Sunteți pe pagina 1din 9

R. L.

Dai
Mem. ASME

Fabrication, Testing, and Modeling of Carbon Nanotube Composites for Vibration Damping
In this research, carbon nanotube (CNT) composite with epoxy resin is fabricated. The dynamic properties of the nanotube composites are evaluated. A testing apparatus for obtaining dynamic properties of composites is set up, and measurement procedures are given. In particular, the loss factors together with stiffness are measured for the specimens with different CNT weight ratios. Experimental results show that CNT additive can provide the composite with several times higher damping as compared with pure epoxy. A composite unit cell model containing a single CNT segment is developed by using the nite element method. Composite loss factors are calculated based on the average ratio of the unit cell energy loss to the unit cell energy input. Calculated loss factors under different strain are compared with experimental data. With the validated model, a parametric study is subsequently performed. Parameters, such as CNT dimension and CNT alignment orientation, are studied. The factors that lead to higher composite damping capacity are identied. DOI: 10.1115/1.3147126 Keywords: carbon nanotube, composite, damping, nite element method

W. H. Liao1
Mem. ASME e-mail: whliao@cuhk.edu.hk Department of Mechanical and Automation Engineering, Chinese University of Hong Kong, Shatin, New Territories, Hong Kong

Introduction

Damping materials play an important role in vibration damping and control systems. Vibration can be reduced via damping materials by transferring the kinetic energy into heat. One of the widely used damping materials is viscoelastic material VEM . In general, the VEM damping effects are signicant when used in vibrating structures due to energy dissipation. However, the stiffness of traditional viscoelastic materials is low, thus large shear strain could be induced while having signicant energy dissipation. The softness property may constrain the VEMs from being used in structures with high strength-to-weight ratio, where the strength and/or transmissibility issues are crucial. On the other hand, epoxy resin is much stiffer than the VEMs and could be used to enhance structural stiffness; but it dissipates little energy from the structure. Considering the above situations, it is desirable to develop alternative means for signicant vibration damping while the stiffness and strength of the material/structure are high. It may be achievable when composites made of carbon nanotubes CNTs and epoxy resins are used. Some foresaw that carbon nanotubes would bring us a revolution in scientic research since the multiwalled carbon nanotubes MWNTs were discovered 1 . It was recognized that CNTs might be the reinforcing materials for developing composites with excellent properties, such as high elastic moduli, on the order of 1 TPa 2 , as well as superior electrical and thermal properties 3 . One issue in making a composite is the selection of base matrix. In this research, the epoxy resin is chosen as the matrix for the composite because of its excellent adhesion, a lower order of shrinkage during cure as compared with other competitive materials, and a good balance of physical and electrical properties. Discussions on epoxy resins entail examination of the basic epoxy polymers, curing agents, and modiers if required.
1 Corresponding author. Contributed by the Technical Committee on Vibration and Sound of ASME for publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received June 28, 2008; nal manuscript received April 21, 2009; published online September 10, 2009. Assoc. Editor: I. Y. Steve Shen.

Researches have been conducted in the eld of nanostructures enhanced polymer resin composites. Most of those researches focused on the strength, toughness, thermal conduction, and electrical conduction enhancement 47 . Recently, researchers started to pay attention to the damping enhancement of the composites due to the CNT addition. The unique damping properties of CNT based structures, such as densely packed CNT thin lms 811 and polymeric composites with dispersed CNT llers 1216 , have been of great interest. In the research of Koratkar et al. 8,9 , the CNTs were densely packed on a silicon substrate before being coated with epoxy. This treatment would bring signicant damping, but fabrication could be costly and difcult. In addition, the weight ratio of CNT is high in this case, which may not be economical. Zhou et al. 12,13 fabricated composites with randomly dispersed CNT addition. The loss factor was measured by applying normal loading on the composite bar. As the damping mechanism was assumed to be the interfacial sliding between CNT segments and base matrix, shear loading would bring out this interaction more easily. In addition, as we can see from ISO 17 and ANSI standards 18 for material damping measurement, shear loading and/or bending loading are commonly used. Zhou et al. rst developed a stick-slip damping model for composites containing well-dispersed single-wall carbon nanotubes SWNTs . Analytical results showed strain-dependent damping enhancement due to stick-slip motion at the interfaces of the SWNTs and the resin. Following Zhou et al.s effort, Liu et al. 15,16 presented an analysis on the structural damping characteristics of polymeric composites containing dilute randomly oriented nanoropes. The analytical results indicated that the loss factor of the composite is sensitive to stress magnitude. Odegard et al. 19 also gave an analytical model to simulate nanotubereinforced polymer composites. Liu and Chen 20 used the boundary element method BEM for analyzing carbon nanotubebased composites. It should be noted that the damping performance of CNT composites was not considered by the latter two research groups. OCTOBER 2009, Vol. 131 / 051004-1

