1. Introduction

The III-nitrides are direct bandgap semiconductors that span the visible wavelengths and have been studied extensively for optoelectronic applications. While commercially available InGaN light-emitting diodes (LEDs) and laser diodes (LDs) are commonly grown on the polar c-plane of the wurtzite crystal structure, it can be advantageous to grow devices on nonpolar and semipolar planes because these nonpolar or semipolar growth planes eliminate or reduce the piezoelectric polarization-induced electric fields in strained heterostructures [1]. In addition, for growth on nonpolar or semipolar planes, compressively strained InGaN quantum wells (QWs) produce optically polarized emission directed normal to the surface [2].

The largest application of polarized light today is in liquid crystal displays (LCDs), which are backlit by white light. In order to create a device with polarized white light emission, we have proposed semipolar device designs that monolithically incorporate electrically injected QWs with blue emission and optically pumped QWs with longer wavelength emission. Optically pumping QWs for long wavelength emission offers several benefits over electrically injecting QWs for long wavelength emission [3,4]. First, devices can incorporate many optically pumped QWs to increase emission intensity. Second, optically pumped QWs can incorporate wide or strain-compensating barriers to prevent relaxation of high indium content InGaN layers. Third, optically pumped QWs have lower carrier densities, which decrease Auger recombination [5,6]. Fourth, band engineering can be used to optimize the emission wavelength of optically pumped QWs [4].

A first report of a monolithic optically pumped and electrically injected device was published by Damilano et al [7]. In this device, the optically pumped, high indium content InGaN QWs for long wavelength emission were grown prior to the electrically injected blue LED, which is not desirable because the LED device layers require higher growth temperatures that may damage high indium content InGaN QWs [8,9]. Additionally, this device was not designed for optically polarized emission, as it was grown on c-plane GaN.

We have previously demonstrated double-sided, optically pumped and electrically injected semipolar device designs for polarized white light emission [3,4]. The devices were grown on double-side-polished (DSP) semipolar substrates. Growth on a semipolar plane enabled optically polarized emission [2]. And use of a DSP substrate allowed for monolithically integrated QWs where a blue LED was grown first and optically pumped QWs for long wavelength emission were grown subsequently on the opposite side of the DSP substrate. Thus, the higher growth temperature steps for the blue LED were performed prior to the growth of the optically pumped QWs to prevent thermal damage to the high indium content InGaN layers [8,9]. However, such a double-sided device design is more complicated as it requires growth on both sides of a substrate and fabrication of a flip-chip device. Additionally, in a double-sided device, the optically pumped QWs are far away from the electrically injected QWs, and a significant portion of light can escape from the sidewalls of a device without passing through the optically pumped QWs.

Alternatively, an improved device design would place the optically pumped QWs for long wavelength emission above the electrically injected LED. Incorporating a tunnel junction (TJ) directly above the p-type layer of the electrically injected LED provides a means of carrier conversion between p- and n-type materials, and when contact is made to the n-side of the LED and the n-side of the TJ, it is possible to electrically inject the buried LED. Figure 1 shows a cross-sectional schematic of the device described in this paper.

 

Fig. 1 Cross-sectional schematic of a $\left(20\overline{2}1\right)$device that used a TJ to grow optically pumped QWs on top of electrically injected QWs. The TJ was grown by MBE and the remaining layers were grown by MOCVD.

Download Full Size | PDF

In this device, the n-type GaN above the TJ is used for current spreading. Current spreading is especially a challenge in devices with buried p-type GaN because the conductivity of p-type GaN is too low to enable current spreading across device areas. LED devices typically employ transparent conducting oxide (TCO) or reflective metal p-contacts to enable current spreading on the p-side of devices, but such contacts not applicable for buried devices. However, in a TJ device, n-type GaN is used to create a TJ contact to the underlying p-type GaN of the buried device. In such a device, the n-type GaN above the TJ also serves as the current spreading layer, and in fact, n-type GaN is a superior current spreading layer. In particular the lower absorptivity of visible wavelengths for GaN compared to TCOs or metals enables a lower absorbance for a GaN current spreading layer [10,11].

