Reduced graphene oxide for room-temperature ... - Semantic Scholar
IOP PUBLISHING
NANOTECHNOLOGY
Nanotechnology 20 (2009) 445502 (9pp)
doi:10.1088/0957-4484/20/44/445502
Reduced graphene oxide for room-temperature gas sensors Ganhua Lu1 , Leonidas E Ocola2 and Junhong Chen1,3,4 1
Department of Mechanical Engineering, University of Wisconsin-Milwaukee, Milwaukee, WI 53211, USA 2 Center for Nanoscale Materials, Argonne National Laboratory, Argonne, IL 60439, USA 3 State Key Laboratory of Pollution Control and Resources Reuse, College of Environmental Science and Engineering, Tongji University, Shanghai 200092, People’s Republic of China E-mail: [email protected]
Received 30 August 2009, in final form 3 September 2009 Published 7 October 2009 Online at stacks.iop.org/Nano/20/445502 Abstract We demonstrated high-performance gas sensors based on graphene oxide (GO) sheets partially reduced via low-temperature thermal treatments. Hydrophilic graphene oxide sheets uniformly suspended in water were first dispersed onto gold interdigitated electrodes. The partial reduction of the GO sheets was then achieved through low-temperature, multi-step annealing (100, 200, and 300 ◦ C) or one-step heating (200 ◦ C) of the device in argon flow at atmospheric pressure. The electrical conductance of GO was measured after each heating cycle to interpret the level of reduction. The thermally-reduced GO showed p-type semiconducting behavior in ambient conditions and was responsive to low-concentration NO2 and NH3 gases diluted in air at room temperature. The sensitivity can be attributed mainly to the electron transfer between the reduced GO and adsorbed gaseous molecules (NO2 /NH3 ). Additionally, the contact between GO and the Au electrode is likely to contribute to the overall sensing response because of the adsorbates-induced Schottky barrier variation. A simplified model is used to explain the experimental observations. (Some figures in this article are in colour only in the electronic version)
makes every carbon atom a surface atom. Graphene has been demonstrated as a promising gas-sensing material [5, 6]; for instance, Schedin et al reported that mechanically-exfoliated graphene can potentially detect gaseous species down to the single molecular level [5]. The gas-sensing mechanism of graphene is generally ascribed to the adsorption/desorption of gaseous molecules (which act as electron donors or acceptors) on the graphene surface, which leads to changes in the conductance of graphene [5]. A carbon nanotube (CNT) [7] can be regarded as a seamless cylinder by rolling up a graphene sheet [8]. Like CNT gas sensors where the contact between the CNT and the metal electrode can contribute to the sensor response [9], the graphene–electrode contact may also introduce additional gas-sensing effects. A wide range of physical and chemical routes can be employed to produce graphene. The mechanical cleavage of graphite was the first successful approach to obtaining single-layer graphene sheets [1]; however, it is difficult to control the number of layers for peeled-off fragments, and
1. Introduction Nanoscaled materials are attractive candidates for gas-sensing elements due to their unique and outstanding properties (e.g., extremely high surface-to-volume ratio) that potentially can lead to novel sensors with exceptional performance while reducing the device size and minimizing the energy consumption. In particular, graphene has become a ‘hot’ nanomaterial since it was first obtained through mechanical cleavage [1] because of its exceptional mechanical [2], thermal [3], and electrical [1, 4] properties. Graphene is a two-dimensional (2D) monolayer comprising sp2 -bonded carbon atoms [4]. It has an enormously high electron mobility at room temperature and its ballistic electron transport remains up to 0.3 μm at room temperature [4]. Electron transport through graphene is highly sensitive to adsorbed molecules owing to the 2D structure of graphene that 4 Author to whom any correspondence should be addressed.
0957-4484/09/445502+09$30.00
1
© 2009 IOP Publishing Ltd Printed in the UK
Nanotechnology 20 (2009) 445502
G Lu et al
in water to produce suspensions of single GO sheets. The existence of oxygen functional groups makes GO sheets strongly hydrophilic; stable aqueous dispersions containing almost entirely 1 nm-thick sheets can be obtained by a mild ultrasonic treatment of graphite oxide in water [18]. The sensing device was fabricated by dispersing the aqueous GO suspension onto Au interdigitated electrodes [27] with both finger-width and inter-finger spacing (source–drain separation) of about 1 μm. These electrodes were fabricated using an e-beam lithography process on a Si wafer with a top layer of thermally-formed SiO2 (200 nm). A few drops of the GO suspension were cast onto Au interdigitated electrodes and a discrete network of GO sheets was left behind on the wafer after water evaporation. The working principle of the sensing device (shown in figure 1) is that the drain–source channel becomes closed after the GO is partially reduced by low-temperature thermal treatments; thus, the conductance of the device varies upon exposure to various gases. Thermal reduction of GO was carried out in a tube furnace (Lindberg Blue, TF55035A-1) by two modes: successive multi-step heating and one-step heating. For successive multistep heating, three cycles in the order of 100, 200, and 300 ◦ C were performed on GO devices; the duration for each heating cycle was one hour (h) and an Ar flow of about 1 lpm was maintained during the process. For one-step heating, GO devices were treated in the furnace at 200 ◦ C in ∼1 lpm Ar flow for 2 h. After heating, samples were quickly cooled to room temperature within ∼5 min with the assistance of a blower. Both two-terminal dc and three-terminal field effect transistor (FET) (figure 1) measurements were performed on GO devices using a Keithley 2602 source meter. Electrical conductance of the GO device was measured by ramping the drain–source voltage (Vds ) and simultaneously recording the drain–source current ( Ids ) to evaluate the influence of thermal treatment on the device characteristics. The bottom of the silicon wafer was used as the back gate electrode in the FET measurements. The sensing performance of as-fabricated GO devices was characterized under practical conditions (i.e., room temperature and atmospheric pressure) against low-concentration NO2 and NH3 diluted in dry air. An air-tight test chamber with an electrical feedthrough was used to house a GO device for gas-sensing characterizations [28]. The chamber volume (6.3 × 10−5 m3 ) was minimized to reduce the capacitive effect. Variations in the electrical conductance of GO were monitored by simultaneously applying a low constant dc voltage (0.1–5 V) and recording the change in current passing through the device when the device was exposed periodically to clean air and NO2 - or NH3 -laden air. A sensing test cycle typically consists of three consecutive steps that include exposures of the device to (i) clean air flow to record a base value of the sensor conductance, (ii) target gas to register a sensing signal, and (iii) clean air flow for sensor recovery. The morphology of GO sheets was characterized using a Hitachi H 9000 NAR transmission electron microscope (TEM), which has a point resolution of 0.18 nm at 300 kV in the phase contrast high-resolution TEM imaging mode. TEM samples were prepared by depositing a few drops of
it is a low-yield process that is undesirable for large-scale production. Epitaxial growth of graphene layers requires high temperatures and an ultrahigh vacuum environment [10], which is expensive and thus may limit the widespread application of graphene. Ambient-pressure chemical vapor deposition (CVD) has recently been reported to produce largearea films of 1 to ∼12 graphene layers [11], but a purification process is needed to eradicate the catalyst Ni particles in order to obtain clean graphene sheets. Li et al [12] recently reported the CVD-growth of predominantly single-layer graphene with lateral sizes up to centimeters on copper foils using methane; however, a high temperature (1000 ◦ C) and vacuum were required for the graphene synthesis. A potentially cost-effective method for mass-producing graphene-based devices is to first produce chemically modified graphene, such as graphene oxide (GO), and then reduce it to obtain graphene for device applications. Large-quantity GO can be easily produced by the chemical exfoliation of graphite through oxidation and the subsequent dispersion in water [13]. The basal plane and edges of GO are decorated with oxygen functional groups [14–16], making GO highly soluble in water. Single-layer GO sheets can be generated by simple sonication of hydrophilic graphite oxide in water. GO is electrically insulating, owing to the disruption of the sp2 bonded graphitic structure by the attachment of electronegative oxygen atoms [17]. It can become conductive by exposing it to reducing agents such as hydrazine [18], NaBH4 [19], through high-temperature treatment [20], or via UV-assisted photocatalysis [21]. Recently, low-temperature annealing reduction of GO has been reported [22]. Hydrazine-reduced GO has shown excellent performance for detecting acetone, warfare agents, and explosive agents at part-per-billion concentrations with greatly reduced noise levels compared with sensors based on CNTs [6]; however, fabrication of the above-mentioned sensor involved toxic chemicals. Chemical reduction using hydrazine was also found to bring extra nitrogen functional groups onto the graphene surface [23], which may decelerate the sensing response [24]. We have recently reported a NO2 sensing device based on partially-reduced GO [25]. Here, we discuss more details about the fabrication and characterization of the high-performance gas sensor using GO that was partially reduced by lowtemperature annealing in Ar at atmospheric pressure. We present the sensing properties of reduced GO devices for detecting low-concentration NO2 and NH3 gases under a practical environment, i.e., atmospheric pressure and room temperature. In particular, the effect of various thermal treatments on the reduction of GO and the subsequent sensing performance is studied. We also include simulation results based on a very simplified model to interpret experimental observations.
