- Email: [email protected]
Cathodoluminescence study of InxGa1−xSb crystals grown by the Bridgman method
ARTICLE IN PRESS
Journal of Crystal Growth 268 (2004) 52–58
Cathodoluminescence study of InxGa1 xSb crystals grown by the Bridgman method M.F. Chioncela,1, C. D!ıaz-Guerraa,*, J. Piquerasa, J. Vincentb, b ! V. Bermudez , E. Die! guezb b
a Departamento de F!ısica de Materiales, Facultad de Ciencias F!ısicas, Universidad Complutense de Madrid, E-28040 Madrid, Spain ! Departamento de F!ısica de Materiales, Facultad de Ciencias, Universidad Autonoma de Madrid, E-28049 Cantoblanco, Madrid, Spain
Received 24 February 2004; accepted 3 May 2004
Communicated by KW Benz
Abstract The nature and the spatial distribution of radiative defects in InxGa1 xSb grown by the vertical Bridgman method have been studied by cathodoluminescence (CL) in a scanning electron microscope. The CL results have been complemented by X-ray microanalysis and backscattered electron imaging to relate the local luminescence properties to the chemical composition. Measurements of the band-gap energy from the CL spectra, supported by X-ray compositional mappings, reveal an effective incorporation of In in the matrix, leading to the formation of the ternary alloy in the whole volume of the ingot. A low gradient of the In content along the growth axis has been found. The CL spectra of the ternary alloy exhibit similar general features to those reported for GaSb. An observed red shift of the near band edge luminescence in InxGa1 xSb, relative to that of GaSb, is due to the reduction in the band gap with increasing x. A band often observed in the CL spectra, peaked at about 20 meV below the band-gap energy, is attributed to the presence in the ternary alloy of an acceptor level that would correspond to the VGa–GaSb acceptor in GaSb. r 2004 Elsevier B.V. All rights reserved. PACS: 78.60.Hk; 68.37.Hk; 61.72.Ff Keywords: A1. Cathodoluminescence; A1. Defects; A2. Bridgman technique; B1. InxGa1 xSb
*Corresponding author. Tel.: +34-91-394-4548; fax: +3491-394-4547. E-mail address: cdiazgue@material.fis.ucm.es (C. D!ıaz-Guerra). 1 Permanent address. Physics Department, Faculty of Chemistry, University of Bucharest, Bd. Elisabeta 14-12, 70346 Bucharest, Romania.
1. Introduction In the past decade, the III–V ternary alloy system GaSb–InSb has become of great interest for optoelectronic devices [1] and high-efficiency thermophotovoltaic (TPV) cells operating in
0022-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2004.05.015
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
2. Experimental method InxGa1 xSb ingots of about 12 mm diameter and a length of 40 mm with a 10% nominal In content were grown by the vertical Bridgman technique. The growth was performed in a conical quartz ampoule that was previously graphitised. No seeds were used. The ingots were prepared filling the ampoule with 1.75 g of In, 19.00 g of Sb and 9.62 g of Ga, all 6 N materials (provided by Goodfellow). The ampoule was evacuated to below 1 10 4 Torr, sealed, heated up to 950 C to melt the powder material, and kept at this temperature during 24 h. The melted material was homogenised by low shaking the furnace for 30 h (+/ 30 rotation around the horizontal position at 60 /h). After homogenisation the furnace was set vertical. The pulling rate was set at 1 mm/h and the growth performed in a temperature gradient of 40 C/cm. After solidification, the furnace was slowly cooled down to room temperature. Several samples were cut at various places along the length of an ingot as shown in Fig. 1. Three disks of 2 mm thickness were cut perpendicular to the growth axis and labelled D1, D2 and D3.
D3
L2
D2
Growth axis
conjunction with low-temperature radiators [2]. TPV cells with band gaps lower than GaSb are expected to be advantageous for low-temperature (o1000 C) non-wavelength selective TPV radiators because they provide more effective absorption of the black body infrared radiation [3]. This alloy covers an interesting low band-gap range, (0.80–0.25) eV, that can be controlled by adjusting the In content to match the requirements. Strong efforts have been undertaken to grow compositionally uniform, crack-free materials. Although InxGa1 xSb crystals have been grown on earth, it is extremely difficult to grow high-quality crystals under terrestrial conditions because, as the densities of the components are different, solute transport occurs due to buoyancy. Many studies have employed microgravity conditions [4–6] or terrestrial condition under high magnetic fields in order to avoid gravity-induced negative effects [7]. However, the growth of bulk good-quality crystals remains an experimentally difficult task due to problems related to chemical segregation and mechanical stresses associated with the significant difference (12.5%) between the tetrahedral radii of Ga and In [8–10]. The properties of the device employing InxGa1 xSb in the structure will ultimately depend on the material quality, but little information is available regarding the influence of defects on the optical and electrical properties of InxGa1 xSb crystals. In fact, very few PL studies of InxGa1 xSb epilayers have been reported in the literature [11–13], while cathodoluminescence (CL) in the scanning electron microscope (SEM)—an important technique to analyse the nature and the spatial distribution of defects in semiconductors—has been to our knowledge only very recently applied to characterize InxGa1 xSb crystals [14]. In this work, the nature of luminescent centres and their spatial distribution along the growth and radial axis of an InxGa1 xSb ingot have been investigated using CL microscopy and spectroscopy in the SEM. The information provided by CL was complemented by wavelength dispersive X-ray microanalysis (WDX) and backscattered electron (BSE) imaging.
