The counteraction between the increased carrier localization effect due to the change of composition nanostructure in the quantum wells (QWs), which is caused by the thermal annealing process, and the enhanced quantum-confined Stark effect in the QWs due to the increased piezoelectric field, which is caused by the increased p-type layer thickness, when the p-type layer is grown at a high temperature on the InGaN/GaN QWs of a high-indium light-emitting diode (LED) is demonstrated. Temperature- and excitation power-dependent photoluminescence (PL) measurements are performed on three groups of sample, including 1) the samples with both effects of thermal annealing and increased p-type thickness, 2) those only with the similar thermal annealing process, and 3) those with increased overgrowth thickness and minimized thermal annealing effect. From the comparisons of emission wavelength, internal quantum efficiency (IQE), spectral shift with increasing PL excitation level, and calibrated activation energy of carrier localization between various samples in the three groups, one can clearly see the individual effects of thermal annealing and increased p-type layer thickness. The counteraction leads to increased IQE and blue-shifted emission spectrum with increasing p-type thickness when the thickness is below a certain value (20-nm p-AlGaN plus 60-nm p-GaN under our growth conditions). Beyond this thickness, the IQE value decreases and the emission spectrum red shifts with increasing p-type thickness.
©2012 Optical Society of America
The green gap in the technology development of light-emitting diode (LED) is caused by the low efficiencies of long-wavelength InGaN-based LED and short-wavelength InAlGaP-based LED. In a nitride-based LED, the material quality of InGaN with high indium content becomes quite poor leading to the low emission efficiency of an InGaN/GaN quantum well (QW). In determining the emission efficiency of a high-indium InGaN/GaN QW, two crucial factors need to be considered, including the quantum-confined Stark effect (QCSE) caused by the internal electrical field built in a QW [1–5] and the carrier localization mechanism due to the formation of indium-rich nano-clustering structures in the QW [6–18]. Both factors originate from the large lattice mismatch between GaN and InN (11% difference).
Basically, there are three types of internal electrical field in an InGaN/GaN QW, including the piezoelectric field due to the strain effect of the GaN barriers and cladding layers on both sides of an InGaN well layer [4, 5, 19], the polarization field due to the difference of the polarization strength between GaN and InGaN , and the built-in field due to the p-i-n junction structure. In a Ga-polarity InGaN/GaN QW, which is usually used in an LED structure, both the piezoelectric and polarization fields are oriented from the p-GaN layer toward the n-GaN layer. The built-in field is oriented in the opposite direction. Also, the polarization field and built-in field are quite small when compared with the piezoelectric field. Therefore, the net internal field is dominated by the piezoelectric field, which determines the strength of the QCSE in the QW. With such an internal electric field across the QW, a potential tilt is formed leading to the separation of the electron and hole wavefunctions and hence the reduction of their overlap integral. In this situation, the radiative recombination rate in the QW is reduced and the effective band gap is decreased (emission spectrum red shifted) . Because the piezoelectric field strength increases with the strain magnitude in the InGaN well layer, the QCSE becomes stronger in a QW of higher indium content. The strong QCSE in a green LED is one of the major reasons for its low emission efficiency. The QCSE can be screened when a large amount of carriers are injected into the QW optically or electrically . The spatial separation of the injected carriers of high density screens the internal electrical field and flattens the potential distribution, leading to the increase of radiative recombination rate and the blue shift of emission spectrum. Based on the screening behavior, the strength of QCSE in a QW can be understood with an excitation power-dependent photoluminescence (PL) measurement. The spectral blue-shift range with increasing PL excitation power in a certain range can be used to represent the QCSE strength. However, it is noted that the carrier screening effect can be weakened if certain nano-scale quantum-dot-like potential traps exist in the QW for localizing the electrons and holes at the same locations.
The other important factor affecting the emission efficiency of a high-indium InGaN/GaN QW is the carrier localization effect due to the formation of indium-rich nano-clustering structures. Again, because of the large lattice mismatch between GaN and InN, their solid miscibility is quite low. When the incorporated indium content is high, the spinodal decomposition process can occur in InGaN to produce sharper indium-rich nano-clusters and milder composition fluctuations [6, 10, 13, 15–18]. The energy states in the indium-rich nano-clusters can deeply localize carriers and are called the localized states. On the other hand, those with milder composition fluctuations lead to shallow carrier trapping and are called the free-carrier states or delocalized states. With certain thermal energy, carriers can move freely among the free-carrier or delocalized states. Carrier localization can help in enhancing emission efficiency of an InGaN layer, particularly in a layer of a high defect density. With high average indium content, the spinodal decomposition process becomes stronger, leading to the nano-clusters of higher density and possibly deeper potentials . Therefore, compared with a blue LED, the carrier localization mechanism plays a more important role in determining the emission behavior of a green LED. At a low temperature (say, 10 K), most carriers are trapped in the localized states. When the sample temperature is increased, carriers with thermal energy can escape from the localized states into free-carrier states [15–18]. In the free-carrier states, carriers can be trapped by defects leading to the trend of decreasing emission efficiency with temperature. During the process of increasing temperature, carriers can be heated such that they are elevated to a higher energy level of the localized states before escaping from the cluster. In this process, one can observe an S-shaped variation of emission spectral peak with increasing temperature [15–18, 21, 22]. The energy required for carrier escape is usually called the activation energy. The activation energy can be calibrated by fitting the temperature-dependent integrated PL intensity variation (the Arrhenius plot) with an equation (discussed later). The structures of nano-clustering and composition fluctuation can be changed through a thermal annealing process. It has been shown that by thermally annealing an InGaN/GaN QW at 800-1000 °C for a few tens min, the aforementioned composition nanostructures can be changed and the emission efficiency can be either enhanced or reduced. The thermal annealing process may change the nano-cluster potential depth and its density for either stronger or weaker carrier localization [13, 15, 16, 23–25]. After thermal annealing, compositional nanostructure variations in such QWs have been observed with transmission electron microscopy and cathodoluminescence measurements.