Journal of Vibration and Acoustics

Copyright 2009 by ASME

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Chemical reagents used in this research WSR618, Blue New Chemical Materials Co. Ltd., China D.E.R. 331, DOW Chemical Co. , epoxy resin Appearance No visible mechanical impurity Epoxide value eq/100 g 0.480.54 Unhydrolytic chlorine eq/100 g 0.0008 Hydrolytic chlorine eq/100 g 0.014 150 C, 40 min volatile % 1.5 40 C viscosity mPa s 2500 WSR6101, Blue New Chemical Materials Co. Ltd., China, epoxy resin Appearance No visible mechanical impurity Epoxide value eq/100 g 0.410.47 Unhydrolytic chlorine eq/100 g 0.0008 Hydrolytic chlorine eq/100 g 0.014 Volatile % 1.0 Softening point C 1220 WSR9804R, Blue New Chemical Materials Co. Ltd., China, modied toughened epoxy resin Appearance No visible mechanical impurity, light yellow Epoxide value eq/100 g 0.450.50 Volatile % 1.0 Specic gravity 1.14 0.03 40 C viscosity mPa s 9001300 WSR6690, Blue New Chemical Materials Co. Ltd., China, modier, active diluent Appearance No visible mechanical impurity Epoxide value eq/100 g 0.260.32 1.0 Volatile % 40 C viscosity mPa s 40100 WSR1224, Blue New Chemical Materials Co. LTD., China, curing agent Appearance No visible mechanical impurity, light brown mg KOH/g 300360 25 C viscosity mPa s 7090 DEH58, DOW Chemical Co., curing agent Appearance Dark brown Amine hydrogen equivalent weight g/eq 2832 25 C viscosity mPa s 85130
The epoxy or amine hydrogen equivalent weight g/eq means the weight of resin in grams, which contains 1 g equivalent of epoxy or amine hydrogen . Another related term is epoxy value eq/100 g .

This research is aimed at studying the damping effect caused by CNT addition. General procedures for fabricating CNT/epoxy composites are given. A set of material testing equipment is designed and constructed. The fabricated composite specimens are evaluated under different loading conditions by varying excitation frequencies and amplitudes. Storage modulus and loss factors are measured for both pure epoxy specimen and CNT composite specimens. In particular, CNT addition is randomly dispersed in fabrication, and shear loading is applied during the damping measurement in this research. A nite element method FEM based model for CNT composite is developed. A composite unit cell model is built. FEM based composite shows the detailed sliding surface status developed between CNT and base epoxy matrix, while none of previous models do. Under the shear loading, simulation is carried out to investigate interfacial sliding under different strain level and CNT orientation. Loss factors can be obtained from the ratio of energy dissipation out of energy input per cycle. The simulation results are compared with experimental data. Based on the FEM model, further parametric studies are performed. Several parameters, such as CNT dimension and CNT alignment orientation, are discussed with respect to composite damping capacity loss factor .

Composite Fabrication

The mechanical properties such as strength and stiffness of the commercially available MWNTs are superior to epoxy resins. The typical Youngs modulus of MWNT is 1 TPa, while that of pure 051004-2 / Vol. 131, OCTOBER 2009

epoxy resin is around 3 GPa. In this research, MWNTs MWNTA-P, Sun Nanotech are chosen as the composite additive. The CNTs were produced through CVD process and their diameter is 1030 nm. SWNTs could have better damping performance as compared with MWNTs. However, as a SWNT is much thinner than a MWNT, the difculty of dispersing CNTs into epoxy resin is much higher, and the price of SWNT is much higher too. In this research, only MWNTs are used for experiments, but the simulation will give a prediction on damping characteristics of the composite with SWNTs. There are numerous types of epoxy resins with different chemical and mechanical properties. Three types of epoxy resins, two types of curing agents hardeners , and one modier toughen diluent, in order to overcome the brittle disadvantage are used in this research. All chemical reagents used in this research are provided by Blue New Chemical Materials Co. Ltd. and DOW Chemical. The detailed information of epoxy resins, curing agents, modier, and abiodiluent are given in Table 1. Because of the nanoscale, the CNTs tend to aggregate. To fabricate a qualied composite, the carbon nanotubes should be dispersed uniformly into the epoxy base matrix. One way is to use chemical functionalization. However, this method disrupts the bonding of the graphene sheet and reduces the mechanical properties of the nal composite 21 . An alternative way to disperse the CNTs into the epoxy resin is sonication by using an ultrasonic processor. It has been reported that while dispersing the CNTs into the epoxy resin by sonication, the energy input will break the Transactions of the ASME

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 1 CNT segments circled in matrix