53
L1
D1 Fig. 1. Scheme of the InxGa1 xSb ingot showing the positions and labels of the samples investigated.
ARTICLE IN PRESS 54
M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
Sample D1 corresponds to the bottom of the ingot, D2 to the middle part and D3 to the top of the crystal, respectively. Samples L1 and L2, of rectangular shape, correspond to longitudinal sections (see Fig. 1). This selection of the samples enables an efficient investigation of the distribution of the luminescent centres along both the growth and radial axis. All the samples investigated were polished with alumina powder to a mirror finish and characterised by secondary electron (SE), BSE, WDX and CL modes of the SEM. Spatial and spectral CL measurements were carried out in a Hitachi S 2500 SEM with a cooled ADC germanium detector. Measurements were performed at 90 K using a 20 kV accelerating voltageU CL spectra were recorded on each sample and deconvoluted using a sum of Gaussian line distributions in order to determine the different CL peaks contributing to the overall emission. The local In content has been determined from the position of the band-gap transition in the ternary alloy. In order to study the composition of the features observed in the CL micrographs and to understand the spectral distribution of the CL emission, mapping of the elements Ga, Sb, In as well as quantitative X-ray microanalysis were performed by WDX in a Jeol JXA-8900 M Superprobe, using an accelerating voltage of 20 kV. The same system was used to obtain the BSE images.
Fig. 2 shows the distribution of Ga and In and the corresponding CL and BSE micrographs of a selected area from sample D3. The lack of contrast in the Sb mapping (not shown) indicates a very uniform distribution of this element. Overall, our results indicate that a higher In content corresponds to a lower Ga concentration and that the Sb content is almost constant. CL panchromatic mappings of the investigated samples show similarities in the distribution of the radiative centres. All the samples display luminescent emission, however, with some areas of diminished or quenched CL emission. Depending on the sample considered, precipitate-like features of elongated shape, showing dark CL contrast, may appear in the bright background (Fig. 3). Similar features have been observed previously in In-doped GaSb samples [15]. A preferential agglomeration of areas of low CL emission can be observed, in all the samples investigated, in regions corresponding to the inner part of the ingot. The decrease of the CL emission towards the inner part of the ingot is more clearly observed in the longitudinal sections, as sample L1. This sample shows the most inhomogeneous distribution of non-radiative
3. Results and discussion The exposed surface of all the samples was mapped by low-magnification SE and panchromatic CL microscopy while several representative areas were also analysed by means of BSE and WDXU All the samples present micro- and macrocracks, but apart of these, SE images display smooth, featureless surfaces. The X-ray compositional mappings show in all the samples a very uniform distribution of Sb and some anti-correlated variation in the Ga and In contents. There is a correspondence between contrast observed in CL, BSE, and X-ray mappings that is not observed in SE images, confirming that the mentioned contrast is not related to surface topography.
Fig. 2. WDX mappings from sample D3 showing the In (a) and Ga (b) distribution. Darker greys correspond to a lower element content. (c) BSE image of the same area. (d) Corresponding CL micrograph.
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
55
Fig. 3. CL image from sample L2 showing dislocation-related dark features in a background of uniform emission.
recombination centres, with extended bright and dark CL domains, and also displays a rich variety of CL structures (Fig. 4). Similar CL structures can be observed to a less extent in all the samples. Inside the grains, high-magnification micrographs of areas of enhanced CL emission show dark, precipitate-like defects and the absence of subgrain boundaries. Some of the precipitate-like features appear randomly oriented, but most of them appear aligned in equally spaced parallel lines (Fig. 4a). The size and the spatial distribution of these CL features, despite the precipitate-like appearance, suggest that they might be related to dislocations. Moreover, X-ray microanalysis of the regions where these dark features appear show no variation of In content. The attributed nature of these features to dislocations is also in agreement with previous results, which indicate that the solubility of In in GaSb is rather high and the formation of In precipitates unlikely [15], while strain-induced dislocations is one of the main problems found in the growth of good-quality InxGa1 xSb crystals [16,17]. The CL images of the areas of diminished CL emission show well-defined dark sub-grain boundaries (see Fig. 4b). These grains have various polygonal shapes, some of them showing hexagonal contours (Fig. 4c). As stated above, our CL microscopy results reveal an
Fig. 4. CL images from samples D2 (a,b) and L1 (c), showing precipitate-like features aligned along parallel lines (a) and different polygonal grain structures and sub-boundaries (b,c).
