Elastic bending modulus of monolayer graphene

Qiang Lua, Marino Arroyob, and Rui Huanga

a
Department of Aerospace Engineering and Engineering mechanics, University of Texas, Austin,

TX 78712, USA
b
Department of Applied Mathematics 3, LaCàN, Universitat Politècnica de Catalunya (UPC),

Barcelona 08034, Spain

Abstract

A new formula for elastic bending modulus of monolayer graphene is derived analytically from

an empirical potential for solid-state carbon-carbon atomic bonds. Two physical origins are

identified for the non-vanishing bending modulus of the atomically thin graphene sheet, one due

to the bond angle effect and the other resulting from the bond order term associated with the

dihedral angles. The analytical prediction compares closely with ab initio energy calculations.

Pure bending of graphene monolayers into cylindrical tubes are simulated by a molecular

mechanics approach, showing slight nonlinearity and anisotropy in the tangent bending modulus

as the bending curvature increases. An intrinsic coupling between bending and in-plane strain is

noted for graphene monolayers rolled into carbon nanotubes.

PACS numbers: 61.48.De, 62.20.de, 81.05.Uw, 82.45.Mp

1
The unique two-dimensional (2D) lattice structure and physical properties of graphene has drawn

tremendous interests recently. In particular, rippling of suspended graphene monolayers has been

observed, with mesoscopic amplitude and wavelength [1]. Imaging of monolayer graphene

sheets on silicon dioxide has also shown structural corrugation [2, 3]. Theoretical studies [4-6]

have suggested that bending stiffness of the monolayer graphene is critical in attaining the

structural stability and morphology for both suspended and supported graphene sheets, which in

turn could have important impacts on their electronic properties. Furthermore, single-walled

carbon nanotubes are essentially graphene monolayers subjected to cylindrical bending along

particular chiral directions. It has been reported that the bending curvature (inverse of the tube

radius) and bending orientation (tube chirality) have dramatic effects on both the mechanical and

electronic properties of carbon nanotubes [7, 8]. An accurate account of the bending modulus of

graphene is thus important for understanding the mechanics of carbon nanotubes.

While the in-plane mechanical properties (elastic modulus and strength) of monolayer

graphene have been deduced from experiments [9, 10], direct measurement of bending stiffness

of monolayer graphene has not been reported in the literature. The often cited experimental value

of 1.2 eV (~0.192 nN-nm) was derived from the phonon spectrum of graphite [11]. Theoretically,

bending modulus of monolayer graphene has been predicted based on empirical potentials [7, 12-

14] and ab initio calculations [15, 16]. The fact that the atomically thin graphene monolayer has

a finite bending modulus is in contrast with classical theories for plates and shells [17]. For

example, the bending modulus of an elastic thin plate scales with the cube of its thickness,

namely,

dM
D= ~ Eh 3 , (1)
dκ

2
where M is the bending moment, κ the bending curvature, h the plate thickness, and E the

Young’s modulus. The linear relationship between the bending modulus and the Young’s

modulus in Eq. (1) is a result of the classical Kirchhoff hypothesis [17], which assumes linear

variation of the strain and stress along the thickness of a thin elastic plate. For a graphene

monolayer, however, its physical thickness cannot be defined unambiguously in the continuum

sense [14], and the Kirchhoff hypothesis simply does not apply. Therefore, different physical

origins must be sought for the bending moment and bending modulus in graphene monolayers.

Based on the first-generation Brenner potential [18], a simple analytical form was derived

for the bending modulus of monolayer graphene under infinitesimal bending curvature [13, 14],

which established a clear connection between the functional form of the empirical potential and

the effective bending modulus. The potential takes the form

Vij = VR (rij ) − bijV A (rij ) , (2)

where i and j are indices labeling the atoms, rij is the bond length, and bij = (bij + b ji ) / 2 is the

bond-order function describing the bonding environment. Near the ground state of graphene, a

cutoff function in the potential restricts the repulsive and attractive interaction functions, VR(r)

and VA(r), to the nearest neighbors, while the bond-order function bij accounts for many-body

