You are on page 1of 8

IOP PUBLISHING Nanotechnology 20 (2009) 245501 (8pp)

NANOTECHNOLOGY doi:10.1088/0957-4484/20/24/245501

Adsorption of ammonia on graphene


Hugo E Romero1 , Prasoon Joshi2 , Awnish K Gupta1 , Humberto R Gutierrez1, Milton W Cole1,3 , Srinivas A Tadigadapa2,3,4 and Peter C Eklund1,3,4
Department of Physics, Pennsylvania State University, University Park, PA 16802, USA Department of Electrical Engineering, Pennsylvania State University, University Park, PA 16802, USA 3 Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA
2 1

E-mail: sat10@psu.edu and pce3@psu.edu

Received 26 January 2009, in nal form 28 April 2009 Published 26 May 2009 Online at stacks.iop.org/Nano/20/245501 Abstract We report on experimental studies of NH3 adsorption/desorption on graphene surfaces. The study employs bottom-gated graphene eld effect transistors supported on Si/SiO2 substrates. Detection of NH3 occurs through the shift of the sourcedrain resistance maximum (Dirac peak) with the gate voltage. The observed shift of the Dirac peak toward negative gate voltages in response to NH3 exposure is consistent with a small charge transfer ( f 0.068 0.004 electrons per molecule at pristine sites) from NH3 to graphene. The desorption kinetics involves a very rapid loss of NH3 from the top surface and a much slower removal from the bottom surface at the interface with the SiO2 that we identify with a Fickian diffusion process. (Some gures in this article are in colour only in the electronic version)

1. Introduction
Graphene is a single at atomic sheet of carbon with the atoms arranged in a two-dimensional (2D) honeycomb conguration. Recent progress in isolating graphene on an insulating substrate (e.g., SiO2 or SiC) now enable this exotic 2D system to be probed experimentally [13]. It has been shown to be a promising building block for novel generation of high speed and sensitive electronic devices [412]. Electron transport experiments on graphene have demonstrated, among other effects, unusual carrier-density-dependent conductivity [1, 13, 14], anomalous quantum Hall effect [1315], minimum quantum conductivity [13], and exceptionally high electron mobilities [16, 17]. These remarkable electronic properties stem from the unique band structure of graphene, which exhibits conduction and valence bands with near-linear dispersion that touch at the Brillouin zone corners to make a zero gap semiconductor. Similar to earlier experiments on carbon nanotubes [18], the transport properties of graphene have been shown to be sensitive to molecules adsorbed on the surface (e.g. NH3 , NO2 , H2 O and CO) [10, 19]. The details of the strength and character of the adsorption (chemi versus physisorption), and the degree of charge transfer between the analyte and graphene is still
4 Authors to whom any correspondence should be addressed.

under debate. Geim and co-workers were the rst to report that a graphene Hall effect sensor device is capable of detecting individual molecules of NO2 [10]. Charge transfer between the graphene and NO2 is thought to be important in this particular case [19, 20]. Here, we report studies of the interaction of NH3 with graphene eld effect transistors (FETs) supported on Si/SiO2 substrates in order to provide further insight into the nature of the moleculegraphene interaction. The SiO2 is used as a gate dielectric and the heavily doped Si substrate as the bottom gate electrode. By sweeping the gate voltage, we can follow the time evolution of the peak in the drain-source resistance (known as the Dirac peak) to monitor the change of the Fermi level in graphene in response to the adsorption and desorption of NH3 . Presumably, this shift of the Dirac peak is dominated by charge transfer effects. The Dirac peak shift and the thermodynamic data for NH3 on graphite are used to determine the effective charge transfer per NH3 molecule ( f ) to the graphene. Our value for f will be compared to recent theoretical calculations for NH3 bound to the surface [20] and to the edges [21] of graphene.

2. Experimental details
The graphene FETs studied here were supported on Si/SiO2 substrates and bottom-gated using the SiO2 (300 nm thermal
1
2009 IOP Publishing Ltd Printed in the UK

0957-4484/09/245501+08$30.00

Nanotechnology 20 (2009) 245501

H E Romero et al

(a)

(b) b)
Graphene Cr/Au Cr/Au Cr/Au Cr/Au

II II
D SiO2 III III Si

I I
10 m 10m 10 m

G II II

Figure 1. (a) Schematic of a graphene device supported on SiO2 with underlying doped Si serving as the back gate (G); S and D refer to the source and drain contacts. (b) Optical micrograph of a graphene device fabricated with TEM grids as a shadow mask. The graphene ake is 3 m wide in this device. The following areas can be identied on the optical image: (I) single graphene layer (also indicated with dashed lines), (II) Cr/Au electrodes, and (III) SiO2 dielectric.