inhomogeneous distribution of non-radiative recombination centres in the volume of the ingot, with areas of diminished or quenched CL emission. Some of these areas may correspond to the presence of a high In content, above x=0.15, leading to an alloy with band gap below the detection limit of about 650 meV of the Ge detector used. The presence of grains, sub-grains and dislocations which probably induced subsequent grain generation is revealed by higher magnification CL microscopy images. Representative CL spectra of samples D1 and D3 are shown in Fig. 5. A shift towards lower energies from the GaSb band-gap value at 90 K, 796 meV, is clearly observed in all the spectra. This observation reveals the formation of the ternary alloy along the whole ingot. CL bands giving rise to these spectra have been determined from the best fits to the experimental data using a sum of Gaussian line distributions. The In content, x, has been calculated from the peak energy of the asfitted band-gap emission [18]. Although some change of the peak position is observed in the
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
56
CL Intensity (arb. units)
Sample D1 Sample D3
660 680 700 720 740 760 780 800 820 840 860
(a)
Photon Energy (meV)
CL Intensity (arb. units)
Sample D1
660 680 700 720 740 760 780 800 820 840 860
(b)
Photon Energy (meV)
CL Intensity (arb. units)
Sample D3
660 680 700 720 740 760 780 800 820 840 860
(c)
Photon Energy (meV)
Fig. 5. (a) CL spectra representative of samples D1 and D3. (b) Gaussian deconvolution of the CL spectrum representative from sample D1. The emission bands appear centred at 761, 740 and 789 meV (dashed lines). The solid line is the best-fit curve while circles correspond to experimental data. (c) Deconvolution of the CL spectrum representative from sample D3. Emission bands are now centred at 752 and 731 meV.
spectra of a given sample, the mean band-gap value over spectra recorded in the same sample could be considered as representative for the
transversal sections of the ingot (samples D1, D2 and D3). The 760 meV band-gap value characteristic for sample D1, corresponding to an In content of x=0.037, decreases to a mean value of 755 meV in sample D3, that corresponds to an In content of x=0.042, while in sample D2 the mean band-gap value is almost the same than in sample D3. This gradient of the In composition of the ternary alloy is also clearly observed in the longitudinal sections (samples L1 and L2) when recording the spectral response along directions parallel to the growth axis. Although the peak positions and the relative intensity of the emission bands may differ from spot to spot, deconvolution of our CL spectra always show similar features. Fig. 5 shows typical spectra recorded from samples D1 and D3, and their corresponding Gaussian deconvolutions. Almost all spectra show two main transitions, one corresponding to the InxGa1 xSb band-gap emission and another one peaked at about 20 meV lower energy. In some spectra a third weak emission, peaked at above band-gap energies, is also observed. Overall, the CL spectra of the ternary alloy exhibit similar qualitative features to those described for GaSb. It is well established, e.g. [19,20], that as-grown undoped GaSb samples show at 90 K two main emissions peaked at about 796 meV (band-gap recombination) and 777 meV (commonly known as band A). Weak emissions peaked at about 756 meV (known as band B) and at above band-gap energies, ranging from 800 to 830 meV, have been also reported [20,21]. The 777 meV emission is due to the transition from the conduction band to the neutral state of the native acceptor level VGa–GaSb [19], while the above band-gap peak is an (e, h) transition including tail states and shallow acceptors [21]. The observed red shift of the band edge luminescence in InxGa1 xSb is due to the reduction of the band-gap energy with increasing x. A corresponding shift with In composition is observed for the lower- and higher-energy Gaussian peaks. A difference of about 20 meV between the InxGa1 xSb band-gap emission and the lowerenergy peak was detected for all the spectra. Contribution to the total emission of a mixture of InxGa1 xSb with slightly different x values, that may coexist in the excitation volume beneath the
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
targeted spot, could in principle generate a similar CL spectral distribution. However, the good chemical homogeneity revealed by WDX measurements and the almost linear variation of the lowerenergy peak position with the solid indium content suggest that this band is a defect band in InxGa1 xSb. Such band is probably related to the neutral state of a native acceptor level, equivalent to the acceptor responsible for the band A emission in GaSb, rather than to the contribution of InxGa1 xSb of higher x content. No evidence of a band similar to the band B of GaSb is detected in our ternary alloy. The latter band, which has been related in GaSb to the VGa–GaSb– VGa triple native acceptor complex, arising from an excess of Ga vacancies [22] or an excess of Ga as a gallium antisite [23], was probably eliminated by the incorporation of In. Some PL studies of InxGa1 xSb layers show only one transition which has been assigned to band-gap recombination [24], while Allegere et al. [12] conclude that a photoluminescence peak related to the acceptor level responsible for the p character of InxGa1 xSb becomes less intense with decreasing x. On the other hand, it was previously proposed [25] that In diffuses in GaSb by a vacancy mechanism on the Ga sublattice. Nevertheless, no comprehensive CL spectroscopic studies of defect bands in bulk InxGa1 xSb crystals have been previously carried out, mainly due to inefficient incorporation of In in the matrix. Furthermore, the only reported CL investigation of In-doped GaSb crystals [15] concluded that indium doping induces a certain reduction of the 777 meV band related to native defects, appropriate concentration of In leading to complete quenching of this luminescence band. Our results suggest that InxGa1 xSb ternary alloy with x values ranging from approximately 0.035–0.055 exhibit qualitatively similar spectral emission as undoped GaSb. The dominant transition corresponds to the band-gap recombination in the ternary alloy. The lower-energy emission appears to be related to a native acceptor of InxGa1 xSb similar to the native acceptor level VGa–GaSb of GaSb. A weak band centered above the band-gap transition energy observed in some spectra can be associated to the corresponding
57
higher-energy transition observed in GaSb [21] related to tail states and shallow acceptors.
