JOURNAL DE P H Y S I Q U E
Colloque C5, suppl6ment au n O l l , Tome 48, novembre 1987
THE USE OF COMPOSITIONALLY GRADED LAYERS TO MINIMIZE SURFACE DEFECTS IN In(AsSb) STRAINED-LAYER SUPERLATTICES
R.M. BIEFELD
Sandia National Laboratories, Albuquerque, NM 87185, U.S.A.
Abstract.--Surface defects have been studied in InAs Sb - /InSb strained-layer euperlat t ices as a tunct ion o f the pr8f iiexof
compositionaly graded buffer layers. Comparisons were made between constant composition, step-graded and continuously graded buffer layers. The use of either constant composition layers or step-graded buffer layers resulted in an increase in surface defects for large lattice mismatch (x>0.1). Surface defects were minimized by the use of continuously graded buffer layers for X = 0.2.
Hgl-xCdxTe solid solutions are widely used for infrared detectors [l].
However, these materials suffer from the diaadvantages of poor lattice atability due to relatively weak bond strengths and large bandgap changes resulting from compositional variations [l]. In(AsSb)
strained-layer superlattices (SLS's) have recently been proposed for use as long wavelength detectors in the 8-12 micron range [l].
Theoretical calculations have shown that the layer strains in
InAs Sb /InAs Sb SLS's decrease the bandgap below that obtainable fromxthk-6ulk I A S S alloys [l]. ~ ~ These s L s s s should exhibit
several advantagesXover the Hg CdxTe materials which include greater bond strengths and smaller ban&& changes from compositional
variations. Both MBE and HOCVD have been used to grow In(AsSb) SLS's [2-61. These structures have been characterized by TEM and x-ray diffraction. The growth of SLS's in the temperature range of 425-475 C is evidence for the superior bond strength and structural stability of these 111-V materials when compared to the Hg-rich 11-V1 materials.
Further theoretical studies indicate that structures with unstrained barrier layers within the unit cells of SLS's can suppress the quantum size effect by up t o 3 0 rev and thus extend the wavelength
response(5).
In order for the In(AsSb) SLS to have a bandgap below that o f the bulk alloys, the quantum well must be under tension [l]. It is well known that when thick, mismatched layers are grown under tenaion cracks will form [7,8]. The cracking has been minimized in
layera grown on QaAs by the use of continuously graded buffer la ers [g]. The preparation of thick In(AsSb) SLS's on constant composition buffer layers for absorbance reaeurerents resulted in samples with large numbers of cracks on the surface. If a euperlattice is grown on an epitaxial layer which is not lattice matched t o the lattice
conmtant of the strained layers then residual strain can be present in the total superlattice 121. Large amounts of misfit between the superlattice and the underlying graded layer can lead to the formation of dislocations and microcracks if the SLS is in tension with respect to the buffer layer [2,7.8]. If cracks are present in the buffer layer, they will be replicated in the SLS. This paper will discuss recent HOCVD growth studies on 1 n A ~ ~ S b ~ - ~ / I n S b SLS's which have nhown
Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphyscol:1987513
JOURNAL DE PHYSIQUE
that the densities of defects and ricrocracks can be minimized by the careful use o f corpositionally graded alloy buffer layers.
2. Experimental procedures.
Two growth systems were used to prepare the samples described in this paper. One system, indicated by the letter V in the sample n u b e r i n g , has been previously described (21, and it consisted of a vertical, atmospheric pressure, glass reaction chamber and a stainless steel gas handling system. In this system the areine was run continuously into a vent line or into the reaction chamber. The other system was a horizontal, atmospheric pressure system with a quartz reaction chamber and a vent-run gas handling system with automatic pressure balancing
(CVD Bquiprent Corp., Uodel 5208). The sources were trimethylindium (TMI), trimethylantirony (TNSb), and arsine, AsH3, in H2. Pd-diffused H was used as the carrier gas. The layers were grown at 475 C on
(900) InSb substrates. The structnres prepared in this work consisted of the InSb substrate, an InSb epitaxial layer, a buffer layer and an uppermost superlattice. The buffer layers were of three types: 1) a constant composition layer of InAs Sb in which X was approximately 1/2 the X in the SLS; 2) a step-gr~deA-fayer consisting of three steps in which the corposition of the final step was identical to the
compostion of the layer in 1); 3) a continuously graded layer where X increased with tire, t, as described by X = x(final)*[l-exp(-t)]. The SLS's considered in this paper consisted of alternating layers of InSb
The superlattices were prepared by a purge-growth :
: : h :
; : : r
?
