AlGaAs modulation-doped FET's with reduced low-frequency noise and thermally stable performance

Journal of Crystal Growth 81 (1987) 359—367 North-Holland, Amsterdam

359

PSEUDOMORPHIC InGaAs/AIGaAs MODULATION-DOPED FEPS WITH REDUCED LOW-FREQUENCY NOISE AND THERMALLY STABLE PERFORMANCE * Shih-Ming LIU and M.B. DAS Electrical Engineering Department, Solid State Device Laboratory and Materials Research Laboratory, The Pennsylvania State University, University Park, Pennsylvania 16802, USA

and C.K. PENG, J. KLEM, T. HENDERSON, W. KOPP and H. MORKO~ Coordinated Science Laboratory, University of Illinois, Urbana, Illinois 61801, USA

A high-performance MODFET structure grown by MBE with the incorporation of a single quantum well In

0 15Ga0 85As layer for the transport of 2D electron gas has been critically examined for its thermal stability at 80 K and low-frequency noise from 10 2 to 108 Hz. Experimental results indicate that the behavior of this device in both these respects is much 2/ssuperior at 80 K when and compared an averagewith carrier the same behavior saturation velocity of conventional of 2 x i0~cm/s MODFET’s. at 300 K A in maximum a 1 ~m gate low-field device carrier clearly mobility indicate that of 29,000 the quality cm of the pseudomorphic quantum well (InGaAs) layer is either comparable or better than that of the usual GaAs buffer layer. The deep level spectra, obtained through photo-FET measurements, and the low-frequency noise spectra at different temperatures obtained for the new pseudomorphic and conventional MODFET’s have clearly indicated that contributions from various deep levels present in the new structure are significantly reduced.

1. Introduction Since its introduction [1,2] as the most attractive amplifying device for both high-frequency analog and high-speed switching applications, the modulation-doped field-effect transistor (MODFET), based on silicon doped A1GaAs on Undoped GaAs buffer, has received considerable attention both for its high performance [3] at 300 K and even superior performance [4] at 77 K under conditions of white-light illumination. Unfortunately, without illumination the device fails to operate at temperatures below 140 K due to hysteresis and even I— V collapse due to a deep level identified as DX centers [5]. The concentration of

DX centers in A1GaAs is a strong function of Al mole fraction x, and it increases rapidly when x exceeds 0.2 and decreases significantly when x 0.15 for heavily Si-doped (> 1018 cm 3) materials [6]. The idea of using a strain-layer InGaAs/GaAs single quantum well was introduced by Rosenberg et al. [7] and this was followed by the AlGaAs/InGaAs/GaAs high transconductance MODFET [8] with stable cryogenic performance. Measurements of low-frequency noise in these devices also mdicated significant reduction of generation—recombination (G—R) type noise sources [9]. The incorporation of InGaAs (undoped) between the AlGaAs and GaAs is capable of forming a desired amount of the conduction band discontinuity even at low mole fractions of In and Al (x 0.15), and fortunately, this is also the appropriate mole fraction necessary to reduce the DX centers in Al—

*

The work done at The Pennsylvania State University in cooperation with GE Electronics Laboratory, Syracuse, N’i’~ was supported by the National Science Foundation under Grant No. EC85-03984. The work done at the University of Illinois was supported by the Air Force Office of Scientific Research and NASA.

.

GaAs. The single quantum-well itself is also a desirable feature to have for the reduction of saturation output (drain) conductance.

0022-0248/87/$03.50 © Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

360

S.M. Liu et aL

/

Pseudomorphic InGaAs/A1GaAs MODFET’s

Initial results reported [8—10]on these devices indicated many attractive features, among them, high transconductance, high transconductance-tooutput conductance ratio, low G—R noise and high-frequency noise, and stable device operation at cryogenic temperatures without white-light illumination are to be noted. In this paper the results of further investigations are presented that particularly probe into the nature of the device thermal stability, the low-frequency noise and the quality of the pseudomorphic InGaAs quantum well,

