Home
Search
Collections
Journals
About
Contact us
My IOPscience
Air-gating and chemical-gating in transistors and sensing devices made from hollow TiO2 semiconductor nanotubes
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2015 Nanotechnology 26 295203 (http://iopscience.iop.org/0957-4484/26/29/295203) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 130.237.165.40 This content was downloaded on 06/08/2015 at 21:17
Please note that terms and conditions apply.
Nanotechnology Nanotechnology 26 (2015) 295203 (9pp)
doi:10.1088/0957-4484/26/29/295203
Air-gating and chemical-gating in transistors and sensing devices made from hollow TiO2 semiconductor nanotubes Yahya Alivov1, Hans Funke1 and Prashant Nagpal1,2,3,4 1
Department of Chemical and Biological Engineering, University of Colorado, Boulder, USA Renewable and Sustainable Energy Institute, University of Colorado, Boulder, USA 3 BioFrontiers Institute, University of Colorado, Boulder, USA 4 Materials Science and Engineering, University of Colorado, Boulder, USA 2
E-mail: [email protected] and [email protected] Received 21 March 2015, revised 25 May 2015 Accepted for publication 4 June 2015 Published 2 July 2015 Abstract
Rapid miniaturization of electronic devices down to the nanoscale, according to Moore’s law, has led to some undesirable effects like high leakage current in transistors, which can offset additional benefits from scaling down. Development of three-dimensional transistors, by spatial extension in the third dimension, has allowed higher contact area with a gate electrode and better control over conductivity in the semiconductor channel. However, these devices do not utilize the large surface area and interfaces for new electronic functionality. Here, we demonstrate air gating and chemical gating in hollow semiconductor nanotube devices and highlight the potential for development of novel transistors that can be modulated using channel bias, gate voltage, chemical composition, and concentration. Using chemical gating, we reversibly altered the conductivity of nanoscaled semiconductor nanotubes (10–500 nm TiO2 nanotubes) by six orders of magnitude, with a tunable rectification factor (ON/OFF ratio) ranging from 1–106. While demonstrated air- and chemical-gating speeds were slow here (∼seconds) due to the mechanical-evacuation rate and size of our chamber, the small nanoscale volume of these hollow semiconductors can enable much higher switching speeds, limited by the rate of adsorption/ desorption of molecules at semiconductor interfaces. These chemical-gating effects are completely reversible, additive between different chemical compositions, and can enable semiconductor nanoelectronic devices for ‘chemical transistors’, ‘chemical diodes’, and very high-efficiency sensing applications. S Online supplementary data available from stacks.iop.org/NANO/26/295203/mmedia Keywords: air-gating, chemical-gating, TiO2 nanotubes 1. Introduction
or ON/OFF ratio) is changed with high switching speeds. However, the large surface area in these nanoscaled devices and the semiconductor interface also offers additional opportunities to control the electronic properties [1]. Here, we show novel air gating and chemical gating in hollow semiconductor nanotubes (10–500 nm titanium-dioxide nanotubes), where changing either the vapor concentration or composition reversibly varies the electronic properties. The semiconductor conductivity was reversibly tuned by six orders of magnitude with chemical gating, and the rectification factor changed from 1 to 106. These additional controls over the electronic
Undesirable effects in planar nanoscaled electronic devices, like leakage current and reduction in the rectification factor of transistors, are being addressed by spatial extension in the third dimension. These spatial three-dimensional (3D) transistors allow larger contact of the gate electrode with the semiconductor channel, thereby improving control of the charge carrier concentration and conductivity of the transistor. Using two degrees of control, gate voltage and channel bias, the channel conductivity and rectification (Iforward bias/Ireverse bias 0957-4484/15/295203+09$33.00
1
© 2015 IOP Publishing Ltd Printed in the UK
Y Alivov et al
Nanotechnology 26 (2015) 295203
properties of a thin semiconductor channel can lead to development of transistors that are modulated using vapor composition, concentration, channel bias, and gate voltage. Moreover, these TiO2 semiconductor nanostructures are key components of devices like solar cells, LEDs, photocatalysts, and photoelectrochemical cells [1–5]. The large surface area-to-volume ratio in nanotubes and nanowires is utilized for efficient light absorption, emission, photovoltaics, photodetection, and photocatalysis in these devices. The chemical gating demonstrated here, using reversible adsorption/desorption of gas molecules on the nanotube surface, can strongly alter the electronic properties of these semiconductor nanostructures and can be utilized in the appropriate design of devices. Adsorbed gas molecules create depleted or filled regions close to the surface in these devices, which alters the conduction channel and therefore the overall electrical resistance of hollow semiconductor nanotubes. This effect can also lead to highly sensitive gas and chemical sensors used to detect various chemical and biological molecules, which are critical in the control of chemical processes, environmental monitoring, agricultural and medical applications [6–10]. Furthermore, nanoscale sensors based on one-dimensional (1D) nanostructures (nanotubes, nanowires, nanobelts, etc) are excellent candidates for use as chemical sensors because of the enhanced sensitivity from very high surface-to-volume ratios and easily functionalizable surfaces to induce specificity. Here we demonstrate a different type of hollow semiconductor TiO2 nanotubes devices—‘air-gated transistors’, ‘chemical diodes’, ‘chemical transistors’, pressure sensors, altitude meters, gas sensors, and alcohol detectors—using: (1) a dramatic change in resistivity with a change in chemical composition and gas pressure; and (2) a strong change in the rectification factor with different gas molecules, reaching 106 for certain gases (ethanol, methanol, water vapor, and acetone).
