NO2 Detection and Real-Time Sensing with Field-Effect Transistors

Aug 28, 2013 - The dynamic response can be described by an analytical model. Implementation in a sensor protocol allows for the fabrication of a funct...
0 downloads 0 Views 3MB Size
Review pubs.acs.org/cm

NO2 Detection and Real-Time Sensing with Field-Effect Transistors Anne-Marije Andringa,†,§ Claudia Piliego,‡ Ilias Katsouras,‡ Paul W. M. Blom,‡ and Dago M. de Leeuw†,‡,* †

Zernike Institute for Advanced Materials, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands Philips Research Laboratories, High Tech Campus 4, 5656 AE Eindhoven, The Netherlands ‡ Max Planck Institute for Polymer Research, Ackermannweg 10, 55128 Mainz, Germany §

S Supporting Information *

ABSTRACT: The huge impact of nitrogen dioxide (NO2) emission on public health and the environment is motivating extensive scientific and technological research in the field of NO2 sensing. Field-effect transistors have emerged as a sensitive and reliable technology for monitoring air quality, due to their amplified sensor response. In this review article, the NO2 detection mechanism with field-effect transistors is discussed. The origin is charge trapping at the gate dielectric, yielding a threshold voltage shift. The dynamic response can be described by an analytical model. Implementation in a sensor protocol allows for the fabrication of a functional demonstrator. The sensor, based on a ZnO field-effect transistor, is capable of detecting concentrations as low as 40 ppb of NO2 in real-time. The sensor operates in ambient air, and apart from drying, no further precautions are taken, showing that the fabricated sensor is selective for NO2. The results reviewed in this paper set a precedent for sensing with field-effect transistors. KEYWORDS: NO2 sensors, field-effect transistors, ZnO, charge trapping, real-time sensors, threshold voltage shift, dynamic read out, electron trapping



INTRODUCTION The industrial development in the past decade together with the improvement of quality of life and transit mobility has increased enormously the amount of toxic gases released into the atmosphere, with dramatic consequences for both the environment and the public health. Air quality monitoring in urban areas has become necessary as a consequence of the daily massive production of harmful gases, ashes, and hydrocarbons. Among poisoning and air polluting gases, nitrogen oxides, NOx, are particularly dangerous.1 This group contains highly reactive gases, such as nitric oxide (NO) and nitrogen dioxide (NO2). Any form of nitrogen oxide (NOx) at levels greater than 1 ppm can cause serious damage to the human respiration system and lung tissues,2 causing or worsening respiratory diseases,3 such as emphysema and bronchitis, or aggravate existing heart conditions. NOx also plays an important role in the chemistry of the atmosphere, contributing to the formation of ozone, which is the major cause of photochemical smog, and acid rain, which damages buildings and pollutes water sources. Nitrogen oxides are mostly released during the combustion of fossil fuels. Both the European Environment Agency (EEA) and the U.S. Environmental Protection Agency (EPA) have identified “on-road” vehicle exhaust as a major source of NOx. In the last report on air pollution by the EEA, 40.5% of the total NOx emissions has been attributed to the road transport sector (see Figure 1).4 NO2 is used as the indicator for the larger © 2013 American Chemical Society

Figure 1. Contribution of different sectors to emissions of nitrogen oxides in 2012 in Europe. Data obtained from the European Environment Agency (EEA).

Special Issue: Celebrating Twenty-Five Years of Chemistry of Materials Received: June 25, 2013 Revised: August 13, 2013 Published: August 28, 2013 773

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

Here NO2 detection with field-effect transistors is reviewed. A wide variety of semiconductors has been applied as a sensing layer, such as amorphous organic semiconductors,39,40 porous silicon,41,42 carbon nanotubes,43−45 silicon,46 and metal oxide nanowires.47 In all cases changes in current upon NO2 exposure have been demonstrated. The detection mechanism, however, was still under debate, and several possible explanations had been suggested. In this review article a thorough study on NO2 detection with FETs is presented. First the key advantages of using FETs for gas detection will be underlined through the comparison with the standard chemiresistor technology. The NO2 detection mechanism in field-effect transistors will be explained by using ZnO as a model semiconductor.48−50 However, it is shown that the detection mechanism is generic and independent of the semiconductor used. The origin is electron trapping at the gate dielectric/semiconductor interface, as determined by delamination of the semiconductor and measurement of the surface potential in the channel. Starting from the explanation of the detection principle, the focus of this review is to describe stepby-step the realization of a functional sensor, able to monitor the partial NO2 pressure in real-time.

group of nitrogen oxides by EPA’s National Ambient Air Quality Standard. The choice of monitoring the level of NO2 is due to its high toxicity. Moreover, NO2 is generated primarily from the liberation of nitrogen contained in fuel as a byproduct of combustion processes.5 The official limit on the amount of this pollutant considered safe to breathe, the national ambient air quality standard for NO2, has been set by EPA at 53 parts per billion (ppb), averaged annually. The huge impact of NO2 emission on the environment and public health has led to scientific and technological interest for NO2 detection.6,7 Furthermore NOx detection can be a diagnostic tool. For instance, detection of nitric oxide8 in exhaled breath is helpful to identify infections of lung tissue.9,10 Before asthma attacks, there is a 3- to 5-fold increase of NO level in exhaled breath. If patients with asthma could monitor their NO level at home with a hand-held device, they could preventively take medicine. Presently, a wide variety of sensors based on chemiluminescent, electrochemical, resistive, and optical detection are commercially available. A brief overview is presented below. In the detection method based on chemiluminescence,11 the light resulting from a chemical reaction is detected. The chemiluminescent sampler for the measurement of NO2 relies on the reaction of NO with O3 to produce an excited form of NO2. As the excited molecule returns to its ground state, fluorescent radiation is emitted. The intensity of the light is proportional to the concentration of NO. A sensor based on this method is sold by Eco Physics.12 A gas converter (for example series CG, M&C TechGroup) can be used to convert NO2 catalytically to NO. Commercial NO2 sensors from Alphasense13 and Dräger14 are based on the use of electrochemical cells.15 The operation principle relies on the electrochemical reduction of NO2 between two electrodes immersed in an electrolyte reservoir. NO2 present in the sample air passes through a capillary diffusion barrier into the reaction cell, where it is reduced at the electrode. The migrating electrons produced by the reaction result in a net current flow, which is proportional to the NO2 concentration. The most common NO2 sensors, however, are chemiresistors based on semiconducting metal oxides16 such as tin oxide, SnO2,17 tungsten oxide, WO3,18−20 or zinc oxide, ZnO.21,22 Porous oxide layers with high specific surface area are used. The gas diffuses into the oxide and modulates the grain boundary resistances by transferring charge carriers from the semiconductor to the adsorbed surface species. Nowadays many companies offer metal-oxide based gas sensors, such as AppliedSensors23 and Synkera Technologies Inc.24 Commercial NO2 sensors are typically either bulky and have a high power consumption or have a low selectivity and limited sensitivity. The stand-alone units can be expensive, which then prohibits point of care use. There is a need for sensitive and selective NO2 sensors that can be fabricated using standard IC technology, which can be miniaturized and used in hand-held applications. Field-effect transistors (FETs)25 have emerged as a worthwhile alternative,26 after a hydrogen sensor based on a Pd-gate transistor was demonstrated in 1975.27 In recent years, FETs have been used to detect a wide range of gases by monitoring the changes in the current.28,29 Examples of semiconductors applied include carbon nanotubes30 and organic31−35 and oxidic36 semiconductors. Gas sensing with various semiconductors has recently been reviewed.37,38