oxide) as the gate dielectric and the degenerately p+ doped cm) as the gate electrode. Si substrate ( 3 103 Our graphene lms were produced by mechanical exfoliation of highly oriented pyrolytic graphite (HOPG Grade ZYHTM , SPI Supplies, Inc., West Chester, PA) using adhesive tape (Scotch Tape, 3M, Inc., Maplewood, MN). Before the transfer of graphene akes, the Si/SiO2 substrates were rst cleaned in isopropanol and later exposed to O2 -plasma. The FET was constructed by the physical vapor deposition of pads of Cr(5 nm) followed by Au(100 nm) for electrical contact at opposite ends of the graphene FET channel (i.e., the source and drain). The Cr and Au were evaporated onto the graphene through square holes in a transmission electron microscopy (TEM) grid (SPI Supplies, Inc., West Chester, PA) used as a shadow mask and consisted of a 10 10 array of 115 115 m2 square holes separated by 10 m bars. By using a micromanipulator while viewing the area through an optical microscope, the TEM grid was carefully positioned on the graphene ake of interest (i.e., natural akes of approximately rectangular shape and more than 10 m long) so that one the TEM bars covers the graphene along its width, leaving its ends uncovered through the apertures of the mask. Then, the TEM grid is temporarily clamped to the substrate during contact evaporation. Our approach for fabricating electrical contacts avoids unnecessary complications that may stem from the exposure of the graphene to lithography chemicals. A schematic drawing of our bottom-gated eld FET that is supported on a Si/SiO2 substrate is shown in gure 1 together with an optical micrograph of the device. The optical image in gure 1 shows a particular FET with two square Cr/Au contacts deposited over each end of a rectangular graphene ake. The number of layers ( N ) in a graphene lm was determined by micro-Raman scattering using the shape of the 2D Raman band at 2700 cm1 and the intensity of the Raman-active G-band scattering at 1585 cm1 [22, 23]. The Raman characterization of N can be summarized as follows: [22] (1) the G-band intensity is approximately linear in N , (2) an N = 1 graphene lm can be easily recognized 2

by the narrow Lorentzian shape of the 2D Raman band, (3) at least one N = 1 lm must be present on the same substrate so that the relative G-band intensity associated with other lms can be measured by a simple translation of the microscope stage while all optics remain xed. Here, we present results on one representative sample that was demonstrated by Raman scattering to be a N = 1 graphene lm of dimensions 10 m 1.5 m. Each Si/SiO2 substrate supporting several graphene FETs was mounted on a ceramic chip carrier that was placed in a socket xed inside a vacuum sealed stainless steel (SS) tube equipped with vacuum and gas (NH3 ) connections. The SS tube was inserted into a tube furnace and evacuated to 5 107 Torr using a turbo-molecular pump. Vacuum-annealing of the samples were performed in situ by heating to 200 C at a constant rate of 2 C min1 to drive out all gases. Once the vacuum-annealing procedure was ceased and graphene reached a fully degassed state, the samples were allowed to cool to room temperature. We assume that graphene is a fully degassed state when there is no further evolution of the Dirac peak shape and/or position with time at 200 C. Samples could then be exposed to anhydrous NH3 (99.99%, H2 O <1 ppm) diluted in ultra high purity He (99.999%). A type-K thermocouple was mounted on the chip carrier near the Si substrate to monitor the local temperature. Electrical measurements were made using a programmable voltage source and digital voltmeter/ammeter (2400 General-Purpose SourceMeter, Keithley Instruments, Inc., Cleveland, OH) interfaced to a computer via LabVIEW (National Instruments Co., Austin, TX). The measurements consisted of the application of a constant sourcedrain voltage Vds = 1 mV to the device, while monitoring the resultant sourcedrain current ( Ids ) as a function of the applied gate bias (Vg ). The sourcedrain resistance was computed as Rds = Vds / Ids .

3. Results and discussion


Since our device fabrication scheme involves shadow masking to make contacts to graphene lms, it therefore has the

Nanotechnology 20 (2009) 245501

H E Romero et al

(a)

90

(b)
60

VDirac (V)

30

Rds (k)

Figure 2. Time evolution of (a) the Rds versus Vg curves at T = 200 C during the rst 8 h of vacuum-annealing of the graphene FET device shown schematically in gure 1. Initially, the Dirac peak is out of range ( Vg > 90 V); after 5 h, the Dirac peak appears at negative Vg and evolves slowly to the nal position at Vg 47 V. This particular sample was vacuum annealed for 72 h, when the position of VDirac stop evolving with time, i.e., sample reached a saturation; (b) the gate voltage for the charge neutrality point, VDirac , during vacuum-annealing of a graphene FET at T 200 C.

advantage that it is a lithography-free process. Recent studies have shown that resist, such as poly(methyl methacrylate) resin, can be quite difcult to remove, requiring heat treatment at 400 C in Ar/H2 ow [24]. These adsorbed molecules may dope (i.e., charge transfer with) the sample, or may block active adsorption sites for analytes and/or might otherwise interfere with the reproducibility of the FET electrical data. The electronic band structure of graphene near the Fermi level ( E F ) is somewhat unique. Near-mirror image valence and conduction bands touch each other at the six corners (K, K points) of a 2D Brillouin zone. These contact points are commonly called Dirac points [1]. This nomenclature stems from the linear dispersion of the electronic energy ( E ) with 2D wavevector (k ), i.e., E = h vF k , where vF is the Fermi velocity of the electrons. In pristine graphene, E F is located where the conduction and valence bands touch, i.e., the Dirac point. A FET built from pristine graphene (i.e., where E F is located near the Dirac point) should exhibit ambipolar conduction. That is, depending on the sign of the gate bias voltage (Vg ), the gate induces E F to rise or fall in the band structure and as a result either electrons or holes dominate the sourcedrain current ( Ids ). Dening the sourcedrain resistance as Rds = Vds / Ids , where Vds is the applied potential difference between source and drain, it can easily be shown that the position (VDirac ) of the peak in Rds is related to n , the 2D net carrier density [1]. Thus, the ambipolar character of graphene FETs refers to the fact that Vg can be used to control n (gate-induced doping). A transition between electron and hole conduction can be shown to occur when Vg positions E F at the Dirac point (n = 0), where the conduction and valence bands touch. Since this is the position in intrinsic graphene, this value of Vg also achieves charge neutrality in the graphene. An approximate relationship between n (in number of electrons cm2 ) and Vg can be used to describe 3

s)) t (hr
Vg (V)

