Deep level defects in the multi-quantum well (MQW) region of InGaN/GaN light emitting diodes (LEDs) were investigated. InGaN quantum well and GaN quantum barrier defect states were distinguished using bias-dependent steady-state photocapacitance and deep level optical spectroscopy, and their possible physical origin and potential impact on LED performance is considered. Lighted capacitance-voltage measurements provided quantitative and nanoscale depth profiling of the deep level concentration within the MQW region. The concentration of every observed deep level varied strongly with depth in the MQW region, which indicates evolving mechanisms for defect incorporation during MQW growth.
© 2012 OSA
InGaN/GaN-based light emitting diodes (LEDs) are an essential component for energy-efficient solid-state lighting (SSL). To fully realize the benefits of SSL, the efficiency of InGaN/GaN LEDs themselves must be improved. Defects have been associated with deleterious phenomena limiting LED performance, such as reduced peak efficiency , decreasing efficiency with increasing emission wavelength (i.e. the green-yellow gap) [2,3] and efficiency roll-off with increasing carrier density [4–6]. However, little is known about the physical origin or electrical and optical properties of defects in InGaN/GaN LEDs. This due to the lack of a deep level defect spectroscopy that is at once quantitative, able to probe deep levels in wide band gap (Eg), blue-emitting InxGa1-xN materials (Eg ~2.9 eV), and able to spatially resolve quantum well (QW) versus quantum barrier (QB) defect states. For example, previous work has applied deep level transient spectroscopy (DLTS) to InGaN/GaN LEDs but was unable to discern InGaN- versus GaN-related defect levels, was unable to quantify the deep level density in the MQW region, and was unable to probe defect states deeper than ~1 eV from the conduction band edge due the reliance of thermal emission . Photocurrent spectroscopy has also been employed to study the influence of defect states in InGaN/GaN LEDs, but this method was unable to observe any deep level directly, leaving the nature of the corresponding defects unknown .
To address these challenges, we applied deep level optical spectroscopy (DLOS) steady-state photocapacitance (SSPC) and lighted capacitance-voltage  (LCV) to In0.13Ga0.87N/GaN LEDs to quantitatively survey deep level defects in the MQW region. The optical nature of DLOS and SSPC enabled probing near-mid-band gap deep levels in (In)GaN, which are commonly expected to be the most efficient non-radiative recombination centers and hence are most pertinent to LED IQE. The bias dependence of SSPC spectra distinguished MQW- from n-GaN bulk-related deep levels , and DLOS quantified the energy of deep levels in the band gap. Differentiation among QW- and QB-related deep levels was then achieved by comparison of LED MQW-related DLOS spectra to previous studies of GaN and InGaN. LCV measurements provided quantitative and nanoscale depth profiling of the deep level concentration (Nt) for individual QWs and QBs and revealed a strong depth-dependence for defect incorporation in the MQW region. These are powerful distinguishing characteristics of DLOS-LCV, especially when applied to InGaN/GaN LEDs because quantifying the energy, density and location of defect levels within a multi-heterostructure is critical to assessing their physical source, impact on device operation, and devising strategies to mitigate their incorporation through material growth optimization. The microscopic origin of observed deep levels and their potential impact on LED operation are considered on the basis of their location in the MQW region and energetic position in the (In)GaN band gap.
InGaN/GaN MQW LEDs emitting at 440 nm were grown on GaN-on-sapphire templates by metal-organic vapor-phase epitaxy (MOVPE). The MQW structure has an n-type GaN layer grown at 1050 °C, followed by a MQW region consisting of five unintentionally doped (UID) 2.5-nm-thick In0.13Ga0.87N QWs grown at 770 °C placed between 7.5-nm-thick Si-doped GaN QBs grown at 850 °C, followed by a 30-nm-thick p-type Al0.15Ga0.85N electron-block layer (EBL) and capped with a 400-nm-thick p-type GaN contact layer. The Si doping target in the QBs was 1 × 1018 cm−3 and 3 × 1018 cm−3 in the n-GaN bulk, and the Mg doping target in the p-layers was 3 × 1019 cm−3. A threading dislocation density of 5.3 × 108 cm−2 was measured from x-ray diffraction (XRD) peak widths . Devices were patterned into 300 μm x 300μm sized mesas using inductively-coupled plasma etching. Ohmic n-type and p-type electrical contacts were formed by evaporating a Ti/Al/Ni/Au and a transparent NiO/Au metal stack, respectively.
