United States

Recently, stretchable strain sensor is considered as the promising technology for various application, such as wearable electronics, healthcare monitoring, and robotic skin/arms, due to its mechanical stretchability and multi-axial sensing capabilities. The recent research for stretchable strain sensors have made tremendous efforts to significantly improve key deterministic parameters, including sensitivity, durability, sensing range, and response time. Although various high-performance strain sensors have been demonstrated using a range of mechanistic device structures, most conventional strain sensors are only capable of detecting uniaxial strain and lack the ability to measure directional variable motions. In order to overcome this challenge, novel sensor devices with geometrically engineered architectures, composed of multidimensionally stacked channels or cross-shaped detecting electrodes have been developed. These architectures were developed to address the challenge of achieving multidirectional strain sensing, by employing isotropic conducting pathways. While these advancements are significant, the complicated and intermixed pathways in the channel region can cause unsatisfactory linearity and make it difficult to determine the strain direction under complex stimulation environments. In addition, it is worth noting that many recent studies have utilized machine learning techniques, such as Neural network to classify data obtained by strain sensor. However, in order for Neural network to be applied effectively in industry, optimization processes such as model lightweighting are necessary. Here, we propose a simply designed array of wearable strain sensor capable of detecting and interpreting the direction and intensity of strain through a neural network algorithm trained through cognitive learning. The wearable strain sensor arrays comprises a heterogeneous surface architecture consisting of a stretchable substrate made of polydimethylsiloxane (PDMS), strain-insensitive rigid electrode regions made of polyethylene terephthalate (PET) film, and a strain-sensing channel region composed of multi-walled carbon nanotube (MWCNT) with carbon black (CB) nanofiber dispersed in a PDMS matrix. We utilized a mechanism based on the change in the number of conducting pathways between MWCNT and CB depending on the direction of strain. The use of a heterogeneous device architecture can reduce the negative effects of electrode strain on electrical interruptions, resulting in more reliable strain sensing characteristics. This wearable strain sensor arrays exhibited high sensitivity with a gauge factor of approximately 8, a wide strain sensing range of 0 to 35%, and excellent reliability after 3,000 stretching cycles. To detect strain from any direction, Strain sensor array was made using three sensors arranged at 45-degree tilted angles. Furthermore, to achieve high-accuracy detection of both the intensity and direction of strain, a machine learning algorithm known as a multi-output neural network was employed. A neural network was constructed for strain sensing, which showed an accuracy of over 95% for both strain intensity and direction. In addition, optimization of the strain sensing system’s neural network was done by adjusting the number of sensors and input data dimension to achieve a lightweight model. By employing this method, not only data obtained from the wearable strain sensor arrays be efficiently handled, but also the strain intensity and direction can be accurately classified.

EMC 2023

Carbon Based Omnidirectional Wearable Strain Sensor Arrays with Optimized Multi-Output Neural Networks

strain sensor

carbon based

neural network

technical paper

The Electronic Materials Conference (EMC) is the premier annual forum on the preparation, characterization and use of electronic materials. This year’s event will feature a plenary session, topical sessions, a poster session and more.

EMC is coordinated with the Device Research Conference (DRC). This recognizes the strong interaction between device and electronic materials research and provides fruitful exchange of information between attendees of both Conferences.

The Electronic Materials Conference (EMC) is the premier annual forum on the preparation, characterization and use of electronic materials. 