-30 0 5 10 15 20 25 30 35

t (hours)

the device response to gate voltage [1],

n 0 (Vg VDirac )/le (Vg VDirac )

(1)

where 0 and l are, respectively, the gate dielectric permittivity and thickness, e is the charge of the electron and VDirac is a constant. Equation (1) is an approximate relationship, in that thermal broadening of the FermiDirac function is not included. Using the well known linear electronic -band dispersion, it is easy to show that the carrier concentration (electrons or 2 holes) E F [25]. The accepted value for the Fermi velocity is vF c/300, where c is the speed of light [13, 14]. Using equation (1), and the values l = 300 nm, and /0 3.9, we can therefore obtain an approximate numerical relationship between E F and VDirac for our graphene FETs, i.e., E F 31.7(meV/ V) VDirac (2) where the positive and negative signs apply, respectively, to hole and electron conduction. This equation, however, ignores the details of the electrostatics in the FET and the SiO2 /graphene interface. When these details are included, the left (or right?) side of equation (1) contains an additional small term that is proportional to n [25]. Our graphene FETs were exposed to ambient conditions for several days before electrical characterization. We found that most of our devices exhibited a Dirac peak in the range +50 V < Vg < +100 V, i.e., they are p-type. A positive shift of the Dirac peak of the same order of magnitude is not uncommon in graphene devices on Si/SiO2 substrates [1, 26, 27]. However, all our samples p-doping can be reversed by vacuum-annealing at 200 C and 5 107 Torr over 10200 h. Over this time, the peak passes through Vg = 0 V and ends up in the range 10 V < Vg < 60 V, depending on the sample (cf gure 2). We have interpreted

Nanotechnology 20 (2009) 245501

H E Romero et al

t (min)
0 -40 10 20 30 40

(a)
-50

(b)

(2) (1)

Rds (k)

VDirac (V)

-60 -70 -80 -90

Vg (V )

n) t (mi

t (hours)

Figure 3. (a) Response of bottom-gated graphene FET (on Si/SiO2 ) to a sudden overpressure of 1 atm of 10% NH3 in He during the rst exposure. The initial Dirac peak corresponds to the FET before NH3 exposure. Data were collected in the range 90 V < Vg < +90 V to avoid breakdown of the gate dielectric. (b) Time dependence of the Dirac voltage VDirac , during NH3 desorption at T = 25 C, after two successive exposuredesorption cycles (1) and (2). The solid lines are t of the data to a double exponential function.

these results [25], to indicate that the p-doping occurs from ambient exposure to air and stems from the adsorption of O2 molecules that remove electrons from the graphene. A clean graphene sample supported on Si/SiO2 should be ntype [25]. This conclusion stems from work function (W ) calculations [25] for graphene and amorphous Si where we found that W (SiO2 ) is 1 eV higher than in graphene, as E F in SiO2 is pinned in a surface state band located just below the conduction band edge. In the present studies of the response of graphene FETs to NH3 , we start with the graphene FET in a fully degassed state, i.e., it is initially n-type. We now discuss the time response of our n-type FETs to a sudden rise in NH3 pressure. We present all data from one device, but the same general behavior was observed in six devices. As described above, the typical device chosen for analysis here was rst degassed in high vacuum at 200 C for 72 h until the Dirac peak remained xed at a negative gate bias Vg 47 V. The particular device for systematic study was then cooled in vacuum to T = 25 C and exposed for 30 min to 1 atm overpressure of 10% NH3 gas diluted in He. The time evolution of Dirac peak to this rst NH3 exposure is shown in gure 3(a). The Dirac peak can be seen to shift rapidly to more negative Vg , eventually moving beyond the safe gate voltage limit for our FETs, i.e., Vg 90 V. The shift of the Dirac peak to more negative Vg shows clearly that NH3 appears to be acting as an n-type dopant, upshifting the position of E F in the electron pockets at the Brillouin zone corners. All six graphene FETs showed this rapid shift of the Dirac peak to negative Vg beyond our safe gate limit. Indeed, the electrical response to NH3 at T = 25 C is so rapid that we propose that it is determined by either adsorption onto the top surface of the graphene, or by top surface chemical reactions whereby the NH3 displaces strongly bound molecules remaining after degassing. We expect therefore that the diffusion of NH3 into the SiO2 /graphene interface is a much slower process. Next, the NH3 -dosed FET was 4