4. Conclusions The Bridgman method has been used to grow InxGa1 xSb ingots with good chemical homogeneity. CL spectral measurements show a good incorporation of In in the matrix, leading to the formation of InxGa1 xSb ternary alloy along the whole ingot. Results provided by WDX and CL techniques also reveal the existence of a slight gradient of the In content along the growth axis. In addition, CL micrographs reveal the presence of grain boundaries and dislocations, while some cracks can be appreciated in SE and BSE images as well. These observations indicate that further work is still necessary in order to grow defect-free, single crystalline InxGa1 xSb and to avoid the problems related to internal stress and straininduced dislocations. The CL spectra of the ternary alloy exhibit similar general features to those reported for GaSb. The red shift of the near band edge luminescence in InxGa1 xSb, relative to that of GaSb, is due to the reduction of the band gap with increasing In content. Another band observed in the CL spectra, peaked at about 20 meV below the band-gap energy, is attributed to the presence in the ternary alloy of a native acceptor level that would correspond to the VGa– GaSb acceptor in GaSb. A weak emission peaked above the band-gap transition energy, observed in some spectra, is attributed to tail states and shallow acceptors.
Acknowledgements This work has been carried out in the frame of the Fifth Framework European Programme for research, HPRN-CT 2001-00199 project. M.F.C. acknowledges European Commission for financial supportU Support from MCYT through project MAT2000-2119 is also acknowledged.
ARTICLE IN PRESS 58
M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
References [1] P.S. Dutta, H.L. Bhat, V. Kumar, J. Appl. Phys. 81 (1997) 5821. [2] T.J. Couts, C.S. Allma, J.P. Benner, Third NREL Conference on Thermophotovoltaic Generation of Electricity, AIP Conference Proceedings, 1997, 401pp. [3] O.V. Sulima, A. Bett, P.S. Dutta, M.G. MauK, R.L. Mueller, Proceedings of the 29th IEEE Photovoltaic Specialists Conference, May 20–24, New Orleans, Louisiana, 2002, 892pp. [4] J.F. Yee, M.C. Lin, K. Sarma, M.R. Wilcox, J. Crystal Growth 30 (1970) 185. [5] Y. Hayakawa, Y. Okano, A. Hirata, N. Imaishi, Y. Kumagiri, X. Zhong, X. Xie, B. Yuan, F. Wu, H. Liu, T. Yamaguchi, M. Kumagawa, J. Crystal Growth 213 (2000) 40. [6] Y. Hayakawa, K. Balakrishnan, H. Komatsu, N. Murakami, T. Nakamura, T. Koyama, T. Ozawa, Y. Okano, M. Miyazawa, S. Dost, H. Le Dao, M. Kumagawa, J. Crystal Growth 237 (2002) 1831. [7] J. Kang, T. Fukuda, Mater. Sci. Eng. B 75 (2000) 149. [8] J.C. Wooley, B.A. Smith, D.G. Lees, Proc. Phys. Soc. London Sect. B 69 (1956) 1339. [9] J.P. Garandet, T. Duffar, J. Favier, JU Crystal Growth 106 (1990) 426. [10] C. Marin, P.S. Dutta, E. Di!eguez, P. Dussere, T. Duffar, JU Crystal Growth 173 (1997) 271.
[11] A.I. Lebedev, I.A. Strenikova, A.E. Yunovich, Sov. Phys. Semicond. 11 (1977) 1246. [12] J. Allegre, M. Averous, R. Jourdan, A. Loullie, J. Lumin. 11 (1976) 339. [13] V. Vankove, A.W.R. Leitch, J.R. Botha, JU Crystal Growth 248 (2003) 279. [14] M.F. Chioncel, C. D!ıaz-Guerra, J. Piqueras, N. Duhanian, T. Duffar, J. Optoelectron. Adv. M. 6 (2004) 237. [15] P. Hidalgo, B. M!endez, J. Piqueras, P.S. Dutta, E. Di!eguez, Semicond. Sci. Technol. 14 (1999) 901. [16] K.J. Bachmann, F.A. Thiel, H. Schrieber, Prog. Cryst. Growth Characterization 2 (1979) 171. [17] P.S. Dutta, A.G. Ostrogorsky, JU Crystal Growth 194 (1998) 1. [18] F.S. Juang, Y.K. Su, T.S. Wu, Solid State Electron. 34 (1991) 1225. [19] M.C. Wu, C.C. Chen, J. Appl. Phys. 72 (1992) 4275. [20] B. M!endez, P.S. Dutta, J. Piqueras, E. Di!eguez, Appl. Phys. Lett. 67 (1995) 2648. [21] G. Benz, R. Conradt, Phys. Rev. B 16 (1977) 843. [22] D. Weler, H. Mehrer, Philos. Mag. A 49 (1984) 309. [23] B. M!endez, J. Piqueras, P.S. Dutta, E. Di!eguez, Mater. Sci. Eng. B 42 (1996) 38. [24] J. Shin, Y. Hsu, T.C. Hsu, G.B. Stringfellow, R.W. Gedridge, J. Electron. Mater. 24 (1995) 1563. [25] D. Mathiot, G. Edelin, Philos. Mag. A 41 (1980) 447.