2. Experimental setup TiO2 nanotubes were grown by anodization in an electrolyte consisting of ethylene glycol +1%NH4F + 2%H2O, using voltage to change the nanotube diameter and growth time to alter the length of the nanotubes (figure 1(a)) [1, 4, 11, 12]. More details on sample preparation and characterization of nanotubes can be found in the supplemental information [1, 11, 12]. As-grown TiO2 semiconductor nanotubes were n-type, as confirmed by scanning tunneling spectroscopy (STS) [11], due to oxygen vacancies (VO) acting as electron donors [1, 5, 11, 12]. The gas detection measurements were conducted in a vacuum chamber with an electrical feedthrough, which was pumped down to ∼1 mTorr base pressure using a rotary pump; the experimental setup is shown in figure 1(b). The gas sensing properties of the TiO2 nanotubes were measured by introducing different partial pressures of ethanol (C2H5OH), methanol (CH3OH), acetone ((CH3)2CO), water vapor (H2O), and carbon monoxide (CO) into the evacuated chamber, ranging from 300 mTorr to the saturation vapor pressures of the different gases at room temperature
Figure 1. (a) Scanning electron micrographs (SEMs) of the top view
of the densely packed TiO2 nanotube arrays grown by electrochemical anodization. (b) Schematic of experimental setup used during gas sensitivity measurements. (c) SEM image of the TiO2 nanotube wire prepared by anodization of Ti wire of 20 μm diameter; magnified area shows nanotube structure of the wire.
(47, 97, 195, and 20 Torr, respectively). Up to five different samples were used in experiments for each type of gas to verify reproducibility of the effects. The gas sensing properties were studied by measuring current–voltage (I–V) characteristics at different pressures, and the response of the sensor was defined as the relative change in current due to the analyte: ΔI = (Ig − Iv)/Iv × 100%, where Ig is the current of the sensor in the target gas and Iv is the current at vacuum (< 10 mTorr). The response time was defined as the time needed for the variation in electrical current to reach 90% of 2
Y Alivov et al
Nanotechnology 26 (2015) 295203
the equilibrium value after injecting the gas, and the recovery time similarly was the time required for the sensor to return to 90% of the original current in air after removing the gas. Sheet resistance of nanotubes (R) was calculated from I–V measurements using Ohm’s law I = VR . Then, resistivity
as the values of the ON/OFF ratio, were fully reproduced with each pumping cycle. To evaluate the effect of different adsorbed gases and identify the mechanism responsible for the air-gating effect, we conducted experiments with several gas molecules. Measurements with pure nitrogen, oxygen, and water vapor (the main components of air) showed that only adsorption of water vapor produced a strong gating effect, indicating that moisture is responsible for the strong air-gating effect (figures 2(d) and (e), compared to figures 3(a)–(c)). The changes in I–V characteristics due to ‘chemical gating’ of the TiO2 nanotubes depended strongly on the chemical nature of the adsorbing molecules, as shown in figures 3(a)–(c) for ethanol, methanol, acetone, and water vapor. Also, using several chemical vapors—ethanol, methanol, acetone, water vapor, nitrogen, and oxygen—we observed different degrees of ‘chemical-gating’ behavior on the observed I–V characteristics of TiO2 nanotubes (figures 3(a)–(c)). While all these gases converted the I–V characteristics from symmetric to asymmetric behavior, increasing the observed current with increasing partial pressure (figure 3(b)) and changing the rectification behavior (figure 3(c)), the saturation current corresponding to the vapor pressure also varied with the chemical nature of the gas. The highest sensitivity was observed for water vapor, followed by ethanol, methanol, and acetone. The ON/OFF ratio varied from 102 to 106, depending on the type of gas and its partial pressure, as seen in figure 3(c). These devices exhibit highly rectifying behavior after exposure to chemical gases and present a new type of diode—‘chemical diode’— operating in the ambient of certain gases, without the need for doping the material in the transistor. The threshold voltage of the diodes was usually in the range 2–4 V, the leakage current was low (varied in the range 10−9–10−6 A, depending on the resistivity of the sample in vacuum), and the breakdown was not seen until −20 V (figure 3(a)). The effect of the gases is highly reversible and stable; repeated cycles for a given gas produced the same results. The high reversibility is seen in figure 3(a), which presents I–V curves for four different gases taken from one nanotube device. After finishing measurements for one gas, the vacuum chamber was evacuated to base pressure, the next gas was introduced into the chamber, and measurements were repeated for the same sample. All these observed effects were tested for five samples for each type of gas. The stability was also tested by leaving the sample exposed to the vapor pressure of water for 3 d and measuring I–V (vacuum and repeated exposures to other gas molecules); comparison with initial I–V showed no notable change in the I–V diode-like behavior. The chemical-gating experiments revealed that the reversible adsorption of different gas molecules and the orders of magnitude change in conductance G is caused by a change in the carrier concentration of the TiO2 nanotubes. The conductance of nanotubes, which is inverse of the resistance R of the semiconducting nanotubes, is expressed as follows [13, 14]:
was calculated using equation R = ρ LA , where L is the nanotube length and A is the total nanotube cross-sectional area. The cross-sectional area A was estimated using the density of the nanotubes, the thickness of the nanotube wall, and the inner and outer diameters of the nanotube, estimated from SEM images. I–V characterization was performed using a Keithley source meter (Keithley 2612A, Tektronix Inc.). To increase sensitivity, ‘wired’ TiO2 nanotubes were prepared by converting Ti wire to TiO2 nanotubes by anodization [1]. A small section (2–3 mm) in the center of a 20 μm diameter Ti wire (ESPI Metals) was anodized to obtain TiO2 nanotubes sandwiched between two ‘naturally formed’ Ti metal electrical contacts (figure 1(c)). The two ends of the Ti-wire were embedded in epoxy resin, leaving the center section exposed for anodization and the ends exposed to form metal contacts for the electrical measurements. During growth, the resistance of the Ti wire was monitored using a multi-meter to identify the time for full transition of the Ti wire to TiO2 nanotubes. This transformation time was identified as the time when a jump in resistance was observed, indicating that all of the metal was oxidized. The wires were then annealed at 773 K in air to obtain the anatase phase of the TiO2. This metal-to-semiconductor transition was further confirmed by the dependence of the wired nanotubes’ resistivity on temperature, which showed an increase of conductivity with temperature, as is typical to semiconductors (increased rate of thermal electron-hole generation with temperature). Well-defined tubular structures of TiO2, similar to ones grown from Ti sheets, were confirmed by SEM (figure 1(c)).
