Report

Nanotube Molecular Wires as Chemical Sensors

See allHide authors and affiliations

Science  28 Jan 2000:
Vol. 287, Issue 5453, pp. 622-625
DOI: 10.1126/science.287.5453.622

Abstract

Chemical sensors based on individual single-walled carbon nanotubes (SWNTs) are demonstrated. Upon exposure to gaseous molecules such as NO2 or NH3, the electrical resistance of a semiconducting SWNT is found to dramatically increase or decrease. This serves as the basis for nanotube molecular sensors. The nanotube sensors exhibit a fast response and a substantially higher sensitivity than that of existing solid-state sensors at room temperature. Sensor reversibility is achieved by slow recovery under ambient conditions or by heating to high temperatures. The interactions between molecular species and SWNTs and the mechanisms of molecular sensing with nanotube molecular wires are investigated.

Carbon nanotubes are molecular-scale wires with high mechanical stiffness and strength. A SWNT can be metallic, semiconducting, or semimetallic, depending on its chirality (1). Utilization of these properties has led to applications of individual nanotubes or ensembles of nanotubes as scanning probes (2, 3), electron field emission sources (4), actuators (5), and nanoelectronic devices (6). Here, we report the realization of individual semiconducting-SWNT (S-SWNT)–based chemical sensors capable of detecting small concentrations of toxic gas molecules.

Sensing gas molecules is critical to environmental monitoring, control of chemical processes, space missions, and agricultural and medical applications (7). The detection of NO2, for instance, is important to monitoring environmental pollution resulting from combustion or automotive emissions (8). Detection of NH3 is needed in industrial, medical, and living environments (9). Existing electrical sensor materials include semiconducting metal oxides (7–9), silicon devices (10, 11), organic materials (12,13), and carbon black–polymer composites (14). Semiconducting metal oxides have been widely used for NO2and NH3 detection (7–9). These sensors operate at high temperatures (200° to 600°C) in order to achieve enhanced chemical reactivity between molecules and the sensor materials for substantial sensitivity (7). Conducting polymers (12) and organic phthalocyanine semiconductors (12,13) have also been investigated for NO2 or NH3 sensing. The former exhibit limited sensitivity (12), whereas the latter tend to have very high resistivity (sample resistance of >10 gigohms) (13). In this report, we show that the electrical resistance of individual semiconducting SWNTs change by up to three orders of magnitude within several seconds of exposure to NO2 or NH3molecules at room temperature. Miniaturized chemical sensors based on individual SWNTs are thus demonstrated. Furthermore, we combine theoretical calculations with experiments to address the underlying fundamental question regarding how molecular species interact with nanotubes and affect their electrical properties.

Semiconducting SWNTs are chiral (m, n) tubes with mn ≠ 3 × integer. The band gap E g of an S-SWNT scales with its diameter d as E g ∼ 1/d (E g ∝ 0.5 eV ford ∼ 1.4 nm) (1). It was previously found that when two metal contacts were used to connect an S-SWNT, the metal/S-SWNT/metal system exhibits p-type transistor characteristics with several orders of magnitude change in conductance under various gate voltages (6, 15, 16). Our nanotube chemical sensors were based on these S-SWNT transistors, obtained by controlled chemical vapor deposition growth of individual SWNTs from patterned catalyst islands on SiO2/Si substrates (Fig. 1A) (16, 17). Gas-sensing experiments were carried out by placing an S-SWNT sample in a sealed 500-ml glass flask with electrical feedthrough and flowing diluted NO2 [2 to 200 parts per million (ppm)] or NH3 (0.1 to 1%) in Ar or air (flow rate of 700 ml/min) through the flask while monitoring the resistance of the SWNT.

Figure 1

Changes of electrical characteristics of a semiconducting SWNT in chemical environments. (A) Atomic force microscopy image of a metal/S-SWNT/metal sample used for the experiments. Nanotube diameter is ∼1.8 nm. The metal electrodes consist of 20-nm-thick Ni, with 60-nm-thick Au on top. (B) Current versus voltage curves recorded before and after exposure to NH3. (C) Current versus voltage curves recorded under V g = +4 V, before and after NO2 exposure.

