Effects of Reduced Exciton Diffusion in Ingan/gan Multiple Quantum Well Nanorods

We investigate the effects of reduced exciton diffusion on the emission properties in InGaN/GaN multiple-quantum-well nanorods. Time-resolved photoluminescence spectra are recorded and compared in dry-etched InGaN/GaN nanorods and parent multiple quantum wells at various temperatures with carrier density in different regimes. Faster carrier recombination and absence of delayed rise in the emission dynamics are found in nanorods. Many effects, including surface damages and partial relaxation of the strain, may cause the faster recombination in nanorods. Together with these enhanced carrier recombination processes, the reduced exciton diffusion may induce the different temperature-dependent emission dynamics characterized by the delayed rise in time-resolved photoluminescence spectra. of various surface textured layer in InGaN LEDs for high light extraction efficiency, " Opt. " High output power density from GaN-based two-dimensional nanorod light-emitting diode arrays, " Appl. Phys. Optical performance of top-down fabricated InGaN/GaN nanorod light emitting diode arrays, " Opt. nanorod array white light-emitting diode, " Appl. light emitting diodes using dislocation-free indium gallium nitride/gallium nitride multiquantum-well nanorods arrays, " Nano Lett. luminescence from strain relaxed InGaN/GaN nanotips for highly efficient light emitters, " Opt. extraction efficiency of GaN-based light-emitting diodes with ZnO nanorod arrays grown using aqueous solution, " Appl. Impact of localized states on the recombination dynamics in InGaN/GaN quantum well structures, " J. Role of self-formed InGaN quantum dots for exciton localization in the purple laser diode emitting at 420 nm, " Appl. Carrier recombination dynamics in InxGa1-xN/GaN multiple quantum wells, " Phys. Quantum-confined stark effect due to piezoelectric fields in GaInN strained quantum wells, " Jpn. Dynamics of polarized photoluminescence in m-plan InGaN/GaN quantum wells, " J. Time-of-flight measurements of charge carrier diffusion in InxGa1-xN/GaN quantum wells, " Phys.component photoluminescence decay caused by carrier transport in InGaN/GaN multiple quantum wells with indium aggregation structures, " Appl. Fabrication of InGaN/GaN nanorod light-emitting diodes with self-assembled Ni metal islands, Two mechanisms of blueshift of edge emission in InGaN-based epilayers and multiple quantum wells, " Appl. Investigation of fast and slow decays in InGaN/GaN quantum wells, " Appl. Influence of free carrier screening on the luminescence energy shift and carrier lifetime of InGaN quantum wells, " J. Time-resolved spectroscopy of InxGa1-xN/GaN multiple quantum wells at room temperature, " Appl. " Observation and modeling of the time-dependent descreening of internal electrical field in a wurtzite GaN/Al0.15Ga0.85N quantum well after high photoexcitation, " Phys. Coexistence of quantum-confined Stark effect and localized states in an (In,Ga)N/GaN nanowire …

In this work, we investigated the nanofabrication impacts on the emission dynamics in InGaN/GaN MQW nanorods with emphasis on the effects of reduced exciton diffusion.We carry out time-integrated (TI) and time-resolved (TR) photoluminescence (PL) experiments on dry-etched MQW nanorod LED structures and the parent MQWs at various temperatures with a broad range of carrier density.We observe different temperature-dependent behaviors of steady and transient emission in nanorods and MQWs.While the emission intensity of TIPL in nanorods decreases monotonically with increasing temperature, the emission intensity in MQWs remains nearly unchanged with temperature up to 120 K.The thermallyactivated carrier transport exhibits delayed-rise in TRPL traces recorded from MQWs when temperature increases [24], but the delayed-rise characteristic is not present in the TRPL traces recorded from nanorods.This difference of temporal evaluation is a result of the reduced exaction diffusion in nanorods.

