How do InAs quantum dots relax when the InAs growth thickness exceeds the
dislocation-induced critical thickness?
J. F. Chen, Y. C. Lin, C. H. Chiang, Ross C. C. Chen, Y. F. Chen, Y. H. Wu, and L. Chang
Citation: Journal of Applied Physics 111, 013709 (2012); doi: 10.1063/1.3675519 View online: http://dx.doi.org/10.1063/1.3675519
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/111/1?ver=pdfcov Published by the AIP Publishing
Articles you may be interested in
Piezoelectric InAs (211)B quantum dots grown by molecular beam epitaxy: Structural and optical properties J. Appl. Phys. 108, 103525 (2010); 10.1063/1.3510490
Presentation and experimental validation of a model for the effect of thermal annealing on the photoluminescence of self-assembled InAs/GaAs quantum dots
J. Appl. Phys. 107, 123107 (2010); 10.1063/1.3431388
Annealing of self-assembled InAs/GaAs quantum dots: A stabilizing effect of beryllium doping Appl. Phys. Lett. 94, 072105 (2009); 10.1063/1.3086298
Dislocation-induced spatial ordering of InAs quantum dots: Effects on optical properties J. Appl. Phys. 91, 5826 (2002); 10.1063/1.1467963
Strain-induced material intermixing of InAs quantum dots in GaAs Appl. Phys. Lett. 77, 1789 (2000); 10.1063/1.1311314
How do InAs quantum dots relax when the InAs growth thickness exceeds
the dislocation-induced critical thickness?
J. F. Chen,1,a)Y. C. Lin,1C. H. Chiang,1Ross C. C. Chen,1Y. F. Chen,1Y. H. Wu,2 and L. Chang2
1
Department of Electrophysics, National Chiao Tung University, Hsinchu, Taiwan 30050 2
Department of Materials Science and Engineering, National Chiao Tung University, Hsinchu, Taiwan 30050 (Received 4 June 2011; accepted 9 December 2011; published online 9 January 2012)
A simple critical thickness for generating lattice misfits is insufficient to describe the onset strain relaxation in InAs quantum dots (QDs). A predominant dot family is shown to relieve its strain by In/Ga interdiffusion, rather than by lattice misfits, at the onset of strain relaxation. This argument is based on photoluminescence spectra, which show the emergence of a fine blueshifted transition at the onset of strain relaxation, along with a low-energy transition from a dot family degraded by lattice misfits. From the analysis of the temperature-dependent blueshift and energy separation between the ground and excited-state transitions, the blueshift is attributed to In/Ga interdiffusion. Transmission electron microscopy suggests a relaxation-induced indium migration from the interdiffused dot family to the dislocated dot family. Post-growth thermal annealing can further relieve strain by inducing more In/Ga interdiffusion in the interdiffused dot family and more dislocations in the dislocated dot family. This study explains the co-existence of strong carrier confinement in the QDs and enormous misfit-related traps in the capacitance-voltage spectra, and an elongated QD electron-emission time.VC 2012 American Institute of Physics.
[doi:10.1063/1.3675519]
I. INTRODUCTION
When the InAs deposition thickness exceeds a critical thickness, compressive strain in InAs self-assembled quantum dots (QDs)1–15is relaxed and lattice misfits near the QDs and threading dislocations in the top GaAs layer are observed.16,17Relaxation in this way is undesirable because the induced defects severely degrade the photoluminescence (PL) spectra of the QDs. By capping the QDs with an InGaAs strain-relieving layer and carefully controlling the InAs growth thickness, the threading dislocations in the GaAs layer can be avoided and the relaxation does not severely degrade the PL spectra. The PL spectra of the QDs exhibit an abnor-mal PL blueshift at the onset of strain relaxation.18,19 The capacitance-voltage (C-V) spectra revealed the confusing co-existence of strong electron confinement in the QD layer and enormous misfit-related traps at about 0.35 eV.20 Electron emission time from the QDs is significantly elongated, com-pared to that from the coherent QDs before relaxation.20 These observations cannot be explained by a simple compres-sive strain relaxation by lattice misfits, which is expected to decrease the bandgap of the QDs and cause a redshift. An understanding of the detailed mechanisms of the onset strain relaxation is necessary in order to tailor the properties of the QDs.
