Using band engineering to tailor the emission spectra of trichromatic semipolar InGaN light-emitting diodes for phosphor-free polarized white light emission

The III-nitrides material system has been researched extensively for optoelectronic applications including light emitting diodes (LEDs) and laser diodes (LDs). Commercially available InGaN LEDs and LDs are commonly grown on the polar c-plane of the wurtzite crystal structure. However, the performance of c-plane devices is affected by piezoelectric polarization-induced electric fields present in strained heterostructures.¹ Alternatively, devices can be grown on the nonpolar or semipolar planes to eliminate or reduce the piezoelectric polarization-induced electric fields,² and high performance nonpolar and semipolar LEDs and LDs have been demonstrated.^3–6 Additionally, while c-plane InGaN LEDs produce isotropic emission normal to the surface, unbalanced compressive in-plane strain in nonpolar and semipolar InGaN layers results in the emission of optically polarized light normal to the surface, which can be useful for certain device applications.^7–13 

By changing the alloy composition, the bandgap of InGaN can correspond to emission of any visible wavelength of light, and white light emission can be achieved by creating a III-nitride device with quantum wells (QWs) that emit different wavelengths of light. However, long wavelength emission from InGaN layers requires high indium contents, which are challenging to achieve in practice. High indium content InGaN layers have a large lattice mismatch with respect to GaN, and when high indium content InGaN layers grown pseudomorphically on GaN reach a critical thickness, stress is relieved by the formation of extended defects at which non-radiative recombination can occur¹⁴ or by a transition to 3D growth.¹⁵ The crystal quality of high indium content InGaN layers is also negatively impacted by the low growth temperatures that are necessary to increase indium incorporation. Low growth temperatures result in decreased adatom diffusion and desorption, which can lead to growth errors, breakdown of surface morphology, and higher impurity incorporation.^16,17 High indium content InGaN layers also have a low thermal budget, as subsequent high temperature steps have been shown to degrade high indium content InGaN layers.^18,19

We report a phosphor-free polarized white light-emitting semipolar device in which a blue LED is used to optically pump monolithically integrated InGaN QWs with longer wavelength emission. Optically pumping high indium content InGaN QWs for long wavelength emission can have several advantages over electrically injecting high indium content InGaN QWs. First, optically pumped QWs benefit from having lower carrier densities than QWs under typical electrical injection conditions. Lower carrier density results in decreased non-radiative Auger recombination, which is only an increasing challenge for devices with long wavelength emission.²⁰ Lower carrier density also results in decreased carrier screening effects. Second, optically pumped QWs benefit from not needing to transport carriers between QWs and not needing to confine QWs within the depletion width of a p-n junction.²¹ This allows for optically pumped QWs to employ wide or strain-compensating barriers between InGaN QWs to help prevent relaxation,^22,23 and there is the opportunity to incorporate an increasing number of optically pumped QWs to produce more long wavelength emission. Additionally, the doping profile can be engineered to tailor the emission wavelength of optically pumped QWs. For example, when a built-in electric field acts in the same direction as the polarization-induced electric field, the total electric field in the QW increases, decreasing the energy of the ground state transitions and red-shifting the emission. Because the QWs are not electrically injected and are therefore unbiased, both the built-in electric field and the total electric field in the QWs will remain largely unchanged during device operation. While carrier screening affects the total electric field in the QWs, the low carrier densities of optically pumped QWs limit the impact of carrier screening.

Band engineering can enable emission at longer wavelengths than can otherwise be achieved by simple manipulation of the alloy composition. Higher InGaN indium contents and/or lower growth temperatures negatively impact the InGaN crystal quality and radiative efficiency. Efforts to reach red emission by increasing the indium content of optically pumped $(202¯1)$ QWs in an n-i-n structure have not yet been successful. However, red emission can be achieved by using a p-i-n structure with the built-in electric field acting in the same direction as the polarization-induced electric field to red-shift emission. Additionally, band engineering can also be used to red-shift the emission wavelength of a QW so that a desired wavelength can be realized using a lower indium content InGaN QW or a thinner QW. The ability to utilize lower indium content InGaN is advantageous, because it enables growth at higher temperatures for material with higher crystal quality. For InGaN layers grown pseudomorphically on GaN, lower indium content InGaN films and thinner QWs also have less stress than higher indium content InGaN films or thicker QWs. Reduced stress is favorable to ensuring that strained layers do not relax.¹⁴ 

Monolithically incorporating electrically injected and optically pumped QWs enables long wavelength emission that can be used to create white light emission from a single III-nitride device. The other advantage of this device design is that unbalanced compressive in-plane strain in semipolar InGaN QWs increases the energy separation of the top two valence bands (VBs) and changes the state character of the VBs, which produces optically polarized emission.^7–13 Polarized white light has important applications in, for example, liquid crystal displays (LCDs), 3D displays, and holograms. LCDs are currently the largest application of polarized white light and typically are backlit using phosphor-converted LEDs, which emit unpolarized light that must be passed through a polarizer. Emission of polarized light offers an opportunity for improving the energy efficiency of devices that utilize polarized light because less light would be lost when passed through a polarizer.

