IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 19 (2008) 115203 (6pp) doi:10.1088/0957-4484/19/11/115203

Carbon nanotube Schottky diode:

an atomic perspective
P Bai1 , E Li1 , K T Lam2 , O Kurniawan1 and W S Koh1
1
Nanoelectronics Research Group, Institute of High Performance Computing,
117528, Singapore
2
Department of Electrical and Computer Engineering, National University of Singapore,
119077, Singapore

E-mail: baiping@ihpc.a-star.edu.sg

Received 20 November 2007, in final form 16 January 2008

Published 18 February 2008
Online at stacks.iop.org/Nano/19/115203

Abstract
The electron transport properties of semiconducting carbon nanotube (SCNT) Schottky diodes
are investigated with atomic models using density functional theory and the non-equilibrium
Green’s function method. We model the SCNT Schottky diode as a SCNT embedded in the
metal electrode, which resembles the experimental set-up. Our study reveals that the
rectification behaviour of the diode is mainly due to the asymmetric electron transmission
function distribution in the conduction and valence bands and can be improved by changing
metal–SCNT contact geometries. The threshold voltage of the diode depends on the electron
Schottky barrier height which can be tuned by altering the diameter of the SCNT. Contrary to
the traditional perception, the metal–SCNT contact region exhibits better conductivity than the
other parts of the diode.
(Some figures in this article are in colour only in the electronic version)

1. Introduction control and understanding are also needed to improve and

Semiconducting carbon nanotubes (SCNTs) have emerged as
a potential building material for future nanoelectronics. Many
prototypical devices, such as transistors [1, 2], diodes [3–5],
sensors [6] and non-volatile memory devices [7, 8], have
been demonstrated using the SCNT. Recently, SCNT Schottky
diodes are successfully fabricated from asymmetric metal–
SCNT contacts [3–5]. This asymmetric metal–SCNT structure
is a potential candidate for high frequency applications owing
to their high carrier mobility [7], small junction areas, high
current transport capability [9] and low capacitances [10]
leading to high cutoff frequency. The diodes are created by
two dissimilar metals at the two ends of the SCNT. One of
the metal electrodes has a lower work function than that of
the SCNT in order to make a Schottky contact. The other
metal electrode at the other end of the SCNT has a higher work
function as compared to the SCNT to form an ohmic contact.
The rectification behaviours were successfully observed by
different groups. However, the measured behaviours of the
diodes were found to differ from one group to another [3–5].
A lack of understanding of the nature of the contact between
the SCNT and the metal electrode is a critical issue [3]. More
stabilize the performance of the diodes and integrate them into
complex circuits.
The performance of the SCNT Schottky diodes is strongly
dependent on the Schottky barrier height formed in the metal–
SCNT contacts. The barrier height depends on several material
parameters, such as the metal contact work function [11, 12],
the environment to which the device is exposed [13], the
diameter of the nanotube [14, 15], as well as the interface
structures [16].
Continuous improvement of the SCNT Schottky diode
performance requires a good understanding of the metal–
SCNT contact. Unfortunately, this understanding is still
lacking despite extensive experimental and theoretical studies
on the topic. This is because the exact atomic arrangement
of the carbon and metal (e.g. Al) atoms at the interface is
hardly known during experiments [3–5]. Theoretical studies
using semi-empirical approaches neglect the microscopic
description (i.e. the electronic structure of the contacts in the
device) [17, 18]. On the other hand, studies found in the
literature using ab initio quantum mechanics approaches were
limited to only a fraction of the device under equilibrium
conditions [15, 19–21].