interactions up to the second nearest neighbors through its dependence on the bond angles, θijk

and θ jik ( k ≠ i, j ), as illustrated in Fig. 1. As shown in [13, 14], at the ground state of a planar

graphene monolayer, with rij = r0 and all the bond angles identically 2π / 3 , the bending

modulus derived from the empirical potential is

3 ⎛⎜ ∂Vij ⎞
⎟ = 1 VA (r0 )b' (2π / 3, 2π / 3) ,
D= (3)
2 ⎜⎝ ∂ cosθ ijk ⎟
⎠g 2

3
where the subscript g denotes derivatives at the ground state and b' denotes the derivative of the

function bij with respect to either one of the two bond angles. Equation (3) reveals that the

physical origin of the bending modulus comes from multi-body interactions of the carbon atoms

through the bond angle effect in the interatomic potential. In other words, any empirical potential

with only the nearest neighbor (two-body) interactions would lead to zero bending modulus of

the monolayer.

Using the second set of the parameters for the Brenner potential [18], the bending

modulus predicted by Eq. (3) is: D = 0.133 nN-nm, or equivalently, 0.83 eV. This prediction

however is considerably lower than that from ab initio energy calculations with D = 3.9 eV-

Å2/atom [16], or equivalently, 0.238 nN-nm (1.5 eV). Applying the same equation for the

second-generation Brenner potential [19] leads to an even lower bending modulus: D = 0.110

nN-nm or 0.69 eV. The discrepancy between the analytical prediction of Eq. (3) and the ab initio

calculations suggests that the bond angle effect in the empirical potentials does not fully account

for the bending stiffness of the monolayer graphene. In this paper we resolve this discrepancy by

showing that, in addition to the bond angle effect, the dihedral angle effect must be included in

the consideration of bending energetics of graphene. With many-body interactions up to the third

nearest neighbors, the dihedral angle effect adds a significant contribution to the bending

stiffness of monolayer graphene and thus the bending energy in carbon nanotubes.

While several empirical potentials (including Tersoff [12] and first-generation Brenner

potentials [18]) consider multibody interactions up to the second nearest neighbors, the second-

generation Brenner potential for carbon includes the third nearest neighbors via a bond order

term associated with the dihedral angles [19]. Taking the same general form as Eq. (2), the bond

order function for the second-generation Brenner potential is

4
 bij = 12 (bijσ −π + bσji −π ) + bijDH + Π ijRC , (4)

where 1
(bσ −π
2 ij
+ bσji −π ) is a function of the bond angles similar to that in the first-generation

Brenner potential, bijDH is a function of the dihedral angles, and Π ijRC represents the influence of

radical energetics and π -bond conjugation on the bond energy. For a perfect graphene lattice

with no vacancy, Π ijRC = 0 , and the dihedral function takes the form

bijDH =
T0
[( )
∑ 1 − cos2 Θijkl fc (rik ) fc (rjl ) ,
2 k ,l ( ≠ i , j )
] (5)

where T0 = −0.00809675 and f c (r ) is the cut-off function that restricts the dihedral function to

the four nearest neighbors (k and l) of the atoms i and j, as illustrated in Fig. 1. The dihedral

angle can be determined by

cos Θijkl = n jik ⋅ nijl , (6)

where n jik and n ijl are the unit normal vectors to the planes of the triangles jik and ijl,

respectively, namely

r ji × rik rij × r jl
n jik = and n ijl = . (7)
rij rik sin θijk rij r jl sin θ jil

Therefore, each C-C bond in the graphene lattice is associated with four dihedral angles, and

each dihedral angle accounts for an interaction between one atom (e.g., atom k) and one of its

third nearest neighbors (e.g., atom l). For a planar graphene monolayer, the dihedral angles are

either 0 or π , and thus bijDH = 0 . However, the dihedral term becomes nonzero upon bending of

the graphene monolayer.

