1. Introduction

Scaling down active nanophotonic devices, namely nanolasers and nanolight-emitting diodes (nanoLEDs), to deep sub-micrometer (sub-µm) sizes, is crucial to achieve ultrasmall (<1 µm²), low energy consumption (<10 fJ/bit), and efficient (>10%) light sources, as needed for compact photonic integrated circuits (PICs) for optical communications [1], biosensing and bioimaging [2–4], and neuromorphic photonic sensing and computing [5,6]. In the last few years, there has been an explosion of novel nanoscale light sources, such as photonic crystal lasers [7–10], metal-dielectric nanolasers [2,11–15], and nanoLEDs [16–18], employing III-V compound semiconductor materials as the active medium. Remarkably, these high refractive index materials, arranged in subwavelength structures, have shown recently to provide efficient ways to manipulate light at the nanoscale through mode interferences and enhancement of both electric and magnetic fields, which enables novel lasers [19], [20], and sensor [21] devices. However, at these small scales, as the surface-to-volume ratio of these nanosources increases substantially, among the numerous challenges, strong non-radiative processes due to defects at the surface and difficulties in extracting the light have been shown to have a detrimental effect on the efficiency of nanoLEDs [18] and nanolasers [22] at room-temperature.

Substantial efforts have been realized on large area LEDs to achieve more efficient light out-coupling, such as substrate modification, use of a scattering medium, microlens arrays, microcavity, or photonic crystals [23,24], and surface plasmon enhanced techniques (for a comprehensive review see [25]), and more recently by deploying metasurfaces [26]. However, some of these methods become challenging to implement when the area of the light sources is reduced to the deep sub-µm². Alternatives for light extraction include to efficiently couple the generated light emission in nanolight sources to nanowaveguides integrated with grating couplers [18], or plasmonic waveguides [16], but thus far the external quantum efficiency at room-temperature of nanoLEDs based on these approaches has been limited to values around 10⁻³ and 10⁻⁷, respectively. Specifically, LED sources with subwavelength sizes usually require metallo-dielectric [18] or plasmonic [16,27] structures leading to undesired device degradation due metal losses effects. Recently, subwavelength optical resonators made of high refractive index materials have been proposed for the GaAs/AlGaAs material system. These do not require the use of metals and take advantage of the strong enhancement of light–matter interactions via the physics of bound states in the continuum (BICs) enabling applications in nonlinear optics [28], nanoscale lasers [29], quantum photonics, and sensors. Despite these progresses, the extraction efficiency from these subwavelength structures has been largely unexplored. Since the GaAs/AlGaAs is a key compound material for photonics providing optical emission and absorption in a wide range of wavelengths, the development of devices from this material with a large extraction efficiency and an ultrasmall size for energy saving applications is highly desirable for high-brightness applications in the deep-red and near-infrared wavelengths.

