- Email: [email protected]
Photoluminescence properties of a near-stoichiometric CuGaSe2 film
Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906 www.elsevier.com/locate/jpcs
Photoluminescence properties of a near-stoichiometric CuGaSe2 film G.L. Gua, B.H. Tsenga,*, H.L. Hwangb a
Institute of Materials Science and Engineering, National Sun Yat-Sen University, Kaohsiung 804, Taiwan, ROC b Department of Electrical Engineering, National Tsing Hua University, Hsinchu 300, Taiwan, ROC
Abstract Epitaxial films of CuGaSe2 were grown on (001)GaAs substrates by an MBE technique. A near-stoichiometric film with chemical compositions consistently varied from Cu- to Ga-rich was prepared by growing the film without the substrate rotation. A series of PL spectra was obtained by directing a focused laser beam point-by-point across the boundary separating the Cu- and Ga-rich regions. Distinct features of these spectra were noted. On the Cu-rich side, optical emissions peaked at 1.71, 1.67, 1.63, and 1.59 eV were observed in a PL spectrum. The peak at 1.71 eV was due to the emission of bound exciton, while the peak at 1.67 eV was caused by the free-to-bound transition. The other two peaks were identified to be the donor-to-acceptor emissions. Further annealing experiments performed in different environments suggested that the peaks at 1.67, 1.63, and 1.59 eV were associated with the optical transitions of CB ! CuGa, Cui ! CuGa, and Cui ! VGa, respectively. On the Ga-rich side, a dominant donor-to-acceptor emission peaked at 1.62 eV was observed, which was determined to be the GaCu ! VCu transition. The two defects with opposite charge states resulted in a highly compensated material with high resistivity. q 2003 Elsevier Ltd. All rights reserved.
1. Introduction CuGaSe2, a member of I – III – VI2 compound semiconductor [1,2], has a direct energy gap of 1.68 eV at room temperature, a very high absorption coefficient up to 105 cm21 at 500 nm and an easily controllable electrical resistivity in a wide range (1021 – 105 Ohm cm). These properties make CuGaSe2 an ideal high bandgap partner with CuInSe2 in a solar-cell structure. Several techniques including liquid phase epitaxy (LPE) [3], metal – organic vapor phase epitaxy (MOVPE) [4], molecular beam epitaxy (MBE) [5], and halogen transport method [6], had been used to grow CuGaSe2 epilayers on ZnSe, GaAs, and GaP substrates. Among the film growth techniques, MBE has favorable features such as a relatively low growth temperature, a low growth rate, and simple growth mechanism. As a result of these advantages, the MBE technique provides an abrupt interface, a smooth surface, and a precise control on composition and thickness. In this work, an MBE technique is developed to * Corresponding author. E-mail address: [email protected] (B.H. Tseng).
grow high-quality CuGaSe2 epitaxial films for structural and optical characterizations. We choose the (001)GaAs wafer as the substrate to grow a CuGaSe2 film because of its availability and relatively low cost. The use of (001) substrate orientation may eliminate the formation of orientation domain structure in the epitaxial film [7]. Like other I –III – VI2 compounds, CuGaSe2 has a wide phase stability region. The Cu2Se – Ga2Se3 pseudobinary phase diagram shows the chalcopyrite phase boundaries extended from 50 to 56 mole % Ga2Se3 [8]. A considerable amount of intrinsic point defects exist in the material as its composition deviated from the stoichiometry. Optical and electrical properties are thus greatly affected by the film composition. The conductivity type of the film can even be changed without the introduction of doping elements. It is essential to identify the energy levels of intrinsic point defects and their abundance in the material. Photoluminescence (PL) spectroscopy is one of the most powerful techniques for such a study. Several research groups have shown the results of PL measurement for bulk CuGaSe2 single crystals. Three donor levels of 10, 80, and 100 meV associated with GaCu, VSe, and Gai, respectively, and three acceptor levels of 15, 55,
0022-3697/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0022-3697(03)00151-3
1902
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
and 95 meV associated with VGa, VCu and Sei, respectively, were assigned by Schon et al. [9]. In addition, PL characterizations of epitaxial CuGaSe2 films were examined by several Japanese groups. A donor level of 108 meV (VSe) and an acceptor level of 35 meV (CuGa) were appointed by Yamada et al. [10]. Due to the complex defect chemistry in CuGaSe2 and the discrepancy of composition measurements, the results of PL measurement were still not consistent. In this work, a near-stoichiometric CuGaSe2 epitaxial film with a composition gradient varying from Cuto Ga-rich was used for the analysis. We also developed a technique for unambiguously determining the film composition using an electron microprobe. The relationship between chemical composition and optical emission spectra is thus established. Further experiments of thermal annealing in various ambient helped the identification of the types of intrinsic point defects which involved in optical emissions.
will meet this requirement. Due to the difficulty in preparing a standard of the ternary chalcopyrite compound with uniform composition, we use pure Cu, GaAs, and ZnSe as elemental standards. The use of elemental standards has an advantage to standardize the measurement procedures and leads to a consistency on the interpretation of experimental results. Composition measurements using the above-mentioned method had been proven successful for determining the chemical composition of CuInSe2 [11]. Our previous composition measurements on thin films of CuInSe2 indicated that there was a composition variation in a microscopic scale. A successful correlation between the PL and EPMA results should have an average of composition data from the same area of optical excitation. Thus, every composition data was average from 5 to 9 detection spots using a beam size of 5 mm.
