- Email: [email protected]
GaAs pseudomorphic quantum wells — growth and thermal stability
Surface Sclencc 22X (1990) ,340~ 341 North-Holland
In_,Ga, _ ,As/ - GROWTH
GaAs PSEUDOMORPHIC QUANTUM AND THERMAL STABILITY
H. NICKEL,
R. LijSCH.
W. SCHLAPP
~-or.vc~hun,~,s,tl.vrrtut der DBP, POB 5000. D 6100 Durn~wulr,
H. LEIER
Recr~\ed
Fed. Rep oj ticnmr~r~~
and A. FORCHEL
6 June 1989; accepted for puhhcation 2X September 1989
We have optimized linewidths
WELLS
the MBE
growth conditions
of InGaAs/GaAs
and high quantum efficiencies. The thermal stability
pseudomorphic
\tructurca In order to ohtan
narrou
cm~svon
of the strained layers is studied using annealing at 7’ c 930°C’.
the evaluation of the luminescence spectra after 30 min annealing wc estimate an In/<&
interdiffusion
From
length of ahout 2 nm .It
9000(‘.
Pscudomorphic semiconductor structures as e.g. ItiGaAs/GaAs have recently become of particular interest. These heterostructures are grown from lattice mismatched materials and allow to tailor e.g. the properties of electronic and optoelectronic devices with much greater flexibility than lattice matched systems. We have grown In,Ga, ,As/GaAs quantum well? (QWs) on undoped (100) oriented GaAs substrates (EPD = 1000 cm-‘) in a Varian MBE 360 system with a rotating substrate holder. InGaAs wells of 30 nm (except for In - contents .Y> 0.12). 10 nm. 5 nm and 2 nm thickness were grown at 520 0 C with the thinnest well uppermost. The wells were cladded by 200 nm GaAs barriers grown at 600” C. The growth temperature was ramped in the GaAs regions within a distance of about 28nm from the interfaces. The increase of the growth temperature within the GaAs barrier layers results in an improvement of the PL linewidths by a factor of 2. For the growth of the InGaAs layers with smooth surfaces the beam equivalent pressure ratio of group V (Asa) versus group III elements had to be increased compared to growth conditions used for conventional GaAs/GaAlAs layers. We used a oo~Y-~o2x/9o/%o~.so (N~)rth-Holland)
”
Elsevier
Saence Pubhshers
1i.V
pressure ratio of - 40. InGaAs quantum wells with In contents of 6%. 12% and 20% were grown. The indium content was determined from separate samples by photoluminescence (PL). ion backscattering (RBS), X-ray analysis (WDX. XRD) and secondary ion mass spectrometry (SIMS) for thick films (l-2 pm) and by PL, XRD and SIMS (calibrated on thick films) for thin layers (- 40 nm). The low temperature (e-hh) luminescence cnergy versus In-content of fully relaxed bulk In ,Ga, , As layers is [1.2]: E”‘(eV)
= 1 Sl5
()I.\-~0.3,
- 1.584x + 0.475.~‘. Ti
(1)
10 K.
A
similar relation for elastically strained In, G>i, ,As bulk films has been derived by adding the strain induced energy shift as outlined in ref. [3]: f?(eV)
= 1.515 - 1.144.u + o.255sJ.
01.u
TI
(2)
10 K.
Fig. 1 shows that both equations fit our experimental values well, which are displayed by horizontal bars. Ey. (2) was used in combination with a conduction to valence band offset ratio of 65 : 35
H. Nickel et ai. / In,Ga I _ .~As / GaAs pseudomorphic quantum wells
Fig. 1. Low temperature (T 1: 10 K) e-hh photoluminescence energy versus indium content of relaxed and strained In,Ga, _,YAs layers on GaAs. The lines labeled “relaxed” and “strained” refer to eqs. (1) and (2), respectively. Experimental results are marked as horizontat bars.
for the evaluation of the PL data of the QW layers. Fig. 2 displays a photolu~nescence spectrum of a sample with an In-content of 6% which contains a set of single quantum wells with thicknesses between 30 and 2 nm. The emission of the 2 nm well occurs at 1.508 eV, i.e., the lowest subband edge is shifted almost entirely up to the GaAs barrier level. The quantum wells give rise to very strong and sharp emission lines. We observe a systematic variation of the emission linewidth with varying In-content and well width. Fig. 3 displays the well width dependence of the emission linewidth for QWs with L, ranging from 1.3
1
.o
In,oeGa,94As/‘GaAs
i
I Lz=30nm
1 46
I .48 energy
i 50 id/)
Fig. 2. Emission spectrum of an Jna.,Ga,,,,As/GaAs sample recorded at ‘7”=52 K (Kr-laser excitation). The emission intensities of the different QWs have been multiplied by the factors indicated.