0957-4484/08/115203+06$30.00 1 © 2008 IOP Publishing Ltd Printed in the UK

Nanotechnology 19 (2008) 115203
In this work, we investigate the electron transport
properties of the SCNT Schottky diodes through atomic
structures, constructed based on experimental set-ups [3, 4],
using density functional theory (within the local density
approximation) and the non-equilibrium Green’s function
(DFT-NEGF) at non-equilibrium conditions. For the first time,
the macroscopic properties of the SCNT Schottky diode, such
as the current–voltage ( I –V ) characteristics, threshold voltage
and conductivity, are associated with the atomic structure
properties. Tuning of the threshold voltage of the diode
can be achieved by altering the diameter of the SCNT. High
conductivity at the section of the SCNT near the Al electrode
is observed. Possible improvements to the rectification
characteristics for the Al–SCNT diode are also discussed.
2. Modelling approach
We first build atomic models of the SCNT Schottky diodes
with reference to some experimental set-ups [3, 4]. A Schottky
barrier in the diode is formed between the Al and the SCNT.
Traditionally, side-contact [20, 21] and end-contact [22] are
mainly used in the theoretical study of the metal–SCNT
contact. These contact models may not reflect the metal–
SCNT contact geometries as described in the experiments. For
instance, we can infer from the device fabrication process in
the experiments [3, 4] that the SCNT is somewhat wrapped by
the Al atoms. Therefore, we model the metal–SCNT contact
structure in the diode with two simplified embedded atomic
structures: (1) a partially embedded model where a SCNT(8, 0)
is partially wrapped by the Al atoms; and (2) a fully embedded
model where the SCNT(8, 0) is fully embedded into the Al bulk
metal. Note that parts of the SCNT in the partially embedded
model is exposed (i.e. not covered by the Al atoms) to mimic
the fact that the SCNT may be lying on a substrate. The
high work function metal electrode (Pd [4] or Au [3]) at the
other end of the SCNT is assumed to be ohmic and is ignored
in our calculations. A partially embedded atomic model
of the Schottky diode is illustrated in figures 1(a) and (b).
The length of the SCNT between the Al–SCNT interface
and the right electrode is chosen to be longer than 30 Å to
minimize the effects of direct charge transport between two
electrodes [23] and the evanescent electron transport modes
in the short SCNT [24]. A sufficient amount of Al atoms is
included in the left electrode to ensure that it represents the
bulk Al. The size of the hole in the Al electrode is chosen
to be as small as possible to accommodate the SCNT(8, 0).
The combined Al–SCNT structure should have a net energy
gain relative to the isolated components to avoid unphysical
configurations. The whole structure is configured as three
parts: left electrode (75 Al + 64 C atoms), conducting channel
(109 Al + 352 C atoms) and right electrode (64 C atoms). The
model temperature is set at 300 K.
Atomic coordinates of the model are optimized with
conjugate-gradient minimization of the total energy with
analytic forces from DFT [25]. The optimized partially
embedded atomic structure is shown in figures 1(a) and (b).
During relaxation, the SCNT(8, 0) is deformed in the radial
direction. The axis size of the SCNT(8, 0) inside the Al
2
P Bai et al
Left Conducting Right
Electrode Channel Electrode
32.82 Å
(a)
q
2.10Å p
2.12Å
Y
2.22Å
X
(b) (c)
Figure 1. (a) Relaxed atomic structure of the partially embedded
model for the SCNT Schottky diode with a SCNT(8, 0) partially
embedded into metal Al. (b) Side view of the model, where key
bonds between Al and C atoms are highlighted and their bond
lengths are shown. (c) The rectification behaviour of the SCNT
Schottky diode calculated from the partially embedded model.
electrode is increased from 6.27 to 6.5 Å in the X direction
and from 6.27 to 6.8 Å in the Y direction due to the attractive
force of the bonds between the Al and the C atoms as shown
in figure 1(b). The final geometry has a similar structure as the
SCNT(8, 0) embedded into metal Pd [19]. The Al and C atoms
that are nearest to each other can be found in cut planes q, p
and the mirror of p highlighted by ovals in figure 1(b). The
distance between them range from 2.10 to 2.22 Å.
The electron transport properties of the atomic diode
model are studied using the DFT-NEGF first principles
method, Atomistix ToolKit (ATK) [26]. This method has been
applied to many systems successfully [27, 28] as the atomic-
level details of the systems are considered self-consistently.
We extend the left and right electrodes of the atomic model in
figure 1(a) to infinity and a bias voltage is applied via the left
and right electrodes. DFT with local density approximation
and NEGF are used to deal with the atomic structure self-
consistently under non-equilibrium conditions [29]. Norm-