Previous work has demonstrated white light emission by monolithically incorporated InGaN layers. However, for many previous devices, the resulting white light was not polarized because these devices were grown on c-plane. Researchers have demonstrated white light produced by electrical injection across vertically stacked InGaN layers,^24–27 as well as laterally distributed InGaN layers with different emission wavelengths.²⁸ Researchers have also demonstrated c-plane device designs that monolithically incorporate optically pumped and electrically injected QWs for white light. Unlike in our device design, the optically pumped QWs for long wavelength emission in these devices were grown prior to the blue LED, which resulted in the high indium content InGaN layers being exposed to high temperature during growth of the blue LED.^29,30 Fellows et al. reported polarized white light, which was achieved by combining emission from a yellow semipolar LED and a blue m-plane LED,³¹ and our group reported an initial demonstration of a monolithic optically pumped and electrically injected device with polarized white light emission.³² This previously published work combined a blue electrically injected LED and yellow optically pumped QWs, and it did not utilize band engineering to red-shift emission from optically pumped QWs.

In this work, we demonstrate a device with polarized white light emission that is achieved by monolithically incorporating electrically injected and optically pumped semipolar InGaN QWs. The doping profile in this device was intentionally engineered to red-shift the emission of one of the optically pumped QWs by creating a built-in electric field in the QW in the same direction as the polarization-induced electric field. A blue LED was first grown on the $(202¯1¯)$ face of a double-side-polished (DSP) bulk GaN substrate, because previous work has demonstrated that blue LEDs with high power, low electrical droop, and small wavelength shift can be grown on this plane.^3,4,33 High indium content InGaN QWs intended for optical pumping and long wavelength emission were subsequently grown on the $(202¯1)$ face. Growing the optically pumped QWs on the opposite side of the DSP substrate was advantageous, because it allowed for the high indium content InGaN QWs to be grown after the higher temperature growth of the blue LED, thus limiting the thermal budget to which the high indium content InGaN was subjected. Additionally, while $(202¯1¯)$ is a favorable growth plane for blue LEDs, stacking faults have been observed during the growth of high indium content InGaN layers on $(202¯1¯).$³⁴ However, prior work has demonstrated that $(202¯1)$ is an ideal plane for growing high indium content InGaN QWs with long wavelength emission.^35–37 In addition, the $(202¯1)$ plane and other planes with a similar sense of polarization can be used to grow a structure in which the doping profile can be engineered to red-shift the emission from optically pumped QWs.^2,33,38,39 To create a built-in electric field in the same direction as the polarization-induced electric field in an InGaN QW grown by metalorganic chemical vapor deposition (MOCVD), the total polarization discontinuity in the growth direction must be antiparallel to that of c-plane (i.e., negative) because MOCVD p-type GaN must be grown after the QW to enable activation of the acceptors by removing the hydrogen from the Mg–H complex. The $(202¯1)$ plane is an appropriate orientation to achieve this, because the total polarization discontinuity in $(202¯1)$ QWs is antiparallel to that of c-plane QWs.^2,33

The ${202¯1}$ III-nitride growth planes, which are inclined by 15° from the nonpolar m-plane, can be used to create LEDs with polarized light emission. In III-nitrides, the x-, y-, and z-directions are defined as the 〈$112¯0$〉 a-direction, 〈$11¯00$〉 m-direction, and the 〈$0001$〉 c-direction, respectively. For growth on $(202¯1)$⁠, we can define the x$′$-, y$′$-, and z$′$-directions where the z$′$-direction is normal to the $(202¯1)$ growth plane, the x$′$-direction is the $[12¯10]$ a-direction, and the y$′$-direction is the $[101¯4¯]$ direction, which is the projection of the –c-direction onto the $(202¯1)$ plane. For the case of InGaN growth on inclined m-planes, the unbalanced in-plane strain results in the top VB being primarily composed of $|X$〉 state.⁷ Thus, for growth on $(202¯1)$⁠, the polarized light emission intensity is at a maximum when the polarizer is aligned along the x$′$-direction, and the polarized light emission intensity is at a minimum when the polarizer is aligned along the y$′$-direction. The optical polarization ratio (ρ) is calculated according to

where $Ix′$ and $Iy′$ are the integrated electroluminescence (EL) intensities with the polarizer aligned along the $[12¯10]$ and $[101¯4¯]$ directions of the sample, respectively.