DLOS, SSPC, dark capacitance-voltage (CV) and LCV were conducted at room temperature to examine deep level defects in the MQW and n-GaN bulk regions of the LEDs. DLOS was performed with a Boonton 7200 capacitance meter, which has a frequency of 1 MHz, for fast sampling of the photocapacitance transients. LCV was performed with an HP4284A meter at 50 kHz. The lower frequency for LCV measurements was chosen to minimize the in-phase element of the impedance. The difference in the capacitance measured at the two frequencies was small for all bias values, indicating no significant series resistance issues. The LEDs had low leakage at −8 V under all illumination conditions. The phase angle for CV and LCV measurements at 50 kHz remained between −88° and −90° under illumination from −8 – 2 V. The phase angle for DLOS measurements at 1 MHz remained between −88° and −83° under illumination from −8 – 0 V. These phase angles were near the −90° ideal angle for a purely capacitive element, confirming that neither leakage nor series resistance influenced the CV, LCV or DLOS measurements. DLOS and SSPC measure the photocapacitance response from deep level photoemission in depleted regions of the diode upon exposure to sub-band gap, monochromatic illumination. DLOS and SSPC measurements were conducted using broadband illumination from a 150 W Xe arc lamp dispersed through a monochromator using appropriate mode sorting filters to provide a photon energy range of 1.20 – 3.60 eV at 25 meV resolution. The photon flux (ϕ) was calculated from the optical power measured using a Newport 1918-C power meter and a Si photodiode detector divided by the photon energy and the focused beam area. The photon flux varied between 1 – 20 × 1016 cm−2s−1 over the scanned range. The saturated photocapacitance was recorded at each photon energy for SSPC. Inflection points in SSPC spectra occur at the onset of deep level photoemission and approximately indicate the energy of the deep level in the band gap. DLOS determined the spectral variation of the deep level optical cross-section (σo) from the time derivative of the photocapacitance transient at t = 0 s, the time when illumination begins, divided by ϕ. Fitting σo to an appropriate model  that accounts for lattice relaxation of the deep level defect precisely determines the optical ionization energy (Eo) and the Franck-Condon energy (dFC). Uncertainty in fitted values of Eo and dFC are estimated to be 0.05 eV from the spread of values obtained by fitting multiple subsets of DLOS data by excluding data points at the high and low photon energy range. DLOS and SSPC measurements were conducted at reverse bias and fixed photon energy by digitizing photocapacitance transients that were simultaneously sampled at 1 ms and 70 ms using two separate recording instruments. This enables capturing of transients with time constants of a few ms to several seconds. Following each DLOS transient, a filling pulse bias Vf was applied in the dark for a 20 s duration to re-populate deep levels. The deep level density was calculated from LCV . Additional LCV details are given in Sect. 4.
3. Deep level characterization
3.1 Distinguishing MQW and n-GaN defects using SSPC
Defects in the MQW and bulk n-GaN were distinguished using depth-dependent SSPC measurements. Depth-dependence of SSPC (and DLOS) arises because only defects located in the depletion region contribute to the photocapacitance response (ΔC). Therefore, it is important to understand how V controls xd in the LED. It will be shown that there is a threshold voltage (Vth) such that MQW deep levels are identified by their appearance in SSPC spectra for V > Vth. Additionally, n-GaN bulk-related deep levels are identified by their appearance in SSPC spectra only when V < Vth. Comparison of MQW deep level spectra to prior DLOS studies of n-GaN [13–15] and UID-InGaN [10,16] in Section 3.2 further differentiates QW- vs. QB-related defects.