Following a similar approach as in Ref. 13, the bending modulus of monolayer graphene

is derived as follows:

5
 ∂ 2W
D= , (8)
∂κ 2

where W is the strain energy density function depending on the bond lengths, bond angles, and

dihedral angles in a unit cell of graphene. Each unit cell contains two carbon atoms and three

inequivalent bonds (relabeled as q = 1-3), and the strain energy density can be obtained by

summing up the interatomic potential energy, namely

1 3
[
W = ∑ VR (rq ) − b(θ q1 ,...,θ q 4 , Θ q1 ,..., Θ q 4 )VA (rq ) − V0 .
S0 q =1
] (9)

where S0 = 3
2
3r02 is the area of the unit cell at the ground state, V0 is the ground-state bond

energy, and the bond-order function has been written explicitly in terms of the four bond angles

and four dihedral angles for each bond (Fig. 1). In the following derivation, it must be checked

that the internal lattice relaxation within the unit cell does not affect the bending modulus, as it

was done in [13] in the absence of the dihedral terms. This has been confirmed to hold in the

present setting as well.

At the ground state of graphene, it can be shown that,

⎛ ∂rq ⎞ ⎛ ∂θ qk ⎞ ⎛ ∂ cos Θ qk ⎞
⎜⎜ ⎟⎟ = 0 , ⎜⎜ ⎟⎟ = 0 , and ⎜⎜ ⎟⎟ = 0 . (10)
⎝ ∂κ ⎠ g ⎝ ∂κ ⎠g ⎝ ∂κ ⎠g

Therefore, the bending modulus of graphene under an infinitesimal bending curvature from the

ground state is

∂W ∂ rq 3 4 ∂W ∂ θ qk ∂W ∂ cos Θ qk
3 2 2 3 4 2
D=∑ + ∑∑ ∂θ ∂κ 2 ∑∑ ∂ cos Θ
+ . (11)
q =1 ∂rq ∂κ ∂κ 2
2
q =1 k =1 qk q =1 k =1 qk

Furthermore, taking derivatives of the strain energy at the ground state, we have

⎛ ∂W ⎞ ⎛ ∂W ⎞ ⎛ ∂W ⎞
⎜ ⎟ = 0, ⎜ ⎟ = − VA (r0 ) (b0σ −π )' , ⎜ ⎟ = ± T0 VA (r0 ) . (12)
⎜ ∂rq ⎟ ⎜ ∂θ qk ⎟ 2S0 ⎜ ∂ cos Θ qk ⎟ S0
⎝ ⎠g ⎝ ⎠g ⎝ ⎠g

6
where (b0σ −π )' denotes the derivative of the function bijσ −π with respect to either one of the two

bond angles at the ground state. Thus, Eq. (11) becomes

VA (r0 ) σ −π 3 4 ⎛⎜ ∂ θ qk ⎞ T V (r ) 3 4 ⎛ ∂ 2 cos Θ qk ⎞
2
D=− (b0 )' ∑∑ ⎟ + 0 A 0 ∑∑ ⎜ cos Θ qk ⎟ . (13)
2S0 ⎜
q =1 k =1 ⎝ ∂κ
2 ⎟ S0 q =1 k =1 ⎜⎝ ∂κ 2 ⎟
⎠g ⎠g

⎛ ∂ 2θ qk ⎞
⎟ = − 9 r02 . Using a method of asymptotic expansion, we
3 4
As shown in Ref. 13, ∑∑ ⎜
⎜
q =1 k =1 ⎝ ∂κ
2 ⎟
⎠g 2 3

find that

⎛ ∂ 2 cos Θ qk ⎞ 2
⎟ = − 21r0 .
3 4
∑∑ ⎜ cos Θqk ∂κ 2
⎜
⎟ 2
(14)
q =1 k =1 ⎝ ⎠g

Finally, we obtain a new analytical form for the bending modulus of monolayer graphene:

VA (r0 ) ⎛ σ −π 14T0 ⎞
D= ⎜ (b0 )'− ⎟. (15)
2 ⎝ 3 ⎠

While the first term on the right-hand side of Eq. (15) is identical to Eq. (3), the second

term results from the effect of dihedral angles in the second-generation Brenner potential. With

the additional term, Eq. (15) predicts that D = 0.225 nN-nm (1.4 eV), very close to the prediction

from the ab initio calculations [16]. It is thus suggested that the effect of dihedral angles plays an

important role in bending of graphene monolayers. This formula clearly shows how the

multibody interactions, including both the second and third nearest neighbors, contribute to the

finite bending stiffness of the atomically thin membrane governed by a bond order potential.

