Molecular beam epitaxy regrowth and device performance of GaAs-based pseudomorphic high electron mobility transistors using a thin indium passivation layer

J O U R N A L O F M AT E R I A L S S C IE N C E: M ATE R IA LS I N EL E C T RO N I C S 1 1 (2 00 0 ) 48 3± 48 7

Molecular beam epitaxy regrowth and device

performance of GaAs-based pseudomorphic high

electron mobility transistors using a thin indium

passivation layer

SHEU-SHUNG CHEN, CHIEN-CHENG LIN

Department of Materials Science and Engineering, National Chiao Tung University,

Hsinchu 300, Taiwan

E-mail: chienlin@cc.nctu.edu.tw

CHIN-KUN PENG

Procomp Informatics Ltd., Hsinchu 300, Taiwan

YI-JEN CHAN

Department of Electrical Engineering, National Central University, Chungli 32054, Taiwan

Using the technique of molecular beam epitaxy, an indium passivation layer as thin as

several tens of

A was implemented to protect underlying III-V epilayers from carbon and

oxygen contamination. After the subsequent desorption of the passivation layer,

GaAs-based pseudomorphic high electron mobility transistors (PHEMTs) were regrown. Negligible

residual carriers were detected at the interface between the regrown PHEMTs and the

underlying layer, resulting in a superior performance. The regrown PHEMTs with a

16100 mm

_{gate demonstrated an extrinsic transconductance g}

me

as high as 330 mS mm

.

Microwave measurements showed that the current gain cut-off frequency f

was 26.5 GHz

and the maximum oscillation frequency f

was up to 48 GHz. A small-signal equivalent

circuit model of the regrown PHEMTs was also evaluated.

# 2000 Kluwer Academic Publishers

1. Introduction

Molecular beam epitaxy (MBE) is a powerful technique in the development of III-V compound semiconductor devices. Among these devices, optoelectronics have attracted a great deal of attention due to their extensive commercial applications. As progress continues, the increasing demands for more multi-functional circuits create an overload in device fabrication. While direct epitaxial growth cannot meet these demands, the regrowth technique becomes an alternative to facilitate the device process and make possible a further improvement in performance. There are several methods for the regrowth of III-V epilayers, including arsenic passivation at low substrate temperatures [1±3], sulfur-based chemical treatments [4±6], and patterned regrowth [7, 8], etc. Each meets certain levels of success and de®ciency. All these studies focused on retaining an interface of high-quality between the originally grown and regrown layers. The high purity at the interface is bene®cial to the regrowth technique. A clean defect-free interface is ideal for the reliability of the regrown devices. Such a regrowth technique can relieve the growth burdens of complex multifunctional opto-electronic devices on the same wafer and improve the ¯exibility in device fabrication.

To develop a simple regrowth technique without any modi®cation of the commercial MBE system is of great interest. Although previous studies indicated that the thin arsenic layers, condensed in situ below room tempera-ture, could lead to good cleanliness at the interface, there were still some limitations to the rapid cooling of the substrate in a modern MBE system design. To obtain a good regrown sample, the passivation layer is crucial in two respects. First, it must be thick enough to prevent the residual carbon and oxygen species from contaminating the underlying epilayers after exposure to the atmosphere for further processing. Secondly, it should be thin enough to be easily removed. Based on these criteria, indium was considered a suitable passivation source, which could be condensed above room temperature.

In the present study, pseudomorphic high electron mobility transistors (PHEMTs) were regrown in situ in the MBE chamber after thermally desorbing the indium passivation layer, which was previously deposited on the buffer layer of the GaAs substrate. The potential contamination of carbon and oxygen from air at the interface was examined using X-ray photoemission spectroscopy (XPS). The C±V depth pro®ler was used to determine the carrier concentration at the interface between the regrown PHEMTs and the underlying layer.

The d.c. and microwave characteristics were examined to evaluate the performance of the regrown PHEMTs. Finally, a small-signal equivalent circuit model of the regrown PHEMTs was simulated.