When the InAs growth thickness is increased, due to size fluctuation certain QDs are larger than others and the first strain is relieved by lattice misfits. The long range strain field due to these structural imperfections can affect the kinetics of adatom migrations.21,22It has been reported that defects such
as dislocations can be energetically favorable sites for indium migration.18,23 Given sufficient kinetics, the dislocated QDs might affect the strain relieving process of the nearby not yet relaxed QDs. This suggests a more complicated strain relaxa-tion process in the QDs. To study the mechanism of onset strain relaxation, we minutely increased the InAs growth thickness to exceed the dislocation-induced critical thickness with a main purpose to establish the origin of the abnormal blueshift. We showed a degraded PL transition in concomi-tance with the emergence of a blueshifted transition at the onset of strain relaxation. Increasing the InAs growth thick-ness could increase the blueshift. From analyzing the energy separation and the temperature dependence of the blueshifted PL spectra, this blueshift is attributed to In/Ga interdiffusion. Hence, a predominant dot family is strain relaxed by In/Ga interdiffusion, in addition to another dot family relaxed by lattice misfits. Cross-sectional transmission electron micros-copy (TEM) images show a predominant dot family with truncated trapezium shape, as observed in coherent QDs before relaxation, and another family of large dislocated dots. Post-growth thermal annealing produces opposite wavelength shifts in the PL spectra of these two dot families, suggesting different strain relaxation processes.
II. EXPERIMENT
The InAs dots were grown by solid source molecular beam epitaxy on top of a nþ-GaAs (100) substrate, a 0.3 lm-thick Si-doped GaAs (7 1016cm3) barrier layer. The QD
layer was realized by depositing an InAs layer from 1.97 to 3.3 monolayer (ML) at 490C (at a rate of 0.26 A˚ /s). After that, a 60 A˚ In0.15Ga0.85As capping layer was grown at
490 C, followed by the growth of a low temperature GaAs
a)Author to whom correspondence should be addressed. Electronic mail:
jfchen@cc.nctu.edu.tw.
layer for 20 s at 490C. Then, the growth was interrupted for 3 min to ramp temperature to 600C for the growth of a 0.2 lm-thick Si-doped GaAs barrier layer to terminate the growth procedure. The dot sheet density was estimated to be about 3 1010cm2. For C-V and deep-level transient
spec-troscopy (DLTS) probing, Schottky diodes were realized by evaporating Al with an area of 5 103cm2.
III. MEASUREMENT AND RESULTS
A. Effect of onset strain relaxation on the PL spectra
Figures1(a)and1(b)show the 50 and 300 K PL spectra of the InAs QDs with different InAs growth thicknesses. When the InAs thickness is increased from 1.97 to 2.7 ML, the spectra display a ground-state dominated transition from a uniform dot family with an expected redshift due to dot size increase. With the InGaAs capping layer to retard In/Ga interdiffusion, this redshift saturates at1300 nm at 300 K (1250 nm at 50 K) with improved uniformity as seen in the 2.7 ML case. These QDs are coherently strained without de-tectable defects. However, with increasing the InAs thick-ness to 2.93 ML, the1300 nm transition drastically reduces its intensity, suggesting generation of dislocations, consistent with the observation of a misfit-related trap at about 0.35 eV by DLTS20and lattice misfits by TEM pictures near the QD layer. Thus, the dislocation-induced critical thickness is between 2.7 and 2.93 ML. Increasing the InAs thickness to 3.06 and 3.3 ML further degrades this 1300 nm transition. When the temperature is lowered from 300 to 50 K, its peak intensity improves by only a factor of about two, suggesting a detrimental effect of non-radioactive defects through which photo-generated carriers are recombined. This feature sug-gests that the related dot family is degraded by lattice misfits and is denoted by dislocated QDs.
