Optical characteristic of NP-LEDs

In order to confirm the efficiency improvement of our NP-LED, the PL internal quantum efficiency (IQE) measurement was performed. A general approach to evaluate the IQE of LEDs is to compare the PL integrated intensity between low and

room temperatures [74]. Figure 4.6 shows the measured quantum efficiency as a function of excitation power at 15 K and 300 K for NP-LEDs and C-LEDs. The efficiency is defined as the collected photon number divided by the injected photon numbers and normalized to the maximum efficiency of low temperature [75].

For the LED grown on planar sapphire substrate and GaN nano-pillar template, one could clearly see that the IQE increased with injected carrier density increased to reach its maximum and decreased as the injected carrier density further increased. The tendency of this two efficiency curves at 15 K and 300 K was quite similar. But under low injection carrier density range, the IQE at 300 K increased more pronounced than it at 15 K. The corresponding density to the peak efficiency for the LED grown on planar sapphire (44.3%) in 300 K is at injected carrier density of about 4×10¹⁷ cm^-3, which is larger than it at 15 K, about 4×10¹⁶ cm^-3. On the other hand, for the LED grown on GaN nano-pillar template, a similar dependence of the IQE on the injected carrier density was observed. Nevertheless, in term of the peak efficiency (55.9%) in 300 K at injected carrier density of 2×10¹⁷ cm^-3, the IQE of LED grown on the NP template was enhanced by ~26.2%. It means that under the same injected power of pumping laser, there is about 26.2% enhancement for the converted photon carriers within the active region, as compared to the conventional LED structure grown on planar sapphire substrate. Moreover, at the corresponding generated carrier density of

8×10¹⁷ cm^-3 (20mW) it can be found that the IQE increase from 41.9% for C-LEDs to 53.1% for NP-LEDs, At this excitation level, we could calculate the corresponding generated carrier density, approximately same level of 20 mA at room temperature in our device. Thus, the better emission efficiency of GaN- NP based LED can be linked directly due to the improvement of IQE and much better crystal quality, attributed to NELOG on nano-pillar template and reduction of threading dislocations, as mentioned in previous section. Besides, we would further discuss about the mechanism responsible for the dependence of IQE on the injected carrier density for LEDs grown on planar sapphire and nano-pillar template.

It was been reported that the collected PL intensity, L, is proportional to the injected carrier density, I, with a power index P. This could be expressed as [76, 77]

L α I^P (4-2)

where parameter P physically reflects the various recombination process. If P equals to 1, it indicates the radiative recombination dominates. On the other hand, P>1, the Shockley-Read-Hall (SRH) recombination occurs, which was related to the presence of nonradiative centers and might be a shunt path to the current. Figure 4.7 (a) and (b) were the relationship between injected carrier density and the PL intensity of both

LED samples grown on planar sapphire and nano-pillar template.

At 15 K for LED sample grown on planar sapphire and nano-pillar template, the PL intensity is almost linearly varied with excitation power density and the power index was equal to 1, indicating that the radiative recombination dominated the recombination process at all injected carrier density range. It was reasonable that the radiative recombination dominated all the recombination process at 15 K, since the nonradiative centers would be quenched generally at such low temperature closed to absolute zero. Nevertheless, under low excitation injected carrier density at 300 K, the superlinear dependence of L on I are observed for both LED samples grown on planar sapphire and nano-pillar template. This means that defect related nonradiative recombination dominates the recombination process in this low carrier injection range at room temperature. But as injected carriers continuously increased, the linear dependence of the PL intensity to the injected carrier density is exhibited. It was noteworthy for both LED samples in 300 K, the value of P decreased from P>1 to P=1 gradually with the increasing of injected carrier density, instead of jump to 1. It means the nonradiative centers are saturated and leads to the gradual suppression of the nonradiative recombination with the injected carrier density. Therefore, the radiative recombination started to dominate the recombination process, resulting in the pronounced increasing of the IQE, as shown in figure 4.6, for the region of injected

carrier density less than 4×10¹⁶ cm^-3.

From the preceding discussion, we concluded that the value of P in the superlinear zone could reflect the different crystalline quality and amout of nonradiative centers (such as defect and threading dislocations) between the two LED samples grown on planar sapphire and nano-pillar template. The value of P for LED samples grown on planar sapphire and nano-pillar template were 1.55 and 1.21, respectively, indicating that LED grown on sapphire had more nonradiative recombination centers than LED grown on nano-pillar template. This result was similar to the crystalline quality analysis in section 4.2.

In addition to the optical characteristic mentioned above, we could get more information from the PL spectrum and further investigated the optical mechanisms which affected the IQE result. We tried to analyze the emission peak energy and the FWHM of spectra as a function of the injected carrier density at 300 K for this two LED samples, LED grown on planar sapphire and nano-pillar template. For the LED grown on planar sapphire substrate, several unique optical properties were observed at 300K, as shown in figure 4.8 (a). First, the emission peak energy gradually decreased with the injected carrier density ranging from 8×10¹⁴ cm^-3 to 4×10¹⁶ cm^-3, and an opposite trend was observed as the injection carriers further increased. Second, the FWHM of spectra shrinks when the injection carrier density ranging from 8×10¹⁴ cm^-3

to 4×10¹⁶ cm^-3, and an opposite trend was observe as the injection carriers further increased.