CNTs into shorter segments if the sonication energy is relatively high 21 . This might affect some of the mechanical properties of the composite such as strength and hardness. It is anticipated that the damping ability would not be affected signicantly because the energy dissipation is assumed to be caused by interfriction between nanotubes and the base matrix 12,13 . The experimental results that will be given in Sec. 3 show that the storage modulus of CNT composite is similar to that of pure epoxy, while the simulation results show that the composite loss factor is insensitive to CNT segment length. In this case, a strong sonication is performed in order to have a relatively homogenous CNT distribution. The viscosity of base epoxy matrix is another issue while dispersing CNTs. At the room temperature, the viscosity of selected epoxy is too high for CNT dispersion about 2000 MPa s . The viscosity drops to about 1000 MPa s when the epoxy is heated above 50 C; however, in this situation, the curing speed is too fast the epoxy solidies almost immediately . Therefore, abiodiluents, such as acetone and ethanol, are used to help disperse carbon nanotubes into epoxy. After dispersion, the diluents are evaporated before the curing agent is added. After the composite is fabricated, scanning electron microscope SEM images are taken in order to observe the status of CNT segments within base matrix. A eld emission scanning electron microscope FESEM, Leo 1530 is used here. The CNT composite specimens with 2% CNT additives are broken in order to expose the internal cross sections. Before the SEM images are taken, gold is sputtered on the samples to make the observing surfaces conductive. Images are taken at different locations of the exposed internal cross section. More than 100 images are taken, and CNT segments are found in less than 20 of them, as the CNT weight ratio is relatively low and the SEM scale factor is set to a high level, in order to observe CNTs with diameter of around 20 nm. One of the SEM images is shown in Fig. 1. The CNTs are broken into shorter segments by sonication as expected, and the dispersion quality is acceptable in general. Journal of Vibration and Acoustics

Measurement of Dynamic Material Properties

In order to evaluate the damping ability of the fabricated CNT/ epoxy composite, a material property testing scheme is proposed. Based on International Standard 17 for testing plastics and ANSI for testing material damping 18 , a method for measuring dynamic material properties of CNT composites is given. The specimens are tested under cyclic nonresonance shear loading. The original standards are generally for VEMs which are much softer and plastics which have much lower damping . The testing scheme and prototyping specimen are modied for suitable measurements. The output of this testing scheme is storage modulus and loss factor. For materials with damping effect, the material property can be modeled by using complex modulus. The storage modulus is related to stiffness, and the loss modulus is related to damping and energy dissipation. It should be noted that the loss factor of composite is the ratio of energy dissipated, as the result of damping during one cycle, to the total energy of the vibrating system over a given period of time. The expression of complex modulus is given as follows: G T, f = G1 + iG2 = G1 1 + i 1

where G1 is the storage modulus, G2 is the loss modulus, = G2 / G1 is the loss factor. More specically, the storage modulus and loss factor can be calculated from the following equations 17 : Ga = FA sA = 1+ L2 h2 tan = tan
G

G E
Ga

cos

Ga

1 ka/k cos

3
Ga

where Ga, FA, sA, L, h, G , and E are apparent shear storage modulus, measured amplitude of the dynamic force, measured OCTOBER 2009, Vol. 131 / 051004-3

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 2 Experimental setup Fig. 3 Measured loss factor versus excitation frequency

amplitude of dynamic displacement, specimen dimension between bonded faces, specimen height, shear modulus, and Youngs modulus, respectively. ka and k are measured stiffness of experimental setup with specimen and measured stiffness of experimental setup with rigid metal bar, respectively. Correction is needed if the compliance of testing apparatus cannot be ignored. Ga and G are measured and corrected phase difference between displacement and force signals, respectively. The value of G / E is set to 0.37 a typical value , while FA, sA, L, and h can be directly measured from experiments. For specimens with a relatively low Youngs modulus and testing apparatus with high stiffness, ka / k tends to be zero i.e., the correction term 1 ka / k cos Ga tends to be 1 , hence, the loss factor can be approximated as = tan
G

tan

Ga

Specimens are fabricated in a dumbbell shape in order to obtain larger strain and stress. It is expected that larger strain induces more complete CNT/matrix sliding thus giving a larger loss factor. Although for a dumbbell shaped specimen both bending load and shearing load are effective, the bending effect is cancelled out by the correction term in bracket in Eq. 2 . The correction considers the length and cross section of the specimen, thus the calculated loss modulus is shear modulus 17 . Pure epoxy specimens and 2% CNT enhanced composite specimens are fabricated using the combination of D.E.R. 331 and WSR1224 for higher matrix stiffness. A testing apparatus with clamps is designed in order to apply controllable and stable shear loading onto specimens. The specimens are inserted into the cylindrical blocks and fastened by glue and screws in order to minimize the displacement between specimens and the clamps. The apparatus are made of aluminum Tshaped clamps and blocks and steel U-shaped clamps . A calibration experiment using rigid specimen made of structural steel was performed to ensure that the clamping boundary and the deections of testing apparatus itself do not bring signicant damping. The experimental results show that the phase difference between input and output sinusoidal curves is very small, which means that the damping from these aspects could be ignored. As we can see from Fig. 2, the U-shaped clamp is mounted on a shaker LW-127500, Labworks , and the T-shaped clamp is xed on the top through a load cell PCB Piezotronics Inc. 208C . The time domain displacement signal of U-shaped clamp is recorded by using a laser vibrometer Polytec OFV 303, 3001 . At the same time, force signal is also recorded through the load cell at the top of the T-shaped clamp. These signals are recorded and processed by a fast Fourier transform FFT analyzer ONO SOKKI CF-3400 . The phase difference between these two sig051004-4 / Vol. 131, OCTOBER 2009

