Extended Exposure of Gallium Nitride Heterostructure Devices to a Simulated Venus Environment Savannah R

Home , University of Arkansas

Extended Exposure of Gallium Nitride Heterostructure Devices to a Simulated Venus Environment Savannah R. Eisner Hannah S. Alpert Caitlin A. Chapin Stanford University Stanford University Stanford University 496 Lomita Mall 496 Lomita Mall 496 Lomita Mall Stanford, CA 94305 Stanford, CA 94305 Stanford, CA 94305 [email protected] [email protected] [email protected]

Ananth Saran Yalamarthy Peter F. Satterthwaite Ardalan Nasiri Stanford University Stanford University University of Arkansas 496 Lomita Mall 496 Lomita Mall 700 Research Center Blvd. Stanford, CA 94305 Stanford, CA 94305 Fayetteville, AR 72701 [email protected] [email protected] [email protected]

Sara Port Simon Ang Debbie G. Senesky University of Arkansas University of Arkansas Stanford University 332 Arkansas Ave 700 Research Center Blvd. 496 Lomita Mall Fayetteville, AR 72701 Fayetteville, AR 72701 Stanford, CA 94305 [email protected] [email protected] [email protected]

AlGaN/GaN heterostructure is suitable for robust, Venus- Abstract—Further development of harsh environment capable electronics. electronics capable of uncooled operation under Venus surface atmospheric conditions (~460°C, ~92 bar, corrosive) would enable future missions to the surface of Venus to operate for up to a year. Wide band-gap gallium nitride (GaN) heterostructure TABLE OF CONTENTS devices are attractive candidates for Venus lander missions due to their ability to withstand high-temperature exposure. Here, 1. INTRODUCTION ...... 1 we present the first assessment of the electrical integrity of GaN- 2. DEVICE FABRICATION AND OPERATION ...... 1 based devices subject to Venus surface atmospheric conditions. Three unique device architectures were fabricated at the 3. EXPERIMENTAL METHODS ...... 4 Stanford Nanofabrication Facility and exposed in a Venus 4. RESULTS AND DISCUSSION ...... 4 simulation chamber for 244 hours at the University of Arkansas 5. CONCLUSIONS ...... 8 Center for Space and Planetary Sciences. The three device architectures tested were InAlN/GaN high electron mobility ACKNOWLEDGEMENTS ...... ERROR! BOOKMARK NOT transistors (HEMTs), InAlN/GaN Hall-effect sensors, and DEFINED. AlGaN/GaN UV photodetectors, which all have potential REFERENCES ...... ERROR! BOOKMARK NOT DEFINED. applications in the collection and readout of sensor data from BIOGRAPHY ...... 11 Venusian landers. After exposure, HEMT threshold voltage had shifted only ~1% and gate leakage current remained on the same order of magnitude, demonstrating stability of the IrOx gate under supercritical CO2 ambient. Fluctuations in drain 1. INTRODUCTION current after exposure are attributed to thermal detrapping and electrically-activated trapping processes. Measurements of the The longevity and scope of proposed missions to the surface InAlN/GaN 2DEG properties in virgin and exposed Hall-effect of Venus are currently limited by the challenge of developing sensors were comparable. Furthermore, the Hall-effect sensors electronics that can survive the ~460°C, ~92 bar corrosive exhibited a maximum change of only +11.4% in current-scaled sensitivity and -6.6% in voltage-scaled sensitivity post-exposure. environment [1], [2]. Despite the use of cooling measures, The UV photodetectors with 362 nm peak wavelength exhibited data transmission from previous Venus landers lasted only an average decrease in responsivity of 38% after exposure, two hours due to failure of the silicon electronics [2]. Long which is thought to be due to strain relaxation or ohmic contact term missions (beyond 2 months) to the surface of Venus are degradation. Similar performance of the InAlN/GaN HEMTs needed to make seismic observations, determine detailed and Hall-effect sensors before and after exposure highlights the mineralogy, and characterize atmosphere-surface viability of this material platform for development of Venus interactions over an extended duration. This data can surface electronics, while the decrease in AlGaN/GaN UV illuminate the origin and diversity of terrestrial bodies (e.g. photocurrent requires further analysis to assess whether the how Venus and Earth diverged in climate and geology), as

978-1-7281-7436-5/21/$31.00 ©2021 IEEE well as the factors that determine the evolution of life on atmospheric conditions has been reported. Here, we describe some planets and not others. Moreover, a surface study of the the first demonstration of the stability of GaN heterostructure runaway greenhouse gas effect on Venus can increase our devices after 244-hour (>10 days) exposure to a simulated understanding of the processes that control climate on Earth- Venus surface environment. like planets [1], [3], [4]. The three distinct GaN heterostructure device types Wide band-gap semiconductors such as silicon carbide (SiC) characterized in this exposure study were chosen due to their and gallium nitride (GaN) have recently emerged as potential sensing and telecommunication applications on promising material platforms for uncooled extreme board a Venus lander. GaN HEMT devices can be used for environment electronics due to their stability at high pressure, chemical, and IR sensing [19]–[23]. Additionally, temperatures, inherent radiation tolerance, and chemical HEMTs have applications in read-out and transmission of resistance [5]–[7]. Furthermore, GaN heterostructures have sensor data to orbiters. Hall-effect sensors are often used for the unique ability to form a polarization-induced two- detecting the position and speed of rotating components (e.g., dimensional electron gas (2DEG) at the interface between gears, motors) aboard a spacecraft, as well as monitoring the GaN and a III-nitride alloy (e.g. AlGaN or InAlN). Unlike reliability of other on-board electronics via non-invasive doped junctions, the 2DEG channel is largely unaffected by current sensing. Ultraviolet (UV) photodetectors have adverse dopant scattering and diffusion effects at elevated applications in radiative heating characterization during temperatures. Previous studies on the high-temperature atmospheric entry. This data can be used to improve thermal operation of GaN heterostructure devices in inert and air protection systems (TPS) on future Venus surface landers. ambient are promising [8]–[16]. InAlN/GaN high electron Furthermore, these three GaN heterostructure device mobility transistors (HEMTs) have been reported to operate architectures have compatible fabrication flows, which for 25 hours at 1000°C in vacuum [16]. would allow for monolithic integration of uncooled sensors with signal readout and transmission. Future maturation of Even with these recent breakthroughs, there have been few this technology could substantially lower mission weight and studies of the effects of long-term exposure to a Venus cost by reducing packaging and shielding requirements. surface environment, where the atmosphere is ~97% supercritical CO2, on wide band-gap electronics [1], [17]. In 2. DEVICE FABRICATION AND OPERATION the most rigorous experiment to date, the stable operation of SiC junction field effect transistor (JFET) integrated circuits The HEMT devices were fabricated at the Stanford (ICs) under Venus surface atmospheric conditions was Nanofabrication Facility on an indium aluminum nitride demonstrated for 60 days [18]. The devices were directly (In0.18Al0.82N)/GaN-on-silicon wafer grown via metal- exposed (e.g. no packaging or cooling) inside the NASA organic chemical vapor deposition (MOCVD) and supplied Glenn Extreme Environment Rig (GEER) and tested in situ by NTT Advanced Technology Corporation. A cross- over the duration. The ICs experienced a small increase in sectional diagram of the device and heterostructure is shown frequency, primarily over the first 20 days of exposure, which in Figure 1(a). Starting from the Si (111) wafer, the stack the authors attributed to burn-in. No comparable assessment consisted of a 300-nm-thick buffer layer, 1-μm-thick GaN of GaN electronics operating under Venus surface layer, 0.8-nm-thick AlN spacer, and a 10-nm-thick InAlN