Regarding the thermal annealing effect, the high-temperature post-growths of the p-AlGaN electron blocking layer and the p-GaN hole transport layer can be an important factor in determining the emission efficiencies of the InGaN/GaN QWs in an LED, particularly that in the green or longer-wavelength range. Research efforts have been made to reduce the growth temperature of the p-type layer for minimizing the thermal annealing effect [26, 27]. However, so far no detailed study has been undertaken for understanding the effects of high-temperature overgrowth of the p-type layer on the QW emission behavior in a high-indium LED. The overgrowth effects may include the thermal annealing process and the increasing piezoelectric field due to the increased p-type layer thickness. In this paper, we demonstrate the effects of the high-temperature overgrowths of the p-AlGaN and p-GaN layers on the InGaN/GaN QW emission behavior in a high-indium LED. The effects include the thermal annealing process of the QWs to change their composition variation structures and the increase of piezoelectric field in the QWs to alter their QCSE. By overgrowing the p-type layers of different thicknesses in a series of LED sample, we observe that the internal quantum efficiency (IQE) of the QWs first increases with increasing p-layer thickness and then decreases with a thicker p-layer. The results of temperature- and excitation power-dependent PL measurements are used for demonstrating the effects of carrier localization and QCSE variations. In section 2 of this paper, we show the sample structures and their preparation procedures. Also, the PL measurement conditions are described. Then, in section 3, the comparisons of the PL measurement results of the samples with different p-layer thicknesses overgrown at a high temperature by metalorganic chemical vapor deposition (MOCVD) are shown. The PL measurement results of the QW samples under different thermal annealing conditions are discussed in section 4. Next, those results of the samples with the capping GaN layers of different thicknesses overgrown by molecular beam epitaxy (MBE) at a significantly lower temperature are illustrated in section 5. Numerical modeling results and discussions are provided in section 6. Finally, the conclusions are drawn in section 7.
2. Sample preparations and measurement conditions
Sixteen samples in total are prepared for measurements in this study. These samples are classified into three groups for comparisons under the conditions of changing different parameters. In the first group, six samples (samples A-F) are grown with MOCVD (Aixtron CCS3x2”FT) on c-plane sapphire substrate. In each sample, after the deposition of ~2-μm n-GaN (at 1100 °C), five InGaN/GaN QW periods with the growth temperatures (thicknesses) of the InGaN wells and GaN barriers at 670 and 820 °C (3 and 12 nm), respectively, are grown. In sample F, the MOCVD growth stops at the top GaN barrier layer. In sample E, a 20-nm p-Al0.2Ga0.8N layer is deposited on the top of the QW structure at 960 °C with the growth duration of 25 sec after temperature ramping from 820 to 960 °C within 140 sec and temperature holding for 30 sec (only with nitrogen flow during the temperature holding stage). Then, in samples D, C, B, and A, a 30, 60, 120, and 180-nm p-GaN layer, respectively, are added to the top of the p-Al0.2Ga0.8N layer with the growth temperature also at 960 °C after temperature holding for 120 sec. The p-GaN growth durations of samples D, C, B, and A are 130, 260, 520, and 780 sec, respectively. It is noted that 960 °C is the lowest growth temperature for maintaining a hole concentration of ~2 x 1017 cm−3 in p-GaN and p-AlGaN in using this MOCVD reactor. After the growth, the sample temperature is ramped down to room temperature (RT) within 35 min. No in situ p-type activation process is applied to any sample until an LED is to be fabricated. The parameters of samples A-F and their measurement results are summarized in Table 1 .
In the second group of sample, five pieces of sample F are thermally annealed with the temperatures and durations (including temperature ramping) the same as the growth conditions (after QW growth) of samples A-E in another MOCVD reactor (Veeco p75) to produce samples FA-FE, respectively. This group of sample will be used for demonstrating the effects of post-growth thermal annealing on the optical property of the InGaN/GaN QWs. We use the Veeco MOCVD reactor, instead of the Aixtron reactor, for thermal annealing because a sample can be heated without rotating the sample carrier in the Veeco reactor. In the Aixtron MOCVD reactor, the sample can be heated only when the sample carrier rotates. In both MOCVD reactors, the sample carriers cannot hold broken sample pieces when they rotate. The parameters of samples FA-FE and their measurement results are summarized in Table 2 . To show the effect of increased piezoelectric field on the QW emission behavior, we overgrow undoped GaN on the structure of sample E for producing the structures similar to samples A-D. For this purpose, the structure of sample E is re-grown with a slightly higher InGaN growth temperature (estimated to be 672 °C, instead of 670 °C for sample E) in the Aixtron MOCVD reactor. The new sample of a shorter QW emission wavelength is designated as sample E’. On sample E’, we use a radio-frequency (RF) plasma-assisted MBE reactor for growing undoped GaN at 680 °C with the thicknesses of about 30, 60, 120, and 180 nm to produce samples E’D, E’C, E’B, and E’A, respectively. The reason for using MBE, instead of MOCVD, for the overgrowth in this group of sample is to take the advantage of the low-temperature GaN growth in MBE for minimizing the thermal annealing effect on the MOCVD-grown InGaN/GaN QWs during the overgrowth stage. Also, we overgrow undoped GaN in this group of sample, instead of p-GaN, for further reducing the MBE growth temperature. Before MBE growth, sample E’ is cleaned with acetone and methanol. Then, the sample is preheated at 200 °C for 100 min in the preparation chamber of the MBE reactor to completely remove the vapor on the sample surface before it is moved into the growth chamber. The MBE growth conditions include the substrate temperature at 680 °C, Ga effusion temperature at 1050 °C, nitrogen mass flow rate at 0.5 sccm, and RF power at 200 W. The growth durations of undoped GaN for samples E’D, E’C, E’B, and E’A are 15, 30, 60, and 90 min, respectively. The parameters of samples E’A-E’D and E’ and their measurement results are summarized in Table 3 .
