Chapter 2 Experimental Instrument Setup
2.4 Cross-section Transmission Electron Microscopy Sample Preparation
Due to the strong interaction between electrons and matter, the specimens have to be rather thin (<<1000 nm) for Transmission electron microscopy (TEM) investigation. Thus, bulk materials have to be thinned enough to make the electron transparent.
Cross-section TEM specimens were prepared using a Tripod polisher with diamond abrasive films. This method is very powerful for the characterization of interface roughness, composition segregation…etc. First of all, the samples were constructed with Si and were grinded mechanically with both two faces. The diamond abrasive films with several particle sizes (30, 15, 6, 3 and 1 m) were utilized on the wet plate. After grinding both faces, the thickness of film was smaller than 1 m. Then, a copper grid with a 2 mm X 2 mm hole was epoxied onto the mirror surface of the mechanically polished specimen. The use of the copper grid prevents from generating cracks in sapphire substrates and it plays the role of supporting layer compared to specimen without copper grid.
For ion milling process, a GATAN 691 Precision Ion Polishing System (PIPS) is used as displayed in Figure 2.7. In this machine, two focused Ar ion beams mill the dimple-ground sample in such a way that a hole results at the desired position. In general, the parameters for the ion milling process are rather specific for the material and have to be optimized. The ion milling rate increases with higher etching angle and higher etching voltage; however, the sample is also more severely damaged. Therefore, the angle as well as the voltage should be kept rather low. Generally, a higher voltage combined with lower angle is less harmful to the sample than lower voltage combined with higher angle. A TEM sample is least contaminated directly after the ion milling process.
Figure 2.1 Different symbols for various transitions. (a) Band-to-band; (b) free exciton (FE) ; (c) donor bound exciton (D0, X) ; (d) acceptor bound exciton (A0, X) ; (e) donor acceptor pair (DAP).
Figure 2.3 Setup of µ-photoluminescence system.
Figure 2.4 Setup of photoluminescence excitation measurement.
Figure 2.5 Apparatus for X-ray measurement.
Figure 2.7 GATAN 691 Precision Ion Polishing System (PIPS).
Chapter 3 Optical Characteristic of InGaN Multi-quantum Well with δ-TMIn Flow 3.1 Introduction
InGaN/GaN multiple quantum wells (MQWs) are used as active layers in high-brightness light-emitting diodes (LEDs) and laser diodes (LDs) in the ultraviolet-blue-green range [31-32]. The blue LEDs, which showed higher efficiency than that of green LEDs, have been widely used in the communication and information technologies.
The Indium (In) mole fraction of 0.1-0.2 and 0.45 is required for blue and green LEDs, respectively. And the large lattice mismatch (~11 % in the c-axis) between InN and GaN was found to give rise to a solid pahse miscibility gap. In this situation, indium aggregation and phase separation can occur in InGaN through the process of spinodal decomposition [4]. In this process, the diffusion of indium atoms lead to the formation of high indium InGaN clusters or phase-separated In-rich nano particles such that the strain distribution can be relax.
At normal growth temperatures, the alloy is unstable over the entire composition. The formation of phase separation leads to a quantum dot-like structure, which is highly beneficial to obtain high external quantum efficiency from the InGaN/GaN MQWs structure.
In our material studies, it was shown that QD-like structures exist around the designated InGaN MQW layers. The cluster structures form spatial potential fluctuations and localized energy states for trapping carriers for effective radiative recombination. It is usually believed that the carrier localization mechanism is the key to the efficient photon emission in such a compound of relatively higher defect density (~ 108 cm-3) [33-34]. Typically, the process of aggregation and hence the effect of carrier localization become stronger with increasing average indium content [35-36]. Owing to carrier localization, many phenomena different from those in conventional III-V semiconductors were observed. For example, the temperature-dependent PL peak energy exhibits an S-shape behavior ( redshift – blueshift – redshift ) which has been explained with temperature–dependent carrier dynamics, associated with carrier localization in potential minima [34, 37]. Meanwhile, a large Stokes’ shift of PL peak with respect to absorption peak was often observed in an InGaN/GaN MQW structure.
The behavior was usually attributed to the carrier localization effect. The Stokes’ shift of a wide range of InGaN epilayers and commercial LEDs demonstrate a linear dependence of Stokes shifts on emission peak energy.
In general, the trimethylindium (TMIn)-flow rate is constant during the growth of QW.
However, due to the larger lattice mismatch between InN and GaN, it is usually difficult to
self-assembled dots-like region would be useful to reduce the broadening and to enhance optical performance in InGaN-based optoelectronics. In this study, we present the improvement in localized effects in InGaN QW by using δ-TMIn flow process during the growth of well layers.