2. Experimental details Graphite oxide was synthesized by the oxidative treatment of purified natural graphite using the modified Hummers method [26]. The graphite oxide was then fully exfoliated 2
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 1. Schematic diagram of the reduced GO device. A reduced GO sheet bridges the source and drain electrodes, which closes the circuit. In FET measurements, the back of the Si wafer is used as the gate electrode.
Figure 2. (a) A TEM image of a single GO sheet spanning across a hole formed by the lacey carbon film and the copper bar of a TEM grid. (b) An SEM image showing a GO sheet (∼1 μm × 2 μm) bridging a pair of Au electrodes; the electrode width of and the gap between neighboring electrodes are both about 1 μm.
such as dielectrophoretic assembly [30] could be used to delicately control the arrangement of GO sheets between electrodes, which may lead to adjustable device characteristics and functionality.
GO suspension on holey-carbon-film-covered copper grids (from Ted Pella, Inc.). A field-emission scanning electron microscope (SEM) (Hitachi S 4800) was used to characterize the fabricated GO devices; the SEM had a resolution of 1.4 nm at 1 kV acceleration voltage.
3.1. Reduction of GO via thermal treatments
3. Results and discussion
We first conducted successive multi-step heating to inspect the effect of low-temperature heating on the reduction of GO. The Ids –Vds curves in figure 3(a) present the typical changes in the conductance of GO devices before and after thermal treatments. GO is normally electrically insulating at room temperature with its resistance on the order of tens of G, as indicated by the almost horizontal Ids –Vds curve A in figure 3(a). This insulating nature of GO can be elucidated by the extensive presence of epoxide and hydroxyl groups on both sides of the basal plane and carbonyl and carboxylic groups on the edges of GO [14], leading to the massive existence of sp3 -hybridized carbon atoms. Although there are regions of unsaturated carbon atoms, i.e., nanometer-sized graphitic domains [31], these regions are separated by vast areas of oxidized carbon atoms. GO stays non-conductive unless more clusters of graphitic atoms (sp2 -hybridized) can be restored by
Figure 2(a) is a TEM image that shows a single GO sheet spanning across a hole formed by the holey-carbon-film and the copper bar of a TEM grid. The lateral dimension of the GO sheet is about 4 μm and the sheet is mostly uniform and smooth, except that there are rolls and folds on its edges. More details on the TEM characterization of GO are presented in our previous paper [29]. SEM was used to verify the presence of GO sheets between electrodes after deposition of the GO suspension on the sensor electrode/wafer. An individual GO sheet (∼1 μm × 2 μm) bridging a pair of neighboring Au electrode fingers is shown in the SEM image of figure 2(b). Similar to the TEM observation, rolls and folds can be seen on the GO sheet in the SEM image. The arrangement of GO sheets on the wafer was quite random and no effort has been made to control the layout of GO sheets; however, techniques 3
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 3. (a) Ids –Vds measurement results of a GO device before and after successive multi-step heating. Curve A: without any thermal treatment; curve B: successively annealed in Ar at 100 and 200 ◦ C for 1 h each; curve C: further annealed in Ar at 300 ◦ C for 1 h. (b) Ids –Vg curve (with Vds = 0.01 V) after 100, 200, and 300 ◦ C annealing, showing p-type semiconductivity of reduced GO in an ambient environment. (c) Ids –Vds curves of a GO device before (curve I) and after (curve II) one-step heating in Ar flow at 200 ◦ C for 2 h.
suggest possible ohmic contact between reduced GO sheets and Au electrodes. The resistance of the GO device further decreased by a factor of ∼10 after 300 ◦ C annealing (curve C in figure 3(a)), indicating most likely further reduction of GO rather than contact improvement. However, additional characterization of the contact between GO sheets and Au electrodes is needed before the reduction of resistance, i.e., the exact contribution of the GO reduction, can be fully understood. Both temperature and duration affect the extent of GO powder reduction, as well as the exfoliated material such as GO sheets. Thermal treatments of GO at 1100 ◦ C in vacuum [32] or Ar/H2 flow [20], or at 400 ◦ C with hydrazine vapor pretreatment [32], yielded significantly reduced material. Figure 3(b) shows the transport characteristics of the GO device after the 300 ◦ C annealing. The behavior is similar to that of a p-type semiconductor. Ids decreases with increasing gate voltage (Vg ), implying that the transport through the reduced GO sheets is dominated by positive charge carriers (holes). This p-type semiconducting behavior agrees with the results reported for graphene sheets prepared by micromechanically cleaving graphite [33], and chemically[34] and thermally- [22] reduced GO when exposed to the ambient environment. The p-type behavior is most likely due to the polarization of adsorbed molecules (e.g., water and O2 ) and/or defects introduced on the graphene sheets during the preparation or reduction process [35], or even the influence of the supporting substrate [36, 37]. In the case of defectinduced changes in electronic properties, the p-type nature of graphene was enhanced with increasing structural defects via
removing oxygen functional groups to decrease the distance between graphitic domains, which results in charge transport via variable range hopping [31], or even to create continuous graphitic ‘paths’ for charge transport. After annealing in Ar at 100 ◦ C for 1 h, the resistance of the device only slightly decreased, which agrees with Jung et al [22]. A slight reduction may have occurred during the 100 ◦ C heating treatment; however, the reduction was insufficient to make GO conductive and introduce enough reactive sites on the GO surface for gas adsorption and sensing. With further annealing at 200 ◦ C for 1 h, the device resistance decreased to ∼750 k, as estimated from the Ids –Vds curve B in figure 3(a), implying that GO sheets in the device were partially reduced. Thermal treatments could pyrolyze and remove oxygen functional groups from GO, which recovers graphitic regions and makes GO more conductive. The release of CO, CO2 , and H2 O has been detected during the thermal reduction of GO [24]. Yang et al [23] reported the partial reduction of GO by exposure to an Ar flow at 200 ◦ C for 30 min and found that the carbon-to-oxygen atomic ratio increased from ∼2.8 to ∼3.9 based on x-ray photoelectron spectroscopy (XPS) measurements. Jung et al demonstrated that the electrical conductivity of individual GO sheets significantly increased after being heated in vacuum at only 125–240 ◦ C, indicating a partial reduction of the GO [22]. Of course, the significant reduction in device resistance could also be partly attributed to the improvement in the electrical contact between GO sheets and Au electrodes during the annealing processes. The linearity and symmetry of the Ids –Vds curve 4
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 4. Room-temperature NO2 sensing behavior of a GO device thermally treated with successive multi-step heating. After annealing in Ar at 100 and 200 ◦ C for 1 h each, (a) the GO device shows repeatable response to 100 ppm NO2 ; (b) the sensing signal is highly dependent on the NO2 concentration; and (c) the GO sensor can detect NO2 at a concentration as low as 2 ppm. (d) 300 ◦ C annealing improved both the sensor sensitivity and response time but lengthened the recovery time compared with 200 ◦ C annealing.
plasma exposure [35]. This is analogous to multiwalled CNTs, whose electronic properties could be altered from metallic to p-type semiconducting with structural deformations [38]. Successive multi-step heating reveals the conductance evolution of GO under various heating conditions, which is instrumental for understanding the reduction process; however, it involves multiple heating cycles that may add complexity and thus hinder practical applications of GO. Aiming at future rapid and large-scale production of reduced GO devices, we employed one-step heating (200 ◦ C in Ar for 2 h) to reduce GO and found that partial reduction of GO also can be achieved through this type of heating. Figure 3(c) compares the conductance of a GO device before (curve I) and after (curve II) the one-step heating previously described. The slope of Ids –Vds curve II is significantly higher than that of curve I, indicating elevated conductance of the GO device and effective GO reduction by the one-step heating. As figures 3(a) and (c) show, the conductance of GO devices may vary within a range after thermal treatments, which can be ascribed to the fluctuation in heating conditions (e.g., heating temperature and time) and the variation of the number density of GO sheets in devices.