X-ray diffraction (XRD) measurement is undertaken with a double-crystal Bede1 system. The Bede software is used to fit the data for obtaining the QW widths and indium compositions of various samples. The PL measurement is performed with top (p-side) excitation and top PL emission monitoring. It is excited by a 406-nm InGaN laser diode (THORLABS, TCLDM9). The excitation power for the temperature-dependent PL measurement is set at 6 mW. For excitation power-dependent PL measurement at RT, the excitation power is varied from 1 through 10 mW.
3. Measurement results of samples A-F
Figure 1 shows the XRD patterns in the (0002) plane of samples A-F. Here, the major peaks and the shoulders on the left correspond to the features of GaN and InGaN, respectively. The oscillation behaviors left to −1000 arcsec originate from the periodical QW structures. The fitting results of the QW widths and indium compositions are shown in rows 4 and 5, respectively, of Table 1. Here, one can see that the variations of the QW width and indium composition are small among samples A-F. Such variations are close to the range of measurement inaccuracy. The QW widths and indium compositions are around 2.6 nm and 22%, respectively, samples A-F. Figure 2 shows the temperature-dependent variations of PL spectral peak energy for samples A-F. Here, every curve demonstrates more or less an S-shaped variation. In each curve, one can see the blue shift or almost zero shift in a certain temperature range. As mentioned in section 1, the S-shaped variation has been used for identifying the carrier localization mechanism in a QW. In Fig. 2, one can also see that the additions of the p-AlGaN and p-GaN layers to the QW structure result in the blue shift of PL spectral peak at RT. A thicker p-type layer leads to a higher emission photon energy until the thickness reaches 60 nm in sample C. Beyond sample C, a thicker p-GaN layer results in a smaller emission photon energy in samples A and B. The variation trend of the emission photon energy at 10 K (low temperature – LT) among samples A-F is similar. In Fig. 3 , we show the variation of normalized integrated PL intensity (integrated over the emission spectrum) with temperature. With the integrated PL intensity normalized to unity at the lowest temperature in the measurement, the value at RT can be regarded as the internal quantum efficiency (IQE) [28, 29]. The separations among those curves in Fig. 3 are larger in the temperature range of 60-210 K, when compared with that at RT, because in this temperature range, a significant amount of localized carriers can escape from the traps to become free carriers. In a sample of weaker carrier localization, more carriers are delocalized and hence a higher non-radiative recombination rate is observed. From Fig. 3, one can see that carrier localization is stronger in samples C and D and that in sample F is the weakest.
The PL spectral peaks at RT and LT and the IQE values of samples A-F are listed in rows 6-8, respectively, of Table 1. Here, starting from the QW structure (sample F), the IQE increases with increasing p-type layer (p-AlGaN and p-GaN) thickness until it reaches a certain value. The IQE value increases from 4.2% in sample F to 7.8% in sample E, 8.1% in sample D, and 13.8% in sample C. Beyond sample C, the IQE value decreases with increasing p-GaN thickness, i.e., deceases from 13.8% in sample C to 11.6% in sample B and 9.1% in sample A. Also, the PL spectral peak first blue shifts with increasing p-type layer thickness. It shifts from 543.9 nm in sample F to 539.9 nm in sample E, 537.6 nm in sample D, and 528.9 nm in sample C. Beyond sample C, the PL spectral peak red shifts with further increasing p-type layer thickness. It red shifts from 528.9 nm in sample A to 532.4 nm in sample B and then 542.9 nm in sample A. The variation trend of the PL spectral peak at LT is the same as that at RT. Also, the variation trend of the emission spectral peak is similar to that of IQE. In other words, the maximum IQE corresponds to the shortest emission wavelength. It is noted that this correspondence may be regarded as a well-known phenomenon, i.e., an InGaN/GaN QW LED of a lower indium content (a shorter emission wavelength) usually has a higher crystal quality and hence a higher IQE. However, in this study, the total indium content is supposed to be the same among samples A-F because the growth conditions for their InGaN/GaN QWs are the same and the total indium content is fixed before the overgrowth of the p-type layers. The differences in emission wavelength and IQE must be due to the changes of material structure and/or electrical property in the QWs caused by the different overgrowth thicknesses or durations at a high temperature.
