Design analysis of phosphor-free monolithic white light-emitting-diodes with InGaN/ InGaN multiple quantum wells on ternary InGaN substrates

Tremendous advancement in III-nitride light-emitting-diodes (LEDs) over the last two decades has prompted LEDs as promising candidate in substituting traditional lamps in myriad lighting applications.^1–4 Specifically, solid state lighting based on white color LEDs is considered as the next-generation illumination system due to the high efficiency and reliable device performance. Conventionally, InGaN-based blue-emitting LED is coated with yellow phosphor to convert part of the blue emission to longer wavelength (λ) for white light generation.^1,2,4,5 However, this phosphor-converted (pc) white LEDs exhibit large Stokes loss (∼10 to ∼30%) due to the wavelength down-conversion.⁵ Additionally, those pc-LEDs suffer from stability issues such as phosphor-aging, packaging cost issues, and dependency on the efficiency of GaN-based LEDs.⁵ Meanwhile, “multichip” approach² has also been proposed for white-color generation by combining red, green and blue (RGB) monochromatic LED chips, which demonstrates several significant advantages over pc-LEDs. However, this approach suffers from significantly increased cost from additional processing steps for chips integration and the need of driver circuitry.

Thus, the pursuit of high-efficiency, cost-effective, and monolithic white LEDs has attracted tremendous attentions.^6–19 Previous works have reported the possibility of fabricating phosphor-free monolithic white LED by stacking multi-color-emitting InGaN/ GaN quantum wells (QWs) on GaN substrate.^6–10 However, it is very challenging to incorporate high In-content into InGaN/ GaN QWs on GaN substrate which is critical for green and yellow emission wavelengths due to charge separation issue from large lattice-mismatch strain.^20,21 On the other hand, nanostructure engineering approaches such as the use of quantum dots,^11,12 nanowires,^13–15 pyramids^16,17 and patterned substrates^18,19 have also been pursued previously to achieve white LEDs. Nevertheless, these methods all require complex fabrication processes which are challenging to be implemented into large scale production. Therefore, alternative cost-effective and high-efficiency solutions are in great demand for white LEDs.

Recent studies have pointed out that the use of ternary InGaN substrates^22,23 for InGaN/ InGaN QW LEDs could lead to promising light output covering the entire visible spectrum including green and yellow, which demonstrated ∼3 times enhanced spontaneous emission rate for λ ∼ 450 nm – 645 nm than those conventional InGaN/ GaN QWs on GaN substrates. Additionally, experimental works reporting successful growth of high In-content ternary InGaN substrate^24–28 would enable the growth of InGaN LEDs on ternary substrate. Specifically, recent works by Hoffbauer and coworkers have demonstrated the growth of high quality InGaN films on sapphire, with In-content up to 40% and film thickness up to 1 μm,^27,28 using Molecular Beam Epitaxy (MBE). Thus, it is anticipated that those ternary substrates would be suitable for InGaN-based green and yellow emitters, which would be of great interest for feasible monolithic white LEDs.

In this letter, we proposed and investigated the use of ternary InGaN substrates for high-efficiency monolithic tunable white LEDs. Nanostructure engineering has been conducted to the multiple QWs (MQWs) active region on ternary substrate to achieve white color illumination. Simulation studies show that the proposed and optimized white LED device structure exhibits ∼2 times larger external quantum efficiency (EQE)⁷ and higher light output power as compare to conventional InGaN/ GaN white LEDs on GaN substrates.^6–10 The chromaticity coordinates and correlated color temperature (CCT) of the proposed white LEDs on ternary substrates also demonstrated comparable results with those nanostructured white LEDs.^14,15

Note that the strain in InGaN/ InGaN QW on ternary InGaN substrate can be reduced by up to ∼75% compared to conventional InGaN/ GaN QW on GaN substrate as pointed out by previous work.²³ The significantly decreased strain resulted in substantial reduction of piezoelectric polarization fields and internal electrostatic fields in the QWs, and consequently led to suppression of the charge separation effect. Thus, the ternary In_0.15Ga_0.85N substrate has been employed in this study for all the proposed LEDs. The characteristics of the proposed InGaN/ InGaN MQWs white LEDs on In_0.15Ga_0.85N substrate are analyzed with APSYS.²⁹ The band structures and radiative recombination rates are solved self-consistently with k•p quantum mechanical solver by taking into consideration carrier transport effect, valence band mixing, strain effect, spontaneous and piezoelectric polarizations, and carrier screening effect. All band parameters used in the simulation were taken from Refs. 30, 31. The band offset ratio (ΔE_c : ΔE_v) for all layers was set as 0.7:0.3, and the surface charge density was set as 50% of the theoretical values.³² The operating temperature was assumed to be 300 K.