Consisting of nearly 80 billion neurons and hundreds of trillions of synapses, the human brain is one of the most complex information-processing systems in the known universe. To understand how the brain functions, it is crucial to be able to measure the spatiotemporal dynamics of the generation and propagation of signals between each of these neurons. Measuring these signals can provide insight into normal function as well as shed light into the signatures or root causes of diseases such as epilepsy or Parkinson’s disease. However, measuring the electrical activity of neurons at such a large scale has proven to be an extremely challenging endeavor. Electrocorticography is one method that is commonly used in the clinic to measure aggregate neural activity. Cutting-edge electrocorticography systems may employ thin-film fabrication of flexible microelectrode arrays with dense wiring schemes to achieve higher channel-count systems. These microelectrode arrays are typically fabricated using polymers such as parylene-C and the microelectrodes are routed individually and conductively bonded to rigid silicon-based acquisition circuits for digitization and downstream signal processing. Recent efforts have demonstrated high-density microelectrode arrays which can achieve thousands of channels; however, it is a major challenge to form reliable interconnects over one thousand channels between the flexible microelectrodes and the rigid acquisition electronics. Taking inspiration from the high-definition flexible display industry, we have developed a 16x16 array of actively multiplexed microelectrodes using dual-gate thin film transistors made of indium gallium zinc oxide and fabricated on flexible polyimide substrates at low temperature. These neural interfaces use on-site multiplexing to help drastically reduce the interconnect bottleneck and thus improve the scalability of the system towards higher channel counts. Furthermore, this marriage of flexible substrates with active electronics is accomplished in a monolithic fashion, allowing for superior performance in a soft and wet biofluid environment. A major challenge with achieving reliability in such a system is the hermeticity of the encapsulation. Active electronics require biasing and control signals, thus any failure in encapsulation, if not properly designed for, can cause charge injection directly into cortical tissue. Thus, we have incorporated conformal atomic-layer-deposited ceramic biofluid barriers in addition to polymer encapsulation layers to help improve the operation of our active electronics in biofluid mediums. We are currently evaluating this encapsulation strategy in accelerated aging soak tests in elevated saline chambers, and have observed functional TFT microelectrode arrays after several weeks of soaking in saline. We expect that this strategy should be compatible with chronic experimental lifetimes of several months and plan to perform animal testing in the rat model after completing benchtop characterization. Another crucial aspect of this work is ensuring that the TFTs are robust over long periods of time. We have found that the contact resistance, mobility, and threshold voltage are all very sensitive to annealing and surface preparation conditions. By incorporating fluorine-based plasma treatment to the IGZO contact areas, and by optimizing the metal deposition conditions, we have found that we can lock in dopants in the IGZO film such that after curing the encapsulating polyimide films in nitrogen ambient, the annealing of these films results in stable and time-invariant electrical properties.

Optimizing Dual-Gate IGZO TFTs for Long-Term Flexible Neural Interfaces

1. Introduction By integrating the 32 graphene FET with antibody modified channel, the new corona virus was detected with 100~1000 times higher sensitivity than the conventional antigen detection kits. The detection time is ~10~20 minutes, which is ~10 times shorter than the PCR system. By using the graphene FET with and without the antibody modification for detecting the signal of virus and for the reference signal, the problems of the signal drift and the physical adsorption of virus were solved. Also by adapting the micro-TAS and automated washing system, the restriction of the Debye length was solved and the impurity such as a saliva could be washed away and the pure signal only from the virus which bind to antibody could be detected. 2. Experimental Process The channel length and width of the fabricated graphene FETs are 10mm and 100mm, respectively, and 32 graphene FET’s are integrated. The 13FET’s in the half of left hand side are modified by antibody through PBASE, and the 13FET’s in the half of right hand side are without antibody modification and used as reference FET. Onto this integrated graphene FET array, the micro total analysis system (m-TAS) made by PDMS was set to introduce the phosphate-buffered saline (PBS) solution from the right hand side inlet and through the graphene FET channel, and to drain from the left hand side outlet. PBS solution was introduced from the reservoir by the computer controlled micropump. In order to control the Debye length, the different concentration of PBS solution should be used for the introduction of virus and for the detection of the charge of virus. 3. Results & Discussions The time dependence of the average of the Dirac point of 13 graphene FET’s with and without the antibody modification was measured using the 0.01xPBS and its Debye length is 7nm. There is the difference I n the Dirac point between with and without the antibody modification, which is attributed by the positive charge of the antibody modified on the graphene surface. 2 time of 1xPBS was introduced into the system without virus to confirm the no influence by the exchange of the high concentration of PBS. Then, the new corona virus with 1x108FFU/mL in high concentration of 1xPBS solution was introduced to the graphene FET. Virus combines well to the antibody in the high concentration of 1xPBS solution. The Debye length at 1xPBS solution is as short as 0.7nm, and the charge of virus could not be detected by the graphene FET as the height of the antibody is as large as ~10nm. Again, the low concentration of 0.01xPBS was introduced to extend the Debye length. Then, the negative charge of the virus was detected and the Dirac point increased positively. The detail of this Dirac point shift was analyzed. Dirac point of the 13FETs with antibody modification shift as large as ~20mV , while that without antibody as small as ~3.5mV after the introduction of virus. The large shift of Dirac point with antibody FET is attributed to three reasons as (1)charge of virus which combine to antibody, (2)charge of the physical adsorbed virus on the graphene, and (3)Drift characteristic of the total measurement system. While the small shift of Dirac point without antibody FET is attributed to two reasons as (2) and (3). The difference of these value, i.e., ~16.5mV(20mV – 3.5mV) is due to the net charge of virus which combined to only antibody. Using the SEM, the new corona virus on the graphene FET channel could be counted. About 10 times larger number of virus binds to graphene surface with antibody modification than without antibody modification. The number of the virus on the channel is linearly proportional to the Dirac point shift. 3. Conclusions Using the integrated graphene FET array and automated washing systems, the new corona virus was detected with the 100~1000times higher sensitivity than the conventional antigen test.