Besides this degraded transition, the PL spectra in Fig.1
show another transition with an abnormal blueshift from 50 to 70 meV with increasing the InAs thickness at the onset of
strain relaxation. When temperature is lowered from 300 to 50 K, this blueshifted transition enhances its intensity by a factor of ten and displays well-resolved ground, first- and sec-ond-excited spectra Figure 2illustrates a similar band-filling effect for the 300 K spectra of the non-relaxed 2.7 ML and relaxed R-3.3 ML (R stands for relaxed), normalized to the ground-state peak. With increasing the excitation power, the ground-state emission begins to saturate and subsequently the first and second excited-state emissions increase their inten-sities. This band-filling effect verifies the blueshifted PL spec-tra are originated from one dot family. It should be noted that this blueshifted transition is narrower than that of the non-relaxed QDs (the 1.97 ML) emitting at a similar wavelength. Hence, the blueshifted transition is relatively high quality and the associated dot family shall not be severely degraded at the onset relaxation. This dot family cannot be relieved mainly by lattice misfits because the blueshift contradicts a compressive strain reduction, which is expected to decrease the bandgap of the QDs and redshift the wavelength.
To establish the origin for the blueshift, Fig.3shows the temperature dependence of the ground-state emission energies for 2.34 and 2.7 ML, and of the blueshifted spectra for R-3.06 and R-3.3 ML, along with the temperature-dependent PL spectra of 2.34 ML and R-3.06 ML in the inset. When temper-ature is lowered from 300 to 50 K, the transition energy is increased by 45 and 46 meV for 2.34 and 2.7 ML and 51 and 53 meV for R-3.06 ML and R-3.3 ML, respectively. Because the ground-state emission energy and the bandgap of the QDs usually follow similar temperature dependence, this result suggests that the bandgap of the blueshifted dot family has a higher temperature-lowering increment rate than that of the coherent QDs before relaxation. This trend is consistent with a reduction of In content (increase of Ga content) in the QDs because, from the Varshni rule, the bandgap of GaAs (or InGaAs) has a higher temperature-lowering increment rate than that of InAs, and the difference between them is 35 meV from 300 to 50 K. The average increment of 7 meV corre-sponds to about 20% loss of indium using a linear proportion between the increments of the ground-state emission energy and the bandgap. This estimated In loss is very rough because
FIG. 1. (Color) (a) The 50 and 300 K PL spectra of the non-relaxed 1.97, 2.34, and 2.7 ML and relaxed 2.93, 3.06, and 3.3 ML QDs. Onset strain relaxation degrades the 1300 nm (at 300 K) transition from the dislocated QDs and generates a blueshifted spectra from the In-outdiffused QDs.
FIG. 2. (Color) 300 K power-dependent band filling for the non-relaxed 2.7 ML and relaxed 3.3 ML QDs, normalized to the ground transition.
it does not consider a detailed relationship between ground-state emission energy and bandgap. However, the trend is consistent with a composition change from InAs toward InGaAs in the blueshifted dot family, relative to the coherent dots just before relaxation. This result along with the PL blue-shift strongly suggests that the blueblue-shifted dot family relieves strain by In/Ga interdiffusion at the onset of relaxation.
Further evidence for In/Ga interdiffusion is provided by a smaller energy separation between the ground and excited emissions, because the occurrence of In/Ga interdiffusion can alter the energy band structure and the excited state would have a wider width (see the simple conduction-band diagram in Fig.2). As shown in Fig. 2, the abnormal blue-shift is 84 meV from 2.7 ML to R-3.3 ML and yet the energy separation between the ground and second-excited emissions is decreased by 21 meV (from 0.132 eV to 0.111 eV). Simi-larly, the energy separation between the ground and first-excited emissions is decreased in the blueshifted dots, as shown by Fig.4. The slight decrease from 2.34 to 2.7 ML is due to dot size enlargement before strain relaxation. A sharp drop (about 8 meV) from 2.7 ML to R-3.06 ML is clearly visible. This energy-separation reduction along with the blueshift is consistent with the effect of In/Ga interdiffusion on the energy band structure of the QDs. On the basis of the above analysis, we believe that, at the onset of strain relaxa-tion, a predominant dot family relieves its strain mainly by In/Ga interdiffusion, rather than by lattice misfits, to retain their fine well-resolved PL spectra. We denote this dot fam-ily as In-outdiffused QDs. Because of size fluctuation, cer-tain QDs are larger than others to relieve strain by lattice misfits at the critical thickness. With further increasing the InAs coverage, one would expect that all the other QDs would eventually relax by lattice misfits. However, on the basis of the above PL analysis, a predominant dot family relieves its strain by In/Ga interdiffusion even when the InAs coverage is increased to 3.3 ML, which is far beyond the critical thickness. Thus, the relaxation by lattice misfits in
certain QDs might affect the adatom migration kinetics and prevent the nearby QDs from strain relaxation by lattice misfits.