Figure 1 shows a CV curve and the associated space-charge profile (ρ) that was measured in the dark to determine Vth. The apparent depletion depth (xd) was determined as xd = Aε/C, where ε is the semiconductor relative permittivity and A is the diode area. Here, ε was taken to be 9.3ε0, which is a weighted average over the QWs and QBs in the MQW region. The spatial resolution of ρ was limited by the effective Debye length (LD), which is difficult to define in a MQW region where the free carrier and dopant concentrations vary abruptly over a new nanometers. Applying the usual expression for LD amid a mean background of ~1 × 1018 cm−3 doping in the MQW region gives LD ~3 nm, which is reasonable since QWs nominally spaced 7.5 nm apart are distinct in Fig. 1(b). It was assumed that xd had negligible extension into the p-AlGaN EBL due to the much heavier (~30x) p-type doping relative to the n-type doping in the QBs. Therefore, xd can be interpreted as the depletion depth relative to the edge of the p-AlGaN EBL. QWs 1 – 3 appear as peaks in ρ, and the adjacent n-GaN region appears as a plateau below the MQW region. It is noted that the measured position of the QWs determined by CV agrees with their position relative to the p-AlGaN EBL expected from the MQW periodicity determined from XRD, confirming that the depletion region does not penetrate significantly into the EBL. Figure 1(b) shows that Vth = −1.7 V, such that V > Vth depletes only the MQW region (xd < 57.5 nm) and V < Vth depletes past the MQW and into the n-GaN bulk (xd > 57.5 nm). Therefore, SSPC measurements of this LED at V ≥ −1.7 V reveal only QW and QB deep levels, while deep levels located in the n-GaN bulk are identified by their emergence in SSPC measurements only when V < 1.7 V. Similarly, defect states in the p-AlGaN EBL were not expected to significantly contribute to SSPC spectra at any Vr due to neglible depletion in the EBL relative to the total depletion region.
With this distinction in mind, Fig. 2(a) presents the SSPC spectra of the LED measured at V = −1.7 V (Vf = 2 V), corresponding to xd ~57 nm, and V = −8 V (Vf = −4 V), corresponding to xd ~100 nm. A Vf of 2 V was used to refill deep levels in the MQW region, and a Vf of −4 V was used to refill deep levels only in the n-GaN bulk. The evolution of the SSPC spectra as a function of V, i.e. xd, provides insight into the location of the observed defects. Before considering variation in SSPC spectra with bias, it is important to realize that the magnitude of ΔC can depend on the spectral intensity of the lamp for the case of thermal carrier re-capture in the depletion region . However, this effect impacts all deep levels equally for a given Vr, so the SSPC line shape is unaffected. Reduced carrier re-capture with increasing Vr did not significantly influence the SSPC results in this study, as it is shown below that the primary impact of Vr on the SSPC spectra is to reduce ΔC.
The −1.7 V SSPC spectrum revealed four deep levels, observed by their photoemission onsets (i.e. distinct changes in the slope of ΔC) at 1.60 eV, 2.05 eV, 2.60 eV and 2.70 eV, and all of the associated defects are located in the MQW region because V = Vth. Sensitivity to the QWs at V = −1.7 V was verified by the peak at 2.85 eV in the SSPC spectrum, which arises from near-band-edge absorption by In0.13Ga0.87N. The −8 V SSPC spectrum evidenced an emergent deep level onset at 3.25 eV that is immediately attributed to the n-GaN bulk because it appears only for V < Vth. Table 1 summarizes the characteristics and location of these deep level defects as determined from SSPC and DLOS analyses described in section 3.2.
