realization of low contact resistance close to theoretical ... · realization of low contact...
TRANSCRIPT
Nano Res
1
Realization of low contact resistance close to
theoretical limit in graphene transistors
Hua Zhong, Zhiyong Zhang(), Bingyan Chen, Haitao Xu, Dangming Yu, Le Huang and
Lian-Mao Peng()
Nano Res., Just Accepted Manuscript • DOI: 10.1007/s12274-014-0656-z
http://www.thenanoresearch.com on November 28 2014
© Tsinghua University Press 2014
Just Accepted
This is a “Just Accepted” manuscript, which has been examined by the peer-review process and has been
accepted for publication. A “Just Accepted” manuscript is published online shortly after its acceptance,
which is prior to technical editing and formatting and author proofing. Tsinghua University Press (TUP)
provides “Just Accepted” as an optional and free service which allows authors to make their results available
to the research community as soon as possible after acceptance. After a manuscript has been technically
edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP
article. Please note that technical editing may introduce minor changes to the manuscript text and/or
graphics which may affect the content, and all legal disclaimers that apply to the journal pertain. In no event
shall TUP be held responsible for errors or consequences arising from the use of any information contained
in these “Just Accepted” manuscripts. To cite this manuscript please use its Digital Object Identifier (DOI®),
which is identical for all formats of publication.
Nano Research
DOI 10.1007/s12274-014-0656-z
TABLE OF CONTENTS (TOC)
Realization of low contact resistance close to
theoretical limit in graphene transistors
Hua Zhong, Zhiyong Zhang*, Bingyan Chen,
Haitao Xu, Dangming Yu, Le Huang and
Lian-Mao Peng*
Peking University, China
High quality graphene/metal junction with contact resistance below 100 Ω μm at
room temperature has been realized.
Realization of low contact resistance close to
theoretical limit in graphene transistors
Hua Zhong, Zhiyong Zhang(), Bingyan Chen, Haitao Xu, Dangming Yu, Le Huang and Lian-Mao
Peng()
Received: day month year
Revised: day month year
Accepted: day month year
(automatically inserted by
the publisher)
© Tsinghua University Press
and Springer-Verlag Berlin
Heidelberg 2014
KEYWORDS
Graphene Field-Effect
Transistors,
Contact Resistance,
Metal-Graphene Interface,
Transfer Length Method
ABSTRACT
Realizing low contact resistance between graphene and metal electrode remains
a well-known challenge for building high-performance graphene devices. In
this work, we attempt to lower the contact resistance in graphene transistors
and further explore the resistance limit between graphene and metal contact.
Through adopting highly pure palladium and high-quality graphene and
controlling the fabrication process so as not to contaminate the interface, the
Pd/graphene contact resistance at room temperate is lowered to sub-100 Ω μm
level both on mechanically exfoliated and chemical-vapor-deposition graphene.
After excluding the parasitic series resistances from the measurement system
and electrodes, the retrieved contact resistance is shown to be systematically
and statistically less than 100 Ω μm, with the minimum value of 69 Ω μm which
is very close to the theoretical limit. Furthermore, the contact resistance shows
no clear dependence on temperature ranged from 77 K to 300 K, and this is
attributed to the saturation of carrier injection efficiency between graphene and
Pd owing to the high quality of the graphene samples used which have
sufficiently long carrier mean-free-path.
Nano Research
DOI (automatically inserted by the publisher)
Research Article
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
95 Nano Res.
1 Introduction
Contact resistance between metal and
semiconductor is crucial for electronic devices,
especially for field-effect transistors (FETs).
Realizing low contact resistance is therefore the
precondition for building high-performance FETs.
For a FET, the source/drain contact resistance
should be much smaller than the channel resistance
to ensure that semiconducting channel dominates
the whole output resistance, and devices with
higher mobility and shorter semiconducting
channel requires lower contact resistance.
Graphene is considered as a competing channel
material for constructing high-speed FETs for
radio-frequency (RF) applications owing to its
ultra-high carrier mobility [1]. Compared with
conventional semiconducting devices with similar
gate length, graphene FETs should have lower
channel resistance per transfer mode due to the
much higher carrier mobility or Fermi velocity, and
thus lower contact resistance than conventional
device between channel and source/drain junction
is required [2]. However realizing low contact
resistance between graphene/metal is a well-known
challenge due to the small density of states (DOS)
of graphene near the Dirac point. In the past five
years, tremendous endeavors have been done to
reduce the contact resistance of graphene FETs. For
example, various kinds of metals, such as titanium
[2], palladium [3], nickel [4], cobalt [5], chromium
[6] and gold [7], have been used to contact
graphene aiming to transfer more charges into
graphene and to increase its carrier density. On the
other hand, various methods for interface
modification, such as oxygen plasma treatment
[8-10], rapid thermal annealing [11], and graphene
contact area patterning [12], have been
demonstrated to lower graphene/metal contact
resistance via increasing the carrier density of
graphene or the carrier injection area/length.
Values of Rc obtained from all these studies ranges
from hundreds to thousands Ω μm, and the lowest
metal/graphene contact resistance reported at room
temperature is around 100 Ω μm with Cu contact
and Au-grain contact [12, 13]. Such a large contact
resistance severely degrades the performance of
graphene FETs (GFETs), limiting such key device
parameters as the on-current and the peak
transconductance [13]. Contact resistance is
therefore becoming the main obstacle for further
performance improvement of GFETs, especially as
channel length scales down to sub-100 nm. Further
reducing contact resistance is currently the most
urgent requirement and the most profitable task for
the development of high performance GFETs.
This work aims to lower the contact
resistance in graphene transistors and explore the
resistance limit. In particular, we reconsider the
Pd/graphene contact experiment and analysis that
were reported in Ref. [3], where a contact resistance
down to 185±20 Ω μm was reliably realized at RT.
