- Email: [email protected]
Ultra-high temperature (>300 °C) suspended thermodiode in SOI CMOS technology
Microelectronics Journal 41 (2010) 540–546
Contents lists available at ScienceDirect
Microelectronics Journal journal homepage: www.elsevier.com/locate/mejo
Ultra-high temperature (4300 1C) suspended thermodiode in SOI CMOS technology S. Santra a,, F. Udrea a, P.K. Guha a,b, S.Z. Ali a, I. Haneef a,c a
Department of Engineering, University of Cambridge, 9JJ Thomson Avenue, Cambridge CB3 0FA, UK School of Engineering, University of Warwick, Coventry CV4 7AL, UK c Institute of Avionics and Aeronautics Air University, E-9, Islamabad 44000, Pakistan b
a r t i c l e in f o
a b s t r a c t
Available online 4 January 2010
This paper reports for the first time on the performance and long-term stability of a silicon on insulator (SOI) thermodiode with tungsten metallization, suspended on a dielectric membrane, at temperatures beyond 300 1C. The thermodiode has been designed and fabricated with minute saturation currents (due to both small size and the use of SOI technology) to allow an ultra-high temperature range and minimal non-linearity. It was found that the thermodiode forward voltage drop versus temperature plot remains linear up to 500 1C, with a non-linearity error of less than 7%. Extensive experimental results on performance of the thermodiode that was fabricated using a Complementary Metal Oxide Semiconductor (CMOS) SOI process are presented. These results are backed up by infrared measurements and a range of 2-D (dimension) and 3-D simulations using ISE and ANSYS software. The on-chip drive electronics for the thermodiode and the micro-heater, as well as the sensor transducing circuit were placed adjacent to the membrane. We demonstrate that the thermodiode is considerably more reliable in long-term direct current operation at high temperatures when compared to the more classical resistive temperature detectors (RTDs) using CMOS metallization layers (tungsten or aluminum). We also compare a membrane thermodiode with a reference thermodiode placed on the silicon substrate and assess their relative performance at elevated temperatures. The experimental results from this comparison confirm that the thermodiode suffers minimal piezo-junction/ piezo-resistive effects. & 2009 Elsevier Ltd. All rights reserved.
Keywords: SOI CMOS Thermodiode RTDs High temperature sensors Smart sensors
1. Introduction Temperature sensors are one of the fastest growing segments in sensors’ market. An integrated temperature sensor for thermal management is a core component in power hungry circuits that tend to operate close to the maximum junction temperature. In such systems, accurate monitoring of the junction temperature is mandatory to optimize the integrated circuits (ICs) performance while maintaining high reliability. Most of the Complementary Metal Oxide Semiconductor (CMOS) processes now target higher junction temperatures to allow increased packing density of transistors, better cost-performance value and more powerful processing. Maximum junction temperatures in bulk CMOS ICs have moved from a conservative level of 125–150 1C [1–4] and even to 175 1C. By using copper or tungsten, which are more resistant to electro-migration, some of these processes can potentially move to 200 1C, provided that they address issues like latch-up and low cross-talk, and overcome reliability problems such as negative bias temperature instability (NBTI), time
Corresponding author. Tel.: + 44 1223 748311; fax: + 44 1223 748348.
E-mail address: [email protected] (S. Santra). 0026-2692/$ - see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2009.12.005
dependent dielectric breakdown (TDDB), etc. Furthermore, the additional use of silicon on insulator (SOI) technology in ICs not only suppresses the latch-up but also minimizes the leakage currents. It also provides an excellent vertical and lateral isolation and thereby allows a further increase in the maximum junction temperatures to 2251 or potentially even 250 1C. Such ICs can be of use in automotive electronics, power supplies, motor control or other power systems. The silicon thermodiodes (i.e. silicon diode temperature sensors) have been reported in the past for diverse applications. These applications include, for example, temperature measurement for cryogenic applications [5–11], flow sensing [12–15], liquid–vapour interface and liquid level point sensing in hydrogen [16], humidity sensing [17–19], pH sensing [20], vacuum sensing [21], PC temperature monitoring [22], thermometry [17,23–25], thermal characterization of thermally conductive underfill for flip-chip packaging [26], pure gas and gas mixtures’ thermal conductivity monitoring [27], temperature-compensation of piezo-resistive stress sensors [28], MEMS bolometers [29], infrared (IR) detectors [30] and IR focal plane arrays [31], etc. Although most of these applications require maximum junction temperature below 200 1C and more often below 150 1C, yet there are some emerging silicon-based sensors for which accurate
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
and reliable temperature monitoring is essential at very high temperatures (i.e. up to 500 1C), well beyond the junction temperatures of standard ICs. The examples of such applications are smart micro-calorimeters [32], resistive gas sensors [33,34] and sensors used in automotive engines, exhausts, etc. Such smart sensors use membrane technologies for thermal isolation of the sensors (e.g. micro-calorimeters) that typically operate at 400 or 500 1C. As a result, while the active sensing element (e.g. gas sensitive layer) suspended on a very thin dielectric membrane would operate at high temperatures for optimal sensing, the onchip electronics can still operate very close to the ambient temperature. For such sensors, accurate monitoring of the temperature at the hot-spot of the membrane is absolutely essential to enhance the sensor sensitivity and selectivity (as it is the case in gas sensors) and last but not least, for reliability assessment. To meet the challenge of temperature monitoring in these sensors, there is a need for IC/CMOS temperature sensors that can provide accurate temperature measurement, are small in size, high in sensitivity, and reliable in performance beyond 300 1C. To the authors’ knowledge, there have been very few reports [5,35–37] on IC temperature sensors that can operate at such high temperatures. In this paper we will report on the use of a membrane thermodiode operating at temperatures well beyond 300 1C [38]. Linearity is preserved up to 500 1C and the maximum temperature, beyond which the saturation current becomes comparable with the drive current of the thermodiode, is around 600 1C. We demonstrate for the first time that at very high temperatures, the membrane thermodiode offers better reliability than equivalent metal resistive temperature detectors (RTDs) using CMOS metals such as aluminum or tungsten while maintaining very high linearity. Furthermore, we have seen no evidence of piezojunction/piezo-resistive effect in the suspended thermodiode, which would otherwise limit the operation of silicon-based resistive sensors at very high temperatures.
Tungsten micro-heater
541
Gas sensing material NMOS
Membrane thermodiode Membrane area
Reference thermodiode
PMOS
N+ N+ P+ P+ Nwell Pwell
Electronics area
Fig. 1. Cross-sectional view of membrane thermodiode, a reference thermodiode and the CMOS electronic cells.
