May_EDFA_Digital

edfas.org 13 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 21 NO. 2 Most studies have been conducted in con- tinuous wavemode. However, microwave power GaN-based amplifiers, such as those used in radar or communication applications, are com- monly operated inpulsedmode. As a result, time- resolved Raman thermography, with a temporal resolution of ∼ 10 ns, was used 10 years ago to study the thermal dynamics of GaN HEMTs. [17] This technique was used to probe the thermal diffusion into the substrate and laterally within the device. In the same way, time-resolved UV micro-Raman spectroscopy has beenused for the temperature measurement of AlGaN/GaN HEMT devices, specifically in the active area. [18] This technique provides a temporal resolution of 200 ns and demonstrates two self- heating stages: First, a fast and high temperature rise due only to active heating under polarization and second, a slower andweaker temperature rise that looks like a heat diffusion in the whole device. [18] ORIGINAL RESULTS FROM GREYC LABORATORY Device technology is in constant evolution. High frequency applications require shorter source-gate and gate-drain distances. Moreover, air bridges and field-plate technologies are often used for power applications. In addition, the metal-to-semiconductor surface ratio is increasingly important. Consequently, it is becomingmore difficult to probe the semiconductors on the front surface by Raman spectroscopy. In response, this article describes the development of a new technique using ceriumoxidemicroparticles depos- ited everywhere on the device surface. This means the particles are placed on the semiconductor and also on the metal surfaces (i.e., the source, drain, and gate surfaces), as shown in Fig. 5. This research verifies that the particles do not modify the electric characteristics of the device. Fig. 5 Location of micro-Raman thermometers on a GaN HEMT surface. Fig. 6 Raman spectrum of cerium oxide deposited on the surface of a GaN HEMT. Fig. 7 Impact of power dissipation on self-heating tem- perature of a GaN HEMT, estimated directly from the GaN layer andmicro-Raman thermometers deposited on GaN and metal surfaces. Instead, they act as micro-Raman thermometers permit- ting temperaturemeasurement on the semiconductor and metal surfaces. They are independent of the mechanical strain eventually present in the heterostructure, which can skew the temperature estimation. Figure 6 presents a typical Raman spectrum obtained in these conditions showing the Raman lines of the CeO 2 microparticles and the GaN semiconductor. In this example, it is possible to determine not only the temperature of the GaN surface using themicroparticles, but also theGaNvolumedepend- ing on the selected wavelength and directly from the GaN lines. This requires a temperature calibration for the CeO 2 microparticles, which was performed for the Stokes Raman shift and the linewidth. Both are totally linear between 300 and 500 K. Figure 7 shows the impact of power dissipation on the self-heating temperature of GaN HEMT estimated directly from the GaN layer (that is, in volume) and from the surface using micro-Raman thermometers deposited on the GaN and metal contacts. The highest temperature is seen on the GaN surface located between the gate and