While the specific potential is used in the above derivation, the procedure and result can be

readily generalized for other empirical potentials.

To validate the analytical expression for the bending modulus of monolayer graphene, we

carry out atomistic simulations in which graphene monolayers are rolled into cylindrical tubes of

7
various diameters. The static molecular mechanics (MM) approach is adopted to calculate the

strain energy, from which the bending moment and bending modulus are deduced. In previous

studies [12, 13, 16], strain energies of fully relaxed carbon nanotubes were calculated as a

function of the curvature ( κ = 1 / R ). It has been noted that, relative to the ground state of planar

graphene, the deformation of a fully relaxed carbon nanotube involves both the bending

curvature and in-plane strain [20, 21]. Consequently, the strain energy of a fully relaxed carbon

nanotubes should include contributions from in-plane strain and thus cannot be written simply as

a quadratic function of the curvature. To achieve pure bending (with zero in-plane strain) of the

graphene monolayers, in our simulations the total potential energy is minimized under the

constraint that the tube radius and length do not change. The constraint on the tube length is

easily applied by the periodic boundary condition along the axial direction of the tube. To

enforce the constraint on the tube radius, the graphene is first rolled up by mapping a 2D sheet of

width a onto a cylindrical tube of radius R = a / 2π . Next, the potential energy is minimized by

internal relaxation between the two sublattices of graphene, with one sublattice held in place and

the other allowed to relax. In this way, the overall tube radius does not change during the energy

minimization step. The resulting tubes from these simulations are not fully relaxed; in other

words, external reaction forces are required to keep the tube dimensions from relaxing, which

include forces in both the axial and radial directions. With the pure bending deformation, the

strain energy density of the constrained tube depends only on the tube radius or curvature. In the

linear elastic regime, we have W = 12 Dκ 2 , and the bending moment is simply,

M = dW / dκ = Dκ , with the bending modulus D as given in Eq. (15).

Figure 2 plots the strain energy per atom as a function of the curvature for graphene

monolayers rolled along the armchair and zigzag directions. For comparison, the results from

8
both the first and second-generation Brenner potentials (B1 and B2) are shown. To further

highlight the effect of dihedral angles, also shown are the results from simulations ignoring the

dihedral term in the second-generation Brenner potential (B2*). Clearly, the strain energy for

potential B2 is systematically higher than the other two. The dihedral term contributes

significantly to the bending energy, especially for large bending curvatures (small nanotubes).

The corresponding bending moments are obtained by numerically differentiating the strain

energy with respect to the bending curvature, as plotted in Fig. 3. In all three cases, the bending

moment increases almost linearly with the curvature up to 2 nm-1, with slight nonlinearity at

large curvatures. By further differentiating the bending moment with respect to the curvature, we

obtain the tangent bending modulus, as plotted in Fig. 4. The tangent modulus at small

curvatures agrees closely with the analytical prediction by Eq. (15) as indicted by the dashed

lines for the three potentials. At large curvatures (carbon nanotubes of small radii), the tangent

bending modulus deviates slightly as a result of nonlinearity. Due to the effect of dihedral angles,

the tangent modulus obtained from the potential B2 is considerably higher than those from B1

and B2*. Interestingly, the tangent bending modulus from B2 decreases slightly as the curvature

increases, while the trend is reversed when the dihedral term is ignored in potentials B2* and B1.

Figure 4 shows that the tangent bending moduli along the zigzag and armchair directions

are essentially identical at small curvatures, but become increasingly different as the curvature

increases. As expected, the monolayer graphene at the ground state is elastically isotropic due to

the hexagonal symmetry of the graphene lattice, and the bending modulus at the linear elastic

regime as predicted by Eq. (15) is independent of the bending direction. However, the lattice

symmetry is distorted by the cylindrical bending deformation and the monolayer graphene

9
becomes slightly anisotropic at the nonlinear regime. Even stronger anisotropy was predicted for

the in-plane elastic moduli of monolayer graphene under finite stretches [20].