B. TEM analysis
To see the microstructures, Figs.5(a)and5(b)show the cross-sectional TEM images of 2.4 ML and R-3.06 ML, respectively. Their statistical distributions of widths and heights are summarized in Fig. 6. The non-relaxed sample shows a uniform trapezium-shaped dot family with an aver-age height of 8.5 nm and an averaver-age base width of 20 nm. The relaxed sample shows a similar trapezium-shaped dot family with an average height of 8.8 nm and an average base width of 27 nm, in addition to a few large-size dots with an
FIG. 3. (Color) Temperature-dependent energy of the ground-state transi-tions of the non-relaxed 2.34 and 2.7 ML QDs and the In-outdiffused QDs of the relaxed 3.06 and 3.3 ML samples. An obvious increase of the temperature-lowering energy increment can be seen in the relaxed samples, suggesting a reduction of In content in the QDs. The inset shows the temperature-dependent PL spectra of the non-relaxed 2.34 ML and the relaxed R-3.06 ML.
FIG. 4. Energy differences between the ground-state and first excited-state transitions of the non-relaxed 2.34 to 2.7 ML QDs and the In-outdiffused QDs of the relaxed R-3.06 ML and R-3.3 ML. The sharp drop from 2.7 to 3.06 ML supports the occurrence of In/Ga interdiffusion.
FIG. 5. TEM images of (a) non-relaxed 2.4 ML and (b) relaxed R-3.06 ML. The 3.06 ML sample shows a predominant family of trapezium shaped dots, and a minor substantially large island-like dot family (indi-cated by arrows).
average height of 16 nm and an average base width of 38 nm [as indicated by the arrows in Fig.5(b)]. A typical large-size dot and trapezium dot are shown in the left and right pictures in the bottom of Fig.5. According to the statis-tics, the number ratio between the trapezium and the large-size dots is 33:6. In terms of the shape and large-size, there are two distinctive dot families: a predominant dot family similar to the coherently strained QDs and a minor dot family of large-size dots. To examine lattice misfits, Fig.7(a)shows the TEM image (the left picture) of a typical trapezium dot in R-3.3 ML and the corresponding Fourier-transformed TEM image (the right picture). The Fourier-transformed TEM reveals no lattice misfits inside the dot but a few misfits under the dot and around the edge, as indicated by circles. This tra-pezium dot family whose contrast is similar to that of the coherent QDs can be correlated to the In-outdiffused QDs. On the other hand, the large-size dots are associated with lattice misfits. The typical large-size dot in Fig.7(b)reveals10 lat-tice misfits inside the dot and8 in the neighboring bottom GaAs layer (two in each of the two small loops and four in the large middle loop) right under the QD. These considerable lat-tice misfits inside the dot can severely degrade the dot and thus the large-size dots are correlated to the dislocated QDs.
On the basis of these TEM observations, we consider the likely onset strain relaxation, as indicated by the sche-matic diagram in Fig.8. The InAs deposition produces QDs in Fig.8(a). The subsequent InGaAs covering can alter the shape of the QDs. Indium atoms can detach from the tops of the QDs and migrate to the base region of the QDs,24,25 lead-ing to the truncated trapezium-shaped QDs. With the InGaAs covering, the emission wavelength can be extended to about 1300 nm by increasing the dot size and retarding In/Ga inter-diffusion. This evolution of growth is generally considered for the coherent QDs. However, when the InAs thickness
exceeds the critical thickness, some QDs are larger than others and are strain relieved by lattice misfits, probably dur-ing the InGaAs coverdur-ing or the initial GaAs coverdur-ing. These dislocated QDs may attract indium adatoms from the nearby still strained QDs,18,23as indicated in Fig.8(b). This indium detachment leads to In/Ga interdiffusion to relieve strain in the In-outdiffused QDs, yielding the abnormal PL blueshift. The base width extension observed in the In-outdiffused
FIG. 6. (Color) Distributions of the widths and heights for the relaxed R-3.06 ML and non-relaxed 2.4 ML samples. The relaxed sample displays two dot families.
