Wet chemical separation of low-temperature GaAs layers from their GaAs substrates

Wet chemical separation of low-temperature GaAs layers from their GaAs substrates

MATERUAIS SCIENCE & ENCINEEMBW ELSEVIER B Materials Scienceand EngineeringB40 (1996) 58-62 Wet chemical separation of low-temperature their GaAs su...

659KB Sizes 0 Downloads 32 Views

MATERUAIS SCIENCE & ENCINEEMBW ELSEVIER

B

Materials Scienceand EngineeringB40 (1996) 58-62

Wet chemical separation of low-temperature their GaAs substrates J. NovAka, M. Morvica, J. Betko”, A. Fbster”, “Institute of Electrical Ellgiheering, Slovak Academy of Sciences, bInstitute fiir Srhichturd lonentechnik, Forschungxentrlrmwn

SK-842 Jiiliclr,

GaAs layers from P. ICordogb

39 Bmtisha, Sled Rephlic D-524175 Jiilich, Germnry

Received4 January 1996

Abstract We reported a method of the wet chemical separation for low-temperature (LT) GaAs epitaxial layers grown by molecular beam. epitaxy from their substrates. Samples with AlAs as well as AlGaAs etch-stop interlayers were used in this study, but better properties have been found for the samples with AlAs. Very interesting dependence of etching rate of the citric acid based etchant on the sample growth temperature is reported. For illustration of usefulness of this method, the temperature dependencies of the resistivity and Hall mobility measured on four separated LT-GaAs layers are presented. Keywords:

Wet chemical separation: Hall mobility; Gallium arsenide

1. Introduction Molecular-beam epitaxial GaA,s layers grown at low temperature (LT) have recently been shown to serve as excellent buffer, active and passivation layers for various device applications. Electrical properties of LTGaAs layers are strongly dependent on the growth and annealing conditions and therefc’re detailed knowledge

of this behaviour is very important for the design and preparation of advanced devices with LT-GaAs [l-5]. The room temperature resistivity of as-grown LTGaAs layers increases strongly with the growth temperature Tg = 200-450 “C and reaches values up to IO7 R cm [4,5]. Many

applications

device structure at temperatures

require

annealing

higher than

Tg.

From this it follows that it is practically impossible to get correct conductivity and Hall-effect parameters of LT-GaAs layers without separating the layers from their substrates. As this fact had not been taken into account in earlier studies of LT-GaAs, the results then needed correction based on the measurement of separated layers [1,2]. However, details about the layer-separation procedure have not yet been reported. In this work, we present a method of the separation of GaAs layers from their GaAs substrates by means of wet chemical etching. The substrate removal procedure allows LT-GaAs layers of a minimum thickness of 1 /tm to be separated. The usefulness of this method is

of a

demonstrated

The

ments on separated LT-GaAs

annealing leads to an increase of the LT-GaAs resistivity, e.g. from about 10 R cm in 200 “C as-grown layers up to lo’-lo7 R cm after their annealing at temperatures higher than 500 “C [3,5]. lY’he sheet resistivity of annealed LT-GaAs layers (usually about 1 /lrn thick) becomes comparable with the resistivity of the substrate

-in the above mentioned cases. On the contrary, in as-grown LT-GaAs samples, thl: hoping conductivity mechanism is dominant at room temperature owing to the extremely high concentration of native defects and the apparent Hall mobility is very low, less than 0.15 cm2V-1s-i in 200-250 ‘C as-grown layers [3].

by conductivity

and Hall-effect

measure-

layers.

2. Substrate removal principles

In general, two different etching techniques can be used for this substrate removal procedure: dry reactive etching (for example, reactive ion etching) or wet chemical etching. Dry reactive etching is usually connected with damaging of the etched surface (up to 50 nm under etched surface) [6,7] therefore the wet chemical etching is more convenient in our case. In addition, the GaAs substrate and the epitaxial layer of LT-GaAs, 0921-5107/96/S15.000 1996- Elsevicr ScienceS.A. All rights reserved