Figure 1. Cross-sectional schematic of the (a) InAlN/GaN HEMTs, (b) InAlN/GaN Hall-effect sensors, and (c) AlGaN/GaN UV photodetectors exposed to the Venus-simulant environment. Interfaces where the 2DEG is present are represented by white dashed line. 2 barrier layer. HEMT active regions were isolated through a readout. The sensitivity of a Hall-effect device with respect BCl3/Cl2 inductively coupled plasma (ICP) reactive ion etch to supply current is referred to as the current-scaled (RIE). Ti/Al/Mo/Au (10/200/40/80 nm) ohmic contacts were sensitivity (Si) and is inversely proportional to sheet carrier electron beam evaporated, patterned via lift-off process, and concentration, rapid thermal annealed (RTA) at 850°C for 35 seconds in N2. Iridium (15 nm) was evaporated and patterned as the = = (2) Schottky contact, followed by Ti/Au (20/200 nm) bond metal 𝑉𝑉𝐻𝐻 𝑟𝑟𝑛𝑛 evaporation and patterning. The Ir gate metal was oxidized at 𝑆𝑆𝑖𝑖 𝐺𝐺𝐻𝐻 𝐼𝐼𝐼𝐼 𝑞𝑞𝑛𝑛𝑠𝑠 300°C for 5 minutes via RTA to form IrOx. The use of IrOx A reduction in the sheet carrier density of the 2DEG increases gates on GaN HEMTs has been reported to reduce leakage the sheet resistance of the device, leading to a drop in current currents compared to as-deposited Ir gates, and high- under constant voltage conditions, and thereby increasing the temperature operation has been demonstrated [26]–[28]. The current-scaled sensitivity. The sensitivity with respect to HEMT gate length was 5 µm, centered in a 20-µm-long by supply voltage, also known as the voltage-scaled sensitivity 100-µm-wide 2DEG channel. (Sv), is proportional to the electron mobility (μH) and is defined: The Hall-effect sensors were fabricated simultaneously on the same InAlN/GaN-on-Si wafer as the HEMT devices and = = = underwent the same isolation and ohmic metal evaporation, (3) 𝑉𝑉𝐻𝐻 𝑟𝑟𝑛𝑛𝐺𝐺𝐻𝐻 𝐺𝐺𝐻𝐻 lift-off, and anneal processes. The sensors were diced and 𝑆𝑆𝑣𝑣 𝜇𝜇𝐻𝐻𝑟𝑟𝑛𝑛 removed from the fabrication run prior to the high- 𝑉𝑉𝑠𝑠𝐵𝐵 𝑅𝑅𝑅𝑅𝑛𝑛𝑠𝑠 𝐿𝐿 � �𝑒𝑒𝑒𝑒𝑒𝑒 temperature oxidation anneal required for IrOx formation on 𝑊𝑊 the HEMTs. A subset of the Hall-effect die was passivated where Vs is supply voltage, R is the device resistance, and (L/W)eff is defined as the effective number of squares, or the with ~20 nm of aluminum oxide (Al2O3) deposited via atomic layer deposition (ALD). Vias were wet chemically etched to ratio of the internal resistance to the sheet resistance [30]. expose the underlying ohmic contacts. Both passivated and unpassivated devices were then patterned with Ti/Au The GaN heterostructure UV photodetectors were fabricated evaporated metal bond pads. A schematic of the passivated on an MOCVD-grown aluminum gallium nitride devices is shown in Figure 1(b). Hall-effect sensor mesas are (Al0.25Ga0.75N)/GaN-on-silicon wafer purchased from octagonal and 100 μm in width. Two of the sensors exposed DOWA, Inc. The cross-sectional device and heterostructure in the Venus chamber (one unpassivated, one passivated) schematic is shown in Figure 1(c). The epitaxial stack were regular octagons fabricated with contacts of equal consists of a 1.5 μm buffer structure on Si (111) followed by length to the non-contacted sides of the mesa. The length of a 1.2 μm GaN layer, 1-nm-thick AlN spacer, 30-nm-thick each side of these devices is 41.4 μm. One unpassivated AlGaN barrier layer and 1-nm-thick GaN cap. Similar to the irregular octagonal sensor with point-like contacts was also HEMT/Hall-effect process, the photodetector exposed. In this device, the length of sides without contacts microfabrication consisted of a BCl3/Cl2 mesa isolation etch is 5.66 times greater than the sides with contacts, which are to define the AlGaN fingers. The result is an array of 2DEG 11.1 μm long. The difference in Hall-effect device interdigitated transducers (IDTs). Ti/Al/Pt/Au ohmic geometries and corresponding effects on sensitivity are metallization (20/100/40/80 nm) was evaporated, patterned reported in detail in earlier work [29]. The sensitivity trends via lift-off, and annealed for 35 s at 850°C in N2 atmosphere. between the three devices in this experiment are consistent It should be noted that the photodetector devices utilize a Pt with the authors’ prior work showing devices with point-like barrier layer in the ohmic metallization stack, while the contacts have the highest current-related sensitivity and HEMT and Hall-effect sensors have a Mo barrier layer. devices with equal sides have the highest voltage-related The photodetectors exhibit high responsivity (R) and low sensitivity [29]. dark current at room temperature [31]. Here, we characterize the photocurrent of the device (I), which is directly The Hall voltage (VH), which is perpendicular to both the proportional to the responsivity, applied current (I) and the external magnetic field (B), is defined as: = (4) 𝐼𝐼 = (1) 𝑅𝑅 𝑃𝑃 𝐼𝐼𝐼𝐼𝑟𝑟𝑛𝑛𝐺𝐺𝐻𝐻 where P is the incident radiant energy. The photocurrent is 𝑉𝑉𝐻𝐻 𝑞𝑞𝑛𝑛𝑠𝑠 also proportional to the gain (G), which is an important figure where