Polarization effects of c-plane and a-plane InGaN MQWs

The 5-period GaN/InGaN MQW samples were simultaneously grown on a-plane GaN/r-plane sapphire and c-plane GaN/c-plane sapphire template layers in metal-organic chemical vapor deposition (MOCVD) reactor. Trimethylgallium, ammonia and disilane were the precursors used to GaN growth. The buffer layer was grown at 550˚C and the epilayer was grown at 1050˚C. After the deposition of GaN template structure InGaN MQWs were grown at 856˚C and 845˚C for two sets of samples. We took two sets of samples (A1, C1 and A2, C2) to compare their optical properties. Samples A1 and C1 are MQWs deposited on a-plane and c-plane GaN template at 856˚C. And A2 and C2 were grown on the other equitant condition 845˚C.

5.2.2 Room temperature photoluminescence spectra

The optical characteristic of the a-plane InGaN/GaN MQW samples were evaluated with a 24mW He-Cd laser operated at 325nm. The laser beam was collimated to a spot of 100um in diameter. Figure 5.2.1 shows the corresponding PL spectra at room temperature for each one. The dominant peak energy (wavelength) for sample A1, C1, A2, C2 is 3.089 eV (401nm), 3.153 eV (393nm), 3.028 eV (410nm) and 3.058 eV (405nm), respectively. The PL peak of those samples and growth temperature are listed in Table 5.2.1. PL spectra show the lower growth temperature, the longer PL peak wavelength. The reason was In tend to thermally diffuses out in higher growth temperature. The PL intensity of a-plane samples were one order magnitude lower compared with c-plane samples. In the PL spectra, the weaker PL intensity of sample A1 in compare of other c-plane samples reveals the presence of GaN emission at 3.4 eV. A-plane MQW exhibited lower intensity and broader half of full width.

Furthermore yellow band emission was observed especially for a-plane MQWs.

Those behaviors can be explained related to high density of structure defects and surface roughness [45]. In Figure 5.2.3 dislocations on sample A1 are seen as sharp-triangular lines propagating parallel to the substrate. These pits are the general from of V defects, which are observed in c-plane InGaN alloy customtly. For a c-plane film, the V defect is an open, inverted pyramid bound by the pyramidal [1010] facets. In the nonpolar orientation, V defect appears as a< defect.

5.2.3 Power dependent photoluminescence

In order to observe the influence of built-in electric field on a-plane and c-plane MQW, the power dependent PL measurement was demonstrated. Figure 5.2.4 shows RT PL spectra for sample C2 as shown in Figure 5.2.4 (a) and sample A2 as shown in Fig. 5.2.4 (b) of InGaN/GaN MQW from 32uw to 22mW. The PL peaks of sample C2

undergo a clear blue shift with increasing power density. But PL peaks of sample A2 exhibit less dependence on excitation power density. The power dependent PL peak shifts from 1.63 W/cm² to 293W/cm² are shown in Fig. 5.2.5. The total shifts of four samples are about 5.5 meV, 13.2 meV, 7.49 meV and 33.02 meV for A1, C1, A2 and C2, respectively. C-plane samples with obviously blue shifts are attributed to the presence of larger built in electric field as the best-known. In particular, a-plane samples show a little blue shift indicates the reducing built-in field existing in a-plane samples, e.g. growing along nonpolar direction and less influenced by internal polarization. The larger shifts in C2 than C1 and A2 than A1 are due to larger strain-influenced internal field. It is prospered that larger strain inducing field resulted from higher In composition. In addition, the excitation power dependence on luminance intensity is analyzed by room temperature dependent PL. Figure 5.2.6 shows the relation between excitation power and luminance intensity for both A2 and C2. In the case of c-plane MQWs, the superlinear relation (I~P^1.446) is demonstrated;

whereas in the case of a-plane sample, the luminance intensity increase linearly with power density (I~P^1.078). The superlinear relation is due to decreasing nonrecombination process screen with photo-injected carriers. In other words, the total internal field is reduced by accumulating carriers in one side of triangular potential when we increase excitation power density which is well know as screen effect. Therefore, it can be concluded that nonrecombination resulted from QCSE play a minor role in nonpolar structure.