After being partially reduced by either successive multistep heating at 100 and 200 ◦ C (1 h each) or one-step heating at 200 ◦ C (2 h), the GO devices became highly responsive to NO2 and NH3 , which was most likely due to the recovery of many graphitic carbon atoms as active sites for target gas adsorption. Possibly vacancies or small holes were created during thermal treatment and these defects may also serve as adsorption sites for gaseous molecules [24]. Figure 4(a) shows a typical dynamic response (current versus time) of a GO device for room-temperature detection of 100 ppm NO2 after being successively heated at 100 and 200 ◦ C in Ar for 1 h each. The sensor was periodically exposed to clean dry air flow (2 lpm) for 10 min to record a base value of the sensor conductance, 100 ppm NO2 diluted in air (2 lpm) for 15 min to register a sensing signal, and clean air flow (2 lpm) again for 25 min to recover the device. Upon the introduction of NO2 , the sensor current went up, i.e., the conductance of the sensor increased; when the NO2 flow was turned off and the air flow restored, the device re-established its conductance in about 30 min. Three cycles were repeated in figure 4(a) and the signal was fairly reproducible. The sensing signal strength (proportional to the spike height with NO2 on) was dependent on the NO2 concentration (shown in figure 4(b)) as it decreased with decreasing NO2 concentrations from 100, to 50, and to 25 ppm. Figure 4(c) shows the sensor response to 2 ppm NO2 , where the conductance increased ∼12% with 40 min NO2 exposure. Assuming a linear relationship between conductance change and NO2 concentration, this sensitivity is comparable to the sensor based on mechanically-cleaved graphene, which showed ∼4.3% increment in the conductance for 1 ppm NO2 [5]. The sensing performance of the devices
3.2. Gas sensing with reduced GO devices Devices with as-deposited GO sheets showed no response to 100 ppm NO2 or 1% NH3 , indicating insignificant change in the electrical transport property of the non-reduced GO. We also found that 1 h 100 ◦ C heating was usually inadequate to make GO devices responsive to gases. 5
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 5. (a) Dynamic sensing behavior of a reduced GO device (successively heated in Ar at 100, 200, and 300 ◦ C for 1 h each) for 1% NH3 detection at room temperature. (b) The recovery process of the device after NH3 sensing was extremely slow. After 50 h air flow, the GO device still did not restore its initial conductance.
reported here is very encouraging for practical applications when considering the simplicity and low cost to fabricate these devices and the potential opportunities for optimization. The sensing performance of the reported device is attributed to the effective adsorption of NO2 on the surface of reduced, p-type GO. NO2 is a strong oxidizer with electron-withdrawing power [39]; therefore, electron transfer from reduced GO to adsorbed NO2 leads to an enriched hole concentration and enhanced electrical conduction in the reduced GO sheet. Figure 4(d) compares the sensitivities of the device for 100 ppm NO2 detection after 200 and 300 ◦ C annealing. The sensor sensitivity is evaluated as the ratio of (G g –G a )/G a , where G a is the sensor conductance in clean air and G g is the sensor conductance in air containing 100 ppm NO2 . With 300 ◦ C annealing, the device showed a sensitivity of ∼1.56 to 100 ppm NO2 , higher than that (∼1.41) after 200 ◦ C annealing. In addition, it had a faster response when exposed to NO2 , as evidenced by a steeper slope upon the exposure to NO2 . This accelerated response is likely due to the creation of more graphitic carbon atoms during the 300 ◦ C annealing because molecular adsorption onto sp2 -bonded carbon (lower binding energy required) is faster than onto defects [6]. However, the sensor recovery after 300 ◦ C annealing became slower because the device did not return to its initial conductance after 30 min exposure to dry air, whereas in the 200 ◦ C annealing case the full recovery was achieved under the same exposure condition. Low-temperature heating and UV illumination could be used to accelerate the sensor recovery. The GO devices responded to NH3 as well, after either successive thermal treatments or one-step heating. Figure 5 is the 1% NH3 sensing data obtained from the same reduced GO device (after 300 ◦ C heating), whose NO2 sensing behavior is presented in figure 4. Upon NH3 exposure, the current passing through the device decreased (figure 5(a)), possibly because adsorption of NH3 (electron donator) can lower the hole concentration in GO, and thus reduce GO conductance. Unlike the relatively fast recovery (about 30 min) after NO2 exposure, the conductance of the GO device could not return to its initial value in air flow (2 lpm) after about 50 h (shown in figure 5(b)). Most of the GO sensors (with one exception to be discussed) recovered very slowly after NH3 sensing.
Further investigation is needed to understand the intrinsic mechanisms associated with the slow recovery of reduced GO from NH3 exposure and to find effective measures to accelerate the recovery process before reduced GO can be used practically for repeatable NH3 detection. Among all of the GO devices fabricated (11 in total), only one device recovered exceptionally fast after NH3 sensing. Figure 6(a) shows three NH3 sensing cycles using this device after it was reduced through one-step heating in Ar at 200 ◦ C for 2 h. This NH3 sensing performance is superior to that shown in figure 5(a); however, an unusual current increase was observed at the beginning of NH3 exposure in each cycle (indicated by the arrows in figure 6(a)), which is against the GO sensing mechanism discussed previously and was not found in other devices. Figure 6(b) shows a magnified view of the area marked by the leftmost arrow in figure 6(a), showing that the current increased with a very sharp slope at the beginning of NH3 exposure. This abnormal increase may be attributed to the unsteadiness of flow field in the chamber during gas switching or some other accidental noise; however, even more strange behavior was observed from this device when using it to sense 1% NH3 again after ∼2 months. As shown in figure 6(c), the device gave a sensing signal (curve I) with a completely opposite trend (i.e., its conductance increased upon NH3 exposure) after 2 months, compared with that obtained freshly after reduction (curve II, obtained by normalizing the curve in figure 6(a)). The current ( I ) is normalized to the initial value ( I0 ) in air flow for the convenience of comparison. This observation suggests competing sensing mechanism(s) may exist besides gas adsorption/desorption on reduced p-type GO for this device. The device was then re-treated in Ar at 200 ◦ C for 2 h; although it then restored to ‘normal’ sensing behavior (conductance decreased upon NH3 exposure), as shown by curve III in figure 6(c), its sensitivity degraded. Figure 6(d) shows the Ids –Vds curve of the device measured 2 months after the first thermal treatment. The curve is asymmetric and nonlinear, which is contrary to the one acquired shortly after the thermal treatment (inset of figure 6(d)) and suggests a non-ohmic contact between the GO and Au electrodes. This contact issue may have played a role in the abnormal NH3 sensing behavior. For instance, Peng et al proposed that the Schottky barrier modulation at the CNT– metal contact significantly contributed to NH3 sensing when 6
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 6. (a) 1% NH3 detection by a GO device treated with one-step heating (in Ar at 200 ◦ C for 2 h). This device recovered faster compared with the one presented in figure 5. (b) Upon exposure to NH3 , the conductance of the device first oddly increased for a few seconds (this increase cannot be explained by a NH3 -adsorption-induced change in GO conductance) before it decreased (as expected). (c) The sensing response for 1% NH3 after 2 months (curve I) is opposite (i.e., the conductance increased upon NH3 exposure) to that observed after freshly heated (curve II, obtained by normalizing the curve in a). The sensor response returned to normal after another heating treatment (in Ar at 200 ◦ C for 2 h) but the sensitivity degraded, as indicated by curve III. (d) The Ids –Vds curve of the device became asymmetric and nonlinear, implying non-ohmic contact between the GO and Au electrodes; the inset shows the Ids –Vds obtained freshly after the first heating.
