Final Report Summary - GECCO (3D GaN for High Efficiency Solid State Lighting)
Executive Summary:
The GECCO consortium is aiming at developing a novel technology for solid state lighting namely 3D core-shell LEDs circumventing many of the limitations of planar LED technology. Due to intense collaboration between partners, substantial progress can be reported for all periods of the project. Some major improvements in 3D GaN technology have been achieved. Major insights into the growth mechanisms of GaN and the necessary processing towards white LEDs have been gained. This includes the fabrication of InGaN quantum wells on m-plane sidewalls with high internal quantum efficiency and high indium content. Special phosphors have been developed in order to fulfil the requirements of 3D blue emitting nanoLEDs processing into white LEDs. The GECCO consortium belongs to the leading groups worldwide working on the realization of 3D GaN technology.
The collaboration between partners has been very efficient and professional, supported by regular project meetings held in Braunschweig (twice), Madrid, Bristol, Brussels and Regensburg and close exchange of information by web-conferences guiding the scientific progress of GECCO. –This has been supported by maintaining a web based project management system which aligned the collaboration and monitoring inside the project. In addition, a risk management inside the R&D work packages was carried out on a regular basis, helping to avoid critical delays.
Project Context and Objectives:
The GECCO consortium is aiming at developing a novel technology for solid state lighting circumventing many of the limitations of planar LED technology by employing 3D core-shell LEDs. This summary describes the main achievements during the project which evolved in accordance to the work plan.
The basis of the project was the selective area growth of first the GaN columns and second the LED related shells. During the course of the project a fundamental understanding of the growth processes was compiled. The growth masks, which are needed for selective area growth, were optimized not only in view of filling homogeneity and reproducibility, but also in view of its insulating properties inside the processed LED-chip. The growth of InGaN shell layers and cores with a high amount of indium incorporation was addressed successfully..
In the second part of the project, GECCO moved from a more epitaxy related focus to device related research. Arrays of 3D LED structures were grown with further optimized processes on patterned templates - fabricated either by employing conventional photo lithography or nano imprint lithography - using selective area growth by MOVPE as well as by MBE. Strategies to combine phosphors with the 3D approach are also demonstrated.
The chip process was further developed and optimized so that complete LED devices can be manufactured from the 3D arrays including a suitable high refractive index (HRI) matrix material which was identified during the first period. High efficiency phosphors for different colors with grain sizes below 1 µm were developed to fit between the rods so that demonstrator device were processed on a chip level which allowed controlling the color and matching the Planckian locus at 7500 K. Additionally the chip process was improved regarding current spreading and functional layers (e.g. insulating layers).
A detailed characterization of the fabricated nanostructures was performed by using a strong combination of nanoanalytical methods, including electron microscopy in transmission and scanning mode (TEM, SEM) as well as cathodoluminescence (CL-SEM), electroluminescence via point contacts inside an SEM and temperature resolved photoluminescence (PL). Avoiding structural defects, which are usually accompanied by non-radiative recombination, as well as insufficient current spreading and non-optimized doping profiles have been identified to be a prerequisite for obtaining high efficacies.
The understanding and optimization of properties of the 3D LEDs was supported by simulations of the local growth conditions, in particular by the temperature distribution inside an array. The characterization of the 3D LEDs was supported by investigating the light extraction by means of ray tracing and by electro thermal simulations inside the LEDs, also including the influence of the packaging.
Even though the evaluated devices based on 3D LEDs did not yet - due to a shift or resources to other work packages - reach the targeted extremely competitive goal of 200 lm/W in electroluminescence, the first demonstration ever of a white, phosphor converted nanorod LED has been a big step forward.
Project Results:
Substantial insight into the growth mechanisms by both MOCVD and MBE could be gained. By optimizing the overall processing, the reproducible growth of arrays of good quality 3D InGaN/GaN nano- and microrods with high aspect ratios and a good yield on whole wafers has become possible. Also InGaN QWs emitting in the red could be realized with surprisingly high PL-IQE values:
3D growth
3D GaN submicron-columns and nanorods have been successfully grown on conductive templates using MOVPE and MBE, respectively. Using MOVPE, the diameter of GaN columns is about 600 nm and the height is under control up to tenths of µm. The structural quality of the 3D columns is very high, no dislocations or stacking faults could be detected by TEM. The doping concentrations in n-type core and the p-type shell, as well as the In concentration in MQW can be varied. Under the utilized growth conditions, up to 100% of the patterned holes can be filled by columns, resulting in a high coverage of about 85% of the patterned region on 2” and 4” wafers by a uniform array. The growth mechanism of GaN 3D columns has been investigated systematically. In particular, the influence of hydrogen, silicon doping, as well as V/III ratio during growth has extensively been studied.
InGaN cores with an Indium incorporation above 20% has been obtained following the optimization process on the MBE growth of InGaN/GaN nanorods.