FIG. 7. (Color) Cross sectional TEM pictures of (a) a typical In-outdiffused QD and its Fourier-transformed image and (b) a typical dislocated QD and its Fourier-transformed image in the R-3.3 ML sample. The dislocated dot has a significant number of lattice misfits inside the dot and in the bottom GaAs layer under the dot.
QDs is characteristic of In/Ga interdiffusion. An increase of the lateral size seen by TEM was previously reported after the InAs QDs were subjected to post-growth thermal anneal-ing and was explained by strain-driven In/Ga interdiffu-sion.26 The lack of height increase (compared with the coherent QDs) is consistent with the deposited indium ada-toms not being accumulated in the In-outdiffused QDs, but migrate to the dislocated dots as shown in Fig.8(c). There-fore, we speculate that the misfit-driven indium migration can detach indium atoms from the In-outdiffused QDs, caus-ing In/Ga interdiffusion to relieve strain.
C. Post-growth thermal annealing
Figure9shows the effect of a post-growth rapid thermal annealing at 650 and 700C for one minute on the 300-K PL spectra of the R-3.3 ML. For a clear comparison, the spectra of the 700C annealing are shifted up by multiplying a factor of 3. As shown, similar to what is observed in coherent QDs,
thermal annealing can blueshift the transition from the In-outdiffused QDs. Hence, thermal annealing induces In/Ga interdiffusion in the In-outdiffused QDs. Because this inter-diffusion is driven by strain, this result suggests retention of considerable strain in the In-outdiffused QDs before thermal annealing. On the other hand, thermal annealing can shift the transition from the dislocated QDs toward low energy, and generate a tail up to 1500 nm after 700C annealing. Hence, the dislocated QDs undergo strain relief not by In/Ga inter-diffusion but by introducing more dislocations to decrease the bandgap of the QDs. Note that the opposite PL shift is consistent with the different nature of strain distribution in the two dot families, supporting the claimed bimodal strain relaxation.
Relative to annealing at 650C, annealing at 700C can produce a low-energy tail up to 1500 nm. To understand this large change, Fig.10shows the DLTS spectra of the 650C and 700C annealing samples, measured for different voltage sweepings. The spectra of the 650C annealing (shifted up for clarity) reveal a misfit-related trap at 0.35 eV near the QD region (approximately 2/3.5 V) while the top GaAs layer is almost defect-free (with a very weak intensity of the misfit-related trap). These spectra are similar to those observed in the as-grown sample.20This result suggests that annealing at 650C does not significantly introduce more dislocations to relieve strain and the redshift is not appreciable. On the other hand, annealing at 700C significantly alters the DLTS spec-tra. A prominent trap at 0.61 eV (r¼ 9.1 1015cm2) with significant intensity is detected in the top GaAs layer with an emission time two orders of magnitude longer than that of the misfit trap. This trap displays a signature of threading disloca-tions from a comparison with previously reported Arrhenius plots.27,28The intensity of this trap shows no saturation even when the filling pulse duration time is increased to 100 ms, characteristic of a threading-dislocation trap, which shall ex-hibit a logarithmic function with filling pulse duration time and is reflective of Coulombic repulsion of the carriers cap-tured at the traps along the linearly arrayed dislocation lines.27 Hence, thermal annealing at 700C provides sufficient ther-mal energy for the generation of threading dislocations in the
FIG. 8. (Color) Schematic diagram of strain relaxation. (a) Initial growth of the InAs QDs, (b) strain relaxation of the right dot by lattice misfits attracts indium migration from the In-outdiffused dot, and (c) the bimodal strain relaxation.