Using n-GaN to create a TJ contact to the underlying p-type GaN of the buried LED is also advantageous because it eliminates the need to etch down to expose the buried p-type layer, as would be required to create metal or TCO contacts to the buried LED device. It is especially beneficial to eliminate the metal or TCO p-contact in a buried device because dry etching techniques, which are required to expose the buried p-type layer, are reported to damage p-type GaN layers. Exposing the surface to energetic ions in a plasma creates nitrogen vacancies, which act as donors and compensate the Mg acceptors [12–14].

Devices with a TJ are advantageous because they can achieve current spreading and ohmic contacts in buried optoelectronic devices. However, creating III-nitride optoelectronic TJ devices is challenging because metalorganic chemical vapor deposition (MOCVD) p-type III-nitride layers must be exposed at the sample surface in order to be activate Mg acceptors during a post-growth anneal that removes hydrogen from the Mg−H complex, which forms during growth and passivates the Mg acceptors [15]. While molecular beam epitaxy (MBE) enables the growth of activated p-type GaN, this growth technique produces InGaN with poor radiative efficiency [16,17]. Therefore, to realize an electrically injected and optically pumped TJ device like the one shown in Fig. 1, our experimental work has focused on a hybrid MOCVD/MBE growth procedure. The use of MBE enabled the growth of activated buried p-type GaN layers and a TJ, while use of MOCVD enabled InGaN with high radiative efficiency. Recent work published by Malinverni et al. first demonstrated a high quality TJ grown by MBE above an InGaN active region grown by MOCVD [18]. In the work by Malinverni et al., a TJ was created by using MBE to grow p-type GaN above MOCVD-grown InGaN QWs and n-type GaN above the p-type layers. Our work builds upon the hybrid MOCVD/MBE TJ growth process detailed in [19], which was also applied in the optoelectronic devices published by Yonkee et al. and Leonard et al. [20,21] In this growth procedure demonstrated by Young et al., a TJ was created by using MBE to grow n-type GaN layers above the p-type GaN of an MOCVD-grown LED structure [19].

In our hybrid MOCVD/MBE growth procedure, NH₃-assisted MBE is used for the growth of the n⁺-type layer of the TJ and enables a device with activated buried p-type GaN [18–21]. Although the optically pumped QWs are subsequently grown by MOCVD, hydrogen has low diffusivity in n-type GaN, which prevents hydrogen from diffusing through the MBE n-type GaN layers to passivate Mg acceptors in the p-type layers of the LED [22]. Use of MBE also is favorable for growth of the TJ because MBE enables a sharp p⁺/n⁺ interface, a high Si-doping concentration, and GaN with high n-type mobility [23,24]. In contrast, the Mg memory effect in MOCVD produces slow turn-off of Mg-doping, and growing GaN with high Si-doping concentrations by MOCVD can result in rough sample surfaces [25]. Lastly, the hybrid MOCVD/MBE growth procedure is advantageous because work published by Young et al. suggests that $\left(20\overline{2}1\right)$ TJs grown by this technique benefit from residual surface oxygen at the regrowth interface, which may decrease the tunneling distance because oxygen acts as an electron donor [19]. Thus, as described below and in [19], the sample was removed from the MOCVD and exposed to ambient prior to being loaded into the MBE for regrowth of the n⁺-type layer of the TJ.

We chose to grow our device on the $\left(20\overline{2}1\right)$ plane because it is an appropriate orientation for achieving optically polarized emission, which was one goal of this work [2]. In addition, prior work has demonstrated that $\left(20\overline{2}1\right)$ is an ideal plane for growing high indium content InGaN QWs with long wavelength emission [26–28], while stacking faults have been observed during the growth of high indium content layers on $\left(20\overline{2}1\right)$ or m-plane [29,30].

