JOURNAL OF APPLIED PHYSICS

VOLUME 83, NUMBER 11

1 JUNE 1998

Effect of wet oxidized Alx Ga12 x As layer on the interdiffusion of InGaAs/GaAs quantum wells Joong-Seon Choe, Sang-Wan Ryu, and Byung-Doo Choe Department of Physics, Seoul National University, Seoul 151-742, Korea

H. Lim Department of Electronic Engineering, Ajou University, Suwon 442-749, Korea

~Received 20 October 1997; accepted for publication 12 February 1998! The effect of wet oxidized AlAs cap layer and AlGaAs interlayer on the thermal stability of In0.2Ga0.8As/GaAs quantum well ~QW! is studied. The QW interdiffusion rate is observed to increase with the Al composition of the Alx Ga12x As interlayer until x reaches about 0.5 and then saturate for x>0.5. When the oxidation is performed at 380 °C for 15 min, the threshold value of x for the enhancement of QW interdiffusion is found to be 0.3. It is also confirmed that the QW interdiffusion can only be explained when the strain effect in InGaAs is taken into account. © 1998 American Institute of Physics. @S0021-8979~98!02510-9#

I. INTRODUCTION

The oxidized AlGaAs layer was demonstrated to be useful as a diffusion mask against Zn to obtain selective impurity-induced layer disordering.10 However, to the best of our knowledge, its feasibility as a dielectric cap layer for the IFIE has not been studied. In this article, we report on the characteristics of Alx Ga12x As as the dielectric cap layer for the IFIE of In0.2Ga0.8As/GaAs QW structure. The existence of the Alx Ga12x As interlayer below the AlAs cap layer is observed to enhance the QW intermixing especially when x is larger than 0.5. This result indicates the IFIE behavior of the QW depends greatly on the structural properties of the oxidized-unoxidized Alx Ga12x As interface.

Recently, many researchers have been interested in the application of the wet oxidized Alx Ga12x As layer with relatively high Al composition for its beneficial optical and electrical characteristics. Low refractive index of the oxidized Alx Ga12x As layer about 1.6 makes it easy to form waveguide in Fabry-Perot laser diode1 and high reflectivity Bragg mirror in vertical cavity surface emitting laser diode.2 Its insulating property is also employed for the current blocking in laser diodes3 and for the gate insulator in III–V metalinsulator-semiconductor field effect transistor.4 On the other hand, selectively enhanced interdiffusion of quantum well ~QW! structures has drawn a great deal of attention because of its possible application for the selective modification of epitaxial structures needed in fabricating novel optoelectronic devices. To enhance the QW interdiffusion selectively, the desired region is ion implanted using materials such as Zn and Si,5 or is encapsulated by a dielectric layer such as SiO2 and Si3N4. 6,7 In the case of ionimplantation-enhanced QW or superlattice intermixing, the damage-induced disordering process is known to play a major role.5 But, the intermixing at the well/barrier can be enhanced by the result of heavy doping even if any implantation-induced damage is not present.8,9 Moreover, the implantation-induced damage can provide nonradiative recombination paths and thus can degrade the optical property of intermixed region. In the case of employing the dielectric cap layer, by contrast, the QW intermixing is enhanced due to the enhanced diffusion of column-III element vacancies from the dielectric into the semiconductor.6 In this impurityfree interdiffusion enhancement ~IFIE! technique using the dielectric cap layer, the degree of intermixing depends on the encapsulating dielectric material and its deposition conditions, as well as the thermal treatment conditions.6 This IFIE technique is therefore free from the influence of implantation-related damages and/or impurities and thus is very promising for the fabrication of novel devices. 0021-8979/98/83(11)/5779/4/$15.00

II. EXPERIMENTS

The In0.2Ga0.8As/GaAs single QW samples were prepared by atmospheric pressure metalorganic chemical vapor deposition on the GaAs substrate. After the growth of 3600Å-thick GaAs buffer layer, 75-Å-thick In0.2Ga0.8As well layer and 720-Å-thick GaAs barrier layer were grown successively. Then a 1600-Å-thick Alx Ga12x As layer and 360Å-thick AlAs top layer were grown in sequence as a capping layer. Each sample is designed to have Alx Ga12x As layer with different Al compositions of x50, 0.3, 0.4, 0.5, 0.6, and 0.8, respectively. In the following, we will denote each sample as GaAs, Al0.3Ga0.7As, or Al0.4Ga0.6As interlayer sample, etc. For the comparison with GaAs interlayer sample, bare In0.2Ga0.8As/GaAs sample ~referred to as bare sample hereafter! without AlAs layer was also prepared. To avoid the formation of poor quality native oxide layer in atmosphere at the AlAs capping layer, wet thermal oxidation was performed at 380 °C for 15 min right after each sample growth. The details of the wet thermal oxidation procedures were reported elsewhere.1,2,11 After the oxidation, the QW interdiffusion was accomplished by a rapid thermal annealing ~RTA! at 900 °C for 15–120 s under the N2 gas ambient. During the RTA process, a piece of GaAs substrate was placed on the sample as a proximity cap supplying As overpressure. Photoluminescence ~PL! measurements were per5779

© 1998 American Institute of Physics

5780

Choe et al.

J. Appl. Phys., Vol. 83, No. 11, 1 June 1998

FIG. 1. PL spectrum of bare, GaAs, Al0.5Ga0.5As samples after RTA of 60 s. Compared with as-grown sample, each shows the blueshift of 10, 23, and 66 meV, respectively.

formed at 11 K before and after RTA to determine the variation of QW transition energy induced by the intermixing. The shape of QW is then simulated so as to give the measured PL peak energy. The diffusivity is finally obtained from the fitting parameters used in the simulation. III. RESULTS AND DISCUSSION

Figure 1 shows typical PL peaks obtained before and after an RTA of 60 s. One can notice that the GaAs interlayer sample @with Al~oxide! cap layer# shows larger blueshift than the bare sample @without Al~oxide! cap layer# by 13 meV after an identical RTA process. This result indicates that the Al~oxide! layer enhances the QW interdiffusion, though the effect is not so large. The Al0.5Ga0.5As interlayer sample shows even larger blueshift after the RTA compared with the GaAs interlayer sample. The intermixed QW through IFIE often shows a PL linewidth broadening due to the introduced defects and interface roughness.12 When silicon nitride is used as a cap layer in the IFIE process, even the reduction of carrier lifetime in addition to the PL peak broadening, is observed in Al0.2Ga0.8As/GaAs QWs.7 As can be seen in this figure, the PL peak broadening is not observable when a wet oxidized AlAs cap layer is used in the IFIE process. Therefore, the Al~oxide! layer formed by a wet oxidation of AlAs might be one of the best cap layer for the IFIE application. The time evolution of the energy shift of the QW PL peaks is plotted in Fig. 2. One can notice that the energy shift of a QW PL peak increases with the RTA time. The blueshift is rapid in the initial stage of the RTA, and then slows down after an RTA longer than about 0.5 min. This is due to the strain-enhanced interdiffusion as will be shown later. This figure shows also that the blueshift of QW PL peak depends on the Al composition in Alx Ga12x As interlayer. The energy shift of the Al0.3Ga0.7As interlayer sample is nearly the same as that of GaAs interlayer sample. The Alx Ga12x As interlayer samples with high Al composition (x>0.5) shows, on the contrary, a significant energy shift. After an RTA of 2

FIG. 2. The energy of QW PL peak after RTA. Samples are bare QW~n!, and QWs with GaAs~.!, Al0.3Ga0.7As ~m!, Al0.4Ga0.6As ~,!, Al0.5Ga0.5As ~j!, Al0.6Ga0.4As ~h!, and Al0.8Ga0.2As ~s!, respectively. As Al composition in AlGaAs interlayer increases, the shift of QW transition energy increases after RTA.

min, the amount of blueshift is about two times larger compared with the GaAs interlayer sample. The Alx Ga12x As interlayer samples with Al composition larger than 0.5 show practically no difference in energy shift, irrespective of the value of x. Al0.4Ga0.6As interlayer sample shows an intermediate energy shift value. These results indicate that the Alx Ga12x As interlayer with high Al composition, as well as the AlAs capping layer, enhances the interdiffusion of In0.2Ga0.8As/GaAs QW. The change of QW shape after the RTA is simulated to give the observed PL peak shift using the following Fick’s equation:

S D

]C ] ]C 5 D . ]t ]z ]z

~1!

FIG. 3. Comparison of two analysis methods: strain effect is not taken into account for dashed curves but included for solid curve. When the strain is not taken into account, the overall behavior of energy shift cannot be fitted well. But the measured data are well fitted when the strain effect is taken into account.

Choe et al.

J. Appl. Phys., Vol. 83, No. 11, 1 June 1998

5781

TABLE I. Diffusivity parameters obtained from the fitting of QW PL peak shift. Al composition D 0 (310217 cm2/s) k

bare

0

0.3

0.4

0.5

0.6

0.8

0.56 19.0

1.1 20.1

1.5 20.2

4.3 20.2

5.7 20.8

5.7 20.8

5.9 20.6

Here C and D denote the indium concentration and the diffusivity, respectively. Figure 3 shows the measured PL peak shift with three fitting curves obtained by different way for the Al0.6Ga0.4As interlayer sample. Two dashed curves are obtained assuming that the diffusivity is constant at a given temperature. Here the employed diffusivity values are 1.5 310215 and 6.3310216 cm2/s, respectively. The strain effect on the diffusivity is taken into account for the solid curve so that the diffusivity depends on the In composition of the QW ~initially 0.2 in this case! as13 D5D 0 exp~ k C ! .

~2!

Here k is a parameter indicating the degree of strain enhancement on the diffusivity. This result shows that the QW intermixing can only be well explained when the straininduced intermixing enhancement is properly taken into account in our IFIE experiment. The values of D 0 and k thus obtained are tabulated in Table I for all the samples employed. The values of D are also plotted in Fig. 4 as a function of Al composition in Alx Ga12x As. As can be seen in Table I, the value of k is nearly constant for all the samples with 2061, which is expected result since k is a property of In0.2Ga0.8As layer itself.13 But the values of D 0 increases drastically at x50.3– 0.5 and then saturates for x>0.5. The saturated diffusivity is larger than that of the bare sample by more than one order. In the wet thermal oxidation process, the Alx Ga12x As layer with high Al composition has a selectivity over a GaAs or Alx Ga12x As layer with low Al composition since the activation energy for the oxidation in-

FIG. 4. Diffusivity of In0.2Ga0.8As/GaAs QW intermixing as a function of Al composition x. Around x50.4, diffusivity increases steeply as Al composition increases. It saturates at about x50.5 and then additional increase in diffusivity is not observable.

creases as the Al composition is decreased.11 This selective oxidation would be prominent in our relatively low temperature oxidation process. Thus the Alx Ga12x As interlayer would not be oxidized for x<0.3 samples.14 In the case of Alx Ga12x As interlayer samples with x>0.5, the electron microscope observation revealed that some region of AlGaAs interlayer near the interface is oxidized. Therefore, te difference of QW intermixing between the bare sample and the Alx Ga12x As interlayer samples with x<0.3 should be mostly due to the role of Al~oxide! cap layer as IFIE provider. And the cause of difference in QW intermixing between the Alx Ga12x As interlayer samples with x <0.3 and x>0.5 should be sought from the role of partially oxidized Alx Ga12x As interlayers in supplying the cation vacancies. After the wet oxidation of AlAs, a great amount of voids are known to exist along the Al~oxide!-Alx Ga12x As (x<0.3! interface.14 This porous nature of the interface results from the high stress load which develops across the interface as a result of volume shrinkage due to the oxidation of AlAs.14 Such a porosity would not be effective to supply the cation vacancies since the cation vacancies should be diffused directly from the dielectrics. Therefore, for the Alx Ga12x As interlayer samples with x<0.3, the diffusivity of the QW intermixing would not differ so much from that of the bare sample due to the limited supply of the cation vacancies across the porous interface ~see Fig. 4!. When some region of AlGaAs interlayer near the AlAs-AlGaAs interface is oxidized, the AlGa ~oxide! itself would supply cation vacancies. Moreover, the transition region between the AlGa ~oxide! and AlGaAs should not contain any pore since the transition region is formed at the oxidation front and the shrinkage of the AlGaAs volume in the course of oxidation must appear at the Al~oxide!-AlGa ~oxide! interface. Therefore, the observed difference of the diffusivity between the low Al composition and the high Al composition interlayer samples should be due to the structural difference between Al ~oxide!-Alx Ga12x As (x<0.3) and Alx Ga12x ~oxide!Alx Ga12x As (x>0.5) interfaces. When the Alx Ga12x As interlayer with x>0.5 is completely oxidized @ i.e., when the AlGa ~oxide!-GaAs interface is porous #, we observed that the diffusivity of QW intermixing is greatly reduced compared with the partially oxidized AlGaAs interlayer sample. This observation indicates that our conjecture about the role of interface voids in reducing the cation vacancy diffusion is correct. Under the oxidation process employed in this work, the AlGa ~oxide! layer would not be thick enough for the Alx Ga12x As interlayer samples with 0.3
5782

