Ref.@15K Ref.@300K 100nmNRA@15K 100nmNRA@300K 200nmNRA@15K 200nmNRA@300K 300nmNRA@15K 300nmNRA@300K
Fig.3-10 IQE of GaN-based LEDs as a function of carrier density at 15K and 300 K
-6 -4 -2 0 2 4 6
Fig.3-11 I-V curves of GaN-based LED grown on sapphire with and without SiO2 NRA
100 200 300 400
Fig. 3-12 L-I curves of GaN-based LED grown on sapphire with and without SiO2 NRA
Fig.3-13 GaN-based LED structure for 2D FDTD simulation
0 50 100 150 200 250 300 0
5 10 15 20
Enhancement in light output power(%)
SiO2 depth
Fig.3-14 Enhancement in light output power as a function of SiO2 NRA depth
Chapter4
InGaN/GaN LEDs grown on free-standing GaN substrate
4-1 Process of free-standing GaN substrate
Fig.4-1 shows the proposed process of fabrication of the free-standing GaN substrate.
In the beginning, we grew an undoped 1.8μm-thick GaN on a 2-in. c-plane sapphire substrate by metal-organic chemical vapor deposition (MOCVD). Then, a 500nm-thick SiO2 layer was deposited on the MOCVD grown-GaN layer by plasma-enhanced chemical vapor deposition (PECVD), followed by a 20nm-thick Ni layer deposition by an e-gun evaporator. The sample was annealed at 850℃ for 60s in nitrogen ambient to form self-assembled Ni nano-cluster on the SiO2 layer. The Ni nano-clusters acted as etching masks and subsequently, the-reactive ion etching and inductive couple plasma dry etching were performed to form GaN NRA. After etching, the GaN nanorod arrays were dipped into hot HNO3 to remove Ni cluster and buffered oxide etching (BOE) to remove SiO2
from the top of GaN nanorod arrays. We again deposited a 200nm-thick SiO2 layer on the GaN nanorod template by PECVD. After GaN nanorods template was coated with PMMA by spin coating, we used RIE and BOE to remove SiO2 on the top of GaN nanorods. Finally, we removed PMMA and grew FS GaN substrate on GaN nanorod-array with SiO2 sidewall passivation by HVPE system. The thermal expansion
coefficient difference between GaN and the sapphire substrate caused the GaN thick layer separating from the sapphire substrate.
4-1.1 Fabrication of GaN nanorods
Fig.4-2 show the SEM images indicated that the fabricated GaN nanorods were
approximately 150nm to 450nm in diameter with a density of 4×109 cm-2 and the etched depth of the GaN nanorods were 1.8μm. Afterward, we again deposited a 200nm-thick
SiO2 layer on the GaN nanorod template by PECVD and then utilized RIE to remove the SiO2 layer in the top surface of GaN nanorods. The Fig.4-3 clearly exhibits the sidewalls of GaN nanorods were surrounded by a thin SiO2 film with thickness of ~50nm.
4-1.2 GaN nanorods self-separated Hydride vapor-phase epitaxy(HVPE)
HVPE was used to regrow a 300μm-thick GaN layer, the main chemical reaction in
HVPE system can be written as
For better coalescence and 2D growth, growth temperature from 950℃ to 1100℃
was selected. Series of growth parameter such as reactor pressure and V/III ratio were varied from 50 to 20 for better crystal quality. The pressure was from 500 mbar to 940 mbar for different growth steps. The Fig.4-4 shows SEM images of the initial stage of
GaN thick films growth on the GaN nanorods and many GaN islands were observed on
the nanorods in the initial growing stage. The thickness of GaN bulk is approximately 3.3μm and the surface is quite rough.
The Fig.4-5 show that these GaN islands were only grown on the top of GaN nanorods, with increasing growth time, these islands coalescence , and flat GaN surfaces appeared. We can observe that GaN regrowth layer was suspended on the nanorod-array template because of the SiO2 sidewall passivation.
4-1.3 Epitaxial characteristics of free-standing GaN substrate
Fig4-6 shows the typical bright field cross-section TEM images of GaN substrate obtained from the nanorod-array. HVPE was used to regrow a 10μm-thick GaN layer. We found few threading dislocation density at GaN top surface.. A number of stacking faults often occurred above GaN nanorods, it is believed that the stacking faults could block the propagation of TDs. From the CL images, the threading dislocation density in the GaN
substrate separated from GaN nanorod-arrays is estimately to be ~107 cm-2 in the Fig4-7.