3. Results and discussion 2. Experimental procedures CuGaSe2 films were grown on (001)GaAs substrates by an MBE technique. A Seiko Seiki SP-400 turbomolecular pump was used to evacuate the MBE system and the background pressure was 4 £ 1029 Torr. High-purity elemental sources of Cu(6N), Ga(7N), and Se(6N) were loaded in the effusion cells made by high-density and highpurity graphite. The substrates were cleaned in acetone and methanol subsequently to remove the organic contaminants and then etched in the NH4OH:H2O2:H2O ¼ 3:1:15 solution for 10 s before loading to the growth chamber. Thin films with different compositions were prepared by varying the temperature of the Ga source, while the temperatures of the Cu and Se sources were kept constant. The films were grown at a substrate temperature of 500 8C for 1 h. PL spectroscopy was employed to study the optical transitions in a CuGaSe2 film. A 514.5 nm emission line of a Coherent Innova-90 Ar ion laser was selected as an excitation source. The laser beam could be focused to a spot with a diameter about 0.2 mm. The sample was fixed in a liquid-He cryostat and could be cooled down to 9 K. The PL signals were focused into an Acton Research SpectraPro 500 monochromator and detected by a Si detector. These signals were processed by a lock-in amplifier with a chopper frequency of 330 Hz and finally collected by a computer. The laser power for PL excitation could be varied by a neutral density filter for the identification of optical transition types. Chemical compositions of CuGaSe2 films were measured by a JEOL JXA-8900R electron microprobe. A simple method by using an electron beam with its energy below 10 keV may prevent the excitation of the substrate material and greatly simplify the analyzing procedures. Normally, a CuGaSe2 film with the thickness about 500 nm
Thin film of CuGaSe2 was grown on an one-inch square GaAs wafer by an MBE technique without a substrate rotation. A proper control of source fluxes would produce a film with both Cu- and Ga-rich regions in one sample. This was the condition chosen for an overall film composition very close to stoichiometry. The composition data obtained from this film is shown in Fig. 1a. As can be seen, the Cu and Ga contents are consistently varied from one side to the other side while the Se content is fairly uniform throughout the film. Careful examinations on the data of the Se content reveals that the Ga-rich region has a value slightly higher than 50 at.% and the Cu-rich region has a value slightly lower than 50 at.%. This result suggests that the film may grow in a two-step process, i.e. binary compounds of Cu2Se and Ga2Se3 form first on the surface and subsequently react to form a ternary chalcopyrite compound [11]. The Cu/Ga ratios of this sample vary from 0.94 to 1.10 as shown in Fig. 1b. The Cu/ Ga ratio at the boundary separating the Cu- and Ga-rich region is about 1. A series of PL spectra were obtained at a temperature of 9 K using a laser beam focused on the positions where the composition measurements were made. Fig. 2 shows the PL spectra and the corresponding composition data. We see that the PL spectra are significantly different in the Cu- and Garich regions. The emission peaks at 1.71, 1.67, 1.63, and 1.59 eV are observed in the Cu-rich region, while a single peak at 1.62 eV is found in the Ga-rich region. A red shift of the peak originally located at 1.62 eV is also noted when the composition is deviated from the stoichiometry. The offstoichiometry composition leads to variation in the lattice constant, which may induce additional mechanical stress and cause a red shift of PL signals. Excitation power dependence on the energy shift and the intensity of individual PL peak had been investigated to
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
1903
identify the transition types of optical emissions [12]. A laser power range of 1 – 400 mW was used for this experiment and the data obtained from two selected spots, one on the Cu-rich region and the other on the In-rich region, were plotted in Fig. 3. Optical emissions originally peaked at 1.62, 1.63, and 1.59 eV show an energy shift about 30 meV as the excitation power increases, whose intensity also show a sublinear excitation power dependence. The characteristics of emission peaks like this is typical for a donor-to-acceptor (D – A) pair transition. On the other hand, the intensity of optical emission peaked at 1.67 eV increases linearly with respect to the excitation power and shows almost no shift in peak energy. These behaviors indicate that it is caused by a free-to-bound (F – B) transition. Finally, the emission peaked at 1.71 eV has a superlinear intensity vs. power dependence and is attributed to the exciton emission. The above analysis indicates that the excitonic and the F– B transitions are dominant in the Cu-rich region, while the D– A pair transitions prevail in the Ga-rich region. Vacancies, interstitials, and antisites are typical intrinsic point defects in CuGaSe2. There are twelve in
Fig. 1. (a)Composition data measured from a near-stoichiometric sample; (b) a plot of the data in term of the Cu/Ga ratio.