It is interesting that the MQW-related deep level onsets at 2.05 eV and 2.60 eV persisted while the MQW-related deep level onsets at 1.60 eV and 2.70 eV were quenched from the SSPC spectrum for V = −8 V < Vth. This can be understood by considering how the extent of xd influences not only which deep levels contribute to ΔC but also the relative magnitudes of their contribution. Equation (1) gives the contribution to ΔC for a portion of the depletion region spanning depths x1 < x2 < xd in the limit where Nt << Nd :
3.2. Deep level characterization using DLOS
While SSPC determined the location of the defects in the LED, DLOS measurements were performed to study σo and quantify Eo for the deep levels. Comparing σo of the MQW-related deep levels to previous studies of n-GaN [13–15] and UID-InGaN  also distinguished QW and QB defects and helped to assess their physical origin. Figure 2(b) presents the DLOS spectra collected simultaneously with the SSPC data of Fig. 2(a). The symbols are the data, and the lines in the spectra are least-squares fits to the model of Pässler .
Considering first the DLOS spectrum taken at V = −1.7 V that was particular to the MQW region, the spectral variation of σo between 1.2 – 2.1 eV and 2.7 – 2.9 eV that correspond to the SSPC onsets at 1.60 eV and 2.70 eV SSPC, respectively, were well fit to the theoretical model to determine Eo values of 1.62 (dFC = 0.32 eV) and 2.76 eV (negligible dFC), respectively. These Eo values are referenced to the conduction band minimum (Ec) because the associated ΔC were positive. The sizeable dFC value for the Ec – 1.62 eV level is consistent with an SSPC onset energy less than Eo due to phonon-assisted photoemission and indicates significant lattice coupling of the corresponding defect center.
Comparing the −1.7 V DLOS spectra of the MQW-related deep levels to previous DLOS study of UID-InGaN helps to identify QW-related defects. Gür et al.  recently reported defect states in UID-In0.2Ga0.8N grown by molecular beam epitaxy (MBE) at Ec – 1.45 eV and Ec – 2.50 eV, whose σo spectra are qualitatively similar to the Ec – 1.62 eV and Ec – 2.76 eV levels reported here. Based on these findings, the Ec – 1.62 eV and Ec – 2.76 eV levels observed in this study are assigned to the LED In0.13Ga0.77N QW layers. The discrepancy in Eo for the deep levels reported here and in  can be understood in light of the difference in indium mole fraction of the respective InGaN layers, which can shift the deep level energy relative to Ec. The similarities in InGaN DLOS spectra reported here and in , despite the difference in indium content and growth method, suggest that UID-InxGa1-xN (x = 0.13 – 0.2) grown in the midst of an n-type background tends to incorporate a common set of defect types over a wide window of growth regimes and growth conditions. Impurities like carbon and intrinsic defects such as the cation vacancy or TDs are speculative but reasonable defect sources given their prevalence and propensity to form deep levels in n-GaN [13–15,18].
The DLOS spectrum at V = −8 V examined n-GaN bulk-related defect states and helped to determine the origin of the 2.05 eV and 2.60 eV SSPC onsets. The −8 V DLOS spectrum shown in Fig. 2(b) is plotted discontinuously at 2.60 eV to better compare its spectral variation to the −1.7 V spectrum between 2.60 – 3.50 eV. The DLOS spectra show an evolution with bias similar to the SSPC spectra. As expected, the QW-related Ec - 1.62 eV and Ec - 2.76 eV deep levels evident at −1.7 V did not appear in the −8 V DLOS spectrum. Without the overlapping spectrum from the Ec – 1.62 eV level, σo between 2.0 – 2.50 eV corresponding to the 2.05 eV SSPC onset could be fit to determine its energetic position at Ec – 2.11 eV with dFC = 0.39 eV. In lieu of the Ec – 2.76 eV level with negligible dFC observed at -1.7 V, a different deep level at Ec – 2.73 eV with sizable dFC = 0.21 eV emerges in the −8 V DLOS spectrum. The Ec – 2.73 eV defect state corresponds to the 2.60 eV SSPC onset at −1.7 V and −8 V, where again the large dFC explains its red-shifted photoemission onset. Despite the similar Eo values, the much broader σo and correspondingly larger dFC of the Ec – 2.73 eV level make clear its distinctness from the QW-related Ec – 2.76 eV defect state. The spectral variation of σo associated with the 3.25 eV onset in the −8 V SSPC spectrum was not well resolved, so this n-GaN bulk deep level was approximated as Ec – 3.25 eV.
