Sensors and Actuators B 156 (2011) 276–282
Contents lists available at ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Room-temperature CO2 sensing using metal–insulator–semiconductor capacitors comprising atomic-layer-deposited La2 O3 thin films K.B. Jinesh a,∗ , V.A.T. Dam a , J. Swerts b , C. de Nooijer a , S. van Elshocht b , S.H. Brongersma a , M. Crego-Calama a a b
Holst Centre/imec-Netherlands, PO Box 8550, 5605 KN, Eindhoven, The Netherlands imec, Kapeldreef 75, B-3001 Leuven, Belgium
a r t i c l e
i n f o
Article history: Received 29 November 2010 Received in revised form 4 April 2011 Accepted 11 April 2011 Available online 20 April 2011
a b s t r a c t Room temperature detection of CO2 using metal–insulator–silicon (MIS) devices is reported. These devices comprise atomic layer deposited La2 O3 thin films as the gas-sensitive dielectric layer and Pt, Pt/Ta and Al as the electrodes. Physical mechanisms that lead to the detection of CO2 at room temperature are discussed. © 2011 Elsevier B.V. All rights reserved.
Keywords: CO2 sensing Atomic layer deposition La2 O3 thin films
1. Introduction Being a greenhouse gas, the detection of carbon dioxide (CO2 ) has been of enormous research interest in recent years. Unlike redox gases such as CO or NOx , CO2 is a rather chemically inert non-redox gas, which hardly has any gas–solid reactions. This is the major challenge in detecting CO2 using standard resistive techniques based on surface electronic conductivity measurements [1]. Since it is a highly infrared active gas, accurate detection of CO2 is possible using spectroscopic techniques. However, spectroscopic techniques are no viable option for miniaturization and therefore are not suitable for integration with standard CMOS technology. In this regard, films of conducting and semiconducting metal oxides such as SnO2 , WO3 are attractive materials for CO2 detection [2,3], but the operating temperatures are far above room temperature, and are therefore challenging to low-power wireless autonomous transducer solutions. CO2 sensing capability of rare-earth metal oxides has been recognized already in 1993 by Sugai and Matsuzawa [4], where they reported that by incorporating rare-earth metal oxides in alkaline metal carbonates, the response time to the CO2 is reduced and the stability of the sensor increased. Rare-earth metals, especially lanthanum (La), are known dopants with semiconducting
∗ Corresponding author. Present address: Energy Research Institute @ NTU (ERI@N), Nanyang Technological University, Research Techno Plaza, Level 5, 50 Nanyang Drive 639798, Singapore. Tel.: +65 83109488. E-mail address:
[email protected] (K.B. Jinesh). 0925-4005/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2011.04.033
oxides such as SnO2 for CO2 sensing [5]. Novel rare-earth metal based compounds such as lanthanum oxychloride (LaOCl) also has been developed for CO2 sensing [6], but the sensing temperature is 300 ◦ C. Recently, chemoresistive measurements with films of rare-earth metal-oxides, where the current through the films varies upon exposure to CO2 [7], have been reported. The disadvantage of measuring electrical current in thick insulating films is that the resistances are too large (∼1 G) for practical applications. Although it is a non-redox gas, CO2 interacts with the basic sites of metal-oxide surfaces. Therefore, unlike current measurements, capacitance measurements are more appropriate for detecting non-redox gases interacting with a semiconducting or a dielectric medium. Capacitance–voltage (C–V) measurements of the metal–insulator–semiconductor (MIS) devices, when exposed to hydrogen and hydrocarbon environments, are known in literature [8] suggesting that work-function changes in the electrode material and the corresponding lowering of the Schottky barrier heights can be exploited to detect gases at room temperature. However, the detection of non-redox gases such as CO2 at room temperature is still far from reality, since these physical principles are not valid without solid–gas interactions. Recent reports show that CO2 adsorbs onto the basic sites of rare-earth metal oxides and the strength of the adsorption is related to the basicity of the oxide [9]. Since the CO2 adsorption on rareearth metal oxides is related to lanthanide contraction [9], La2 O3 would be an ideal material for CO2 sensing. This article reports for the first time the room-temperature detection of CO2 by means of capacitive sensing techniques using ultra-thin La2 O3 films. A comparative study of CO2 sensing using oxides of other rare-earth