exposed to vacuum desorption conditions, i.e., the NH3 supply is isolated and the sample chamber was rapidly evacuated in 10 s, i.e., the NH3 overpressure therefore drops very rapidly compared to the characteristic time for the ensuing changes in the FET resistance. The kinetics for the rst desorption cycle at T = 25 C can only be followed once the Dirac peak recovers below our gate voltage limit of Vg = 90 V. As can be seen in the gure, the desorption kinetics at T = 25 C are considerably slower than estimated for the adsorption response (gure 3(b), curve 1); we observed that this particular device recovers during desorption from being very strongly n-type (i.e., when the Dirac peak appears at gate voltages beyond 90 V) to exhibit a new stationary value VDirac 52 V in 5 h which is somewhat more negative than the initial vacuumdegassed FET state (VDirac = 47 V). Each N = 1 device behaves a little differently in the details of the kinetics and nal equilibrium values of the Dirac peak. The NH3 response of this particular device is typical of the general character of all devices measured. Next, the FET was further degassed in vacuum at T = 100 C until Rds (Vg ) no longer changed with time, after which it was cooled down to room temperature, and exposed to the same 10% NH3 /He mixture at p = 1 atm a second time. The Dirac peak was then observed to rapidly shift again toward more negative Vg , disappearing behind our (device-safe) gate dielectric limit of 90 V. As can be seen in gure 3(b), the device response to a second desorption (degas) cycle (curve 2) at T = 25 C is similar to the rst desorption (degas) cycle (curve 1). Curve 2 shows the Dirac peak rapidly upshifting from below 90 V to a new steady state value of 38 V at T = 25 C in vacuum which is less negative than was observed after the rst cycle. The solid lines in gure 3(b) represent ts to the desorption data by a double exponential function,

VDirac (t ) = V + V1 exp(t /1 ) + V2 exp(t /2 )

(3)

where t is the time and 1 and 2 are effective time constants. The constant V in equation (3) is the steady state position

Nanotechnology 20 (2009) 245501

H E Romero et al

40 30 20

(3)

(2)
Rds (k)

40 30 20

40 30 20 10

(1)

10

-80

-40

0 Vg (V)

40

80

Figure 4. Room temperature sourcedrain resistance as a function of gate voltage for three successive NH3 exposuredesorption cycles (1)(3). Three representative Rds (Vg ) curves are plotted: before NH3 exposure (solid lines), after NH3 desorption at T = 25 C (dotted lines), and after vacuum-degassing at T = 100 C (dashed lines). The arrows indicate the position of the Dirac peaks.

of the Dirac peak after long-term desorption. As can be seen from the gure, equation (3) provides a good description of the experimental observation. Of course, for the double exponential t to be meaningful, we expect to obtain two very different values for 1 and 2 . Indeed, the slow and fast components of the NH3 desorption process are (rst desorption) 1 = 2.1 0.2 h and 2 = 0.22 0.02 h (i.e., 1 /2 = 9 2); (second desorption) 1 = 30 1 min and 2 = 4 1 min (i.e., 1 /2 = 8 3). The values returned for the prefactors are, respectively, (rst desorption) V1 = 18.3 0.4 V and V2 = 12.1 0.7 V, (second desorption) V1 = 30.2 0.8 V and V2 = 13.5 0.5 V. Our results indicate that the FET desorbs somewhat faster the second time. We are not sure what the physical or chemical implications are for this change in response times. The fact that the ratio 1 /2 9 is found to be the same for both desorption cycles is discussed below and appears to stem from Fickian diffusion. All samples showed the general behavior exhibited by VDirac (t ) with the kinetics described by a combination of fast and slow processes (equation (3)). In one other sample, we did a tting of equation (3) to the data and again found the ratio 1 /2 = 9.5 0.6. In gure 4, we show the representative Rds (Vg ) curves corresponding to the rst (bottom, curves 1), second (middle, curves 2), and third (top, curves 3) NH3 exposuredesorption cycles. The Dirac peaks are coded in the gure to indicate when they were collected: before NH3 exposure (solid curve), after NH3 desorption at T = 25 C (dotted curve) and after NH3 desorption at T = 100 C for 16 h in vacuum (dashed curve). As shown in gure 4, the amplitude and 5

position of the Dirac peak were not observed to recover exactly to their original vacuum-degassed FET state values after a T = 25 C vacuum desorption. However, after a T = 100 C vacuum desorption, the amplitude, but not the position, of the Dirac peak recovers to the original (degassed state) value; the Dirac peak shift appears at a slightly less negative Vg value after this treatment. We attribute this result to the possibility of displacement chemical reactions in the SiO2 /graphene interface whereby the NH3 molecules displace some species that weakly charge transfer, thereby partially neutralizing the ambient p-doped SiO2 surface. Despite much effort by several groups, the interaction of NH3 with SWNTs remains controversial. A few years ago, it was reported that individual semiconducting nanotubes (sSWNTs) could be used to detect small concentrations of NH3 with high sensitivity by measuring changes of the nanotube conductance and the FET threshold voltage [18]. The authors reported that the electrical conductance of a p-doped nanotubeFET decreased by approximately two orders of magnitude after exposure to a ow of 1% NH3 at room temperature, and they suggested that a charge transfer of electrons from NH3 to SWNTs was responsible for the change in properties of the SWNTs. On the other hand, theoretical calculations on isolated defect-free nanotubes predicted that NH3 molecules should physisorb to the tube wall without signicant change of the band structure, although a very small charge transfer of f 0.03 0.04e per adsorbed NH3 molecule was predicted to occur [18, 2831]. Therefore, from this work on SWNTs it appears that experimental and theoretical results are consistent with the NH3 acting as a weak n-type dopant, overriding or compensating the initial p-type behavior deduced from the sign of the gate voltage at threshold. Theoretical studies of gas adsorption on a pristine graphene surface have also been reported recently by that indicate that NH3 molecules can act as donors with a small charge transfer of f 0.03e [20]. This is consistent with calculations on defect-free SWNTs [18, 2831]. More recent theoretical investigations have explored the adsorption of NH3 on nanotubes near wall defects, e.g., Stone-Wales defects [32]. The results are possibly relevant to our study in graphene and can be summarized as follows: (i) chemisorption of NH3 at a wall defect is an exothermic process with a relatively low activation barrier; therefore chemisorption of NH3 near these defects should readily happen at room temperature; (ii) NH3 adsorption at defects is accompanied by a high charge transfer, i.e., f = 0.18e; (iii) the presence of pre-dissociated oxygen near or at the defect enhances the NH3 chemisorption process which becomes nearly spontaneous. These signicant electron transfers from chemisorbed NH3 at defects, rather than low charge transfer at more numerous pristine sites might also explain the large drops in electrical conductance of s-SWNTs observed experimentally [18, 33]. It is also relevant to our results here that recent temperature-programed-desorption (TPD) measurements of NH3 on SWNTs indicate that NH3 can be strongly bound to the SWNT surface (i.e., chemisorbed) [34]. Perhaps this conclusion is consistent with a weak charge transfer that enhances the binding energy of the NH3 to the tube