341
Fig. 3. Well width dependence of the luminescence linewidth at 7= 2 K for InGaAs/GaAs QWs with In contents of 6% (squares), 12% (circles) and 20% (triangles). Dashed lines are guides for the eye only.
to 30 nm for samples with 6% In content (squares), 12% In content (circles) and 20% In content (triangles). For decreasing L, the half width of the emission goes through a maximum at about 2-5 nm. For very thin and very wide quantum wells we observe rather narrow linewidths (< 0.4 meV for Xl” = 0.06). The L, variation of the linewidth can be explained by small well width fluctuations. Using the finite potential well model including the strain in the pseudomo~~c structure we calculate the derivative of the subband energy with respect to the well widths dE/dL,. The corresponding energy variation for an interface roughness of one monolayer amounts to 1.5,3 and 5 meV for L, = 5 nm and In = contents of 6%, 12% and 20%, respectively. The experimental values are significantly lower than these estimates. This indicates a high quality of the interfaces under the present growth conditions. Partial relaxation in strained In,YGa, ~ ,As/ GaAs QWs is generally accompanied in PL measurements by a low energetic satellite to the excitonic line, because relaxed areas have a lower band gap [4]. We have grown strained In,Ga, __s As/GaAs QWs of 6% and 12% In with widths up to 30 nm and of 20% In with widths up to 10 nm,
742
1 420
1415 fl
ICI ‘jY
(e’dj
14IO
I c40
I 435 i:r1erqy
(c’d)
Fig. 4. Emiwon spectra of an ln,,,Ga,,,,Aa/GaA~ QW (I,, = 10 nm) after annealing at different temperatures.
all showing intense and very sharp emission lines with no sign of satellites, i.e. no sign of relaxation. However. as the absence of satellite lines in PL is not sufficient to prove the absence of relaxation, we have investigated our layer structures by annealing. In the presence of metastability or at the onset of relaxation a thermal treatment should cause a significant relaxation [5,6] and hence give rise to a significant change of the PL spectrum. For the annealing a SIN, cap was sputter deposited on the samples. The samples were annealed in evacuated quartz ampoules in a conventional thermal annealing system at temperatures between 500 and 930 o C (annealing time: 30 min). After annealing the cap layer was removed by buffered HF. As a typical example, fig. 4 shows luminescence spectra of an In,,.,,Ga,,,,As/GaAs QW (L,, = 10 nm) after thermal processing at different temperatures. The pseudomorphic structure displays a high thermal stability. For annealing temperatures below 800°C the emission spectra coincide with the emission of the as grown layer. At 800” C we observe the onset of a small shift of the emission to higher energy, which increases as the annealing temperature is raised to 8.50” C. For annealing at 900 o C a very strong shift (21 meV) occurs. Fig. 5 displays the energy shift of the emission lines of In,,zGa,,,As/GaAs QWs after annealing at various temperatures up to 930” C. Above 800 o C all lines shift to higher energies. The mag-
Fig. 5. bergy shift of the PL emision lines of In,,LGa,,,As/ GaAa QWs (L, = 2. 5 , 10 nm) versus annealing temperature (30 min annealing time).
nitude of the shift increases with increasing L,. due to the finite potential barrier, and ranges after annealing at 930 o C from 25 meV (L, = 2 nm) to 100 meV (L, = 10 nm). The shift of the emission lines to higher energy generally can be due to two effects. Firstly, due to the exchange of Ga and In at the interface between well and barrier the effective width of the quantum well is reduced in the energy range of the lowest subbands. Secondly, the interdiffusion may increase the Ga-content at the center of the well. The latter effect is important only if the interdiffusion length is comparable to the quantum well thickness.
r 4
,
*,
,.
”
.?