&
~
%
; has been described previously [2] . In the purge- growth technique the reaction chamber is purged with hydrogen between layer growth and the reactants are either turned off or allowed to flow into a vent line. The composition of the buffer layers and the superlattices were determined from x-ray diffraction by the use of Vegard's Law. The layer thicknesses were determined from
photomicrographs of shallow angle grooves and in sore instances from cross-sectional TBU and x-ray diffraction.
3. Results and discussion.
The surface morphology of four typical InAs Sb /InSb SLS samples is shown in figure 1. The sample in figure l~~ia'ii~ten period
A Sb /InSb SLS (V170) with 38.0 nr thick layers. This SLS was grow#.& ? b f of an 800 nm thick InAs Sb constant corposition buffer layer and a 400 nr thick i n i t f d layer. Transmission electron ricroscopy (TEM) of this sample revealed a number o f dielocations at the interface between the InSb layer and the
InAsO-lSb g buffer (2). However, there were no dislocations at the buffer l a g i r - ~ ~ ~ interface or within the SLS 121. Pigure 1A shows the presence of the usual cross-hatched pattern that is associated with mismatched epitaxial layers, but there are also several rather distinct features which are cracks on the surface of this sample.
These cracks are due to the thickness of the InAsO &SbOs9 buffer layer. Matthews and Klokholr [7] have shown that t ere is a critical layer thickness beyond which epitaxial layers that are in tension will form cracks or fractures if the mismatch is not relieved by
dislocations. For an InAs Sb layer on InSb this critical layer thickness is 130 nm. 0.1 0.9
Pigure 2B shows the surface of a 20 period InAeo.2Sbo.8/InSb SLS (CVD98) with 25.0 nm thick layers. This SLS was grown on a step- graded layer which consisted of an initial 120 nr InSb layer, a 500 nr InAs Sb layer, a 500 nr I n A s 0 ~ 0 6 S b O ~ g d layer and a final 500 nm 1n~sO'O8b 0-92ayer. These buffer layers wer chosen to be thick enou#h1s00t2at they would decouple from the substrate by forming dislocations and yet thin enough s o that they would not fracture and
form microcracks. No microcracks can be seen in figure 2B. However, similar structures uith thicker SLS's on step-graded buffer layers have microcracks on the surface and their surface morphology looks similar to the morphology of figure 1A. If the tension in the step- graded buffer is not completely relieved by the formation of
dislocations then cracks can form in either the buffer layer or the SLS [2,7]. The critical layer thickness for the formation of
dislocations or cracks is a necessary but not a sufficient condition for their formation [7,10]. Under conditions in which the mismatch in the atep-graded buffer layers is completely relieved by the formation of dislocations, cracks will not form and dislocation and crack free SLS's can be grown on top of step-graded buffer layers.
Pig. l--Surface morphology of Fig. 2--A transmission electron InAs Sbl-x/InSb SLS's grown on micrograph of CVD 71.
diffgrent types of buffer layers.
Figures 1C and 1D illustrate the surface morphologies o f SLS's grown on top of continuoualy graded buffer layers. Figure 1C is the surface of a 20 period InAso 2Sb0 8/InSb SLS (CVD 104) with 14.9 nm thick layers. This SLS was qrown'on an 1800 nm thick, continuously graded buffer layer with a final composition of InAs
structure in figure 1D has an SLS of the same c o r p o s ~ t tSb onO&?th The 11.5 nm thick layers and a similar buffer layer which is only 900 nm thick (CID 71). The surface morphology o f figure 1C shows no cracks while the morphology of figure 1D does indicate the presence of some cracks.