5.3 Investigation of localization effect in a-plane MQW

5.3.1 Low temperature and temperature dependent photoluminescence

To further study the emphasis of localization effect on our samples, temperature dependent PL was applied. The low temperature PL spectra of four samples at 20k are

shown in Fig 5.3.1. The spectra show two separated peaks for a-plane MQWs obviously. Two peaks for A1 and A2 are at 3.189, 3.295 and 3.138, 3.246, respectively. The observation of separated two peaks is suggested due to localized tail states. Typically, so-called localized states originated from spatial In segregation. For III-V semiconductor, the recombination of localized exciton at the potential minima is responsible for the luminance mechanism. In InGaN alloy, the possible origin of the In rich disorder are likely due to the defect-resistance nature [46]. The defective structures consists V-defect structure with threading dislocation (TD). They are observed associated with In-rich cluster witch is responsible for the localization effect.

The evidence and behavior of exciton localization in our investigation will be discussed later.

Fig 5.3.2 illustrates the temperature dependent PL spectra. At 20K, the two peaks reveal almost equal intensity. However, the high energy peak exhibits remarkable thermal quenching. After 160K, it hardly can be recognized, but lower energy peak dominate the luminance until room temperature. In general case, the temperature dependent quench in luminance can be explained by thermal emission of the carriers with activation energy out of the confining potential. Since activation of high energy peak is measured about 27meV and much less than the band offsets of conduction or/and valance band, the thermal quenching effect is not resulted from the carriers run out of the well into barrier. Otherwise, the most possible reason related to the luminance mechanism is originated from potential fluctuation. The higher energy peak is suggested due to the localization state with deep potential. At low temperature, excitons are randomly trapped by the spatial localized state. As increasing temperature, the thermally activated excitons obtain energy then delocalizated. Finally, they can be trapped by the nonrecombination center during transfer or relax in deeper potential minima. As a result, the high energy peak experience fast quenching and low

energy peak contribute the dominant luminance at high temperature as we observed.

5.3.2 Photoluminescence Excitation

Photoluminescence excitation (PLE) can provide us the information of absorption spectrum and understand the distribution of energy state. Our PLE spectra along with PL measurement of four samples are shown in Fig.5.3.3. The measurements are carried out at 20K in order to get efficient luminance. The dispersed light comes from the monochrometer in front of a Xe lamp with continuous radiation. All of the PL spectra are measured by the excitation of dispersed light at 325nm come from Xe lamp. Acquire for the quantitative comparison, all of the PLE spectra are normalized to unity at their maximum. In order to analyze the Stokes’ shift 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 the sigmoidal formula

Eq. (5.2.2)

where I0 is the maximum intensity, Eeff is the effectve band gap, ΔE is the boarding factor which is related the distribution of absorption state. In Fig 5.3.3, a sharp peak for each sample at about 3.49eV is the signal from GaN barrier, clearly seen for all the PLE spectra. As can be seen at low temperature PL spectra exciting by He-Cd laser, two separated peaks also presented at the excitation of Xe lamp. Now we focus on the low energy peak because it dominant in the luminance mechanism from 20K to 300K.

The experimental and fitting results are list in table 5.3.1. The stoke shift of A1, C1, A2 and C2 are 236, 341, 224 and 275meV, respectively. The Stokes shift in the InGaN QW is resulted from the separation of wavefunction overlap due to the built-in

internal field and the presence of localized states [47-48]. From the above observation, the power dependent experiments indicate the influence on built-in internal field can be neglected in a-plane QW. Furthermore, in the same growth condition, the observations show that a-plane sample exhibited larger stokes shift than c-plane sample. This phenomenon indicates obvious localization effect in a-plane samples and the result may be corresponding to the behavior of defect associated with In rich disordered clusters reported for polar [0001] InGaN/GaN QWs [46].