3. Results and discussion One of the most striking observations was the ‘air gating’ in these semiconductor nanotubes. Figure 2(a) shows the I–V characteristics of TiO2 nanotubes before (red curve) and after (blue curve) pumping the vacuum chamber. The resistivity of these nanotubes increased by more than two orders of magnitude when the chamber was evacuated from atmospheric pressure (630 Torr in Boulder, CO) to 1 mTorr. Even more striking was the change in the rectification factor. The I–V curve exhibits a rectifying diode-like behavior at atmospheric pressure (rectification factor or ON/OFF ratio at ± 5 V = I+5V/I−5V) is 103, (figure 2(b)), and at low pressures it is symmetric (ON/OFF ratio 1 at 1 mTorr). The rectification decreases with decreasing pressure (figure 2(b)). The current as a function of pressure at +5 and −5 V increases linearly under reverse bias, and it increases super-linearly under forward bias (figure 2(c)). The dependence of the I–V characteristics on pressure was highly reversible; the transition from diode-like behavior to symmetric and vice versa, as well
G=
3
1 neμπd 2 = R L
(1)
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 2. (a) I–V characteristics of TiO2 nanotubes in logarithmic and linear (inset) scale. Red curve corresponds to I–V at atmospheric pressure, blue curve at vacuum (∼1 mTorr). Comparison of these curves makes a clear resistivity increase in vacuum and rectifying behavior. (b) I–V characteristics of TiO2 nanotubes as a function of pressure. Inset shows rectification factor (ON/OFF ratio) at 5 V. (c) Currents at reverse −5 V bias (blue) and forward 5 V bias (green) as a function of pressure. Dependence at reverse bias is linear, while the one at forward bias is super-linear, consistent with the rectification behavior of I–V at higher pressures. (d), (e) I–V characteristics of TiO2 nanotubes in (d) nitrogen and (e) oxygen ambient as a function of pressure. No effect of these gases is observed on TiO2 nanotube conductivity.
carrier concentration Δns:
where n is the initial carrier concentration, e the electronic charge, μ the mobility of the electrons, and d and L are the radius and length of the nanowire channel, respectively. During gas sensing, the change in conductance (ΔG) of the semiconducting nanotubes will result from a change in the
ΔG =
4
Δneμπd 2 L
(2)
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 3. Chemical gating: (a) I–V characteristics of TiO2 nanotube sensor in ambient of water vapor (curve 1), ethanol (curve 2), methanol (curve 3), and acetone (curve 4) at vapor pressures, which were 47, 93, 200, and 20 Torr, respectively. Highly pronounced diode-like behavior is observed with the ON/OFF ratio ranging from 103–106, depending on type of gas. (b) Current at 5 V for ethanol (red), methanol (green), and acetone (purple), corresponding to vapor pressure. (c) Rectification factor (ON/OFF ratio) as a function of partial pressure of different gases. (d) Carrier concentration as a function of pressure.
Hence, the sensitivity of the sensors can be defined as: ΔG Δn = . G n
(3)
The sensitivity of the sensor is linearly proportional to the change in carrier concentration Δn. The carrier concentration n was calculated analyzing I–V curves in the intermediate bias regime (see references [1] and [11] for details), while mobility was calculated using the equation: ρ=
1 1 ; →μ= neμ ρen
(4) Figure 4. STM spectra for TiO2 nanotubes before (red curve) and after (blue curve) exposing to ethanol. Both spectra were taken from the same point on TiO2 nanotube films. The shift of Fermi level toward the conduction band is clearly seen due to an increase of electron density after exposing to ethanol.
where ρ is resistivity, which was calculated from the sheet resistance R = V/I and the nanotube dimensions. The carrier concentration was 9.6 × 1012 cm−3 at the lowest base pressure, and it increased with gas pressure, reaching at vapor pressures
5
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 5. (a) I–V characteristics of undoped TiO2 nanotubes at different CO flow rates; (b) Current as a function of CO concentration; (c) I–V characteristics of Nb-doped TiO2 nanotubes at different CO flow rates; (d) Current as a function of CO concentration.
3.2 × 1019, 6.8 × 1018, 3.6 × 1018, and 1.12 × 1019 cm−3, respectively, for water vapor, ethanol, methanol, and acetone at saturation vapor pressure (figure 3(d)). The carrier mobility calculated using equation (4) showed a modest decrease from ∼7.1 to ∼4.7 cm2 V−1 s−1 when changing the pressure from the vapor pressure of the respective solvents to 2 mTorr (lowest pressure) and did not depend on the type of gas. The shift of the Fermi level of the nanotubes monitored using scanning tunneling microscopy (STM), before and after exposing to chemical gas (figure 4) (Fermi-level is shifted after exposure to ethanol), was also consistent with the observed change in carrier concentration n calculated using the Fermi energy [13]: n = NC e
E F − EC kT
where me* is the effective mass and h is the Planck’s constant. At 1 mTorr, Nc was 1.54 × 1014 cm−3 and increased to 2.5 × 1019 cm−3 after exposure to ethanol; these values are close to those calculated using equation (4). Furthermore, the chemical-gating effect, much like the field effect, provides an additional control by changing the chemical composition and gas pressure to the individual hollow nanotube channel. Since the charges still conduct under application of a voltage (VSD or V), but the carrier concentration is modulated by the partial pressure and chemical nature, the change in current can be related to chemical effect mobility, applied voltage (V), and nanotube dimensions as [13] ΔI = ΔQA/t and t = L/v and v = μCET E = μCET × V/L
(5)
ΔI =
where NC is the effective density of states in the conduction band, k is the Boltzmann constant, and T is the absolute temperature. The STS measurements in different gas environments were performed for the same position of the STM tip to prevent changes related to surface states, etc. The effective density of states NC is equal to ⎡ 2πm * kT ⎤3/2 e ⎥ NC = 2 ⎢ ⎣ h2 ⎦
μCET ΔQdV L
(7)
where μCET is the chemical-effect mobility and ΔQ is chemically induced charge carriers per unit area (A). Using the change in observed nanotube current with concentration (figures 3(b) and (d)), like field-effect transistors, chemical gating has a linear region for all gases (mlin, partial pressure < ∼ 10 Torr) and a saturation region (> ∼ 10 Torr) (figures 3(b)–(d)). While the mobility of chemically induced charge carriers (due to reversible adsorption of gases) was found to be similar to the field-induced charges
(6)
6
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 6. I–V characteristics of vacuum-annealed sample measured
in air (blue curve) and in low pressure (red curve).