2. Device structure and fabrication

tance versus gate voltage when the device was biased below its threshold voltage at 300 K and the temperature was lowered to 80 K under the same bias stress. This approach eliminated the possibility of electron trapping, and hence the threshold voltage (VTH) remained near that obtamable at 300 K. As the gate bias was allowed to increase the conductance curve (indicated by arrow going up), approached towards the curve obtained under positive bias stress condition (indicated by arrow going down). The net shift of VTH as a result of the positive and negative gate bias stress can be interpreted as a measure of an effective trap concentration of the n-AlGaAs layer under the gate electrode when its thickness is known. Calculations for the test samples with x 0.3, 0.23 and 0.2; with AlGaAs layer thickness of 380 A and a permittivity value of 1.1 pF/cm, give the trap (DX center) concentration values of 8.5 x iO’~,3.9 X iO’~,and 1.1 x 1017 cm3, respectively. The pseudomorphic SQW device, on the other hand (see fig. id), shows negligible threshold voltage shift under the bias stress conditions, this =

The pseudomorphic single-quantum-well (SQW) AlGaAs/InGaAs structures studied were grown by MBE on semi-insulating GaAs substrates. The structure begins with a ten-period 20 A/50 A AlAs/GaAs superlattice. This was followed by an undoped 1 ~im thick GaAs layer. Above this buffer layer were grown the following layers in the order indicated: a 200 A In 0 15Ga085 As layer, a 30 A Al0 15G085AS spacer 3) A1layer, a 350 A Si-doped (n =2 x 1018 cm 015Ga085As layer, and a 200 A Si-doped GaAs cap layer. The top layer provided good ohmic contacts for the source and the drain, Field-effect transistors were fabricated by defifling and etching of mesa isolation patterns using standard UV lithographic and etching techniques.

3. The effects of bias stress at 80 K Initially, a postive gate bias voltage stress was applied at 300 K keeping the source and drain voltage nearly at zero, such that there would be neutral n-AlGaAs region under the gate electrode. The temperature was lowered to 80 K under this bias stress condition and the channel conductance was measured. These tests were performed using several conventional A1GaAs/GaAs MODFET’s with different Al mole-fraction (x) values, and the new AlGaAs/InGaAs pseudomorphic MODFET’s and the results are presented in figs. ia—id. The data presented also include the channel conduc-

—

clearly suggests a dramatic reduction of trapping due to DX centers as would be expected for the low Al mole-fraction value used. In order to identify whether there are any effects of other possible traps that may be present either in AlGaAs or in the InGaAs layer, the drain I/V characteristics of the new MODFET’s were measured at 300, 80, and 12 K and they are presented in fig. 2.

4. Photo-FET deep level spectra For a rapid determination of the relative importance of the deep level trap distributions in the conventional A1GaA5/GaAs MODFET’s and that in the A1GaA5/InGaAs MODFET’s, the photo-FET method [11] of measurement was used. The change in the drain current (below the saturation region) with nearly full 2D electron gas channel was monitored using a lock-in technique when the device was subjected to a monochromatic source of light excitation from wavelength of 3.2 to 1.3 ~sm. The experimental arrangement was similar to that of Heuken et al. reported recently [12].

S.M. Liu eta!.

/

Pseudomorphic InGaAs/AIGaA5 MODFET’s

361

200

-240 80 K

/:$m;

23~:530~O)

V~ CV)

V

6~ (V)

23720

__

V~ IV)

V;s

CV)

Fig. 1. Threshold voltage shift indicated by the shift of channel conductance (Gd~)versus gate voltage curves after gate bias stress at 80 K for conventional AIGaAs/GaAs device with x = 0.3, 0.23, and 0.2. The pseudomorphic MODFET (device 2329*1) shows negligible threshold voltage shift.

The activation energies of deep levels can be deduced by observing the threshold or transitions in the ~ ‘DS versus photon energy plots as presented in fig. 3. An increase in ‘~‘DS implies a donor-like level (D) and a decrease indicates an acceptor-like level (A). The existence of a strong donor emission starting from 0.41 eV and extending up to 0.55 eV, in the high Al mole-fraction sample (x 0.35) (curve a), is most likely due to =

the DX centers. The emission spectra of the device with x 0.25 (curve b) in this range of energies is much weaker and divided into several identifiable discrete levels and is extended only up to 0.48 eV. In the case of AlGaAs/InGaAs structure with low Al and In mole fractions (x 0.15 in both cases) (curve c), there are several discrete identifiable levels from 0.413 to 0.46 eV in contrast to the continuous contribution of the DX centers in curve =

=

/

S.M. Liu eta!.

362

Pseudomorphic InGaAs/AIGaA5 MODFET’s

5 ~1

I

8ZK

—

I

I

I

I

~

~1ZK

.3.