It is noted that, due to the constraint on the tube radius and length, the strain energy of

pure bending (Fig. 2) is slightly higher than the corresponding strain energy in fully relaxed

carbon nanotubes [13]. To illustrate the effect of the constraint, Fig. 5 plots the strain energy as a

function of the tube radius for a (10, 0) carbon nanotube. The tube radius R is gradually

increased in the MM simulations, while the tube length remains fixed. Only the second-

generation Brenner potential with the dihedral term is used here. As shown in Fig. 5, the strain

energy decreases until it reaches a minimum at R/R0 ~ 1.013, where R0 = 0.397 nm is the tube

radius before relaxation. The minimum energy is a few percent lower than the pure bending

energy and it compares closely with the corresponding strain energy by MM calculations without

imposing any constraint on the tube radius. Therefore, relaxation of the radial constraint alone

leads to increase of the tube radius by about 1.3% or an in-plane strain ε = 0.013 in the

circumferential direction of the tube. Further relaxation in the axial direction would result in

even lower strain energy along with slightly different tube radius and length for a fully relaxed

nanotube [21].

To understand the energy reduction and radius increase in the relaxed tube, one may

consider the total strain energy as the sum of the bending energy and the in-plane membrane

strain energy. As the tube radius increases, the bending energy, WB = Dκ 2 / 2 = D /(2 R 2 ) ,

decreases and the membrane strain energy, WM = Cε 2 / 2 , increases, where D and C are the

elastic modulus for bending and in-plane stretch, respectively. However, with D = 0.225 nN-nm

and C = 289 N/m for the second-generation Brenner potential [13, 20], the competition between

the two energy terms leads to a minimum strain energy at ε ~ 0.005 for the (10, 0) nanotube,

10
much smaller than that in Fig. 5. More accurately, we expand the strain energy with respect to

the bending curvature and membrane strain to the leading orders, namely

W (κ , ε ) ≈ Dκ 02 + M (κ − κ 0 ) + σε + D(κ − κ 0 )2 + Cε 2 ,
1 1 1
(16)
2 2 2

where κ 0 = 1 / R0 is the curvature before relaxation, M = (∂W / ∂κ )ε =0 is the bending moment,

and σ = (∂W ∂ε )ε = 0 is the in-plane membrane force in the circumferential direction for the un-

( )( )
relaxed tube. The strain energy then has a minimum at ε = Dκ 02 − σ / C + 3Dκ 02 . For the (10,0)

nanotube, we find that, with σ = −2.38 N/m, Eq. (16) agrees closely with the MM calculations in

Fig. 5 for the in-plane strain up to a few percent. Therefore, the tube is subject to a compressive

membrane force along the circumferential direction before relaxation. The compressive

membrane force may be qualitatively understood as a result of shortening of the bond lengths in

the constrained tube, relative to the bond length at the ground state of graphene. The imposed

constraint over the tube radius effectively applies an external pressure onto the tube, balancing

the internal membrane force. The applied external pressure may be estimated from the Laplace-

Young equation, namely, p = σ / R0 = 6.0 GPa, which compares closely with ab initio

calculations of carbon nanotubes under hydrostatic pressure [22]. We note that the presence of

the membrane force before relaxation is in clear contrast with the classical plate theory which

predicts zero membrane force under the pure bending condition. This suggests an intrinsic

coupling between bending and in-plane strain due to the discrete nature of the graphene lattice.

In summary, we have shown that the non-vanishing bending modulus of the atomically

thin graphene monolayers results from multibody atomistic interactions up to the third nearest

neighbors, as accounted for in a bond-order empirical potential by the effects of bond angles

(second nearest neighbors) and the dihedral angles (third nearest neighbors). A new analytical

11
expression is derived for the elastic bending modulus at the ground state, which compares

closely with ab initio calculations. Slight nonlinearity and anisotropy are noted for tangent

bending modulus obtained from molecular mechanics based atomistic simulations. An intrinsic

coupling between bending and in-plane strain is suggested, which leads to a compressive

membrane force under pure bending and reduction of the strain energy in relaxed carbon

nanotubes.