nals is calculated. In this experiment, both pure epoxy and CNT composite specimens 2% CNT epoxy composite are tested. First, a frequency sweep test is performed. The frequency range is from 80 Hz to 400 Hz. It should be noted that frequency around 280 Hz is skipped to avoid resonance a nite element analysis was performed in advance to obtain structural resonant frequencies . Loss factor is calculated from the phase difference of the input/output signals. It can be seen from the experimental results in Fig. 3 that the composite loss factor is greatly enhanced by the CNTs as compared with that of pure epoxy. In particular, the loss factor is enhanced by up to 400% 0.2 versus 0.05 around 300 400 Hz . Experimental results also show that the damping due to CNTs is frequency dependent. The difference of the loss factors between two specimens is bigger while sweeping to higher frequency within the testing frequency range. Further testing on the composites is performed with xed excitation frequency at 300 Hz. This frequency is away from a structural resonance and signicant damping from the CNT composite was found within the observed range using the existing equipment. In this experiment, excitation amplitude displacement of the shaker that is related to shear strain is varied. The gain of the power amplier for the shaker is gradually increased, and the corresponding strain ranges from 0.00022 to 0.001, which are calculated from the shaker tip displacement and specimen dimensions. It can be seen from Fig. 4 that the difference on the damping between the pure epoxy and CNT composite specimens is inconspicuous when the input strain is 0.00022. When shear strain is above 0.0005, we can see that more than 350% damping enhancement is obtained. The sixth order polynomial is used here for curve tting the data.

Fig. 4 Measured loss factor versus strain

Transactions of the ASME

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 6 Finite element model of composite unit cell

Fig. 5 Measured storage modulus versus strain

In addition, specimens with the same cross-sectional area as the epoxy and CNT composite specimens containing 3M ISD112 VEM are fabricated and tested. It can be seen from Fig. 4 that the measured VEM loss factor is higher than that of the CNT composite. By taking an average of lost factors for strain level ranging from 0.0005 to 0.001, the average loss factor is 0.35 versus 0.25 VEM versus CNT composite . Another experiment is performed to measure the storage modulus under different strain levels. The experimental results are shown in Fig. 5. From the gure, we can see that the addition of 2 wt % CNTs does not signicantly affect the material storage modulus of the epoxy. Based on the interfacial sliding assumption, a higher strain induces a larger sliding area between CNT segments and base matrix; therefore, more energy is dissipated. The storage modulus of the CNT composite is at a similar level as compared with the pure epoxy under different strain. On the other hand, it can be seen from the gure that the VEM storage modulus is much lower than those of both pure epoxy and CNT composite. The storage modulus of a typical VEM 3M ISD 112 at room temperature/300Hz is about 7 MPa, while it is over 2 GPa for the epoxy resin/CNT composite. As discussed above, the CNT composite is much stiffer 2 GPa versus 7 MPa than the VEM, while its loss factor is comparable given a certain level of strain 0.25 versus 0.35 . It is promising that the CNT composites could be used in the structural vibration applications when both damping and stiffness are important.

3. Interfacial friction transfers mechanical energy to thermal energy thus dissipates energy out of system. 4. The CNTs in composite are assumed to be straight. 5. CNT segments are well separated, and every single CNT segment is isolated within a small epoxy resin cubic, which is called the composite unit cell. The dimension of this unit cell is calculated from the CNT volume ratio. This assumption would be acceptable when the damping mainly comes from CNT/epoxy sliding but not CNT/CNT friction. 6. CNTs in matrix are randomly oriented and uniformly distributed so that we can average the energy loss over all CNT segments. Based on assumptions 4 and 5, the CNTs in the composite are modeled as straight segments isolated in composite unit cells. It should be noted that the ultrasonic process breaks CNTs into shorter segments while at the same time reduces the segment curvature. For CNT composite fabricated in this research, it can be observed from the SEM image that the CNT dispersion quality is good in general. The aspect ratio of CNTs is also found to be decreased as no segment longer than 800 nm is observed in these images. In addition, considering CNT waviness will greatly increase modeling difculty to calculate the average energy dissipation. The composite unit cell cube-shaped containing single CNT segment is modeled in 3D, as shown in Fig. 6. Extra blocks are mounted on the top and bottom of this unit cell in the model in order to facilitate applied loading. All DOFs are xed on one of the extra blocks, while vertical loading calculated from desired average strain and the dimension of composite unit cell and moment on the opposite direction in order to cancel out the bending effect from vertical loading are applied on the other block. The dimension of this composite unit cell is obtained from the CNT weight ratio. Higher CNT weight ratio gives smaller CNT composite unit cell. Detailed dimension and properties are given in Table 2. The CNT segment is modeled as a hollow tube with rounded ends, and a thin sheath tube is built in order to simulate the contact/bonding surface between CNT and matrix see Fig. 7 . This sheath layer plays a very important role in damping simulation. Some of the over stressed elements in this layer will be deleted in order to simulate the interfacial sliding mechanism. The Youngs modulus and Poissons ratio of sheath layer elements are set to be the same as epoxy resin, but a failure strain 0.0016 is set in order to simulate the interfacial sliding mechanism. The value of this critical sliding strain can be determined from experiments and numerical analysis, which will be discussed in Sec. 5. The model is rst constructed with boundary conditions and the initial loading applied. It is then solved by using ANSYS implicit sparse solver and DYNA explicit solver for highly nonlinear cases such as large displacement. As CNT is much stiffer than base matrix, strain near CNT/ matrix interface is larger than the average strain in composite unit cell. Once the critical sliding strain is determined, the sheath elements over this critical strain during each load step are detected OCTOBER 2009, Vol. 131 / 051004-5