Fig. 6. Annealing temperature dependence of emission intensity (open circles) and luminescence halfwidth (full circles) at 2 As/GaAs QW with L, = 5 nm. Dashed K for an In,,,,,Ga,,, lines are guides for the eye only.
H. Nickel et al. / In ,Ga ,
1As / GaAs pseudomorphic quantum wells
We have used the energetic shift of the quantum well emission after annealing to estimate the In/Ga interdiffusion length. For annealing temperatures I 900’ C the energetic shifts due to annealing are small compared to the depth of the potential well in the as grown structure and we consider only a change of the effective QW width. Due to the thick barrier layers of our structures and the weak changes of the width, we further assume that eq. (2) is valid prior and after annealing. For the evaluation we determine the equivalent rectangular QW width, which accounts for the energetic shift of the emission at the different annealing temperatures. Within this approximation the width of the 10 nm QW is reduced after annealing at 900 o C by 2.6 nm for x,” = 0.06 and interdiffusion 3.2 nm for x,,, = 0.12. The In/Ga lengths amount to approximately 50% of these values. The corresponding In/Ga interdiffusion coefficients are in the order of 1 X lo-l7cm2/s. The values estimated here for elastically strained InGaAs layers are comparable with the lowest data reported for the Al/Ga interdiffusion in GaAs/GaAlAs QWs [7]. The present evaluation indicates an increase of the In/Ga interdiffusion coefficient with the In content. The high thermal stability of the pseudomorphic layers results in comparatively weak changes of the emission intensity and the emission line width due to the annealing. As shown in fig. 6 for QW with L, = 5 nm the an In O.OhGao.94As/GaAs linewidth of the as-grown structure of 0.75 meV is essentially maintained up to annealing temperatures of 800°C. Even for an annealing temperature of 900°C the linewidth does not exceed 1 meV. The emission intensity is slightly more sensitive to the annealing conditions. Compared to the as grown structure the emission intensity drops by about 30% and 60% for annealing at 800” C and 900 o C, respectively (annealing time 30 min). All strained layer structures investigated above are highly stable against relaxation processes, as shown by our annealing studies. We would like to point out that this implies significantly higher
343
critical thicknesses than expected according to the Matthews-Blakeslee “single kink” dislocation mechanism [8]. For example, the sample with x,,, = 0.12 containing a 30 nm InGaAs strained layer exceeds the critical value by a factor of about 2, even if the thick GaAs barrier layers are taken into account [5]. In summary, we have grown pseudomorphic quantum wells with narrow In,Gai x As/GaAs emission linewidths and high quantum efficiencies. Particularly small linewidths are observed for very narrow (- 2 nm) and broad QWs. This indicates that the PL linewidth in our structures is mainly due to small well width fluctuations. Our studies of the thermal stability indicate only small interdiffusion effects in the pseudomorphic layers for T< 900°C. We are grateful to J. Hommel for experimental assistance and to A. Packer, K. Miethe, W. Koschig and H. Schwinn for the measurements of the In content. The financial support of the work at Stuttgart University by the Deutsche Forschungsgemeinschaft is gratefully acknowledged.
References 111G. Ji, U.K. Reddy,
D. Huang, TX Henderson and H. Morkoc, Superlattices Microstruct. 3 (1987) 539. 121K.H. Goetz, D. Bimberg, H. Jiirgensen, J. Selders, A.V. Solomonov. G.F. Glinskii and M. Razeghi, J. Appl. Phys. 54 (1983) 4543. Z.G. Chen, V.D. Kulakovskii. A. Uddin [31 T.G. Andersson, and J.T. Vallin, Phys. Rev. B 37 (1988) 4032. 141 J.P. Reithmaier. H. Cerva and R. Lijsch, Appl. Phys. Lett. 54 (1989) 48. [51 J.Y. Tsao and B.W. Dodson, AppI. Phys. Left. 53 (19X8) 848; B.W. Dodson and J.Y. Tsao. Appl. Phys. Lett. 51 (1988) 1325. [61 T.E. Zipperian, E.D. Jones, B.W. Dodson, J.F. Klem and P.L. Gourley, IEEE GaAs IC Symposium 1988, Technical Digest, p. 251. [71 H. Leier, H. Rothfritz, A. Forchel and G. Weimann, J. Cryst. Growth 95 (1989) 277. PI J.W. Matthews and A.E. Blakeslee. J. Cryst. Growth 27 (1974) 118.