J. Novrik

et al. / Materials

Science

which should be separated, have the same chemical composition and it is very difficult to perform the etching process with such accuracy that it can be stoped just at the substrate-layer interface. This problem can be overcome by using another GaAs related material of different etching behaviour, (e.g. AlAs or AlGaAs) as an etch-stop interlayer [8]. Consequently, a two-step etching treatment can be used to precisely control the removal process: at first, a GaAs substrate is etched off without affecting the etch-stop interlayer (step I), and then etching the interlayer leaving LT-GaAs layer unaffected (step II). In general, the etching proceeds by an oxidation-reduction reaction at the semiconductor surface, followed by dissolution of the oxidized material. A typical etchant for this purpose consists of two separate components, one of which is an oxidizing agent (usually H,O,), whereas the other component dissolves the resulting oxide (for example by NH,OH or an acid). Selective etching of GaAs (step I) in relation to the AlGaA and AlAs has been investigated in the past with respect to the etching selectivity of different materials in the epitaxial multilayer structures. For these purposes NH,0H/H202 [9], or its combination with HCl [lo] or citric acid based etchants [II] were found as etchants with sufficient selectivity. Common selective AlGaAs etchants (step II) are boiling HCl [12], KI:I,:H,O solution (so-called I, etchant) [13], as well as diluted hydrofluoric acid [8]. The extreme selectivity of diluted hydrofluoric acid (higher as 107) between the AlAs release layer and AlGaAs (or GaAs) layer can be utilized for under cutting large area epitaxial structures [8]. Unfortunately, this technique requires that the corners of an epitaxial structure be lifted up during the undercutting to permit the outdiffusion of dissolved H2 gas from the etching zone (HZ is the most critical product of the chemical reaction). It is known that as-grown LT-GaAs layers have relatively high lattice misfit, and thus the additional strain may lead to total distortion of the structure. Therefore, this lift-off technique cannot be used in our case.

and Engineering

B40 (1996)

58-62

59

The removal procedure of GaAs substrates consists of following steps: - lapping of the sample from the substrate side; - epoxy the sample, layer down, to a piece of microscopic glass; - polishing etch of the GaAs substrate; - etching down the rest of the substrate; - etching of the interlayer, and - rinsing of the separated layer in DI water. We found that the second step, using epoxy as the sample, is very important to carry out the separation procedure successfully. It is necessary to use highly firm and temperature stable epoxy resin, which can substitue for the role of the substrate instead of removed GaAs. The application of commonly used butylphtalate resin has been found unsuitable because the strain incorporated in the epitaxial layer can cause total distortion or deformation of the epitaxial layer as it is demonstrated in Fig. 1. The total crushing of an LT-GaAs layer attached by improper resin (buthylphtalate) to the glass holder is shown in Fig. l(a). This adhesive is sufficiently

(4

3. Experimental GaAs layers used in this study were grown in a Varian Mod GEN II MBE system on indium-freemounted 2 inch (100) semi-insulating GaAs substrates with the resistivity of 1.5 x lo* R cm. The growth temperature was adjusted by a calibrated thermocouple in the range between 200 and 600 “C. An As,/Ga beamequivalent pressure ratio of 19 and a growth rate of 1 pm h-’ were used to grow 2 and 6 pm thick epitaxial layers. An AlAs or AlGaAs etch-stop interlayer of 0.5 ,Ltrn thickness was grown between the substrate and the LT-GaAs.

Fig. 1. (a) Crushing of separated LT-GaAs epitaxial layer grown at Tr = 200 “C as a consequence of non-compensated strain using improper resin. (b) Deformation of LT-GaAs epitaxial layer grown at T, = 200 “C after removing of substrate up to thickness of 10 ,um.

60

J. Novcik et al. / Materials Seieuce

resistive to various echants but it is very soft and cannot substitute for the substrate forces at LT-GaAs epitaxial layers grown under 300 “C. Fig. l(b) shows the mechanical deformation of the epitaxial layer/substrate system caused by different lattice constants of these two materials, In this case, a 2 pm thick LTGaAs epitaxial layer was grown at T, = 200 “C (i.e. Aa/a = 1.3 x 10W3) and the GaAs semi-insulating substrate was thinned to the total thickness of 10 pm. The 6 mm long and 2 mm wide samples were observed to bend up to 0.3 mm. The samples were initially epoxied to a glass holder, then they were lapped down from the substrate side to the thickness of about 80 pm by means of common abrasives, For further thinning of the samples down to 20 pm a polishing etch of the GaAs substrate in H,SO,:H,O:H,O, (3:l:l) was performed. Remaining parts of the substrate were then selectively etched off in a solution based on citric acid. This etching stops at the AlAs (or AlGaAs) etch-stop interlayer automatically. The separation procedure was finished by removing the interlayer with diluted hydrofluoric acid. Samples with AlAs as well as AlGaAs etch-stop interlayers were used in this study, but better properties have been found for the samples with AlAs. The main advantage of the AlAs etch-stop layer and the citric acid etching system is that after etching off the remaining parts of the GaAs substrate and removing of AlAs etch stop layer by diluted hydrofluoric acid, the surface of the LT-GaAs 1a:yer is unaffected and clean. From this point of view, AlAs, compared with AlGaAs, can be removed from the LT-GaAs surface by hydrofluoric acid with selectivity higher as 107. All AlGaAs etchants have a lower selectivity to GaAs and the etching is influenced by imperfections in the epitaxial layer, Consequently, tb: total and planar removal of AlGaAs layer without damaging of LTGaAs surface is very difficult. In addition, the colour of AlAs differs from that of GaAs, while the colours of GaAs and AlGaAs are similar. This colour difference is very important because it allows the etching process to be directly observed and controlled, if it is necessary. The main advantage of the citric etching system is the sufficient selectivity and the planar etching of GaAs unaffected by crystal imperfections. It is true that the selectivity of this etching system to GaAs/AlAs is low (only about !j) in comparison with that to GaAs/Al,,,Ga,,,As (higher as 110) [lo], but considering all advantages and problems mentioned above, we have decided to use the combination of AlAs. etch-stop layer and citrc acid/H,O, etchant (5: 1).