2. Experiment

Samples were homoepitaxially grown on a 7.5 mm × 7.5 mm free-standing $\left(20\overline{2}1\right)$ bulk GaN substrate supplied by Mitsubishi Chemical Corporation. First, an LED was grown by MOCVD. The device structure consisted of a 1.4 μm Si-doped n-type GaN layer, a 6 nm unintentionally doped (UID) GaN layer, a 3 period 3 nm/4 nm UID InGaN/GaN nm multiple quantum well (MQW) active region, a 10 nm UID GaN layer, a 12 nm Mg-doped AlGaN electron blocking layer (EBL), a 145 nm Mg-doped p-type GaN layer, and a 13 nm Mg-doped p⁺⁺-type contact layer. The sample was removed from the MOCVD, exposed to ambient, and the p-type GaN was activated by annealing at 600 °C in atmosphere. Then MBE was used to grow a 20 nm n⁺-type layer above the p⁺-GaN layer to form a TJ, followed by a 100 nm n-type GaN capping layer was also grown by MBE. Further details on TJ growth by our hybrid MCOVD/MBE growth technique are in [19]. The sample was removed from the MBE and exposed to ambient before MOCVD was used to grow the optically pumped MQW active region. The structure consisted of a 440 nm Si-doped n-type GaN layer, a 35 nm UID GaN barrier, a 2.5 period 6 nm/35 nm UID InGaN/GaN MQW stack, a 35 nm UID GaN barrier, and a 23 nm n-type GaN layer. The growth of the optically pumped MQWs was performed at a relatively low temperature after the higher temperature growth of the first 200 nm of n-type GaN.

Devices were fabricated by etching a 1.8 μm high mesa to define a 0.1 mm² circular injection area and to expose the n-type GaN below the blue LED. The etched area was a ring such that the optically pumped QWs were not etched in the field between the electrically injected devices. A separate 300 nm deep etch exposed the n-type GaN above the TJ. The etched area was a circle, which had a radius of 32 μm and which was positioned at the center of the 0.1 mm² mesa. A Ti/Al/Ni/Au (20/100/200/300 nm) metal stack was deposited by electron-beam evaporation to make contact to the n-side of the LED and the n-side of the TJ. The metal contact to the n-side of the LED was a ring surrounding the mesa. The metal contact to the n-side of the TJ was a circle with a radius of 25 μm. The n-contact metal stack was annealed at 450 °C in nitrogen. The backside of the device was polished to an optically smooth surface to prevent scattering and preserve the optical polarization of light emitted by the strained semipolar InGaN QWs. A schematic of this device structure is shown in Fig. 1.

3. Results and discussion

Figure 2 shows the current−voltage characteristics for the device shown in Fig. 1. The device was operated under DC bias at room temperature. Though measurements of the contacts made to the n-side of the LED indicate that the contacts were ohmic, the turn-on voltage was higher than the turn-on voltages reported by Young et al. for LED and p-n diode devices with TJ contacts that were grown by the hybrid MOCVD/MBE growth technique [19]. We expect that improved electrical characteristics for optically pumped and electrically injected TJ devices can be achieved by optimizing the TJ and n-contacts.

 

Fig. 2 Current−voltage characteristic of the device in Fig. 1.

Download Full Size | PDF

Figure 3 shows the electroluminescence (EL) emission spectrum of the device shown in Fig. 1. The spectral measurements were made at a current density of 60 A/cm². Light emitted from the $\left(20\overline{2}1\right)$ face was collected using a 0.45 numerical aperture 20 × microscope objective. The experimental setup is detailed in [31]. The relatively narrow peak at 450 nm is emission from the electrically injected_{LED, and the peak at 560 nm is emission from the optically pumped QWs. For the data presented in Fig. 3, no polarizer was in the optical path. Similar to the device reported in [3], the emission color was spatially non-uniform. Blue light that has not been extracted or absorbed can excite the optically pumped QWs away from the electrically injected area. Thus, as the measurement location was moved farther from the center of the electrically injected device, the relative intensity of yellow emission from the optically pumped QWs increased compared to blue emission from the LED. For the data presented in Figs. 3 and 4, the measurements were made 2.8 mm away from the center of the circular LED geometry described above.}

 

Fig. 3 EL emission spectrum for the device shown in Fig. 1 without a polarizer in the optical path. The peak at 450 nm is from the electrically injected LED and a peak at 560 nm is from the optically pumped QWs.

Download Full Size | PDF

 

Fig. 4 EL emission spectra with the polarizer aligned along $\left[1\overline{2}10\right]$ ($x^{\prime }$-direction) and with the polarizer aligned along $\left[10\overline{1}\overline{4}\right]$ ($y^{\prime }$-direction).