Integrated Graphene FET Array for High Sensitive Detection of New Corona Virus with Automated Solution Exchange System

The IV-VI semiconductor family of PbSe-SnSe alloys spans narrow direct bandgap materials in the Pb-rich cubic rocksalt phase, and indirect bandgap semiconductors in the Sn-rich layered orthorhombic phase. In recent years, there is growing interest in harnessing electronic and photonic properties of these orthorhombic phases that merit their study as epitaxial thin films. Specifically, there is excitement in harnessing in-plane ferroelectricity in SnSe demonstrated in the single or odd number bilayer limit [1]. Additionally, research groups are showing a large contrast in electronic and optical properties for phase-change devices across the orthorhombic to rocksalt structural transition that can not only be driven by temperature but also most by direct application of intense light-fields [2,3]. The development of growth recipes to fabricate high quality thin films of not only SnSe but also Pb1-xSnxSe offers new compositional routes to tune these properties. In this work, we synthesize Pb1-xSnxSe epitaxial films in the orthorhombic phase across a large composition range by molecular beam epitaxy, on technologically relevant III-V GaAs substrates. Importantly, we find it possible to stabilize metastable single phase compositions beyond what is expected in the bulk. A Riber Compact 21 MBE system with compound PbSe and SnSe cells is used to grow films. (001)-oriented semi-insulating GaAs substrates are prepared by desorbing the surface oxide under a Se overpressure, followed by PbSe surface preparation steps. We find that smooth, out-of-plane single oriented SnSe films result both from samples grown with a PbSe dose, and samples grown with a PbSe dose and buffer layer. We expand on the successful growth of SnSe on GaAs and use the identified surface preparation methods to grow orthorhombic Pnma space group Pb1-xSnxSe alloys on GaAs [4]. The Pb1-xSnxSe alloy composition is determined by the relative fluxes of the PbSe and SnSe compound cells. XRD results are used to extract orthorhombic lattice constants as a function of alloy composition, at room temperature. The difference in unit cell volume between alloys around XSn=0.5 and the predicted unit cell volume of similar composition Pb1-xSnxSe alloys in the cubic phase is small, implying a small energy cost to a phase transition at these intermediate compositions. In addition to structural properties, we will be reporting on the optical bandgap as a function of alloy composition. Based on the bulk phase diagram of the PbSe-SnSe system, we expect alloy compositions from XSn=0.4-0.8 to be two phase; both cubic and orthorhombic. The XSn=0.64 sample is phase separated. However, we have managed to grow thin film samples at XSn=0.55 and XSn=0.74 which are purely orthorhombic. Moreover, these are not the only metastable sample that we have produced in this materials system. We have synthesized room-temperature stable orthorhombic Cmcm phase of Pb1-xSnxSe , a phase typically only stable above 700 K. We will be exploring the structural properties of these samples further, including temperature cycling experiments to better understand the stability of these phases in the thin film limit. [1] C. Kai, et al. …S.S. Parkin, Nano letters, 20, 6590,(2020). [2] T. Katase et al. ... T. Kamiya, Science advances, 7(12), (2021). [3] Y. Huang, ... M. Trigo, Physical Review X, 12(1), (2022). [4]B. Haidet et al. ... K. Mukherjee, Phys. Rev. Mater., 4, 033402 (2020).