Fig. 2. A series of PL spectra obtained from the near-stoichiometric sample.
Fig. 3. The excitation power dependence of (a) energy shifts and (b) intensity of the PL peaks.
1904
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
total. Chemical composition of the film as well as the formation energy of point defect determines what kinds of defects are dominant in the film. Theoretical calculations on the defect formation energy have been done by Wei et al. [13,14] and Neumann [15]. The results are summarized in Table 1. As can be seen, Cu interstitial (Cui), Cu on Ga antisite (CuGa), Ga on Cu antisite (GaCu), Cu vacancy (VCu), Ga vacancy (VGa), Se vacancy (VSe), and defect pair of (GaCu þ VCu) have relatively low formation energies. For a Cu-rich film, the dominant defect types will be Ga vacancy (acceptor), Cu interstitial (donor), CuGa antisite (acceptor), and Se vacancy (VSe). A similar argument can be applied to the Ga-rich film. Thermal annealing in various ambient may result in a change of defect distributions. This will help to identify the defect levels involved in optical transitions. Vacuum annealing will cause thermal decomposition from the film surface in the form of Cu2Se and Ga2Se3. It has been known that more Ga2Se3 escapes than Cu2Se. This leads to a surface composition deficient in the Se content and also shifts the Cu/Ga ratio to a higher value. To prohibit thermal decomposition, the sample should be annealed in a close tube filled with Se vapor to suppress surface decomposition. An alternate method is to anneal the sample in the MBE chamber under the exposure to a Se-beam flux. This is a convenient operation as compared with the close-tube annealing. Although the annealing is done in an open system, a quasi-equilibrium state can be maintained. In this work, the film was annealed at 500 8C for 30 min in a Sebeam flux of 1.5 £ 1014 atoms/cm2 s. Fig. 4 shows the composition data measured throughout a near-stoichiometric film before and after annealing. We see an overall increase in the Se content up to 50.2 at.% and a very slight change in the Cu/Ga ratios of this film. Our previous TEM observations of the as-grown CuInSe2 and CuGaSe2 films showed that antiphase domain boundaries (APB’s) prevailed in the film. Thermal annealing in the presence of a Se-beam flux
Table 1 A partial of formation energy of intrinsic point defects in CuGaSe2 Defect types
Formation energy (eV)
Reference
Cu vacancy (VCu) Ga vancancy (VGa) Se vacancy (VSe) Cu vacancy (VCu) Ga vancancy (VGa) Cu interstitial (Cui) Cu on Ga antisite (CuGa) Ga on Cu antisite (GaCu) Defect pair (GaCu þ 2VCu)
3.43 2.72 3.07 0.66 2.83 1.91 1.41 2.04 0.20
[15] [15] [15] [13] [13] [13] [13] [13] [13]
Other defect types not listed here have much lager formation energy.
Fig. 4. Chemical composition data of a near-stoichiometric film annealed at 500 8C for 30 min in a MBE chamber under the exposure to a Se-beam flux.
did eliminate the APB’s and considerably reduced the antisite defects in the film [7]. Fig. 5 shows the PL spectra of the as-grown sample with chemical composition in the Cu-rich region and the same sample after annealing in a Se vapor flux. A dramatic decrease in the intensity of the optical emission peaked at 1.67 eV and a relatively slight decrease in the PL intensity peaked at 1.63 eV are observed after annealing. The annealing process was further conducted subsequently by depositing a 20 nm-thick Cu layer on the film surface at room temperature and then by heating the sample in vacuum at 500 8C for 30 min. The intensities of both peaks are increased after annealing. As mentioned above, two acceptors (CuGa, VGa) and two donors (Cui, VSe) are prevailing in the Cu-rich region. The defect concentration of VSe, CuGa, and Cui may decrease after annealing in a Se vapor flux, while an increase of CuGa and Cui may occur after annealing with the film surface capped by a thin Cu layer. It implies that these two peaks are associated with CuGa and Cui. The PL peak at 1.67 eV has been identified to be caused by the F– B transition. It may be assigned to be the transition from conduction band to an acceptor level of CuGa according to our previous work on CuInSe2 [7]. Another peak at 1.63 eV is an emission caused by the D– A transition and may be attributed to the transition from a donor level of Cui to an acceptor level of CuGa. Finally, a D– A emission peaked at 1.59 eV did not have a significant change in
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
1905
Fig. 5. A series of PL spectra obtained from the Cu-rich region: (a) as-grown, (b) Se-annealing, and (c) Cu-annealing.