Download Full Size | PDF

Figure 4 shows the EL emission spectra for the $\left(20\overline{2}1\right)$ device with the polarizer aligned along $\left[1\overline{2}10\right]$ ($x^{\prime }$-direction) and with the polarizer aligned along $\left[10\overline{1}\overline{4}\right]$ ($y^{\prime }$-direction), corresponding to the maximum and minimum intensities that can pass through the polarizer, respectively [2,4]. The optical polarization ratio is calculated according to $\rho =\left(I_{x\text{'}}-I_{y\text{'}}\right)/\left(I_{x\text{'}}+I_{y\text{'}}\right)$, where $I_{x\text{'}}$ and $I_{y\text{'}}$ are the integrated intensities of the emission spectra when the polarizer is aligned along the $x^{\prime }$-direction and $y^{\prime }$-direction, respectively. For the data presented in Fig. 4, emission from both the optically pumped and electrically injected QWs was polarized with an overall polarization ratio of 0.28. In terms of the individual MQW active regions, the blue emission from the electrically injected QWs had an optical polarization ration of 0.20, and the yellow emission from the optically pumped QWs had an optical polarization ratio of 0.36.

Figure 5 summarizes optical polarization ratios reported in the literature as a function of the peak emission wavelength of the dominant component. The open symbols correspond to polarization ratios reported in the literature for electrically injected $\left(20\overline{2}1\right)$ InGaN QWs [32–34]. As determined from the measured data in Fig. 4, the solid red diamond points in Fig. 5 indicate the polarization ratios of the optically pumped yellow QWs and the electrically injected blue QWs of the $\left(20\overline{2}1\right)$ TJ device. Figure 5 shows that the polarization ratios of the blue and yellow peaks in Fig. 4 agree well with other reported values. The polarization ratio of the yellow QWs is greater than the polarization of blue QWs because increasing the InGaN indium content increases strain, which results in separation of the valence bands.

 

Fig. 5 Optical polarization ratios reported in the literature as a function of the peak emission wavelength for electrically injected $\left(20\overline{2}1\right)$ InGaN QWs. The red diamond points indicate the polarization ratios of the optically pumped yellow QWs and electrically injected blue QWs from the $\left(20\overline{2}1\right)$ TJ device measured in Fig. 4.

Download Full Size | PDF

Overall, the emission peaks from this device are similar to the emission peaks from commercial phosphor converted LEDs for white light, which combine a blue LED pumping a yellow phosphor. The emission spectrum for a phosphor converted LED consists of a narrow blue peak, which is emission from the blue LED, and a broad yellow peak, which is emission from the phosphor. However, the emission spectrum in Fig. 3 does not correspond to white light because the ratio of blue to yellow light is too high. As shown in the 1931 CIE x, y chromaticity diagram in Fig. 6, the emission spectrum in Fig. 3 corresponds to a point at (0.23, 0.21) and the lies at the end of the Planckian locus. Increasing the ratio of yellow to blue emission in future devices should enable spatially uniform phosphor-free polarized white light. Moreover, increasing the absorbance, radiative efficiency, and/or number of optically pumped QWs will increase the ratio of yellow to blue emission. Incorporating a dichroic coating with high blue reflectance and high yellow transmittance also will allow for control over the ratio of yellow to blue emission. Extraction engineering may also improve the efficiency of future devices. For example, a photonic crystal can increase extraction while preserving optical polarization [35].

 

Fig. 6 CIE x, y chromaticity diagram indicating the chromaticity coordinates corresponding to the spectrum in Fig. 3.

Download Full Size | PDF

4. Conclusion

In summary, we have grown and fabricated a $\left(20\overline{2}1\right)$ III-nitride device in which a TJ is used to monolithically incorporate optically pumped QWs for long wavelength emission and an electrically injected blue LED, which is used as the excitation source for the optically pumped QWs. The use of a TJ allows the optically pumped, high indium content InGaN QWs to be grown after the blue LED and avoid exposure to high growth temperatures, which have been shown to degrade high indium content InGaN. Use of NH₃ MBE enabled the growth of the low resistance TJ in this device, while use of MOCVD enabled the growth of InGaN with high radiative efficiency. Our initial device produced emission peaks at 450 nm from the electrically injected QWs and at 560 nm from optically pumped QWs. Optically polarized emission was measured, with a polarization ratio of 0.28. By increasing the ratio of yellow to blue emission, future devices can be used to produce phosphor-free polarized white light, which has important applications and other technologies.