Adsorption/desorption and electrically controlled flipping of ammonia molecules on graphene

In this paper, we evaluate the adsorption/desorption of ammonia molecules on a graphene surface by studying the Fermi level shift. On the basis of a physically plausible model, the adsorption and desorption rates of ammonia molecules on graphene have been extracted from the measured Fermi level shift as a function of exposure time. An electric-field-induced flipping behavior of the ammonia molecules on graphene is suggested based on field effect transistor (FET) measurements.


Introduction
Electrical transport experiments on graphene have demonstrated that graphene has carrierdensity-dependent conductivity [1], the quantum Hall effect [2], minimum quantum conductivity [3] and high carrier mobility [4]. Because of these characteristics, graphene is considered to be a promising new material for memory, logic, analogue, opto-electronic and sensor devices and potentially many more applications [5]- [11].
Controlling the intrinsic electrical property and being able to locally change the carrier density are important for graphene devices. It has been shown that graphene is sensitive to molecular adsorbates (e.g. NH 3 , H 2 O, NO 2 and CO) [12]. The Dirac cone band structure of graphene allows the control of both the carrier type and the carrier concentration induced by adsorbates owing to charge transfer from adsorbed molecules to graphene. A graphene Hall effect device is capable of sensing individual molecules of NO 2 [12]. However, the details of the strength of the adsorption and the degree of charge transfer for different adsorbates are still debated [13,14]. In this paper, we report an experimental study of the adsorption/desorption and likely 'flipping' of ammonia molecules on synthetic, large-area graphene [15] by detecting the Fermi level shift of a graphene field effect transistor (FET).

Experiment
Large-area graphene films grown by chemical vapor deposition (CVD) on Cu foils of 25 µm thickness (Alfa Aesar, item no. 13382) [15] were used to study the adsorption/desorption of NH 3 molecules. The surface of the graphene-on-Cu was first coated with poly-methyl methacrylate (PMMA). After the Cu substrate was dissolved in a Fe(NO 3 ) 3 solution (1 M l −1 ), PMMA-graphene was removed from the solution and transferred to a SiO 2 /Si substrate (p + doped, ρ ∼ 0.002-0.005 cm; Addison Engineering) [11]. Finally, PMMA was removed by rinsing in acetone at room temperature. Graphene FET devices were constructed by physical vapor deposition of Au films (∼ 500 nm), used as source and drain electrodes, on both sides of the graphene film. Figure 1(a) shows a schematic diagram of the graphene FET used for the transport measurement. Typically, the transport channels defined by the two electrodes deposited on graphene films were 5 mm wide and 1 mm long.
The quality and number of stacking layers of the graphene films were determined by micro-Raman spectroscopy (WITec Alpha300, 532 nm laser). Figure 1(b) shows an optical image (taken at the center of the graphene FET) of graphene on a SiO 2 /Si wafer. The 300 nm SiO 2 /Si wafers are almost ideal substrates for optically imaging graphene [16]. The uniformity of the color contrast in the optical image indicates uniform graphene thickness, although some small cracks were observed that were likely formed during the transfer process. The Raman spectrum (figure 1(c)) shows the following features typical of monolayer graphene: (i) a G-to-2D intensity ratio of ∼ 0.5 and (ii) a symmetric two-dimensional band centered at ∼ 2680 cm −1 with a full-width at half-maximum of ∼ 33 cm −1 [15,17]. The D band scattering from our sample, if present at all, was lower than the detection limit of the Raman system used.
Graphene FET devices on SiO 2 /Si substrates were clamped on a ceramic heater, which was then placed in a high-vacuum chamber equipped with electrical and gas feed-throughs. The vacuum chamber can be evacuated to a pressure of 5 × 10 −8 torr using a turbo pump (Turbo-V 81-M, Varian). Vacuum-annealing of the samples was performed in situ by heating the graphene on SiO 2 /Si samples at 150 • C for 2 h at 5 × 10 −8 torr in order to eliminate pre-existing adsorbates (i.e. H 2 O, O 2 molecules). After this vacuum-annealing, the samples were then exposed to NH 3 gas (99.99%; Airgas) for known exposure times. FET measurements were made by a programmable voltage source (2611A, Keithley) and a digital voltmeter/ammeter (6221 and 6514, Keithley). A back-gate bias (V gs ) ranging from −100 to +100 V was applied on the Si side of the SiO 2 /Si substrate. To improve the signal-to-noise ratio, a relatively high source-drain voltage (V ds ) of 1.5 V was applied to the device, while the source-drain current (I ds ) was monitored as a function of the applied back-gate bias.