K.B. Jinesh et al. / Sensors and Actuators B 156 (2011) 276–282
metals including Ce, Gd and Er were carried out, from which we observed that La2 O3 exhibits the highest response to CO2 (results not shown here). Metal–insulator–silicon (MIS) capacitor devices comprising atomic-layer-deposited La2 O3 as the dielectric layer are employed for gas detection. La2 O3 was chosen as the CO2 sensing layer because it is known that at elevated temperatures, lanthanum oxide-doped SnO2 films show sensitivity to CO2 [10] and CO [11], in addition to methane [12], formaldehyde [13] and ethanol [14]. Further in this article, the physical mechanism by which La2 O3 acts as a CO2 sensing layer besides its catalytic activities is discussed. 2. Experimental 2.1. Synthesis of La2 O3 thin films La2 O3 thin films (15 nm) were deposited by means of atomic layer deposition on standard low-Ohmic p-type silicon substrates using an ASM-Pulsar3000TM system. The La-precursor used was tris(2,2,6,6-tetramethyl-3,5-heptanedionato) lanthanum or (La(thd)3 ) and O3 was the oxidizing agent. The deposition details have been reported earlier [15]. 2.2. Device fabrication Different electrode materials such as Pt, Pt with Ta adhesion layer, and Al were sputter deposited on the as-deposited films and patterned using standard lithography to form several electrode configurations such as squares and inter-digitated structures. Reactive ion etching (RIE) was employed to selectively etch the electrodes, while the La2 O3 film was kept unetched over the Si surface. Mass-spectroscopy monitoring in RIE ensures that the etching of the metal electrodes is stopped at the La2 O3 surface. 2.3. Electrical characterization The measurements were performed at 10 kHz (30 mV) in a 3 L custom-built probe-station (MDC) to enable gas-exposure in a controlled environment. Subsequently, the C–V measurements were repeated in an environment with a controlled amount of CO2 mixed with N2 (calibrated using mass-spectroscopy). Mass-spectrometry measurements show that the saturation time of the gas in the system is approximately 10 min. The electrical measurements were carried out using an Agilent B1500A for current and capacitance measurements in function of voltage and CO2 exposure. Fig. 1(a) and (b) shows the schematic diagram of the metal/La2 O3 /Si device and the scanning electron microscope (SEM) image of the typical electrode structure employed to measure the capacitance or current of the devices. Fig. 1(c) illustrates the schematic diagram of the measurement setup, where the analyte gases are connected to the measurement setup through mass-flow-controllers (MFC’s). The saturation time of the gases in the chamber was separately measured using mass-spectrometer, and is nearly 10 min, due to the large chamber volume (3 L). The probe-station is connected to the gas-tight chamber, while the gas flows through the chamber. To ensure a complete mixing of the analyte gas (here CO2 ) with the carrier gas (N2 or air), the dimensions of the tubes and the flow rate of the gas are optimized. 3. Results and discussion Prior to the electrical measurements, the metal/La2 O3 /Si devices were kept in N2 ambient and the capacitance–voltage (C–V) measurements were performed at room temperature in order to characterize the electrical behavior of the devices. For initial characterization of the devices, capacitors with square electrodes of
277