X-ray diffraction of CVD 71 indicates that there is residual strain in both the buffer layer and the SLS. This residual strain in the buffer
layer occurrs when not enough dislocations or cracks form to relieve the mismatch between the buffer layer and the substrate. Residual strain in the SLS occurrs when the lattice constant of the SLS is not the s-e ae that of the underlying buffer layer. If the total
thickness of the superlattice exceeds the critical layer thickness for the mismatch between the buffer layer and the SLS then dislocations
CS-84 JOURNAL DE PHYSIQUE
and cracks could form in the SLS. For CV0 71 the measured parallel lattice constant for the buffer layer is 0.64574 nr. The calculated parallel lattice constant for a free standing SLS for CID 71 would be 0.64363 nm. The SLS has a residual strain o f 0.0034. For this strain the critical layer thicknesses for dislocation and crack formation are 200 and 6 0 0 nm, respectively. Since the SLS has a total thickness of 460 nm some dislocations might form but cracks would not be expected to form. The mismatch which is accorodated by the buffer layer is 0.0065 for which the critcal layer thickness for dislocation and crack formation are 100 and 160 nm, respectively. Since the buffer layer is 900 nm both dislocations and cracks would be expected to form in the buffer layer which explains the surface morphology in figure 10. The difference between the structures of figure 1C and 10 is the thickness of the SLS and the buffer layers. The x-ray analysis of CV0 104 indicates that the buffer layer is completely relaxed with a parallel lattice constant of 0.64381 nr. This gives a residual strain of 0.0004 and the corresponding critical layer thicknesses of 2110 and 48710 nm for dislocation and crack formation [7,10]. These numbers are consistent with the absence of cracks in figure 1C. Apparently the thicker buffer layer of CVD 104 causes dislocations t o form during the layer growth and reduces the riafit so that cracks do not form.
Figure 2 is a cross sectional transmission electron micrograph of CVD 71. There are a number of dislocations at the interface between the InSb layer and the start of the InAs Sb graded layer. There are no dislocations in the SLS in this pgrticglar view, but there were a few in other areas of this sample. This micrograph was chosen to indicate that there are large areas of superlattice that do not contain dislocations and to show the uniformity of the superlattice and the sharpness o f the interface..
4. Conclusions.
The compositional profile of the underlying epitaxial graded layer was found to strongly affect the density of surface defects.
The surface defects can be minimized by the careful grading of the buffer layers. Thick buffer layers which are continuously graded yield the surfaces with the fewest defects. Dislocation free S L S J s which should be useful for device applications have been grown in the InAsxSbl-x/InSb system up to X = 0.2.
*This work was performed at Sandia National Laboratories supported by the 11.8. Department of Energy under contract number DB-AC04-76DP00789.
References
[l] OSBOURN, G. C., J. Vac. Sci. fech. (1984) 176.
[2] BIBFBLD, R. M., J. Cryst. Growth, 77 (1986) 369.
[3] DAWSON, L. R., J. Vac. Sci. Tech. (1985) 598.
[4] LBB, G. S., L O B Y., LIN, Y. F., BBDAIR, S. H., LAIDIG, W. D..
Appl. Phys. Lett. 47 (1985) 1219.
[5] OSBOURN, G. C. , in Proc. U. S. Workshop on the Physics and Chemistry o f BgCdfe, 1986.
[6] YEN, U. Y., LBVINB, B. F., BBTHBA, C. G., CHOI, X. X., CHO, A. Y., Appl. Phya. Lett. 50 (1987) 927.
[7] HATTRBWS, J. W., KLOKHOLM, B., Mat. Res. Bull. 7 (1972) 213.
[8] OLSBN, Q. H., ABEAEAMS, M. S., ZANEROWSKI,
T. J., J. Hlectrochu. Soc. 121 (1974) 1650.
[g] XASANO, E., HOSOKI, S., J. Appl. Phys. 46 (1975) 394.
1101 UATTHBWS, J. W., J. Vac. Sci. fecbnol. 1 2 (1975) 126.