By using highly pure palladium and high-quality
graphene, and controlling the device fabrication
process and thus maintaining the clean interface,
the palladium/graphene contact resistance at RT is
lowered to sub-100 Ω μm, which has never been
systematically obtained before. After excluding the
parasitic series resistances from measurement
system, the retrieved contact resistance is shown to
be systematically and statistically less than 100 Ω
μm with the lowest value of 69 Ω μm, which is
very close to the theoretical limit. Furthermore, on
contrast to the linear dependence on temperature
reported in Ref. [3], the graphene/metal contact
resistance in our experiments does not show
obvious temperature dependence. We attribute this
temperature independent contact resistance to the
saturation of injection efficiency between graphene
and Pd owing to the sufficiently long
mean-free-path of carriers in clean graphene. In
addition, the ultra-low contact resistance also
benefits from the usage of purer Pd layer, which
should induce heavier hole transfer between Pd
and graphene than in previous experiments. 2 Experimental and Results Graphene flakes were mechanically exfoliated onto
heavily-doped silicon substrate covered with 285
nm silicon oxide. Single layer graphene flakes were
identified by optical microscope, and further
confirmed by Raman spectroscopy before device
fabrication and by atomic force microscopy (AFM)
Address correspondence to Zhiyong Zhang, [email protected]; Lian-Mao Peng, [email protected]
| www.editorialmanager.com/nare/default.asp
96 Nano Res.
after electronic measurements. Typical Raman
spectrum and AFM image are presented in Figures
S1 and S2 (see supporting information). The
identified monolayer graphene flakes were firstly
etched into strips with 2 μm width, and then metal
electrodes (Pd/Au 30/50 nm) were patterned and
deposited to contact graphene through standard
electron-beam lithography (EBL) and
electron-beam evaporation (EBE) processes. It is
worth mentioning that the metal Pd with high
purity (99.95%) was evaporated in high vacuum
with base pressure of about 1×10-8 Torr. Pd is one
of the best contact metal for graphene due to the
high work function of Pd and its wetting property
on graphene. The high work function provides
high density of carrier transferred from Pd to the
graphene, and then leads to large number of
conducting modes. Meanwhile the good wetting
between graphene and Pd can provide large
transmission efficiency for carrier. Metal contact
length was designed as 1 μm, which is much larger
than the transfer length of carriers between the
metal contact and graphene and is sufficient for
injection of carriers [3, 14]. A group of devices were
fabricated on the same graphene strip, with
channel length arranging from 1 μm to 6 μm and
with a step of 1 μm. A scanning electron
microscope (SEM) image showing a group of
GFETs is presented as the inset of Figure 1a. Total
resistance (Rtotal) of each GFET was measured in
vacuum at RT as a function of back-gate voltage
(Vbg), with the source-drain bias (Vds) being set to
0.05 V. Transfer characteristics of a typical group of
GFETs are plotted in Figure 1a, in which the largest
resistance points (corresponding to the Dirac point
voltage Vdirac) are found to locate at -8.3±1.3 V,
corresponding to a residual carrier density
fluctuation of less than 1×1011 cm-2. The narrow
distribution of Vdirac reflects excellent uniformity of
the group of fabricated devices, while such a low
residual carrier density indicates that the graphene
quality remained very high even after device
fabrication. Under large gate bias Vbg=Vdirac±50 V,
the graphene channel is heavily electrical doped
with an average carrier density of up to 3.78×1012
cm-2, and effects due to residual carrier density
fluctuation are becoming negligible since the
residual carrier density is much smaller than the
total carrier density. Standard transfer length
method (TLM) is used here to extract contact
resistance, by calculating the total resistance of
graphene channel as
Rtotal=Rsheet/W×L+2Rc/Wc (1)
where Rsheet is the sheet resistance of graphene, L
and W are respectively the channel length and
width, and Rc is the contact width (Wc) normalized
contact resistance. In our experiments, the channel
width (W) is designed to equal the contact width
(Wc). By linearly fitting Rtotal of GFETs with
different channel lengths, 2Rc/Wc and Rsheet/W can
be obtained from the intercept (Rint) and slope
simultaneously. The retrieved Rsheet/W (black
squares) and Rc (red squares) at various back gate
voltages (Vbg) are plotted in Figure 1b, in which
both Rsheet and Rc decline with increasing net gate
voltage (|Vbg-VDriac|). The gate voltage dependence
of Rsheet and Rc are both due to field doping via gate,
which shifts the Fermi level of graphene in the
channel and beneath the metal contact and
modifies the density of states of graphene channel
[3]. It is worth mentioning that the minimum Rc at
the p branch Rcp approaches 104.7±12.9 Ω μm at
Vbg=Vdirac-50 V (the left end of measurement range),
which is smaller than the 185±20 Ω μm through Pd
contact in Ref. [3], and even smaller than the
published minimum metal/graphene contact
resistance [12, 13]. Similar to previous results, the
contact resistances at p-branch and n-branch are
asymmetric. For example the Rc at back-gate
voltage of Vdirac+50 V at n-branch is 396.3±35.7 Ω
μm, which is almost three times larger than that at
p-branch. Asymmetry found in Rc originates from
the metal-induced doping, i.e. electrons transfer
into or out of graphene under the metal contact.
The measured asymmetric Rc demonstrates that
graphene is heavily p-doped by Pd, as Rc in
p-branch is always smaller than that in n-branch.
The more asymmetric Rc between n and p branch
found here than that reported in Ref. [3] indicates
heavier doping in our Pd/graphene contact. In
addition, the net charge density induced by gate
voltage [given by Cbg(Vbg-VDirac)/e], is about 3.80×1012
cm-2 for Rc =104.7 Ω μm in our graphene channel,
while the corresponding value reported in Ref. [3]
for Rc =185 Ω μm is 7.18×1012 cm-2 (with -30 V net
gate voltage and 90 nm gate oxide). Realization of
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
97 Nano Res.
smaller contact resistance at lower gate-induced
carrier density here indicates that the contact
resistance improvement is originated from the
improved carrier injection at metal/graphene
interface, rather than the increased carrier
concentration and conduct modes in graphene
channel.