3.2 Sample Structure
The epitaxial growth of InxGa1-xN/GaN MQWs on c-plane (0001) sapphire (Al2O3) substrates was performed by metalorganic vapor phase epitaxy (MOPVE). The precursors of Ga, In and N were trimethylgallium (TMGa), trimethylindium (TMIn) and ammonia (NH3).
Bis-cyclopentadienyl magnesium (Cp2Mg) and silane (SiH4) were used as p-type and n-type doped. Prior to the deposition of a GaN nucleation layer, the sapphire wafer was pre-baked at 1100 ℃ with H2 ambient for 10 min. A 30-nm-thick GaN nucleation layer was grown at 550
℃, followed by a 4-um-thick Si-doped n-type GaN layer at the elevated temperature of 1060
℃. The 8-pairs of InxGa1-xN/GaN MQWs were grown at 770 ℃. The InxGa1-xN/GaN MQWs had been capped with a 50-nm-thick Mg-doped p-type AlGaN layer as electron blocking layer grown at 1050 ℃. And finally, a 100-nm Mg-doped p-type GaN layer to prevent surface recombination. For the 8-pairs InxGa1-xN/GaN MQWs active region, each pair consists of a 2.5-nm thick InxGa1-xN well layer and a 13-nm thick GaN barrier layer. The sample structure was shown in Figure 3.1.
In order to investigate the influence of δ-TMIn flow on optical and structural properties, two different samples were fabricated. Noting that the growth condition of these two samples were the same, except for the differently initial TMIn-flow rate (fTMIn) in the well layer. For sample A, the fTMIn was fixed at 230 sccm for overall growth of InGaN layers. But for sample B, the initial fTMIn in each InGaN well layer was 400 sccm persisting for a 10% growth time of an InGaN layer, and was then switched to 230 sccm. Schematic fTMIn variations during growth for sample A and B were shown in Figure 3.2.
3.3 Material analysis
3.3.1 TEM image of InGaN/GaN MQW
Figure 3.3 showed cross-section bright-field high-resolution TEM image obtained form the InGaN/GaN MQWs structure of sample A and sample B. The barriers and wells in the MQWs can be easily distinguished by fluctuations in the indium composition. The well and barrier thickness of these two sample were estimated about 2.2 nm and 13.0 nm, respectively,
which was good agreement with the structural design. Figure 3.3 (a) and (b) both showed a very sharp interface, which indicated the good quality of InGaN/GaN MQWs structure even for the sample B with δ-TMIn flow rate.
We can observe that a number of obvious dark spots were found in the well regions. The distinct dark spots were considered as In-rich regions of the compositional inhomogeneity and acted as a QD or a quantum mesodot if the potential gap is large enough to confine particles laterally [9, 21, 34, 36]. In general, the QD-like regions and In-rich regions were as a result of low miscibility between GaN and InN. And these regions considered as the origin of the emission in the well layer and can increase the internal quantum efficiency.
3.3.2 High Resolution X-ray Diffraction of InGaN/GaN MQW
Figure 3.4 showed the HRXRD diffraction pattern for the (0004) reflection of θ-2θ spectra measured of the samples A and B. The θ-2θ diffraction patterns were simulated using a computer program based on dynamical theory. The strong peak in each spectrum originated from the GaN epilayers. SL satellite peaks were marked as SL-1, SL-2…. Additionally, both spectra showed higher order diffraction peaks, which indicated the good layer periodicity and the structural quality.
In principle, the average indium composition in the well layer and the period (well and barrier) can be determined from the relative positions of the 0-th and higher-order peaks in the HRXRD spectra. The period (D) is given by[39]
(sin sin 0 )
2 n
D= nλ θ − θ th (3-1)
where n the order of the satellite peaks, λ the wavelength of x-ray radiation, θn the diffraction angle, and θ0th the angle of the 0th-order peak. According to the best fit to the measured spectra, we obtained average indium composition (x) of the well layers and the approximate value of D. D for both samples were similar, which were about 14.7 nm respectively and consistent with the measured values from the TEM images. And indium composition (x) of sample A and B were similar, both about 0.2. The composition of samples A and B was similar but the FWHM of SL higer order peak of both samples was quite different. As shown in Table 1, for both sample, the FWHM of the higher-order SL satellite peaks broaden. And the FWHM of sample B with δ-TMIn flow rate was always larger than sample A. This broadening may be caused by spatial variation of alloy composition fluctuation [40].
Table 1 The FWHM of Satellite peaks of sample A and sample B
Typical 10K spectrum of sample A was compared with that of sample B in Figure 3.5. An InGaN-related emission band was observed in each sample. The peak position in sample A and B was about 2.79 and 2.81 eV with a full linewidth at half-maximum (FWHM) of about 80 and 53 meV, respectively. It can be seen that the PL peak position of sample B blue shifts toward higher photon energy side while its FWHM of PL peak was smaller than that of sample A.