Nanotechnology 20 (2009) 245501

H E Romero et al

wall, i.e., over and above that associated with physisorption. In these TPD experiments, most of the NH3 gas desorbed only at elevated temperatures (500700 K). In addition, and in contrast to the previous work on NH3 -SWNTs, Bradley et al [35] found that vacuum-degassed nanotube-FETs were insensitive to NH3 . They suggested that for NH3 gas to be detected in the FET response, the NH3 must rst dissolve in a H2 O monolayer which forms on the nanotube-FETs under ambient lab conditions. They proposed that the NH3 molecule in this environment can become a cation with the charge being compensated in the SWNTs, i.e., act as an ndopant. As a result, it was proposed that the nanotube-FETs respond to NH3 gas only when they are exposed in ambient (humid) conditions. Feng et al [32], on the other hand, reported experimental evidence of the extreme sensitivity of NH3 adsorption on thermal-annealing prehistory. Their results revealed a systematic decrease of NH3 adsorption on SWNTs after they were annealed to successively higher temperatures from 500 up to 1400 K. However, they did not identify the decrease in NH3 doping on a reduction in H2 O bound to the surface, but rather to large charge transfer associated with less numerous sites near functional groups (e.g., chemisorbed O2 ) and defects on the surface that can be removed by vacuumannealing. Upon exposure of our degassed graphene FET to 0.1 atm of NH3 , the Dirac peak rapidly shifted past our safe gate voltage limit and we were therefore only able to partially observe the Rds (Vg ) peak, i.e., the peak itself and its negative voltage tail were beyond the gate limit. Since we did not observe any signicant change in the shape of the Rds (Vg ) curves as they shift with adsorption/desorption, we then t this partial Rds (Vg ) curve to obtain the peak position, i.e., VDirac 140 V. Before NH3 exposure, the degassed device exhibited a Dirac peak at 47 V; we therefore observed a negative shift in the Dirac peak of VDirac 90 V due to equilibrium adsorption of NH3 under our experimental conditions. The same basic behavior was observed in ve other graphene FETs. The shift VDirac was detectable in three other samples where the tail of the Dirac peak was still visible within our gate voltage limit, as described above. The estimated values of VDirac determined as described above by a rigid shift of a typical Dirac peak, are 100 V, 89 V, and 90 V, respectively. In average, VDirac = 92 5 V. We can interpret the 92 V Dirac peak shift due to equilibrium coverage of NH3 under our experimental conditions in terms of the charge transfer of electrons between adsorbed NH3 and the graphene. For simplicity, we can begin to analyze this process by assuming a monolayer coverage of NH3 in a closed-packed hexagonal conguration [36], and for the use the collision diameter for NH3 , i.e., d 4.09 A effective molecular diameter [37]. The effective graphene area per NH3 molecule is therefore Am = 3( 3 + 1)2 d 2 /14. The relationship between the charge transfer per molecule f , Am and the excess charge per unit area on the graphene ( n ) is given by f = n Am , which by equation (1) can be expressed as f = VDirac Am . Using this equation, we can estimate f 0.01e for a monolayer coverage of NH3 on graphene. This value of f is approximately a factor of 3 lower than that predicted by theory [20]. 6