Samples were homoepitaxially grown by atmospheric pressure metalorganic chemical vapor deposition (MOCVD) on a 7.5 mm $×$ 7.5 mm free-standing, DSP $(202¯1¯)$/$(202¯1)$ substrate supplied by Mitsubishi Chemical Corporation. First, an LED was grown on the $(202¯1¯)$ face. The device structure consisted of a 3 μm Si-doped n-type GaN layer, a 30 nm unintentionally doped (UID) GaN layer, a 12 nm UID InGaN single QW active region, a 30 nm UID GaN layer, a Mg-doped AlGaN electron blocking layer (EBL), a 150 nm Mg-doped p-type GaN layer, and a Mg-doped p⁺⁺-type GaN contact layer. Subsequently, high indium content InGaN QWs were grown on the $(202¯1)$ face. The structure consisted of a 450 nm Si-doped n-type GaN layer, a 60 nm n-type GaN layer with [Si] = 7.5 × 10¹⁸cm⁻³, a 35 nm UID GaN barrier, a 6 nm UID InGaN QW, a 35 nm UID GaN barrier, a 110 nm n-type GaN layer with [Si] = 7.5 × 10¹⁸cm⁻³, a 35 nm UID GaN barrier, a 6 nm UID InGaN QW, a 35 nm UID GaN barrier, and a 60 nm p-type GaN layer with [Mg] = 1.0× 10¹⁹cm⁻³. The QW situated in a p-i-n structure on $(202¯1)$ was grown at a growth temperature 10 °C colder than the QW situated in an n-i-n structure. A Pd/Ag/Ni/Au (3/2000/2000/3000 Å) circular contact with a 180 μm radius was deposited by electron beam evaporation on the Mg-doped p⁺⁺-type contact layer of the $(202¯1¯)$ LED to form the p-contact. The n-contact was formed by scribing the $(202¯1¯)$ face and soldering indium to the exposed n-type layer of the LED. A schematic of the device structure is shown in Fig. 1.

Figure 2 shows simulated energy band diagrams for the optically pumped $(202¯1)$ QWs at a 0 V bias. The simulation was performed with commercial package SiLENSe version 5.8.⁴⁰ In Fig. 2(a), the QW in the n-i-n structure consisted of 6 nm of In_0.27Ga_0.73N, and the QW in the p-i-n structure consisted of 6 nm of In_0.29Ga_0.71N. The change in the indium content for the different QWs corresponds to the different growth temperatures used in growing the optically pumped QWs in the experimental sample. For all of the modeling work, the QWs were assumed to consist of 6 nm of InGaN, the n-type GaN layers were doped with [Si] = 7.5× 10¹⁸cm⁻³, the p-type GaN layer was doped with [Mg]= 1.0 × 10¹⁹cm⁻³, and the UID GaN layers were assumed to have 1.0 × 10¹⁷cm⁻³ donors. Table I summarizes the material properties used in the modeling.^41,42 Figures 2(b) and 2(c) consider In_0.27Ga_0.73N and In_0.29Ga_0.71N QWs in either the n-i-n or p-i-n structures, respectively. The electron–heavy hole ground state transitions are indicated on these band diagrams, and Table II shows the calculated energies for the electron–heavy hole, electron–light hole, and electron–split-off hole ground state transitions for the In_0.27Ga_0.73N and In_0.29Ga_0.71N QWs in either an n-i-n or p-i-n structure.

As can be seen in Fig. 2(a), the p-i-n built-in electric field acts in the same direction as the polarization-induced electric field on $(202¯1)$⁠, resulting in an increased electric field in the QW situated in a p-i-n structure compared to the QW in an n-i-n structure. This increased electric field is expected to red-shift the emission wavelength, and the simulations calculate that a red-shift of 76 nm is expected when comparing emission from the In_0.29Ga_0.71N QW in the p-i-n structure to emission from the In_0.27Ga_0.73N QW in the n-i-n structure. In addition, these simulations allow us to separately consider the impact that changing the alloy content had on emission wavelength and the impact that band engineering had on emission wavelength. Both increased InGaN indium content and increased total electric field in the QW contribute to red-shifting emission. The modeling work was able to quantify that, in this case, band engineering accounted for a red-shift of 57 nm and the change in alloy content accounted for a red-shift of 19 nm. This illustrates that band engineering can have a significant impact on the emission wavelength of optically pumped QWs.

