Comparative study of the photoluminescence of ... - Semantic Scholar
Microelectronics Journal 36 (2005) 374–378 www.elsevier.com/locate/mejo
Comparative study of the photoluminescence of InGaP layers grown on GaAs substrates by LPE and MOVPE techniques Tatiana Prutskija,*, Claudio Pelosib, Raul A. Brito-Ortac a
Instituto de Ciencias, BUAP, Apartado Postal 207, 72000 Puebla, Pue., Me´xico b IMEM/CNRParco Area delle Scienze 37/A-43010, Parma, Italy c Instituto de Fı´sica ‘Luis Rivera Terrazas’, UAP, Apartado Postal J-48, 72570 Puebla, Pue., Me´xico Available online 17 March 2005
Abstract We make a comparative study of the luminescent properties of InGaP films grown on GaAs substrates by two different growth techniques: liquid phase epitaxy (LPE) and metal-organic vapor phase epitaxy (MOVPE). The grown InxGa1KxP (xz0.5) films were nearly lattice matched to GaAs and had the same thickness of approximately 0.5 mm. Photoluminescence (PL) measurements were performed in a wide directions. Observations suggest that the temperature (10–300 K) range for polarization of the emitted radiation along the [011] and ½011 InxGa1KxP layers in the structures grown by MOVPE present ordering, while in the layers grown by LPE no ordering was observed. q 2005 Elsevier Ltd. All rights reserved. Keywords: Liquid phase epitaxy; MOVPE; III–V Semiconductors; Photoluminescence
1. Introduction It is known that in InGaP layers lattice-matched to a GaAs substrate and grown by LPE the group III atoms, In and Ga, are randomly arranged within their face-centered cubic sublattice, while in layers of the same material grown by MOVPE these atoms are frequently ordered in the CuPt structure. This means that in MOVPE layers a superlattice structure made of alternate In-rich and Ga or ð111Þ planes with interleaving planes rich diagonal ð111Þ of P atoms is spontaneously formed during the growth. The ordering of the group III atoms leads to a reduction of the bandgap energy and to the splitting of the valence band [1]. The degree of ordering depends on a number of factors, such as the growth temperature, the growth rate, and the orientation of the substrate. Also, the ordering is not uniform, but occurs only in some domains of the film, and this is not good from a practical point of view. One of the techniques that has been used to characterize InGaP layers is photoluminescence spectroscopy. The temperature behavior of the PL emission maximum of * Corresponding author. Tel./fax: C52 222 242 1072. E-mail address: [email protected] (T. Prutskij).
0026-2692/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2005.02.056
ordered and disordered InGaP layers grown by MOVPE were compared in Ref. [2]. For disordered layers, the highenergy peak was identified as an excitonic emission, while for ordered layers, the dominant high-energy peak at low temperature was found to be strongly dependent on excitation intensity, its energy position decreased first and then increased with increasing temperature and it was identified as not due to excitonic emission but to transitions from below-gap states. An excitonic PL peak has been observed at low temperatures in high purity InGaP layers grown by LPE, together with three additional peaks which are identified as LO phonon replicas by analyzing their temperature dependencies [3]. Polarized spectroscopy is one of the basic characterization tools used to study atomic ordering in InGaP layers. Anisotropy in the PL emission polarized parallel to [011] crystallographic directions of the layer (these and ½011 directions are respectively parallel to the cleavage facets of the structures) is frequently observed for layers with ordering. This means that a difference between the peak positions and their intensities in the luminescence for the two polarizations is observed. The temperature dependence of the energy of the peaks and the ratio of the integrated PL intensities of the luminescence in the two polarization directions were studied experimentally and explained using
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
the k$p model applied to the superlattice structure of the layer in Ref. [4]. We report here a comparative study of the photoluminescence and polarized photoluminescence spectra obtained in a wide range of temperatures to characterize InGaP layers, lattice matched to a GaAs substrate, grown by LPE and by MOVPE. After a brief section concerning the growth of the samples there is a section exposing, analyzing, and comparing the PL and polarized PL spectra for both types of samples. Conclusions are given in a final section.