5.4 Conclusion

In conclusion, the optical properties on polarization effect of a-plane InGaN/GaN MQWs and c-plane MQWs were investigated. Compare with c-plane MQWs, a-plane MQWs exhibited less power dependent blue shift due to not effected by quantum confine stark effect (QCSE). Moreover, the power dependent luminance intensity was demonstrated, too. The superlinear relation (I~P^1.446) of c-plane MQWs revealed the recombination efficiency increase by gradually screen of the bending potential. But the proportion relation (I~P^1.078) in a-plane sample indicated the flat band structure witch agrees with the absence of the build in field along the direction of QWs.

Finally, the localization effect of a-plane MQWs was investigated by temperature dependent PL and PLE experiment. We observed two separated peaks at 20k in a-plane sample; whereas only one peak in c-plane sample. In a-plane sample, high energy peak experience dramatic thermal quenching but low energy peak kept the luminance until room temperature. This temperature dependent experiment indicated the low energy peak result from the localization state with stable luminance efficiency.

Analysis of PLE experience show the apparent stoke shift also conclude the obvious localization in a-plane MQWs.

2.4 2.6 2.8 3.0 3.2 3.4 3.6

Fig. 5.2.1 Room temperature PL of sample A1,C1,A2 and C2

A1

Table 5.2.1 List of PL peaks and the growth temperature

Fig. 5.2.2 : SEM image of sample A1

2.4 2.6 2.8 3.0 3.2 3.4 3.6

Fig. 5.2.3 Excitation power dependent PL of (a) sample C2 and (b) sample A2

1 10 100 1000

Fig. 5.2.4 Dependence of peak energy of the PL spectrum on excitation power density (a) A1/C1 and (b) A2/C2

0.1 1 10 100 1E-4

1E-3 0.01 0.1 1 10 100

c2 a2

I~P

I~P

Intensity (a.u.)

Power density(W/cm²)

Fig. 5.2.5 PL intensity dependence on excitation power density of a-plane(A2) and c-plane sample(C2)

2.8 3.0 3.2 3.4 3.6 0.0

0.2 0.4 0.6 0.8 1.0

A1 C1

Normalized intensity (a.u.)

Photon energy (eV)

2.6 2.8 3.0 3.2 3.4 3.6 3.8

GaN A2

C2

Intensity (a.u.)

Photon energy (eV)

Fig. 5.3.1 Low temperature PL spectrum at 20 K (a) A1/C1 (b) A2/C2

Fig. 5.3.2 Temperature dependence PL spectrum at 20 K

2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2

Fig 5.3.3 PL and PLE spectra of a-plane and c-plane InGaN/GaN MQWs

Table 5.3.1 Experimental and fitting factors of PLE measurement of a-plane and c-plane InGaN/GaN MQWs

A1

A2

C1

C2

3.210

3.088

3.168

3.036

14 33

61 65 3.446

3.429

3.392

3.311

236

341

224

275

PL peak (eV) Eg (eV) Stoke shift (meV) E (meV)

Chapter 6 Trenched laterally over-grown a-plane GaN with low dislocations density

6.1 Introduction

The quantum efficiency of light emitting diodes (LEDs) and laser diodes (LDs) based on group-III nitride heterostructures is reduced due to the presence of built-in electric fields. The polarization-related electric fields contributed by the spontaneous and piezoelectric polarizations cause electron and hole separation in quantum wells that lead to inclining of band structure, poor recombination efficiencies, reduced oscillator strength, and red shift in emission wavelength. The spontaneous and piezoelectric polarizations paralleled to [0001] c direction of GaN based devices are caused by the arrangement of atoms and strain in MQW, respectively. As a result, the performances of III-nitride devices are limited by the polarization-related internal electric fields.

Without polarization effects, non-polar GaN is currently the subject of intense research due to the potential to improve the internal quantum efficiency (IQE) of GaN optoelectronic devices. To eliminate such polarization effects, growth along nonpolar orientations has been respectively explored for [1120] a-plane GaN on

r-plane sapphire [49] and a-plane SiC, [50] and ]

2 1 10

[ [1010]m-plane GaN on [100]

LiAlO2 substrates [51-52]. According to the recent studies of a-plane and m-plane AlInGaN based quantum wells demonstrate that it is possible to avoid such polarization fields effect by growing device structures along these nonpolar orientations.