For the first level of template preparation three different types of mask pattern were used for photolithography at TUBS. All of them were full-area masks with hexagonal structures arranged in the closest packaging. The diameter D varies from 0.8 µm to 2 µm and the pitch P is either 3 µm or 4 µm. Templates composed of a 2" silicon or sapphire substrate with a SiO2 masking layer were available. In addition a fourth 4" mask was available for the second level. The progress beyond the first level is also an optimized fabrication process so that amount of filled patterns could be strongly increased.
The third level mask design has been developed at OOS. A new 4" mask was designed and fabricated which has an optimized structure with respect to the desired high aspect ratio of the nano- and microrods, and to support the phosphor deposition experiments. Nano-imprint lithography (NIL) was chosen for the fabrication of the templates in order to provide the necessary throughput for serving the growth experiments; one master stamp was used for the template fabrication at OOS. The pattern (i.e. diameter of 600 nm and 800 nm and a pitch of 3 µm and 4 μm) was distributed across the 4" wafer in a way that accommodates the non-symmetric deposition pattern of the wafer due to the MOVPE reactor geometry. These optimized conditions based on the achievements of the project are necessary for core-shell nano- and microrods with highest internal quantum efficiencies.
The defect density of the GaN rods grown by MOVPE at TUBS could be reduced in two respects. Firstly, the density of large particles which emerge on the GaN columns and which would be harmful for device fabrication was reduced by changes in the growth procedure. Secondly, by structural optimization of the 3D LEDs, both in core and in shell, the defect density inside the QWs was reduced considerably, as judged by PL-IQE. The new structures show an enhanced PL-IQE which indicates less non-radiative recombination centers and thus a reduced dislocation density. Maximum PL-IQE values of around 60% have been achieved.
3D GaN structures with an aspect ratio above 5 were achieved by an array of GaN columns grown by selective area growth on patterned templates using MOVPE. Successively also 3D GaN based LED structures with an increase in active area ratio when compared to a planar LED of active area (AA) above 5 were achieved. For this purpose, metal organic vapor phase epitaxy (MOVPE) selective area growth of an array of n-doped GaN rods (core) was followed by the shell growth of a 3- or 5-fold InGaN/GaN multiquantum well (active region) and a p-doped cap layer.
3D GaN structures with defect densities below 104 cm-2, a diameter distribution below 10% and a height distribution below 10% are achieved by periodic arrays of GaN columns with an aspect ratio of about 10 and higher. These arrays are grown on most of the patterned area by selective area MOVPE on whole 2” templates (GaN buffer layer on Sapphire) fully masked by patterned SiOx. After the MOVPE growth a GaN column is present at each intended position in the array. The density of larger particles is below mid 104 cm-2. By TEM several columns are proven to be free of threading dislocation and stacking faults.
Core-shell structures including an InGaN SQW or MQW showing a PL-IQE of about 60% were achieved by reducing the effect of Si-doping of the n-GaN core by a spacer shell. Investigation of complete LED structures showed a reduction of the PL-IQE which is assigned to diffusion of Mg atoms into the active region - it has been suggested that this can be controlled by an undoped spacer between the QW and the p-GaN shell.
The approach of 3D LEDs is also found to be appropriate in order to reduce the material consumption for the production of light emitting diodes. Since the defect density of the GaN cores will be reduced due to the small footprint and the large aspect ratio, no thick GaN buffers with high crystal quality are required. We could already prove that half of the GaN layer thickness is a sufficient basis for the growth of a homogenous LED array. Due to the high active area ratio of the 3D LEDs, an increased current and therefore an increased lumen light output per wafer area compared to SOA LEDs can be achieved. Both of the two effects will reduce the material consumption per lumen output drastically.
Light Conversion, Light Extraction and Benchmarking
The work regarding light conversion, light extraction and benchmarking covered the development of suitable phosphors, the identification, testing and implementation of high refractive index materials, the development of conversion and deposition technologies, the development of chip processes of 3D structures with high out-coupling efficiencies and the benchmark of 3D GaN devices.
Within the phosphor development suitable phosphor materials for the realization of highly efficient coupling between 3D GaN and light conversion layers were developed. Green and yellow micro-grained garnets could be successfully synthesized, followed by a post-processing step to further reduce the primary particle size. To realize warm-white 3D core-shell InGaN/GaN microrod LEDs red micro-grained phosphors were investigated as well.
A feasible chip process for 3D LEDs has been developed and LED operation at blue and also at longer wavelengths has been demonstrated. The initial process developed within GECCO has been revised several times with respect to improved high current performance, light extraction and efficiency. Strategies for further developments were retrieved. State-of-the-art methods for increasing the light extraction efficiency have to be taken into account. All steps are compatible with high volume processing.