In this work, we report a strong enhanced signal emission in single vertical-emitting AlGaAs/GaAs/AlGaAs nanopillars, composed by a double barrier quantum well (DBQW) nanostructure in the central GaAs region. When the emitting nominal area is decreased to the sub-µm scale we show a reproducible, robust, and long-term enhanced signal, exhibiting for the best samples a 37-fold enhancement of the photoluminescence (PL) intensity on pillars with dimensions of ∼300 nm without deploying high-Q dielectric cavities nor plasmonic effects. This results in an estimated extraction efficiency ∼0.6 and approaches previous record efficiency values of 0.72 in single vertical nanowires in III-V materials [30], and 0.79 achieved using nanorods in GaN materials [31]. However, these works employed different semiconductor compound material systems and exploited other methods such as back reflectors to further control the emission. For example, the work of Claudon et al. [30] employs nanowires with an in-plane oriented quantum-dot emitter at cryogenic temperatures, and takes advantage of carefully tailored nanowire ends (back reflector in the substrate and top tapper) to funnel more efficiently the spontaneous emission into a given mode. Our results show that a pronounced extraction effect can be achieved in sub-µm pillars using III-V spectrally broad emitters (quantum wells or bulk) operating at room temperature, covering a wide range of visible and infrared wavelengths, despite the presence of surface recombination effects, which is typically the situation of many LED sources. Finite difference time domain (FDTD) simulation results of both micro- and nanopillar structures confirm the efficient light emission in nanopillars is a result of a large extraction efficiency of emitted light in sub-µm pillars, whereas in micropillars most of the light is trapped in the semiconductor via total internal reflection which limits the light out-coupling to ∼2%. Time-resolved µ-PL (TRPL) using time-correlated single-photon counting (TCSPC) measurements of the emission dynamics at room-temperature allow us to estimate the surface recombination velocity of ∼10⁵ cm/s and ∼2.5 × 10⁵ cm/s for passivated and unpassivated samples, respectively. With these results, the external quantum efficiency (EQE) of fabricated pillars is estimated. We conclusively show that, in the best case scenario of unity injection efficiency and taking as a reference the best values of surface recombination reported in [32], the EQE in vertical-emitting nanopillar-based LEDs can potentially reach a high external efficiency ∼45%, whereas the external efficiency of µm-based LED pillars saturates at ∼2%. The light extraction efficiency reported here is achieved based on the nanopillars morphology and ultrasmall size, leading to strong reduction of light trapping via mode suppression. Importantly, it does not require further nanostructuring e.g. of high-Q dielectric cavities, or array structures, nor encapsulation techniques, which typically are challenging to integrate into nanoscale devices. Our results pave the way for the development of high-performance nanoscale optoelectronic devices such as nanoLEDs and nanolasers for applications in nearfield spectroscopy and sensing, integrated optical interconnects, optical neural networks, solid-state lighting, and optical communications.

2. Fabrication of subwavelength vertical-emitting GaAs/AlGaAs nanopillars

A systematic experimental study was performed to investigate the size effect on AlGaAs/GaAs/AlGaAs nanopillars. The layer stack, shown schematically in Fig. 1(a), is composed from top to bottom by 150 nm of AlGaAs (30% Al), ∼50 nm of a GaAs-based compound material consisting of GaAs (20 nm)/AlAs(3nm)/GaAs (6 nm)/AlAs (3nm)/GaAs (20 nm) double barrier quantum well (DBQW) nanostructure, 150 nm of AlGaAs (30% Al), and 300 nm of GaAs, all not intentionally doped, and grown by molecular beam epitaxy (MBE) on an undoped GaAs substrate. The selection of the GaAs-DBQW nanostructure is mainly motivated by its quantum resonant properties that can be exploited to achieve nanoLEDs with electro-optical pulse responses of interest for neuromorphic nanophotonic computing [6]. However, in this work we are mainly interested on the optical properties of these layers to achieve efficient light emission.

 

Fig. 1. Semiconductor epilayer stack and examples of representative micro- and nanopillars. (a) Schematic of an encapsulated AlGaAs/GaAs/AlGaAs nanopillar with width d. (b)-(e) Scanning electron microscopy images of: (b) circular nanopillar with a height of ∼2.5 µm, (c) circular nanopillar with a height of ∼0.6 µm and slopped sidewalls, (d) circular micropillar with similar slopped sidewalls, and (e) pillar array of circular nanopillars ranging from 200 nm to 800 nm (central row).

Download Full Size | PDF

The fabrication of the micro- and nanopillars involved electron beam lithography (EBL) using a Vistec 5200 ES 100 kV tool where the pillars were patterned using a negative e-beam resist of 500 nm thickness with a 200 nm SiO_x hard mask deposited by plasma enhanced chemical vapor deposition (PECVD) at 300 °C. After exposure of the resist using EBL the sample was developed transferring the pattern to the e-beam resist. The next step involved transferring the pattern from the resist to the hard mask using reactive ion etching (etch rate of ∼594 nm/min). Once the pattern was transferred, the remaining resist was removed by oxygen plasma followed by deoxidation in NH₄OH:H₂O. The following step was to etch the pillars by dry etching using inductively coupled plasma (ICP) in an SPTS ICP machine to etch the pillars until ∼0.6 µm depth (etchings up to ∼2.5 µm were also successfully realized although in this work we have focused our analysis in pillars with heights ranging from 0.5 µm–0.8 µm). After, the remaining SiO_x hard mask was etched with hydrofluoric (HF) acid in a vapour etcher tool followed by one cycle of oxygen plasma treatment to clean the surface of the pillars. The last step was a deoxidation step using a solution of NH₄OH:H₂O. The samples were then passivated and coated with a 50 nm thick dielectric material (SiO_x). For a complete description of the fabrication, passivation and dielectric coating methods, see Supplementary Information and Fig. S1.