(∼4 cm2 V−1 s−1), the sensitivity (ΔQ / ΔP) varies strongly in the linear region (from water (∼1000), ethanol (∼100), methanol (∼10), to acetone (∼1)) and the saturation regime. When the sensor is exposed to reducing gas molecules like water, ethanol, methanol, and acetone, molecules react with the pre-existing oxygen ions at the TiO2 nanotube surface to form CO2 and H2O, according to the following equations, and the electrons are released back to the TiO2 nanotubes [1, 10, 14]: Water: H 2 O(gas) + 2Ti Ti + O O ↔ ( Ti Ti − OH) + (OH)O+ + e−
(8)
H 2 O + 2Ti Ti + O ↔ 2 ( Ti Ti − OH) + V..O + 2e−
(9)
Figure 7. (a) I–V characteristics in logarithmic plot as a function of ethanol partial pressure. (b) Induced current signal by ethanol as a function of pressure; the currents correspond to forward bias +5 V. The sensitivity was increased to 8 mTorr compared to 20 mTorr for the planar structure.
Ethanol: CH 3CH 2 OH(gas) → CH 3CH 2 OH(ads)
(10)
effect due to the strongly charge-depleted nanotube surface under vacuum. Similar studies of CO chemical gating and gas sensing by placing the sample in the flowing mixed gas CO and N2 showed very small changes compared to reducing gases (resistivity increases and carrier concentration decreases; see figures 5(a) and (b)). To increase the sensitivity of these nanotubes for oxidizing gases or electron acceptors like CO, either we can electrically inject charges, or the nanotubes can be doped with electron dopants. We used Niobium (Nb) doping [12] to sensitize the nanotubes towards oxidizing gases (figures 5(c) and (d), compared to figures 5(a) and (b)), which leads to about a six-fold higher chemical gating and sensing from CO. It was observed that at lower CO concentrations (8 × 104 ppm), the current starts saturating. The current saturation can be explained by the limited number of surface states for CO molecules to accept electrons, as the TiO2 nanotube surface is already ‘exhausted’ with a small number of free electrons after removal of adsorbed oxygen/water vapor atoms. These oxygen vacancies are nominally formed in the as-grown TiO2 nanotubes during annealing at 500 °C for 1 h to convert from the amorphous to the anatase phase. During this annealing, oxygen species are adsorbed on the TiO2 nanotube and fill all available vacant sites on TiO2 nanotubes [10, 14]. To test this
−
CH 3CH 2 OH(ads) + 6O (ads) → 2CO2 (gas) + 3H 2 O(gas) + 6e−
(11)
Methanol: CH 3OH(gas) → CH 3OH(ads) CH 3OH + O− → CO2 (gas) + 2H 2 O(gas) + 3e−
(12) (13)
Acetone: CH 3COCH 3(gas) + O− → CH 3C+O + CH 3O− + e− (14) CH 3C+O → C+H 3 + CO CO + O− → CO2 + e−
(15) (16)
This leads to an increase in carrier concentration in the TiO2 nanotube walls and a decrease in the surface depletion layer width [1], and, therefore, to an increase of the electrical current in the TiO2 nanotube sensors. Methanol and acetone adsorption reactions are similar (see equations (12)–(16)), where each methanol molecule donates three electrons 3e− to TiO2 nanotubes, and acetone donates one electron e−, which can explain lower currents for these components (figure 3). However, other oxidizing gases like carbon monoxide (CO) on TiO2 nanotubes, do not produce a striking chemical-gating 7
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 8. (a) Dynamic response of the TiO2 nanotube for ethanol gas. (b) Results of ‘blowing experiments’. I–V characteristics of TiO2
nanotubes in ethanol vapor pressure after removal of vacuum chamber cap and after blowing with warm air for 1 min. Dramatic reduction of current is seen after each procedure. (c) Induced maximum current at vapor pressures for different gases—water vapor, ethanol, methanol, and acetone. (d) Results of alcohol testing experiments: red curve corresponds to the sum of ethanol and water data, when water and ethanol vapors were injected consecutively in the vacuum chamber; blue curve corresponds to the I–V when a mixture of liquid ethanol and water was used as a vapor source.
hypothesis, we annealed TiO2 nanotubes in ultra-high vacuum (base pressure ∼10−7 Torr at 500 °C for 3 h) to desorb all adsorbed molecules and to recover/generate oxygen vacancies, which act as shallow donors [1, 5, 11, 12]. The results of I–V characterization of vacuum-annealed samples showed no chemical gating or gas sensing with pressure on exposure to any gases studied in this work (figure 6). This result shows that adsorption of oxygen to TiO2 nanotubes is crucial for chemical gating and gas detection. (further discussion can be found in supplemental information) [1]. While these studies were performed using one electrode connected to the metallic Ti sheet and the second to the TiO2 nanotubes from the top, leading to formation of a Schottky junction, another device formed by converting Ti wire to TiO2 nanotubes was also fabricated to evaluate the effects of symmetric metal contacts (figure 1(c)). In this device, both ‘naturally formed’ electrical contacts are identical, and the I– V curves were symmetrical at all gas pressures (figure 7(a)). The symmetric I–V behavior is explained by a back-to-back Schottky barrier [1] configuration of Ti-TiO2-Ti, and the wired TiO2 nanotubes showed a 2.5 times increase in
sensitivity compared to the asymmetric planar structure, as shown in figure 7(b). (The minimum detected analyte concentration was 12 ppm compared to 31 ppm for the planar structure.) The response of the TiO2 nanotube array sensors to gases was very fast and reproducible for repeated cycles, as shown in our experiments. Figure 8(a) presents results for response and recovery studies (response time ∼16 s, recovery time ∼10 s) for ethanol, limited by the size of our experimental setup and the vacuum pump. The small size of the nanoscaled tubes can be used to achieve much faster switching and sensing times, limited by the adsorption/desorption of gas molecules on the nanotube surface. The fast response of TiO2 nanotubes was demonstrated by ‘blowing’ experiments, where the exposed nanotube sensor recovered instantly (the current dropped by four orders of magnitude in
Search
Collections
Journals
About
Contact us
My IOPscience
Air-gating and chemical-gating in transistors and sensing devices made from hollow TiO2 semiconductor nanotubes
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2015 Nanotechnology 26 295203 (http://iopscience.iop.org/0957-4484/26/29/295203) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 130.237.165.40 This content was downloaded on 06/08/2015 at 21:17
Please note that terms and conditions apply.