N+
Contacts to Metal 1
P
P+
Thermodiode and micro-heater
2. Thermodiode design, fabrication and on-chip circuitry In chemical sensor, the micro-hotplate (MHP) is a region of the sensor placed on a thin dielectric membrane, which contains a micro-heater. The temperature in this region is quasi-uniform and considerably higher than that outside the membrane. We designed our MHP to contain a tungsten micro-heater and placed the drive and signal processing electronics outside the membrane. For accurate monitoring of the MHP temperature we have placed a thermodiode, which we refer to as the membrane thermodiode, right in the centre of the membrane under the MHP. An additional thermodiode is placed outside the membrane to monitor the junction temperature of the IC chip (which in our case, where the IC power consumption is negligible when compared to that of the sensor, is often close to the ambient temperature). A schematic cross-section of the membrane thermodiode, a reference thermodiode and the CMOS electronic cells are shown in Fig. 1. The MHP, SOI thermodiode was designed in Cadence 5.0 software. The Cadence layout of the thermodiode is shown in Fig. 2(a). The diameter of the diode was 34 mm. Note that the shapes of both the membrane and the MHP are chosen to be circular to minimize the mechanical stress at the membrane edge. The thermodiodes integrated micro-hotplates were fabricated using a commercial 6 inch 1.0 mm SOI CMOS process. This process features a 0.25 mm active silicon layer, a 1.0 mm buried oxide layer and triple high temperature metallization based on tungsten. The use of tungsten metallization allows operation of the MHP at high temperatures (up to 700 1C). The tungsten layer was used for the resistive micro-heater (metal 1 layer), contacts of the two
SOI dielectric membrane
50 μm
Fig. 2. (a). Cadence layout of the SOI p+ /p/n+ thermodiode with diameter of 34 mm. (b). The optical micrograph of a fabricated micro-hotplate with SOI thermodiode temperature sensor embedded under the hotplate, within the oxide membrane.
terminal diodes (metal 1, 2 layer) and interdigitated electrodes (metal 3 layers). These interdigitated electrodes are used specifically for gas sensing material’s resistance measurement. After CMOS fabrication, the wafers underwent a single deep reactive ion etching (DRIE) back-etch step at a separate micro electrical mechanical systems (MEMS) foundry to form the thin oxide–nitride membrane (ca. 5 mm) in specific areas. The details of the micro-heater design and characterization have been reported elsewhere [39]. An optical microscope picture of the fabricated MHP (with the membrane thermodiode and micro-heater) is shown in Fig. 2(b). The diameter of the membrane on which micro-heater and thermodiode are embedded at its centre, is 300 mm. As already mentioned, a reference thermodiode was placed on the silicon substrate to monitor the ambient temperature and also assess the piezo-junction effect. The structure of the reference thermodiode is identical to that of the membrane
542
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
25
20
I
Power (mW)
I
+
15
10
IA
5 – Reference thermodiode
Membrane thermodiode
0 0
200
Fig. 3. Instrumentation amplifier (IA) circuit for temperature measurement.
thermodiode. To eliminate the effect of the ambient temperature, a simple differential circuit, as shown in Fig. 3, was implemented on-chip. The current source and an instrumentation amplifier (IA), whose gain was controlled externally, were integrated on-chip (outside the membrane area) for good matching characteristics.
600
400 Temperature (°C)
Fig. 4. Power versus temperature plot of the micro-heater which was taken at different positions of the wafer.
3. Results and discussion
Tungsten micro-heater
3.1. Micro-hotplate characterization 3.1.1. Power versus temperature and transient response The tungsten resistive micro-heater was calibrated to verify its temperature coefficient of resistance (TCR) using a high precision (Signatones model S-1060R QuieTemp DC) hot chuck system. The hot chuck system has 1 1C resolution and can operate at temperatures up to 300 1C. A Hewlett Packard 4142B modular DC source/monitor and a Signatones model S-1160 probe station was used for electrical measurements (I–V characterization). The MHP exhibits an ultra-low power consumption (16 mW for 600 1C), very short transient times (2 ms to 600 1C), and extremely good reproducibility within a wafer, and from wafer to wafer. A power versus temperature plot (measured at different positions of the wafer) calculated for a tungsten micro-heater using its I–V data and TCR is shown in Fig. 4. The fast transient response of the device enables heating up and cooling down of the MHP very quickly. 3.1.2. Temperature rise/distribution: ANSYS simulations and infrared imaging Fig. 5 shows the temperature distribution over the MHP simulated using ANSYS. Infrared thermal imaging of the membrane was also carried out using a Quantum Focus Instruments Infrascope II IR imaging scope to verify the temperature distribution across the surface of the membrane. The IR image of a micro-hotplate at 250 1C is shown in Fig. 6 (the 250 1C limit was imposed by the detection range of the IR system). The IR imaging results (shown in Fig. 6) confirm that the temperature distribution is uniform in the heater region (deviation within 5%) while it decreases rapidly beyond that, across the rest of the membrane. This is also confirmed by the ANSYS simulations. We can therefore, assume that the temperature is approximately constant within the MHP and hence we can ignore the temperature gradients inside the membrane thermodiode.
800
Membrane
25
200
350
600
780 °C
Fig. 5. Temperature distribution over the membrane as determined by ANSYS simulations.
3.2. SOI thermodiode performance 3.2.1. V–T plot up to 300 1C and piezo-junction effect A diode can be used as a temperature sensor when it is driven by a constant forward current. In this operation scheme, the forward voltage of the thermodiode decreases linearly with the increase in temperature. The forward voltage versus temperature (V–T) measurements of the membrane thermodiode and the reference thermodiode were performed up to 300 1C using the hot chuck. The V–T plot for both on-membrane and off-membrane thermodiodes is shown in Fig. 7. A comparison of the V–T plot for the two thermodiodes confirms that there is very little difference between their characteristics indicating that there is minimal piezo-junction effect. It also indicates that the residual and thermally induced mechanical stress in the membrane has negligible influence over the forward voltage temperature characteristics of the thermodiode. The slope of the thermodiodes is found to be 1.3 mV/1C at 65 mA forward current. The reason we chose such low current ( mA) for the temperature sensor operation is to avoid any self heating effect.
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
543
1.0 Experimental
Thermodiode and micro-heater
0.8 Forward voltage drop (V)
Tracks
Numerical Analytical
0.6 65 μA
0.4 1 μA
0.2
525
385
14 nA
285
0.0 0 Membrane
0.9 Reference thermodiode Suspended thermodiode
Forward voltage drop (V)
400 Temperature (°C)
600
800
Fig. 8. Experimental, numerical and analytical V–T plot.
Fig. 6. IR image of the microhotplate at 250 1C. Note that the temperature profile is more accurate in the micro-heater area as beyond this the dielectric membrane is transparent to IR.
0.8
0.7
0.6
0.5 0
200
100
200 Temperature (°C)
300
Fig. 7. Forward voltage drop versus temperature (V–T) plot of the suspended and reference thermodiode.