Figures 1(b)–1(e) show representative examples of fabricated semiconductor pillars without dielectric capping. The samples comprised pillar structures with circular, Figs. 1(b)–1(d), square, and rectangular (not shown) geometries. Figure 1(b) shows the scanning electron microscope (SEM) picture of a d∼360 nm wide circular nanopillar, where d is the pillar diameter. The samples contained pillars with dimensions ranging from 200 nm to 8 µm width organized in arrays, Fig. 1(e), spaced by at least 10 µm so that the emission from pillars could be collected and analyzed individually. Figure 1(d) shows an example of a large micropillar (8 µm width). Micro- and nanopillars as the one shown in Fig. 1(b) typically displayed slightly slopped sidewalls. Structures with more pronounced tapering effect were also fabricated and tested. An example is shown in Fig. 1(c) showing a SEM picture of a nanopillar with ∼0.6 µm deep etch. Although the tapering effect has not been systematically analyzed here, previous works [30,33] suggest that tapered top facets of long nanowires, as we decrease diameter, can have a role in increasing the light transmission resulting in efficient adiabatic coupling of the photons from the semiconductor to air. This method could be used to further optimize the light management and extraction of our nanopillar structures.

3. Characterization and simulation of the light extraction efficiency

3.1 Photoluminescence

Semiconductor AlGaAs/GaAs/AlGaAs pillars with diameters ranging from 200 nm to 8 µm lateral width were characterized at room-temperature using a micro-photoluminescence (PL) microscope setup covering the visible range of the spectrum and using a λ=561 nm laser excitation (Supplementary information), aiming at a comprehensive study of the PL quality of the pillars. Firstly, we analyze the emission properties of the AlGaAs layer stack, with a similar high refractive index of the GaAs. The measurements were taken for micro- and nanopillars with i) circular, ii) square, and iii) rectangular shapes with a height of ∼0.6 µm, similar to Figs. 1(c) and 1(d). Figure 2(a) presents examples of confocal microscope images showing the emission for both optically pumped micropillars (top) and nanopillars (bottom) for the circular geometry (the respective intensity profiles are overlaid). The image data was recorded with spectral sensitivity, such that for each pillar the respective PL spectrum was collected (Supplementary information). The respective integrated intensity was plotted as a function of the pillar size and the results are summarized in Fig. 2(b), while representative spectra are shown in Fig. 3.

 

Fig. 2. Photoluminescence of single nanopillars. (a) Continuous-wave micro-PL measurement results for circular micro- and nanopillars at room-temperature showing the confocal microscopy images and intensity profiles from (top) micropillars and (bottom) nanopillars. (b) Integrated intensity plot as a function of diameter (the dash line shows a ${d^2}$ dependence). For the case of the rectangular pillars the value d is the length of the pillar.

Download Full Size | PDF

 

Fig. 3. Measured photoluminescence spectra of the AlGaAs for micro- and nanopillars as a function of the pillar diameter.