Results and discussion
All graphene FETs were measured at room temperature. Figure 2 shows the typical response of the I ds to the gate bias for the as-prepared graphene FET under ambient conditions. A linear I ds -V gs curve is observed across the gate bias range used for the as-prepared graphene FET samples. With the gate bias ramped from −80 to 80 V, the I ds decreased from 2.0 to 0.8 mA, indicating that the as-prepared graphene was heavily p-typed; this may be due to the adsorption of water molecules from air or PMMA residue from the transfer process [18] or both. Thus, the Dirac point was outside the range of the gate biases that were studied. To minimize the presence of other adsorbates prior to exposure to NH 3 , the samples were heated at 150 • C for 2 h under vacuum at 5 × 10 −8 torr. After annealing, a V-shaped gate response of the I ds is observed from the graphene FET, as shown in figure 2. The in situ annealing thus yielded a Dirac point that is closer to zero gate bias, demonstrating the removal of (at least some) adsorbates and the recovery of the intrinsic bias dependence of graphene.  To study the effect of adsorption of NH 3 on the electrical response of graphene, the FET devices were exposed to NH 3 (g) after vacuum-annealing. The vacuum-annealed graphene FET was exposed to 10 torr of NH 3 gas for a total time of 30 min. Figure 3(a) shows the time evolution of the I ds versus V gs during this exposure to NH 3 gas. Initially, the Dirac point is close to +3 V back-gate bias; after 5 min of exposure, the Dirac point appears at −18 V and then gradually shifts to its final position at about −30 V. These results suggest that ammonia molecules are adsorbed on the graphene surface and cause a shift in the Fermi level in graphene from the Dirac point to the conduction band.
Studies of individual semiconducting carbon nanotube (CNT) sensors have been perfomed based on resistivity changes attributed to molecular adsorption on CNTs and partial electron transfer to the CNTs [19]. A recent first-principles calculation on graphene predicts that dipolar molecules can act as donors or acceptors with a small charge transfer χ (e.g. 0.027e for NH 3 ) [13]. This is consistent with calculations on defect-free CNTs [19]. Based on this calculated charge transfer of electrons from ammonia to graphene, the number density of the ammonia molecules, n, on the graphene surface can be estimated as where C is the capacitance of the graphene FET, V D is the back-gate voltage shift of the Dirac point relative to that of vacuum-annealed graphene, S is the surface area, e is the electron charge, ε r is the dielectric constant of SiO 2 and d is the thickness of SiO 2 . The values of n have been plotted as a function of exposure time in figure 3(b). The inset of figure 3(b) shows the saturated molecular density, with this assumed charge transfer, as a function of the NH 3 gas pressure.
The adsorption/desorption of the molecules can be understood based on a 'physically plausible' model. Denoting the number density of the molecules in the gas phase as n 0 (a constant in our experiments) and the number density on the graphene surface as n(t), the dynamics of the adsorption can be expressed as a rate equation, where t is the duration of exposure of graphene to NH 3 gas, p 0 is the adsorption rate and p is the desorption rate. In this model, the number of molecules that are adsorbed on (dn(t) ad /dt) and desorbed from (dn(t) de /dt) the surface per unit time is assumed to be proportional to the concentration of gas-phase (n 0 ) and surface (n) molecules, respectively. The adsorption/desorption rates are assumed to be dependent only on the temperature. We also assume that the density of molecules on the surface n(t) = 0 at t = 0, meaning that vacuumannealing removes all adsorbates. By solving equation (2), the density of the molecules on the surface as a function of exposure time is obtained as A curve fitted to the n(t) data using equation (3) is shown in figure 3 as well. The extracted values of p and n 0 p 0 are 0.0027 and 2.1 × 10 10 cm −2 , respectively, and hence with the assumptions mentioned above, at a NH 3 gas pressure of 10 torr, about 2.1 × 10 10 NH 3 molecules are adsorbed on 1 cm 2 of graphene and 0.27% molecules desorb in 1 s. It is worth noting that the values of p suggest a sensing response time (t 0 = 1/ p) if graphene FET is considered as an NH 3 (g) sensor.
