Results and Discussion

We then performed the temperature dependent EL measurements on the prepared UV LEDs on PSS and flat substrate sample. The APSYS simulation and an equivalent circuit model will be also developed. The details will be discussed in the following sections.

5-3.1 Temperature dependent EL characteristics

Temperature dependence EL at various injection current levels has been measured from 30 to 300 K. Before we start to discuss the variation of both samples under different temperature and injected current, we could take a look at the spectrums at several specific conditions. The three-dimensional (3D) plots of the EL spectrums at three injection currents of 0.1, 1, and 20 mA with different environment temperature were chosen and shown in figure 5.4 (a) to (c), respectively. First, when the injected current was as low as 0.1 mA in figure 5.4(a), we could observe a maximum at lowest temperature of 30 K in both samples. The intensity gradually decreased with the increasing temperature. Since the injected current maintains constant through whole measurement period, we could realize it as a reduced efficiency while the temperature increased. This result could be understood by relating to the behavior of nonradiative recombination process. The increased temperature would activate the nonradiative centers and deteriorate the device lighting efficiency. Since it is determined by the numbers of nonradiative centers and we had confirmed the PSS’s effect to improve IQE, therefore, it is not surprising to observe a sharper slope of efficiency in flat sapphire sample.

When the injected current level increased to 1 mA, we notice that the curves had similar tendency with that at 0.1 mA, but the slope becomes smoother. It is then reasonable to assume there must exist some additional factor to influence the efficiency behavior and contribute to an opposite trend to the nonradiative centers. If we further increased the injection current to 20 mA, we found the efficiency peaks moved from the lowest temperature to about 80-100K.

It means this unexpected factor lowering the efficiency at low temperature becomes stronger at high injection level. Therefore, the next step for our discussion is to find out what the mechanism is to cause the opposite trend, and realize the efficiency trend comes from at different current and environment temperature condition.

Temperature dependent EL at constant low current injection

In order to realize the origin of the efficiency curves, we discussed the whole results by four different conditions, that is temperature dependent EL at constant low and high current injection and current dependent EL at constant low and high environment temperature, respectively. Here, we started the discussion from the condition of a temperature dependent curve at constant low current injection of 0.1 mA.

As shown in figure 5.5(a), the efficiency decreased as the increasing temperature in both samples. Generally, thermal quenching phenomenon was observed in InGaN-based structures.

At low injection current, the carrier injected into quantum well, and then the carriers can be confined due to the localized states in the In-rich region. In InGaN-based structures, the emission came from the localized states, which could trap carriers within this potential minimum. We could refer to the model of InGaN material system in figure 4.12. When the temperature increased, the carrier received activation energy to be thermalized and escaped from localized centers to defect or nonradiative centers resulting in the EL efficiency decreased. This mechanism caused a decrement in efficiency with increasing temperature as the red line in figure 5.5(b). However, we find not only the efficiency, but the voltage has an obvious increment with decreasing temperature. This forward voltage increasing of InGaN/GaN UV LEDs is believed due to the hole concentrations and mobility decrement [5]

at low temperature operation. The authors in [5] had studied the variation of electron and hole concentration and mobility as shown in figure 5.6(a) and (b). From figure 5.6(a), the electron mobility becomes a half to a third from room temperature to 100 K but the hole mobility drops to one order of magnitude lower than that in room temperature as shown in figure 5.6(b).

Lower the environment temperature, poorer the activated hole concentration and mobility.

Hence, this will cause an observation of efficiency decrement as the environment temperature was cooling down as represented in the green line in figure 5.5(b). Finally, we could realize the efficiency tendency as the blue line. We note that, since the operation voltage shown in figure 5.5(c) does not vary that serious with the decreasing temperature, the efficiency suppression by green line is not that obvious. That is the reason we could still observe an efficiency peak at lowest temperature.

Temperature dependent EL at constant high current injection

We then move to the condition of constant high current injection condition, 20 mA, with a varied temperature as shown in figure 5.7. Compared to low injection case, the peak shifts to higher temperature. It is then interesting to find out what causes the difference. Basically, the influence of nonradiative centers did not change with the injection current, so the curve of

nonradiative centers remains the same with the condition of low injection as the red line in figure 5.7. Followed by our previous assumption, the greatly reduced hole mobility suppress the efficiency especially at low temperature. This phenomenon could be realized and quantized by observing the voltage variation from room temperature to low temperature.

