Photoluminescence study of SiGe quantum well broadening by rapid thermal annealing

,. . . . . . . . C R Y S T A L QROWTH

ELSEVIER

Journal of Crystal Growth 157 (1995) 57-60

Photoluminescence study of SiGe quantum well broadening by rapid thermal annealing H. Lafontaine a, *, D.C. Houghton a, N. Rowell b, G.C. Aers a, R. Rinfret b Institute for Microstructural Sciences, National Research Council, Ottawa, Canada, K1A OR6 b Institute for National Measurement Standards, National Research Council, Ottawa, Canada, K1A OR6

Abstract

Photoluminescence was used to measure the broadening of thin (t = 12-77 ,~) SiGe quantum wells (QW) under typical RTA conditions. An anneal time of 300 s and temperatures ranging from 800 to 1000°C were used. "No phonon" SiGe transition energy shifts of up to 65 meV are measured. Results are analyzed taking into account the initial diffusion during growth, the increase in QW bandgap due to intermixing and the decrease in quantum confinement. Interdiffusivity values showing an Arrhenius behavior and an activation energy of 2.7 eV are obtained.

1. I n t r o d u c t i o n

There has been increasing interest in the silicon germanium system for applications in both electronics [1,2] and photonics [3] devices and circuits. The addition of a single SiGe layer can significantly increase device speeds compared to similar "silicon only" designs. However, the fabrication of integrated circuits (IC) usually involves rapid thermal annealing (RTA) steps, e.g. for dopant activation, oxidation, etc. [4]. The thermal budget becomes critical when SiGe epitaxial layers are used [5], due to diffusion effects and strain relaxation. In order to define an upper limit for acceptable thermal exposure in a fabrication process, it is important to know precisely the diffusivity of germanium and silicon in thin SiGe layers under typical anneal conditions. Photoluminescence [6,7] allows the measurement

* Corresponding author. Fax: hugues@ m50sci.lan.nrc.ca.

+1

613 941 4667; E-maih

of changes in the width of quantum wells at temperatures significantly below those needed to produce structural changes which can be observed with other techniques such as X-ray diffraction and Rutherford backscattering (RBS). Recently, photoluminescence was used as an optical technique to study interdiffusion during furnace annealin~ of SiGe quantum wells with well widths of 68-73 A [8,9]. In this work, we use photoluminescence to study the effect of rapid (300 s), low temperature (800-1000°C) anneals, similar to those typically used in IC fabrication, on SiGe quantum wells with widths between 12 and 77 ,~ and Ge fractions of 0.17 to 0.38.

2. E x p e r i m e n t a l p r o c e d u r e

Growth was done by conventional, solid-source MBE without intentional doping, in a VG Semicon V80 system using procedures described earlier [10]. The heterostructures were grown on 100 mm Si(100) substrates at constant temperatures ranging from 590

0022-0248/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0 0 2 2 - 0 2 4 8 ( 9 5 ) 0 0 3 6 3 - 0

58

H, Lafontaine et al. /Journal of Crystal Growth 157 (1995) 57-60

Table 1 Details of the samples used in this study; all samples have spacer and cap layers of 200 ,~, Sample 1 2 3 4 5

Growth

Germanium

Well thickness

temperature (°C)

fraction

(,~)

595 591 591 799 792

0.17 0.38 0.38 0,17 0.17

43 22 12 68 77

' '

_

'

'

'

S,Ge'.P

\

,

_ _

L

as grown.