Having developed suitable micrograin phosphors, the next step in the process chain was the embedding of these phosphors inside the microrod array by using an appropriate matrix material. This matrix material had to fulfil several requirements like optical transparency, high refractive index, high stability vs. high optical, mechanical and thermal load as well as good process ability. Due to the different geometry and particle sizes of the core-shell microrod structures and the micro-grain phosphor compared to conventional planar technology the matrix materials have to be evaluated especially for their filling behaviour. From a choice of several different materials a special high refractive index (HRI) silicone was finally selected. Also the phosphor deposition process was evaluated. Most standard deposition processes for silicones like spin coating or spraying yielded acceptable results, but at the end the electrophoretic deposition (EPD) technique was chosen because it showed excellent filling properties and the best uniformity after it has been tailored to the special material properties of the phosphor filled silicone.
After epitaxy, a chip process is needed to establish electric contacts and to provide planar bond pads for wire bonding. However, it also has to facilitate a sufficient current spreading and a high light extraction. Especially high aspect ratios remain a major challenge for the chip process due to increasing fragility. Furthermore, attention must be given to possible current leakage pathways and light extraction losses. Several different routes have been tested, adapted and refined, leading to the successful demonstration of the first electrical driven white 3D-LED already in month 18 of this project.
However, due to the necessary and early focus on verifying the process capability inside a high volume production environment which used up more resources than anticipated, the original efficacy target of 200 lm/W was not yet reached within the time frame of the project. The IQE values which have been evaluated within the benchmarking (up to 60%) are still somewhat below state-of-the-art planar LEDs. On the other hand, fast progress has been made in solving critical questions and gaining important insight in the fundamental issues and chances of a novel 3D core-shell technology, which can now be the basis for further device development.
A further iteration of the chip process is currently under testing, and the next generation - which will benfit from the “Lessons Learned” of this project - is now under development within an internal follow-up project at OSRAM Opto Semiconductors. Looking at the fast progress after focusing on individual technological issues which reduced efficiency in the first demonstrator devices, it is reasonable to expect significant improvements within the next few months.
Characterisation
By employing TEM, FIB, SEM and FESEM structural and morphological characterisation of 3D Ga(In)N materials was performed giving valuable insights into the growth processes of MOVPE and MBE. For characterization of optical properties different setups for cathodoluminescence and temperature resolved photoluminescence were employed.
The benchmarking of 3D LEDs and detailed performance test on demonstrators have been obtained by two different approaches. The first consisted in measuring the radiant flux of a microrod LED by comparing the emission intensity of the microrod device to that of a planar reference LED of similar chip size, wavelength and lambertian angular power distribution. In a second approach, an efficiency investigation of single 3D LEDs was performed using a manipulator setup inside the CL-SEM and in this way comparing the electroluminescence from three different samples contacted by a probe tip.
Therefore, an experimental setup to determine quantitatively the electro-optical behavior (I-V, P-I, heating effects) of single (individual) and arrays of core-shell InGaN/GaN nano and microrods as well as benchmark LEDs has been set in place and already been used extensively. The equipment enables full characterization of blue and white LEDs addressing e.g. possible leakage pathways, the effect of indium concentration gradient along the microrod height, current density dependence of detected emission bands and quantitative determination of quantum efficiencies.
Modelling
Thermal simulation of 3D GaN LEDs under operation both as the semiconductor structure only and as a device in the 3D package have been performed with the aid of ANSYS software.
In the case of device oriented modelling, the elaborated electro-thermal model incorporating physical heat dissipation sub-model has been used mainly for analyses of single nanowire structure. The series of numerical simulations allows estimating an influence of different geometrical, physical, operational parameters on temperature distribution within the 3D GaN structure. The highest temperature values are reported at the top of the nanowire; hence the thermal resistance of the modelled structure depends directly on dimensions of core-shell structure.
In the case of package oriented modelling, the analysis mainly focused on identification of highest temperature gradient locations along the heat flow paths from the semiconductor structure to the ambient. Possible improvements have been identified and pointed out. Moreover, results of these simulations have been used to formulate boundary conditions for device oriented modelling.
Electro-thermal simulations have been performed with the use of commercial software SENTAURUS at electro-thermal models of 3D LEDs structures using a numerical analysis of their features. Since Sentaurus was elaborated for simulations of silicon structures and next adopted for the simulation of other semiconductors by adding additional procedures and new material libraries, including GaN, its usefulness and correctness in the application to GaN structures is not obvious. Therefore, firstly, some test simulations for simple GaN devices covering quantum wells has been performed. They allowed identifying the software weaknesses and the possibility of its use as the CAD tool supporting numerical analyses of optoelectronics GaN device and general, they confirmed the usefulness of SENTAURUS for the planned investigations by delivering consistent results.
As the first 3D LED structure, in order to test the simulation package, a planar diode presented in detail in available literature was simulated. The comparison of the simulated and experimentally measured characteristics proves the correctness of the developed numerical approach to GaN LED diode simulation.