CHEMIRESISTORS Conventional metal oxide sensors are two-terminal chemiresistors.51 The adsorption of gas molecules on the surface of the metal oxide semiconductor changes the resistance of the device. By monitoring the change in resistance, detection of gases in the ppm regime is possible. The operation mechanism is well established.52,53 The sensors are usually operated in air, at temperatures between 200 and 400 °C. When a sensor is heated to a high temperature in the absence of oxygen, free electrons easily flow through the grain boundaries of the metal oxide semiconductor film. In an oxygen atmosphere, oxygen is adsorbed onto the semiconductor surface, forming a potential barrier at the grain boundaries. At the temperature used, the oxygen traps electrons from the bulk of the material. As a consequence, the free charge carrier concentration is reduced, and a depletion layer is formed. This impedes the flow of electrons and thus increases the resistance. When the sensor is exposed to an atmosphere containing a reducing gas, such as H2, CH4, CO or H2S, the reverse process takes place. The thickness of the depletion layer and the potential energy barrier decreases, allowing the electrons to flow and thus reducing the electrical resistance. In case of strong oxidizing gases, such as NO2, the molecules interact with the surface of the metal oxide through surface adsorbed oxygen ions, thus increasing the potential barrier at the grain boundaries.54 As a consequence, the thickness and resistance of the depletion layer increases. In this manner, the sensor acts as a variable resistor whose value is a function of the type of gas and its concentration.55 The electrical transport is dominated by Schottky barriers at the grain boundaries. The sensitivity of the sensor depends on the size of the depletion layer relative to the thickness of the bulk semiconductor.56 Commercial sensors therefore typically consist of thick porous films of a metal oxide. The lack of selectivity of the discrete devices is usually addressed by combining a number of sensors in an array. Drawbacks that limit the functionality are the low sensitivity, typically about 1 ppm, and the poor reversibility, especially after exposure to high gas concentrations. The necessary technological improvements are the enhancement of the sensitivity and the reduction of the power 774

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

consumption due to the high operating temperatures. To this end, other semiconductors, such as phthalocyanines,57−59 carbon nanotubes,60 and magnetic particle films61 are being investigated.



FIELD-EFFECT TRANSISTORS AS GAS SENSORS Field-effect transistors (FETs) have emerged as a worthwhile alternative to chemiresistors. An FET is a three-terminal device in which an additional gate electrode modulates the charge carrier density in the semiconductor. In the standard configuration, a semiconductor layer is deposited between two electrodes, the so-called source and drain. The third electrode, the gate, is electrically insulated from the semiconductor layer by a dielectric oxide layer, thus creating a parallel plate capacitor. When a bias voltage is applied to the gate with respect to the source electrode, charge carriers electrostatically accumulate or deplete in the semiconductor at the semiconductor/gate dielectric interface. Owing to the fieldeffect, the charge carrier density in the semiconductor can be modulated. The current through the active layer can be varied over orders of magnitude. The key advantage of a transistor over a chemiresistor is the amplified sensor response due to the current modulation by the extra gate electrode. It has been reported that FETs might offer more sensing parameters than just the current density. For instance, the extracted charge carrier mobility can change with gas exposure.28,29 Furthermore, the sensor response of FETs can be enhanced by combining discrete FETs into ring oscillators and other adaptive amplifier circuits.62 NO2 detection with FETs has been investigated using different semiconductors. Several possible sensing mechanisms have been suggested. The adsorbed NO2 is a strong electronacceptor that influences the charge carrier density in the channel. The change in current has been explained by doping of the semiconductor due to the oxidizing NO2 or by depletion of the electrons from the conduction channel.63 For carbon nanotube sensors,64,65 Zhang et al. postulated that the NO2 molecules change the metal work function, leading to a change in the Schottky barrier height at the carbon nanotube/metal interface.66 Kong et al. proposed molecular gating of the nanotubes by the polar NO2 molecules as the sensing mechanism.67 However, Andringa et al. elucidated the mechanism by reporting a detailed analysis of NO2 detection using ZnO-based transistors.68 The observed current decrease in the transistor originates from a gradual shift of the threshold voltage due to charge trapping. The same behavior was observed by replacing ZnO with n-type,69 p-type,70 and ambipolar organic semiconductors.71 The description provided for the NO2 detection in field-effect transistors is therefore generic and independent of the semiconductor. Only the nature of the gate dielectric is crucial. In the following sections, the charge trapping dynamics are described, a sensor model is developed, and a real-time sensor is demonstrated.

Figure 2. Typical transfer curve of a ZnO field-effect transistor measured with a drain bias of 2 V. In the insets, two schematic presentations of the transistor are shown. At a negative gate bias the semiconductor is depleted and the transistor is off. At a positive bias, electrons are accumulated in the semiconductor at the ZnO/SiO2 interface and current is flowing. Adapted from ref 68.

channel. Upon applying a source-drain bias, current is flowing. The magnitude is proportional to the induced charge carrier density and the field-effect mobility. At negative gate bias any electron in ZnO is depleted. Holes cannot be injected and are immobile in ZnO. Hence, as schematically depicted in the upper inset of Figure 2, in depletion no current is flowing. The transistor is measured in the linear region, where the drain bias is much smaller than the gate bias; hence, the charge density is uniform throughout the channel. Figure 2 shows that the current starts to flow around zero gate bias. This onset is the pinch-off voltage, which is assumed to be equal to the threshold voltage, Vth. The ZnO transistors were characterized in NO2 ambient at room temperature. When the gate is floating, the current does not change with NO2 content. The device is stable. However, large changes occur when a gate bias is applied. The process is illustrated in Figure 3a. A ZnO transistor kept in vacuum is turned on by applying a positive gate bias, and the current is measured as a function of time. The current is stable, until NO2 gas is admitted. The current immediately drops, and within 10 min the current has decreased by more than 3 orders of magnitude. To elucidate the origin of the current drop, the transfer curves were measured at the indicated times in Figure 3a. Figure 3b shows that in vacuum the transistor switches on at 0 V gate bias (red curve). After NO2 exposure and upon continuously applying a positive gate bias, the transfer curve shifts toward the applied gate bias (green and blue). The main effect of NO2 exposure is a shift of the threshold voltage, Vth. Since ZnO transistors without applied gate bias are chemically stable in NO2 ambient, doping of the semiconductor by oxidizing NO2 can be ruled out. To elucidate the role of a finite gate field during NO2 exposure, a continuous gate bias was applied for a longer time, with grounded source-drain electrodes. The resulting threshold voltage shift was monitored by measuring the transfer curves as a function of time. In Figure 4a, the linear transfer curves measured at a 2 V drain bias are presented as a function of stress time, the time at which the gate bias is applied. Figure 4a shows that the transfer curve shifts with stress time toward the applied gate bias, in this case 30 V. The arrows indicate transfer curves measured after 1 to 1000 s. The shape of the curves does not change; therefore, the mobility does not change. There is only a shift in threshold voltage, which is taken to be similar to the shift in pinch-off voltage. The threshold voltage shift, with



THRESHOLD VOLTAGE SHIFT The NO2 detection mechanism is described by using transistors with ZnO as semiconductor. A typical transfer curve of a ZnO transistor,72,73 where the current is measured as a function of gate bias at a constant drain bias, is presented in Figure 2. ZnO is an intrinsic undoped n-type semiconductor. At zero gate bias there are no charges accumulated and, hence, the current is negligible. At positive gate bias electrons are accumulated in the 775

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

respect to the reference value measured in vacuum, is presented in Figure 4b as a function of time on a semilogarithmic scale. Initially a fast change is observed, while at longer times the threshold voltage shift saturates at the value of the applied gate bias. This rate of change resembles the temporal behavior of a stretched-exponential relaxation. Hence, the threshold voltage shift has been fitted to the following: ⎧ ⎡ ⎛ t ⎞ β ⎤⎫ ⎢ − ⎜ ⎟ ⎥⎪ ⎬ ΔVth(t ) = V0⎨ − 1 exp ⎪ ⎣ ⎝ τ ⎠ ⎦⎭ ⎩ ⎪



(1)

where τ is a relaxation time, β is a dispersion parameter equal to T/T0, and V0 = VG − Vth0, where VG is the applied gate bias and Vth0 is the threshold voltage at the start of the experiment. A good fit is obtained, as shown in Figure 4b. The threshold voltage shifts only when the gate bias is applied. The drain bias has no influence, in contrast with classical chemiresistors. This has been demonstrated68 by keeping the ZnO transistor in NO 2 atmosphere and intermittently grounding the drain. The decrease in current is irrespective of the time that the drain has been grounded, demonstrating that the drain bias has no effect. Implicitly, it also shows that the current itself is not the origin of the current decay. Therefore, contact degradation74 can be ruled out as the sensing mechanism. From these experiments several conclusions about the detection mechanism can be drawn. Doping of the ZnO semiconductor by oxidizing NO2 can be ruled out. The transistors are stable in NO2 atmosphere when the gate is floating. Moreover, the shape of the transfer curve does not change during the measurements in NO2 at different gate biases; hence, the material properties remain unaffected. Also contact degradation can be excluded as the sensing mechanism because the current itself is not the origin of the current decay. The temporal behavior of the threshold voltage has been monitored as a function of the NO2 concentration and of the layer thickness. The transistors were measured during exposure to different concentrations of NO2. For each measurement, the threshold voltage shift has been determined as a function of stress time. Figure 5a shows the extracted threshold voltage shifts as a function of time and NO2 concentration. For all concentrations, the threshold voltage shifts to the applied gate bias. The main effect of exposing the transistor to different concentrations is a change in the time dependence. All data can be fitted with a stretched-exponential, as shown by the fully drawn curves. For all concentrations, a good agreement is obtained between the experimental data and the fit curves. Figure 5b shows that the main change is in the characteristic relaxation time τ, which varies over orders of magnitude by changing the NO2 concentration. The slope is unity; the relaxation time is inversely proportional to the NO2 partial pressure:

Figure 3. (a) Drain current of a ZnO field-effect transistor as a function of time using an applied gate bias of 30 V and a drain bias of 2 V. The measurement started in vacuum, and after 4 min 250 ppb of NO2 was admitted. (b) Transfer characteristics corresponding to the indicated times during the measurement in Fig. (3a). The red curve was measured in vacuum, the green and blue curves during NO2 exposure. Reprinted with permission from ref 68. Copyright 2011 Wiley-VCH.

τ≈

1 pNO

2

Figure 4. (a) Transfer curves of a ZnO transistor exposed to 320 ppb NO2 as a function of time under a continuously applied gate bias of 30 V. The arrows indicate transfer curves measured after 1−1000 s. (b) The threshold voltage shift obtained from Fig. (4a) presented as a function of time on a semilogarithmic scale. The solid line is a fit to a stretched-exponential time dependence. Adapted from ref 68.

(2)

As shown in Figure 5b, extremely low NO2 concentrations, down to the ppb level, can be detected. The ZnO layer thickness can also be varied, to investigate the vast parameter space of field-effect transistors. Again, the dispersion parameter hardly changes; the main effect is a change in relaxation time. The relaxation time increases exponentially with increasing layer thickness.68 776

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

Upon applying a positive gate bias, electrons are accumulated in the channel. Some get trapped at, for example, defect sites or acceptor molecules. The trapped charges do not contribute to the current, but they shift the threshold voltage. When the trapped charges fully compensate the applied gate bias, there is no driving force left for charge accumulation. The threshold voltage shift has saturated at the applied gate bias; a steady-state has been reached, where the gate bias is completely compensated by immobile trapped charges. The threshold voltage in a transistor without trapped charges is set by the difference in work function between the electrodes and the semiconductor. When a ZnO transistor is turned “on”, electrons are accumulated and some of them are trapped. These trapped charges do not contribute to the current, but they are still part of the electrostatic charge that shields the gate voltage. The threshold voltage shift can be quantified as eQf/Cox, where e is the elementary charge, Cox is the gate dielectric capacitance, and Qf are the fixed charges.75 A shift of the threshold voltage with time is due to an increase of the fixed charges, Qf. With time the effective gate field decreases and the threshold voltage saturates at the applied gate bias. When this point is reached (steady-state), no current flows. Experimentally it was determined that the threshold voltage shift as a function of time and partial NO2 pressure can be described by a stretchedexponential time dependence. This decay, called William-Watts or Kohlrausch relaxation, is observed in a wide class of disordered systems, such as relaxation of glasses toward equilibrium, dielectric relaxation in a charge density wave system, and dispersive transport in photoconductors.76 It holds for all dispersive transport processes in an exponential distribution of states.77 The measurements used to determine the threshold voltage shift resemble those used to determine the operational stability of amorphous semiconductors due to gate bias stress.78−80 When a transistor is “on” for a long time, the threshold voltage shifts to the applied gate bias according to a stretchedexponential time dependence. Without NO2, the ZnO transistor shift is small; the operational stability does not hinder the experiment. With NO2, however, a dramatic shift of the threshold voltage is observed. The shift is orders of magnitude faster, but the functional dependence is the same. Experimentally, the temporal behavior of the threshold voltage as a function of partial NO2 pressure can be described by a stretched-exponential decay. The characteristic relaxation time, τ, is inversely proportional to the partial NO2 pressure and scales exponentially with the ZnO layer thickness. Both dependences can be rationalized by assuming that the NO2 detection mechanism relies on charge trapping with an anomalously wide distribution of trapping times, due to a time dependent diffusion coefficient. The NO2 concentration is low (in the ppb range). Assuming a Langmuir isotherm for the NO2 absorption, the density of absorbed NO2 is then proportional to the partial vapor pressure. Simultaneously, the density of absorbed NO2 decreases exponentially with the ZnO layer thickness, with a decay constant determined by the morphology.40 These realistic assumptions are sufficient to explain the experimentally observed dependences of the characteristic relaxation time τ on pressure and layer thickness, τ ∼ 1/p(NO2) and τ ∼ exp dZnO.

Figure 5. (a) Threshold voltage shift of a ZnO transistor as a function of time upon exposure to different concentrations of NO2. The continuously applied gate bias was 30 V. The extracted shifts were fitted to a stretched-exponential time dependence. (b) Relaxation time, τ, as a function of the NO2 concentration. The slope of the fit is unity. Reprinted with permission from ref 68. Copyright 2011 Wiley-VCH.



CHARGE TRAPPING MECHANISM The mechanism responsible for the threshold voltage shift can be described by charge trapping mediated by the NO2. The charge trapping process is schematically depicted in Figure 6.



Figure 6. Schematic representation of the charge trapping process in a transistor. With time, mobile electrons are trapped until the steadystate is reached. The gate bias is then completely compensated by immobile trapped charges.

DYNAMICS OF TRAPPING AND RECOVERY The transport in a field-effect transistor changes upon exposure to NO2. The threshold voltage shifts toward the applied gate 777

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

transistor and a normalization factor for the NO2 concentration. To measure the recovery, first the threshold voltage was set at 30 V by applying a continuous gate bias. With the NO2 still present, the stressed transistors were then recovered by grounding both gate and drain electrodes. Figure 7 shows that detrapping becomes much faster with increasing temperature. The extracted relaxation time, τ, is presented as a function of reciprocal temperature (red). The activation energy for detrapping amounts to 1.2 ± 0.1 eV and an attempt-toescape frequency, ν, is derived of 1011±1 Hz. Hence the recovery can be quantitatively described as well with three parameters, ν, T0, and Ea, which are constant for a given transistor. The influence of the applied biases on the recovery has been investigated. Both the drain and the gate bias do not affect the relaxation time. Only when a positive gate bias is applied, the final threshold voltage changes; it saturates at that applied gate bias. Furthermore, the recovery could not be enhanced by putting the stressed transistor in deep depletion by applying a negative gate bias. A parameter that influences the kinetics is the presence of light. It has been reported that the recovery properties of ZnO and SnO2 sensors were improved remarkably by UV light irradiation.83,84 Preliminary experiments show that the relaxation time under UV illumination in the ZnO FET decreases by orders of magnitude. The different time constants for charge trapping and recovery can be explained by a deep trap filling process. The trap filling is essentially controlled by the number of unoccupied trap states and by the free carrier density induced by the gate bias, while the recovery is due to thermally activated emission. Initially, the transistor is not stressed. The traps are empty, and the free carrier concentration is relatively high. The capture rate, cn, is directly proportional to the free carrier concentration, n, and to the number of unoccupied states:85

bias. The origin of the shift is charge carrier trapping. The change in current or threshold voltage with time depends on the NO2 concentration. Hence, a field-effect transistor can in theory be used as an NO2 sensor. At room temperature, however, the charge trapping is irreversible. The time scale for release of charge carriers, or recovery, is larger than months. In other words, at room temperature the transistor is not a sensor but an integrator as it monitors the accumulation of trapped charge carriers. The response of a sensor to a gas should be reversible, which implies tuning of the time scales for trapping and release of charge carriers. To that end, trapping and recovery are investigated as a function of temperature. Charge trapping hardly depends on temperature.81 The relaxation time for charge trapping is presented in Figure 7 as a

Figure 7. Relaxation time, τ, as a function of reciprocal temperature for charge trapping and recovery. The solid lines show that τ is thermally activated for both stress and recovery with activation energies of 0.1 and 1.2 eV, respectively. The attempt-to-escape frequencies are 1 Hz and 1011 Hz, respectively. Reprinted with permission from ref 81. Copyright 2012 Elsevier B.V.