2. Growth Two different growth techniques were used to obtain InGaP/GaAs structures: LPE and MOVPE. The InGaP layers fabricated by the LPE technique (which will be referred to as structures PR-521 and PR-638) were grown using a conventional carbon sliding boat in a horizontal LPE system. For both structures the same growth temperature of 740 8C was used, but two different temperature regimes were applied: for structure PR-521 the solution was cooled down at a rate of 28/min until the growth temperature of 740 8C was reached. After that the layer grew during 1 min. This structure was grown on a (100) GaAs semi-isolated substrate misoriented by 28 towards the [110] direction. Structure PR-638 was grown on a (111)B GaAs semiisolated substrate at the same temperature, but a superquickcooling regime was used, which means that the rate of decrease of the temperature was of the order of 10 8/s, i.e. much greater than in the usual case. The thickness of the InGaP layer is 0.5 mm for PR-521 and 0.35 mm for PR-638. The InGaP layers grown by MOPVE were grown at a temperature of 600 8C, with a growth rate of w1 mm/h, a V/III ratio of 160 and under a pressure value of 60 mbar. The main reagents were TMG, TMI, and phosphine in a 10% mixture with hydrogen. One of the samples, which we call structure PE-86, is a single, 0.5 mm thick, InGaP layer Si-doped (8!1017 cmK3), and grown on a GaAs substrate. The other sample, structure PE-202, is a multilayer structure consisting of three layers: a 0.1 mm AlGaAs layer grown on a Ge substrate, a 0.3 mm GaAs layer grown on top of the first one, and a last layer, a 0.6 mm InGaP layer. The layer thicknesses were taken from the TEM image of the crosssection of this structure. After the growth, the cross section morphology was examined by TEM, and the lattice match between the substrate and the layer was determined from rocking curves measured by high-resolution X-ray diffraction.
375
cryostat, a 500M SPEX monochromator, and a CCD detector. The excitation wavelength was 514.5 nm and the laser-spot area at the sample surface was of about 2 mm2. For the polarized detection PL spectra measurements, a depolarizer was installed at the entrance slit of the monochromator. Fig. 1 shows the dependence on In content of the spectral position of the peak attributed to band-to-band transitions at 300 K. The In content was obtained from the measured rocking curves using Vegard’s law. In the same figure, our results are compared with a curve giving the empirical dependence of the band gap on In content at 300 K taken from Ref. [5]. Note that despite of the fact that this curve was found for LPE grown InGaP samples, it gives gap values lower than those found for our LPE samples. Nevertheless, the gap values of our samples are close to 1.9 eV, which is a well-established value for the In0.5Ga0.5P alloy. However, we compare with the empirical curve because this curve is sometimes used to determine the ordering parameter h for layers grown by MOVPE [6] by means of hh
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi DE0 ; 471 meV
where DE0 is the difference between the gap of the MOVPE sample and the gap value given by the curve for the same In content, and 471 meV is the band-gap reduction for the completely ordered material. The ordering parameter is equal to 0.28 for structure PE-86 and to 0.49 for structure
3. Photoluminescence Photoluminescence spectra were measured in a wide temperature range (10–300 K) using a conventional experimental set-up, which includes an ArC ion laser, a He
Fig. 1. Dependence of the energy position of the 300 K photoluminescence peak due to band-to-band transitions with respect to the In content of the layers as calculated from the diffraction curves. The continuous curve gives the empirical dependence of the band gap on In content at 300 K [5].
376
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
E1
10 K
PR-521 E2
300 K
10 K
PE-86 Eg
300 K
Eg
Eg E1
1.7 1.9 2.0 photon energy, eV
1.8
2.1
1.8 1.9 photon energy, eV
2.0
E2
Fig. 2. Temperature behavior of the energy position of the PL emission peaks of structure PR-521. The inserts show the PL spectra at 10 and 300 K.
PE-202. Thus, according to this ordering criterion, the major grade of ordering occurred in structure PE-202. At low temperatures, the PL spectra of our samples have one, two or three peaks, while at room temperature, there is only one peak in each spectra. The temperature behavior of the spectral position of the peaks for each structure is shown in Figs. 2–5. The experimental points in these figures correspond to the energy position of all maxima in the PL spectra. The continuous curve in each figure is the best fit to the energy position of the peak located at the higher energy in the spectrum. This fit was found using the Varshni relation. The two inserts in each figure show, respectively, the spectra measured at 10 and 300 K. It is worth mentioning that the LPE grown samples have a higher emission intensity than the MOVPE samples. This and the smaller full-width at half-maximum (FWHM) of the PL
Fig. 4. Temperature behavior of the energy position of the PL emission peaks of structure PE-86. The inserts show the PL spectra at 10 and 300 K.
peaks indicates the high crystalline quality of the LPE samples. The PL spectra of the LPE grown structures, PR-521 and PR-638, can be seen in the inserts of Figs. 2 and 3, respectively. At 10 K, both spectra have one peak at 1.985 eV due to band-to-band transitions (peak Eg), and another peak (E1) at approximately 1.938 eV, due to band˚ ) transitions. Peak E1 is related to the to-acceptor (eKA unintentional carbon doping which came from the graphite boat [7]. The temperature behavior of the two peaks (Eg and E1) for both structures is very similar (notice, in particular, that the Varshni expressions for the Eg peaks of both structures have the same parameters). The similarity in energy positions and temperature behavior is very remarkable if one keeps in mind that these structures were grown using two different growing regimes. There are, however,
10 K PL intensity, u. a.
PR-638
PE-202
Eg
10 K E1
300 K
300 k
Eg
Eg E1
1.8
1.9 2.0 photon energy, eV
2.1
Fig. 3. Temperature behavior of the energy position of the PL emission peaks of structure PR-638. The inserts show the PL spectra at 10 and 300 K.
1.7
1.8 1.9 photon energy, eV
2.0
Fig. 5. Temperature behavior of the energy position of the PL emission peaks of structure PE-202. The inserts show the PL spectra at 10 and 300 K.
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
some differences between the spectra of these samples. First, the relative intensities of peaks Eg and E1 are different. As it may be seen in the upper inserts of Figs. 2 and 3, peak E1 dominates in the spectrum of PR-521, while for the PR638 spectrum the main peak is Eg. This means that the impurity concentration in PR-638 is lower. Second, there is a peak, E2, in the spectrum of PR-521, believed to be the ˚ transition, which is not seen in phonon replica of the eKA the spectrum of the other sample. The PL spectra of the layers grown by MOPVE have only one peak which, at high temperatures, can be identified as due to band-to-band transitions (Eg). In these structures, the spectral position of the PL maximum depends on excitation intensity and shifts towards higher energy with increasing excitation intensity. The experimental points shown in Figs. 4 and 5 were obtained using a low excitation intensity. As it may be seen in the figures, the temperature dependence of the spectral position of peaks for both samples is the same. As mentioned before, the continuous curves are the best fits found using the Varshni relation. Notice that, besides the different Eg values at TZ0, the value of the other parameters is the same for both curves. At low temperature, the temperature dependence of the energy position of the peak is not well represented by the Varshni curve. Furthermore, it may be seen from the upper inserts in Figs. 4 and 5 that, at 10 K, the peak of the PL spectrum of structure PE-86 (55 meV) has a full-width at half-maximum (FWHM) which is approximately twice the FWHM of the peak of structure PE-202 (22 meV). The width of the PE-86 peak at 10 K and its form suggest that this peak could be in reality a superposition of two peaks. One of these peaks, the higher-energy peak, is probably due to band-to-band recombination, and the other one, which dominates at low excitation intensity, may be due to a nonzero density of below-gap states [2]. Or, both peaks could be due to the contribution of band-to-band transitions within domains with different ordering parameters. The PL emission was also analyzed in two orthogonal directions of polarization parallel, respectively, to the [011] crystallographic directions of the substrate. The and ½011 polarization of the exciting laser beam was, in all crystallographic direcmeasurements, parallel to the ½011 tion of the sample. For the LPE samples no difference between the spectra for both polarizations was observed. The results for the MOVPE samples are shown in Figs. 6 and 7. Fig. 6 shows the PL spectra for both polarizations for structure PE-86 at two temperatures, 10 and 300 K. Fig. 7 shows the same but for structure PE-202. A difference between both, the peak spectral positions and the peak intensities, for the two orthogonal polarizations was observed in the spectra of both structures at 300 K. This is in agreement with previous findings [4]. In fact, at 300 K, the peak energy for the [011] polarization is about 10 meV polarization for structure (5 meV) lower than for the ½011 PE-86 (PE-202), and this difference decreases for decreasing temperature. Within the model proposed in [4],
PE-86
377
[011] [0-11]
10 K
300 K
Fig. 6. The photoluminescence emission for two orthogonal polarizations for structure PE-86 at two temperatures, 10 and 300 K. Full lines correspond to the ½011 polarization and dashed lines to the [011] polarization.
the difference between the energy position of the peaks for the two polarizations is related to the splitting of the valence band and the decrease in peak energy separation is caused by a relative decrease in occupation of the light-hole band with decreasing temperature. However, the ratio of intensities of the luminescence with orthogonal polarizations I½011 =I½011 is 1.7 at 300 K and 2.8 at 10 K for both structures, i.e. it increases with decreasing temperature. The fact that almost the same values were found for both structures is contradictory with the very different ordering parameters. The fact that the sample with higher ordering [011] [0-11]
PE-202
10 K 300 K
Fig. 7. The photoluminescence emission for two orthogonal polarizations for structure PE-202 at two temperatures, 10 and 300 K. Full lines correspond to the ½011 polarization and dashed lines to the [011] polarization.
378
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
parameter has the smaller separation between peak energy position for the two polarizations (which indicates a smaller valence band splitting) is also contradictory. These two facts can be probably explained by the difficulty in determining the gap value, and consequently the ordering parameter h, in samples with ordering by using only PL measurements. [2] This also can be the result of the action of two different factors: valence-band splitting as a result of ordering and valence-band splitting as a result of the strain introduced into the layer by the no-coincidence of the lattice parameters of the layer and the substrate [8]. The layers obtained by MOVPE were grown on different substrates and, therefore, they can have different amounts of strain. In the TEM images of PE-202 some amount of dislocations is observed and, probably, this layer may be more relaxed than the layer in PE-86. From the experimental point of view it is not easy to separate the contributions of each effect. However, on strain relaxed layers with a thickness of more than 1 mm, one can expect that effects of strain will be minimized and only the ordering will be responsible for the separation of the peaks in the PL polarized spectra. Further investigations are, therefore, needed to determine the possible existence of strain and its effects on the valence-band splitting in the MOVPE samples.
4. Conclusions We compared the temperature behavior of the PL emission of layers of InGaP nearly lattice-matched to a GaAs substrate grown by two different epitaxial techniques (LPE and MOVPE) and with different degree of ordering. The LPE grown layers did not show any effect related to ordering of the group III atoms within their crystalline sublattice. Using the energy position of the PL peak at room temperature, we found different values of the ordering parameter for the two MOVPE samples studied. Also, a difference between the peak spectral positions and the peak intensities for the PL emission in two orthogonal
polarizations was observed in the spectra of these samples at 300 K. The value of the valence-band splitting in each sample suggested by the peaks separation are, however, not consistent with the values of the ordering parameters.
Acknowledgements This work was supported in a bilateral collaboration between Conacyt (Mexico) and CNR (Italy). Support was also received from the VIEP-BUAP, grant III 65-04/ING/G.
References [1] A. Mascarenhas, S. Kurtz, A. Kibbler, J.M. Olson, Polarized and-edge photoluminescence and ordering in Ga0.52In0.48P, Phys. Rev. Lett. 63 (1989) 2108–2111. [2] M.C. DeLong, D.J. Mowbray, R.A. Hogg, M.S. Skolnick, M. Hopkinson, J.P.R. David, P.C. Taylor, S.R. Kurtz, J.M. Olson, Photolumninescence, photoluminescence excitation and resonant Raman spectroscopy of disordered and ordered Ga0.52In0.48P, J. Appl. Phys. 73 (1993) 5163–5172. [3] G.C. Jiang, Y. Chang, L.B. Chang, Y.D. Juang, W.L. Lu, L.S. Lu, K.H. Chang, Photoluminescence of three phonon replicas of the bound exciton in undoped InGaP prepared by liquid phase epitaxy, J. Appl. Phys. 78 (1995) 2886–2888. [4] G.W. ‘t Hooft, C.J.B. Riviere, M.P.C. Krijin, C.T.H.F. Liedenbaumt, A. Valster, Ordering induced splitting of light-hole and heavy-hole bands in GaInP grown y organometallic vapor-phase epitaxy, Appl. Phys. Lett. 61 (1992) 3169–3171. [5] G.B. Stringfellow, The importance of lattice mismatch in the growth of GaxIn1KxP epitaxial crystals, J. Appl. Phys. 43 (1972) 3455–3460. [6] P. Ernst, C. Geng, F. Scholz, H. Schweizer, Y. Zhang, A. Mascarehnas, Band-gap reduction and valence-band splitting of ordered GaInP2, Appl. Phys. Lett. 67 (1995) 2347–2349. [7] J.B. Lee, S.D. Kwon, I. Kim, Y.H. Cho, B.-D. Choe, The characteristics of an In0.5Ga0.5P and In0.5Ga0.5P/GaAs heterojunction grown on a (100) GaAs substrate by liquid-phase epitaxy, J. Appl. Phys. 71 (1992) 5016– 5021. [8] T. Prutskij, P. Dı´az-Arencibia, R.A. Brito-Orta, A. Mintairov, T. Kosel, J. Merz, Luminescence anisotropy of InGaP layers grown by liquid phase epitaxy, J. Phy. D: Appl. Phy. 37 (2004) 1563–1568.
Comparative study of the photoluminescence of InGaP layers grown on GaAs substrates by LPE and MOVPE techniques Tatiana Prutskija,*, Claudio Pelosib, Raul A. Brito-Ortac a
Instituto de Ciencias, BUAP, Apartado Postal 207, 72000 Puebla, Pue., Me´xico b IMEM/CNRParco Area delle Scienze 37/A-43010, Parma, Italy c Instituto de Fı´sica ‘Luis Rivera Terrazas’, UAP, Apartado Postal J-48, 72570 Puebla, Pue., Me´xico Available online 17 March 2005
Abstract We make a comparative study of the luminescent properties of InGaP films grown on GaAs substrates by two different growth techniques: liquid phase epitaxy (LPE) and metal-organic vapor phase epitaxy (MOVPE). The grown InxGa1KxP (xz0.5) films were nearly lattice matched to GaAs and had the same thickness of approximately 0.5 mm. Photoluminescence (PL) measurements were performed in a wide directions. Observations suggest that the temperature (10–300 K) range for polarization of the emitted radiation along the [011] and ½011 InxGa1KxP layers in the structures grown by MOVPE present ordering, while in the layers grown by LPE no ordering was observed. q 2005 Elsevier Ltd. All rights reserved. Keywords: Liquid phase epitaxy; MOVPE; III–V Semiconductors; Photoluminescence
1. Introduction It is known that in InGaP layers lattice-matched to a GaAs substrate and grown by LPE the group III atoms, In and Ga, are randomly arranged within their face-centered cubic sublattice, while in layers of the same material grown by MOVPE these atoms are frequently ordered in the CuPt structure. This means that in MOVPE layers a superlattice structure made of alternate In-rich and Ga or ð111Þ planes with interleaving planes rich diagonal ð111Þ of P atoms is spontaneously formed during the growth. The ordering of the group III atoms leads to a reduction of the bandgap energy and to the splitting of the valence band [1]. The degree of ordering depends on a number of factors, such as the growth temperature, the growth rate, and the orientation of the substrate. Also, the ordering is not uniform, but occurs only in some domains of the film, and this is not good from a practical point of view. One of the techniques that has been used to characterize InGaP layers is photoluminescence spectroscopy. The temperature behavior of the PL emission maximum of * Corresponding author. Tel./fax: C52 222 242 1072. E-mail address: [email protected] (T. Prutskij).
0026-2692/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.mejo.2005.02.056
ordered and disordered InGaP layers grown by MOVPE were compared in Ref. [2]. For disordered layers, the highenergy peak was identified as an excitonic emission, while for ordered layers, the dominant high-energy peak at low temperature was found to be strongly dependent on excitation intensity, its energy position decreased first and then increased with increasing temperature and it was identified as not due to excitonic emission but to transitions from below-gap states. An excitonic PL peak has been observed at low temperatures in high purity InGaP layers grown by LPE, together with three additional peaks which are identified as LO phonon replicas by analyzing their temperature dependencies [3]. Polarized spectroscopy is one of the basic characterization tools used to study atomic ordering in InGaP layers. Anisotropy in the PL emission polarized parallel to [011] crystallographic directions of the layer (these and ½011 directions are respectively parallel to the cleavage facets of the structures) is frequently observed for layers with ordering. This means that a difference between the peak positions and their intensities in the luminescence for the two polarizations is observed. The temperature dependence of the energy of the peaks and the ratio of the integrated PL intensities of the luminescence in the two polarization directions were studied experimentally and explained using
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
the k$p model applied to the superlattice structure of the layer in Ref. [4]. We report here a comparative study of the photoluminescence and polarized photoluminescence spectra obtained in a wide range of temperatures to characterize InGaP layers, lattice matched to a GaAs substrate, grown by LPE and by MOVPE. After a brief section concerning the growth of the samples there is a section exposing, analyzing, and comparing the PL and polarized PL spectra for both types of samples. Conclusions are given in a final section.
2. Growth Two different growth techniques were used to obtain InGaP/GaAs structures: LPE and MOVPE. The InGaP layers fabricated by the LPE technique (which will be referred to as structures PR-521 and PR-638) were grown using a conventional carbon sliding boat in a horizontal LPE system. For both structures the same growth temperature of 740 8C was used, but two different temperature regimes were applied: for structure PR-521 the solution was cooled down at a rate of 28/min until the growth temperature of 740 8C was reached. After that the layer grew during 1 min. This structure was grown on a (100) GaAs semi-isolated substrate misoriented by 28 towards the [110] direction. Structure PR-638 was grown on a (111)B GaAs semiisolated substrate at the same temperature, but a superquickcooling regime was used, which means that the rate of decrease of the temperature was of the order of 10 8/s, i.e. much greater than in the usual case. The thickness of the InGaP layer is 0.5 mm for PR-521 and 0.35 mm for PR-638. The InGaP layers grown by MOPVE were grown at a temperature of 600 8C, with a growth rate of w1 mm/h, a V/III ratio of 160 and under a pressure value of 60 mbar. The main reagents were TMG, TMI, and phosphine in a 10% mixture with hydrogen. One of the samples, which we call structure PE-86, is a single, 0.5 mm thick, InGaP layer Si-doped (8!1017 cmK3), and grown on a GaAs substrate. The other sample, structure PE-202, is a multilayer structure consisting of three layers: a 0.1 mm AlGaAs layer grown on a Ge substrate, a 0.3 mm GaAs layer grown on top of the first one, and a last layer, a 0.6 mm InGaP layer. The layer thicknesses were taken from the TEM image of the crosssection of this structure. After the growth, the cross section morphology was examined by TEM, and the lattice match between the substrate and the layer was determined from rocking curves measured by high-resolution X-ray diffraction.
375
cryostat, a 500M SPEX monochromator, and a CCD detector. The excitation wavelength was 514.5 nm and the laser-spot area at the sample surface was of about 2 mm2. For the polarized detection PL spectra measurements, a depolarizer was installed at the entrance slit of the monochromator. Fig. 1 shows the dependence on In content of the spectral position of the peak attributed to band-to-band transitions at 300 K. The In content was obtained from the measured rocking curves using Vegard’s law. In the same figure, our results are compared with a curve giving the empirical dependence of the band gap on In content at 300 K taken from Ref. [5]. Note that despite of the fact that this curve was found for LPE grown InGaP samples, it gives gap values lower than those found for our LPE samples. Nevertheless, the gap values of our samples are close to 1.9 eV, which is a well-established value for the In0.5Ga0.5P alloy. However, we compare with the empirical curve because this curve is sometimes used to determine the ordering parameter h for layers grown by MOVPE [6] by means of hh
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi DE0 ; 471 meV
where DE0 is the difference between the gap of the MOVPE sample and the gap value given by the curve for the same In content, and 471 meV is the band-gap reduction for the completely ordered material. The ordering parameter is equal to 0.28 for structure PE-86 and to 0.49 for structure
3. Photoluminescence Photoluminescence spectra were measured in a wide temperature range (10–300 K) using a conventional experimental set-up, which includes an ArC ion laser, a He
Fig. 1. Dependence of the energy position of the 300 K photoluminescence peak due to band-to-band transitions with respect to the In content of the layers as calculated from the diffraction curves. The continuous curve gives the empirical dependence of the band gap on In content at 300 K [5].
376
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
E1
10 K
PR-521 E2
300 K
10 K
PE-86 Eg
300 K
Eg
Eg E1
1.7 1.9 2.0 photon energy, eV
1.8
2.1
1.8 1.9 photon energy, eV
2.0
E2
Fig. 2. Temperature behavior of the energy position of the PL emission peaks of structure PR-521. The inserts show the PL spectra at 10 and 300 K.
PE-202. Thus, according to this ordering criterion, the major grade of ordering occurred in structure PE-202. At low temperatures, the PL spectra of our samples have one, two or three peaks, while at room temperature, there is only one peak in each spectra. The temperature behavior of the spectral position of the peaks for each structure is shown in Figs. 2–5. The experimental points in these figures correspond to the energy position of all maxima in the PL spectra. The continuous curve in each figure is the best fit to the energy position of the peak located at the higher energy in the spectrum. This fit was found using the Varshni relation. The two inserts in each figure show, respectively, the spectra measured at 10 and 300 K. It is worth mentioning that the LPE grown samples have a higher emission intensity than the MOVPE samples. This and the smaller full-width at half-maximum (FWHM) of the PL
Fig. 4. Temperature behavior of the energy position of the PL emission peaks of structure PE-86. The inserts show the PL spectra at 10 and 300 K.