where G(t) is the change in the conductance of the sensor; G max is the maximum change in the device conductance with sufficiently long gas exposure; c is the gas concentration; K is the binding equilibrium constant (assuming Langmuir adsorption); and k is the surface reaction rate (depending on the properties of both the sensing element and analyte molecules). Note that the sensing process starts at t = 0 for equation (1). If we only focus on the trend of sensing responses, neglect the details of all constants, and take into account that I (change in the measured current I of the sensor) is proportional to G , we could obtain a very simplified expression for normalized change in the measured current I (t)/I0 as,
using CNT sensors at room temperature [9]. Therefore, the influence of graphene–metal contact on the sensor performance is worthy of future study. We also could not exclude the connection between the abnormal NH3 sensing behavior and the unusually fast recovery. Understanding this odd NH3 sensing behavior may lead to measures that can be used to engineer sensing device properties, e.g., recovery rate and gas selectivity. 3.3. Simplified modeling of strange NH3 sensing behavior We have developed a very simplified model to shed light on the unusual NH3 sensing behavior presented in figure 6 and hope to inspire more in-depth analysis in the future. We assume that there are two NH3 sensing mechanisms ( M 1 and M 2) that can lead to competing responses from the device (i.e., M 1 leads to increased conductance and M 2 leads to decreased conductance; see figure 7) and that the overall sensing behavior is simply the linear combination of the two mechanisms. Due to the close relation between the CNT and the graphene, it could be a reasonable starting point to evaluate GO sensors with existing models developed for CNT sensors. Taking advantage of the CNT sensing model proposed by Lee et al [40], the GO sensing response due to gas molecule adsorption could be written as cK , G(t) = G max (1) 1 + cK 1 − exp − 1+cK kt K
I (t)/I0 = A exp(−Ct) + D,
(2)
where I0 is the current measured before the NH3 exposure; A, C , and D are constants determined by factors such as the sensor properties, gas concentration, and properties of target gas molecules. Assigning suitable values to A, C , and D , we can obtain equations to describe M 1 and M 2 and use them to simulate the experimental results presented in figure 6. As shown in figure 7, we assume that M 1 causes the device conductance to increase upon the NH3 exposure and that the normalized change in the measured current I (t)/I0 due to M 1 can be expressed as,
M 1: I (t)/I0 = 5 − 4 exp[−0.033 · (t − 600)]. 7
(3)
Nanotechnology 20 (2009) 445502
G Lu et al
and the metal electrode makes the sensing response more complex and deserves further investigation. The simple and low-cost manufacturing process and the wide availability of GO could lead to cost-effective graphene-based gas sensors and other opportunities for graphene.
Acknowledgments This work was financially supported by the NSF (CMMI0900509 and CBET-0803142) and by a catalyst grant from the University of Wisconsin-Milwaukee Research Foundation. The authors thank R S Ruoff and D A Dikin for providing GO suspensions, M Gajdardziska-Josifovska for providing TEM access, and D Rosenmann for assistance in metal sputtering. The SEM imaging was conducted at the UWM Electron Microscope Laboratory. The e-beam lithography was performed at the Center for Nanoscale Materials of Argonne National Laboratory, which is supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357.
Figure 7. Simulation of the observed abnormal NH3 sensing response using a simplified model. M 1 causes the conductance to increase while M 2 causes the conductance to decrease upon NH3 exposure. Curve I and curve II are experimental data (the first cycles of corresponding curves shown in figure 6(c)). Linear combinations of M 1 and M 2 could be used to qualitatively explain observed NH3 sensing responses.
References (Note that the NH3 exposure starts at t = 600 s.) Similarly, we could use the following equation to represent the sensing signal resulting from M 2, which decreases the conductance,
M 2: I (t)/I0 = exp[−0.012(t − 600)].
[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 Science 306 666–9 [2] Frank I W, Tanenbaum D M, Van der Zande A M and McEuen P L 2007 J. Vac. Sci. Technol. B 25 2558–61 [3] Balandin A A, Ghosh S, Bao W Z, Calizo I, Teweldebrhan D, Miao F and Lau C N 2008 Nano Lett. 8 902–7 [4] Geim A K and Novoselov K S 2007 Nat. Mater. 6 183–91 [5] Schedin F, Geim A K, Morozov S V, Hill E W, Blake P, Katsnelson M I and Novoselov K S 2007 Nat. Mater. 6 652–5 [6] Robinson J T, Perkins F K, Snow E S, Wei Z Q and Sheehan P E 2008 Nano Lett. 8 3137–40 [7] Iijima S 1991 Nature 354 56–8 [8] Baughman R H, Zakhidov A A and de Heer W A 2002 Science 297 787–92 [9] Peng N, Zhang Q, Chow C L, Tan O K and Marzari N 2009 Nano Lett. 9 1626–30 [10] Berger C et al 2006 Science 312 1191–6 [11] Reina A, Jia X, Ho J, Nezich D, Son H, Bulovic V, Dresselhaus M S and Kong J 2009 Nano Lett. 9 30–5 [12] Li X et al 2009 Science 324 1312–4 [13] Ruoff R 2008 Nat. Nanotechnol. 3 10–1 [14] Cai W W et al 2008 Science 321 1815–7 [15] Lerf A, He H Y, Forster M and Klinowski J 1998 J. Phys. Chem. B 102 4477–82 [16] He H Y, Klinowski J, Forster M and Lerf A 1998 Chem. Phys. Lett. 287 53–6 [17] Park S and Ruoff R S 2009 Nat. Nanotechnol. 4 217–24 [18] Stankovich S, Piner R D, Chen X Q, Wu N Q, Nguyen S T and Ruoff R S 2006 J. Mater. Chem. 16 155–8 [19] Muszynski R, Seger B and Kamat P V 2008 J. Phys. Chem. C 112 5263–6 [20] Wang X, Zhi L J and Mullen K 2008 Nano Lett. 8 323–7 [21] Williams G, Seger B and Kamat P V 2008 ACS Nano 2 1487–91 [22] Jung I, Dikin D A, Piner R D and Ruoff R S 2008 Nano Lett. 8 4283–7 [23] Yang D et al 2009 Carbon 47 145–52 [24] Jung I, Dikin D, Park S, Cai W, Mielke S L and Ruoff R S 2008 J. Phys. Chem. C 112 20264–8
(4)
Normally, M 2 is dominant in GO sensors for detecting NH3 while M 1 is negligible, which explains the typical NH3 sensing behavior of the GO sensors shown in figure 5(a); however, an abnormal sensing response could appear when M 1 cannot be neglected, even if M 2 still dominates. For instance, if the total signal consists of mostly M 2 (e.g., 87%) and a small contribution from M 1 (e.g., 13%), the combination (figure 7) fits the first cycle of the experimental curve II in figure 6(c) fairly well; I (t)/I0 first increases with the start of NH3 flow and then decreases. If the contribution of M 1 increases to 83.8% due to certain unidentified reasons (e.g., enhanced GO-contact influence or degraded GO), the simulated result matches with the curve I (from figure 6(c)), which is against the expected response from GO sensors. Of course, the model presented here is overly simplified and more sophisticated simulations are desired for the in-depth theoretical analysis.
4. Conclusion In summary, GO was partially reduced via thermal treatments in Ar flow with temperatures as low as 200 ◦ C. After the treatment, the electrical conductance of the GO sheets increased due to the removal of oxygen functional groups. The thermally-reduced GO demonstrated transport characteristics typical of a p-type semiconductor, which can be used to fabricate molecular adsorption-type gas sensors. Miniaturized gas sensors based on the reduced GO were fabricated and exhibited room-temperature sensing properties under atmospheric pressure. The electrical contact between the GO 8
Nanotechnology 20 (2009) 445502
G Lu et al
[32] Becerril H A, Mao J, Liu Z, Stoltenberg R M, Bao Z and Chen Y 2008 ACS Nano 2 463–70 [33] Romero H E, Shen N, Joshi P, Gutierrez H R, Tadigadapa S A, Sofo J O and Eklund P C 2008 ACS Nano 2 2037–44 [34] Gilje S, Han S, Wang M, Wang K L and Kaner R B 2007 Nano Lett. 7 3394–8 [35] Kim K, Park H J, Woo B C, Kim K J, Kim G T and Yun W S 2008 Nano Lett. 8 3092–6 [36] Adam S, Hwang E H, Galitski V M and Das Sarma S 2007 Proc. Natl Acad. Sci. USA 104 18392–7 [37] Fratini S and Guinea F 2008 Phys. Rev. B 77 195415 [38] Martel R, Schmidt T, Shea H R, Hertel T and Avouris P 1998 Appl. Phys. Lett. 73 2447–9 [39] Leenaerts O, Partoens B and Peeters F M 2008 Phys. Rev. B 77 125416 [40] Lee C Y and Strano M S 2005 Langmuir 21 5192–6
[25] Lu G H, Ocola L E and Chen J H 2009 Appl. Phys. Lett. 94 083111 [26] Park S, An J, Piner R D, Jung I, Yang D, Velamakanni A, Nguyen S T and Ruoff R S 2008 Chem. Mater. 20 6592–4 [27] Lu G H, Huebner K L, Ocola L E, Gajdardziska-Josifovska M and Chen J H 2006 J. Nanomater. 2006 60828 [28] Lu G H, Ocola L E and Chen J H 2009 Adv. Mater. 21 2487–91 [29] Lu G H, Mao S, Park S, Ruoff R S and Chen J H 2009 Nano Res. 2 192–200 [30] Vijayaraghavan A, Sciascia C, Dehm S, Lombardo A, Bonetti A, Ferrari A C and Krupke R 2009 ACS Nano 3 1729–34 [31] Gomez-Navarro C, Weitz R T, Bittner A M, Scolari M, Mews A, Burghard M and Kern K 2007 Nano Lett. 7 3499–503
9
NANOTECHNOLOGY
Nanotechnology 20 (2009) 445502 (9pp)
doi:10.1088/0957-4484/20/44/445502
Reduced graphene oxide for room-temperature gas sensors Ganhua Lu1 , Leonidas E Ocola2 and Junhong Chen1,3,4 1
Department of Mechanical Engineering, University of Wisconsin-Milwaukee, Milwaukee, WI 53211, USA 2 Center for Nanoscale Materials, Argonne National Laboratory, Argonne, IL 60439, USA 3 State Key Laboratory of Pollution Control and Resources Reuse, College of Environmental Science and Engineering, Tongji University, Shanghai 200092, People’s Republic of China E-mail: [email protected]