Nanotechnology Nanotechnology 26 (2015) 295203 (9pp)
doi:10.1088/0957-4484/26/29/295203
Air-gating and chemical-gating in transistors and sensing devices made from hollow TiO2 semiconductor nanotubes Yahya Alivov1, Hans Funke1 and Prashant Nagpal1,2,3,4 1
Department of Chemical and Biological Engineering, University of Colorado, Boulder, USA Renewable and Sustainable Energy Institute, University of Colorado, Boulder, USA 3 BioFrontiers Institute, University of Colorado, Boulder, USA 4 Materials Science and Engineering, University of Colorado, Boulder, USA 2
E-mail: [email protected] and [email protected] Received 21 March 2015, revised 25 May 2015 Accepted for publication 4 June 2015 Published 2 July 2015 Abstract
Rapid miniaturization of electronic devices down to the nanoscale, according to Moore’s law, has led to some undesirable effects like high leakage current in transistors, which can offset additional benefits from scaling down. Development of three-dimensional transistors, by spatial extension in the third dimension, has allowed higher contact area with a gate electrode and better control over conductivity in the semiconductor channel. However, these devices do not utilize the large surface area and interfaces for new electronic functionality. Here, we demonstrate air gating and chemical gating in hollow semiconductor nanotube devices and highlight the potential for development of novel transistors that can be modulated using channel bias, gate voltage, chemical composition, and concentration. Using chemical gating, we reversibly altered the conductivity of nanoscaled semiconductor nanotubes (10–500 nm TiO2 nanotubes) by six orders of magnitude, with a tunable rectification factor (ON/OFF ratio) ranging from 1–106. While demonstrated air- and chemical-gating speeds were slow here (∼seconds) due to the mechanical-evacuation rate and size of our chamber, the small nanoscale volume of these hollow semiconductors can enable much higher switching speeds, limited by the rate of adsorption/ desorption of molecules at semiconductor interfaces. These chemical-gating effects are completely reversible, additive between different chemical compositions, and can enable semiconductor nanoelectronic devices for ‘chemical transistors’, ‘chemical diodes’, and very high-efficiency sensing applications. S Online supplementary data available from stacks.iop.org/NANO/26/295203/mmedia Keywords: air-gating, chemical-gating, TiO2 nanotubes 1. Introduction
or ON/OFF ratio) is changed with high switching speeds. However, the large surface area in these nanoscaled devices and the semiconductor interface also offers additional opportunities to control the electronic properties [1]. Here, we show novel air gating and chemical gating in hollow semiconductor nanotubes (10–500 nm titanium-dioxide nanotubes), where changing either the vapor concentration or composition reversibly varies the electronic properties. The semiconductor conductivity was reversibly tuned by six orders of magnitude with chemical gating, and the rectification factor changed from 1 to 106. These additional controls over the electronic
Undesirable effects in planar nanoscaled electronic devices, like leakage current and reduction in the rectification factor of transistors, are being addressed by spatial extension in the third dimension. These spatial three-dimensional (3D) transistors allow larger contact of the gate electrode with the semiconductor channel, thereby improving control of the charge carrier concentration and conductivity of the transistor. Using two degrees of control, gate voltage and channel bias, the channel conductivity and rectification (Iforward bias/Ireverse bias 0957-4484/15/295203+09$33.00
1
© 2015 IOP Publishing Ltd Printed in the UK
Y Alivov et al
Nanotechnology 26 (2015) 295203
properties of a thin semiconductor channel can lead to development of transistors that are modulated using vapor composition, concentration, channel bias, and gate voltage. Moreover, these TiO2 semiconductor nanostructures are key components of devices like solar cells, LEDs, photocatalysts, and photoelectrochemical cells [1–5]. The large surface area-to-volume ratio in nanotubes and nanowires is utilized for efficient light absorption, emission, photovoltaics, photodetection, and photocatalysis in these devices. The chemical gating demonstrated here, using reversible adsorption/desorption of gas molecules on the nanotube surface, can strongly alter the electronic properties of these semiconductor nanostructures and can be utilized in the appropriate design of devices. Adsorbed gas molecules create depleted or filled regions close to the surface in these devices, which alters the conduction channel and therefore the overall electrical resistance of hollow semiconductor nanotubes. This effect can also lead to highly sensitive gas and chemical sensors used to detect various chemical and biological molecules, which are critical in the control of chemical processes, environmental monitoring, agricultural and medical applications [6–10]. Furthermore, nanoscale sensors based on one-dimensional (1D) nanostructures (nanotubes, nanowires, nanobelts, etc) are excellent candidates for use as chemical sensors because of the enhanced sensitivity from very high surface-to-volume ratios and easily functionalizable surfaces to induce specificity. Here we demonstrate a different type of hollow semiconductor TiO2 nanotubes devices—‘air-gated transistors’, ‘chemical diodes’, ‘chemical transistors’, pressure sensors, altitude meters, gas sensors, and alcohol detectors—using: (1) a dramatic change in resistivity with a change in chemical composition and gas pressure; and (2) a strong change in the rectification factor with different gas molecules, reaching 106 for certain gases (ethanol, methanol, water vapor, and acetone).
