 MPa, aan 5 “True ures (MPa) -s 0” 01 02-03 04 05-06-0708 Troe Stain Figure 3-13: Comparison between experimental data and eight-chain predicted behav- ior of a natural rubber filled with 24 vol% carbon black tested in uniaxial compression. 07; a rr — shear sin ates 0.0% (esti 60-06 TTT SES mei" » natura rubber 24 vol carbon black N3ST simple shear Wet2 MPa, AALS 00s 01 ONS 02025 03 «01S «4 OS Engineering Suan, E, Figure 3-14: Comparison between experimental data and eight-chain predicted be- havior of a natural rubber filled with 24 vol% carbon black tested in simple shear. 100Chapter 4 Equilibrium Behavior of Filled Elastomers 4.1 Introduction In this chapter the influence of filler particles on the equilibrium response will be ex- plicitly considered. From the experimental data presented in Chapter 2 and a number of experimental investigations (Meinecke and Taftaf 1988; Smallwood 1944; Mullins and Tobin 1965) it is well known that adding small amounts of filler particles to an elastomer can have a strong influence on its mechanical response. In particular, the incorporation of filler particles is known to (1) increase the stiffness of the material, (2) change the strain history dependence of the stiffness (commonly referred to as the Mullins effect), and (3) alter time-dependent aspects of material behavior such as hysteresis and stress relaxation. While filler particles are known to alter these impor- tant aspects of the macroscopic stress-strain behavior of elastomeric materials, the mechanism by which the alterations occur is still a subject of debate. For example, the increase in stiffness is considered to result. from two primary contributions: (i) the continuum level explanation whereby the stiffness of a composite will be some weighted combination of the stiffnesses of the individual constituent materials, de- pending on the exact microstructure; and (ii) the molecular level explanation that the filler acts both to effectively increase the crosslink density of the material by 101providing additional crosslinking sites at the particle-matrix interface (Bueche 1960, 1961) and also to reduce the segmental mobility close to the filler particles (Kraus 1978). In addition to the volume fraction of filler particles, a number of state variables have been proposed to influence the magnitude of these contributions: (1) the size, type, and shape of the fillers (Mullins 1950); (2) the filler aggregate structure (Small- wood 1944; Guth and Gold 1938; Meinecke and Taftaf 1988) (here the term structure is used in the sense of Medalia and Kraus (1994), i.e, structure is the property specify- ing the bulkiness of the aggregates); and (3) the polymer-filler interface area (Medalia and Kraus 1994). Of these three parameters, the particle size and the interface area have received perhaps the most attention in the literature (Mullins 1950; Paipetis, Melanitis, and Kostopoulos 1988; Medalia and Kraus 1994); this seems somewhat unjust since an early paper by Smallwood (1944) convincingly demonstrated that the main influence of this internal length scale is to change the aggregate structure of the filler particles. The hypothesis that the filler acts at a molecular level to ef- fectively increase crosslinking sites has been exploited by Blanchard and Parkinson (1952), Bueche (1960, 1961), and more recently, by Govindjee and Simo (1991) to explain and model the Mullins effect, which is the term given to the phenomenon whereby an elastomer experiences a loss of stiffness during its first few load excur- sions. The loss in stiffness is found to depend on the stress level the material had experienced in prior loadings. In modeling this phenomenon, Bueche (1960, 1961) as well as Govindjee and Simo (1991) attributed the loss in stiffness to be a result of the breaking of the matrix molecular chain/particle bond and the effective loss in crosslinking density. While the Bueche-type model was successfully fitted to Mullins effect data as shown in Govindjee and Simo (1991), the debonding of the matrix from the particle does not appear to be a successful mechanistic explanation for the loss in stifiness. As shown in Harwood, Mullins, and Payne (1965) as well as Harwood and Payne (1966a), the Mullins effect is observed in both filled and unfilled elastomers. Indeed, Harwood and Payne (1966a) have shown that for a given elastomer, the loss of stiffness can be correlated to the maximum stress achieved in prior loadings and is, independent of the volume fraction of carbon black. They postulate that the presence 102of the carbon black acts to amplify the matrix strain (over that of the applied strain), thus making the Mullins effect more apparent at smaller applied strains in the filled elastomers than in the corresponding unfilled elastomers. In other words, the carbon black acts as a continuum level reinforcement and the strain history dependence of stiffness (Mullins effect) is due to a molecular level event occurring in the matrix material (not at the matrix/particle interface). In addition to stiffness, fillers are also observed to alter time-dependent aspects of the stress-strain behavior including strain rate dependence of the stress-strain behavior, hysteresis, and stress relaxation behavior. Filled elastomers appear to exhibit a higher rate-dependence and larger hysteresis than their corresponding unfilled elastomer. However, this may simply be a result of the amplification of the matrix strain (as in the case of the Mullins effect as described above) or there may be additional dissipative mechanisms introduced due to the matrix/particle interface. This background suggests that in order to clarify the relationship between filler content and its influence on mechanical behavior (stiffness, Mullins effect, hysteresis, time-dependence), continuum level modeling of the interaction between filler and ma- trix should be further explored. Prior to investigating the effectiveness of continuum level modeling in capturing the dependence of the Mullins effect and time-dependent aspects on filler content, the ability of continuum level modeling to predict the de- pendence of stiffness on filler content will first be explored and is the topic of this chapter. Note, continuum modeling can only be considered as a candidate for ex- plaining filler effects on strain history and time-dependent aspects of behavior if it: can successfully predict the stiffness. Below, experimental results of the dependence of stiffness on filler content will first be reviewed. The ability of various models to predict the observed dependencies will then be discussed. Micromechanical modeling of the particle filled elastomer will then be presented using both two-dimensional and three-dimensional finite element simulations. The micromechanical model predictions will then be compared to experimental data as well as composite models. 1034.2 Experimental Data To enable a direct comparison and evaluation of different proposed constitutive models for the isothermal equilibrium response of particle filled elastomers, it is crucial to have accurate experimental data. A number of experiments on both chloroprene rubber and natural rubber were presented in Chapter 2. Some of this data is shown again in Figure 4-1 illustrating the resulting stress-strain behavior! for three different carbon black filled chloroprene elastomers differing only in volume fraction filler. Besides 25 vol ch (test id: 43-02) 15 vole eb test 3-09), Tale cb (etd: 04-11), ~chforoprene ruber caubor black: N6OD nial compression “005 “Or “ous “92 Figure 4-1: Stress-strain behavior of a chloroprene rubber filled with different levels of carbon black. demonstrating the strong influence of filler particle concentration, Figure 4-1 also shows the pronounced amount of hysteresis and the negligible amount of residual strain that is commonly observed in elastomers. It is also clear that the stress during the relaxation periods relaxes towards an equilibrium state and that this equilibrium level is only dependent on the current strain state. From this data it is possible to estimate the equilibrium stress-strain curve for each filler concentration, as shown Note that true stress (the force divided by the current cross-sectional area) and true strain (the natural logarithm of the current height divided by the original height) are the stress and strain measures used. 104earlier in Figure 2-30 and repeated in Figure 4-2. . ‘True Suess (MP2) “or 02-03 _0s 0S 06-07-08 Tre Sain Figure 4-2: Comparison between the equilibrium loci for three different. chloroprene rubbers. ‘The experimentally predicted equilibrium behavior of natural rubber filled with different. concentrations of carbon black particles is shown in Figure 4-3. A direct comparison between Figures 4-2 and 4-3 demonstrate that the carbon black particles in the chloroprene rubber have a stronger influence on the stiffness than the carbon black particles in the natural rubber. The reason for the different. reinforcing behavior is due to the different particle dispersions in the two cases. As was shown in Figures 2.2, 2-14 and 2-22 the particles in the chloroprene rubber have a strong tendency to cluster into elongated aggregates whereas the filler particles in the natural rubber shown in Figures 2-42, 2-46 and 2-59 are more uniformly dispersed. 4.2.1 Initial Modulus ‘The most direct evaluation of the influence of filler particles on the mechanical re- sponse of elastomers is to consider the small strain Young’s modulus versus volume fraction particles, as shown in Figure 4-4. In this figure the experimental data for a number of different elastomers and filler systems are summarized. Although there is 1052 volwed 17 vols cb vale cb To cb ‘filed tural evbber cexrbonblick: N3ST uniaxial sompression predicted equittrium “8 ‘Tree Stain “05 “08 Figure 4-3: Comparison between the equilibrium loci for the four different natural rubbers tested. a scatter in the data due to the different elastomers and fillers used by the various investigators, the figure indicates in what range the normalized modulus is likely to be for a given volume fraction and particle type. This information will be used in the next section as the reference to which several proposed constitutive models will be compared. 4.3 Theoretical Models In this section the dependence of the stress-strain behavior on the volume fraction filler particles will be studied. As a starting point, a comparison between the experimental data presented in the previous section and a number of constitutive models proposed in the literature for the small strain equilibrium modulus will first be presented. Then, the large-strain predictions of a few selected models will be considered in more detail. ‘The purpose of this approach is to first. sort out which models are promising and then examine these models more carefully. Note that no attempt has been made to exhaustively compare all proposed models or to summarize in any detail the different 106Meincke-Tata (198) tension of SBR (N330 cabo back) Meincke-Tafta (1980, compesion of SBR (N330 carbo lack) arwood-Payne (1966, tension of NR (HAF curbon black) ‘Siawood (194) tension of NR (chanel Black) Smallwood (1948), tension of NR (P33) Smallwood (1948), easion of NR (Thera) Saltwond (1948), tension of NR Kado) Smallwood (1944), tension of NR (XX zn oxide) Smallwood (144), tesion of NR (ides whiting) Salwood (1948), nson of NR (Calo ely) “Matiae-Tobin (1965), esion of NR (MT eaton Mack) Malas-Tobin (1968) tesion of NR (HAF carton lack) Berguro-Boyee (1998), compression of CR (N6O0 carbon Hak) Beso Boye (1998, compesion of WR (N35 co Bat) so8xo0o ER, o voeesapa = ‘Normalized Young's modulus, ‘0.05 On 0.15 02 0.25 03 Volume Fraction Filler Figure 4-4: Experimentally determined normalized Young’s modulus E/E, as a func- tion of volume fraction particles. 107approaches used. Instead, a selection of commonly used and/or quoted models have been chosen and a final constitutive equation, simplified to a uniaxial compression loading situation, is stated. For more details the interested reader is referred to the original papers. In the following, the effective volume average composite properties will be denoted by o=y fom, (41) and macroscopic properties will be denoted in the same way as for an unfilled material, for example the effective Young’s modulus will be denoted E. Similarly, volume averages in the matrix phase wil! be denoted by (4.2) where V is the total volume of the unit cell, and Vr, the volume of the matrix phase. The prediction of the equilibrium locus is a composite theory problem that can be approached in two different ways: one can derive rigorous bounds for the behavior or one can try to estimate the overall behavior. One of the first and simplest. estimations is the dilute estimate of Einstein (1906, 1911) who derived the increase in viscosity caused by a suspension of spherical particles in a viscous fluid. The same approach was later applied by Smallwood (1944) to predict the small strain Young’s modulus of patticle-filled solids: E = Em(1 + 2.5v,), where Ey is the Young’s modulus of the matrix material and vy is the volume fraction filler. This estimate however is only good at very low (ie. dilute) filler concentrations. A number of attempts have been made to incorporate interactions between neighboring particles to allow for predictions also at higher volume fractions. Most of these models add one or more terms to a polynomial series expansion of the amplification factor. One of the most cited models of this class is the Guth-Gold model (1936, 1938, 1945) E = Ey(1 + 2.50, + 14.107). (4.3) f 108,A number of variations of this model have also been developed in which the coefficients in the expansion are slightly different (Vand 1948; Simha 1940). A different method to incorporate particle-particle interactions into the Smallwood estimate has been de- veloped by Mooney (1951). In this modei the first order interactions are incorporated through a crowding factor k that needs to be experimentally determined: 2.5uy T= hy E= Bmexp [ (4.4) Note that if this relationship is Taylor expanded it also conforms to the same class as the Guth-Gold model. A similarly structured model has been developed by Leighton and Acrivos (Leighton 1985; Acrivos et al. 1994) for the relative viscosity of suspen- sions, when applied to a linear elastic solid this model can be written: (45) In addition to these models that are restricted to spherical particles, theories have also been developed for other shapes. One of the most used models of this type is the Guth (1945) model E = Ep (1+ 0.67gyu, + 1.62(g;04)"] (4.6) developed for rod-like shapes characterized by, gy, the ratio of the length to the wid. of the particles or the filler aggregate structures into which the particles cluster. This model attempts to account for the fact that particle aggregation (clustering) has a significant impact on stiffness at higher volume fractions (ie. vy Z 0.15); in practice, 94 is typically between 4 and 10 which essentially makes this model similar to the Guth-Gold model but with a much higher coefficient on the quadratic term, thus further amplifying the aggregation or “shape” effect at higher volume fractions. As shown in Figure 4-5, the filled elastomer models of Equations (4.3) to (4.6) do a reasonably good job in predicting the dependence of modulus on filler content when compared to Figure 4-4. In addition to these models specifically developed for filled 109Voigt (1889) upper bound Hashin—Sherikman (1963) upper bound Ponte Castaieda (1989) Guth (1945) (gf=4.7) Mooney (1991) (e=1 63) 7] Budiansky self-consistent estimate (1965) a ‘Acrivos et al (1998) Bergstrim-Boyce (b=34) ‘Guth~Gold (1938) Hashin~Shtrikman (1963) lower bound Dilute estimate (1906) Reuss lower bound (1929) Govindjee~Simo (1991) Normalized Young's modulus, ° ‘0.05 1. O4S 02 0.25 03 Volume Fraction Filer Figure 4-5: Theoretical predictions of the initial Young's modulus, Ey/Em = 100. 