Download Full Size | PDF

For the case of micropillars, top panel of Fig. 2(a), clearly the light emission is reduced as the diameter d decreases, following a typical scaling area-law, ${d^2}$, of planar LEDs. However, as d is reduced from the micrometer size down to 0.2 µm, bottom panel of Fig. 2(a), although the nominal emission area is reduced by a factor of more than 100, the intensity is reduced only by ∼10 times. For example, the integrated PL intensity, in the range of 300 nm < d < 900 nm (see also Fig. S2), is comparable with the intensity found for pillars with sizes of d∼1 µm, Fig. 2(b). This strongly deviates from the ${d^2}$ dependence observed in micropillars and results in a typical enhancement ranging from 20 to 37-fold for the pillars ranging from 500 nm > d > 300 nm, Fig. 2(b). Noteworthy, a 37-fold enhancement of the emission was achieved, as determined for the d = 360 nm circular pillar for the case of best samples. Although this result was the maximum observed enhancement and is not a statistical result, this emission peak was reproducibly measured in multiple samples for all nanopillars within the range of 300 nm < d < 400 nm for both circular, square and rectangular geometries (see results of Figs. S2 and S3 in Supplementary Material). The observed emission peak in this diameter range may indicate the occurrence of a resonant effect, although neither pronounced Fabry-Pérot nor whispering gallery mode resonances are predicted for these pillar sizes (also confirmed in FDTD simulations, section 3.2) due to the small height and width of the pillars, typically smaller than the emission wavelength. Further, as seen is Fig. 3, the spectral properties are almost independent of the pillar size or shape with a constant full width half maximum (FWHM) of ∼20 nm and emission peak at ∼670 nm, and therefore without exhibiting relevant resonant effects. This is markedly different from the case of the 2D array configuration where the Q-factor and emission wavelength are strongly dependent on the size and pitch of the individual elements of the array. Therefore, we attribute this enhancement observed all nanopillars with diameters ∼360 nm to a relatively strong mode overlap between the mode and the active medium for this optimum nanopillar diameter, which provides a high transmittance for light extraction, a situation also observed in our simulations (section 3.2) and also in GaN QW nanorods [31].

In order to analyze the observed light enhancement, we first determine the extraction efficiency for planar structures, which are typically dominated by the escape cone of internal light, given by ${\eta _{ex}}\sim 1/4{n^2}$, where n is the refractive index of the semiconductor. Effectively, for the case of the material analyzed here (n=3.5), a small extraction efficiency of 2% is estimated as a consequence of the large value of the refractive index. Therefore, only a small fraction of the emitted light can be directly radiated into air. Taking this value as a reference for the light out-coupling in micropillars, one can estimate the respective enhancement observed for the nanopillars. We consider the measured µ-PL intensity is expressed as PL intensity ∝ P_pumpAη_absη_intη_lensη_ex, where P_pump is the pump power, A is the active area, η_int is the internal quantum efficiency and η_abs and η_lens are the absorption efficiency and the collection efficiency of the PL by the objective, respectively. Here, we assume P_pump, η_abs and η_lens to be constant since the incident flux of the excitation P_pump, and η_lens are kept the same for all the measurements. For the P_pump used no major changes of the PL emission peak position and spectral shape were observed, Fig. 3, indicating similar absorption efficiency for all pillars. Lastly, we assume that η_int is also kept constant. As shown in supplementary information and Fig. S5, for the typical optical pumping levels used in the experiments and considering the respective carrier densities as a function of active volume, effectively η_int∼0.05 across the pillar sizes analyzed here. Therefore, by plotting the integrated intensity, Fig. 2(a), per effective emitting area as a function of pillar size one can estimate the extraction efficiency (see Fig. S4 for more information). Considering the largest micropillar (8 µm) correspond to η_ex∼2%, we estimate typical extraction efficiencies in the range of 33-57% for nanopillars <500 nm width (Fig S4). Effectively, the strong enhancement observed here is a result of large extraction efficiency, which is particularly remarkable taking into account the drastic reduction of the nominal emission area and the strong non-radiative effects typical of GaAs/AlGaAs materials, as analyzed in section 3.3.