Looking to the large body of work on defect levels in GaN provides insight into possible physical origins of the MQW-related deep levels at Ec – 2.11 eV and Ec – 2.73 eV and the n-GaN bulk-related defect state at Ec – 3.25 eV. The Ec – 3.25 eV is similar to defect states widely reported for bulk MOVPE-grown n-GaN that could be related to either unintentional Mg through reactor memory effects  or carbon contamination . The energy level and broad absorption spectrum for the MQW-related Ec – 2.73 eV defect state resembles deep levels previously reported for carbon impurities [14,18] and also VGa in n-GaN , suggesting that this defect state is associated with the QB layers. Given the prevalence of the potential defect sources for the Ec – 2.73 and Ec – 3.25 eV levels, it is likely that these levels exist in both the GaN QBs and bulk n-GaN regions. The MQW-related level at Ec – 2.11 eV has not been previously reported to our knowledge. The absence of this defect state in prior DLOS studies of UID-InGaN  and bulk n-GaN grown by MOVPE at Tg ~1000 °C [13–15] prompts a tentative association of the Ec – 2.11 eV level to the n-GaN QBs that were grown at Tg ~850 °C for these LEDs. The emergence of additional defect levels with reduced Tg is reasonable, as reducing Tg can increase the propensity for point defect incorporation. The physical origin of the Ec – 2.11 eV deep level remains under further investigation.
The importance of understanding the physical sources of deep levels in LEDs is matched by the need to understand their potential impact on LED operation. Knowledge of the energetic position of deep levels in the (In)GaN band gap and their spatial distribution in the MQW-region obtained from DLOS affords invaluable insight into their roles as electronic traps and non-radiative recombination centers. The Ec – 1.62 eV InGaN-related deep level stands out as a potentially effective non-radiative recombination center given its physical location in the QWs and energetic position near the middle of Eg,InGaN. The QB-related defect level at Ec – 2.11 eV may also behave as non-radiative recombination center. Wave functions of electrons and holes located in the QW can penetrate significantly into the QB because of the strong polarization-induced electric fields in LEDs grown on the c-plane. This would allow for carriers in the QWs to non-radiatively recombine via QB defect states, at the expense of IQE. Additionally, trapping by the remaining QB and QW defect states can impede the transport of electrons and especially holes through the MQW region, which would degrade LED external quantum efficiency.
4. Depth dependence of deep level density
Having determined the location of deep level defects, the LCV technique was used to quantitatively profile their depth distribution. The density of defects within a particular layer of the MQW region or adjacent n-GaN bulk can be determined from the difference in V (ΔV) required to achieve the same xd (and therefore C) with deep levels fully occupied or empty of electrons . This voltage shift is illustrated in the inset of Fig. 1(a). From the Poisson equation, V depends on xd asEq. (2) that the deep level is sufficiently deep in the band gap to be fully occupied throughout the entire depletion region in the absence of light at the measurement temperature (> 1eV for 293 K) and thus. In this case, the ΔV required to reach xd when defects are emptied by illumination (nt = 0) compared to when the defects are fully occupied (nt = Nt) is
Using this method, Nt(x) was calculated for the deep levels using Eq. (3) by performing sequential LCV scans at selected deep-level-dependent photon energies. For example, [Ec – 1.62 eV] (brackets indicate the density of the deep level defect) was measured from ΔV under 2.10 eV illumination relative to the dark CV. Referring to Fig. 2(a), at 2.10 eV, only the Ec – 1.62 eV is optically stimulated. Likewise, [Ec – 2.11 eV] was calculated from ΔV under 2.60 eV illumination relative to 2.10 eV illumination, and [Ec – 2.76 eV] was calculated from ΔV under 2.85 eV relative to 2.60 eV illumination. A net Nt was calculated for the remaining defect states at Ec – 2.73 eV and 3.25 eV from ΔV at 3.30 eV illumination relative to ΔV at 2.85 eV.