The causes for different emission wavelengths and IQE values among samples A-F include at least the thermal annealing process on the QWs during the high-temperature (960 °C) overgrowth and the variation of QCSE due to different p-type layer thicknesses. As mentioned in section 1, the strength of QCSE can be estimated with the carrier screening process through an excitation power-dependent PL measurement. The PL spectral peak energy variations with excitation power level of samples A-F are shown in Fig. 4 . We list the blue shift ranges of PL spectral peak when the excitation power increases from 1 through 10 mW for samples A-F in row 9 of Table 1. It is interesting to see that the shift range of spectral peak also increases first and then decreases with increasing p-type layer thickness. The shift range of spectral peak increases from 4.9 meV in sample F to 6.0 meV in sample E, 6.3 meV in sample D, and then 12.9 meV in sample C. Then, from sample C, it decreases to 11.8 meV in sample B and 7.5 meV in sample A. Again, the variation trend is similar to other parameters described above. Sample C has the largest shift range of spectral peak among samples A-F. Here, QCSE alone cannot systematically explain the observed PL measurement results because a stronger QCSE should lead to a lower IQE. The other important factor, i.e., the thermal annealing process, must be considered. The thermal annealing process can change the structures of nano-clusters in the QWs for altering the carrier localization behaviors. The strength of carrier localization can be estimated by calibrating the activation energies through the fitting of the Arrhenius plot based on the equation of [30–32]Fig. 5 . The parameters of α, β, EA1, and EA2 used for fitting are listed in rows 10-13 of Table 1. Among the four parameters, the most important one is the activation energy EA2. From Table 1, one can see that EA2 increases first with increasing p-type layer thickness from 51.0 meV in sample F to 77.1 meV in sample C. Then, as the p-type layer thickness further increases, EA2 drops to 66.3 meV in sample B and 57.8 meV in sample A. The level of EA2 can be used to represent the strength of carrier localization. A larger EA2 corresponds to the condition of stronger carrier localization.
The results in Figs. 2-5 and Table 1 can be summarized with two variation trends: When a p-type layer is overgrown at a high temperature on the top of the QWs, the thermal annealing process changes the composition variation structures, particularly the nano-clustering structures, for enhancing carrier localization. A thicker p-type layer or a longer thermal annealing duration leads to a stronger carrier localization effect (a larger EA2). This variation trend is reversed when the p-type layer thickness is larger than 60 nm. Meanwhile, the increased p-type layer thickness produces a stronger piezoelectric field and hence a stronger QCSE. In this situation, the spectral blue shift range increases in the excitation power-dependent PL measurement. The trend of increasing spectral blue shift range with increasing p-type layer thickness is also reversed when the thickness is larger than 60 nm. The enhancement of carrier localization through the thermal annealing process can result in the increase of IQE. On the other hand, the enhancement of QCSE leads to IQE reduction.
4. Measurement results of samples FA-FE
To confirm the effect of increasing carrier localization through the thermal annealing process, we show the temperature-dependent and excitation power-dependent PL measurement results of samples FA-FE in Figs. 6 -8 and Table 2. Meanwhile, with the XRD measurement, the fitting results of the QW widths and indium compositions of those samples are shown in rows 4 and 5, respectively, of Table 2. For comparison, the results of sample F are also shown in these figures and Table 2. Different from samples A-E, samples FA-FE are thermally annealed (post-growth thermal annealing) but have no extra piezoelectric fields caused by the overgrown p-type layers. Again, one can see that the variations of the QW width and indium composition from the XRD measurement are small among samples FA-FE and F. The QW widths and indium compositions are around 2.6 nm and 22.8%, respectively. From Fig. 6, one can again see the S-shaped variation of temperature-dependent PL spectral peak energy in each sample. The temperatures for the minimum spectral peak energies are quite different among those samples. In particular, the temperature for the minimum spectral peak energy of sample F at 240 K is significantly larger than those of samples FA-FE (90 K for sample FE, 90-120 K for sample FD, 120 K for sample FC, and 180 K for both samples FA and FB). The similar differences can also be observed among samples A-F, as shown in Fig. 2. Such differences reflect the different nano-clustering structures among those samples of different thermal annealing conditions. Figure 7 shows the variations of integrated PL intensity with temperature for samples FA-FE and F. The IQE values of those samples are shown in row 8 of Table 2. Here, we can see that the IQE monotonically increases with increasing thermal annealing duration, from 4.2% in sample F to 8.5% in sample A. In this table, one can also see the monotonically blue-shifting trend of PL spectral peak wavelength at RT with increasing thermal annealing duration. In Fig. 8, we show the variations of PL spectral peak energy of samples FA-FE and F when the excitation power is increased from 1 through 10 mW. The shift ranges of spectral peak are listed in row 9 of Table 2. Here, one can see that the shift range slightly varies between 4.1 and 5 meV in a random manner. All the shift range values are significantly smaller than those of samples A-E (see Table 1), clearly indicating that the QCSEs in samples FA-FE are about the same as that of sample F. In other words, the post-growth thermal annealing process does not significantly affect the strength of QCSE. The figures for showing the fittings to the Arrhenius plots for samples FA-FE and F are not shown in this paper. However, the fitting parameters are listed in rows 10-13 of Table 2. Here, one can see the monotonically increasing trend of EA2 with increasing annealing duration, indicating the increasing trend of carrier localization strength with post-growth thermal annealing. With stronger carrier localization, the IQE is increased. The blue shift trend with increasing thermal annealing duration can be interpreted as follows: During thermal annealing, indium atoms move toward the nano-clusters to form even higher-indium clusters. In this situation, the background indium content for the free-carrier states is reduced, leading to the emission blue shift trend at RT. The increased indium contents in the nano-clusters to form deeper traps can be identified from the longer PL spectral peak wavelengths at LT in samples FD and FE, when compared with that in sample F (see row 7 of Table 2). However, with a longer thermal annealing duration, the indium atoms can be redistributed within a nano-cluster to form a multiple-minimum potential trap such that the majority of localized states have higher energy levels than the counterparts in sample F. Therefore, the PL spectral peak wavelengths either at RT or at LT of samples FA-FC become shorter than that of sample F.
By comparing the results in Table 2 with those in Table 1, one can see that the emission wavelength of sample FA (the sample with the longest post-growth thermal annealing process) is longer than that of sample C. Also, the IQE and EA2 of sample FA are smaller than the corresponding values of sample C. These comparisons indicate that as far as the change of nano-clustering structure for increasing the carrier localization effect is concerned, post-growth thermal annealing is not as effective as in situ thermal annealing.