The light extraction enhancement was also analyzed for the GaAs DBQW material and the results are shown in Fig. 4. Here, pillars down to ∼600 nm lateral width were characterized at room-temperature using a micro-photoluminescence (PL) microscope coupled to a spectrometer and covering the near-infrared region of the spectrum (Supplementary information). A laser excitation at λ=532 nm was used in the experiments. The spectra for pillars ranging from 0.59 µm to 1 µm are shown in Fig. 4(a). Similarly to the AlGaAs, the overall emission properties are almost independent of the pillar size or shape with similar FWHM nm and in this case an emission peak at ∼860 nm. The emission from nanopillars below 600 nm is not shown due to the GaAs substrate emission that limits the measurements for the smallest pillars. The integrated intensity for circular, square and rectangular shapes is plotted as a function of the pillar size in Fig. 4(b). Similarly as in the case of the AlGaAs layer stack, the collected emission also deviates from the ${d^2}$ dependence observed for micropillars. Assuming identical extraction efficiency of 2% for planar GaAs-based structures, the estimated extraction efficiency for the best case of the 0.59 nm pillar is 10-fold. The strong enhancement observed in this work for both GaAs and AlGaAs layer materials in single vertical-emitting nanopillars is reached without the complex design of high-Q cavities, such as in photonic crystal structures [23], nor requiring plasmonic effects [34]. Noteworthy, the extraction efficiencies reported here for the best cases are comparable to previous reports showing 73% extraction efficiency [23], or a 16-fold enhancement of the emission [35], for GaN 2D pillar array structures. However, we note these previous works rely on 2D array structures, where light scattering in the array plays a key role in the light extraction. Lastly, the estimated extraction efficiency ∼60% approaches previous record efficiencies in single structures including the work of Claudon et al. in single vertical nanowires in quantum dot materials [30] and the work of Kuo et al. in single nanorods in GaN QW materials [31]. However, these works employed semiconductor material systems with high internal quantum efficiencies. Our results demonstrate that this substantial extraction effect in sub-µm pillars can be extended to III-V spectrally broad emitters operating at room temperature without using further methods such as back reflectors as reported in the work of Claudon et al.

 

Fig. 4. Measured photoluminescence spectra of the GaAs for micro- and nanopillars. (a) Measured photoluminescence spectra of the GaAs for micro- and nanopillars with a circular geometry as a function of the pillar diameter. (b) Integrated intensity plot as a function of diameter (the dash line shows a d² dependence). For the case of the rectangular pillars the value d is the length of the pillar.

Download Full Size | PDF

3.2 Simulations of the light extraction efficiency

In the following, we analyze the light extraction efficiency observed experimentally as a function of lateral size of the pillar structures using FDTD simulations (Supplementary Information). In the FDTD simulations we used a 3D FDTD method based on Yee’s algorithm with a perfectly matched layer boundary condition. We assumed an identical epilayer stack semiconductor structure as the one reported experimentally, Fig. 1(a). In the FDTD, a single dipole source is positioned at the center of the x-y plane of either the AlGaAs or GaAs DBQW regions in the vertical direction (see Fig. 5). The dipole source (with a Gaussian shape) is polarized with in-plane direction orientation (x-direction). Also, this way we avoid surface-dependent effects typically found in dipoles close to the periphery [36].

 

Fig. 5. 3D FDTD simulation results of the electric field distribution for (i) micro- and (ii) nanopillar structures analyzed for both (a) AIGaAs and (b) GaAs materials. Electric field intensity distribution for (i) a micropillar with a diameter of 3 µm and (ii) a nanopillar width of 360 nm. A dipole excitation (indicated with an arrow) was used with an in-plane polarization oriented in the x-direction and placed in the pillar axis. The color scale bar represents the relative strength of the electric field intensity. The dashed lines inside the pillars indicate the different semiconductor layers, corresponding to, from top to bottom, AlGaAs/GaAs/AlGaAs/GaAs. Red is the maximum intensity (normalized to 1).