Experimental procedures
The LED samples used in this study are grown on sapphire substrates by metal-organic chemical vapor deposition.The InGaN/GaN MQW LED structures consist of a GaN buffer, an n-type GaN layer, five pairs of 2.5 nm-thick In 0.1 Ga 0.9 N quantum wells sandwiched between 13 nm GaN barriers, a p-type of GaN layer, as well as a p-type capping layer.The InGaN/GaN NRs with average diameters of 200 nm are fabricated by inductively-coupledplasma etching using a novel mask of self-assemble ITO-based nanodots [25].Scanning electron image of the nanorods and transmission electron image of a single nanorod are shown in Fig. 1.Frequency doubled optical pulses generated from a Ti: Sapphire regenerative amplifier (Libra, Coherent, 1 kHz, 90 fs) are used for pumping.The pumping wavelength at 400 nm only excites the InGaN well layers from backside with excitation flux in the range of 7-117 µJ/cm 2 (corresponding to carrier density of 10 16 -10 18 cm −3 ).Due to different scattering or reflection in the MQWs and NRs, the equal excitation density does not mean necessarily the same carrier density in the two samples.Time-integrated emission spectra are measured with a fiber spectrometer (USB 2000 + , Ocean-Optics).The technique of optical Kerr gating is utilized to do TRPL studies with Kerr medium of 5 mm thick carbon disulfide cell.The temporal resolution is about 2 ps, which is sufficient to study the dynamic feature of reduced exciton diffusion in the first 200 ps temporal window.A PMT (5784-20, Hamamatsu) was used to record the TRPL traces with band pass optical filters at 450 nm with 40 nm bandwidth or with wavelength selection by a monochrometer.The slow component of emission decay is checked with a fast oscilloscope with a fast PMT with nanosecond resolution.The emission spectra at different decay time are recorded with a 0.5 m spectrograph (Sp 2500i, Princeton Instruments) equipped a liquid nitrogen cooled charge coupled devices.A liquid helium cryostat (MicroCryostatHe, Oxford) with a transparent sample holder is used to do temperature-dependent measurements from 5 K to 300 K.

Results and discussion
TIPL spectra in nanorods show some common emission features of InGaN/GaN heterostructures (Fig. 2).With increasing excitation density, the emission shifts to shorter wavelength band in the spectra recorded at room temperature [Fig.2(a)].This blue-shift has been explained by the combination of band-tail filling effect and reversal QCSE with high density of photo-excited carriers [18,[26][27][28][29].With decreasing temperature, PL intensity increases dramatically with slight blue-shift of the emission band [Fig.2(b)].Compared with the parent MQWs, emission wavelength is shorter [Fig.2(c)] and emission intensity is much higher in NRs [Fig.2(f)] under same experimental configurations (i.e., the same excitation power and temperature).Temperature-dependence of emission properties show diverges in two samples as compared in Figs.2(c) and 2(d).With increasing temperature, the emission peak shifts to the red side monotonically in NRs [Fig.2(c)].In MQWs, the emission peak shifts to the blue side with temperature above 150 K, but further increasing temperature leads to a red-shift of emission peak [Fig.2(c)].While the emission intensity in nanorods decreases monotonically with increasing temperature, the emission intensity in MQWs remains nearly unchanged with temperature up to 120 K [Fig.2(d)].To understand the detailed mechanisms, we probe the carrier dynamic behaviors with TRPL spectra.Figure 3 displays the TRPL results recorded from the InGaN nanorods at room temperature.We first check the difference of TRPL traces recorded with the band pass optical filter and wavelength selected at the emission peak with a monochrometer [inset, Fig. 3(a)].The difference between the two curves is small, so that the curve recorded with the band pass filter is a good average of the dynamics at the emission peak in the sample.Compared to the wavelength selection measurement, the signal to noise ratio of the curves recorded with the (C) 2012 OSA optical filter is much better.In the following sections, we'll present these results as a good approximation to illustrate the carrier dynamics.