rn is the scattering factor of the material, GH is the of merit for photodetectors and is defined as the ratio of shape factor, q is the electronic charge, and ns is the sheet charge carriers to the photon flux: carrier density. In most applications, the ideal Hall-effect sensor would exhibit a high sensitivity with respect to supply = (5) current and supply voltage to maximize the signal for ℎ𝑐𝑐 𝐼𝐼 𝐺𝐺 3 𝑒𝑒𝑒𝑒 𝑃𝑃 where λ is the wavelength of incident radiation, h is Planck’s Venus surface atmospheric conditions were simulated within contact, and e is the electronic charge. a 500 mL capacity stainless steel pressure vessel (Supplier: Parr Instruments) at the University of Arkansas Center for Space and Planetary Science. The vessel, referred to as the 3. EXPERIMENTAL METHODS Venus simulation chamber throughout this paper, is equipped with an internal thermocouple and exterior heating sleeve. As The HEMTs were characterized in air on a Signatone Inc. the chamber is heated, chamber pressure increases probe station equipped with a heated chuck. Current-voltage correspondingly. In this experiment, the chamber was (ID-VDS, ID-VGS, IG-VGS) measurements were taken on a pumped down to 37 mbar at room temperature and flushed semiconductor parameter analyzer (Agilent Technologies with CO2 at ~4 bar to remove residual gas in the chamber. B1500A). Using this setup, the HEMT devices were More CO2 was introduced into the chamber until the desired characterized at room temperature before and after the Venus starting pressure, calculated using the ideal gas law, was simulation chamber exposure. Additionally, circular reached. The chamber was sealed and the heating sleeve was transmission line measurements (CTLM) of the ohmic set. The chamber took ~45 minutes to reach the temperature contacts were taken. To compare the effects of the high- setpoint of 460°C and shortly afterwards the pressure temperature supercritical CO2 environment with high- stabilized to ~96 bar. Under these conditions, the CO2 is in a temperature air exposure, a virgin 1 µm gate-length HEMT supercritical state. During the 244-hour run, the temperature was characterized from room temperature to 600°C in air. In and pressure were monitored and remained stable at 460°C ± this experiment, the heated chuck was ramped from 25°C to 0.5% and 96 bar ± 2%. 600°C in 100°C increments. The chuck was held at each temperature for 10 minutes before device data was acquired. After the run was complete, the heating sleeve was turned off After cooling back down to 25°C, the device was measured and the hot gas was vented. The chamber was pumped down again. to 30 mbar and flushed with N2 at ~4 bar to remove any residual CO2. The chamber was filled with ~41 psi N2 and the The Hall-effect sensors were wire-bonded to a PCB and samples were retrieved from inside the chamber once the tested within a tunable 3D Helmholtz coil in an experimental temperature reached 23°C. It should be noted that N2 and setup described elsewhere [32]. Hall voltage measurements trace constituents of the Venus atmosphere were not included were conducted at room temperature. Current was sent across in the chamber during the exposure test. the device between two opposing contacts via a sourcemeter ESULTS AND DISCUSSION (Keithley 2400) and the resulting Hall voltage was measured 4. R across the other set of contacts using a multimeter (Agilent HEMT—Reliable operation of InAlN/GaN HEMTs at high 34410A). The current-spinning technique described in [29] temperature is frequently limited not by the heterostructure was employed to reduce the offset voltage. A switching itself, but by premature failure of the contact metallization or matrix (U2715A) alternated measurement of the eight phases. passivation scheme. Reported high-temperature degradation The sensors and coil were contained within MuMetal® mechanisms associated with the contacts include gate magnetic field shielding cannisters. The Hall-effect sensors sinking, metal ball-up, gold diffusion, and electromigration were tested under a magnetic field of 2 mT and biased with [10], [12], [16], [34], [35]. Figure 2(a) and (b) show optical current ranging from 60 uA to 0.9 mA. Additionally, microscope images of the HEMT before and after exposure. measurements of mobility, sheet resistance, and 2DEG sheet After 10-day exposure in the Venus chamber, the ohmic density were taken of the samples that had been exposed to contacts and bond pads on the HEMT were drastically darker the Venus-analogue environment, as well as equivalent virgin (Fig. 2(b)). Color changes across the surface indicate lateral samples, using the Lakeshore 8404 Hall measurement tool. metal interdiffusion. The brown, purplish color on the gold bond pads is indicative of the formation of Au-Al Six UV photodetectors were illuminated with 365 nm UV light at the open-air probe station and the photocurrent was measured on the semiconductor parameter analyzer. The photocurrent measurements were taken at room temperature.