Obviously, the reliability of retrieved Rc and
Rsheet through TLM relies on the uniformity of
graphene channel and graphene/metal contacts
among all GFETs in one group. In an ideal situation,
the Rc and Rsheet obtained from GFETs in one group
should be identical, and Rtotal of these devices
should depend linearly on channel length at fixed
bias. However, absolute uniformity among GFETs
is impossible to achieve in reality since discrepancy
among devices, such as those caused by charged
impurity (manifested by the Dirac point voltages
fluctuation) and introduced during device
fabrication process, is inevitable. In general, the
more uniform the graphene channel is, the more
reliable the retrieved Rc and Rsheet are. In the group
of GFETs as shown in Figure 1a, excellent
uniformity is evident as manifested by the small
value of VDirac and its narrow distribution, and
retrieved results through TLM are thus quite
reasonable and reliable. Reliability of the retrieved
results is further verified through the following
data analysis. We randomly pick out i (=2, 3, 4, 5, 6)
GFETs with different L from the group of six GFETs,
and linearly fit Rtotal at Vbg=VDirac-50 V to discern the
divergence between each other. For all possible
combinations (∑𝐶6𝑖=57, for i =2, 3, 4, 5, 6), Rint are all
below 120 Ω. Distribution of Rint is given in Figure
S3, showing that 85 percent of them lie between
100±20 Ω. Figure 1c gives five fitting curves with
five different i, each of them corresponds to the
largest intercept under fixed i. Therefore the up
limit of Rint is no more than 120 Ω, which testifies
the rationality of the minimum Rc about 104.7±12.9
Ω μm retrieved through TLM in our experiments.
As another evidence to support our claim is
that it is indeed possible to obtain such a low
contact resistance. We also fabricated a short
channel GFET (with 500 nm gate length) on
mechanically exfoliated graphene using Pd as
metal contacts. As shown in Figure 1d, the device
exhibits an on-state current (Ion) as large as 324 μA
at source/drain bias voltage of 0.05 V and back-gate
voltage of -60 V, corresponding to an on-state
resistance (Ron) of 154 Ω. The resistance of
graphene channel (W=2 um and Leffect=0.3 μm),
taking charge transfer effect into account, is
estimated to be about 60 Ω by assuming Rsheet of
400 Ω/□, leading to a single Pd/graphene junction
Rc ~ 94 Ω μm [16]. In a former report, Guo et al.
also claimed that Rc in a palladium-contacted
100-nm top-gated GFET is less than 100 Ω μm [17].
Therefore it is reasonable to lower resistance of
Pd/graphene contact down to sub-100 Ω μm. By
the way, the 500 nm GFET presents obvious
n-branch suppression and positive Dirac point
voltage different from that in longer channel
devices. It is mainly because the graphene channel
could be influenced more easily by the charges
transferred from the contact metal compared to the
long-channel-length transistors.
Strictly speaking, contact resistance Rc should
contain four components, i.e. metal/graphene
interface junction resistance (Ri), resistance of metal
upon the interface (Rm), resistance of graphene
under the interface (Rg), and potential junction
resistance in graphene between the contact and
channel parts (Rj), i.e. Rc=Ri+Rm+Rg+Rj. However, in
an actual measurement (depicted in Figure 2a), the
measured Rtotal of a GFET also contains a parasitic
series resistance (Rp) from measurement system,
which includes interconnect metal line resistance
and probe/pad contact resistance, and the real Rtotal
should be expressed as
Rtotal=Rsheet/W×L+2Rc/Wc+Rp. (2)
Therefore the retrieved Rc (via, e.g. TLM) by just
removing the channel length dependent sheet
resistance of graphene channel also contains the
parasitic series resistance (Rp), which exists in all
measurement systems and is always neglected in
metal/graphene contact resistance measurements
and analysis. To obtain the intrinsic Rc of a GFET,
the parasitic series resistance (Rp) should be
measured and subtracted from the obtained Rint,
which means the intrinsic contact resistance
expression is
Rc=(Rint-Rp)×Wc/2. (3)
Note, Rp is a system parameter and independent of
contact width, which should be subtracted before
contact resistance normalization. In our
| www.editorialmanager.com/nare/default.asp
98 Nano Res.
experiments, we develop a method to estimate Rp
by measuring the resistances of metal ‘devices’,
whose geometry are exactly as the same as the
GFET, but with the graphene channel replaced by a
metal line with width of 4 μm. SEM image of the
calibration structure is shown in the inset of Figure
2b, and a full view of the device patterning is
shown in Figure S5. Two sets of metal ‘devices’ are
fabricated and measured, and all 12 devices show
resistances around 25 Ω as presented in Figure 2b.
The resistances of those 12 devices have no obvious
dependence on channel length, suggesting that the
“metal” channel resistance is negligible. We thus
conclude that the parasitic series resistance Rp of
our measurement system is around 25 Ω. After
subtracting the effect of parasitic series resistance
Rp (here we used Rp=20 Ω in Eq. (3) to avoid
possible overestimation of contact resistance) from
the retrieved Rc in Figure 1c, the gate voltage
dependent intrinsic contact resistance then is
obtained and shown in Figure 2c. Obviously the
minimum contact resistance at p-branch (Vbg= VDirac
-50 V) is about 82.7 Ω μm, which is much smaller
than all published values of contact resistance for
graphene/metal junction. Moreover the contact
resistance at p-branch is very close to the
theoretical limit, which is calculated based on
Landauer formula for an ideal metal/graphene
contact on 285 nm silicon oxide and plotted as the
red line in Figure 2c [3].