IV. CONCLUSION

We have studied the effect of Al ~oxide! and Alx Ga12x ~oxide! layers on the interdiffusion of In0.2Ga0.8As/GaAs QWs. When x is larger than 0.5, the QW intermixing is greatly enhanced after the RTA. This large enhancement is attributed to the IFIE action of the AlGa ~oxide! interlayer. When x is smaller than 0.3, the diffusion enhancement effect is rather small and the IFIE is believed to be provided by the Al ~oxide! cap layer. The wet oxidized AlAs and the Alx Ga12x As ~especially with x>0.5! layers are believed to be a good candidate as a cap layer for the IFIE application since these layers do not degrade the optical property of intermixed QW. ACKNOWLEDGMENTS

Parts of this work were supported by the BSRI program of the Korea Ministry of Education and the KOSEF through the SPRC of Jeonbuk National University. 1

Choe et al.

J. Appl. Phys., Vol. 83, No. 11, 1 June 1998

F. A. Kish, S. J. Caracci, N. Holonyak, Jr., P. Gavrilovic, K. Meehan, and J. E. Williams, Appl. Phys. Lett. 60, 71 ~1992!.

2

A. L. Holmes, M. R. Islam, R. V. Chelakara, F. J. Ciuba, R. D. Dupuis, M. J. Ries, E. I. Chen, S. A. Maranowski, and N. Holonyak, Jr., Appl. Phys. Lett. 66, 2831 ~1995!. 3 G. M. Yang, M. H. MacDougal, and P. D. Dapkus, Electron. Lett. 31, 886 ~1995!. 4 E. I. Chen, N. Holonyak, Jr., and S. A. Maranowski, Appl. Phys. Lett. 66, 2688 ~1995!. 5 P. Gavrilovic, D. G. Deppe, K. Meehan, N. Holonyak, Jr., J. J. Coleman, and R. D. Burnham, Appl. Phys. Lett. 47, 130 ~1985!. 6 L. J. Guido, N. Holonyak, Jr., K. C. Hsieh, R. W. Kaliski, W. E. Plano, R. D. Burnham, R. L. Thornton, J. E. Epler, and T. L. Paoli, J. Appl. Phys. 61, 1372 ~1987!. 7 W. J. Choi, S. Lee, Y. Kim, D. Woo, S. K. Kim, S. H. Kim, J. I. Lee, K. N. Kang, J. H. Chu, S. K. Yu, J. C. Seo, D. Kim, and K. Cho, Appl. Phys. Lett. 67, 3438 ~1995!. 8 W. Walukiewicz, Phys. Rev. B 50, 5221 ~1994!. 9 K. Ogawa and M. Kawabe, Jpn. J. Appl. Phys., Part 1 29, 1240 ~1990!. 10 J. M. Dallesasse, N. Holonyak, Jr., N. El-Zein, T. A. Richard, F. A. Kish, A. R. Sugg, R. D. Burnham, and S. C. Smith, Appl. Phys. Lett. 58, 974 ~1991!. 11 H. Nickel, J. Appl. Phys. 78, 5201 ~1995!. 12 K. Rammohan, D. H. Rich, M. H. MacDougal, and P. D. Dapkus, Appl. Phys. Lett. 70, 1599 ~1997!. 13 S.-W. Ryu, I. Kim, and B.-D. Choe, Appl. Phys. Lett. 67, 1417 ~1995!. 14 S. Guha, F. Agahi, B. Pezeshki, J. A. Kash, D. W. Kisker, and N. A. Bojarczuk, Appl. Phys. Lett. 68, 906 ~1996!.