However, based on thermodynamic data for NH3 on graphite [38, 39], a surface coverage of much less than a monolayer is expected. Our estimate for f must be rened to take this lower coverage into account. The well known thermodynamic relationship between surface coverage and binding energy E b is given by [40]

surface = gas exp

Eb kB T

(4)

where surface and gas are, respectively, the three-dimensional (3D) number density on the surface and in the gas phase. This relationship is derived assuming that only one molecule can occupy a site with this binding energy and there is no interaction among the adsorbed molecules. The fractional coverage is dened as = Am dsurface . Using the experimental thermodynamic value E b = 0.18 eV [38, 39], we nd = 0.15 for p = 0.1 atm (partial pressure of NH3 ) and T = 25 C which is indeed much less than a monolayer. We then correct our estimation of the charge transfer f taking into account the fractional coverage = 0.15. With this coverage we obtain f 0.068 0.004 electrons per NH3 molecule. This value is in excellent agreement with the recent theoretical study of the physisorption of NH3 on defect-free graphene discussed above [20]. It is important to notice that since our graphene was fully degassed in vacuum at 200 C for long periods of time, we do not expect any signicant interference from H2 O and/or O2 on the surface, as discussed above in relationship to Bradley et als work with NH3 on s-SWNT FETs [35]. To make sure the NH3 gas (stated to be <1 ppm H2 O) did not bring signicant H2 O into our experimental chamber, we carried out one additional experiment rst introducing a KOH trap between the gas cylinder and the chamber. Essentially the same experimental results were obtained. Similar to SWNTs, recent theoretical calculations [21] showed that NH3 can strongly adsorb on defect sites located at a graphenes edge and signicantly inuence the electronic and transport properties of narrow ribbons. The charge transfer at these edge sites is high, f = 0.27e [21]. Consequently, it is important to consider the possibility that defect sites might affect NH3 detection by graphene FETs. Using equation (1), we can write f = VDirac w/2, where w = 1.5 m is 1 the average width of our graphene ake and = 0.47 A is the defect concentration on a graphene ake having two free armchair-shaped edges. Assuming full decoration of these sites with NH3 , we estimate that a charge transfer of f = 0.27e corresponds to a Dirac peak shift of VDirac 2 V. Although this value is easily measurable; it is too small to explain the 90 V shift we observe due to NH3 exposure. That large a shift must come from basal plane adsorption. We now discuss the adsorption/desorption kinetics of NH3 on graphene as observed from our experiments. To explain our results, we assume that for a particular hexagon or adsorption site of the graphene, the molecule can be adsorbed either on top (free surface) or on the bottom (next to the SiO2 substrate) of the graphene. It is not likely, however, that both top and bottom sites associated with the same hexagon can be simultaneously occupied, i.e., the second adsorption would be expected to have

Nanotechnology 20 (2009) 245501

H E Romero et al

a much lower binding energy. However, this assumption is not critical to our arguments about the kinetics. Since both sides of the graphene sheet are not equally accessible to the NH3 molecules, we identify two distinct kinetic processes that control the overall adsorption/desorption: (i) rapid adsorption/desorption involving the top surface of the graphene and (ii) slow adsorption/desorption involving the bottom surface of the graphene. In (ii), these processes are kinetically hindered by a slow diffusion into and out of 2D slit pores dened at the interface between the SiO2 and the graphene whose dimensions correlate with the surface roughness of the SiO2 and the ability of the graphene lm to follow the SiO2 texture [24, 41, 42]. The thickness of the SiO2 /graphene interface in graphene devices [24]. pores has been estimated to be 9 A We expect that molecules will be more readily adsorbed on the top surface of the graphene and VDirac will therefore change rapidly due to charge transfer between the graphene and these molecules. Bottom sites could be lled more slowly by diffusion of NH3 into the interface and eventually the fractional occupation of top and bottom sites on the graphene will be approximately equal if the nearby SiO2 surface does not significantly inuence the binding energy on the bottom sites. In other words, the adsorption-induced kinetics of the Dirac peak is dominated by the rapid top surface adsorption leading to fractional coverage top . At much longer times (equilibrium), we expect the top and bottom coverage to be almost equal, i.e., top = bottom. These two fractional coverages sum up to yield the thermodynamic coverage associated with the particular site binding energy, i.e., = top + bottom. Unfortunately, the top surface adsorption/desorption rates are so fast that we cannot measure them accurately; the effective time constants are much less than 1 s and data collection for a Rds (Vg ) curve takes 15 s. However, on desorption, the NH3 molecules bound to the bottom must diffuse out of the interface and this process is slow enough to be measured. Let us consider desorption from the NH3 -saturated surface at p = 1 atm and T = 25 C. As discussed above, our initial condition for the fractional coverage is top = bottom and = top + bottom. Removing the 1 atm chamber overpressure of 10% NH3 /He mixture, requires that top goes to zero rapidly. Consequently, a positive shift in the Dirac peak of VDirac 45 V was observed, i.e., the device immediately exhibited a Dirac peak near 90 V. We then measured the slower time evolution of bottom dictated by diffusion and starting from the initial coverage of /2, as shown in gure 3(b). This diffusion problem is of general importance to the kinetics of gas storage and has been solved analytically [43]. We simplify the analysis by working in one dimension (1D). This approximation is motivated by the fact that the width of our samples is much narrower than the length. The important diffusion is therefore primarily perpendicular to the lateral edges of the graphene akes. Ficks second law in 1D is then used to solve for the time evolution of the molecular concentration [43]:

effective molecular diffusion coefcient. Initially, the NH3 concentration C0 within the pore is uniform. However, the concentration of NH3 at the edges of the graphene drops to zero at the edge as the NH3 evaporates into vacuum. Mathematically, we therefore have to solve equation (5) subject to the boundary conditions C (x , 0) = C0 and C ( L /2, t ) = 0. The solution has the analytical form [43]:

M (t ) =

L /2 + L /2

C (x , t ) dx = M +

8 LC0 2 (6)

exp(t / ) +

1 exp (9t / ) 9

where = L 2 / 2 D and M is the number of gas molecules remaining in the SiO2 /graphene interface (slit pores) at innite time. Note that desorption kinetics is described by a double exponential function with a fast and a slow term (decreasing higher order terms are ignored). The value of depends on a microscopic model for D . Nevertheless, the form of this solution to Ficks equation is interesting in our case because it supports the experimental observation of 1 /2 9. We take this agreement as indirect evidence for the association of the slow desorption with rate limiting diffusion of NH3 from underneath the graphene ake.

