Degradation mechanisms of GaAs PHEMTs in high humidity conditions

Microelectronics Reliability 45 (2005) 1894–1900 www.elsevier.com/locate/microrel

Degradation mechanisms of GaAs PHEMTs in high humidity conditions Takayuki Hisaka a,*, Yasuki Aihara a, Yoichi Nogami a, Hajime Sasaki a, Yasushi Uehara b, Naohito Yoshida a, Kazuo Hayashi a a

High Frequency & Optical Device Works, Mitsubishi Electric Corporation, 4-1, Mizuhara, Itami, Hyogo 664-8641, Japan b Advanced Technology R&D Center, Mitsubishi Electric Corporation, 8-1-1, Tsukaguchi-Honmachi, Amagasaki, Hyogo 661-8661, Japan Received 10 December 2004; received in revised form 21 January 2005 Available online 10 May 2005

Abstract We have studied the degradation mechanisms of AlGaAs/InGaAs pseudomorphic HEMTs (PHEMTs) under high humidity conditions (85
C, 85% relative humidity). The degraded samples under high humidity conditions show a decrease in maximum drain current (Imax) and a positive shift in threshold voltage (Vth). Cross-sectional transmission electron microscopy (TEM) images from the deteriorated devices reveal an existence of damaged recess surface region and a peeling of a passivation film (SiNx). The secondary ion mass spectrometry (SIMS) depth profile at the interface between the passivation film and AlGaAs surface also indicates the diffusion of gallium (Ga), arsenic (As) and aluminum (Al) into the passivation film. The degradation of PHEMTs arises from mainly two mechanisms: (1) the positive shift in Vth due to stress change under the gate caused by the peeling of passivation films, and (2) the decrease in Imax due to the net carrier concentration reduction of the AlGaAs carrier supply layer caused by the combination of surface degradation at the AlGaAs recess regions and diffusion of Ga, As and Al at the interface between the passivation film and AlGaAs surface. A special treatment just prior to the deposition of SiNx films on the devices effectively suppresses the degradation of PHEMTs under high humidity conditions without degradation of the high frequency performance.  2005 Elsevier Ltd. All rights reserved.

1. Introduction In order to reduce the cost of high frequency GaAs devices, the use of non-hermetic packages has been pursued with intensity in the last few years. In order to sur-

*

Corresponding author. Tel.: +81 72 784 7239; fax: +81 72 780 2690. E-mail address: [email protected] (T. Hisaka).

vive inside non-hermetic packages, GaAs devices must exhibit resistance to humidity. The reliability of GaAs devices under high humidity conditions has been investigated [1–3], however the device degradation mechanisms under high humidity conditions are not well understood. Thicker passivation films and other polymer coatings (e.g. polyimide and benzo cyclo butane (BCB) [4,5]) on the devices without the hermetic sealing have conveniently been used as a primitive method to protect the device surface, despite the fact that these films and coatings degrade the high-frequency characteristics due to

0026-2714/$ - see front matter  2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.microrel.2005.01.017

T. Hisaka et al. / Microelectronics Reliability 45 (2005) 1894–1900

2. Samples and measurement The devices studied in this work are single-recessed AlGaAs/InGaAs PHEMTs, as shown in Fig. 1. They have 0.25 lm long and 150 lm wide T-shaped Ti/Al gates. The devices are passivated by silicon nitride (SiNx) films deposited by plasma-enhanced chemical vapor deposition (PECVD). Typical values of maximum drain current (Imax), peak transconductance (gm), and unity current gain frequency (ft) are 450 mA/mm, 600 mS/mm, and 65 GHz, respectively. The devices are mounted on packages without a cap. Unbiased humidity tests were carried out under 85
C, 85% relative humidity (85
C/85% test), and 121
C, 100% relative humidity at 2 atm (pressure cooker test, PCT). A 1000 h 85
C/85% test is standard in product qualification. We also use PCT for more accelerated test by higher temperature and humidity to shorten the test time in addition to 85
C/85% test.