Journal of Crystal Growth 268 (2004) 52–58
Cathodoluminescence study of InxGa1 xSb crystals grown by the Bridgman method M.F. Chioncela,1, C. D!ıaz-Guerraa,*, J. Piquerasa, J. Vincentb, b ! V. Bermudez , E. Die! guezb b
a Departamento de F!ısica de Materiales, Facultad de Ciencias F!ısicas, Universidad Complutense de Madrid, E-28040 Madrid, Spain ! Departamento de F!ısica de Materiales, Facultad de Ciencias, Universidad Autonoma de Madrid, E-28049 Cantoblanco, Madrid, Spain
Received 24 February 2004; accepted 3 May 2004
Communicated by KW Benz
Abstract The nature and the spatial distribution of radiative defects in InxGa1 xSb grown by the vertical Bridgman method have been studied by cathodoluminescence (CL) in a scanning electron microscope. The CL results have been complemented by X-ray microanalysis and backscattered electron imaging to relate the local luminescence properties to the chemical composition. Measurements of the band-gap energy from the CL spectra, supported by X-ray compositional mappings, reveal an effective incorporation of In in the matrix, leading to the formation of the ternary alloy in the whole volume of the ingot. A low gradient of the In content along the growth axis has been found. The CL spectra of the ternary alloy exhibit similar general features to those reported for GaSb. An observed red shift of the near band edge luminescence in InxGa1 xSb, relative to that of GaSb, is due to the reduction in the band gap with increasing x. A band often observed in the CL spectra, peaked at about 20 meV below the band-gap energy, is attributed to the presence in the ternary alloy of an acceptor level that would correspond to the VGa–GaSb acceptor in GaSb. r 2004 Elsevier B.V. All rights reserved. PACS: 78.60.Hk; 68.37.Hk; 61.72.Ff Keywords: A1. Cathodoluminescence; A1. Defects; A2. Bridgman technique; B1. InxGa1 xSb
*Corresponding author. Tel.: +34-91-394-4548; fax: +3491-394-4547. E-mail address: cdiazgue@material.fis.ucm.es (C. D!ıaz-Guerra). 1 Permanent address. Physics Department, Faculty of Chemistry, University of Bucharest, Bd. Elisabeta 14-12, 70346 Bucharest, Romania.
1. Introduction In the past decade, the III–V ternary alloy system GaSb–InSb has become of great interest for optoelectronic devices [1] and high-efficiency thermophotovoltaic (TPV) cells operating in
0022-0248/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2004.05.015
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
2. Experimental method InxGa1 xSb ingots of about 12 mm diameter and a length of 40 mm with a 10% nominal In content were grown by the vertical Bridgman technique. The growth was performed in a conical quartz ampoule that was previously graphitised. No seeds were used. The ingots were prepared filling the ampoule with 1.75 g of In, 19.00 g of Sb and 9.62 g of Ga, all 6 N materials (provided by Goodfellow). The ampoule was evacuated to below 1 10 4 Torr, sealed, heated up to 950 C to melt the powder material, and kept at this temperature during 24 h. The melted material was homogenised by low shaking the furnace for 30 h (+/ 30 rotation around the horizontal position at 60 /h). After homogenisation the furnace was set vertical. The pulling rate was set at 1 mm/h and the growth performed in a temperature gradient of 40 C/cm. After solidification, the furnace was slowly cooled down to room temperature. Several samples were cut at various places along the length of an ingot as shown in Fig. 1. Three disks of 2 mm thickness were cut perpendicular to the growth axis and labelled D1, D2 and D3.
D3
L2
D2
Growth axis
conjunction with low-temperature radiators [2]. TPV cells with band gaps lower than GaSb are expected to be advantageous for low-temperature (o1000 C) non-wavelength selective TPV radiators because they provide more effective absorption of the black body infrared radiation [3]. This alloy covers an interesting low band-gap range, (0.80–0.25) eV, that can be controlled by adjusting the In content to match the requirements. Strong efforts have been undertaken to grow compositionally uniform, crack-free materials. Although InxGa1 xSb crystals have been grown on earth, it is extremely difficult to grow high-quality crystals under terrestrial conditions because, as the densities of the components are different, solute transport occurs due to buoyancy. Many studies have employed microgravity conditions [4–6] or terrestrial condition under high magnetic fields in order to avoid gravity-induced negative effects [7]. However, the growth of bulk good-quality crystals remains an experimentally difficult task due to problems related to chemical segregation and mechanical stresses associated with the significant difference (12.5%) between the tetrahedral radii of Ga and In [8–10]. The properties of the device employing InxGa1 xSb in the structure will ultimately depend on the material quality, but little information is available regarding the influence of defects on the optical and electrical properties of InxGa1 xSb crystals. In fact, very few PL studies of InxGa1 xSb epilayers have been reported in the literature [11–13], while cathodoluminescence (CL) in the scanning electron microscope (SEM)—an important technique to analyse the nature and the spatial distribution of defects in semiconductors—has been to our knowledge only very recently applied to characterize InxGa1 xSb crystals [14]. In this work, the nature of luminescent centres and their spatial distribution along the growth and radial axis of an InxGa1 xSb ingot have been investigated using CL microscopy and spectroscopy in the SEM. The information provided by CL was complemented by wavelength dispersive X-ray microanalysis (WDX) and backscattered electron (BSE) imaging.
