Monolayer control of chemical beam etching

CRYSTAL GROWTH

ELSEVIER

Journal of Crystal Growth 145 (1994) 680—686

______________________

Monolayer control of chemical beam etching W.T. Tsang

a,*, T.H. Chiu b R.M. Kapre a, J.F. Ferguson b TBeII Laboratories, Murray Hill, New Jersey 07974, USA b AT&T Bell Laboratories, Holmdel, New Jersey 07733, USA

a A T&

Abstract An etching process with real-time monitoring of each monolayer removed is demonstrated. This etching capability which we refer to as monolayer chemical beam etching (ML-CBET) is achieved by employing in-situ reflection high energy electron diffraction (RHEED) intensity oscillation monitoring during etch. The etching is accomplished by injecting AsC13 or Pd3 gas directly into the growth chamber at a temperature comparable to the typical growth temperature in the same system. This permits instant switching from etching to growth and vice versa in the same experimental run. The combination of growth and etch in a single chamber with both controllable at atomic scale presents a very powerful processing method for the fabrication of intergrated devices that require multiple steps of etch and regrowth. We have also identified the roughening mechanism during etch and present the methods to obtain excellent morphology through a migration enhanced smoothing mechanism.

1. Introduction Chemical beam epitaxy (CBE) is a versatile semiconductor epitaxial growth technique [1,21 that utilizes beams of chemicals directly impinging on a heated substrate surface. Under properly controlled conditions, decomposition of metal-alkyls on the surface resulted in the in corporation of the desired elements to produce the epitaxial film in a layer-by-layer fashion. By introducing a chemical that reacts with the substrate materials at growth temperature to form volatile species, the growth is immediately converted into a chemical beam etching (CBET) process. If the etching also preceeds in a layer-by-layer fashion, this _______

*

Corrsponding author,

process can be viewed as an exact reversal of the chemical beam epitaxy. As in the case of growth, in principle, etching control at monolayer scale is feasible by CBET. For the development of advanced optoelectronic devices where multiple steps of etch and regrowth are needed, such an in situ etching method has the potential of providing an accurate etch depth control, a clean interface without ion damage and fewer processing steps. In-situ etching inside a metalorganic chemical vapor deposition (MOCVD) growth chamber has been reported using HCI [3] or PCI3 [4J. The etch rate in the atmospheric or low pressure (—~100 Torr) reactors is quite sensitive to the system pressure, the gas flow rate and the substrate temperature. Control of the etch rate is at best of the order of a few nm which is still subjected to the reproducibility of substrate temperature. In

0022-0248/94/$07.00 © 1994 Elsevier Science B.V. All rights reserved SSDI 0022-0248(94)00391-X

W T. Tsang et al. /Journal of Crystal Growth 145 (1994) 680—686

this work we show that etching in a high vacuum growth chamber can be monitored in real-time by

SUBSTRATE TEMP = 580°C TEGa SWITCH IN

,~

means of reflection high energy electron diffraction (RHEED) intensity oscillation. This makes possible etch rate control at monolayer scale. Furthermore, the surface morphology obtained at an etching temperature comparable to the growth temperature of InP or GaAs is shown to be dominated by the surface cation diffusion mechanism.

-

The etching experiments were carried out in a CBE system with a gas manifold equipped with a growth—vent switching system, a precision goniometer stage, and 20 keV RHEED gun. The intensity of the specular beam of the RHEED pattern was measured by using a photodector with an entrance aperture. AsC13 was used with H2 carrier gas in a concentration of 5% and a low-temperature (—‘ 50°C) injector was used to introduce the gas into the chamber. The (100)oriented GaAs or InP substrate was used. Etching of InP was done at substrate temperatures from 450 to 570°C. For the RHEED measurements, etching was done on GaAs at substrate temperatures from 450 to 650°C with cracked AsH3 flux. The chamber pressure during etching was typically 1 x 10—i Torr. The surface morphology was studied by Normarski or atomic force microscopy (AFM).

