RAPID COMMUNICATIONS

PHYSICAL REVIEW B 73, 081401共R兲 共2006兲

First-principles study of work functions of double-wall carbon nanotubes

Bin Shan
Department of Applied Physics, Stanford University, Stanford, California 94305-4040, USA

Kyeongjae Cho*
Department of Mechanical Engineering, Stanford University, Stanford, California 94305-4040, USA
共Received 27 November 2005; revised manuscript received 11 January 2006; published 6 February 2006兲

Using first-principles density functional calculations, we investigated work functions 共WFs兲 of thin double-
walled nanotubes 共DWNTs兲 with outertube diameters ranging from 1 to 1.5 nm. The results indicate that a
work function change within this diameter range can be up to 0.5 eV, even for DWNTs with the same outer
diameter. This is in contrast with single-walled nanotubes, which show negligible WF changes for diameters
larger than 1 nm. We explain the WF change and the related charge redistribution in DWNTs using the charge
equilibration model. The predicted work function variation of DWNTs indicates a potential difficulty in their
nanoelectronic device applications.

DOI: 10.1103/PhysRevB.73.081401 PACS number共s兲: 31.15.Ar, 72.80.Rj, 73.30.⫹y

Double-walled nanotubes 共DWNTs兲 have attracted con- the inner shell of DWNTs.13 Nanotubes with diameters
siderable attention due to their electronic and mechanical ⬃4 Å have already been observed in confined channels or as
properties.1 Compared to single-walled nanotubes 共SWNTs兲, the inner shell of multiwall nanotubes.14–16 The inner tubes
they have higher mechanical stiffness and greater thermal in thin DWNTs usually feature high surface curvature and
stability, which may be of benefit in field emission devices.2 are shown to have large variations in work functions.17 There
Their double-shell structure offers the possibility of shielding have been some previous theoretical investigations on charge
the inner tube from external perturbations.3 Such advantages redistribution in selected DWNTs.13,18 However, as far as we
make them very interesting for certain device applications. know, there has been no systematic study of the work func-
The utilization of DWNTs in earlier days was largely hin- tions of DWNTs and it remains unclear how the encapsula-
dered by the unwanted carbon nanomaterials generated at the tion of ultrasmall inner tubes affects the work functions of
same time with DWNTs during fabrication processes. Over DWNTs.
the past few years, substantial improvements have been In this study, we report first-principles work function cal-
achieved in fabrication and purification techniques of culations for two series of DWNTs consisting of either
DWNTs. Synthesis of high-quality DWNTs in a large scale 共m , n兲@共n , n兲 pairs or 共m , 0兲@共n , 0兲 pairs 共m ⬍ n兲, with the
has been realized using the high-temperature pulsed arc dis- first chiral index representing the inner tube and the latter
charge technique4 and the catalytic chemical vapor representing the outer tube. The diameters of the DWNTs we
deposition.5 In a more recent experiment, the production of studied range from 1 to 1.5 nm, within which work function
DWNTs with the experiment yields more than 95% of what variations are most significant. It was found that the work
has been reported.6 Thin DWNTs have also been success- function variations can be up to 0.5 eV, even for DWNTs
fully fabricated by filling C60 molecules in SWNTs followed with similar outer diameters. This can be a significant value
by electron beam irradiation, yielding DWNTs with diameter in device physics. The origin of work function modulations
distribution around 13–14Å.7 In addition, it has now become in DWNTs, as well as the charge redistribution, are shown to
possible to identify atomic correlations between adjacent be correlated with the type of the inner shell, and can be
graphene layers using high-resolution transmission electron qualitatively understood within the framework of the charge
microscopy,8 enabling an unambiguous classification of equilibration model 共CEM兲. Some issues in the characteriza-
DWNTs. All these point toward a possible application of tion and measurements of thin DWNTs are also discussed.
DWNTs in nanoelectronic devices. The work function was calculated using the standard pro-
Recently, the potential use of DWNTs as channel materi- cedure by taking the difference between the vacuum level ␾
als for field-effect transistors 共FETs兲 has been successfully and the Fermi level E f . The vacuum level ␾ is determined
demonstrated,9 with better subthreshold swing factors as from the average potential at the center of the vacuum region
compared to SWNTs. Also, due to different inner and/or where it approaches a constant. We have excluded the slowly
outer tube combinations and interlayer interactions, DWNTs- decaying exchange-correlation part of the potential in the
based FET yields richer transistor characteristics.10 For a fur- vacuum region to achieve better convergence.19 E f was
ther large-scale integration of DWNT-based FETs, an under- placed at the midgap in the case of a semiconducting DWNT.
standing of the DWNT work function is essential, since even The calculation was done using the Vienna ab-initio simula-
work functions that change on the order of 0.1 eV may lead tion package20 within the local density approximation
to substantial changes in device characteristics.11 共LDA兲. 30 k points were used along the nanotube’s one-
The encapsulation of molecules such as C60 inside dimensional Brillouin zone, which was tested to give good
SWNTs are known to modify the electronic properties of convergence. In all simulations, an orthorhombic unit cell
carbon nanotubes.12 Due to their concentric double layer with cross-sectional dimensions of 共35⫻ 35兲 Å2 was used.
structures, ultrasmall diameter tubes can be encapsulated as With this unit cell size, the minimum separation between