40~

I

..‘

.3

‘

GS

~30 -~ E

—.‘

300K

. -.

‘~.‘

.2

.

.2 —

.1

—

// /.vV

10-

.1

~

/~_

~~~0~

0 --

I

I

1.0

I

.~

(v)

~~‘l)S

Fig. 2. The drain I/V characteristics of a pseudomorphic MODFET (device 12329 * 1) at 300, 82 and 12 K.

I

+

I

I

I

I

-~4-~--~ /

-

/

~//

/

\xY }//I/ ./+

// / xl

I)

n

.855

.785

I

1.4

I

A\\

.61 \i

II

A

/

(/

‘~‘ \ \ .582

-11) ~

—,

x4 x4 /

//

c

.550

A $ I) 4

A I.419 .413 I) I [l~A

4

I) I

1.8

I

I

2.2 Wavelength

Eci~

U —

.437

.465

2.6

I

I

3.0

(sm)

Fig. 3. Photo-FET deep level spectra of MODFET’s (D = donor, A = acceptor type traps, and indicated activation energies are in eV). Curve a: A1GaA5/GaAs device with x = 0.35; VGS = 0.3 V, VDS = 0.4 V, and ~ = 19.5 mA. Curve b: A1GaAs/GaAs device with x 0.25; VDS = 0.3 V, V 0~= 0.5 V, and ‘Ds = 16 mA. Curve C: Pseudomorphic InGaAs device with x = 0.15; VGS = 0.25 V, VDS = 0.3 V, and I~s= 18 mA.

SM. Liu eta!.

/ Pseudomorphic InGaAs/A lGaAs MODFET’s

a from 0.41 to 0.55 eV. This is a strong indication of the elimination of the DX centers in the new structure.

4. Low-frequency noise spectra In MODFET’s and MESFET’s there are two primary sources of low-frequency noise [13], namely, the generation—recombination or G—R noise due to a variety of deep level traps and a background 1/f noise, possibly due to the very nature of the crystalline structure of the material involved. Recently it has been suggested that the 1/f noise originates from the atoms of crystal as quantum radiation [14], however, no direct relationship between the device parameter and the noise spectral density has been developed except for an empirical parameter called Hooge’s parameter [15]. However, theoretically calculated values of Hooge’s parameter have not yet been experi-

363

mentally validated using presently available devices. Hence, we have taken a different approach in identifying the low-frequency noise generators (under current saturation conditions of operation) by concentrating on the measurement of equivalent gate input noise noise voltage representing both 1/f an G—R noise components as discussed elsewhere [9]. In detailed equivalent gate noise voltage spectra are presented in fig. 4, showing behavior of a conventional MODFET and that of two InGaAs SQW MODFET samples. A large noise bulge in the vicinity of 10 kHz in the conventional device (curve a) is due to the DX center. By observing the movement of this noise bulge along the f axis at different temperatures, we were able to characterize the related deep level trap [17] with activation energy, E~ 0.41 eV. The low-frequency noise level in the InGaAs SQW FET structure, as shown in fig. 5 (curves b, c, and d), is the lowest we have observed in any comparable size MOD—

—3 _____________________________________________________________________________

10

I

I

I

I

I

I

I

I

I

300K

IN

102

100

102 lrc(IIICI1CV

106

( H7.)

Fig. 4. Low-frequency noise spectra of MODFET’s. Curve a: A1GaA5/GaAs device with x = 0.3, VG 5 = 0 V, ~ = 1.4 V, and I~s= 42 mA. Curve b: Pseudomorphic InGaAs device (2329*3); V0~ = 0.4 V, ~ = 0.9 V, and ‘DS = 34 mA. Curve C: Pseudomorphic InGaAs device (2329* 1); VGS = 0.3 V, ~ = 0.9 V, and Irs = 33.5 mA. Curve d: Device (2329* 1), VGS = 0.2 V, VDS = 0.45 V, and ‘Ds = 24 mA.

108

/ Pseudomorphic InGaAs/AlGaAs MODFET’s

SM. Liu eta!.

364

1°

I

1o

I

~

id11

~

i

ió~2

-13

10 101

-

i

__

-

~

102

i04 freq.

io~

(liz)

Fig. 5. Low-frequency noise spectra of pseudomorphic device (2329*3) at different temperatures. Curve a: 246 K; curve b: 217 K; curve C: 180 K. Curve d: 157 K, and curve e: 117 K, bias points in all cases were VGS = 0.3 V, VDS = 0.7 V, and ‘DS = 26 mA (‘~~ for curve e was 28 mA).

