LETTERS
Detection of individual gas molecules
adsorbed on graphene
F. SCHEDIN1, A. K. GEIM1, S. V. MOROZOV2, E. W. HILL1, P. BLAKE1, M. I. KATSNELSON3
AND K. S. NOVOSELOV1*
1Manchester Centre for Mesoscience and Nanotechnology, University of Manchester, Manchester, M13 9PL, UK
2
Institute for Microelectronics Technology, 142432 Chernogolovka, Russia
3
Institute for Molecules and Materials, University of Nijmegen, 6525 ED Nijmegen, Netherlands
*e-mail: Konstantin.Novoselov@manchester.ac.uk
Published online: 29 July 2007; doi:10.1038/nmat1967
The ultimate aim of any detection method is to achieve such
a level of sensitivity that individual quanta of a measured
entity can be resolved. In the case of chemical sensors, the
quantum is one atom or molecule. Such resolution has so far
been beyond the reach of any detection technique, including
solid-state gas sensors hailed for their exceptional sensitivity1–4.
The fundamental reason limiting the resolution of such sensors
is fluctuations due to thermal motion of charges and defects5,
which lead to intrinsic noise exceeding the sought-after signal
from individualmolecules, usually bymany orders ofmagnitude.
Here, we show thatmicrometre-size sensorsmade fromgraphene
are capable of detecting individual events when a gas molecule
attaches to or detaches from graphene’s surface. The adsorbed
molecules change the local carrier concentration in graphene one
by one electron, which leads to step-like changes in resistance.
The achieved sensitivity is due to the fact that graphene is an
exceptionally low-noise material electronically, which makes it
a promising candidate not only for chemical detectors but also
for other applications where local probes sensitive to external
charge, magnetic field or mechanical strain are required.
Solid-state gas sensors are renowned for their high sensitivity,
which—in combination with low production costs and miniature
sizes—have made them ubiquitous and widely used in many
applications1,2. Recently, a new generation of gas sensors has
been demonstrated using carbon nanotubes and semiconductor
nanowires (see, for example, refs 3,4). The high acclaim received
by the latter materials is, to a large extent, due to their exceptional
sensitivity allowing detection of toxic gases in concentrations
as small as 1 part per billion (p.p.b.). This and even higher
levels of sensitivity are sought for industrial, environmental and
military monitoring.
The operational principle of graphene devices described below
is based on changes in their electrical conductivity, σ, due to gas
molecules adsorbed on graphene’s surface and acting as donors
or acceptors, similar to other solid-state sensors1–4. However, the
following characteristics of graphene make it possible to increase
the sensitivity to its ultimate limit and detect individual dopants.
First, graphene is a strictly two-dimensional material and, as
such, has its whole volume exposed to surface adsorbates, which
maximizes their effect. Second, graphene is highly conductive,
exhibitingmetallic conductivity and, hence, low Johnson noise even
in the limit of no charge carriers6–9, where a few extra electrons
can cause notable relative changes in carrier concentration, n.
Third, graphene has few crystal defects6–10, which ensures a low
level of excess (1/f ) noise caused by their thermal switching5.
Fourth, graphene allows four-probe measurements on a single-
crystal device with electrical contacts that are ohmic and have
low resistance. All of these features contribute to make a unique
combination that maximizes the signal-to-noise ratio to a level
sufficient for detecting changes in a local concentration by less than
one electron charge, e, at room temperature.
The studied graphene devices were prepared by
micromechanical cleavage of graphite at the surface of oxidized
Si wafers7. This allowed us to obtain graphene monocrystals of
typically 10 µm in size. By using electron-beam lithography, we
made electrical (Au/Ti) contacts to graphene and then defined
multiterminal Hall bars by etching graphene in an oxygen
plasma. The microfabricated devices (Fig. 1a, upper inset) were
placed in a variable temperature insert inside a superconducting
magnet and characterized by using field-effect measurements
at temperatures, T , from 4 to 400K and in magnetic fields, B,
up to 12 T. This allowed us to find the mobility, µ, of charge
carriers (typically, ≈5,000 cm2 V−1 s−1) and distinguish between
single-, bi- and few-layer devices, in addition to complementary
measurements of their thickness carried out by optical and atomic
force microscopy6–9. Figure 1a, lower inset, shows an example
of the field-effect behaviour exhibited by our devices at room
temperature. This plot shows that longitudinal (ρxx) and Hall
(ρxy) resistivities are symmetric and antisymmetric functions
of gate voltage, Vg, respectively. ρxx exhibits a peak at zero Vg,
whereas ρxy simultaneously passes through zero, which shows that
the transition from electron to hole transport occurs at zero Vg
indicating that graphene is in its pristine undoped state6.
To assess the effect of gaseous chemicals on graphene devices,
the insert was evacuated and then connected to a relatively large
(5 l) glass volume containing a selected chemical strongly diluted
in pure helium or nitrogen at atmospheric pressure. Figure 1b
shows the response of zero-field resistivity, ρ= ρxx(B= 0)= 1/σ,
to NO2, NH3, H2O and CO in concentrations, C, of 1 part per
million (p.p.m.). Large easily detectable changes that occurred
within 1min and, for the case of NO2, practically immediately after
letting the chemicals in can be seen. The initial rapid response
was followed by a region of saturation, in which the resistivity
changed relatively slowly. We attribute this region to redistribution
652 nature materials VOL 6 SEPTEMBER 2007 www.nature.com/naturematerials
' 2007 Nature Publishing Group

LETTERS
0.1 1
C (p.p.m.)
∆
n

(
1
0
1
0

c
m
–
2
)
10
1
2
5
10
20
50
–20 0
V
g
(V)
0
2
4
20



a
n
d



x
y

(
k
Ω
)
ρ
ρ
ρ
ρ
ρ
ρ
~ ~
0 500 1,000
–4
–2
0
2
4
t (s)
∆



/



(
%
)
I II III IV
NH
3
CO
H
2
O
NO
2
a
b
xy
Figure 1 Sensitivity of graphene to chemical doping. a, Concentration, 1n, of
chemically induced charge carriers in single-layer graphene exposed to different
concentrations, C, of NO2. Upper inset: Scanning electron micrograph of this device
(in false colours matching those seen in visible optics). The scale of the micrograph
is given by the width of the Hall bar, which is 1 µm. Lower inset: Characterization of
the graphene device by using the electric-field effect. By applying positive (negative)
Vg between the Si wafer and graphene, we induced electrons (holes) in graphene in
concentrations n= αVg. The coefficient α≈ 7.2×1010 cm−2 V−1 was found from
Hall-effect measurements6–9. To measure Hall resistivity, ρxy , B= 1 T was applied
perpendicular to graphene’s surface. b, Changes in resistivity, ρ, at zero B caused
by graphene’s exposure to various gases diluted in concentration to 1 p.p.m. The
positive (negative) sign of changes is chosen here to indicate electron (hole) doping.
Region I: the device is in vacuum before its exposure; II: exposure to a 5 l volume of
a diluted chemical; III: evacuation of the experimental set-up; and IV: annealing at
150
◦C. The response time was limited by our gas-handling system and a
several-second delay in our lock-in-based measurements. Note that the annealing
caused an initial spike-like response in ρ, which lasted for a few minutes and was
generally irreproducible. For clarity, this transient region between III and IV
is omitted.
of adsorbed gas molecules between different surfaces in the insert.
After a near-equilibrium state was reached, we evacuated the
container again, which led only to small and slow changes in ρ
(region III in Fig. 1b), indicating that adsorbed molecules were
–40 –20 0 20 40
0
1
2
V
g
(V)
σ
(
k
Ω
–
1
)
Figure 2 Constant mobility of charge carriers in graphene with increasing
chemical doping. Doping increased from zero (black curve) to ∼1.5×1012 cm−2
(red curve) due to increasing exposure to NO2. Conductivity, σ , of single-layer
graphene away from the neutrality point changes approximately linearly with
increasing Vg and the steepness of the σ (Vg ) curves (away from the neutrality point)
characterizes the mobility, µ (refs 6–9). Doping with NO2 adds holes but also
induces charged impurities. The latter apparently do not affect the mobility of either
electrons or holes. The parallel shift implies a negligible scattering effect of the
charged impurities induced by chemical doping. The open symbols on the curves
indicate the same total concentration of holes, n
t
(∼2.7×1012 cm−2), as found
from Hall measurements. The practically constant σ for the same n
t
yields
that the absolute mobility, µ= σ/n
t
e, as well as the Hall mobility are
unaffected by chemical doping. For further analysis and discussions, see the
Supplementary Information.
strongly attached to the graphene devices at room temperature.
Nevertheless, we found that the initial undoped state could be
recovered by annealing at 150 ◦C in vacuum (region IV). Repetitive
exposure–annealing cycles showed no ‘poisoning’ effects of these
chemicals (that is, the devices could be annealed back to their initial
state). A short-time ultraviolet illumination offered an alternative
to thermal annealing.
To gain further information about the observed chemical
response, we simultaneously measured changes in ρxx and ρxy
caused by gas exposure, which allowed us to find directly (1)
concentrations, 1n, of chemically induced charge carriers, (2)
their sign and (3) mobilities. The Hall measurements revealed that
NO2, H2O and iodine acted as acceptors, whereas NH3, CO and
ethanol were donors. We also found that, under the same exposure
conditions, 1n depended linearly on the concentration, C, of an
examined chemical (see Fig. 1a). To achieve the linear conductance
response, we electrically biased our devices (by more than ±10V)
to higher-concentration regions, away from the neutrality point, so
that both σ=neµ andHall conductivity, σxy =1/ρxy =ne/B, were
proportional to n (see Fig. 1a, lower inset)6–9. The linear response
as a function of C should greatly simplify the use of graphene-based
sensors in practical terms.
Chemical doping also induced impurities in graphene in
concentrations Ni = 1n. However, despite these extra scatterers,
we found no notable changes in µ even for Ni exceeding
1012 cm−2. Figure 2 shows this unexpected observation by showing
the electric-field effect in a device repeatedly doped with
NO2. V-shaped σ(Vg) curves characteristic for graphene6–9
nature materials VOL 6 SEPTEMBER 2007 www.nature.com/naturematerials 653
' 2007 Nature Publishing Group