2. Experimental setup TiO2 nanotubes were grown by anodization in an electrolyte consisting of ethylene glycol +1%NH4F + 2%H2O, using voltage to change the nanotube diameter and growth time to alter the length of the nanotubes (figure 1(a)) [1, 4, 11, 12]. More details on sample preparation and characterization of nanotubes can be found in the supplemental information [1, 11, 12]. As-grown TiO2 semiconductor nanotubes were n-type, as confirmed by scanning tunneling spectroscopy (STS) [11], due to oxygen vacancies (VO) acting as electron donors [1, 5, 11, 12]. The gas detection measurements were conducted in a vacuum chamber with an electrical feedthrough, which was pumped down to ∼1 mTorr base pressure using a rotary pump; the experimental setup is shown in figure 1(b). The gas sensing properties of the TiO2 nanotubes were measured by introducing different partial pressures of ethanol (C2H5OH), methanol (CH3OH), acetone ((CH3)2CO), water vapor (H2O), and carbon monoxide (CO) into the evacuated chamber, ranging from 300 mTorr to the saturation vapor pressures of the different gases at room temperature
Figure 1. (a) Scanning electron micrographs (SEMs) of the top view
of the densely packed TiO2 nanotube arrays grown by electrochemical anodization. (b) Schematic of experimental setup used during gas sensitivity measurements. (c) SEM image of the TiO2 nanotube wire prepared by anodization of Ti wire of 20 μm diameter; magnified area shows nanotube structure of the wire.
(47, 97, 195, and 20 Torr, respectively). Up to five different samples were used in experiments for each type of gas to verify reproducibility of the effects. The gas sensing properties were studied by measuring current–voltage (I–V) characteristics at different pressures, and the response of the sensor was defined as the relative change in current due to the analyte: ΔI = (Ig − Iv)/Iv × 100%, where Ig is the current of the sensor in the target gas and Iv is the current at vacuum (< 10 mTorr). The response time was defined as the time needed for the variation in electrical current to reach 90% of 2
Y Alivov et al
Nanotechnology 26 (2015) 295203
the equilibrium value after injecting the gas, and the recovery time similarly was the time required for the sensor to return to 90% of the original current in air after removing the gas. Sheet resistance of nanotubes (R) was calculated from I–V measurements using Ohm’s law I = VR . Then, resistivity
as the values of the ON/OFF ratio, were fully reproduced with each pumping cycle. To evaluate the effect of different adsorbed gases and identify the mechanism responsible for the air-gating effect, we conducted experiments with several gas molecules. Measurements with pure nitrogen, oxygen, and water vapor (the main components of air) showed that only adsorption of water vapor produced a strong gating effect, indicating that moisture is responsible for the strong air-gating effect (figures 2(d) and (e), compared to figures 3(a)–(c)). The changes in I–V characteristics due to ‘chemical gating’ of the TiO2 nanotubes depended strongly on the chemical nature of the adsorbing molecules, as shown in figures 3(a)–(c) for ethanol, methanol, acetone, and water vapor. Also, using several chemical vapors—ethanol, methanol, acetone, water vapor, nitrogen, and oxygen—we observed different degrees of ‘chemical-gating’ behavior on the observed I–V characteristics of TiO2 nanotubes (figures 3(a)–(c)). While all these gases converted the I–V characteristics from symmetric to asymmetric behavior, increasing the observed current with increasing partial pressure (figure 3(b)) and changing the rectification behavior (figure 3(c)), the saturation current corresponding to the vapor pressure also varied with the chemical nature of the gas. The highest sensitivity was observed for water vapor, followed by ethanol, methanol, and acetone. The ON/OFF ratio varied from 102 to 106, depending on the type of gas and its partial pressure, as seen in figure 3(c). These devices exhibit highly rectifying behavior after exposure to chemical gases and present a new type of diode—‘chemical diode’— operating in the ambient of certain gases, without the need for doping the material in the transistor. The threshold voltage of the diodes was usually in the range 2–4 V, the leakage current was low (varied in the range 10−9–10−6 A, depending on the resistivity of the sample in vacuum), and the breakdown was not seen until −20 V (figure 3(a)). The effect of the gases is highly reversible and stable; repeated cycles for a given gas produced the same results. The high reversibility is seen in figure 3(a), which presents I–V curves for four different gases taken from one nanotube device. After finishing measurements for one gas, the vacuum chamber was evacuated to base pressure, the next gas was introduced into the chamber, and measurements were repeated for the same sample. All these observed effects were tested for five samples for each type of gas. The stability was also tested by leaving the sample exposed to the vapor pressure of water for 3 d and measuring I–V (vacuum and repeated exposures to other gas molecules); comparison with initial I–V showed no notable change in the I–V diode-like behavior. The chemical-gating experiments revealed that the reversible adsorption of different gas molecules and the orders of magnitude change in conductance G is caused by a change in the carrier concentration of the TiO2 nanotubes. The conductance of nanotubes, which is inverse of the resistance R of the semiconducting nanotubes, is expressed as follows [13, 14]:
was calculated using equation R = ρ LA , where L is the nanotube length and A is the total nanotube cross-sectional area. The cross-sectional area A was estimated using the density of the nanotubes, the thickness of the nanotube wall, and the inner and outer diameters of the nanotube, estimated from SEM images. I–V characterization was performed using a Keithley source meter (Keithley 2612A, Tektronix Inc.). To increase sensitivity, ‘wired’ TiO2 nanotubes were prepared by converting Ti wire to TiO2 nanotubes by anodization [1]. A small section (2–3 mm) in the center of a 20 μm diameter Ti wire (ESPI Metals) was anodized to obtain TiO2 nanotubes sandwiched between two ‘naturally formed’ Ti metal electrical contacts (figure 1(c)). The two ends of the Ti-wire were embedded in epoxy resin, leaving the center section exposed for anodization and the ends exposed to form metal contacts for the electrical measurements. During growth, the resistance of the Ti wire was monitored using a multi-meter to identify the time for full transition of the Ti wire to TiO2 nanotubes. This transformation time was identified as the time when a jump in resistance was observed, indicating that all of the metal was oxidized. The wires were then annealed at 773 K in air to obtain the anatase phase of the TiO2. This metal-to-semiconductor transition was further confirmed by the dependence of the wired nanotubes’ resistivity on temperature, which showed an increase of conductivity with temperature, as is typical to semiconductors (increased rate of thermal electron-hole generation with temperature). Well-defined tubular structures of TiO2, similar to ones grown from Ti sheets, were confirmed by SEM (figure 1(c)).