To further confirm the reliability of contact
resistance below 100 Ω um in graphene/metal
junction, we fabricated 6 sets of GFETs (denoted as
S1 to S6) based on mechanically exfoliated
graphene flakes. The original contact resistances
are extracted by TLM and plotted in Figure 2d, in
which the p-branch contact resistance Rcp (green
squares) at Vbg= VDirac-50 V and n-branch contact
resistance Rcn (red dots) at Vbg= VDirac+50 V are
respectively shown together with their standard
errors. The retrieved contact resistances of the 6
sets devices are quite uniform, i.e. Rcp locates
around 100 Ω μm and Rcn is distributed about 400
Ω μm. The minimum contact resistance Rc, i.e Rcp in
Figure 2d, are found to be between 89 to 140 Ω μm,
and 2 of them are below 100 Ω μm. After
calibration, i.e. removing 20 Ω parasitic series
resistance, the intrinsic minimum contact resistance
Rc (black stars) are seen to distribute between 69 Ω
μm to 120 Ω μm, and half of them show Rc below
100 Ω μm. In addition, we fabricated two more sets
of GFETs using chemical-vapor-deposition (CVD)
graphene, and the extracted minimum Rc is also
found to be about 100 Ω μm (Figure 2d, denoted as
C1 and C2). Details of the CVD graphene samples
are presented in the supporting information
(detailed in Figure S4 and S5). The experimental
results both on mechanically exfoliated and CVD
graphene showed that contact resistance between
graphene and metal can be lowered down to
sub-100 Ω μm at RT, and sometime even close to
the theoretical limit.
Temperature dependence of contact
resistance in graphene devices is controversial. For
example, Xia et al. reported that Rc is linearly
dependent on temperature and drops from 185 Ω
μm to 120 Ω μm as temperature decreases from 300
K to 6 K [3]. However, completely different results
were reported by other group, showing that Rc of
optimized graphene/metal contact is independent
of temperature [18]. Therefore it is necessary to
check the relation between temperature and contact
resistance of our graphene devices. We measured
the transfer characteristics of the fabricated GFETs
from RT (300 K) down to liquid nitrogen
temperature (77 K), and transfer characteristics of a
typical GFET with channel length of 1 um in
sample S1 are shown in Figure 3a. As expected, the
transfer characteristics become sharper near the
Dirac point voltage as the temperature is decreased,
which may be attributed to the enhanced carrier
mobility at lower temperature. Field-effect mobility
of holes is found to increase averagely from about 4,
000 cm2V-1s-1 to near 10, 000 cm2V-1s-1 as
temperature decreased from 300 K to 77 K owing to
the suppressed surface phonon scattering on SiO2
substrate [19]. As a comparison, fitted mobility (μFT,
after decoupling the effect of contact resistance) is
also found to increase from about 4, 500 cm2V-1s-1 to
near 12, 000 cm2V-1s-1 as shown in Figure 3b. It
should be noted that the fitted mobility is larger
than field-effect mobility owing to its decoupled
contact resistance, and the difference becomes
more obvious at lower temperature, showing that
the effect due to contact resistance is becoming
more important as temperature decreased. The bias
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
99 Nano Res.
dependent contact resistance Rc is retrieved
through TLM at each temperature (except for room
temperature, all the transfer characteristics were
measured continuously from 77 K to 250 K) and
shown in Figure 3c. It should be noted the Rc
values at different temperatures are tending to
converge to a constant at large gate bias, indicating
that the minimum contact resistance is in fact
independent of temperature. It is also found that Rc
at several temperatures becomes negative near the
Dirac point voltage, suggesting that the TLM
method is not valid in this region any more. The
reason behinds the failure of TLM could be the
contact-induced doping effect [20], i.e. the
gate-induced carrier density is not sufficient to
screen out charges transferred from metal contact
into graphene channel. The independence between
contact resistance and temperature is further
demonstrated by Figure 3d, which gives Rc-T
relation of all 6 groups of grapheme devices. It is
obvious that none of the 6 groups show clear
temperature dependence, which differs from the
Rc-T relation reported in Ref. [3] while similar to
that found in Ref. [18].
3 Discussion
Although using the same metal (Pd) to contact
graphene, this work is different from that reported
by Xia et al. [3] in two important aspects. (1) The
contact resistance is found here to be independent
of temperature, while it was reported to be linearly
dependent on temperature by Xia et al.; (2) The
reported values of graphene/metal contact
resistance by Xia et al. is much larger than that
realized in this work. To find out why our results
are so different from that reported by Xia et al., we
compare some key parameters between our results
and Ref. [3] and list them in Table 1. In our opinion,
quality of graphene, work function difference,
interaction between metal and graphene, and the
properties of interface are those decisive factors.
Firstly, high quality graphene samples are
used in this work which are found to remain high
quality even after device fabrication. Carrier
mobility and mean-free-path are used to
characterize the quality of graphene samples. Hole
mobility of our samples all above 4,000 cm2V-1s-1
(Figure 4a) at RT, which are much higher than the
mobility of about 2,000 cm2V-1s-1 at RT and 3,600
cm2V-1s-1 at 6 K in Ref. [3]. Furthermore our
samples have a lower density of residual charges
(n0) and much smaller Vdirac which is closer to zero
(besides tox is different), both of which reflect the
cleanness and high-quality of the graphene
samples used in this work. The averaged n0 (3.0~4.0
×1011 cm-2) and μFE (4,000~4,500 cm2V-1s-1) of our
samples are plotted in Figure 4a. And the surface of
our graphene samples is seen to be very clean (see
the AFM image displayed in Figure S2 of
supporting information). Higher carrier mobility
indicates longer mean-free-path, which in turn
suggests better carrier injection efficiency between
graphene and metal. The mean-free-path Lm of hole
in a typical device is about 80 nm at 300 K, and
increases to about 125 nm at 78 K as shown in
Figure 4b. The mean-free-path of hole in graphene
is one of the important facts that dominate the
transmission efficiency TMG between graphene and
metal [9]. The relation between transmission
efficiency Tmg and Lm as given in Figure 4c (details
on the calculation are given in Methods) shows that
a decrease of temperature leads to an increase in Lm
and thus an increase in Tmg. It should be noted that
the gradient of Tmg becomes smaller and smaller
with increasing Lm, indicating that the dependence
of Tmg (or Rc) on Lm will become very weak if Lm is
sufficient large. Owing to the long Lm realized in
our graphene samples, TMG varies only from 0.81 at
300 K to 0.87 at 78 K, suggesting only 7% change
for Tmg over the entire temperature range in our
measurements. With smaller Tmg as reported in Ref.