990

1030

1070

1110

1150

Energy (meV)

Fig. l, PL spectra obtained after increasing anneal temperatures on sample 5 of Table 1.

function of anneal temperature for all 5 growths of Table 1, after 300 s anneals at temperatures ranging from 800 to 1000°C. The general direction of the energy shifts is the same for all samples (i.e. only blue shifts are observed). However, significant differences can be seen between samples in the importance of the shifts and the temperature at which they are occurring. These differences are resulting from a combination of factors affecting the diffusivity of the material, such as growth temperature, germanium fraction, and structural effects, predominantly well width. For instance, the effect of growth temperature can be appreciated by comparing samples 1, 2, and 3, which were grown at 591 to 595°C, and samples 4

~

60

43 A

o o ×

50

~A

12A

sea

/

/0 o/ /

Anneal time = 300 s

t,

"'

~2

/

3. Results and discussion

Fig. 1 shows typical photoluminescence spectra taken on samples from growth 5 (see Table 1), after anneals at increasing temperatures. Two SiGe peaks can be seen, the " n o phonon" (NP) peak and its "transverse optical" (TO) replica [6]. As the anneal temperature increases, the shift towards higher energy becomes more pronounced and also the intensity ratio between NP andTO peaks is modified. Fig. 2 shows the shifts in the NP peak energy as a

,

.-~

950

to 800°C. The multiquantum well structures (3 or 10 periods) were grown with a germanium fraction x ranging between 0.17 and 0.37, with well thicknesses (t = 12-77 A) below the critical value predicted by Matthews and Blakeslee [11] for relaxation ignoring kinetics barrier. The composition and thickness were determined by double crystal rocking curve analysis and cross-sectional TEM. Sample data are given in Table 1. Samples were cleaved out of a single piece and rapid thermal anneals (RTA) were performed using an A.G. Associates, Heatpulse 410 system. A reducing gas mixture with a hydrogen content of 4% diluted in nitrogen was introduced into the anneal chamber several minutes before temperature cycles, in order to purge the system of any residual oxygen. Photoluminescence spectra were recorded using a Fourier transform infrared spectrometer with the samples immersed in liquid helium ( T - - 2 K). The excitation wavelength was 458 nm (Ar + laser) and the power density at the sample was ,-, 1 W / c m 2. The luminescence was measured by a germanium detector at 77 K.

'

lO00°Canoea,

30

lo o "10

i,,I

750

....

800

I . . . .

850

i

,

900

,

j

i

I

i

950

i

,

,

I

i

1000

i

i

i

1050

Temperature (°C) Fig. 2. Energy shifts as a funcdon of anneal temperature for five

different well thicknesses corresponding to samples of Table 1.

59

H. Lafontaine et al. / Journal of Crystal Growth 157 (1995) 57-60

and 5, grown at almost 800°C. In spite of differences in composition and well widths, it is clear that the energy shifts measured are much smaller for the last two samples, as no significant change in the NP peak position is seen at temperatures below 850°C. In contrast, samples 1, 2, and 3 show large NP emission energy shifts at an anneal temperature of 800°C, e.g. 20 meV for sample 2. We attribute the lower energy shift in the case of samples 4 and 5 to an initial "in situ" anneal during MBE growth at higher temperature (i.e. 800 versus 600°C). This caused significant interdiffusion and well broadening during growth, therefore limiting further evolution afterwards. In fact, the PL transition energies for the as-grown samples 4 and 5 were consistent with significant diffusion during growth. Due to limited data, we can only examine the combined effect of germanium fraction and well width here, by comparing the results obtained for the first three samples. Sample 1 shows relatively low energy shifts, due to its lower germanium fraction (x = 0.17) and large well width (t = 43 A). Assuming a simple diffusion law behavior (Fick equation), one will expect the changes in composition and well width under annealing to be relatively limited. Sample 2 exhibits the largest energy shifts, due to its high germanium content and intermediate well width. Shifts of up to 65 meV in the NP transition energy are obtained. We attribute the magnitude of these shifts to large changes in the bandgap energy while comparatively small changes in the quantum confinement are anticipated. On the other hand, sample 3, with a well width of 12 A, presents a large initial confinement effect which, in turn, reduces the energy shifts measured on this sample. In order to obtain diffusivity values, we used a model that takes into account the initial diffusion during growth and the shift in the bandgap as well as the quantum confinement. The diffusion equation is solved numerically and a well profile is obtained, for which the calculated transition energy matches the photoluminescence data obtained from "as grown" samples. Further diffusion is then introduced into the simulation, with experimental anneal temperatures for a 300 s anneal time. The simulation proceeds until it matches corresponding PL data, and diffusivity values, D, are obtained. Fig. 3 shows the interdiffusivity values obtained for all 5 samples studied, as o