We observed that the conductance of S-SWNT samples can be substantially increased or decreased by exposure to NO2or NH3. A current versus voltage (I-V) curve recorded with an S-SWNT sample after a 10-min exposure to NH3 showed an ∼100-fold conductance depletion (Fig. 1B). Exposure to NO2 molecules increased the conductance of the SWNT sample by about three orders of magnitude (Fig. 1C) when the SWNT sample was initially depleted by a back-gate voltage (V g) of +4 V (6,15, 16). The SWNT is a hole-doped semiconductor, as can be gleaned from the current versus gate voltage (I-V g) curve shown in Fig. 2 (middle curve), where the conductance of the SWNT is observed to decrease by three orders of magnitude under positive gate voltages (6, 15, 16). TheI-V g curve recorded after the S-SWNT sample was exposed to NH3 exhibits a shift of −4 V (Fig. 2, left curve). In contrast, theI-V g curve was shifted by +4 V after NO2 exposure (Fig. 2, right curve). The low resistance (∼360 kilohms) of the SWNT under zero gate voltage suggests substantial hole carriers existing in the p-type nanotube at room temperature. Exposure to NH3 effectively shifts the valence band of the nanotube away from the Fermi level, resulting in hole depletion and reduced conductance. For the NO2 case, exposure of the initially depleted sample to NO2 resulted in the nanotube Fermi level shifting closer to the valence band. This caused enriched hole carriers in the nanotube and enhanced sample conductance. These results show that molecular gating effects are capable of shifting the Fermi level of S-SWNTs and modulating the resistance of the sample by orders of magnitude.

Figure 2

Chemical gating effects to the semiconducting SWNT. Current versus gate voltage curves before NO2 (circles), after NO2(triangles), and after NH3 (squares) exposures. The measurements with NH3 and NO2 were carried out successively after sample recovery.

The conductance of the SWNT sample increased sharply by about three orders of magnitude after 200 ppm of NO2 was introduced (Fig. 3A). We investigated five S-SWNT samples and found that the response times (defined as time duration for resistance change by one order of magnitude) of the samples to 200 ppm of NO2 were in the range of 2 to 10 s. The sensitivity [defined as the ratio between resistance after (R after) and before (R before) gas exposure] is ∼100 to 1000. After the NO2 flow was replaced by pure Ar, the conductance of the SWNT samples was observed to slowly recover, and the typical recovery time was ∼12 hours. This suggests slow molecular desorption from the nanotube sample and that the SWNT chemical sensors can be reversibly used. Heating the exposed sample in air at 200°C led to recovery in ∼1 hour. For comparison, a high-performance metal oxide sensor (Cd-doped SnO2) operates at 250°C for detecting 100 ppm of NO2 with a response time of ∼50 s, a recovery time of ∼8 min, and a sensitivity of ∼300 (8, 18). A polypyrole-conducting polymer sensor can detect 0.1% NO2by an ∼10% resistance change in ∼5 to 10 min at room temperature (12). Thus, the S-SWNT sensors have the advantage of room temperature operation with sensitivity up to 103 over these materials.

Figure 3

Electrical response of a semiconducting SWNT to gas molecules. (A) Conductance (underV g = +4 V, in an initial insulating state) versus time in a 200-ppm NO2 flow. (B) Data for a different S-SWNT sample in 20- and 2-ppm NO2flows. The two curves are shifted along the time axis for clarity. (C) Conductance (V g = 0, in an initial conducting state) versus time recorded with the same S-SWNT sample as in (A) in a flow of Ar containing 1% NH3. (D) Data recorded with a different S-SWNT sample in a 0.1% NH3 flow. Read 3.5e-7, for example, as 3.5 × 10−7.

NH3-sensing results were obtained with the same SWNT sample after recovery from NO2 detection (Fig. 3C). The conductance of the SWNT sample was observed to decrease ∼100-fold after exposure to a 1% NH3 flow. The response times to 1% NH3 for five S-SWNT samples were ∼1 to 2 min, and the sensitivity was ∼10 to 100. For comparison, metal oxide NH3 sensors typically operate at 300° to 500°C, with a response time of ∼1 min and a sensitivity of ∼1 to 100 toward 200 ppm to 1% NH3 (8). Conducting polymer sensors can detect 1% NH3 with a response time of ∼5 to 10 min by an ∼30% resistance change at room temperature (12).

For the S-SWNT samples, lowering the NO2 concentration to 20 and 2 ppm led to response times of ∼0.5 to 1 min and ∼5 min, respectively (Fig. 3B). Lowering the concentration of NH3to 0.1% led to a response time of ∼10 min (Fig. 3D). Thus, for detecting an ∼10-fold resistance change of individual S-SWNT samples within minutes of gas exposure, the lower concentration limit is ∼2 ppm for NO2 and ∼0.1% for NH3. Similar sensing results were obtained when Ar or air was used as the carrier gas. This suggests that NH3 or NO2 dominates the response of the SWNT samples over molecules in the ambient environment. Over time, repeated sensing and recovery experiments with the S-SWNT samples obtained reproducible results.

