Reliability in room-temperature negative differential ... - icorlab

APPLIED PHYSICS LETTERS 97, 181109 共2010兲

Reliability in room-temperature negative differential resistance characteristics of low-aluminum content AlGaN/GaN double-barrier resonant tunneling diodes C. Bayram, Z. Vashaei, and M. Razeghia兲 Center for Quantum Devices, Department of Electrical Engineering and Computer Science, Northwestern University, Evanston, Illinois 60208, USA

共Received 1 June 2010; accepted 22 October 2010; published online 5 November 2010兲 AlGaN/GaN resonant tunneling diodes 共RTDs兲, consisting of 20% 共10%兲 aluminum-content in double-barrier 共DB兲 active layer, were grown by metal-organic chemical vapor deposition on freestanding polar 共c-plane兲 and nonpolar 共m-plane兲 GaN substrates. RTDs were fabricated into 35-␮m-diameter devices for electrical characterization. Lower aluminum content in the DB active layer and minimization of dislocations and polarization fields increased the reliability and reproducibility of room-temperature negative differential resistance 共NDR兲. Polar RTDs showed decaying NDR behavior, whereas nonpolar ones did not significantly. Averaging over 50 measurements, nonpolar RTDs demonstrated a NDR of 67 ⍀, a current-peak-to-valley ratio of 1.08, and an average oscillator output power of 0.52 mW. © 2010 American Institute of Physics. 关doi:10.1063/1.3515418兴 Resonant tunneling diodes 共RTDs兲 possess a unique quantum-interference characteristic called negative differential resistance 共NDR兲, which is macroscopically identified in device I-V characteristics as the decrease in current with the increase in voltage.1 NDR, an example of quantummechanical tunneling phenomenon in ultrathin structures,2 is employed in many electronic devices 共such as in radios兲 as well as in the generation and detection of terahertz waves.3 RTDs, compared to other negative resistance devices such as 共Esaki兲 tunnel and transferred-electron devices 共i.e., Gunn diode, Thyristor, and impact ionization avalanche transittime diode兲, possess lower junction capacitance 共due to relatively lower doping levels兲, enabling them to operate at higher oscillating frequencies. However, the current bottlenecks of conventional 共such as GaAs-based兲 RTDs are the upper frequency limit, output power, and operating temperature.4 Recently, III-nitrides have gained interest for intersubband 共ISB兲 devices. This is because large electron effective mass 共mⴱ ⬃ 0.2– 0.3⫻ m0兲 and longitudinal optical phonon energy 共⬃90 meV兲 enable ultrafast ISB relaxation offering very high speed devices. Specifically, wide bandgap, large conduction band discontinuity 关⬃2.1 eV in AlN/GaN 共Ref. 5兲兴, high carrier mobility, and thermal stability promise high power, high frequency room-temperature operation for GaNRTDs. Demonstrating NDR and understanding the transport in AlGaN/GaN double-barrier 共DB兲 RTDs will enable roomtemperature terahertz oscillators and quantum cascade lasers.6 Despite these promises, there are no reports of reliable NDR in GaN-based DB heterostructures. Various groups have studied AlN/GaN double-barrier heterostructures grown by molecular beam epitaxy.7–11 Recently, we have studied similar structures by metal-organic chemical vapor deposition 共MOCVD兲.12,13 In all these works, NDR behavior degraded after the initial electrical measurements. Attempts to increase the reproducibility of NDR ina兲

Electronic mail: [email protected].

0003-6951/2010/97共18兲/181109/3/$30.00

cluded decreasing the template dislocation density,12–14 improving active layer quality,14 and decreasing the device mesa size7–9,11–14兲. However, most of these works were on sapphire substrate that is highly lattice-mismatched to GaN, leading to a high density of dislocations 共ⱖ108 cm−2兲. Moreover, these works employed high aluminum content 共ⱖ70%兲 barrier designs, leading to large lattice-mismatch at the heterointerfaces. One way to improve the material quality and NDR behavior is by using low-aluminum content AlGaN/GaN double-barrier heterostructures grown on lowdislocated substrates 共such as freestanding GaN兲—which has not yet been studied. In this work, we investigate the reliability and reproducibility of NDR in low-aluminum content polar and nonpolar AlGaN/GaN DB RTDs. As such, very low dislocated 共⬃105 cm−2兲 polar and nonpolar freestanding GaN substrates,15 and AlGaN/GaN double heterostructures,16 composed of 20% and 10% aluminum content in barriers, were employed in the active layers of RTDs. The material was grown in an MOCVD reactor.16 First, i- and n-GaN carrier concentrations were determined via Hall-effect measurements as −6 ⫻ 1016 and −3 ⫻ 1019 cm−3, respectively. Then, the quality, thickness, and aluminum content of the double-barrier active layers were calibrated via AlGaN/GaN superlattice growths followed by x-ray diffraction, photoluminescence, and atomic force microcopy 共AFM兲 studies.12,13,16 After the material calibrations, the growths of RTDs started with 3 ␮m i-GaN regrowth. Then, 400-nm-thick n-GaN was grown to act as the bottom contact. The DB active layer of RTD is composed of a narrower bandgap material sandwiched between wider bandgap materials; thus, an inherit lattice-mismatch between barrier and well materials is inevitable. The critical thickness 共above which dislocations are generated兲 is inversely proportional to the aluminum content for GaN homoepitaxial growth.17 Thus, our DB active layer design employs low-aluminum content 共20% or 10%兲 in the barrier to prevent the lattice relaxation via dislocation formation as the barrier is grown