Finite Element Modeling

The nite element method is used for modeling the CNT composites. The chosen platform is ANSYS 10.0/MULTIPHYSICS. 3D solid element SOLID185 is used. SOLID185 is dened by eight nodes having three degrees of freedom at each node: translations in the x, y, and z directions 22 . This element can be used in relatively large deformation cases, and there is no redundant degree of freedom that brings extra computational load. It was assumed that the energy dissipation in CNT composite mostly comes from interfacial sliding between CNT and base matrix. In addition, the base matrix i.e., epoxy resin itself also gives a certain contribution to the composite damping, which is assumed to be constant with respect to strain, and the loss factor of the base matrix is equal to that of pure epoxy resin. Several assumptions are made in order to model CNT composites. 1. Damping consists of two parts: contribution from epoxy resin matrix and contribution from CNT/matrix interfacial sliding. 2. CNT and base matrix are perfectly bonded at initial state before loading is applied. As the interfacial strain increases with applied shear loading, sliding occurs when the critical strain is reached. Journal of Vibration and Acoustics

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 2 CNT segment Length Outer diameter Thickness Youngs modulus Poissons ratio 100 nm 20 nm 4 nm 1000 GPa 0.3

Model parameters Sheath layer Thickness Youngs modulus Poissons ratio Critical sliding strain 2 nm 3 GPa 0.37 0.0016

Epoxy resin unit cell Cubic edge length Youngs modulus Poissons ratio Loading 300 nm 3 GPa 0.37 Shear

and removed from the model in order to simulate the debonding between nanotube segment and base epoxy matrix. Then, another set of loading is applied after the preset time interval. The model is solved again, and more elements are removed in the new round of calculation. The number of removed sheath tube elements is counted after each loading cycle in order to calculate the cyclic energy dissipation. After the solver nishes computing, the energy loss during one cycle is calculated by adding the strain energy of all deleted elements Uloss =
1 2

cr crv m

where v, m, cr, and cr are volume of sheath element, number of deleted sheath elements, critical sliding stress, and critical sliding strain, respectively. The total element energy storage element strain energy during this cycle is also calculated from the model. By using these two energies, we can calculate composite loss factor. More specically, from simulated strain/stress distribution, the stored energy energy input of a composite unit cell per loading cycle, Ucell can be calculated as follows: Ucell =
1 2

cell cellV n

where V, n, cell, and cell represent the volume of element in the composite unit cell, the total number of elements in the composite unit cell, and the calculated von Mises equivalent stress and strain 22 , respectively. The loss factor is the ratio of energy dissipated, as the result of damping during one cycle, to the total energy of the vibration system under cyclic loading. The loss factor of the unit cell could be obtained as follows: Uloss = Ucell
1 2

cr crv m

Eepoxy =
m

2 crv

1 2

cell cellV n n

Ecell

2 cellV

Simulation Results

where the sliding occurs on CNT/matrix interface and how the sliding develops. These conditions have not been considered by previous researches. Figure 8 shows interfacial sliding examples between CNT segment and base matrix in two different CNT orientations. The zenith angle and azimuth angle of the CNT segment are denoted by two parameters and , as shown in Fig. 8 to describe CNT orientation with respect to global coordinate system. The upper two show the sliding status when CNT segment is perpendicular to the loading surface = 90 deg and = 90 deg . The lower two in Fig. 8 illustrate the sliding status when CNT segments zenith angle and azimuth angle are 30 deg and 90 deg, respectively. The sheath layer elements gray , shown in the middle of Fig. 8, are the ones over critical strain, which are removed from the model in order to simulate the sliding effect during each loading cycle. More studies on CNT segments zenith and azimuth alignment angles will be given in Sec. 6. Other than CNT orientation, the interfacial sliding area is also related to the applied strain level. Figure 9 shows CNT segment sliding status under different strain levels, i.e., 0.0001, 0.0002, 0.0003, 0.0005, 0.0006, 0.0007, 0.0008, and 0.001, respectively, from a to h . The sliding area is developed along with strain level higher strain induces larger sliding area. Shear loading is applied on the composite unit cell. CNT orientation parameters and range from 0 deg to 90 deg, while the step size is 7.5 deg, which leads to 13 13= 169 different cases built in the rst octant for each loading level. The angular range covers one eighth of the 3D coordinate system so that it is sufcient for our study due to geometric symmetry. The number of deleted sheath elements is recorded in terms of and under different strain levels. In this model, the total number of elements in the sheath layer nmax = 1973, which corresponds with the maximum number of elements involved in interfacial sliding. In other words, at most 1973 elements could be removed during sliding to result in the largest possible energy dissipation. In general, higher strain level leads to larger sliding area that induces more energy dissipation; the different CNT alignment causes a different sliding area size and location under the same loading. Based on simulation results for 169 cases, the average number of sliding sheath elements is calculated for each loading level. The average energy loss of each composite unit cell can be further