Fig. 6. A series of PL spectra obtained from the Ga-rich region: (a) as-grown and (b) Cu-annealing.
intensity after annealing. It may be caused by the transition from a donor level of Cui to an acceptor level of VGa, because the concentration of VGa may decrease after annealing with a Cu cap layer but do not increase after annealing in a Se vapor flux (the Cu/Ga ratio of the film decreases from 1.05 to 1.02 after annealing). The PL spectrum obtained from the Ga-rich region of a near-stoichiometric film show a main peak at 1.62 eV (see Fig. 6), which was characterized to be caused by the D– A transition. Intrinsic point defects of VCu and GaCu are the majority in the Ga-rich region according to the composition data. The peak at 1.62 eV is attributed to the GaCu –VCu transition. After annealing the sample with a thin Cu layer capped on the surface, the PL spectrum shows a decrease in the peak intensity at 1.62 eV and an increase in the intensity of the emission peaked at 1.66 eV. This annealing process may cause a decrease in the number of VCu and GaCu. Because of the relatively low formation energy of VCu, the number of VCu may diminish faster than that of GaCu. This implies that the peak at 1.66 eV is associated with the emission from a donor level of GaCu to the valence band. There is also one broad peak below 1.50 eV detected after annealing, which is probably caused by some deep levels. The origins of this emission need further investigation. Four-point probe was used to measure the resistivity of the as-grown sample. The result is shown in Fig. 7. As can
be seen, the resistivity decreases drastically by several orders of magnitude as composition varies from the Ga-rich region to the Cu-rich region. The low resistivity measured in the Cu-rich region is probably due to the abundance of
Fig. 7. The resistivity data measured point-by-point from a nearstoichiometric film.
1906
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
the CuGa antisites (acceptors) in that area. On the other hand, high compensation of VCu (acceptor) and GaCu (donor) in the Ga-rich region leads to high resistivity of the film. The resistivity increased as the film composition became more Ga-rich. When the Cu/Ga ratio was less than 0.9, the resistivity was too high and out of the measurement range of a four-point probe. It is found that the resistivity data is consistent with the PL results.
4. Conclusions A specific sample of CuGaSe2 with chemical composition very close to the stoichiometry was prepared. A boundary separating the Cu-rich and Ga-rich region could be clearly observed and the composition measured by EPMA using elemental standards indicated the Cu/Ga ratio about 1 along the boundary. The EPMA measurements also showed a consistent composition variation across the boundary. Further characterizations indicated that the resistivity and PL results fitted quite well with the composition data. The technique for composition measurements developed in this work can be easily applied to the thin-film samples and also avoids the use of compound standards which may cause the discrepancy in composition measurements. In this work, we did the annealing experiments on this sample and tried to identify the intrinsic point defects which caused the optical transitions. The assignments were based on chemical composition, calculated formation energy of the defects, and the characteristics of PL emission peaks. The PL peaks at 1.71, 1.67, 1.63, and 1.59 eV were observed on the Cu-rich side of the sample. Further experiments suggested that the peaks at 1.67, 1.63 and 1.59 eV were associated with the optical transitions of CB ! CuGa, Cui ! CuGa, and Cui ! VGa, respectively. On
the Ga-rich side, a dominant donor-to-acceptor emission peaked at 1.62 eV was observed, which was determined to be the GaCu ! VCu transition.
References [1] L. Shay, S. Wagner, H.M. Kasper, Appl. Phys. Lett. 27 (1975) 89. [2] L.I. Kazmerski, F.R. White, G.K. Morgan, Appl. Phys. Lett. 29 (1975) 268. [3] H. Asai, K. Sugiyama, Jpn. J. Appl. Phys. 20 (1981) 1401. [4] S. Chichibu, Y. Harada, M. Uchida, T. Wakiyama, S. Matsumoto, S. Shirakata, S. Isomola, H. Higuchi, J. Appl. Phys. 76 (1994) 3009. [5] A. Yamada, Y. Makita, S. Niki, A. Obara, P. Fons, H. Shibata, M. Kawai, S. Chichibu, H. Nakanishi, J. Appl. Phys. 79 (1996) 4318. [6] O. Igarashi, J. Crystal, Growth 143 (1994) 213. [7] B.H. Tseng, S.B. Lin, G.L. Gu, W. Chen, J. Appl. Phys. 79 (1996) 1391. [8] J.C. Mikkelsen, J. Electron. Mater. 10 (1981) 541. [9] J.H. Schon, K. Fess, K. Friemelt, C. Kloc, E. Bucher, Inst. Phys. Conf. Ser. 152 (1998) 59. [10] A. Yamada, P. Fons, S. Niki, H. Shibata, A. Obara, Y. Makita, H. Oyanagi, J. Appl. Phys. 81 (1997) 2794. [11] G.Y. Chiang, The effects of composition variation on luminescence properties of CuInSe2 epitaxial films, MS Thesis, National Sun Yat-Sen University, 1998. [12] S. Niki, Y. Makita, A. Yamada, A. Obara, O. Igarashi, S. Misawa, M. Kawai, H. Nakanishi, Y. Taguchi, N. Kutsuwada, Solar Energy Mater Solar Cells 35 (1994) 141. [13] S.H. Wei, S.B. Zhang, A. Zunger, Appl. Phys. Lett. 72 (1998) 3199. [14] S.B. Zhang, S.H. Wei, A. Zunger, H. Kadayama-Yoshida, Phys. Rev. B 57 (1998) 9642. [15] H. Neumann, Crystal Res. Technol. 18 (1983) 901.