97, 181109-1

© 2010 American Institute of Physics

Downloaded 05 Nov 2010 to 165.124.165.59. Redistribution subject to AIP license or copyright; see http://apl.aip.org/about/rights_and_permissions

181109-2

Bayram, Vashaei, and Razeghi

Appl. Phys. Lett. 97, 181109 共2010兲

FIG. 1. 共Color online兲共2 ␮m ⫻ 2 ␮m兲 AFM images of completed 共a兲 polar and 共b兲 nonpolar DB RTD structures. Vertical scales are 2 nm. Respective crystallographic directions are shown as insets.

thinner 共1.5 or 1.6 nm兲 than the critical thickness 共⬃1.8 or ⬃3.0 nm兲.17 The polar DB active layer was composed of 2.0 nm i-GaN, 1.5 nm i-Al0.20Ga0.80N, 1.25 nm i-GaN, 1.5 nm i-Al0.20Ga0.80N, and 2.0 nm i-GaN, whereas nonpolar RTD active layer was composed of 2.6 nm i-GaN, 1.6 nm i-Al0.10Ga0.90N, 1.6 nm i-GaN, 1.6 nm i-Al0.10Ga0.90N, and 2.6 nm i-GaN. This arrangement enables a conduction band offset of 0.42 共0.16兲 eV in AlGaN barrier, giving a single and discrete electronic level of 0.32 共0.10兲 eV in the polar 共nonpolar兲 GaN well. The devices were finalized via top contact layer of 250-nm-thick n-GaN 关Fig. 2共a兲兴. The NDR is very sensitive to surface roughness and dislocations and can be observed only for very high material quality.18 A typical 共2 ␮m ⫻ 2 ␮m兲 AFM scan of the completed polar and nonpolar RTD structure is shown in Figs. 1共a兲 and 1共b兲, respectively. The surfaces 共a兲 and 共b兲 possess a root-mean-square roughness of 1.2 and 1.6 Å, respectively. Well-ordered parallel atomic steps are observed in Figs. 1共a兲 and 1共b兲 with no dislocation terminations demonstrating the excellent material quality. Figure 2共a兲 shows the side-view schematic of the fabricated RTD device. Figures 2共b兲 and 2共c兲 show the top-view optical and scanning electron micrograph 共SEM兲 of the fabricated RTD device. The mesa diameter is 35 ␮m 关Fig. 2共b兲 inset兴, which is connected to the top contact pads via the bridge structure. The bottom contact is surrounding the mesa partly to enable efficient carrier injection 关Figs. 2共b兲 and 2共c兲兴. Fabrication of the RTDs was realized via conventional semiconductor methods and tools.12 It started with the top mesa formation by dry etching. This was followed by the dry etching of excess n-type material 共point A in Fig. 2兲. This step was to minimize the leakage current and parasitic capacitance between contact layers. Then, the bottom contact 共points F and G in Fig. 2兲 composed of 400 Å Ti/1500 Å Au was deposited. This was followed by the deposition of 300nm-thick silicon dioxide as passivation. This passivation layer was removed from the top contact region of the device mesa 共i.e., point D in Fig. 2兲 and bottom contact pads 共i.e., point G in Fig. 2兲 via wet etching. The RTD was completed

FIG. 2. 共Color online兲共a兲 Side-view schematic of the fabricated device. 共b兲 Top-view optical and 共c兲 scanning electron micrograph of fabricated device. Inset of 共b兲 shows the SEM bird’s eye view of the mesa. Points A–G shown in the figures correspond to side-view 共a兲 and top-views 关共b兲 and 共c兲兴 of the indicated device locations.