However, nonpolar a-plane GaN base material grown on r-plane sapphire substrates which always accompany with a wavy, stripe-like growth feature possess a large density of threading dislocations and stacking faults. It is as result of the serious

aniostropic in-plane strain between different crystal axis [13]. Recently, successful epitaxial lateral overgrowth (ELOG) of a-plane GaN on r-plane sapphire has been reported. ELOG not only improves significantly the material quality by reducing the density of threading dislocations but also alleviates the strain-related surface roughening and faceting [14]. Despite the ELOG assisted morphology and quality improvements in a-plane GaN over r-plane sapphire, the study of the epilayer quality and dislocations distribution in the ELOG epilayer is quite not lucid. In this letter, we successfully improve [1120] a-plane GaN quality by using TELOG and the optical and structural properties is presented explicitly.

6.2 Sample preparation

Fig.6.2.1 shows the flow chart of the process sequence of the a-GaN template and subsequent TELOG. At first, the a-plane GaN templates with 1.5μm thickness were grown with low pressure-metalorganic vapor phase epitaxy (LP-MOVPE) on r-plane Al O2 3 substrates using conventional two-step growth technique. After a serial of conventional photolithography techniques, a 2 μm/18 μm (window/wing) TELOG stripe pattern orientation was chosen parallel to the [1100]direction to realize vertical c-plane sidewalls. Mask patterning was followed by etching of SiO2 using inductively coupled plasma etching through the windows to the GaN epitaxial film.

GaN stripes were etching through the mask openings, down to the r-plane sapphire substrate, thus forming Ga-face [0001] and N-face[0001]planes on the sidewalls and exposed r-plane sapphire at the bottom of the trenches by reactive ion etching. To simplify the growth process, the SiO2 mask was removed by hydrofluoric acid and followed by depositing a-plane GaN TELOG film using single-step growth process.

The growth condition in this study were described as follow: growth temperature

~1190 C low pressure and low V/III ratio, ~700-800. o

6.3 SEM image

To observe the growth behavior and mechanism, we stop the process before coalescence. As shown in Fig.6.3.1, the Cross-sectional SEM images of TELOG by MOCVD, the Ga-face growth rate was faster than N-face growth rate with a thin GaN layer about 0.2μm grown on the bottom of the trenches. The thin GaN layer will influence the quality and growth of TELOG. Furthermore, the influence by the thin GaN layer to Ga-face is more obvious than N-face so that the ratio of Ga-face growth rate / N-face growth rate is about twice, not as high as the reports of UCSB, an order of magnitude [15].

6.4 High-resolution X-Ray measurement

High-resolution X-Ray rocking curves, as shown in Fig.6.4.1, revealed that the a-plane GaN templates posses a serious anisotropic structural characteristics: the

FWHM of as-grown a-GaN 1.5 μm bulk layer in [1100] direction is almost twice as large as that in [0001] direction. It shows that the strains between the different crystal axis, for instance c-axis and m-axis, are quite different and enhance the formation of line defect. Moreover, the surface geometry becomes a wavy, stripe-like growth feature while the nucleation layer was not optimized or the thickness is thicker and thicker as result of some pits on the thin film. However, after lateral overgrowth, the stresses of TELOG layer were released in both c-axis and m-axis and thus the crystal quality was enhanced especially in the [1100] direction so that the FWHM was reduced from 1810 arcsec to 351 arcsec. On the other hand, since the strip of the TELOG layer did not coalesce, we observed obvious wing tilt phenomenon along

[0001] direction which leads to the X-Ray rocking curve broadening effect. Unlike the symmetric wing tilt in c-plane TELOG GaN [53-54], the wing tilt in a-plane TELOG GaN is asymmetric so that the X-Ray rocking curve shows the asymmetric profile. This results in asymmetric overgrown wings[55] which possess the different lateral growth rates of the window GaN in the [0001] and the [0001] directions.

