Thermal Conductivity Measurements of GaN and AlN Group-III nitride semiconductors (e.g. GaN, AlN, InN) have attracted considerable attention in recent years for their utilization in high power optoelectronic and electronic devices [14]. These high power/high temperature uses make thermal behavior, particularly the thermal conductivity (k), an extremely important material property. The material s ability to dissipate heat directly affects its device performances. The thermal conductivity is a function of both intrinsic and extrinsic effects [5]. The latter are related to thermal wave (i.e. phonon) scattering by generic crystal defects (dislocations, impurities, process-induced damage, etc.), while the former is due to anharmonic thermal wave-thermal wave scattering. For materials with free carrier concentrations #1019 cm-3 the free carrier contribution to k is negligible. Thus k provides a measure of the material s quality. Therefore, the thermal conductivity is important from both applied (device modeling, sample quality) and fundamental (basic mechanisms of thermal wave propagation) perspectives. However, despite the considerable body of work, both experimental and theoretical, on the electronic, optical, and structural properties of the group-III nitride semiconductors, relatively few investigations have been reported on thermal conductivity measurements [616] or theory [7,1719]. Thermal Conductivity Measurement Methods There are a number of ways to evaluate k including steady-state longitudinal heat flow [5], modified Angstrom s method [5], optical pump-probe [20], laser flash [13], third-harmonic electrical method [9], scanning thermal microscopy (SThM) [8,1012, 1416], etc. With the exception of SThM most of these approaches require either contacts (destructive) and/or thick samples (>100 m), and do not have high spatial resolution. On the other hand, SThM is essentially nondestructive, and has a micron range spatial/depth resolution. The latter is important for examining material fabricated by new, low defect density techniques such as lateral epitaxial overgrowth (LEO) and pendeo-epitaxy, in addition to mapping spatial variations of k across a wafer. In SThM [21] the probe consists of a "V" shaped (radius of curvature 1 m) resistive element (Pt/10%Rh wire) incorporated at the end of a flexible cantilever. The probe tip is positioned using an atomic force microscopy (AFM) arrangement (laser, mirrors, and detector) as shown in a. The resistive element forms one leg (Rprobe) of a Wheatstone bridge (Figure b). There is a feedback loop, which adjusts the bridge voltage in order to keep the bridge balanced thus maintaining a constant probe temperature. The SThM tip is operated at a preset value 4050C above the specimen temperature. When the heated probe contacts the sample heat flows from the probe to the material (c), which tends to cool the tip. The feedback senses this shift and increases the voltage, thus returning the tip s resistance (temperature) to its preset point. The raw data from the SThM instrumentation provides only the voltage applied to the bridge (Uout). However, k of the specimen is proportional to the heat flow (or Uout2) when the tip is making contact. A simple calibration method was developed and implemented for absolute k measurements [10,22]. The spatial/ depth resolution of this technique is estimated to be in the 23 m range for GaN and AlN. GaN and AlN Thermal Conductivity Results: Overview 1. GaN Material Sichel and Pankove [6] performed the first measurement of k of "bulk" GaN (400 m of material fabricated by HVPE) as a function of temperature (25K < T < 360K) with a room temperature k 1.3 W/cm.K (along the c-axis), a value less than Slack s estimation of 1.7 W/cm.K [17]. More recently, Vaudo et al. [13] reported room temperature values of 2.20.2 W/cm.K on 200500 m thick free-standing HVPE GaN using the laser flash method. The Brooklyn College group, in collaboration with Emcore, measured the room temperature k on partially and fully coalesced LEO GaN/sapphire (0001) using SThM [12]. A correlation between low/high threading dislocation density and high/low k was established. High k, comparable to the values subsequently reported by Vaudo et al. [13], were found on the overgrown regions of these sam ples. The solid squares in show the SThM profile and the solid circles depict an AFM profile on two neighboring stripes of a LEO GaN/sapphire (0001) sample. The overgrown and over-the-mask regions are also indicated on the AFM data. The SThM shape reveals high and low thermal conductivity regions. A direct correlation between the thermal conductivity values and the details of the overgrown patterning is established by plotting the SThM and AFM profiles on the same length scale. High k readings, in the 2.002.10 W/cm.K range, correspond to the upper portion of the overgrown regions (low defect density) while the low k values are found in the over-the-mask sections (high defect density). Using a third-harmonic electrical technique Luo et al. determined a room temperature thermal conductivity on LEO material of at least 1.55 W/cm.K [9]. The thermal conductivity in HVPE-grown n-GaN/sapphire (0001) has been correlated with Si doping levels and sample thickness using SThM by the Brooklyn College group in conjunction with R.J. Molnar of MIT Lincoln Laboratories [10,11]. Plotted in are k values as a function of doping level in two sets of HVPE n-GaN. In both groups the thermal conductivity decreased linearly with log n, the variation being about a factor of two decrease in k for every decade increase in n (in the measured range). Reference 11 has also reported the variation of k (within 23 m of the surface) with film thickness. There is a general trend of increasing thermal conductivity with thickness, consistent