|Home | About | Journals | Submit | Contact Us | Français|
This article provides a comprehensive review of recent (2008 and 2009) progress in gas sensors based on semiconducting metal oxide one-dimensional (1D) nanostructures. During last few years, gas sensors based on semiconducting oxide 1D nanostructures have been widely investigated. Additionally, modified or doped oxide nanowires/nanobelts have also been synthesized and used for gas sensor applications. Moreover, novel device structures such as electronic noses and low power consumption self-heated gas sensors have been invented and their gas sensing performance has also been evaluated. Finally, we also point out some challenges for future investigation and practical application.
Semiconducting metal oxides have been known for decades to be good gas sensing materials. Ethanol sensors based on SnO2 thick films have been commercialized for years. In 1991, Yamazoe demonstrated that reduction in crystal size would significantly increase the sensor performance . This is because nanosized grains of metal oxides are almost depleted of carriers (most carriers are trapped in surface states) and exhibit much poorer conductivity than microsized grains in ambient air, hence, when exposed to target gases, they exhibit greater conductance changes as more carriers are activated from their trapped states to the conduction band than with microsized grains. Thus, the technological challenge moved to the fabrication of materials with nanocrystals which maintained their stability over long-term operation at high temperature . The exploration of one-dimensional (1D) oxide nanostructures has been stimulated and facilitated by the convenience of obtaining large amounts of single crystalline nanowires/nanobelts via the vapor transport  and vapor-liquid-solid (VLS) methods . The Sberveglieri  and Yang  groups initiated the investigation of gas sensing properties of SnO2 nanobelts. Sberveglieri et al. demonstrated the use of SnO2 nanowires as sensor materials showing prominent current changes towards ethanol and CO, respectively, in a synthetic air environment , while Yang et al. demonstrated the first photochemical NO2 nanosensors (based on individual SnO2 nanoribbons) operating at room temperature . In 2004, our group reported high-performance ZnO nanowire sensors with a low detection limit of 1 ppm ethanol at 300 °C . Ever since then the number of reports on gas sensors based on 1D metal oxide nanostructures have been growing exponentially every year.
The article presents a comprehensive perspective on research efforts made in gas sensor fabrication and testing based on one-dimensional metal oxide nanostructures in recent years, during which electrospun oxide nanofibers have been gaining attention in gas sensor applications [8-15]. The data on film type nanosensors reviewed in this article are mainly published in 2008 and 2009. In Section 2, gas sensor configurations and measurements, performance parameters, as well as theoretical fundamentals of gas sensors based on 1D nanostructures will be introduced, while the material system, cited are focused on undoped metal oxide nanowires or nanobelts. Section 3 features modified 1D metal oxide nanostructures, as well as heterostuctures. In Section 4, novel gas sensors based on novel operation principles such as the “electronic nose”, the self-heated gas sensor and the optical gas sensor will be elaborated on. Finally, Section 5 summarizes the whole article and indicates possible future developments in one-dimensional metal oxide nanosensors.
Up to now, one-dimensional metal oxide nanostructures sensors have been characterized in three ways: conductometric, field effect transistor (FET) and impedometric ones. Conductometric sensors are based on resistance changes caused by exposure of the sensor surface to a target species. So far, two types of conductometric nanowire gas sensors have been mainly fabricated: one is the film type, in which a film composed of nanowires is contacted by pairs of metal electrodes on a substrate (Figure 1a) or a ceramic tube (Figure 1b); the other is the single nanowire type in which a single nanowire bridges two metal electrodes on a heavily doped silicon substrate covered with SiO2 acting as insulating layer between the nanowire/electrode combinations and the conducting silicon (Figure 2). In the fabrication of film type nanosensors, nanowires products are usually pulverized to a pulp state and either directly painted or screen-painted  onto the substrates or tubes. But other approaches are reported. Sometimes nanowire growth is integrated into device fabrication [17-20]: SiO2/Si substrates with patterned metal coatings were used to catalyze the growth of the metal oxide nanowires and the coating also acts as electrodes contacting the sensing material. This type of sensor has lower contact resistance compared to the previous one because the nanowire growth process is integrated into device fabrication. Well-aligned nanowire arrays have been fabricated into nanosensors to explore benefits brought about by orderliness.