3. Results and discussion 3.1. Degradation of PHEMTs under high humidity conditions Fig. 2 and Table 1 show typical DC characteristics before and after the humidity test (85
C/85% test for 1000 h). There are two changes: a positive Vth shift (+0.06 V) and a decrease in maximum drain current (Imax) around 12%. TEM images of the PHEMT before and after the humidity test are compared in Fig. 3. A denatured surface region and peeling of the passivation film are observed after the test. As reported previously in [6], under high temperature large signal operation, a damaged region at the recess surface of AlGaAs/InGaAs PHEMTs caused by an electrochemical reaction leads to the reduction of Imax. Based on this results, the Imax reduction more than 10% needs the surface damaged layer more than 10 nm, assumed that the Imax reduction was originated only by the surface damage. However the level of damage from the humidity test, as seen in the TEM image, is smaller than the expected level, and is not enough to reduce Imax by more than 10% measured in this study. 500 Imax decrease

450 400 350 Ids (mA/mm)

their parasitic capacitance. In this paper, we have investigated the physical mechanisms behind degradation of AlGaAs/InGaAs PHEMTs under high humidity conditions (85
C, 85% relative humidity). We have identified two mechanisms. One is the reduction of a drain current (Ids) due to the combination of surface degradation and diffusion of Ga, As and Al at the interface between SiNx and AlGaAs. The other is a Vth shift caused by piezoelectric effect resulting from the peeling of the SiNx film. In this paper, we present this work and we also discuss how to mitigate these degradation mechanisms.

1895

Before humidity test

300 250 200 150 After humidity test

100

Vth shift

50 0 -1.5

-1.0

-0.5 Vgs(V)

0.0

0.5

Fig. 2. Typical drain current (Ids) as a function of gate bias (Vgs) before (dashed line) and after (solid line) 85
C/85% test for 1000 h. Drain bias is 3.0 V.

Table 1 DC characteristics after 85
C/85% test for 1000 h

Fig. 1. Schematic cross-section of the AlGaAs/InGaAs PHEMT under test.

Idss [mA/mm] Imax [mA/mm] Vth [V] BVgd [V] BVgs [V] Rs [X mm] Rd [X mm]

Initial

After humidity test

200 420 0.63 7.9 7.3 0.35 0.53

180 370 0.57 7.5 6.8 0.37 0.54

1896

T. Hisaka et al. / Microelectronics Reliability 45 (2005) 1894–1900

Fig. 3. TEM photographs of device (a) before and (b) after humidity test.

In [6], we reported that the damaged recess region under large signal operation was concentrated on the drain side, which indicates that the surface degradation was strongly accelerated by electric field. This difference in the damage level may be attributed to the difference in the electrochemical reaction which is strongly dependent on the electric field. In order to survey the additional origin to explain the measured Imax reduction, we investigated the reaction between semiconductor and passivation films in more detail. The diffusion phenomena of Ga and As at the interface between dielectric films and GaAs under humidity conditions has been studied by means of secondary ion mass spectrometry (SIMS) [7]. We also used SIMS to investigate the diffusion phenomena at the interface between the AlGaAs recess surface and the passivation film of samples especially fabricated with a large recess region (50 m · 70 lm). Fig. 4 shows the SIMS depth profiles of SiNx/AlGaAs interface of samples before and after PCT. During

PCT, oxygen from the surface diffuses through the passivation film and piles up at the interface between SiNx and AlGaAs. In addition, Ga, As and Al diffuse into the SiNx film from the AlGaAs surface. Since this diffusion was not observed under high temperature storage (121
C, 96 h, nitrogen (N2) ambient, not shown in this paper), the diffusion must have been accelerated by the existence of oxygen and/or moisture. We consider that a reaction of the AlGaAs surface with oxygen or H2O causes the diffusion of III/V elements. Defects in the AlGaAs surface layer formed by the outdiffusion of III/V elements are considered to act as electron traps. Consequently, the formation of this type of defect may be responsible for the reduction of Imax in addition to the denatured surface. Regarding the Vth shift, it is difficult to attribute it to a surface degradation, since shifts in Vth are caused by the electrical changes in layers under the gate. Vth shifts after several types of accelerated tests have been investigated, and several mechanisms have been reported: gate