peaks indicates the high crystalline quality of the LPE samples. The PL spectra of the LPE grown structures, PR-521 and PR-638, can be seen in the inserts of Figs. 2 and 3, respectively. At 10 K, both spectra have one peak at 1.985 eV due to band-to-band transitions (peak Eg), and another peak (E1) at approximately 1.938 eV, due to band˚ ) transitions. Peak E1 is related to the to-acceptor (eKA unintentional carbon doping which came from the graphite boat [7]. The temperature behavior of the two peaks (Eg and E1) for both structures is very similar (notice, in particular, that the Varshni expressions for the Eg peaks of both structures have the same parameters). The similarity in energy positions and temperature behavior is very remarkable if one keeps in mind that these structures were grown using two different growing regimes. There are, however,
10 K PL intensity, u. a.
PR-638
PE-202
Eg
10 K E1
300 K
300 k
Eg
Eg E1
1.8
1.9 2.0 photon energy, eV
2.1
Fig. 3. Temperature behavior of the energy position of the PL emission peaks of structure PR-638. The inserts show the PL spectra at 10 and 300 K.
1.7
1.8 1.9 photon energy, eV
2.0
Fig. 5. Temperature behavior of the energy position of the PL emission peaks of structure PE-202. The inserts show the PL spectra at 10 and 300 K.
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
some differences between the spectra of these samples. First, the relative intensities of peaks Eg and E1 are different. As it may be seen in the upper inserts of Figs. 2 and 3, peak E1 dominates in the spectrum of PR-521, while for the PR638 spectrum the main peak is Eg. This means that the impurity concentration in PR-638 is lower. Second, there is a peak, E2, in the spectrum of PR-521, believed to be the ˚ transition, which is not seen in phonon replica of the eKA the spectrum of the other sample. The PL spectra of the layers grown by MOPVE have only one peak which, at high temperatures, can be identified as due to band-to-band transitions (Eg). In these structures, the spectral position of the PL maximum depends on excitation intensity and shifts towards higher energy with increasing excitation intensity. The experimental points shown in Figs. 4 and 5 were obtained using a low excitation intensity. As it may be seen in the figures, the temperature dependence of the spectral position of peaks for both samples is the same. As mentioned before, the continuous curves are the best fits found using the Varshni relation. Notice that, besides the different Eg values at TZ0, the value of the other parameters is the same for both curves. At low temperature, the temperature dependence of the energy position of the peak is not well represented by the Varshni curve. Furthermore, it may be seen from the upper inserts in Figs. 4 and 5 that, at 10 K, the peak of the PL spectrum of structure PE-86 (55 meV) has a full-width at half-maximum (FWHM) which is approximately twice the FWHM of the peak of structure PE-202 (22 meV). The width of the PE-86 peak at 10 K and its form suggest that this peak could be in reality a superposition of two peaks. One of these peaks, the higher-energy peak, is probably due to band-to-band recombination, and the other one, which dominates at low excitation intensity, may be due to a nonzero density of below-gap states [2]. Or, both peaks could be due to the contribution of band-to-band transitions within domains with different ordering parameters. The PL emission was also analyzed in two orthogonal directions of polarization parallel, respectively, to the [011] crystallographic directions of the substrate. The and ½011 polarization of the exciting laser beam was, in all crystallographic direcmeasurements, parallel to the ½011 tion of the sample. For the LPE samples no difference between the spectra for both polarizations was observed. The results for the MOVPE samples are shown in Figs. 6 and 7. Fig. 6 shows the PL spectra for both polarizations for structure PE-86 at two temperatures, 10 and 300 K. Fig. 7 shows the same but for structure PE-202. A difference between both, the peak spectral positions and the peak intensities, for the two orthogonal polarizations was observed in the spectra of both structures at 300 K. This is in agreement with previous findings [4]. In fact, at 300 K, the peak energy for the [011] polarization is about 10 meV polarization for structure (5 meV) lower than for the ½011 PE-86 (PE-202), and this difference decreases for decreasing temperature. Within the model proposed in [4],
PE-86
377
[011] [0-11]
10 K
300 K
Fig. 6. The photoluminescence emission for two orthogonal polarizations for structure PE-86 at two temperatures, 10 and 300 K. Full lines correspond to the ½011 polarization and dashed lines to the [011] polarization.
the difference between the energy position of the peaks for the two polarizations is related to the splitting of the valence band and the decrease in peak energy separation is caused by a relative decrease in occupation of the light-hole band with decreasing temperature. However, the ratio of intensities of the luminescence with orthogonal polarizations I½011 =I½011 is 1.7 at 300 K and 2.8 at 10 K for both structures, i.e. it increases with decreasing temperature. The fact that almost the same values were found for both structures is contradictory with the very different ordering parameters. The fact that the sample with higher ordering [011] [0-11]
PE-202
10 K 300 K
Fig. 7. The photoluminescence emission for two orthogonal polarizations for structure PE-202 at two temperatures, 10 and 300 K. Full lines correspond to the ½011 polarization and dashed lines to the [011] polarization.