Indeed, the power consumption of the diode, when it is driven with 65 mA driving current, is only 0.12% of that of the MHP (operated at 500 1C) introducing errors of less than 1 1C. 3.2.2. V–T plot up to 780 1C: Experimental versus analytical and numerical results To characterize the thermodiode beyond 300 1C, the onmembrane tungsten micro-heater was used. The forward voltage of the thermodiode at different temperature was measured at three forward current levels (65 mA, 1 mA and 14 nA) from 25 to
780 1C. To obtain such high temperatures, we have used tungsten metallization (instead of the standard aluminum) and thin SOI layers (0.25 mm). This choice minimizes the depletion region volume and thus obtains a very low value for the diode saturation current (Is). From the V–T characteristics of the thermodiode shown in Fig. 8, it can be clearly observed that the voltage versus temperature slope linearity is maintained up to high temperatures (e.g. 500 1C at 65 mA driving current), which is due to this very low Is (Is few fA at room temperature). At high temperatures, the saturation current becomes comparable or even higher than the driving current and the consequence of this is that the V–T characteristics become non-linear. It is important to note that the V–T slope and the maximum temperature beyond which the saturation current becomes comparable with the drive current depend entirely on the driving current. The V–T slope changes from ( 2.2) to ( 1.3) mV/1C due to the change in driving current from 14 nA to 65 mA, respectively. A higher driving current leads to a lower sensitivity but maintains the linearity of the diode up to a higher temperature. Here it is interesting to mention that in earlier reports on high temperature operation of thermodiodes, Boltovets et al. [5] and Shwarts et al. [36] presented silicon diode temperature sensors operating up to 327 1C. Ang [35] showed the electrical characteristics of n+ /n/p+ silicon diodes in the temperature range 27–310 1C. In all the three reports [5,35,36] the thermodiodes were used as a discrete device, and were not integrated with onchip CMOS circuitry. Kimura and Toshima [37] have also reported a thermistor-like p–n junction temperature sensor that can cover very wide temperature range ( 200 to 500 1C). However, they have operated the diode in the forward bias-voltage mode only for the temperature range 200 to 150 1C whereas beyond that, for 150–500 1C, they used the diode in the reverse bias-voltage mode. The temperature measurement of the reverse current is not only difficult given its very small level, but also not reliable as often this current varies very significantly from wafer to wafer and lot to lot. Also its linearity is relatively poor. In our case we have used a CMOS temperature sensor in forward biased mode with an onchip constant current drive. This allows a more accurate measurement, and easier detection and post-processing of the results. To understand the V–T characteristics of the SOI thermodiode fully, extensive finite element analysis (FEM) in ISE TCAD software
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
was also carried out. The thermodiode was also characterized by analytical modeling. These numerical and analytical predictions have also been compared with the experimental results in Fig. 8, and all three are in excellent agreement with each other. For further details of the analytical calculations and an analysis of the thermodiode’s non-linear behaviour at very high temperatures, the readers may refer to [40]. It is worth mentioning that we also picked up thermodiodes randomly from different locations across one wafer and from different wafer lots. These thermodiodes were then evaluated experimentally to verify the repeatability of the results. The results were found to be highly uniform and repeatable. As mentioned earlier, the thermodiodes were driven by three different driving currents (i.e. 14 nA, 1 mA and 65 mA). Corresponding to each driving current, the maximum temperatures beyond which the thermodiodes’ response became non-linear were calculated assuming the maximum error of 10% of the non-linear terms. The results shown in Fig. 9 point out that the non-linearity starts dominating at lower temperatures for the lower driving current. This is because in this case, the saturation current approaches the value of driving current at a lower temperature level. 3.3. Long term performance stability of thermodiode To evaluate its long-term stability, the thermodiode was operated at 400 and 500 1C for 100 h using the tungsten microheater on the membrane. The tungsten micro-heater (i.e. metal RTD) was operated using a constant current while the thermodiode was driven at a forward current of 65 mA. As the heating occurred in a localized isolated membrane area (within the onchip tungsten micro-heater), the on-chip electronic circuits were not affected. At the same time, the case of an open cavity package (for gas sensing) was also considered, and therefore it is assumed that the materials used in sensor die packaging are not exposed to the high temperature levels. However, for this study, the indirect effect of the mechanical stress caused by the thermal cycling during operation on both the membrane and the die attachment to the package has not been taken into account. It was also assumed that the moisture effects will be negligible at such high temperatures. The initial measurements indicated that the maximum deviations in the thermodiode forward voltage over the 100 h time
period were 10 mV ( 7 1C) and 48 mV ( 40 1C) for 400 and 500 1C, respectively. These of course were unacceptable. However, a more thorough investigation revealed that actually, this voltage drift was not due to a change in the thermodiode parameters. Instead, this was caused by an increase in the resistance (R) of the tungsten micro-heater (as shown in Fig. 10), which in turn led to a corresponding increase in the actual temperature. Taking this into account, it was found that the drift in the thermodiode forward voltage (DV) was only 1–2 mV ( 1 1C), as shown in Fig. 11. The comparison of the relative change in tungsten microheater resistance (DR/R) and thermodiode voltage (DV/V) is shown in Fig. 12(a) and (b) at 400 and 500 1C, respectively. It was found that the DR/R were 1% and 4% at 400 and 500 1C, respectively after 100 h of operation, whereas DV/V were close to zero. So it confirms that the performance of the thermodiode temperature sensor is more reliable than that of tungsten RTDs,
200 Tungsten micro-heater resistance (Ω)
544
500 °C
196 194 192 190
400 °C
188 0
20
40 60 Time (hour)
80
100
Fig. 10. Tungsten micro-heater resistance change with time.
0.5 400 °C
0.4 Forward voltage drop (V)
50 65 μA
Nonlinearity relative error (%)
198
1 μA
40
14 nA
30
20
500 °C
0.3
0.2
0.1 Measured forward voltage drop Corrected forward voltage drop
10 285
385
525
0.0 0
0 0
200
400 Temperature (°C)
Fig. 9. Non-linearity relative error versus temperature plot.
600
20
40 60 Time (hour)
80
100
Fig. 11. Long term continuous operation of thermodiode at 400 and 500 1C for 100 h. The first curve (solid) is the measured forward voltage drop across the thermodiode. The second curve (dotted) has been corrected by taking into account the temperature change (due to the change in micro-heater resistance) to represent the thermodiode characteristic more accurately.
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
545
I. Haneef would like to acknowledge financial support by Higher Education Commission of Pakistan for his Ph.D. studies.
4 Change in resistance of tungsten micro-heater
3
Change in voltage of the thermodiode
ΔR/R, ΔV/V (%)
References
2
1
400 °C
0
-1
-2 0
20
40
60 Time (hour)
80
100
20 Change in resistance of tungsten micro-heater Change in voltage of the thermodiode
ΔR/R, ΔV/V (%)
10 500 °C
0
-10
-20 0
20
40 60 Time (hour)
80
100
Fig. 12. Relative change in tungsten micro-heater resistance and thermodiode voltage with time at (a) 400 1C and (b) 500 1C.
which might suffer from electro-migration and piezo-resistive effect (as the stress increases with temperature).
4. Conclusion In this paper the performance and long-term stability of a SOI thermodiode beyond 300 1C has been reported. The thermodiode is capable of operating linearly up to very high temperatures (up to 500 1C) and its characteristics match well with numerical and analytical predictions. The piezo-junction effect within the thermodiode was shown to be negligible. The thermodiodes are fairly reliable even after 100 h of continuous operation at 400 and 500 1C, and the maximum deviations were only 1–2 mV ( 1 1C). It was also shown that the membrane thermodiodes are more stable than metal RTDs (placed on the membrane) at high temperatures.
Acknowledgement The work was supported by Engineering and Physical Sciences Research Council (EPSRC), UK under the Project no. EP/F004931/1.
