Discriminating Gas Concentrations in Extreme Temperature Environments

B. J. D. Furnival, N. G. Wright, A. B. Horsfall School of Electrical, Electronic and Computer Engineering

Newcastle University Newcastle upon Tyne, England

E-Mail: B.J.D.Furnival@ncl.ac.uk

Abstract — In this paper a Pt/HfO2/SiO2/SiC MIS capacitor is reported, which unlike previously documented Pt/SiO2/SiC and Pd/TiO2/SiO2/SiC devices shows no response to O2 ambients. As a result, if the capacitor is used within a sensor array it could significantly improve selectivity between H2 and O2. The use of SiC as the substrate material also enables this device and the potential array to operate in extreme conditions, where standard electronics fail. Examination of the devices sensing mechanisms reveal that in contrast to previous reports, a positively charged dipole layer is not formed during exposure to O2. Whilst the passivation of trapping states at the SiO2/SiC interface have little influence on the response to either H2 or O2.

I. INTRODUCTION The ability to monitor the gaseous constituents of hostile

environments is of interest for a variety of applications over a range of industrial sectors. These include; the storage of nuclear waste and the decommissioning of nuclear facilities, improving the efficiency of combustion engines by advanced monitoring of emissions and the study of extraterrestrial atmospheres during exploration of the interior planets. A large proportion of research into resilient sensing has incorporated the use of wide band-gap semiconductors such as silicon carbide (SiC). As in the case of the 4H polytype, the band-gap is 3.23eV, almost 3 times than that of silicon (Si). The benefit of wide band-gap semiconductors is their superior temperature resilience, as whilst devices fabricated on Si fail at below 175°C, SiC can potentially operate at temperatures in excess of 900°C [1]. Additionally, the strong silicon to carbon bonds in the SiC lattice, provide enhanced chemical inertness and radiation tolerance when compared with Si.

The Metal-Insulator-Semiconductor (MIS) capacitor is a popular structure for gas sensing and has been shown to be sensitive to many gas species, including H2, O2, H2S and CO, at temperatures of greater than 300°C [2,3,4]. The response of a device to a mixed gas ambient is therefore complex and influenced by multiple stimuli, which cannot be individually quantified. Using a sensor array significantly improves

selectivity and statistical techniques enable accurate discrimination between gases [5]. To produce a sensor array, multiple devices with a range of sensitivities are required, which could be achieved by varying the catalyst used for the gate contact [6]. However, materials suitable for this, such as platinum (Pt), palladium (Pd) and iridium (Ir) are expensive and it would be more convenient and offer an extra design parameter, if it were possible to achieve this be means of modifying the dielectric stack.

It has previously been revealed that a Pt/SiO2/SiC device has sensitivity to both H2 and O2, using a detection mechanism that combines the formation of a dipole layer below the gate contact and passivation of interface states [7]. A different study indicates that incorporating a high-κ dielectric layer produces similar results, as a Pd/TiO2/SiO2/SiC device gives a substantial response to both H2 and O2 [2]. However, in this paper a Pt/HfO2/SiO2/SiC structure is demonstrated that shows no sensitivity to an O2 ambient. The device still gives a large response to H2 and its detection mechanism is found not to incorporate passivation of interface states.

II. EXPERIMENTAL Research grade 4H-SiC with a 2μm Si face epitaxial layer and N-Type doping density of 2.03x1016 cm-3 was purchased from CREE. After RCA cleaning, the sample was oxidised at 1150°C, in 80 SCCM of dry O2 for 200 minutes, producing a 20nm thick SiO2 film on the Si-face. A stack of 5nm Ti/100nm Ni was deposited on the C-Face by DC sputtering and vacuum annealed for 200 seconds at 1050°C, creating a Ni2Si based Ohmic contact. Previous results have shown that after this contact has been formed, it can be removed and the material will still retain its Ohmic properties [8]. Therefore, using a combination of wet chemical etching and low power oxygen glow discharge the contact was removed, before 60nm of hafnium was DC sputtered onto the SiO2. Rapid Thermal Processing (RTP) of the hafnium for 3 minutes at 550°C was used to form HfO2. Contacts comprising of 5nm Ti/75nm Pt were then DC sputtered onto the HfO2 and to enable wire

978-1-4244-9289-3/11/$26.00 ©2011 IEEE

Figure 1. The structure of a Pt/HfO2/SiO2/SiC

MIS capacitor

bonding, annular rings of 5nm Cr/185nm Au were deposited on top of these by E-Beam. The sensors were completed with a back contact of 5nm Ti/60nm Ni/5nm Cr/60nm Au that was deposited onto the Ohmic contact region [8]. A schematic cross section of the completed sensor is given in Fig. 1.

After fabrication, devices with a contact area of 4.91x10-4 cm2 were mounted into a ceramic DIL chip and placed in a gas test chamber. Initially they were annealed in N2 at 400°C for several hours to stabilise their characteristics. They were then heated to 300°C and the capacitance - voltage (C-V) and conductance - voltage (G-V) characteristics were measured in different ambients, using a HP4284 LCR bridge. The chamber was filled with a combination of pure N2 and 2.5% H2 in N2 or 2.0% O2 in N2, with the flow rate of each being managed by a mass flow controller (MFC). During this investigation, the chamber pressure was maintained at 11 Torr, whilst the input rate of H2 was varied between 0ppm and 2300ppm and O2 between 0ppm to 1800ppm.