rind

Eugineeritg B40 (1996) 58-62

growthtemperature(“C) Fig. 2. Comparison of etchingrate (ci), latticemismatch (A) and electricalconductivity(I) of LT-GaAs epitaxiallayersgrown at differenttemperatures Tp.Latticemismatch An/aisshownin %, scale is on the right Y-axis,inside.

3. Results and discussion The etching rate of the citric acid/H,O, etchant composition (5:l) over the semiinsuilating GaAs substrate is near 0.3 /tm min- I, Like all other properties of LT-GaAs, the etching rate of this material is also influenced by its nonstoichiometry, which is a function of the growth temperature Ts. Fig. 2 illustrates as the etching rate depends on Tg. At first the etching rate decreases from 0.24 ,tlrn min- ’ for 7’, = 200 “C to the minimum value of 0.14-o. 15 jtm min - ’ for Tg = 300400 “C and then the etching rate increases up to 0.2 pm min- ’ for r, = 550 “C. All samples were etched at the same time in the same etching bath at the temperature of 25 “C to exclude possible mistakes in the determination of the etching rate. It is generally believed that special properties of LT-GaAs are connected with a high amount of As incorporated into the GaAs lattice [2,11) 121 causing remarkable nonstoichiometry. This nonstoichiometry leads to the lattice mismatch in layers grown at T, lower than 350 “C, as it is demonstrated in Fig. 2. The lattice mismatch depends on the growth temperature and it decreases from An/ a = 1.3 x low3 for r, = 200 “C to zero for the layer grown at T, = 400 “C or higher. The character of this T, decrease is similar to that of the etching rate vs. Tg for the temperature interval between 200 and 400 “C. But for higher r,, the lattice mismatch An/a goes to

J. AJo&k

et al. 1 Materials

S&me

zero while the etching rate increases. Therefore, we can deduce the etching along the misfit dislocations is not the main starting mechanism of the LT-GaAs etching. On the contrary, it is interesting to compare the plots of the etching rate and electrical conductivity of the epitaxial layers vs. Ts. As it follows from Fig. 2, both dependences have a very similar characteristic. Therefore, it may be expected that both the etching rate and the electrical conductivity are initialized by the same mechanism related to the material structure of the epitaxial layer. The arsenic excess (about 1 at.O/) is uniformly distributed in the crystal lattice and leads to the creation of deep levels which are responsible for high resistivity of GaAs layers grown by MBE at low and intermediate temperatures. The role of the arsenic precipitates is less significant [13,14], which can be observed in layers grown at 200-300 “C after annealing at temperature 600 “C or higher [14,15]. Therefore, we suppose that the etching rate is also related to the arsenic excess. The method of wet chemical separation presented in this study was applied to different MBE epitaxial structures, Separated LT-GaAs epitaxial layers were characterised by using a high impedance system and the van der Pauw method. Ohmic contacts were made by rubbing pure gallium into the surface of the samples. The main advantage of gallium contacts is that they are liquid in the whole temperature range in which the electrical parameters were measured. This feature is very important because epitaxial layers without substrate support are very frail and a minimal additional stress may lead to the fatal distortion of a structure measured. As shown in Fig. 3, Ga/LT-GaAs contacts are ohmic, having linear I-V characteristics in the whole measured range. The Ga/LT-GaAs contacts are suitable for both high resistive as well as low resistive

-12

-a

T'

'

-4 "

0 I

I

4 7

ma

Etzgineeritzg

2,ox1o4-

./

n

A--A---

A-A-

IO'

l’ O

/I@’

-O-8077/420-as-grown -0-8077/420-annealed -*-6161/250-as-grown -*-6161/250-annealed

IO2

A--A

A

A-

r,



1 L ”



“‘1

3,0

23

2,o

3,5

inverse measurement temperature (10” K-l) Fig. 4. Temperature dependence of the resistivity of two LT-GaAs epitaxial layers grown at T, = 250 “C and rg = 420 “C before and after annealing at T= 590 “C.

epitaxial layers. For illustration, temperature dependencies of the resistivity and Hall mobility measured on four separated LT-GaAs layers are shown in Figs. 4 and 5, respectively. The layers were grown at Tg = 250 and 420 “C, both were separated before (as grown) and after annealing (590 “C). The resistivity of both epitaxial layers is comparable with or lower than the substrate resistivity, and it may be obtained by the differential van der Pauw measurement without separation too. On the contrary, the Hall parameters cannot be evaluated from the measurement on the unseparated layer. As it follows from Fig. 5, the separation it makes it possible to obtain reasonable results even in the case of extremely

12

$=~=k@kQk-~-azzz-~

1x10-Q

""I

? ’ O3f LT MBE GaAs separated > ,//-I N E 102: 23 */ a g A/ -o-8077/420-as-grown 2 IO’?