[1] G.C.M. Meijer, Thermal sensors based on transistors, Sensors Actuator 10 (1986) 103–125. [2] R.A. Bianchi, F. Vinci Dos Santos, J.M. Karam, B. Courtois, F. Pressecq, S. Sifflet, CMOS—compatible smart temperature sensors, Microelectron. J. 29 (1998) 627–636. [3] G. Wang, G.C.M. Meijer, Temperature characteristics of bipolar transistors fabricated in CMOS technology, Sensors Actuator A—Phys. 87 (2000) 81–89. [4] C. Falconi, M. Fratini, CMOS microsystems temperature control, Sensors Actuator B—Chem. 129 (2008) 59–66. [5] N.S. Boltovets, V.V. Kholevchuk, R.V. Konakova, V.F. Mitin, E.F. Venger, Ge-film resistance and Si-based diode temperature microsensors for cryogenic applications, Sensors Actuator A—Phys. 92 (2001) 191–196. [6] M. Bose, S.B. Ota, Study of forward characteristics of a cryogenic temperature sensor diode, Rev. Sci. Instrum. 67 (1996) 4176–4178. [7] Y. Farah, J. H. Sondericker, Precision cryogenic temperature data acquisition system, in: Advances in Cryogenic Engineering, Cambridge, MA, USA, 1986, 1315–1320. [8] P.C. McDonald, Silicon diode temperature sensors for process systems, Cryogenics 35 (1995) 199–201. [9] S.B. Ota, P.K. Mishra, V.C. Sahni, M.R. Singh, J. Bascunan, Study of magnetic properties of a cryogenic temperature sensor diode, Rev. Sci. Instrum. 71 (2000) 1873–1875. [10] Y.M. Shwarts, V.L. Borblik, N.R. Kulish, V.N. Sokolov, M.M. Shwarts, E.F. Venger, Silicon diode temperature sensor without a kink of the response curve in cryogenic temperature region, Sensors Actuator A—Phys. 76 (1999) 107–111. [11] J. Veprek, P. Strnad, Stability of silicon diodes as temperature sensors in the range 4.2–273 K, Cryogenics 24 (1984) 245–248. [12] R. Kersjes, J. Eichholz, A. Langerbein, Y. Manoli, W. Mokwa, An integrated sensor for invasive blood-velocity measurement, Sensors Actuator A—Phys. 37–38 (1993) 674–678. [13] R. Kersjes, W. Mokwa, Fast liquid flow sensor with thermal isolation by oxide-filled trenches, Sensors Actuator A—Phys. 47 (1995) 373–379. [14] L. Lofdahl, G. Stemme, B. Johansson, Turbulence measurements using sensors based on silicon technology, in: ICIASF Record, International Congress on Instrumentation in Aerospace Simulation Facilities, 1989, 95–103. [15] M. Schiffer, E. Obermeier, F. Grewe, A. Ebner, H.H. Fernholz, A new height-adjustable AeroMEMS surface fence probe fabricated in SOI technology for high resolution wall shear stress measurement in turbulent flows, in: Digest of Technical Papers-International Conference on Solid State Sensors and Actuators and Microsystems, Transducers ‘05, Seoul, 2005, 601–604. [16] P.J. Dempsey, R.H. Fabik, Using silicon diodes for detecting the liquid–vapor interface in hydrogen, in: Proceedings of the International Instrumentation Symposium, Las Vegas, NV, USA, 1992, 287–299. [17] S. Banerjee, V.R. Bhave, G.N. Navaneeth, Electronic wet and dry, direct and differential recording thermometer, Indian J. Environ. Health 22 (1980) 93–95. [18] S. Wu, H. Zhou, S. Yang, R. Wu, Y. Chen, Development of a new temperature and humidity sensitive integrated sensor, Guti Dianzixue Yanjiu Yu Jinzhan/ Res. Prog. Solid State Electron. 18 (1998) 349–353. [19] M. Kimura, K. Kikuchi, Thermistor-like pn diode temperature sensor and a new method to measure the absolute humidity using these temperature sensors combined with a microheater, in: Proceedings of the 2nd IEEE Conference on Sensors, Toronto, Canada, 2003, 636–641. [20] Y.L. Chin, J.C. Chou, T.P. Sun, W.Y. Chung, S.K. Hsiung, A novel pH sensitive ISFET with on chip temperature sensing using CMOS standard process, Sensors Actuator, B—Chem. 76 (2001) 582–593. [21] M. Kimura, F. Sakurai, H. Ohta, T. Terada, Proposal of a new structural thermal vacuum sensor with diode-thermistors combined with a micro-air-bridge heater, Microelectron. J. 38 (2007) 171–176. [22] R.K. Gorkhali, PC temperature alarm, Electron. World 114 (2008) 36–38. [23] C. Babilya, Digital thermometer uses silicon diodes to sense rapid temperature changes, Electron. Des. 29 (1981) 265–266. [24] J. Barnes, Differential-temperature sensor is very inexpensive, Electron. Des. News 18 (1973) 90–91. [25] D.C. Elmore, Diodes act as a temperature sensor in remote temperaturemeasuring circuit, Electron. Des. 24 (1976). [26] W.S. Lee, I.Y. Han, J. Yu, S.J. Kim, K.Y. Byun, Thermal characterization of thermally conductive underfill for a flip-chip package using novel temperature sensing technique, Thermochim. Acta 455 (2007) 148–155. [27] C.K. Saul, J.N. Zemel, Diode-based microfabricated hot-plate sensor, Sensors Actuator A—Phys. 65 (1998) 128–135. [28] Z. Wang, K. Tian, Y. Zhou, L. Pan, C. Hu, L. Liu, A high-temperature silicon-oninsulator stress sensor, J. Micromech. Microeng. 18 (2008) 045018. [29] P. Neuzil, Y. Liu, H.H. Feng, W. Zeng, Micromachined bolometer with singlecrystal silicon diode as temperature sensor, IEEE Electron. Dev. Lett. 26 (2005) 320–322.
546
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
[30] J.K. Kim, C.H. Han, A new uncooled thermal infrared detector using silicon diode, Sensors Actuator A—Phys. 89 (2001) 22–27. [31] M. Ueno, Y. Kosasayama, T. Sugino, Y. Nakaki, Y. Fujii, H. Inoue, K. Kama, T. Seto, M. Takeda, M. Kimata, 640 480 pixel uncooled infrared FPA with SOI diode detectors, in: Proceedings of SPIE–5783, Article no. 63, Orlando, FL, 2005, 566–577. [32] M. Graf, D. Barrettino, K.U. Kirstein, A. Hierlemann, CMOS microhotplate sensor system for operating temperatures up to 500 1C, Sensors Actuator B—Chem. 117 (2006) 346–352. [33] S.Z. Ali, P.K. Guha, C.C.C. Lee, F. Udrea, W.I. Milne, T. Iwaki, J. Covington, J. W. Gardner, S. Maeng, J. Park, High temperature SOI CMOS tungsten microheaters, in: Proceedings of the 5th IEEE Conference on Sensors, Daegu, Korea, 2006, 847–850. [34] P.K. Guha, S.Z. Ali, C.C.C. Lee, F. Udrea, W.I. Milne, T. Iwaki, J.A. Covington, J.W. Gardner, Novel design and characterisation of SOI CMOS micro-hotplates for high temperature gas sensors, Sensors Actuator B—Chem. 127 (2007) 260–266.
[35] S.S. Ang, High-temperature characteristics of epitaxial high-low junction n + – n–p+ silicon diodes, Microelectron. J. 26 (1995) 375–382. [36] Y.M. Shwarts, V.L. Borblik, N.R. Kulish, E.F. Venger, V.N. Sokolov, Limiting characteristics of diode temperature sensors, Sensors Actuator A—Phys. 86 (2000) 197–205. [37] M. Kimura, K. Toshima, Thermistor-like pn junction temperature-sensor with variable sensitivity and its combination with a micro-air-bridge heater, Sensors Actuator A—Phys. 108 (2003) 239–243. [38] S. Santra, P.K. Guha, M.S. Haque, S.Z. Ali, F. Udrea, Si diode temperature sensor beyond 300 1C, in: Proceedings of the IEEE International Semiconductor Conference, CAS, Sinaia, Romania, 2007, 415–418. [39] S.Z. Ali, F. Udrea, W.I. Milne, J.W. Gardner, Tungsten based SOI microhotplates for smart gas sensors, J. Microelectromech. Syst. 17 (2008) 1408–1417. [40] S. Santra, P.K. Guha, S.Z. Ali, I. Haneef, F. Udrea, SOI diode temperature sensor operated at ultra high temperatures—a critical analysis, in: Proceedings of the 7th IEEE Conference on Sensors, Lecce, Italy, 2008, 78–81.