FIG. 9. (Color) 300 K PL spectra of the relaxed R-3.3 ML sample subjected to a post-growth rapid thermal annealing at 650C and 700C for one mi-nute. Thermal annealing can blueshift the spectra of the In-outdiffused QDs and distend the spectra of the dislocated QDs toward a low-energy side.
FIG. 10. (Color) DLTS spectra of the relaxed 3.3 ML sample after 650C
and 700C annealing. The spectra of 650C annealing display the misfit trap near the QD region (at approximately3/3.5 V) while the spectra of 700C annealing reveal a threading-dislocation trap at 0.61 eV in the top
top GaAs, layer probably by the gliding process through the lattice misfits near the QDs due to elastic acting as a shear stress.28 The generation of these threading dislocations can cause significant strain relieving in the QDs to decrease the bandgap of the QDs, leading to the redshift toward 1500 nm. Therefore, a pure strain relieving by lattice misfits (without In/Ga interdiffusion) near the QDs does not significantly relieve strain to decrease the bandgap, and only by the genera-tion of threading dislocagenera-tions in the GaAs layer can strain sig-nificantly be relieved to reduce the bandgap for observing an apparent redshift. This can explain why the dislocated dots emit a wavelength similar as that of the coherent dots before relaxation, as shown in Fig.1.
D. C-V spectra
The lattice misfits at 0.35 eV (relative to the GaAs conduction-band edge) are effective electron traps that can compensate the ionized impurity in the bottom GaAs layer and cause drastic carrier depletion.29 This compensation extends to the regions under the In-outdiffused QDs, forming a completely connected depletion layer in the bottom GaAs layer. This can be seen by the 300 K apparent-carrier concen-tration profiling (converted from C-V spectra) of the R-3.3 ML sample in Fig.11. Although the top GaAs layer displays a normal depletion near the QDs with a designed background ND¼ 9 1016cm3, the bottom GaAs layer displays a nearly
double depletion width. On the basis of a simple Schottky depletion modelV¼ ðq=2eÞNDL2, a double depletion width
would require a decrease of the background concentration from 9 1016cm3to 2.25 1016cm3, yielding an average
trapped electron concentration Nt¼ 6.75 1016cm3.
Detailed differential capacitance analysis has yielded a simi-lar concentration.29 As shown in Fig. 11, the CV spectra show a sharp electron-confinement peak related to the In-outdiffused QDs and the drastic carrier depletion. Further decreasing reverse voltage can push down the Fermi level to intersect with the misfit-related traps, and more reverse volt-age is needed for the Fermi level to sweep out the trapped
electrons and yield the misfit-related C-V plateau (from3.5 to4 V), as shown in the inset of Fig.11. In order to sweep out the trapped electrons, the edge of the depletion region needs to move a width of 0.358 0.328 ¼ 0.03 lm, deter-mined from C¼ 175 to 160 pF in the misfit plateau. This width and the trapped electron concentration (6.75 1016cm3) yields trapped electrons of 2 1011cm2. This
is about one order of magnitude higher than the QD density (about 3 1010cm2) and is comparable to the total
elec-trons confined in the In-outdiffused QDs.
Because the drastically depleted region can suppress tun-neling, electrons in the In-outdiffused QDs need a substantial time to emit to the bottom GaAs layer. The long emission time can be easily measured by the resolvable frequency dis-persion in Fig. 12. To obtain the electronic band, we use the high-frequency capacitance at 100 kHz (where the QD elec-trons cannot follow to be modulated) to obtain the depletion-region width L, and the confinement energyE (relative to the GaAs conduction-band edge) of the Fermi-level probed elec-trons in the QDs, from a simple Schottky depletion model, E¼ ðq=2eÞNDðL 0:2Þ
2
þ ðkT=qÞInðNC=NDÞ, where ND is
the compensated concentration in the bottom GaAs layer, 0.2 lm is the distance of the QD layer from the surface andNc
is the effective density of states in the GaAs conduction band. Thus, each reverse bias can be converted to anE. The corre-sponding electron density of states in the QDs (at thisE), CQ,
can be obtained from the low-frequency capacitance (at 3 kHz) using CL¼
ðC1þCQÞC2
ðC1þCQÞþC2, where C1¼ e=ðL 0:2Þ and
C2¼ e=0:2 are the geometric capacitance from the QD layer
to the edge of the depletion region and from the sample sur-face to the QD layer, respectively.30 The obtained electronic band is shown in the inset of Fig.12, which can be correlated to the capping layer, first-excited, and ground sates of the In-outdiffused QDs and dislocated QDs by a comparison with the PL spectra. The dislocated QDs peaked at 0.3 eV are 47 meV below the ground state of the In-outdiffused QDs. Comparing with the difference of 71 meV in the 300 K PL spectra (1.019 and 0.948 eV from the In-outdiffused and dislo-cated QDs, respectively), the ratio for the band-offsets between electrons and holes is 6.6:3.4, a value close to that
FIG. 11. (Color) Apparent-carrier concentrations of the R-3.3 ML sample, measured at 90 K, 30 kHz, and 300 K, 70 kHz, respectively. These concen-trations are converted from C-V spectra, as is the one measured at 80 K and 1 kHz in the inset.