100 × 100 m2 area were used. Fig. 2 shows an example of such C–V curves obtained for these devices for forward and reverse voltage sweeps, when exposed to a very high concentration of CO2 (2.5%) compared to the normal levels in the environment (∼300 ppm). Either by extrapolating the depletion capacitance to the voltage axis, or more correctly by plotting Mott plots (1/C2 versus V) and extrapolating 1/C2 to the voltage axis, the flat-band voltage (VFB ) can be estimated from the C–V measurements [16]. Prior to CO2 exposure, the flat-band voltage (VFB ) of the device in N2 is nearly −0.3 V and has very small hysteresis. The saturation capacitance is approximately 68 pF for 15 nm La2 O3 films used in this device. Fig. 2(a) and (b) shows that on exposure to CO2 , the C–V behavior of the device changes significantly. The C–V curves in CO2 were measured keeping the device in saturated gas atmosphere for 20 min. It should be noted that the time constant of the current system is determined mainly by the large volume of the gas-exposure setup (3 L) and thus the observed response time is not related to the La2 O3 –CO2 interaction. When switching back to the pure carrier gas (N2 ) atmosphere, a complete recovery of the device was observed, with comparable time response of the detection. The recovery is faster when the sample was heated to 70 ◦ C. Three major changes can be observed in C–V curves upon CO2 exposure: firstly, as Fig. 2(c) shows, the saturation capacitance was reduced by approximately 12%. Secondly, as shown in Fig. 2(d), the hysteresis of the device increased considerably and thirdly, the flat-band voltage shifts approximately by 1 V as shown in Fig. 2(d). The C–V behavior of a MIS device provides information about various types of oxide charges in the device. These oxide charges are mainly of four types: fixed oxide charges, oxide trap charges, interfacial oxide charges and mobile ions [16]. The band-bending in silicon is influenced by the presence of these oxide charges and the work-function difference between the metal electrode and the silicon substrate. Therefore, the flat-band voltage (VFB ) of the device can be expressed as [16] VFB = MS − (Qf + Qit )
1 d − ε0 k ε0 k
d
[m (x) + ot (x)]xdx,
(1)
0
where k is the dielectric constant of the insulating medium, MS is the difference of work-function of the electrode and silicon substrate, and (Qf + Qit ) is the total charge (fixed oxide charge and interface trap charge) that comes mainly from the imperfections of film deposition on silicon. The terms inside the integral over the film thickness d represent the contribution of the mobile charge density (m (x)) and oxide trap charges (ot (x)) to the flat-band shift. The hysteresis in the C–V curves usually originates from the mobile charges present in the dielectric film [16]. The change in flat-band voltage of the device upon exposure to CO2 is an indication of the change of net charges in the La2 O3 film introduced by gas molecules. According to the conventional interpretation of C–V data [16], the hysteresis is a measure of the mobile charges in the film. However, any dissipative and irreversible process induced by the electric field (estimated as the ratio between the applied voltage across the film and the film thickness) can generate a hysteresis in C–V measurements. The large hysteresis in the forward and reverse bias sweep in the presence of CO2 points out such an electric field-induced dissipative process in the device. The changes in VFB and hysteresis together imply three possible scenarios of explaining the origin of the extra charges in the film: (1) CO2 molecules that diffused into the dielectric layer through the periphery of the electrodes to the active capacitor region form dipoles at the metal–La2 O3 interface under the strong electric field (1–3 MV/cm) that results in modification of the net electric field in the La2 O3 film; (2) gas molecules react with La2 O3 forming La2 (CO3 )3 or La2 O2 CO3 in the film (as proposed in Ref. [7]); (3) a combination of both these mechanisms.
278
K.B. Jinesh et al. / Sensors and Actuators B 156 (2011) 276–282
Fig. 1. (a) Structure of the metal/La2 O3 /Si device used for CO2 detection; (b) top-view of the lithographically patterned inter-digitated metal electrode with Pt, Al and Pt/Ta; (c) the schematics of the gas setup used to measure the electrical properties of the devices in controlled gas atmosphere. The gas flow is controlled accurately using mass-flow controllers (MFC’s) and the saturation time of the gas in the chamber was measured separately using mass-spectrometer.