Contents lists available at ScienceDirect
Microelectronics Journal journal homepage: www.elsevier.com/locate/mejo
Ultra-high temperature (4300 1C) suspended thermodiode in SOI CMOS technology S. Santra a,, F. Udrea a, P.K. Guha a,b, S.Z. Ali a, I. Haneef a,c a
Department of Engineering, University of Cambridge, 9JJ Thomson Avenue, Cambridge CB3 0FA, UK School of Engineering, University of Warwick, Coventry CV4 7AL, UK c Institute of Avionics and Aeronautics Air University, E-9, Islamabad 44000, Pakistan b
a r t i c l e in f o
a b s t r a c t
Available online 4 January 2010
This paper reports for the first time on the performance and long-term stability of a silicon on insulator (SOI) thermodiode with tungsten metallization, suspended on a dielectric membrane, at temperatures beyond 300 1C. The thermodiode has been designed and fabricated with minute saturation currents (due to both small size and the use of SOI technology) to allow an ultra-high temperature range and minimal non-linearity. It was found that the thermodiode forward voltage drop versus temperature plot remains linear up to 500 1C, with a non-linearity error of less than 7%. Extensive experimental results on performance of the thermodiode that was fabricated using a Complementary Metal Oxide Semiconductor (CMOS) SOI process are presented. These results are backed up by infrared measurements and a range of 2-D (dimension) and 3-D simulations using ISE and ANSYS software. The on-chip drive electronics for the thermodiode and the micro-heater, as well as the sensor transducing circuit were placed adjacent to the membrane. We demonstrate that the thermodiode is considerably more reliable in long-term direct current operation at high temperatures when compared to the more classical resistive temperature detectors (RTDs) using CMOS metallization layers (tungsten or aluminum). We also compare a membrane thermodiode with a reference thermodiode placed on the silicon substrate and assess their relative performance at elevated temperatures. The experimental results from this comparison confirm that the thermodiode suffers minimal piezo-junction/ piezo-resistive effects. & 2009 Elsevier Ltd. All rights reserved.
Keywords: SOI CMOS Thermodiode RTDs High temperature sensors Smart sensors
1. Introduction Temperature sensors are one of the fastest growing segments in sensors’ market. An integrated temperature sensor for thermal management is a core component in power hungry circuits that tend to operate close to the maximum junction temperature. In such systems, accurate monitoring of the junction temperature is mandatory to optimize the integrated circuits (ICs) performance while maintaining high reliability. Most of the Complementary Metal Oxide Semiconductor (CMOS) processes now target higher junction temperatures to allow increased packing density of transistors, better cost-performance value and more powerful processing. Maximum junction temperatures in bulk CMOS ICs have moved from a conservative level of 125–150 1C [1–4] and even to 175 1C. By using copper or tungsten, which are more resistant to electro-migration, some of these processes can potentially move to 200 1C, provided that they address issues like latch-up and low cross-talk, and overcome reliability problems such as negative bias temperature instability (NBTI), time
Corresponding author. Tel.: + 44 1223 748311; fax: + 44 1223 748348.
E-mail address: [email protected] (S. Santra). 0026-2692/$ - see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2009.12.005
dependent dielectric breakdown (TDDB), etc. Furthermore, the additional use of silicon on insulator (SOI) technology in ICs not only suppresses the latch-up but also minimizes the leakage currents. It also provides an excellent vertical and lateral isolation and thereby allows a further increase in the maximum junction temperatures to 2251 or potentially even 250 1C. Such ICs can be of use in automotive electronics, power supplies, motor control or other power systems. The silicon thermodiodes (i.e. silicon diode temperature sensors) have been reported in the past for diverse applications. These applications include, for example, temperature measurement for cryogenic applications [5–11], flow sensing [12–15], liquid–vapour interface and liquid level point sensing in hydrogen [16], humidity sensing [17–19], pH sensing [20], vacuum sensing [21], PC temperature monitoring [22], thermometry [17,23–25], thermal characterization of thermally conductive underfill for flip-chip packaging [26], pure gas and gas mixtures’ thermal conductivity monitoring [27], temperature-compensation of piezo-resistive stress sensors [28], MEMS bolometers [29], infrared (IR) detectors [30] and IR focal plane arrays [31], etc. Although most of these applications require maximum junction temperature below 200 1C and more often below 150 1C, yet there are some emerging silicon-based sensors for which accurate
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
and reliable temperature monitoring is essential at very high temperatures (i.e. up to 500 1C), well beyond the junction temperatures of standard ICs. The examples of such applications are smart micro-calorimeters [32], resistive gas sensors [33,34] and sensors used in automotive engines, exhausts, etc. Such smart sensors use membrane technologies for thermal isolation of the sensors (e.g. micro-calorimeters) that typically operate at 400 or 500 1C. As a result, while the active sensing element (e.g. gas sensitive layer) suspended on a very thin dielectric membrane would operate at high temperatures for optimal sensing, the onchip electronics can still operate very close to the ambient temperature. For such sensors, accurate monitoring of the temperature at the hot-spot of the membrane is absolutely essential to enhance the sensor sensitivity and selectivity (as it is the case in gas sensors) and last but not least, for reliability assessment. To meet the challenge of temperature monitoring in these sensors, there is a need for IC/CMOS temperature sensors that can provide accurate temperature measurement, are small in size, high in sensitivity, and reliable in performance beyond 300 1C. To the authors’ knowledge, there have been very few reports [5,35–37] on IC temperature sensors that can operate at such high temperatures. In this paper we will report on the use of a membrane thermodiode operating at temperatures well beyond 300 1C [38]. Linearity is preserved up to 500 1C and the maximum temperature, beyond which the saturation current becomes comparable with the drive current of the thermodiode, is around 600 1C. We demonstrate for the first time that at very high temperatures, the membrane thermodiode offers better reliability than equivalent metal resistive temperature detectors (RTDs) using CMOS metals such as aluminum or tungsten while maintaining very high linearity. Furthermore, we have seen no evidence of piezojunction/piezo-resistive effect in the suspended thermodiode, which would otherwise limit the operation of silicon-based resistive sensors at very high temperatures.
Tungsten micro-heater
541
Gas sensing material NMOS
Membrane thermodiode Membrane area
Reference thermodiode
PMOS
N+ N+ P+ P+ Nwell Pwell
Electronics area
Fig. 1. Cross-sectional view of membrane thermodiode, a reference thermodiode and the CMOS electronic cells.
