1. Introduction

Colloidal compound quantum dots (QDs) have recently been introduced to white LED technology as a new family of phosphor materials with many superior properties [1–4]. Due to strong quantum confinement, semiconductor QDs, such as core/shell CdSe/(Zn,Cd)S QDs, are characterized by sharp exciton absorption features, high luminescence efficiency, and size-tunable emission color spanning the entire visible spectrum [5]. QDs of the same chemical composition and different size can therefore be employed to provide multiple spectral components in the white LED output, with improved color qualities and aging performance [6]. Perhaps the most significant potential of QD phosphors lies in the recent discovery that there exists a path for indirect injection of electron-hole pairs into QDs (for radiative recombination and thus band-edge emission from QDs) by non-contact, nonradiative energy transfer from a proximal InGaN quantum well (QW) [7]. This indirect, nonradiative energy transfer path is considered to be the consequence of dipole-dipole interactions associated with QW-QD coupling and the extremely fast intraband relaxation in colloidal QDs (subpicosecond time scales). It is fundamentally different from the multi-step ‘down conversion’ scheme and removes several of the intermediate steps involved in color conversion, thereby eliminating energy losses associated with these steps and increasing the efficiency [8–10].

While the QD approach holds great potential for solid state lighting, the fabrication of efficient QD phosphor-based LEDs cannot be accomplished with convential nitride LED configurations. In designing QD-QW coupled white LEDs, it is essential to keep the QDs in the immediate vicinity of the QW surface as the non-radiative energy transfer rate, $\gamma ={\tau }_{ET}{}^{-1}$, scales as d⁻⁴, where d is the QD-QW distance [2]. This requirement is in conflict with the fact that a doped GaN contact layer needs to cap the InGaN QW to complete the p-i-n LED structure. Recently, Achermann et al. observed color conversion in an electrically pumped light-emitting diode (LED) using nonradiative energy transfer between an InGaN/GaN QW and a monolayer of CdSe/ZnS QDs [8]. Spectroscopic measurements revealed that 13% of the radiative power of the QW was transferred to red emission from the QDs when the QW and QDs were separated by a 5nm-thick n-type capping layer for close proximity. The ultra thin n-type capping layer, however, was not sufficiently conductive for current spreading, which, in turn, reduced the LED efficiency. In another study [11], GaN-based LEDs were surface-patterned with deep-etched elliptical holes to accommodate colloidal QDs (acceptors) such that the nanocrystals were within close vicinity of the active layers (donors). Efficient nonradiative energy transfer was achieved via a short donor-acceptor distance, leading to a 43% increase in QD phosphor-emission over that provided by the conventional color down-conversion mechanism.

We present in this communication an alternative QD-QW coupling approach to address the conflict between the need for close proximity of QWs and QDs and the use of a sufficiently thick contact layer with a low resistance for LED operation. The p-i-n InGaN/GaN MQW heterostructure was patterned and dry-etched to form dense arrays of nanopillars using a novel etch mask consisting of self-assembled In₃Sn clusters. Colloidal QD phosphors have been deposited into the gaps between the InGaN/GaN nanopillar LEDs, leading to sidewall coupling between the QDs and InGaN MQW emitters. The immediate QW-QD contact and low-resistance design of the LED contact layer can therefore be achieved simultaneously. Strong non-radiative energy transfer was observed from the InGaN MQW to the colloidal QD phosphors, which led to a 263% enhancement in effective internal quantum efficiency for the QDs incorporated in nanopillar LEDs in comparison with those deposited over planar LED structures. Time-resolved photoluminescence was used to characterize the energy transfer process between the QWs and QDs. The measured rate of non-radiative QD-QW energy-transfer agrees well with the value calculated from the quantum efficiency data for the QDs deposited in the nanopillar LEDs.