Acknowledgments

The authors (RH and QL) gratefully acknowledge funding of this work by the US Department of

Energy through Grant No. DE-FG02-05ER46230. MA acknowledges the support received

through the grant DPI2007-61054 of the Ministerio de Ciencia e Innovacion and through the

prize “ICREA Academia” for excellence in research, funded by the Generalitat de Catalunya.

12
References

1. J.C. Meyer, A.K. Geim, M.I. Katsnelson, K.S. Novoselov, T.J. Booth, and S. Roth, The

structure of suspended graphene sheets. Nature 446, 60-63 (2007).

2. M. Ishigami, J.H. Chen, W.G. Cullen, M.S. Fuhrer, E.D. Williams, Atomic structure of

graphene on SiO2. Nano Letters 7, 1643-1648 (2007).

3. E. Stolyarova, K.T. Rim, S. Ryu, J. Maultzsch, P. Kim, L.E. Brus, T.F. Heinz, M.S.

Hybertsen, G.W. Flynn, High-resolution scanning tunneling microscopy imaging of

mesoscopic graphene sheets on an insulating surface. Proc. Natl. Acad. Sci. U.S.A. 104,

9209-9212 (2007).

4. A. Fasolino, J. H. Los, and M. I. Katsnelson, Intrinsic ripples in graphene. Nature Materials

6, 858-861 (2007).

5. N. Abedpour, M. Neek-Amal, R. Asgari, et al., Roughness of undoped graphene and its

short-range induced gauge field. Phys. Rev. B 76, 195407 (2007).

6. E.-A. Kim and A.H.C. Neto, Graphene as an electronic membrane. Eur. Phys. Lett. 84,

57007 (2008).

7. Z.-C. Tu, Z.-C. Ou-Yang, Single-walled and multiwalled carbon nanotubes viewed as elastic

tubes with the effective Young's modulus dependent on layer number. Phys. Rev. B 65,

233407 (2002).

8. R. Saito, R.G. Dresselhaus and M.S. Dresselhaus, Physical Properties of Carbon Nanotubes.

Imperial College Press, London 1998.

9. C. Lee, X.D. Wei, J.W. Kysar, and J. Hone, Measurement of the elastic properties and

intrinsic strength of monolayer graphene. Science 321, 385-388 (2008).

13
10. J.S. Bunch, A.M. van der Zande, S.S. Verbridge, I.W. Frank, D.M. Tanenbaum, J.M. Parpia,

H.G. Craighead, and P.L. McEuen, Electromechanical resonators from graphene sheets.

Science 315, 490-493 (2007).

11. R. Nicklow, N. Wakabayashi, and H.G. Smith, Lattice dynamics of pyrolytic graphite. Phys.

Rev. B 5, 4951-4962 (1972).

12. J. Tersoff, New empirical approach for the structure and energy of covalent systems. Phys.

Rev. B 37, 6991-7000 (1988).

13. M. Arroyo and T. Belytschko, Finite crystal elasticity of carbon nanotubes based on the

exponential Cauchy-Born rule. Phys. Rev. B 69, 115415 (2004).

14. Y. Huang, J. Wu, and K.C. Hwang, Thickness of graphene and single-wall carbon nanotubes.

Phys. Rev. B 74, 033524 (2006).

15. D. Sanchez-Portal, E. Artacho, J.M. Soler, A. Rubio, and P. Ordejon, Ab initio structural,

elastic, and vibrational properties of carbon nanotubes. Phys. Rev. B 59, 12678-12688

(1999).

16. K.N. Kudin, G.E. Scuseria, and B.I. Yakobson, C2F, BN, and C nanoshell elasticity from ab

initio computations. Phys. Rev. B 64, 235406 (2001).

17. S. Timoshenko and S. Woinowsky-Krieger, Theory of Plates and Shells. McGraw-Hill, New

York 1987.