function of reciprocal temperature (black). A straight line is obtained, showing that the relaxation time is thermally activated as follows:

τ=v

−1

⎛ E ⎞ exp⎜ a ⎟ ⎝ kBT ⎠

cn = n{Nt [1 − f (Et )]}vthσn

where Nt[1 − f(Et )] gives the density of empty states, Nt is the density of filled traps, Et is the trap depth, νth the thermal velocity, and σn is the capture cross-section. As the threshold voltage shifts toward the applied gate bias, the free carrier density decreases, which slows down the capture rate. Equation 5 shows that the capture rate is temperature independent. Only the thermal velocity depends on the square root of the absolute temperature. In conclusion, the electron capture process is a temperature independent, as is experimentally observed by an almost negligible activation energy of 0.1 eV. The attempt-to-escape frequency of 1 Hz extracted from the threshold voltage dynamics upon charging is orders of magnitude lower than that of a typical phonon frequency of 1012 Hz. The reason is that, depending on the microscopic mechanism, the extracted escape-frequency contains temperature independent prefactors. The process determining the rate is probably the diffusion of NO2 at the gate dielectric−ZnO interface. During recovery all free carriers are exhausted. Therefore, retrapping can be disregarded. The emission rate from the traps is now mostly determined by the trap depth and by the number of filled traps, Nt. The emission rate is given by the following:

(3)

where Ea is the activation energy, kB the Boltzmann constant, and ν is the so-called attempt-to-escape frequency. A low activation energy, Ea, of 0.10 ± 0.04 eV and an attempt-toescape frequency, ν, of 1 ± 1 Hz were measured. The attemptto-escape frequency is equal to the phonon frequency multiplied by the probability of electron−phonon colocalization.82 The extracted parameters can be used to quantitatively describe the charging dynamics. To account for the NO2 dependence, eq 3 can be normalized as follows: τ=

* pNO

2

pNO

2

⎛ E ⎞ v−1 exp⎜ a ⎟ ⎝ kBT ⎠

(5)

(4)

where the normalization constant p*NO2 is the partial pressure at which the frequency factor has been determined. Using eq 1, the complete charging dynamics, viz. the time, bias, temperature and NO2 dependence can be described by only three parameters, ν, T0, and Ea, which are constant for a given 778

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials ⎛ E − Ei ⎞ en = [Nt f (Et )]vthσnni exp⎜ t ⎟ ⎝ kBT ⎠

Review

(6)

Contrary to the temperature independent charge trapping, the recovery is strongly thermally activated. From the temperature dependent recovery experiments, an activation energy of 1.2 eV has been extracted. Detrapping can be described as a simple phonon-assisted process. The value of the extracted attempt-to-escape frequency of about 1011 Hz is comparable to that of a typical phonon frequency, and the capture cross-section of 10−16 cm−2 is in line with that of a neutral trap. To confirm the presence of trapped charge carriers and to determine the trap depth, thermally stimulated current measurements (TSC) have been performed. The TSC curves have been analyzed, and a trap depth around 1 eV has been obtained. The value is in good agreement with the activation energy derived from the threshold voltage dynamics.81

Figure 8. Schematic of the exfoliation experiment to localize trapped charge carriers. (a) The transistor after applying a positive gate bias in NO2. The trapped charges are either in the semiconductor (I) or at the interface with the gate dielectric (II). In both cases, when there is no screening, the trapped charges give rise to a negative surface potential equal to the threshold voltage shift. (b) Potential profiles after the exfoliation process. In case I, the exfoliation will remove the semiconductor including the trapped charge carriers. The resulting surface potential is then zero. In case II, the trapped charges will stay behind at the gate dielectric interface and the negative surface potential remains. Adapted from ref 90.



LOCALIZATION OF TRAPPED CHARGE CARRIERS The threshold voltage dynamics are due to charge trapping and release. The question is now where the traps are located. The electron trapping can be either in the bulk of the semiconductor or at the interface between the semiconductor and the gate dielectric. The exact location cannot easily be determined because the gate dielectric interface is buried under the semiconductor. By exfoliating the semiconductor, the revealed gate dielectric can be made accessible for characterization with Scanning Kelvin Probe Microscopy (SKPM).86,87 This technique was earlier applied to reveal the location of trapped charges due to gate bias stress88 and to link the threshold voltage shift in a transistor with a self-assembled monolayer (SAM)-modified gate dielectric to charges trapped by the SAM.89 To apply the exfoliation technique, organic semiconductors have been selected. Due to weak van der Waals forces, an organic semiconductor layer can be completely removed at once by simply using adhesive tape. The transistors were charged by applying a gate bias in an NO2 ambient. In situ measurements of the surface potential within the transistor channel were performed before and after exfoliation, using SKPM.90 The experiment to find the exact location of the trapped charges is schematically depicted in Figure 8 for an n-type semiconductor. The transistor with the trapped electrons exhibits a positive threshold voltage. The trapped charges are either in the semiconductor (I) or at the dielectric (II). In both cases, because there is no screening by the unipolar bulk n-type semiconductor, the trapped charges give rise to a negative surface potential with a magnitude equal to the value of the threshold voltage shift. A distinction can be made after exfoliation of the semiconductor. In case I, the exfoliation will remove the semiconductor including the trapped charge carriers. The resulting surface potential is then zero. In case II, the trapped charges will stay behind in the dielectric, and the negative surface potential remains. For the transistor fabrication, N,N-dialkylsubstituted(1,7&1,6)-dicyanoperylene-3,4:9,10-bis(dicarboximide) derivative (Polyera ActivInk N1400),91,92 a well-established air-stable n-type semiconductor, was used. The transistor was exposed to NO2, and a gate bias of 20 V was applied for 60 s. The surface potential, measured as quickly as possible after charge trapping, is presented as the black curve in Figure 9a. The negative

Figure 9. (a) Comparison of surface potential before and after delamination. Surface potential profiles of an N1400 ActivInk transistor after applying a 20 V gate bias for 60 s in NO2. The black curve shows the potential profile with the semiconductor still present, and the red curve shows the potential profile after delamination. The surface potentials are identical both with and without semiconductor, which demonstrates that the charges are not trapped in the semiconductor but at the gate dielectric interface. (b) The actual exfoliation process, using adhesive tape and tweezers. Reprinted with permission from ref 90. Copyright 2012 American Institute of Physics.

surface potential indicates the presence of trapped charges. To locate the charges, the exfoliation technique was applied. The transistor is exposed to NO2 using the same procedure. However, now the semiconductor is delaminated after stressing using adhesive tape and tweezers, as shown in Figure 9b. The measured potential profile after exfoliation is shown as the red curve. The surface potentials are similar with and without semiconductor, which demonstrates that the charges are not trapped in the semiconductor but at the gate dielectric interface. The minor differences are due to recovery in ambient light before the SKPM measurements start, when the semiconductor is still present.93 The potentials measured are not generated by the peeling process. The reproducibility of the measured potentials and the correspondence of the threshold 779

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

is no recovery, the sensor can only be used once and has to be reset into the pristine state after each measurement. For a dynamic read out the time scales for charging and recovery should be comparable, which can be achieved using higher operating temperatures. To set the operating temperature and the cycle times, the temporal behavior of the threshold voltage needs to be modeled. The response of the transistor can be calculated with experimentally determined parameters, without the use of any additional fit parameters. The model uses the temporal behavior of the threshold voltage as a function of temperature and NO2 pressure as derived in the previous sections. The threshold voltage shifts for trapping and release follow a stretched-exponential time dependence (eq 1). The relaxation times τ are thermally activated (eq 3). The experimentally determined values for activation energy, attempt-to-escape frequency, and characteristic temperature for both charging and recovery remain fixed. In order to calculate the dependence on NO2 pressure, the relaxation time for trapping is taken inversely proportional to the NO2 pressure (eq 2). The temporal behavior of the sensor can now be analytically calculated. The operation mechanism of the sensor is elucidated in Figure 10. For this example the driving protocol consists of

