Documente Academic
Documente Profesional
Documente Cultură
200303405
1 Introduction AlGaN/GaN high electron mobility transistors (HEMTs) have received much attention
for their ability to operate at high-power levels because of a high breakdown field in the wide-bandgap
semiconductor [1, 2]. The off-state breakdown voltage (BVoff) of the device is important because it de-
termines the maximum output power, Pmax ~ Imax × BVoff/8 for class A operation. The temperature de-
pendence measurement is useful in understanding the breakdown mechanism [3, 4]. Dang et al. reported
that the BVoff had a positive temperature coefficient by using the drain-current injection technique [3].
Their results suggest that impact ionization in the channel is a dominant mechanism of the breakdown.
On the other hand, Tan et al. described that thermal runaway at the surface, rather than impact ionization,
was the mechanism of breakdown because the reverse breakdown voltage had a negative temperature
coefficient [4]. The results are contradictory and understandings of the breakdown mechanism in
AlGaN/GaN HEMTs are not sufficient.
In this paper, we have studied the mechanism of off-state breakdown in AlGaN/GaN HEMTs in detail
by investigating the temperature dependence of BVoff and the relation with the gate leakage current (Ileak).
The mechanism of the breakdown has been discussed based on the leakage current dependence of BVoff.
2 Experimental AlGaN/GaN HEMTs used in this work were fabricated on an AlGaN/GaN het-
erostructure grown by MOCVD on a (0001) sapphire substrate [5]. The epitaxial layer structure was
i-AlGaN (5 nm)/n-AlGaN (10 nm, 4 ×1018 cm–3)/i-AlGaN (5 nm)/i-GaN (3 µm)/i-AlN (4 nm). The AlN
mole fraction of the AlGaN layer was 0.3. The gate length (LG) was varied from 1 to 10 µm. The gate-
source and gate-drain spacings were 1.5 and 2 µm, respectively. The gate width was 20 µm. The pinch
off voltage was about – 3 V.
BVoff was estimated by measuring ID – VDS characteristics at off-state bias as shown in Fig. 1. Here,
LG = 1.5 µm and VGS = – 5 V (off-state). ID increased steeply at VDS above 73 V. BVoff was defined as VDG
at which the increase in ID was 1 µA (0.05 mA/mm). In this case, BVoff was estimated to be 84 V.
*
Corresponding author: e-mail: yohno@nuee.nagoya-u.ac.jp; Phone: +81 52 789 5387, Fax: +81 52 789 5232
7.0
ID (µA)
6.5 0.05
mA/mm
6.0
Breakdown
5.5
0 20 40 60 80 100 120
VDS (V)
3 Results and discussion Figure 2 shows BVoff as a function of temperature. BVoff increased from 71 to
84 V with increasing temperature from 250 to 350 K. The device has a positive temperature coefficient
of 0.12 V/K. This suggests that impact ionization in the channel is a dominant mechanism of breakdown
in the present device [3].
Figure 3 shows BVoff of the devices fabricated by two methods as a function of LG. Closed and open
circles are BVoff for the conventional devices without plasma surface treatment and for the devices with
plasma treatment, respectively. The plasma treatment was done on the AlGaN surface just under the gate
metal before the gate metal deposition by using a C2F6 ECR-plasma. The details were reported in Ref.
[6]. This is effective to reduce the gate leakage current (Ileak) by two to three orders of magnitude. BVoff
in the devices with small Ileak fabricated by plasma treatment were higher than those without plasma
treatment. This suggests that electrons injected from the gate caused the impact ionization. This explana-
tion is also supported by the fact that BVoff was lower in the long-gate device, Ileak of which was larger as
described below.
Next, we investigated the origin of the gate leakage current. Figure 4 shows Ileak at VDS = 50 V and
VGS = – 5 V as a function of LG for the conventional device without plasma treatment. Here, we used Ileak at
VDS = 50 V, for convenience, in the following discussion because Ileak was almost independent of VDS in a
voltage regime lower than BVoff. Ileak was proportional to LG, hence the area of the gate. This indicates
that the contribution of the surface leakage current was small and that the dominant component of Ileak
was due to electrons injected from the gate metal into the channel. Ileak normalized by the value at 300 K
is shown in Fig. 5 by closed circles as a function of temperature. The dashed line is the reverse current of
90 150
140
85
130
80 120
BVoff (V)
BVoff (V)
Fig. 2 Temperature dependence of BVoff. A positive Fig. 3 LG dependence of BVoff. Closed circles in the
temperature coefficient of 0.12 V/K was obtained. figure are BVoff for the conventional devices without
plasma surface treatment before the gate metal deposi-
tion, and open circles are those for the devices with
plasma treatment.