4. Summary and conclusions


We have attempted to elucidate the mechanism of the adsorption/desorption of NH3 on graphene. We conclude that because of exposure to air, graphene FETs initially behave as a p-type system. Consistent with recent work function calculations for graphene and SiO2 [25], these FETs can be rendered n-type by a long-term vacuum-degassing at 200 C. With subsequent in situ exposure to NH3 , these degassed graphene FETs exhibit a large (90 V) shift in the Dirac peak shift in the position of the Dirac peak that we have identied with the small transfer of additional electrons into the graphene. The molecules adsorb and desorb rapidly on the outer surface of the graphene, and much more slowly from the inner surface that is in contact with the SiO2 gate dielectric. Using a 15% sub-monolayer coverage predicted by thermodynamic data, and a 92 V shift in the Dirac peak, we estimate that a charge transfer f 0.07 electrons per NH3 to the graphene occurs upon adsorption, in good agreement with theoretical estimates for binding at pristine sites. The total number of defect sites at the graphenes edges is too low to account for the observed Dirac peak shifts. The slow NH3 desorption rate from supported graphene FETs is consistent with the diffusion of molecules in the SiO2 /graphene interface and supports the experimental observation of a double exponential desorption with 1 /2 9. Desorption in the second and subsequent cycles is a faster by a factor of 1000, but still preserves the ratio 1 /2 9. The increase in desorption rate is not fully understood.

2 C (x , t ) = D 2 C (x , t ) t x

(5)

Acknowledgment
This work was supported by the NSF NIRT Grant No. ECS0609243. 7

where x is the spatial variable running from x = 0 (graphene ribbon midpoint) to x = L /2 (ribbon edge) and D is the

Nanotechnology 20 (2009) 245501

H E Romero et al

References
[1] Novoselov K S, Geim A K, Morozov S V, Jiang D, Zhang Y, Dubonos S V, Grigorieva I V and Firsov A A 2004 Electric eld effect in atomically thin carbon lms Science 306 6669 [2] Novoselov K S, Jiang D, Schedin F, Booth T J, Khotkevich V V, Morozov S V and Geim A K 2005 Two-dimensional atomic crystals Proc. Natl Acad. Sci. USA 102 104513 [3] Berger C et al 2004 Ultrathin epitaxial graphite: two-dimensional electron gas properties and a route toward graphene-based nanoelectronics J. Phys. Chem. B 108 199126 [4] Bunch J S, van der Zande A M, Verbridge S S, Frank I W, Tanenbaum D M, Parpia J M, Craighead H G and McEuen P L 2007 Electromechanical resonators from graphene sheets Science 315 4903 [5] Lemme M C, Echtermeyer T J, Baus M and Kurz H 2007 A graphene eld-effect device IEEE Electron Device Lett. 28 2824 [6] Williams J R, Dicarlo L and Marcus C M 2007 Quantum hall effect in a gate-controlled pn junction of graphene Science 317 63841 [7] Standley B, Bao W, Zhang H, Bruck J, Lau C N and Bockrath M 2008 Graphene-based atomic-scale switches Nano Lett. 8 33459 [8] Stoller M D, Park S, Zhu Y, An J and Ruoff R S 2008 Graphene-based ultracapacitors Nano Lett. 8 3498502 [9] Stampfer C, Schurtenberger E, Molitor F, Guttinger J, Ihn T and Ensslin K 2008 Tunable graphene single electron transistor Nano Lett. 8 237883 [10] Schedin F, Geim A K, Morozov S V, Hill E W, Blake P, Katsnelson M I and Novoselov K S 2007 Detection of individual gas molecules adsorbed on graphene Nat. Mater. 6 6525 [11] Wang X, Zhi L and Muellen K 2008 Transparent conductive graphene electrodes for dye-sensitized solar cells Nano Lett. 8 3237 [12] Ponomarenko L A, Schedin F, Katsnelson M I, Yang R, Hill E W, Novoselov K S and Geim A K 2008 Chaotic Dirac billiard in graphene quantum dots Science 320 3568 [13] Novoselov K S, Geim A K, Morozov S V, Jiang D, Katsnelson M I, Grigorieva I V, Dubonos S V and Firsov A A 2005 Two-dimensional gas of massless Dirac fermions in graphene Nature 438 197200 [14] Zhang Y B, Tan Y W, Stormer H L and Kim P 2005 Experimental observation of the quantum Hall effect and Berrys phase in graphene Nature 438 2014 [15] Novoselov K S, Jiang Z, Zhang Y, Morozov S V, Stormer H L, Zeitler U, Maan J C, Boebinger G S, Kim P and Geim A K 2007 Room-temperature quantum Hall effect in graphene Science 315 1379 [16] Bolotin K I, Sikes K J, Jiang Z, Klima M, Fudenberg G, Hone J, Kim P and Stormer H L 2008 Ultrahigh electron mobility in suspended graphene Solid State Commun. 146 3515 [17] Du X, Skachko I, Barker A and Andrei E Y 2008 Approaching ballistic transport in suspended graphene Nat. Nanotechnol. 3 4915 [18] Kong J, Franklin N R, Zhou C, Chapline M G, Peng S, Cho K and Dai H 2000 Nanotube molecular wires as chemical senbelsors Science 287 6225 [19] Wehling T O, Novoselov K S, Morozov S V, Vdovin E E, Katsnelson M I, Geim A K and Lichtenstein A I 2008 Molecular doping of graphene Nano Lett. 8 1737 [20] Leenaerts O, Partoens B and Peeters F M 2008 Adsorption of H2 O, NH3 , CO, NO2 , and NO on graphene: a rst-principles study Phys. Rev. B 77 125416 [21] Huang B, Li Z, Liu Z, Zhou G, Hao S, Wu J, Gu B-L and Duan W 2008 Adsorption of gas molecules on graphene nanoribbons and its implication for nano-scale molecule sensor J. Phys. Chem. C 112 134426