3. RHEED measurements In the experiment, a clear (2 x 4) As-rich reconstructed surface RHEED pattern was first established for the incident beam of the [110] azimuth after oxide desorption and annealing at 600°C under cracked AsH3. Fig. 1 shows the trace of RHEED intensity oscillation from the specular beam immediately after the initiation of GaAs growth by switching in the triethylgallium (TEGa) flow. The period of oscillation corresponds exactly to a single (Ga + As) layer. This provides a real-time growth thickness control with

GROWTH STARTS GROWTH

AsC~ 3IN ETCHING STARTS STOP

TEGa=0.51 sccm

ETCHING

-

STOP

AsC~3=0.15sccm

t-AsCt

-

,J ~

~-TEGaOUT

ONE MONOLAYER OF GaAs GROWN

0

2. Experimental procedure

681

I

I

10

20

3OUT

J~

ONE MONOLAYER OF GaAs REMOVED I

I

I

I

30 40 50 60 70 TIME (sec) Fig. 1. Intensity oscillations of the specular beam in the (2 X 4) RHEED pattern after the commencement of GaAs growth by switching in TEGa (left portion), and of etching by switching in AsCI3 (right portion). Note the abruptness in starting and termination of growth or etch achieved.

atomic layer precision. On switching out the TEGa flow, the intensity oscillation stops abruptly and is followed by a recovery which is associated with a surface smoothing effect due to the surface cation diffusion that minimizes the step density. Next, AsCl3 was switched in from vent. Fig. 1 shows that a damped intensity oscillation starts immediately in a way very similar to the growth situation. The period of oscillation also corresponds to exactly the time required to remove one monolayer of GaAs. The uniform periodicity indicates a constant etching rate. When AsCl3 is switched out, the oscillation also stops abruptly indicating a clean termination of the etching process. The intensity recovery indicates a smoothing process is taking place. As in the growth case, the oscillatory effect in RHEED suggests that etching occurs via a layer-by-layer process. Thus, we have demonstrated that the chemical beam etching (CBET) process is just the reverse of the chemical beam epitaxy (CBE) [5].The use of RHEED oscillation is a very powerful method to control the etching with atomic scale precision. We will differentiate this etching process from others by referring to it as monolayer chemical beam etching (ML-CBET). This technique should be applicable to other growth-etch chamber, for example, the MOCVD reactor where oscillatory behavior in the optical reflectance spectroscopy has been

682

W. T. Tsang et a!. /Journal of Crystal Growth 145 (1994) 680—686 SUBSTRATE TEMPERATURE (°C) 560

10.0

527

496

467

3-

441

PCI

3

417

354

0 66

‘11 a, -J a,

I— (3 z

3: I-. Fa,

0

~oI9~~\

22.7 kCaIImoIe)

.650°C ~580 A 550

-

WI

0525 1

00

0.1I

0.2 0.3 0.4 I I I AsC~FLUENCE (sccm)

0.5 I

Fig. 2. The etch rate as a function of pure AsCI3 fluence at several different substrate temperature.

associated with monolayer growth of GaAs [61. It should be possible to apply this technique in Si etching because RHEED oscillation is present in Si CBE [71. RHEED oscillations also provides a very fast and accurate in -situ measurement of etch rates versus etching parameters which provide very useful insight regarding the etching mechanisms. Fig. 2 shows the etch rates as a function of the pure MCI3 (% of AsCl3 x total flow) fluences at several substrate temperatures. The linear dependence suggests a reactant supply-limited mode of etching. A weak temperature dependence can be

00

1.15

1.2

1.25

1.3

1.35

100011 (K) 1.4 sub 1.45

1.5

Fig. 4. The InP INVERSE etch rateTEMPERATURE, as a function of substrate temperature for a fixed PCI 3 fluence of 0.66 SCCM and a PH3 flow of 3 SCCM.