After that, a core-shell 3D GaN LED structure with parameters from experimental groups of the project was used and a numerical model has been developed. Taking into account the symmetry in the designed core-shell structure, the domain of worked out model has been limited to one cylindrical nano-pillar with its surroundings. The three-dimensional modelling has been achieved due to the use of cylindrical coordination system in the simulations. The model has been used to numerically investigate the core-shell geometry for different design parameters giving an insight on their influence on the device operating conditions.
The modelling and simulation of MOVPE growth processes covered MOVPE reactor models of heat and mass transfer as well as models of 3D structure growth. It dealt with the elaboration of a new 3D model of the reactor chamber taking into account gas flow and temperature field analyses. Its benefits are the reduction of the model size with concurrent incorporation of the showerhead geometry and heat transfer radiation mechanisms (both these elements has been neglected in hitherto studies).
Emission from arrays of 3D GaN pillars and single columns have been successfully simulated using ray tracing. The results are compared to results from experiments and explain significant features of the emission pattern observed by an optical microscope.
Furthermore, in order to evaluate the numerical models, both experimental and numerical investigations of gas flow and temperature distribution in the MOVPE reactor have been performed. The experiments were carried out using the TUL PECVD reactor that has been adopted to work in the conditions typical for the processes of film growth from the gas phase during semiconductors fabrication on the one hand and to create appropriate conditions for the considered measurements, on the other hand. Among others, a set-up was created to carry out temperature distribution measurements inside the reactor chamber using IR camera and thermocouple. The numerical investigations were aimed at the creation of numerical models of heat and gas flow inside the reactor chamber of the TUL PECVD. They allowed the verification of different numerical approaches, e.g. boundary conditions, in the case of PECVD models as well as to perform their quality and quantity evaluation taking the advantage of temperature measurements inside the chamber under different operating conditions.
For the development of electro- thermal models of 3D-LEDs both electrical and thermal phenomena inside the devices were taken into account. The models of both planar and 3D structures of GaN LEDs have been developed using commercial CAD package Sentaurus. The usefulness of the models to investigate optoelectronics GaN heterostructure covering layers of multi quantum wells (MQW) has been examined.
The worked out models for both investigated reactors PECVD (in TUL) and MOVPE (in TUBS) shown that all the heat transport mechanisms: conduction, convection, and radiation must be incorporated in the final models. Some simplifications of the geometry are possible mainly due to symmetry.
To summarize, GECCO has led a substantial progress concerning the experimental and theoretical understanding as well as realization of 3D core-shell nano- and micro LEDs for future solid state lighting technologies.
Potential Impact:
The research initiated and performed by the GECCO team will be taken up by the Epitaxy Competence Center ec², a joint collaboration between TUBS and OOS. This research center, opened officially in 2015 (www.ec2.tu-bs.de) has been implemented in order to focus on application oriented research in the field of nitride-based semiconductors, with particular emphasize on 3D core-shell nanorod LEDs. The next generation of the chip processing from arrays of 3D-LEDs is already under development within an internal follow-up project at OSRAM Opto Semiconductors. The European GaN research now has access to 3D GaN technology - via the epitaxy competence center ec² - so that this fascinating novel technology can jointly be brought forward.
The ec² also has access to advanced analytical capabilities for 3D nanostructures, including a new field emission scanning electron microscope with cathodoluminescence system, which has been designed to suit the demand of nondestructive 3D-LED-analytics on 2” and 4” wafers. Due to the field emission gun it enables also a high spatial resolution by simultaneously investigating CL and electron beam induced current (EBIC).
The detected challenges regarding the growth and processing were addressed and solved during the GECCO project, opening the possibility to a novel technology for solid state lighting. Hence it is expected that, by using the knowledge and developed processes, an increased efficacy as well as a production costs will be achieved by devices of 3D-LEDs. As the external properties of such devices are similar to those planar LEDs it can replace the conventional structures pushing the affordability of modern lighting.
The results published by the GECCO consortium had and still have a high impact on the international scientific community. Numerous invited talks have been held by the GECCO members at international conference. One of the highlights concerning dissemination is the spring 2014 conference of the E-MRS in Lille, France, where GECCO organized the symposium K on "Challenges for group III nitride semiconductors for solid state lighting and beyond".
The dissemination and training also comprised respective internal consortium's activities. Exchange visits of PhD students and post-docs were realised as well as a Workshop for PhD students and post-docs was organised.
The main achievements of the GECCO project are published by many scientific papers and conference contributions. An overview of those is also given on the GECCO homepage (www.gecco.tu-bs.de) three issues of the annual project newsletters are provided there, too.