10 -13

10-14 1015 ~-

F 10-16 1017 ~ 10-18 I 6

i

i

i

i

I

,

6.5

,

,

,I

,

7

,

,

,

I

. . . .

I

i

7.5 8 104/T (k)

i

jl

]1

8.5

i

j

j

I

9

~

i

i

9.5

Fig. 3. Interdiffusivity of SiGe, for 5 different well thicknesses, compared to Si *, Ge, B and As in Si.

a function of anneal temperature. Assuming Arrhenius behavior, an activation energy 2.7 eV is calculated, which is close to the value obtained by Shiraki and co-workers [9]. An expression for diffusivity is thus obtained: DSiGe =

D Oexp( -

2.7/kBT),

where k B is the Boltzmann constant (8.617 X 10 -s eV K - l ) , T is the process temperature (K), DsiGe is the diffusivity as a function of temperature and D O is the pre-factor expressing the diffusivity at infinite temperature. From the data of Fig. 3, we can extract a value for D O of 2.8 X 10 -5 cm 2 s -1. For comparison purposes, the diffusion coefficients of Si * selfdiffusion Ge, B and As impurities in Si [4] are also plotted in Fig. 3. The interdiffusivity of SiGe for 300 s anneals appears to be intermediate between extrapolated data for As and Si * impurity diffusion in Si. These data indicate that intermixing of SiGe is not strongly enhanced by the strain gradient which accompanies the Ge profile.

4. Conclusion The broadening of SiGe quantum wells during rapid thermal annealing was investigated by photoluminescence spectroscopy. Energy shifts as a function of anneal temperature are in qualitative agreement with predictable effects of growth temperature, Ge composition and well width. A model which takes

60

H. Lafontaine et al./ Journal of Crystal Growth 157 (1995) 57-60

these last three variables into account allowed us to calculate interdiffusivity values at temperatures as low as 800°C. Photoluminescence appears to be a very sensitive and reliable technique for measuring the effect of low temperature RTA on very thin SiGe wells.

References [1] B.S. Meyerson, K.E. Ismail, D.L. Harame, F.K. LeGoues and J.M.C. Stork, Semicond. Sci. Technol. 9 (1994) 2005. [2] U. KSnig, Mater. Res. Soc. Symp. Proc. 281 (1993) 387. [3] R.A. Soref, Proc. of the IEEE 81 (1993) 1687.

[4] S.M. Sze, Physics of Semiconductor Devices (Wiley-Interscience, New York, 1981). [5] D.C. Houghton, J. Appl. Phys. 70 (1991) 2136. [6] J. Weber and M.I. Alonso, Phys. Rev. B 40 (1989) 5683. [7] N. Rowell, J.P. Noel, D.C. Houghton, A. Wang, L.C. Lenchyshyn, M. Thewalt and D.D. Perovic, J. Appl. Phys. 74 (1993) 2790. [8] H. Sunamura, S. Fukatsu, N. Usami and Y. Shiraki, Appl. Phys. Lett. 63 (1993) 1651. [9] H. Sunamura, S. Fukatsu, N. Usami and Y. Shiraki, Jap. J. Appl. Phys. 33 (1994) 2344. [10] D.C. Houghton, D.J. Lockwood, M.W.C. Dharma-Wardana, E.W. Fenton, J.M, Baribeau and M.W. Denhoff, J. Crystal Growth 81 (1987) 434. [11] J. Matthews and A.E. Blakeslee, J. Crystal Growth 27 (1974) 118.