2. Device structure and fabrication

Figure 1 shows schematically a nanopillar LED structure infiltrated with QD phosphors. Each nanopillar is configured with an InGaN/GaN MQW-embedded p-i-n heterojunction, comprising a sapphire substrate, an n⁺ (N_d ≈5 × 10¹⁸cm⁻³) GaN buffer layer (~2µm), an n-type (N_d ≈1 × 10¹⁸cm⁻³) GaN layer (~0.1 µm), three pairs of 10nm-thick In_0.1Ga_0.9N quantum wells sandwiched between 50nm-thick GaN barriers, a p-type (N_a ≈1 × 10¹⁸cm⁻³) GaN layer (~0.2 µm), as well as a p-type (N_a ≈1 × 10¹⁹cm⁻³) GaN capping layer (~0.2 µm). Silicon and magnesium are the n- and p-type dopants in the LED heterojunction, respectively. The nanopillars are ~200nm in diameter, sharing the common base of the n-type GaN buffer layer, and are fabricated with a Ti/Al n-type bottom contact and a slanted indium-tin-oxide (ITO) nanowire structure serving as the p-type top contact [12]. Finally, the sidewall surface of the nanopillars is conformally coated with QD phosphors.

 

Fig. 1 (a) Schematic of the nanopillar LED for the MQW-QD coupling configuration. (b) Cross-sectional SEM image of a nanopillar LED device.

Download Full Size | PDF

The InGaN/GaN MQW nanopillar LEDs were fabricated by inductively-coupled-plasma (ICP) etching using a novel etch mask of self-assembled In₃Sn nanodots, which is described as follows: A SiO₂ layer was first deposited on the top surface of the LED wafer by plasma-enhanced chemical vapor deposition (PECVD), and an indium tin oxide (ITO) layer was subsequently deposited onto the SiO₂ layer with electron beam evaporation. The as-deposited ITO film can be assumed as an oxygen-poor ITO layer, which has a mixture of metallic phase and ITO phase, since the ITO films were deposited by using electron beam evaporator [13]. Next, the sample was dipped into 3% HCl solution to dissolve the ITO phase, which left self-assembled metallic clusters on the SiO₂ layer. The self-assembled metallic clusters were analyzed as In₃Sn clusters by X-ray diffraction and energy-dispersive spectroscopy in cross-sectional transmission electron microscopy. The SiO₂ layer was then patterned with a fluorine-based dry etching process using In₃Sn clusters as the etch mask. Finally, the InGaN/GaN nanopillar heterostructure was formed by dry-etching the LED wafer via the patterned SiO₂ hard mask. By controlling the size of the In₃Sn clusters and the ICP etching conditions, nanopillars of diameters ~200nm and ~50% filling factor were fabricated successfully. An etching depth of ~0.6 µm was designed to reach the n-type layer of the LED heterojunction and to laterally expose the emissive quantum wells along the sidewall of the nanopillars.

After the nanopillar structure was created, the oxide hard mask was stripped with buffered oxide etch (BOE), and Ti/Al (5nm/200nm) contact was patterned and deposited over the bottom of the nanopillars as n-type electrode. For the formation of the p-type electrode, two types of devices were fabricated. One type employed p-type GaN surfaces as the electrode, the other type deposited ITO on top of p-type GaN as the electrode. To make the ITO electrode, a slanted ITO nanowire structure was deposited on top of the nanopillars by the oblique-angle deposition method using electron beam evaporation. During the deposition, the sample stage was tilted at a large angle such that the vaporized source atoms reach the top surface of the nanopillars at a grazing angle. The high-aspect-ratio nanopillars block the vapor from reaching the lower pillar sections and the gap in between via the “shadowing effect”, preventing electrical shorting of the device. In so doing, zigzagged ITO nanowires were produced on top of the nanopillars using two consecutive deposition angles (85 and −85°), as illustrated with a cross-sectional scanning electron microscopy (SEM) image (Fig. 1(b)). This step was followed by a normal-angle ITO deposition (0°) to connect the nanowires with a continuous blanket ITO film and annealing in nitrogen to enhance the conductivity and transparency of the ITO.