[3], i.e. 0.75 even at 6 K, Tmg shows more sensitive
dependence on Lm. Therefore the negligible
temperature independence found in our
experiments on metal/graphene contact resistance
is mainly resulted from the use of higher quality
graphene samples, which is consistence with those
reports on metal/graphene interfaces with long Lm.
Secondly, higher metal-induced doping
concentration in the graphene under contact is
observed in this work than that in Ref. [3] owing to
the clean interface and high-quality Pd film
deposited in high vacuum. While it is difficult to
give a quantitative description of this character, we
may deduce the picture from the following two
| www.editorialmanager.com/nare/default.asp
100 Nano Res.
observations. (1) Heavier doping leads to higher
degree of asymmetry in contact resistances, which
may be measured by the resistance ratio
Rratio=Rcn/Rcp, where Rcn and Rcp are respectively the
contact resistance of n- and p-branch at fixed gate
voltage. It is obviously that Rratio is as large as about
4 (Figure 2d) at |Vbg-VDirac|=50 V. As a comparison,
the ratio is much smaller in Ref. [3] (about 1.4, from
Figure 2c of Ref. [3]). Larger contact resistance at
n-branch than at p-branch is originated from the
Pd-induced hole doping concentration in graphene
under the metal, and the heavier of p-doping, the
larger of the asymmetric Rratio [15]. (2) There are
two contact resistance peaks in Figure 4d. The main
peak locates at |Vbg-VDirac|=0, corresponding to the
Dirac point of graphene in the channel, and the
second peak locates at Vbg-VDirac>0, corresponding
to the charge neutral point of graphene under Pd
contact. For the observation of the second
resistance peak, the Fermi level shift (ΔEF, using the
Dirac point as the reference) or charge doping
concentration must be sufficiently large for the
minor peak to be distinguished from the main peak.
Simultaneously, coupling strength cannot be too
strong to broaden the second peak from
observation. It was supposed the appearance of the
second peak means a Fermi level shift of more than
125 meV [21], while ΔEF is about 80 meV in Ref. [3].
The different coupling strength, also affected by the
metal contact, quality of graphene and interface
properties, could be the main reason, in our
judgment, that finally leads to above two
remarkable differences.
4 Conclusions
In conclusion, we attempted to lower the contact
resistance in graphene transistors and explored the
limit resistance between graphene and metal.
Through adopting highly pure Pd and high-quality
graphene, companied with clean interface, the
palladium/graphene contact resistance at RT has
been lowered to sub-100 Ω μm both on
mechanically exfoliated and CVD graphene. After
excluding the parasitic series resistances from
measurement systems and electrodes, the retrieved
contact resistances are shown to be systematically
and statistically below 100 Ω μm with the
minimum value of 69 Ω μm which is very close to
the theoretical limit. Furthermore the contact
resistances shows no clear dependence on
temperature ranged from 77 K to 300 K, and this is
attributed to the saturation of injection efficiency
between graphene and Pd owing to the sufficiently
long mean-free-path of carrier in the graphene
samples used in this work.
Methods: [Experiments]
Graphene flakes used in this work were
mechanically exfoliated onto the silicon substrate
with 285 nm thick of silicon oxide. In the devices
fabrication process, graphene flakes were firstly
etched into a 2-μm width strips using oxygen
plasma reactive-ion etching (RIE) with shield of
polymethyl methacrylate (PMMA), defined by
traditional electron-beam-lithography (EBL)
process. The windows for source/drain contacts
are patterned through another EBL process on
PMMA. Then, Pd/Au (30 nm/ 50 nm) metal layer
was deposited using electron-beam-evaporation
(EBE) under base vacuum of 1 × 10-8 Torr,
followed by a lift-off process. Electronic
measurements were carried out using Keithley
4200 SCS in vacuum (under pressure about 1×
10-3 Torr) at RT. Low temperature measurements
were performed in a cryogenic probe station
(Lakeshore TTP-4) using liquid nitrogen as
cooling gas.
[Theoretical limit of graphene/metal
contact resistance]
The theoretical limit of graphene/metal
contact resistance is calculated based on
Landauer formula, which is G=(4e2/h)TM. h is
Planck constant, e is elementary charge, and T
and M are separately the carrier transmission
efficiency and number of quantum modes. For an
ideal metal/graphene contact on 285 nm silicon
oxide, the transmission efficiency T=1. Number of
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
101 Nano Res.
quantum modes can be derived from the Fermi
level shift ΔEF, that is M=2(ΔEF/πhVF), VF is the
Fermi velocity. As the carrier density of graphene
channel is controlled by the 285 nm silicon oxide
back-gate, which is determined by ρ
=-Cox(Vbg-VDirac)=4πeΔEF|ΔEF |/(h2VF2). So, for a
certain back-gate voltage, there is certain net
carrier density which corresponding to a certain
number of quantum modes and limited
conductance or resistance.
[Parameter calculation]
Field-effect mobility is calculated using
μFE=gmL/(WVdsCox), where gm is the
transconductance, Cox is the back-gate capacitance,
Vds is the source-drain bias voltage, and L and W
are respectively the channel length and width.
Averaged mobility is the average of mobilities of
GFETs with different channel length in the same
set. Actual mobility of graphene could be slightly
higher than μFE, as contact resistance degrades
transconductance which leads to a degraded
mobility value. The usually used model [22],
fitting the experimental data using a single
mobility value μFT, takes the contact resistance
into account, but the results are compromised
between Rc and μFT, and the mobility is always
overestimated [23].
Mean free path Lm is calculated according to
Lm=σh/2e2kF, σ is the conductivity which is
obtained from the linear fit of transfer length
method, h is Planck constant, e is elementary
charge, and 𝑘𝐹 = √𝜋𝑛 is the Fermi wave vector,
where n is the net carrier density.
Transmission efficiency 𝑇𝑚𝑔 =
√𝐿𝑚/(𝐿𝑚 + 𝐿𝑚𝑔) is the carrier tunneling
probability between metal contact and graphene,
where Lmg=hvF/2π2η is the effective coupling
length and coupling strength η is assumed to be 5
meV.