FET and MESFET devices. The appearance of an extended i/f noise region in the vicinity of 1 Hz is a direct indication that in this region the device’s true 1/f noise is dominant compared to any G—R noise contributions. A bulge near 250 Hz and a smeared out spectral behavior of noise up to 10 MHz following nearly i/f noise slope clearly indicate that the G—R noise producing deep level traps are present only at moderate concentrations. Due to the involvement of multiple deep levels and their low concentrations, sharp peaks were not observed in the noise spectra even when its magnitude was multiplied by the frequency (f), as depicted in fig. 5. For clarity, results obtained only at selected temperatures are presented. These data can be used by focusing on the movement of a specific spectral tail along the same meansquared noise voltage level as indicated by horizontal lines crossing the various curves in fig. 5. The results gave two trap levels at E~ 0.48 eV and E~ 0.266 eV. The 82 K noise spectra in fig. 4 is seen to indicate a near ideal 1/f noise. Thus, —

—

because of a general reduction of the deep level trap concentrations in the new InGaAs SQW FET structure, that involves a low In and Al mole fractions, it appears that the device’s low-frequency noise performance has improved by more than an order of magnitude.

5. Carrier mobility and saturation velocity It has been reported elsewhere [18] that determination of carrier mobility in the 2D electron gas in short gate-length MODFET’s is possible by directly eliminating the parasitic resistance effect. This method is based on an open-circuit measurement of the channel perturbation voltage (vdS) when the gate is excited by an AC signal voltage (v~~) and the device is current-source biased (Jo) in its linear I/V characteristic operating region. Results of such measurements carried out on a InGaAs SQW device at 300 and 80 K are presented in fig. 6. The mobility variation with

SM. Liu eta!.

/ Pseudomorphic InGaAs/A !GaAs MODFETs

I

(pF) Cg1.0 1.5 .5 C -~- .4

I

~

I

~-~i30~K 0

.2

V

3~

365

I

30,000

)JBOKI 300

.2

\ ~) Cg \ I 80K

I

.4

20,000 10.000 0

(V)

Fig. 6. Low-field channel carrier mobility and gate capacitance dependence on gate bias voltages at 300 and 80 K for a pseudomorphic MODFET (device 2329*1).

gate bias voltage is typical of MODFET structure, As the mole fraction of In is low (x = 0.15), the mobility of the 2d electrons in the SQW is not significantly different from that in undoped GaAs as would be expected [19]. The carrier saturation velocity, ~‘sat’ is another

40

I

I

I

important material parameter that critically determines the high frequency performance of MODFET’s, Since this parameter in short-length InGaAs samples can have high values [20], it was considered important to determine the parameter from direct measurement through analytical con-

F

I

I

I

V(,s ~

2329 #1

Vl)S

(~ol:r)

Fig. 7. Measured and calculated drain I/V characteristics of a pseudomorphic MODFET (device 2329*1) at 300 K.

5. M. Liu et a!. / Pseudomorphic InGaAs/A lGaAs MODFET’s

366

siderations. Results of such measurements and calculations are presented in fig. 7. These calculations were performed using a three-section model for the channel. The model effectively matches the measured ‘Ds/ VDS characteristics by utilizing the intrinsic device channel resistance (rCH) data as obtained in course of determination of p.s. In the first section of the channel near the source, Peff was approximated as ILeff

=

(1)

~

As the field ~ reaches a high value ~ at which velocity overshoot occurs, and the second section of the channel begins. In this region the model assumes a constant ~ even though the field increases monotonically. When the field reaches a value of E~ the space-charge-like (SCL) third region begins. The following approximation [21] was used for the calculation of voltage across the SCL regions: sinhi—~), I\ iTL 2a /\

~

(2)

