Chapter 3 Experiment setup & Theory
3.4 Raman spectroscopy
Raman spectroscopy is a spectroscopic technique used to study the phonon modes of material. When laser light incident into the sample, photons will collide with atoms in material. For elastic collide, the energy of photon will not be changed, and the frequency of photon is same, this effect is called Reyleigh scattering. But for inelastic collide, the energy of photon will exchange with it of atom, then the frequency of photon will be changed, this effect is called Raman scattering. In the case of Stokes scattering, the photon losses energy.
On the contrary is called anti-Stokes scattering. The diagram of the Stokes scattering and the anti-scattering is shown in Fig. 3.4.1. The Stokes scattering is usually much weak than the anti-Stokes scattering, so Stokes scattering is usually observed.
Raman spectroscopy measures the change of photon energy called Raman shift (cm-1), which is expressed as below:
hC
∆E
=
−
=
∆σ σ σ' (3.4.1)
where σ and 'σ are the energy of incident photon and scattering photon, respectively.
From the Raman spectra, the phonon vibration modes, composition, stress, and quality of material can be study.
The Raman-active modes scattering efficiency is given by:
2 i
Sαεs•χ•ε (3.4.2)
where εs and εi denote polarization for the incident and scattering light, respectively, and χ is the Raman tensor of the scattering process. From above equation we can know the
Raman active modes is related to the polarization, incident direction and scattering direction for incident and scattering light. Table 3.4.1 show the corresponding confiquration for GaN material. For a(b,c)d shown in Table 3.5.1, the a and d are direction for incident and scattering light, respectively. And b and c are polarization for incident and scattering light, respectively.
Fig. 3.4.2 shows the experimental setup schematically. The Raman scsttering was measured with an Ar-ion laser (Coherent INNOVA 90) as an excitation source with a wavelength of 488 nm. The scattered light was collected by a camera lens and imaged onto
the entrance slit of the Spex 1877C. After reflected by two mirrors, the laser light was focus by a lens which focal length is 15 cm, and the scattering signal was collected into monochromator (Spex 1877C) by two convex lens. The probed light was dispersed by monochromator (Spex 1877C) equipped with 1200 grooves/mm grating. The Raman scattering is detected by cooling CCD with temperature of 150 K.
3.5
3.6
Atomic force microscope (AFM)
Atomic force microscope (AFM) images the surface of sample by scanning a sharp tip (10 µm in long and small than 10 nm in diameter) over and measuring the deflection of the tip.
The fundamental setup of AFM is illustrated in Fig. 3.5.1. A piezoelectric controller moves the sample in x-y direction under the tip. The position of the tip is measured by reflecting laser from the backside of the cantilever to a split photodiode. Depending on the distance between the tip and the sample so that the force acting on the tip is repulsive, the AFM work in contact mode. In non-contact or taping mode the tip is further away from the sample and in the region of an attractive force [70]. The cantilever is set to vibration close to its resonant frequency and change in the surface morphology lead to changes in the frequency, which can be measured sensitively.
High resolution X-ray diffraction (HRXRD)
X-ray diffraction is a non-destructive method to study the structural properties of
material, which can be used to judge the crystal quality, thickness and the composition of compound semiconductor. And the degree of lattice mismatch between an epitaxial layer and the substrate also can be charactized. XRD utilizes two successive Bragg reflections from two independent crystal planes. From Bragg’s law, the condition for constructive interference is:
λ
θ n
dhklcsin B =
2 (3.6.1)
where dhklc is the reciprocal lattice spacing and can be expressed as
2
where a is the crystal lattice constant, (h, k, l, c) is known as the Miller indices of the plane,
θB is the incident angle, n is an integer representing the diffraction order, and λ is the wavelength of the incident radiation. For a fixed incident X-ray wavelength λ, each crystalline material has a characteristic X-ray diffraction pattern associated with it, which yields a accurate information on its crystal structure and lattice space. The ω, ω-2θ scan and reciprocal space mapping (RSM) measurements of high resolution x-ray diffraction (HRXRD) made with a Philips MRD X’pert PRO diffractometer using CuKα1 radiation were applied to analyze the crystal quality and strain in InGaN QW.
E c
E D
E A E v
Fig. 3.1.1 Radiative recombination transitions in semiconductor.
Fig. 3.1.2 Setup of temperature dependent PL and TRPL system.
Fig. 3.3.1 Setup of µ-PL system.
Table 3.4.1 Selection rules of wurtzite structure
Fig. 3.4.2 Setup of Raman system.
Fig. 3.5.1 Setup of AFM
Chapter 4 Optical properties of InGaN/GaN multiple quantum well light emitting diodes grown on sapphire substrate with different misorientation angle
4.1 Introduction
Generally, GaN-based optoelectronic devices grown on c-plane (0001) sapphire substrate, but due to the large mismatch of lattice constant between GaN and sapphire, huge density of threading dislocations up to 109 cm-2 present in nitride layers, deteriorating the surface morphology and crystal quality of GaN and InGaN active layers. To obtain high-quality LDs, a highly uniform and smooth surface of the expitaxal GaN film is required, because surface macrostep patterns sometimes produce a large propagation loss. Surface morphology of the epitaxial GaN film depends on various growth conditions, such as nitridation of the substrate [45], growth conditions of the buffered layer [5][46], ramping conditions before GaN epitaxial growth [47][48]. Some groups have studied the influences of using sapphire substrate with slight misorientation toward [1120] direction on growth of GaN [14][15][16][17]. (Diagram of sapphire substrate with misorientation angle toward [1120] direction is shown in Fig. 4.1.1) From AFM and XRD results, they found that the surface macrostep patterns of GaN greatly depend on the slight misorientation angle of the sapphire substrate. Moreover, the improvement of the crystal quality and the surface morphology of GaN by using slight misorientation angle have been observed. T. Yuasa et al. proposed these are due to appropriate atomic steps of the substrate formed by the slight misorientation, then
relieve the lattice mismatch between GaN and sapphire [17]. However, the optical properties of InGaN MQW LEDs grown on sapphire substrate with different misorientation angle have not been studied so far. The influences of misorientation angle on optical properties of InGaN QW are still unclear.
In this work, the optical properties of InGaN/GaN LED using sapphire substrate with misorientation angle of 0o, 0.2o, 0.35o and 1o toward [1120] direction have been studied.
From temperature dependent PL measurement, the anomalous emission energy shift as function of temperature, which is so-called S-curve, has been observed [18]. By using Eliseev et al. proposed model [19], the degree of potential fluctuation of the InGaN as function of misorientation angle has been characterized. Moreover, the recombination dynamic of carriers and the dimensionality of nanostructure in active region have also been studied by temperature dependent TRPL measurement.
On the other hand, the material properties of samples have been investigated by atomic force microscope (AFM), high resolution transmission electron microscopy (HRTEM), high resolution X-ray Diffraction (HRXRD) and Raman spectroscopy.
4.2 Sample preparation
The sample structure in this study is shown in Fig. 4.2.1. The InGaN/GaN MQW LEDs in this work were grown by a commercial LP-MOCVD system with a vertical reactor. The
MO compounds of TMGa, TMIn, TMAl and gaseous NH3 were employed as the reactant source materials for Ga, In, Al and N, respectively, and H2 and N2 were used as the carrier gas.
All samples were grown on c-plane (0001) sapphire substrate with different misorientation angle of 0o, 0.2o, 0.35o and 1o toward [1120] direction, consisting of 50-nm-thick AlN nucleation layer, a 3.5 µm Si-doped n-type GaN, and an unintentionally doped active layer with InxGa1-xN/GaN MQWs, and 0.4 µm Mg-doped p-type GaN. The doped concentration of n- and p-type GaN is nominally 5 x 1018 and 1 x 1019 cm-3, respectively. For conventional MQW LEDs, the MQWs layer comprise six periods In0.2Ga0.8N well (~ 3 nm) and GaN barrier (~ 14 nm).
After epitaxial growth, the fabrication of devices was carried out as follows. First, the top ohmic contacts of In0.2Ga0.8N/GaN MQW LEDs were formed using a metal system of Ni/Au for p+-GaN contact layer. Next, Ti/Ai systems were evaporated onto the exposed n-GaN epilayer. Finally, In0.2Ga0.8N/GaN MQW LEDs were cut into square pieces with the dimension of 300 × 300 µm2.
Material properties of InGaN/GaN MQW LEDs
4.3
4.3.1 Atomic force microscope image
The AFM images (5µm x 5µm) of the p-type GaN are shown in Fig. 4.3.1 and the root mean square (RMS) of surface roughness measured by AFM is shown in Fig. 4.3.2. From
images we can observe that the surface morphology becomes smoother as misorientation angle increases from 0o to 0.2o, but it is deteriorated as misorientation angle increase above 0.2o. Moreover, except the 0.2o sample, we can see some pits formed in p-type GaN and become more deteriorative in the angle above 0.2o. The formation of pits on p-GaN is due to large dislocation density produced from underlying layer. The pits will degrade the surface morphology and causes rough surfaces with higher RMS values.
4.3.2 High resolution transmission electron microscopy
In order to observe the homogeneity of InGaN/GaN MQWs and the dislocation density in the device, the structural analysis was performed by means of cross-section HRTEM. Fig.
4.3.3 is the cross-section HRTEM observation of the LED grown on sapphire substrate with misorientation angle of 0o, 0.2o and 1o. The n-type GaN, p-type GaN and six QWs can be clearly identified in the images. For 0o and 1o sample, we can see that many dislocations produced from underlying GaN layer penetrate through the InGaN/GaN MQWs and reach upper p-type GaN layer. The dislocation density extracted from TEM image is shown in Fig.
4.3.5. By using misorientation angle of 0.2o, the producing of dislocations is suppressed and the quality of device is improved. About its origins, some articles have proposed that the growth mode of GaN will be shifted from spiral-dominated to step-flow as sapphire substrate with slight misorientation angle about 0.2o [16]. From Burton, Cabrera and Frank (BCF)
theory, spiral growth is dislocations controlled, and it competes with step flow growth on a misorientation substrate [53]. The condition for spiral annihilation can be simplified to:
S
MC ω
ω < (4.3.1)
where ωMC is the misorientation terrace width which decrease with increase of misorientation angle; ωS is width of the terrace on the two interlocking spiral ramps that create a spiral hillock in GaN. When the misorientation angle increases to satisfy the above condition, the misorientation step would annihilate the spiral steps faster than the spiral could spread on the surface. Therefore, due to suppression of spiral growth, the GaN grown on misorientation angle of 0.2o can effectively decrease dislocation produced in underlying GaN layer. On the other hand, the incorporation of group III atoms, the epitaxial growth of III-nitrides, is closely related to the density of dangling bonds at step edges [72]. Therefore, when the misorientation angle is too large, the density of step edges is too high and it becomes difficult to grow with smooth surface, and the strain relaxation in the interface increases, resulting in increasing of dislocation density.
The Fig. 4.3.4 shows the HRTEM images for three samples, the six pairs InGaN/GaN QWs are clear observed, and the well and barrier layer thickness are estimated to be about 3 nm and 14 nm, respectively. In the detailed structure of the MQWs, a lot of dark spots are observed in the wells, the diameters of these dark spots are about 5 nm. This dark spots have
been observed for several articles [26][68], which is the indium-rich clusters due to large discrepancy in atomic size between indium and gallium and the large lattice mismatch of 11%
between InN and GaN. Besides, we also could see that thickness fluctuation existing in MQWs. And from figure we can see the composition and thickness homogeneity of MQWs is improved as miorientation angle increases from 0o to 0.2o, but it deteriorated as misorientation increases above 0.2o.
There are many reports said that indium incorporation is dislocation sensitive [54][71].
From Fig. 4.3.3, the dislocations penetrating through the InGaN/GaN MQWs structures could significantly increase the indium compositional fluctuation around the dislocation. The presence of dislocation in the underlying MQW layer strongly effects the indium incorporation into the upper MQW layers.
Moreover, the indium incorporation is also significantly sensitive to the surface morphology of an underlying layer. The article indicates that the indium incorporation efficiency is drastically higher on the facets than on a smooth surface [55]. The deposition of the indium atoms is attracted at facets rather on a smooth surface to reduce the surface energy due to many dangling bonds at facets. Therefore, the indium phase separation in InGaN QWs layer can easily occur on the rough surface of epilayer. The dislocations will also degrade the surface morphology and cause rough surface with higher RMS values. Therefore, due to fewer dislocation and smoother surface of underlying GaN layer for 0.2o sample, the
homogeneity of InGaN/GaN MQWs is improved.
On the other hand, the dislocations penetrate into p-type GaN will degrade the surface morphology and cause rough surfaces with higher RMS values, which may result in a depletion of Mg from the surface and cause a reduced doping level, deteriorating the contact resistant for InGaN/GaN MQWs LED [56][57]. Moreover, the high dislocation density will induce anomalous high leakage current, then, the performance of LED is degraded [69].
4.3.3 High resolution X-ray diffraction
HRXRD rocking curve (XRC) measurement for symmetrical GaN (002) reflections was applied to evaluate the crystal quality of GaN thin films. The FWHM of XRC for GaN (002) was plotted as shown in Fig. 4.3.6. We can observe that the FWHM is decreased as misorientation angle increases from 0o to 0.2o, but it increases as misorientation angle increase above 0.2o. The FWHM of the symmetrical GaN (002) is correlated to the threading dislocations density [49]. Therefore, it is ascertained that dislocation density tends to decrease as sapphire substrate with slight misorientation angle about 0.2o, but the crystal quality is deteriorated as misorientation angle increases above 0.2o, which is agree with our HRTEM analysis results.
Fig. 4.3.7 shows the HRXRD ω-2θ diffraction pattern for the LEDs with misorientation of 0o, 0.2o, 0.35o, and 1o, respectively. As can be seen, the HRXRD diffraction pattern shows
one periodical structures, which can be attributed to the In0.2Ga0.8N/GaN MQWs. In addition, the third order satellite peak of the diffraction pattern for all samples can be clearly observed, suggesting that the grown structure has sharp interfaces and good periodicity between GaN barrier and InGaN well. Additionally, the inset shows the variation of MQWs interface roughness (IRN) with misorientation angle of sapphire substrate. The interface roughness value of MQWs can be obtained from the following equation [63].
⋅Λ
where n is the order of the satellite; Λ and σ/Λ are thickness of the period and IRN, respectively; ∆θM is the angle distance between adjacent satellite peaks; W0 and Wn are the FWHM of zeroth- and nth-order peaks, respectively. The results are shown in Fig. 4.3.7. From figure we can found that IRN of InGaN/GaN MQWs decreases with increasing misorientation angle of substrate, and reaches minimum which is about 1.35% for 0.2° misorientation sapphire substrate was used. Increasing the misorientation angle to 1° increases the IRN of MQWs. The previous reports indicated that the IRN of MQWs is affected by defect, microstructure and phase separation in MQWs [63]. Based on x-ray analysis, we can prove that the defect density of 0.2o sample is lower than other samples, and can also concluded that defect density of 0.2o sample was reduced dramatically, which are consistent with the results observed from HRTEM images.
Fig. 4.3.9 shows the RSM results for InGaN/GaN MQW LEDs grown on (a) 0°, (b) 0.2°,
(c) 0.35° and (d) 1° misorientation sapphire substrate, respectively. The diffraction pattern was obtained from the asymmetric (1015) reflection of the GaN epilayer. Compared with sample 0o, 0.35o and 1o, the reciprocal lattice point (RLP) of GaN layer of 0.2o sample is circular, indicating that the spread of RSM intensity for 0.2o sample is narrower than that of RSM intensity for other samples. In general, the spread of diffraction intensity is related to the orientation distribution and decrease in the coherency of the scattering along the structure [64].
The asymmetrical feature for RLP of GaN indicates the out-of-plane lattice mismatch formed in GaN layer. This means that the GaN epilayer grown on 0.2° substrate has better crystallization quality. Besides, the RLP of satellite peaks can be clearly observed for 0.2o sample and the satellite peaks up to four orders, while the reciprocal lattice point of satellite peaks became relatively weak for 0o sample and 0.35o sample or vanished for 1o sample, suggesting that the InGaN/GaN MQWs of good quality were grown. Therefore, based on the above-mentioned results, the crystalline quality of 0.2o sample is relatively good in contrast to the quality of other samples, and the noticeable improvement in the quality of GaN epilayer could be responsible for achievement of high quality InGaN/GaN MQWs.
On the other hand, the overall strain state of MQWs with respect to the GaN epilayer can be examined by RSM. S. Pereira et al. proposed the method to investigate the strain and composition distributions within wurtzite InGaN/GaN layers [50]. The information can be extracted from the elongation of broadened RLP in asymmetric x-ray reflections. The
schematic diagram illustrating the effect of strain and composition gradients in the symmetric and asymmetric RLP of InxGa1-xN is shown in Fig. 4.3.10.
The RSM pattern in the Fig. 4.3.9 represent the Qx and Qy position of GaN and MQWs satellite peaks. The distance between GaN epilayer and MQWs satellite peaks in the Qx
direction indicates the degree of lattice relaxation, εxx, and the relationship between εxx and Qx
for all samples can be described by the following equation [73]:
−1 layer and InGaN layer, respectively. Hence, the degree of lattice relaxation for 0
GaN
Qx QxInGaNs
o, 0.2o, 0.35o
and 1o sample is about 1.28×10-4, 7.4×10-5, 8×10-5 and 3.64×10-4, respectively. We can find that the 0.2o sample have the smallest degree of lattice relaxation. For heteroepitaxial structure, the defect density, such as threading dislocation, V defect and so on, of MQWs will increases with increasing the lattice relaxation degree of MQWs. In other words, the InGaN/GaN MQWs of 0.2o sample have the lowest defect density compared with other samples. This think can be proved by our AFM and HRTEM images, the dislocation density decreases as misorientation agnle increases from 0o to 0.2o, but it increases as misorientation angle increases above 0.2o.
4.3.4 Raman spectroscopy
To determine the stress for the LED grown on different misorientation angle, the Raman spectroscopy was performed. In this study, we chose z(y,y)z configuration to probe Raman spectra. For this scattering configuration, the A1(LO) and the E2(high) phonon modes are observed, which correspond to the lattice vibration paralleling and perpendicularing to the c-axis [0001] direction, respectively. Fig 4.3.11 shows the Raman spectra for the InGaN/GaN MQW LEDs grown on different misorientation angle. The two clear active phonon bands are observed at around 568 cm-1 and 735 cm-1, corresponding to the E2(high) and A1(LO) phonon modes of GaN [51], and no InGaN relatived phonon bands are observed in the figure. The E2(high) mode is sensitive to changes in the elastic properties of the material, which is generally observed to estimate the biaxial stress in material. From inset of Fig. 4.3.12, we can observe the position of E2(high) peak shifts slightly with increasing misorientation angle. To extract the accurate position of E2(high) peak, the Lorentzian function was used to fit E2(high) mode peak, the results are shown in Fig. 4.3.12. We can observe that the Raman shift increases as misorientation angle increases from 0o to 0.2o, but it decreases as misorientation angle increase above 0.2o. In general, the Raman shift as a function of the biaxial stress can be expressed as:
σ ω
ω
ωE = E − =C
∆ 2 2 0 (4.3.4)
where ωE2 and ω0 are the Ramsn shift of E2(high) mode for strained sample and strain free
where ωE2 and ω0 are the Ramsn shift of E2(high) mode for strained sample and strain free