For instance, Figure 3 shows the sensor device structure that has been adopted in exploring gas sensing properties of ZnO nanorods , in which ZnO nanorod arrays are sandwiched between the silicon substrate supporting their growth and an indium thin film that forms ohmic contact with the nanorods and the copper electrode. Another interesting approach to assemble nanosensors involves combining two vertically aligned CuO nanowire arrays into a three-dimensional nanostructure, in which two pieces of nanowire arrays were attached to the copper plate and the micromanipulator tip (copper wire) and the distance between the two arrays can be adjusted. The sensor built in this style was reported to be capable of detecting air-diluted H2S at the parts per billion level . However, these aligned nanowire arrays are grown by “bottom up” methods, and the orderliness is not as good as those fabricated via “top to bottom” approaches. Francioso and Son et al. employed microelectronics processes such as photolithography and plasma etching (demonstrated in Figure 4) to fabricate TiO2  and ZnO  parallell nanowire arrays, respectively, and investigated their gas sensor behaviors. Son et al.  developed an alternative technique to construct well-aligned nanowire arrays, which takes advantage of the two facts: 1) the step edges of terraces are energetically favorable for the nucleation of adatoms; 2) a low deposition rate is favorable for producing localized nucleation at the step edges of terraces. They prepared uniform terraces on (0001) sapphire substrates by annealing a miscut sapphire substrate and minimized the ZnO deposition rate by using low laser pulse repetition rate as well as a shadow mask which blocks direct ZnO plume generated by laser ablation pulsed laser deposition (PLD).
Basically, the working principle of a FET type sensor is that the species adsorbed onto the channel surface can work as an extra virtual gate bias and hence cause changes in the apparent threshold voltage. However, the FET configuration of a sensor does not guarantee it work in the simple ideal way. For example, Andrei et al.  discovered that their SnO2 nanobelt FET can be modeled as two Schottky diodes connected back-to-back with a series resistance from the nanobelt separating the diodes and only work as a FET in the presence of hydrogen. Another interesting phenomenon observed by Zhang et al. is that the gate effect typical of a FET was substantially weakened when their In2O3 nanowire transistors were exposed to high concentrations of NH3 (10%) . They proposed that these NH3 molecules residing on the nanowire surface can be charged and discharged by sweeping the gate bias and hence effectively work as charge traps screening the electric field induced by the gate bias.
Impedometric sensors are based on impedance changes and are operated under alternating voltage upon exposure to target species. Like conductometric sensors, there are two types of impedometric ones: film type  and single nanowire type . But this group of sensors has not attracted as much attention as the conductometric sensor yet. The rest of Section 2 will be mainly focused on conductometric sensors.
Semiconducting oxides generally owe their conductivity to their deviation from stoichiometry. Defects such as interstitial cation or anion vacancies also play an important role in their conductivity. Target species can be classified into two groups: oxidizing gases or electron acceptors such as NO2, which produce a decrease in the conductance of n-type semiconducting materials (i.e., electrons are the major carriers, such as ZnO, SnO2, In2O3) and an increase in the conductance of p-type semiconducting materials (i.e. holes are the major carriers, such as CuO); reducing gases or electron donors such as H2S, CO, H2 and water vapor, which act in a reverse manner. Interestingly, sensors based on TiO2 nanofibers have recently been reported not to undergo conductance decrease towards oxidizing gas NO2 as they normally do .
As shown in Figure 5, at higher concentration ranges the device resistance plummeted before NO2 stimulus was removed. The authors attributed such an abnormal response behavior to the conduction type inversion (n-to-p) of the sensing material whose conduction is surface-trap limited, owing to the high surface-to-volume ratio of this material.