Although the FWHM was only reduced from 973 arcsec to 816 arcsec, the quality would be better while the TELOG film fully coalesced and without wing tilt phenomenon.

6.5 Micro-photoluminescence measurement

This phenomenon was confirmed by micro-photoluminescence (μ-PL) of a-plane TELOG. We use a scanning near-field optical microscope (SNOM) to scan a 25

μm×25 μm confocal image. The sample is excited by a He-Cd laser operating on 325 nm with 25 mW. Using a 40×objective, He-Cd laser is focused into a spot with 1 μm in diameter on sample. The photoluminescence is collected in a fiber with 25 μm in the diameter and detected by a Photo Multiplier Tube (PMT). Furthermore, μ-PL spectra is dispersed by a 320mm monochromator (Jobin-Yvon Triax 320). The wavelength resolution is about 1nm by using 300 grooves/mm grating and the slit of 0.1 mm. Although the Fabry-Perot effect make the multi peaks be found in the μ-PL spectra, we still can ascertain the quality of GaN by the μ-PL intensity. As shown in Fig.6.5.1, five different regions were easily distinguished in the TELOG sample.

Comparing the μ-PL data with SEM image, the stripped a-GaN seed stands at region 2 and the best crystal quality area, region 1, the strongest μ-PL intensity area, stands at the N-face GaN wing. On the contrary, the worst crystal quality area, region 5, the lowest u-PL intensity area, stands at the un-coalesced widows region so that a-GaN

can grow on r-sapphire without nucleation layer but has a very texture surface and not thicker than 0.3 μm as Fig. 2 shows. Another good quality area, region 3, stands at the initial region of the Ga-face GaN Wing, region 4. We suggest that the crystal quality of Ga-face GaN possesses higher growth rate easily is effected by the thin a-GaN layer grown in the un-coalesced windows but not obviously at the beginning of re-growth while the thin a-GaN layer was not formed.

6.6 Cross-sectional transmission electron microscopy (TEM) measurement

The distributions and types of dislocations were investigated by cross-sectional transmission electron microscopy (TEM) shown as Fig.6.6.1. According to the g=(1120) and g=(0002) two beam bright field images, most of the threading dislocations are in contrast in both of the g=(1120) and (0002) two beam condition, indicating that these dislocations are mixed a+c type dislocations. The distribution of dislocations fully consists with the u-PL results. The threading dislocations densities (TDD) of stripped GaN seed were more than 1×10¹⁰ cm⁻². TDD of Ga-face GaN wing which was influenced by the thin layer is about 3×10⁸ cm⁻² and TDD of N-face GaN wing about 3×10⁷ cm⁻², three orders of magnitude lower than planar films, is not influenced as obviously as Ga-face GaN wing. This is a stronger evidence to support the suggestion that Ga-face GaN is easier to overgrowth and is influenced by the thin GaN layer on the bottom of trench. As a result, a better quality a-plane TELOG GaN can be established on a pattern into the sapphire surface over 0.3 μm depth.

6.7 Cathodoluminescence (CL) image

To realize the relationship of optical property and TDD, the Cross-sectional Cathodoluminescence (CL) image of a-plane TELOG was photographed, shown as

Fig.6.7.1. The CL intensity distribution is almost the same as the results investigated by TEM. So that the dislocations still perform as the non-radiative center in a-plane GaN film and become the principle problem what should be solved immediately. According the above results, we perform the same process with a 3 μm/7 μm stripe pattern, shown as the Fig.6.7.2. The coalescence process can fully be completed below 10μm, it is a useful technique to overcome the thickness problem

Fig.6.7.1. The CL intensity distribution is almost the same as the results investigated by TEM. So that the dislocations still perform as the non-radiative center in a-plane GaN film and become the principle problem what should be solved immediately. According the above results, we perform the same process with a 3 μm/7 μm stripe pattern, shown as the Fig.6.7.2. The coalescence process can fully be completed below 10μm, it is a useful technique to overcome the thickness problem