List of Websites:
www.gecco.tu-bs.de
Prof. Dr. Andreas Waag
TU Braunschweig
Institute of Semiconductor Technology
Hans-Sommer-Strasse 66
38106 Braunschweig, Germany
a.waag@tu-bs.de
phone +49-531-391-3773
fax +49-531-391-5844
The GECCO consortium is aiming at developing a novel technology for solid state lighting namely 3D core-shell LEDs circumventing many of the limitations of planar LED technology. Due to intense collaboration between partners, substantial progress can be reported for all periods of the project. Some major improvements in 3D GaN technology have been achieved. Major insights into the growth mechanisms of GaN and the necessary processing towards white LEDs have been gained. This includes the fabrication of InGaN quantum wells on m-plane sidewalls with high internal quantum efficiency and high indium content. Special phosphors have been developed in order to fulfil the requirements of 3D blue emitting nanoLEDs processing into white LEDs. The GECCO consortium belongs to the leading groups worldwide working on the realization of 3D GaN technology.
The collaboration between partners has been very efficient and professional, supported by regular project meetings held in Braunschweig (twice), Madrid, Bristol, Brussels and Regensburg and close exchange of information by web-conferences guiding the scientific progress of GECCO. –This has been supported by maintaining a web based project management system which aligned the collaboration and monitoring inside the project. In addition, a risk management inside the R&D work packages was carried out on a regular basis, helping to avoid critical delays.
Project Context and Objectives:
The GECCO consortium is aiming at developing a novel technology for solid state lighting circumventing many of the limitations of planar LED technology by employing 3D core-shell LEDs. This summary describes the main achievements during the project which evolved in accordance to the work plan.
The basis of the project was the selective area growth of first the GaN columns and second the LED related shells. During the course of the project a fundamental understanding of the growth processes was compiled. The growth masks, which are needed for selective area growth, were optimized not only in view of filling homogeneity and reproducibility, but also in view of its insulating properties inside the processed LED-chip. The growth of InGaN shell layers and cores with a high amount of indium incorporation was addressed successfully..
In the second part of the project, GECCO moved from a more epitaxy related focus to device related research. Arrays of 3D LED structures were grown with further optimized processes on patterned templates - fabricated either by employing conventional photo lithography or nano imprint lithography - using selective area growth by MOVPE as well as by MBE. Strategies to combine phosphors with the 3D approach are also demonstrated.
The chip process was further developed and optimized so that complete LED devices can be manufactured from the 3D arrays including a suitable high refractive index (HRI) matrix material which was identified during the first period. High efficiency phosphors for different colors with grain sizes below 1 µm were developed to fit between the rods so that demonstrator device were processed on a chip level which allowed controlling the color and matching the Planckian locus at 7500 K. Additionally the chip process was improved regarding current spreading and functional layers (e.g. insulating layers).
A detailed characterization of the fabricated nanostructures was performed by using a strong combination of nanoanalytical methods, including electron microscopy in transmission and scanning mode (TEM, SEM) as well as cathodoluminescence (CL-SEM), electroluminescence via point contacts inside an SEM and temperature resolved photoluminescence (PL). Avoiding structural defects, which are usually accompanied by non-radiative recombination, as well as insufficient current spreading and non-optimized doping profiles have been identified to be a prerequisite for obtaining high efficacies.
The understanding and optimization of properties of the 3D LEDs was supported by simulations of the local growth conditions, in particular by the temperature distribution inside an array. The characterization of the 3D LEDs was supported by investigating the light extraction by means of ray tracing and by electro thermal simulations inside the LEDs, also including the influence of the packaging.
Even though the evaluated devices based on 3D LEDs did not yet - due to a shift or resources to other work packages - reach the targeted extremely competitive goal of 200 lm/W in electroluminescence, the first demonstration ever of a white, phosphor converted nanorod LED has been a big step forward.
Project Results:
Substantial insight into the growth mechanisms by both MOCVD and MBE could be gained. By optimizing the overall processing, the reproducible growth of arrays of good quality 3D InGaN/GaN nano- and microrods with high aspect ratios and a good yield on whole wafers has become possible. Also InGaN QWs emitting in the red could be realized with surprisingly high PL-IQE values:
3D growth
3D GaN submicron-columns and nanorods have been successfully grown on conductive templates using MOVPE and MBE, respectively. Using MOVPE, the diameter of GaN columns is about 600 nm and the height is under control up to tenths of µm. The structural quality of the 3D columns is very high, no dislocations or stacking faults could be detected by TEM. The doping concentrations in n-type core and the p-type shell, as well as the In concentration in MQW can be varied. Under the utilized growth conditions, up to 100% of the patterned holes can be filled by columns, resulting in a high coverage of about 85% of the patterned region on 2” and 4” wafers by a uniform array. The growth mechanism of GaN 3D columns has been investigated systematically. In particular, the influence of hydrogen, silicon doping, as well as V/III ratio during growth has extensively been studied.
InGaN cores with an Indium incorporation above 20% has been obtained following the optimization process on the MBE growth of InGaN/GaN nanorods.