GalLT-GaAs Nr.6161

-

61

t

IO‘)

a

58-62

LT MBE GaAs separated

IO'

4,0x106 -

340 (‘1996)

as grown

annealed

E 7J ; i?? x g

I

-1x1o-g bias (Vj



Fig. 3. I-V characteristics of Ga ohmic contacts to as grown and annealed LT-GaAs epitaxial layer grown at rp = 250 “C.

/ lOOr

AlA

/-A

: 3 : . -

-o-8077/420-annealed -~-6161/250-as-grown --l-6161/250-annealed -.-----A

10-l

:

3

A~~~--

300

350

400

450

measurement temperature (K)

Fig. 5. Temperature dependence of the Hall mobility of two LTGaAs epitaxial layers grown at TS = 250 “C and I-P = 420 “C before and after annealing at T= 590 “C.

62

J. Novrik et al. / Materials Science and Engineerhg B40 (1996) 58-62

low Hall mobility connected with. a hopping conductivity mechanism. In all samples investigated, the Hall coefficient had the negative sign (according to the convention: the Hall coefficent is negative for electrons and positive for holes). Of course, the dependencies obtained need further analysis with regard to a possible mixed conductivity regime and the two-layer problem (hopping conductivity, band conductivity, etc.). Results of this analysis will be presented elsewhere.

5. Conclusions We have presented the method of wet chemical separation of LT-GaAs layers. Citric acid/H,O, etch system was found as a proper etchant. Described dependence of the etching rate on the growth temperature is an important information needed at various applications of LT-GaAs epitaxial layers in device technology. From the dependence of the etching rate and electrical conductivity on the layer growth temperature T,, we assume that the same mechanism is responsible both for the change in the electric conductivity and the change in the etching rate vs. Tg. The successful epitaxial layer separation enabled the measurement of the transport properties of LT-GaAs MBE samples with high resistivity and very low Hall mobility.

References J.R. Sizelove andC.E. Stutz,in C.J. 111D.C. Look,G.D. Robinson,

Miner (ed.), Se~~~ikdating III- V Moterids, Ixtapa, Adam Hilger,Bristol,1992. PI DC. Look, D.C. Waiters,G.D. Robinson,J.R. Sizelove,M.G. Mier andC.E. Stutz,J. Appl. Phys., 74 (1993) 306. 131 J. Betko,P. KordoS,S. Kuklovskp,A. Forster,D. GreguSova andH. Liith, Mater. Sci. Eng., B28 (1994)147. [41 P. Kordos,J. Betko,A. Forster,S. Kuklovskg,Ch. Diekerand F. Rtiders,21stSy~p. on Compowd Semicorductors, SanDiego, 1994. [51 P.Kordos,A. Fiirster,J. Betko,M. Morvic andJ. Novak,Appl. Phys. Lett, 67 (1995) 973.

161P.CollotandC. Gaonach,Semicond. Sci. ‘?‘ec/~rroL,5 (1990)237. [71 L. He andW.A. Anderson,Solid Stnte Electr., 35 (1992)151. PI E. Yablonovitch,T. Gmitter,J.P.HarbisonandR. Bhat,Appl. Phys. Lett., 51 (1987)2222. C. Besombes, C. Corbet,F. Heliotand L91 D. Ankri, A. Savennec, J. Riou,Appl. Phys. Lett,, 40 (1982) 816. PO1H. Nobuhara,0. WadaandT. Fuji, Electrorf. L&t., 21 (1985) 718. [111G.C. DeSalvo,W.F. TsengandJ. Comas,J. Electroch~. Sot.,

139(1992)831.

WI X.S. Wu, L.A. ColdrenandJ.L. Merz, Electrorz L&t., 21 (1985) 558. 1131 A. Malag,J. Ratajczakand J. Gazecki,Motcr Sci. Eug., B20

(1993)332. andS. O’Hagan,J. Appl, Phys., 75 (1994) 3396. [I41 M. Missous 1151 Z. Lilienthal-Weber, ,.-I X.W. Lin, J. Washburnand W. Schaff, Appl. Phys. Lett., 66 (1995)2086.