110elastomers, a number of general composite theory models can be used. For example, the traditional Voigt (1889) upper bound for a linear elastic isotropic material is given by E=vE,+(1-0)Em (4.7) wheiz Ey is the Young’s modulus of the filler particles; the Hashin-Shtrikman upper bound (Hashin and Shtrikman 1963; Hashin 1983) for the case when the filler particles are almost incompressible? can be en as: 5Em + 30/(Ey — Em) SEy — 2v,;(Ey — Em)” E=K (48) These upper bounds provide little value in bounding the behavior of particle-filled elastomer, :. seen in Figure 4-5, because the filler particle modulus is very high compared to the matrix, and the load must transfer to the filler through deformation of the matrix. Similarly, the corresponding Reuss (1929) lower bound E= (49) and the Hashin-Shtrikman (1963, 1983) lower bound for stiff almost. incompressible particles 3E my + 2Ey + 30,(By — Pm) E = Eq et dt SN To) (4.10) BB, + 2Ey — 2ug(Ey — Em) oe are not very restrictive. A number of other composite theory estimations have also been developed: a self consistent estimate developed by Budiansky (1965) which for the special case of rigid This derivation assumes the bulk modulus of the filler particles, Ky, to be much larger than the shear modulus of the filler particles, yy. The end result without this assumption is nearly the same although the equation is substantially more invoived. 1particles in an incompressible matrix can be written as: En B 2.5" (4.1) A different self-consistent estimate for rigid particles in a neo-Hookean matrix has been developed by Ponte Castafieda (1989): E = 30; E (4.12) which can also be rewritten to take the basic amplification form of Equation (4.3). This model predicts the data reasonably well as shown in Figure 4-5, but. begins to overpredict the stiffness at filler concentrations greater than 0.25. The Mori and Tanaka (1973) estimate for spherical particles isotropically dispersed in an elastic matrix can be written uy(Ky — Km) Km K= Kat eee (4.13) Taek) KaiKn Waa + Km a Uy (Hy = Hm) Hm (4.14) + 1s = tnd _—_ (1 = Uy) (v4) ~ Hn) Say + Hm where K is the bulk modulus and 1 the shear modulus. Note that when this esti- mate is applied to stiff almost incompressible particles it coincides with the Hashin- Shtrikman lower bound. Another approach, based on the concept of amplified strain, has been developed by Govindjee and Simo (1991) which for the case of rigid particles in a neo-Hookean matrix can be written v,/2 =u E (4.15) which is weaker than the Reuss lower bound as is shown in Figure 4-4, and, indeed, is acknowledged to be a “loose lower bound” in Govindjee (1997). 124.3.1 Large Strains and Strain Amplification A serious drawback of most of these models is that they only predict the behavior at infinitesimal strains. To predict the behavior at large strains Mullins and Tobin (1965) introduced the notion of amplified strain which in the case of uniaxial loading they quantified as A = 1+ X(A—1) where 2 is the applied axial stretch, A is the amplified stretch, and X is a constant depending on the volume fraction filler (for the Guth-Gold model X = 1 + 2.5vy + 14.1v}, and for the Guth model X = 1+ 0.679;u; + 1.629703.) Following their procedure, the stress-strain relationship for a particle filled elastomer is then obtained by replacing A by A in the constitutive relationship for the matrix.’ For a neo-Hookean matiix this provides the stress- stretch relationship ¢ = m(A? — 1/A). In the casz of the Govindjee-Simo model, the large strain behavior is obtained by inserting the amplified deformation gradient (E)m = ((F) — y(R))/(1— vp) into the expression for the strain energy density of the matrix material. For uniaxial loading the expression for the composite strain energy density becomes (y=) |e Ay {o-mts2(H-s) }-3]. (4.16) giving the stress-stretch relationship o = pa- vy) - . (4.17) ala-™) Each of the above models contains some ambiguity in how a general strain state is to be amplified; for example Govindjee and Simo suggest amplifying the total deformation gradient, whereas Mullins and Tobin discuss amplification of the uniaxial strain. Here, a new approach (Bergstrém and Boyce 1999) is proposed whereby the first invariant of the stretch, J, = (A} + 43 + 3)'/? (which can be considered to be an 3As noted by Govindjee (1997), however, this procedure does not clearly define which stress ‘measure is obtained from the modified constitutive relationship. Here, the true stress has been used. 113average scalar measure of the overall strain) is the strain measure amplified. For a neo-Hookean matrix with semi-rigid particles this approach gives the average strain energy density (W) = (1— 49)(W)m = (1 0)" (3), (4.18) where (L:)m = X(/i—3)+3 is the /,-amplification relationship with X taken to be the amplification factor. This gives an initial modulus of E = Em(1— vy)X. The exact form of X is not clear, however, the initial modulus results of others as discussed earlier suggest X = 1-+ avy + bu?. Using the Smallwood end Einstein derivations for modulus amplification E = Ey,(1+2.5v,) for dilute concentrations gives a = 3.5. Then based on the results of Guth and Guth-Gold, the quadratic coefficient 6 is estimated to be 34 for the chloroprene rubber tested. This gives: W= (1-9) + 3.5uy + bv9)[h — 3), (4.19) and Hm(1 — vy)[L + 3.5u, + bvj] (*- ). (4.20) From Equation (4.19) it is clear that ,-amplification is equivalent to what might be interpreted as modulus amplification for the neo-Hookean material. However, the proposed model has a further advantage in that it can be applied not only to a neo- Hookean matrix but to any higher order J-based hyperelastic material model (Arruda and Boyce 1993; Boyce 1996; Yeoh 1993). To obtain the corresponding relationship for an eight-chain material start with the strain energy density for the matrix phase: (W() ms S.8)) oy = PH Pm n (282) + 5 © [J = 17 + pocO[1 — In 6] + Wo, (4.21) 114where (*)m is the average chain stretch in the matrix material obtained by amplifi- cation OO) im = (4.22) and where B =21(Ge) : (4.23) ‘The Cauchy stress is then given by 11 OW dey prs (4.24) J 3° aw as Inserting (4.21) into (4.24) then gives 8 to LX) m/ A) PaO XG Tq PUI) dev[B*] + (1— vy) [J — 1]. (4.25) ‘The ability of this model to capture the effect of filler content on initial modulus is shown in Figure 4-5 along with the models discussed earlier. This new approach of amplifying 1, also provides the advantage of easily being used to describe any arbitrary imposed deformation as will be demonstrated later. In order to provide a better quantitative feel for how well each of the large strain models (Guth and Guth-Gold with Mullins-Tobin kinematic assumption, Govindjee- Simo, Ponte Castafieda, and Bergstrém-Boyce model using I)-amplification of the eight-chain mode!) work, they have been compared in Figures 4-6 to 4-8 to the equilib- rium uniaxial compression data of the chloroprene rubber presented in Chapter 2. In Figure 4-6 is shown the stress-strain behavior when filled with 7 vol% filler. The figure shows that all four models give good predictions at small strains, but at large strains the model by Guth predicts too fast an increase in stiffness with applied strain. The corresponding situation when filled with 15 vol% filler is shown in Figure 4-7. From this figure it is clear that the Govindjee-Simo model prediction is much more com- 115uh estimate afa47) ‘etgarom-Boye (4, anLock=25) Past Coated © Ehloropree rubber (7 vol cb N60) y 2 Gbuinleessina os} iil com 1, -57 MPa 01 -02~O SOS True Sain Figure 4-6: Comparison between theoretical predictions and experimental data for a chloroprene rubber with 7 vol% carbon black. pliant than the actual data. This observation is in agreement with Figure 4-5 which demonstrated that the small strain Young’s modulus as predicted by the Govindjee- Simo model is much lower than what is observed in most elastomers. Figure 4-7 also shows that the Guth model very nicely predicts the small strain behavior but is much too stiff at large strains. Both the models by Ponte Castafieda and Bergstrém-Boyce, on the other hand, give fair predictions at all levels of applied strain. In Figure 4-8 is shown the behavior when filled with 25 vol% filler. As in Figure 4-7, the Govindjee- Simo model is too compliant and both the Guth and the Guth-Gold estimates are too stiff at large strains. In this case the model by Ponte Castaiieda also predicts an overly stiff response at large strains, but the Bergstrém-Boyce model still provides a good prediction. A direct comparison between the Bergtrém-Boyce model and the three different. chloroprene rubbers tested is shown in Figure 4-9 illustrating very good agreement between the Bergstrim-Boyce model and the experimental data both for different filler concentrations and finite strain deformations. The figure also shows that the transition of the experimentally observed equilibrium stress-strain response 116‘Guth estimate (gf=4.7) / ecm? Sera rnin, Ponte Cutateda aw “or 02-03. aS OS 078 True Stain Figure 4-7: Comparison between theoretical predictions and experimental data for a chloroprene rubber with 15 vol% carbon black. with filler concentration from being strictly convex to having a more sigmoidal shape is not captured by the proposed model. This change in behavior, which is as also discussed in Section 2.11, is not captured by the model because the underlying eight- chain model does not capture it and the proposed model is based on a amplification of the strain state in the eight-chain model. To further test the applicability of the large strain models (Guth and Guth-Gold with Mullins-Tobin kinematic assumption, Govindjee-Simo, Ponte Castafieda, and Bergstrim-Boyce model using 1)-amplification of the eight-chain model) they are compared in Figures 4-10 to 4-12 to the equilibrium compression data of the natural rubber presented in 4-3. In Figure 4-10 is shown the stress-strain behavior when filled with 7 vol% filler. The figure shows that the small strain behavior is well predicted by all models, but that for larger strains the Govindjee-Simo model is too compliant and that the Guth estimate is too stiff. The behavior when filled with 17 vol% filler is shown in Figure 4-11 illustrating that only the Bergstrém-Boyce model and the Ponte Castafieda model are in good agreement with the experimental data. In Figure 4-12 174 uniaxial compression oO BTM Gthestinate a7) Sah Gord ehoe Bergrem=Boyee (3, antock=2 5) tirnene ber (2 we NOD) ovine: So. 4-05-06 OS OB ‘Tre Stain Figure 4-8: Comparison between theoretical predictions and experimental data for a chloroprene rubber with 25 vol% carbon black. is shown the behavior when filled with 24 vol% filler. In this case the model by Ponte Castafieda also predicts an overly stiff response at large strains, but the Bergstrém- Boyce model still provides a good prediction. A comparison between the experimental data for the natural rubber and predictions from the Bergstrém-Boyce model, shown in Figure 4-13, demonstrate good agreement at all three volume fractions of filler. As has been shown in Figures 4-6 to 4-8 and Figures 4-10 to 4-12, the model by Govindjee and Simo does not give good predictions of the observed stress-strain data. ‘The work by Govindjee (1997) further includes the development of a model for the average strain in the matrix which has been shown to be in good agreement with two- dimensional finite element analysis. As will be shown in the next section, however, the effective amplified strain is different in two- and three-dimensional simulations. From Figures 4-6 to 4-8 and Figures 4-10 to 4-12 it is also clear that the strain amplification concept used in both the Guth and Guth-Gold estimates is not working very well at large strains. Figures 4-9 and 4-13 furthermore show that the experimental data for both the chloroprene rubber and natural rubber supports the Bergstrém-Boyce 118Figure 4-9: Comparison between theoretical predictions and experimental data for three different chloroprene rubber filled with carbon black N60. J,-amplification concept. The amplification of /; permits the study of other strain states. In Figure 4-14, the Bergstrém-Boyce prediction of the dependence of strain for both uniaxial and plane-strain compression are shown. The experimental data in the figure is the predicted equilibrium stress and is in good agreement with the predicted equilibrium behavior. One last example of J amplification is shown in Figure 4- 15 illustrating the simple shear response at three different filler concentrations. As demonstrated in the figure, the agreement between predicted and experimental data is very good. 1193 Guth estimate (4.7) ‘Guth-Cold extinate (6225, lamLocks?.1) as} iN ‘natural rabber (vol cb N3S1) Govindjee-Simo i Figure 4-10: Comparison between theoretical predictions and experimental data for a natural rubber with 7 vol% carbon black. Gurestinae(gf=8.7) | GuGoldecamate otra robber (17 volt cb N3S1), Bergsuom-Boyce(b=29,amLack=2.1) 9 Ponte Cartateda Govingje-Simo wi ogre “1 -02~-03 os OSC OSCSC OB ‘Tree Stain Figure 4-11: Comparison between theoretical predictions and experimental data for a natural rubber with 17 vol% carbon black. 120“True Suess (MPa) Guth eximat gf=8.7) ‘GuhGold esumate Ponte Casaheda Bergsut Boyce (b=25,lamLock=2 1) iatral aber (24 vt cb SST) Govindje-Simo Figure 4-12: Comparison between theoretical predictions and experimental data for a natural rubber with 24 vol% carbon black. L & aca ance tos BAS Bm? 4 o € z-25| a & -s 7 wie. 5 vat compen cuss or aT) ay ea ae Troe Sin Figure 4-13: Comparison between theoretical predictions and experimental data for three different natural rubber filled with carbon black N351. 121upc da: lanes loading Planes ce pedi feild ayee prediction: unital loading in loading Exper dan Betgstom-Bc 4 a4 a g 7 oa bes oo fa “ “4 a ctropen ber os son yates ne 1-05 Pte 9 a a Tne iain Figure 4-14: Model prediction vs. experimental data for the stress-strain behavior in both uniaxial and plane-strain loading situations. 07, | 17 vos cb “Tre Stress, T,, (MPa) ‘natural rubber simple shear Hy =05 MPa, lam s 12 03 oF 05 Enginering Stan, E,, Figure 4-15: Model prediction vs. experimental data for the stress-strain behavior in simple shear. 