1098-0121/2006/73共8兲/081401共4兲/$23.00 081401-1 ©2006 The American Physical Society

 RAPID COMMUNICATIONS

B. SHAN AND K. CHO PHYSICAL REVIEW B 73, 081401共R兲 共2006兲

This value of interwall spacing is consistent with other first-

principles calculations22 and experimental values from scan-
ning tunneling microscope measurements.23 Even though in
real fabrication processes, DWNTs with nonoptimal inter-
wall spacing are also expected to exist, those with optimal
interwall spacing are more energetically favorable and would
prevail in the final product. Moreover, similar correlation to
the inner tube chirality was seen in DWNTs with nonoptimal
interwall spacing. Thus, in the following, we will focus our
discussion on work functions of DWNTs with optimal inter-
wall spacing.
Figure 1共b兲 summarizes work functions of various
DWNTs with optimal interwall spacing, plotted against the
diameter of the outer tube. Also shown on the graph are work
functions of SWNTs, indicated by the dotted line. As can be
seen from the graph, there is essentially no diameter depen-
dence of work functions for SWNTs larger than 1 nm 共type-I
SWNT兲. However, different work functions of DWNTs can
be observed in the diameter range from 1 to 1.5 nm, for up to
0.5 eV. More specifically, 共m , 0兲@共n , 0兲 DWNTs show a
general decrease trend in work functions while those of
共m , m兲@共n , n兲 pairs show little variation. This indicates that
the work function difference is primarily due to the presence
of the inner ultrasmall diameter tube 共type-II SWNT兲, which
is known to have substantial differences in work functions.17
In particular, when the inner shell of the DWNTs is of 共m , 0兲
zigzag type, the overall work function of the DWNT was
increased due to the higher work functions of the inner tube.
The qualitative feature of the work function variations in
DWNTs can be understood by the CEM, which has been
FIG. 1. 共a兲 Heat of formation per unit cell for the DWNTs; solid successfully applied to molecular dynamics simulations.24
square for 共3,3兲@共n,n兲 DWNTs and empty circle for 共4,0兲@共n,0兲
Generally speaking, when two material systems with differ-
DWNTS. 共b兲 Work functions for DWNTs 共dots兲 and SWNTs 共tri-
ent chemical potentials 共minus of work functions兲 come into
angles兲 of different diameters.
contact, chemical potentials are equalized by electron trans-
periodic tube images was ⬎18 Å, which ensures that the fer from a higher chemical potential system to a lower po-
corresponding work function converges to a constant value. tential system. The total electrostatic energy of such a system
Kohn-Sham single-electron wave functions were expanded can be written, neglecting higher order terms, as

