Result and Discussions

The fabricated laser waveguide had a width of 20 μm. The lengths of the two sections were 300 μm and 850 μm. By injecting different amount of currents in the two sections, we were able to control the lasing mode. Either section can be used as the gain region or the absorbing region. In other words, either section can be used to control the mode switching.

Pure ground state lasing, excited state lasing or both state lasing are achievable by merely adjusting the current ratio of the two sections.

Fig. 4.8 shows the lasing characteristics as functions of the currents injected into the two sections, the horizontal axis being the current injected into the shorter section and the vertical axis being the current injected into the longer section. The emission wavelength of the ground state transition is 1294 nm while that for the excited state transition is 1200 nm. Different lasing modes take place in different regions in this two-dimensional plot. The boundary of the data in each region indicates the threshold condition. Clearly two boundaries can be identified, one for the ground state lasing and the other for the excited state lasing. The allowed lasing region is where the currents are higher than the boundary. The overlap area of the two allowed regions is where dual state lasing takes place. We notice that the threshold boundaries bend inwards at high currents, especially for the ground state threshold curve. In other words, if we increase the current injected into one of the sections, the current going to the other section needs to be increased also in order to reach the threshold condition. This is in contrary to what one would expect because the optical gain in each section should increase with the injection current. However, when we map the threshold condition shown in Fig.4.8, the currents used are relatively high compared to what one would normally use to operate a laser. The effect of heating causes the gain to drop because of the Fermi distribution is a function of temperature.

So higher currents are needed in order to reach the threshold. This phenomenon is particularly obvious for the ground state lasing. This is because the gain at a lower lasing energy is

affected more by the Fermi function, causing the saturated gain to drop.

This current dependent lasing mode distribution can be understood as follows: The gain due to ground state transition is limited (by the dot density) and saturates at high currents. If the cavity loss is low, the ground state will lase first. But if the current injected to one of the section is not enough or even below transparency, the gain required from the other section has to be increased in order for the total loop gain to reach the threshold condition. This required gain may go beyond the saturation value of the ground state. So in this case the ground state can not lase. The only state can lase is the excited state, which, because of a higher density of states, can provide a higher gain as long as one of the section has a high enough current. This is why that the sole excited state lasing occurs when one section has a very high current while the other one has a low current. The dual state lasing happens when currents to both sections are high and it happens in a very wide range. The origin of dual state lasing has been discussed previously and has been attributed to partial clamping of the carriers in the ground state after threshold [59-62].

Based on the result shown in Fig.4.8, one can clearly see that wavelength switching among various QD lasing modes is possible through the variation of currents applied to the separate sections. If we fix the total current and simply change the ratio of the currents applied to the two sections, it is possible that we can switch among ground state lasing, excited state lasing and dual state lasing. Fig. 4.9 shows the spectra at three different current ratios along the line with a constant total current of 225 mA. Three different lasing modes were obtained at three different current ratios. The amount of wavelength change is around 100 nm.

Ground state lasing

FIG. 4.8. Lasing mode distribution vs input currents to the two sections of the laser cavity.

1294 nm 1200 nm

1.18 1.20 1.22 1.24 1.26 1.28 1.30 1.32 wavelength (nm)

1.18 1.20 1.22 1.24 1.26 1.28 1.30 1.32 wavelength (nm)

1.18 1.20 1.22 1.24 1.26 1.28 1.30 1.32 wavelength (nm)

FIG. 4.9. Lasing mode switching by adjusting the ratio of the currents injected into the two sections. The three spectra, corresponding to ground state, dual state, and excited lasing, were measured along a constant current line with a total current of 225 mA.

4.6.4 Conclusion

We demonstrated two-state switching, between the ground state ~1.3 μm and the excited state ~1.2 μm, of an InAs quantum dot laser using a two-section quantum dot laser. Mode switching was achieved by adjusting the gain of each section by the current injected into that section. With a constant total current, we were able to switch between 1.2 μm emission and 1.3 μm emission simply by adjusting the current ratio applied to the two sections.

Chapter 5

The Growth Behavior of InAs QDs on InAlAs/GaAs Metamorphic Buffer Layer

5.1 Introduction

The utilization of self-assembled InAs quantum dots grown on metamorphic buffer has extended the lasing wavelength beyond 1.3 μm [63, 64]. For further laser applications, this topic has attracted considerable interest in the past years. The metamorphic buffer, compositionally graded layers, is made by changing the composition of epilayers along growth direction of epilayers. This buffer can accommodate the large lattice mismatch between the active layer and the GaAs substrate by forming misfit dislocation. They can isolate these misfit dislocations and then prevent their propagation into the active layer. In addition, due to the arbitrary final composition, the metamorphic buffer provides a virtual substrate with arbitrary lattice constant. A high quality metamorphic buffer has characteristics of strain relaxation and efficient misfit dislocation filtering. Up to date, various metamorphic buffers have been grown on GaAs substrates such as InGaAs [65], InAlAs [66], InGaP [67], AlGaInAs [68] and AlGaAsSb [69].