2. Experimental

Fig. 1 illustrates the regrowth sequences for the PHEMT device using a MBE system (Model 32P, Riber). The epi-ready semi-insulating (100) GaAs substrate was loaded into the MBE system. The water vapor was desorbed at 350_{C in the transfer chamber, followed by a normal}

degassing in the growth chamber from a preset temperature (300_{C) to 630}_{C with a slow heating}

rate. After growing a 0.5 mm-thick GaAs buffer layer at 580_{C under an As-stabilized condition, gallium and}

arsenic sources were abruptly shuttered to terminate the growth and the substrate temperature was lowered to 50_{C for the further deposition of a thin indium layer, as}

shown in Fig. 1a. The In-passivated sample was then removed from the MBE system and preserved in air for weeks before being reloaded for regrowth. Fig. 1b represents the thermal desorption of the indium

passivation layer at 630_{C for 30 min. During the}

thermal desorption process, indium, carbon and oxygen were monitored using a residual gas analyzer (RGA). When none of these species were able to be detected the clear GaAs(100) ÿ (2 6 4) surface reconstruction was obtained under an As-stabilized condition monitored by re¯ection high energy electron diffraction (RHEED), indicating that the indium layer was completely removed with no contamination of carbon and oxygen in the underlying layers. Fig. 1c reveals that the regrown PHEMTs consist of a 0.5 mm undoped GaAs buffer layer, a 15 nm undoped In_0:15Ga_0:85As pseudomorphic channel layer, a 3 nm undoped Al_0:25Ga_0:75As spacer layer, a

50 nm Si-doped Al_0:25Ga_0:75As gate layer with

261018_cmÿ3 _{doping concentration, and a 20 nm}

Si-doped GaAs cap layer with 361018_cmÿ3 _doping

concentration. The PHEMTs were regrown at 580_C,

except that the In_0:15Ga_0:85As channel layer was processed at 540_{C. The regrown PHEMTs had a typical}

oval-defect density ranging from 102_{to 10}5_cmÿ2_{and the}

featureless specular surface was similar to that of the traditional process with no passivation layer. Mesa etching was then performed using a phosphoric-based etchant, and the source and drain ohmic metal patterns were de®ned photo-lithographically, followed by the evaporation and lift-off of AuGe/Ni/Au metals. The Al gate was formed after the gate recess process.

The surface compositional pro®les of the In-passivated sample were measured using electron spectroscopy for chemical analysis (ESCA: model PHI-5400, Perkin-Elmer). The MgK_aX-ray source was operated at 300 W and 15 keV, while the specimen was sputtered by argon ions. The C±V pro®les were plotted using an electro-chemical system (Model PN-4300, Bio-Rad) to determine the carrier concentration at the interface between the original buffer layer and the regrown PHEMTs.

Based on the Hall effect, with a van der Pauw con®guration, the carrier concentration and the electron

mobility in the regrown PHEMTs with 465 mm2

dimensions were measured at a magnetic ®eld of 0.5 T and an electric current of 100 mA. The device d.c. characteristics with a 1650 mm2_{gate and a source±drain}

spacing of 3 mm were explored using a semiconductor parameter analyzer (Model 4145B, Hewlett Packard). The microwave scattering parameters (S-parameter) were measured in the range of 0.45 to 26.5 GHz with a microwave probe station (CASCADE) and an automatic network analyzer (Model 8510C, Hewlett Packard). A small-signal equivalent circuit was also simulated based on the measured S-parameters with a microwave design system (MDS: Hewlett Packard).

3. Results and discussion

The electron energies of As_3d, Ga_2p, In_3d, O_1s, and C_1s core levels were monitored by XPS analyzes. Fig. 2

Figure 1 Regrowth sequences of PHEMTs. Both the In deposition and desorption were completed in the MBE growth chamber.

Figure 2 XPS spectra near the As3dcore energy for the surfaces of the

as-grown In-passivated sample (m) and that after being sputtered for 3 min (solid line).

shows the XPS spectra near the As3dcore energy for the

surfaces of the as-grown In-passivated sample and the one after being sputtered for 3 min, respectively. The oxide peak almost completely disappeared after being sputtered for 3 min. Details of the resultant depth pro®les are listed in Table I. It is shown that carbon and oxygen existed only near the outermost surface. Both elements were effectively prevented from contaminating the underlying epilayers due to the indium layer. The

simultaneous appearance of In_3d and Ga_2p signals

indicates that the indium layer was not as smooth as expected. The early appearance of the As_3d signal was probably caused by a high arsenic background during the MBE growth of the indium layer.