53
L1
D1 Fig. 1. Scheme of the InxGa1 xSb ingot showing the positions and labels of the samples investigated.
ARTICLE IN PRESS 54
M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
Sample D1 corresponds to the bottom of the ingot, D2 to the middle part and D3 to the top of the crystal, respectively. Samples L1 and L2, of rectangular shape, correspond to longitudinal sections (see Fig. 1). This selection of the samples enables an efficient investigation of the distribution of the luminescent centres along both the growth and radial axis. All the samples investigated were polished with alumina powder to a mirror finish and characterised by secondary electron (SE), BSE, WDX and CL modes of the SEM. Spatial and spectral CL measurements were carried out in a Hitachi S 2500 SEM with a cooled ADC germanium detector. Measurements were performed at 90 K using a 20 kV accelerating voltageU CL spectra were recorded on each sample and deconvoluted using a sum of Gaussian line distributions in order to determine the different CL peaks contributing to the overall emission. The local In content has been determined from the position of the band-gap transition in the ternary alloy. In order to study the composition of the features observed in the CL micrographs and to understand the spectral distribution of the CL emission, mapping of the elements Ga, Sb, In as well as quantitative X-ray microanalysis were performed by WDX in a Jeol JXA-8900 M Superprobe, using an accelerating voltage of 20 kV. The same system was used to obtain the BSE images.
Fig. 2 shows the distribution of Ga and In and the corresponding CL and BSE micrographs of a selected area from sample D3. The lack of contrast in the Sb mapping (not shown) indicates a very uniform distribution of this element. Overall, our results indicate that a higher In content corresponds to a lower Ga concentration and that the Sb content is almost constant. CL panchromatic mappings of the investigated samples show similarities in the distribution of the radiative centres. All the samples display luminescent emission, however, with some areas of diminished or quenched CL emission. Depending on the sample considered, precipitate-like features of elongated shape, showing dark CL contrast, may appear in the bright background (Fig. 3). Similar features have been observed previously in In-doped GaSb samples [15]. A preferential agglomeration of areas of low CL emission can be observed, in all the samples investigated, in regions corresponding to the inner part of the ingot. The decrease of the CL emission towards the inner part of the ingot is more clearly observed in the longitudinal sections, as sample L1. This sample shows the most inhomogeneous distribution of non-radiative
3. Results and discussion The exposed surface of all the samples was mapped by low-magnification SE and panchromatic CL microscopy while several representative areas were also analysed by means of BSE and WDXU All the samples present micro- and macrocracks, but apart of these, SE images display smooth, featureless surfaces. The X-ray compositional mappings show in all the samples a very uniform distribution of Sb and some anti-correlated variation in the Ga and In contents. There is a correspondence between contrast observed in CL, BSE, and X-ray mappings that is not observed in SE images, confirming that the mentioned contrast is not related to surface topography.
Fig. 2. WDX mappings from sample D3 showing the In (a) and Ga (b) distribution. Darker greys correspond to a lower element content. (c) BSE image of the same area. (d) Corresponding CL micrograph.
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
55
Fig. 3. CL image from sample L2 showing dislocation-related dark features in a background of uniform emission.
recombination centres, with extended bright and dark CL domains, and also displays a rich variety of CL structures (Fig. 4). Similar CL structures can be observed to a less extent in all the samples. Inside the grains, high-magnification micrographs of areas of enhanced CL emission show dark, precipitate-like defects and the absence of subgrain boundaries. Some of the precipitate-like features appear randomly oriented, but most of them appear aligned in equally spaced parallel lines (Fig. 4a). The size and the spatial distribution of these CL features, despite the precipitate-like appearance, suggest that they might be related to dislocations. Moreover, X-ray microanalysis of the regions where these dark features appear show no variation of In content. The attributed nature of these features to dislocations is also in agreement with previous results, which indicate that the solubility of In in GaSb is rather high and the formation of In precipitates unlikely [15], while strain-induced dislocations is one of the main problems found in the growth of good-quality InxGa1 xSb crystals [16,17]. The CL images of the areas of diminished CL emission show well-defined dark sub-grain boundaries (see Fig. 4b). These grains have various polygonal shapes, some of them showing hexagonal contours (Fig. 4c). As stated above, our CL microscopy results reveal an
Fig. 4. CL images from samples D2 (a,b) and L1 (c), showing precipitate-like features aligned along parallel lines (a) and different polygonal grain structures and sub-boundaries (b,c).