Figure 3(a) shows the normalized white light EL spectrum of the device from Fig. 1. Light emitted from the $(202¯1¯)$ face was collected using a 0.45 numerical aperture 20$×$ microscope objective. The experimental setup is detailed in Ref. 43. For the data presented in Fig. 3(a), there was no polarizer in the optical path, and the device was operated at 10 A/cm² under DC bias at room temperature. Similar to the device reported in Ref. 32, the emission color was spatially non-uniform, making it possible to tune the color by changing the measurement location. To obtain white light emission as shown in Fig. 3(a), the measurements were made 680 μm away from the center of the circular p-contact. Moving away from the center of the electrically injected device increased the relative intensity of long wavelength emission from the optically pumped QWs compared to blue emission from the electrically injected QW because blue light that did not escape on the first pass was reflected internally and could excite the optically pumped QWs away from the electrically injected area.

In Fig. 3(a), the relatively narrow peak at 450 nm is emission from the electrically injected $(202¯1¯)$ LED. The peak at 520 nm is emission from the optically pumped $(202¯1)$ QW with a slightly InGaN lower indium content in an n-i-n structure, and the peak at 590 nm is emission from the optically pumped $(202¯1)$ QW with a slightly higher InGaN indium content in a p-i-n structure, corresponding to a 70 nm shift in wavelength relative to the QW in an n-i-n structure. As shown in Fig. 2 and Table II, this red-shifted emission from the QW in the p-i-n structure compared to the QW in the n-i-n structure was due to increased indium content InGaN as well as the effect of the built-in electric field. The electron—heavy hole ground state transitions indicated in Fig. 2 and Table II are of slightly lower energy than the emission peaks because the emission peaks are also composed of higher energy transitions. The experimentally observed 70 nm red-shift in the peak emission wavelength of the higher indium content InGaN QW in the p-i-n structure compared with the lower indium content InGaN QW in the n-i-n structure agrees well with the simulation results, which calculated a 19 nm red-shift attributed to the increase in the indium content of the InGaN QW and a 57 nm red-shift attributed to the band engineering.

We previously demonstrated a similar white light emitting device by combining emission from monolithic electrically injected and optically pumped QWs.³² The device combined blue emission from an electrically injected $(202¯1¯)$ QW and yellow emission from three optically pumped $(202¯1)$ QWs in an n-i-n structure. The EL spectrum from this device is shown in Fig. 3(b) and can be compared to the emission spectrum in Fig. 3(a). To facilitate the comparison of the emission spectra from the optically pumped QWs in the two cases, each spectrum was normalized to the corresponding maximum of the blue emission peak. The device measured in Fig. 3(a) will be referred to as the red/green/blue (RGB) device, and the device measured in Fig. 3(b) will be referred to as the blue/yellow (BY) device.

The spectrum depicted in Fig. 3(a) from the RGB device has several advantages over the spectrum in Fig. 3(b) from the BY device for both general illumination and display technology. By producing light that is more widely distributed across the visible spectrum, the RGB device is able to render colors more faithfully than the BY device, which is important for general illumination. The RGB device would also be more energy efficient than the BY device for backlighting displays. Creating red, green, and blue emission peaks would align the white light spectrum with the transmission spectra of display filters that are typically designed to transmit red, green, and blue light, resulting in less light being lost to absorption in the filters.

Figure 4 is a 1931 Commission Internationale de l'Eclairage (CIE) x, y chromaticity diagram, which indicates the location of the spectrum shown in Fig. 3(a). The CIE x- and y-chromaticity coordinates are (0.33, 0.35). This point lies close to the Planckian locus and corresponds to white light emission with a correlated color temperature (CCT) of 5604 K and a color rendering index (CRI) of 70.

In addition to demonstrating white light emission, this device also exhibited polarized white light emission from the electrically injected and optically pumped semipolar QWs. Figure 5 shows EL emission spectra with the polarizer aligned along $[12¯10]$ (x$′$-direction) and with the polarizer aligned along $[101¯4¯]$(y $′$-direction). The optical polarization ratio, calculated using integrated intensities according to Eq. (1), is 0.30. Previous experimental results have demonstrated significant differences in growth and performance of devices on $(202¯1)$ and $(202¯1¯),$^9,33 including significantly higher optical polarization ratios for $(202¯1¯)$ QWs compared to $(202¯1)$ QWs.^8–11 For the spectra shown in Fig. 5, the optical polarization ratio for the $(202¯1¯)$ QW is similar to the optical polarization of the $(202¯1)$ QWs. However, measurements at different locations revealed that the optical polarization ratio of the blue emission decreased as the measurement location was moved farther from the center of the p-contact. These results indicate that the blue emission was composed of an increasing proportion of scattered light with increasing distance from the electrically injected area. This occurred because the intensity of the blue light that was both optically polarized and had an angle of incidence within the escape cone decreased with increasing distance from the p-contact and an increasing number of reflections.

