


Figure 1. A unit cell of the C_{60} crystal optimized by the extended bond order potential; the calculated lattice constant a = 14.4 Å. 
With the general implementation of nonbond interactions, various transport properties can also be readily calculated. Among the mechanical transport phenomena, thermal conductivity is one of the most difficult quantity to calculate, especially for solid state materials, because of the long phonon mean free path. However, under certain conditions, the thermal current correlation functions can be evaluated without extending the simulation system. Weak bonded solids and materials under high temperature have relatively short phonon mean free path that enables the numerical evaluation of thermal conductivity using small simulation cell. In addition to atomistic simulation, Boltzmann equation for transport phenonmena can also be used as a model to calculate material transport properties. Here, we will only demonstrate the ablility of the implemented potential to calculate thermal conductivity through atomistic simulation. The fundamental method is to make use of the GreenKubo relation,
where j(t) is the thermal current,
where h_{i} is the site i energy density. In our simulation, it represents particle energy density. In general, thermal conductivity is a tensor. For an isotropic material, the average thermal conductivity is measured in experiments. At room temperature, the thermal conductivity of C_{60} crystal is 0.4 W m^{1} K^{1}. Experiments also showed that the thermal conductivity of C_{60} crystal was nearly independent of temperature till 260 K. In Fig.2, we show the thermal currentcurrent correlation function.
Figure 2. The thermal currentcurrent correlation function at room temperature for the C_{60} fcc crystal. 
The average thermal conductivity is obtained from the integration of the correlation function. The integration curve is depicted in Fig.3. The integration converges to a constant value with certain fluctuation due to the fluctuation of correlation function. The simulated result for the C_{60} thermal conductivity at room temperature is 0.4 W m^{1} K^{1} with the fluctuation of 20%.
Figure 3. The integration curve for the C_{60} crystal thermal current correlation function. The final value converges to 0.5 (differing from the thermal conductivity by a constant factor). 
Besides the crystal packing of nano structure, the accurate energetic calculation of a single walled nanotube (SWNT) also requires accurate NB interactions to be considered. Nanotube has been of great interest for many people since its discovery.^{10} Both theoretical predictions and experimental results indicate that nanotubes are one class of the strongest material known to date. The pure Brenner potential has been used for SWNT.^{4} However, since it has no NB interactions, it can not be applied to multiwalled nanotube (MWNT) directly. For MWNT, partition method can be employed to impose NB interactions only across different layers. The drawback of this method is preventing chemical reactions between atoms that belong to different layers. Even for SWNT, NB interactions can greatly affect the structures and energetics. Gao et. al.^{10} demonstrated that large SWNTs have collapsed form that is a result of intralayer van der Waals interaction. Our implementation allows NB interactions to exist within each SWNT and across different tubes for MWNTs. Fig.4 shows the energetics of a (70,70) SWNT as a function of different conformations. For each structure, we applied constraints to restrict the distance between two atoms on the opposite sides across the diameter of the nanotube. Then each structure is optimized under the constraints. The energetics are depicted in Fig.4. The red line is obtained from implemented potential, and the blue line is calculated from the original Brenner potential. As we can see from the figure, the implemented potential gives the global optimal structure to be a collapsed form, and circular form is a metastable structure. The original bond order potential only predicts a single optimal structure that is circular form. The physical reason for the collapsed structure is that the van der Waals attraction between two flat area can lower the total energy. This flat region is an analog to graphite crystal stacking.
Figure 4. The energetics of a singlewalled nanotube (70,70) as a function of the tube conformation. The xaxis is the distance of constraints. 
For MWNTs, the inclusion of NB interactions is more important for many material properties. A double walled nanotube (DWNT) has an inter tube rotational mode. This mode is also a result of long range van der Waals interactions. Fig.5 plots the potential as a function of the relative rotational angle between two nanotubes. The outer tube is a (15,15) SWNT, and the inner one is a (10,10) tube. The space between them is 3.48 Å, which is close to the graphite inter layer distance (3.3545 Å). This motion is a very low frequency collective mode. When only original form of Brenner potential is used, the relative rotation resembles a completely free motion, since each tube is not able to feel the existence of another one. Only when are NB interactions considered, the correct dynamics can be extracted. From Fig.5, the rotational barrier is 10 cal/mol/layer. The potential has 5 repeating units, which is the maximum common divider of (10,10) tube and (15,15) tube repeating unit number.
Figure 5. The rotational potential as a function of relative rotational angle between inner and outer tubes The period for the rotational potential is 72°. 