conserving pseudopotentials are adopted to describe the
electron–ion interactions. The valence wavefunctions are
expanded through a nonlinear combination of single zeta
atomic orbitals. The effects from semi-infinite leads are
corrected through self-energy [30]. Charge distribution,
density of states and transmission function are calculated from
the self-consistent Green’s function. The Landauer formula is
used to calculate the current flowing through the conducting
channel.
Nanotechnology 19 (2008) 115203
Figure 2. Transmission functions as a function of energy and bias.
Larger transmission coefficients are found in the conduction band
(right) than in the valence band (left). The two shaded triangle areas
are the transmission windows under forward and reverse bias,
respectively.
3. Rectification behaviour
The rectification behaviour of the SCNT Schottky diode can
be observed from the calculated current–voltage ( I –V ) curve
with the partially embedded model shown in figure 1(c). To
understand the origin of this rectification behaviour, we study
the electron transmission function of the model, which governs
the electric fluxes due to the electrons from the two electrodes.
Figure 2 shows the contour plot of the transmission
function as a function of energy and bias. Current flow
across the SCNT Schottky diode is formed by delocalized
electrons from the orbitals around the Fermi level ( E f ). The
delocalized electron energies within the transmission window
are shown as two shaded triangles in figure 2. Note that the
width of the transmission window (e|Vbias |) is the difference
in electrochemical potential between the two electrodes and
it is a function of the applied bias (Vbias ). At forward bias,
the energy gap (central part in figure 2) is shifted to the lower
energy and the conduction band edge moves closer to or even
crosses E f if the bias is large enough. For a sufficiently large
bias (i.e. the transmission coefficients can be found in the
transmission window), the electrons have a finite probability
to tunnel through the Schottky barrier into conduction bands
and form the forward current. Larger available states and
transmission coefficients in the conduction bands are translated
into more conducting channels and make the SCNT become
conducting. For reverse bias, the valence band edge will
shift closer to E f . However, there are fewer conducting
channels, which contribute to the reverse current as there are
fewer available states and poorer transmission coefficients as
illustrated in figure 2.
The rectification behaviour of the diode (caused by
the asymmetric transmission coefficient distribution in the
conduction and valence bands and the shift of the band edges
under applied bias) can also be explained through the band
diagrams of the CNT Schottky diode as shown in figure 3. At
forward bias, the conduction bands are shifted to lower energy
and overlap with the transmission window. When the overlap
occurs, electrons are able to drift over the small barrier into the
3
P Bai et al
Figure 3. Band diagrams of the CNT Schottky diode. (a) Forward
bias. (b) Reverse bias.
Figure 4. Transmission functions of three SCNT Schottky diode
models: side-contact (top inset), fully embedded (middle inset) and
partially embedded (figure 1(a)). The transmission functions are
shown as a function of the difference between energy E and the
Fermi level E f .
conduction bands of the CNT and form the forward current. At
reverse bias, the valence bands overlap with the transmission
window. However, the reverse current is much smaller than
the forward current due to fewer states there. Note that this
current mechanism is different from the p-type Si Schottky
diode where the forward current is caused by the majority
carriers (holes) flowing from semiconductor to metal over a
smaller potential barrier [31].
The asymmetric transmission function could be caused
by the geometries of the metal–SCNT contact besides
the metal material and the SCNT nature as the metal–
SCNT interface geometry has a significant impact on the
electron conductance [22]. Figure 4 shows the transmission
function of three different SCNT Schottky diode configurations
(traditional side-contact [20, 21], fully embedded, and partially
embedded (figure 1(a))) under zero bias. The atomic structures
of the side-contact and fully embedded diode models are shown
in the insets of figure 4. From the figure, the calculated energy
gap is about 0.92 eV under zero bias with small tails caused by
metal-induced gap states due to the tight coupling between Al
and C atoms. The transmission coefficients in the conduction
bands are larger than those in the valence band for all three
different models. However, the ratios of the transmission
coefficients in the conduction bands to those in the valence
bands for the three models are different. The fully embedded