The GaN-on-silicon die were mounted onto an alumina substrate for loading into the chamber. In order to attach the die, a thick-film gold layer (3066/3068N wire-bondable Au Conductors, Ferro Corporation) was screen printed on 98% alumina substrate (CoorsTek Inc.). After the firing process,

8835 gold cermet conductor (ESL 8835 (520C) 9802A, (a) (b) Electro-science Laboratories) was screen printed on the Figure 2. Optical image of the microfabricated HEMT alumina substrate. Fabricated chips were attached to the (a) before and (b) after the 10-day simulated Venus printed gold paste followed by a 4-bar pressure to remove the exposure. air traps [33]. 4 intermetallic compounds. Au-Al compound formation parameters are reported in Table 1. Maximum drain (referred to as “purple plague”) can lead to voiding and saturation current (ID,sat) and maximum transconductance localized areas of high resistance across the contact. Despite (gm,max) both increased after Venus chamber exposure (Table the appearance of the bond pads, TLM measurements show 1). This behavior is attributed to thermal detrapping in the the specific contact resistivity remained on the same order of InAlN/GaN heterostructure. While trapping phenomenon is magnitude (2.2 × 10-5 Ω∙cm2 before exposure and 3.7 × 10-5 still not fully understood in InAlN/GaN, thermal treatment Ω∙cm2 after exposure). This indicates that the Mo layer was has been demonstrated to be effective for detrapping an effective Al-Au diffusion barrier in the ohmic purposes [35]. Moreover, it is possible that supercritical CO2, metallization region not overlapping with the bond pad. which can penetrate deeply into the GaN heterostructure, could be interacting with residual gases in the chamber to The DC output characteristics (ID−VDS) and transconductance anneal defects in the lattice [36], [37]. However, drain current (gm) of the 5 μm gate-length HEMT before and after exposure relaxation back towards pre-exposure values became more are shown in Figure 3(a) and (b), respectively. HEMT pronounced as the negative gate voltage was increased in the

(a) (b)

Figure 3. ID−VDS (a), transconductance (b), ID−VGS (c), and IG−VGS (d) of the 5 μm gate-length HEMT before and after exposure in the Venus chamber.

Table 1. HEMT electrical parameters before and after exposure in the Venus chamber. Before After Parameter Symbol Unit Exposure Exposure

Saturation Current ID,sat A/mm 0.20 0.26 -9 -9 Off current ID,off A/mm 3.92 × 10 5.35 × 10 7 7 On/Off Ratio ID,on/ID,off - 3.36 × 10 3.28 × 10

Threshold Voltage VTH V -3.18 -3.22

Max Transconductance gm,max mS/mm 65 84 Subthreshold Swing SS mV/dec 89 90

test routine post exposure (Fig. 3(a)). During ID−VDS measurements, VGS was stepped from 0 V to -3 V sequentially. The percent change from the pre-exposure drain current was +32% for VGS = 0 V, but only +25% for VGS = - 1 V. The trend of decreasing percent change in drain current continues with further negative applied gate voltage. This drain current relaxation can be explained by the accumulation of negative charge trapped on defects induced by gate voltage steps [38]. In other words, the Venus chamber exposure likely led to detrapping or defect annealing in the lattice, causing a brief increase in drain current that was counteracted by re- trapping as the device was electrically characterized post- exposure. In the future, characterization of the trap states in the InAlN/GaN HEMTs before and after exposure could be accomplished by deep level transient spectroscopy (DLTS).

Figure 3(c) and (d) illustrate ID−VGS and IG−VGS before and

Figure 4. Post-exposure Auger depth profile over the after exposure. The extracted off-current (ID,off) and on/off IrOx gate on the InAlN/GaN HEMT. ratio (ID,on/ID,off) before and after exposure are comparable and on the order of 10-9 A/mm and 107, respectively (Table 1). The gate leakage current (IG) of the HEMT remains on the (a) same order of magnitude after exposure (Fig. 3(c)), while the drain current exhibits the small increase explained above (Fig. 3(d)). The threshold voltage (VTH) of the HEMT is extrapolated in the linear regime (Table 1). VTH was reduced by only ~1.1% after exposure, further proof of the high thermal stability of the gate after subjection to Venus surface conditions. An Auger electron spectroscopy (AES) depth profile over the gate confirms that the composition remains IrOx and the gate has not sunk down into the underlying substrate (Fig. 4).

The transfer characteristics ID−VDS and ID−VGS of the virgin 1 μm gate-length HEMT characterized at high-temperature in air (as described under Experimental Methods) are displayed in Figure 5(a) and (b). When the temperature was increased (b) to 300°C and 400°C, measurements show the occurrence of current collapse, which is consistent with the presence of deep traps in the HEMT [38]. The decrease in drain current with increasing temperature is expected and due to increased carrier scattering at elevated temperatures. However, permanent degradation of HEMT performance upon cooling back down to 25°C is also observed. VTH is shifted by nearly +1 V and maximum drain saturation current is reduced substantially (Fig. 6(a) and (b)). The non-recoverability of HEMT characteristics after only short-term (10 minutes at each temperature point) high-temperature exposure is attributed to oxidation reactions in air. Instability of the IrOx gate in air is thought the be the cause of the degraded device performance because ungated InAlN/GaN Hall-effect sensors from the same fabrication run exhibited near full Figure 5. (a) ID−VDS (a) and (b) ID−VGS of the virgin 1 recoverability of 2DEG properties after multiple thermal μm gate-length HEMT characterized at high- cycles up to 576°C and back in air [15]. These data highlight

temperature in air. the need for careful consideration of the atmosphere (e.g. supercritical CO2 vs. air) of the intended high-temperature application of unpassivated HEMT devices.

Hall-effect Sensors—Table 2 summarizes the 2DEG properties of InAlN/GaN Hall-effect sensors post-exposure as well as the properties of equivalent virgin devices. The differences in mobility, sheet carrier density, and sheet between the virgin and exposed samples are within the tolerances of die-to-die variation and measurement setup limitations, suggesting that the 2DEG remained largely unaffected by prolonged exposure to Venus surface atmospheric conditions.