Photoluminescence properties of a near-stoichiometric CuGaSe2 film G.L. Gua, B.H. Tsenga,*, H.L. Hwangb a
Institute of Materials Science and Engineering, National Sun Yat-Sen University, Kaohsiung 804, Taiwan, ROC b Department of Electrical Engineering, National Tsing Hua University, Hsinchu 300, Taiwan, ROC
Abstract Epitaxial films of CuGaSe2 were grown on (001)GaAs substrates by an MBE technique. A near-stoichiometric film with chemical compositions consistently varied from Cu- to Ga-rich was prepared by growing the film without the substrate rotation. A series of PL spectra was obtained by directing a focused laser beam point-by-point across the boundary separating the Cu- and Ga-rich regions. Distinct features of these spectra were noted. On the Cu-rich side, optical emissions peaked at 1.71, 1.67, 1.63, and 1.59 eV were observed in a PL spectrum. The peak at 1.71 eV was due to the emission of bound exciton, while the peak at 1.67 eV was caused by the free-to-bound transition. The other two peaks were identified to be the donor-to-acceptor emissions. Further annealing experiments performed in different environments suggested that the peaks at 1.67, 1.63, and 1.59 eV were associated with the optical transitions of CB ! CuGa, Cui ! CuGa, and Cui ! VGa, respectively. On the Ga-rich side, a dominant donor-to-acceptor emission peaked at 1.62 eV was observed, which was determined to be the GaCu ! VCu transition. The two defects with opposite charge states resulted in a highly compensated material with high resistivity. q 2003 Elsevier Ltd. All rights reserved.
1. Introduction CuGaSe2, a member of I – III – VI2 compound semiconductor [1,2], has a direct energy gap of 1.68 eV at room temperature, a very high absorption coefficient up to 105 cm21 at 500 nm and an easily controllable electrical resistivity in a wide range (1021 – 105 Ohm cm). These properties make CuGaSe2 an ideal high bandgap partner with CuInSe2 in a solar-cell structure. Several techniques including liquid phase epitaxy (LPE) [3], metal – organic vapor phase epitaxy (MOVPE) [4], molecular beam epitaxy (MBE) [5], and halogen transport method [6], had been used to grow CuGaSe2 epilayers on ZnSe, GaAs, and GaP substrates. Among the film growth techniques, MBE has favorable features such as a relatively low growth temperature, a low growth rate, and simple growth mechanism. As a result of these advantages, the MBE technique provides an abrupt interface, a smooth surface, and a precise control on composition and thickness. In this work, an MBE technique is developed to * Corresponding author. E-mail address: [email protected] (B.H. Tseng).
grow high-quality CuGaSe2 epitaxial films for structural and optical characterizations. We choose the (001)GaAs wafer as the substrate to grow a CuGaSe2 film because of its availability and relatively low cost. The use of (001) substrate orientation may eliminate the formation of orientation domain structure in the epitaxial film [7]. Like other I –III – VI2 compounds, CuGaSe2 has a wide phase stability region. The Cu2Se – Ga2Se3 pseudobinary phase diagram shows the chalcopyrite phase boundaries extended from 50 to 56 mole % Ga2Se3 [8]. A considerable amount of intrinsic point defects exist in the material as its composition deviated from the stoichiometry. Optical and electrical properties are thus greatly affected by the film composition. The conductivity type of the film can even be changed without the introduction of doping elements. It is essential to identify the energy levels of intrinsic point defects and their abundance in the material. Photoluminescence (PL) spectroscopy is one of the most powerful techniques for such a study. Several research groups have shown the results of PL measurement for bulk CuGaSe2 single crystals. Three donor levels of 10, 80, and 100 meV associated with GaCu, VSe, and Gai, respectively, and three acceptor levels of 15, 55,
0022-3697/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0022-3697(03)00151-3
1902
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
and 95 meV associated with VGa, VCu and Sei, respectively, were assigned by Schon et al. [9]. In addition, PL characterizations of epitaxial CuGaSe2 films were examined by several Japanese groups. A donor level of 108 meV (VSe) and an acceptor level of 35 meV (CuGa) were appointed by Yamada et al. [10]. Due to the complex defect chemistry in CuGaSe2 and the discrepancy of composition measurements, the results of PL measurement were still not consistent. In this work, a near-stoichiometric CuGaSe2 epitaxial film with a composition gradient varying from Cuto Ga-rich was used for the analysis. We also developed a technique for unambiguously determining the film composition using an electron microprobe. The relationship between chemical composition and optical emission spectra is thus established. Further experiments of thermal annealing in various ambient helped the identification of the types of intrinsic point defects which involved in optical emissions.
will meet this requirement. Due to the difficulty in preparing a standard of the ternary chalcopyrite compound with uniform composition, we use pure Cu, GaAs, and ZnSe as elemental standards. The use of elemental standards has an advantage to standardize the measurement procedures and leads to a consistency on the interpretation of experimental results. Composition measurements using the above-mentioned method had been proven successful for determining the chemical composition of CuInSe2 [11]. Our previous composition measurements on thin films of CuInSe2 indicated that there was a composition variation in a microscopic scale. A successful correlation between the PL and EPMA results should have an average of composition data from the same area of optical excitation. Thus, every composition data was average from 5 to 9 detection spots using a beam size of 5 mm.