The excitation-power dependent TRPL traces are present in Fig. 3(a).Due to the limited length of the optical delay line, the maximum temporal window of TRPL traces recorded by the method of optical Kerr gating is about 3.5 ns in our study.To check the emission evolution in a longer dynamic range, a TRPL trace recorded with a fast PMT is shown in Fig. 3(b).The TRPL traces recorded from the InGaN nanorods show highly non-exponential decay, especially, in the early stage [Fig.3(c)].The recombination rate is strongly dependent on the excitation densities.To interpret the nonexponential decay dynamics, several models have been proposed in the last decade including a bi-dimensional donor-acceptor like recombination model [30], a charge-separated dark state model [16], and a frequently-used phenomenological stretched exponential (SE) function [31].For simplicity, we will use the SE function to describe the carrier dynamics in this work.The SE function in the form of 0 ( / ) 0 ( ) = is a good mathematic description of the TRPL traces.In the function, τ 0 is the SE decay lifetime, and β expresses the distribution of rates with values between 0 (broad distribution) and 1 (narrow distribution).
Figures 3(d) and 3(e) compare the best fitted parameters of the TRPL traces in the NRs and MQWs at room temperature with different excitation densities.The carrier recombination rate is generally faster in nanorods than in MQWs [Figs.3(d), 3(f) and 3(g)].As discussed in previous reports [2,32], the surface trapping effect may be a reason for the faster recombination since the etching process can cause surface damages in nanorods.If we checked the internal efficiency as the ratio of emission intensity at room temperature to that at 5 K approximately, the internal efficiency in the nanorods is lower than that in the parent QWs, which might be another signature of increased surface recombination.Also, other effects induced during nanofabrication can change the dynamical behaviors of the PL emission in nanorods.The QCSE associated to the strain relaxation after nano-texturing may also lead to the faster recombination in nanorods.The QCSE, induced by the strain-induced internal electrical field, causes the spatial separation of confined electrons and holes in InGaN/GaN heterostructures and slows down the carrier recombination [18,29,33].In the nanorods, the strain may be partially relaxed [1,2,8,9] and less internal electrical field will be present, leading to the faster relaxation in nanorods [2], which is consistent with the powerdependent TIPL measurement as shown in Fig. 2(e).With increasing excitation density, blueshift of the emission peak, which is frequently considered as a signature of reversal QCSE (free carrier screening), is observed in both MQWs and nanorods.However, smaller blueshifts were observed in the nanorod samples.These results indicate that the strain-induced QCSE still exists in the nanorods, in agreement with a recent study [34], but the strain may be weaker in comparison to the parent MQWs.
As shown in the Figs.3(d) and 3(e), the carrier recombination dynamics shows strongly dependence on the excitation density.The recombination lifetime decreases rapidly with carrier density below a critical value but the change becomes smaller and converges to a value at high carrier density in both samples.The reversal QCSE and the band-tail filling effect are frequently assigned to the power-dependent variations of recombination rate.With increasing excitation power, the internal electrical field may be screened and compensated by photon-excited carriers, leading to the reversal QCSE with the observed decrease of carrier recombination lifetime [18,29].This effect also exhibits the time-dependent descreening phenomena with red-shift of PL spectra over the decay time [33].Initially, the high density carriers induced by the intense laser pulses compensate the internal field efficiently.The carrier density decreases over the delay time so that the QCSE becomes more important and the recombination slows down [18,29,33].In Figs.3(f) and 3(g), we compare the counterplotted image of the intensity as function of delay time and emission wavelength in the nanorods and MQWs, respectively, with excitation density of 117 . Less red-shift of the emission over time can be observed in nanorods, indicating less strain in nanofabricated samples.Band tail filling effect may also cause the temperature-dependent variation of recombination rate [11,24,26].In general, this effect should not cause the discrimination of carrier dynamics in nanorods and the parent MQWs, since the band structure change is neglectable in the nanorods with diameters much larger than the effective Bohr radius (~3 nm) [27].Nevertheless, considering the confinement of carrier transport, the filling of lower energy levels can lead to the different recombination rates between the nanorods and MQWs.In InGaN MQWs, it has been observed that excitation transfers from the weakly-localized states to the strongly-localized state with increasing temperature [11,24].This thermallyactivated carrier transport drives the excitons filling the low energy levels prior to the interband transition.The diameters of nanorods are close or smaller than the carrier diffusion length [21][22][23], so that the carrier transport will be partially eliminated in nanorods.The influence of carrier transport confinement on the emission properties in nanorods is evidenced by the transient spectral results.With relatively low carrier density, the temperature-dependent TRPL spectra recorded from the MQWs and nanorods are compared in Figs.4(a) and 4(b).Instead of immediate decay post-excitation, slight rise of the emission over time can be distinct in the TRPL curves recorded from MQW samples.This delayed-rise becomes more explicit with increasing the temperature up to 160 K.It is observable till the temperature goes up to 250 K, above which the non-radiative decay process dominates.The delayed rise originates from intra-well carrier transport from weakly-localized states (higher potential minima) to strongly-localized states (lower potential minima) [11,24].This temperature dependence indicates the carrier transport is activated thermally.The carrier transport process, in a time scale of sub-nanosecond [21][22][23], enables that photon-excited electrons and holes to recombine from the strongly-localized states in MQWs.In nanorods, the size effect has profound impacts on carrier transport process and the temporal evolution behavior of the emission in nanorods diverges from that in MWQs.As shown in Fig. 4(b), with increase of the temperature, the delayed-rise is not observable in the TRPL traces in nanorods.This difference is a clear signature that the thermally-activated carrier transport process [11,24] has been strongly modified and partially eliminated by the nanofabrication processes.
In principle, the delayed rise appears when the population accumulation at the emitting levels caused by the carrier transport can overcome the removal process including radiative and non-radiative recombination processes [11,24,32].At the low temperature (5 K), thermal energy is insufficient for carriers to conquer the weak potential barriers.When the temperature increases, thermal energy will assist the carriers to "escape" from the weaklylocalized states and transport to the strongly-localized states, which causes the delayed-rise in the TRPL curves in the MQWs.To understand the absence of the delayed rise in nanorods, we will analyze the mechanisms on both the carrier accumulation and removal processes.
For carrier transport in InGaN samples, the exciton diffusion length is a critical scale.The diffusion lengths of 60 nm [22], 100 nm [20], and 200-500 nm [21] have been evaluated in the previous studies with techniques of polarized photoluminescence and spatially-resolved catholuminescence.These dimensions are in a same scale of the average diameter of the nanorods.Assuming a diffusion length of ~100 nm and a uniform distribution of the carrier density, the distance from localization to the nearest boundary will be shorter than the diffusion length.In consequence, the carrier diffusion may be significantly reduced in the nanorods.The carriers generated in the nanorods cannot diffuse to the strongly localized states if they exist originally outside the boundary.The population accumulation at the emitting levels in the parent MQWs originally contributed by the states outside the boundary is eliminated in the nanorods.Thus, the residual carrier transport in the nanorods may be not sufficient to support the carrier accumulation, which could be one reason of the absence of delayed rise in the TRPL traces.
Nevertheless, if the reduced exciton diffusion is the only origin, the distribution of emission centers in the MQWs should be narrower than that in the nanorods.However, the distribution factor [Fig. 2(e)] doesn't show significant difference between the two samples, indicating that other factors may be also involved in the absence of delayed rise in the nanorods.The enhanced removal processes either radiatively or non-radiatively may be the reason.As we discussed above, the surface state recombination and the reduced QCSE will lead to faster recombination in the nanorods.These processes may also cause the absence of delayed rise in the nanorods.
Hence, multiple mechanisms may be involved in the different emission dynamics observed in nanorods and MQWs with faster PL recombination and different temperaturedependent temporal evolutions in nanorods.Many effects, including the surface damages, the partial relaxation of strain and the band filling effect, could contribute to the faster recombination in nanorods.Besides these enhanced carrier removal processes, the discrimination of temperature dependence, characterized by the delayed-rise in the TRPL traces, should be also attributed to the reduced exciton diffusion.With higher power excitation, free carriers dominate in both MQWs and nanorods, so that the delayed rise process is absent in TRPL traces recorded from both samples.