The current-scaled and voltage-scaled sensitivities of the three InAlN/GaN Hall-effect sensors before and after exposure are shown in Figure 6. Table 3 and Table 4 report their average current-related and voltage-related sensitivities and the corresponding percent change, correspondingly. The Figure 6. Voltage-related versus current-related current-scaled sensitivity increased in all three devices after sensitivity for the three Hall-effect sensors before and exposure, with the largest increase of 11.4% seen in the after exposure. unpassivated device with contacts equal in length to the non-

contacted sides. The increase in current-scaled sensitivity, Table 2. Virgin and post-exposure Hall-effect sensor which is inversely proportional to sheet carrier density (Eq. 2DEG properties. 2), indicates a small, permanent decrease in ns in both the passivated and unpassivated sensors after long-term exposure RSH μ ns Sample to the simulated Venus environment. (Ω/□) (cm2/V-1s-1) (cm-3) Unpassivated, The small fluctuations in voltage-scaled sensitivity before 272 1124 2.03 × 1013 Virgin and after exposure in both unpassivated Hall-effect sensors Unpassivated, (equal-sided device: -1.7%; point-like contacts: +2.2%) are 302 1076 1.92 × 1013 Post-Exposure within the measurement limitations of the test setup (Table 3). However, the decrease in voltage-scaled sensitivity of the Passivated, 252 1031 2.41 × 1013 passivated device is larger (-6.6%), consistent with a small Virgin decrease in the device mobility after exposure (Eq. 3). Passivated, Though the InAlN/GaN heterostructure is lattice matched 248 1026 2.45 × 1013 Post-Exposure and therefore not prone to mechanical strain, the addition of an Al2O3 passivation layer adds tensile stress to the system Table 3. Current-related sensitivities of the [40]. High-temperature strain relaxation likely accounts for Hall-effect sensors before and after exposure. the decrease in mobility due to increased dislocation Si,pre Si,post % change scattering, although further studies are needed to study the Device (V/V/T) (V/V/T) Si effects of supercritical CO2 on ALD Al2O3. Unpassivated, 24.1 26.9 +11.4 UV Photodetectors—Figure 7 shows the photocurrent of the Equal Sides AlGaN/GaN UV photodetectors before and after simulated Unpassivated, Venus chamber exposure. The photocurrent across the 34.9 37.6 +7.6 Point-like devices decreased by 38±2.9% on average after exposure (Fig. 7). As an experimental setup control measure, a virgin Passivated, 20.9 21.9 +5.0 photodetector die was tested simultaneously as the exposed Equal Sides die. The photocurrent of the four virgin photodetectors Table 4. Voltage-related sensitivities of the averaged only 1.47% less than the photocurrent of the six Hall-effect sensors before and after exposure. detectors prior to Venus exposure. This small difference in average photocurrent is due to die-to-die variation, indicating Sv, pre Sv, post % change Device that the decrease in photocurrent of the detectors after (V/A/T) (V/A/T) Sv exposure is indeed real and not an artifact of changes in the Unpassivated, test setup. 0.0552 0.0542 -1.7 Equal Sides The decrease in photocurrent post-exposure can be attributed Unpassivated, 0.0377 0.0385 +2.2 to strain relaxation in the AlGaN/GaN heterostructure. It is Point-like well established that mechanical stress in the AlGaN/GaN Passivated, structure significantly influences the 2DEG carrier 0.0521 0.0487 -6.6 concentration and mobility. High-temperature induced strain Equal Sides

relaxation can produce dislocations and cracks, which in turn 7

structures can determine if the specific contact resistivity is increased after extended Venus surface atmospheric condition exposure.

5. CONCLUSIONS

In this study, we have characterized three unique GaN heterostructure device architectures before and after 10 days of exposure to simulated Venus atmospheric conditions of 465°C, ~96 bar, and supercritical CO2 ambient. Increases in HEMT drain saturation current post-exposure are attributed to thermal detrapping, while subsequent drain relaxation under bias conditions is due to trapping. The stability of the threshold voltage, gate leakage current, and on/off ratio in the InAlN/Gan HEMTs make these devices excellent candidates Figure 7. Photocurrent reponse of the 6 UV for Venus lander missions. It was found that 2DEG properties photodetectors under 365 nm illumination before of InAlN/GaN Hall-effect sensors were not significantly and after Venus chamber exposure. altered after Venus chamber exposure compared to virgin devices. Furthermore, the Hall-effect sensors exhibited a maximum average shift of +11.4% in current-scaled sensitivity and -6.6% in voltage-scaled sensitivity, suggesting near full recovery of sensing performance after the Venus environment exposure. However, the AlGaN/GaN UV photodetectors suffered a large decrease in photocurrent of ~38% after the exposure, which is thought to be due to strain relaxation or degradation of the ohmic contact metallization. The stability of the InAlN/GaN HEMTs and Hall-effect sensors after 10-day exposure to a simulated Venus surface environment makes these devices attractive candidates for deployment on future Venus lander missions, as well as in other harsh environments such as CO2 sequestration and storage. Diminished photocurrent in the AlGaN/GaN UV

Figure 8. SEM image of UV photodetector after photodetector after simulated Venus exposure requires exposure in the Venus chamber. Contact with Al further analysis to determine if the strain-dependence of the wire bond bump is shown on the right. 2DEG properties precludes this class of devices from prolonged Venus surface operation. This study indicates reduce the 2DEG mobility via increased scattering [41]–[45]. Ti/Al/Mo/Au is a reliable ohmic contact to GaN-based Decrease in 2DEG sheet carrier density, and the subsequent devices for extreme environment operation, while increase in sheet resistance, can explain the decrease in Ti/Al/Pt/Au is questionable. photocurrent after exposure. Future work will examine the behavior of these devices in Another possible explanation for the decrease in photocurrent situ under Venus surface atmospheric conditions in the is degradation of the ohmic contact metallization. Scanning NASA GEER chamber, and the reliability will be quantified electron microscopy (SEM) images of the UV photodetectors through extended lifetime testing. We plan to increase the after exposure in the Venus chamber show a halo of diffused fidelity of the simulated Venus environment through metal surrounding the contacts and bleeding into the 2DEG inclusion of N2 and trace amounts of SO2 in the chamber. IDT region (Fig. 8). Lateral diffused metal could be acting as Finally, these devices will be subjected to radiation testing to a reflection coating over the IDTs, thereby reducing investigate their potential for use in other space applications. photocurrent. Alternatively, if vertical diffusion occurred as well as lateral diffusion, the migrated metal could be degrading the 2DEG mobility through impurity scattering. ACKNOWLEDGEMENTS Moreover, the SEM images reveal substantial morphological differences between the ohmic contacts with and without The authors would like to thank Dr. Karen Dowling for residual Al wire bond bumps. Changes to the surface insightful discussions. This work was supported in part by the morphology are consistent with a non-uniform contact- National Aeronautics and Space Agency through the High barrier layer interface, which can cause localized increased Operating Temperature Technology program under grant contact resistance and decrease the effective photocurrent number NNX17AG44G. Part of this work was performed at readout. Future characterization of Ti/Al/Pt/Au CTLM the Stanford Nano Shared Facilities (SNSF), supported by the 8