The dipole formation of gas molecules at the metal–dielectric interface is well-known for decades and Poteat et al. have noticed that the interface state density distribution in the silicon band-gap of the MIS devices does not change with gas exposure [8]. Thus, the flat-band shifts in most of the MIS devices exposed to various gases are assigned solely to changes in the work-function of the electrode material. However, rare-earth metal oxides are known to have several basic sites, and are vulnerable to carbonate formation [8]. The Temperature-Programmed-Desorption (TPD) analysis reported by Sato et al. [9] clearly shows that the strength of the basic sites increases with decreasing atomic number in the lanthanide series, which is explicitly related to the well-known lanthanide contraction. The same authors report that La2 O3 has the largest (structural) carbonate formation among other rare-earth metal oxides. Within the three aforementioned scenarios responsible for CO2 sensitivity, the influence of the interface polarization effects can be examined by studying the CO2 response of metal/La2 O3 /Si devices with different metal electrodes. To examine the influence of the electrode material on the response of these MIS devices, three different electrodes such as Pt, Pt/Ta and Al have been deposited on different samples of 15 nm La2 O3 deposited in the same batch. For this, either 100 nm Pt or 100 nm of Pt with 10 nm of Ta adhesion layer or 200 nm Al was sputter deposited on the La2 O3 surface and patterned with standard lithography and subsequent etching. The C–V curves of these devices were measured before and exposing to controlled CO2 concentrations. Fig. 3(a), (c) and (d) shows the flat-band voltage shifts (VFB ) of the devices when exposed to CO2 mixed with N2 for different electrode materials namely, Pt, Al and Pt/Ta. Fig. 3(b) is the response of Pt/La2 O3 /Si device when exposed to CO2 mixed with dry air. The responses of the devices to CO2 depend strongly on the electrode material. While the devices with Pt electrode exhibit a reduction in VFB , the devices with Al and Pt/Ta exhibit an increase in VFB . If the (oxy)carbonate formation in La2 O3 is the only reason for the change
in VFB , then VFB should change in the same direction regardless of the electrode material on top. While the possibility of the flat-band shift due to the work-function change of the Pt, Al or Pt/Ta electrodes can be completely neglected, the electrode-dependence of VFB in Fig. 3 shows the prominent role of the metal–La2 O3 interface in determining the flat-band shift of the devices in addition to the carbonate formation in the oxide film itself. In a similar device configuration, where high-permittivity dielectrics such as BaTiO3 and SrTiO3 mixed with CuO is exposed to CO2 , Ishihara et al. reports the changes in the capacitance values with CO2 exposure and the direction of which depends on the devices used [17]. They also have arrived at a similar conclusion that the direction of the capacitance change is related to the semiconducting properties of the mixed oxides they use. The response to CO2 in dry air is comparable to that in N2 , though the different behavior of the response curves could be due to a cross-sensitivity to other gases such as oxygen in normal air. However, response to these gases is not a dominating factor in these devices. Cross-sensitivity to relative humidity is significant, but can be distinguished from other gases, since the charge transfer mechanisms of CO2 and water to rare-earth metal-oxides are expected to be different [7]. The influence of other gases on the CO2 sensitivity of the devices is currently under investigation. To understand the role of the electrode materials in CO2 sensing, the density of states (Dit measured in cm−2 eV−1 ) in the silicon band-gap were extracted from the parallel conductance–voltage measurements as shown in Fig. 4. Density of states in metal/dielectric/silicon system refers to the density of states in the band-gap of silicon, approximately 3KB T/q (KB is the Boltzmann constant, T is the temperature and q is the electronic charge) from both edges of the midgap [16]. Variation in admittance of the device when exposed to a gas shows the recombination and generation of the mobile carriers, which reflects directly in the Dit distribution in silicon [16]. Being an inert metal, Pt does not react with the La2 O3 surface. On the contrary, Al and Ta have
K.B. Jinesh et al. / Sensors and Actuators B 156 (2011) 276–282
279
Fig. 2. (a) C–V curves of the Al/La2 O3 /Si devices in N2 and after exposing to 2.5% CO2 mixed with N2 . The arrows indicate the direction of the voltage sweep. (b) Parallel resistance of the device measured simultaneously with capacitance in N2 and CO2 mixed with N2 . (c) Normalized capacitance of the Al/La2 O3 /Si devices; (d) Hysteresis of the C–V curves and (e) flat-band voltage shift (VFB ) of the devices as a function of the CO2 exposure time. The capacitance, hysteresis and VFB values are saturating nearly after 16 min gas exposure. Note that the response time is dominated by the saturation time of the 3-L gas-exposure system, which is approximately 10 min.