voltage shift with the surface potential exclude static charges by contact electrification or tribo-charging. The almost complete delamination of the semiconductor was confirmed by photoexcitation experiments. With the semiconductor still present, the surface potential and the threshold voltage are recovered by turning on the light of the microscope. In this case the trapped charges are released. However, after delamination the surface potential does not change. The photoexcited carriers cannot percolate to the contacts; the surface charges remain trapped. To support the assignment of trapped charges to the gate dielectric, the experiment was repeated with two other semiconductors, viz. poly(perylene bisimide acrylate) (PPerAcr)94,95 and polytriarylamine (PTAA).96,97 Both semiconductors can be completely removed by stripping. The SKPM measurements confirmed that the trapped electrons are located at the gate dielectric. The microscopic mechanism of the charge trapping is different from that in chemiresistors. When the resistor is exposed to NO2, the NO2 adsorbs on the surface of the metal oxide and redox reactions take place, as explained in a previous section. The extraction of electrons from the metal oxide results in an increase in the width of the depletion layer and of the corresponding potential barriers at the grain boundaries. Then the resistance increases, or the current decreases. Transistors, however, are stable in NO2 without applying a gate bias. The source-drain current only changes when a positive gate bias is applied, i.e. when electrons are accumulated. The consequent shift in threshold voltage is due to trapped charges at the gate dielectric and not due to a change in grain boundary resistance. Therefore, the detection mechanism in a transistor cannot simply be trapping of an electron by an isolated NO2 molecule, as it might be expected from the large electronegativity of NO2. This reduction process should take place no matter if the NO2 molecule is in the bulk of the semiconductor or at the gate dielectric interface. However, the exfoliation experiments unambiguously show that the trapped electrons are located at the gate dielectric. This finding explains why the type of semiconductor is irrelevant. It is not the type of semiconductor but the nature of the gate dielectric that is crucial.



ANALYTICAL SENSOR MODEL An NO2 sensor should respond in real-time to the partial NO2 pressure. When a transistor biased in accumulation is exposed to NO2, the threshold voltage changes. The change of the threshold voltage with time depends on the partial NO2 pressure. Hence, the fabrication of an NO2 sensor based on a field-effect transistor seems straightforward. However, development of a sensor protocol that allows for a dynamic response is not trivial. With the application of a continuous gate bias, the threshold voltage shifts to the applied gate bias. The time needed to reach the final state is dependent on the partial NO2 pressure. The final state itself is irrespective of the partial NO2 pressure and the temperature. Hence, with a fixed continuous gate bias, a transistor cannot be used as a sensor. The gate has to be turned on intermittently. When a gate bias is applied, the transistor is charged and the threshold voltage shifts. When the transistor is turned off by applying 0 V gate bias, the transistor recovers. By optimizing the cycle time and temperature, the final threshold voltage depends on the partial NO2 pressure, yielding a real-time NO2 sensor. At room temperature a practical sensor cannot be realized, since the recovery time is in the order of months. Because there

Figure 10. Calculated threshold voltage response to repetitive cycles of 10 s charging at gate bias 30 V followed by 10 s recovering at gate bias 0 V upon exposure to 100 and 1000 ppb NO2. (a) At room temperature there is no recovery, and for every concentration the threshold voltage saturates at the maximum applied gate bias. (b) At 470 K, the response is reversible and the threshold voltage shifts upward and downward upon charging and recovering. After a number of cycles, a dynamic equilibrium is reached. The maximum and minimum threshold voltages in equilibrium depend on the partial NO2 pressure. Reprinted with permission from ref 81. Copyright 2012 Elsevier B.V.

repetitive cycles of 10 s charging at a gate bias of 30 V followed by 10 s recovering at a gate bias of 0 V, as presented by the dashed lines. The calculated threshold voltage response is shown for 100 and 1000 ppb NO2. In the calculation, the bias, V0, is taken as the difference between the applied gate bias and the threshold voltage at the start of the pulse. 780

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

The inset of Figure 11 shows the time calculated to reach dynamic equilibrium for the specified concentration range. The equilibrium is reached faster at higher temperatures and at higher NO2 pressure. The reason can be that at higher pressures the rate of charging increases, leading to a faster equilibrium. The calculated equilibrium times have to be compared with those using a driving protocol with full recovery. At 470 K full recovery will take an hour, which is incompatible with any practical sensor application. However, a full recovery is not needed to achieve an NO2 concentration dependent response. A functional NO2 sensor can be made by intermittently addressing the gate bias. The analytical description can be used to optimize the driving protocol. The input parameters follow from the threshold voltage dynamics. Only three parameters for charging and three parameters for recovery are necessary, ν, T0, and Ea, and a normalization factor for the NO2 concentration. The parameters are constant for a given transistor. Without any fit constants, the sensor response can be predicted. By adjusting the driving protocol and operating temperature, the response time can be optimized and the sensitivity can be maximized for the desired partial NO2 pressure window.

The calculated temporal behavior presented in Figure 10a shows that a functional sensor cannot be realized at room temperature. Upon application of the first gate bias pulse, the threshold voltage shifts. Upon switching off the gate bias, the threshold voltage remains unchanged. At room temperature there is no relaxation. In the next charging pulse the threshold shifts further. After infinite time, the threshold voltage is the same as the applied gate bias, irrespective of the NO2 concentration. The only difference is the time needed to reach the final stage. The transistor acts as an integrator. The threshold voltage irreversibly saturates at the maximum applied gate bias. At high temperature a functional sensor can be realized. Figure 10b shows that, by adjusting the operation temperature, a reversible response can be obtained. The threshold voltage shifts upward and downward upon charging and recovering. The selected driving protocol yields a dynamic response where the final threshold voltage is a function of the NO2 pressure. The calculated temporal behavior can be understood as follows. The driving force for the change in threshold voltage is the difference between the gate bias and the threshold voltage at the start of the pulse. With an increasing number of cycles, the threshold voltage shifts to the applied gate bias. The driving force for charging decreases and, at the same time, the driving force for recovery increases. Figure 10b shows that, after a number of cycles, ΔVth charging is equal to ΔVth recovery and a dynamic equilibrium is reached. The maximum and minimum threshold voltages in equilibrium increase with partial NO2 pressure. The concentration dependence follows from the fact that the rate of charging increases with NO2 pressure while the rate of recovery remains the same. Using this analytical description, the final threshold voltage for each driving protocol, temperature, and NO2 pressure can be calculated. As an example, in Figure 11 the calculated



FUNCTIONAL NO2 SENSOR AND REAL-TIME DETECTION In the previous section, the temporal behavior of the threshold voltage has been modeled to set the operating temperature and cycle time in order to obtain a reversible sensor. The model uses as input the threshold voltage dynamics as a function of temperature and NO2 pressure as measured for both charging and recovery. The response of the transistor can then be calculated with experimentally determined parameters. Here the phenomenological model is implemented into a hardware demonstrator sensor. It is shown that the partial NO2 pressure in ambient air can be monitored in real-time. Concentrations as low as 40 ppb have been detected. The measurements are verified by comparing the sensor response with a calibrated commercial NO2 detector. The sensor cell was fabricated from Teflon and equipped with feedthroughs for the electrical source, drain, and gate contact of the ZnO field-effect transistor. For demonstration purposes, a large box of 0.4 L was used. A photograph is presented in the inset of Figure 12. The transistor was placed onto a ceramic plateau, fitting around a substrate heater and contacted. To maintain visibility of the sensor but to prevent UV irradiation from distorting the measurement, the polycarbonate lid of the flow cell was covered with an orange UV blocking foil. Ambient air was dried using molecular sieves. No other precautions were taken. The cell was flushed for 1 h. The humidity of the exhausted air was measured to be 40 ppm. As NO2 source, a permeation tube was chosen with a calibrated emission rate. The partial NO2 pressure was set by the temperature of the permeation tube, and the air flows through the sensor cell and a bypass. The measurement protocol, including the operating temperature, pulse time, and duty cycle, was set using the sensor model previously discussed in order to obtain a reversible sensor. In the measurement protocol, the gate is intermittently turned on and off and the NO2 concentration is varied. When a gate bias is applied, the transistor is charged and the threshold voltage shifts. When the transistor is turned off, the transistor recovers. By pulsing the gate, the threshold voltage oscillates between a maximum voltage after charging and a minimum

Figure 11. Maximum threshold voltage in equilibrium after charging as a function of NO2 content for 420 and 470 K using the driving protocol as in Figure 10. The solid lines are calculated. The experimentally determined threshold voltage verifies the calculations. In the inset the equilibration time is plotted for the specified concentration range. Reprinted with permission from ref 81. Copyright 2012 Elsevier B.V.