3. Results and discussion One of the most striking observations was the ‘air gating’ in these semiconductor nanotubes. Figure 2(a) shows the I–V characteristics of TiO2 nanotubes before (red curve) and after (blue curve) pumping the vacuum chamber. The resistivity of these nanotubes increased by more than two orders of magnitude when the chamber was evacuated from atmospheric pressure (630 Torr in Boulder, CO) to 1 mTorr. Even more striking was the change in the rectification factor. The I–V curve exhibits a rectifying diode-like behavior at atmospheric pressure (rectification factor or ON/OFF ratio at ± 5 V = I+5V/I−5V) is 103, (figure 2(b)), and at low pressures it is symmetric (ON/OFF ratio 1 at 1 mTorr). The rectification decreases with decreasing pressure (figure 2(b)). The current as a function of pressure at +5 and −5 V increases linearly under reverse bias, and it increases super-linearly under forward bias (figure 2(c)). The dependence of the I–V characteristics on pressure was highly reversible; the transition from diode-like behavior to symmetric and vice versa, as well
G=
3
1 neμπd 2 = R L
(1)
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 2. (a) I–V characteristics of TiO2 nanotubes in logarithmic and linear (inset) scale. Red curve corresponds to I–V at atmospheric pressure, blue curve at vacuum (∼1 mTorr). Comparison of these curves makes a clear resistivity increase in vacuum and rectifying behavior. (b) I–V characteristics of TiO2 nanotubes as a function of pressure. Inset shows rectification factor (ON/OFF ratio) at 5 V. (c) Currents at reverse −5 V bias (blue) and forward 5 V bias (green) as a function of pressure. Dependence at reverse bias is linear, while the one at forward bias is super-linear, consistent with the rectification behavior of I–V at higher pressures. (d), (e) I–V characteristics of TiO2 nanotubes in (d) nitrogen and (e) oxygen ambient as a function of pressure. No effect of these gases is observed on TiO2 nanotube conductivity.
carrier concentration Δns:
where n is the initial carrier concentration, e the electronic charge, μ the mobility of the electrons, and d and L are the radius and length of the nanowire channel, respectively. During gas sensing, the change in conductance (ΔG) of the semiconducting nanotubes will result from a change in the
ΔG =
4
Δneμπd 2 L
(2)
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 3. Chemical gating: (a) I–V characteristics of TiO2 nanotube sensor in ambient of water vapor (curve 1), ethanol (curve 2), methanol (curve 3), and acetone (curve 4) at vapor pressures, which were 47, 93, 200, and 20 Torr, respectively. Highly pronounced diode-like behavior is observed with the ON/OFF ratio ranging from 103–106, depending on type of gas. (b) Current at 5 V for ethanol (red), methanol (green), and acetone (purple), corresponding to vapor pressure. (c) Rectification factor (ON/OFF ratio) as a function of partial pressure of different gases. (d) Carrier concentration as a function of pressure.
Hence, the sensitivity of the sensors can be defined as: ΔG Δn = . G n
(3)
The sensitivity of the sensor is linearly proportional to the change in carrier concentration Δn. The carrier concentration n was calculated analyzing I–V curves in the intermediate bias regime (see references [1] and [11] for details), while mobility was calculated using the equation: ρ=
1 1 ; →μ= neμ ρen
(4) Figure 4. STM spectra for TiO2 nanotubes before (red curve) and after (blue curve) exposing to ethanol. Both spectra were taken from the same point on TiO2 nanotube films. The shift of Fermi level toward the conduction band is clearly seen due to an increase of electron density after exposing to ethanol.
where ρ is resistivity, which was calculated from the sheet resistance R = V/I and the nanotube dimensions. The carrier concentration was 9.6 × 1012 cm−3 at the lowest base pressure, and it increased with gas pressure, reaching at vapor pressures
5
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 5. (a) I–V characteristics of undoped TiO2 nanotubes at different CO flow rates; (b) Current as a function of CO concentration; (c) I–V characteristics of Nb-doped TiO2 nanotubes at different CO flow rates; (d) Current as a function of CO concentration.
3.2 × 1019, 6.8 × 1018, 3.6 × 1018, and 1.12 × 1019 cm−3, respectively, for water vapor, ethanol, methanol, and acetone at saturation vapor pressure (figure 3(d)). The carrier mobility calculated using equation (4) showed a modest decrease from ∼7.1 to ∼4.7 cm2 V−1 s−1 when changing the pressure from the vapor pressure of the respective solvents to 2 mTorr (lowest pressure) and did not depend on the type of gas. The shift of the Fermi level of the nanotubes monitored using scanning tunneling microscopy (STM), before and after exposing to chemical gas (figure 4) (Fermi-level is shifted after exposure to ethanol), was also consistent with the observed change in carrier concentration n calculated using the Fermi energy [13]: n = NC e
E F − EC kT
where me* is the effective mass and h is the Planck’s constant. At 1 mTorr, Nc was 1.54 × 1014 cm−3 and increased to 2.5 × 1019 cm−3 after exposure to ethanol; these values are close to those calculated using equation (4). Furthermore, the chemical-gating effect, much like the field effect, provides an additional control by changing the chemical composition and gas pressure to the individual hollow nanotube channel. Since the charges still conduct under application of a voltage (VSD or V), but the carrier concentration is modulated by the partial pressure and chemical nature, the change in current can be related to chemical effect mobility, applied voltage (V), and nanotube dimensions as [13] ΔI = ΔQA/t and t = L/v and v = μCET E = μCET × V/L
(5)
ΔI =
where NC is the effective density of states in the conduction band, k is the Boltzmann constant, and T is the absolute temperature. The STS measurements in different gas environments were performed for the same position of the STM tip to prevent changes related to surface states, etc. The effective density of states NC is equal to ⎡ 2πm * kT ⎤3/2 e ⎥ NC = 2 ⎢ ⎣ h2 ⎦
μCET ΔQdV L
(7)
where μCET is the chemical-effect mobility and ΔQ is chemically induced charge carriers per unit area (A). Using the change in observed nanotube current with concentration (figures 3(b) and (d)), like field-effect transistors, chemical gating has a linear region for all gases (mlin, partial pressure < ∼ 10 Torr) and a saturation region (> ∼ 10 Torr) (figures 3(b)–(d)). While the mobility of chemically induced charge carriers (due to reversible adsorption of gases) was found to be similar to the field-induced charges
(6)
6
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 6. I–V characteristics of vacuum-annealed sample measured
in air (blue curve) and in low pressure (red curve).