5. Measurement results of samples E’A-E’D and E’
To confirm the increasing QCSE variation with increasing overgrowth thickness under the condition of minimized thermal annealing effect, we show the temperature-dependent and excitation power-dependent PL measurement results of samples E’A-E’D and E’ in Figs. 9 -11 and Table 3. Meanwhile, with the XRD measurement, the fitting results of the QW widths and indium compositions of those samples are shown in rows 4 and 5, respectively, of Table 3. Again, the variations of the QW width and indium composition from the XRD measurement are small among samples E’A-E’D and E’. The QW widths and indium compositions are around 2.6 nm and 21.8%, respectively. The indium content of sample E’ is slightly lower than that of sample E. However, such a small difference does not affect our conclusions in this paper. In Fig. 9, the dependencies of PL spectral peak energy on temperature show clear S-shaped variations in all samples. The variations of integrated PL intensity with temperature are shown in Fig. 10 . Their results are summarized in Table 3. Here, in rows 6 and 7, one can see the monotonically red-shifting trend of the PL spectral peak with increasing overgrowth thickness at both RT and LT. Although the IQE values of samples E’A-E’D do not show a clear variation trend, they are mutually quite close and are significantly smaller than that of sample E’. The excitation power dependencies of PL spectral peak of those samples are shown in Fig. 11. The blue-shift ranges of those samples are listed in row 9 of Table 3. Here, the spectral shift range increases monotonically with increasing overgrowth thickness although their values are smaller than those of samples A-E. The generally smaller spectral shift ranges (weaker QCSEs) in samples E’A-E’D and E’, when compared with samples A-E, can be attributed to their shorter emission wavelengths (by about 10 nm) or their lower average indium contents. In a lower-indium InGaN/GaN QW, the QCSE is weaker. In this situation, the effect of overgrowth is also expected to be weaker. The monotonically increasing trend of spectral shift range with increasing overgrowth thickness clearly indicates the effect of increasing piezoelectric field or QCSE. With a stronger QCSE, the IQE is reduced and the emission spectrum is red shifted. However, the variation of IQE is quite small and does not show a clear trend. The figures for showing the fitting curves to the Arrhenius plots of samples E’A-E’D and E’ are not shown in this paper. The fitting parameters are listed in rows 10-13 of Table 3. Here, one can see that although EA2 values do not show a clear variation trend, they are significantly larger than that of sample E’. These results imply that with the low-temperature (680 °C) overgrowth of GaN, the effective carrier localization effect becomes stronger. This result has two possible causes: First, with the low-temperature but long-duration (30-180 min) MBE overgrowth, a certain thermal annealing effect is still produced. Second, the increase of piezoelectric field or the change of strain condition can also lead to the change of nano-clustering structure for stronger carrier localization. This issue deserves further investigation. It is noted that although stronger carrier localization can result in a higher IQE, the dominating QCSE still leads to the decrease of IQE.
6. Modeling results and discussions
To further confirm the QCSE variation trend among samples A-F, we use the commercial software APSYS (Crosslight Software Inc.) to simulate the structures of those samples [33–35]. In this software, the factors of various internal fields due to piezoelectric polarization and spontaneous polarization, and built-in fields are included. However, the factor of nano-clustering and composition fluctuation structures or the carrier localization effect in an InGaN/GaN QW is not included. Therefore, the simulation results can only be used for comparing the QCSEs among different samples, particularly for understanding the QCSE dependence on the overgrowth thickness. In particular, without the factor of QW nanostructures being included in the software, the simulated energy states or band gaps can be oversimplified. In Fig. 12 , we show the simulation results of potential diagrams of samples A-F. In the simulations, the Mg doping concentration in the p-type layers is set at 1.2 x 1018 cm−3 with the acceptor activation energy set at 170 meV. The Si doping concentration in the n-type layer is set at 5 x 1018 cm−3 with the donor activation energy set at 10 meV. That in the intrinsic GaN layers is set at 1 x 1016 cm−3. The indium content of the InGaN well layer is assumed to be 22%. In the diagram of Fig. 12, the zero energy is set at the conduction-band minimum of the n-type layer. Five QWs with tilted potential distributions can be clearly seen. The differences of QW energy level between different samples decrease with increasing QW number counted from the p-type side. In LED operation, the QW closest to the p-type layer dominates the emission behavior because of the low mobility of hole in GaN . In our PL measurements with top excitation, the first QW absorbs the strongest excitation light and hence makes the most important contribution to the measured PL results. Although the measured PL results are based on the mixed contributions of different QWs, we will use the simulation results of the first QW to compare with the PL measurement results. The results are summarized in rows 4 and 5 of Table 4 . Here, the band gap is defined as the energy difference between the conduction-band minimum and the valence-band maximum. Therefore, this energy can be significantly smaller than those of the major emitted photons. The internal electric field in the QW corresponds to the slope of the energy diagram within the QW. In Table 4, one can see that the band gap decreases with increasing p-type layer thickness, but saturates in samples A-C. The decreasing trend of band gap with increasing overgrowth thickness is generally consistent with the red-shift trend of emission wavelength in samples E’A-E’D and E’. The variation of internal electric field shown in Table 4 (row 5) demonstrates the increased potential tilt or QCSE with increasing p-type layer thickness . However, the variation of internal electric-field strength saturates when the p-GaN layer thickness is larger than 120 nm (sample B).