maximum threshold voltage in equilibrium is presented as a function of NO2 pressure for 420 and 470 K. The driving protocol comprised a charge and reset time of 10 s each. The solid lines are calculated. The experimentally determined threshold voltages are presented by the squares. The values coincide with the calculated line. The perfect agreement verifies the methodology developed. 781

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

To perform dynamic measurements, a transistor was mounted in the sensor cell and the heater was set at 200 °C (inset of Figure 12).98 The system was flushed with air for 1 h. The measurement protocol was started, and the permeation tube was heated to deliver a certain amount of NO2. The signal from the sensor was processed and recorded. The flow was switched intermittently from “air with NO2” to “air without NO2”. The NO2 content was varied by adjusting the temperature of the permeation tube. The dynamic sensor response, the NO2 concentration as a function of time, is presented in Figure 12 as the black curve. The discretization of the determined threshold voltage was set at 0.1 V, which yields a resolution in NO2 concentration of about 20 ppb. The lowest stable concentration achieved, set by the minimum quantity the permeation tube can deliver, was 40 ppb. The baseline is about 10 ppb. This background is due to degassing of the walls of the sensor cell. The sensor is capable of detecting these minimal changes in concentrations. The sensor was tested up to an NO2 level of 1 ppm. Above 1 ppm, the threshold voltage is almost at the applied gate bias and the concentration can no longer be resolved. When a higher concentration range is desired, this saturation can be addressed by reducing the stress time or increasing the temperature. The response time of the sensor at 200 °C lies between tens to hundreds of seconds. A new value of the threshold voltage is recorded every 22 s and compared to the calibration curve calculated in equilibrium. When the concentration in the system has changed, a new equilibrium has to be obtained by performing multiple measurement cycles. The equilibrium is reached faster at higher temperatures and higher NO2 concentrations. The response time can be further optimized by adjusting the measurement protocol. However, there will be an inherent trade-off between sensitivity and response time. In addition, the enormous dead volume of the current sensor cell prevents fast intermixing and limits the time resolution to around 100 s. To verify the sensor response, the NO2 content was measured simultaneously with the EcoPhysics NO sensor system and converter. The recorded response is presented in Figure 12 as the red curve. A good agreement between the measured and reference data is obtained, which validates the methodology. Only a minor drift in the NO2 concentration measured by the ZnO FET sensor was observed. The nature of the drift is unclear; possible reasons are slight variations in temperature or influence of residual water. The EcoPhysics NO sensor is a compact expensive standalone system used for environmental monitoring, industrial applications such as waste incineration, and exhaled breath analysis in hospitals. The price of such devices prohibits point of care use. The experiments reported show that similar performance can be obtained with a single field-effect transistor as NO2 transducer. The transistor is fabricated using standard IC technology, which can easily be miniaturized and used in a variety of hand-held applications. The detection is not optical but only electrical. The implementation of the transducer can therefore lead to a major cost reduction of an NO2 detection system. The present sensor is sensitive to water: operating the sensor in wet air causes instability observed as a baseline drift. The absorption of water onto a gate dielectric such as SiO2 in transistors causes threshold voltage shifts. Surface passivation of silanol groups by organic primers (e.g., HMDS or OTS) has been shown to strongly reduce these instabilities.99,100

Figure 12. Sensor verification. The ZnO field-effect transistor sensor was heated, turned on, and equilibrated. Then the NO2 concentration was set by adjusting the temperature of the permeation tube. The flow was switched by the 4-port 2-way valve between sensor and bypass. The measured NO2 concentration as a function of time is presented by the black line. The red line represents the NO2 content as simultaneously measured with a calibrated commercial reference sensor. Inset: Photograph of the Teflon sensor cell containing a ZnO field-effect transistor mounted on a heater. Adapted with permission from ref 98. Copyright 2013 Elsevier B.V.

voltage after recovery. After a number of cycles, equilibrium is reached, in which the threshold voltage at the end of the charging pulse corresponds to the NO2 concentration. The flow chart to determine the NO2 partial pressure from the dynamic electrical measurement of the sensor is as follows. First a lookup table is created. The three parameters, ν, T0, and Ea, describing the charging dynamics and the three parameters describing the recovery are used as input. Then the gate bias, the measurement protocol, and the temperature are set. For a given NO2 pressure, the threshold voltage is calculated as a function of time using eqs 1 and 3 or 4 for recovery or charging, respectively. In the calculation, a correction for probing the threshold voltage is implemented by taking into account the average charging and recovery time of the transfer curve measurement. The calculations are continued until convergence is reached, indicating that the dynamic equilibrium has been obtained. The minimum and maximum extracted voltages are stored together with the set NO2 pressure. The lookup table is completed for the relevant NO2 pressures and used as calibration curve. The gate bias protocol and temperature chosen in the model are applied to the transistor. As measurement protocol, the following parameters were used: 10 s gate bias pulse at 30 V for charging and 10 s at zero gate bias for recovery. The temperature was set at 200 °C. On the basis of the activation energies of charge trapping and release, the sensor is reversible at this temperature. The threshold voltage is experimentally extracted at the end of each charging and recovery pulse. To this end, a transfer curve is recorded, by measuring the sourcedrain current with increasing gate bias at a low drain bias of 2 V. To reduce charge trapping during the probing of the threshold voltage, the transfer curves were measured using a short integration time and a compliance stop. The threshold voltage is translated into NO2 concentrations using the lookup table. The discretization of the threshold voltage sets the NO2 concentration resolution. Finally, in the case that the transistor is sensitive to conventional gate bias stress, due to, for example, high background humidity, the lookup table has to be adapted. 782

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials



Optimization of the gate dielectric might solve this problem and should be the objective of future research in the field. In the experiments reported, molecular sieves were used to dry the ambient air. This procedure is sufficient to yield a reliable sensor. The ZnO transistor was stable in NO. The sensor in combination with the developed measurement protocol did also not show baseline drifts in compressed dry air, indicating that the sensor is insensitive to oxygen. Although the sensor dynamics and extracted parameters were determined from measurements performed in nitrogen, they could be successfully used to determine the NO2 concentration in the ambient air dried by the molecular sieves. The shelf life of the sensors in ambient was found to be over a year, and under dry conditions the sensor has been operated for multiple days without any drop in performance.

Review

ASSOCIATED CONTENT

S Supporting Information *

Video showing the sensor demonstrator detecting NO2 in realtime (MPEG). This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We would like to thank J. R. Meijboom, S. G. J. Mathijssen, W. S. C. Roelofs, J. H. Klootwijk, N. Vlietstra, and M. Spijkman from Philips Research Laboratories, E. C. P. Smits from TNO/ Holst, M. Kemerink from TU/E, and H. L. Gomes from the University of Faro.



CONCLUSIONS This paper presents an overview of NO2 sensing with fieldeffect transistors. The detection mechanism has been elucidated, and it is based on charge trapping. The charges are located at the gate dielectric. By using ZnO as model semiconductor, the steps for the realization of a functional NO2 sensor for real-time detection have been presented. First the dynamic response of the sensor was calculated using a phenomenological charge trapping model. The analytical model was then implemented in the sensor protocol to create a demonstrator. By simultaneously measuring the NO2 content with a calibrated reference sensor, the methodology was validated. The sensor is capable of detecting concentrations of NO2 as low as 40 ppb. Moreover, the sensor operates in ambient air. Apart from drying the air, no further precautions were taken, showing that the fabricated sensor is selective for NO2. The real-time detection of low concentrations of NO2 down to the ppb level sets a precedent for sensing with field-effect transistors. The results collected show that FETs are able to achieve excellent detection limits, fast response times, and appreciable selectivity. These encouraging findings will stimulate further research to improve sensing capabilities of FETs for NO2 detection. Several critical issues remain to be addressed, such as a deeper understanding of the interactions between NO2 and the semiconductor/dielectric layers at the microscopic level. Localization of the traps in the dielectric suggests that, for the trapping mechanism, the nature of the gate dielectric is crucial while the type of semiconductor is irrelevant. Identifying the microscopic nature of the traps, by investigating different dielectrics and surface treatments, should be one of the future areas of research. A full understanding of the phenomena, such as adsorption/desorption behavior and temperature dependence of the trapping, will further improve performance of NO2 field-effect transistor sensors. From the application point of view, the challenges are related to selectivity, temperature control, and signal kinetics.101 The selectivity is particularly important for monitoring vehicle exhaust emissions, which contain many species of vapors and gases that can interfere with the NO2 detection. In summary, field-effect transistors will provide the technology of choice for low cost, low power, highly sensitive, and fast response NO2 sensors. The results reviewed in this paper constitute a starting point and suggest that further future advancements will provide the “ideal” NO2 sensing transistor able to combine high sensitivity and selectivity, fast response and recovery, repeatability, and reliability.