Received 30 August 2009, in final form 3 September 2009 Published 7 October 2009 Online at stacks.iop.org/Nano/20/445502 Abstract We demonstrated high-performance gas sensors based on graphene oxide (GO) sheets partially reduced via low-temperature thermal treatments. Hydrophilic graphene oxide sheets uniformly suspended in water were first dispersed onto gold interdigitated electrodes. The partial reduction of the GO sheets was then achieved through low-temperature, multi-step annealing (100, 200, and 300 ◦ C) or one-step heating (200 ◦ C) of the device in argon flow at atmospheric pressure. The electrical conductance of GO was measured after each heating cycle to interpret the level of reduction. The thermally-reduced GO showed p-type semiconducting behavior in ambient conditions and was responsive to low-concentration NO2 and NH3 gases diluted in air at room temperature. The sensitivity can be attributed mainly to the electron transfer between the reduced GO and adsorbed gaseous molecules (NO2 /NH3 ). Additionally, the contact between GO and the Au electrode is likely to contribute to the overall sensing response because of the adsorbates-induced Schottky barrier variation. A simplified model is used to explain the experimental observations. (Some figures in this article are in colour only in the electronic version)
makes every carbon atom a surface atom. Graphene has been demonstrated as a promising gas-sensing material [5, 6]; for instance, Schedin et al reported that mechanically-exfoliated graphene can potentially detect gaseous species down to the single molecular level [5]. The gas-sensing mechanism of graphene is generally ascribed to the adsorption/desorption of gaseous molecules (which act as electron donors or acceptors) on the graphene surface, which leads to changes in the conductance of graphene [5]. A carbon nanotube (CNT) [7] can be regarded as a seamless cylinder by rolling up a graphene sheet [8]. Like CNT gas sensors where the contact between the CNT and the metal electrode can contribute to the sensor response [9], the graphene–electrode contact may also introduce additional gas-sensing effects. A wide range of physical and chemical routes can be employed to produce graphene. The mechanical cleavage of graphite was the first successful approach to obtaining single-layer graphene sheets [1]; however, it is difficult to control the number of layers for peeled-off fragments, and
1. Introduction Nanoscaled materials are attractive candidates for gas-sensing elements due to their unique and outstanding properties (e.g., extremely high surface-to-volume ratio) that potentially can lead to novel sensors with exceptional performance while reducing the device size and minimizing the energy consumption. In particular, graphene has become a ‘hot’ nanomaterial since it was first obtained through mechanical cleavage [1] because of its exceptional mechanical [2], thermal [3], and electrical [1, 4] properties. Graphene is a two-dimensional (2D) monolayer comprising sp2 -bonded carbon atoms [4]. It has an enormously high electron mobility at room temperature and its ballistic electron transport remains up to 0.3 μm at room temperature [4]. Electron transport through graphene is highly sensitive to adsorbed molecules owing to the 2D structure of graphene that 4 Author to whom any correspondence should be addressed.
0957-4484/09/445502+09$30.00
1
© 2009 IOP Publishing Ltd Printed in the UK
Nanotechnology 20 (2009) 445502
G Lu et al
in water to produce suspensions of single GO sheets. The existence of oxygen functional groups makes GO sheets strongly hydrophilic; stable aqueous dispersions containing almost entirely 1 nm-thick sheets can be obtained by a mild ultrasonic treatment of graphite oxide in water [18]. The sensing device was fabricated by dispersing the aqueous GO suspension onto Au interdigitated electrodes [27] with both finger-width and inter-finger spacing (source–drain separation) of about 1 μm. These electrodes were fabricated using an e-beam lithography process on a Si wafer with a top layer of thermally-formed SiO2 (200 nm). A few drops of the GO suspension were cast onto Au interdigitated electrodes and a discrete network of GO sheets was left behind on the wafer after water evaporation. The working principle of the sensing device (shown in figure 1) is that the drain–source channel becomes closed after the GO is partially reduced by low-temperature thermal treatments; thus, the conductance of the device varies upon exposure to various gases. Thermal reduction of GO was carried out in a tube furnace (Lindberg Blue, TF55035A-1) by two modes: successive multi-step heating and one-step heating. For successive multistep heating, three cycles in the order of 100, 200, and 300 ◦ C were performed on GO devices; the duration for each heating cycle was one hour (h) and an Ar flow of about 1 lpm was maintained during the process. For one-step heating, GO devices were treated in the furnace at 200 ◦ C in ∼1 lpm Ar flow for 2 h. After heating, samples were quickly cooled to room temperature within ∼5 min with the assistance of a blower. Both two-terminal dc and three-terminal field effect transistor (FET) (figure 1) measurements were performed on GO devices using a Keithley 2602 source meter. Electrical conductance of the GO device was measured by ramping the drain–source voltage (Vds ) and simultaneously recording the drain–source current ( Ids ) to evaluate the influence of thermal treatment on the device characteristics. The bottom of the silicon wafer was used as the back gate electrode in the FET measurements. The sensing performance of as-fabricated GO devices was characterized under practical conditions (i.e., room temperature and atmospheric pressure) against low-concentration NO2 and NH3 diluted in dry air. An air-tight test chamber with an electrical feedthrough was used to house a GO device for gas-sensing characterizations [28]. The chamber volume (6.3 × 10−5 m3 ) was minimized to reduce the capacitive effect. Variations in the electrical conductance of GO were monitored by simultaneously applying a low constant dc voltage (0.1–5 V) and recording the change in current passing through the device when the device was exposed periodically to clean air and NO2 - or NH3 -laden air. A sensing test cycle typically consists of three consecutive steps that include exposures of the device to (i) clean air flow to record a base value of the sensor conductance, (ii) target gas to register a sensing signal, and (iii) clean air flow for sensor recovery. The morphology of GO sheets was characterized using a Hitachi H 9000 NAR transmission electron microscope (TEM), which has a point resolution of 0.18 nm at 300 kV in the phase contrast high-resolution TEM imaging mode. TEM samples were prepared by depositing a few drops of
it is a low-yield process that is undesirable for large-scale production. Epitaxial growth of graphene layers requires high temperatures and an ultrahigh vacuum environment [10], which is expensive and thus may limit the widespread application of graphene. Ambient-pressure chemical vapor deposition (CVD) has recently been reported to produce largearea films of 1 to ∼12 graphene layers [11], but a purification process is needed to eradicate the catalyst Ni particles in order to obtain clean graphene sheets. Li et al [12] recently reported the CVD-growth of predominantly single-layer graphene with lateral sizes up to centimeters on copper foils using methane; however, a high temperature (1000 ◦ C) and vacuum were required for the graphene synthesis. A potentially cost-effective method for mass-producing graphene-based devices is to first produce chemically modified graphene, such as graphene oxide (GO), and then reduce it to obtain graphene for device applications. Large-quantity GO can be easily produced by the chemical exfoliation of graphite through oxidation and the subsequent dispersion in water [13]. The basal plane and edges of GO are decorated with oxygen functional groups [14–16], making GO highly soluble in water. Single-layer GO sheets can be generated by simple sonication of hydrophilic graphite oxide in water. GO is electrically insulating, owing to the disruption of the sp2 bonded graphitic structure by the attachment of electronegative oxygen atoms [17]. It can become conductive by exposing it to reducing agents such as hydrazine [18], NaBH4 [19], through high-temperature treatment [20], or via UV-assisted photocatalysis [21]. Recently, low-temperature annealing reduction of GO has been reported [22]. Hydrazine-reduced GO has shown excellent performance for detecting acetone, warfare agents, and explosive agents at part-per-billion concentrations with greatly reduced noise levels compared with sensors based on CNTs [6]; however, fabrication of the above-mentioned sensor involved toxic chemicals. Chemical reduction using hydrazine was also found to bring extra nitrogen functional groups onto the graphene surface [23], which may decelerate the sensing response [24]. We have recently reported a NO2 sensing device based on partially-reduced GO [25]. Here, we discuss more details about the fabrication and characterization of the high-performance gas sensor using GO that was partially reduced by lowtemperature annealing in Ar at atmospheric pressure. We present the sensing properties of reduced GO devices for detecting low-concentration NO2 and NH3 gases under a practical environment, i.e., atmospheric pressure and room temperature. In particular, the effect of various thermal treatments on the reduction of GO and the subsequent sensing performance is studied. We also include simulation results based on a very simplified model to interpret experimental observations.