The In concentration in quantum well is assumed to fluctuate spatially, thus forming deep cusps or QDs-like regions in the energy gap. This kind of quantum well will not have a smooth, sloping band structure that was necessary for the quantum confined Stark effect (QCSE). The QCSE increases the separation of the electron and hole wave functions and reduces the emission intensity. The carriers will reside in localized states created by the fluctuations or QDs-like regions instead of separating to opposite sides of the well. Thus, the QCSE will be reduced or even eliminated in wells with large indium fluctuations or QDs-like regions. Moreover, the confinement provided by the regions mentioned above, which effectively forms QDs, can overcome the negative effect of the polarization field. As a result, a PL blue shift observed from samples B with a narrower linewidth might be due to the increase in indium fluctuations or QDs-like regions. Exciton pairs are confined in the local minima, and the cusps or QDs-like regions operate as excellent radiative recombination centers. According to the reports of other group, the InGaN/GaN MQWs with QDs-like regions in the well layers show high emission efficiency.
It is usually believed that the carrier localization mechanism is the key to the efficient photon emission in such a compound of relatively higher defect density (~108 cm-3). Typically, the process of aggregation and hence the effect of carrier localization becomes stronger with increasing average content. With the effect of localization, it has been widely observed that
PL spectral peak showed an S-shape variation with temperature [33, 36]. This temperature-dependence behavior originates from the localization of thermalized carriers and hence a blue shift of PL spectral peak in a certain temperature range. The same results will be shown later in my thesis.
3.4.2 Temperature Dependent of Photoluminescence 3.4.2.1 S-shifted Behavior
Figure 3.6 (a) and (b) displayed the temperature-dependent PL spectra of sample A and B in the range from 10 to 300 K. Both sample A and sample B, the PL line shape remained symmetric only up to 190 K, and above this temperature it became inhomogeneously broadened.
Generally, band-gap energies of semiconductors decreased with increasing temperature following the Varshni empirical equation [41]
2 where T temperature in Kelvin, Eg(0) the band gap at 0 K, and α and β known as Varshni’s fitting parameters. In an alloy, the emission line was redshifted with respect to the Varshni equation. However, anomalous temperature dependence had been observed in our InGaN LED structure as shown in Figure 3.7, where luminescence peak energies made a blue shift with respect to the values predicted by the Varshni’s formula. Eliseev et al. [42] reported that such blue shifting behavior can be interpreted by the effect of localized tail states assuming that the density of state (DOS) of excitons induced by potential fluctuation. This leads to a statistical distribution of excitionic transition energies, which are assumed to have a Gaussian distribution with a standard deviation σ. The dependence of σ on composition, x, is described by [43],
where γ is a factor smaller than one, accounting for the quantum mechanical averaging of the excitonic wave function. Vc(x) is the smallest volume in which a change in composition may occur and aex(x) is the exciton Bohr radius. When Vc(x) is the volume of the primitive unit
broadening is σ2/kBT. By introducing this term into Eq. (3-2), the redshifted peak position Epeak(T) versus temperature can be written as [43]
2 2
usually σ shows the energy representing the degree of localization due to the composition fluctuation and kB is Bolzmann’s constant. The value of σ is larger, the localization effect is stronger.
The values of σ were fitted to be 14.7 and 17.9 meV for sample A and B respectively, indicating that the localization effect of sample B was stronger than that of sample A. This results showed that a pulsed variation of δ-TMIn flow in the well layer can improve the localization effect in the InGaN/GaN MQWs structures.
3.4.2.2 Activation energy
Generally, the internal quantum efficiency can be evaluated by the temperature dependence of the integrated PL intensity [45, 46]. An Arrhenius plot of the normalized integrated PL intensity for the InGaN-related PL emission over the temperature range under investigation was displayed in Figure 3.8.
The analysis of these data had been carried out using well-known thermal activation relation [47]
where C the constant, Ea the activation energy, and kB the Bolzmann’s constant. At T > 80 K, the integrated PL intensity is thermally activated with an activation energy of about 35.3 meV and 42.8 meV, respectively, to samples A and B.
It has been suggested that the measured activation energy Ea in InGaN samples represents the localization energies of excitons, resulting from band edge fluctuations. Generally, the quenching of the luminescence with temperature can be explained by thermal emission of the carriers out of a confining potential with an activation energy correlated with the depth of the confining potential. So we can infer that the δ-TMIn flow rate resulted in higher localization effect owing to the formation of In-rich QDs structure.