seen in Fig. 3 which shows that the etching rate peaks at about 525°C.The etch rate stays relatively constant even up to 650°C.The decrease in etch rate with increasing temperature is more pronounced at high etching rate. This is similar, but at a lesser extent, to the case of InP etched by PC13. The etch rate of InP shown in Fig. 4 is determined from cross-section thickness measurements. Below about 500°C,an Arrhenius behavior with an activation energy of 0.989 eV is

(a)

_______

AsC~FLUENCE (sccm) 2

N

0.3

-

~

—~-_-__~__--—a-

I

(b) growth

0.2

(31

z

-.0—-C

0

0

0—

0—0.1

o

I

H W

________

I

500

I

I

550 600 650 SUBSTRATE TEMPERATURE (°C) Fig. 3. The GaAs etch rate plotted as a function of substrate temperatures.

__ ____

_____

(C)

etch

Fig. 5. Evolution of a semiconductor surface with (a) tn-level step edges. Under growth (b), the adsorbed group III atoms tend to anchor at the step edge and smooth out the surface. Under etch (c), the adsorbed gas molecule will etch away the embedded group III atom and leave a hole behind.

W. T Tsang et aL/Journal of Crystal Growth 145 (1994) 680—686

observed which indicates the desorption of InCl is the rate limiting step [8]. At temperatures above 530°Cthe InP etch rate shows a noticible decrease. Compared to the case of GaAs, this is expected because the desorption of PC13 from InP surface is easier. For GaAs etching, the diffi-

683

culty to obtain meaningful RHEED oscillation below 500°C is in part due to the charging-up effect in the present setup. It is also possible that the etching no longer proceeds in a layer-by-layer mode and the etch rate becomes limited by the desorption of GaCl.

a—

Fig. 6. The surface morphologies of lnP under various etching conditions: (a) 450°C.1.3 A/s; (b)450°C,6.5 A/s; (c)550°C, 11 A/s; (d) 530—570°C,1.2 A/s; (e) 550°C,10 A/s, with TMIn.

684

W. T Tsang et a!. /Journal of Crystal Growth 145 (1994) 680—686

4. Surface morphology It should be noticed that CBET etching of InP or GaAs at low temperature where the etch rate is limited by the desorption of clorinated species, a rough morphology is easily observed. The development of rough morphology in an etching process can be conceptually understood in terms of the simple schematic shown in Fig. 5. A surface with tn-level step edges under growth mode has a tendency to evolve into a surface with bi-level steps as shown in Fig. Sb because the incoming group III atom diffuses around and anchors at the step edges. On the other hand, the incoming etching gas molecules will remove the group III atom and leave a hole behind as shown in Fig. Sc, unless the neighboring group III atoms on the surface can move around to fill the hole. The surface morphology of InP etched by PCI3 is very consistent with the argument that smoothing is due to the migration of surface group III atom. Figs. 6a and 6b shows that a rough surface is in general produced for etching at low substrate temperature. At high temperature, very smooth and featureless surfaces are obtained at etch rate less than 6 A/s (Fig. 6c). At high etch rates, the surface was still excellent, but with detectable background texture when viewed under a Nomarski optical microscope (Fig. 6d). Upon the addition of trimethylindium (TMIn) flow equivalent to a growth rate of of 1 A/s during etching, a dramatically improved morphology is obtained even at a high net etch rate of 10 A/s. Under this condition at high temperature, apparantly a smoothing mechanism takes place because of the added In atoms on the surface during etching resulting in a surface indistinguishable from that of an unetched substrate as determined by optical microscopy. To further illustrate the migration enhanced surface diffusion that smooths out an etched surface, the following etching experiments has been conducted on InP using AsC13. A moderate substrate temperature of 520°Cis chosen which ensures that the mCi desorption is no longer a limiting step meanwhile the thermal damage is not serious if the surface is not subjected to a cracked PH3 beam for a short period of time.

________________________________________

Fig. 7. The InP surface morphologies scanned under AFM of (a) a bare substrate and (b) a sample of 2100 A etched depth by AsCl3. The vertical scale is 1000 A/div.