With the observations above, we can now try to explain the results of samples A-F shown in Table 1. From the results based on the measurements of samples FA-FE, we can see that to a certain extent, a longer thermal annealing duration leads to a shorter emission wavelength, a stronger carrier localization effect, and a higher IQE. On the other hand, from the results based on the measurements of samples E’A-E’D and E’, one can see that under a certain condition, the QCSE increases with increasing overgrowth thickness. In this situation, the emission wavelength is red shifted and the IQE is reduced. Meanwhile, carrier localization can be enhanced. The increased QCSE with increasing overgrowth thickness is confirmed by the simulation results. Therefore, the variations of emission behaviors in samples A-F can be attributed to the counteraction between the carrier localization effect and the QCSE. In samples C-F, the thermal annealing effect dominates. In other words, as the p-type layer thickness is increased, the thermal annealing process changes the indium composition structures and hence enhances the carrier localization effect. In this situation, the activation energy is enlarged, the IQE is increased, and the emission wavelength is blue shifted with increasing p-type layer thickness. At the same time, the shift range of spectral peak is increased with increasing p-type layer thickness, indicating the increased piezoelectric field. It is noted that the piezoelectric field dominates over the spontaneous polarization field and the built-in field in an InGaN/GaN QW . However, among samples A-C, the further increase of p-type layer thickness leads to the following two behaviors: 1) The effect of the thermal annealing process for increasing carrier localization saturates such that the EA2 values of samples A and B become smaller, when compared with that of sample C. It is noted that the in situ thermal annealing process in samples A-E and the post-growth thermal annealing process in samples FA-FE should result in different nano-clustering structures and hence different carrier localization effects even though their thermal annealing conditions are individually the same. In particular, the mixed process of thermal annealing and increasing piezoelectric field should lead to different results from those of only thermal annealing. Therefore, although the saturation behavior of thermal annealing is unclear in samples FA-FE, this behavior can exist in samples A-E. 2) The increasing trend of QCSE saturates when the p-type layer thickness is larger than 60 nm. Although the saturation behavior is unclear in samples E’A-E’D, it can be clearly seen in the simulation results. The decreasing trend of spectral-peak shift range with further increasing p-type layer thickness in samples A-C can also be attributed to the carrier localization effect, which prevents the electrons and holes from moving toward the individual low-potential sides for producing the screening effect. Hence, the measurement of spectral-peak shift range cannot provide us with the real QCSE strength. In samples A-C, the increased QCSE dominates over the increased carrier localization effect such that as the p-type layer thickness is further increased, the IQE is reduced and the emission wavelength is red shifted.
LEDs based on samples A-D have been fabricated after a p-type activation process of 820 °C for 15 min in a rapid thermal annealing chamber. The LED characteristics have been presented in an earlier publication . The p-type activation process is expected to change little the indium-rich nano-clustering structures. The variation trends of the electroluminescence wavelength and the emission intensity of the four LEDs are the same as the PL spectral peak and the IQE of the corresponding samples, respectively.
In summary, we have demonstrated the counteraction between the increased carrier localization effect due to the change of composition nanostructure in the QWs, which was caused by the thermal annealing process, and the enhanced QCSE in the QWs due to the increased piezoelectric field, which was caused by the increased p-type layer thickness, when the p-type layer was grown at a high temperature on the InGaN/GaN QWs of a high-indium LED. Temperature- and excitation power-dependent PL measurements were performed on three groups of sample, including: the samples with both effects of significant thermal annealing and increased p-type thickness, those only with the similar thermal annealing process, and those with increased overgrowth thickness and minimized thermal annealing effect. From the comparisons of emission wavelength, IQE, spectral shift with increasing excitation level, and calibrated activation energy of carrier localization between various samples in the three sample groups, one could clearly see the individual effects of thermal annealing and increased p-type layer thickness. The counteraction resulted in increased IQE and blue-shifted emission spectrum with increasing p-type thickness when the thickness was below a certain value (20-nm p-AlGaN plus 60-nm p-GaN under our growth conditions). Beyond this thickness, the IQE decreased and the emission spectrum red shifted with further increasing p-type thickness. The fabricated LEDs based on the samples under study showed the same variation trends in emission wavelength and efficiency. The results discussed in this paper are useful for optimizing the design of a high-indium LED.
This research was supported by National Science Council, Taiwan, The Republic of China, under the grants of NSC 99-2221-E-002-123-MY3, 100-2622-E-002-008- CC2, 100-2221-E-002-170, by NTU Excellent Research Project (10R80908-B), by Epistar Corporation (100H31025), and by US Air Force Scientific Research Office under the contract of AOARD-11-4114. One of the authors (Wenyu Cao) wishes to thank the MediaTek Fellowship for sponsoring her exchange study at National Taiwan University.
