In situ characterization of rare earth-CdTe heterostructures by ion beam analysis

In situ characterization of rare earth-CdTe heterostructures by ion beam analysis

266 Thin Solid Films, 249 (1994) 266-270 In situ characterization of rare earth-CdTe heterostructures by ion beam analysis P. Gros, G. Fiat, D. Bru...

415KB Sizes 0 Downloads 37 Views

266

Thin Solid Films, 249 (1994) 266-270

In situ characterization of rare earth-CdTe heterostructures

by ion beam analysis P. Gros, G. Fiat, D. Brun, B. Daudin, J. Eymery, E. Ligeon and A. C. Chami* CEA / DOpartement de Recherche Fondamentale sur la MatiOre Condenske/SP2M ]PI, BP 85 X, 38041 Grenoble Cedex, France (Received December 3, 1993; accepted March 1, 1994)

Abstract A molecular beam epitaxy (MBE) system used for epitaxial growth of rare earth (Eu, Sm and Nd) compounds on CdTe is described. Owing to the high reactivity of the films, it appeared fruitful to connect this apparatus to a 2.5 MeV van de Graaff accelerator, allowing an in situ structural characterization of the samples by Rutherford backscattering spectrometry and ion channelling analysis. Effusion cell flux calibration, homogeneity and thickness layer measurements are presented, For the first steps of the rare earth deposition, the correlation of thickness measurements to reflection high energy electron diffraction patterns is shown. Interfacial defect studies by ion channelling during CdTe growth on Cdt _xZnxTe are also reported.

1. Introduction During the past ten years, molecular beam epitaxy (MBE) of I I - V I semiconductors has been widely studied, due to the great interest in the potential applications of these materials as infrared detectors (HgCdTe) or luminescent devices. The practical application of these materials to electronic systems implies first the control of both n- and p-doping, and, second, the realization of metal-semiconductor heterojunctions. The choice of the deposited metal is not trivial, due to the high reactivity of Te with most metals and to interdiffusion processes. With regard to this last point, it was underlined by Sands et al. [I] that the best result was obtained when the metal or metallic compound was in epitaxial relation with the semiconductor, as the topological defects leading to a fast diffusion path and/or recombination centres are minimized. In the particular case of CdTe, it has been shown recently that the rare earth (RE) metals were promising candidates for the realization of metal-semiconductor heterostructures [2, 3]. However, as the RE (and their compounds) rapidly oxidize in the open air, the characterization (chemical composition, defect density etc.) of these materials must be carried out in the MBE chamber. In addition, the optical properties of real electronic systems are strongly dependent on the crystalline quality of the semiconductor epitaxial film. For example, the growth of CdTe on (001) Cd~ _xZnxTe (x ~ 0.03), leads to the formation of misfit dislocations for thicknesses *Permanent address: Institut de Physique, USTHB, BP 32, El Alia, Bab Ezzouar, Alger, Algeria.

0040-6090/94/$7.00 SSDI 0040-6090(94) 06121 -Z

larger than the critical thickness (,-~400 nm) [4]. For a (11 l)-oriented substrate a large number of defects are observed at the interface (substrate/layer) owing to twin formation [5]. Therefore, the necessity to optimize growth conditions has motivated numerous structural studies. In particular, defect density measurements in I I - V I semiconductor layers grown by MBE have been carried out using ion beam analysis. Rutherford backscattering spectrometry (RBS) and ion channelling (IC) have been used to determine the layer thickness, the lattice location of impurities in sublattices and the defect density [6]. Usually, structural measurements were performed e x s i t u since the MBE and ion beam analysing systems are generally not coupled, but, as emphasized above, this is a disadvantage for samples easily oxidized or sensitive to surface contamination. To overcome these difficulties, it appeared useful to design an MBE system on line with a 2.5 MeV van de Graaff accelerator. The aim of this paper is to describe such a system and to discuss some aspects of the use of in s i t u RBS and ion channelling to characterize the rate deposition, effusion cell flux measurements, layer homogeneity and the interfacial defect density during the CdTe growth.