The etch rate is set at 3 A/s. The surface under continuous etching together with a stabilizing PH3 flux of 2 SCCM for 720 s shows a rough appearance which is observable even under optical microscopy. Fig. Th shows the micrograph of surface roughness viewed under AFM. An averaged roughness of about 400 A deep is measured

W. T Tsang et al. /Journal of Crystal Growth 145 (1994) 680—686

685

etched by a 6 s pulse followed by a 4 s recovery without any PH3 flux, i.e. annealed under vacuum. A very small roughness of about a few A deep is only detectable under AFM. With optical microscopy, such a surface is completely indistinguishable from the unetched substrate. It is well known that growth interruption during a growth process will enhance the smoothness because of the diffusion of group III atoms. The surface diffusion length increases with decreasing group V overpressure. It is clear from the present ex-

periments that the same mechanism is operative during the etching process at temperatures comparable to that of the growth process.

5. Conclusions ‘—V

~

-

~

...~ ~

We have developed an etching process with real-time counting of each monolayer removed, thus achieving etching with monolayer precision

•~.

4

I

and control. This is an exact reversal of CBE process. The new etching capability which we refer to as ML-CBET is achieved by employing

in-situ RHEED intensity oscillation monitoring during etch. Having both epitaxial growth and etching integrated in the same process and both

‘~

u

,~~

I i~’~o

-

k. ~

~

~..

.;~,

~

“i.

~

~

c~):.r .

.

I

* ______

~

r

1

capable of ultimate control downpowerful to the atomic layer precision represents a very combination. This permits instant switching from growth to etching and vice versa, clean regrown interfaces critical for device applications. The surface morphology is found to be dominated by the surface cation diffusion mechanism. With this

Fig. 8. The Inp surface morphologies scanned by AFM for a total etched depth of 2100 A. Sample (a) is etched by a 6 s MCI3 pulse and followed by a 10 s anneal under 2 SCCM PH3. Sample (b) is etched by a 6 s AsCl3 pulse and followed by a 4 s anneal under vacuum.

understanding, novel methods that takes the advantage of enhanced migration of surface cations have been shown to greately improve the surface morphology.

References for a total etched thickness of 2100 A. Fig. 8a shows a much improved surface when etching is performed under pulsed mode, i.e. a 6s etch followed by a 10 s recovery under a beam of cracked PH3 (repeated 120 times). However, the best morphology is obtained when the surface is ‘—

[1] W.T. Tsang, in: VLSI Electronics Microstructure Science, Vol. 21, Eds. N.G. Einspruch, S.S. Cohen and R.N. Singh (Academic Press, New York, 1989). [2] W.T. Tsang, J. Crystal Growth 120 (1992) i; 105 (1990) 1. [3] C. Caneau, R. Bhat, M. Koza, J.R. Hayes and R. Esagui, J. Crystal Growth 107 (1991) 203.

686

W~T Tsang et al. /Journal of Crystal Growth 145 (1994) 680—686

[4] B. Henle, R. Rudeloff, H. Bolay and J. Scholz, in: Proc. 4th Conf. on InP and Related Materials, 1992, p. 159. [5] W.T. Tsang, T.H. Chiu and R.M. Kapre, J. Crystal Growth i35 (1994) 377. [6] D.E. Aspnes, R. Bhat, C. Caneau, E. Colas, L.T. Florez, S. Gregory, J.P. Harbison, I. Kamiya, V.G. Deramidas, MA. Koza. M.A.A. Pudensi, W.E. Quinn, S.A. Schwarz,

mt.

M.C. Tamargo and H. Tanaka, J. Crystal Growth 120 (1992) 71. [71SM. Mokier, W.K. Liu, N. Ohtani and BA. Joyce, J. Crystal Growth 120 (1992) 290. [81 W.T. Tsang, R. Kapre and P.F. Sciortino, Jr., Appl. Phys. Lett.62 (1993) 2084.

I

I