Single-Phase Orthorhombic SnSe-PbSe Alloy Thin Films Stabilized on GaAs by Molecular Beam Epitaxy

III-V materials nowadays are becoming a promising building block for nonlinear optics thanks to their significant nonlinear properties. MBE and MOCVD have been widely used to grow epitaxial films on various substrates. However, the difficulty in lattice matching, the constraint in substrate selection, the issue in epitaxial transfer aligning, and the cost in fabricating, become the barrier to applications. Here, we introduce a new thin-film epitaxial growth method, templated liquid phase (TLP) growth, to directly grow III-V semiconductor film on any substrate with single crystallinity, high uniformity, and high carrier mobility. As a demonstration, we choose lithium niobate (LiNbO3), a promising material widely used in nonlinear optics and acoustoelectric fields, as the substrate to grow patterned InAs layer. With this technology, we grow InAs in line-shape patterns with widths varying from 3 micron to 10 micron and lengths varying from 10 micron to 100 micron. Based on our measurements, the electron mobility reaches ~3500 cm2/Vs at 100K and ~2400 cm2/Vs at 300K. In prospect, TLP growth method can construct a platform to fabricate any desired patterns on any substrates in applications of nonlinear optics profit from the superior nonlinear properties of III-V semiconductors.

Monolithic III-V on LiNbO3 for Nonlinear Optics Application

Over the last few years, tremendous improvements of device performance and reliability have been demonstrated in devices grown on blanket Si substrates. Recently, electrically pumped QD lasers deposited in narrow oxide pockets in a butt-coupled configuration and on-chip coupling have been realized on patterned Si photonic wafers. However, the device yield and reliability, which ultimately determines the scalability of such technology, are limited by the material uniformity. Here, we perform the detailed analysis on the material anisotropy induced by the pocket geometry. Both the QD morphology and defect configuration anisotropy suggest that all pockets should be aligned to the [1 -1 0] direction of the III-V crystal. The 300 mm growth template consists of rectangular oxide pockets both “parallel” and “orthogonal” with respect to the Si substrate step edges, shown in Figure 1(a) and (b). The room temperature photoluminescent (PL) spectra of the as-grown in-pocket QD laser material, shown in Figure 1(d), indicate that QD quality is higher in the “orthogonal” and wider pockets. The corresponding laser performance, shown in Figure 1(e), suggesting the gain from the QDs in the narrower “parallel” pockets is not high enough for lasing. Representative PL intensity maps of a “half-laser” structure and the corresponding atomic force microscopy (AFM) image of the exposed active region dots are shown in Figure 1(f). The drastic discrepancy of the emission uniformity and QD density between the materials in the wider “orthogonal” pocket and the narrower “parallel” pockets are attributed to asymmetric indium adatom diffusion, with [1 -1 0] being the fast diffusion direction. To map the crystal orientation onto the pocket orientation, faceted trenches (FTs), which would form when a GaAs layer experience abrupt tensile relaxation, were introduced in the pocket with the test structure shown in Figure 2(a). Nomarski images show that FTs elongate along the “orthogonal” pocket, confirming that “orthogonal” pockets are in the [] orientation with respect to the III-V crystal, shown in Figure 2(b) and (c). Thus, aligning the pocket to the fast diffusion orientation would prevent indium adatoms from diffusing onto the edges. Since the edge of the III-V strip is stress free, it is also expected that the residual thermal stress state of the in-pocket materials would also be highly anisotropic with the direction along the short edge of the strip experiencing lower residual tension. It has previously been confirmed that the misfit dislocations (MDs) generated from the residual thermal stress is responsible for device degradation. Such stress anisotropy introduced by the pock geometry could then be utilized to reduce the MD density. Two TEM foils were prepared parallel (“‖”) and perpendicular (“⊥”) to a 20 µm [1 -1 0] oriented pocket, shown in Figure 2(d). Comparing to Figure 2(e) where foil oriented parallel to the pocket, the foil perpendicular to the pocket, shown in Figure 2(f), has a significantly lower density of misfit dislocations, even much lower than those measured in blanket substrate materials. It has previously been found that the MDs in the [1 -1 0] direction formed via tensile relaxation have a Ga-core structure, which resides closer to the middle of the GaAs/AlGaAs bandgap than the MDs in the [1 1 0] direction Though the MDs along the long edge of the pocket would be reduced regardless of the alignment of the pocket with respect to the III-V crystal, all pockets should be aligned to the [1 -1 0] direction to effectively lower the density of the most detrimental defects. The in-pocket devices would then potentially surpass the performance of the state-of-the-art devices grown on blanket Si substrate.