378
T. Prutskij et al. / Microelectronics Journal 36 (2005) 374–378
parameter has the smaller separation between peak energy position for the two polarizations (which indicates a smaller valence band splitting) is also contradictory. These two facts can be probably explained by the difficulty in determining the gap value, and consequently the ordering parameter h, in samples with ordering by using only PL measurements. [2] This also can be the result of the action of two different factors: valence-band splitting as a result of ordering and valence-band splitting as a result of the strain introduced into the layer by the no-coincidence of the lattice parameters of the layer and the substrate [8]. The layers obtained by MOVPE were grown on different substrates and, therefore, they can have different amounts of strain. In the TEM images of PE-202 some amount of dislocations is observed and, probably, this layer may be more relaxed than the layer in PE-86. From the experimental point of view it is not easy to separate the contributions of each effect. However, on strain relaxed layers with a thickness of more than 1 mm, one can expect that effects of strain will be minimized and only the ordering will be responsible for the separation of the peaks in the PL polarized spectra. Further investigations are, therefore, needed to determine the possible existence of strain and its effects on the valence-band splitting in the MOVPE samples.
4. Conclusions We compared the temperature behavior of the PL emission of layers of InGaP nearly lattice-matched to a GaAs substrate grown by two different epitaxial techniques (LPE and MOVPE) and with different degree of ordering. The LPE grown layers did not show any effect related to ordering of the group III atoms within their crystalline sublattice. Using the energy position of the PL peak at room temperature, we found different values of the ordering parameter for the two MOVPE samples studied. Also, a difference between the peak spectral positions and the peak intensities for the PL emission in two orthogonal
polarizations was observed in the spectra of these samples at 300 K. The value of the valence-band splitting in each sample suggested by the peaks separation are, however, not consistent with the values of the ordering parameters.
Acknowledgements This work was supported in a bilateral collaboration between Conacyt (Mexico) and CNR (Italy). Support was also received from the VIEP-BUAP, grant III 65-04/ING/G.
References [1] A. Mascarenhas, S. Kurtz, A. Kibbler, J.M. Olson, Polarized and-edge photoluminescence and ordering in Ga0.52In0.48P, Phys. Rev. Lett. 63 (1989) 2108–2111. [2] M.C. DeLong, D.J. Mowbray, R.A. Hogg, M.S. Skolnick, M. Hopkinson, J.P.R. David, P.C. Taylor, S.R. Kurtz, J.M. Olson, Photolumninescence, photoluminescence excitation and resonant Raman spectroscopy of disordered and ordered Ga0.52In0.48P, J. Appl. Phys. 73 (1993) 5163–5172. [3] G.C. Jiang, Y. Chang, L.B. Chang, Y.D. Juang, W.L. Lu, L.S. Lu, K.H. Chang, Photoluminescence of three phonon replicas of the bound exciton in undoped InGaP prepared by liquid phase epitaxy, J. Appl. Phys. 78 (1995) 2886–2888. [4] G.W. ‘t Hooft, C.J.B. Riviere, M.P.C. Krijin, C.T.H.F. Liedenbaumt, A. Valster, Ordering induced splitting of light-hole and heavy-hole bands in GaInP grown y organometallic vapor-phase epitaxy, Appl. Phys. Lett. 61 (1992) 3169–3171. [5] G.B. Stringfellow, The importance of lattice mismatch in the growth of GaxIn1KxP epitaxial crystals, J. Appl. Phys. 43 (1972) 3455–3460. [6] P. Ernst, C. Geng, F. Scholz, H. Schweizer, Y. Zhang, A. Mascarehnas, Band-gap reduction and valence-band splitting of ordered GaInP2, Appl. Phys. Lett. 67 (1995) 2347–2349. [7] J.B. Lee, S.D. Kwon, I. Kim, Y.H. Cho, B.-D. Choe, The characteristics of an In0.5Ga0.5P and In0.5Ga0.5P/GaAs heterojunction grown on a (100) GaAs substrate by liquid-phase epitaxy, J. Appl. Phys. 71 (1992) 5016– 5021. [8] T. Prutskij, P. Dı´az-Arencibia, R.A. Brito-Orta, A. Mintairov, T. Kosel, J. Merz, Luminescence anisotropy of InGaP layers grown by liquid phase epitaxy, J. Phy. D: Appl. Phy. 37 (2004) 1563–1568.