There are two types of adsorption: physisorption, the first step of the association of the gas species with the sensor surface, and chemisorptions, which involves exchange of electrons between the adsorbed species and the material surface. The major difference between these two processes is that physisorption is exothermic while chemisorption is endothermic, precisely an activated process whose activation energy can be supplied by thermal or non-equilibrium ones such as illumination . This leads to the fact that physisorption predominates in low temperature range whereas chemisorption dominates in higher temperature range. The sensing characteristics of metal oxides are widely considered to be related with chemisorbed oxygen and water, which can act as intermediates catalyzing the charge transfer processes between gas species and the bulk and which complicates the study of gas sensing mechanisms. The major way they interfere with the gas sensing process is through fluctuations in the concentration and the charges of oxygen vacancies. Ahn et al.  investigated the effect of oxygen-vacancy-related defects on gas-sensing properties of their single ZnO nanowire gas sensors and found that the gas sensitivity towards NO2 was linearly proportional to the photoluminescence intensity of oxygen-vacancy-related defects. Their work proves evidence of the role that oxygen vacancies play in gas sensing.
Recent development in the synthesis of single crystalline nanowires or nanobelts has stimulated research into their gas sensing properties, which reveal important information about the reactions between target species and metal oxide surfaces free from complications caused by grain boundaries. For example, single crystalline SnO2 nanobelts provided Moskovits et al. the opportunity to study surface reaction kinetics between the individual nanobelt surface and CO and fit the experimental data to the analytical model they derived . Another example is that single crystalline SnO2 nanobelts with well-defined facets [with exposed (1 0 1) and (0 1 0) surfaces] also give Yang et al. the model to verify their results from numerical investigation into surface interactions between SnO2 with NO2 species: through first-principle density functional theory (DFT) calculations. They found unexpectedly that most stable adsorbed species involve an unexpected NO3 group doubly bonded to Sn centers, which was confirmed by their X-ray absorption spectroscopy studies on nanoribbons .
Nanowire/nanobelt diameter is usually on the order of several nanometers and is comparable to the Debye length and this often results in much larger sensitivity than their thin film or bulk counterparts. The size dependent characteristic has also been studied by some researchers. Liao et al. found that thin nanorods have a significantly better sensing performance than thick nanorods in the detection of C2H5OH and H2S (100 ppm) in air . The gas performance of film type gas sensors can be limited not just by surface reaction processes, but also by the morphology and microstructure of the films. Contact barriers among nanowires can also affect the gas sensing properties via affecting the resistance of the bulk material . Generally, researchers use the power law, S = a + bcρ, to fit the concentration-sensitivity curves of film type nanosensors, and Langmuir adsorption isotherm to fit the sensitivity-concentration curves of single nanowire gas sensors .
Sensitivity, response and recovery time, linear range, as well as limit of detection (LOD) are important performance parameters for gas sensors. The sensitivity of conductometric sensors is defined as the ratio of the device's resistance when exposed to target species to that in ambient air, exactly Rg/Ra (where R represents resistance, the subscript ‘g’ represents target gas, and ‘a’ represents ambient air) if the target gas is an oxidizing one, or Gg/Ga (G represents conductance) if it is a reductive one. Response (recovery) time is defined as the time period needed for the device to undergo resistance changing from 10% (90%) to 90% (10%) of the value in equilibrium upon exposure to an oxidizing (reducing) analyte. According to its definition, the estimation of LOD is done via extrapolating the Rg/Ra versus concentration curve to 3σ/Ra (σ is the standard deviation of Ra), but very few references mention to have done it in this way [31,37]. This is mainly because of the morphological complexity of the porous sensor surface and lack of efficient model to fit the sensitivity-concentration curves.
There are few problems in this field of study which can severely hinder real applications of metal oxide 1D nanostructures. First, the researchers in this community do not abide by a unified LOD when claiming the detection limit of their gas sensor reaches some ppb or ppm level. Generally they just label the lowest concentration of the analyte used in their test as the detection limit of their gas sensors. The second is lack of uniformity in the working temperature selected. Almost half the publications report a working temperature setting of 300 °C, and one quarter at 400 °C. The optimum working temperature is not always explored. Third, as water vapor produces resistance changes for metal oxides, it is important to present humidity information. Fourth, very few researchers (except [30,38-40]) have worked outside the linear range or selectivity to facilitate industrial applications.