Fig. 3 shows the C±V depth pro®le of the regrown PHEMTs with a total thickness of 0.6 mm, which covered the regrowth layers. Near the outermost surface layer, the n-type carrier concentration was 361018_cmÿ3_{, which}

was very close to the expected value. The concentration then rapidly decreased to as low as 261014_cmÿ3 _{at the}

interface between the original buffer layer and the regrown PHEMTs, similar to the concentration in the semi-insulating substrate, indicating that no residual carriers existed at the interface. This C±V depth pro®le shows no perceivable difference from that of conven-tional PHEMTs. Both XPS and C±V results reveal that a high-quality interface was preserved by the indium passivation layer.

Van der Pauw Hall measurements on the regrown samples showed two-dimensional electron concentra-tions of 4:461012 _{and 4:0610}12_cmÿ2_{, and electron}

mobilities of 4300 and 15 400 cm2_Vÿ1_sÿ1 _{at room}

temperature and 77 K, respectively. These values are consistent with those of the traditionally grown PHEMTs. The high carrier concentration and good electron mobility ensure the superior performance of PHEMTs as described later.

Fig. 4a shows the d.c. characteristics for regrown PHEMTs with a 1:0650 mm2_{gate at 300 K, indicating a}

good pinch-off at a gate bias of ÿ 1:0 V. The extrinsic transconductance gme of 330 mS mmÿ1 was obtained at

Vdsˆ 1:75 V and Idsˆ 136 mA mmÿ1, as shown in Fig.

4b. With a measured source resistance Rs of 0.2 O mm,

the estimated intrinsic transconductance gmi was

353 mS mmÿ1_{. These results were better than those of}

the conventional PHEMTs mentioned in a previous study [9].

Microwave S-parameters for the devices with a 16100 mm2 _{gate were measured at frequencies ranging}

from 0.45 to 26.5 GHz biased at V_dsˆ 2:0 V and

V_gsˆ ÿ 0:25 V. Fig. 5 displays the maximum available gain (MAG), maximum stable gain (MSG), and current gain jh₂₁j2, calculated from the measured S-parameters. The current gain cut-off frequency f_t was estimated at

T A B L E I XPS depth pro®les of the In-passivated samples with a 0.5 mm-thick GaAs buffer layer.

Sputter time (min) C1s O1s Ga2p3 As3d In3d5

0 21.3 30.5 Ð 14.5 33.6 1 Ð 9.5 28.7 26.2 35.5 3 Ð Ð 62.3 31 6.6 8 Ð Ð 66.7 31.3 1.8 10 Ð Ð 67.9 31.5 0.5 15 Ð Ð 69.3 30.6 Ð 25 Ð Ð 69.3 30.6 Ð

Figure 3 C±V depth pro®les of the regrown PHEMTs.

(a)

(b)

Figure 4 D.c. I±V characteristics of the regrown PHEMT device with a 1.0 mm gate length at 300 K. (a) Idsÿ Vds; (b) Idsÿ gmeÿ Vgs.

26.5 GHz by the unity gain intercept of the jh21j2values.

The maximum oscillation frequency fmax was about

48 GHz from the unity gain intercept of the MSG/MAG curves using a rate of 6 dB per octave roll-off.

Fig. 6 shows the small-signal equivalent circuit model of the regrown PHEMTs using the MDS to ®t the measured S-parameters. Fig. 7 indicates an excellent agreement between the simulated and measured S-parameters ranging from 0.45 to 26.5 GHz. Table II lists the ®nal element values for the equivalent circuit in

Fig. 6. The simulated g_mi …334 mS mmÿ1_{† and R}

s

(0.1996 O mm) were consistent with the results obtained from the d.c. measurements (Fig. 4). The input capacitance C_gss …1:97 pF mmÿ1_{† was small enough to}

ensure a better intrinsic transconductance g_mi and

consequently improved the current gain cut-off

fre-quency ft. Moreover, the feedback capacitance

Cgd …0:25 pF mmÿ1† was so small that this device

should be able to operate more stably. Meanwhile, the output conductance g_o…ˆ 1=R_ds† was 14:02 mS mmÿ1_,

resulting in a gain ratio g_mi=g_o of 23.8. Based on this small-signal equivalent circuit model, the microwave devices and other related sub-systems, including the matching network of this regrown PHEMT, could be better designed and correctly applied.