For the first level of template preparation three different types of mask pattern were used for photolithography at TUBS. All of them were full-area masks with hexagonal structures arranged in the closest packaging. The diameter D varies from 0.8 µm to 2 µm and the pitch P is either 3 µm or 4 µm. Templates composed of a 2" silicon or sapphire substrate with a SiO2 masking layer were available. In addition a fourth 4" mask was available for the second level. The progress beyond the first level is also an optimized fabrication process so that amount of filled patterns could be strongly increased.
The third level mask design has been developed at OOS. A new 4" mask was designed and fabricated which has an optimized structure with respect to the desired high aspect ratio of the nano- and microrods, and to support the phosphor deposition experiments. Nano-imprint lithography (NIL) was chosen for the fabrication of the templates in order to provide the necessary throughput for serving the growth experiments; one master stamp was used for the template fabrication at OOS. The pattern (i.e. diameter of 600 nm and 800 nm and a pitch of 3 µm and 4 μm) was distributed across the 4" wafer in a way that accommodates the non-symmetric deposition pattern of the wafer due to the MOVPE reactor geometry. These optimized conditions based on the achievements of the project are necessary for core-shell nano- and microrods with highest internal quantum efficiencies.
The defect density of the GaN rods grown by MOVPE at TUBS could be reduced in two respects. Firstly, the density of large particles which emerge on the GaN columns and which would be harmful for device fabrication was reduced by changes in the growth procedure. Secondly, by structural optimization of the 3D LEDs, both in core and in shell, the defect density inside the QWs was reduced considerably, as judged by PL-IQE. The new structures show an enhanced PL-IQE which indicates less non-radiative recombination centers and thus a reduced dislocation density. Maximum PL-IQE values of around 60% have been achieved.
3D GaN structures with an aspect ratio above 5 were achieved by an array of GaN columns grown by selective area growth on patterned templates using MOVPE. Successively also 3D GaN based LED structures with an increase in active area ratio when compared to a planar LED of active area (AA) above 5 were achieved. For this purpose, metal organic vapor phase epitaxy (MOVPE) selective area growth of an array of n-doped GaN rods (core) was followed by the shell growth of a 3- or 5-fold InGaN/GaN multiquantum well (active region) and a p-doped cap layer.
3D GaN structures with defect densities below 104 cm-2, a diameter distribution below 10% and a height distribution below 10% are achieved by periodic arrays of GaN columns with an aspect ratio of about 10 and higher. These arrays are grown on most of the patterned area by selective area MOVPE on whole 2” templates (GaN buffer layer on Sapphire) fully masked by patterned SiOx. After the MOVPE growth a GaN column is present at each intended position in the array. The density of larger particles is below mid 104 cm-2. By TEM several columns are proven to be free of threading dislocation and stacking faults.
Core-shell structures including an InGaN SQW or MQW showing a PL-IQE of about 60% were achieved by reducing the effect of Si-doping of the n-GaN core by a spacer shell. Investigation of complete LED structures showed a reduction of the PL-IQE which is assigned to diffusion of Mg atoms into the active region - it has been suggested that this can be controlled by an undoped spacer between the QW and the p-GaN shell.
The approach of 3D LEDs is also found to be appropriate in order to reduce the material consumption for the production of light emitting diodes. Since the defect density of the GaN cores will be reduced due to the small footprint and the large aspect ratio, no thick GaN buffers with high crystal quality are required. We could already prove that half of the GaN layer thickness is a sufficient basis for the growth of a homogenous LED array. Due to the high active area ratio of the 3D LEDs, an increased current and therefore an increased lumen light output per wafer area compared to SOA LEDs can be achieved. Both of the two effects will reduce the material consumption per lumen output drastically.
Light Conversion, Light Extraction and Benchmarking
The work regarding light conversion, light extraction and benchmarking covered the development of suitable phosphors, the identification, testing and implementation of high refractive index materials, the development of conversion and deposition technologies, the development of chip processes of 3D structures with high out-coupling efficiencies and the benchmark of 3D GaN devices.
Within the phosphor development suitable phosphor materials for the realization of highly efficient coupling between 3D GaN and light conversion layers were developed. Green and yellow micro-grained garnets could be successfully synthesized, followed by a post-processing step to further reduce the primary particle size. To realize warm-white 3D core-shell InGaN/GaN microrod LEDs red micro-grained phosphors were investigated as well.
A feasible chip process for 3D LEDs has been developed and LED operation at blue and also at longer wavelengths has been demonstrated. The initial process developed within GECCO has been revised several times with respect to improved high current performance, light extraction and efficiency. Strategies for further developments were retrieved. State-of-the-art methods for increasing the light extraction efficiency have to be taken into account. All steps are compatible with high volume processing.