Two monolithic white LED structures with vertical injection configuration shown in Fig. 1 [LED (A) and LED (B)] are proposed and studied in this letter. The LED device without any nanostructure engineering [LED (C)] is used as a reference. LED (A) is composed of two blue-emitting QWs and two yellow-emitting QWs; while LED (B) and LED (C) are composed of three blue-emitting QWs and one yellow-emitting QW. The 0.4-μm thick n-In_0.15Ga_0.85N layer (n-doping = 5 × 10¹⁸ cm⁻³) is used as the substrate of the devices, followed by four periods of 3-nm In_xGa_1−xN QWs (x = 0.2 or 0.25 for blue emission and x = 0.38 for yellow emission) with 6-nm In_0.15Ga_0.85N quantum barriers (QBs), and a 200-nm p-In_0.15Ga_0.85N (p-doping = 1.2 × 10¹⁸ cm⁻³) layer. The background doping in the active region is assumed as n = 5 × 10¹⁶ cm⁻³. For both LED (A) and LED (B), a 1-nm thin GaN barrier has been inserted to surround the blue-emitting QW (in this case, the 6-nm In_0.15Ga_0.85N QB is replaced with 5-nm In_0.15Ga_0.85N QB and 1-nm GaN thin barrier) located closer to the p-InGaN layer for improved carrier confinement in the blue-emitting QWs.

As a result, Fig. 2 shows the simulated band structures and carrier distributions of LED (A), LED (B) and LED (C) at J = 100 A/cm². Fig. 2(a) plots the conduction bands and electron concentrations, and Fig. 2(b) plots the valence bands and hole concentrations for all devices. The corresponding bandgaps of GaN, In_0.15Ga_0.85N, In_0.2Ga_0.8N, In_0.25Ga_0.75N, and In_0.38Ga_0.62N are calculated as 3.437 eV, 2.839 eV, 2.654 eV, 2.475 eV and 2.044 eV respectively. As can be observed from Fig. 2, majority of the carriers are populating the yellow QW for LED (C) with peak electron and hole concentrations of 82 × 10¹⁸ cm⁻³ and 114 × 10¹⁸ cm⁻³ respectively while negligible amount of carriers are populating the blue QWs. Since the blue QW has smaller effective barrier height (h_c, h_v) compared to the yellow QW, the latter forms a strong local minimum in the MQWs active region, which is localizing carriers strongly and leading to poor carrier concentrations in the blue QWs. The insertion of the GaN thin barriers [the case of LED (A) and LED (B)] solves this issue by increasing h_c of electrons from 259 meV in LED (C) to 490 meV in both LED (A) and LED (B), and h_v of holes from 223 meV in LED (C) to 371 meV in both LED (A) and LED (B). As a result, the carrier population in the blue-emitting QWs has been enhanced significantly as shown in Fig. 2.

The peak electron and hole concentrations in the blue-emitting QW is 40 × 10¹⁸ cm⁻³ and 48 × 10¹⁸ cm⁻³ respectively for LED (A) and 32 × 10¹⁸ cm⁻³ and 41 × 10¹⁸ cm⁻³ respectively for LED (B), and the peak electron and hole concentrations in the yellow-emitting QW is 30 × 10¹⁸ cm⁻³ and 54 × 10¹⁸ cm⁻³ respectively for LED (A) and 27 × 10¹⁸ cm⁻³ and 45 × 10¹⁸ cm⁻³ respectively for LED (B). This balanced carrier concentration in the MQWs active region is expected to lead to balanced light output in both blue and yellow spectral regimes, which would enable proper color-mixing to achieve white illumination.

To demonstrate the optical properties of the proposed white LEDs, Fig. 3 plots the simulated spontaneous emission spectral of LEDs (A), (B) and (C) at J = 100 A/cm². The insertion of the GaN thin barrier effectively leads to dual-wavelength emissions in blue and yellow. Specifically, both LEDs (A) and (B) show the blue peak spontaneous emission wavelength (λ_peak) ∼ 480 nm due to the enhanced carrier populations in the blue QWs. The yellow peak for both LEDs is obtained with λ_peak ∼ 580 nm. Note that as a comparison, LED (C) shows a single green emission peak with λ_peak ∼ 560 nm and the full width half maximum (FWHM) ∼ 62 nm, which can be attributed from the recombination of excited states since considerably large amount of carriers are collected by the yellow QW. The spontaneous emission spectrum from LED (C) also proves that it is very important to have the GaN thin barrier insertion in order to achieve possible white color illumination. Furthermore, the design of the MQWs active region from LEDs (A) and (B) shows flexibility in terms of combining the blue and the yellow QWs, which would be ideal for large scale device manufacturing purpose.

The simulated light output power, I-V characteristic and relative EQE of the investigated monolithic white LEDs as a function of current density (J) are shown in Fig. 4. Here, the extraction efficiency, monomolecular coefficient A, and Auger coefficient C are employed as 70%, 10⁶ s⁻¹, and 10³⁴ cm⁶s⁻¹, respectively.³² In general, all three LEDs on ternary InGaN substrate exhibit promising light output power and EQE due to significantly reduced internal electrostatic field. Superficially, the light output power reaches ∼134 mW, ∼170 mW and ∼139 mW for LED (A), LED (B) and LED (C) respectively at J = 250 A/cm² as shown in Fig. 4(a). Note that the increased turn on voltage (∼ 2.7 V) for both LEDs (A) and (B) are due to increased series resistance as compared with LED (C) (∼ 2.1 V), which is attributed from the larger barrier height from the insertion of the GaN thin barrier. From the relative EQE plot in Fig. 4(b), it shows that both LEDs (A) and (B) on ternary substrate have the peak EQE (∼ 50%) with J ∼ 10 – 15 A/cm², which is significantly higher than that of InGaN/ GaN monolithic white LED on conventional GaN substrate (EQE ∼ 20%).⁷ Despite the fact that LED (C) shows a higher EQE peak (∼ 68%) at J ∼ 25 A/cm², it has a much severe droop towards higher current densities, which is attributed from poorer carrier confinement without any of the GaN thin barrier design. More importantly, only green color illumination can be obtained from LED (C) while white color illumination can be achieved by LEDs (A) and (B) with comparable output power and EQE.

Fig. 5 shows the locations of the light emission for the investigated monolithic white LEDs on ternary substrate on the Commission Internationale de l’Eclairage (CIE) 1931 color space chromaticity diagram at various injection current levels. As indicated on the CIE diagram, the coordinates for LED (A) and LED (B) are essentially in the white region at various current injection levels while the coordinate for LED (C) is located in the green region. This again implies that the insertion of thin GaN barrier surrounding the blue QW adjacent to the yellow QW is essential for white color generation. The CIE coordinates are relatively constant, approximately (0.30, 0.28), for LED (A) with current density up to 150 A/cm² whereas for LED (B), it moves from (0.30, 0.28) to (0.26, 0.22) as current increases from 50 A/cm² to 150 A/cm². Both LED (A) and LED (B) exhibit minor blue-shift with increased injection current due to quantum confined Stark effect.

For the investigated monolithic white LED structures in this work, the corresponding CCTs at 50 A/cm² are obtained as ∼ 8200 K for LED (A) and ∼ 8800 K for LED (B), which are appropriate for general illumination purpose.^1,5 The CCT values from LEDs (A) and (B) are also comparable with those of nanostructure white LEDs (∼ 4500 - 6500 K).^14,15 For comparison purpose, Fig. 5 also summarizes the CIE coordinates for InGaN/ GaN-based white LED on GaN substrate [(0.32, 0.41) at 50 A/cm²]⁹ and nanostructure white LEDs [(0.35, 0.37) and (0.29, 0.37) respectively for dot-in-wire LED¹⁴ and disk-in-wire LED¹⁵ at 50 A/cm²]. These results suggest that our proposed monolithic white InGaN/ InGaN MQW LED structures can achieve stable white color illumination and promising efficiency, while only require standard device fabrication processes.

In summary, high-efficiency phosphor-free monolithic white LEDs with InGaN/ InGaN QWs on ternary InGaN substrates are proposed and analyzed. Simulation studies show that by integrating blue- and yellow-emitting InGaN/ InGaN MQWs with engineered structures on ternary InGaN substrate, large output power (∼170 mW) and high EQE (∼ 50%) are achieved for stable white illumination at various current injections at room temperature. The results also demonstrate the importance of nanostructure engineering for InGaN/ InGaN MQWs LEDs on ternary InGaN substrate for white light emission. The chromaticity coordinates around (0.30, 0.28) and CCT ∼ 8200 K can be obtained at J = 50 A/cm² for the tunable white LED structures with engineered active region in this study. Thus, it is expected that the monolithic InGaN/ InGaN MQW white LED device based on InGaN ternary substrates would serve as a promising candidate for high-efficiency and cost-effective solid state lighting applications. Further experimental studies to investigate the optimized structure as well as the optimized growth is required for enabling the advantages presented in this study for InGaN/ InGaN MQWs white LEDs on ternary InGaN substrate.

This work is supported by the Kate Gleason endowed professorship fund from Rochester Institute of Technology.