National Science Foundation under award ECCS-1542152. Al2O3/AlGaN/GaN MIS-HEMTs and HEMTs at 600 °C in air,” Semicond. Sci. Technol., vol. 31, no. 11, p. 115017, Oct. 2016, doi: 10.1088/0268- REFERENCES 1242/31/11/115017. [12] D. Maier et al., “Testing the Temperature Limits of [1] J. L. Hall et al., “Final Report of the Venus Science and GaN-Based HEMT Devices,” IEEE Trans. Device Technology Definition Team, Venus Flagship Mission Mater. Relib., vol. 10, no. 4, pp. 427–436, Dec. 2010, Study,” NASA Jet Propulsion Laboratory, Apr. 2009. doi: 10.1109/tdmr.2010.2072507. [Online]. Available: [13] H. Lu, P. Sandvik, A. Vertiatchikh, J. Tucker, and A. https://vfm.jpl.nasa.gov/files/Venus+Flagship+Mission Elasser, “High temperature Hall effect sensors based on +Study+090501-compressed.pdf AlGaN∕GaN heterojunctions,” Journal of Applied [2] J. Cutts et al., “Venus Technology Plan,” National Physics, vol. 99, no. 11, p. 114510, Jun. 2006, doi: Aeronautics and Space Administration, May. 2014. 10.1063/1.2201339. [Online]. Available: [14] M. De Vittorio et al., “High temperature https://www.lpi.usra.edu/vexag/reports/Venus- characterization of GaN-based photodetectors,” Sensors Technology-Plan-140617.pdf and Actuators A: Physical, vol. 113, no. 3, pp. 329–333, [3] R. Herrick et al., “Goals, Objectives, and Investigations Aug. 2004, doi: 10.1016/j.sna.2004.04.016. for Venus Exploration,” National Aeronautics and [15] H. S. Alpert et al., “Sensitivity of 2DEG-based Hall- Space Administration, May. 2014. [Online]. Available: effect sensors at high temperatures,” Review of https://www.lpi.usra.edu/vexag/reports/GOI- Scientific Instruments, vol. 91, no. 2, p. 025003, Feb. 140625.pdf 2020, doi: 10.1063/1.5139911. [4] J. O’Rourke et al., “Venus Goals, Objectives, and [16] D. Maier et al., “InAlN/GaN HEMTs for Operation in Investigations,” National Aeronautics and Space the 1000°C Regime: A First Experiment,” IEEE Administration, Sep. 2019. [Online]. Available: Electron Device Lett., vol. 33, no. 7, pp. 985–987, Jul. https://www.lpi.usra.edu/vexag/reports/VEXAG_Venus 2012, doi: 10.1109/led.2012.2196972. _GOI_Current.pdf [17] P. G. Neudeck, R. D. Meredith, L. Chen, D. J. Spry, L. [5] P. G. Neudeck, R. S. Okojie, and Liang-Yu Chen, M. Nakley, and G. W. Hunter, “Prolonged silicon “High-temperature electronics - a role for wide bandgap carbide integrated circuit operation in Venus surface semiconductors?,” Proc. IEEE, vol. 90, no. 6, pp. 1065– atmospheric conditions,” AIP Advances, vol. 6, no. 12, 1076, Jun. 2002, doi: 10.1109/jproc.2002.1021571. p. 125119, Dec. 2016, doi: 10.1063/1.4973429. [6] P. Hazdra and S. Popelka, “Radiation resistance of [18] P. G. Neudeck et al., “Operational Testing of 4H-SiC wide-bandgap semiconductor power transistors,” Phys. JFET ICs for 60 Days Directly Exposed to Venus Status Solidi A, vol. 214, no. 4, p. 1600447, Oct. 2016, Surface Atmospheric Conditions,” IEEE J. Electron doi: 10.1002/pssa.201600447. Devices Soc., vol. 7, pp. 100–110, 2019, doi: [7] S. J. Pearton, E. A. Douglas, R. J. Shul, and F. Ren, 10.1109/jeds.2018.2882693. “Plasma etching of wide bandgap and ultrawide [19] D. Gajula, I. Jahangir, and G. Koley, “High bandgap semiconductors,” Journal of Vacuum Science Temperature AlGaN/GaN Membrane Based Pressure & Technology A, vol. 38, no. 2, p. 020802, Mar. 2020, Sensors,” Micromachines, vol. 9, no. 5, p. 207, Apr. doi: 10.1116/1.5131343. 2018, doi: 10.3390/mi9050207. [8] P. Herfurth et al., “Ultrathin Body InAlN/GaN HEMTs [20] C. A. Chapin, R. A. Miller, K. M. Dowling, R. Chen, for High-Temperature 600°C Electronics,” IEEE and D. G. Senesky, “InAlN/GaN high electron mobility Electron Device Lett., vol. 34, no. 4, pp. 496–498, Apr. micro-pressure sensors for high-temperature 2013, doi: 10.1109/led.2013.2245625. environments,” Sensors and Actuators A: Physical, vol. [9] Y. Cai, Z. Cheng, Z. Yang, C. W. Tang, K. M. Lau, and 263, pp. 216–223, Aug. 2017, doi: K. J. Chen, “High-Temperature Operation of 10.1016/j.sna.2017.06.009. AlGaN/GaN HEMTs Direct-Coupled FET Logic [21] T. Lalinský et al., “AlGaN/GaN based SAW-HEMT (DCFL) Integrated Circuits,” IEEE Electron Device structures for chemical gas sensors,” Procedia Lett., vol. 28, no. 5, pp. 328–331, May 2007, doi: Engineering, vol. 5, pp. 152–155, 2010, doi: 10.1109/led.2007.895391. 10.1016/j.proeng.2010.09.402. [10] D. Maier et al., “High temperature stability of nitride- [22] R. Sokolovskij et al., “Hydrogen sulfide detection based power HEMTs,” Universität Ulm, 2011, doi: properties of Pt-gated AlGaN/GaN HEMT-sensor,” 10.18725/OPARU-3904. Sensors and Actuators B: Chemical, vol. 274, pp. 636– [11] A. J. Suria, A. S. Yalamarthy, H. So, and D. G. 644, Nov. 2018, doi: 10.1016/j.snb.2018.08.015. Senesky, “DC characteristics of ALD-grown