model has the smallest transmission coefficient ratio, while the
Nanotechnology 19 (2008) 115203
other two models have a similar ratio. This is because the
fully embedded model, where the whole SCNT is wrapped
by the Al atoms in the contact region, has much stronger
coupling between the SCNT and the Al atoms than the other
two models. We can correlate the strength of the coupling with
the atomic bond length between the SCNT and the Al atoms
by comparing the shortest Al–C bonds in the three different
contact geometries. The shortest Al–C bond lengths for the
side-contact, partially embedded, and fully embedded models
are 3.00 Å, 2.10 Å and 2.03 Å, respectively, and we can see
that the fully embedded model has the smallest Al–C bond
length, indicating that there is strong coupling between the
SCNT and the Al electrode. Thus, the transmission coefficient
ratio, which is proportional to the diode rectification ratio, is
inversely proportional to the strength of the coupling between
the SCNT and the Al electrode.
4. Threshold voltage
According to the band diagrams shown in figure 3, the
threshold voltage depends on the electron Schottky barrier
height (ESBH), the energy difference between Fermi level and
the conduction band edge in our case. The ESBH is mainly
decided by the energy gap of the SCNT embedded in the Al
electrode and the Fermi level alignment to the conduction band
edge. For the sake of simulation time, in our calculation we
use smaller SCNT(8, 0) which has a diameter of 0.63 nm. The
calculated ESBH is 0.46 eV, half of the energy gap. Note that
the Fermi level is assumed to be aligned at the middle of the
energy gap in our calculation (figure 4). We can estimate the
ESBH for larger SCNTs to compare with the SCNTs used in
the experiments as the energy gap of the SCNT is inversely
proportional to the diameter of the SCNT [10]. For instance,
the ESBH of a SCNT with a diameter of 1.5 nm (close to the
size of the SCNT used in the experiment [4]) is 0.25 eV. Thus,
by tuning the ESBH, we can correlate the threshold voltage of
a larger diameter SCNT with a small SCNT(8, 0).
The tuning of the ESBH in the calculations is
straightforward. The calculations are performed in two steps:
electrode calculations and device calculations. In the electrode
calculations, we modify the calculated Fermi level of the right
electrode (SCNT), which would directly change the Fermi
level alignment in the transmission function of the diode in
the device calculations. In this way, we can calculate the I –V
curve (including the threshold voltage) of the diode with
smaller ESBH, which correlates to the larger diameter of
SCNTs.
Figure 5 illustrates the I –V characteristics of the partially
embedded model for the SCNT Schottky diode with different
ESBHs. The calculated results show that the threshold
voltage decreases as the ESBH decreases. Comparing to the
experiment where an SCNT (with diameter around 1.3 nm) is
embedded in Al (left electrode) and Pd (right electrode) [4],
we see that the threshold voltage for ESBH = 0.25 eV
(equivalent to an SCNT with diameter 1.5 nm) is a good fit
to the experiment results. The I –V curves for SCNTs with
ESBH = 0.46 and 0.15 eV, corresponding to SCNTs with
4
P Bai et al
Figure 5. Current–voltage characteristics of the partially embedded
model with the electron Schottky barrier heights (ESBH) of 0.46,
0.25 and 0.15 eV. The measured result [4] is also shown.
diameter 0.63 and 3.0 nm, are also plotted in figure 5 for
comparison.
Besides an excellent fit for the threshold voltage, it is also
encouraging to see that the magnitude of the current up to
1 V is within one order of magnitude. This shows that our
simulation methodology is sufficiently good to predict the I –V
characteristics of an SCNT Schottky diode by using ESBH to
correlate the I –V characteristics to the diameter of the SCNT
measured in experiments. The minor difference in the I –V
characteristics may be due to the SCNT–Pd contact and
additional defects that are probably introduced at the Al–SCNT
contacts in the experiment [4] which we did not model here.
5. Conductivity inside the diode
Besides the threshold voltage and I –V characteristics,
the conductivity within the SCNT Schottky diode is also
an important macroscopic property which is important to
consider. In order to study the conductivity of different parts
in the partially embedded diode model (i.e. the Al–SCNT
contact region and the SCNT outside the contact region), we
calculate the voltage drop inside the diode. The voltage drop is
calculated based on the difference of the electrostatic potential
of the biased structure and the zero-biased structure.
The contour plot of the voltage drop on the cross section
shown by line p in figure 1(b) is plotted in figure 6(a), while its
integrated values in the plane perpendicular to the transmission
direction is given in figure 6(b.) It is observed that the voltage