inhomogeneous distribution of non-radiative recombination centres in the volume of the ingot, with areas of diminished or quenched CL emission. Some of these areas may correspond to the presence of a high In content, above x=0.15, leading to an alloy with band gap below the detection limit of about 650 meV of the Ge detector used. The presence of grains, sub-grains and dislocations which probably induced subsequent grain generation is revealed by higher magnification CL microscopy images. Representative CL spectra of samples D1 and D3 are shown in Fig. 5. A shift towards lower energies from the GaSb band-gap value at 90 K, 796 meV, is clearly observed in all the spectra. This observation reveals the formation of the ternary alloy along the whole ingot. CL bands giving rise to these spectra have been determined from the best fits to the experimental data using a sum of Gaussian line distributions. The In content, x, has been calculated from the peak energy of the asfitted band-gap emission [18]. Although some change of the peak position is observed in the
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
56
CL Intensity (arb. units)
Sample D1 Sample D3
660 680 700 720 740 760 780 800 820 840 860
(a)
Photon Energy (meV)
CL Intensity (arb. units)
Sample D1
660 680 700 720 740 760 780 800 820 840 860
(b)
Photon Energy (meV)
CL Intensity (arb. units)
Sample D3
660 680 700 720 740 760 780 800 820 840 860
(c)
Photon Energy (meV)
Fig. 5. (a) CL spectra representative of samples D1 and D3. (b) Gaussian deconvolution of the CL spectrum representative from sample D1. The emission bands appear centred at 761, 740 and 789 meV (dashed lines). The solid line is the best-fit curve while circles correspond to experimental data. (c) Deconvolution of the CL spectrum representative from sample D3. Emission bands are now centred at 752 and 731 meV.
spectra of a given sample, the mean band-gap value over spectra recorded in the same sample could be considered as representative for the
transversal sections of the ingot (samples D1, D2 and D3). The 760 meV band-gap value characteristic for sample D1, corresponding to an In content of x=0.037, decreases to a mean value of 755 meV in sample D3, that corresponds to an In content of x=0.042, while in sample D2 the mean band-gap value is almost the same than in sample D3. This gradient of the In composition of the ternary alloy is also clearly observed in the longitudinal sections (samples L1 and L2) when recording the spectral response along directions parallel to the growth axis. Although the peak positions and the relative intensity of the emission bands may differ from spot to spot, deconvolution of our CL spectra always show similar features. Fig. 5 shows typical spectra recorded from samples D1 and D3, and their corresponding Gaussian deconvolutions. Almost all spectra show two main transitions, one corresponding to the InxGa1 xSb band-gap emission and another one peaked at about 20 meV lower energy. In some spectra a third weak emission, peaked at above band-gap energies, is also observed. Overall, the CL spectra of the ternary alloy exhibit similar qualitative features to those described for GaSb. It is well established, e.g. [19,20], that as-grown undoped GaSb samples show at 90 K two main emissions peaked at about 796 meV (band-gap recombination) and 777 meV (commonly known as band A). Weak emissions peaked at about 756 meV (known as band B) and at above band-gap energies, ranging from 800 to 830 meV, have been also reported [20,21]. The 777 meV emission is due to the transition from the conduction band to the neutral state of the native acceptor level VGa–GaSb [19], while the above band-gap peak is an (e, h) transition including tail states and shallow acceptors [21]. The observed red shift of the band edge luminescence in InxGa1 xSb is due to the reduction of the band-gap energy with increasing x. A corresponding shift with In composition is observed for the lower- and higher-energy Gaussian peaks. A difference of about 20 meV between the InxGa1 xSb band-gap emission and the lowerenergy peak was detected for all the spectra. Contribution to the total emission of a mixture of InxGa1 xSb with slightly different x values, that may coexist in the excitation volume beneath the
ARTICLE IN PRESS M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
targeted spot, could in principle generate a similar CL spectral distribution. However, the good chemical homogeneity revealed by WDX measurements and the almost linear variation of the lowerenergy peak position with the solid indium content suggest that this band is a defect band in InxGa1 xSb. Such band is probably related to the neutral state of a native acceptor level, equivalent to the acceptor responsible for the band A emission in GaSb, rather than to the contribution of InxGa1 xSb of higher x content. No evidence of a band similar to the band B of GaSb is detected in our ternary alloy. The latter band, which has been related in GaSb to the VGa–GaSb– VGa triple native acceptor complex, arising from an excess of Ga vacancies [22] or an excess of Ga as a gallium antisite [23], was probably eliminated by the incorporation of In. Some PL studies of InxGa1 xSb layers show only one transition which has been assigned to band-gap recombination [24], while Allegere et al. [12] conclude that a photoluminescence peak related to the acceptor level responsible for the p character of InxGa1 xSb becomes less intense with decreasing x. On the other hand, it was previously proposed [25] that In diffuses in GaSb by a vacancy mechanism on the Ga sublattice. Nevertheless, no comprehensive CL spectroscopic studies of defect bands in bulk InxGa1 xSb crystals have been previously carried out, mainly due to inefficient incorporation of In in the matrix. Furthermore, the only reported CL investigation of In-doped GaSb crystals [15] concluded that indium doping induces a certain reduction of the 777 meV band related to native defects, appropriate concentration of In leading to complete quenching of this luminescence band. Our results suggest that InxGa1 xSb ternary alloy with x values ranging from approximately 0.035–0.055 exhibit qualitatively similar spectral emission as undoped GaSb. The dominant transition corresponds to the band-gap recombination in the ternary alloy. The lower-energy emission appears to be related to a native acceptor of InxGa1 xSb similar to the native acceptor level VGa–GaSb of GaSb. A weak band centered above the band-gap transition energy observed in some spectra can be associated to the corresponding
57
higher-energy transition observed in GaSb [21] related to tail states and shallow acceptors.