In the region of the injected carrier density from 8×10¹⁴ cm^-3 to 4×10¹⁶ cm^-3 in figure 4.8 (a), the emission energy showed the red-shift and the FWHM shrank with increasing injected carrier density. As can be seen in figure 4.7 (a) in this section, the parameter P in this region is larger than 1, indicating the nonradiative recombination dominated the carrier recombination process. Under this circumstance, the excited carriers tended to recombine at higher energy extended states before reaching into lower energy localized states [78]. However, with the increasing of the injected carrier density, the nonradiative recombination was gradually bleached out and the carriers could reach into lower energy localized states easier. This is why the emission energy showed the red-shift and the FWHM shrank with increasing injected carrier density.

On the other hand, as the injected carrier density further increasing, other mechanisms dominated the optical characteristic. In general, there are two mechanisms responsible for the blue-shift of emission energy with increasing injected carrier density. The first is the so-called coulomb screening of the quantum-confined Stark effect (QCSE). This effect occurred in the general InGaN/GaN MQWs because the internal field direction is parallel to the MQWs growth direction. As a result, the devices with the general InGaN/GaN MQWs often endure a strong QCSE. It caused a

band tilting and a separation of wavefunction between electrons and holes, which resulting in a wavelength redshift and recombination efficiency reduction.

Nevertheless, the increasing of injected carrier density could weaken the QCSE, that is why we called this effect as the coulomb screening of the QCSE. Therefore, an increasing trend of transition energy and efficiency might be observed if this effect dominates. Moreover, as the screening effect dominates the emission process, it accompanied a reduction in FWHM, due to QCSE was restrained, resulting in the effect of band tilting had become quenched and further overlapping the wavefuction between electrons and holes.

The second is band filling effect of localized states, which is an effect for carriers filled at a higher energy level while the injected carrier density increased continuously.

For the InGaN/GaN material, indium composition inhomogeneity and monolayer thickness fluctuation were occurred commonly in the InGaN MQWs. Such self-organized In-rich region was generated in InGaN active region, resulting in potential fluctuation of the energy bandgap [79-81]. As further increasing the injected carrier density, the filling effect of high energetic localized centers started interfering and become dominated, that also induced a blue-shift of emission energy. However, different from the effect of QCSE, this effect accompanied the broadening of FWHM.

In figure 4.8 (b), we showed the emission energy and the FWHM as function of

injected carrier density at 300 K for the LED grown on GaN nano-pillar template.

Compared with the LED grown on planar sapphire, it was apparent that the variation in both the emission energy and the FWHM were smaller. We concluded this phenomenon into three parts. First, at low injected carrier density (8×10¹⁴ cm^-3 to 4×

10¹⁶ cm^-3), the LED grown on nano-pillar with the smaller parameter P (1.21 and 1.55 for NP-LED and C-LED, respectively) had less threading dislocations and nonradiative centers. This result also decreased higher energy extended states and carriers could reach into lower energy localized states without the influence of nonradiative centers. This is why the variation of the emission energy red-shift and FWHM for NP-LED were smaller than C-LED. Second, for the region of the injected carrier density from 4×10¹⁶ cm^-3 to 2×10¹⁷ cm^-3, the radiative recombination become dominated (P=1), leading to coulomb screening of the QCSE. However, the Raman spectra of bulk GaN grown on nano-pillar template, as shown in figure 4.3 in section 4.1, demonstrated that the residual strain was almost free for bulk GaN grown on nano-pillar template. As a result, we could expect that the coulomb screening of the QCSE was moderate for NP-LED. This expectation was also confirmed by the peak energy and FWHM as a function of injected carrier density. Third, as the injected carrier density increasing continuously, as mentioned before, the band filling effect of localized states dominated. The reason for moderate variation in both peak energy and

FWHM for NP-LED might because more uniform for the indium composition inhomogeneity and monolayer thickness fluctuation. On the other hand, the C-LED with larger residual strain might bring about different size of indium clusters, resulting in inhomogeneity of indium composition and mono layer thickness.

All the information provided above could be the evidence to explain the origin of higher IQE for the LED grown on nano-pillar template. Besides, we tried to use another method to analyze about this. Here, we collected the temperature dependent PL intensity and plot the normalized intensity with inversed temperature which is understood as Arrhenius plot in figure 4.9. Then, this temperature dependent curve could be fitted by the following equation to get the activation energy:

IT = I0 / [1+A exp(-Ea / kT)+B exp(-Eb / kT)] (4-3)

Where IT, I0 are the integrated PL intensity for T and 0 K, A and B are constants, k is the Boltzmann constant, T is the temperature, Ea is the activation energy for PL quenching, and Eb is generally associated to the free exciton binding energy [82]. The fit activation energy for the LED grown on the planar sapphire and nano-pillar template are 79 meV and 62 meV, respectively. In general, the activation energy can be explained by the ability to confine the carriers within the potential minima. One

could understand it through the schematic of activation energy in figure 4.10. It had been understood that because of the indium inhomogeneous in InGaN material system, there are a large indium fluctuation and a various localized states. The average of this localization could be quantized as an effective localized state. On the other hand, because of a large density of dislocation which results in a nonradiative recombination, we could also summarize this state as a effective defect state. The difference between these two states could be regarded as the so-called activation energy. One could also realize it as one kind of energy to enable a confined electron trapped in the localized state escaping from it. As a result, we could conclude that higher value of the activation energy indicates the stronger confinement of injected carriers and that certainly promises the higher IQE.

We demonstrated the PL internal quantum efficiency (IQE) measurement and found out that NP-LED had larger IQE than C-LED. To analyze the variation of IQE for both NP-LED and C-LED, we further compared the difference of peak energy and FWHM as a function of infected carrier density and temperature dependent PL. All the evidences revealed that NP-LED with higher IQE had better performance in optical analysis.