Download Full Size | PDF

Figure 5 presents FDTD simulations showing the electric field distribution of light emitted for a micropillar (d=3 µm), panel (i), and a nanopillar (d=360 nm), panels (ii). For the case of the micropillar, the electromagnetic field intensity exhibits a significantly stronger backward emission than forward emission. As a result, due to total internal reflection effect, it is challenging to achieve efficient light extraction into air from micropillar structures without further light management methods. However, as size is reduced in nanopillars a strong suppression of guided modes is achieved (there is only a few guided modes left, namely HE₁₁ and TE₀₁ [36], to trap light inside the semiconductor), leading to a preferred forward light propagation which is a key aspect to achieve an increase of light extraction. Furthermore, the radiation pattern can deviates substantially from planar LEDs and become more directional. We have simulated the radiation patterns of a representative nanopillar (360 nm), Fig. 6(a). The results indicates the top emission into air for nanopillars as a more collimated propagation along the z-direction. This does not follow the typical Lambertian pattern of planar LEDs [dashed curve in Fig. 6(a)] and instead is more directional, which can be used to tailor the radiation pattern of nanoscale LED sources.

In order to quantitatively compare the achieved emitted PL in the experiments, we have simulated the light extraction efficient in a few representative structures where the light enhancement was more pronounced in the experiments, specifically in the range from 300–500 nm. We have compared the situation of a single dipole as described previously, and the case of an ensemble of dipoles by running a series of simulations using a dipole source in various positions and orientations (see Supplementary Information). This represents the more realistic case of our bulk and quantum well type of materials with homogeneous broadening consisting of an atomic ensemble of identical incoherent (wide) solid-state emitters at room temperature. The results are shown in Fig. 6(b) showing efficiency ranging from 0.6–0.84 for calculations using a single dipole source and values ranging from 0.4–0.7 assuming a more realistic case of an ensemble of dipoles randomly distributed across the active material with random polarizations. These values are in line with the values achieved experimentally for the same diameter range, Fig. S4. We notice a peak enhancement in the efficiency in pillar sizes at d=400 nm. We attribute this to a relatively strong mode overlap between the mode and the active medium for this optimum nanopillar diameter, specifically for the case of dipole sources oriented in the z-direction, similarly as observed in the experiments, Fig. 2(b). It is expected that further optimization of the extraction efficiency can be achieved by carefully tailoring the morphology properties of the pillars such as pillar height and tapering effect [30,33].

 

Fig. 6. (a) Simulation of the emission patterns for a representative nanopillar of 360 nm (solid green line). The result is compared with the wide angle Lambertian distribution of a plain slab structure representative of a large micropillar LED (dashed black line). All plots are normalized to the maximum directional emission. (b) Light extraction efficiency as a function of pillar diameter d. The calculations assumed a single dipole located on the pillar axis (red circles), and the case of an ensemble of dipoles averaged over all positions and dipole orientations (blue circles).

Download Full Size | PDF

Lastly, we note that no significant Purcell enhancement of the spontaneous emission is expected for the nanopillars analyzed here. For the case of spectrally broad emitters (QWs or bulk), the Purcell enhancement scales as F_p∼1/V, where V is the mode volume. As extensively discussed in previous works [37,38], in the case of broad emitters embedded in ultrasmall cavities (metal or plasmonic) for strong mode confinement, only in the case of aggressive scaling (∼100 nm) of the sources some Purcell enhancement is expected. The Purcell enhancement factor was estimated in our pillar structures (Supplementary Information). We calculate a Purcell enhancement of around 3 for a 360 nm wide pillar in the mode of interest in the ideal case of a monochromatic dipole, but this factor is reduced to a value below one, that is no enhancement, when taking into account the homogeneous and inhomogeneous broadening in the active region [37]. These results are in line with previous work for the case of a 350 nm nanopillar LED using InGaAs material operating at room-temperature [18] that reports a Purcell enhancement of the spontaneous emission close to one, even in in the presence of a metallic cavity.

3.3 Surface recombination velocity

To further characterize the optical properties of the pillar structures, we performed time-resolved photoluminescence spectroscopy (TRPL) measurements (Supplementary Information) using a time-correlated single-photon counting (TCSPC) method to investigate the carrier dynamics in AlGaAs/GaAs/AlGaAs pillars. The measurements included both untreated and passivated pillars. For the passivation, the samples were submerged in a solution of (NH₄)₂S for 10 min followed by deposition of SiO_x via PECVD forming a 50 nm thick protecting capping layer (see Supplementary Information).