In this research, the interfacial sliding between CNT segments and base epoxy resin is modeled. The model shows when and

Fig. 7 CNT model with sheath layer

Fig. 8 Deleted sheath elements during CNT/epoxy sliding

051004-6 / Vol. 131, OCTOBER 2009

Transactions of the ASME

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 9 CNT/epoxy interfacial sliding status under different strain level, = 7.5 deg, = 90 deg, cell = 0.0001, 0.001

obtained. The average number of sliding elements, number of maximum sliding elements, number of minimum sliding elements, and standard deviation of sliding number w.r.t. CNT segment orientation associated with strain range 0.0002, 0.001 are given in Table 3. The loss factor of the CNT composite thus can be calculated from Eq. 7 . The calculated loss factor of the CNT composite is compared with the experimental data, as shown in Fig. 10. We can see from this gure that the calculated and measured loss factors of the CNT composite are in the similar trend and level with respect to strain. We can also see from the gure that both experimental data and simulation results show peaks around strain range 0.00060.0008. It could be explained that most CNT segments are fully debonded when strain is above this level, nearly the full sheath layer is deleted over 1940 elements are deleted . The interfacial sliding is almost complete. In this case, regardless of the CNT alignment, the energy loss of the composite unit cell nearly reaches its upper bound and the composite almost gives maximum energy loss Uloss. However, as total energy input Ucell keeps going up as the strain level is further increased beyond the energy loss upper bound, the loss factor calculated from Eq. 7 would drop. In this simulation, the CNT/matrix interfacial critical sliding strain is set to 0.0016. The critical strain is determined from experimental data. However, from available experimental data and other references 23,24,16 , the critical strain can only be estimated. A range of critical strain is varied to nd the one that gives the loss factor curve the best t with experimental data. Simulation results for different critical strain values are shown in Fig. 11, where critical strain value is set to 0.0012, 0.0014, 0.0016, and 0.0018. We can see from this gure that the simulation with c = 0.0016 ts experimental data the best, and thus this value is used for further simulation in this research. It can also be seen from Fig. 10 that simulated loss factor is smaller than the one measured from experiment for a given strain. It should be noted that only the contributions from epoxy resin matrix and CNT/matrix interfacial sliding were considered when modeling damping, as stated in the rst assumption. Some other mechanisms could also dissipate energy. CNT/CNT stick-slip motion could be one of these mechanisms while energy dissipa-

tion from bubbles/cavities inside epoxy matrix might be another. These mechanisms could be further considered for more accurate models.

Parametric Study

The validated model could be used for predicting CNT composite dynamic properties and guiding composite fabrication. Several factors could affect the energy dissipation ability of the CNT composite, such as the dimensions of CNTs and the alignment orientations of CNTs inside the matrix. These key factors will be studied in order to provide some references for material selection and composite fabrication. There are several commercially available carbon nanotubes. A brief summary on geometrical properties of some major MWNTs is given in Table 4. As shown in the table, typical diameter of commercially available MWNTs ranges from 5 nm to 50 nm while the length ranges from 200 nm to 20 m. For the studies in this research, CNT diameter range 10 nm, 50 nm and length range 100 nm, 600 nm are set as they cover most commonly used MWNTs. Another reason for the chosen CNT length range is that an ultrasonic process was performed during CNT dispersion. High energy sonication could break CNTs down to small segments, thus the selection of the length range 100 nm, 600 nm would be reasonable. CNT models with different CNT length while keeping the same CNT diameter are built. Loss factor versus strain with respect to different CNT lengths are shown in Fig. 12. These simulation results are obtained based on the same epoxy matrix and CNT weight ratio while assuming randomly oriented dispersion of CNTs in composite. It is observed that, given the same CNT diameter, CNT addition with a shorter segment length induces slightly higher composite damping when the strain is over a certain level around 0.00065 for the case shown in Fig. 12 , and longer CNTs bring out higher damping before this strain level. However, composite loss factor curves with different CNT length are generally close to each other. In other words, composite loss factor is not sensitive to CNT length. As it is hard to keep CNTs

Table 3 Number of deleted sheath elements under different strain level Strain Average Max Min Std. dev. Loss factor 0.0002 92 352 0 95 0.095 0.0003 441 746 157 167 0.141 0.0004 957 1174 707 136 0.159 0.0005 1333 1755 1033 159 0.149 0.0006 1586 1945 1220 155 0.133 0.0007 1725 1966 1365 143 0.117 0.0008 1810 1973 1463 123 0.104 0.0009 1911 1973 1577 84 0.096 0.001 1961 1973 1694 55 0.089