2. Experimental set-up and results The samples are prepared using an MBE system designed by our laboratory and built by MECA-2000. This system is equipped with eight effusion cells (Cd, Te, CdTe, Eu, Sm, Nd, Au and In) (Fig. 1). The crucibles are in boron nitride except that for Nd, which

© 1994 - - Elsevier Science S.A. All rights reserved

P. Gros et al. / In situ characterization

Ion B e a m

4 0 0 0 ~ , ~ "~ 3000 M '~ 2000

(a) 200

CdTe

1oo0 o

"1/

~ Eu . 500 600 channel

o

400

~

267

o f rare earth-CdTe heterostructures

9

T(*C) 350300

"~

700

200

100

lO

4

I

:::1

"-•1

1

8

|

:I m

7 o ¢9

Ion B e a m

(b)

Fig. 1. (a) Schematic diagram of the molecular beam epitaxy system coupled to the van de Graaff ion accelerator, h deposition chamber; 2: introduction chamber; 3: differential pump (the two diaphragms are 1 mm in diameter, in order to guarantee a base pressure of 10 -1° Torr in the deposition chamber and 10 -7 Torr in the accelerator line); 4: ionic pump; 5" Ti sublimator; 6: ionic pump plus Ti sublimator; 7: turbomolecular pump; 8: cryogenic pump; 9, 10 and 11: valves. (b) Schematic diagram of the deposition chamber, h sample holder with xyz translations and O-q~ rotations; 2: RHEED electron gun (35 keV); 3: RHEED fluorescent screen; 4: Si surface barrier detector protected by a shutter to prevent contamination during growth (the detection angle is 157°); 5: effusion cells.

is in Ta. The growth conditions and surface crystallinity are controlled by reflection high energy electron diffraction ( R H E E D ) . The thickness of the deposited layers and their crystallinity are studied by RBS (2.5 MeV van de Graaff accelerator on line with the MBE via a differential pumping system). The bakeable surface barrier detector is protected by a shutter to avoid any contamination during growth. Consequently, RBS measurements are performed only during growth interruptions.

2.1. Layer thickness and thickness homogeneity The substrates used in this study were commercial (100) and (I11) C d , _ x Z n x T e wafers (x =0.04). Prior to their introduction in the MBE chamber, they were chemically etched in a bromine methanol solution. The thermal treatment in the chamber consisted of annealing at 230 °C for about 30 min, followed by heating at 340 °C. CdTe growth was initiated at this temperature

2

I

[

15

III

2'0 2'5 3'0 19000/kT (eV-t)

35

Fig. 2. Top: Rutherford backscattering spectrometry spectrum for 2 MeV 4He incident on 10 t6 Eu atoms deposited on (001) CdTe at 340 °C. Bottom: amount of incorporated Eu atoms measured by RBS vs. CdTe substrate temperature. Each experimental point is a limit between two growth regimes characterized by RHEED patterns (along the [110] CdTe azimuth) shown in the inset. The lower region is related to a monophase alloy formation ( C d - T e - E u ) while the upper one shows the presence of at least two different compounds.

for about 10 min. The substrate temperature was then lowered to 320 °C for the growth of a 1.5-2 lam thick CdTe buffer in order to relieve the mismatch with the Cdt_,.Zn.,.Te. (The typical growth rate was 0.7 lam h - J . ) When depositing a rare earth metal-(Sm, Nd or Eu) on CdTe for a substrate temperature in the range of 100-350 °C, a reaction-diffusion process takes place, which results in the formation of specific compounds due to the high chemical reactivity between the RE and the I I - V I semiconductors [2, 3, 6]. R H E E D analysis shows at first that the interfacial layer (up to a few nm thick) is matched to the CdTe lattice, but then a drastic change in the R H E E D patterns occurs. The total amount of incorporated rare earth is measured in situ by RBS, leading to the determination of the "phase diagram" shown in Fig. 2. The experimental line (i.e. the limit between the monophase alloy formation and the growth of two or more phases) is associated with extra spots appearing on the characteristic R H E E D patterns of the thin interfacial layer. These extra spots probably correspond to the nucleation of 3D metallic rare earth or another rare-earth phase. Two regions are observed as a function of temperature: in the low substrate temperature region, the rare earth incorporation rate was nearly constant, whereas at high temperature it was found to increase as a function of Ts [31.