Finally, a layer of colloidal CdSe/CdS core/shell QDs (QSP-620, Ocean Nanotech LLC) with photoluminescence (PL) emission centered at 620nm was deposited over the surface of the nanopillars by soaking the nanopillar device in the QD solution for ~12 hours. Figure 2(a) and (b) show, respectively, the SEM image and z-scan 3D confocal fluorescent microscope image of the QD-coated nanopillars (before slanted ITO deposited), confirming the conformal deposition of QDs over the sidewalls of the nanopillar structure.

 

Fig. 2 (a) SEM image and (b) Z-scan confocal fluorescent microscope image of the QD-coated nanopillars.

Download Full Size | PDF

3. Results and discussion

For LED charaterization, electrical pumping of the devices was implemented by forward-biasing the p-i-n junction of the nanopillars with a Keithley 4200 semiconductor parameter analyzer. The emission of the nanopillar LEDs was characterized with an integrating sphere-measurement to take into account the non-Lambertian radiation pattern of the nanopillar structure. The nanopillar sample was placed inside the integrating sphere (Thornlabs MA189), and the output emission of the sample was diffusely reflected by the barium sulfate-coated inner surface of the sphere and redistributed isotropically into all solid angles; the spectral and intensity detection of the light exiting from a small aperture at the sphere surface resulted in an accurate determination of the total number of photons from the nano-structured emitter. The output of the integrating sphere was coupled, via an optical fiber, to a spectrometer (Spectropro, ~0.1 nm spectral-resolution) equipped with a p-i-n photodetector. Both the dispersion of the spectrometer grating and the response of the photodetector were calibrated for the detection linearity. A barium sulfate-coated baffle was placed in front of the fiber coupler at the interior surface of the integration sphere to prevent the direct illumination of the optical fiber during the optical pumping. For photoluminescence characterization, the second harmonic of the output (λ = 400nm) of a Ti:sapphire femtosecond amplifier (800 nm, 1 kHz, 80fs, Coherent Libra system) was introduced into the integrating sphere to excite the sample optically.

Figure 3 compares the photoluminescence (PL) spectra of the GaN heterojunction before and after the nanopillar processing. The PL intensity of the InGaN QWs was spectrally integrated from 410nm to 490nm, showing a ratio of 1.17:1 between the PL emission of the nanopillar and planar structure. The observed PL enhancement for the InGaN/GaN nanopillars can be interpreted considering two mechanisms. First, the radiative recombination efficiency of (In, Ga)N is not susceptible to the surface states created during the dry etching-formation of nanopillars due to the ionic nature of the material [14]. Secondly, light extraction efficiency is substantially improved in the nanopillars by means of the reduction in total internal reflection. Hence the nanopillar LED structure with augmented brightness provides us with a proper template to investigate the nonradiative recombination between colloidal QDs and InGaN QWs for high color conversion efficiency.

 

Fig. 3 Photoluminescence spectra of the nitride LED heterostructure before and after the nanopillar processing.

Download Full Size | PDF

The output spectra of a nanopillar LED prior to and after QD deposition was plotted in Fig. 4(a) . The LED was forward biased with a current density of 62.5mA/cm². The shaded areas represent, respectively, the portion of nitride (blue) emission absorbed by the QD coating and the red emission from the same layer of QDs. The color-conversion quantum efficiency (η_c), defined as the ratio of the photon count from the QD emission ($N_{QD}^{QW\leftrightarrow QD}$) upon QW-QD coupling to that from the net QW emission without QD coupling ($N_{}^{QW}$), ${\eta }_c=N_{QD}^{QW\leftrightarrow QD}/N_{}^{QW},$ was calculated to be ~12.4%. In addition, an effective internal quantum efficiency of QD emission,${\eta }_{eqi}=N_{QD}^{QW\leftrightarrow QD}/\left(N_{}^{QW}-N_{QW}^{QW\leftrightarrow QD}\right),$ was defined in the present study, wherein $N_{QW}^{QW\leftrightarrow QD}$ represents the photon count of blue emission from the coupled QW-QD system. An effective internal quantum efficiency of η_eqi = 43.6% was calculated from the output spectra of the nanopillar LED (Fig. 4(a)).