1224.4 Finite Element Simulations of Particle-Filled Elastomers From the previous section it is clear that only a few models can adequately pre- dict the observed small strain uniaxial behavior of different particle-filled elastomers. ‘The corresponding situation for large strain deformation including considerations of nonlinear effects such as hysteresis and time-dependence is even less satisfactory. In this section the potential for using finite element (FE) simulations as a tool in the development of better constitutive models will be discussed. The starting point for this investigation is the evaluation of the ability to predict the equilibrium modulus using FE micromechanical representations of the filled elastomer. Only if it can be demonstrated that this simple loading situation can be accurately simulated will it be justified to consider additional nonlinear effects using these techniques. Even for a simple loading situation such as small strain uniaxial compression, a number of inter- esting issues can be investigated. For example, it is possible to quantify the influence of different filler structures. If the particles are taken to be spherical and perfectly dispersed then the simulation results should coincide with the Smallwood (1944) and Guth-Gold (1938) predictions, provided the filler concentration is small enough. This can then be compared to a random distribution of particles having a higher intrin- sic structure. It is also possible to evaluate the strain amplification concept as has been done by Govindjee (1997) by use of two-dimensional FE simulations. As will be shown in the following section, however, there is a significant difference between two- and three-dimensional simulations. The possibility to achieve large strain predictions and generalizations can further be facilitated by the detailed information available from this type of simulation. As a first step in this investigation it is necessary to determine what type of FE model is most appropriate. The larger the simulation, in terms of both the number of filler particles modeled as well as the number of finite el- ements used, the better the results become; however, the computational expense will eventually restrict the simulation size that can be run in a realistic amount of time. In the next section a number of different types of simulations have been compared, 123the main objectives of this exploration were: (1) to quantify the difference in behav- ior between a unit cell containing one particle and a unit cell containing a number of randomly distributed particles, when both unit cells have the same area/volume fraction particles; (2) to quantify the difference in behavior between two-dimensional and three-dimensional loading situations at the same area/volume fraction filler parti- cles; (3) to quantify the influence of area/volume fraction particles on the mechanical response; (4) to quantify the influence of different particle shapes, where shape is an issue due to the tendency of particles to cluster into irregularly shaped aggregates; (5) to quantify the influence of different boundary conditions; and finally, (6) to try to determine whether the stiffening effect of filler particles can be explained using this approach. By comparing the predictions from the different numerical techniques with the known experimental behavior presented above, it is possible to determine the level of ‘sophistication’ that is needed in order to, if possible, reproduce the experimental data. Results from several models that did not correlate well with experimental data are also included in order to illuminate the problems that can occur with simulations of this kind. The presentation of the results starts with trivially simple FE simulations and concludes with large expensive three-dimensional stochastic simulations. In the simulations, unless otherwise stated, the matrix is modeled as a neo-Hookean material with a Young’s modulus of Em = 2 MPa and the filler particles are modeled as a linear clastic material with B;/Em = 100, and Poisson’s ratio 1 = 0.3. The elements used in the simulations are based on quadratic interpolation functions, and the composite Young's modulus was approximated by the secant modulus at an applied strain of «= -0.002. 4.4.1 Two-dimensional Simulations The simplest possible FE approach to model a particle-filled composite is to consider a two-dimensional representative volume element (RVE) containing one particle. One of the more important issues that needs to be addressed in this case is the choice of particle shape. A first example is shown in Figure 4-16 illustrating the influence of 124different particle shapes on the stress-strain behavior for a very coarse mesh, and in Figure 4-17 is shown how the normalized Young's modulus depends cn the area fraction filler for a more refined mesh. In both of these figures the boundary of the RVE was constrained to remain rectangular. In all simulations the stress was “is a2 Troe Stn “os Figure 4-16: Influence of different particle shapes on two dimensional simulations of a RVE containing one particle. Plane stress loading, area fraction filler ay = 0.15. calculated by tf. T=7f, ;bet+tenld, (4.28) V Joy 2 where V is the volume, p is the position vector, and ¢ the traction vector. In the case of uniaxial loading this equation reduces to the total reaction force on the top surface divided by the area of the top surface. Results from the coarse mesh are included as a comparison basis with the more refined mesh, and also, the coarse mesh will be used later as a building block in stochastic simulations with a RVE containing many particles. The two figures show that particles with larger radius of gyration give a stiffer response, and that the initial modulus is rather strongly dependent on 125Normalized Young's Modulus. or 15 02 025, ‘Area Practon Filler Figure 4-17: Influence of different particle shapes for a more refined mesh. Plane stress loading. the particle shape. It is further clear that these models significantly underpredict the mechanical stiffness compared to the experimental data in Figure 4-1, where for example, a real elastomer filled with 15 vol% filler is typically about 100% stiffer than when unfilled, whereas the two-dimensional simulations in Figure 4-17 show at most a 40% increase in stiffness. One of the input parameters that is needed in the simulations is the stiffness of the filler particles. It is well known that most filler particles that are used (e.g. carbon black) are significantly stiffer than the elastomeric matrix. The actual value that is assigned to the filler particle stiffness is not very important in this case, as is shown in Figure 4-18, as long as it is chosen large enough. Since the two-dimensional simu- lations with a RVE containing one particle underpredict the stiffness of the material (compared to experimental data), it is of interest to examine the behavior of cor- responding stochastic simulations with a RVE containing many particles. In Figure 4-19 is shown the influence of mesh density on the normalized Young’s modulus for a mesh with quadrilateral particles. The mesh density used in Figure 4-19 is defined as 126er i Figure 4-18: Influence of particle stiffness on the stress-strain behavior. Plaue stress loading, shape parameter C; = 0.8, area fraction filler ay = 0.15. Las 14 g 135 ER, boo [Normalized Young’s Modulus, ump fat sep nth lipase 03) SOS (aomty 6) (iy 00) ‘0200400600 8001000 1200 1400 160018002000 Mesh Density Figure 4-19: Influence of mesh density on the normalized Young’s modulus. Plane stress loading, quadrilateral particles, area fraction particles ay = 0.15. 127the total number of possible particle sites per unit RVE. From this figure it is clear that the size the RVE is not very important as long as it is larger than a lower limit. In Figure 4-19 and in all stochastic simulations to follow (unless otherwise stated) each data point corresponds to a phase space average from 50 Monte Carlo simulations in a manner similar to that used by Govindjee (1997), and the errorbars represent the uncertainty in the predicted data calculated from Evrror /((E*) — (E)")/N, where (-) symbolizes an ensemble average. In all stochastic simulations repeated boundary conditions allowing deformed lateral sides were used as illustrated in the inset of Figure 4-19. The influence of different boundary conditions were also inves- tigated, the predicted stiffness for the different cases, however, converged as the size of the RVE increased. The repeated boundary conditions illustrated in Figure 4-19 impose the least additional constraint and therefore allows for accurate predictions for smaller representative volume elements than the case when the sides, for example, are constrained to be straight. To investigate the importance of particle shape, octalateral particles were also simulated. The predictions in Figure 4-20 show that the cell with octalateral particles has a slightly more compliant response than the cell with quadrilateral particles. The main reason for the more compliant response is that the mesh is more refined around the particles allowing for easier deformation. By comparing the type of one-particle simulation shown in Figure 4-18 for which E/Ey, * 1.66 at the area fraction ay = 0.24 with the results from the RVE containing many octalateral particles shown in Figure 4-20 in which E/Eyy = 1.78++0.01 at the same area fraction, it is clear that the many- particle simulation predicts a higher stiffness than the simple one-particle simulation. ‘The reason for this is that in many-particle simulations some of the rubber becomes occluded and cannot deform, thus making the area fraction of particles effectively larger. This is illustrated in Figure 4-21 by the contours of the first stretch invariant dt, = +3 +3. The white regions in the graph are virtually undergoing no deformation due to the geometric constraints resulting from the particle aggrogation. Finally note that the predicted modulus in all the two-dimensional simulations are much lower than the experimental data presented earlier. This is the reason it is not 128ek, Normalized Young's Modulus, 0 005 a O15) 02 025 ‘Area Fraction Particles Figure 4-20: Influence of area fraction of particles on the normalized Young’s modulus. Plane stress loading. Figure 4-21: Example contours of the first strain invariant /,. Plane stress loading, octalateral particles, shape parameter C2 = 0.8, area fraction filler ay = 0.24, applied strain € = —0.2. 129necessary to refine the mesh further in the stochastic two-dimensional simulations, since refinement would only make the response even more compliant. 44.2 Three-Dimensional Simulations From the previous section it is clear that two-dimensional simulations do not correlate well with experimental data. The next level of simulations are three-dimensional using an axisymmetric unit cell of the type shown in Figure 4-22. Note that in == cts = Guih-Gols eximae Te cisio sst}[-S-_ cio? 107 nom ‘Normalized Young's Modulus, E/E, t 5 a +8 b * ° 005 On 015 02 025 Figure 4-22: Axisymmetric three-dimensional simulation with one particle illustrating the influence of volume fraction of particles on the normalized Young's modulus. The parameter Cy is defined in Figure 4-17. these simulations the particles are ellipsoidal and the RVE is a cylinder which is a shape that cannot be packed to fill space. The figure shows that the simulation result for spherical particles agrees well with the Guth-Gold model (1938) up to a very large filler concentration. And as will be shown in Figure 4-26 when compared with the experimental data, the Guth-Gold model can be used as a practical lower bound for the reinforcement independent of filler used. In Figure 4-22 also note 130that the axisymmetric simulation examples with non-spherical particles obviously possess anisotropic stiffness properties. This observation holds in general when the particles are anisotropic in shape even though the composite might contain a random distribution of particles. Anisotropic composite stiffness properties can also occur even if the particles have isotropic shapes if the particle aggregation is anisotropic. In Figure 4-23 are shown results both from the two-dimensional (plane stress) and the three-dimensional (axisymmetric) simulations. Here it is clear that the Ter Relfpamentc C1510 <3. Sethe pate (0.7) Normalized Young's Modulus, E/E, Figure 4-23: Comparison between predicted Young’s modulus for two-dimensional and three-dimensional simulations. two-dimensional simulations greatly underpredict the stiffness when compared to the three-dimensional simulations; this is a direct result of the area nature of the two- dimensional volume fraction as discussed in Appendix B. The effect of the randomly distributed nature of the filler particles is explored in the stochastic three-dimensional simulations shown in Figure 4-24 ill'strating the pre- dicted behavior of a RVE containing cubic particles, a RVE containing cubocahedral particles, and the axisymmetric data from Figure 4-22. As before, the simulations summarized in Figure 4-24 were performed with repeated boundary conditions allow- 131eerie isi) oy ame eas) Normalized Young's Modulus, E/E, mn 015, 02 025 Volume Fraction Filler Figure 4-24: Predicted influence of volume fraction particles on the normalized Young’s modulus from stochastic three-dimensional simulations. Each data point is the stochastic average of 10 simulations. 132ing the sides to deform. The figure shows that the predicted stiffness is significantly higher than the axisymmetric prediction in Figure 4-22. This is reasonable since randomly distributed particles will cluster into irregularly shaped aggregates causing more of the rubber matrix to become occluded then for the case of perfectly dispersed particles. The figure furthermore shows that the cubocahedral particles give a more compliant response than the cubic particles especially at high volume fractions, the reason for this being the more refined mesh. The influence of different mesh refine- Dir Ba “BOT A 905 os aT -O0e GH OY ON ‘Tre Sean Figure 4-25: Mesh refinement study of a RVE containing cubocahedral particles, vy = 0.1. ments is more systematically shown in Figure 4-25 where the stress-strain behavior of a RVE containing a fixed distribution of octalateral particles with the mesh of the matrix phase being refined to different degrees. The figure shows that the stress-strain behavior converges as the matrix phase is refined; the medium level refinement was deemed sufficient for this study. A third type of simulation with rhombic dodecahe- dron‘ shaped particles of the type illustrated in the inset in Figure 4-26 have also been performed. A direct comparison between the different stochastic three-dimensional simulation results and the experimental data, illustrated in Figure 4-26, reveal that “The rhombic dodecahedron is the dual polyhedron of the cubocahedron and is a space filling Archimedean solid. 133the simulations with cubic particles produce predictions that approximate the upper limit of reinforcement observed in the experiments, whereas the cubocahedral and thombic dodecahedral particles give a slightly more compliant response. The results also demonstrate that consideration of a random distribution of particles (as opposed to perfect dispersion) is also necessary to adequately account for the effect of filler on stiffness. The data in Figure 4-26 suggest that the main reason for the increased = Cubic panictes == Cubocahedral particles ‘Rhombic dodecahedral particles ~~ Guth-Gold prediction © Experimental data oe [Normalized Young's modula, 005 O ous 02 025 ‘Volume Fraction Particles Figure 4-26: Comparison between FE simulations and the experimental data for filled elastomers presented in Figure 4-4. Each data point is the stochastic average of 10 simulations. stiffness observed in particle-filled elastomers is due to the composite interaction be- tween the fillers and the matrix and not due to changes of the matrix properties due to the fillers. As discussed earlier, the impact of fillers on elastomeric behavior is also often discussed in terms of some form of strain amplification factor. ‘The concept of strain amplification can be investigated with the micromechanical models by ex- tracting measures of the average strain in the matrix. Here we select two measures to explore: the first measure is the average first invariant of stretch in the matrix 134(hi)m = (tt[B])m = (43 + 43 + A3)m as a function of applied strain; the second av- erage measure consists of the average principal stretches in the matrix (A1)m; (A2)m: (As)m a5 a function of applied strain. These quantities have been extracted from the micromechanical models and plotted in Figures 4-27 to 4-32 for the case when vy = 0.15. The amplification of J, is shown in Figures 4-27 to 4-29 for uniaxial com- pression, plane strain compression, and simple shear, respectively. The Bergstrém- Boyce model shown in the figures was proposed earlier in Equation (4.18) and was based directly on /1-amplification: (hi)m = [1 + 3.5uy + bog] (, - 3) +3. (4.27) A direct comparison between the micromechanical simulations and the -amplification model indicate excellent agreement in all three loading modes, further supporting the newly proposed approach. The average principal stretches are shown in Figures 4-30 Beare ga To- SOREN mb fodedtedal particles) NNowmlized in the Marx (,), 1 Uniaxial Compression ‘0 00s 01 —-0NS 02-0253 035-04 ‘Tre Stain Figure 4-27: J,-amplification in the matrix as a function applied strain. Uniaxial compression, vy = 0.15 to 4-32 for uniaxial compression, plane strain compression, and simple shear, respec- 135Nomad inthe Mati. z 5 Plane Surin Loading ~005 4, “as 02 “eas T Tne Senin Figure 4-28: J,-amplification in the matrix as a function applied strain. Plane strain compression, uy = 0.15. 