冋 冉 冊 冉 冊册
by 285 768 plane waves with an energy cutoff of 286.6 eV. N
⳵E 1 ⳵ 2E 2
For the LDA exchange-correlation potential, we have used a Etot共q1,q2, ¯ ,qN兲 = 兺 E0i + qi + q
functional form fitted to the Monte Carlo results of homoge- i ⳵ qi 2 ⳵ q2i i
neous electron gas.21 A conjugate-gradient method was used
1
for both the electronic structure calculation and geometry
optimization. The DWNTs were assumed to be fully relaxed
+ 兺 Vij共qi,q j兲
2 i⫽j
共1兲
when the force on each atom was less than 0.05 eV/ Å.
We first discuss the stability and optimal interwall spacing N
1
for DWNTs. The energetics with respect to interwall spacing = 兺 共E0i + ␹0i qi + 21 J0i q2i 兲 + 兺 Vij共qi,q j兲, 共2兲
was calculated by fixing the innertube diameter and varying i 2 i⫽j
the diameter of the outer tube. Figure 1共a兲 shows the heat of
where E0i is the energy of neutral atom i , −eqi is the excess
formation for selected DWNTs 关⌬E = E共DWNT兲
electrons on atom i , ␹0i is the electronegativity 共minus of
− E共SWNT兲outer − E共SWNT兲inner兴 with different interwall spac-
work functions兲, J0i is the atomic hardness, and Vij is the
ing. It was found that the most energetically favorable outer Coulombic interaction between atoms i and j. In the context
tube for 共3,3兲 and 共4,0兲 tubes are 共8,8兲 and 共13,0兲 tubes, of DWNTs, we can view the DWNT as a giant “molecule,”
respectively. For the two series of DWNTs we studied within with inner and outer tubes being two artificial “atoms” A and
the diameter range, the most stable inner-outer combination B. Taking the derivative of Etot with respect to qi, we arrive
is consistently 共m , m兲@共m + 5 , m + 5兲 for armchair pairs and at the following equations for the chemical potential and
共m , 0兲@共m + 9 , 0兲 for zigzag pairs. The optimal interwall charge transfer:
spacings corresponding to such combinations are close to the
interlayer spacing of graphite 共3.35 Å兲, possibly due to struc- q
␹A共q兲 = ␹A0 + JA0 q + , 共3a兲
tural similarities between graphene sheets and nanotubes. 2C

081401-2
 RAPID COMMUNICATIONS

FIRST-PRINCIPLES STUDY OF WORK FUNCTIONS OF … PHYSICAL REVIEW B 73, 081401共R兲 共2006兲

TABLE I. Work functions of selected DWNTs.

Outer Inner WF WF
Tube shell shell from from
diameter WF WF CEM DFT
DWNT 共Å兲 共eV兲a 共eV兲a 共eV兲 共eV兲

共4,0兲@共13,0兲 10.32 4.70 5.95 5.32 5.13

共5,0兲@共14,0兲 11.11 4.66 5.10 4.88 5.00
共6,0兲@共15,0兲 11.91 4.68 4.99 4.84 4.80
共7,0兲@共16,0兲 12.70 4.69 5.10 4.90 4.90
共8,0兲@共17,0兲 13.49 4.69 4.80 4.74 4.77
共3,3兲@共8,8兲 11.00 4.68 4.50 4.59 4.64
共5,5兲@共10,10兲 13.75 4.68 4.63 4.66 4.64
aWork functions of SWNTs from Ref. 17

q
␹B共− q兲 = ␹B0 − JB0 q − , 共3b兲
2C

where C = 2␲␧ / ln共b / a兲 is the unit length capacitance be-

tween two coaxial cylinders, with a and b being the inner-
and outertube diameter, respectively. The dielectric constant
of the nanotube is set to 1.25 Under equilibrium, the electro-
chemical potential of the inner tube and outer tube must be
equal. By equating ␹A共q兲 and ␹B共−q兲, we have the following
first-order solution for the final chemical potential and charge
transfer for DWNTs:

FIG. 3. Site-projected DOS for 共4,0兲@共13,0兲, 共8,0兲@共17,0兲, and

共3,3兲@共8,8兲 DWNTs, respectively. Empty circles for the outer tube
and filled triangles for the inner tube.