Up to now, various metal oxides 1D nanostructures (SnO2 nanowhiskers, In2O3 nanowires, ZnO nanorods, WO3 nanowires, TeO2 nanowires, CuO nanoribbons, CdO nanowires etc.) have been fabricated into film type nanosensors. As shown in Table 1, the most widely studied substances are SnO2 and ZnO, probably due to the convenience of obtaining large quantities of SnO2 or ZnO nanowires  via thermal evaporation or a vapor-liquid-solid method. Actually this Table serves as a supplementary to another one in reference , which provides additional information on reported gas sensor properties. It's noteworthy that, in agreement with intuition, gas sensitivities of single nanowire gas sensors are invariably far less than those of nanowire film gas sensors, but the significance of single nanowire gas sensors is their potential application in microarray electronic noses .
The metrics of humidity senor involves more complicated procedures than sensors for other target species in order to obtain reliable data. In the latter, the ambient gas is usually switched between air and a target gas diluted in air, which simulates real applications; in the former, the target gas (water vapor) has to be injected into a highly dry environment which requires pretreatment such as evacuation to remove water adsorbent in the chamber . When testing is conducted under high vacuum, the usual concept, the concentration, used to represent the quantity of gas species present is replaced by an alternative one, the gas pressure of target species.
Hierarchical, dendritic, or branched nanostructures (Figure 6) are potential benefits brought about by their unique morphology that have also been explored. Table 2 shows the performance of gas sensors made from such nanostructures. Here we should point out that unlike the other three types of structures which are homogeneous, SnO2 nanobrushes are composed of heavily doped back-bone nanowires, which exhibit metallic conduction and provide more electron paths in the three-dimensional porous film, and semiconducting undoped SnO2 nanowires which act as efficient sensing components.
Hundreds of publications have reported on the gas sensor behaviors of metal oxide 1D nanostructures since 2002. Since the lowest detection concentration is a very important performance index for sensors, it is important to make a summary on the level achieved towards common target species (CO, NO2, NH3, ethanol, H2, H2S) of sensors fabricated from different metal oxide 1D nanostructures. Table 3 is the result of such effort. This can be referred by those who wish to publish new data in this realm.
It is widely accepted that the presence of noble metal elements (Pt, Pd, Au, etc.) on the surface of a metal oxide enhances the interaction of reducing gases with the absorbed oxygen on the surface, hence modified oxide nanowires have also attracted some research attention. In 2002, Arbiol et al. reported the synthesis and structural study of Pd nanoparticles on SnO2 nanowhiskers and believed such system had enhanced sensor performance . It was not until 2005 that such systems began to arouse attention for sensor fabrication and measurement (Figure 7). Kolmakov et al.  performed in-situ deposition of Pd nanoparticles in the same reaction chamber where the gas sensing measurements were carried out to ensure that the observed behavioral alteration was due to the Pd functionalization rather than the properties variation from one nanowire to another.
Figure 7a shows the conductance changing during the whole deposition process, in which Schottky barrier-type junctions resulted in the formation of electron depletion regions within the nanowire and constricted the effective conduction channel. Their Pd-functionalized SnO2 nanostructures exhibited a dramatic improvement in sensitivity toward oxygen and hydrogen, compared to pristine SnO2 nanostructures, due to the enhanced catalytic dissociation of the molecular adsorbate on the Pd nanoparticle surfaces and the subsequent diffusion of the resultant atomic species to the oxide surface (Figure 7b) .
Wang et al., instead, studied the performance of sensor based on collective Pd nanocluster (sputtered) modified multiple ZnO nanorods and also found them to have enhanced sensitivity in detecting hydrogen: changes in room-temperature resistance of approximately a factor of 5 larger than unmodified ZnO nanorods upon exposure to hydrogen concentrations in N2 of 10–500 ppm .