1 a SSB Gatti Podedneda prices 1 & 8 Normalize I inthe Matix, 8 ‘Simple Shear Loading, ‘0 or 02 4 05 06 03, ct nginering Stain, E Figure 4-29: J-amplification in the matrix as a function applied strain. Simple shear, vy = 0.15. 136tively. The Govindjee and Simo model is based on the amplification of principal Uniaxial Compression 000s —~COY SSCS C20 SOS “Tre Strain Figure 4-30: Amplified principal stretches in the matrix, uniaxial loading, uy = 0.15. Five simulations with dodecahedral particles (mesh density 96). stretches in the matrix where the amplified stretches are given by: Ary Ode = , £21,238. (4.28) ‘The Govindjee-Simo model predictions for amplified principal stretches as a func- tion of applied strain are compared to the micromechanical results in Figure 4-30. While there is good agreement for the first principal stretch (A1)m, there is poor agreement with the other two principal stretches. Indeed, since the Govindjee-Simo model provides a direct amplification of the macroscopic principal stretches, two of the amplified stretches are predicted to be the same for this case of uniaxial compres- sion and these predictions are in error. Indeed, the micromechanical results suggest ‘a more interesting picture as to the local flow of the elastomer, revealing the effect, of the particles to locally somewhat constrain the flow in one local direction causing the local deformation state to be close to plane strain. These results further support, 137a ‘Average Principal Sueices inthe Matrix. 0 005 “ar ~015 =z “02s True Strain Figure 4-31: Amplified principal stretches in the matrix, plane-strain loading, vy = 0.15. Three simulations with dodecahedral particles (mesh density 96). ‘Average Principal Suetches in the Matrix Simple Shear om 02, 05 06 03, rn Engincering Strain, E, Figure 4-32: Amplified principal stretches in the matrix, simple shear loading, vy = 0.15. Two simulations with dodecahedral particles (mesh density 96). 138the use of amplifying an effective stretch quantity such as (/)m as proposed in this paper. I: is also possible to study the average rotations of the filler particles during the deformation. Govindjee and Simo in developing their model assumed that the filler particle rotations are affine, an assumption that can be tested using the RVE shown in Figure 4-33. In part (a) of the figure is shown the mesh used in the simulation, part ZS S54 LSS S> 2S (a) (b) Figure 4-33: Uniaxial deformation of a RVE containing 15 vol% filler. The matrix is modeled with the eight-chain model. (a) Mesh used in the simulation, (b) filler particle distribution. (b) shows the distribution of filler particles inside the RVE. The deformed particle distribution when tested to a final strain of —0.2 is shown in Figure 4-34. The dif- ference between the average rotation tensor of the filler particles (R); obtained from the micromechanical simulation and the applied rotation tensor R. (which here is the identity tensor since the applied deformation was uniaxial) are shown in Figures 4-35 and 4-36. Figure 4-35 shows three components {ARj, ARis, ARos} of the tensorial difference and Figure 4-36 shows the norm of the difference. ‘The figures show that on average the rotations of the filler particles are very small. One example of the distribution of one rotation tensor component, here (Rj2)/, is shown in Figure 4-37. ‘The figure shows that there is significant variation in the rotations of different filler 139Figure 4-34: Deformed particle distribution of a RVE containing 15 vol% filler par- ticles tested in uniaxial compression. The matrix is modeled with the eight-chain model. 0.05 -R 2 & $ Component of (R) 00s} vor] . aR . aR an Uniaxial Compression sy “005 =n ~ais “2 True Stain Figure 4-35: Components of the tensorial difference between the average rotation te! sor of the filler particles (R), and the applied rotation tensor R. Uniaxial compression of a RVE containing 15 vol% filler particles. 140ons = singular velue Indy norm os Frobenius norm 02s} § Tensor NomiR},— R) oot 0.005 2 Uniaxial Compression 0 005 01 ‘Tre Sin “18 “a2 Figure 4-36: Tensor norms of the difference between the average rotation tensor of the filler particles (R), and the applied rotation tensor R. Uniaxial compression of a RVE containing 15 vol% filler particles. ou [Uniaxial Compression ou ou on] Relaive Populaion gos g 04] 0.02] bis 01 OSS Rotation, R,), Figure 4-37: Distribution in (Riz), for a RVE containing 15 vol% fillers tested in uniaxial compression to a final strain of —0.2. 141particles but that on average the rotations almost cancel out. Using the same type of micromechanical simulation it is also possible to study the behavior in simple shear. ‘The FEM mesh and particle distribution for the case when the volume fraction filler Particles is 15% is shown in Figure 4-38, and the displaced mesh when deformed to an engineering strain of Ey2 = 0.61 is shown in Figure 4-39. The resulting stress-strain response is presented in Figure 4-40 together with the response of the corresponding unfilled system. The amplification of the Young’s modulus is this case becomes 2.22, in good agreement with the )-based amplification concept presented in Equation (4.25) giving X = 2.29 for b = 34 and the experimental data in Figure 4-4. The difference between the filler rotations and the applied rotations, as shown in Figures 4-41 and 4-42, is relatively small. The distribution in (Rj2),, as exemplified in Figure 4-43, again illustrates that different filler particles undergo rather different rotations. It is not necessary, however, to explicitly take the distribution of filler particle rota- tions into consideration when predicting the macroscopic stress-strain response. The continuum level homogenization based on J;-amplification, which implicitly takes the rotations into consideration, is sufficient for predicting the behavior. A final illustration of the FE predicted behavior is shown in Figure 4-44 where the finite strain stress-strain predictions for three different volume fractions of filler par- ticles are depicted. In all micromechanical simulations rhombic dodecahedral shaped particles were used. As illustrated in the graph, the simulations of the 7 and 15 vol% filled elastomers are in very good agreement with the experimental data. However, the simulation of the elastomer filled with 25 vol% filler is stiffer than what was ob- served in the experiments. One reason for this could be that when this highly filled material is subjected to finite strains, the matrix will undergo severe deformations which require a very fine finite element mesh in order to be accurately modeled. Also, as the particle density increases, the overall size of the RVE must increase (i.e. must contain many more particles than included here) in order to adequately address the random dispersion of particles and thus accurately account for the continuity of ma- trix regions. Also, at this high filler concentration the simulated system wi!l be close to the percolation limit which for spherical particles has been shown to be in the 142© Figure 4-38: FEM mesh and particle distribution used to study the simple shear Fesponse of a RVE containing 15 vol% filler particles, (a) perspective view showing the mesh, (b) perspective view showing the particle distribution, (c) side view showing the particle distribution. 143Figure 4-39: Perspective view of the deformed particle distributions. fem (15 vale fen Eyton ani ‘Tre Suess, 0, hy 02023 03 035 ua vas aos 01 Ons Engineering Stain, E,, Figure 4-40: Stress-strain response in simple shear of a RVE continuing 15 vol% filler particles, pr = 0.67 MPa and Al = 2. 144x10” R23 Simple Shear 00s 0101S 02 025 03 035 04 OMS Engineering Shear Stain, E,. 20) 0 Figure 4-41: Components of the tensorial difference between the average rotation tensor of the filler particles (R)y and the applied rotation tensor R. Simple shear of a RVE containing 15 vol% filler particles. 02s i oot 00s} 0005 01 015 02 025 03 035 On ous Engineering Shear Sinn, Figure 4-42; Tensor norms of the difference between the average rotation tensor of the filler particles (R), and the applied rotation tensor R. Simple shear of a RVE containing 15 vol% filler particles. 145028” 3 4 9 Simple Shear od 0.05} Sos 0 06s 01 als 02 025 03 035 04 04s Rotation, (R,) Figure 4-43: Distribution in (Rj), for a RVE containing 15 vol% fillers tested in simple shear, Riz = 0.20. range (0.22, 0.28] (Kliippel and Heinrich 1995). 146Hoa hi incre Sea | igre ime inn For Seema Pema Sea emia ee ay Y -002 008 -006 008 01-012 -014 016 018 -02 ‘True Sain Figure 4-44: Example stress-strain response for a RVE with rhombic dodecahedral particles. Each data point is the stochastic average of 8 simulations, p= 0.57 MPa, lock = 1474.5 Conclusions ‘The incorporation of filler particles in elastomers is a widely known and used method to tailor the stiffness properties of the material. Although numerous investigations have been performed in an effort to deepen the understanding of the mechanisms caus- ing the change in behavior, a broad foundation between micromechanisms, continuum models and experimental observations is far from complete. This first modeling at- tempt has focused on the equilibrium behavior, recognizing that the investigation of additional nonlinear effects such as time and temperature dependence is greatly facil- itated by a clear understanding and successful modeling of the equilibrium behavior. The mechanical response, predominantly in the small deformation regime, has been studied by using the finite element method. It has been shown that it is possible to get quantitative agreement between experimental data and FE simulations. In particular, it has been shown that the correlation between volume fraction filler par- ticles and the initial Young's modulus of the material can be accurately modeled by stochastic simulations of three-dimensional representative volume elements contain- ing many filler particles. For example, the stiffening effect of low volume fractions of reinforcing fillers, which are known to follow the Guth-Gold (1938) relationship, can be accurately simulated by a system containing perfectly dispersed particles. Further- more, the microstructure corresponding to the experimentally observed upper limit of reinforcement for a given volume fraction can be represented by a random distribution of particles. These observations suggest that the main stiffening effect from the filler particies is due to overall composite behavior and not due to a change of the elastic properties of the matrix material due to the incorporation of the filler particles. The need to introduce additional crosslinking sites due to the filler particles as has been proposed in some work in order to fit experimental data seems not to be justified as the simulations presented in this chapter agree well with the experimental data. As a final conclusion, both Guth’s suggestion (1945) to characterize the filler ag- gregates with a shape factor and Medalia’s (1994) occluded rubber concept can be considered valid approaches for the modeling of the small strain equilibrium behavior 148of particle-filled elastomers. The large-strain equilibrium behavior can similarly be modeled by a concept similar to that of strain amplification suggested by Mullins and Tobin (1957). The new concept is based on J,-amplification instead of strain amplification and has been shown to give better overall predictions of the dependence of stress-strain behavior on filler content from small to large applied strains and has also been dirzctly shown to capture the behavior in modes of deformation other than uniaxial loading (plane strain and simple shear). Fully three-dimensional finite el- ement based micromechanical simulations have further been shown to support the Jy-amplification model and to be in good agreement with experimental data. These results suggest that the concept of /,-amplification might also prove useful in predict- ing the dependence of Mullins effect, hysteresis and other time dependencies on filler content. Furthermore, as demonstrated in this chapter, the ability to successfully capture the dependence of stiffness on filler content through fully three-dimensional micromechanical models of randomly dispersed filler particles lays the groundwork for using micromechanical modeling to study the more complex issues of large strain nonlinear rate and temperature dependent behavior as well as hysteresis. 149Chapter 5 Time-Dependence of Unfilled Elastomers 5.1 Introduction ‘As was shown in Chapter 2, the mechanical behavior of elastomeric materials is time- dependent, often causing significant hysteresis when subjected to cyclic loading. In some cases, particularly when the interest is focused on long time scales, it is sufficient to model the equilibrium response of the elastomer using a hyperelastic framework of the type discussed in the last two chapters. But frequently it is desirable to have a more detailed model that can also predict the time-dependent response. As a first step towards this goal, the time-dependence of unfilled elastomers will be discussed and modeled, and the more complicated problem of time-dependence of particle filled elastomers will be left for Chapter 6. To predict the observed mechanical behavior numerous constitutive models focus- ing on different aspects of the elastomeric response have been developed by different investigators throughout the years. Most early work focused on the isothermal equi- librium behavior but there has been a continuous effort also on modeling the time- dependent behavior. The micromechanical understanding and constitutive modeling of the various non-linearities pertaining to time-dependence, however, are not nearly as well developed as for the equil ium response. Many models have therefore been 150developed focusing on subsets of the observed elastomeric response such as Mullins effect (Govindjee and Simo 1991) and small strain oscillatory loading (e.g. Ward 1983), but some workers (Lion 1996; Lion 1998; Bergstrém and Boyce 1998; Kim et al. 1997) have also tried to model the time-dependent behavior when subjected to arbitrary strain histories. The objective of this work was to develop a micromechanism inspired theory that successfully captures many of the time-dependent characteristics. In order to motivate the proposed model, experimental data on a chloroprene rubber will be presented in Section 5.2. In Section 5.3, the Bergstrém and Boyce (1998) model will be presented. ‘An extension to the model will then be presented that allows for different time- dependent properties during loading and unloading. A procedure to determine the material constants used in the model is presented in Section 5.4, and the chapter is ended with conclusions in Section 5.5. 