3. Results and discussion 2. Experimental procedures CuGaSe2 films were grown on (001)GaAs substrates by an MBE technique. A Seiko Seiki SP-400 turbomolecular pump was used to evacuate the MBE system and the background pressure was 4 £ 1029 Torr. High-purity elemental sources of Cu(6N), Ga(7N), and Se(6N) were loaded in the effusion cells made by high-density and highpurity graphite. The substrates were cleaned in acetone and methanol subsequently to remove the organic contaminants and then etched in the NH4OH:H2O2:H2O ¼ 3:1:15 solution for 10 s before loading to the growth chamber. Thin films with different compositions were prepared by varying the temperature of the Ga source, while the temperatures of the Cu and Se sources were kept constant. The films were grown at a substrate temperature of 500 8C for 1 h. PL spectroscopy was employed to study the optical transitions in a CuGaSe2 film. A 514.5 nm emission line of a Coherent Innova-90 Ar ion laser was selected as an excitation source. The laser beam could be focused to a spot with a diameter about 0.2 mm. The sample was fixed in a liquid-He cryostat and could be cooled down to 9 K. The PL signals were focused into an Acton Research SpectraPro 500 monochromator and detected by a Si detector. These signals were processed by a lock-in amplifier with a chopper frequency of 330 Hz and finally collected by a computer. The laser power for PL excitation could be varied by a neutral density filter for the identification of optical transition types. Chemical compositions of CuGaSe2 films were measured by a JEOL JXA-8900R electron microprobe. A simple method by using an electron beam with its energy below 10 keV may prevent the excitation of the substrate material and greatly simplify the analyzing procedures. Normally, a CuGaSe2 film with the thickness about 500 nm
Thin film of CuGaSe2 was grown on an one-inch square GaAs wafer by an MBE technique without a substrate rotation. A proper control of source fluxes would produce a film with both Cu- and Ga-rich regions in one sample. This was the condition chosen for an overall film composition very close to stoichiometry. The composition data obtained from this film is shown in Fig. 1a. As can be seen, the Cu and Ga contents are consistently varied from one side to the other side while the Se content is fairly uniform throughout the film. Careful examinations on the data of the Se content reveals that the Ga-rich region has a value slightly higher than 50 at.% and the Cu-rich region has a value slightly lower than 50 at.%. This result suggests that the film may grow in a two-step process, i.e. binary compounds of Cu2Se and Ga2Se3 form first on the surface and subsequently react to form a ternary chalcopyrite compound [11]. The Cu/Ga ratios of this sample vary from 0.94 to 1.10 as shown in Fig. 1b. The Cu/ Ga ratio at the boundary separating the Cu- and Ga-rich region is about 1. A series of PL spectra were obtained at a temperature of 9 K using a laser beam focused on the positions where the composition measurements were made. Fig. 2 shows the PL spectra and the corresponding composition data. We see that the PL spectra are significantly different in the Cu- and Garich regions. The emission peaks at 1.71, 1.67, 1.63, and 1.59 eV are observed in the Cu-rich region, while a single peak at 1.62 eV is found in the Ga-rich region. A red shift of the peak originally located at 1.62 eV is also noted when the composition is deviated from the stoichiometry. The offstoichiometry composition leads to variation in the lattice constant, which may induce additional mechanical stress and cause a red shift of PL signals. Excitation power dependence on the energy shift and the intensity of individual PL peak had been investigated to
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
1903
identify the transition types of optical emissions [12]. A laser power range of 1 – 400 mW was used for this experiment and the data obtained from two selected spots, one on the Cu-rich region and the other on the In-rich region, were plotted in Fig. 3. Optical emissions originally peaked at 1.62, 1.63, and 1.59 eV show an energy shift about 30 meV as the excitation power increases, whose intensity also show a sublinear excitation power dependence. The characteristics of emission peaks like this is typical for a donor-to-acceptor (D – A) pair transition. On the other hand, the intensity of optical emission peaked at 1.67 eV increases linearly with respect to the excitation power and shows almost no shift in peak energy. These behaviors indicate that it is caused by a free-to-bound (F – B) transition. Finally, the emission peaked at 1.71 eV has a superlinear intensity vs. power dependence and is attributed to the exciton emission. The above analysis indicates that the excitonic and the F– B transitions are dominant in the Cu-rich region, while the D– A pair transitions prevail in the Ga-rich region. Vacancies, interstitials, and antisites are typical intrinsic point defects in CuGaSe2. There are twelve in
Fig. 1. (a)Composition data measured from a near-stoichiometric sample; (b) a plot of the data in term of the Cu/Ga ratio.
Fig. 2. A series of PL spectra obtained from the near-stoichiometric sample.