One of the most important properties for nanotube is its high strength. Theoretical calculations and experimental results suggest that the Young's modulus for nanotube is on the order of 1TPa. As a comparison, the Young's modulus for diamond, the most known strong material, is about 1TPa. In our previous papers,^{2,3} we have reported the Young's modulus for SWNT and DWNT is around 940 GPa in the elastic regime. In future applications of nanotubes, people are interested the mechanical properties not only in the elastic regime, but in the plastic regime and the regime where mechanical failure occurs. For a SWNT, the uniaxial stretching mostly involves the covalent bond stretching and bending. In general, chemical bonds are orders of magnitude stronger than NB interactions, therefore, NB forces only contribute negligibly to the overall deformation. However, for MWNTs, condensed phase of SWNTs, and composite materials, NB interactions can be very important when the deformation involves the inter tube sliding, crossing, and compression. As an example in this paper, we only consider the mechanical behavior of a SWNT and a DWNT under dynamic stretching. For DWNT, no inter tube slipping is allowed. Therefore, NB interactions are not very crucial in the calculation.
A single nanotube of 20 layers long is used in the calculation. For a (10,10) tube, it contains 800 atoms, and (15,15)/(10,10) DWNT consists of 2000 atoms. One end of the tube is held stationary, and a constant velocity stretching (constant strain rate) is applied to the other end. Fig.6 shows the force and energy of the (10,10) SWNT during the stretching process. The red and blue curves are calculated from the implemented potential, and the black and green curves are obtained form the original Brenner potential. From the force evolution, it is clear that the tube undergoes elastic deformation first and then plastic deformation. The discontinuity occurs when the SWNT suffers the mechanical failure (disruption). As we can see from Fig.6, both the implemented and the original Brenner potential give very similar results. The implemented potential requires larger force and more energy because there is a small potential barrier created by NB interactions before disruption happens. However, the difference is small. More interestingly, Fig.6 shows that force constant starts to increase from 12ps as the system approaches the failure point. At a first glance, it is very similar to workhardening. However, it is caused by the cutoff function f_{ij} in Eq.(2) and Eq.(3). The function artificially increases the force constant while it is turned on from R_{ij}^{(1)}. Physically, the system should approach the failure point from plastic deformation regime without going through "workhardening" process. This is another shortcoming in the Brenner potential. The cutoff function restricts the covalent interaction within 2 Å. Physically, it is not always the case. When a covalent bond is stretched, the two electrons in the bonding orbital are still spinpaired even two nuclei are separated by more than 2 Å. In other words, the covalent bonding still exists although it is weakened. In order to overcome this problem, a new scheme of calculating bonding information rather than a simple distance criterion is necessary.
Figure 6. The force and energy as a function of dynamic stretching time for (10,10) SWNT. 
The same dynamic stretching is also applied to a (15,15)/(10,10) DWNT. Since NB interactions are not important in this process, one can expect that the energy and force are simple addition of a (15,15) SWNT and a (10,10) SWNT. Fig.7 plots the energy and force as a function of time for the DWNT. Since the stretching is constant strain rate, the strain is proportional to its time scale. Because the stretching velocity is high, nanotubes do not have time to rearrange themselves to form defect structures. From Fig.6 and Fig.7, the disruption occurs around 18ps. It is much shorter than the time scale to form defected structure that is on the order of nanosecond. Therefore, this type of dynamic stretching can be considered as an adiabatic process.
Figure 7. The force and energy as a function of dynamic stretching time for (15,15)/(10,10) DWNT. 
From the figures, we can also see that the disruption occurs at very short time scale. Once the tube starts to break, the complete breakage happens extremely fast. Fig.8 and Fig.9 show the time evolution of the dynamic stretching processes.
In this paper, we applied an extended bond order potential to nanotube and fullerenes simulation. The extended potential considers NB interactions properly, and offers a generic scheme to implement bond order potential with long range interactions. Moreover, the modification of bond terms and NB terms can be separable. In other words, one can further improve bond potential or NB potential without worrying too much about the correlation effects. For example, we have discussed the problem of 2 Å cutoff in bond terms in the above section. To overcome this drawback, one might use other schemes which have much longer bonding range, the NB terms can remain untouched during the modification. Therefore, this is a robust scheme of include NB interactions. For the switchingon method, it needs to be reformulated or reparameterized significantly if one alters bond terms. Using the new extended potential, we also discussed the conditions under which NB interactions are important and the situations where NB terms are not crucial. Through examining the bond order potential in more detail, we pointed out a few shortcomings of the existing potential and the directions that can be taken. We believe that bond order potentials are very powerful tools for classical simulations of material properties. A semiclassical bond order potential that has more quantum mechanical concepts will be the next generation of reactive potential.
This research was funded by a grant from NASA Computational Nanotechnology and by a grant from LLNL. The facilities of the MSC are also supported by grants from NSF (ASC 9217368 and CHE 9112279), ARO (MURI), ARO (DURIP), ONR (DURIP), Chevron Petroleum Technology Co., Asahi Chemical, OwensCorning, Exxon, Chevron Chemical Co., Asahi Glass, Chevron Research Technology Co., Avery Dennison, BP America, and Beckman Institute.
Home About Foresight Blog News & Events Roadmap About Nanotechnology Resources Facebook Contact Privacy Policy Foresight materials on the Web are ©1986–2016 Foresight Institute. All rights reserved. Legal Notices. 
Web site developed by Stephan Spencer and Netconcepts; maintained by James B. Lewis Enterprises. 