(∼4 cm2 V−1 s−1), the sensitivity (ΔQ / ΔP) varies strongly in the linear region (from water (∼1000), ethanol (∼100), methanol (∼10), to acetone (∼1)) and the saturation regime. When the sensor is exposed to reducing gas molecules like water, ethanol, methanol, and acetone, molecules react with the pre-existing oxygen ions at the TiO2 nanotube surface to form CO2 and H2O, according to the following equations, and the electrons are released back to the TiO2 nanotubes [1, 10, 14]: Water: H 2 O(gas) + 2Ti Ti + O O ↔ ( Ti Ti − OH) + (OH)O+ + e−
(8)
H 2 O + 2Ti Ti + O ↔ 2 ( Ti Ti − OH) + V..O + 2e−
(9)
Figure 7. (a) I–V characteristics in logarithmic plot as a function of ethanol partial pressure. (b) Induced current signal by ethanol as a function of pressure; the currents correspond to forward bias +5 V. The sensitivity was increased to 8 mTorr compared to 20 mTorr for the planar structure.
Ethanol: CH 3CH 2 OH(gas) → CH 3CH 2 OH(ads)
(10)
effect due to the strongly charge-depleted nanotube surface under vacuum. Similar studies of CO chemical gating and gas sensing by placing the sample in the flowing mixed gas CO and N2 showed very small changes compared to reducing gases (resistivity increases and carrier concentration decreases; see figures 5(a) and (b)). To increase the sensitivity of these nanotubes for oxidizing gases or electron acceptors like CO, either we can electrically inject charges, or the nanotubes can be doped with electron dopants. We used Niobium (Nb) doping [12] to sensitize the nanotubes towards oxidizing gases (figures 5(c) and (d), compared to figures 5(a) and (b)), which leads to about a six-fold higher chemical gating and sensing from CO. It was observed that at lower CO concentrations (8 × 104 ppm), the current starts saturating. The current saturation can be explained by the limited number of surface states for CO molecules to accept electrons, as the TiO2 nanotube surface is already ‘exhausted’ with a small number of free electrons after removal of adsorbed oxygen/water vapor atoms. These oxygen vacancies are nominally formed in the as-grown TiO2 nanotubes during annealing at 500 °C for 1 h to convert from the amorphous to the anatase phase. During this annealing, oxygen species are adsorbed on the TiO2 nanotube and fill all available vacant sites on TiO2 nanotubes [10, 14]. To test this
−
CH 3CH 2 OH(ads) + 6O (ads) → 2CO2 (gas) + 3H 2 O(gas) + 6e−
(11)
Methanol: CH 3OH(gas) → CH 3OH(ads) CH 3OH + O− → CO2 (gas) + 2H 2 O(gas) + 3e−
(12) (13)
Acetone: CH 3COCH 3(gas) + O− → CH 3C+O + CH 3O− + e− (14) CH 3C+O → C+H 3 + CO CO + O− → CO2 + e−
(15) (16)
This leads to an increase in carrier concentration in the TiO2 nanotube walls and a decrease in the surface depletion layer width [1], and, therefore, to an increase of the electrical current in the TiO2 nanotube sensors. Methanol and acetone adsorption reactions are similar (see equations (12)–(16)), where each methanol molecule donates three electrons 3e− to TiO2 nanotubes, and acetone donates one electron e−, which can explain lower currents for these components (figure 3). However, other oxidizing gases like carbon monoxide (CO) on TiO2 nanotubes, do not produce a striking chemical-gating 7
Y Alivov et al
Nanotechnology 26 (2015) 295203
Figure 8. (a) Dynamic response of the TiO2 nanotube for ethanol gas. (b) Results of ‘blowing experiments’. I–V characteristics of TiO2
nanotubes in ethanol vapor pressure after removal of vacuum chamber cap and after blowing with warm air for 1 min. Dramatic reduction of current is seen after each procedure. (c) Induced maximum current at vapor pressures for different gases—water vapor, ethanol, methanol, and acetone. (d) Results of alcohol testing experiments: red curve corresponds to the sum of ethanol and water data, when water and ethanol vapors were injected consecutively in the vacuum chamber; blue curve corresponds to the I–V when a mixture of liquid ethanol and water was used as a vapor source.
hypothesis, we annealed TiO2 nanotubes in ultra-high vacuum (base pressure ∼10−7 Torr at 500 °C for 3 h) to desorb all adsorbed molecules and to recover/generate oxygen vacancies, which act as shallow donors [1, 5, 11, 12]. The results of I–V characterization of vacuum-annealed samples showed no chemical gating or gas sensing with pressure on exposure to any gases studied in this work (figure 6). This result shows that adsorption of oxygen to TiO2 nanotubes is crucial for chemical gating and gas detection. (further discussion can be found in supplemental information) [1]. While these studies were performed using one electrode connected to the metallic Ti sheet and the second to the TiO2 nanotubes from the top, leading to formation of a Schottky junction, another device formed by converting Ti wire to TiO2 nanotubes was also fabricated to evaluate the effects of symmetric metal contacts (figure 1(c)). In this device, both ‘naturally formed’ electrical contacts are identical, and the I– V curves were symmetrical at all gas pressures (figure 7(a)). The symmetric I–V behavior is explained by a back-to-back Schottky barrier [1] configuration of Ti-TiO2-Ti, and the wired TiO2 nanotubes showed a 2.5 times increase in
sensitivity compared to the asymmetric planar structure, as shown in figure 7(b). (The minimum detected analyte concentration was 12 ppm compared to 31 ppm for the planar structure.) The response of the TiO2 nanotube array sensors to gases was very fast and reproducible for repeated cycles, as shown in our experiments. Figure 8(a) presents results for response and recovery studies (response time ∼16 s, recovery time ∼10 s) for ethanol, limited by the size of our experimental setup and the vacuum pump. The small size of the nanoscaled tubes can be used to achieve much faster switching and sensing times, limited by the adsorption/desorption of gas molecules on the nanotube surface. The fast response of TiO2 nanotubes was demonstrated by ‘blowing’ experiments, where the exposed nanotube sensor recovered instantly (the current dropped by four orders of magnitude in