References and links
1. S. F. Chichibu, A. C. Abare, M. S. Minsky, S. Keller, S. B. Fleischer, J. E. Bowers, E. Hu, U. K. Mishra, L. A. Coldren, S. P. DenBaars, and T. Sota, “Effective band gap inhomogeneity and piezoelectric field in InGaN/GaN multiquantum well structures,” Appl. Phys. Lett. 73(14), 2006–2008 (1998). [CrossRef]
2. P. Riblet, H. Hirayama, A. Kinoshita, A. Hirata, T. Sugano, and Y. Aoyagi, “Determination of photoluminescence mechanism in InGaN quantum wells,” Appl. Phys. Lett. 75(15), 2241–2243 (1999). [CrossRef]
3. T. Takeuchi, S. Sota, M. Katsuragawa, M. Komori, H. Takeuchi, H. Amano, and I. Akasaki, “Quantum-confined Stark effect due to piezoelectric fields in GaInN strained quantum wells,” Jpn. J. Appl. Phys. 36(Part 2, No. 4A), L382–L385 (1997). [CrossRef]
4. T. Takeuchi, C. Wetzel, S. Yamaguchi, H. Sakai, H. Amano, I. Akasaki, Y. Kaneko, S. Nakagawa, Y. Yamaoka, and N. Yamada, “Determination of piezoelectric fields in strained GaInN quantum wells using the quantum-confined Stark effect,” Appl. Phys. Lett. 73(12), 1691–1693 (1998). [CrossRef]
5. C. F. Huang, C. Y. Chen, C. F. Lu, and C. C. Yang, “Reduced injection current induced blueshift in an InGaN/GaN quantum well light-emitting diode of prestrained growth,” Appl. Phys. Lett. 91(5), 051121 (2007). [CrossRef]
6. I. H. Ho and G. B. Stringfellow, “Solid phase immiscibility in GaInN,” Appl. Phys. Lett. 69(18), 2701–2703 (1996). [CrossRef]
7. Y. Narukawa, Y. Kawakami, S. Fujita, S. Fujita, and S. Nakamura, “Recombination dynamics of localized excitons in In0.20Ga0.80N/In0.05Ga0.95N multiple quantum wells,” Phys. Rev. B 55(4), R1938–R1941 (1997). [CrossRef]
8. S. Chichibu, K. Wada, and S. Nakamura, “Spatially resolved cathodoluminescence spectra of InGaN quantum wells,” Appl. Phys. Lett. 71(16), 2346–2348 (1997). [CrossRef]
9. L. K. Teles, J. Furthmuller, L. M. R. Scolfaro, J. R. Leite, and F. Bechstedt, “First-principles calculations of the thermodynamic and structural properties of strained InxGa1-xN and AlxGa1-xN alloys,” Phys. Rev. B 62(4), 2475–2485 (2000). [CrossRef]
10. Y.-S. Lin, K.-J. Ma, C. Hsu, S.-W. Feng, Y.-C. Cheng, C.-C. Liao, C. C. Yang, C.-C. Chou, C.-M. Lee, and J.-I. Chyi, “Dependence of composition fluctuation on indium content in InGaN/GaN multiple quantum wells,” Appl. Phys. Lett. 77(19), 2988–2990 (2000). [CrossRef]
11. C. K. Choi, B. D. Little, Y. H. Kwon, J. B. Lam, J. J. Song, Y. C. Chang, S. Keller, U. K. Mishra, and S. P. DenBaars, “Femtosecond pump-probe spectroscopy and time-resolved photoluminescence of an InxGa1-xN/GaN double heterostructure,” Phys. Rev. B 63(19), 195302 (2001). [CrossRef]
12. A. Kaneta, K. Okamoto, Y. Kawakami, S. Fujita, G. Marutsuki, Y. Narukawa, and T. Mukai, “Spatial and temporal luminescence dynamics in an InxGa1−xN single quantum well probed by near-field optical microscopy,” Appl. Phys. Lett. 81(23), 4353–4355 (2002). [CrossRef]
13. Y.-S. Lin, K.-J. Ma, C. Hsu, Y.-Y. Chung, C.-W. Liu, S.-W. Feng, Y.-C. Cheng, C. C. Yang, M.-H. Mao, H.-W. Chuang, C.-T. Kuo, J.-S. Tsang, and T. E. Weirich, “Quasiregular quantum-dot-like structure formation with postgrowth thermal annealing of InGaN/GaN quantum wells,” Appl. Phys. Lett. 80(14), 2571–2573 (2002). [CrossRef]
14. M. Rao, D. Kim, and S. Mahajan, “Compositional dependence of phase separation in InGaN layers,” Appl. Phys. Lett. 85(11), 1961–1963 (2004). [CrossRef]
15. S. W. Feng, T. Y. Tang, Y. C. Lu, S. J. Liu, E. C. Lin, C. C. Yang, K. J. Ma, C. H. Shen, L. C. Chen, J. Y. Lin, and H. X. Jiang, “Cluster size and composition variations in yellow and red light-emitting InGaN thin films upon thermal annealing,” J. Appl. Phys. 95(10), 5388–5396 (2004). [CrossRef]
16. Y. C. Cheng, E. C. Lin, C. M. Wu, C. C. Yang, J.-R. Yang, A. Rosenauer, K.-J. Ma, S.-C. Shi, L. C. Chen, C.-C. Pan, and J.-I. Chyi, “Nanostructures and carrier localization behaviors of green-luminescence InGaN/GaN quantum-well structures of various silicon-doping conditions,” Appl. Phys. Lett. 84(14), 2506–2508 (2004). [CrossRef]
17. H. C. Wang, S. J. Lin, Y. C. Lu, Y. C. Cheng, C. C. Yang, and K. J. Ma, “Carrier relaxation in InGaN/GaN quantum wells with nanometer-scale cluster structures,” Appl. Phys. Lett. 85(8), 1371–1373 (2004). [CrossRef]
18. H. C. Wang, Y. C. Lu, C. Y. Chen, and C. C. Yang, “Carrier capture times of the localized states in an InGaN thin film with indium-rich nanocluster structures,” Appl. Phys. Lett. 89(1), 011906 (2006). [CrossRef]
19. V. Fiorentini, F. Bernardini, F. Della Sala, A. Di Carlo, and P. Lugli, “Effects of macroscopic polarization in III-V nitride multiple quantum wells,” Phys. Rev. B 60(12), 8849–8858 (1999). [CrossRef]