FIG. 12. (Color) 90 K C-V spectra of the R-3.3 ML sample, which shows the electron confinements of the capping layer, the first-excited and ground states of the In-outdiffused QDs and the dislocated QDs, as illustrated by the converted electronic band in the inset.
previously reported.31 Hence, the electronic band displays a bimodal structure. Also, the C-V plateau at around3 V can be attributed to the electrons in the In-outdiffused QDs even though its emission time is five orders of magnitude longer than that of the coherent QDs before relaxation. This analysis explains the co-existence of strong dot-carrier confinement and enormous misfit-related traps. The elongation of the emis-sion time is due to the alteration of the electron emisemis-sion pro-cess in the In-outdiffused QDs.
IV. CONCLUSIONS
This work presents a detailed study of onset strain relax-ation in InAs QDs. Evidence is presented to show a bimodal strain relaxation in which a minor dot family is relaxed by lattice misfits and a predominant dot family is relaxed by In/ Ga interdiffusion. This bimodal strain relaxation is supported by TEM pictures, which display a predominant family of tra-pezium shaped dots, similar to those observed in the coher-ent QDs, and a minor dot family of large-size dislocated dots. A likely mode of strain relaxation is discussed to explain the observed PL blueshift. Thermal annealing can produce opposite wavelength shift in the two dot families. This bimodal strain relaxation can explain the elongated electron emission time in the relaxed QD samples.
ACKNOWLEDGMENTS
The authors would like to thank the National Science Council of the Republic of China, Taiwan (Contract No. NSC-97-2112-M-009-014-MY3) and for financially support-ing this research. Dr. J. Y. Chi and R. S. Hsiao are com-mended for preparing the samples.
1
F. Heinrichsdorff, M. H. Mao, N. Kirstaedter, A. Krost, and D. Bimberg,
Appl. Phys. Lett.71, 22 (1997).
2D. J. Eaglesham and M. Cerullo,Phys. Rev. Lett.64, 1943 (1990). 3
D. Leonard, K. Pond, and P. M. Petroff,Phys. Rev. B50, 11683 (1994).
4
J. M. Moison, F. Houzay, F. Barthe, and L. Leprince,Appl. Phys. Lett.64, 196 (1994).
5C. W. Snyder, J. F. Mansfield, and B. G. Orr,Phys. Rev. B46, 551 (1992). 6H. Shoji, K. Mukai, N. Ohtsuka, M. Sugawara, T. Uchida, and H.
Ishi-kawa,IEEE Photonics Technol. Lett.7, 1385 (1995).
7
G. Yusa and H. Sakaki,Electron. Lett.32, 491 (1996).
8J. C. Campbell, D. L. Huffaker, H. Deng, and D. G. Deppe,Electron. Lett.
33, 1337 (1997).
9
C. M. A. Kapteyn, F. Heinrichsdorff, O. Stier, R. Heitz, M. Grundmann, and P. Werner,Phys. Rev. B60, 14265 (1999).
10P. N. Brunkov, A. Patane, A. Levin, L. Eaves, P. C. Main, Y. G. Musikhin,
B. V. Volovik, A. E. Zhukov, V. M. Ustinov, and S. G. Konnikov,Phys. Rev. B65, 085326 (2002).
11
J. M. Garcia, J. P. Silveira, and F. Briones, Appl. Phys. Lett. 77, 409 (2000).
12X. Letartre, D. Stievenard, and M. Lanoo,J. Appl. Phys.