2. Experimental details Graphite oxide was synthesized by the oxidative treatment of purified natural graphite using the modified Hummers method [26]. The graphite oxide was then fully exfoliated 2
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 1. Schematic diagram of the reduced GO device. A reduced GO sheet bridges the source and drain electrodes, which closes the circuit. In FET measurements, the back of the Si wafer is used as the gate electrode.
Figure 2. (a) A TEM image of a single GO sheet spanning across a hole formed by the lacey carbon film and the copper bar of a TEM grid. (b) An SEM image showing a GO sheet (∼1 μm × 2 μm) bridging a pair of Au electrodes; the electrode width of and the gap between neighboring electrodes are both about 1 μm.
such as dielectrophoretic assembly [30] could be used to delicately control the arrangement of GO sheets between electrodes, which may lead to adjustable device characteristics and functionality.
GO suspension on holey-carbon-film-covered copper grids (from Ted Pella, Inc.). A field-emission scanning electron microscope (SEM) (Hitachi S 4800) was used to characterize the fabricated GO devices; the SEM had a resolution of 1.4 nm at 1 kV acceleration voltage.
3.1. Reduction of GO via thermal treatments
3. Results and discussion
We first conducted successive multi-step heating to inspect the effect of low-temperature heating on the reduction of GO. The Ids –Vds curves in figure 3(a) present the typical changes in the conductance of GO devices before and after thermal treatments. GO is normally electrically insulating at room temperature with its resistance on the order of tens of G, as indicated by the almost horizontal Ids –Vds curve A in figure 3(a). This insulating nature of GO can be elucidated by the extensive presence of epoxide and hydroxyl groups on both sides of the basal plane and carbonyl and carboxylic groups on the edges of GO [14], leading to the massive existence of sp3 -hybridized carbon atoms. Although there are regions of unsaturated carbon atoms, i.e., nanometer-sized graphitic domains [31], these regions are separated by vast areas of oxidized carbon atoms. GO stays non-conductive unless more clusters of graphitic atoms (sp2 -hybridized) can be restored by
Figure 2(a) is a TEM image that shows a single GO sheet spanning across a hole formed by the holey-carbon-film and the copper bar of a TEM grid. The lateral dimension of the GO sheet is about 4 μm and the sheet is mostly uniform and smooth, except that there are rolls and folds on its edges. More details on the TEM characterization of GO are presented in our previous paper [29]. SEM was used to verify the presence of GO sheets between electrodes after deposition of the GO suspension on the sensor electrode/wafer. An individual GO sheet (∼1 μm × 2 μm) bridging a pair of neighboring Au electrode fingers is shown in the SEM image of figure 2(b). Similar to the TEM observation, rolls and folds can be seen on the GO sheet in the SEM image. The arrangement of GO sheets on the wafer was quite random and no effort has been made to control the layout of GO sheets; however, techniques 3
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 3. (a) Ids –Vds measurement results of a GO device before and after successive multi-step heating. Curve A: without any thermal treatment; curve B: successively annealed in Ar at 100 and 200 ◦ C for 1 h each; curve C: further annealed in Ar at 300 ◦ C for 1 h. (b) Ids –Vg curve (with Vds = 0.01 V) after 100, 200, and 300 ◦ C annealing, showing p-type semiconductivity of reduced GO in an ambient environment. (c) Ids –Vds curves of a GO device before (curve I) and after (curve II) one-step heating in Ar flow at 200 ◦ C for 2 h.
suggest possible ohmic contact between reduced GO sheets and Au electrodes. The resistance of the GO device further decreased by a factor of ∼10 after 300 ◦ C annealing (curve C in figure 3(a)), indicating most likely further reduction of GO rather than contact improvement. However, additional characterization of the contact between GO sheets and Au electrodes is needed before the reduction of resistance, i.e., the exact contribution of the GO reduction, can be fully understood. Both temperature and duration affect the extent of GO powder reduction, as well as the exfoliated material such as GO sheets. Thermal treatments of GO at 1100 ◦ C in vacuum [32] or Ar/H2 flow [20], or at 400 ◦ C with hydrazine vapor pretreatment [32], yielded significantly reduced material. Figure 3(b) shows the transport characteristics of the GO device after the 300 ◦ C annealing. The behavior is similar to that of a p-type semiconductor. Ids decreases with increasing gate voltage (Vg ), implying that the transport through the reduced GO sheets is dominated by positive charge carriers (holes). This p-type semiconducting behavior agrees with the results reported for graphene sheets prepared by micromechanically cleaving graphite [33], and chemically[34] and thermally- [22] reduced GO when exposed to the ambient environment. The p-type behavior is most likely due to the polarization of adsorbed molecules (e.g., water and O2 ) and/or defects introduced on the graphene sheets during the preparation or reduction process [35], or even the influence of the supporting substrate [36, 37]. In the case of defectinduced changes in electronic properties, the p-type nature of graphene was enhanced with increasing structural defects via
removing oxygen functional groups to decrease the distance between graphitic domains, which results in charge transport via variable range hopping [31], or even to create continuous graphitic ‘paths’ for charge transport. After annealing in Ar at 100 ◦ C for 1 h, the resistance of the device only slightly decreased, which agrees with Jung et al [22]. A slight reduction may have occurred during the 100 ◦ C heating treatment; however, the reduction was insufficient to make GO conductive and introduce enough reactive sites on the GO surface for gas adsorption and sensing. With further annealing at 200 ◦ C for 1 h, the device resistance decreased to ∼750 k, as estimated from the Ids –Vds curve B in figure 3(a), implying that GO sheets in the device were partially reduced. Thermal treatments could pyrolyze and remove oxygen functional groups from GO, which recovers graphitic regions and makes GO more conductive. The release of CO, CO2 , and H2 O has been detected during the thermal reduction of GO [24]. Yang et al [23] reported the partial reduction of GO by exposure to an Ar flow at 200 ◦ C for 30 min and found that the carbon-to-oxygen atomic ratio increased from ∼2.8 to ∼3.9 based on x-ray photoelectron spectroscopy (XPS) measurements. Jung et al demonstrated that the electrical conductivity of individual GO sheets significantly increased after being heated in vacuum at only 125–240 ◦ C, indicating a partial reduction of the GO [22]. Of course, the significant reduction in device resistance could also be partly attributed to the improvement in the electrical contact between GO sheets and Au electrodes during the annealing processes. The linearity and symmetry of the Ids –Vds curve 4
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 4. Room-temperature NO2 sensing behavior of a GO device thermally treated with successive multi-step heating. After annealing in Ar at 100 and 200 ◦ C for 1 h each, (a) the GO device shows repeatable response to 100 ppm NO2 ; (b) the sensing signal is highly dependent on the NO2 concentration; and (c) the GO sensor can detect NO2 at a concentration as low as 2 ppm. (d) 300 ◦ C annealing improved both the sensor sensitivity and response time but lengthened the recovery time compared with 200 ◦ C annealing.
plasma exposure [35]. This is analogous to multiwalled CNTs, whose electronic properties could be altered from metallic to p-type semiconducting with structural deformations [38]. Successive multi-step heating reveals the conductance evolution of GO under various heating conditions, which is instrumental for understanding the reduction process; however, it involves multiple heating cycles that may add complexity and thus hinder practical applications of GO. Aiming at future rapid and large-scale production of reduced GO devices, we employed one-step heating (200 ◦ C in Ar for 2 h) to reduce GO and found that partial reduction of GO also can be achieved through this type of heating. Figure 3(c) compares the conductance of a GO device before (curve I) and after (curve II) the one-step heating previously described. The slope of Ids –Vds curve II is significantly higher than that of curve I, indicating elevated conductance of the GO device and effective GO reduction by the one-step heating. As figures 3(a) and (c) show, the conductance of GO devices may vary within a range after thermal treatments, which can be ascribed to the fluctuation in heating conditions (e.g., heating temperature and time) and the variation of the number density of GO sheets in devices.