REFERENCES

(1) Environmental Protection Agency (EPA). Air Pollution, 2013, www.epa.gov/air/nitrogenoxides/ (accessed 06.06.2013). (2) American Lung Association. Nitrogen Dioxide, www.lung.org/ healthy-air/outdoor/resources/nitrogen-dioxide.html (accessed 06.06.2013). (3) Environmental Protection Agency (EPA). Air Pollution, 2013 Air, Asthma trigger, www.epa.gov/asthma/no2.html (accessed 06.06.2013). (4) European Environmental Agency (EEA). Air quality in Europe, Report, 2012, www.eea.europa.eu/publications/air-quality-in-europe2012. (5) Wetchakuna, K.; Samerjaia, T.; Tamaekonga, N.; Liewhirana, C.; Siriwonga, C.; Kruefua, V.; Wisitsoraatb, A.; Tuantranontb, A.; Phanichphanta, S. Sens. Actuators, B 2011, 160, 580. (6) Inoue, T.; Ohtsuka, K.; Yoshida, Y.; Matsuura, Y.; Kajiyama, Y. Sens. Actuators, B 1995, 25, 388. (7) Di Franco, C.; Elia, A.; Spagnolo, V.; Scamarcio, G.; Lugarã, P. M.; Ieva, E.; Cioffi, N.; Torsi, L.; Bruno, G.; Losurdo, M.; Garcia, M.; Wolter, S.; Brown, A.; Ricco, M. Sensors 2009, 9, 3337. (8) Kuzmych, O.; Allen, B. L.; Star, A. Nanotechnology 2007, 18, 375502. (9) Taylor, D. R. J. Allergy Clin. Immunol. 2006, 117, 259. (10) Puckett, J. L.; George, S. C. Respir. Physiol. Neurobiol. 2008, 163, 166. (11) Swanson, J. J.; Watts, W. F.; Newman, R. A.; Ziebarth, R. R.; Kittelson, D. B. Environ. Sci. Technol. 2013, 47, 4521. (12) www.ecophysics.com/. (13) www.alphasense.com/industrial-sensors/index.html. (14) www.draeger.com/sites/en_uk/pages/industry/productselector. (15) Barsan, N.; Koziej, D.; Weimar, U. Sens. Actuators, B 2007, 121, 18. (16) Franke, M. E.; Koplin, T. J.; Simon, U. Small 2006, 2, 36. (17) McCue, J. T.; Ying, J. Y. Chem. Mater. 2007, 19, 1009. (18) Kim, I.-D.; Rothschild, A.; Lee, B. H.; Kim, D. Y.; Jo, S. M.; Tuller, H. L. Nano Lett. 2006, 6, 2009. (19) Meng, D.; Yamazaki, T.; Shen, Y.; Liu, Z.; Kikuta, T. Appl. Surf. Sci. 2009, 256, 1050. (20) Shen, Y.; Yamazaki, T.; Liu, Z.; Meng, D.; Kikuta, T.; Nakatani, N. Thin Solid Films 2009, 517, 2069. (21) Barsan, N.; Weimar, U. J. Electroceram. 2001, 7, 143. (22) Baratto, C.; Sberveglieri, G.; Onischuk, A.; Caruso, B.; Stasio, S. D. Sens. Actuators, B 2004, 100, 261. (23) www.appliedsensor.com/products/air_classification_module. (24) www.synkerainc.com/index.php?option=com_content&view= article&id=37:synkera-technologies. 783

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

(25) Sze, S. M. Physics of Semiconductor Devices; Wiley-Interscience: New York, 1981. (26) Roberts, M. E.; Sokolov, A. N.; Bao, Z. J. Mater. Chem. 2009, 19, 3351. (27) Lundstrom, K. I.; Shivaraman, M. S.; Svensson, C. M. J. Appl. Phys. 1975, 46, 3876. (28) Torsi, L.; Dodabalapur, A.; Sabbatini, L.; Zambonin, P. G. Sens. Actuators, B 2000, 67, 312. (29) Torsi, L.; Dodabalapur, A. Anal. Chem. 2005, 77, 380A. (30) Goldoni, A.; Petaccia, L.; Lizzit, S.; Larciprete, R. J. Phys.: Condens. Matter 2010, 22, 8. (31) Bouvet, M. Anal. Bioanal. Chem. 2006, 384, 366. (32) Li, B.; Lambeth, D. N. Nano Lett. 2008, 8, 3563. (33) See, K. C.; Becknell, A.; Miragliotta, J.; Katz, H. E. Adv. Mater. 2007, 19, 3322. (34) Huang, J.; Miragliotta, J.; Becknell, A.; Katz, H. E. J. Am. Chem. Soc. 2007, 129, 9366. (35) Mabeck, J. T.; Malliaras, G. G. Anal. Bioanal. Chem. 2006, 384, 343. (36) Dutta, S.; Dodabalapur, A. Sens. Actuators, B 2009, 143, 50. (37) Afzal, A.; Cioffi, N.; Sabbatini, L.; Torsi, L. Sens. Actuators, B 2012, 171, 25. (38) Tricoli, A.; Teleki, A.; Righettoni, M. Angew. Chem., Int. Ed. 2010, 49, 7632. (39) Das, A.; Dost, R.; Richardson, T.; Grell, M.; Morrison, J. J.; Turner, M. L. Adv. Mater. 2007, 19, 4018. (40) Marinelli, F.; Dell’Aquila, A.; Torsi, L.; Tey, J.; Suranna, G. P.; Mastrorilli, P.; Romanazzi, G.; Nobile, C. F.; Mhaisalkar, S. G.; Cioffi, N.; Palmisano, F. Sens. Actuators, B 2009, 140, 445. (41) Barillaro, G.; Diligenti, A.; Nannini, A.; Strambini, L. M.; Comini, E.; Sberveglieri, G. IEEE Sens. J. 2006, 6, 19. (42) Barillaro, G.; Strambini, L. M. Sens. Actuators, B 2008, 134, 585. (43) Mattmann, M.; Helbling, T.; Durrer, L.; Roman, C.; Hierold, C.; Pohle, R.; Fleischer, M. Appl. Phys. Lett. 2009, 94, 183502. (44) Peng, S.; Cho, K. J.; Qi, P. F.; Dai, H. J. Chem. Phys. Lett. 2004, 387, 271. (45) Helbling, T.; Pohle, R.; Durrer, L.; Stampfer, C.; Roman, C.; Jungen, A.; Fleischer, A.; Hierold, C. Sens. Actuators, B 2008, 132, 491. (46) McAlpine, M. C.; Ahmad, H.; Wang, D. W.; Heath, J. R. Nat. Mater. 2007, 6, 379. (47) Zhang, D. H.; Liu, Z. Q.; Li, C.; Tang, T.; Liu, X. L.; Han, S.; Lei, B.; Zhou, C. W. Nano Lett. 2004, 4, 1919. (48) Ferro, R.; Rodriguez, J. A.; Bertrand, P. Thin Solid Films 2008, 516, 2225. (49) Jing, Z. H.; Zhan, J. H. Adv. Mater. 2008, 20, 4547. (50) Liewhiran, C.; Phanichphant, S. Sensors 2007, 7, 185. (51) Lonergan, M. C.; Severin, E. J.; Doleman, B. J.; Beaber, S. A.; Grubbs, R. H.; Lewis, N. S. Chem. Mater. 1996, 8, 2298. (52) Barsan, N.; Weimar, U. J. Phys.: Condens. Matter 2003, 15, R813. (53) Wilson, D. M.; Hoyt, S.; Janata, J.; Booksh, K.; Obando, L. IEEE Sens. J. 2001, 1, 256. (54) Shouli, B.; Liangyuan, C.; Dianqing, L.; Wensheng, Y.; Pengcheng, Y.; Zhiyong, L.; Aifan, C.; Liu, C. C. Sens. Actuators, B 2010, 146, 129. (55) Kanan, S. M.; El-Kadri, O. M.; Abu-Yousef, I. A.; Kanan, M. C. Sensors 2009, 9, 8158. (56) Barsan, N.; Hubner, M.; Weimer, U. Sens. Actuators, B 2011, 157, 510. (57) Capone, S.; Mongelli, S.; Rella, R.; Siciliano, P.; Valli, L. Langmuir 1999, 15, 1748. (58) Kilinc, N.; Atilla, D.; Ozturk, S.; Gurek, A. G.; Ozturk, Z. Z.; Ahsen, V. Thin Solid Films 2009, 517, 6206. (59) Paul, S.; Joseph, M. Sens. Actuators, B 2009, 140, 439. (60) Cantalini, C.; Valentini, L.; Lozzi, L.; Armentano, I.; Kenny, J. M.; Santucci, S. Sens. Actuators, B 2003, 93, 333. (61) Read, D. H.; Martin, J. E. Adv. Funct. Mater. 2010, 20, 1577. (62) Crone, B. K.; Dodabalapur, A.; Sarpeshkar, R.; Gelperin, A.; Katz, H. E.; Bao, Z. J. Appl. Phys. 2002, 91, 10140. (63) Katz, H. E. Electroanalysis 2004, 16, 1837.