Acknowledgements
This work was supported by the Ministry of
Science and Technology of China (Grant Nos.
2011CB933001 and 2011CB933002), and National
Science Foundation of China (Grant Nos. 61322105,
61271051, 61321001 and 61390504), and Beijing
Municipal Science and Technology Commission
(Grant Nos. Z131100003213021 and
D141100000614001).
Electronic Supplementary Material:
Supplementary material (Raman spectroscopy
measurements, AFM imaging, linearity
demonstration, details of CVD samples and SEM of
calibration structure) is available in the online
version of this article at
http://dx.doi.org/10.1007/s12274-***-****-*
(automatically inserted by the publisher).
References
[1] Novoselov, K. S.; Fal’ko, V. I.; Colombo, L.; Gellert, P.
R.; Schwab, M. G.; Kim, K. A roadmap for graphene. Nature
2012, 490, 192-200.
[2] Nagashio, K.; Nishimura, T.; Kita, K.; Toriumi, A.
Metal/graphene contact as a performance killer of ultra-high
mobility graphene-analysis of intrinsic mobility and contact
resistance. IEDM 2009, 565-568.
[3] Xia, F. N.; Perebeinos, V.; Lin, Y. M.; Wu, Y. Q.; Avouris,
P. The origins and limits of metal-graphene junction
resistance. Nat. Nanotechnol. 2011, 6, 179-184.
[4] Venugopal, A.; Colombo, L.; Vogel, E. M. Contact
resistance in few and multilayer graphene devices. Appl.
Phys. Lett. 2010, 96, 013512.
[5] Nouchi, R.; Shiraishi, M.; Suzuki, Y. Transfer
characteristics in graphene field-effect transistors with Co
contacts. Appl. Phys. Lett., 2008, 93, 152104.
[6] Nagashio, K.; Nishimura, T.; Kita, K.; Toriumi, A.
Systematic investigation of intrinsic channel properties and
contact resistance of monolayer and multilayer graphene
field-effect transistor. Jpn J. Appl. Phys. 2010, 49, 051304.
[7] Malec, C. E.; Davidović, D. Electronic properties of
Au-graphene contacts. Phys. Rev. B 2011, 84,033407.
[8] Robinson, J. A.; LaBella, M.; Zhu, M.; Hollander, M.;
| www.editorialmanager.com/nare/default.asp
102 Nano Res.
Kasarda, R.; Hughes, Z.; Trumbull, K.; Cavalero, R.; Snyder,
D. Contact graphene. Appl. Phys. Lett. 2011, 98, 053103.
[9] Choi, M. S.; Lee, S. H.; Yoo, W. J. Plasma treatments to
improve metal contacts in graphene field effect transistor. J.
Appl. Phys. 2011, 110, 073305.
[10] Li, W.; Liang, Y. R.; Yu, D. M.; Peng, L. M.; Pernstich,
K. P.; Shen, T.; Hight Walker, A. R.; Cheng, G. J..; Hacker, C.
A.; Richter, C. A.; Li, Q. L. ; Gundlach, D. J.; Liang, X. L.
Ultraviolet/ozone treatment to reduce metal-graphene
contact resistance. Appl. Phys. Lett. 2013, 102, 183110.
[11] Balci, O.; Kocabas, C. Rapid thermal annealing of
graphene-metal contact. Appl. Phys. Lett. 2012, 101, 243105.
[12] Smith, J. T.; Franklin, A. D.; Farmer, D. B.;
Dimitrakopoulos, C. D. Reducing contact resistance in
graphene devices through contact area patterning. ACS Nano
2013, 7, 3661-3667.
[13] Parrish, K. N.; Akinwande, D. Impact of contact
resistance on the transconductance and linearity of graphene
transistors. Appl. Phys. Lett. 2011, 98, 183505.
[14] Xu, H. T.; Wang, S.; Zhang, Z. Y.; Wang, Z. X.; Xu, H.
L.; Peng, L. M. Appl. Phys. Lett. 2012, 100, 103501.
[15] Huard, B.; Stander, N.; Sulpizio, J. A.;
Goldhaber-Gordon, D. Evidence of the role of contacts on
the observed electron-hole asymmetry in graphene. Phys.
Rev. B 2008, 78, 121402(R).
[16] Assumption of Rsheet=400 Ω/□ is dependent on the
experimental data as shown in Fig. 1b, where Rsheet/W=200
Ω μm and W=2 μm. For a rough estimation, we have ignored
the effect of charges that transferred from metal contact to
graphene channel near the contact edges, if included, the Rc
should lie between 50 Ω μm and 154 Ω μm depending on the
scale of contact-induced doping region. As a further
reasonable estimate, the effective graphene channel length
(LEff) can be estimated through L-LT, where LT is the charge
tranfer length in graphene, with order about of 100 nm. And
then the LEff is then about 300 nm.
[17] Guo, Z. L.; Dong, R.; Chakraborty, P. S.; Lourenco, N.;
Palmer, J.; Hu, Y. K.; Ruan, M.; Hankinson, J.; Kunc, J.;
Cressler, J. D.; Berger, C.; de Heer, W. A. Record maximum
oscillation frequency in C-face epitaxial graphene transistors.
Nano Lett. 2013, 13, 942.
[18] Wang, L.; Meric, I.; Huang, P. Y.; Gao, Q.; Gao, Y.;
Tran, H.; Taniguchi, T.; Watanabe, K.; Campos, L. M.;
Muller, D. A.; Guo, J.; Kim, P.; Hone, J.; Shepard, K. L.;
Dean, C. R. One-dimensional electronic contact to a
two-dimensional material. Science 2013, 342, 614-617.
[19] Chen, J. H.; Jang, C.; Xiao, S. D.; Ishigami, M.; Fuhrer,
M. S. Intrinsic and extrinsic performance limits of graphene
devices on SiO2. Nat Nano 2008, 3, 206-209.
[20] Nouchi, R.; Saito, T.; Tanigaki, K. Observation of
negative contact resistance in graphene field-effect
transistors. J. of Appl. Phys. 2012, 111, 084314.