␹B0 − ␹A0
q= , 共4兲
JA0 + JB0 + 1/C

1 1
JB0 + 2C JA0 + 2C
␹= 1 ␹A
0
+ 1 ␹B .
0
共5兲
FIG. 2. Contour plots of charge accumulation 共left column兲 and JA0 + JB0 + JA0 + JB0 +
C C
depletion region 共right column兲 for 共4,0兲@共13,0兲, 共8,0兲@共17,0兲, and
共3,3兲@共8,8兲 DWNTs, respectivly. The contour plots are shown in a It can be seen that the direction and the amount of charge
plane normal to the tube axis, with black dots indicating positions transfers are directly related to the work function difference
of carbon atoms. The contour value spacings are set to be equal. between the inner and outer tubes, which is confirmed by

081401-3
 RAPID COMMUNICATIONS

B. SHAN AND K. CHO PHYSICAL REVIEW B 73, 081401共R兲 共2006兲

subsequent charge redistribution analysis. The final work density of states 共DOS兲, i.e., the local DOS integrated within
function of the DWNT is a linear combination of work func- spheres of the Wigner-Seitz radius centered on each ion, for
tions of the inner and outer tubes. Due to the small interlayer the above three DWNTs. In 共m , 0兲@共n , 0兲 DWNTs such as
spacing and the resulting large 1 / C term, the work function 共4,0兲@共13,0兲, the outer tube is semiconducting with low WF
of the DWNT can be roughly approximated by the average and the inner tube is metallic with high WF due to ␴ − ␲
of the work function of inner and outer tubes. Table I lists hybridization.26 When they form DWNT, substantial charge
work functions predicted by the CEM as well as those from transfer from the outer tube to the inner tube leads to hole
first-principles calculations. The general trend of the work doping of the outer tube. Even though the band gap of the
function change predicted from CEM is in good agreement semiconducting outer tube is largely retained despite the in-
with the first-principles calculations, with the largest error terwall interaction, the Fermi level is below its valence band
⬃0.2 eV. edge, giving rise to a finite DOS at the Fermi energy 关Fig.
The charge redistribution in DWNTs is calculated by sub- 3共a兲兴. For those DWNTs with ultrasmall zigzag SWNTs as
tracting the charge distribution of isolated inner and outer inner shells, their work functions are thus expected to be
tubes from the self-consistent charge distribution, ⌬␳ 0.2–0.5 eV higher than larger diameter DWNTs. As the
= ␳共DWNT兲 − ␳共inner兲 − ␳共outer兲. Figures 2共a兲, 2共c兲, and 2共e兲 DWNT diameter gets larger, such as for a 共8,0兲@共17,0兲 pair,
and Figs. 2共b兲, 2共d兲, and 2共f兲 show the charge accumulation the work functions of the inner and outer tube are compa-
regions and charge depletion regions, respectively. In rable and their individual electronic structures are less per-
共4,0兲@共13,0兲 DWNT, the work function of the inner tube is turbed. The DWNT is a semiconductor with a reduced band
⬃1 eV higher than that of the outer tube. Due to this large gap as compared to band gaps of both the outer and inner
difference in work functions, the amount of charge transfers tubes 关Fig. 3共b兲兴. 共3,3兲@共8,8兲 DWNT presents an interesting
is also more significant than in other DWNTs, as indicated case. Even though it has almost the same diameter as
by the dense contour lines in Figs. 2共a兲 and 2共b兲. The direc- 共4,0兲@共13,0兲 one and is also metallic, its electronic structure
tion of charge transfer can also be clearly identified. The around the Fermi level is considerably different. Due to the
accumulated charge mainly locates on the innertube 共4,0兲 presence of interwall interactions, states at the Fermi level
region rather in in the empty intertube region. Such charge are mostly localized in the inner tube with tails extending to
transfer patterns have also been observed in 共5,0兲@共14,0兲 the outer shell 关Fig. 3共c兲兴. Its work function is about 0.5 eV
tubes. This indicates that provided a large enough work func- lower than that of 共4,0兲@共13,0兲 DWNT, due to the higher
tion difference, the direction of charge transfer can be pre- WF of the inner tube.17 These differences are likely to affect
dicted from SWNT work functions, consistent with one’s the transport and scanning tunneling microscope measure-
physical intuition. In the case of 共8,0兲@共17,0兲 and ments of DWNTs.
共3,3兲@共9,9兲 DWNTs, the work function difference of inner In conclusion, we have systematically studied the work
and outer tubes is not as large and the direction of charge functions of the two series of DWNTs. It was found that the
transfer is more subtle due to the presence of interwall inter- work function differences of DWNTs in a narrow diameter
action. The depleted charge from the ␲ electron system ac- distribution 共1.0–1.5 nm兲 can be up to 0.5 eV. The general
cumulates in the intertube region, with characteristics similar trend of the work function change and charge redistribution
to the interlayer state in graphite systems. Such interlayer are successfully explained by the CEM. These findings pro-
states in DWNTs have also been observed in other first- vide useful insight into the work function change in DWNTs
principles studies.18 The amount of charge transfer is also that may be used to develop new ways of engineering elec-
smaller as evident by the sparse contour lines in Figs. tronic structures of DWNTs.
2共c兲–2共f兲. This work was supported by NSF grant on Network for
Finally, we discuss some possible issues in the character- Computational Nanotechnology 共NCN兲. Part of the calcula-
ization of thin DWNTs. Figure 3 shows the site-projected tions were done at San Diego Supercomputing Center.