While the injected current is 0.1 mA, the operation voltage varied about 0.5 V from room temperature to low temperature but it increased about 1.4 V at 20 mA current injection. This implies the influence of the poorer hole mobility becomes stronger than low injection level.

So, the curve of the hole concentration and mobility could be expressed as the green line in figure 5.7(b) and the efficiency curve is represented as the blue line.

In figure 5.7(c), the forward voltage increased about 1 and 1.4 V for PSS and flat sapphire UV LED, respectively. Therefore, we conclude that the variation of operation voltage plays an important role in temperature dependence EL efficiency, especially, the EL efficiency may be affected when decreasing temperature at high injection current. In high injection current, it appears that the carriers are effectively captured by active centers in the MQW under the application of lower forward voltage at room temperature. But, forward voltage increased while decreasing temperature, they are rather transferred to nonradiative recombination centers as a result of escape from the MQW region, thus reducing the EL efficiency. This is because the carriers can escape out of the well region due to the external field effects as shown in figure 5.8. We also demonstrated that the higher field existing in the well under the higher forward voltage decreases the radiative recombination rate, which also causes the reduced EL intensity.

Current dependent EL at constant low environment temperature

We had discussed about the conditions of temperature dependent results at low and high current injection. Next, we tried to understand the current dependent EL at a low and room temperature case. We first plotted the efficiency curves from 0.1 to 100 mA at low temperature, 30 K, environment as shown in figure 5.9(a). The EL quantum efficiency at 30K increased slightly with increasing injection current before injection current is below 1 mA.

Recall the IQE measurement form PL method; we conclude the initial stage with an increased IQE is belonging to the compensation of the nanradiative recombination centers. Though we cooling down the temperature to about 77 K, it is believed that is not sufficient to suppress all inherent dislocation. Therefore, the injected carriers were attributed to compensate the nonradiative recombination centers, and reach the maximum at about 1 mA. We find the black line, PSS UV LED, is higher than without PSS UV LED in red line before 1 mA. It could be

explained as the PSS sample has fewer dislocations than the flat sample, so a higher and smoother IQE curve is observed.

While further increasing injection current, the EL efficiency drops rapidly, we observed that the reduction of EL efficiency of PSS InGaN/GaN UV LED is very similar to without PSS structure. The results indicated that the reduction of dislocation density didn’t affect the efficiency seriously at high injection current. Therefore, we could understand the effect of nonradiative centers as a red curve shown in figure 5.9(b). Then, we look forward the role of hole concentration and mobility. Since the environment temperature was fixed in this case, the amount of operation voltage should be proportional to the injected current only. Therefore, the efficiency should be decreased with the increasing current. The EL efficiency reduction at high injection currents is coming from the higher forward voltage which results in carriers escape from the quantum well and overflow to p-GaN. This influence could be represented as the green line in figure 5.9(b) and the efficiency is black one. Relative the device operated at room temperature, the voltage variation is much serious as injected current increased as shown in figure 5.9(c). Therefore, the efficiency slope is sharper than in room temperature case.

Current dependent EL at room temperature

At the last case, we move to the case of injected current dependent EL efficiency at room temperature. The measured efficiency curves were shown in figure 5.10(a). The EL efficiency if both samples increased at the initial stage with a peak at about 8 mA and drops as the current continuously increased. Similar to above cases, we could understand the curved by dividing into two stages. First is the influence of dislocation at initial stage and second is hole concentration and mobility. Since the device is operated at room temperature, the nonradiative recombination strongly dominated the efficiency performance and that is the reason of PSS sample enjoyed a higher IQE than flat sample. This could undoubtedly be presented as the red line in figure 5.10(b). For the hole performance, since room temperature has little effect to the amount of activated hole concentration, it is quite flat during whole current region as the green line in figure 5.10(b). Hence, the efficiency is combined as the black line. One thing should be noted is the operation voltage difference between PSS and flat sample is negligible as shown in figure 5.10(c). So the high current region remains identical tendency to the low injection region, we could observe a higher EL IQE curve in PSS than flat sample in the whole current injection period.

5-3.2 Equivalent circuit analysis for temperature dependent EL efficiency

We summarized the temperature dependent EL efficiency curves in figure 5.11(a) and (b), which acted as a function of injection current of PSS and without PSS InGaN/GaN UV LED, respectively. At first glance, one could find that every temperature, as the injected current increased, has peak efficiency. However, this peak position varied as the temperature changed.