[23] V. J. Gokhale, Y. Sui, and M. Rais-Zadeh, “Novel mobility transistors,” Appl. Phys. Lett., vol. 96, no. 26, uncooled detector based on gallium nitride p. 263515, Jun. 2010, doi: 10.1063/1.3458700. micromechanical resonators,” in Infrared Technology [35] Y. Wu and J. A. del Alamo, “Electrical Degradation of and Applications XXXVIII, May 2012, doi: InAlN/GaN HEMTs Operating Under ON Conditions,” 10.1117/12.920450. IEEE Trans. Electron Devices, vol. 63, no. 9, pp. 3487– [24] H. Chen, K. Liu, L. Hu, A. A. Al-Ghamdi, and X. Fang, 3492, Sep. 2016, doi: 10.1109/ted.2016.2594034. “New concept ultraviolet photodetectors,” Materials [36] T.-C. Chang, P.-H. Chen, C.-Y. Lin, and C.-C. Shih, Today, vol. 18, no. 9, pp. 493–502, Nov. 2015, doi: “Low temperature defect passivation technology for 10.1016/j.mattod.2015.06.001. semiconductor electronic devices—supercritical fluids [25] E. Monroy, F. Omn s, and F. Calle, “Wide-bandgap treatment process,” Materials Today Physics, vol. 14, p. semiconductor ultraviolet photodetectors,” Semicond. 100225, Aug. 2020, doi: Sci. Technol., vol. 18, no. 4, pp. R33–R51, Mar. 2003, 10.1016/j.mtphys.2020.100225. doi: 10.1088/0268-1242/18/4/201. [37] W.-C. Hung et al., “Leakage Current in Fast Recovery [26] T. Lalinský et al., “Iridium oxides based gate interface Diode Suppressed by Low Temperature Supercritical of AlGaN/GaN high electron mobility transistors Fluid Treatment Process,” IEEE Electron Device Lett., formed by high temperature oxidation,” Applied vol. 41, no. 10, pp. 1540–1543, Oct. 2020, doi: Surface Science, vol. 283, pp. 160–167, Oct. 2013, doi: 10.1109/led.2020.3015287. 10.1016/j.apsusc.2013.06.069. [38] A. Y. Polyakov et al., “Trapping Phenomena in [27] C. M. Jeon and J.-L. Lee, “Investigation of IrO2 and InAlN/GaN High Electron Mobility Transistors,” ECS RuO2 Schottky contacts on AlGaN/GaN J. Solid State Sci. Technol., vol. 7, no. 2, pp. Q1–Q7, heterostructure,” Journal of Applied Physics, vol. 95, 2018, doi: 10.1149/2.0131802jss. no. 2, pp. 698–704, Jan. 2004, doi: 10.1063/1.1627454. [39] P. V. Raja et al., “Deep-Level Traps in AlGaN/GaN- [28] C. M. Jeon, H. W. Jang, and J.-L. Lee, “Thermally and AlInN/GaN-Based HEMTs With Different Buffer stable Ir Schottky contact on AlGaN/GaN Doping Technologies,” IEEE Trans. Electron Devices, heterostructure,” Appl. Phys. Lett., vol. 82, no. 3, pp. vol. 67, no. 6, pp. 2304–2310, Jun. 2020, doi: 391–393, Jan. 2003, doi: 10.1063/1.1536246. 10.1109/ted.2020.2988439. [29] H. S. Alpert et al., “Effect of Geometry on Sensitivity [40] S. Ozaki et al., “Effects of air annealing on DC and Offset of AlGaN/GaN and InAlN/GaN Hall-Effect characteristics of InAlN/GaN MOS high-electron- Sensors,” IEEE Sensors J., vol. 19, no. 10, pp. 3640– mobility transistors using atomic-layer-deposited 3646, May 2019, doi: 10.1109/jsen.2019.2895546. Al2O3,” Appl. Phys. Express, vol. 10, no. 6, p. 061001, [30] U. Ausserlechner, “Closed form expressions for sheet May 2017, doi: 10.7567/apex.10.061001. resistance and mobility from Van-der-Pauw [41] B. Shen, T. Someya, and Y. Arakawa, “Influence of measurement on 90° symmetric devices with four strain relaxation of the AlxGa1−xN barrier on transport arbitrary contacts,” Solid-State Electronics, vol. 116, properties of the two-dimensional electron gas in pp. 46–55, Feb. 2016, doi: 10.1016/j.sse.2015.11.030. modulation-doped AlxGa1−xN/GaN heterostructures,” [31] P. F. Satterthwaite, A. S. Yalamarthy, N. A. Scandrette, Appl. Phys. Lett., vol. 76, no. 19, pp. 2746–2748, May A. K. M. Newaz, and D. G. Senesky, “High 2000, doi: 10.1063/1.126463. Responsivity, Low Dark Current Ultraviolet [42] M. Hou et al., “Degradation of 2DEG transport Photodetectors Based on Two-Dimensional Electron properties in GaN-capped AlGaN/GaN heterostructures Gas Interdigitated Transducers,” ACS Photonics, vol. 5, at 600 °C in oxidizing and inert environments,” Journal no. 11, pp. 4277–4282, Oct. 2018, doi: of Applied Physics, vol. 122, no. 19, p. 195102, Nov. 10.1021/acsphotonics.8b01169. 2017, doi: 10.1063/1.5011178. [32] K. M. Dowling et al., “THE EFFECT OF BIAS [43] D. Chen et al., “High-Temperature Characteristics of CONDITIONS ON AlGaN/GaN 2DEG HALL Strain in AlGaN/GaN Heterostructures,” Jpn. J. Appl. PLATES,” presented at the 2018 Solid-State, Actuators, Phys., vol. 45, no. 1A, pp. 18–20, Jan. 2006, doi: and Microsystems Workshop, May 2018, doi: 10.1143/jjap.45.18. 10.31438/trf.hh2018.54. [44] Z. Feng, Y. Zhou, S. Cai, and K.-M. Lau, “Doping [33] A. Nasiri et al., “High-Temperature Electronics Concentration and Structural Dependences of the Packaging for Simulated Venus Condition,” Journal of Thermal Stability of the 2DEG in GaN-Based High- Microelectronics and Electronic Packaging, vol. 17, no. Electron-Mobility Transistor Structures,” Jpn. J. Appl. 2, pp. 59–66, Apr. 2020, doi: 10.4071/imaps.1115241. Phys., vol. 44, no. 1, pp. L21–L23, Dec. 2004, doi: [34] C. Ostermaier et al., “Metal-related gate sinking due to 10.1143/jjap.44.l21. interfacial oxygen layer in Ir/InAlN high electron [45] D. J. Chen et al., “Temperature-dependent strain relaxation of the AlGaN barrier in AlGaN∕GaN 10