Quantum Dot Morphology and Defect Configuration Anisotropy for III-V Laser Material Grown on Patterned Si Photonic Wafers

Recent research efforts have unlocked substantial improvements in the performance and reliability of InAs quantum dot lasers on silicon by mitigating the effects of active-region-adjacent misfit dislocations through the introduction of defect trapping layers (TLs). These TLs enable lifetimes of lasers on silicon comparable with those grown on native GaAs substrates. Even so, current state-of-the-art lifetimes still fall somewhat short of requirements for more widespread adoption as both types of lasers exhibit gradual degradation in performance with aging. In this work, we demonstrate that this gradual degradation is a result of the generation of point defect populations during aging, which, over time, coalesce into dislocation microloops which we characterize using scanning transmission electron microscopy (STEM). We selected a single high-performing laser from each of three large aging sets of QD lasers (1) on silicon with TLs, (2) on silicon without TLs (no-TL)), and (3) on native GaAs. Each selected device was aged for between 8700 and 12,000 h at 80 °C at a current density between 1050 and 1900 A cm-2, depending on the initial threshold current and output power. After 2700 h, the no-TL device had doubled its initial threshold current (a common metric for device failure), and after 5700 h it failed to lase entirely. The TL device and the GaAs device, however, show threshold current increases of just 35% and 46% respectively after 8700 h with extrapolated lifetimes of 45,000 h and 75,000 h. Note, however, that these numbers are not directly comparable due to somewhat harsher aging conditions for the GaAs device. After aging, the devices were imaged using backside electroluminescence to check for signs of inhomogeneous degradation. For all lasers, degradation appeared largely homogeneous aside from dark line defects associated with misfit dislocations in the devices on silicon. There are also regions of decreased luminescence throughout the structure for the two devices on silicon, which we attribute to misfit dislocations lying parallel to the laser ridge for the no-TL device and to misfit dislocations escaped from the TL for the TL device, as confirmed by plan-view STEM of these regions. With sufficient thinning of the plan-view foils, dislocation microloops with diameters ranging from 5 to 20 nm are visible both in the [001] zone imaging condition overlaid with the QD contrast and more clearly in various g-3g STEM weak beam imaging modes where competing QD contrast is minimized. In the no-TL device in regions within 100-150 nm of the heavily climbed misfit dislocations, there is a dramatic increase in loop density on the order of 1010 cm-2. We believe this to be the result of a large density of point defects expelled by the misfit dislocations as they undergo climb. Thus, degradation caused by climbing misfit dislocations is multifold: first an increase in total misfit length and second the generation of a high density of point defects which coalesce into microloop defects surrounding the misfit dislocation, all of which substantially enhance non-radiative recombination processes. Meanwhile, the TL and GaAs device active regions, though largely or entirely free of climbed misfit dislocations, are still observed to contain microloop densities on the order of 108–109 cm-2, which likely accounts for the gradual, homogeneous degradation observed in these devices. These microloops coalesce from elevated point defect populations in the active region, and while there are many remaining questions, such as whether these point defects are pre-existing or are generated during aging, this work points the way for future routes to improve device and material quality.

Gradual Degradation via Dislocation Microloop Formation in InAs Quantum Dot Lasers on Si and GaAs

InGaAs quantum wells (QW) grown by MOVPE are used to extend the absorption edge of GaAs photovoltaics into the infrared, increasing the current density with minimal loss in voltage as compared to a bulk GaAs device. GaAsP strain compensation is used to provide a tensile offset to the compressive strain in the wells, allowing growth of 50 pairs of strain balanced wells which did not exhibit detrimental effects of lattice relaxation/defect formation on photovoltaic figures of merit. Significant work was performed to optimize the growth and interfaces of these films to prevent step bunching and poor surface morphology and device performance when grown on offcut substrates, which are necessary for e.g. growth on non-polar germanium for a bottom junction or atomic ordering of InGaP for a top junction. A distributed Bragg reflector (DBR) is incorporated below the device, effectively increasing the optical path through the QWs and nearly doubling their sub-GaAs-bandgap collection per well. Dual-junction InGaP/GaAs devices were designed using a drift-diffusion model incorporating the optical effects of each subcell, connecting tunnel junction, and anti-reflective coating (ARC) along with the expected contribution of a varying number of QWs. These parameters were co-optimized to design control and QW-enhanced designs which were each nominally current-matched under AM0. Dual-junction devices were grown with an InGaP top junction coupled to either a control GaAs bottom cell without quantum wells or to the previously developed 50-QW enhanced GaAs bottom cell. Once fabricated, the QW-enhanced device provided an extra 2.6 mA/cm2 from sub-gap absorption with a minimal loss in voltage (11 mV) compared to a control device without QWs or DBR, providing an absolute efficiency increase of 3.6% under AM0.