N+
Contacts to Metal 1
P
P+
Thermodiode and micro-heater
2. Thermodiode design, fabrication and on-chip circuitry In chemical sensor, the micro-hotplate (MHP) is a region of the sensor placed on a thin dielectric membrane, which contains a micro-heater. The temperature in this region is quasi-uniform and considerably higher than that outside the membrane. We designed our MHP to contain a tungsten micro-heater and placed the drive and signal processing electronics outside the membrane. For accurate monitoring of the MHP temperature we have placed a thermodiode, which we refer to as the membrane thermodiode, right in the centre of the membrane under the MHP. An additional thermodiode is placed outside the membrane to monitor the junction temperature of the IC chip (which in our case, where the IC power consumption is negligible when compared to that of the sensor, is often close to the ambient temperature). A schematic cross-section of the membrane thermodiode, a reference thermodiode and the CMOS electronic cells are shown in Fig. 1. The MHP, SOI thermodiode was designed in Cadence 5.0 software. The Cadence layout of the thermodiode is shown in Fig. 2(a). The diameter of the diode was 34 mm. Note that the shapes of both the membrane and the MHP are chosen to be circular to minimize the mechanical stress at the membrane edge. The thermodiodes integrated micro-hotplates were fabricated using a commercial 6 inch 1.0 mm SOI CMOS process. This process features a 0.25 mm active silicon layer, a 1.0 mm buried oxide layer and triple high temperature metallization based on tungsten. The use of tungsten metallization allows operation of the MHP at high temperatures (up to 700 1C). The tungsten layer was used for the resistive micro-heater (metal 1 layer), contacts of the two
SOI dielectric membrane
50 μm
Fig. 2. (a). Cadence layout of the SOI p+ /p/n+ thermodiode with diameter of 34 mm. (b). The optical micrograph of a fabricated micro-hotplate with SOI thermodiode temperature sensor embedded under the hotplate, within the oxide membrane.
terminal diodes (metal 1, 2 layer) and interdigitated electrodes (metal 3 layers). These interdigitated electrodes are used specifically for gas sensing material’s resistance measurement. After CMOS fabrication, the wafers underwent a single deep reactive ion etching (DRIE) back-etch step at a separate micro electrical mechanical systems (MEMS) foundry to form the thin oxide–nitride membrane (ca. 5 mm) in specific areas. The details of the micro-heater design and characterization have been reported elsewhere [39]. An optical microscope picture of the fabricated MHP (with the membrane thermodiode and micro-heater) is shown in Fig. 2(b). The diameter of the membrane on which micro-heater and thermodiode are embedded at its centre, is 300 mm. As already mentioned, a reference thermodiode was placed on the silicon substrate to monitor the ambient temperature and also assess the piezo-junction effect. The structure of the reference thermodiode is identical to that of the membrane
542
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
25
20
I
Power (mW)
I
+
15
10
IA
5 – Reference thermodiode
Membrane thermodiode
0 0
200
Fig. 3. Instrumentation amplifier (IA) circuit for temperature measurement.
thermodiode. To eliminate the effect of the ambient temperature, a simple differential circuit, as shown in Fig. 3, was implemented on-chip. The current source and an instrumentation amplifier (IA), whose gain was controlled externally, were integrated on-chip (outside the membrane area) for good matching characteristics.
600
400 Temperature (°C)
Fig. 4. Power versus temperature plot of the micro-heater which was taken at different positions of the wafer.
3. Results and discussion
Tungsten micro-heater
3.1. Micro-hotplate characterization 3.1.1. Power versus temperature and transient response The tungsten resistive micro-heater was calibrated to verify its temperature coefficient of resistance (TCR) using a high precision (Signatones model S-1060R QuieTemp DC) hot chuck system. The hot chuck system has 1 1C resolution and can operate at temperatures up to 300 1C. A Hewlett Packard 4142B modular DC source/monitor and a Signatones model S-1160 probe station was used for electrical measurements (I–V characterization). The MHP exhibits an ultra-low power consumption (16 mW for 600 1C), very short transient times (2 ms to 600 1C), and extremely good reproducibility within a wafer, and from wafer to wafer. A power versus temperature plot (measured at different positions of the wafer) calculated for a tungsten micro-heater using its I–V data and TCR is shown in Fig. 4. The fast transient response of the device enables heating up and cooling down of the MHP very quickly. 3.1.2. Temperature rise/distribution: ANSYS simulations and infrared imaging Fig. 5 shows the temperature distribution over the MHP simulated using ANSYS. Infrared thermal imaging of the membrane was also carried out using a Quantum Focus Instruments Infrascope II IR imaging scope to verify the temperature distribution across the surface of the membrane. The IR image of a micro-hotplate at 250 1C is shown in Fig. 6 (the 250 1C limit was imposed by the detection range of the IR system). The IR imaging results (shown in Fig. 6) confirm that the temperature distribution is uniform in the heater region (deviation within 5%) while it decreases rapidly beyond that, across the rest of the membrane. This is also confirmed by the ANSYS simulations. We can therefore, assume that the temperature is approximately constant within the MHP and hence we can ignore the temperature gradients inside the membrane thermodiode.
800
Membrane
25
200
350
600
780 °C
Fig. 5. Temperature distribution over the membrane as determined by ANSYS simulations.
3.2. SOI thermodiode performance 3.2.1. V–T plot up to 300 1C and piezo-junction effect A diode can be used as a temperature sensor when it is driven by a constant forward current. In this operation scheme, the forward voltage of the thermodiode decreases linearly with the increase in temperature. The forward voltage versus temperature (V–T) measurements of the membrane thermodiode and the reference thermodiode were performed up to 300 1C using the hot chuck. The V–T plot for both on-membrane and off-membrane thermodiodes is shown in Fig. 7. A comparison of the V–T plot for the two thermodiodes confirms that there is very little difference between their characteristics indicating that there is minimal piezo-junction effect. It also indicates that the residual and thermally induced mechanical stress in the membrane has negligible influence over the forward voltage temperature characteristics of the thermodiode. The slope of the thermodiodes is found to be 1.3 mV/1C at 65 mA forward current. The reason we chose such low current ( mA) for the temperature sensor operation is to avoid any self heating effect.
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
543
1.0 Experimental
Thermodiode and micro-heater
0.8 Forward voltage drop (V)
Tracks
Numerical Analytical
0.6 65 μA
0.4 1 μA
0.2
525
385
14 nA
285
0.0 0 Membrane
0.9 Reference thermodiode Suspended thermodiode
Forward voltage drop (V)
400 Temperature (°C)
600
800
Fig. 8. Experimental, numerical and analytical V–T plot.
Fig. 6. IR image of the microhotplate at 250 1C. Note that the temperature profile is more accurate in the micro-heater area as beyond this the dielectric membrane is transparent to IR.
0.8
0.7
0.6
0.5 0
200
100
200 Temperature (°C)
300
Fig. 7. Forward voltage drop versus temperature (V–T) plot of the suspended and reference thermodiode.