T. Hisaka et al. / Microelectronics Reliability 45 (2005) 1894–1900 SiNx

AlGaAs

5

GaAs

1.E+05

4

Ga +

[0 -1 -1]

+

∆Vth(%)

As Intensity (cps)

3 2

1.E+04

Si +

1.E+03

+

1.E+02

Al

+

1897

1 0 -1 -2

H

[0 -1 1]

-3 1.E+01

-4

O+

-5 0

1.E+00

0

20

40 depth (nm)

(a)

SiNx

60

AlGaAs

80

200

400

600

TIME[Hr]

Fig. 5. Vth change of devices with normal gate orientation [0 1 1] and with a 90
different orientation [0 1 1] during 85
C/85% test.

GaAs

1.E+05

Ga

+

1.E+04 Intensity (cps)

As + Si

1.E+03

O

+

+ +

+

H

1.E+02

Al

1.E+01

1.E+00

(b)

0

20

40 depth (nm)

60

80

Fig. 4. SIMS profiles of the interface between SiNx and AlGaAs (a) before and (b) after humidity test (PCT).

sponds to an increase of tensile stress of the gate [12]. To investigate the cause of the stress change, we independently measured the stress value of both the metal and passivation films deposited on whole wafers. As deposited, the gate metal (Ti/Al) has a tensile stress ( 2.6 · 109 dyn/cm2) and the passivation layer has a compressive stress (4.9 · 109 dyn/cm2). The tensile stress of the gate metal usually leads to a positive Vth shift. On the contrary, the compressive stress of the passivation leads to a negative Vth shift. Before the humidity test, these two stresses in combination tend to compensate each other. After the humidity test, the peeling of the passivation films increases the stress under the gate due to disappearance of the stress compensation effect, which consequently causes a positive Vth shift. 3.2. Supression of the degradation

metal sinking during high temperature stress [8,9], reversible Vth change due to traps under gate via electric field stress [10,11], and Vth shift by piezoelectric effect [12] due to stress change near the gate by hydrogen poisoning [13]. Gate sinking or other damage under the gate was not observed in our TEM images. The Vth shift of the degraded samples was permanent and thus was not related to an electron trapping effect. In order to investigate the presence of a piezoelectric effect, we additionally performed humidity tests on PHEMTs with a gate orientation of [0 1 1] which is 90
different from the normal gate orientation of [0 1 1]. Fig. 5 shows the Vth change of PHEMTs with a gate orientation of [0 1 1] and [0 1 1]. As one can see from this graph, the sign of the Vth shift depends on the gate orientation. These results indicate the Vth shift is caused by piezoelectric effect, due to stress change near the gate. We postulate that the Vth shift corre-

Based on these finding, suppressing the diffusion of III/V elements, reducing the tensile stress of the gate metal and improving of adhesion of the passivation film are important approaches to improve humidity tolerance. We postulate that the surface reaction between SiNx and semiconductor, and SiNx adhesion are strongly dependent on the surface condition of the semiconductor before the SiNx deposition. Thus, we investigated the impact of a pre-deposition treatment before the passivation film deposition. As reported previously [6], a pre-deposition treatment was found to effectively suppress the surface degradation under high temperature and electrical field stress. The treatment was performed in a baking furnace under H2 gas flow. To also suppress the degradation under high humidity test, we investigated an identical predeposition treatment before passivation film deposition.