[21] Knoch, J.; Chen, Z. H.; Appenzeller, J. Properties of
metal-graphene contacts. IEEE Trans. Nanotechnol. 2012,
11, 513-519.
[22] Kim, S.; Nah, J.; Jo, I.; Shahrjerdi, D.; Colombo, L.;
Yao, Z.; Tutuc, E.; Banerjee, S. K. Realization of a high
mobility dual-gated graphene field-effect transistor with
Al2O3 dielectric. Appl. Phys. Lett. 2009, 94, 062107.
[23] Zhang, Z. Y.; Xu, H. L.; Zhong, H.; Peng, L. M. Appl.
Phys. Lett. 2012, 101, 213103.
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
95 Nano Res.
Table 1 Comparison of key parameters retrieved with that reported in Ref. [3].
tox(nm) μ
(cm2V-1s-1)
n0
(×1011 cm-2)
Vdirac
(V)
ΔEF
(meV)
Rration Rc(Ω μm)
Ref. [3] 90 2000 5.0 23 < 100 1.2 185
This work 285 4000 3.5 -8.3 >125 3.5 104 (84)
| www.editorialmanager.com/nare/default.asp
Nano Res.
Figure 1. Measurements of contact resistance of palladium/graphene interface and parameter retrieval. (a) A set of
Rtotal from a group of GFETs versus back-gate voltage. Inset: typical SEM image showing the channel of a group
of GFETs. Channel width is 2 μm and channel length varies from 1 μm to 6 μm, with step of 1 μm, and the scale
bar denotes 2 μm. (b) Rsheet/W of graphene channel (black squares) and Rc extracted by TLM (red squares),
together with the standard errors, versus back-gate voltage after shifting the Dirac point voltage to zero. Rsheet is
divided by the channel width for convenient illustration. (c) Linearity of Rtotal of GFETs in set S1. Fitting curves of
i (i=2, 3, 4, 5 ,6) GFETs that define the up limit of Rint. (d) Transfer characteristics of a 500-nm GFET. The largest
on-current Imax is 323.9 μA, corresponding to a minimum total resistance Rmin of 154.4 Ω. L=0.5 μm, W =2 μm,
and Vds=0.05 V.
0 1 2 3 4 5 60
200
400
600
800
1000
1200
1400 6 GFETs
2 GFETs
3 GFETs
4 GFETs
5 GFETs
Rtotal
c
Rto
tal(
L(m)
At 300 K
Vbg
-Vdirac
=-50 V
-40 -20 0 20 40 600
200
400
600
800
1000
1200
1400
1600
Vbg
-Vdirac
(V)
Rs
he
et/W
(
m)
b
Rsheet
/W
Rc
0
200
400
600
800
1000
1200
1400
1600
Rc
At 300 K
-60 -40 -20 0 20 40 600
2000
4000
6000
8000
10000
1um
2um
3um
4um
5um
6um
Rto
tal (
)
Vbg
(V)
a
Set S1
At 300 K
Vds=0.05 V
-60 -40 -20 0 20 40 6050.0µ
100.0µ
150.0µ
200.0µ
250.0µ
300.0µ
350.0µ
Vbg
(V)
I ds(A
)
100
200
300
400
500
600
700
Rto
tal (
)
Imax=323.9 A
Rmin=154.4
d
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
Nano Res.
Figure 2. Model and retrieved parameter statistics. (a) A schematic of the measured Rtotal, Rtotal=2Rc/Wc+
Rsheet/W×L+Rp. Definition of Rc is shown in the red dash rectangle. (b) Measured Rp of metal ‘devices’ with
different channel length. Inset: SEM image of the metal ‘channels’. Scale bar denotes 2 μm. (c) Comparison
of experimental contact resistance to the ideal G/M graphene. Red line, contact resistance in an ideal
metal-graphene contact on 285 nm oxide. (d) Minimum Rc in the n branch (red dots) and p branch (black
squares) from all sets of our graphene samples, including exfoliated graphene (denoted as S) and CVD
graphene (denoted as C) samples. The black starts represent the final Rc in p branch after excluding Rp in
DC measurement system. The dash lines are 400 Ω μm (red) and 100 Ω μm (black) respectively.
1 2 3 4 5 60
5
10
15
20
25
30
35
40
45
50
Rp(
)
Lm(um)
Rp1
Rp2
b
-40 -20 0 20 400
400
800
1200
1600
Rc(
m
)
Vbg
-VDirac
(V)
experiment
simulation
c
S1 S2 S3 S4 S5 S6 C1 C20
100
200
300
400
500
600
Rcp
Rcn
Rc
Rc(
m
)
Graphene Samples
d
At 300 K
Vbg
-Vdirac
=50 V
| www.editorialmanager.com/nare/default.asp
Nano Res.
Figure 3. Temperature dependence of graphene/metal contact resistance. (a) Typical back-gate transfer
characteristics of a 1-μm GFET under different temperatures. (b) Averaged μFE (red dots) and μFT (black
squares) versus T. (c) Rc as a function of back-gate voltage under different temperatures. (d) Rc from all sets
of GFETs under finite back-gate voltage bias versus temperature T.
-60 -40 -20 0 20 40 600
500
1000
1500
2000
2500 300 K
250 K
200 K
150 K
110 K
77.5 K
Rto
tal(
)
Vbg
(V)
L=1 m
Vds
=0.05V
a
-60 -40 -20 0 20 40 60-400
-200
0
200
400
600
800
1000
1200
C
300K
250K
200K
150K
110K
77.5K
Rc(
m
)
Vbg
-Vdirac
(V)
S1
50 100 150 200 250 300
4000
6000
8000
10000
12000
14000
FT
FE
cm
2V
-1s
-1)
T (K)
b
50 100 150 200 250 30060
80
100
120
140
160
180
200
S1 S2 S3
S4 S5 S6R
c(
m
)
T (K)
d
Vbg
=VDirac
-50V
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
Nano Res.