268

P. Gros et a L / In situ characterization o f rare earth - CdTe heterostructures

The uniqueness of the RBS technique is to provide absolute measurements of very small quantities. For example, we show in Fig. 2 the RBS spectrum (2 MeV, 4He) corresponding to an Eu-based interfacial layer evaporated at 613 K. The analysis of the Eu peak area gives an absolute value of 10 t6 Eu atoms, the detection level limit for heavy atomic mass materials being much less than one monolayer. Moreover, such an in situ analysis is required owing to the great oxidizability of these very thin RE metal layers. RBS analysis was also used to determine the thickness homogeneity of the elaborated films. The relative standard deviation measured on the surface sample (20 mm × 20 mm) is found to lie between 5% and 10% depending on the position of the effusion cell in the vacuum chamber. 2.2. Deposition rate o f the materials

The deposition rate (in at. cm -2 s -1) is also measured by RBS. The experimental procedure consists in evaporating the material onto a substrate at fixed temperature and measuring the amount of deposited atoms for different cell temperatures. It was checked that, for a fixed cell temperature, the deposition rate of the materials listed in Table 1 was nearly constant and independent of the substrate temperature (in the room temperature range). For a sticking coefficient equal to 1, the deposition rates are related to the effusion cell fluxes by a geometrical factor. Moreover, the growth rate (nm s -l) can be calculated assuming that the thin film density is known. We have reported in Fig. 3, for the eight cells, the flux values qJ of the atoms sticking to the substrate vs. 1 / k T (where T is the cell temperature). The slope In 4~/(1/T) is independent of the absolute values of the sticking coefficient provided it is constant and independent of the incident flux in the range under consideration. According to the Knudsen law, the vaporization (or sublimation) enthalpy AH is expressed by A H = - R l n ~ / ( 1 / T ) + R T / 2 . From the slope measurements (Fig. 3), we deduce the AH values reported in

TABLE 1. Measured enthalpies of vaporization or sublimation (eV at. -~) deduced from Fig. 3 and compared with literature data Material

This work

Trange (K)

Literature

T range (K)

Ref.

Au Cd CdTe Eu In Nd Sm Te2

3.4 1.0 1.6 1.5 2.2 3.1 1.6 1.4

1373-1648 443- 503 493-623 623-823 1063-1163 1083-1533 723- 893 523-653

3.6 1.1 1.6 1.8 2.4 3.2 2.1 1.7

1100-1300 400- 500 731-922 693-900 1050-1150 1135-1500 500-1000 723-785

7 7 8 7 7 7 7 8

T ("C)_ 5+00 ~'lO

I+S0

15

E

CdTe ~Tez \.Cd

A

]

'

w, I0 n

,:

m

lO g . . . . . . . . . . . . . . . . . . . . . . . . . 4

8

12

i/kT

16

20

24

28

(eV -I)

Fig. 3. Flux measurements(at. cm-2 s -z) of the elementsstuck on the substrate, determined by RBS, vs. l/kT (where T is the cell temperature). The slope In ~/(l/T) of each curve allows the determinationof the enthalpies of vaporization or sublimation.