1898

T. Hisaka et al. / Microelectronics Reliability 45 (2005) 1894–1900

SiNx Ion Intensity (counts)

10 10 10 10 10 10

AlGaAs

5

After test without pre-deposition treatment

4

3

Before test

2

After test with pre-deposition treatment

1

0

0

20

40

60

SiNx Ion Intensity (counts)

10

10

10

10

10

AlGaAs

4

After test without pre-deposition treatment

3

Before test

2

After test with pre-deposition treatment

1

Table 2 Gate metal stress

0

0

(b)

80

depth (nm)

(a)

20

40

60

pre-deposition treatment. The diffusion of Ga, As and Al are apparently suppressed by the treatment. The special treatment effectively reduces the reaction of the AlGaAs surface with oxygen or H2O. We also found that this treatment can reduce the stress change of the gate metal during humidity test. Table 2 shows the stress of gate metal (Ti/Al) with and without the pre-deposition treatment. The gate metal initially has small compressive stress after the deposition. The stress of the gate metal changes to a large tensile stress after the PECVD process, which is the highest temperature process (about 300
C) after gate metal evaporation. The pre-deposition treatment reduces the change of metal stress during the PECVD process. Thus we believe that the treatment suppresses the recrystallization of the gate metal. We also found that the treatment suppresses the peeling of the passivation film, as confirmed by TEM imaging of a sample with the treatment after humidity test in Fig. 7. XPS analysis showed that oxidation of Ga, As and Al were reduced after the treatment (not shown). We therefore postulate that the treatment improves the adhesion of the passivation film by reducing the oxidation of Ga and As. In addition, the reduction of the gate stress change during PECVD

80

Stress (dyn/cm2)

depth (nm)

Fig. 6. Ga and As profile at the interface between SiNx and AlGaAs of the samples with and without pre-deposition treatment after humidity test (PCT).

Fig. 6 shows the SIMS depth profile of the SiNx/AlGaAs interface after PCT, for the devices with and without the

Gate metal (as depo.) Gate metal (after PCVD process)

2.00E+08 Without pre-deposition treatment With pre-deposition treatment

2.60E+09 1.76E+09

Fig. 7. TEM photographs of device with pre-deposition treatment after humidity test (PCT). The dot lines indicate the interface between the SiNx and AlGaAs surface.

T. Hisaka et al. / Microelectronics Reliability 45 (2005) 1894–1900

1899

30

With pre-deposition

20

treatment 10

∆ Idss (%)

0 -10 -20 -30 -40

Without pre-deposition

-50

treatment -60 0

20

40

60

80

100

Time (hr)

Fig. 8. Idss change of samples with and without pre-deposition treatment during humidity test (PCT).

process when the pre-deposition treatment is applied might also affect the suppression of the peeling of the passivation film. Fig. 8 shows the change in Idss of the samples both with and without the treatment during the humidity test. It is apparent that the treatment suppresses the decrease in Idss. Fig. 9 shows Idss–Vgs curves before and

Ids (mA/mm)

400

(a)

300

Before PCT

after the humidity test. Both Vth shift and Imax change are suppressed by the treatment. Fig. 10 shows the MAG/MSG of the samples both with and without predeposition treatment. As shown in this graph, the treatment causes no harm to high frequency characteristics. This excellent stability of the devices in the DC characteristics correlated to improve adhesion of SiNx and reduce the diffusion of III/V elements from AlGaAs surface. Our work therefore demonstrates that the surface treatment constitutes an effective way to improve resistance to humidity without degrading the electrical performance of the devices.

200 After 96hr PCT

100 0 –1

30

–0.5

0 Vg (V)

25

Before PCT

200 After 96hr PCT

100 0 –1

(b)

MAG/MSG(dB)

Ids (mA/mm)

400 300

–0.5

Without predeposition treatment

0.5

0

0.5

Vg (V)

Fig. 9. Ids–Vgs curve of the samples (a) with and (b) without pre-deposition treatment before and after humidity test (96 h PCT).

With predeposition treatment

20 15 10 5 0 1

10 Frequency (GHZ)

100

Fig. 10. MAG/MSG of the samples with and without predeposition treatment (Wg = 80 lm, Vds = 2.0 V).