0 4
0
10
-100 Measured
-2
10
-150
-4
10
-200 -6
10
0 2 4 6 8 10 12 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2
LG (µm) 1000/T (K-1)
the Schottky contact calculated by using the thermionic emission theory assuming a barrier height of
φB = 1.5 eV [7]. The temperature dependence of the measured Ileak was much smaller than that of the
calculated Ileak. These results suggest that the dominant component of Ileak is the tunneling current from
the gate electrode to the channel. Although, at present, the mechanism of the large tunneling current has
not been fully understood, a donor-like high-density surface state [8, 9] combined with the polarization
field might decrease the depletion layer thickness of AlGaN layer under the gate metal, resulting in the
electron tunneling.
From the above results, the mechanism of off-state breakdown can be described as follows; (1) Leakage
electrons tunnel from the gate electrode to the channel, (2) with increasing VDG, the impact ionization
coefficient increases, resulting in an abrupt increase in the drain current triggered by the injected electrons.
BVoff shown in Fig. 3 and plotted as a function of Ileak in Fig. 6. BVoff decreases monotonically with
increasing Ileak. Based on the breakdown mechanism described above, an analytical formula, which de-
scribes the relation between BVoff and Ileak, was derived. Wolff’s analytical expression of impact ioniza-
tion coefficient α(E) is given as follows [10]:
qE −3E p Ei
α (E ) = exp 2
, (1)
Ei (qE λ )
where E is the electric field, Ei the threshold energy for electron–hole pair generation, Ep the effective
energy of the phonons involved in the scattering process, λ the phonon scattering mean free path, and q
150
with plasma treat.
140 w/o plasma treat.
Eq. (3)
130
120
BVoff (V)
110
100
90
80 Fig. 6 BVoff of various devices plotted as a function of Ileak.
70 The dashed line is given by Eq. (3).
-7 -6 -5 -4 -3
10 10 10 10 10
Ileak @ VDS=50V, VGS=-5V (A)
the elementary charge. Assuming that impact ionization occurs in the narrow high field region (Emax) and
that Ileak acts as triggering electron flux, the impact ionization current is given by
I ii = α ( Emax ) I leak l , (2)
where l is the length of the maximum-field region. Emax is approximately equal to VDG/l. When BVoff is
defined as VDG at which the increase in drain current (∆ID) due to Iii becomes a certain value (0.05
mA/mm in the present case), BVoff is given by
Ei ∆I D l
2
4 Summary We have investigated the temperature dependence of BVoff. It has a positive temperature
coefficient of 0.12 V/K, and increased by reducing the gate leakage current. These results suggest that
off-state breakdown was not due to thermal runaway but due to impact ionization in the channel, which
was caused by gate leakage electrons.
Acknowledgements This work was supported in part by a Grant-in-Aid for General Scientific Research of the
Ministry of Education, Culture, Sports, Science and Technology, and by Industrial Technology Research Grant
Program in ’02 from NEDO of Japan.
References
[1] Y. F. Wu et al., Appl. Phys. Lett. 69, 1438 (1996).
[2] M. Asif Khan et al., Solid State Electron. 41, 1555 (1997).
[3] X. Z. Dang et al., Electron. Lett. 35, 602 (1999).
[4] W. S. Tan et al., Appl. Phys. Lett. 80, 3207 (2002).
[5] M. Akita et al., Electron. Lett. 36, 1736 (2000).
[6] S. Mizuno et al., J. Appl. Phys. 41, 5125 (2002).
[7] A. J. Sierakowski and L. F. Eastman, J. Appl. Phys. 86, 3398 (1999).
[8] M. Konishi et al., Abst. Int. Symp. Compound Semiconductors (Tokyo, Japan, 2001), p. 10.
[9] J. P. Ibbetson et al., Appl. Phys. Lett. 77, 250 (2000).
[10] P. A. Wolff, Phys. Rev. 95, 1415 (1954).
[11] M. Reigrotzki et al., J. Appl. Phys. 86, 4458 (1999).
[12] R. L. Aggarwal et al., Solid State Commun. 117, 549 (2001).
[13] Y. Ohno et al., Int. Phys. Conf. Series 170, 119 (2002).
[14] H. S. Nalwa (Ed.), Handbook of Advanced Electronic and Photonic Materials and Devices, Vol. 2: Semicon-
ductor Devices (Academic Press, New York 2001), p. 73.
[15] B. K. Ridley, J. Phys.: Condens. Matter 8, 511 (1996).