Journal of Vibration and Acoustics

OCTOBER 2009, Vol. 131 / 051004-7

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 10 Comparison of loss factor between simulation and experiment

Fig. 12 Loss factor versus strain with respect to different CNT length, CNT diameter D = 20 nm; L represents CNT segment length in nanometers

with a high aspect ratio instead of a low one, this observation would keep the cost of composite fabrication relatively low. Another study is performed with respect to CNT diameter in range 10 nm, 50 nm . Other geometrical parameters, including CNT segment length, diameter, and the dimension of composite unit cell, are kept the same. Loss factor versus strain with different CNT diameters are shown in Fig. 13. It can be seen that the effect of CNT diameter on the composite loss factor is signicant. More specically, larger CNT diameter gives more energy dissipation for a single composite unit cell as increasing the CNT diameter enlarges the surface area of the CNT segment. A larger composite loss factor is thus induced when the number of CNT segments in composite is the same. Additional analysis on the loss factor with different CNT diameters is performed, while the CNT weight ratio is xed at 2%. The results are shown in Fig. 14. It can be observed from the gure that thinner CNTs induce higher composite damping. The maximum composite loss factor with 10 nm CNT additives is 0.4, while the one with 50 nm CNT additives is 0.16. The reason is as

follows: although a smaller diameter gives smaller surface area for a single CNT segment, a larger total sliding surface is induced as the number of CNT segments in composite increases when CNT diameter decreases, given the same CNT weight ratio. Therefore, if the weight ratio of CNT additives is the same, CNTs with a smaller diameter can be chosen to obtain higher composite damping. If cost is not a big concern, we can even use SWNTs, which have a much smaller diameter only a few nanometers . As mentioned, CNT/epoxy interfacial sliding status is related to CNT alignment orientation in matrix. The model is used here to

Fig. 13 Loss factor versus strain with respect to different CNT diameter, D represents CNT diameter and CNT segment length L = 200 nm

Fig. 11 Simulated composite loss factors with different critical strains Table 4 Geometrical properties of commercial MWNTs Supplier NanoLab Type MWNT 1 MWNT 2 MWNT 3 MWNT 4 MWNT 1 MWNT 2 MWNT 3 MWNT MWNT Length 15 m 5 20 m 15 m 5 20 m 200 nm 300 nm 500 nm 400 nm 1 m 2 m Diameter 15 5 nm 15 5 nm 3015 nm 3015 nm 6.5 nm 12 nm 20 nm 1030 nm 250 nm

Rosseter

SunNano n-Tec

Fig. 14 Loss factor versus strain with respect to different CNT diameter, D represents CNT diameter and CNT segment length L = 200 nm, CNT weight ratio is xed for each case

051004-8 / Vol. 131, OCTOBER 2009

Transactions of the ASME

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 15 CNT/epoxy interfacial sliding status by varying = 90 deg

, = 0 deg, 90 deg,

study the relationship between CNT orientation and composite damping ability. Figure 15 shows simulation results regarding CNT/matrix interfacial sliding status in different CNT segment orientation, while CNT segments are aligned on the x-z plane, swept from the negative z toward the positive x = 0 deg, 90 deg , = 90 deg . As we can see from this gure, interfacial sliding occurs along the CNT longitudinal direction when = 0 deg. The sliding surface location moves to both ends of the segment while CNT rotating around y axis with = 52.5 deg.

8 9

Conclusion

In this paper, CNT/epoxy composite specimens with different CNT weight ratios were fabricated. Dynamic testing apparatus and method for evaluating damping and stiffness of CNT/epoxy composites were also introduced. The loss factors were measured and analyzed under different frequencies and strains. Experimental results showed that the CNT additives greatly enhanced more than 350% enhancement the damping capacity of pure epoxy resins. Fabricated composite has the same level of storage modulus as epoxy resin and a similar level of loss factor as compared with VEMs, when the strain is over a certain level. The observed phenomena match our anticipation and the results show promise in potential applications when both damping and stiffness are desired. FEM composite unit cell model was developed, and the loss factor of the CNT composite was calculated and compared with the experimental results. Simulation curve ts with experimental data in general. Parameters, such as CNT length, diameter, and alignment orientation, were investigated. It was found that the composite loss factor is insensitive to CNT segment length but highly sensitive to CNT diameter and CNT alignment.

10

11

12

13

14 15

16

17

Acknowledgment
This work was supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China Project No. CUHK4144/07E and CUHK Postdoctoral Fellowship Scheme 07/ERG/12 .