3.4.3 Excitation Dependent of Photoluminescence
In most InGaN epilayers, the energies of photo- and electroluminescence transitions in
InGaN quantum wells exhibit a characteristic blueshift with increasing pumping power. This effect has been attributed either to band-tail filling, or to screening of piezoelectric field. And Figure 3.9 shows the excitation power-dependent of PL spectra of sample A and B respectively, and they were measured at 10 K. We didn’t observe the strong blue shift in power range from 1.74W/cm2 to 85 W/cm2 in both sample A and B. The strong piezoelectric field in InGaN epilayer comes from the large lattice mismatch induced strain between InN and GaN. And the formation of In-rich QDs can release parts of strain stored in InGaN LED structure. So there were no significant blueshift with increasing excitation power density as observed in previous report [11, 48]
3.4.4 Photoluminescence Excitation and Light Output Performance
Photoluminescence excitation (PLE) can provide the information of absorption spectrum and understand the energy state distribution. Figure 3.10 summarizes the PL and PLE of the InGaN/GaN MQWs of sample A and B at the temperature of 10 K. The PLE detection energy is set at the main InGaN-related PL peak.
In order to analyze the Stokes’ shift, which is defined as the difference in energy between the effective band gap and the emission peak energy, it is essential to have an accurate description of the absorption edge that includes the effects of broadening. A PLE measurement was performed to get the absorption edge. Martin et al. suggested that by fitting the PLE spectra to sigmoidal formula [49]
0
where α0 the constant, Eeff the effective band gap, and ∆E the broadening parameter which indicates a distribution of absorption states , and E the excitation energy at which the intensity of emission.
A large Stokes’ shift of PL peak with respect to PLE absorption peak was observed. The Stokes’ shift was often attributed to carrier localization in disordered systems, such as InGaAs/GaAs QDs or InGaN/GaN MQW structures. Stokes’ shift is 186 meV and 225 meV for sample A and sample B, respectively. And ∆E is the broadening parameter representing the degree of composition fluctuation. The larger value indicates the large inhomogeneity. The ∆E is 65 meV and 66 meV for sample A and sample B respectively. The localization within the
increasingly large dots, caused by fluctuations resulting form variation in dot size or shape.
The fabricated LED samples were tested for their light outputs as a function of injection current (L-I) as shown in Figure 5.11. As can see the emission power intensity of sample B is higher than that of sample A for overall driving-current range. At lower driving current of 20 mA, the sample B have a light output power of ~ 3.6 mW 16% greater than ~ 3.1 mW for the sample A. The enhancement of light output increases with the driving current up to ~24% at 60 mA
3.5 Summary
In summary, the effects of δ-TMIn-flow process with an initial fTMIn of 400 sccm during the well layer growth on the optical properties of InGaN/GaN MQWs were investigated. The HRXRD θ-2θ spectra and HRTEM images indicate the good layer periodicity and the structural quality of the InGaN/GaN MQW. And in both sample A and B, there exists the In-rich clusters in the InGaN/GaN MQW layers whether the δ-TMIn flow or not. But we can observe from PL spectra, the PL peak energy was different at 10 K even though the same composition extracted from XRD measurement. From the FWHM result of PL measurement, In-rich clusters were more uniform in size of sample as compared to sample A. And according to the PL and PLE measurement result, the larger values of σ, Ea and Stokes’ shift in sample B indicate that the δ-TMIn flow resulted in the increase the composition fluctuation in InGaN MQW region and shows the stronger carrier localization effect. And the light output of the GaN LEDs with the δ-TMIn-flow process is increased up to 24% without obvious deterioration of interfacial abruptness.
Figure 3.1 The structure of InGaN/GaN MQW.
Figure 3.2 Schematic diagrams of fTMIn variation over time in InGaN QWs for sample A
(a) (b)
(c) (d)
(a) (b)
(c) (d)
Figure 3.3 Cross-sectional TEM image of the InGaN/GaN MQWs of (a) sample A and (b) sample B. High magnification TEM image of the InGaN/GaN MQWs of (c) sample A and (d) sample B.
Figure 3.4 HRXRD spectra for (0004) reflection from the InGaN/GaN MQW structure of (a)
Figure 3.5 The normalized PL emission spectra of samples A (solid line) and B (solid circle) at 10K.
Figure 3.6 Temperature-dependent PL spectra of (a) sample A and (b) sample B.
Figure 3.7 The diagram of peak energy versus temperature of (a) sample A and (b) sample B.
Figure 3.8 Normalized PL intensity as a function of T-1 for (a) sample A and (b) sample B.
Figure 3.10 PL and PLE spectra of the InGaN/GaN MQWs of (a) sample A and (b) sample B at 10K.
Fig 3.11 L-I characteristics for the LEDs of samples A and B
Chapter 4 Optical Characteristic of GaN-Quantum-dots grown on AlN Nanoholes
Chapter 4 Optical Characteristic of GaN-Quantum-dots grown on AlN Nanoholes