In the TRPL experiments, the pillars were optically pumped using a picosecond pulsed laser at 561 nm with a full-width half-maximum (FWHM) of 80 ps and a repetition rate of 50 MHz. The PL was filtered and detected by an avalanche photo detector (APD) with a temporal resolution of ∼ 225 ps according to the instrument response function (IRF) shown in the inset of Fig. 7(a). In Fig. 7(a) the measured decay curves are shown for circular micropillars from passivated samples with a width of 8 µm, 6 µm and 4 µm. In the inset a comparison with an unpassivated 8 µm pillar is shown. The decay curves follow a typical size-dependence, which scales as ${{4{\upsilon _s}} / d}$, where ${\upsilon _s}$ is the surface recombination velocity and d is the width of the pillar. We note that TRPL decay curves of pillars <4 µm were not measured here due to the fast carrier lifetimes <200 ps of the smallest pillars, faster than the temporal resolution of our setup. However, as shown in [39], surface recombination is a material dependent parameter, and in the case of low injection conditions its value is approximately the same for both micro- and nanopillars. The TRPL decay curves are fitted using a single exponential decay function to obtain the values of the carrier recombination lifetime. The rate of the measured PL decay defined as $\tau _{PL}^{ - 1}$, is related to the radiative, $\tau _r^{}$, and nonradiative, $\tau _{nr}^{}$, carrier lifetimes by the expression $\tau _{PL}^{ - 1} = \tau _{nr}^{ - 1} + \tau _r^{ - 1}$, where the nonradiative rate may have a bulk and a surface contribution. Assuming that the surface-related nonradiative recombination rate scales as ${{4{\upsilon _s}} / d}$, the surface recombination velocity can be estimated directly from the size-dependent determined carrier lifetimes in the low injection regime [40]:

 

Fig. 7. Time resolved photoluminescence (TPRL) of passivated and unpassivated nanopillars and estimation of associated inverse carrier lifetimes. (a) Experimental TRPL decay curves measured at room temperature for 8 µm, 6 µm, and 4 µm wide micropillars after sulfur passivation treatment followed by silicon oxide coating. In the inset is shown TRPL decay curves for the 8 µm wide micropillar before and after passivation. The respective carrier recombination lifetimes are also displayed which are extracted from the measured TRPL curves fitted with a single exponential decay function. The IRF is shown inset in (a) in pink. (b) Inverse carrier lifetime, τ_PL⁻¹, estimated from the TRPL measurements versus the inverse pillar width, d⁻¹, before (blue dots) and after passivation (black dots). Also shown is the corresponding linear fit (dashed grey curves) of the experimental data using Eq. (1) allowing us to estimate the corresponding surface recombination velocity, υ_s, values.