Figure 3 plots ΔV against xd for the QW-related Ec – 2.76 eV deep level. The contribution to ΔV for each QW is marked by horizontal lines. Plateaus in ΔV occur when xd reaches just past the QWs. Application of Eq. (3) yields the distribution of [Ec – 2.76 eV] in the MQW region for both LEDs as shown in Fig. 4 . Individual Nt values could not be measured for QW4 and QW5 because they remained within the depletion region when the applied bias (forward in this case) turned on the diode. In this case, an average Nt was calculated from the net ΔV accumulated by both QWs. The same analysis was applied to ΔV data for the remaining deep levels in the MQW region, and the results are also shown in Fig. 4.
Figure 4 demonstrates a strong depth-dependence skewed toward the n-side of the MQW region for every observed deep level. This observation is similar to previous work in AlGaAs/GaAs MQWs, where it was found that the growth of AlGaAs/GaAs heterointerfaces improved the structural and optical quality of subsequently grown QWs . For the case of AlGaAs/GaAs, it was suggested that reduced defect incorporation in the optically active region improved QW luminescence due to impurity trapping at the underlying heterointerfaces and/or strain-induced defect gettering . Previous studies have also reported reduced deep level incorporation and improved IQE in InGaN/GaN MQW with the insertion of an 50 nm thick In0.04Ga0.96N underlying layer (UL) below the MQW region, where it was suggested that the indium atoms in the UL itself were responsible for defect reduction . It is not clear for the LED under study if the depth dependence of Nt occurs due to defect gettering by InGaN/GaN heterointerfaces, the InGaN QW layers, or some other mechanism. Determining the impact of InGaN ULs and InGaN/GaN heterointerfaces on defect incorporation and optical quality in the MQW region of LEDs will the subject of future work.
Figure 4 also reveals several trends for deep level formation versus physical location in the MQW region and energetic position in the band gap. Overall, the In0.13Ga0.87N QW layers incorporate larger Nt compared to the GaN QBs, despite the QBs being grown ~200 °C below the typical Tg ~1050 °C for n-GaN. This implies that defect-induced non-radiative recombination is most likely to occur in the QWs. Comparatively large Nt in the QWs could result from the relatively low Tg for In0.13Ga0.87N, strain, or In0.13Ga0.87N thermal degradation from exposure to elevated temperatures during subsequent GaN growth (which are ~80 °C higher than QW Tg). Increasing the indium mole fraction in the QWs to achieve longer wavelength LED emission (i.e. green and yellow) is likely to further exacerbate all of these possible mechanisms for QW defect incorporation. Therefore, QW-related deep levels are anticipated to contribute to the green-yellow gap problem for InGaN/GaN LEDs [2,3]. For the case of near-mid-band gap levels, [Ec – 1.62 eV] in the QWs is ~10x greater than [Ec – 2.11 eV] in the QBs throughout the MQW region. This observation reinforces the concern that the Ec – 1.62 eV deep state could be the chief source of defect-related non-radiative recombination in the MQW region of the LED. The deep level concentration of the Ec – 1.62 eV level in the QW and the Ec – 2.11 eV level in the QB follow a qualitatively similar trend versus depth, decreasing by ~7x from the n-side to the p-side of the MQW region. This is a much stronger depth-dependence than was exhibited by the Ec – 2.76 eV QW-related defect state and the Ec – 2.73 eV and Ec – 3.25 eV levels in the QBs, whose energies are all within ~0.1 – 0.7 eV of the valence band minimum. It appears that the mechanism for reducing defect incorporation as MQW proceeds is fortuitously effective at sequestering defects that form near-mid-band gap deep levels, which would be quite beneficial to LED efficiency, especially for the case where the QW closest to the p-side produces the majority of LED electroluminescence .