Chang et al. [75-77] employed a chemical-physical approach to decorate the nanowire surface with noble metal nanoparticles for film type nanosensor application: 1) immerse the nanowires sample in salt solution (PdCl2, HAuCl4); 2) place the sample under UV illumination for a period of time. During the illumination period the noble metal cations adsorb onto the nanowire surface and are reduced into atoms (from interaction with incident high energy photons) which coalesce and grow into nanoparticles. Moreover they used a patterned substrate to induce the growth of ZnO nanowires so that the decorated nanowires can self assembled into a nanofilm sensor (Figure 8). No further pulverization or grinding was involved as in standard film type nanosensor fabrication that may disintegrate the nanoparticle-nanowire structure. The thus-fabricated device exhibited enhanced response towards ethanol compared to non-decorated ZnO nanofilm sensors: increased from 18.5% to 44.5% at 170 °C and from 36.0% to 61.5% at 230 °C.
Pd-functionalized SnO2 nanowire film type gas sensors have also been fabricated and shown to have a sensitivity of 253 towards 2000 ppm H2 gas at 100 °C . Gas sensors based on Pd nanocrystal-modified CeO2 nanowires were shown to be more sensitive than unmodified CeO2 nanowires towards CO; moreover, its response to CO is highly selective among other species such as H2, ethanol, gasoline, and H2S .
Doping is another technique utilized to improve gas sensing properties of metal oxides, where the dopant atoms are believed to act as activators for surface reactions. Ru-doped SnO2 nanowires were synthesized and demonstrated to undergo conductance changes upon exposure to liquefied petroleum gas (LPG), which is seldom used as target gas for measuring the gas sensing properties of SnO2 nanowires . Our group fabricated film type nanowire sensors based on Sb-doped SnO2 single crystalline nanowires and found that they exhibited faster response and recovery compared to sensors based on undoped SnO2 nanowires.
Heterostructures based on metal oxide nanostructures has also been synthesized and fabricated into gas sensors. Similar to those of noble metal decorated nanowires, an accepted mechanism of such systems is that the target gas molecules chemisorbs onto the hetero-interface which induces charge transferring process and modulates the barrier height and hence produce apparent resistance changing. Table 4 displays research efforts made in this aspect.
Several prototypes of novel gas sensors operating under different physical and even statistical principles have been fabricated. One is based on existent technology, the basic idea of which is to make use of the fact that a metal oxide has different calibration curves (sensitivity-concentration relationship) or temperature characteristics (sensitivity-temperature relationship under specific concentration) towards different chemical species. Another makes use of the Joule heat produced by the sensor itself to achieve the optimum temperature rather than an external heater and hence reduce the total power consumption. The third one is based on a novel phenomenon that correlates the photoluminescence quenching effect metal oxides exhibit when exposed to NO2, whose mechanism remain unresolved.
The idea of “electronic noses” is inspired by the olfactory systems of humans and mammals. It discriminates different gas species by discerning the conductivity pattern of its various elements produced upon exposure to these gas species and hence is strongly dependent on data analysis and model building-up.
The most well-known way to fabricate a highly selective “electronic nose” is to integrate different recognition molecules on the same chip. However, this has been proven to not work well with metal oxides probably because the reactions do not vary much between a specific target species and different metal oxide surfaces. Chen et al. demonstrated a hybrid “electronic nose” composed of In2O3, ZnO, SnO2 nanowires and single-walled carbon nanotubes combined with statistical method principal component analysis (PCA) which cannot effectively distinguish H2 and ethanol molecules . A general approach for fabricating metal oxides “electronic nose” is to separate one monolithic metal oxide thin film into sensor segments by parallel electrode strips used for measuring the conductance of the segments while a gas-permeable membrane layer with varying thickness deposited on top of the segments, along with a temperature gradient maintained along the segment arrays, forming two mechanisms to modify the selectivity of different segments .