Download Full Size | PDF

3.4 External quantum efficiency

In the following, from the experimental results achieved for both the extraction efficiency and surface recombination velocity, the theoretical $EQE = {\eta _{ex}}{{\tau _r^{ - 1}} / {({\tau_r^{ - 1} + \tau_{nr}^{ - 1}} )}}$ of either micro- or nanopillars can be estimated (Supplementary Information), and their potential for high-performance nanoscale LEDs is discussed. Figure 8 shows the calculated EQE results as a function of carrier density for the case of (a) a micropillar (8 µm) using a low value of extraction efficiency, ${\eta _{ex}} = 0.02$, and (b) the case of a 360 nm wide nanopillar assuming a value of extraction efficiency for the best case scenario reported, here ${\eta _{ex}} = 0.57$. We also consider two different surface recombination values ${\upsilon _s} = {10^5}$ cm/s (black curve) and ${\upsilon _s} = {10^3}$ cm/s (red curve), respectively. The ${\upsilon _s} = {10^5}$ cm/s was taken for the best passivated samples observed here, Fig. 7(b), while the value ${\upsilon _s} = {10^3}$ cm/s was taken from works reported elsewhere [32,40]. As shown in Fig. 8(a) for the case of the micropillars, even in the best case scenario of a small ${\upsilon _s}$, the values of EQE are limited to a maximum of 0.02, fixed by the extraction efficiency. Remarkably, in the case of moderate or high injection conditions (carrier density n>10¹⁸ cm^-3), where typically LED and laser sources are operated, the EQE of a nanopillar at room temperature can reach similar values of 0.02 for ${\upsilon _s} = {10^5}$ cm/s and a much higher value of 0.45 for ${\upsilon _s} = {10^3}$ cm/s (values taken at n=2 × 10¹⁸ cm^-3), Fig. 8(b). We note these calculations do not take into account temperature effects, which may become relevant at high carrier injection. Also, we assumed the case of unity injection efficiency and negligible Auger recombination effect. Nevertheless, for the carrier injection levels analyzed here, the Auger recombination does not play a strong role in the values of the EQE, excluding the case of very large carrier densities (>10¹⁹ cm^-3) – see Supplementary Information Fig. S6. Compared to recently reported nanoLED devices with a similar or larger footprint than the structures analyzed here, namely photonic crystal LEDs [17], plasmonic LEDs [16], and metal-cavity LEDs [18], we note these suffer from low external quantum efficiencies at room temperature in the range of 10⁻⁷ –10⁻³. Only in the case of optically pumped metallic nanostructures in the deep sub-wavelength (<<100 nm) quantum efficiencies exceeding 50% were shown [34], while realization of electrical injection is still to be demonstrated. Noteworthy, while the focus in nanoLEDs has been mostly in strategies to couple light to waveguides, in this work we show that outcoupling of the emission to air in nanopillars can provide large benefits in terms of EQE enhancement without requiring the complex fabrication of nanocavities to exploit the Purcell effect nor encapsulation techniques to optimize the extraction efficiency.

 

Fig. 8. Estimated EQE at room-temperature as a function of carrier density for (a) micropillar (d = 8 µm), and (b) nanopillar (d = 360 nm). The traces in black show the calculated EQE assuming the value of nonradiative recombination taken from the surface recombination velocity for the measured passivated samples, Fig. 7. The traces in red show the calculated EQE in the best case scenario of a low surface recombination velocity of 10³ cm/s reported elsewhere [32,40].

Download Full Size | PDF

4. Conclusion

In summary, a large improvement of light-extraction with an enhancement factor of up to 37 in sub-wavelength high-index material vertical-emitting nanopillars is achieved. This enhancement results in light out-coupling ∼60%, without requiring high-Q cavities, plasmonic effects, large dense arrays, back reflectors, nor encapsulation techniques. This enables highly efficient nanolight-emitting sources reaching performances comparable to micrometer-sized devices, in both deep-red and near-infrared emission. These results compare with previous record efficiency of 0.72 using single vertical nanowires in III-V quantum dot materials operating at cryogenic temperatures [30] and 0.79 achieved using nanorods in GaN materials [31]. Our results demonstrate that the extraction effect in sub-µm pillars can be extended to III-V spectrally broad emitters (quantum wells or bulk), despite the presence of strong surface recombination effects, which is typically the case of many LED sources operating at room temperature. Noteworthy, we show that in the best case scenario of unity injection efficiency and taking as a reference the best values of surface recombination reported in [32,40], the EQE in vertical-emitting nanopillar-based LEDs, taking advantage of the large extraction efficiency, can potentially reach a high external efficiency ∼45%, whereas the external efficiency of large µm size LED pillars without further methods saturates at ∼2%. It is also expected that further optimization of the extraction efficiency can be achieved by tailoring the morphology properties of the pillars such as pillar height and tapering effect [30,33]. The development of energy-efficient and compact nanostructured materials and nanodevices with the light management presented here will play a crucial role for the development of high-performance nanoscale optoelectronic devices not only in ultra-high speed information and communication technologies, but also in single-photon sensing, biosensing and bioimaging, and efficient photovoltaics.