4. Conclusions The Bridgman method has been used to grow InxGa1 xSb ingots with good chemical homogeneity. CL spectral measurements show a good incorporation of In in the matrix, leading to the formation of InxGa1 xSb ternary alloy along the whole ingot. Results provided by WDX and CL techniques also reveal the existence of a slight gradient of the In content along the growth axis. In addition, CL micrographs reveal the presence of grain boundaries and dislocations, while some cracks can be appreciated in SE and BSE images as well. These observations indicate that further work is still necessary in order to grow defect-free, single crystalline InxGa1 xSb and to avoid the problems related to internal stress and straininduced dislocations. The CL spectra of the ternary alloy exhibit similar general features to those reported for GaSb. The red shift of the near band edge luminescence in InxGa1 xSb, relative to that of GaSb, is due to the reduction of the band gap with increasing In content. Another band observed in the CL spectra, peaked at about 20 meV below the band-gap energy, is attributed to the presence in the ternary alloy of a native acceptor level that would correspond to the VGa– GaSb acceptor in GaSb. A weak emission peaked above the band-gap transition energy, observed in some spectra, is attributed to tail states and shallow acceptors.
Acknowledgements This work has been carried out in the frame of the Fifth Framework European Programme for research, HPRN-CT 2001-00199 project. M.F.C. acknowledges European Commission for financial supportU Support from MCYT through project MAT2000-2119 is also acknowledged.
ARTICLE IN PRESS 58
M.F. Chioncel et al. / Journal of Crystal Growth 268 (2004) 52–58
References [1] P.S. Dutta, H.L. Bhat, V. Kumar, J. Appl. Phys. 81 (1997) 5821. [2] T.J. Couts, C.S. Allma, J.P. Benner, Third NREL Conference on Thermophotovoltaic Generation of Electricity, AIP Conference Proceedings, 1997, 401pp. [3] O.V. Sulima, A. Bett, P.S. Dutta, M.G. MauK, R.L. Mueller, Proceedings of the 29th IEEE Photovoltaic Specialists Conference, May 20–24, New Orleans, Louisiana, 2002, 892pp. [4] J.F. Yee, M.C. Lin, K. Sarma, M.R. Wilcox, J. Crystal Growth 30 (1970) 185. [5] Y. Hayakawa, Y. Okano, A. Hirata, N. Imaishi, Y. Kumagiri, X. Zhong, X. Xie, B. Yuan, F. Wu, H. Liu, T. Yamaguchi, M. Kumagawa, J. Crystal Growth 213 (2000) 40. [6] Y. Hayakawa, K. Balakrishnan, H. Komatsu, N. Murakami, T. Nakamura, T. Koyama, T. Ozawa, Y. Okano, M. Miyazawa, S. Dost, H. Le Dao, M. Kumagawa, J. Crystal Growth 237 (2002) 1831. [7] J. Kang, T. Fukuda, Mater. Sci. Eng. B 75 (2000) 149. [8] J.C. Wooley, B.A. Smith, D.G. Lees, Proc. Phys. Soc. London Sect. B 69 (1956) 1339. [9] J.P. Garandet, T. Duffar, J. Favier, JU Crystal Growth 106 (1990) 426. [10] C. Marin, P.S. Dutta, E. Di!eguez, P. Dussere, T. Duffar, JU Crystal Growth 173 (1997) 271.
[11] A.I. Lebedev, I.A. Strenikova, A.E. Yunovich, Sov. Phys. Semicond. 11 (1977) 1246. [12] J. Allegre, M. Averous, R. Jourdan, A. Loullie, J. Lumin. 11 (1976) 339. [13] V. Vankove, A.W.R. Leitch, J.R. Botha, JU Crystal Growth 248 (2003) 279. [14] M.F. Chioncel, C. D!ıaz-Guerra, J. Piqueras, N. Duhanian, T. Duffar, J. Optoelectron. Adv. M. 6 (2004) 237. [15] P. Hidalgo, B. M!endez, J. Piqueras, P.S. Dutta, E. Di!eguez, Semicond. Sci. Technol. 14 (1999) 901. [16] K.J. Bachmann, F.A. Thiel, H. Schrieber, Prog. Cryst. Growth Characterization 2 (1979) 171. [17] P.S. Dutta, A.G. Ostrogorsky, JU Crystal Growth 194 (1998) 1. [18] F.S. Juang, Y.K. Su, T.S. Wu, Solid State Electron. 34 (1991) 1225. [19] M.C. Wu, C.C. Chen, J. Appl. Phys. 72 (1992) 4275. [20] B. M!endez, P.S. Dutta, J. Piqueras, E. Di!eguez, Appl. Phys. Lett. 67 (1995) 2648. [21] G. Benz, R. Conradt, Phys. Rev. B 16 (1977) 843. [22] D. Weler, H. Mehrer, Philos. Mag. A 49 (1984) 309. [23] B. M!endez, J. Piqueras, P.S. Dutta, E. Di!eguez, Mater. Sci. Eng. B 42 (1996) 38. [24] J. Shin, Y. Hsu, T.C. Hsu, G.B. Stringfellow, R.W. Gedridge, J. Electron. Mater. 24 (1995) 1563. [25] D. Mathiot, G. Edelin, Philos. Mag. A 41 (1980) 447.