The state-of-the-art multielectrode KAMINA platform is based on the second idea. Kolmakov et al. combined KAMINA technology with “bottom-up” SnO2 nanowire mats as the sensing elements . They applied the statistical method linear discriminant analysis (LDA) to transfer sensor signals collected from all 38 channels into an optimized coordinate system of lower dimensionality equal to the number of training gases minus one (Figure 9).
The more different (separated) the conductivity patterns for the various trained gases, the better the discrimination power of the e-nose. Two points can be drawn from the Figure: (1) The isothermal (taken without thermal gradient) conductivity patterns of the nanowire sensor array are already sufficient to obtain substantially different signal patterns for the various gases; (2) If a temperature gradient is additionally applied along the microarray, the discrimination power is significantly enhanced. Their results suggested that the nanowire density variation over the segments alone can provide substantial differences in the conductivity patterns of the gradient microarray.
Kolmakov et al.  investigated the self-heating effect on the operation of single SnO2 nanowire gas sensors towards oxygen or hydrogen in oxygen and their sensor was able to operate without a heater, consuming only a few microwatts of power. Prades et al.  further tested the idea of optimizing the sensing conditions for the detection of gaseous species without the aid of an extra heater, which helps reducing the power consumption of the devices. They fabricated individual SnO2 nanowire devices and measured their sensing properties towards a specific concentration of NO2 under various input current and heating temperature respectively. Then they compared the calibration curve of each measuring method (Figure 10a) and generate a curve (Figure 10b) that correlate the input current and the effective temperature (roughly estimated through the response and recovery times obtained under each operating conditions) it achieves which can help further improvement of future prototype performances. Their devices operated optimally with less than 20μW for NO2 sensing to both bias and self-heating, and is significantly lower than the 140 mW required for the external microheater.
Faglia et al. observed such phenomenon in SnO2 and ZnO nanowires: exposure to a few ppm concentrations of nitrogen dioxide significantly quenched the visible photoluminescence (PL) emission of nanowires (Figure 11), which was not affected by variation of relative humidity and by other interfering gases [93-95].
The response is highly selective toward humidity and other polluting species, such as CO and NH3. They believed the results foresaw the development of a class of selective metal oxide gas sensors and pushed the lowest detection concentration of this novel configuration to 1ppm of nitrogen dioxide. Figure 12 can be viewed as the calibration curve of such optical gas sensor, whose sensitivity is defined as the ratio of PL intensity of SnO2 nanowires exposed to target gas (NO2) to the PL intensity of SnO2 nanowires exposed to blank sample (dry air).
In order to resolve the mechanism beneath, they studied the recombination dynamics based on time-resolved photoluminescence spectroscopy and numerical regression of the data by means of different decay functions (exponential, stretched exponential and double-exponential). Their findings suggest that adsorption of NO2 acts in de-activating the emitting transitions the same manner as in the static quenching mechanisms holding in bimolecular interactions .
This article has reviewed recent progress in the past two years in gas sensors based on semiconducting metal oxide one-dimensional nanostructures. During the past two years, although metal oxide nanowires/nanobelts have been exploited for detecting biological species such as glucose , urea , and L-tyrosine , or functioning merely as transducers in FET biosensors , numerous new data on their sensor properties towards common target species such as NO2, ethanol, H2, H2S has been published. Moreover, more modified or doped metal oxide nanowires/nanobelts have also been synthesized and their sensor properties towards these chemical gases investigated. Additionally, new device structures such as the “electronic nose” and the low power consumption self-heated gas sensor have been designed and their sensor response has also been evaluated. In the review, we have pointed out the lack of professionalism in many sensor performance measurements in several aspects. Specifically, there is the abuse of the concept of LOD, widespread failure to label the linear range or to assess the selectivity of the reported film type nanosensors, which make the results of little technological significance. Future development of this field would hopefully witness improvements in these two aspects.
This work was supported by the National Natural Science Foundation of China (50602014, 10874042), Hunan Provincial Natural Science Foundation of China (07jj107), Program for New Century Excellent Talents in University (NCET-06-0699), and the “973” National Key Basic Research Program of China (Grant No. 2007CB310500).