18. D.W. Brenner, Empirical potential for hydrocarbons for use in simulating the chemical vapor

deposition of diamond films. Phys. Rev. B 42, 9485-9471 (1990).

19. D.W. Brenner, O.A. Shenderova, J.A. Harrison, S.J. Stuart, B. Ni, and S.B. Sinnott, A

second-generation reactive empirical bond order (REBO) potential energy expression for

hydrocarbons. Journal of Physics-Condensed Matter 14, 783-802 (2002).

14
20. Q. Lu and R. Huang, Nonlinear mechanics of single-atomic-layer graphene sheets. Journal of

Computational and Theoretical Nanoscience, in press.

21. J. Wu, K.C. Hwang, Y. Huang, An atomistic-based finite deformation shell theory for single-

wall carbon nanotubes. J. Mech. Phys. Solids 56, 279-292 (2008).

22. S. Reich, C. Thomsen, P. Ordejon, Elastic properties of carbon nanotubes under hydrostatic

pressure. Phys. Rev. B 65, 153407 (2002).

15
 k
i Θq3
Θq2
θijk Θq4
Θq1
θjil
j
l

θq4
θq3
θq1 θq2

FIG. 1. (color online) Illustration of bond angles and dihedral angles in graphene lattice. Each

dihedral angle is represented by three bonds connecting four carbon atoms (e.g., i, j, k, l), and

each carbon-carbon bond is associated with four bond angles and four dihedral angles.

16
 Strain energy per atom (eV/atom) 0.15
B1, armchair
B1, zigzag
B2, armchair
0.1 B2, zigzag
B2*, armchair
B2*, zigzag

0.05

0
0 0.5 1 1.5 2 2.5 3

Bending curvature (1/nm)

FIG. 2. (color online) Strain energy per atom in graphene tubes as a function of bending
curvature obtained from different empirical potentials: B1 for the first-generation Brenner
potential, B2 for the second-generation Brenner potential, and B2* for the second-generation
Brenner potential without considering the dihedral term. Results from molecular mechanics
simulations are shown for pure bending of monolayer graphene along armchair and zigzag
directions, while the quadratic function, W = Dκ 2 / 2 , is plotted as the dashed line using the
analytical bending modulus for each potential.

17
 0.6
B1, armchair

Bending moment per length (nN)

0.5 B1, zigzag
B2, armchair
0.4 B2, zigzag
B2*, armchair
0.3 B2*, zigzag

0.2

0.1

0
0 0.5 1 1.5 2 2.5
Bending curvature (1/nm)

FIG. 3. (color online) Bending moment in graphene tubes versus bending curvature along
armchair and zigzag directions, obtained from different empirical potentials: B1 for the first-
generation Brenner potential, B2 for the second-generation Brenner potential, and B2* for the
second-generation Brenner potential without considering the dihedral term. The linear elastic
bending moment-curvature relation, M = Dκ , is plotted as the dashed line using the analytical
bending modulus for each potential.

18
 0.25
Bending modulus (nN−nm)
0.2

0.15

0.1 B1, armchair

B1, zigzag
B2, armchair
0.05 B2, zigzag
B2*, armchair
B2*, zigzag
0
0 0.5 1 1.5 2 2.5
Bending curvature (1/nm)

FIG. 4. (color online) Tangent bending modulus of monolayer graphene as a function of bending
curvature along armchair and zigzag directions, obtained from different empirical potentials: B1
for the first-generation Brenner potential, B2 for the second-generation Brenner potential, and
B2* for the second-generation Brenner potential without considering the dihedral term. The
analytical prediction for the linear elastic bending modulus (independent of curvature) is plotted
as the dashed line for each potential.

19
 0.118
Strain energy per atom (eV/atom)
0.117

0.116

0.115

0.114

0.113
0 0.005 0.01 0.015 0.02
Strain ε = R/R0 − 1

FIG. 5. (color online) Relaxation of the strain energy for a (10, 0) carbon nanotube as its radius
increases. The atomistic calculations are shown by open circles and the prediction by Eq. (16) is
plotted as the solid curve. The dashed line indicates the strain energy from an atomistic
simulation without imposing any constraint on the tube radius.

20