SSPC and DLOS measurements identified several deep levels associated with the QW, QB and n-GaN bulk regions of a 440 nm emitting InGaN/GaN LED. The location of deep level defects in the MQW region was determined by bias-dependent SSPC-DLOS and comparing the observed defect optical cross-sections with previous studies of InGaN and GaN films. The role of deep levels to reduce LED efficiency as non-recombination centers or trapping states was considered based on their energetic position in the band gap and spatial position in the MQW region. Near-mid-band gap levels were observed in both QW and QB layers, and the former is expected to be a major source of non-radiative recombination. The deep level density in the MQW region was measured using LCV. LCV analysis showed that the density of every observed deep level was significantly higher on the n-side of the MQW region, indicative of rapidly evolving mechanisms for defect incorporation during MQW growth. Overall, Nt in the QW layers was much higher than in the QBs. This observation suggests that defect-related non-radiative recombination in the LEDs occurs primarily in the QWs. It is posited that QW-related defects contribute to the green-yellow gap because the changes in QW growth conditions that are required to increase indium incorporation are anticipated to also enhance QW defect incorporation.
The authors would like to thank Andrew A. Allerman for helpful discussions. This work was supported by Sandia’s Solid-State Lighting Science Energy Frontier Research Center, sponsored by the Department of Energy Office of Basic Energy Science. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Company, for the United States Department of Energy's National Nuclear Security Administration under Contract DE-AC04-94AL85000.
References and links
1. M. F. Schubert, S. Chhajed, J. K. Kim, E. F. Schubert, D. D. Koleske, M. H. Crawford, S. R. Lee, A. J. Fischer, G. Thaler, and M. A. Banas, “Effect of dislocation density on efficiency droop in GaInN/GaN light-emitting diodes,” Appl. Phys. Lett. 91(23), 231114 (2007). [CrossRef]
2. S. F. Chichibu, A. Uedono, T. Onuma, B. A. Haskell, A. Chakraborty, T. Koyama, P. T. Fini, S. Keller, S. P. Denbaars, J. S. Speck, U. K. Mishra, S. Nakamura, S. Yamaguchi, S. Kamiyama, H. Amano, I. Akasaki, J. Han, and T. Sota, “Origin of defect-insensitive emission probability in In-containing (Al,In,Ga)N alloy semiconductors,” Nat. Mater. 5(10), 810–816 (2006). [CrossRef] [PubMed]
3. H. Yamada, K. Iso, M. Saito, H. Masui, K. Fujito, S. P. DenBaars, and S. Nakamura, “Compositional Dependence of Nonpolar m-Plane InxGa1-xN/GaN Light Emitting Diodes,” Appl. Phys. Express 1, 041101 (2008). [CrossRef]
4. J. Hader, J. V. Moloney, and S. W. Koch, “Temperature-dependence of the internal efficiency droop in GaN-based diodes,” Appl. Phys. Lett. 99(18), 181127 (2011). [CrossRef]
5. T. Mukai, M. Yamada, and S. Nakamura, “Characteristics of InGaN-Based UV/Blue/Green/Amber/Red Light-Emitting Diodes,” Jpn. J. Appl. Phys. 38(Part 1, No. 7A), 3976–3981 (1999). [CrossRef]
6. B. Monemar and B. E. Sernelius, “Defect related issues in the “current roll-off” in InGaN based light emitting diodes,” Appl. Phys. Lett. 91(18), 181103 (2007). [CrossRef]