69, 7336 (1991).
13
H. Drexler, D. Leonard, W. Hansen, J. P. Kotthaus, and P. M. Petroff,
Phys. Rev. Lett.73, 2252 (1994).
14S. Sauvage, P. Boucaud, F. H. Julien, J.-M. Gerard, and J.-Y. Marzin,J. Appl. Phys.82, 3396 (1997).
15
H. L. Wang, F. H. Yang, S. L. Feng, H. J. Zhu, D. Ning, H. Wang, and X. D. Wang,Phys. Rev. B61, 5530 (2000).
16J. S. Wang, J. F. Chen, J. L. Huang, P. Y. Wang, and X. J. Guo,Appl. Phys. Lett.77, 3027 (2000).
17
J. F. Chen and J. S. Wang,J. Appl. Phys.102, 043705 (2007).
18
N. A. Cherkshin, M. V. Marksimov, A. G. Makarov, V. A. Shehukin, V. M. Ustinov, N. V. Lukovskaya, Y. G. Musikhin, G. E. Cirlin, N. A. Bert, Z. I. Alferov, N. N. Ledentsov, and D. Bimberg,Semiconductors37, 861 (2003).
19
J. F. Chen, R. S. Hsiao, Y. P. Chen J. S. Wang, and J. Y. Chi,Appl. Phys. Lett.87, 141911 (2005).
20J. F. Chen, Y. Z. Wang, C. H. Chiang, R. S. Hsiao, Y. H. Wu, L. Chang, J.
S. Wang, T. W. Chi, and J. Y. Chi,Nanotechnology18, 355401 (2007).
21
C. W. Snyder, B. G. Orr, D. Kessler, and L. M. Sander,Phys. Rev. Lett.
66, 3032 (1991).
22
M. V. Maximov, A. F. Tsatsulnikov, B. V. Volovik, D. S. Sizov, Y. M. Shernyakov, I. N. Kauander, A. E. Zhukov, A. R. Kovsh, S. S. Mikhrin, V. M. Ustinov, Z. I. Alferov, R. Heitz, V. A. Shchukin, N. N. Ledentsov, D. Bimberg, Y. G. Musikhin, and W. Neumann, Phys. Rev. B 62, 16671 (2000).
23
T. S. Shamirzaev, A. M. Gilinsky, A. K. Kalagin, A. I. Toropov, A. K. Gutakovski, and K. S. Zhuravlev, Semicond. Sci. Technol. 21, 527 (2006).
24N. N. Ledentsov, J. Bohrer, D. Bimberg, I. V. Kochnev, M. V. Maximov,
P. S. Kopev, Z. I. Alferov, A. O. Kosogov, S. S. Ruvimov, P. Werner, and U. Gosele,Appl. Phys. Lett.69, 1095 (1996).
25F. Ferdos, S. Wang, Y. Wei, A. Larsson, M. Sadeghi, and Q. Zhao,Appl. Phys. Lett.81, 1195 (2002).
26
A. Babinski, J. Jasinski, R. Bozek, A. Szepielow, and J. M. Baranowski,
Appl. Phys. Lett.79, 2576 (2001).
27T. Wosinski,J. Appl. Phys.65, 1566 (1989). 28
Y. Uchida, H. Kakibayashi, and S. Goto,J. Appl. Phys.74, 6720 (1993).
29
F. Chen, Ross C. C. Chen, C. H. Chiang, M. C. Hsieh, Y. C. Chang, and Y. F. Chen,J. Appl. Phys.108, 063705 (2010).
30J. F. Chen, Ross C. C. Chen, C. H. Chiang, Y. F. Chen, Y. H. Wu, and L.
Chang,Appl. Phys. Lett.97, 092110 (2010).
31
S. D. Lin, V. V. IIchenko, V. V. Marin, K. Y. Panarian, A. A. Buyanin, and O. V. Tretyak,Appl. Phys. Lett.93, 103103 (2008).