1900

T. Hisaka et al. / Microelectronics Reliability 45 (2005) 1894–1900

4. Conclusions In summary, the degradation of AlGaAs/InGaAs PHEMTs with SiNx passivation films under high humidity condition is mostly caused by two mechanisms: (1) a Vth shift is due to stress change under the gate, and (2) a decrease in Imax is due to a combination of surface degradation at AlGaAs recess region and the diffusion of III/V elements at the interface between SiNx and AlGaAs. A pre-deposition treatment just prior to SiNx deposition effectively suppresses the degradation of PHEMTs under high humidity conditions without using a thicker passivation film or other polymer coatings. Acknowledgements The authors wish to thank Ms. A. A. Villanueva and Prof. J. A. del Alamo of MIT, and Dr. Sonoda for helpful discussions. We also wish to thank Mr. Takesue for humidity test and Mr. Shiga for performing the experiments and for useful discussions. References [1] Magistrali F, Ogliari D, Sangalli M, Vanzi M. Degradation mechanism of GaAs MESFET devices in high humidity conditions, ISTFA89. p. 141–51. [2] Roesch WJ, Winters RA, Rubalcava AL, Ingle B. Humidity resistance of GaAs ICs, GaAs IC Symposium 1994. p. 251–5. [3] Ersland P, Jen H-R, Yang X. Lifetime acceleration model for HAST tests of a pHEMT process, GaAs Reliability Workshop 2003. p. 3–6.

[4] Sugitani S, Ishii T, Tokumitsu M. Three-dimensional interconnect with excellent moisture resistance for MMICs, Technical report of IEICE ED2001-200, 2002. p. 19–26. [5] Kaleta T, Varmazis C, Chinoy P, Carney JP, Jansen N, Loboda M. A two layer hermetic-like coating process for on-wafer encapsulation of GaAs MMICÕs, GaAs IC Symposium, 1995. p. 128–31. [6] Hisaka T, Nogami Y, Sasaki H, Hasuike A, Yoshida N, Hayashi K, Sonoda T, Villanueva AA, del Alamo JA. Degradation mechanism of PHEMT under large signal operation, GaAs IC Symposium. 2003. p. 67–70. [7] Shiramizu T, Tanimura J, Kurokawa H, Sasaki H, Abe S. Diffusion phenomena at the interface between dielectric films and compound semiconductors, Fourth International Symposium on Atomic Level Characterizations for New Materials and Devices, 2003. p. P1–30. [8] Canali C, Castaldo F, Fantini F, Ogliari D, Umena L, Zanoni E. Gate metallization ‘‘Sinking,’’ into the active channel in Ti/W/Au metallized power MESFETÕs. Electron Dev Lett 1986;EDL-7(3):185–7. [9] Chou YC, Grundbacher R, Leung D, Lai R, Liu PH, Kan Q, et al. Physical identification of Gate metal interdiffusion in GaAs PHEMTs. Electron Dev Lett 2004;25(2): 64–6. [10] Villanueva AA, del Alamo JA, Hisaka T, Hayashi K. Electrical degradation mechanisms of RF power GaAs PHEMTs. IEDM 2003:30.4.1–4. [11] Meneghesso G, Canali C, Cova P, De Bortoli E, Zanoni E. Trapped charge modulation: a new cause of instability in AlGaAs/InGaAs pseudomorphic HEMTÕs. Electron Dev Lett 1996;17:232–4. [12] Asbeck PM, Lee CP, Chang MF. Piezoelectric effects in GaAs FETÕs and their role in orientation-dependent device characteristics. Trans Electron Dev 1984;ED-31(10): 1377–80. [13] Alamo JA, Blanchard RR, Mertens SD. Hydrogen degradation of InP HEMTs and GaAs PHEMTs. IEICE Trans. Electron 2001;E84-C(10):1289–93.