After being partially reduced by either successive multistep heating at 100 and 200 ◦ C (1 h each) or one-step heating at 200 ◦ C (2 h), the GO devices became highly responsive to NO2 and NH3 , which was most likely due to the recovery of many graphitic carbon atoms as active sites for target gas adsorption. Possibly vacancies or small holes were created during thermal treatment and these defects may also serve as adsorption sites for gaseous molecules [24]. Figure 4(a) shows a typical dynamic response (current versus time) of a GO device for room-temperature detection of 100 ppm NO2 after being successively heated at 100 and 200 ◦ C in Ar for 1 h each. The sensor was periodically exposed to clean dry air flow (2 lpm) for 10 min to record a base value of the sensor conductance, 100 ppm NO2 diluted in air (2 lpm) for 15 min to register a sensing signal, and clean air flow (2 lpm) again for 25 min to recover the device. Upon the introduction of NO2 , the sensor current went up, i.e., the conductance of the sensor increased; when the NO2 flow was turned off and the air flow restored, the device re-established its conductance in about 30 min. Three cycles were repeated in figure 4(a) and the signal was fairly reproducible. The sensing signal strength (proportional to the spike height with NO2 on) was dependent on the NO2 concentration (shown in figure 4(b)) as it decreased with decreasing NO2 concentrations from 100, to 50, and to 25 ppm. Figure 4(c) shows the sensor response to 2 ppm NO2 , where the conductance increased ∼12% with 40 min NO2 exposure. Assuming a linear relationship between conductance change and NO2 concentration, this sensitivity is comparable to the sensor based on mechanically-cleaved graphene, which showed ∼4.3% increment in the conductance for 1 ppm NO2 [5]. The sensing performance of the devices
3.2. Gas sensing with reduced GO devices Devices with as-deposited GO sheets showed no response to 100 ppm NO2 or 1% NH3 , indicating insignificant change in the electrical transport property of the non-reduced GO. We also found that 1 h 100 ◦ C heating was usually inadequate to make GO devices responsive to gases. 5
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 5. (a) Dynamic sensing behavior of a reduced GO device (successively heated in Ar at 100, 200, and 300 ◦ C for 1 h each) for 1% NH3 detection at room temperature. (b) The recovery process of the device after NH3 sensing was extremely slow. After 50 h air flow, the GO device still did not restore its initial conductance.
reported here is very encouraging for practical applications when considering the simplicity and low cost to fabricate these devices and the potential opportunities for optimization. The sensing performance of the reported device is attributed to the effective adsorption of NO2 on the surface of reduced, p-type GO. NO2 is a strong oxidizer with electron-withdrawing power [39]; therefore, electron transfer from reduced GO to adsorbed NO2 leads to an enriched hole concentration and enhanced electrical conduction in the reduced GO sheet. Figure 4(d) compares the sensitivities of the device for 100 ppm NO2 detection after 200 and 300 ◦ C annealing. The sensor sensitivity is evaluated as the ratio of (G g –G a )/G a , where G a is the sensor conductance in clean air and G g is the sensor conductance in air containing 100 ppm NO2 . With 300 ◦ C annealing, the device showed a sensitivity of ∼1.56 to 100 ppm NO2 , higher than that (∼1.41) after 200 ◦ C annealing. In addition, it had a faster response when exposed to NO2 , as evidenced by a steeper slope upon the exposure to NO2 . This accelerated response is likely due to the creation of more graphitic carbon atoms during the 300 ◦ C annealing because molecular adsorption onto sp2 -bonded carbon (lower binding energy required) is faster than onto defects [6]. However, the sensor recovery after 300 ◦ C annealing became slower because the device did not return to its initial conductance after 30 min exposure to dry air, whereas in the 200 ◦ C annealing case the full recovery was achieved under the same exposure condition. Low-temperature heating and UV illumination could be used to accelerate the sensor recovery. The GO devices responded to NH3 as well, after either successive thermal treatments or one-step heating. Figure 5 is the 1% NH3 sensing data obtained from the same reduced GO device (after 300 ◦ C heating), whose NO2 sensing behavior is presented in figure 4. Upon NH3 exposure, the current passing through the device decreased (figure 5(a)), possibly because adsorption of NH3 (electron donator) can lower the hole concentration in GO, and thus reduce GO conductance. Unlike the relatively fast recovery (about 30 min) after NO2 exposure, the conductance of the GO device could not return to its initial value in air flow (2 lpm) after about 50 h (shown in figure 5(b)). Most of the GO sensors (with one exception to be discussed) recovered very slowly after NH3 sensing.
Further investigation is needed to understand the intrinsic mechanisms associated with the slow recovery of reduced GO from NH3 exposure and to find effective measures to accelerate the recovery process before reduced GO can be used practically for repeatable NH3 detection. Among all of the GO devices fabricated (11 in total), only one device recovered exceptionally fast after NH3 sensing. Figure 6(a) shows three NH3 sensing cycles using this device after it was reduced through one-step heating in Ar at 200 ◦ C for 2 h. This NH3 sensing performance is superior to that shown in figure 5(a); however, an unusual current increase was observed at the beginning of NH3 exposure in each cycle (indicated by the arrows in figure 6(a)), which is against the GO sensing mechanism discussed previously and was not found in other devices. Figure 6(b) shows a magnified view of the area marked by the leftmost arrow in figure 6(a), showing that the current increased with a very sharp slope at the beginning of NH3 exposure. This abnormal increase may be attributed to the unsteadiness of flow field in the chamber during gas switching or some other accidental noise; however, even more strange behavior was observed from this device when using it to sense 1% NH3 again after ∼2 months. As shown in figure 6(c), the device gave a sensing signal (curve I) with a completely opposite trend (i.e., its conductance increased upon NH3 exposure) after 2 months, compared with that obtained freshly after reduction (curve II, obtained by normalizing the curve in figure 6(a)). The current ( I ) is normalized to the initial value ( I0 ) in air flow for the convenience of comparison. This observation suggests competing sensing mechanism(s) may exist besides gas adsorption/desorption on reduced p-type GO for this device. The device was then re-treated in Ar at 200 ◦ C for 2 h; although it then restored to ‘normal’ sensing behavior (conductance decreased upon NH3 exposure), as shown by curve III in figure 6(c), its sensitivity degraded. Figure 6(d) shows the Ids –Vds curve of the device measured 2 months after the first thermal treatment. The curve is asymmetric and nonlinear, which is contrary to the one acquired shortly after the thermal treatment (inset of figure 6(d)) and suggests a non-ohmic contact between the GO and Au electrodes. This contact issue may have played a role in the abnormal NH3 sensing behavior. For instance, Peng et al proposed that the Schottky barrier modulation at the CNT– metal contact significantly contributed to NH3 sensing when 6
Nanotechnology 20 (2009) 445502
G Lu et al
Figure 6. (a) 1% NH3 detection by a GO device treated with one-step heating (in Ar at 200 ◦ C for 2 h). This device recovered faster compared with the one presented in figure 5. (b) Upon exposure to NH3 , the conductance of the device first oddly increased for a few seconds (this increase cannot be explained by a NH3 -adsorption-induced change in GO conductance) before it decreased (as expected). (c) The sensing response for 1% NH3 after 2 months (curve I) is opposite (i.e., the conductance increased upon NH3 exposure) to that observed after freshly heated (curve II, obtained by normalizing the curve in a). The sensor response returned to normal after another heating treatment (in Ar at 200 ◦ C for 2 h) but the sensitivity degraded, as indicated by curve III. (d) The Ids –Vds curve of the device became asymmetric and nonlinear, implying non-ohmic contact between the GO and Au electrodes; the inset shows the Ids –Vds obtained freshly after the first heating.