5.2 Experimental Data of elastomeric materials To illustrate the complicated response that is characteristi subjected to general strain histories, a representative experimental example will be presented. Although the data is for a chloroprene rubber, the qualitative response illustrated is quite general and is observed in most other elastomers as well (see Chapter 2). The example, shown in Figures 5-1 and 5-2, illustrates the experimentally observed large strain time-dependent behavior of the chloroprene rubber. In this case the elastomeric sample was subjected to a constant strain-rate loading-unloading cycle, and in the test a number of hold periods during which the applied strain was held constant were inserted, as shown in the strain history in Figure 5-la. The stress response due to the imposed deformation is shown in Figure 5-1, illustrating that the stress during the relaxation periods relaxes towards an equilibrium state and that this equilibrium level only depends on the current state of strain. This observation suggests that there exists a unique equilibrium response that is approached in the limit as the applied strain rate goes to zero, as illustrated in Figure 5-2b. This means 151A hloropene rubber “Tvl N6OD exon Back uniaxial compression ‘iain rate=-DOV8 ~~ Or~SCO2SCSSSCOSOGOD chloroprene rubber 7 vols N6O0 carbon black uniaxial compression sain ate = 001, 100150200 —as0~—«00~=S=CSDSC ‘Tre Stain (b) Figure 5-1: (a) Stress-strain response, é = —0.1/s, relaxation time=20 s. (b) Stress- time response. 152aha — (eats: 07) al -06| ol chlooprene ruber al welt eatbon back NSO ‘nial compression “I 02—~OS OCC “True Suain (a) $028[ pp enie on Trae Ses. 0 (MP é g ons a a a "Tre Sain (b) Figure 5-2: (a) Estimated equilibrium stress as a function of strain. (b) Time- dependent stress as a function of strain. 153that the stress in the elastomer can be decomposed into two parts: an equilibrium stress 4 and a time-dependent deviation from the equilibrium state og. The total stress in the system is then given by o = 04 +0, or graphically, the total stress in Figure 5-1a is equal to the sum of the curves in Figures 5-2a and 5-2b. The micromechanisms causing this type of response of elastomers is the topic of the next section were a constitutive framework for modeling the behavior is presented. More experimental data is then presented in Section 5.3.1 together with predictions from the model presented in the next section. 5.3 A Mechanism for Time-Dependence in Elas- tomers To explain the proposed mechanism for time-dependence consider first a lightly crosslinked elastomer consisting of long strands of polymers between crosslinks. Sup- pose for the moment that the network contains one free chain es schematically shown in Figure 5-3a. In this figure a cross-section of the network is shown and intersect- ing chains are indicated by dots. Due to the topological constraints from the other chains the allowed conformations of the free chain are confined into a tube-like region. The is of the confining tube can be defined as the chain conformation between the end-points having the shortest contour length. Such a path represents a group of conformations which are accessible to each other without viclating the topological constraints and is called the primitive path. Consider now a situation in which the network in Figure 5-3a is deformed at a high rate. Due to the imposed deformation the free chain will also deform more or less affinely with the network. Hence the entropy of the free chain is decreased and the free chain contributes additional deformation resistance to the system. If the applied strain is then held constant in the deformed state then the free chain will slowly return to a more relaxed configuration. Note that the reconfigurational motion of the chain will be different at the chain ends and in the central region of the chain. Close to the ends the chain will displace itself along its 154(2) Undeformed Network (© Deformed and Figure 5-3: One free chain in a network. 155contour length by reptational motion. The mechanism of displacement in the central region is different and not really of reptational type. In the central region the chain is highly stretched and any Brownian displacement that relaxes the stretched state will increase its entropy and hence be energetically favorable. The additional displacement freedom supplied from the reptation at the ends will therefore be consumed and used to create and modify the axis of the constraining tube. That is, the additional slack carried by the reptating segments will be consumed in the central region of the chain. Note that the direction of the reptational displacement can be in both directions along the constraining tube but the propagating away from the central region will cause an increase in the free energy and hence occurs with much lower frequency. In most real elastomers the number of free chains of the type illustrated in Figure 5-3 is very small and the proposed mechanism can therefore not explain the large magnitude of stress. relaxation that is observed in experiments. One way to generalize the approach is to also consider dangling chain ends of the type indicated in Figure 5-4. Note that the proposed mechanism just discussed can be applied unmodified for the case of a chain attached to the main network only at one end. The subchain CD, in this figure, is Figure 5-4: A dangling chain end in a network. highly stretched causing a force on the segment located at C and therefore promote the slippage through the entanglement point C. During a rapidly applied stretch, 156the effective chain stretch of subchain CD deformation. During the subsequent relaxation the total subchain length between the initially increased due to the applied entanglement points C’ and D is increased causing a reduction of the chain stretch. The mechanism by which this occurs is through the reptaticnal displacement of sub- chain AC. Point C in Figure 5-4 is an entanglement point which can be considered to act like a slip-ring through which the free chain can slip. It has been shown by Santagelo and Roland (1992), however, that elastomers prepared in such a way that the number of dangling chain ends is severely restricted still exhibit significant stress relaxation. This means that the two mechanisms for time-dependence discussed so far still might not be sufficient for explaining the observed behavior. The situation can be remedied by adding one additional level of abstraction allowing also for time- dependence in end-crosslinked networks by realizing that there will be chain regions which become highly stretched during fast applied macroscopic deformation but has deficient connectivity to the main load bearing network and hence have the capability to reach a lower energy state if given enough time. The mechanism proposed above then applies by identifying highly stretched chain regions with the stretched central part of the chain in Figure 5-4, and regions with loose connectivity to the main load bearing network with the chain end. Also in this case, the relaxation is facilitated by the reconfigurational displacement of chains from low stretched regions to high stretched regions. 5.3.1 Constitutive Modeling In order to implement the mechanisms presented above into a constitutive framework consider first the average distance a chain segment reptates as a function of time. This distance is given by the reptational dynamics scaling law (u) = C3(u?)'/? = C3 /9(t)- For the special case of a free chain in a fixed network the mean square displacement. 157of a chain segment has been shown to scale as (Doi and Edwards 1986): OW, ten th, mSt 0 and C, should be in the vicinity of the interval (0.125,0.5]. More specifically, since the relaxation process is governed by the finite displacement of chain segments, the time scale of interest. should be of the order of the tube disengagement time suggesting that C, should be close to 0.5. The effective distance a segment travels as a function of time should then scale as L = C3t. The effective length of the relaxing chain can therefore be written Leg(t) = Lo + L(t) giving the effective chain stretch as Xj = 1+ Cyt. This equation can be written in rate form as -1 =0s(%%-1)™, (5.2) ] (Cu-/c where Cs > 0, and Cs & 1. Note that Equation (5.2) is for a fixed stress level, but since the reconfigurational displacement is energy activated with the energy barrier corresponding to the inter- action with the surrounding chains it is necessary to explicitly incorporate a stress dependence. In lack of a more detailed representation this stress dependence is here taken to follow a power law relationship giving 35 = 01 0 - 1) (2)", (5.3) 158where C, = C,/#™ > 0, Cy = —1, and m > 0. Equation (5.3) shows how the effective stretch rate of a relaxing chain depends on its current state. The time-dependent behavior of a real elastomeric material can then be approximated by using the ex- perimental observation discussed in Section 5.2, that the mechanical behavior can be decouple into two parts: an equilibrium response and a time-dependent deviation from equilibrium, suggesting that the material can be modeled as two polymer net- works acting in parallel. The first network A captures the equilibrium response of the material and can be modeled by any model based on hyperelasticity. Network B is here modeled as a perfect network in series with a time-dependent element having the characteristics of Equation (5.3) which acts to to relieve the strain in the net- work with time and capture the experimentally observed behavior, see the rheological ir, Tr Figure 5-5: One dimensional rheological representation of the constitutive model. representation in Figure 5-5. By using the nomenclature of Gurtin (1981) and Silhavy (1997), the total defor- mation gradient acting on both network A and B can be written F = F, = Fy = x(p,1), see Figure 5-6. The deformation gradient F can furthermore be decom- posed into distortional and dilatational parts: F = J'/3F* where J = det(F). The Cauchy stress acting on network A can be obtained from the eight-chain network representation (see Section 3.2) dev [B*] + x{J —1]2 (5.4) 159‘Current configuration Relaxed configuration Figure 5-6: Multiplicative decomposition of deformation. FFT, and ¥* = /ir(B*)/3 is: the effective distortional chain stretch based on the eight-chain topology assumption. where 1%, Alc, « € Ry are material constants, B' ‘The deformation gradient on network B can be decomposed into elastic and viscous parts: Fy = F%F}. The Cauchy stress acting on network B can be obtained in the same manner as for network A: dev [Bg] + [J - 12 (5.5) where Jg = det(Fy), and Xf = /tr(BG)/3 The total stress in the system then simply becomes T, + Tp. ‘The total velocity gradient of network 3, Lp = FgF5\, can similarly be decom- posed into elastic and viscous components: Ly = Lg +FQL}F% ' = Ly +Ly, where Ly = FFy" = Dy + Wy and Ly = Dy + We. ‘The unloading process relating the deformed state with the intermediate state is not uniquely defined since an arbitrary rigid body rotation of the intermediate state still leaves the state stress free. The intermediate configuration can be made unique in different ways (Boyce et al. 1989), one particularly convenient approach is 160to prescribe Ws = 0. This will, in general, result in elastic and inelastic deformation gradients which both contain rotations. The rate of shape change of network B can now be constitutively prescribed by Di =iaNe, (5.6) where Np gives the direction of the driving stress state of the relaxed configuration convected to the current configuration and +p is an effective deformation rate. Noting that Tp is computed in the loaded configuration, the driving stress state on the relaxed configuration convected to the current configuration is then given by Ty = dev(Tg] and by defining an effective stress by the Frobenius norm rg = ||T’s|lr = (tr[T,(T’p)")"”, the direction of the driving stress becomes Np = T'p/||T’s|lr- The effective deformation rate is given by Equation (5.3) =r DR] [Z]” 7) where Ci, C2, 7 and m are material constants. In Equation (5.7), the term [Xj — 1], which was motivated by reptational dynamics, captures a stretch dependence of the effective viscosity and is a key component to capturing the rate-dependent behavior; the term (#)” is phenomenologically motivated by thermal activation arguments. Note that in the proposed model the constants C, = C,/7” and m are positive; and C2 is a constant that from reptational dynamics considerations is likely to be close to -1. The procedure by which the material constants used in the model are determined is discussed in detail in Section 5.4. To evaluate the effectiveness of the proposed model, it is compared in Figures 5-7 to 5-25 to experimental data of both the chloroprene rubber and natural rubber discussed in Chapter 2. In Figure 5-7 is shown the behavior when loaded to different final true strains. The figure shows that the predicted behavior both during uploading and unloading are in good agreement with the experimental data. In Figures 5-8 and 5-9 is shown the behavior for a constant strain-rate load cycle interrupted by hold periods during which the applied strain was held constant allowing the stress to relax. 161“Tre Suess (MPa) chlorowene rubber 7 vols carton black N6OO ‘unianal compression stain aie=-0015 0 lOO) Tre Sain Figure 5-7: Uniaxial compression to different final strains, chloroprene rubber filled with 7 vol% carbon black N600. Material constants used: 1% = 0.75 MPa, ph = 0.95 MPa, AWgck = Ag = 2.5, Cy = 0.6 s“(MPa)-™, Cp = —1, m And in Figure 5-10 is shown the stress in the time-dependent network as a function of the applied strain for the same test. A direct comparison demonstrates that the model predictions of the overall stress levels as well as the rate of stress relaxation at all strain levels are in very good agreement with the experimental data. And in Figure 5-11 is shown the behavior at two very different strain rates. Both the slow strain rate experiment and the uploading behavior during the fast experiment are well predicted. The stress during the unloading in the fast experiment, however, is underpredieted. The reason for the underprediction in stress is that the proposed constitutive model predicts the same strain rate dependence during both loading and unloading, and the chloroprene rubber used in the experiment exhibits lower rate dependence during unloading. A procedure by which the model can augmented to allow for different behavior during loading and unloading is presented in Section 5.3.3. Another test case is shown in Figures 5-12 and 5-13. Here the sample was first compressed with a constant strain rate to a final strain of —0.4, and then subjected 162‘popes esti o-07) hloropene ruber 7 vols carbon lack NGO 01S SOS Figure 5-8: Stress relaxation test on a chloroprene rubber filled with 7 vol% carbon black. Material constants used: 1%, = 0.75 MPa, y9 = 0.95 MPa, AWge* = algek = 2.5, C, = 0.6 s“\(MPa)-™, C, = -1,m=4. = ‘exon sme tee -on chloroprene rubber “vol carbon black NGO uniaxial compression strain te = ~0.01/, ‘elaxation time = 20 8 a a ee a ime (5) Figure 5-9: Stress relaxation test showing the dependence on relaxation time for a chloroprene rubber filled with 7 vol% carbon black. Material constants used: 1% = 0.75 MPa, u = 0.95 MPa, Ak = Ng = 2.5, C, = 0.6 s“!(MPa)“™, Co m=4. 163uni’ compression | Ne teint = 0.005 1 alata ime = 20 ON or 02S OO ‘True Suain Figure 5-10: Stress relaxation test showing the stress in the time-dependent network for a chloroprene rubber filled with 7 vol% carbon black. Material constants used: 19, = 0.75 MPa, p19, = 0.95 MPa, AW = Al = 2.5, Cy = 0.6s"!(MPa)-™, Cp = —1, m=4, strain rate=-0.15 (st id 1-10) Drei da sain rate=-0.18 rain rte=-.002/ (etd BI-11) Dredicied data sain ate=-0.0025 hloroprene rubber To catbon back N6OO ‘uniaial eompression a re Sin Figure 5-11: Uniaxial compression with different strain rates, chloroprene rubber filled with 7 vol% carbon black N600. Material constants used: .