For example, it is around 8 mA when the temperature is above 120 K and gradually moved to 1 mA when temperature was decreased to 80 K. Moreover, the maximum in whole the measurement curves happened at 100 K for flat sample as shown in figure 5.11(a) and 80 K for PSS sample in figure 5.11(b), instead of the lowest temperature. Here, we adopted the equivalent circuit model proposed by Shuji Nakamura and Steven P. Denbaars et al. [6] for better understanding of the temperature dependence EL efficiency tendency. They claimed four components putting in the equivalent circuit model illustrated in figure 5.12(a). This circuit model ensures all injected carriers are traced and are not unintentionally lost. The resistor R₁ represents current leakage paths, such as extended crystal defects and sample surface, for example. This current component is not considered to involve carrier recombination, hence is purely carried by electrons. There are two diodes in this model. Diode D₁ is responsible for current flow due to radiative recombination. This current component results in photon emission and detection upon recombination. Diode D₂ is responsible for nonradiative recombination current. Such recombination occurs via nonradiative recombination centers (NRCs) and does not emit photons within the wavelength range of interest. These two types of recombination are not limited within or near the active region.

They are distinguished by whether a photon is emitted and detected within the wavelength range of interest upon a recombination event. Another resistor R2 combined with a switch represents carrier overflow. This component is considered to be the electron unipolar current, i.e., is carried by electrons that do not recombine with holes and exit the system to the p-type contact. Current division between ID1 (through D1) and ID2 (through D2) is defined as:

where r_R is the radiative recombination rate and r_NR is the nonradiative recombination rate. EL efficiency is defined in this current range as

1

equation are taken to identify the fundamental behavior of:

V is the applied voltage, ID0 is the diode saturation current, q is the unit charge, n is the ideality factor, k is the Boltzmann constant, and T is the absolute temperature. Equations (5-3) and (5-4) can be related via V to obtain from eq. (5-2) as

By substituting β=1 and ID0=10 pA into equation 5-5, we could estimate the efficiency number with different R₁ as shown in figure 5.12(b). Since the efficiency is inverse proportional to R1, it reaches the ideal maximum, ~1, while R1 is infinite. On the other hand, the efficiency decreased if R1 becomes smaller. Therefore, we can observe the entire curve gradually shifts upwards towards EL efficiency ~ 1 as a result of nonradiative recombination centers deactivation when temperature is reduced. In this model, the EL efficiency increased when increasing injection current, and the EL efficiency curve bending less sharp while temperature is decreasing. The phenomenon indicated that the significant leakage current existed, and the leakage current is reduced when temperature decreased. It also coincides with the results we found in figure 5.9(a), that is, the PSS sample has less dislocation density, i.e., the larger R1 and contributes to a higher IQE.

We finally summarized the EL efficiency as a function of injection current at 77 and 300 K and normalized to the maximum number at 77 K as shown in figure 5.13. The EL efficiency is 60.3%, and 41.6% at 20 mA of PSS and without PSS InGaN/GaN UV LED, respectively.

From this comparison results, we could reasonably assume that we could also get an IQE number by this method once we could estimate and get the peak number. The next problem will be, since we had established a method for IQE estimating by PL method, what the difference between these two methods. Therefore, the comparison between these two methods will be discussed.

5-3.3 Comparison of IQE Measurement by PL and EL Methods

We performed the IQE measurement by PL method and compared the results between

these two, PL and EL, as shown in figure 5.14. Here we transformed the laser power in PL and injected current in EL method into carrier density in a unit of particles per centimeter cubic. For current and carrier density transformation, we need to calculate the active area and assume the current spreading in whole active region uniformly. Then, we could put the EL and PL results in a figure for comparison. The IQE tendency of EL and PL at low and room temperature are showing there. The solid lines represent the results from EL method and hallow lines represent the results from PL method. The normalized peak position does not have obvious difference between these two methods; however, the curves at room temperature have an apparent difference. We could understand it by dividing into low temperature and room temperature cases.