by 400 Å Ti/1500 Å Au top contact metal deposition that formed the top contact bridge to the device mesa and top contact pads. The transmission line model measurements of the top and bottom contacts of polar and nonpolar RTDs were carried out, and highly Ohmic19 and Schottky contact behaviors 共with a barrier height of 1.5 eV兲 for polar and nonpolar devices were identified, respectively. The ohmicity difference between the contacts on polar and nonpolar planes could be related to the different surface energies of polar and nonpolar orientations as the same metal contacts were employed and need further investigation. RTD electrical measurements were realized under continuous wave at room-temperature. The voltage polarity refers to that applied to the top electrode 关see Fig. 2共a兲兴. All I-V curves were measured using an HP4155A semiconductor parameter analyzer configured to input a voltage sweep while measuring current. Figures 3共a兲–3共c兲 show I-V characteristic of the fabricated 共a兲 polar and 关共b兲 and 共c兲兴 nonpolar devices. NDR is clearly observed in both devices at NDR voltage onset 共VP兲 FIG. 3. 共Color online兲 Electrical characterizations of 共a兲 polar and 关共b兲 and 共c兲兴 nonpolar RTDs are shown. I-V curve measurements show 共a兲 the degradation and 共b兲 the reliability in NDR behavior of polar and nonpolar devices. 共c兲 Hysteresis curve of the nonpolar RTD is displayed.


181109-3

Appl. Phys. Lett. 97, 181109 共2010兲

Bayram, Vashaei, and Razeghi

of 2.3 共polar one兲 and 6.2 V 共nonpolar one兲. The higher VP of nonpolar device is attributed to contact Schottky barrier. The reliability and reproducibility of NDR were studied via tens of I-V scans. Figure 3共a兲 shows the first through 20th scan of the polar device. Although polar Al0.2Ga0.8N / GaN-based device possesses more reproducible NDR than their AlN/GaN-based counterparts,12–14 their NDR behavior was observed to suffer from degradation as more scans were carried out. The enhanced NDR reproducibility up to 20 scans is attributed to the lower aluminum content in the double-barriers.19 Lower aluminum content in the barriers decreases the lattice-mismatch to GaN and thereby dislocation density and lowers piezoelectric field at the AlGaN/ GaN heterointerfaces. The lack of reliability—the decrease in NDR onset voltage and current-peak-to-valley ratio with consecutive scans—is attributed to interface dislocations trapping charges. These trapped charges lower the effective barrier height and alter the dominant transport mechanism. The fact that NDR degrades with increasing scan number suggests that these dislocations do not release the trapped charges and implies further that material improvements are required for more reliable NDR in polar RTDs. Figures 3共b兲 and 3共c兲 plot the I-V curves of nonpolar RTDs. These diodes showed more reproducible and reliable NDR behavior than the polar ones. The first, 30th, and 50th scan are plotted in Fig. 3共b兲. No significant degradation in the NDR behavior is observed. This suggests a decrease of aluminum content in DB active layer down to 10%, and elimination of polarization fields20 increased the NDR reproducibility and reliability. This is expected as barriers with lower aluminum content have lower lattice-mismatch to GaN wells and generate lower strain and polarization fields at the AlGaN/GaN heterointerfaces. As shown in Fig. 3共c兲, when in resonance—the energy of the electron states in the electrode aligns with the discrete energy level of the well—the peak current 共IP ⬇ 83.41 mA兲 is achieved. This corresponds to the peak voltage 共VP ⬇ 6.24 V兲. The expected VP was ⬃0.2 V. The relatively large observed VP in our device is attributed to the contact Schottky barrier 共⬃1.5 V兲 and the small series resistance that becomes relatively important at high currents 共⬃80 mA兲. These suggest that further improvements in material growth and device fabrication are necessary for ideal performance. With further increase in bias, the emitter electron energy level falls below the edge of the conduction band into the gap and the current is minimized. In this case, the current and voltage are labeled as valley current 共IV ⬇ 75.36 mA兲 and voltage 共VV ⬇ 6.62 V兲. Nonpolar devices were further characterized as NDR behavior was reproducible and reliable. For hysteresis measurements 关i.e., difference in I-V behavior between upward and downward scan, seen in Fig. 3共c兲兴, the voltage sweep was adjusted for a loop. The hysteresis identified in the I-V curve 关Fig. 3共c兲兴 is attributed to the charge trapping associated with defects. Defects, charged during the upward scan, broaden the discrete electronic state in the well, preventing the observation of NDR in the downward scan. The fact that these charged defects can be emptied effectively shows that our material and device quality is very high, and the reproducibility of NDR in III-nitride double-barrier heterostructures can be improved by decreasing the aluminum content in the DB active layer, dislocation density, and polarization fields.

Typical figures of merit in NDR devices are the negative differential resistance 共R ⬇ ⌬V / ⌬I兲, current-peak-to-valleyratio 共CPVR= IP / IV兲, and average oscillator output power 共PMAX兲共given by PMAX ⬇ 3 / 16 ⫻ ⌬I ⫻ ⌬V兲.21 The average values of the R, CPVR, and PMAX over 50 scans are 67⫾ 5 ⍀, 1.08⫾ 0.02, and 0.52⫾ 0.01 mW, respectively. These figures of merit, achieved for MOCVD-grown 35-␮m-diameter diodes and at room-temperature, are comparable to the state-of-the-art GaAs-based RTDs 共Ref. 4兲 and demonstrate the promise of low-aluminum content nonpolar AlGaN/GaN double-barrier RTDs. In conclusion, the employment of low dislocation density substrate, polarization-free design, and low-aluminum content active layer approach was shown to increase reliability and reproducibility of NDR in RTDs. GaN-based RTDs possessed an R of −67 ⍀, a CPVR of 1.08, and a PMAX of 0.52 mW at room-temperature. Our work motivates further research toward low-aluminum content polarization-free AlGaN/GaN DB-structures grown on low dislocation density substrates. The authors thank Dr. J. Zavada of ARO, Dr. J. Mangano, and Dr. Scott Rodgers of DARPA and Mr. Jerry Speer of the U.S. Army RDECOM for their interest and encouragement, and C.B. acknowledges IBM and the Link Foundation Energy Fellowships. L. L. Chang, L. Esaki, and R. Tsu, Appl. Phys. Lett. 24, 593 共1974兲. L. Esaki, Phys. Rev. 109, 603 共1958兲. 3 T. C. L. G. Sollner, W. D. Goodhue, P. E. Tannenwald, C. D. Parker, and D. D. Peck, Appl. Phys. Lett. 43, 588 共1983兲. 4 M. Asada, S. Suzuki, and N. Kishimoto, Jpn. J. Appl. Phys. 47, 4375 共2008兲. 5 G. Martin, S. Strite, A. Botchkarev, A. Agarwal, A. Rockett, H. Morkoc, W. R. L. Lambrecht, and B. Segall, Appl. Phys. Lett. 65, 610 共1994兲. 6 V. D. Jovanović, D. Indjin, Z. Ikonić, and P. Harrison, Appl. Phys. Lett. 84, 2995 共2004兲. 7 A. Kikuchi, R. Bannai, K. Kishino, C. M. Lee, and J. I. Chyi, Appl. Phys. Lett. 81, 1729 共2002兲. 8 C. T. Foxon, S. V. Novikov, A. E. Belyaev, L. X. Zhao, O. Makarovsky, D. J. Walker, L. Eaves, R. I. Dykeman, S. V. Danylyuk, S. A. Vitusevich, M. J. Kappers, J. S. Barnard, and C. J. Humphreys, Phys. Status Solidi C 0, 2389 共2003兲. 9 M. Hermann, E. Monroy, A. Helman, B. Baur, M. Albrecht, B. Daudin, O. Ambacher, M. Stutzmann, and M. Eickhoff, Phys. Status Solidi C 1, 2210 共2004兲. 10 M. V. Petrychuk, A. E. Belyaev, A. M. Kurakin, S. V. Danylyuk, N. Klein, and S. A. Vitusevich, Appl. Phys. Lett. 91, 222112 共2007兲. 11 S. Leconte, S. Golka, G. Pozzovivo, G. Strasser, T. Remmele, M. Albrecht, and E. Monroy, Phys. Status Solidi C 5, 431 共2008兲. 12 C. Bayram, Z. Vashaei, and M. Razeghi, Appl. Phys. Lett. 96, 042103 共2010兲. 13 Z. Vashaei, C. Bayram, and M. Razeghi, J. Appl. Phys. 107, 083505 共2010兲. 14 S. Golka, C. Pflügl, W. Schrenk, G. Strasser, C. Skierbiszewski, M. Siekacz, I. Grzegory, and S. Porowski, Appl. Phys. Lett. 88, 172106 共2006兲. 15 Z. Vashaei, E. Cicek, C. Bayram, R. McClintock, and M. Razeghi, Appl. Phys. Lett. 96, 201908 共2010兲. 16 Z. Vashaei, C. Bayram, P. Lavenus, and M. Razeghi, Appl. Phys. Lett. 97, 121918 共2010兲. 17 A. D. Bykhovski, B. L. Gelmont, and M. S. Shur, J. Appl. Phys. 81, 6332 共1997兲. 18 M. Razeghi, A. Tardella, R. A. Davies, A. P. Long, M. J. Kelly, E. Britton, C. Boothroyd, and W. M. Stobbs, Electron. Lett. 23, 116 共1987兲. 19 C. Bayram, Z. Vashaei, and M. Razeghi, Appl. Phys. Lett. 97, 092104 共2010兲. 20 F. Sacconi, A. D. I. Carlo, and P. Lugli, Phys. Status Solidi A 190, 295 共2002兲. 21 S. C. Kim and A. Brandli, IRE Trans. Circuit Theory 8, 416 共1961兲. 1 2