InGaAs/GaAsP Multiple Quantum Well Enhanced Multijunction Photovoltaics

Traditional lock-in amplifiers employ a simple IIR filter which can have long measurement times to produce acceptable noise reduction. Using AC field Hall Effect measurements, we demonstrate that a combination of digital lock-in output stage filters, FIR and IIR low pass filters, can produce improved noise reduction and measurement time over the traditional approach of independent IIR filters or FIR filters. We will present step response, equivalent noise bandwidth and autocorrelation times for a n pole IIR filter (n <=4) followed by a filter response of any finite length. We will demonstrate that the performance of the hybrid output filter, with time measured in units of FIR time, depend only on the ratio of the IIR time constant to the FIR time. Finally, we will present criteria for statistically independent samples for the various filter configurations. AC field Hall Effect characterization is carried out with very low AC field frequency, typically 100 mHz, to avoid inductive pick-up in the measurement. Measurement times for a single voltage reading are as long as 60 seconds. The standard protocol for the Hall effect measurement includes current reversal and two different geometries which increases the number of voltage readings to 6 and results in a 6-minute total observation time to measure a single Hall voltage. Multiple Hall voltage readings are needed to calculate the of standard error of the Hall voltage and increases the measurement time in excess of 1 hour. An effective hybrid filter approach for AC field Hall effect measurement promises shorter measurement times or lower noise in the same observation window. Because it is a well understood material, an InAs sample was chosen as a model system for performance testing of the hybrid output filter in an AC field Hall effect measurement. Using currents of 10 microamps, the Hall voltage will be on the order of a few microvolts for this InAs sample depending on the field magnitude. In this work, an InAs sample with 6.52e16 cm-3 carrier density was first measured in a Lake Shore Cryotronics 8407 using the AC field option of the 8400 system. Using a 10 microamp excitation current and 0.644 T AC field amplitude, the magnitude of the measured Hall voltage was 2.44 microvolts with a 119 nanovolt standard deviation. The measurement electronics were subsequently replaced with the Lake Shore Cryotronics Model M81 Synchronous Source Measure platform equipped with a balanced current source (BCS-10) and voltmeter (VM-10) and the Hall effect measurement repeated with the same magnet and power supply. The M81 lock-in was configured with a hybrid output filter consisting of an IIR filter with a 4 second time constant and 12 dB roll-off combined with a FIR filter with 40 second averaging time which results in a total settle time of the filter of 60 seconds. The measured Hall voltage was 2.47 microvolts with a standard deviation of the Hall voltage of 37 nanovolt – a factor 3.2 improvement in measurement uncertainty over the 8400 system.

Improved AC Field Hall Measurements Using Hybrid Filters

Blue and green InGaN/GaN light-emitting diodes (LEDs) are crucial for solid-state lighting and display applications. Violet–blue LEDs have achieved a high EQE of ~80%. However, InGaN/GaN LEDs still suffer from strong efficiency loss in the green region, which is usually referred to as the “green gap.” The green gap issue has been related to Auger recombination, the quantum-confined Stark effect (QCSE), carrier leakage, random alloy fluctuations, and defects and carrier localization, among other causes. To achieve longer emission wavelengths, the indium composition must be increased, leading to several problems. First, high indium content increases lattice mismatch and introduces additional point defects during metalorganic chemical vapor deposition (MOCVD). High indium content also leads to strong polarization-related electric fields in the quantum wells (QWs), lowering the wave function overlap and decreasing the radiative and non-radiative recombination rates (i.e., from the QCSE). An important question is how much of the change in efficiency reduction from blue to green is due to intrinsic effects (i.e., wave-function overlap and phase-space filling (PSF)) vs. extrinsic effects (i.e., material degradation). In this work, we use small-signal electroluminescence (SSEL) to study the carrier dynamics in commercial c-plane LEDs with various emission wavelengths ranging from blue (~470 nm), to cyan (~500 nm), to green (~530 nm). The LED series consists of three wafers grown by Lumileds with identical active region designs (3X, 3 nm QWs) but different indium compositions. LEDs with mesa diameters of 100 µm and ground-signal-ground electrodes were fabricated. SSEL was then used to study the carrier dynamics of the LEDs. Each element in an equivalent circuit model was acquired via simultaneous fitting of the impedance and modulation responses, leading to the injection efficiency and total carrier recombination lifetime. Combining the total carrier lifetime with internal quantum efficiency (IQE) enables the separation of the radiative and non-radiative recombination lifetimes and rates as a function of carrier density (n) and current density (J). The results reveal that efficiency loss at longer wavelength is related to reductions in injection efficiency and radiative recombination, and an increase in Auger recombination. The interplay between the QCSE and carrier-current density relationship is shown to determine the relative importance of the radiative (B(n)*n2) and non-radiative recombination rates (A(n) and C(n)*n3) as the indium content is increased. Both the SSEL experiment and Schrödinger-Poisson calculations showed C(n)/B(n) is similar for the three wafers at high n, indicating that intrinsic material degradation is not significant. In addition, compared with QCSE, PSF, and material degradation, the increase of the carrier density at a given J for longer wavelengths is the dominant reason for the green gap. Thus, methods to decrease the carrier density in each QW are suggested to increase the EQE of green LEDs.