In_,Ga, _ ,As/ - GROWTH
GaAs PSEUDOMORPHIC QUANTUM AND THERMAL STABILITY
H. NICKEL,
R. LijSCH.
W. SCHLAPP
~-or.vc~hun,~,s,tl.vrrtut der DBP, POB 5000. D 6100 Durn~wulr,
H. LEIER
Recr~\ed
Fed. Rep oj ticnmr~r~~
and A. FORCHEL
6 June 1989; accepted for puhhcation 2X September 1989
We have optimized linewidths
WELLS
the MBE
growth conditions
of InGaAs/GaAs
and high quantum efficiencies. The thermal stability
pseudomorphic
\tructurca In order to ohtan
narrou
cm~svon
of the strained layers is studied using annealing at 7’ c 930°C’.
the evaluation of the luminescence spectra after 30 min annealing wc estimate an In/<&
interdiffusion
From
length of ahout 2 nm .It
9000(‘.
Pscudomorphic semiconductor structures as e.g. ItiGaAs/GaAs have recently become of particular interest. These heterostructures are grown from lattice mismatched materials and allow to tailor e.g. the properties of electronic and optoelectronic devices with much greater flexibility than lattice matched systems. We have grown In,Ga, ,As/GaAs quantum well? (QWs) on undoped (100) oriented GaAs substrates (EPD = 1000 cm-‘) in a Varian MBE 360 system with a rotating substrate holder. InGaAs wells of 30 nm (except for In - contents .Y> 0.12). 10 nm. 5 nm and 2 nm thickness were grown at 520 0 C with the thinnest well uppermost. The wells were cladded by 200 nm GaAs barriers grown at 600” C. The growth temperature was ramped in the GaAs regions within a distance of about 28nm from the interfaces. The increase of the growth temperature within the GaAs barrier layers results in an improvement of the PL linewidths by a factor of 2. For the growth of the InGaAs layers with smooth surfaces the beam equivalent pressure ratio of group V (Asa) versus group III elements had to be increased compared to growth conditions used for conventional GaAs/GaAlAs layers. We used a oo~Y-~o2x/9o/%o~.so (N~)rth-Holland)
”
Elsevier
Saence Pubhshers
1i.V
pressure ratio of - 40. InGaAs quantum wells with In contents of 6%. 12% and 20% were grown. The indium content was determined from separate samples by photoluminescence (PL). ion backscattering (RBS), X-ray analysis (WDX. XRD) and secondary ion mass spectrometry (SIMS) for thick films (l-2 pm) and by PL, XRD and SIMS (calibrated on thick films) for thin layers (- 40 nm). The low temperature (e-hh) luminescence cnergy versus In-content of fully relaxed bulk In ,Ga, , As layers is [1.2]: E”‘(eV)
= 1 Sl5
()I.\-~0.3,
- 1.584x + 0.475.~‘. Ti
(1)
10 K.
A
similar relation for elastically strained In, G>i, ,As bulk films has been derived by adding the strain induced energy shift as outlined in ref. [3]: f?(eV)
= 1.515 - 1.144.u + o.255sJ.
01.u
TI
(2)
10 K.
Fig. 1 shows that both equations fit our experimental values well, which are displayed by horizontal bars. Ey. (2) was used in combination with a conduction to valence band offset ratio of 65 : 35
H. Nickel et ai. / In,Ga I _ .~As / GaAs pseudomorphic quantum wells
Fig. 1. Low temperature (T 1: 10 K) e-hh photoluminescence energy versus indium content of relaxed and strained In,Ga, _,YAs layers on GaAs. The lines labeled “relaxed” and “strained” refer to eqs. (1) and (2), respectively. Experimental results are marked as horizontat bars.
for the evaluation of the PL data of the QW layers. Fig. 2 displays a photolu~nescence spectrum of a sample with an In-content of 6% which contains a set of single quantum wells with thicknesses between 30 and 2 nm. The emission of the 2 nm well occurs at 1.508 eV, i.e., the lowest subband edge is shifted almost entirely up to the GaAs barrier level. The quantum wells give rise to very strong and sharp emission lines. We observe a systematic variation of the emission linewidth with varying In-content and well width. Fig. 3 displays the well width dependence of the emission linewidth for QWs with L, ranging from 1.3
1
.o
In,oeGa,94As/‘GaAs
i
I Lz=30nm
1 46
I .48 energy
i 50 id/)
Fig. 2. Emission spectrum of an Jna.,Ga,,,,As/GaAs sample recorded at ‘7”=52 K (Kr-laser excitation). The emission intensities of the different QWs have been multiplied by the factors indicated.