20. F. Bernardini and V. Fiorentini, “Nonlinear macroscopic polarization in III-V nitride alloys,” Phys. Rev. B 64(8), 085207 (2001). [CrossRef]
21. Y. H. Cho, G. H. Gainer, A. J. Fischer, J. J. Song, S. Keller, U. K. Mishra, and S. P. DenBaars, “S-shaped temperature- dependent emission shift and carrier dynamics in InGaN/GaN multiple quantum wells,” Appl. Phys. Lett. 73(10), 1370–1372 (1998). [CrossRef]
22. P. Ruterana, S. Kret, A. Vivet, G. Maciejewski, and P. Dluzewski, “Composition fluctuation in InGaN quantum wells made from molecular beam or metalorganic vapor phase epitaxial layers,” J. Appl. Phys. 91(11), 8979–8985 (2002). [CrossRef]
23. I. K. Park, M. K. Kwon, J. O. Kim, S. B. Seo, J. Y. Kim, J. H. Lim, S. J. Park, and Y. S. Kim, “Green light-emitting diodes with self-assembled In-rich InGaN quantum dots,” Appl. Phys. Lett. 91(13), 133105 (2007). [CrossRef]
24. Y. H. Cho, Y. P. Sun, H. M. Kim, T. W. Kang, E.-K. Suh, H. J. Lee, R. J. Choi, and Y. B. Hahn, “High quantum efficiency of violet-blue to green light emission in InGaN quantum well structures grown by graded-indium-content profiling method,” Appl. Phys. Lett. 90(1), 011912 (2007). [CrossRef]
25. Y.-Y. Chung, Y.-S. Lin, S.-W. Feng, Y.-C. Cheng, E.-C. Lin, C. C. Yang, K.-J. Ma, C. Hsu, H.-W. Chuang, C.-T. Kuo, and J.-S. Tsang, “Quantum-well-width dependencies of postgrowth thermal annealing effects of InGaN/GaN quantum wells,” J. Appl. Phys. 93(12), 9693–9696 (2003). [CrossRef]
26. J. B. Limb, W. Lee, J. H. Ryou, D. Yoo, and R. D. Dupuis, “Comparison of GaN and In0.04Ga0.96N p-layers on the electrical and electroluminescence properties of green light emitting diodes,” J. Electron. Mater. 36(4), 426–430 (2007). [CrossRef]
27. J.-H. Ryou, W. Lee, J. Limb, D. Yoo, J. P. Liu, R. D. Dupuis, Z. H. Wu, A. M. Fischer, and F. A. Ponce, “Control of quantum-confined Stark effect in InGaN/GaN multiple quantum well active region by p-type layer for III-nitride-based visible light emitting diodes,” Appl. Phys. Lett. 92(10), 101113 (2008). [CrossRef]
28. S. Watanabe, N. Yamada, M. Nagashima, Y. Ueki, C. Sasaki, Y. Yamada, T. Taguchi, K. Tadatomo, H. Okagawa, and H. Kudo, “Internal quantum efficiency of highly-efficient InxGa1−xN-based near-ultraviolet light-emitting diodes,” Appl. Phys. Lett. 83(24), 4906–4908 (2003). [CrossRef]
29. A. Sasaki, S. Shibakawa, Y. Kawakami, K. Nishizuka, Y. Narukawa, and T. Mukai, “Equation for internal quantum efficiency and its temperature dependence of luminescence, and application to InxGa1-xN/GaN multiple quantum wells,” Jpn. J. Appl. Phys. 45(11), 8719–8723 (2006). [CrossRef]
30. E. M. Goldys, M. Godlewski, R. Langer, A. Barski, P. Bergman, and B. Monemar, “Analysis of the red optical emission in cubic GaN grown by molecular-beam epitaxy,” Phys. Rev. B 60(8), 5464–5469 (1999). [CrossRef]
31. M. Leroux, N. Grandjean, B. Beaumont, G. Nataf, F. Semond, J. Massies, and P. Gibart, “Temperature quenching of photoluminescence intensities in undoped and doped GaN,” J. Appl. Phys. 86(7), 3721–3728 (1999). [CrossRef]
32. H. P. D. Schenk, M. Leroux, and P. de Mierry, “Luminescence and absorption in InGaN epitaxial layers and the van Roosbroeck–Shockley relation,” J. Appl. Phys. 88(3), 1525–1543 (2000). [CrossRef]
33. X. Ni, Q. Fan, R. Shimada, Ü. Özgür, and H. Morkoç, “Reduction of efficiency droop in InGaN light emitting diodes by coupled quantum wells,” Appl. Phys. Lett. 93(17), 171113 (2008). [CrossRef]
34. Y. K. Kuo, J. Y. Chang, M. C. Tsai, and S. H. Yen, “Advantages of blue InGaN multiple-quantum well light-emitting diodes with InGaN barriers,” Appl. Phys. Lett. 95(1), 011116 (2009). [CrossRef]
35. S. Chiaria, E. Furno, M. Goano, and E. Bellotti, “Design criteria for near-ultraviolet GaN-based light-emitting diodes,” IEEE Trans. Electron Dev. 57(1), 60–70 (2010). [CrossRef]
36. S. Fujita, M. Funato, D. C. Park, Y. Ikenaga, and S. Fujita, “Electrical characterization of MOVPE-grown p-type GaN:Mg against annealing temperature,” MRS Proc. 537, G6.31 (1998).
37. F. Bernardini and V. Fiorentini, “Spontaneous versus piezoelectric polarization in III-V nitrides: Conceptual aspects and practical consequences,” Phys. Status Solidi B 216(1), 391–398 (1999). [CrossRef]
38. C. H. Liao, C. Y. Chen, H. S. Chen, K. Y. Chen, W. L. Chung, W. M. Chang, J. J. Huang, Y. F. Yao, Y. W. Kiang, and C. C. Yang, “Emission efficiency dependence on the overgrown p-GaN thickness in a high-indium InGaN/GaN quantum-well light-emitting diode,” IEEE Photon. Technol. Lett. 23(23), 1757–1759 (2011). [CrossRef]