For low temperature case, the PL efficiency is higher than EL efficiency at low carrier injection level, that means there exists more leakage paths while EL injection. More leakage current occurred when electrical injected carrier into quantum well to be captured by nonradiative recombination centers in n-GaN. Figure 5.15(a) demonstrated that some leakage current appeared in n-GaN and capture the injected electrons to deteriorate the device efficiency. On the other hand, it would not happen when the electrons were resonant pumped by a 390 nm laser and being able to well confined in MQWs without being absorbed by defects in n-GaN. Further increasing injected carrier density, the PL and EL efficiency decreased due to carrier began to overflow from quantum well to p-GaN layer as shown in figure 5.15(b). And we found that the EL efficiency decreased more rapidly at high injected carrier density, it can be attributed to the forward voltage increased when increasing injected carrier density and pushing carriers escaping from the quantum well and overflow to p-GaN resulting an efficiency reduction.

For room temperature side, we observed a higher IQE curve in both PSS and flat sample from PL results than EL case. We depicted the schematic of electron an hole injection recombination as shown in figure 5.16. As we know, both electron and hole concentration and mobility decreased when decreasing temperature, but the degree of decrement in holes is nearly one magnitude serious than electrons. This fact caused that hole concentration distribution is extremely non-uniform and insufficient to fill up whole the active region.

Therefore, some electrons migrate to the last few pairs of MQWs near p-GaN and recombine at the surface finally. This could fully explain the lower IQE curves from EL results than PL.

5-3.4 APSYS simulation

From the discussion above, we could conclude that there are two main factors

influencing the device efficiency under current injection condition, the density of dislocation and hole concentration and mobility variation. The dislocation density has been verified by TEM and IQE measurement in previous chapter; however, we had not been able to verify the effect of hole characteristics. In this section, we tried to verify it by the help of packaged simulation software, APSYS. Since the activated hole concentration and hole mobility would be strongly reduced under a low temperature condition, we simulated a condition of three order of magnitude reduction in concentration and mobility. The simulated I-V curves were shown in figure 5.17. The solid line represents our experimental results and the dash line stands for the simulation results. It is quite match between our simulated and experimental results. can see the simulation of I-V curve is similar to experimental results. Furthermore, based on this parameters, we simulated the electron and hole distribution inside the MQWs as shown in figure 5.19(a) and (b), respectively. The black and red line represents the condition at 300 and 77K, respectively. The electron concentration distribution is quite uniform, but still a higher (~five times) electron concentration is observed near p-GaN than near n-GaN as shown in figure 5.18(a). This investigation also coincides with our contention before for some electrons would transport to p-GaN easily due to the non-uniform hole distribution. Figure 5.18(b) displays the hole concentration distribution in MQW, we found a more serious non-uniform distribution of hole concentration in the MQWs. The hole concentration in MQW near p-GaN is over one order of magnitude than near n-GaN. The shortage of injected holes failed to effectively recombine with the electrons, so that some of the injected electrons reach the p-GaN and exit the system without being able to recombine. Then, figure 5.19

influencing the device efficiency under current injection condition, the density of dislocation and hole concentration and mobility variation. The dislocation density has been verified by TEM and IQE measurement in previous chapter; however, we had not been able to verify the effect of hole characteristics. In this section, we tried to verify it by the help of packaged simulation software, APSYS. Since the activated hole concentration and hole mobility would be strongly reduced under a low temperature condition, we simulated a condition of three order of magnitude reduction in concentration and mobility. The simulated I-V curves were shown in figure 5.17. The solid line represents our experimental results and the dash line stands for the simulation results. It is quite match between our simulated and experimental results. can see the simulation of I-V curve is similar to experimental results. Furthermore, based on this parameters, we simulated the electron and hole distribution inside the MQWs as shown in figure 5.19(a) and (b), respectively. The black and red line represents the condition at 300 and 77K, respectively. The electron concentration distribution is quite uniform, but still a higher (~five times) electron concentration is observed near p-GaN than near n-GaN as shown in figure 5.18(a). This investigation also coincides with our contention before for some electrons would transport to p-GaN easily due to the non-uniform hole distribution. Figure 5.18(b) displays the hole concentration distribution in MQW, we found a more serious non-uniform distribution of hole concentration in the MQWs. The hole concentration in MQW near p-GaN is over one order of magnitude than near n-GaN. The shortage of injected holes failed to effectively recombine with the electrons, so that some of the injected electrons reach the p-GaN and exit the system without being able to recombine. Then, figure 5.19