4. Conclusions

Using a thin indium layer for surface passivation, the regrowth of PHEMTs was carried out in situ without any modi®cation of a commercial MBE system. XPS analyzes showed that this thin indium layer was thick enough to protect the underlying epilayers from carbon and oxygen contamination during exposure to the air. The C±V-depth pro®les revealed that no residual carriers were distributed at the interface between the original buffer layers and the regrown PHEMTs, resulting in high performance of regrown PHEMTs.

Figure 6 A typical small-signal equivalent circuit model of the regrown PHEMT device simulated by the HP MDS.

Figure 7 The measured S-parameters (solid line) and simulated data (n) from the small-signal equivalent circuit of the regrown PHEMT device ranging from 0.45 to 26.5 GHz. In the upper polar chart, the full scales of coordinate for S12 and S21 are one and ®ve (arbitrary units), respectively. In the lower Smith chart, both S11 and S22 have the same co-ordinate scale equal to one (arbitrary unit).

Figure 5 Microwave characteristics of the regrown PHEMT device with a 1.0 mm-long gate at the frequencies ranging from 0.45 to 26.5 GHz at 300 K.

T A B L E I I The ®nal values of small-signal equivalent circuit elements for the regrown PHEMTs with a 16100 mm2_gate

Equivalent circuit element Final value

Gate resistance Rs…O† 6.747

Source resistance Rd…O† 1.996

Drain resistance Rd…O† 6.011

Output resistance Rds…O† 713.1

Intrinsic input resistance Rd…O† 8.936

Input capacitance Cgds…pF† 0.197

Feedback capacitance Cgd…pF† 0.025

Output capacitance Cds…pF† 0.014

Input inductance Linp…nH† 0.058

Source inductance Ld…nH† 0.00357

Output inductance Lout…nH† 0.0808

t …ps† 1.26

The regrown devices exhibited an extrinsic

transcon-ductance g_me as high as 330 mS mmÿ1_{. Microwave}

measurements showed that the current gain cut-off frequency f_twas 26.5 GHz and the maximum oscillation

frequency f_max was up to 48 GHz. The S-parameters

obtained from the simulation of a small-signal equiva-lent circuit model were consistent with the measured ones. This simulation is essential to the design of microwave devices of the regrown PHEMTs. Based on the excellent transport and device characteristics, the In-passivated technique could result in an ideal interface for the regrown PHEMTs, applicable to a variety of different device structures.

Acknowledgment

The authors thank the ®nancial support by the National Science Council, Taiwan, under Grant No. NSC 87-2216-E009-014.

References

1. S . P. KOWA L C Z Y K, D . L . M I L L E R, J . R . WA L D RO P, P. G . N E W M A NandR . W. G R A N T, J. Vac. Sci. Technol. 19 (1981) 255. 2. D . L . M I L L E R, R . T. C H E N, K . E L L I OT T and S . P.

KOWA L C Z Y K, J. Appl. Phys. 57 (1985) 1922.

3. P. E T I E N N E,P. A L N OT,J . F. RO C H E T T E andJ . M A S S I E S, J. Vac. Sci. Technol. B 4 (1986) 1301.

4. C . J . S A N D RO F F,R . N . N OT E N B U R G,J . C . B I S C H O F FandR . B H AT, Appl. Phys. Lett. 51 (1987) 33.

5. Y. N A N N I C H I,J . FA N,H . O I G AWAandA . KO M A, Jpn. J. Appl. Phys. 27 (1988) 2367.

6. J . WA N G,T. C H E N G,P. H . B E TO N,J . J . H A R R I S,E . B OW S E R andC . T. FO X O N, Semicond. Sci. Technol. 8 (1993) 2101 7. A . Y. C H OandW. C . B A L L A M Y, J. Appl. Phys. 46 (1975) 783. 8. Y. C . PAO,J . F R A N K L I N andC . Y U E N, J. Cryst. Growth 127

(1993) 892.

9. C . K . P E N G,W. H . L A N,S . L . T U,S . J . YA N G,S . S . C H E Nand C . C . L I N, Mater Chem. Phys. 45 (1996) 92.

Received 7 January