18

19

20

References
1 Iijima, S., 1991, Helical Microtubules of Graphitic Carbon, Nature London , 354, pp. 5658. 2 Yu, M. F., Lourie, O., and Dyer, M. J., 2000, Strength and Breaking Mechanism of Multiwalled Carbon Nanotubes Under Tensile Load, Science, 287, pp. 637640. 3 Thostenson, E. T., Ren, Z. F., and Chou, T. W., 2001, Advances in the Science and Technology of Carbon Nanotubes and Their Composites: A Review, Compos. Sci. Technol., 61, pp. 18991912. 4 Ajayan, P. M., Stephan, O., Colliex, C., and Trauth, D., 1994, Aligned Carbon 21 22 23

24

Nanotube Arrays Formed by Cutting a Polymer Resin-Nanotube Composite, Science, 265 5176 , pp. 12121214. Qian, D., Dickey, E., Andrews, C. R., and Rantell, T., 2000, Load Transfer and Deformation Mechanisms in Carbon Nanotube-Polystyrene Composites, Appl. Phys. Lett., 76, pp. 28682870. Ruan, S. L., Gao, P., Yang, X. G., and Yu, T. X., 2003, Toughening High Performance Ultrahigh Molecular Weight Polyethylene Using Multiwalled Carbon Nanotubes, Polymer, 44 19 , pp. 56435654. Lau, K. T., Shi, S. Q., Zhou, L. M., and Cheng, H. M., 2003, Micro-Hardness and Flexural Properties of Randomly-Oriented Carbon Nanotube Composites, J. Compos. Mater., 37 4 , pp. 365376. Koratkar, N., Wei, B., and Ajayan, P. M., 2002, Carbon Nanotube Films for Damping Applications, Adv. Mater., 14 1314 , pp. 9971000. Koratkar, N., Wei, B., and Ajayan, P. M., 2003, Multifunctional Structural Reinforcement Featuring Carbon Nanotube Films, Compos. Sci. Technol., 63, pp. 15251531. Lass, E. A., Koratkar, N. A., Ajayan, P. M., Wei, B. Q., and Keblinski, P., 2002, Damping and Stiffness Enhancement in Composite Systems With Carbon Nanotubes Films, Mater. Res. Soc. Symp. Proc., 740, pp. 371376. Lass, E., Ajayan, P., and Koratkar, N., 2003, Engineered Connectivity in Carbon Nanotube Films for Damping Applications, Proc. SPIE, 5052, pp. 141150. Zhou, X., Shin, E., Wang, K. W., and Bakis, C. E., 2004, Interfacial Damping Characteristics of Carbon Nanotube Based Composites, Compos. Sci. Technol., 64 15 , pp. 24252437. Zhou, X., Wang, K. W., and Bakis, C. E., 2004, Effects of Interfacial Friction on Composites Containing Nanotube Ropes, Proceedings of the 44th AIAA/ ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Paper No. AIAA-2004-1784. Suhr, J., Koratkar, N., Keblinski, P., and Ajayan, P., 2005, Viscoelasticity in Carbon Nanotube Composites, Nature Mater., 4, pp. 134137. Liu, A., Huang, J. H., Wang, K. W., and Bakis, C. E., 2006, Effects of Interfacial Friction on the Damping Characteristics of Composites Containing Randomly Oriented Carbon Nanotube Ropes, J. Intell. Mater. Syst. Struct., 17, pp. 217229. Liu, A., Wang, K. W., Bakis, C. E., and Huang, J. H., 2006, Analysis of Damping Characteristics of a Viscoelastic Polymer Filled With Randomly Oriented Single-Wall Nanotube Ropes, Proc. SPIE, 6169, pp. 275291. International Standard, 1996, PlasticsDetermination of Dynamic Mechanical PropertiesShear VibrationNon-Resonance Method, Paper No. ISO 6721-6. American National Standard, 1998, Resonance Method for Measuring the Dynamic Mechanical Properties of Viscoelastic Materials, Paper No. ANSI S2.22. Odegard, G. M., Gates, T. S., Wise, K. E., Park, C., and Siochi, E. J., 2002, Constitutive Modeling of Nanotube-Reinforced Polymer Composites, Report No. NASA/CR-2002-211760. Liu, Y. J., and Chen, X. L., 2003, Continuum Models of Carbon NanotubeBased Composites Using the Boundary Element Method, Electronic Journal of Boundary Elements, 1, pp. 316335. Andrews, R., and Wiesenberger, M. C., 2004, Carbon Nanotube Polymer Composites, Curr. Opin. Solid State Mater. Sci., 8, pp. 3137. ANSYS Inc., 2005, Theory Reference 10.0. Yu, M. F., Lourie, O., and Dyer, M. J., 2000, Strength and Breaking Mechanism of Multiwalled Carbon Nanotubes Under Tensile Load, Science, 287, pp. 637640. Wong, M., Paramsothy, M., Xu, X. J., Ren, Y., Li, S., and Liao, K., 2003, Physical Interactions at Carbon Nanotube-Polymer Interface, Polymer, 44, pp. 77577764.

Journal of Vibration and Acoustics

OCTOBER 2009, Vol. 131 / 051004-9

Downloaded 27 Jan 2011 to 117.211.100.18. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

S-ar putea să vă placă și