*Electronic address: kjcho@stanford.edu 14 Z. M. Li et al., Phys. Rev. Lett. 87, 127401 共2001兲.
1 C. Wei et al., Phys. Rev. B 67, 115407 共2003兲. 15 L. C. Qin et al., Nature 408, 50 共2000兲.
2 Y. W. Son et al., Nanotechnology 16, 125 共2005兲. 16 N. Wang et al., Nature 408, 51 共2000兲.
3 M. Endo et al., Nano Lett. 4, 1451 共2004兲. 17 B. Shan and K. Cho, Phys. Rev. Lett. 94, 236602 共2005兲.
4
T. Sugai et al., Nano Lett. 3, 769 共2003兲. 18
Y. Miyamoto et al., Phys. Rev. B 65, 041402共R兲 共2002兲.
5 E. Flahaut et al., Chem. Commun. 共Cambridge兲 12, 1442 共2003兲. 19 T. C. Leung et al., Phys. Rev. B 68, 195408 共2003兲.
6
M. Endo et al., Nature 433, 476 共2005兲. 20
G. Kresse and J. Furthmuller, Comput. Mater. Sci. 6, 15 共1996兲.
7 B. W. Smith et al., Nature 共London兲 396, 323 共1998兲. 21 D. M. Ceperley and B. J. Alder, Phys. Rev. Lett. 45, 566 共1980兲.
8
A. Hashimoto et al., Phys. Rev. Lett. 94, 045504 共2005兲. 22
J. C. Charlier and J. P. Michenaud, Phys. Rev. Lett. 70, 1858
9 T. Shimada et al., Appl. Phys. Lett. 84, 2412 共2004兲.
共1993兲.
10 S. Wang et al., J. Phys. Chem. B 109, 17361 共2005兲. 23 M. H. Ge and K. Sattler, Science 260, 515 共1993兲.
11 24
J. Guo and M. S. Lundstrom, IEEE Trans. Electron Devices 49, A. K. Rappe and W. A. Goddard, J. Phys. Chem. 95, 3358
1897 共2002兲. 共1991兲.
12 Y. Cho et al., Phys. Rev. Lett. 90, 106402 共2003兲. 25 F. Leonard and J. Tersoff, Appl. Phys. Lett. 81, 4835 共2002兲.
13 Y. L. Mao et al., Phys. Rev. B 71, 033404 共2005兲. 26 X. Blase et al., Phys. Rev. Lett. 72, 1878 共1994兲.

081401-4