341
Fig. 3. Well width dependence of the luminescence linewidth at 7= 2 K for InGaAs/GaAs QWs with In contents of 6% (squares), 12% (circles) and 20% (triangles). Dashed lines are guides for the eye only.
to 30 nm for samples with 6% In content (squares), 12% In content (circles) and 20% In content (triangles). For decreasing L, the half width of the emission goes through a maximum at about 2-5 nm. For very thin and very wide quantum wells we observe rather narrow linewidths (< 0.4 meV for Xl” = 0.06). The L, variation of the linewidth can be explained by small well width fluctuations. Using the finite potential well model including the strain in the pseudomo~~c structure we calculate the derivative of the subband energy with respect to the well widths dE/dL,. The corresponding energy variation for an interface roughness of one monolayer amounts to 1.5,3 and 5 meV for L, = 5 nm and In = contents of 6%, 12% and 20%, respectively. The experimental values are significantly lower than these estimates. This indicates a high quality of the interfaces under the present growth conditions. Partial relaxation in strained In,YGa, ~ ,As/ GaAs QWs is generally accompanied in PL measurements by a low energetic satellite to the excitonic line, because relaxed areas have a lower band gap [4]. We have grown strained In,Ga, __s As/GaAs QWs of 6% and 12% In with widths up to 30 nm and of 20% In with widths up to 10 nm,
742
1 420
1415 fl
ICI ‘jY
(e’dj
14IO
I c40
I 435 i:r1erqy
(c’d)
Fig. 4. Emiwon spectra of an ln,,,Ga,,,,Aa/GaA~ QW (I,, = 10 nm) after annealing at different temperatures.
all showing intense and very sharp emission lines with no sign of satellites, i.e. no sign of relaxation. However. as the absence of satellite lines in PL is not sufficient to prove the absence of relaxation, we have investigated our layer structures by annealing. In the presence of metastability or at the onset of relaxation a thermal treatment should cause a significant relaxation [5,6] and hence give rise to a significant change of the PL spectrum. For the annealing a SIN, cap was sputter deposited on the samples. The samples were annealed in evacuated quartz ampoules in a conventional thermal annealing system at temperatures between 500 and 930 o C (annealing time: 30 min). After annealing the cap layer was removed by buffered HF. As a typical example, fig. 4 shows luminescence spectra of an In,,.,,Ga,,,,As/GaAs QW (L,, = 10 nm) after thermal processing at different temperatures. The pseudomorphic structure displays a high thermal stability. For annealing temperatures below 800°C the emission spectra coincide with the emission of the as grown layer. At 800” C we observe the onset of a small shift of the emission to higher energy, which increases as the annealing temperature is raised to 8.50” C. For annealing at 900 o C a very strong shift (21 meV) occurs. Fig. 5 displays the energy shift of the emission lines of In,,zGa,,,As/GaAs QWs after annealing at various temperatures up to 930” C. Above 800 o C all lines shift to higher energies. The mag-
Fig. 5. bergy shift of the PL emision lines of In,,LGa,,,As/ GaAa QWs (L, = 2. 5 , 10 nm) versus annealing temperature (30 min annealing time).
nitude of the shift increases with increasing L,. due to the finite potential barrier, and ranges after annealing at 930 o C from 25 meV (L, = 2 nm) to 100 meV (L, = 10 nm). The shift of the emission lines to higher energy generally can be due to two effects. Firstly, due to the exchange of Ga and In at the interface between well and barrier the effective width of the quantum well is reduced in the energy range of the lowest subbands. Secondly, the interdiffusion may increase the Ga-content at the center of the well. The latter effect is important only if the interdiffusion length is comparable to the quantum well thickness.
r 4
,
*,
,.
”
.?
Fig. 6. Annealing temperature dependence of emission intensity (open circles) and luminescence halfwidth (full circles) at 2 As/GaAs QW with L, = 5 nm. Dashed K for an In,,,,,Ga,,, lines are guides for the eye only.