Many studies as well as theoretical works [13,20] have reported that graphene has to be functionalized to achieve its impressive gas-sensing performance. Ab initio studies of the gas adsorption on graphene corroborate the role of impurities or vacancies, thus demonstrating a stronger gas adsorption at sites of atomic substitutions or defects [20]. The high sensitivity obtained on reduced graphene oxide gas sensors also supports the importance of functionalization [21]. Recently, experiments compared the electrical gas-sensing performance of 'dirty' and intrinsic graphene devices [14]. It has been reported that the responses of intrinsic graphene devices are surprisingly small even upon exposure to a strong analyte such as ammonia vapor. The unintentionally 'functionalized' residual polymer layer from the lithographic resist served to help concentrate the gas molecules or possibly enhance charge transfer. In our studies, a certain degree of Fermi level shift due to ammonia molecules was detected on performing an FET measurement at room temperature. These results may be due to the PMMA residue from the transfer process [11] and defects in the graphene grain boundary. Our previous works demonstrate that the typical grain size of CVD-grown graphene is ∼ 10 µm, which is two orders lower than the length of graphene devices [22].
Another main point of interest is that the value of the charge transfer rate χ between dipolar molecular adsorbates and graphene may depend strongly on molecular orientation with respect to the graphene surface [13]. The NH 3 molecule could, among other possibilities, orient with the N end of the molecule closest to the surface and the C 3v axis essentially perpendicular to the surface ('u' for 'up') or, alternatively, with the H atoms adjacent to the surface and the C 3v axis again perpendicular to the surface ('d' for 'down'). To determine whether there could be an effect from molecular orientation of adsorbed NH 3 molecules, a scan at low gate bias from −20 to +20 V was carried out immediately after applying a high-gate-bias pulse (+100 or −100 V) for 5 s to the graphene FET devices. Figure 4 shows four sequential I ds -V gs measurements in each with the application of one pulse before the scan. The four pulses were applied in the sequence of −100, +100, −100 and +100 V. Due to the interaction between the electric field and the molecular dipoles, we suggest that the high-positive-gate-bias pulse aligns NH 3 molecules along the d orientation and that the high-negative-bias pulse flips NH 3 molecules to the u orientation. The low-gate-bias scan is assumed to have a smaller effect on the orientation of the molecules. The measured curves clearly show that back-gate voltages of the Dirac point are (repeatably) −6 V (for −100 V pulses) and +3 V (for +100 V pulses). Compared with the vacuum-annealed result, V D for the u orientation is about −9 V, while a negligible shift occurs for the d orientation. This result indicates that almost all the NH 3 molecules flip from the u orientation to the d orientation when a +100 V pulse is applied, and the u orientation has a relatively large charge transfer ratio, which is consistent with a prediction based on the asymmetry of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) of the ammonia molecule [13]. The u orientation is energetically favored [13] and this explains the donor character observed at zero bias in our experiment. The flipping of some molecular dipoles on graphene could be the reason for the hysteretic behavior reported in other electric field effect measurements [18].

Summary and conclusions
In this paper, we have reported our study of the adsorption/desorption and flipping behavior of NH 3 molecules on a graphene surface through observation of the Fermi level shift, inferred to be from a partial charge transfer from the NH 3 molecules to graphene. A simple model has been used to evaluate the rates of adsorption and desorption of NH 3 molecules on graphene from the measured shift of the Fermi level as a function of exposure time. An electric-field-induced flip of the molecular dipoles (i.e. NH 3 molecules) is suggested from measured back-gate voltage shifts in the Dirac point after electric field pulses were applied via the gate bias.