During the HVPE cooling process, the 300μm-thick GaN substrate self-separates
from the underlying host sapphire substrate as a result of the thermal expansion coefficient difference between GaN and the sapphire substrate. In order to release the thermal stress during HVPE cooling process, the GaN nanorods were broken and resulted
in the self-separation of GaN from sapphire substrate. Fig.4-8 shows the results of as-grown 300μm GaN films separated from GaN nanorod-arrays template and flat GaN template, respectively. A complete 2 in. self-separated free-standing GaN substrate grown on GaN nanorod-arrays template was demonstrated in Fig.4-8(a). In contrast, as shown in figure 4-8(b), the GaN grown on flat GaN template smashed into several pieces because the thermal stress cannot be released in the HVPE cooling process.
4-2 InGaN/GaN multiple quantum wells grown on free-standing GaN substrate
The material epitaxial growth uses low pressure metal-organic chemical vapor
deposition (MOCVD). A 25nm of low temperature GaN nucleation layer followed by a 1.6μm u-GaN buffer layer were grown on sapphire substrate or free-standing GaN substrate. The LED device structure was 3.2μm n-GaN, 10 pairs of
n-In0.08Ga0.92N/GaN(2nm/2nm) pre-strain layer, 5 pairs of In0.18Ga082N/GaN (2.5nm/9nm) quantum wells, p-In0.16Ga084N(100nm) electron blocking layer, and 40 pairs of p-AlGaN/GaN (2.5nm/2.5nm) cap layer Ni/Au(3nm/7nm) was and Cr/Au(100nm/250nm) were subsequently evaporated onto the sample surface to serve as transparent conductive layer and p-type electrode, respectively. Ti/Al/Ni/Au contact was deposited onto the exposed n-GaN layer to serve as the n-type electrode, as shown in Fig 4-9. Finally, the epitaxial wafers were scribed to the fabrication of 300×300 μm2 LEDs.
4-3 Measurement and discussion
4-3.1 Raman spectrum
In order to further identify the difference strain in the free-standing GaN substrate separated from nanorod-array template and flat GaN template grown on sapphire, we perform the measurement of Raman scattering. The E2-high phonon modes of free-standing GaN and flat GaN template were located at 566.8 cm-1 and 569.6 cm-1, respectively, as shown in the Fig4-10. The E2-high peak for the GaN substrate separated from the GaN nanorod-array template was very close to that of the strain-free GaN substrate, which is believed to be 567.1 ±0.1 cm-1.[30] Therefore, the residual stress in the GaN substrate obtained from the nanorod-array template was negligible. The residual stress could be calculated by the following equation:
where ωγand ω0 represent the Raman peaks of GaN template grown on sapphire and GaN
substrate separated from GaN nanorod-array template. The estimated value of stress is about 1.02GPa by adopting a theoretical Kγ value of 2.56cm-1/GPa report by Wanger and Bechstedt [31].
4-3.2 Electroluminesence measurement
The comparison of L-I characteristic between LED grown on FS GaN substrate and
xx
LED grown on sapphire are shown in Fig.4-12. The total light output power of LED grown on FS GaN substrate is higher than that grown on sapphire in all injection regimes.
At an injection current of 20mA, the light output power of LED grown on FS GaN substrate was enhanced by a factor of 12.3％ compared to that grown on sapphire. The dislocation density of LED grown on FS GaN substrate was effectively suppressed, which increase IQE. The figure shows the light output power of GaN LED grown on FS GaN substrate does not saturate for the injection currents up to 550mA due to GaN substrate provides the better heat dissipation. The thermal resistances were measured to analysis the heat dissipation of these two samples. The thermal resistance of LED grown on FS GaN and that grown on sapphire is 20.2℃/W and 52.39℃/W, respectively. The smaller thermal resistance might be induced from better heat dissipation for LED regrowth on FS GaN substrate.
To further improve device performance, it is essential to understand the effects of these defects on the carrier injection mechanisms and recombination processes. Several groups have reported the dominance of carrier tunneling in GaN-based light emitting diodes[32,33]. Recently, Chitinis et al. [34] demonstrated a high-quality AlInGaN/InGaN multiple quantum well (MQW) LED structure by incorporation of a small amount of In into the AlGaN barrier layers. The electrical characteristics showed an ideality factor of
2.28, which was close to the theoretical prediction for carrier recombination in the space-charge region. In Fig.4-12, we present a study of electrical characteristics of commercially available GaN-based LED on different substrate. The ideal factor of
GaN-based could be calculated by the following equation:
qv qv
E nkT
s s
I = I e = I e
where Is is the reverse saturation current, q is the elementary charge, k is the Boltzmann constant, T is the absolute temperature, and n is the ideality factor.
The ideal factor of LED grown on sapphire and that grown on FS GaN were 3.96 and 2.11, respectively. The ideal factor of LED grown on FS GaN is smaller than that on sapphire may be related to the better crystalline quality.[35]
We have the case of injected current dependent EL spectrum at room temperature.