With the above discussions on carrier dynamics, we can get a better understanding of the mechanisms on TI emission properties in nanorods.The nanofabrication influences the carrier recombination processes and causes the faster carrier recombination with a shorter emission wavelength in the nanorods [2,11,24,26], resulting in different peak shifts of PL emission in the nanorods and MQWs with increasing excitation density or decreasing temperature.The blue shift of the emission peak with excitation power [Fig.2(e)] has been regarded as a result of many effects including the QCSE and band filling effect.In general, these effects co-exist in InGaN samples.The slope of excitation-density-dependence may indicate different mechanisms.In nanorods, the rapid shift at the relatively low power may be attributed to cocontribution of the QCSE and band filling effect while the slow shift at the high power may be a signature of the full compensation of QCSE.Similar results have been observed in the previous literatures [18].With the measured data, one can estimate the internal electrical field by solving the Poisson equation and the Schrödinger equation ( [18] and ref. in) or more accurately with numerical simulation [2].These detailed numerical tasks are beyond the scope of this study.Roughly, we evaluate the internal electrical field as the saturated compensation field induced by photon-induced carriers.The magnitude of electrical field with accuracy in order is estimated to be ~200 kV/cm in the nanorods and more than ~400 kV/cm in the parent samples, which is lower than the samples with thick wells.With an increasing temperature, the thermally-activated carrier transport enables most of carriers to recombine from the strongly-localized states in MQWs.This carrier transport may compensate the increasing non-radiative recombination with temperature up to 120 K [Fig.2(d)].In nanorods, the carrier transport process is partially eliminated and the emission intensity drops monotonically with increasing temperature.This causes a stronger temperature dependence of PL intensity in nanorods.The modification of the carrier transport in nanorods could cause relatively low quantum efficiency in nanorods at room temperature [Fig.2(d)].Nevertheless, the light extraction efficiency is significantly improved [1,2,10].Higher power of PL emission is observed in nanorods [Fig.2(f)], even though the number of photo-excited carriers is much less since the filling factor is about 50% in nanorods, which suggests the nanorod structure can be an alternative approach for roughening the LED devices.Since the total output power is the ultimate goal of LED design, optimizing NR diameters is of promise to improve the device performance.

Conclusions
In conclusion, we have performed a systematic study on emission properties of InGaN/GaN nanorod LED structures by TRPL and TIPL experiments at different power densities and temperatures.We observe different temperature behaviors of steady and transient emission properties in nanorods and MQWs.Besides the contributions of surface states and strain relaxation effect, the reduced exciton diffusion may also play an important role to these differences.The impacts of nanofabrication on carrier dynamics revealed in this study are key factors for the nano-LED design.Device performance can be improved with a tradeoff between internal quantum efficiency and light extraction efficiency in nanorods.

Fig. 1 .
Fig. 1.Scanning electron microscopy image of the morphologies of the InGaN/GaN nanorods.Inset shows the transmission electron microscopy image of the cross section of a single nanorod.

Fig. 2 .
Fig. 2. Emission properties of InGaN/GaN nanorods.(a) Room-temperature emission spectra of InGaN/GaN nanorods recorded with different excitation densities; (b) Emission spectra of InGaN/GaN nanorods recorded at different temperature with excitation density of 25 µJ/cm 2 ;(c) Photon energy of Emission peaks from nanorods and MQWs is plotted as functions of temperature with excitation densities of 25 µJ/cm 2 and 117 µJ/cm 2 , respectively; (d) Intensities of emission from nanorods and MQWs are plotted as functions of temperature with excitation densities of 25 µJ/cm 2 and 117 µJ/cm 2 , respectively.(e) and (f) plot the photon energy of emission peak and the emission intensity recorded in the samples of nanorods and MQWs as function of excitation density at 5 K and 300 K, respectively.

Fig. 4 .
Fig. 4. Temperature-dependent TRPL traces recorded from MQWs (a) and nanorods (b) with excitation density of 25 µJ/cm 2 .The curves are vertically shifted for clarity.The dashed lines are horizontal references to clarify the delayed rise of TRPL traces recorded from the MQWs.