[22] Gupta A, Chen G, Joshi P, Tadigadapa S and Eklund P C 2006 Raman scattering from high-frequency phonons in supported N -graphene layer lms Nano Lett. 6 266773 [23] Ferrari A C et al 2006 Raman spectrum of graphene and graphene layers Phys. Rev. Lett. 97 187401 [24] Ishigami M, Chen J H, Cullen W G, Fuhrer M S and Williams E D 2007 Atomic structure of graphene on SiO2 Nano Lett. 7 16438 [25] Romero H E, Sheng N, Joshi P, Gutierrez H R, Tadigadapa S A, Sofo J O and Eklund P C 2008 n-Type behavior of graphene supported on Si/SiO2 substrates ACS Nano 2 203744 [26] Tan Y W, Zhang Y, Bolotin K, Zhao Y, Adam S, Hwang E H, Das Sarma S, Stormer H L and Kim P 2007 Measurement of scattering rate and minimum conductivity in graphene Phys. Rev. Lett. 99 246803 [27] Das A et al 2008 Monitoring dopants by raman scattering in an electrochemically top-gated graphene transistor Nat. Nanotechnol. 3 21015 [28] Chang H, Lee J D, Lee S M and Lee Y H 2001 Adsorption of NH3 and NO2 molecules on carbon nanotubes Appl. Phys. Lett. 79 386365 [29] Zhao J, Buldum A, Han J and Lu J P 2002 Gas molecule adsorption in carbon nanotubes and nanotube bundles Nanotechnology 13 195200 [30] Bauschlicher C W and Ricca A 2004 Binding of NH3 to graphite and to a (9,0) carbon nanotube Phys. Rev. B 70 115409 [31] Lu J, Nagase S, Maeda Y, Wakahara T, Nakahodo T, Akasaka T, Yu D, Gao Z, Han R and Ye H 2005 Adsorption conguration of NH3 on single-wall carbon nanotubes Chem. Phys. Lett. 405 902 [32] Andzelm J, Govind N and Maiti A 2006 Nanotube-based gas sensors-role of structural defects Chem. Phys. Lett. 421 5862 [33] Robinson J A, Snow E S, Badescu S C, Reinecke T L and Perkins F K 2006 Role of defects in single-walled carbon nanotube chemical sensors Nano Lett. 6 174751 [34] Ellison M D, Crotty M J, Koh D, Spray R L and Tate K E 2004 Adsorption of NH3 and NO2 on single-walled carbon nanotubes J. Phys. Chem. B 108 79383 [35] Bradley K, Gabriel J-C P, Briman M, Star A and Gruner G 2003 Charge transfer from ammonia physisorbed on nanotubes Phys. Rev. Lett. 91 218301 [36] Rowntree P, Scoles G and Xu J 1990 The structure of ammonia overlayers ohysisorbed onto the surface of single crystal graphite, determined by means of atomic beam diffraction J. Chem. Phys. 92 38537 [37] Halpern A M and Glendening E D 1996 Estimating molecular collision diameters using computational methods THEOCHEM 365 912 [38] Avgul N N and Kiselev A V 1970 Physical adsorption of gases and vapors on graphitized carbon blacks Chem. Phys. Carbon 6 1124 [39] Lakhli A and Killingbeck J P 2005 Dynamic and spectroscopic studies of single molecules physisorbed on graphite substrates. 2. Application to the ammonia molecule J. Phys. Chem. B 109 1132231 [40] Bruch L W, Cole M W and Zaremba E 1997 Physical Adsorption: Forces and Phenomena (Mineola, NY: Dover) p 340 [41] Stolyarova E, Rim K T, Ryu S, Maultzsch J, Kim P, Brus L E, Heinz T F, Hybertsen M S and Flynn G W 2007 High-resolution scanning tunneling microscopy imaging of mesoscopic graphene sheets on an insulating surface Proc. Natl Acad. Sci. USA 104 920912 [42] Stoeberl U, Wurstbauer U, Wegscheider W, Weiss D and Eroms J 2008 Morphology and exibility of graphene and few-layer graphene on various substrates Appl. Phys. Lett. 93 051906 [43] Crank J 1975 The Mathematics of Diffusion 2nd edn (Oxford: Clarendon) p 414

You might also like