Fig. 3. The excitation power dependence of (a) energy shifts and (b) intensity of the PL peaks.
1904
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
total. Chemical composition of the film as well as the formation energy of point defect determines what kinds of defects are dominant in the film. Theoretical calculations on the defect formation energy have been done by Wei et al. [13,14] and Neumann [15]. The results are summarized in Table 1. As can be seen, Cu interstitial (Cui), Cu on Ga antisite (CuGa), Ga on Cu antisite (GaCu), Cu vacancy (VCu), Ga vacancy (VGa), Se vacancy (VSe), and defect pair of (GaCu þ VCu) have relatively low formation energies. For a Cu-rich film, the dominant defect types will be Ga vacancy (acceptor), Cu interstitial (donor), CuGa antisite (acceptor), and Se vacancy (VSe). A similar argument can be applied to the Ga-rich film. Thermal annealing in various ambient may result in a change of defect distributions. This will help to identify the defect levels involved in optical transitions. Vacuum annealing will cause thermal decomposition from the film surface in the form of Cu2Se and Ga2Se3. It has been known that more Ga2Se3 escapes than Cu2Se. This leads to a surface composition deficient in the Se content and also shifts the Cu/Ga ratio to a higher value. To prohibit thermal decomposition, the sample should be annealed in a close tube filled with Se vapor to suppress surface decomposition. An alternate method is to anneal the sample in the MBE chamber under the exposure to a Se-beam flux. This is a convenient operation as compared with the close-tube annealing. Although the annealing is done in an open system, a quasi-equilibrium state can be maintained. In this work, the film was annealed at 500 8C for 30 min in a Sebeam flux of 1.5 £ 1014 atoms/cm2 s. Fig. 4 shows the composition data measured throughout a near-stoichiometric film before and after annealing. We see an overall increase in the Se content up to 50.2 at.% and a very slight change in the Cu/Ga ratios of this film. Our previous TEM observations of the as-grown CuInSe2 and CuGaSe2 films showed that antiphase domain boundaries (APB’s) prevailed in the film. Thermal annealing in the presence of a Se-beam flux
Table 1 A partial of formation energy of intrinsic point defects in CuGaSe2 Defect types
Formation energy (eV)
Reference
Cu vacancy (VCu) Ga vancancy (VGa) Se vacancy (VSe) Cu vacancy (VCu) Ga vancancy (VGa) Cu interstitial (Cui) Cu on Ga antisite (CuGa) Ga on Cu antisite (GaCu) Defect pair (GaCu þ 2VCu)
3.43 2.72 3.07 0.66 2.83 1.91 1.41 2.04 0.20
[15] [15] [15] [13] [13] [13] [13] [13] [13]
Other defect types not listed here have much lager formation energy.
Fig. 4. Chemical composition data of a near-stoichiometric film annealed at 500 8C for 30 min in a MBE chamber under the exposure to a Se-beam flux.
did eliminate the APB’s and considerably reduced the antisite defects in the film [7]. Fig. 5 shows the PL spectra of the as-grown sample with chemical composition in the Cu-rich region and the same sample after annealing in a Se vapor flux. A dramatic decrease in the intensity of the optical emission peaked at 1.67 eV and a relatively slight decrease in the PL intensity peaked at 1.63 eV are observed after annealing. The annealing process was further conducted subsequently by depositing a 20 nm-thick Cu layer on the film surface at room temperature and then by heating the sample in vacuum at 500 8C for 30 min. The intensities of both peaks are increased after annealing. As mentioned above, two acceptors (CuGa, VGa) and two donors (Cui, VSe) are prevailing in the Cu-rich region. The defect concentration of VSe, CuGa, and Cui may decrease after annealing in a Se vapor flux, while an increase of CuGa and Cui may occur after annealing with the film surface capped by a thin Cu layer. It implies that these two peaks are associated with CuGa and Cui. The PL peak at 1.67 eV has been identified to be caused by the F– B transition. It may be assigned to be the transition from conduction band to an acceptor level of CuGa according to our previous work on CuInSe2 [7]. Another peak at 1.63 eV is an emission caused by the D– A transition and may be attributed to the transition from a donor level of Cui to an acceptor level of CuGa. Finally, a D– A emission peaked at 1.59 eV did not have a significant change in
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
1905
Fig. 5. A series of PL spectra obtained from the Cu-rich region: (a) as-grown, (b) Se-annealing, and (c) Cu-annealing.