% = 0.75 MPa, pi%, = 0.95 MPa, Nek = Agee = 2 0.6 s'(MPa)-", Cz = —1, m= 4. 164to a sinusoidal oscillation in the applied strain with frequency 0.02 Hz and strain amplitude 0.1 (the applied strain history is shown in the inset of Figure 2-10). The figures show that all features of the stress response, such as tangent modulus, stress amplitude, and unloading behavior, are all very well predicted. Another example of chloropene rubber 7 vol carbon back N6OO ‘lax i stain ra =-0015 frequency = 0.02 Hz strain amplitude = 0.1 ‘numberof eyes = 5 “as ‘Tree Stain Figure 5-12: Oscillatory test with frequency f = 0.02 Hz showing stress as a function of applied strain. Material constants used: % = 0.75 MPa, u9 = 0.95 MPa, alge = Aiget = 2.5, Cy = 0.6 s-!(MPa)-™, Cp = -1, m= 4. the same type of experiment is shown in Figures 5-14 and 5-15 but here the applied frequency was 0.25 Hz. In this case the uploading behavior and the tangent modulus of the stress-strain loops are well predicted. But the stress amplitude predicted by the modulus is larger than what is experimentally observed. One last example of chloroprene rubber filled with 7 vol% carbon black is shown in Figure 5-16. In :his figure is shown results both from plane strain compression and uniaxial compression tests illustrating fair agreement between experimental data and model predictions. In Figure 5-17 is shown results for a chloroprene rubber filled with 15 vol% carbon black tested at two different strain rates. Also here the predictions have good quality except for the unloading at high rates where the model predicts too high rate depen- 165Senin) Applied True & 4 hloroprene rubber 7 Vol8 carbon back N600 ‘iaxial compression 166 stmin ate = 00175 {frequency = 0.02 He stain amplitude = 0.1(esti: 6-17) ‘Model predictions & E-o3 3-26] ol 1 Tel carbon back NEOO iaxial compression strain ae =-D01/s -04 frequency =025 He strain amplitude = 0.1 ‘numberof eyeles = 5 % “or 323" Tre Sinin Figure 5-14: Oscillatory test with frequency f .25 Hz showing stress as a function of applied strain. Material constants used: 4% = 0.75 MPa, 3, = 0.95 MPa, Ale = Aigek = 2.5, C, = 0.6 s-!(MPa)-™, Cp = —1, m= 4. 167got ~ wey i ee " 02 7 vol% carbon black N600 Satoeeat Sinan ON umber of eyeles = 5 Sk ee Figure 5-15: Oscillatory test with frequency f = 0.25 Hz showing stress as a function of time. Material constants used: 5 = 0.75 MPa, 1%, = 0.95 MPa, Algck = Algtk = 2.5, C, = 0.6 s“(MPa)-, Cp = -1,m=4. plane suai coms Blane sain FT chloroprene ubber “vol carbon back N600 sein rate = 008 os} i Tre Sain Figure 5-16: Plane strain response of a chloroprene rubber filed with 7 vol% carbon black. Material constants used: 4% = 0.75 MPa, 1%, = 0.95 MPa, A'gc = Algek = 2.5, Cy =0.6 s(MPa)-", Cp = -1, m=4. 168dence. The prediction of the relaxation behavior shown in Figure 5-18, however, is in strain rte=-.05/ (esti: d1-12) (redicted dat sin rao=-0.08/s ‘fin rire O00/ (ist a>18) 2 prodiced dain sain rate=- O00 chloroprene rubber 15 vol erbon Back NSDO uniaial eomptesion 0 Ol -02~—~OSS OSC OGSCOTCOB ‘True Sain Figure 5-17: Effect of different strain rates on the stress-strain response of a chloro- prene rubber filled with 15 vol% carbon black. Material constants used: 1% = 1.2 MPa, 19 = 2.6 MPa, Ale = Ng = 1.9, Cy = 0.1 s-\(MPa)-®, C2 = —1, m= excellent agreement with the experimental data. In Figure 5-19 is shown results for a chloroprene rubber filled with 25 vol% carbon black when tested both in uniaxial compression and plane strain compression. When comparing the model predictions with the experimental data it is clear that the model does a good job in predicting the magnitude of the hysteresis both in uniaxial and plane strain compression. ‘The large strain resnonse of an unfilled natural rubber is shown in Figure 5-20 and the influence of different strain rates is shown in Figure 5-21, Both of these figures show good agreement between model and experiments. Data for simple shear of a natural rubber to a shear strain of 0.5 is shown in Figure 5-22. ‘The experimental data shown in the figure exhibit very little hysteresis. The small amount of hysteresis, may simply be related to the magnitude of the applied strain where little hysteresis is observed in uniaxial compression to a final strain of 0.2 as shown in Figure 2-64. This behavior also could be interpreted as an indication that the hysteresis might 1693 experimental dat (tei: priced daa esi B-0) hloropene rubber os 15 vol carbon black N6OO uniaxial compression Sain rate = 00 ‘relaxation time = 408 150 Time(s) 0 30 100 200) 20 300 Figure 5-18: Stress relaxation test showing stress as a function of time for a chloro- prene rubber filled with 15 vol% carbon black. Material constants used: y% = 1.2 MPa, { = 2.6 MPa, Algck = Algek = 1 = 0.1 s~!(MPa)-", C, = -1,m=4. plane suai, relax ime=t0 s (test id 2542) = bbe rere s Gest: 3-08) ~7- thal prediction “5 Tre Stain Figure 5-19: Plane strain response of a chloroprene rubber filled with 25 vol% carbon black. Material constants used: 4 = 1.95 MPa, y% = 8 MPa, Algck = Algek = C, = 0.01 s“!(MPa)"™, C2 = -1, m=4. 170os] unfilled natural rubber uniaxial compression “32 oa 6 08 ‘Tree Sian ara a6 Figure 5-20: Large strain response of an unfilled natural rubber. Material constants used: #4 = 0.5 MPa, p3 = 0.4 MPa, Alt = Ak = 2.1, C, = 80 s-!(MPa)-™, GQ =-I,m=4. Sun rate=-D TA ent m-e-07) Strain te=-0 01 (pei da) a Tre Sues (Ps) & & -02| or uafiled nara ubber ‘ial compression OO) OE OS sas Tre Sain Figure 5-21: Effect of different strain rates on the stress-strain response for an unfilled natural rubber. Material constants used: 1% = 0.5 MPa, u%, = 0.4 MPa, Alek = Agek = 2.1, Gy = 80 s-'(MPa)-™, Cy = -1, m=4. 17be pressure dependent. A more detailed examination of other shear experiments (to be shown in Figures 5-23 and 5-25) es well as uniaxial tension versus compression data from Lion (1997) however suggest that if there is a pressure dependence then in general this dependence is very weak. Therefore pressure dependence of hysteresis has not been included in the current model framework. shea sin rate=0.04 (et: ns sgesugigate-0.045 Gest nr-s103) 035 04 “Tre Suess, 7, (MPa) ed natural rubber simple shear 0a os 06 0 oF 02 03, Engine Sn, E,. Figure 5-22: Comparison between simple shear predictions and experimental data, unfilled natural rubber. Material constants used: 4, = 0.5 MPa, 1, = 0.4 MPa, Ags = Alek = 92.1, E, = 80 s-'(MPa)-™, C, =—1, m= 4. Another simple shear example is shown in Figure 5-23 for a natural rubber filled with 17 vol% carbon black N315. The figure demonstrate better agreement between the experimental data and the model prediction, ‘The last two figures, 5-24 and 5-25, are for a natural rubber filled with 24 vol% carbon black tested in uniaxial compression and simple shear, respectively. Both of the figures show that the model predictions and the experimental data are in reasonable agreement, although the magnitude of the predicted time-dependent stress during the unloading phases is higher than what is experimentally observed, ‘This discrepancy between predictions and experimental data and how the model can be improved is the topic of Section 5.3.3. 17203) + ey aii ten Or est sO atu uber 17 volt carbon beck N3SI simple shear o 005 oT on 02 025 03 Engineering Susi, E,, Figure 5-23: Comparison between simple shear predictions and experimental data, natural rubber filled with 17 vol% carbon black. Material constants used: 1% = 1.0 MPa, ni = 1.4 MPa, Ag* = Ag = 1 = 20 s“(MPa)-", C; = -1, m jn te. ei TST HAR Sa en wa, preted fran ce0.15) ZiT BOSSES SIE SB) atl rubber 24 vol carbon bck N3ST ‘uniaxial compression ge, “acs 01 -OIS “02 G35 93-0 -Oa 0 as re Strain Figure 5-24: Effect of different strain rates on the stress-strain response of a natural rubber filled with 24 vol% carbon black N351. Material constants used: .% = 1.2 MPa, p12, = 1.9 MPa, Aloe = Algck = 1.5, Cy = 0.5 s-!(MPa)-", C, = -1, m= 173= Begegupe mer cenit 0-209 ‘True Suess T,, (MPa) sf & natural ubber ‘carbon black: 24 vol N3ST simple shear 2 000501 015 02028 03 035 04 04s aS Engineering Suan, E, Figure 5-25: Simple shear behavior of a natural rubber filled with 24 vol% N351. Material constants used: :% = 1.2 MPa, 9 = 1.9 MPa, alge = alge = 1.5, C, = 0.5 "(MPa)", C) = -1, m=4. 5.3.2 Verification of the Clausius-Duhem inequality To verify that the developed constitutive model does not violate the second law of thermodynamics it will here be derived under what conditions it satisfies the Clausius- Duhem inequality. Start by considering a general material in which Y(-), T(-),9(),4(-) tepend on the state variables {F, 0, g,F'}. The Clausius-Duhem inequality pb-T-L+ p+ 28 0. 5.11 5 Now let Fg = (J§)'/*F¥, and from (5.5) 8 aerlBe) 4 ale ies, Te _i0te ae ime Ty = 48 dev[By] +x[J§—1]1, and E* = 422 = 79H? _ deve), ce ie YoU ~ TT Pe giving Z ' Yenex ' Ts-L’= bie dey fBe -dev[B] + 7 [J& — 1] dev[B*]-1. (5.12) ot = te den deve] + REE 5 —Jdev"] 1. (12) Note, dev[B§] -1 = tr(dev[B*}) =0 and dev[Bi] - dev[By] > 0 175giving Ty - LY > 0 if and only if u > 0 and CG, > 0. By taking q = K(F,6)g where K is a positive semi-definite heat conductivity tensor the dissipation inequality is automatically satisfied for all thermomechanical processes, 5.3.3 One Approach to Improve the Unloading Predictions ‘The model discussed in the Section 5.3.1 predicts the same time-dependence during both loading and unloading, but from the experimental data in Figures 5-7 to 5-25 and the experimental investigation in Chapter 2 it is clear that many elastomers exhibit, weaker time-dependence during unloading. The reason for the different behavior during loading and unloading can be understood if one considers the two network representation discussed earlier. Within this framework the time-dependent network Bis deformed due to its interaction with the equilibrium network A, i.e. network B is deformed due to its interaction with its closest neighboring chains which constitutes its environment. Consider now a situation in which the system is deformed in tension. Due to the applied deformation both network A and B will be stretched out in the loading direction. If now the applied deformation is reversed causing the system to unload, network A will still stay in a tensile stress state whereas network B will go into a compressive stress state. That is, the chains in network B during the loading phase experience an environment that have the same characteristics as itself. But during the unloading phase, network B will experience an environment having a very different behavior. This argument suggests that the loading and unloading behaviors can be fundamentally different indicating that to achieve better predictions of both loading and unloading it might be necessary to modify the model presented in the previous section. One possible approach for doing this in an ad hoc fashion is to allow for different values of the scale factor C; during loading and unloading. To implement this in an unique way it is necessary to have a clear definition of when a material is loading and when it is unloading. One way to do this is to introduce a load flag € 176defined by €=sign(T’ -D*), (6.13) loading is then said to occur when € = 1. Examples of how this approach works are shown in Figures 5-26 to 5-30 for a chloroprene rubber filled with 7 vol% carbon black. In all examples Ci.down = 100Ci.up- By comparing the predicted data with the experimental results in these figures is clear that this approach not only improves the unloading predictions at high rates but also the overall quality of the model, of course at the expense of introducing one additional material constant. 4 exparimencal data (ist ids: b1-07, b1-06, 61-05) predicted dat a -15} “Tre Stress (MPa) os} chloroprene rubber 7 vals arbor black N6OD ‘nial compression ruin rate=-00Us a Figure 5-26: Uniaxial compression to different final strains, chloroprene rubber filled with 7 vol% carbon black N600. Material constants used: 43, = 0.75 MPa, py, = 0.95 MPa, Age = algek = 2.5, Grup = 0.6 s-!(MPa)-™, Cy = —I, m= 4. 17strain ate=-0.1/5 (testis bl 10) Dredicied data sain rate=~O.1 Frain rauca-0.002/ (est id: bI<11) Predicied dat: sain ate=-0.0025 y -os| hloroprene rubber 7 vols carbon black N6OO ‘nina compression or 02-03 0s SSCS OTC True Sain Figure 5-27: Uniaxial compression with different strain rates, chloroprene rubber filled with 7 vol% carbon black N600. Material constants used: 9% = 0.75 MPa, p% = 0.95 MPa, Aigck = Algsk = 2.5, Chup = 0.6 s~!(MPa)-™, Cp = -1,m=4. hloropene rubber 17 vols earbon black N6OO ‘uniatial compression 001K 2208 Figure 5-28: Stress relaxation test on a chloroprene rubber filled with 7 vol% carbon black. Material constants used: 9, = 0.75 MPa, 1%, = 0.95 MPa, Aig = Ag = 2.5, Cup = 0.6 s-'(MPa)-™, Cp = —1, m= 4. 178é Soa} BP ol oa] hloroprene rubber vol carbon black N6OO amin rate = 001s 0 30 100 150 Time (6) Figure 5-29: Stress relaxation test showing the dependence on relaxation time for a chloroprene rubber filled with 7 vol% carbon black. Material constants used: 1% 0.75 MPa, y3, = 0.95 MPa, Age = Alot = 2.5, Chup = 0.6 s“!(MPa)-™, Cp chloroprene rubber 1 vole carbon back N6OO ‘uniaxial compression 0 100 200 300) 0 300 oo Times) Figure 5-30: Oscillatory test with frequency = 0.02 Hz showing stress as a function of time. Material constants used: 9 = 0.75 MPa, 19% = 0.95 MPa, Ae* = AWgek = 2.5, Cup = 0.6 2-!(MPa)-™, Cp = -1, m= 4. 1795.4 Determination of Material Constants In an effort to further illustrate how the model works, in this seciion will be pre- sented one procedure for determining the material constants. Depending of what type of experimental data is available, different procedures for determining the ma- terial constants can be constructed. Since perhaps the most common experimental setup is uniaxial deformation the particular example shown in this section is based on uniaxial stress-strain data at two different strain rates of the type illustrated in Figure 5-31. Note, that the experimental data in Figure 5-31 include the behavior TEE stain atee-9 fast estid: bt-10), Te- Sain rate=-0.00zs Glow) GesudBIo1) hloroprene ruber Tol NOOO extbon back ‘uniaxial eampession “a “38 ‘Tre Sin Figure 5-31: Experimental data used to determine the material constants in the model. during a complete load cycle. As discussed in Section 2.15, many elastomers exhibit different time-dependence during loading and unloading, and to get good predictions of, for example, oscillatory loading situations it is helpful to include this non-linear effect in the determination of the material constants. The first step in the procedure is to determine the equilibrium behavior, i.e. the material constants 4, \'g and x. A procedure by which this can be done was discussed in Section 3.3. ‘To facilitate the determination of the material constants for the time-dependent 180network it is helpful to separate out the response of the time-dependent, network by subtracting the equilibrium response from the total stress shown in Figure 5-31 giving the curve in Figure 5-32. The shear modulus of network B can then be approximated “ - as eee 7 Ay, destacany O.1F chioroprene rubber 27 a. Papasan ows} eee aon «ma 025 a2 03a 0S SCOT “True Sain Figure 5-32: Stress in network B as a function of applied strain. from Aca, AX Hy = @yret (5.14) ‘The value of Xigc* can in principle be determined by the same procedure as used to determine A'¢, The strain in the elastic part of network B, however, never becomes very large making it difficult to estimate 4g* through this procedure. But this also means that the actual value of \¥g‘* has very little influence on the predicted stres strain behavior, and for convenience can simply be chosen as Aig = Al. The next. three constants to be determined are: C,, Cp and m. To determine these consider the flow-rule for the time-dependent network B, Equation (5.7), specialized to uniaxial deformation i BRC) - 1) lon( tl” sign(on()). (6.15) a(t) 181‘The function op (t) can be obtained by replotting the data in Figure 5-32 as a function of time as done in Figures 5-33 and 5-34. Since both the material constants of the 2 chorpe ber eo 77 vol% carbon Hack N6OO Os) i ‘Uniasil campresion \ fore) oa ' seine 20001 2 005 i j 4 2 os ‘ é 2 a ‘ * ous) Sya0" wo 100012001400 Too Time(s) f Figure 5-33: Stress in the time-dependent network as a function of time. elastic part of network B and the stress o9(t) are now known, it is straightforward to invert the eight-chain relationship and obtain ¢(t). The strain in the viscous part of network B can then be obtained by subtracting e%(t) from the prescribed strain e(t) giving the data in Figures 5-35 and 5-36. By choosing ty,, ty, tae, tay such that Chaltis) = eBpltie)s and ops(tas) = ony(tay)s (5.16) the constants C2, m and C, can be calcuiated from In [ey (tos) /ébo(tos)] Cp ee 5, * In [By (ey) = 1) / OB. (6.7) In [Eb p(tay)/ebs(trs)] Infoasip)/ons(ts))" (5.18) ball . 