Indeed, the power consumption of the diode, when it is driven with 65 mA driving current, is only 0.12% of that of the MHP (operated at 500 1C) introducing errors of less than 1 1C. 3.2.2. V–T plot up to 780 1C: Experimental versus analytical and numerical results To characterize the thermodiode beyond 300 1C, the onmembrane tungsten micro-heater was used. The forward voltage of the thermodiode at different temperature was measured at three forward current levels (65 mA, 1 mA and 14 nA) from 25 to
780 1C. To obtain such high temperatures, we have used tungsten metallization (instead of the standard aluminum) and thin SOI layers (0.25 mm). This choice minimizes the depletion region volume and thus obtains a very low value for the diode saturation current (Is). From the V–T characteristics of the thermodiode shown in Fig. 8, it can be clearly observed that the voltage versus temperature slope linearity is maintained up to high temperatures (e.g. 500 1C at 65 mA driving current), which is due to this very low Is (Is few fA at room temperature). At high temperatures, the saturation current becomes comparable or even higher than the driving current and the consequence of this is that the V–T characteristics become non-linear. It is important to note that the V–T slope and the maximum temperature beyond which the saturation current becomes comparable with the drive current depend entirely on the driving current. The V–T slope changes from ( 2.2) to ( 1.3) mV/1C due to the change in driving current from 14 nA to 65 mA, respectively. A higher driving current leads to a lower sensitivity but maintains the linearity of the diode up to a higher temperature. Here it is interesting to mention that in earlier reports on high temperature operation of thermodiodes, Boltovets et al. [5] and Shwarts et al. [36] presented silicon diode temperature sensors operating up to 327 1C. Ang [35] showed the electrical characteristics of n+ /n/p+ silicon diodes in the temperature range 27–310 1C. In all the three reports [5,35,36] the thermodiodes were used as a discrete device, and were not integrated with onchip CMOS circuitry. Kimura and Toshima [37] have also reported a thermistor-like p–n junction temperature sensor that can cover very wide temperature range ( 200 to 500 1C). However, they have operated the diode in the forward bias-voltage mode only for the temperature range 200 to 150 1C whereas beyond that, for 150–500 1C, they used the diode in the reverse bias-voltage mode. The temperature measurement of the reverse current is not only difficult given its very small level, but also not reliable as often this current varies very significantly from wafer to wafer and lot to lot. Also its linearity is relatively poor. In our case we have used a CMOS temperature sensor in forward biased mode with an onchip constant current drive. This allows a more accurate measurement, and easier detection and post-processing of the results. To understand the V–T characteristics of the SOI thermodiode fully, extensive finite element analysis (FEM) in ISE TCAD software
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
was also carried out. The thermodiode was also characterized by analytical modeling. These numerical and analytical predictions have also been compared with the experimental results in Fig. 8, and all three are in excellent agreement with each other. For further details of the analytical calculations and an analysis of the thermodiode’s non-linear behaviour at very high temperatures, the readers may refer to [40]. It is worth mentioning that we also picked up thermodiodes randomly from different locations across one wafer and from different wafer lots. These thermodiodes were then evaluated experimentally to verify the repeatability of the results. The results were found to be highly uniform and repeatable. As mentioned earlier, the thermodiodes were driven by three different driving currents (i.e. 14 nA, 1 mA and 65 mA). Corresponding to each driving current, the maximum temperatures beyond which the thermodiodes’ response became non-linear were calculated assuming the maximum error of 10% of the non-linear terms. The results shown in Fig. 9 point out that the non-linearity starts dominating at lower temperatures for the lower driving current. This is because in this case, the saturation current approaches the value of driving current at a lower temperature level. 3.3. Long term performance stability of thermodiode To evaluate its long-term stability, the thermodiode was operated at 400 and 500 1C for 100 h using the tungsten microheater on the membrane. The tungsten micro-heater (i.e. metal RTD) was operated using a constant current while the thermodiode was driven at a forward current of 65 mA. As the heating occurred in a localized isolated membrane area (within the onchip tungsten micro-heater), the on-chip electronic circuits were not affected. At the same time, the case of an open cavity package (for gas sensing) was also considered, and therefore it is assumed that the materials used in sensor die packaging are not exposed to the high temperature levels. However, for this study, the indirect effect of the mechanical stress caused by the thermal cycling during operation on both the membrane and the die attachment to the package has not been taken into account. It was also assumed that the moisture effects will be negligible at such high temperatures. The initial measurements indicated that the maximum deviations in the thermodiode forward voltage over the 100 h time
period were 10 mV ( 7 1C) and 48 mV ( 40 1C) for 400 and 500 1C, respectively. These of course were unacceptable. However, a more thorough investigation revealed that actually, this voltage drift was not due to a change in the thermodiode parameters. Instead, this was caused by an increase in the resistance (R) of the tungsten micro-heater (as shown in Fig. 10), which in turn led to a corresponding increase in the actual temperature. Taking this into account, it was found that the drift in the thermodiode forward voltage (DV) was only 1–2 mV ( 1 1C), as shown in Fig. 11. The comparison of the relative change in tungsten microheater resistance (DR/R) and thermodiode voltage (DV/V) is shown in Fig. 12(a) and (b) at 400 and 500 1C, respectively. It was found that the DR/R were 1% and 4% at 400 and 500 1C, respectively after 100 h of operation, whereas DV/V were close to zero. So it confirms that the performance of the thermodiode temperature sensor is more reliable than that of tungsten RTDs,
200 Tungsten micro-heater resistance (Ω)
544
500 °C
196 194 192 190
400 °C
188 0
20
40 60 Time (hour)
80
100
Fig. 10. Tungsten micro-heater resistance change with time.
0.5 400 °C
0.4 Forward voltage drop (V)
50 65 μA
Nonlinearity relative error (%)
198
1 μA
40
14 nA
30
20
500 °C
0.3
0.2
0.1 Measured forward voltage drop Corrected forward voltage drop
10 285
385
525
0.0 0
0 0
200
400 Temperature (°C)
Fig. 9. Non-linearity relative error versus temperature plot.
600
20
40 60 Time (hour)
80
100
Fig. 11. Long term continuous operation of thermodiode at 400 and 500 1C for 100 h. The first curve (solid) is the measured forward voltage drop across the thermodiode. The second curve (dotted) has been corrected by taking into account the temperature change (due to the change in micro-heater resistance) to represent the thermodiode characteristic more accurately.
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
545
I. Haneef would like to acknowledge financial support by Higher Education Commission of Pakistan for his Ph.D. studies.
4 Change in resistance of tungsten micro-heater
3
Change in voltage of the thermodiode
ΔR/R, ΔV/V (%)
References
2
1
400 °C
0
-1
-2 0
20
40
60 Time (hour)
80
100
20 Change in resistance of tungsten micro-heater Change in voltage of the thermodiode
ΔR/R, ΔV/V (%)
10 500 °C
0
-10
-20 0
20
40 60 Time (hour)
80
100
Fig. 12. Relative change in tungsten micro-heater resistance and thermodiode voltage with time at (a) 400 1C and (b) 500 1C.
which might suffer from electro-migration and piezo-resistive effect (as the stress increases with temperature).
4. Conclusion In this paper the performance and long-term stability of a SOI thermodiode beyond 300 1C has been reported. The thermodiode is capable of operating linearly up to very high temperatures (up to 500 1C) and its characteristics match well with numerical and analytical predictions. The piezo-junction effect within the thermodiode was shown to be negligible. The thermodiodes are fairly reliable even after 100 h of continuous operation at 400 and 500 1C, and the maximum deviations were only 1–2 mV ( 1 1C). It was also shown that the membrane thermodiodes are more stable than metal RTDs (placed on the membrane) at high temperatures.
Acknowledgement The work was supported by Engineering and Physical Sciences Research Council (EPSRC), UK under the Project no. EP/F004931/1.