where L~is the length of the SCL region, and a is the effective distance of the mobile electron charge sheet in the same region from the gate electrode. The calculations for a given drain I/V curve were repeated until consistent values for L and a could be obtained. This self consistency of data demonstrates the validity of our approach. An average carrier saturation velocity (Vsat) can be obtained from the calculated distribution for a given current. The estimated value of 0sat was 2 X iO~cm/s. ~,

—

6. Conclusions The modulation-doped AlGaAs/ InGaAs/ GaAs pseudomorphic heterostructure with low Al and In mole fractions has been demonstrated to provide thermally stable device I/V characteristics and a reduced level of low-frequency noise which is at least an order of magnitude below that of conventional A1GaA5/ GaAs MODFET’s. The thermal stability of the I/V characteristics of this device at 80 K has been found to be excellent, and

even at 12 K the device continues to operate with some degradation of the current saturation characteristics (possibly due to ohmic contact degradation). By photo-FET measurements, deep level spectra of the device were obtained and compared with that of conventional MODFET’s, These results have clearly indicated that the effect of DX centers can indeed be greatly reduced in a pseudomorphic MODFET structure so that the device may be free from the I/V collapse and other undesirable behavior. Acknowledgements The authors are grateful to Dr. Sander Weinreb for providing a cryogenic system for measurements at 12 K. They are also indebted to Drs. P.C. Chao and K.H.G. Duh for useful discussions concerning MODFET structures.

References [1] T. Mimura, K. Josbin, S. Hiyamizu, K. Hikosaka and M. Japan. J. Appl.and Phys. 20 Linh, (1981)IEEE L598.Trans. Electron [2] Abe, D. Delagebeaudeuf NT. Devices ED-29 (1982) 955. [3] K.H.G. Duh, P.C. Chao, P.M. Smith, L.F. Lester, B.R. Lee and J.C.M. Hwang, Millimeter-Wave Low-Noise HEMT’s, Device Research Conf., Amherst, MA, June Paper IIA-2. [4] 1986, P. Solomon and H. Morkoç, IEEE Trans. Electron Dcvices ED-31 (1984) 1015 [5] DV. Lang, R.A. Logan and H. Jaros, Phys. Rev. B19 (1979) 1015. [6] S. Dhar, W.-P. Hong, P.k. Bhattaycharya, Y. Nashimoto and F.-Y. Juang, IEEE Trans. Electron Devices ED-33 (1986) 698. [7] J.J. Rosenberg, M. Benlamri, PD. Kichner, J.M. Woodal and G.D. Pettit, IEEE Electron. Device Letters EDL-6 (1985) 491. [8] W.T. Massehnk, A. Ketterson, J. Klem, W. Kopp and H. Morkoç, Electron. Letters 21 (1985) 937. [9] S.M. Liu, M.B. Das, C.K. Peng, J. Klem, T.S. Henderson, W. Kopp and H. Morkoç, IEEE Trans. Electron Devices ED-33 (1986) 576. [10] A.A. Ketterson, W.T. Masseink, J.S. Gedymin, J. Klem, C-K. Peng, W. Kopp, H. Morkoç and KR. Gleason, IEEE Trans. Electron Devices ED-33 (1986) 564. [11] F.J. Tegude and K. Heime, presented at 11th Intern. Symp. on GaAs and Related Compounds, Biarritz, 1984.

S.M. Liu eta!.

/ Pseudomorphic InGaAs/A !GaAs MODFET’s

[12] M. Heuken, L. Loreck, K. Heime, K. Ploog, W. Schlapp and G. Weimann, IEEE Trans. Electron Devices ED-33 (1986) 693. [13] SM. Liu and MB. Das, in: Proc. UGIM Symp., A&M University of Texas, May 1983, p. 169. [14] PH. Handel, Phys. Rev. A22 (1980) 795. [15] A. Van der Zeil, PH. Handel, X. Zhu and K.H.G. Duh, IEEE Trans. Electron Devices ED-32 (1985) 667. [16] SM. Liu, M.B. Das, W. Kopp and H. Morkoç, IEEE Electron Device Letters EDL-6 (1985) 453.

367

[17] SM. Liu, MB. Das, W. Kopp and H. Morkoç, in: Proc. UGIM Symp., Auburn University, June 1985, p. 202. [18] SM. Liu, MB. Das, W. Kopp and H. Morkoç, IEEE Electron Device Letters EDL-6 (1985) 594. [19] I.J. Fritz, L.R. Dawson and T.E. Zippenan, J. Vacuum Sci. Technol. Bi (1983) 387. [20] BR. Nag, SR. Ahmed and M. DebRoy, IEEE Trans. Electron Devices ED-33 (1986) 788. [21] A.B. Grebene and S.K. Ghandhi, Solid State Electron. 12 (1969) 573.