5.2 Strain Relaxation of the Metamorphic Buffers

In our work, ternary InxAl1-xAs alloy was used to obtain the graded metamorphic buffer layer. Optimizing the characteristics of metamorphic buffers (i.e. strain relaxation and efficient misfit dislocation filtering) can provide a wide range of indium (In) content with fully relaxed InxAl1-xAs hetero-structure. The indium content was usually chosen at the range of 25-60%. The relaxation of residual strain in the buffer layer is usually incomplete. The

incomplete relaxation result in the increase of residual strain in the top (active) layers. This residual strain of the top layer can affect the device performance.

The inverse step drop technique was used to effectively relax residual strains in the metamorphic buffers by growing a graded buffer layers with excessive indium content relative to the indium content of the top layers. Thus, the inverse step drop technique is possibly a good method to produce better relaxation rates in the top layers. Nevertheless, the amount of decreased indium content has to be optimized with respect to the relaxation of the underlying graded layer to avoid compressive strain (slightly indium decrease) or tensile strain (significant indium decrease) in the top layers. In this study, we grow InAlAs graded buffer layers on GaAs with different excessive indium contents with respect to the content in the top layers. We then investigated the growth behavior of InAs QDs on these metamorphic structures.

5.3 Sample Growth

In this study, metamorphic structure and InAs quantum dots (QDs) were grown by solid source molecular beam epitaxy. The metamorphic structure of epi-layer was grown on si-doping (100) GaAs substrates.Prior to the growth, surface oxide was desorbed under the suitable As2 flux from the GaAs substrate until the reflective high-energy electron diffraction (RHEED) pattern showed a clear 2×4 surface reconstruction. A 300 nm GaAs buffer layer was grown on the substrate first to obtain a smooth GaAs surface, followed by a linear-graded buffer consisted of undoped InxAl1-xAs layers. The indium content of this linear-graded buffer was obtained by increasing the indium cell temperature and decreasing the aluminum cell temperatures in order to make growth rate constant. As a result, the growth started with indium content close to 2% in the InxAl1-xAs linear-graded buffer, and the average growth rate in the graded buffer layer was kept at 0.6 μm/hr. After the linear-graded buffer layer was formed, a thin inverse layer (InAlAs) was grown. The indium composition in the inverse layer

was lower than the final indium content in the linear-graded buffer layer. Next, a 200 nm InGaAs layer and a 2.6-monolayer (ML) InAs QD layer were grown above the inverse layer such that InAs QDs were capped by a 500 nm InGaAs layer. Finally, a 2.6-ML InAs QD layer was deposited on the top of the surface for the purpose of indirectly “viewing” the growth condition of QD underneath by observing the surface morphology using atomic force microscope (AFM). After the growth of the linear-graded buffer layer, the indium content above the buffer layer (i.e., the top layer (InAlAs/InGaAs)) was kept constant. In this work, the final indium content in the linear-graded buffers were chosen at the range of 30 to 35%.

The indium content in the top layer was kept at 25%. Fig. 5.1 shows the schematic drawing of the metamorphic structure of the samples and their indium composition profile.

FIG. 5.1. The schematic drawing of the metamorphic structure of the samples and their indium composition profile.

Table 5.1 Final In content and growth temperature in graded buffer layer, In composition in the InAlAs inverse layer and InGaAs layers for all samples.

Final In% for the

Table 5.1 shows the final indium composition and the growth temperature in a linear-graded buffer layer and the indium composition in the InAlAs inverse layer and in InGaAs layers for all samples. The linear-graded buffer layers of all samples were grown at low temperature (330-450°C). The growth at low temperature was followed by the growth at high temperature (500°C) for the growth of an additional step and InAs QDs layers. The time from low temperature to high temperature was three minutes. High V/III is necessary for metamorphic layer growth. To achieve this goal, As2 beam-equivalent pressure BEPAs2 was tuned to 1.2x10^-6 Torr such that V/III was higher than 20.