Having developed suitable micrograin phosphors, the next step in the process chain was the embedding of these phosphors inside the microrod array by using an appropriate matrix material. This matrix material had to fulfil several requirements like optical transparency, high refractive index, high stability vs. high optical, mechanical and thermal load as well as good process ability. Due to the different geometry and particle sizes of the core-shell microrod structures and the micro-grain phosphor compared to conventional planar technology the matrix materials have to be evaluated especially for their filling behaviour. From a choice of several different materials a special high refractive index (HRI) silicone was finally selected. Also the phosphor deposition process was evaluated. Most standard deposition processes for silicones like spin coating or spraying yielded acceptable results, but at the end the electrophoretic deposition (EPD) technique was chosen because it showed excellent filling properties and the best uniformity after it has been tailored to the special material properties of the phosphor filled silicone.
After epitaxy, a chip process is needed to establish electric contacts and to provide planar bond pads for wire bonding. However, it also has to facilitate a sufficient current spreading and a high light extraction. Especially high aspect ratios remain a major challenge for the chip process due to increasing fragility. Furthermore, attention must be given to possible current leakage pathways and light extraction losses. Several different routes have been tested, adapted and refined, leading to the successful demonstration of the first electrical driven white 3D-LED already in month 18 of this project.
However, due to the necessary and early focus on verifying the process capability inside a high volume production environment which used up more resources than anticipated, the original efficacy target of 200 lm/W was not yet reached within the time frame of the project. The IQE values which have been evaluated within the benchmarking (up to 60%) are still somewhat below state-of-the-art planar LEDs. On the other hand, fast progress has been made in solving critical questions and gaining important insight in the fundamental issues and chances of a novel 3D core-shell technology, which can now be the basis for further device development.
A further iteration of the chip process is currently under testing, and the next generation - which will benfit from the “Lessons Learned” of this project - is now under development within an internal follow-up project at OSRAM Opto Semiconductors. Looking at the fast progress after focusing on individual technological issues which reduced efficiency in the first demonstrator devices, it is reasonable to expect significant improvements within the next few months.
Characterisation
By employing TEM, FIB, SEM and FESEM structural and morphological characterisation of 3D Ga(In)N materials was performed giving valuable insights into the growth processes of MOVPE and MBE. For characterization of optical properties different setups for cathodoluminescence and temperature resolved photoluminescence were employed.
The benchmarking of 3D LEDs and detailed performance test on demonstrators have been obtained by two different approaches. The first consisted in measuring the radiant flux of a microrod LED by comparing the emission intensity of the microrod device to that of a planar reference LED of similar chip size, wavelength and lambertian angular power distribution. In a second approach, an efficiency investigation of single 3D LEDs was performed using a manipulator setup inside the CL-SEM and in this way comparing the electroluminescence from three different samples contacted by a probe tip.
Therefore, an experimental setup to determine quantitatively the electro-optical behavior (I-V, P-I, heating effects) of single (individual) and arrays of core-shell InGaN/GaN nano and microrods as well as benchmark LEDs has been set in place and already been used extensively. The equipment enables full characterization of blue and white LEDs addressing e.g. possible leakage pathways, the effect of indium concentration gradient along the microrod height, current density dependence of detected emission bands and quantitative determination of quantum efficiencies.
Modelling
Thermal simulation of 3D GaN LEDs under operation both as the semiconductor structure only and as a device in the 3D package have been performed with the aid of ANSYS software.
In the case of device oriented modelling, the elaborated electro-thermal model incorporating physical heat dissipation sub-model has been used mainly for analyses of single nanowire structure. The series of numerical simulations allows estimating an influence of different geometrical, physical, operational parameters on temperature distribution within the 3D GaN structure. The highest temperature values are reported at the top of the nanowire; hence the thermal resistance of the modelled structure depends directly on dimensions of core-shell structure.
In the case of package oriented modelling, the analysis mainly focused on identification of highest temperature gradient locations along the heat flow paths from the semiconductor structure to the ambient. Possible improvements have been identified and pointed out. Moreover, results of these simulations have been used to formulate boundary conditions for device oriented modelling.
Electro-thermal simulations have been performed with the use of commercial software SENTAURUS at electro-thermal models of 3D LEDs structures using a numerical analysis of their features. Since Sentaurus was elaborated for simulations of silicon structures and next adopted for the simulation of other semiconductors by adding additional procedures and new material libraries, including GaN, its usefulness and correctness in the application to GaN structures is not obvious. Therefore, firstly, some test simulations for simple GaN devices covering quantum wells has been performed. They allowed identifying the software weaknesses and the possibility of its use as the CAD tool supporting numerical analyses of optoelectronics GaN device and general, they confirmed the usefulness of SENTAURUS for the planned investigations by delivering consistent results.
As the first 3D LED structure, in order to test the simulation package, a planar diode presented in detail in available literature was simulated. The comparison of the simulated and experimentally measured characteristics proves the correctness of the developed numerical approach to GaN LED diode simulation.