Table 1. Our results (except for Sm) agree rather well (within a deviation of 20%) with the published data. The smaller discrepancies can be assigned to differences in the measurement temperature range, to a temperature gradient along the cells or to a deviation from the Knudsen law due to the particular cell geometry. The Sm discrepancy is quite large and may be due to hydrogen incorporation in the metal: the H2 residual pressure in the evaporation chamber directly depends on the temperature of the Sm cell. Sm hydride stoichiometry embraces a considerable range of composition [9] and the cohesion energy of the compound can be considerably altered. It is quite difficult to decrease H content because the dissociation of Sm hydrides occurs at about the same temperature as Sm vaporization. 2.3. Structural defect measurements

During the growth process, numerous types of dislocation (misfit, threading) as well as antiphase boundaries can appear in the growing layer. Furthermore, for the [111] growth direction twin boundaries can occur. These various types of extended crystallographic defects can be studied in situ by RBS in channelling geometry [10], the energy dependence of the channelling giving the nature of the defects (the results can be confirmed by high resolution electron microscopy). Furthermore, the defect density corresponding to the different stages of the growth can be Obtained on a single sample which remains in the UHV chamber. The occurrence of misfit dislocations during the growth of (100) CdTe on Cd0.96Zno.oaTe (Aa/a ~ 3 x 10 - 3 between the two materials) is illustrated in Fig. 4. For such a misfit, the critical thickness is about 300 nm ( c f details of the method in Ref. 4). Curve (a) is the RBS spectrum in the (110) channelling plane for a CdTe thickness of about 210 nm, smaller than the critical thickness (strained CdTe layer). Curve (b) corresponds to a thickness of about 600 nm (after the relaxation of

P. Gros et al. /In situ characterization of rare earth-CdTe heterostructures

269

2000 !

(c)

1500

..

1500 "0 "~

(a) ':.z.





1000

~

.

(b)

500 0 o

......

600

• 700

......

~ ....

800

900

~ m

O, 400

Channel

Fig. 4. 4He RBS spectra (E=2.4 MeV) for (001) CdTe/(001) Cdo.96Zno.o4Te.Curve (a): aligned spectrum through the (110) plane before the relaxation of a 210 nm thick CdTe layer. Curve (b): appearance of misfit defects (see step) due to the relaxation of a 600 nm thick CdTe layer. Curve (c): random spectrum.

the CdTe layer). The step height on this spectrum is related to the occurrence of misfit dislocations at the interface, inducing a dechannelling component. The dislocation length per unit surface No at the interface is proportional to the dechannelling probability P• = (ZD -- ZV)/(1 -- ZV), where 1 -- Xv and 1 - ZD represent the number of channelled particles before and after the interface respectively. PD also measures the product NDaD, where ao is the dechannelling crosssection per unit length of dislocations. For a planar channelling of 2.4 MeV 4He in CdTe, aD is around 20 nm [4]. The sensitivity of the misfit dislocation density measurements determined from planar channelling can then be estimated. For a mean distance d between two successive misfit dislocations in a square array, the dechannelling probability is PD = 2aD/d. Assuming we can detect a step corresponding to a minimum dechannelling ratio o f 5% (Po = 5%) we deduce that the minimum distance between two dislocations must be d = 800 nm. This corresponds to the misfit dislocation density o f a fully relaxed layer having a lattice misfit o f about 3 × 1 0 - 4 with respect to its substrate. It is worth noting that the above method is suitable for a quick determination o f critical thickness larger than about a few tens of nm. RBS coupled to MBE allows to carry out interesting new experiments, such as the study of kinetic strain relaxation o f layers o f various heterostructures. Figure 5 refers to the homoepitaxy of (111) CdTe. It is well known that twins are formed in this growth direction. Although channelling experiments along the [111] direction do not allow us to distinguish the twinned domains, the twinned boundaries induce a dechannelling effect. This effect leads to the step observed on the aligned spectra shown in Fig. 5. It is noticeable that a large density of twin boundaries takes place at the interface for the first stages of growth (see spectrum (b)). On spectra (c) and (d), it is shown that

500

600

700

Channel.

