Power dependent PL measurement

4-4.1 Carrier localization effect

Due to composition inhomogeneity and monolayer thickness fluctuation Of InGaN QWs self-organized In-rich region is generated in InGaN active region, resulting in potential fluctuation of energy bandgap. AS shown in Fig.4.4.1.1 As injected carrier density increases further, an occupation of high energy stats of localized centers will be enhanced. And the band filling effect will make the carriers more easily escape from localized states to extendend states which decrease IQE.

From the Fig 4.4.1.2 we could found that LED without prestrain layer structure has much higher FWHM under the same carrier density. At carrier density above 1x10¹⁸(#/cm³), the FWHM increase fast and the band filling effect is start to dominate.

From Fig 4.4.2, we can see sample III has larger FWHM than others, which means In distribution is the most inhomogeneous and more nonradiative center in quantum well.

On the other hand, the LT-GaN and InGaN/GaN prestrain sample has similar band resulting a decreasing in degreed of wave function overlap which is called the QCSE.

The internal electric filed in the QW can be screened by photogenerated carriers.

Consequently, the QCSE effect become weaker when the carrier density increased, resulting in the emission peak wavelength buleshift and IQE enhanced at low injected carriers region. . Figure 4.4.2.1 shows the emission peak energy under different input power. When the input power was increase, the peak energy of the emission spectra of the LEDs changed to larger energy. The blue shift of three sample with increasing

input power before 10mW may be explained by the carrier screening of the QCSE resulting from piezoelectric fields.[45][46] When the prestrain layer was inserted between n-GaN and MQW, the bule shift by screen effect are decrease from 20meV to only 10meV. We now analyze the spontaneous PL in the InGaN samples in more detail. We explain the PL behavior differences of the LED with and without prestrain layer by quantum confinement and the predominant strong piezoelectric field in thin quantum wells. Indeed, in hexagonalnitride MQWs, the quantum-confined Stark effect arises due to the piezoelectric field.[46] We analyzed the influence of nonequilibrium carriers on the position of PL spectra in the InGaN/GaN MQWs using a triangular well model. The photoexcited carriers screen the internal field. For an idealized case, neglecting thermal distribution in the bands, the emitted quantum energy hn for band-to-band recombination in a quantum well in the presence of dependent forbidden gap, which, taking into account band-gap renormalization, can

be expressed as ³

with maximum field strength F(0 )and static relative dielectric constant ε, which has been taken as 10; and E_e,h is the difference of the lowest-energy level from the

triangular well bottom for electrons and holes, respectively, and can be calculated

In order to compare our experimental results with the calculations, we expressed hv as a function of excitation power density P (in MW/cm²), which for the case of

predominantly square-law recombination are related as[48] n =n P hv , where α is the absorption coefficient for laser light and has been taken as 1.5X10⁵ cm^-2, and γ

is the square-law recombination coefficient and has been taken as[49] 4.8x10^-11 cm³ s^-1.

The results of these calculations for InGaN/GaN MQWs are also illustrated in Figure 4.4.2.1 by the dotted line. This rather crude model gives remarkably good agreement between experimental data and theoretical estimations for more than four orders of excitation power density (up to 10 kW/cm²). Note that a number of other effects were not included in our model (we neglected the two-dimensional nature of the system, carrier distribution in the barriers and the wells, excitonic and nonradiative recombination channels, possible recombination coefficient change due to separation of the carriers in wells, etc.). The F₀ of reference was found to be equal to 1.21 V/cm, which is quite similar to other evaluations in similar semiconductor structures.[48] The estimation of the piezoelectric field of sample I and sample II are

the value of 0.68 and 0.73MV/cm, which is much smaller than sample III. From the above experiment result and data analysis, we can conclude that the use of LT-GaN prestrain layer release the biaxial strain in quantum well for a certainty.

4-5 Temperature dependent PL measurement

In order to further clarify the influence of the carrier confinement ability of LED with and without prestrain layer, temperature dependent PL measurements were performed. The Fig. 4.5.1 shows Arrhenius plots of the normalized integrated PL intensity for the InGaN-related PL emission over the temperature range from 20K to 300K. It was found that the integrated PL intensity dropped slowly with temperature during the low-temperature region, whereas it decreased rapidly during the high-temperature region. The best fitting gives three activation energies of about 51, 45, and 32 meV for InGaN/GaN, LT-GAN, and without prestrain layer, respectively.

In general, 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 [49]. It has suggested that the localization of carriers operates as excellent radiative recombination centers. In other words, high localization energies of excitons will provide deep potential wells that suppress the diffusion of electrical carriers toward various non-radiative defects. The

carrier localization in the active layer also has a significant effect on the performance of LEDs, resulting in an increase in radiative recombination efficiencies[50].

4-6 The measurement of internal quantum efficiency

From the preview result, we know that LT-GaN prestrain did change the property of active region. For this study, we used the IQE method which S. Watanabe et al.

proposed to determine the IQE of InGaN/GaN MQW LEDs to study the IQE of different structure. The internal quantum efficiency can be calculated by

EX affected by mostly carrier injection efficiency by laser, light extraction and correction efficiency of PL, and does not depend on either excitation power density or measurement temperature. First, we measured the excitation power dependent PL intensity at low and room temperature, and then the relative PL quantum efficiency curves can be obtained by using equation 4.6.1. And the constant C would be canceled out by normalizing the curves to the peak value at the lowest temperature, because it is independent on temperature or excitation power. From this normalization, the PL

efficiency curves will not depend on carrier injection efficiency by laser, light extraction and correction efficiency of PL. Therefore, the PL efficiency would be find out from this model.

In tradition, the IQE is estimated by assuming that IQE is 100 % at low temperature regardless of excitation power density. However, IQE is strongly dependent on injected carrier density. Consequently, it is more reasonable to assume the peak of PL efficiency at lowest temperature is equal 100 %, and then the IQE curves as a function of excitation power and temperature can be understand.

Moreover, to avoid the absorption of GaN, the frequency doubled femtosecond pulse Ti: sapphire laser of 390 nm was used to excite sample, the excitation power density was changed from 0.01 to 80 mW, and calculated injection carrier density is ab out 2 .0 × 1 0^{1 5} t o 1. 6 × 1 0^{1 8} cm^{- 3} b y u s i n g t h e eq u at i o n b el ow : repetition rate of laser, d_active is the active layer thickness, α_GaN is the absorption of GaN, α_InGaN is absorption of InGaN, R is reflection of sample surface, and Loss_objective is transmission loss of objective.

Fig. 4.6.1 shows the IQE of InGaN/GaN MQW LEDs as a function of injected

carrier density at 15K and 300 K. We can observe that the IQE increases with increasing injected carrier density to reach its maximum. As injected carrier density further increases, then the IQE decreases. The tendency of two efficiency curves at 15 K and 300 K is very similar. But under low injected carrier density region, the IQE curve at 300 K increases obviously than it at 15 K. The results indicated that the IQE at 15 K saturated more easily than it at 300 K.

The experimental results indicated that the IQE are about 61.9 %, 41.4 % and 53.6% at injected carrier density is 4.7 × 10¹⁷ #/cm³(~20mA) for sample I, sample II and sample III, respectively. Table 4.6.1 shows the value of IQE, We believe the higher internal quantum efficiency for the LED is due to the better crystalline quality, attributed to reducing of threading dislocations and releasing strain from sapphire ,so enhance the carrier confinement ability as studied in previous work [51]. However, when carrier density increase to 1.6 × 10¹⁸ #/cm³ ,the IQE are about 33.9 %,and 51.1

% for LED with InGaN/GaN and LT-GaN prestrain layer. Under similar quality and strain release order, the LED with LT-GaN insertion layer has much higher efficiency than InGaN/GaN is due to better In distribution and less nonradiative center.

4-7Current dependent intensity and efficiency discussion

Fig.4.7.1 shows the light output power versus injection current (L–I) characteristics of these LED samples. When a 20-mA current injection was applied to the LEDs with emitting wavelength of 460 nm, the output powers enhancement of LED-I and LED-II were approximately 11% and 18%, respectively. Even InGaN/GaN LED has higher output power enhancement at 20mA, as injection current increase to 200mA, the LT-GaN LED has larger output power enhancement than InGaN/GaN LEDs. In addition to the improvement of light output power for the LED. Note that three samples, regardless of having InGaN/GaN, LT-GaN or without prestrain layer between active region and n-type GaN show almost the same current density dependency of forward voltage. As shown in Fig. 4.7.2, the EQE rapidly decreases as the forward current increases up to 200mA, resulting in the serious EQE droop of ~ 54%. However, the MQW LEDs with LT-GaN prestrain layer show very different dependency of the EQE on the forward current density, EQE very slowly decreases with increasing the forward current density, showing the magnitude of EQE droop of approximately 36% at J_f of 200 mA. Table 4.7.1 shows the value of output power enhancement and efficiency decrease. These results indicate that LT-GaN prestrain layer facilitate the suppression of In inhomogeneous distribution in the active region , and successfully enhance the power and efficiency at high carrier density.

Figure.. 4.2.1 The schematic drawing of sample structure

Figure.. 4.2.2 The schematic drawing of fabrication processes of LEDs

Figure. 4.3.1 Top view CL images of (a) LT-GaN prestrain layer, (b)InGan/GaN prestrain (c) w/o prestrain structure at corresponding emission peak

wavelength

Figure. 4.4.1.1 Schematic band diagrams of localized state due to In fluctuation

Figure. 4.4.1.2 FWHM of three samples at different carrier density

Figure. 4.4.2.1 Emission energy of (a)LT-GaN (b)InGan/GaN (c)w/o prestrain structure at different power density.

Figure. 4.5.1 The Arrhenius plot of the integrated PL intensity obtained from the main emission peak over the temperature range from 10 to 300 K.

Figure. 4.6.1 IQE as a function of excitation power at 15K and 300 K.

Figure. 4.7.1 Emission power as a function of current density of three sample

Figure. 4.7.2 Emission efficiency as a function of current density

Table 4.6.1 IQE of InGaN/GaN, LT-GaN and without prestrain layers

InGaN/GaN LED at 10 and 80mW

Sample PL IQE @10mW(%) PL IQE @80mW(%)

LT-GaN InGaN/GaN w/o prestrain

~ 61.4%

~ 61.9 %

~53.6%

~51.1%

~33.9%

~26.0%

Table 4.7.1 Output power enhance and efficiency decrease

Sample Output power enhance (%) Efficiency droop(%)

20mA 200mA Max~200mA

LT-GaN ~11% ~39% 36%

InGaN/GaN ~18% ~28% 49%

Chapter 5 Analysis of the reduce efficiency droop by Graded quantum wells structure

5-1 Introduction

In recent years, great efforts have been made to reduce the efficiency droop.

Most of them are focus on minimizing the carrier overflow by reduce or eliminate the polarization field in the active region, such as using polarization matched multiple quantum wells (MQWs) [52, 53], staggered InGaN quantum wells [54], and non-polar or semi-polar GaN substrate [55]. But for improving hole distribution, only several approaches, such as p-type MQWs [56] or coupled quantum wells [57], are explored.

However, in the p-type MQWs, the Mg-dopant is very likely to diffuse into wells, while in the coupled quantum wells, electrons are tend to overflow by using thin barriers. These will result in reduction of radiative efficiency. In this research, we designed and grew a new LED structure with graded-thickness multiple quantum wells (GQWs) by using metal-organic chemical vapor deposition (MOCVD). Better hole distribution in such graded-thickness designed MQWs were demonstrated by APSYS simulation as well as the electroluminescence (EL) measurements.

5-2 Sample structure and fabrication

The LED structures were grown on c-plane sapphire substrates by MOCVD. A

buffer layer, ten-pair InGaN/GaN superlattice were grown on the top of sapphire.

After that, six-pair MQWs were grown with 10-nm-thick GaN barriers. For our designed experiment, the thicknesses of In0.15Ga0.85N quantum wells for GQW LED structure, controlled by growth time, are 1.5, 1.8, 2.1, 2.4, 2.7, 3 nm along [0001]

direction. While the reference LED structure has a unique well-thickness of 2.25 nm.

It’s worth noting here that the total volumes of active region for the two samples are the same. Finally, a 20-nm-thick electron blocking layer with Al0.15Ga0.85N and a 120-nm-thick p-GaN layer were grown to complete the epi-structure. The sample structure is shown in Fig. 5.2.1. For EL measurements, the LED chips were fabricated

by regular chip process with ITO current spreading layer and Ni/Au contact metal, and the size of mesa is 300×300 μm².

5-3 APSYS simulation of electron and hole concentration distribution

Based on our experimental structures, we built up the model of the reference and GQW LED structures. The typical LED structure was composed of 2-μm-thick n-type GaN layer (n-doping=2E18 cm^-3), six pairs of In_0.15Ga_0.85N/GaN MQWs with 10-nm-thick GaN barriers, 20-nm-thick p-Al_0.15Ga_0.85N electron blocking layer (p-doping=5E17 cm^-3), and 120-nm-thick p-type GaN layer (p-doping=1E18 cm^-3. Other material parameters of the semiconductors used in the simulation can be found

in Ref. [58]. Commonly accepted Shockley-Read-Hall recombination lifetime parameters (several nano-seconds) are used in the simulations. Figure 5.3.1 and Figure 5.3.2 shows the simulated hole distribution and radiative recombination distribution along MQWs at 100 A/cm². For reference LED structure, it can clearly be seen that holes mostly concentrate in the QW nearest p-side (denoted as the first QW), so does the radiative recombination. This phenomenon coincides with the optical measurement result in ref. [59], which is mainly due to poor transportation of holes.

While in the case of GQW LED structure, the hole concentration decreases in the first QW by about 16%, but increases in the second, third, and fourth QWs by 7%, 94%, and 175%, respectively, as compared with reference LED. It indicates that the holes are more capable of transporting across the first QW, consisted with our hypothesis.

On the other hand, electrons are relatively not being affected due to their high mobility. Therefore, more wells will participate in the recombination process, as illustrated by the radiative recombination distribution in Fig.5.3.2 Moreover, due to the relative low carrier densities in the first QW and more uniform of carrier distribution, the possibility of Auger scattering and carrier overflow can be lower.

Accordingly, alleviation of efficiency droop can be expected.

5-4 Current dependent electroluminescence measurement and analysis

The electroluminescence EL spectra of the LEDs were easured using a pulsed

current source with 1% duty cycle and 2us pulse width, to eliminate the heating effect.

Light was collected by an optical fiber placed above the diode and connected to a computer controlled sepectromemter equipped with a charge coupled device detector.

Different well width has different emission energy, so the GQW LED has larger band with than reference LED at whole current range. And which has been mentioned before is LED with thicker QW will emission longer wavelength due to QCSE effect.

So, The emission energy is of GQW LED is smaller than reference LED at small current density, as the injection increase, the peak energy will increase with injection current due to band filling effect, and the GQW LED has larger emission energy at

According to the simulated results mentioned above, more holes distribute in the narrower wells in GQW LED structure. Once more carriers radiatively recombine in

narrower well, the intensity of shorter-wavelength part in emission spectrum will rise.

Thus, the symmetry of spectrum might be changed, as can be seen from the power dependent of EL spectrum in Fig. 5.4.1. To investing EL spectrum in detail, the asymmetry factor (AsF) was calculated. As illustrated inset of Fig.5.4.3, it can be defined as the distance from the center line of the peak to the back slope (AB) divided by twice the distance from the center line of the peak to the front slope (2AC), with all measurements made at 50% of the maximum peak height. The calculated AsF under every injection level for both samples are summarized in Fig. 5.4.2. It can clearly be seen that, AsF of reference LED decreases slightly from 1.04 to be about 0.98 when injection current increases from 1 A/cm² to 100 A/cm². While GQW LED showes larger variation, the AsF starts at 1.05 (0.1 A/cm²) and saturates at about 0.89 (after 20 A/cm²). According to the definition of AsF, if the bluer light emits from narrower wells, the symmetry of spectrum would be interrupted and smaller than 1.

To further study the mechanisms responsible for the variation of the graded quantum structure, more electrical properties were investigated as below. Fig.5.4.3 shows the emission peak energy (a) and FWHM (b) of spectra as a function of the injected carrier density at room temperature for both LEDs of graded guantum well and reference. In Fig.5.4.3 several unique optical properties were observed. First, the emission peak energy gradually increases with the injected carrier density, this is due

to screening effect and band filling effect. And we can see, at low injection current, GQW LED has smaller emission peak energy and larger FWHM. The reason is when LED is under low injection current, only the quantum well which closest to p-type GaN generate light, and the GQW one has wider well width than reference. However, at high current density, we can observe the emission energy of GQW is larger than reference, it is due to carrier transport to thinner well which close to n-type GaN and emission short wavelength of GQW LED. The rapidly increase in FWHM of GQW also prove this penominace. Therefore, we can conclude that GQW does have superior radiative recombination distribution, which leads to the EL spectrum blueshifts and broadens significantly with increasing the injection current.

5-5 Analysis of injection carrier density dependence EL efficiency and efficiency droop

computer controlled sepectromemter equipped with a charge couple d device detector. It can clearly be seen that the light output power enhancement at current density of 22 A/cm2 and 244 A/cm2 for GQW LED is 36% and 71%. This indicates that even with wider wells (worse wave function overlap for electrons and holes) near p-side, the overall efficiency for GQW LED is still higher than reference, and the utilization rate of MQWs is improved. More importantly, LED without any structure,

the relative efficiency reaches its high as 33.3 A/cm² and efficiency shows only a slight decrease 17% at 244 Acm2, as shown in Fig 5.5.1. The quantum efficiency

peaks at a higher current density and shows a lesser decrease as current density increase. This improvement could be mainly attributed to the superior hole distribution and radiative recombination distribution, and also the reduction of Auger scattering resulting from the lower carrier concentration in QW nearest p-side. As the

peaks at a higher current density and shows a lesser decrease as current density increase. This improvement could be mainly attributed to the superior hole distribution and radiative recombination distribution, and also the reduction of Auger scattering resulting from the lower carrier concentration in QW nearest p-side. As the