Based on the observed dependence of the optical polarization ratio on measurement position, it is expected that efforts to improve color uniformity and extraction efficiency should also improve the optical polarization ratio of the emission. Future work to improve color and achieve a device with uniform white light emission must focus on increasing the intensity of long wavelength emission relative to the intensity of blue emission. There are several approaches to improve color uniformity. First, an important area of ongoing work is developing growth conditions to achieve a larger number of optically pumped QWs. InGaN QW solar cell research has shown that tens of QWs are needed to achieve significant absorption.^44,45 Increasing the number of optically pumped QWs will absorb more blue light and emit more light at long wavelengths. Second, continued optimization of growth conditions and epitaxial design can increase the radiative efficiency and absorbance of the optically pumped QWs. Improved material quality and larger overlap of the electron and hole wavefunctions will result in better performance of the optically pumped QWs. In particular, because band engineering can enable the use of lower indium content InGaN or thinner QWs in achieving a desired emission wavelength, ongoing work is investigating the larger design space that is enabled by band engineering.

Another approach that will improve color uniformity is depositing a dichroic mirror with high reflectance of blue light and high transmittance of longer wavelengths. Dichroic mirrors will reflect blue light for additional passes through the optically pumped QWs, which will increase the probability that blue light is absorbed and increase the long wavelength emission. Another extraction engineering technique that can be applied to future devices is incorporating a photonic crystal. This device is polished on both sides, and while this preserves optical polarization, it results in low extraction efficiency. The emission intensity of devices can be increased by incorporating a photonic crystal to increase extraction while preserving optical polarization, as has been demonstrated on m-plane LEDs.⁴⁶ 

Another consideration is that carrier densities of the optically pumped QWs in future devices will be higher than the carrier densities in this initial device demonstration, where measurements were made far from the electrically injected device and at a relatively low current density of 10 A/cm². However, the carrier density of an optically pumped QW is always expected to be low relative to the carrier density in electrically injected QWs because the carrier density in an optically pumped QW is limited by absorbance of an InGaN QW. Therefore, though higher carrier densities will result in increased carrier screening, carrier screening is not expected to have a large impact on emission wavelength, and future devices can slightly adjust band engineering or alloy content to compensate for the blue-shift produced by carrier screening.

In summary, we have grown and fabricated a polarized white light-emitting device that monolithically integrates a blue LED grown on the $(202¯1¯)$ face of a bulk GaN substrate with optically pumped high indium content InGaN QWs for long wavelength emission grown on the $(202¯1)$ face. The doping profile was intentionally engineered to red-shift the emission of one of the optically pumped QWs by creating a built-in electric field in the QW in the same direction as the polarization-induced electric field. Emission peaks were observed at 450 nm from the electrically injected QW and at 520 nm and 590 nm from the In_0.27Ga_0.73N and In_0.29Ga_0.71N optically pumped QWs, which were grown the same thickness in n-i-n and p-i-n structures, respectively. Emission with CIE x- and y-chromaticity coordinates of (0.33, 0.35), a CCT of 5604 K, a CRI of 70, and an optical polarization ratio of 0.30 was reported. This monolithic electrically injected and optically pumped device produced polarized white light with an emission spectrum that would be desirable for either general illumination or display technology.

This work was supported by the Solid State Lighting and Energy Electronics Center (SSLEEC) and the KACST-KAUST-UCSB Solid State Lighting Program (SSLP). A portion of this work was done in the UCSB nanofabrication facility, part of the NSF funded National Nanotechnology Infrastructure Network (ECS-03357650). This work made use of the MRL Central Facilities at UCSB supported by the MRSEC Program of the NSF under Award No. DMR 1121053. C. D. Pynn was supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE 1144085. The authors would like to thank the contributors to the open source Python color science package Colour, which is freely distributed under the New BSD License and was used to generate the CIE diagram as well as calculate CIE coordinates, CCT, and CRI. The authors also thank Dr. Tom Mates for performing SIMS characterization and Karthik Krishnaswamy for helpful discussion.