Carrier Dynamics in Blue, Cyan and Green Commercial InGaN/GaN LEDs Measured by Small-Signal Electroluminescence

Nonradiative recombination centers (NRCs) are known to negatively affect the efficiency of light-emitting diodes (LEDs). NRCs contribute not only to Shockley–Read–Hall (SRH) recombination, but also to trap-assisted Auger recombination (TAAR). The relationship between NRCs and the TAAR processes has been predicted theoretically and observed experimentally to a limited degree. However, some details about TAAR are still under debate, including its magnitude relative to intrinsic Auger and SRH recombination, its dependence on carrier density (e.g., ∝n2 or ∝n3), and its role in the green gap, particularly in high-quality commercial-grade LEDs. TAAR is typically assumed to be linearly dependent on trap density, directly tying the magnitude of TAAR to the quality of the sample. Previous studies have examined TAAR using time-resolved photoluminescence and optical differential lifetime measurements, which do not capture certain effects related to electrical injection. In this work, we report a quantitative study of TAAR in c-plane commercial-grade green LEDs grown by metal-organic chemical vapor deposition (MOCVD). We investigate a series of three wafers with identical active regions (3X, 3-nm-thick quantum wells) but different NRC trap densities (4.4e14 cm-3 < Nt < 1.5e15 cm-3). We use small-signal electroluminescence (SSEL) and deep-level optical spectroscopy (DLOS) to study the carrier dynamics and density of NRCs in the samples, respectively. SSEL is used to determine the carrier lifetime, which is used in conjunction with the internal quantum efficiency (IQE) to decouple the radiative and nonradiative recombination rates. The total recombination lifetime in higher quality growth is found to be longer than that in lower quality growth and is attributed to the longer nonradiative recombination time in higher quality samples. On the other hand, the radiative recombination rates for different growth qualities at a given carrier density (n) are identical, as expected from samples with the same active region design. We extract the relationship between the density of NRCs and the corresponding nonradiative recombination rates at a given carrier density (n). The difference in nonradiative recombination between the samples offers information about both SRH and TAAR with respect to growth quality. The linear dependence between the SRH recombination rate and the density of NRCs at low n was confirmed. However, TAAR was found to be sublinear with the density of NRCs, possibly due to the relatively small changes (~3X) in defect density (Nt) in these samples. We find that TAAR is best fit to an n2 dependence on carrier density. This is supported by the observation that all three samples have similar IQE at high carrier density, which would not be consistent with an n3 process. We also show that the TAAR recombination coefficients (D) are on the order of ~1e13 cm3 s-1, which is similar to that reported for violet LEDs and shows the weak dependence of TAAR on bandgap. Using the extracted TAAR coefficients, we calculate that TAAR accounts for about 10% of the total nonradiative recombination rate at moderate carrier densities. Thus, intrinsic Auger dominates the nonradiative recombination and TAAR is not a major contributor to the green gap in these commercial LEDs. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.

Premium content

Carbon Based Omnidirectional Wearable Strain Sensor Arrays with Optimized Multi-Output Neural Networks

Downloads

Next from EMC 2023

Optimizing Dual-Gate IGZO TFTs for Long-Term Flexible Neural Interfaces

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES