f Noise Characteristics of Surface-Treated Normally-Off ... - IEEE Xplore

1 downloads 0 Views 524KB Size Report
Young-Woo Jo, Student Member, IEEE, Jung-Hee Lee, Senior Member, IEEE, and Yoon-Ha Jeong, Fellow, IEEE. Abstract—Normally off Al2O3/GaN MOSFETs ...
IEEE ELECTRON DEVICE LETTERS, VOL. 36, NO. 3, MARCH 2015

229

1/ f Noise Characteristics of Surface-Treated Normally-Off Al2O3/GaN MOSFETs SungHwan Sakong, Sang-Hyun Lee, Student Member, IEEE, Taiuk Rim, Member, IEEE, Young-Woo Jo, Student Member, IEEE, Jung-Hee Lee, Senior Member, IEEE, and Yoon-Ha Jeong, Fellow, IEEE

Abstract— Normally off Al2 O3 /GaN MOSFETs are fabricated with a tetramethylammonium hydroxide (TMAH) treatment as a postgate recess etch. The effects of the surface treatment on the etched GaN surface are investigated using low-frequency (1/ f ) noise and capacitance–voltage (C–V ) measurements. For a quantitative comparison with conventional devices, the oxide trap density (N ot ) is extracted using the unified 1/ f noise model, whereas the interface trap density ( Dit ) is extracted using the high–low-frequency C–V method. After the TMAH treatment, Not is found to have decreased from 5.40 × 1019 to 2.50 × 1019 eV−1 cm−3 , whereas Dit is decreased from 2.8 × 1012 to 1.1 × 1011 eV−1 cm−2 , as compared with conventional devices. The surface treatment is thus shown to lower trap density in the Al2 O3 /GaN MOSFETs by smoothing the surface and suppressing plasma damage in the recessed GaN surfaces. Index Terms— GaN, MOSFET, normally-off, tetramethylammonium hydroxide (TMAH), interface, trap density, 1/f , low frequency noise.

I. I NTRODUCTION -NITRIDE semiconductors demonstrate superior performance in high-power, high-frequency, and high-temperature applications due to their remarkable material and physical properties, such as high saturation velocity, electron mobility, and breakdown voltage and good thermal conductivity [1], [2]. In particular, AlGaN/GaN heterostructure field-effect transistors (HFETs) based on the 2DEG are among the most promising power devices, operating in the normallyoff mode. However, these normally-off AlGaN/GaN HFETs suffer from a relatively low threshold voltage with a limited gate voltage swing, which can interfere with the on/off error during switching, and significant gate leakage current due to the high interface trap density caused by the instability of Schottky gate contacts [3].

III

Manuscript received December 30, 2014; revised January 14, 2015; accepted January 18, 2015. Date of publication January 21, 2015; date of current version February 20, 2015. This work was supported by the IT Consilience Creative Program through NIPA under Grant NIPA-2014-H0201-14-1001. The review of this letter was arranged by Editor J. A. del Alamo. S. Sakong is with the Division of IT-Conversions Engineering, Pohang University of Science and Technology, Pohang 790-784, Korea (e-mail: [email protected]). S.-H. Lee and Y.-H. Jeong are with the Department of Electrical Engineering, Pohang University of Science and Technology, Pohang 790-784, Korea (e-mail: [email protected]). T. Rim is with the Future IT Innovation Laboratory, Pohang University of Science and Technology, Pohang 790-784, Korea. Y.-W. Jo and J.-H. Lee are with the School of Electrical Engineering and Computer Science, Kyungpook National University, Daegu 702-201, Korea. Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LED.2015.2394373

The deposition of high-quality gate dielectrics (such as Si3 N4 [4] and Al2 O3 [5]) on the recessed GaN surface effectively enables the fabrication of the GaN MOSFET. In these MOSFET structures, the gate dielectrics decrease the gate leakage current and permit a substantial gate voltage swing. Nevertheless, the surface roughness caused by plasma damage during the recess etch increases the interface trap density and gate leakage current. However, the TMAH solution provides an anisotropic etchant of GaN with very slow etch rates [6] for smoothing effectively the surface with plasma damages. In this letter, we investigate a TMAH treatment for reducing plasma damage and its effect on normally-off Al2 O3 /GaN MOSFETs. We compare devices fabricated with and without the surface treatment using low frequency (1/f ) noise [7] and capacitance-voltage (C-V) measurements [8]. The oxide trap density (Not ) and interface trap density (Dit ) values extracted indicate that the TMAH-treatment of an etched GaN surface effectively decreases the trap density in Al2 O3 /GaN MOSFETs and improves DC performance, in particular the sub-threshold slope, breakdown voltage, and drive current Ids , under the same bias conditions. II. D EVICE FABRICATION The AlGaN/GaN heterostructure was grown on a sapphire substrate by metal-organic chemical vapor deposition (MO-CVD). The active region of the MOSFET was etched by a transformer-coupled-plasma reactive-ion etching (TCP-RIE) to a depth of 250 nm using a BCl3 /Cl2 gas mixture. The gate region was formed as a 3 μm × 50 μm rectangle and fully recessed with a photoresist mask by completely removing the AlGaN layer by TCP-RIE. This was directly followed by the surface treatment using a 25% concentration of the TMAH solution at 85°C for 10 min without removing the photoresist, whereas conventional devices undergo no TMAH treatment. Immediately after etching, a gate insulator with a 30-nm Al2 O3 layer was grown by the atomic layer deposition (ALD) on both THAH-treated and conventional GaN surfaces. Ti/Al/Ni/Au ohmic contacts were deposited as the source and drain electrodes by an electron beam evaporator and annealed using rapid thermal processing at 850°C for 30 sec in a N2 ambient. Finally, the Ni/Au gate electrode was deposited on the Al2 O3 gate insulator. Fig. 1(a) shows a schematic cross-section of the device and an atomic force microscopy (AFM) image of the recessed GaN surface. From the AFM image after the

0741-3106 © 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

230

IEEE ELECTRON DEVICE LETTERS, VOL. 36, NO. 3, MARCH 2015

Fig. 2. Four sets of data measured from TMAH-treated and conventional Al2 O3 /GaN MOSFETs: (a) Hooge parameter vs. gate overdrive, (b) (S I d /Id 2 ) vs. Id at f = 10 Hz. The data are compared with the corresponding plots of (constant × (gm /Id )2 ) vs. Id (solid line). (c) The measured (SV g )1/2 (symboles) versus (Vgs − Vt ), fitted by solid lines. (d) The N and μ fluctuations extracted from SV g .

Fig. 1. (a) Schematic cross-section of the Al2 O3 /GaN MOSFET and AFM image of the interfaces. (b) Breakdown voltages. (c) 300 ns pulsed I -V measurements for gate lag effect. (d) Transfer curves. (e) Drain current spectral density (S I d ) of the TMAH-treated Al2 O3 /GaN MOSFET (red) and conventional Al2 O3 /GaN MOSFET (blue) at Vds = 1 V and Id = 0.35 mA.

TMAH treatment, the recessed GaN surface exhibits fewer and blunter protrusions and pits than the surface that was not treated. The improved surface is directly related to the dielectric qualities. For example, the TMAH-treated devices show enhanced breakdown voltage from reduced crowding effects and pits (Fig. 1(b)). And the devices also show the reduced current lagging due to reduced current collapse with decreased interface traps (Fig. 1(c)) [9]. III. R ESULTS AND D ISCUSSION The transfer characteristics of both TMAH-treated and conventional Al2 O3 /GaN MOSFETs with L g = 3 μm, L w = 50 μm, and gate-to-drain length (L gd ) = 13 μm are shown in Fig. 1(d). The normally-off operation was successfully demonstrated in both devices but the TMAH-treated devices exhibit greater maximum transconductance (gmmax ) and maximum drain current (Idmax ) due to reduced mobility degradation resulting from the surface roughness between the Al2 O3 and GaN channel. Fig. 1(e) compares the frequency dependence of the measured drain-current noise spectral density (SI d ) in the TMAH-treated and conventional Al2 O3 /GaN MOSFETs biased at Vds = 1 V in the linear region with Id = 0.35 mA. The data was obtained from eight samples of each device

and was averaged. This is necessitated by the large sample-to-sample variations in Id and gm . For instance the dispersion of SI d ranges from 2.4 × 10−17 to 5.0 × 10−17 and 1.7 × 10−17 to 2.6 × 10−16 A2 /Hz at f = 10 Hz in surface treated and conventional devices, respectively. The value of S I d is 3.2 × 10−17 and 1.0 × 10−16 A2 /Hz at f = 10 Hz for devices with and without TMAH treatments, respectively. Thus SI d in the TMAH-treated device is smaller with reduced dispersion, compared with the conventional device. Fig. 2(a) shows the Hooge parameter (α H ) versus gate overdrive voltage (Vgs − Vt ) in both devices. The α H value as extracted via α H = f W LC ox (Vgs − Vt )SI d /q Id2 [10] ranges from 10−1 to 10−3 . This value is shown the higher than the typical value of 10−6 in Si MOSFETs by orders of magnitude. Evidently the high-k dielectric (Al2 O3 ) and the low quality GaN substrate with the complicated epi-structures [11] are the primary causes. The large α H is an indication that carrier number fluctuation (N) is involved. Fig. 2(b) shows the average normalized drain-current noise spectral density (SI d /Id2 ) at f = 10 Hz as a function of Id at Vds = 1 V for the two device types. As clear from the fitted curve there is a good match between SI d /Id2 and (gm /Id )2 except for the highId region where the fitted curve mismatch is observed. The slope deviation between SI d /Id2 and (gm /Id )2 explicitly points to the mobility fluctuation strongly involved in the noise data as well. Fig. 2(c) presents the gate-voltage referred noise spectral density (SV g ) versus (Vgs -Vt ). In the unified 1/ f noise model, SV g is specified by SV g = SV F B [1 + αμ0 Cox |Vgs − Vt |]2 with VF B noise density given by 2 ). Here α is the scattering SV F B = (λkT q 2 Not )/( f γ L W L g Cox coefficient, μ0 (μ0T M AH = 155, μ0conv = 96 cm2 ·V−1 ·s−1 ) 1/2

is extracted using Y (Vg ) = Id /gm = (WC ox μ0 Vd /L)1/2 (Vgs − Vt ) at Vd = 1 V, Cox is the gate capacitance

SAKONG et al.: 1/ f NOISE CHARACTERISTICS OF SURFACE-TREATED NORMALLY OFF Al2 O3 /GaN MOSFETs

TABLE I E XTRACTED E FFECTIVE T RAP D ENSITY AND S CATTERING C OEFFICIENT OF THE Al2 O 3 /GaN MOSFETs

231

the results. This is due to the trap distribution peaked at the oxide surface and Dit being healed directly via the surface treatment. IV. C ONCLUSION

(2.65 × 10−7 F/cm2 ), λ is the tunneling attenuation length, γ is the frequency exponent, and Not is the oxide trap density (eV−1 cm−3 ) [7]. The parameter γ has been determined to be unity from Fig. 1(e), and λ as represented by λ = [4π(2m ∗  B )1/2 /h]−1 has been used with m ∗ denoting the effective electron mass and  B the oxide barrier height. The parameter values used are taken from [12], that is, m ∗ = 0.16m 0 (m 0 is the electron mass) and  B = 3.45 eV, so that λ ≈ 0.11 nm. Fig. 2(d) shows the relative magnitude of N and μ thus extracted from SV g . The result indicates N as a primary source of the noise driven by the direct or FN tunneling of electrons in conduction band into or out of oxide traps. But μ occurring concurrently via the Coulomb interaction between trapped and the image charges should also be an equally important source as manifested by the large sample to sample variation and the good fit between SI d and Id . The fast trapping and detrapping of electrons could affect both N and μ, the quantitative and detailed description of which 1/2 appears to be in order. Note from Fig. 2(c) that SV g is linear with respect to (Vgs − Vt ), indicating thereby that the series resistance does not significantly affect the drain current. If it 1/2 affects Id appreciably, SV g is proportional to (Vgs − Vt )2 [7]. 1/2

Also, the slope of SV g vs. Vg curve in surface treated device is smaller than that of conventional devices, suggesting that surface treated device exhibits weaker dependence on Vg . Thus the data presented in Fig. 2 shows that N and μ are suppressed significantly in TMAH-treated device due to the reduced trapping and detrapping of electrons in oxide traps and the concomitantly suppressed Coulomb interaction between trapped charges and their image charges. The results obtained are summarized in Table I, namely Not , scattering coefficient (α), the relative magnitudes of μ, N as determined by the unified 1/ f model, and Dit extracted by the high-low frequency C-V (HLCV) method. The superior noise characteristics of the surface treated device can be attributed to the improved morphology of the recess GaN surface via reduced plasma damage. Finally Dit more efficiently improved compared with Not is consistent with

We have fabricated normally-off Al2 O3 /GaN MOSFETs (L g = 3 μm and L w = 50 μm) both with and without using a TMAH treatment as a post-gate recess etch. The impact of the surface treatment has been characterized, using the low frequency 1/ f noise and HLCV measurements and various parameters entailed. The Not and Dit extracted from the surface-treated devices are noticeably less than those in conventional devices. Thus the simple TMAH surface treatment improves the quality of both the dielectric and interface of Al2 O3 /GaN MOSFETs by mitigating surface damage in the GaN recesses caused by plasma etching. From both a performance and reliability perspective, this surface treatment is attractive for use in Al2 O3 /GaN-based power device applications. R EFERENCES [1] T. P. Chow and R. Tyagi, “Wide bandgap compound semiconductors for superior high-voltage unipolar power devices,” IEEE Trans. Electron Devices, vol. 41, no. 8, pp. 1481–1483, Aug. 1994. [2] O. Aktas et al., “High temperature characteristics of AlGaN/GaN modulation doped field-effect transistors,” Appl. Phys. Lett., vol. 69, no. 25, pp. 3872–3874, Dec. 1996. [3] C. Sanabria et al., “The effect of gate leakage on the noise figure of AlGaN/GaN HEMTs,” IEEE Electron Device Lett., vol. 27, no. 1, pp. 19–21, Jan. 2006. [4] G. Simin et al., “Large periphery high-power AlGaN/GaN metal-oxidesemiconductor heterostructure field effect transistors on SiC with oxidebridging,” IEEE Electron Device Lett., vol. 22, no. 2, pp. 53–55, Feb. 2001. [5] T. Hashizume, S. Ootomo, and H. Hasegawa, “Suppression of current collapse in insulated gate AlGaN/GaN heterostructure field-effect transistors using ultrathin Al2 O3 dielectric,” Appl. Phys. Lett., vol. 83, no. 14, pp. 2952–2954, Oct. 2003. [6] Z. Yang et al., “GaN-on-patterned-silicon (GPS) technique for fabrication of GaN-based MEMS,” Sens. Actuators A, Phys., vols. 130–131, no. 14, pp. 371–378, Aug. 2006. [7] R. Jayaraman and C. G. Sodini, “A 1/f noise technique to extract the oxide trap density near the conduction band edge of silicon,” IEEE Trans. Electron Devices, vol. 36, no. 9, pp. 1773–1782, Sep. 1989. [8] R. Engel-Herbert, Y. Hwang, and S. Stemmer, “Comparison of methods to quantify interface trap densities at dielectric/III-V semiconductor interfaces,” J. Appl. Phys., vol. 108, no. 12, p. 124101, 2010. [9] S. C. Binari, P. B. Klein, and T. E. Kazior, “Trapping effects in GaN and SiC microwave FETs,” Proc. IEEE, vol. 90, no. 6, pp. 1048–1058, Jun. 2002. [10] J.-M. Peransin et al., “1/f noise in MODFETs at low drain bias,” IEEE Trans. Electron Devices, vol. 37, no. 10, pp. 2250–2253, Oct. 1990. [11] L. K. J. Vandamme and F. N. Hooge, “What do we certainly know about 1/f noise in MOSTs?” IEEE Trans. Electron Devices, vol. 55, no. 11, pp. 3070–3085, Nov. 2008. [12] V. Di Lecce et al., “Metal-oxide barrier extraction by Fowler–Nordheim tunnelling onset in Al2 O3 -on-GaN MOS diodes,” Electron. Lett., vol. 48, no. 6, pp. 347–348, Mar. 2012.