Fig. 3. Flat-band shift as a function of CO2 concentration of (a) Pt/La2 O3 /Si device after exposure to CO2 in N2 , the dotted grey line is a liner fit to the data before it saturates; (b) Pt/La2 O3 /Si device when CO2 is mixed in dry air; (c) Al/La2 O3 /Si device in CO2 mixed with N2 and (d) Pt/Ta/La2 O3 /Si device when CO2 mixed with N2 .
280
K.B. Jinesh et al. / Sensors and Actuators B 156 (2011) 276–282
Fig. 4. Interface density of states in silicon band-gap of the Pt/La2 O3 /Si and Al/La2 O3 /Si devices as a function of CO2 concentration (in N2 ).
a tendency to form oxides or alloys at the interface with La2 O3 . Therefore, the response of the Pt/La2 O3 interface to CO2 should be different from Al/LaAlOx /La2 O3 or Pt/Ta/LaTaOx /La2 O3 configurations. The formation of additional charges in the oxide will be reflected in the density of states (Dit ) in the silicon band-gap. Fig. 4 shows Dit measured for the devices with different electrodes. The Dit values of the Pt/La2 O3 /Si devices remain constant on exposure to CO2 , while exhibiting a significant flat-band shift. This indicates that the change in VFB is solely related to small changes in the effective electric field in the devices, rather than chemical changes in the oxide. The dipole formation of the CO2 molecules at the Pt/La2 O3 interface would be responsible for the reduction of the net electric field in the oxide film. In the case of Al/La2 O3
or Ta/La2 O3 , a clear interface is absent due to the oxidation or alloy formation of Al and Ta with La2 O3 , where the inter-diffusion and subsequent oxidation of Al into La2 O3 generally occurs even at room temperature [15]. Due to this, a well-defined interface between Al and La2 O3 is absent in the Al/La2 O3 /Si system. The change in Dit of the Al/La2 O3 /Si devices indicates the changes in the chemical environments in the film more than interface polarization or dipole formation of the gas molecules. The possible explanation for this is that CO2 molecules interact with the oxygen vacancies created when Al and Ta form their oxides or mixed alloys with La2 O3 . But the percentage of the (oxy)carbonate formation in the film is much less to induce any observable change in the dielectric permittivity of La2 O3 . Therefore, the reduction of the saturation capacitance should be resulting from the influence of the molecules on the internal electric field of the devices rather than change in the dielectric permittivity, and the hysteresis arises from changes in the molecules in the strong electric field existing in the device, which are energy dissipative processes. From the slope of the linear regime of the data in Fig. 3, we find that Pt/La2 O3 /Si devices have a sensitivity of 71.5 V/ppm in terms of flat-band shift, while Al/La2 O3 /Si devices have a sensitivity of 293 V/ppm and Pt/Ta/La2 O3 /Si devices have 88.4 V/ppm. Though Al/La2 O3 /Si devices appear to have a higher sensitivity, the devices are saturated faster than the Pt/La2 O3 /Si devices. This quick saturation also suggests that chemical change in the film is responsible for the sensitivity in Al/La2 O3 /Si devices, where the interfacial polarization is dominant in Pt/La2 O3 /Si devices. If the sensing mechanism involves chemical interaction of the gas molecules with the defects (base-sites) in the oxide, an apparent change would be visible in the leakage current through the devices. Therefore, measuring the leakage current through the device is an alternate option for gas detection. The leakage cur-
Fig. 5. (a) I–V behavior of the Pt/La2 O3 /Si in N2 and 1000 ppm CO2 mixed with N2 ; (b) Child’s regime, where the current is proportional to V2 ; (c) I–V behavior of the Al/La2 O3 /Si in N2 and 1000 ppm CO2 mixed with N2 ; (d) current measured at 1 V as a function of time (averaged over five measurements), when the device is exposed to different concentrations of CO2 mixed with N2 . The arrows indicate the time when the gas exposure starts. The sensor recovers at room temperature when it is exposed only to N2 , and it comes back to the base-line (current before the gas exposure) at around 7000 s.
K.B. Jinesh et al. / Sensors and Actuators B 156 (2011) 276–282
rent through a dielectric film follows different laws according to the electric field across the film. These conduction schemes can be summarized into four regimes: (1) Ohms regime, where at low electric fields (typically