7 (5.19) Piatt) — 1) [Cons( ts)!” sign (oot) . 182hloreprene rubber 7 vals earbon black N6OD ‘niasalcompresion fastest) strain rte = 005 10 1s 20 25 30 Times) in the time-dependent network as a function of time. 03 aes ‘lorie ruber f° ryote eb tack NOOO a , ‘ras oading ‘Sow est) so of ib siainsaes 0.0005 A gf ¥ +s 2 2 | te = -04 Ba a z 03 0 0700400 «0 600 10001200 14001600 Times) Figure 5-35: Inelastic strain in network B as a function of time. 183hloroprene rubber sng. Tvol carbon black N6OO al ns ‘ial loading 7 fast tes () Nin rac = 0.05 & Inelastic sain in Network Be 0 5 10 15 70 2s 30 Time s) Figure 5-36; Inelastic strain in network B as a function of time. — 2 %, — Ney)? + 2/A% 5 = Cal eae, and Yi; = jb +2, (5.20) Note, all quantities in Equations (5.17) to (5.19) can be directly obtained from the where graphs in Figures 5-33 to 5-36 giving the material constants used earlier. Also note that the solutions to (5.16) are not unique, and that different solutions will result in different value for the material constants. 5.5 Conclusions To accurately predict the mechanical response of elastomeric materials is a difficult, problem requiring much more than simply using a hyperelastic material model. Most elastomeric materials in industrial applications often exhibit Mullins effect, strain-, rate-, and temperature-dependence causing significant amounts of hysteresis, some- thing that cannot be predicted by a hyperelastic model. For this reason a number of viscoelasticity based models have been proposed addressing different aspects of 184experimentally observed elastomeric behavior. In this chapter the focus has been on developing a new model that captures the time-dependent large strain behavior of unfilled elastomers. The foundation of the new model is that the mechanical behavior can be decomposed into two parts: an equilibrium network corresponding to the state that is approached in long time stress relaxation tests; and a second network captur- ing the non-linear rate-dependent deviation from the equilibrium state. The stress state in both networks is calculated with the eight-chain I;-based rubber elasticity model presented in Chapter 3. The time-dependence of the second network is taken to be governed by stress-assisted reptational displacement of chain segments undergo- ing a reconfigurational motion due to the slippage of untrapped entanglements. The proposed model is then incorporated into a finite strain continuum framework. A sys- tematic procedure to obtain the material constants needed in the model is presented, and numerical simulations of uniaxial, plane strain and simple shear deformation with different imposed strain histories are compared with known experimentai data of both chloroprene rubber and natural rubber. The comparisons indicate that the proposed model can accurate predict the experimentally observed time-depenucace in most ex- perimental conditions. The only exception being the experimentally observed weaker strain-rate dependence during unloading at high strain rates for certain elastomers. To allow for accurate predictions also under these conditions an ad hoc procedure to extend the proposed model based on different prefactors in the stress assisted repta- tion based relationship during loading and unloading has been developed. By direct. comparison with experimental data it is shown that this simple approach gives the correct response also for the unloading behavior at high scrain rates. 185Chapter 6 Time-Dependence of Filled Elastomers 6.1 Introduction The previous two chapters discussed the equilibrium behavior of filled elastomers and the time-dependence of unfilled elastomers. ‘The goal of the current chapter is to combine these two subproblems into a unified theory—the time-dependence of filled elastomers. Many different attempt strategies can be used when developing models for the time-dependent behavior of a particle filled elastomer. Perhaps the simplest approach is to use a model developed for an unfilled elastomer, one example being the model presented in the previous chapter, and then find effective material constants for the composite system composed of both the elastomeric matrix and the semi-rigid filler particles. That this type of approach can be useful was exemplified in the last chapter for both chloroprene rubber and natural rubber filled with different volume fractions of filler particles. One obvious limitation of this method, however, is that the effective material constants must be determined for each elastomeric system of interest. In the design of new components using elastomeric materials itis often desirable to be able to tailor the mechanical response of the elastomeric material for the particular application at hand, and one of the most common ways to accomplish this is by 186changing the amount of filler particles added to the elastomer. To make the design phase efficient it is of great importance to be able to quantitatively predict how a particular elastomeric system will behave before the material is actually made, and the goal of the current chapter is to develop a model that allows for this type of prediction. The framework of the proposed model is presented in Section 6.2 and the chapter is ended with conclusions in Section 6.3. 6.2 Modeling To model the time-dependence of a filled elastomer the framework developed in the last two chapters will be combined into a unified theory. As was shown in Chapter 2, the response of both unfilled and particle-filled elastomers can be decomposed into equilibrium and time-dependent components. The structure of the framework developed for the time-dependence of unfilled elastomers can therefore also be used in this case. ‘The one-dimensional rheological representation shown in Figure 5-5 remains unchanged but the constitutive equations for the different components need to be changed to explicitly take into consideration also the volume fraction of particles. Based on the discussion in Chapter 4, the two hyperelastic network components can simply be modeled with the /,-amplified eight-chain representation that was presented in Equation (4.25): = Hy 2% )m/ AR) PM TES PO dev[B"] + (1 vg) [J — 1], (6.1) LOG) XG) Pee) dev[B§] + (1- vy) «[J§ - 1]1, (6.2) 187where (63) (6.4) (6.5) and where }* is the effective chain stretch (based on the eight-chain assumption) of the macroscopic composite material; (A*)m is the volume average chain stretch in the matrix phase; Xg is the effective chain stretch associated with the elastic part of the deformation gradient of network B, F%, for the macroscopic composite material; and finally, (Am is the volume average of Xj in the matrix phase of the composite system. Note that in deriving Equations (6.1) to (6.5), the result from Chapter 4 that the matrix phase does not change properties when filler particles are added has been used. ‘The behavior of the time-dependent component, however, needs to be studied in more detail. The governing mechanism for the time-dependence is still modeled by the reptation-based scaling law which for the case of an unfilled elastomer can be written \y = 1+Cyt. Fora filled elastomer the corresponding relationship becomes (Xp)m = 1+ Cat, (6.6) where as before (-)m denotes volume average in the matrix phase. Time differentiating (6.6) and adding energy activation gives $m = Cs (Cm 9° [8 7) where (ron = (Ialle)g = (esl) (628) Bim = VXOR = 1) +1. (69) 188The average effective shear stress in the matrix phase, (rp)m, cannot be obtained using the same /,-amplification approach that was used for the effective chain stretches. Instead a different amplification procedure will be proposed and verified through micromechanical simulations. Equation (6.7) shows how the time derivative of (X})m depends on the current state but when implementing the model the quantity that is needed is the time derivative of the macroscopic Xp associated with the RVE. By applying the chain rule on (6.9), Equation (6.7) can be transformed into: (O5)m — 1) fala)” : (6.10) The definition of (rs)m is given by (6.8) but since the distribution of Ty is not known it is not possible to evaluate (ra)m directly using (6.8). One way to determine (7a)m is to use micromechanical simulations of the type discussed in Chapter 4 in which the matrix material is represented with the time-dependent model presented in Chapter 5. Simulation results using this approach are shown in Figures 6-1 to 6-4. A first example, shown in Figure 6-1, illustrates how (ra)m depends on applied strain for a RVE containing 7 vol% dodecahedron shaped particles. In the figure is also shown how 7» for an unfilled RVE depends on the applied strain. When implementing into the constitutive framework, however, the relationship that is needed is between (Tp)m and Tp of the composite material (i.e. the rT of the RVE), not between (rp)m and applied strain. The problem here is that rg, the effective shear stress in network B, is an internal state variable that cannot be directly obtained from the simulations in the same way it cannot be directly obtained fiom a traditional mechanical experiment. Simulation of relaxation tests would have to to be conducted in order to attempt to indirectly obtain 7p of the RVE. From the same type of micromechanical simulation it is also possible to extract the three stretch quantities Xg, Xj, \* as shown in Figures 6-2, 6-3, 6-4, respectively. From Figure 6-2 it is clear that (X$)m is much larger than XG for the unfilled RVE. ‘The corresponding difference between (Xp)m and Xp for the unfilled RVE is much smaller, as illustrated in Figure 6-3. Finally, as shown in Figure 6-4, the amplification 189012; sinc compression ‘ain ae=-O.1s 7 vol files a hombic dodecahedton shaped paricles g Ty 8 § Tq (unfiled) Effective Shear Suess t, (MPA) 0.03} "9-902 008-005 008 91 012 014 O16 O18 02 “Applied Tue Sun Figure 6-1: Effective shear stress rg as a function of the applied strain, uniaxial ion, 7 vol% fillers. Matrix material constants: 1%, = 19, = 0.5 MPa, Xlgck = Age = 2.1, Cy = 20 s-\(MPa)-™, Cp = ~1, m= 4. relationship in (6.3) is in good agreement with the micromiechanical simulation results. As mentioned above, it is not possible to use this type of simulation to directly obtain Tg for a filled RVE as a function of the applied strain. However, microme- chanical simulations using a time-independent matrix can be used to obtain both the macroscopic norm, 7, and the average matrix norm. Both 7 and (7)m can be extracted as a function of the applied strain, as shown in Figures 6-5 to 6-7. Figure 6-5 shows results from a RVE containing 7 vol% particles deformed in uniaxial com- pression. The figure illustrates that the macroscopic cell is larger than the average matrix (r)m and that to a good approximation + Xd — 0) (7)m (6.11) Similar results are shown in Figure 6-6 for a RVE containing 17 vol% fillers tested in uniaxial compression, and in Figure 6-7 for a RVE containing 17 vol% fillers tested 190uniaxial compression sarin te=-0.1s ‘Tole ler: ‘thombic dodecahedron shaped particles “ear 008 005 008 01 012 014 O16 O18 02 ‘Applied True Suain Figure 6-2: Elastic chain stretch of network B as a function of the applied strain, uniaxial compression, 7 vol% fillers. Matrix material constants: 9 = 9, = 0.5 MPa, Nock = algek = 2.1, G, = 20 s“"(MPa)“™, Cy = —1, m= 4. ‘uniaxial compression sirin lems 7 vol fillers thombic dodecahedron shaped particles Inelastic Chain Sueich of Network B ‘0 002-004 -006 -008 -o1 -012 -0l4 -016 -o18 -02 ‘Applied True Suan Figure 6-3: Inelastic chain stretch in network B as a function of the applied strain, uniaxial compression, 7 vol% fillers. Matrix material constants: 4% = 1%, = 0.5 MPa, Algek = ylgck = 2.1, CG) = 20 s"'(MPa)-™, Cp = -1,m=4. 1911.03 ‘unianial compression simin ate=-0.15, 1.02577 vol fillers ‘thombic dodecahedion shape particles ue 102 ¥ tos} oH 1.005} 0 0M 008 005 -008 01 012 014 016 018 02 ‘Applied True Strain 4 Figure 6-4: Chain stretch 3* as a function of the applied strain, uniaxial compression, 7 vol% fillers. Matrix inaterial constants: ', = y% = 0.5 MPa, \¥% = Ng = 2.1, Cy = 20 s“\(MPa)-™, C2 = -1,m=4. in simple shear. These results suggest that (rg)m can be approximated from (ra)m = Xa (6.12) which is the relationship that has been used in the simulations to follow. As an aside, it is interesting to point out that using this simulation technique it is also possible to study a number of other properties of the RVE, as shown in Figures 6- 8 to 6-13. The first example, shown in Figure 6-8, illustrates the strain energy density (W) of the cell, (W/)m of the matrix and (W), of the filler particles as a function of the applied strain. In the figure is also plotted (1 — vs)(W)m which is shown to be close to (W), in good agreement with Equation (4.18). In Figure 6-9 is shown that the volume average J both in the matrix and filler phases are very close to unity. ‘The pressures (P), (P)m, (P)s are shown in Figure 6-10. From the definition of these quantities, they are directly correlated by (P) = (1 v)(P)m + v/(P)j. The axial stress component Tz and corresponding deviatoric stress Ts, are examined in Figures 192==T &, 10-90) 0.02] uniaxial compression ight-chain matrix material °=0.5 MPa A***=2.1 “oo 008-006 008 SCO ‘Tre Stain Figure 6-5: Effective stress 7 as a function of applied strain, uniaxial compression, 7 vol% fillers, 04s; at i Be TOW) 035} 3 03 é 3 2221 vol filers 0.0s| ania eompression . eight-chain matrix material p9<0.5 MPa, 2422.1 0-00-00 006-008 81-0 01-6 Figure 6-6: Effective stress 7 as a function of applied strain, uniaxial compression, 17 vol% filler: 1932 TZ &, oe bxec-vmy shear Ss) € 2 @ 8 simple shear 17 vol fillers cight-chain maui material: u?=0.5 MPa, 2!**> 2.1 oa Cr a ‘Tree Stain Figure 6-7: Effective stress r as a function of applied strain, simple shear, 17 vol% fillers. Sun Energy Dey a) = 8 2 5 0-002 008 006 -008 01 -0.2 -Ol4 O16 -0l8 -02 ‘Applied Tre Stain Figure 6-8: Strain energy density as a “unction of applied strain, uniaxial compression, 7 vol% fillers. 1941.005; 1.004} 1003} 1002 1.001 Volume Defomation} g . 0.958 997} uniaxial compression 17 v008 fillers oss} tombic dodecahedron shaped paces ight-chaie matrix material: n°=0.5 MP X'=2.1 gsh—e ‘0-002 004 006 -008 -01 012-014 0.16 -O18 02 “Applied True Stain Figure 6-9: Volumetric deformation J as a function of applied strain, uniaxial com- pression, 7 vol% fillers. ong ntl compreion sae ee / oa} tle dette pepe 4 tight-chain matrix material: u*=0.5 MPa,2!"*=2.1 oe an Zo , Bos 7 5 : food cay Oe e “ oy . —@ . TR ~ °y

