A Study of Temperature Distribution in SOI- Smart Power Devices in Transient Conditions by Optical Interferometry a a a a a b N. Seliger , D. Pogany , C. Fürböck , P. Habaš , E. Gornik and M. Stoisiek a Institute for Solid State Electronics, TU Vienna Floragasse 7, A-1040 Vienna, Austria b Siemens Corporate Research and Development ZFE T KM6 Otto-Hahn-Ring 6, D-81739 Munich, Germany Abstract Thermally induced changes in the optical reflectivity of the Fabry-Perot resonator formed by the SOI layer structure are used to evaluate the temperature variation in SOI power devices under switching bias conditions. Measurements of the temperature distribution inside and outside the active device region in the SOI well are presented and compared to thermal simulation. 1. Introduction Silicon-On-Insulator (SOI) by Direct Wafer Bonding  has become an attractive technique for the fabrication of smart power devices. Self-heating effects in such structures are, however, more critical compared to bulk devices due to a reduced heat removal across the buried and trench oxides [2,3]. Power dissipation could cause localized temperature increase in the device active region, which may influence the device performance and reliability. In monolithic chip technology, the amount of heat laterally spread from the power devices across the sided trench oxides is also important as it may influence the characteristics of a nearby CMOS control circuit. In this paper we present an analysis of the transient temperature variations inside and outside the SOI well of smart power devices obtained by optical interferometry measurements. 2. Measurements The measurements have been carried out on trench isolated Lateral Double-diffused (LD)MOSFETs . The SOI wafer structure consists of a highly doped p+-substrate, the buried oxide of 3µm, the Si layer of 20µm thickness and the SiO2 passivation layer. The temperature variation in the Si layer is monitored by the intensity change of an infrared laser beam (λ=1.3µm) which is focused on the device from the top side (Fig.1a). A homogeneous temperature increase ∆ T in the Si layer of the thickness L causes an increase in the optical thickness ∆ LOPT due to the thermooptical effect: dn ∆LOPT = ⋅ ∆T ⋅ L , (1) dT dn ≈ 1.5·10-4 K-1 [5,6] is the temperature coefficient of the refractive index. Light dT absorption in the highly doped substrate causes that only the temperature-induced variation in the optical thickness of the SOI layer is detected via the reflectivity changes of the Fabry-Perot (F-P) resonator  formed by SiO2 passivation-layer/Si-layer/SiO2 buried oxide/Si-substrate (Fig.1). A homogeneous vertical temperature distribution in the well is assumed, which is reasonable for positions far from the MOS channel. Heating in the Si active layer is induced by applying short pulses of 20µs to 100µs from 0 to 12V to the gate under various drain-to-source biases. The duty cycle is chosen sufficiently long to assure device cooling between the pulses. In addition, the temperature evolution within and around the well is calculated by finite- element-simulation  using the measured Fig.1: a) Cross section of LDMOSFET with the laser beam indicated. The device is symmetric with respect to the drain metallization. b) Top view of the device with marked laser beam measurement positions (beam diameter 8µm, positions 1 to 6 are located in the center of the n--drift region). power dissipated in the device as input. 3. Results and discussion Fig.2: Intensity signal measured on different positions (see Fig.1b) inside and outside the LDMOS device. The intensity signal during heating pulse on and off (Fig.2) was measured at different 2,0 LDMOS: P=16W for 100µ s Intensity signal (a.u.) 1,5 9 4 8 1,0 0,5 7 position 3 Heating Cooling 0,0 0 100 200 300 400 Time (µ s) positions (see Fig.1b) of the laser beam inside and outside the well. The transient temperature increase within the device was calculated from the measured intensity signal using the properties of the F-P reflectivity-temperature function (a result from a calculation is shown in the inset of Fig.3). T(t)=5.77 x (t-6.97x10-3 t2) 200 2,0 Fig.3: Intensity signal measured on LDMOSFET and the corresponding temperature function Temperature increase (K) evaluated from the experiment. The inset shows the calculated Fabry-Perot reflectivity as function of temperature for the SiO2/Si/SiO2-multilayer inherent in LDMOSFET.160 Intensity signal (a.u.) 1,5 140 LDMOS: P=16W for 100µs 120 Temperature increase (K) 0,8 120 position 1 1,0 100 0,6 simulation pos.1 80 Reflectivity position 7 80 0,4 position 9 0,5 60 0,2 40 40 0,0 0 50 100 150 200 250 300 350 400 Temperature increase (K) 0,0 20 0 0 10 20 30 40 50 60 0 Time (µ s) 0 100 200 300 400 Time (µ s) Fig.4: Temperature increase extracted from measured intensity signals in Fig.2. The time evolution of the temperature at position 1 is compared to results obtained from finite element simulation. The time evolution of the temperature at different measurement positions (Fig.4) is then found from the corresponding intensity functions. The temperature increase obtained from optical measurements is in good agreement with the simulation results (Fig.4). The decay of the temperature after pulse turn-off is governed by the thermal time constant of the system, which is found from measurements to be about 70µs, in good agreement with simulation. Maximum temperature increase (K) Maximum temperature increase (K) 160 160 2 3 LDMOS: P=16W for 100µs 140 4 120 measurements 120 simulation position 1 5 80 position 1 100 7 80 40 8 9 10 6 60 0 LDMOS: P=16W for 100µ s heating pulse 40 0 100 200 300 400 500 -50 0 50 100 150 200 250 Distance from device center (µ m) Lateral distance from trench oxide (µm) Fig.5: Peak temperature measured at the end of the Fig.6: Peak temperature measured and simulated heating pulse (at time 100µs) at different positions at different lateral positions (see Fig.1b) outside along the device length (Fig.1b). The laser beam is the well. located at the center of the drift region. From spatial measurements along the device length we found an almost homogeneous distribution of the peak temperature in the well (Fig.5). A significant decrease in the well temperature is only measured in the vicinity of the trench oxide, as there is no heat generation at the device ends and the heat is also removed across the trenches. Optical measurements outside the well for a typical pulse width of 100µs (Fig.6) show that the maximum temperature at a distance 45µm and 90µm far from the trench oxide is reduced by a factor of 5.5 and 10, respectively, in comparison with the temperature inside the well. 4. Conclusions Time domain reflectivity measurements provide a simple, contactless and non-invasive (no sample preparation is needed) method to study transient heating effects in thick-film SOI smart power devices. The thermal time constant and the temperature distribution inside and outside the trench-isolated device obtained from the experiments are in a good agreement with the results of numerical simulation. Acknowledgements This work is financially supported by the Austrian Federal Ministry of Science and the Society for Microelectronics, Austria. References 1. W. P. Maszara, J. Electrochem. Soc. (138), 341 (1991). 2. M. B. Kleiner, S. A. Kühn and W. Weber, IEEE Trans. Electron Devices (43), 1602 (1997). 3. J. Korec, Mat. Sc. and Eng. B29, 1 (1995). 4. M. Stoisiek, K.-G. Oppermann, U. Schwalke and D. Takacs, in Proc. ISPSD´95, 325 (1995). 5. H. Icenogle, B. Platt and W. Wolfe, Applied Optics (10), 2348 (1976). 6. N. Seliger, P. Habaš and E. Gornik, in Proc. of ESSDERC´96, 847 (1996). 7. M. Born and E. Wolf, „Principles of Optics“, Pergamon Press, New York (1985). 8. ANSYS 5.0, „User’s Manual“, Swanson Analysis Systems, Inc. (1992).
Pages to are hidden for
"A Study of Temperature Distribution in SOI- Smart Power Devices in"Please download to view full document