Fig. 6. A series of PL spectra obtained from the Ga-rich region: (a) as-grown and (b) Cu-annealing.
intensity after annealing. It may be caused by the transition from a donor level of Cui to an acceptor level of VGa, because the concentration of VGa may decrease after annealing with a Cu cap layer but do not increase after annealing in a Se vapor flux (the Cu/Ga ratio of the film decreases from 1.05 to 1.02 after annealing). The PL spectrum obtained from the Ga-rich region of a near-stoichiometric film show a main peak at 1.62 eV (see Fig. 6), which was characterized to be caused by the D– A transition. Intrinsic point defects of VCu and GaCu are the majority in the Ga-rich region according to the composition data. The peak at 1.62 eV is attributed to the GaCu –VCu transition. After annealing the sample with a thin Cu layer capped on the surface, the PL spectrum shows a decrease in the peak intensity at 1.62 eV and an increase in the intensity of the emission peaked at 1.66 eV. This annealing process may cause a decrease in the number of VCu and GaCu. Because of the relatively low formation energy of VCu, the number of VCu may diminish faster than that of GaCu. This implies that the peak at 1.66 eV is associated with the emission from a donor level of GaCu to the valence band. There is also one broad peak below 1.50 eV detected after annealing, which is probably caused by some deep levels. The origins of this emission need further investigation. Four-point probe was used to measure the resistivity of the as-grown sample. The result is shown in Fig. 7. As can
be seen, the resistivity decreases drastically by several orders of magnitude as composition varies from the Ga-rich region to the Cu-rich region. The low resistivity measured in the Cu-rich region is probably due to the abundance of
Fig. 7. The resistivity data measured point-by-point from a nearstoichiometric film.
1906
G.L. Gu et al. / Journal of Physics and Chemistry of Solids 64 (2003) 1901–1906
the CuGa antisites (acceptors) in that area. On the other hand, high compensation of VCu (acceptor) and GaCu (donor) in the Ga-rich region leads to high resistivity of the film. The resistivity increased as the film composition became more Ga-rich. When the Cu/Ga ratio was less than 0.9, the resistivity was too high and out of the measurement range of a four-point probe. It is found that the resistivity data is consistent with the PL results.
4. Conclusions A specific sample of CuGaSe2 with chemical composition very close to the stoichiometry was prepared. A boundary separating the Cu-rich and Ga-rich region could be clearly observed and the composition measured by EPMA using elemental standards indicated the Cu/Ga ratio about 1 along the boundary. The EPMA measurements also showed a consistent composition variation across the boundary. Further characterizations indicated that the resistivity and PL results fitted quite well with the composition data. The technique for composition measurements developed in this work can be easily applied to the thin-film samples and also avoids the use of compound standards which may cause the discrepancy in composition measurements. In this work, we did the annealing experiments on this sample and tried to identify the intrinsic point defects which caused the optical transitions. The assignments were based on chemical composition, calculated formation energy of the defects, and the characteristics of PL emission peaks. The PL peaks at 1.71, 1.67, 1.63, and 1.59 eV were observed on the Cu-rich side of the sample. Further experiments suggested that the peaks at 1.67, 1.63 and 1.59 eV were associated with the optical transitions of CB ! CuGa, Cui ! CuGa, and Cui ! VGa, respectively. On
the Ga-rich side, a dominant donor-to-acceptor emission peaked at 1.62 eV was observed, which was determined to be the GaCu ! VCu transition.
References [1] L. Shay, S. Wagner, H.M. Kasper, Appl. Phys. Lett. 27 (1975) 89. [2] L.I. Kazmerski, F.R. White, G.K. Morgan, Appl. Phys. Lett. 29 (1975) 268. [3] H. Asai, K. Sugiyama, Jpn. J. Appl. Phys. 20 (1981) 1401. [4] S. Chichibu, Y. Harada, M. Uchida, T. Wakiyama, S. Matsumoto, S. Shirakata, S. Isomola, H. Higuchi, J. Appl. Phys. 76 (1994) 3009. [5] A. Yamada, Y. Makita, S. Niki, A. Obara, P. Fons, H. Shibata, M. Kawai, S. Chichibu, H. Nakanishi, J. Appl. Phys. 79 (1996) 4318. [6] O. Igarashi, J. Crystal, Growth 143 (1994) 213. [7] B.H. Tseng, S.B. Lin, G.L. Gu, W. Chen, J. Appl. Phys. 79 (1996) 1391. [8] J.C. Mikkelsen, J. Electron. Mater. 10 (1981) 541. [9] J.H. Schon, K. Fess, K. Friemelt, C. Kloc, E. Bucher, Inst. Phys. Conf. Ser. 152 (1998) 59. [10] A. Yamada, P. Fons, S. Niki, H. Shibata, A. Obara, Y. Makita, H. Oyanagi, J. Appl. Phys. 81 (1997) 2794. [11] G.Y. Chiang, The effects of composition variation on luminescence properties of CuInSe2 epitaxial films, MS Thesis, National Sun Yat-Sen University, 1998. [12] S. Niki, Y. Makita, A. Yamada, A. Obara, O. Igarashi, S. Misawa, M. Kawai, H. Nakanishi, Y. Taguchi, N. Kutsuwada, Solar Energy Mater Solar Cells 35 (1994) 141. [13] S.H. Wei, S.B. Zhang, A. Zunger, Appl. Phys. Lett. 72 (1998) 3199. [14] S.B. Zhang, S.H. Wei, A. Zunger, H. Kadayama-Yoshida, Phys. Rev. B 57 (1998) 9642. [15] H. Neumann, Crystal Res. Technol. 18 (1983) 901.