[1] G.C.M. Meijer, Thermal sensors based on transistors, Sensors Actuator 10 (1986) 103–125. [2] R.A. Bianchi, F. Vinci Dos Santos, J.M. Karam, B. Courtois, F. Pressecq, S. Sifflet, CMOS—compatible smart temperature sensors, Microelectron. J. 29 (1998) 627–636. [3] G. Wang, G.C.M. Meijer, Temperature characteristics of bipolar transistors fabricated in CMOS technology, Sensors Actuator A—Phys. 87 (2000) 81–89. [4] C. Falconi, M. Fratini, CMOS microsystems temperature control, Sensors Actuator B—Chem. 129 (2008) 59–66. [5] N.S. Boltovets, V.V. Kholevchuk, R.V. Konakova, V.F. Mitin, E.F. Venger, Ge-film resistance and Si-based diode temperature microsensors for cryogenic applications, Sensors Actuator A—Phys. 92 (2001) 191–196. [6] M. Bose, S.B. Ota, Study of forward characteristics of a cryogenic temperature sensor diode, Rev. Sci. Instrum. 67 (1996) 4176–4178. [7] Y. Farah, J. H. Sondericker, Precision cryogenic temperature data acquisition system, in: Advances in Cryogenic Engineering, Cambridge, MA, USA, 1986, 1315–1320. [8] P.C. McDonald, Silicon diode temperature sensors for process systems, Cryogenics 35 (1995) 199–201. [9] S.B. Ota, P.K. Mishra, V.C. Sahni, M.R. Singh, J. Bascunan, Study of magnetic properties of a cryogenic temperature sensor diode, Rev. Sci. Instrum. 71 (2000) 1873–1875. [10] Y.M. Shwarts, V.L. Borblik, N.R. Kulish, V.N. Sokolov, M.M. Shwarts, E.F. Venger, Silicon diode temperature sensor without a kink of the response curve in cryogenic temperature region, Sensors Actuator A—Phys. 76 (1999) 107–111. [11] J. Veprek, P. Strnad, Stability of silicon diodes as temperature sensors in the range 4.2–273 K, Cryogenics 24 (1984) 245–248. [12] R. Kersjes, J. Eichholz, A. Langerbein, Y. Manoli, W. Mokwa, An integrated sensor for invasive blood-velocity measurement, Sensors Actuator A—Phys. 37–38 (1993) 674–678. [13] R. Kersjes, W. Mokwa, Fast liquid flow sensor with thermal isolation by oxide-filled trenches, Sensors Actuator A—Phys. 47 (1995) 373–379. [14] L. Lofdahl, G. Stemme, B. Johansson, Turbulence measurements using sensors based on silicon technology, in: ICIASF Record, International Congress on Instrumentation in Aerospace Simulation Facilities, 1989, 95–103. [15] M. Schiffer, E. Obermeier, F. Grewe, A. Ebner, H.H. Fernholz, A new height-adjustable AeroMEMS surface fence probe fabricated in SOI technology for high resolution wall shear stress measurement in turbulent flows, in: Digest of Technical Papers-International Conference on Solid State Sensors and Actuators and Microsystems, Transducers ‘05, Seoul, 2005, 601–604. [16] P.J. Dempsey, R.H. Fabik, Using silicon diodes for detecting the liquid–vapor interface in hydrogen, in: Proceedings of the International Instrumentation Symposium, Las Vegas, NV, USA, 1992, 287–299. [17] S. Banerjee, V.R. Bhave, G.N. Navaneeth, Electronic wet and dry, direct and differential recording thermometer, Indian J. Environ. Health 22 (1980) 93–95. [18] S. Wu, H. Zhou, S. Yang, R. Wu, Y. Chen, Development of a new temperature and humidity sensitive integrated sensor, Guti Dianzixue Yanjiu Yu Jinzhan/ Res. Prog. Solid State Electron. 18 (1998) 349–353. [19] M. Kimura, K. Kikuchi, Thermistor-like pn diode temperature sensor and a new method to measure the absolute humidity using these temperature sensors combined with a microheater, in: Proceedings of the 2nd IEEE Conference on Sensors, Toronto, Canada, 2003, 636–641. [20] Y.L. Chin, J.C. Chou, T.P. Sun, W.Y. Chung, S.K. Hsiung, A novel pH sensitive ISFET with on chip temperature sensing using CMOS standard process, Sensors Actuator, B—Chem. 76 (2001) 582–593. [21] M. Kimura, F. Sakurai, H. Ohta, T. Terada, Proposal of a new structural thermal vacuum sensor with diode-thermistors combined with a micro-air-bridge heater, Microelectron. J. 38 (2007) 171–176. [22] R.K. Gorkhali, PC temperature alarm, Electron. World 114 (2008) 36–38. [23] C. Babilya, Digital thermometer uses silicon diodes to sense rapid temperature changes, Electron. Des. 29 (1981) 265–266. [24] J. Barnes, Differential-temperature sensor is very inexpensive, Electron. Des. News 18 (1973) 90–91. [25] D.C. Elmore, Diodes act as a temperature sensor in remote temperaturemeasuring circuit, Electron. Des. 24 (1976). [26] W.S. Lee, I.Y. Han, J. Yu, S.J. Kim, K.Y. Byun, Thermal characterization of thermally conductive underfill for a flip-chip package using novel temperature sensing technique, Thermochim. Acta 455 (2007) 148–155. [27] C.K. Saul, J.N. Zemel, Diode-based microfabricated hot-plate sensor, Sensors Actuator A—Phys. 65 (1998) 128–135. [28] Z. Wang, K. Tian, Y. Zhou, L. Pan, C. Hu, L. Liu, A high-temperature silicon-oninsulator stress sensor, J. Micromech. Microeng. 18 (2008) 045018. [29] P. Neuzil, Y. Liu, H.H. Feng, W. Zeng, Micromachined bolometer with singlecrystal silicon diode as temperature sensor, IEEE Electron. Dev. Lett. 26 (2005) 320–322.
546
S. Santra et al. / Microelectronics Journal 41 (2010) 540–546
[30] J.K. Kim, C.H. Han, A new uncooled thermal infrared detector using silicon diode, Sensors Actuator A—Phys. 89 (2001) 22–27. [31] M. Ueno, Y. Kosasayama, T. Sugino, Y. Nakaki, Y. Fujii, H. Inoue, K. Kama, T. Seto, M. Takeda, M. Kimata, 640 480 pixel uncooled infrared FPA with SOI diode detectors, in: Proceedings of SPIE–5783, Article no. 63, Orlando, FL, 2005, 566–577. [32] M. Graf, D. Barrettino, K.U. Kirstein, A. Hierlemann, CMOS microhotplate sensor system for operating temperatures up to 500 1C, Sensors Actuator B—Chem. 117 (2006) 346–352. [33] S.Z. Ali, P.K. Guha, C.C.C. Lee, F. Udrea, W.I. Milne, T. Iwaki, J. Covington, J. W. Gardner, S. Maeng, J. Park, High temperature SOI CMOS tungsten microheaters, in: Proceedings of the 5th IEEE Conference on Sensors, Daegu, Korea, 2006, 847–850. [34] P.K. Guha, S.Z. Ali, C.C.C. Lee, F. Udrea, W.I. Milne, T. Iwaki, J.A. Covington, J.W. Gardner, Novel design and characterisation of SOI CMOS micro-hotplates for high temperature gas sensors, Sensors Actuator B—Chem. 127 (2007) 260–266.
[35] S.S. Ang, High-temperature characteristics of epitaxial high-low junction n + – n–p+ silicon diodes, Microelectron. J. 26 (1995) 375–382. [36] Y.M. Shwarts, V.L. Borblik, N.R. Kulish, E.F. Venger, V.N. Sokolov, Limiting characteristics of diode temperature sensors, Sensors Actuator A—Phys. 86 (2000) 197–205. [37] M. Kimura, K. Toshima, Thermistor-like pn junction temperature-sensor with variable sensitivity and its combination with a micro-air-bridge heater, Sensors Actuator A—Phys. 108 (2003) 239–243. [38] S. Santra, P.K. Guha, M.S. Haque, S.Z. Ali, F. Udrea, Si diode temperature sensor beyond 300 1C, in: Proceedings of the IEEE International Semiconductor Conference, CAS, Sinaia, Romania, 2007, 415–418. [39] S.Z. Ali, F. Udrea, W.I. Milne, J.W. Gardner, Tungsten based SOI microhotplates for smart gas sensors, J. Microelectromech. Syst. 17 (2008) 1408–1417. [40] S. Santra, P.K. Guha, S.Z. Ali, I. Haneef, F. Udrea, SOI diode temperature sensor operated at ultra high temperatures—a critical analysis, in: Proceedings of the 7th IEEE Conference on Sensors, Lecce, Italy, 2008, 78–81.