To understand the chemical gating effects and the nanotube gas-sensing mechanism, we first considered the fact that S-SWNT samples appear to be hole doped (p-type) before the molecular sensing experiments. Hole doping in S-SWNTs has been observed by several groups (6, 15, 16). Possible hole-doping mechanisms include metal electrode–tube work function mismatch (6) and electrostatic effects due to charged species existing on the SiO2 surface or bulk (19). Because our nanotubes are long (>3 μm), we suggested a hole-doping mechanism (for example, charged chemical groups on SiO2) operating throughout the nanotube length. As a result of the hole doping, the Fermi level of an S-SWNT is typically located at ∼25 meV above the valence band (19), which is responsible for the observed conductance of S-SWNT samples at room temperature (typical resistance of 300 kilohms to 5 megohms). Next, we considered the chemical nature of the molecules. NO2 has an unpaired electron and is known as a strong oxidizer. Upon NO2 adsorption, charge transfer is likely to occur from an SWNT to NO2 because of the electron-withdrawing power of the NO2 molecules. NH3, on the other hand, is a Lewis base with a lone electron pair that can be donated to other species. However, it is necessary to investigate whether these qualitative pictures represent the correct mechanisms of molecular sensing with SWNTs.

We carried out first-principles calculations on molecule-SWNT complexes using density functional theory (20). NO2 is found to bind with a semiconducting (10, 0) tube with an adsorption energy E a ∼ 0.9 eV (18.6 kcal/mol) and 0.1 electron charge transfer from the tube to a NO2molecule. Charge transfer from the nanotube to NO2 should be the mechanism for increased hole carriers in an S-SWNT and enhanced conductance. For the NH3-SWNT system, calculations found no binding affinity between NH3 molecules and the (10, 0) tube. We suggest two possible indirect routes through which NH3 molecules may affect S-SWNTs. The first is that NH3 binds to hydroxyl groups on the SiO2substrate (21), which could partially neutralize the negatively charged groups on the SiO2 surface and lead to positive electrostatic gating to the S-SWNT. Second, interactions may exist between NH3 molecules and an SWNT through other species. It was previously found that NH3 can interact strongly with adsorbed oxygen species on graphite (22). Preadsorbed oxygen species on a nanotube could interact with NH3 and affect its electrical properties. These possible mechanisms require further experimental and theoretical investigations.

We also investigated the electrical properties of metallic SWNTs in various chemical environments. A metallic tube was identified by small changes in the conductance with gate voltage (a typical resistance of ∼20 to 200 kilohms) (16). We found that, for a typical metallic SWNT, exposure to NO2 or NH3 increased or decreased, respectively, the conductance of the sample by ≤30%. The explanation for these small changes is that, for a metallic SWNT, small shifts of the Fermi level do not result in a substantial change in the density of states at the Fermi level and, thus, in the charge carriers in the nanotube.

The interactions between NO2 and NH3 with graphite have been previously investigated (22–25). For SWNTs, molecular interaction effects have been studied in the case of Br and I intercalation with bulk samples of SWNT ropes (26,27). The intercalation leads to substantially enhanced sample conductance (26, 27). Our report is concerned with molecular interactions with individual semiconducting and metallic SWNTs. We have also investigated the effects of NO2 and NH3 on the electrical properties of mats of SWNT ropes made from as-grown laser ablation materials. In a 200-ppm NO2 flow, the resistance of an SWNT mat is found to decrease from R = 150 to 80 ohms (R before/R after ∼ 2) in ∼10 min (Fig. 4A). In a 1% NH3 flow, the resistance of a second SWNT mat increases from 120 to 170 ohms (R after/R before ∼ 1.5) in ∼10 min (Fig. 4B). In these bulk SWNT samples, the molecular interaction effects are averaged over metallic and semiconducting tubes. Also, the inner tubes in SWNT ropes are blocked from interacting with NO2 and NH3 because the molecules are not expected to intercalate into SWNT ropes. This explains the small resistance change of bulk SWNT mats by gas exposure compared to that of an individual S-SWNT.

Figure 4

Electrical response of bulk SWNT mats to NO2 and NH3 molecules. (A) Conductance versus time data recorded with an SWNT mat in 200 ppm of NO2. (B) Conductance versus time recorded with an SWNT mat in 1% NH3.

The main feature of individual S-SWNT sensors, besides their small sizes, is that they operate at room temperature with sensitivity as high as 103. An individual nanotube sensor can be used to detect different types of molecules. The selectivity is achieved by adjusting the electrical gate to set the S-SWNT sample in an initial conducting or insulating state. The fast response of a nanotube sensor can be attributed to the full exposure of the nanotube surface area to chemical environments. Thus, nanotube molecular wires should be promising for advanced miniaturized chemical sensors.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: hdai{at}chem.stanford.edu

REFERENCES AND NOTES

View Abstract

Navigate This Article