(64) Helbling, T.; Hierold, C.; Durrer, L.; Roman, C.; Pohle, R.; Fleischer, M. Phys. Status Solidi B 2008, 245, 2326. (65) Li, J.; Lu, Y. J.; Ye, Q.; Cinke, M.; Han, J.; Meyyappan, M. Nano Lett. 2003, 3, 929. (66) Zhang, J.; Boyd, A.; Tselev, A.; Paranjape, M.; Barbara, P. Appl. Phys. Lett. 2006, 88, 3. (67) Kong, J.; Franklin, N. R.; Zhou, C. W.; Chapline, M. G.; Peng, S.; Cho, K. J.; Dai, H. J. Science 2000, 287, 622. (68) Andringa, A.-M.; Meijboom, J. R.; Smits, E. P. C.; Mathijssen, S. G. J.; Blom, P. W. M.; de Leeuw, D. M. Adv. Funct. Mater. 2011, 21, 100. (69) Piliego, C.; Jarzab, D.; Gigli, G.; Chen, Z. H.; Facchetti, A.; Loi, M. A. Adv. Mater. 2009, 21, 1573. (70) Veres, J.; Ogier, S. D.; Leeming, S. W.; Cupertino, D. C.; Khaffaf, S. M. Adv. Funct. Mater. 2003, 13, 199. (71) Smits, E. C. P.; Anthopoulos, T. D.; Setayesh, S.; van Veenendaal, E.; Coehoorn, R.; Blom, P. W. M.; de Boer, B.; de Leeuw, D. M. Phys. Rev. B 2006, 73, 205316. (72) Ong, B. S.; Li, C.; Li, Y.; Wu, Y.; Loutfy, R. J. Am. Chem. Soc. 2007, 129, 2750. (73) Bashir, A.; Wobkenberg, P. H.; Smith, J.; Ball, J. M.; Adamopoulos, G.; Bradley, D. D. C.; Anthopoulos, T. D. Adv. Mater. 2009, 21, 2226. (74) Richards, T.; Sirringhaus, H. Appl. Phys. Lett. 2008, 92, 023512. (75) Pernstich, K. P.; Haas, S.; Oberhoff, D.; Goldmann, C.; Gundlach, D. J.; Batlogg, B.; Rashid, A. N.; Schitter, G. J. Appl. Phys. 2004, 96, 6431. (76) Kakalios, J.; Street, R. A.; Jackson, W. B. Phys. Rev. Lett. 1987, 59, 1037. (77) Chen, Y. F.; Huang, S. F. Phys. Rev. B 1991, 44, 13775. (78) Mathijssen, S. G. J.; Colle, M.; Gomes, H.; Smits, E. C. P.; de Boer, B.; McCulloch, I.; Bobbert, P. A.; de Leeuw, D. M. Adv. Mater. 2007, 19, 2785. (79) Gomes, H. L.; Stallinga, P.; Dinelli, F.; Murgia, M.; Biscarini, F.; de Leeuw, D. M.; Muccini, M.; Mullen, K. Polym. Adv. Technol. 2005, 16, 227. (80) Jackson, W. B.; Marshall, J. M.; Moyer, M. D. Phys. Rev. B 1989, 39, 1164. (81) Andringa, A.-M.; Vlietstra, N.; Smits, E. C. P.; Spijkman, M.-J.; Gomes, H. L.; Klootwijk, J. H.; Blom, P. M. W.; de Leeuw, D. M. Sens. Actuators, B 2012, 171, 1172. (82) Jones, D. A.; Lee, J. U. Nano Lett. 2011, 11, 4176. (83) Lu, G. Y.; Xu, J.; Sun, J. B.; Yu, Y. S.; Zhang, Y. Q.; Liu, F. M. Sens. Actuators, B 2012, 162, 82. (84) Comini, E.; Faglia, G.; Sberveglieri, G. Sens. Actuators, B 2001, 78, 73. (85) Muller, R. S.; Kamins, T. I. Device electronics for integrated circuits, 2nd ed.; John Wiley & Sons: New York, 1986. (86) Pingree, L. S. C.; Reid, O. G.; Ginger, D. S. Adv. Mater. 2009, 21, 19. (87) Palermo, V.; Palma, M.; Samori, P. Adv. Mater. 2006, 18, 145. (88) Mathijssen, S. G. J.; Spijkman, M.; Andringa, A.-M.; van Hal, P. A.; McCulloch, I.; Kemerink, M.; Janssen, R. A. J.; de Leeuw, D. M. Adv. Mater. 2010, 22, 5105. (89) Gholamrezaie, F.; Andringa, A.-M.; Roelofs, W. S. C.; Neuhold, A.; Kemerink, M.; Blom, P. W. M.; de Leeuw, D. M. Small 2012, 8, 241. (90) Andringa, A.-M.; Roelofs, W. S. C.; Sommer, M.; Thelakkkat, M.; Kemerink, M.; de Leeuw, D. M. Appl. Phys. Lett. 2012, 101, 153302. (91) Boudinet, D.; Benwadih, M.; Altazin, S.; Gwoziecki, R.; Verilhac, J. M.; Coppard, R.; Le Blevennec, G.; Chartier, I.; Horowitz, G. Org. Electron. 2010, 11, 291. (92) Facchetti, A.; Chen, Z.; Zheng, Y.; Quinn, J. R.; Newman, C. R.; Yan, H.; Nga, Ng, T.; Sambandan, S.; Lujan, R.; Arias, A.-C. Proc. SPIE 2009, 7487, 748702. (93) Charrier, D. S. H.; Kemerink, M.; Smalbrugge, B. E.; de Vries, T.; Janssen, R. A. J. ACS Nano 2008, 2, 622. 784

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785

Chemistry of Materials

Review

(94) Huttner, S.; Sommer, M.; Thelakkat, M. Appl. Phys. Lett. 2008, 92, 093302. (95) Lindner, S. M.; Thelakkat, M. Macromolecules 2004, 37, 8832. (96) Veres, J.; Ogier, S.; Lloyd, G.; de Leeuw, D. M. Chem. Mater. 2004, 16, 4543. (97) Brondijk, J. J.; Spijkman, M.; van Seijen, F.; Blom, P. W. M.; de Leeuw, D. M. Phys. Rev. B 2012, 85, 165310. (98) Andringa, A.-M.; Smits, E. C. P.; Klootwijk, J. H.; de Leeuw, D. M. Sens. Actuators, B 2013, 181, 668. (99) Mathijssen, S. G. J.; Kemerink, M.; Sharma, A.; Coelle, M.; Bobbert, P. A.; Janssen, R. A. J.; de Leeuw, D. M. Adv. Mater. 2008, 20, 975. (100) Bobbert, P. A.; Sharma, A.; Mathijssen, S. G. J.; Kemerink, M.; de Leeuw, D. M. Adv. Mater. 2012, 24, 1146. (101) Potyrailo, R. A.; Surman, C.; Nagraj, N.; Burns, A. Chem. Rev. 2011, 111, 7315.

785

dx.doi.org/10.1021/cm4020628 | Chem. Mater. 2014, 26, 773−785