H. Nickel et al. / In ,Ga ,
1As / GaAs pseudomorphic quantum wells
We have used the energetic shift of the quantum well emission after annealing to estimate the In/Ga interdiffusion length. For annealing temperatures I 900’ C the energetic shifts due to annealing are small compared to the depth of the potential well in the as grown structure and we consider only a change of the effective QW width. Due to the thick barrier layers of our structures and the weak changes of the width, we further assume that eq. (2) is valid prior and after annealing. For the evaluation we determine the equivalent rectangular QW width, which accounts for the energetic shift of the emission at the different annealing temperatures. Within this approximation the width of the 10 nm QW is reduced after annealing at 900 o C by 2.6 nm for x,” = 0.06 and interdiffusion 3.2 nm for x,,, = 0.12. The In/Ga lengths amount to approximately 50% of these values. The corresponding In/Ga interdiffusion coefficients are in the order of 1 X lo-l7cm2/s. The values estimated here for elastically strained InGaAs layers are comparable with the lowest data reported for the Al/Ga interdiffusion in GaAs/GaAlAs QWs [7]. The present evaluation indicates an increase of the In/Ga interdiffusion coefficient with the In content. The high thermal stability of the pseudomorphic layers results in comparatively weak changes of the emission intensity and the emission line width due to the annealing. As shown in fig. 6 for QW with L, = 5 nm the an In O.OhGao.94As/GaAs linewidth of the as-grown structure of 0.75 meV is essentially maintained up to annealing temperatures of 800°C. Even for an annealing temperature of 900°C the linewidth does not exceed 1 meV. The emission intensity is slightly more sensitive to the annealing conditions. Compared to the as grown structure the emission intensity drops by about 30% and 60% for annealing at 800” C and 900 o C, respectively (annealing time 30 min). All strained layer structures investigated above are highly stable against relaxation processes, as shown by our annealing studies. We would like to point out that this implies significantly higher
343
critical thicknesses than expected according to the Matthews-Blakeslee “single kink” dislocation mechanism [8]. For example, the sample with x,,, = 0.12 containing a 30 nm InGaAs strained layer exceeds the critical value by a factor of about 2, even if the thick GaAs barrier layers are taken into account [5]. In summary, we have grown pseudomorphic quantum wells with narrow In,Gai x As/GaAs emission linewidths and high quantum efficiencies. Particularly small linewidths are observed for very narrow (- 2 nm) and broad QWs. This indicates that the PL linewidth in our structures is mainly due to small well width fluctuations. Our studies of the thermal stability indicate only small interdiffusion effects in the pseudomorphic layers for T< 900°C. We are grateful to J. Hommel for experimental assistance and to A. Packer, K. Miethe, W. Koschig and H. Schwinn for the measurements of the In content. The financial support of the work at Stuttgart University by the Deutsche Forschungsgemeinschaft is gratefully acknowledged.
References 111G. Ji, U.K. Reddy,
D. Huang, TX Henderson and H. Morkoc, Superlattices Microstruct. 3 (1987) 539. 121K.H. Goetz, D. Bimberg, H. Jiirgensen, J. Selders, A.V. Solomonov. G.F. Glinskii and M. Razeghi, J. Appl. Phys. 54 (1983) 4543. Z.G. Chen, V.D. Kulakovskii. A. Uddin [31 T.G. Andersson, and J.T. Vallin, Phys. Rev. B 37 (1988) 4032. 141 J.P. Reithmaier. H. Cerva and R. Lijsch, Appl. Phys. Lett. 54 (1989) 48. [51 J.Y. Tsao and B.W. Dodson, AppI. Phys. Left. 53 (19X8) 848; B.W. Dodson and J.Y. Tsao. Appl. Phys. Lett. 51 (1988) 1325. [61 T.E. Zipperian, E.D. Jones, B.W. Dodson, J.F. Klem and P.L. Gourley, IEEE GaAs IC Symposium 1988, Technical Digest, p. 251. [71 H. Leier, H. Rothfritz, A. Forchel and G. Weimann, J. Cryst. Growth 95 (1989) 277. PI J.W. Matthews and A.E. Blakeslee. J. Cryst. Growth 27 (1974) 118.