7. L. Rigutti, A. Castaldini, and A. Cavallini, “Anomalous deep-level transients related to quantum well piezoelectric fields in InyGa1−yN/GaN-heterostructure light-emitting diodes,” Phys. Rev. B 77(4), 045312 (2008). [CrossRef]
8. L. Rigutti, A. Castaldini, M. Meneghini, and A. Cavallini, “Photocurrent spectroscopy evidence for stress-induced recombination centres in quantum wells of InGaN/GaN-based light-emitting diodes,” Semicond. Sci. Technol. 23(2), 025004 (2008). [CrossRef]
9. A. Armstrong, A. R. Arehart, and S. A. Ringel, “A method to determine deep level profiles in highly compensated, wide band gap semiconductors,” J. Appl. Phys. 97(8), 083529 (2005). [CrossRef]
10. A. Armstrong, M. H. Crawford, and D. D. Koleske, “Quantitative and depth-resolved investigation of deep-Level defects in InGaN/GaN heterostructures,” J. Electron. Mater. 40(4), 369–376 (2011). [CrossRef]
11. S. R. Lee, A. M. West, A. A. Allerman, K. E. Waldrip, D. M. Follstaedt, P. P. Provencio, D. D. Koleske, and C. R. Abernathy, “Effect of threading dislocations on the Bragg peakwidths of GaN, AlGaN, and AlN heterolayers,” Appl. Phys. Lett. 86(24), 241904 (2005). [CrossRef]
12. R. Passler, “Photoionization cross-section analysis for a deep trap contributing to current collapse in GaN field-effect transistors,” J. Appl. Phys. 96(1), 715–722 (2004). [CrossRef]
13. A. Armstrong, A. R. Arehart, B. Moran, S. P. DenBaars, U. K. Mishra, J. S. Speck, and S. A. Ringel, “Impact of carbon on trap states in n-type GaN grown by metalorganic chemical vapor deposition,” Appl. Phys. Lett. 84(3), 374–376 (2004). [CrossRef]
14. P. B. Klein, J. A. Freitas Jr, S. C. Binari, and A. E. Wickenden, “Observation of deep traps responsible for current collapse in GaN metal–semiconductor field-effect transistors,” Appl. Phys. Lett. 75(25), 4016–4018 (1999). [CrossRef]
15. A. Hierro, D. Kwon, S. A. Ringel, M. Hansen, J. S. Speck, U. K. Mishra, and S. P. DenBaars, “Optically and thermally detected deep levels in n-type Schottky and p+-n GaN diodes,” Appl. Phys. Lett. 76(21), 3064–3066 (2000). [CrossRef]
16. E. Gür, Z. Zhang, S. Krishnamoorthy, S. Rajan, and S. A. Ringel, “Detailed characterization of deep level defects in InGaN Schottky diodes by optical and thermal deep level spectroscopies,” Appl. Phys. Lett. 99(9), 092109 (2011). [CrossRef]
17. P. Blood and J. W. Orton, The Electrical Characterization of Semiconductors: Majority Carriers and Electron States (Academic, London, 1990), Chap. 7.
18. J. L. Lyons, A. Janotti, and C. G. Van de Walle, “Carbon impurities and the yellow luminescence in GaN,” Appl. Phys. Lett. 97(15), 152108 (2010). [CrossRef]
19. S. D. Brotherton, “Measurement of deep-level spatial distributions,” Sol. St. Elec. 19(4), 341–342 (1976). [CrossRef]
20. P. M. Petroff, R. C. Miller, A. C. Gossard, and W. Wiegmann, “Impurity trapping, interface structure, and luminescence of GaAs quantum wells grown by molecular beam epitaxy,” Appl. Phys. Lett. 44(2), 217–219 (1984). [CrossRef]
21. T. Akasaka, H. Gotoh, T. Saito, and T. Makimoto, “High luminescent efficiency of InGaN multiple quantum wells grown on InGaN underlying layers,” Appl. Phys. Lett. 85(15), 3089–3091 (2004). [CrossRef]
22. A. David, M. J. Grundmann, J. F. Kaeding, N. F. Gardner, T. G. Mihopoulos, and M. R. Krames, “Carrier distribution in (0001) InGaN/GaN multiple quantum well light-emitting diodes,” Appl. Phys. Lett. 92(5), 053502 (2008). [CrossRef]