III. RESULTS The C-V and G-V results presented have been corrected for the effect of series resistance, using the method described by

Figure 2. The C-V response of a Pt/HfO2/SiO2/SiC MIS capacitor to

various concentrations of H2

Figure 3. The G-V response of a Pt/HfO2/SiO2/SiC MIS capacitor to

various concentrations of H2

Nicollian and Brews [9]. The data in Fig. 2 shows the C-V response of a Pt/HfO2/SiO2/SiC MIS capacitor exposed to differing concentrations of H2 at 300°C. As the concentration of H2 increases, the flatband voltage (VFB) reduces from 1.45V to 1.05V and the oxide capacitance (COX) increases from 69.8pF to 72.9pF. It has previously been reported that a reduction in VFB is the result of H2 absorption near the metal/dielectric interface, which causes the formation of a negative dipole layer [6]. This dipole layer reduces the metal - semiconductor barrier height and so a smaller bias is required to achieve the flatband condition. However, a growth in the value of COX also indicates that oxide thickness is reducing as the H2 concentration increases. This is a result of the dipole layer forming at the dielectric side of the metal/dielectric interface, which effectively results in a thinner oxide layer.

Fig. 3 shows the conductance/angular frequency - voltage (G/ω-V) characteristics, recorded for the Pt/HfO2/SiO2/SiC MIS capacitor during exposure to different concentrations of H2. From the data it can be observed that for greater concentrations of H2, the peak of the G/ω plot moves towards lower bias voltages and the amplitude increases. The change in position of the G/ω peak towards lower bias is similar to that observed for the variation of VFB. This is because the peaks are centered at midgap, which occurs at the same bias as VFB. The amplitude increase of the G/ω peaks can be described using the C-V data, as G/ω has previously been observed to have a strong link with COX [10]. Therefore, as the value of COX increases at higher concentrations of H2, so does the amplitude of the G/ω peak. The similarities found between the information available from the C-V and G/ω-V data means that either data set can be used to interrogate the sensor in a H2 rich ambient.

The data in Fig. 4 shows the C-V response of the sensor after exposure to different concentrations of O2 in N2. As the quantity of O2 in the chamber is varied, no change to either VFB or COX is observed. This indicates that unlike the previously reported data on Pt/SiO2/SiC and Pd/TiO2/SiO2/SiC devices [2,6], the Pt/HfO2/SiO2/SiC structure is not sensitive

185nm Gold / 5nm Chrome

75nm Platinum / 5nm Titanium

60nm Hafnium Dioxide

20nm Silicon Dioxide

4H Silicon Carbide

Ohmic Contact

Figure 4. The C-V response of a Pt/HfO2/SiO2/SiC MIS capacitor to various concentrations of O2

to O2. Furthermore, the data in Fig. 5 shows the G/ω-V characteristics recorded for each of the O2 concentrations. As the relationship given earlier between C-V and G-V characteristics predicts, the G/ω peak remains centered at the flat band voltage for all data sets and the observed change in peak amplitude is within measurement error. This again indicates that the Pt/HfO2/SiO2/SiC structure is not sensitive to O2. A device with properties such as this would be ideal for use within a sensor array, as paired with an O2 sensitive device it could become possible to uniquely identify the concentration of oxygen in a mixture.

In order to further examine the H2 and O2 sensitivity of the Pt/HfO2/SiO2/SiC MIS capacitor, a plot of sensitivity vs. bias voltage – flatband voltage is shown in Fig. 6. The sensitivity is calculated using equation 1, where C(Baseline) is the capacitance recorded during exposure to pure N2 and C(Measured) is the capacitance recorded within the desired ambient.

= ∗ ∗ 100%

Figure 5. The G-V response of a Pt/HfO2/SiO2/SiC MIS capacitor to

various concentrations of O2

Figure 6. The sensitivity of a Pt/HfO2/SiO2/SiC MIS capacitor to various

concentrations of H2 and O2

The data in Fig. 6 shows that for each ambient the peak sensitivity occurs at the flatband voltage and the amplitude of this peak increases with growing H2 concentration. So in order to optimise the device for H2 detection, it should be biased at a point as close to the midgap as possible. At a H2 concentration of 600ppm the peak sensitivity is around 20% and for the maximum H2 concentration of 2300ppm this grows to around 40%. As expected, the other H2 concentrations form peaks that are positioned between these and their spacing demonstrates the devices excellent resolution. However, Fig. 6 also reveals that sensitivity to O2 does not peak at VFB and instead remains constantly near zero, further confirming the Pt/HfO2/SiO2/SiC structure is not sensitive to O2.

As discussed earlier the most popular model for describing gas sensitivity in MIS capacitors includes two mechanisms; 1) the formation of a charged dipole layer at the metal/dielectric interface 2) the modification of trapping states at the SiO2/SiC interface [6]. Using equation 2 the first of these mechanisms can be considered.

= − + +

Where φms is the metal-semiconductor work function, Qf the fixed oxide charge, Qm the mobile charge and Qot the oxide trapped charge. Assuming that the charge in the oxide stays constant during gas exposure, equation 2 allows the shift in φms with H2 and O2 concentration to be extracted. The data in Fig. 7 shows that as the concentration of H2 increases and the charged dipole layer grows, the value of φms reduces. As φms is the difference between the work function of the metal (φm) and the semiconductor (φs) and it can be assumed that φs stays constant as the gas concentration varies. The change in φms shown in Fig. 7 must be caused by a reduction of φm, which is similar to previous reports that changes in the metals barrier height is a major part of the gas sensing mechanism [2,6]. However, during O2 exposure, φms only shows a small change, which recovers it back to its original value before H2

(2) (1)

Figure 7. The metal-semiconductor work function (φms) of a

Pt/HfO2/SiO2/SiC MIS capacitor in various concentrations of H2 and O2

exposure was performed. As no increase of φms is observed, this suggests that in contrast with the reports for Pt/SiO2/SiC and Pd/TiO2/SiO2/SiC devices a positive dipole is not formed in the Pt/HfO2/SiO2/SiC structure during O2 exposure [2,6].

The second of the sensing mechanisms proposed is the passivation of trapping states at the SiO2/SiC interface. In order to examine this for the Pt/HfO2/SiO2/Si device, the Hill-Coleman technique was used to extract the density of interface states (DIT) for each concentration of H2 and O2. In contrast to previous reports, the data shows that in all cases DIT was approximately 4.5x1011eV-1cm-2. This demonstrates that at midgap, exposure of the sensor to either H2 or O2 has little effect on DIT. However, by studying the devices C-V characteristics for stretch out, the DIT over a wider range of the bandgap can be considered. The data shown in Fig. 2 and Fig. 4 reveals that during both H2 and O2 exposure, little change is observed in the shape of the C-V response and therefore it seems interface states have no effect on the detection mechanism of the Pt/HfO2/SiO2/SiC structure.

IV. CONCLUSIONS In this paper we have shown for the first time that, unlike

previously reported Pt/SiO2/SiC and Pd/TiO2/SiO2/SiC MIS capacitors, which respond to both H2 and O2, a Pt/HfO2/SiO2/SiC structure is only sensitive to H2. This suggests that if the structure were used as a component of a sensor array, selectivity between H2 and O2 could be improved considerably. We have demonstrated that both the C-V and G-

V characteristics can be used to monitor the concentration of H2. Extraction of the H2 sensitivity for all of the available bias voltages reveals a peak occurs at midgap for any H2 concentration, therefore optimization of a device for H2 sensitivity requires biasing at flatband voltage. Finally, the sensing mechanisms of the structure were examined. Through extraction of φms by assuming oxide charge remains constant, it has been shown that φm reduces with increasing H2 concentration. This supports previous reports that formation of a charged dipole layer at the metal/dielectric interface enables gas sensitivity. However, by studying the shape of C-V characteristics, no stretch-out has been found and extracted values of DIT at midgap stay constant with varying gas concentration. This is a contrast with the second previously reported mechanism of trapping state passivation at the SiO2/SiC interface.

ACKNOWLEDGMENT This work was supported by EPSERC (platform grant

EP/D068827/1), ONE North East and a CASE studentship from the sensors KTN and John Caunt Scientific.

REFERENCES [1] A. Lloyd Spetz, A. Baranzahi, P. Tobias, and I. Lundstr�m, Phys.

Stat. Sol. A 162, 493 (1997). [2] M. H. Weng, R. Mahapatra, A. B. Horsfall, and N. G. Wright, IEEE

Sens. J. 7 (10), 1395 (2007). [3] M. H. Weng, R. Mahapatra, N. G. Wright, and A. B. Horsfall,

Meas. Sci. Technol. 19, 024002 (2008). [4] S. Nakagomi, P. Tobias, A. Baranzahi, I. Lundstr�m, P.

Mårtensson, and A. Lloyd Spetz, Sensor Actuat. B - Chem. 45, 183 (1997).

[5] P. Zhang, C. Lee, H. Verweij, S. A. Akbar, G. Hunter, and P. K. Dutta, Sensor Actuat. B - Chem. 123, 950 (2007).

[6] M. L�fdahl, M. Eriksson, M. Johansson, and I. Lundstr�m, J. Appl. Phys. 91 (7), 4275 (2002).

[7] P. Tobias, B. Golding, and R. N. Ghosh, IEEE Sens. J. 2 (5), 543 (2003).

[8] B. J. D. Furnival, K. Vassilevski, N. G. Wright, and A. B. Horsfall, Mater. Sci. Forum 679-680, 469 (2011).

[9] E. H. Nicollain and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and Technology. (Wiley Inter-Science, 2003), p.906.

[10] J. Piscator, B. Raeissi, and O. Engstr�m, Appl. Phys. Lett 94, 213507 (2009).