Figure 4. Retrieved materials and device parameters. (a) Average residual carrier density n0 and
field-effect mobility μFE of our GFETs at room temperature. (b) The temperature dependent hole
mean-free-path in a typical graphene device at Vbg=VDirac -50 V. (c) Mean-free-path dependent Tmg. (d) Rc
of GFETs in set S2 as a function of (Vbg-Vdirac). The red arrow points to the second peak in the red box of
the Rc-(Vbg-Vdirac) curve, which corresponds to the Dirac point of graphene under metal contact.
-60 -40 -20 0 20 40 600
200
400
600
800
1000
1200
S2
Rc(
m
)
Vbg
-VDirac
d
S1 S2 S3 S4 S5 S63,500
4,000
4,500
5,000
5,500
uF
E(c
m2V
-1s
-1)
Graphene samples
S1 S2 S3 S4 S5 S6
3.0
3.5
4.0
4.5
5.0
n0(1
01
1cm
-2)
a
0 50 100 150 2000.0
0.2
0.4
0.6
0.8
1.0
TMG
=0.81
RT
Tm
g
Lm(nm)
TMG
=sqrt(m+)
Xia's at 6 K
TMG
=0.87
77K
c
50 100 150 200 250 30080
90
100
110
120
130
T (K)
Lm (
nm
)
|Vbg
-VDirac
|=50 V
n3.7811012
cm-2
d
| www.editorialmanager.com/nare/default.asp
Nano Res.
Electronic Supplementary Material
Realization of low contact resistance close to
theoretical limit in graphene transistors
Hua Zhong, Zhiyong Zhang(), Bingyan Chen, Haitao Xu, Dangming Yu, Le Huang and Lian-Mao
Peng()
Supporting information to DOI 10.1007/s12274-****-****-* (automatically inserted by the publisher)
RAMAN SPECTRUM:
Figure S1. Raman spectrum obtained from a mechanically exfoliated graphene. The
intensity ratio of 2D peak to G peak is about 2 and no D peak is observed, implying that
the graphene sample is single layer and of high quality without defects. Laser with
wavelength of 633 nm is used.
1000 1500 2000 2500 30000
500
1000
1500
2000
Inte
nsit
y (
a.u
.)
Raman Shift (cm-1)
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
Nano Res.
ATOMIC FORCE MEASUREMENT:
LINEARITY DEMONSTRATION:
Figure S3. Due to variations from device to device, linear fit of different GFETs leads to
different intercepts. For graphene sample S1, 6 GFETs in total, there exist 57 possible
combinations, most of the intercepts (48/57) locate around 100±20 Ω. Only a few of them
are far away from the center, for just 2 or 3 GFETs are used for linear fit. The concentrated
distribution shows that GFETs in device group S1 are quite uniform.
0.1
1
10
40
70
95
99.520 40 60 80 100 120
20 40 60 80 100 1200
4
8
12
16
20
24
Co
un
ts
Rint
()
Cu
mu
lati
ve C
ou
nts
Figure S2. AFM image showing the graphene channel. Height of the graphene layer is
about 1 nm, confirmed the single layer property of our graphene samples. Though, AFM is
the last characterization step, the surface of the graphene channel is still quite smooth and
clean. Width of the graphene channel is about 2 μm.
| www.editorialmanager.com/nare/default.asp
Nano Res.
CHEMICAL-VAPOR-DEPOSITION GRAPHENE SAMPLES: graphene was grown on polycrystalline
platinum surface and then transferred into the SiO2/Si substrate using polymethyl methacrylate (PMMA)
600K as sacrificial layer.[1,2] GFETs were fabricated using the same process flow as described in the main text,
etched into 2-μm width strip first and then the deposition of metal contact (Pd/Au 30/20 nm). The transfer
characteristics of 5 GFETs were measured in vacuum at room temperature (see left of Figure S4). Quality of
CVD graphene is great, though not as well as mechanically exfoliated graphene, Dirac point voltages are
around 0.7±1.1 V, and field-effect mobility of holes is above 3,000 cm2V-1s-1. Contact resistance were
extracted by transfer length methods as well, as plotted in the right of Figure S4. The minimum Rc is 124.6±
37.9 Ω μm in the p branch and 238.2±27.1 Ω μm in the n branch.
WHOLE VIEW OF THE MEASURED GFETS AND CALIBRATION STRUCTURE:
Figure S5. SEM images of the measured GFETs and calibration structure. Left: A set of
GFETs with channel with of 2 μm. Right: Calibration structure as the same as measured
GFETs only with the 2-μm graphene channel replaced with 4-μm metal line. All those
metal layers were deposited at same time. Scale bar is 50 μm.
Figure S4. CVD graphene sample. Left: transfer characteristics of 5 GFETs based on CVD
graphene sample, source-drain bias voltage was 0.05 V. Right: The extracted Rc (black
squares) as a function of back-gate voltage. Red caps shown the standard errors.
-60 -40 -20 0 20 40 600
100
200
300
400
500
600
700
800
900
1000
1100
Rc(
m)
Vbg
-Vdirac
(V)-60 -40 -20 0 20 40 60
0.0
20.0µ
40.0µ
60.0µ
80.0µ
100.0µ
120.0µ
140.0µ
160.0µ
180.0µ
I ds(A
)
Vbg
(V)
1um
2um
3um
4um
5um
Vds
=0.05 V
www.theNanoResearch.com∣www.Springer.com/journal/12274 | Nano Research
Nano Res.
[1] Gao, L. B.; Ren, W. C.; Xu, H. L.; Jin, L.; Wang, Z. X.; Ma, T.; Ma, L. P.; Zhang, Z. Y.; Fu, Q.; Peng, L. M.; Bao, X. H.; Cheng, H.
M., Repeated growth and bubbling transfer of graphene with millimeter-size single-crystal grains using platinum. Nat. Commun. 2012, 3,
699.
[2] Xu, H. L.; Zhang, Z. Y.; Shi, R. B.; Liu, H. G.; Wang, Z. X.; Wang, S.; Peng, L. M., Batch-fabricated high-performance graphene
Hall elements. Sci. Rep. 2013, 3.
Address correspondence to Zhiyong Zhang, [email protected]; Lian-Mao Peng, [email protected]