Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
6 Device Testing for High-Temperature Electronic Materials This chapter concentrates on devices made from silicon and silicon carbide, as these are the two materials systems for which the most data are available. The data are encouraging and suggest that electronic systems based on either of these two materials will operate successfully at elevated temperatures. There are three areas of testing that are discussed in this chapter: (1) short-term, constant-temperature tests, (2) constant-temperature life tests, and (3) thermal-cycling tests. The short-term and constant-temperature tests are encouraging. There is no known thermal-cycling data, however, and more research is required. A few comments at the conclusion of the chapter are devoted to the subjects of packaging and future testing SHORT-TERM CONSTANT-TEMPER\TURE TESTS Short-term constant-temperature tests are the first type done on each device intended for use at elevated temperatures and thus the one test for which there is the most data. In a typical test of this kind, a device or integrated circuit is placed in a test fixture that can be varied in temperature. Often, the device under test is also in a special, controlled environment, such as dry nitrogen or vacuum. Figures 6-1 and 6-2 are examples of two such tests. A representative set of conditions, with some associated comments, Is given in Table 6-1. Figure 6-1 illustrates the variation of threshold voltage with temperature for state- of-the-art silicon MOSFETs. Both n- and p-channe! devices change from enhancement mode to depletion mode at about 350 °C. Figure 6-2 illustrates the drain characteristics of a SiC MOSFET at 650 °C. The 61 ~ n 0.8 0.6 S o.4 - 0 ~ 0.2 co o > o o it, -0.2 o4 -0.6 -0.8 -1 .o Is c ~v .c cl 2 e-Channel Devices _~ - - 1 1 ~1 1 1 1 1 ~I I I I I I 0 50 100 150 200 250 300 350 400 Temperature (°C) FIGURE 6-1 Variations in threshold voltage for p- and e-type silicon MOSFETs with temperature. SOURCE: Grzybowski and Tyson (1993). T = 923 K _~ 9V 6V 3V OV 1 1 1 ~ 0 5 10 15 20 25 Drain Voltage (V) FIGURE 6-2 Drain characteristics of a SiC inversion-mode MOSFET at 650 °C. SOURCE: Palmour et al. (1991).
Materials for High-Temperature Semiconductor Devices TABLE 6-1 Short-Term Constant-Temperature Tests Temp. Attachment Device Type (°C) Comments Substrate Method Reference SiC 25-650, Gain increases BN plate Not reported Palmour et al., 1991 transistors unpacked with temp. SiC 25-350, Prototypes Not reported Not attached Palmour, 1993 transistors packaged available (wafer probe) SiC diode 25-550 Very low Not reported Not reported Ghezzo et al., 1992 leakage currents SiC pressure 25-400 Piezo-resistive None Not reported Shor et al., 1994 sensors bridge devices SiC 25-350 (as a Small gain Alumina or Braze Tomana et al., 1993 amplifiers hybrid IC) change with glass ceramic temp. Silicon transistors -65-450 (both FETs and bipolar) Gain decreases Alumina Silver glass with temp. Grzybowski, 1991; Grzybowski and Tyson, 1993 characteristics are normal for MOSFETs, with low leakage and relatively constant gain. These results clearly indicate that devices in both materials systems have been made that operate at temperatures well above the current limits of fielded electronic systems, typically about 125° C. The results are also consistent with the fundamental limits of device operation, based on the characteristics of each material of energy gap, effective mass, minority carrier lifetime, and leakage currents. CONSTANT-TEMPERATURE L1~E TESTS Constant-temperature life tests are also encouraging (Table 6-2), although much work remains to be done in this area. The times on tests at the temperatures indicated are too small to be conclusive, however. Effects such as metal electromigration, impurity diffusion in the semiconductor, and phase changes occurring in contact regions, have not yet been adequately tested. A possible exception is the last result, which was based on silicon integrated circuits with conventional metallization and feature sizes that are large (about 7.5 micrometers) by current standards. This result suggests that silicon-based 62 circuits with conventional metallization may be useful to temperatures well above the current limit of 125 °C. THERMAL-CYCLING TESTS Thermal-cycling tests logically follow after the short- term and constant-temperature tests. Thermal-cycling tests are absolutely essential for evaluating suitability of devices intended for use at elevated temperatures. It is well known, and is a major limitation in conventional temperature electronics, that thermal cycling results in failures that are not discovered by constant-temperature tests. One class of these failures is associated with the work-hardening of parts of the device, its metallization, the package attachment, and the package itself. In flip- chip technology, for example, the work-hardening of the solder bumps that are used to hold the device in place and form the interconnects is a well-known limitation. In fact, a substantial amount of development has been devoted to minimizing this problem. For elevated-temperature electronics, these mechanisms will almost certainly be more of a limitation, since the temperature swings will be larger.
Device Testing for High-Temperature Electronic Materials TABLE 6-2 Constant-Temperature Life Tests Time Temp. Attachment Device Type (hrs) (°C) Substrate Method Reference SiC diode rectifiers1,000 350 Hermetic glass Not reported Edmond et al., 1991 packages Silicon MOSFETS1,000 200 Alumina AlSi eutectic Palmer and Heckman, 1978 Silicon bipolar quad500 300 Not reported Not reported Beasom and op-amp Patterson, 1982 Silicon-based ring4,000 250 Not reported Not reported Migitaka and oscillators Kurachi, 1994 These limitations may demand new and innovative techniques for device attachment and connection, such as the use of graded attachment techniques or compliant attachment methods, in which the differences in thermal expansion are taken up by a relatively flexible part of the attachment. The use of different attachment materials may also be useful. For example, the use of amalgams has been explored on a preliminary basis. The use of these materials would allow one to make the rigid attachment at a temperature that is intermediate between the extremes, thus minimizing the stress that occurs at either extreme. FUTURE REQUIREMENTS FOR HIGH TEMPERAlllRE TESTING The testing of devices, circuits, and systems intended for high-temperature operation is more difficult than testing for lower-temperature situations. For these lower- temperature applications, the concepts of step-stress testing and accelerated aging are established. In these two approaches, the device under test is subjected to increasingly higher temperatures and the failure rates noted. In a well-behaved test, the resulting failure rates 63 will allow the calculation of an activation energy, which will in turn allow the prediction of failure rates at lower temperatures. The situation, however, is less clear for higher- temperature devices. For example, the mere testing at higher temperatures is a challenge, with a lack of equipment available for these higher-temperature tests. Also, a key assumption for accelerated testing is that the mechanisms of failure are the same for the accelerated test as for the application. This assumption has not been shown to be valid, and may not be true for many tests. For example, it is likely that the material used to mount the devices to the substrates would melt in accelerated testing, thus introducing a new failure mechanism and invalidating the test. A possible solution to this problem is the continuous- variables testing method. This method involves the measurement of parameters with very high resolution. In these tests, the device conditions are similar to those of the intended application. By careful monitoring of the device parameters, failure mechanisms can be detected at early stages. Use of this technique eliminates the problems associated with step-stress and accelerated testing, in which new failure modes may be introduced.