You might also like

• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  Rating: 4 out of 5 stars

  4/5 (5810)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brené Brown

  Rating: 4 out of 5 stars

  4/5 (1092)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  Rating: 4.5 out of 5 stars

  4.5/5 (843)
• • Magazines
  • Podcasts
  • Sheet music
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  Rating: 4 out of 5 stars

  4/5 (608)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  Rating: 4.5 out of 5 stars

  4.5/5 (1716)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  Rating: 4 out of 5 stars

  4/5 (590)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  Rating: 4 out of 5 stars

  4/5 (1104)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  Rating: 4 out of 5 stars

  4/5 (897)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  Rating: 4.5 out of 5 stars

  4.5/5 (540)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  Rating: 4.5 out of 5 stars

  4.5/5 (2104)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  Rating: 4.5 out of 5 stars

  4.5/5 (346)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  Rating: 4 out of 5 stars

  4/5 (1018)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  Rating: 4.5 out of 5 stars

  4.5/5 (474)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  Rating: 4 out of 5 stars

  4/5 (822)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  Rating: 4 out of 5 stars

  4/5 (1850)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  Rating: 4.5 out of 5 stars

  4.5/5 (271)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  Rating: 4.5 out of 5 stars

  4.5/5 (122)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  Rating: 4.5 out of 5 stars

  4.5/5 (441)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Toibin

  Rating: 3.5 out of 5 stars

  3.5/5 (1946)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  Rating: 3.5 out of 5 stars

  3.5/5 (401)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  Rating: 3.5 out of 5 stars

  3.5/5 (2259)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  Rating: 4.5 out of 5 stars

  4.5/5 (4610)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  Rating: 4 out of 5 stars

  4/5 (4203)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  Rating: 4.5 out of 5 stars

  4.5/5 (1929)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  Rating: 4.5 out of 5 stars

  4.5/5 (806)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  Rating: 4 out of 5 stars

  4/5 (98)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  Rating: 4.5 out of 5 stars

  4.5/5 (266)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  Rating: 3.5 out of 5 stars

  3.5/5 (2521)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  Rating: 3.5 out of 5 stars

  3.5/5 (231)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  Rating: 4 out of 5 stars

  4/5 (1898)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  Rating: 4.5 out of 5 stars

  4.5/5 (234)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  Rating: 3.5 out of 5 stars

  3.5/5 (738)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  Rating: 4 out of 5 stars

  4/5 (3811)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  Rating: 4.5 out of 5 stars

  4.5/5 (2409)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  Rating: 4 out of 5 stars

  4/5 (74)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M.L. Stedman

  Rating: 4.5 out of 5 stars

  4.5/5 (789)
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  Rating: 4 out of 5 stars

  4/5 (45)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  Rating: 3.5 out of 5 stars

  3.5/5 (792)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  Rating: 3.5 out of 5 stars

  3.5/5 (104)
• Dytran 2012 Example Problem Manual PDF

  

  Document538 pages

  Dytran 2012 Example Problem Manual PDF

  AdamD

  No ratings yet
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  Rating: 3.5 out of 5 stars

  3.5/5 (137)
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  Rating: 4 out of 5 stars

  4/5 (104)
• Srs Syn Dsine

  

  Document12 pages

  Srs Syn Dsine

  AdamD

  No ratings yet
• Abaqus Example Problems Manual - 6

  

  Document1 page

  Abaqus Example Problems Manual - 6

  AdamD

  No ratings yet
• Sinclair Interplanetary SSC07-X-3

  

  Document9 pages

  Sinclair Interplanetary SSC07-X-3

  AdamD

  No ratings yet
• Modeling and Analysis of Spacecraft Stru AcousticExcit PDF

  

  Document8 pages

  Modeling and Analysis of Spacecraft Stru AcousticExcit PDF

  AdamD

  No ratings yet
• Comparison of Nonlinear Random Response Using Equivalent Linearization and Numerical Simulation

  

  Document14 pages

  Comparison of Nonlinear Random Response Using Equivalent Linearization and Numerical Simulation

  AdamD

  No ratings yet
• Global Stiffness Analysis of Biw Structure: Sharanbasappa, E Sujith Prasd, Praveen Math

  

  Document6 pages

  Global Stiffness Analysis of Biw Structure: Sharanbasappa, E Sujith Prasd, Praveen Math

  AdamD

  No ratings yet
• JUI-RPWI-ESA-EBOX-RP-Mechanical Analysis Report RPWI - Issue - 2 - Rev - 0 - Signed

  

  Document134 pages

  JUI-RPWI-ESA-EBOX-RP-Mechanical Analysis Report RPWI - Issue - 2 - Rev - 0 - Signed

  AdamD

  No ratings yet
• Patran 2019: Release Guide

  

  Document34 pages

  Patran 2019: Release Guide

  AdamD

  100% (1)
• Nastran 2019 Doc Release

  

  Document116 pages

  Nastran 2019 Doc Release

  AdamD

  No ratings yet
• ANSYS 7.0 CAD Geometry Connections, Geometry Plugins, and Geometry Readers

  

  Document6 pages

  ANSYS 7.0 CAD Geometry Connections, Geometry Plugins, and Geometry Readers

  AdamD

  No ratings yet
• RBEs MPCs in MSC - Nastran

  

  Document91 pages

  RBEs MPCs in MSC - Nastran

  AdamD

  No ratings yet
• Constitutive Modeling of Elastomers - Accuracy of Predictions and Numerical Efficiency

  

  Document19 pages

  Constitutive Modeling of Elastomers - Accuracy of Predictions and Numerical Efficiency

  AdamD

  No ratings yet
• Chapter 27: Large Sliding Analysis of A Buckle

  

  Document12 pages

  Chapter 27: Large Sliding Analysis of A Buckle

  Dan Wolf

  No ratings yet