Fig. 5. 4He RBS spectra (E = 2 MeV) for CdTe(lll) homoepitaxy. Curve (a): aligned spectrum on nominal (111) CdTe. The steps on curves (b), (c) and (d) indicate that interfacial defects are buried respectively by 140, 340 and 700 nm of CdTe,

the defective region is buried under a CdTe overlayer. The overlayer crystalline quality increases with thickness. For a layer thickness larger than about 1 lam, the CdTe crystallinity (determined by the channelling) is comparable to that of the bulk. We stress that the magnitude of planar dechannelling due to twin boundaries is in the range of that introduced by one surface monolayer (around 1015 at. cm-2). Assuming that twin boundaries induce a dechannelling similar to that of amorphous regions located between (110) planes and that the thickness of the twinned layer lies 50 nm down, we can estimate that the average distance between two twin boundaries (in the (111) growth plane) is about 200 nm [10]. The amount of twinned region has been previously studied by Foti et al. [11] using dechannelling along different orientations. To summarize this section, we want to stress that RBS and channelling are efficient tools for characterizing in situ the epitaxial growth, with quantitative analysis requiring knowledge of the scattering cross-sections for each type o f defect.

3. C o n c l u s i o n

We have described an MBE apparatus coupled with a 2.5 MeV van de Graaff accelerator. This apparatus displays two major advantages. First, the in situ characterization of highly reactive thin films is allowed. As no encapsulation is needed, direct correlation between R H E E D patterns and the amount of incorporated rare earth (as low as 10 '5 at. cm -2) is made possible. Secondly, the possibility of achieving on a single sample in situ growth characterization (determination of critical thicknesses and of extended defect density) results in a significant gain of time compared with classical ex situ experiments. The results obtained demonstrate that the original association of MBE and ion beam analysis is a

270

P. Gros et aL/ In situ characterization of rare earth - CdTe heterostructures

powerful tool for the in situ study of highly reactive heterostructures and of growth defects.

Acknowledgments The authors wish to thank G. Demoment, G. Henrissat and A. Vandrot for their technical help.

References 1 T. Sands, C. J. Palmstram, J. P. Harbison, V. G. Keramidas, N. Tabatabaie, T. L. Cheeks, R. Ramesh and Y. Silberberg, Mater. Sci. Rep., 5 (1990) 99. 2 P. Gros, A. C. Chami, B. Daudin and E. Ligeon, Appl. Phys. Lett., 61 (1992) 1335.

3 B. Daudin, P. Gros, E. Ligeon and A. C. Chami, Appl. Surf. Sci., 65/66 (1993) 821. 4 A. C. Chami, E. Ligeon, R. Danielou, J. Fontenille, G. Lentz, N. Magnea and H. Mariette, Appl. Phys. Lett., 52 (1988) 1874. 5 J. Cibert, Y. Gobil, K. Saminadayar, S. Tatarenko, A. C. Chami, G. Feuillet, L. S. Dang and E. Ligeon, AppL Phys. Lett., 54 (1989) 828. 6 A. C. Chami, B. Daudin, J. Fontenille, P. Gros and E. Ligeon, J. Appl. Phys., 74 (1993) 237. 7 R. Hultgren, R. L. Orr, P.D. Anderson and K. K. Kelley, in Selected Values of Thermodynamic Properties of Metals and Alloys, Wiley, New York, 1963. 8 K. Mills, in Thermodynamic Sulphides, Selenides and Tellurides, Butterworth, London, 1974. 9 O. Greis, P. Knappe and H. Muller, J. Solid State Chem., 39 (1981) 49. 10 L. C. Feldman, J. W. Mayer and S. T. Picraux, in Materials Analysis by Ion Channeling, Academic Press, New York, 1982. 11 G. Foti, L. Csepregi, E. F. Kennedy, P. P. Pronko and J. W. Mayer, Phys. Lett., A, 64 (1977) 265.