After that, a core-shell 3D GaN LED structure with parameters from experimental groups of the project was used and a numerical model has been developed. Taking into account the symmetry in the designed core-shell structure, the domain of worked out model has been limited to one cylindrical nano-pillar with its surroundings. The three-dimensional modelling has been achieved due to the use of cylindrical coordination system in the simulations. The model has been used to numerically investigate the core-shell geometry for different design parameters giving an insight on their influence on the device operating conditions.
The modelling and simulation of MOVPE growth processes covered MOVPE reactor models of heat and mass transfer as well as models of 3D structure growth. It dealt with the elaboration of a new 3D model of the reactor chamber taking into account gas flow and temperature field analyses. Its benefits are the reduction of the model size with concurrent incorporation of the showerhead geometry and heat transfer radiation mechanisms (both these elements has been neglected in hitherto studies).
Emission from arrays of 3D GaN pillars and single columns have been successfully simulated using ray tracing. The results are compared to results from experiments and explain significant features of the emission pattern observed by an optical microscope.
Furthermore, in order to evaluate the numerical models, both experimental and numerical investigations of gas flow and temperature distribution in the MOVPE reactor have been performed. The experiments were carried out using the TUL PECVD reactor that has been adopted to work in the conditions typical for the processes of film growth from the gas phase during semiconductors fabrication on the one hand and to create appropriate conditions for the considered measurements, on the other hand. Among others, a set-up was created to carry out temperature distribution measurements inside the reactor chamber using IR camera and thermocouple. The numerical investigations were aimed at the creation of numerical models of heat and gas flow inside the reactor chamber of the TUL PECVD. They allowed the verification of different numerical approaches, e.g. boundary conditions, in the case of PECVD models as well as to perform their quality and quantity evaluation taking the advantage of temperature measurements inside the chamber under different operating conditions.
For the development of electro- thermal models of 3D-LEDs both electrical and thermal phenomena inside the devices were taken into account. The models of both planar and 3D structures of GaN LEDs have been developed using commercial CAD package Sentaurus. The usefulness of the models to investigate optoelectronics GaN heterostructure covering layers of multi quantum wells (MQW) has been examined.
The worked out models for both investigated reactors PECVD (in TUL) and MOVPE (in TUBS) shown that all the heat transport mechanisms: conduction, convection, and radiation must be incorporated in the final models. Some simplifications of the geometry are possible mainly due to symmetry.
To summarize, GECCO has led a substantial progress concerning the experimental and theoretical understanding as well as realization of 3D core-shell nano- and micro LEDs for future solid state lighting technologies.
Potential Impact:
The research initiated and performed by the GECCO team will be taken up by the Epitaxy Competence Center ec², a joint collaboration between TUBS and OOS. This research center, opened officially in 2015 (www.ec2.tu-bs.de) has been implemented in order to focus on application oriented research in the field of nitride-based semiconductors, with particular emphasize on 3D core-shell nanorod LEDs. The next generation of the chip processing from arrays of 3D-LEDs is already under development within an internal follow-up project at OSRAM Opto Semiconductors. The European GaN research now has access to 3D GaN technology - via the epitaxy competence center ec² - so that this fascinating novel technology can jointly be brought forward.
The ec² also has access to advanced analytical capabilities for 3D nanostructures, including a new field emission scanning electron microscope with cathodoluminescence system, which has been designed to suit the demand of nondestructive 3D-LED-analytics on 2” and 4” wafers. Due to the field emission gun it enables also a high spatial resolution by simultaneously investigating CL and electron beam induced current (EBIC).
The detected challenges regarding the growth and processing were addressed and solved during the GECCO project, opening the possibility to a novel technology for solid state lighting. Hence it is expected that, by using the knowledge and developed processes, an increased efficacy as well as a production costs will be achieved by devices of 3D-LEDs. As the external properties of such devices are similar to those planar LEDs it can replace the conventional structures pushing the affordability of modern lighting.
The results published by the GECCO consortium had and still have a high impact on the international scientific community. Numerous invited talks have been held by the GECCO members at international conference. One of the highlights concerning dissemination is the spring 2014 conference of the E-MRS in Lille, France, where GECCO organized the symposium K on "Challenges for group III nitride semiconductors for solid state lighting and beyond".
The dissemination and training also comprised respective internal consortium's activities. Exchange visits of PhD students and post-docs were realised as well as a Workshop for PhD students and post-docs was organised.
The main achievements of the GECCO project are published by many scientific papers and conference contributions. An overview of those is also given on the GECCO homepage (www.gecco.tu-bs.de) three issues of the annual project newsletters are provided there, too.
List of Websites:
www.gecco.tu-bs.de
Prof. Dr. Andreas Waag
TU Braunschweig
Institute of Semiconductor Technology
Hans-Sommer-Strasse 66
38106 Braunschweig, Germany
a.waag@tu-bs.de
phone +49-531-391-3773
fax +49-531-391-5844