The surface morphology was characterized by a Nanoscope III (Digital Instruments Inc.) AFM in the tapping mode. The scanned areas of the examined samples were 5x5 μm² and 2x2 μm². The depth resolution was less than 1 nm. The normal scan rate was one line per second and data were taken at 512 points per line and 512 lines per scan area. Specimens for cross-sectional Transmission Electron Microscopy (TEM) were prepared by mechanical grinding and polishing followed by ion milling. The TEM studies were carried out on a JEOL 2000 microscope operated at 300 kV. Images for quantitative analysis of the (002) dark-field

intensities were taken with a slow-scan charge coupled device Gatan camera providing a wide dynamic range and high linearity. We also performed low-temperature (25 K) photoluminescence measurement using the 514.5 nm line of an argon ion laser. The signal was collected using an InGaAs detector and lock-in technique.

5.4

The optimized metamorphic hetero-structure is that there is no residual strain in the top layer above the graded metamorphic buffer. The residual strain in the top layer is affected by the amount of excessive indium content of metamorphic buffer over the indium content of the InAlAs inverse layer. The residual strain can be relax completely in the inverse layer when the amount of decreased indium content in the inverse layer is optimized. For most of cases, the final indium composition in the buffer layer was fixed and usually around 20% to 60%.

The indium content in the inverse layer can be decreased by the method mentioned earlier for the purpose of lowering strain. The amount of decreased indium content in the inverse layers was typically 10%~15% relative to the indium content of the buffer layer and depending on the indium composition in the buffer layer. For example, the indium content of the top layers is 35% when the final indium composition of the buffer layer is 45% to 50%. Here, we compared the effect of the residual strain in the growth behavior of QDs between two samples (i.e., sample A versus sample B). Sample A and sample B had the same indium content (i.e., 25%) in the top layer but different indium content in the buffer layer. The final indium compositions of the buffer layer in samples A and B are 35% and 30%, respectively. Because the indium composition in the top layer is 25%, the amounts of decreased indium content in the inverse layer with respect to the graded buffer layer for sample A and B are 10% and 5%, respectively. Different indium composition in the buffer layer also means that the relaxation rates of the buffer layer in sample A and sample B are different.

Fig. 5.2 shows AFM images of sample with 5x5 μm² scanned area (Fig. 5.2(a)) and an

enlarge image (Fig 2(b), 2x2 μm² scanned area) of area “a” in Fig. 5.2 (a). Fig. 5.3 shows AFM images of sample B. Fig. 5.3(a) shows AFM images with 5x5 μm² scanned area. Fig.

5.3(b) is the AFM images of 2x2 μm² taken on the positions ‘a’, as indicated in Fig. 5.3(a).

Fig. 5.3(c) shows surface height profile in Fig. 5.3(b).

The surface morphology of both samples reveals the modulation in the dot distribution.

The density of the dots modulates along the [110] direction. QDs were selectively grown on the surface. We observed that the selective growth was more obvious for sample B than sample A. One possible reason is that the non-uniform strain of the top layer caused the surface modulation. The origin of the non-uniform strain is not clear, but it is related to the formation of underlying non-uniform indium composition. When the inverse layer cannot relax the residual strain of the graded buffer layer, it results in the formation of the non-uniform indium composition in the inverse layer. This phenomenon could enhance the formation of non-uniform strain in the top layer. For sample B, the amount of decreased indium content in the inverse layer with respect to the buffer layer was insufficient. The inverse layer did not effectively relax residual strain in the metamorphic buffer and suffered from stronger compressive strain. Therefore, the modulation was more severe. Thus the distribution of selective growth of InAs QDs was sensitive to the modulation. That the amount of modulation depended on the amount of decreased indium composition had been reported by many research groups [70, 71]. Fig 5.3(b) shows QD size is inhomogeneous in the modulated period surface. Fig. 5.3 also shows InAs QDs are not easy to grow on the valley of the modulated surface.

(a)

(b)

FIG. 5.2. (a)AFM images with 5x5 μm² scanned area. (b) the AFM images of 2x2 μm² taken on the positions ‘a’, as indicated in Fig. (a).

a

(a)

(b)

a

(c)

Fig. 5.3. (a) AFM images with 5x5 μm² scanned area. (b)The AFM images of 2x2 μm² taken on the positions ‘a’, as indicated in (a). (c) surface height profile in (b).

Fig. 5.4(a) shows the AFM result of sample C with 5x5 μm² scanned area. Fig. 5.4(b) is the enlarged AFM images of 2x2 μm² scanned area taken from area ”a” in Fig. 5.4(a). The metamorphic structure of sample B is the same as sample A, but the growth rate in the In0.25Al0.75As inverse layer and In0.25Ga0.75As layer is different; 0.15 μm/hr in sample B versus 0.1 μm/hr in sample A. From Fig.5.4 (a) and (b), the period modulated along the [110]

direction are also observed, but the selective growth do not occur in this sample. The distribution of the QDs are uniformly grown on the surface.