with the observed general reduction of both extended dislocations and point defects with film thickness. The thermal conductivity behavior before and after plasma-induced effects was studied on a series of four HVPE n-GaN/sapphire (0001) samples by SThM in a joint project between Brooklyn College investigators and the Adesida group at the University of Illinois in Champaign-Urbana [14]. Before treatment k was measured to be in the 1.701.75 W/cm.K range, similar to that of HVPE GaN material with comparable carrier concentration and thickness [10,11]. The solid squares in (a) show k for the four samples measured before processing. The samples were processed under constant Ar gas flow and pressure for a fixed time period, the only variable processing parameter being the DC bias voltage (125500 V). At 125 V the thermal conductivity was only slightly less (k 1.65 W/cm.K) compared to the untreated case. After the initial 125 V procedure k exhibited a linear decrease with the DC voltage in the investigated range, with values 0.3 W/cm.K for the 500 V situation [14]. Shown in (b) is k measured after each individual processing procedure as a function of the processing voltage. The observed linear decrease of k at room temperature with increasing DC bias voltage could be attributed to the scattering of the thermal waves associated with additional "defects" being induced by the bias voltage. More defects are being generated with increasing bias voltages. Using SThM the topographical variations in k were correlated with plan-view transmission electron microscopy (TEM) and atomic force microscopy (AFM) imaging on a series of HVPE n-GaN samples [15]. Relatively low values of k (1.2 W/cm.K) were found in the regions near the edge dislocations, as identified by both TEM and AFM. On the other hand relatively high k regions ( 1.45 W/cm.K) were found in the sections between these defects. A theoretical prediction for k for GaN at room temperature was initially estimated by Slack to be 1.7 W/cm.K [17]. A simple expression to calculate the thermal conductivity of perfect dielectric crystals at "high temperatures" was derived by Witek [18]. Using this approximation he estimated the room temperature thermal conductivity of GaN to be 4.10 W/cm.K. Subsequently Berman confirmed this value by evaluating the 300K thermal resistivity of GaN [19]. 2. AlN Material Slack et al. [7] measured the thermal conductivity of "single crystals" of AlN (340 ppm by weight oxygen) as a function of temperature (0.4 K < T < 1800 K) and found k 2.85 W/cm.K at 300 K, less than his own estimate of 3.19 W/cm.K for pure AlN (no oxygen). Using SThM, Florescu et al, in collaboration with TDI, Inc. (Gaithersburg, MD), determined the 300 K thermal conductivity on free standing AlN samples (thicknesses in the range of 300800 m) grown by HVPE on Si(111) substrates [16]. After growth the Si substrates were removed by chemical etching. The "as-grown surfaces" exhibited triangular patterns of about 10 m on a side, probably due to misfit dislocations. At the center of these triangular portions the measured k was in the range of 3.03.3 W/cm.K , higher than the 2.85 W/cm.K reported by Slack [7]. Slack estimated 300 K values of approximately 3.19 W/cm.K for the pure (no oxygen) AlN thermal conductivity [7]. More recently, Witek derived the room temperature thermal conductivity value for AlN finding a maximum value of 5.9 W/cm.K [18]. Summary This article has presented an overview of the experimental thermal conductivity (mainly at room temperature) of "bulk-like" and thin film GaN and "bulk-like" AlN as measured by a variety of techniques including nondestructive, high spatial/depth resolution SThM. The highest reported values were (a) for GaN k 2.1 W/cm.K found on both the overgrown regions of LEO films and free standing thick HVPE samples while (b) for AlN k 3.3 W/cm.K on free standing material. In GaN the influence on k of dislocation density, n-doping, film thickness, and process-induced effects has been evaluated. Acknowledgement This work was supported by Office of Naval Research contract #N00014-99-C-0663 monitored by Dr. Colin Wood and the New York Office of Science, Technology, and Academic Research through its Centers for Advanced Technology program. References [1] See, for example, S. Nakamura, MRS Bull. 22, 29 (1997); also, M. Shur and M. Khan, MRS Bull. 22, 44 (1997) [2] See, for example, Group III Nitride Semiconductor Compounds, ed. B. Gil (Claredon, Oxford, 1998) [3] J. Zolper et al., Compound Semiconductor 5(5), 29 (1999) [4] R. Sharp, Compound Semiconductor 5(5), 26 (1999) [5] See, for example, Thermal Conduction in Semiconductors by C. Bhandari and D. Rowe (Wiley, New York, 1988) [6] E. Sichel and J. Pankove, J. Phys. Chem. Solids 38, 330 (1977) [7] G. Slack et al., J. Phys. Chem. Solids 48, 641 (1987) [8] V. Asnin et al., Appl. Phys. Lett. 75, 1240 (1999) [9] C. Luo et al., Appl. Phys. Lett. 75, 4151 (1999) [10] D. Florescu et al., Mat. Res. Soc. Symp. Proc. 595, W3.89.1 (2000) [11] D. Florescu et al., J. Appl. Phys. 88, 3295 (2000) [12] D. Florescu et al., Appl. Phys. Lett. 77, 1464 (2000) [13] R. Vaudo et al., presented at the Int. Workshop on Nitride Semiconductors, Nagoya, Japan 2000 [14] D. Florescu et al., to be published in Mat. Res. Soc. Symp. Proc. 2000, Symposium G [15] D. Florescu et al., presented at the 2000 Int. Conf. Comp. Semicond., Monterey [16] D. Florescu et al., private communication [17] G. Slack, J. Phys. Chem. Solids 34, 321 (1973) [18] A. Witek, Diamond and Related Materials 7, 962 (1998) [19] R. Berman, Diamond and Related Materials 8, 2016 (1999) [20] W. Capinski et al., Appl. Phys. Lett. 71, 2109 (1997) [21] ThermoMicroscopes/Park Instruments; 1171 Borregas Avenue; Sunnyvale, CA 94089 [22] F. Ruiz et al., Appl. Phys. Lett. 73, 1802 (1998)