heterostructures with and without Si3N4 surface Peter Satterthwaite received his passivation,” Appl. Phys. Lett., vol. 88, no. 10, p. B.S./M.S. from Stanford in 2018. His 102106, Mar. 2006, doi: 10.1063/1.2186369. has previously done research in nano- structured materials for solar fuel BIOGRAPHY synthesis, graphene devices, and novel sensor concepts leveraging carbon Savannah Eisner (n e Benbrook) nanotubes and GaN technology. He is received her B.S. in Electrical interested in how novel materials and Engineering from 𝑒𝑒́ Villanova nano-fabrication can enable next-generation NEMS University in 2017 and her M.S. in devices. He is currently pursuing a PhD at MIT under Electrical Engineering from Stanford Prof. Farnaz Niroui. University in 2020. She has been a Ardalan Nasiri received his B.S. Ph.D candidate in the EXtreme degree in Applied Physics from Shiraz Environments Microsystems Lab University, Shiraz, Iran, in 2008, and (XLab) in the Aeronautics & Astronautics department at his M.S. degree in Optoelectronics Stanford University since 2017. Her research interests Engineering from Malek-Ashtar include developing high-temperature electronics that can University of Technology, Isfahan, enable long-duration missions to the surface of Venus. Iran in 2012. Also, he got his second M.S. and Ph.D. degrees in Electrical Hannah Alpert received her B.S. in Engineering from the University of Physics and Astronomy from Yale Arkansas, Fayetteville, AR in 2020. His research interests University and her M.S. and Ph.D. in include microelectronic fabrication and electronic Aeronautics and Astronautics from packaging. He is currently a Senior Electronic Packaging Stanford University in the EXtreme Engineer at GlobalFoundries Inc., NY. Environments Microsystems Lab. She

currently works on entry, descent, and Sara Port graduated from Stony Brook landing technology at NASA. Her University with a Bachelor's of research interests include designing Science in Astronomy and Physics. She and building sensors for space applications. obtained a Ph.D. in Space and Planetary Science at the University of Caitlin Chapin received the B.S. Arkansas where she studied the degree in mechanical engineering formation of "metal frost" on the from the Georgia Institute of highlands of Venus through Technology in 2011, and the Ph.D. experiments and thermodynamic modeling. She is degree from the Extreme currently a Postdoctoral Researcher at NASA Glenn Environments Microsystems Lab, Research Center. Stanford University, in 2018. She is currently a Post-Doctoral Scholar at Simon S. Ang received the B.S.E.E. Lawrence Livermore National Laboratory. Her research degree from the University of interests are two-fold: developing sensors using gallium Arkansas, Fayetteville, AR, USA, nitride to enable operation in harsh environments and the M.S.E.E. degree from the characterizing the effect of mechanical stimuli and Georgia Institute of Technology, electronic traps on device performance. Atlanta, GA, USA, and the Ph.D. Ananth Saran Yalamarthy received degree in electrical engineering from the B.Tech degree in Mechanical Southern Methodist University, Engineering from the Indian Institute Dallas, TX, USA. He was a Professor of electrical of Technology, Madras, India, in 2014 engineering and the Director of the High-Density and the Ph.D. degree from the Electronics Center, University of Arkansas. He has department of Mechanical authored or coauthored over 315 journal and proceeding Engineering at Stanford University, articles. He is the principal author of power-Switching Stanford, CA, in 2019. His research Converters—Third Edition (CRC Press, 2010). Dr. Ang interests are in the intersection of energy, electronics and was a fellow of the Institution of Engineering and MEMS devices, with a special focus on wide-band gap Technology, U.K., the Electrochemical Society, and the materials. City and Guilds of LondonInstitute, U.K.

Debbie G. Senesky received the B.S. degree from the University of Southern California, Los Angeles, in 2001, and the M.S. and Ph.D. degrees from the University of California, Berkeley, in 2004 and 2007, respectively, all in mechanical engineering. She was a MEMS Design Engineer for GE Sensing from 2007 to 2008. She is currently an Associate Professor with the Aeronautics and Astronautics Department, Stanford University. Her research interests include the development of nanomaterials for extreme harsh environments, high- temperature electronics, and robust instrumentation for space exploration.