(a)

(b)

Fig. 5.4. (a) AFM images with 5x5 μm² VOF. (b)The AFM images of 2x2 μm² taken on the positions ‘a’, as indicated in (a).

a

We took TEM images (1-10) with bright field cross-section in order to identify that the embedded QDs have selective growth. Figs. 5.5 (a)-(c) show the TEM images of sample B.

Fig. 5.5(b) and (c) were performed using higher magnification. Apparently, the TEM images reveal the selectively growth of QDs. In Fig. 5.5 (b), QDs do not appear in the embedded layer and surface. In Fig. 5.5 (c), QDs are found in both the embedded layer and the surface and they have shown the same distribution. It indicates that the modulated surface had existed in the inverse layer. The rest of the top layer remains the modulation behavior. However, the selective growth behavior of QDs identical in either embedded layers or on the surface.

Fig. 5.6 (a) and (b) show the TEM images of sample C. The TEM images indicate that there is no selective growth of QDs in sample C. Fig. 5.6 (b) shows the image with higher magnification. Clearly, the QDs distribution in the embedded layer is the same as the distribution on the surface. AFM and TEM images show the consistent results.

A high density of dislocations was found in the lower part of the graded buffer layer. The density of dislocations rapidly fell down in the upper part of the graded buffer layer. No dislocation is detected in the inverse step. TEM photographs reveal the perfect crystal in the inverse layer. It indicates that the buffer layer efficiently filtered the threading dislocation.

(a)

(b)

(c)

FIG. 5.5. (a) The TEM images for sample B. The image reveals the selective growth of InAs QDs. The embedded QDs layer still exist QDs. The QD distribution for embedded layer is the same as the surface. (b) the region without QDs under higher magnitude. The embedded QDs layer don’t exist QDs. (c) the region with QDs using higher magnification.

(a)

(b)

FIG. 5.6. (a) The TEM images for sample C. The image reveals QDs wiyhout selectively growth. The QDs distribution for embedded layer is same as the surface. (b) TEM image with higher magnification.

Fig. 5.7 shows photoluminescence (PL) results taken at the low temperature. The emission peak shift slightly between samples that the emission peak energy of sample B is higher than that of samples A and C. It indicates that the emission peak depended on the relaxation rate in the inverse layer. The high-energy emission peak is due to an incomplete relaxation in the inverse layer, and InAs QDs suffered from high strain in the underlying layer. PL results indicate that the relaxation rate in the metamorphic structure only affected the selective growth behavior of QDs.

Fig. 5.7. The low temperature photoluminescence measurement off sample A, B and C.

5.5 The Growth behavior of the InAs QDs with Different Growth Temperatures of the Buffer Layer

If metamorphic samples have the same structure and composition profile, the behavior of the period modulation should be inconsistent due to the various amount of residual strain in the buffer layer. The growth conditions of the buffer layer can affect its amount of residual

strain. One of the growth conditions is the growth temperature. In this subsection, we present how the growth temperature of the buffer layer affects the amount of the residual strain in the metamorphic structure.

Both samples B and D have the same structure and growth condition except the growth temperature of the buffer layer. For samples B and D, the growth temperature of the buffer layer is 380 and 450^oC, respectively. The AFM results of sample D were shown in Fig.5.8 including an AFM image with 5x5 μm² scanned area (Fig. 5.8(a)) and the profile of the surface height (Fig. 5.8(b)). The AFM images also show the selective growth when the growth temperature of the metamorphic graded buffer increased to 450^oC. The profile of the surface height indicated that the QDs did not grow on the valley of the modulated surface. It also indicated that high growth temperature of the buffer layer could not improve the strain relaxation rate of the buffer layer when samples B and D have the same graded indium profile

Both samples B and D have the same structure and growth condition except the growth temperature of the buffer layer. For samples B and D, the growth temperature of the buffer layer is 380 and 450^oC, respectively. The AFM results of sample D were shown in Fig.5.8 including an AFM image with 5x5 μm² scanned area (Fig. 5.8(a)) and the profile of the surface height (Fig. 5.8(b)). The AFM images also show the selective growth when the growth temperature of the metamorphic graded buffer increased to 450^oC. The profile of the surface height indicated that the QDs did not grow on the valley of the modulated surface. It also indicated that high growth temperature of the buffer layer could not improve the strain relaxation rate of the buffer layer when samples B and D have the same graded indium profile