drop increases as the SCNT extends to the right electrode.
There is little voltage drop in the part where the SCNT is
partially wrapped with Al atoms. This can be understood by
considering electrons flowing through the Al electrode rather
than the SCNT. However, we see a small voltage drop at the
Al–C atom interfaces from figure 6(a). It tells us that electrons
also flow through the SCNT. This means that the metal–SCNT
contact region and the SCNT near the contact region have
better conductivity than the part close to the right electrode.
Contrary to the general belief that the metal–SCNT interface
has a large resistance, our finding shows otherwise. We must
emphasize that the voltage drop obtained in figure 6(b) is the
Nanotechnology 19 (2008) 115203
(a)
(b)
Figure 6. Forward bias voltage drop and the average conductivity
inside the conducting channel along transmission direction ( Z ) in the
partially embedded model. (a) Contour plots of the voltage drop on
the cross section shown by line p in figure 1(b). (b) Integrated values
of the voltage drop in the plane perpendicular to the transmission
direction.
average or effective value calculated from the 3D real space
effective potential in the simulation.
To understand the mechanism of the voltage drop along
the diode, we perform an analysis of charge transfer between
Al and C atoms in the atomic partially embedded model.
Figure 7 shows the difference between the charge density of the
combined Al–SCNT structure (figure 1(a)) and the sum of the
charge density of the isolated SCNT and pure metal Al atoms,
the latter two having the same structures as in figure 1(a).
When the SCNT is tightly bonded to the Al atoms, there is
a substantial hybridization between the electron states of the
Al atoms and those of the SCNT, which results in the metal-
induced inhomogeneous charge distribution in the SCNT [20].
From figure 7 we see that the electrons are depleted from the
Al and C atoms and move to the interface between the Al and
the C atoms, where the C atoms are in closer contact with
the Al atoms as circled in figure 1(b). This agrees well with
previous studies [19, 20]. The charge movement leads to the
formation of adsorption dipoles and the generation of a local
electric field, which results in energy-level re-alignment [15].
The charge transfer from the Al to the broadened SCNT levels
leads to a lowering of their energy and an increased binding
energy because of charge–image charge interactions [21]. The
metal-induced inhomogeneous charge distribution is localized
by nature. As a result, the SCNT near the Al–SCNT interface
becomes more conductive.
6. Conclusions
We have investigated the electron transport of the SCNT
Schottky diode through the atomic structures built based on
experimental set-ups using the first principles calculations.
Embedded contacts, (inferred from the device fabrication
process) are used to connect the metal and the SCNT
to build the atomic models of the diode. The I –V
5
P Bai et al
Figure 7. Contour plots of charge density difference in the partially
embedded model on the cross sections indicated by line p in
figure 1(b).
characteristic, threshold voltage and conductivity of the diode
are analysed from an atomic perspective by correlating them
to the atomic structure properties. Our calculations reveal
that the rectification behaviour of the diode is caused by
the asymmetric transmission coefficient distribution in the
conduction and valence bands and the shift of the band
edges under applied bias. A higher transmission coefficient
ratio between the conduction and the valence bands, which
is proportional to a better diode rectification ratio, could
be achieved by a loosely coupled metal–SCNT interface
configuration. The threshold voltage of the diode depends on
the electron Schottky barrier height which can be tuned by
varying the diameter of the SCNT. Contrary to general belief,
we have found that the metal–SCNT contact region inside the
diode has better conductivity than the region far away from
the Al electrode. This enhanced conductivity is attributed to
the localized hybridization between the electron states of the
carbon and the metal atoms which possibly lowers the potential
barrier and induces extra states. Our finding may play an
important role in improving the performance of the SCNT
Schottky diodes which have wide applications in nanoscale RF
microwave systems.
Acknowledgments
This work is supported by Singapore A*STAR SERC TSRP
grant no. 052 117 0029. We gratefully acknowledge
the beneficial discussion with Gengchiau Liang from NUS
Singapore and Zhenhua Yao’s group from Atomistix A/S,
Denmark.
References
[1] Tans S J, Verschueren A R M and Dekker C 1998 Nature
393 49
[2] Weitz R T, Zschieschang U, Effenberger F, Klauk H,
Burghard M and Kern K 2007 Nano Lett. 7 22
Nanotechnology 19 (2008) 115203 P Bai et al