The Fig.4-13 shows the EL spectra of LED grown sapphire substrate and FS GaN substrate, measured at different injection current. As the injection current increases from 1 to 120 mA, the peak wavelength of LED grown on FS GaN substrate shifts toward shorter wavelength from 442.9nm to 441.7 nm. But for LED grown on sapphire substrate, as the injection current increases, its peak wavelength just exhibits a large blue-shift of 4 nm, as shown in Fig 4-14. As the injection current increases, the number of free electrons
and holes in the conduction and valence band increases. Furthermore, the electrons and holes are spatially separated by the piezoelectric field. And this effect can produce a free-carrier-induced field. Due to the direction of this field is opposite to the piezoelectric field, the two electronic fields tend to compensate with each other. So the QCSE becomes smaller, and the transition energy will become larger, which leads to a blue-shift of peak wavelength. Moreover, the magnitude of QCSE can be scaled by the blue-shifts depending in injection current. Thus the large blue-shift of LED grown on sapphire substrate reveals that the QCSE is much stronger than that grown on FS GaN substrate.
Because the blue-shift of LED grown on FS GaN substrate is small, the QCSE of it is small too
We performed the current-dependent EL measurement at low temperature (LT) and room temperature (RT) and plot the efficiency curves of both LED on sapphire substrate and free-standing GaN substrate in the Fig.4-15 after normalizing to its peak at low temperature. The corresponding efficiency ~61.6％ in RT is at injected current of 20mA.
For the GaN LED grown on GaN substrate, a similar dependence of the IQE on the injected current was observed. However in term of the efficiency ~78.1％ in RT is at injected current of 20mA. The IQE of LED grown on GaN substrate was enhancement by 26.8％, as compare to the LED grown on sapphire substrate. We believe that the higher
IQE for the LED grown on GaN substrate is due to the better crystalline quality , attributed to low threading dislocation and strain relaxation that can reduce quantum confined Stark effect (QCSE).
The GaN-based LED efficiency generally is highest at low currents—typically a few milliamperes—and as the injection current increases, the efficiency decreases gradually.
This well-known phenomenon, called efficiency droop must be solved for devices operating at high powers. A solution to the droop problem has not yet been provided, and, considering that different explanations were proposed including Auger recombination [36,37], electron leakage [38] , carrier delocalized[39] and lack of hole injection[40].
Fig.4-16 shows the normalized external quantum efficiency as a function of forward voltage. Investigation of figure reveals a varied efficiency droop behavior for LED grown on different substrates. A much less efficiency droop was observed for the LED grown on FS GaN substrate. LED grown on FS GaN substrate can suppress the QCSE and provide better thermal dissipation, the ability of carrier confinement is better at high injection current. For eliminating the influence of thermal dissipation, the injection pulse current was used. A more efficiency droop for the LED grown on sapphire substrate was also observed due to strain-induced poor QW quality and the strong QSCE. We demonstrate the suppression of efficiency droop in GaN-based LED by utilizing GaN substrate. The
relaxation of compressive stress in GaN epilayers reduces the piezoelectric polarization, which in turn reduces the QCSE. Therefore, the efficiency droop is significantly suppressed at high injection current
Formation of Ni Clusters by thermal
annealing
SiO2film deposition by PECVD
PMMA by spin coating GaN nanorods formed
by ICP-RIE
Removing SiO2on the top of nanorods by RIE and BOE
Nanorods passivation by SiO2
Thick GaN growth by HVPE
SiO2film deposition by PECVD
PMMA by spin coating GaN nanorods formed
by ICP-RIE
Removing SiO2on the top of nanorods by RIE and BOE
Nanorods passivation by SiO2
Thick GaN growth by HVPE
GaN separation from the sapphire
Fig. 4-1 Schematic illustrations of the fabrication of free-standing GaN substrate by nanorods-assisted separation
Fig.4-2 SEM image of GaN nanorods
Fig.4-3 SEM image for GaN nanorods with sidewall passivation
Fig.4-4 SEM images for the initial stage of HVPE regrowth with a growth time of 2 min
Fig.4-5 SEM images of HVPE regrowth with a growth time of 5min
Fig.4-6 TEM images of GaN substrate by nanorods-assisted separation
Fig4-7 CL images of GaN substrate by nanorods-assisted separation
Fig.4-8 Results of GaN thick films obtained from(a)GaN nanorod arrays and (b) flat GaN
substrate
Fig.4-9 The schematic of GaN-based LED structure
500 550 600 650
Fig4-10 Raman analysis of GaN grown on sapphire and free-standing GaN