where G(t) is the change in the conductance of the sensor; G max is the maximum change in the device conductance with sufficiently long gas exposure; c is the gas concentration; K is the binding equilibrium constant (assuming Langmuir adsorption); and k is the surface reaction rate (depending on the properties of both the sensing element and analyte molecules). Note that the sensing process starts at t = 0 for equation (1). If we only focus on the trend of sensing responses, neglect the details of all constants, and take into account that I (change in the measured current I of the sensor) is proportional to G , we could obtain a very simplified expression for normalized change in the measured current I (t)/I0 as,
using CNT sensors at room temperature [9]. Therefore, the influence of graphene–metal contact on the sensor performance is worthy of future study. We also could not exclude the connection between the abnormal NH3 sensing behavior and the unusually fast recovery. Understanding this odd NH3 sensing behavior may lead to measures that can be used to engineer sensing device properties, e.g., recovery rate and gas selectivity. 3.3. Simplified modeling of strange NH3 sensing behavior We have developed a very simplified model to shed light on the unusual NH3 sensing behavior presented in figure 6 and hope to inspire more in-depth analysis in the future. We assume that there are two NH3 sensing mechanisms ( M 1 and M 2) that can lead to competing responses from the device (i.e., M 1 leads to increased conductance and M 2 leads to decreased conductance; see figure 7) and that the overall sensing behavior is simply the linear combination of the two mechanisms. Due to the close relation between the CNT and the graphene, it could be a reasonable starting point to evaluate GO sensors with existing models developed for CNT sensors. Taking advantage of the CNT sensing model proposed by Lee et al [40], the GO sensing response due to gas molecule adsorption could be written as cK , G(t) = G max (1) 1 + cK 1 − exp − 1+cK kt K
I (t)/I0 = A exp(−Ct) + D,
(2)
where I0 is the current measured before the NH3 exposure; A, C , and D are constants determined by factors such as the sensor properties, gas concentration, and properties of target gas molecules. Assigning suitable values to A, C , and D , we can obtain equations to describe M 1 and M 2 and use them to simulate the experimental results presented in figure 6. As shown in figure 7, we assume that M 1 causes the device conductance to increase upon the NH3 exposure and that the normalized change in the measured current I (t)/I0 due to M 1 can be expressed as,
M 1: I (t)/I0 = 5 − 4 exp[−0.033 · (t − 600)]. 7
(3)
Nanotechnology 20 (2009) 445502
G Lu et al
and the metal electrode makes the sensing response more complex and deserves further investigation. The simple and low-cost manufacturing process and the wide availability of GO could lead to cost-effective graphene-based gas sensors and other opportunities for graphene.
Acknowledgments This work was financially supported by the NSF (CMMI0900509 and CBET-0803142) and by a catalyst grant from the University of Wisconsin-Milwaukee Research Foundation. The authors thank R S Ruoff and D A Dikin for providing GO suspensions, M Gajdardziska-Josifovska for providing TEM access, and D Rosenmann for assistance in metal sputtering. The SEM imaging was conducted at the UWM Electron Microscope Laboratory. The e-beam lithography was performed at the Center for Nanoscale Materials of Argonne National Laboratory, which is supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357.
Figure 7. Simulation of the observed abnormal NH3 sensing response using a simplified model. M 1 causes the conductance to increase while M 2 causes the conductance to decrease upon NH3 exposure. Curve I and curve II are experimental data (the first cycles of corresponding curves shown in figure 6(c)). Linear combinations of M 1 and M 2 could be used to qualitatively explain observed NH3 sensing responses.
References (Note that the NH3 exposure starts at t = 600 s.) Similarly, we could use the following equation to represent the sensing signal resulting from M 2, which decreases the conductance,
M 2: I (t)/I0 = exp[−0.012(t − 600)].
[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 Science 306 666–9 [2] Frank I W, Tanenbaum D M, Van der Zande A M and McEuen P L 2007 J. Vac. Sci. Technol. B 25 2558–61 [3] Balandin A A, Ghosh S, Bao W Z, Calizo I, Teweldebrhan D, Miao F and Lau C N 2008 Nano Lett. 8 902–7 [4] Geim A K and Novoselov K S 2007 Nat. Mater. 6 183–91 [5] Schedin F, Geim A K, Morozov S V, Hill E W, Blake P, Katsnelson M I and Novoselov K S 2007 Nat. Mater. 6 652–5 [6] Robinson J T, Perkins F K, Snow E S, Wei Z Q and Sheehan P E 2008 Nano Lett. 8 3137–40 [7] Iijima S 1991 Nature 354 56–8 [8] Baughman R H, Zakhidov A A and de Heer W A 2002 Science 297 787–92 [9] Peng N, Zhang Q, Chow C L, Tan O K and Marzari N 2009 Nano Lett. 9 1626–30 [10] Berger C et al 2006 Science 312 1191–6 [11] Reina A, Jia X, Ho J, Nezich D, Son H, Bulovic V, Dresselhaus M S and Kong J 2009 Nano Lett. 9 30–5 [12] Li X et al 2009 Science 324 1312–4 [13] Ruoff R 2008 Nat. Nanotechnol. 3 10–1 [14] Cai W W et al 2008 Science 321 1815–7 [15] Lerf A, He H Y, Forster M and Klinowski J 1998 J. Phys. Chem. B 102 4477–82 [16] He H Y, Klinowski J, Forster M and Lerf A 1998 Chem. Phys. Lett. 287 53–6 [17] Park S and Ruoff R S 2009 Nat. Nanotechnol. 4 217–24 [18] Stankovich S, Piner R D, Chen X Q, Wu N Q, Nguyen S T and Ruoff R S 2006 J. Mater. Chem. 16 155–8 [19] Muszynski R, Seger B and Kamat P V 2008 J. Phys. Chem. C 112 5263–6 [20] Wang X, Zhi L J and Mullen K 2008 Nano Lett. 8 323–7 [21] Williams G, Seger B and Kamat P V 2008 ACS Nano 2 1487–91 [22] Jung I, Dikin D A, Piner R D and Ruoff R S 2008 Nano Lett. 8 4283–7 [23] Yang D et al 2009 Carbon 47 145–52 [24] Jung I, Dikin D, Park S, Cai W, Mielke S L and Ruoff R S 2008 J. Phys. Chem. C 112 20264–8
(4)
Normally, M 2 is dominant in GO sensors for detecting NH3 while M 1 is negligible, which explains the typical NH3 sensing behavior of the GO sensors shown in figure 5(a); however, an abnormal sensing response could appear when M 1 cannot be neglected, even if M 2 still dominates. For instance, if the total signal consists of mostly M 2 (e.g., 87%) and a small contribution from M 1 (e.g., 13%), the combination (figure 7) fits the first cycle of the experimental curve II in figure 6(c) fairly well; I (t)/I0 first increases with the start of NH3 flow and then decreases. If the contribution of M 1 increases to 83.8% due to certain unidentified reasons (e.g., enhanced GO-contact influence or degraded GO), the simulated result matches with the curve I (from figure 6(c)), which is against the expected response from GO sensors. Of course, the model presented here is overly simplified and more sophisticated simulations are desired for the in-depth theoretical analysis.
4. Conclusion In summary, GO was partially reduced via thermal treatments in Ar flow with temperatures as low as 200 ◦ C. After the treatment, the electrical conductance of the GO sheets increased due to the removal of oxygen functional groups. The thermally-reduced GO demonstrated transport characteristics typical of a p-type semiconductor, which can be used to fabricate molecular adsorption-type gas sensors. Miniaturized gas sensors based on the reduced GO were fabricated and exhibited room-temperature sensing properties under atmospheric pressure. The electrical contact between the GO 8
Nanotechnology 20 (2009) 445502
G Lu et al
[32] Becerril H A, Mao J, Liu Z, Stoltenberg R M, Bao Z and Chen Y 2008 ACS Nano 2 463–70 [33] Romero H E, Shen N, Joshi P, Gutierrez H R, Tadigadapa S A, Sofo J O and Eklund P C 2008 ACS Nano 2 2037–44 [34] Gilje S, Han S, Wang M, Wang K L and Kaner R B 2007 Nano Lett. 7 3394–8 [35] Kim K, Park H J, Woo B C, Kim K J, Kim G T and Yun W S 2008 Nano Lett. 8 3092–6 [36] Adam S, Hwang E H, Galitski V M and Das Sarma S 2007 Proc. Natl Acad. Sci. USA 104 18392–7 [37] Fratini S and Guinea F 2008 Phys. Rev. B 77 195415 [38] Martel R, Schmidt T, Shea H R, Hertel T and Avouris P 1998 Appl. Phys. Lett. 73 2447–9 [39] Leenaerts O, Partoens B and Peeters F M 2008 Phys. Rev. B 77 125416 [40] Lee C Y and Strano M S 2005 Langmuir 21 5192–6
[25] Lu G H, Ocola L E and Chen J H 2009 Appl. Phys. Lett. 94 083111 [26] Park S, An J, Piner R D, Jung I, Yang D, Velamakanni A, Nguyen S T and Ruoff R S 2008 Chem. Mater. 20 6592–4 [27] Lu G H, Huebner K L, Ocola L E, Gajdardziska-Josifovska M and Chen J H 2006 J. Nanomater. 2006 60828 [28] Lu G H, Ocola L E and Chen J H 2009 Adv. Mater. 21 2487–91 [29] Lu G H, Mao S, Park S, Ruoff R S and Chen J H 2009 Nano Res. 2 192–200 [30] Vijayaraghavan A, Sciascia C, Dehm S, Lombardo A, Bonetti A, Ferrari A C and Krupke R 2009 ACS Nano 3 1729–34 [31] Gomez-Navarro C, Weitz R T, Bittner A M, Scolari M, Mews A, Burghard M and Kern K 2007 Nano Lett. 7 3499–503
9