 

Fig. 4 (a) Electroluminescence spectra of the nanopillar LED sample with and without QD coupling; (b) Electroluminescence spectra of a planar control sample with and without QD coupling. Both samples are from the same LED heterostructure.

Download Full Size | PDF

For the sake of comparison, a control sample was prepared by depositing QDs on the top surface of a planar LED structure containing an identical p-i-n heterojunction as well as a planar top ITO contact. The deposition condition was carefully tailored to ensure that approximately the same number of QDs were deposited on both the planar and nanopillar LEDs, which was verified by optical density (OD) characterization. In the control sample, QDs are physically separated from the InGaN quantum wells by a 240nm-thick p-type GaN capping layer as well as a 200nm-thick ITO coating. The color conversion in the control sample can be implemented only via the processes of absorption and re-emission that are typical for the PL phenomenon. The emission spectra of the planar control sample prior to and after the QD deposition are shown in Fig. 4(b). The internal quantum efficiency of the QDs in the planar control sample was calculated to be η_eqi ≈12.0%, which is consistent with the intrinsic PL quantum yield (η_i) of the QD films that was measured with the integrating sphere technique.

Compared with QDs in the control sample, the effective internal quantum efficiency of the QDs incorporated in a nanopillar LED exhibits a remarkable augmentation of 263%. Since the intrinsic quantum yield of the QDs (η_i) does not exhibit any dependence on the substrate topography, the dramatically enhanced effective internal quantum efficiency of the QDs in nanopillar LEDs can be interpreted only by the presence of substantial nonradiative energy transfer between the QDs and QWs at the sidewall of the nanopillars.

Finally, a time-resolved photoluminescence study was conducted to confirm the existence of the nonradiative energy transfer. The QWs in the nanopillar sample were photo excited with 80fs-pulses of 400nm-wavelength and 1 KHz repetition rate. The nonlinear Kerr shutter technique was employed with a temporal resolution of ~1.5 ps [15, 16]. The photoluminescence from the InGaN QWs (λ~450nm) was collected and detected with a photomultiplier tube. As illustrated in Fig. 5 , a decay lifetime of ~150 ps was measured in the nanopillars without QDs, whereas the lifetime drops to 125 ps upon the QD deposition over the surfaces of the nanopillars, indicating the carriers in the InGaN QWs decay faster with the presence of QDs. Since the deposition of QDs does not change the intrinsic carrier dynamics of the InGaN QWs, this faster decay suggests that an additional relaxation channel is introduced for the carriers in the InGaN QWs, which can only be the nonradiative energy transfer from QWs to QDs. The rate of nonradiative energy transfer, $r_{transfer}^{QW\to QD}$, can therefore be determined directly as$r_{transfer}^{QW\to QD}=\frac{1}{125}-\frac{1}{153}=1.46{\text{ns}}^{-1}$. This value becomes comparable to the rate of radiative recombination, suggesting the important role of non-radiative energy transfer in the color conversion process of QD-coupled nanopillar LEDs.

 

Fig. 5 Time-resolved photoluminescence characterization of a nanopillar LED sample with and without QD overcoating. The nanopillars exhibit an average diameter of ~200nm.

Download Full Size | PDF

The rate of nonradiative energy transfer can also be obtained and confirmed from the electroluminescence results in Fig. 4. The luminescence of the QDs in the nanopillar structure can be expressed as

In the present work, $r_{transfer}^{QW\to QD}$can be calculated from Eq. (3) using the measured parameters, η_eqi~43.6.0%, η_i~12.0%, (1-T) ~25.2%, as well as the reported value for $r_r^{QW}$~2.0 ns⁻¹, which gives $r_{transfer}^{QW\to QD}~1.32$ ns⁻¹. This result agrees well with the energy transfer rate (1.46 ns⁻¹) obtained from the aforementioned time-resolved photoluminescence study, which adds further evidence to the presence of nonradiative energy transfer between InGaN QWs and CdSe/ZnS QDs.