[3] Lu C, An L, Fu Q, Liu J, Zhang H and Murduck J 2006 Appl. [15] Shen B and Cho K 2004 Phys. Rev. B 70 233405
Phys. Lett. 88 133501 [16] Tung R T 2001 Phys. Rev. B 64 205310
[4] Yang M H, Teo K B K, Milne W I and Hasko D G 2005 Appl. [17] Xue Y and Ratner M A 2004 Phys. Rev. B 70 205416
Phys. Lett. 87 253116 [18] Jimenez D, Cartoixa X, Miranda E, Sune J, Chaves F A and
[5] Manohara H M, Wong E W, Schlecht E, Hunt B D and Roche S 2007 Nanotechnology 18 025201
Siegel P H 2005 Nano Lett. 5 1469 [19] Zhu W and Kaxiras E 2006 Nano Lett. 6 1415
[6] Meng S, Maragakis P, Papaloukas C and Kaxiras E 2007 Nano [20] Okada S and Oshiyama A 2005 Phys. Rev. Lett. 95 206804
Lett. 7 45 [21] Dag S, Gulseren O, Ciraci S and Yildirim T 2003 Appl. Phys.
[7] Durkop T, Getty S A, Cobas E and Fuhrer M S 2004 Nano Lett. Lett. 83 3180
4 35 [22] Deretzis I and Magna A La 2006 Nanotechnology 17 5063
[8] Jeong W K and Jiang Q 2007 Nanotechnology 18 095705 [23] Zhang W, Lu W and Wang E G 2005 Phys. Rev. B 72 075438
[9] Javey A, Guo J, Paulsson M, Wang Q, Mann D, [24] Pomorski P, Roland C and Guo H 2004
Lundstrom M and Dai H 2004 Phys. Rev. Lett. 92 106804 Phys. Rev. B 70 115408
[10] Dresselhaus M S, Dresselhaus G and Avouris P 2001 Carbon [25] Sugino O and Oshiyama A 2002 Phys. Rev. Lett. 68 1858
Nanotubes: Synthesis, Structure, Properties, and [26] Brandbyge M, Mozos J, Ordejon P, Taylor J and
Applications (Berlin: Springer) Stokbro K 2002 Phys. Rev. B 65 165401
[11] Moon S, Lee S, Song W, Lee J S, Kim N, Kim J and [27] Zhou Y H, Zheng X H, Xu Y and Zeng Z Y 2006 J. Chem.
Park N 2007 Appl. Phys. Lett. 90 092113 Phys. 125 244701
[12] Nosho Y, Ohno Y, Kishimito S and Mizutani T 2006 [28] Wang Z, Kadohira T, Tada T and Watanabe S 2007 Nano Lett.
Nanotechnology 17 3412 7 2688
[13] Cui X, Freitag M, Martel R, Brus L and Avouris P 2003 Nano [29] Taylor J, Guo H and Wang J 2001 Phys. Rev. B 63 245407
Lett. 3 783 [30] Datta S 2000 Superlatt. Microstruct. 28 253
[14] Tseng Y C, Phoa K, Carlton D and Bokor J 2006 Nano Lett. [31] Singh J 2001 Semiconductor Devices: Basic Principle
6 1364 (New York: Wiley)