Steve Denbaars | University of California Santa Barbara: Can InGaN red hold efficiency at the 80 °C and 500 A/cm² displays need?
07:03 - 09:17
Other snippets from this talk
Summary of the clip:
Can InGaN red hold efficiency at the 80 °C and 500 A/cm² displays need?
Red InGaN micro-LEDs have reached external quantum efficiencies as high as 20%, but only at very low current densities. Droop worsens from blue to green to red: green peaks near 38% and falls to about 17%. For projection displays driven at 500 A/cm², blue is no problem, while green and red are. Direct-view displays avoid this. Wall-plug efficiency reaches 15%, from 5 to 50 micron devices.
A key InGaN red advantage is integration: red, green and blue can sit on a single chip, so no dicing and no placement of separate red die is needed. The speaker calls this really the only advantage of the red now. That monolithic RGB approach removes pick-and-place of tiny red chips and changes assembly economics for full-colour micro-LED displays.
Temperature behaviour is where InGaN red pulls ahead. From room temperature to 200 °C, external quantum efficiency holds up well against gallium arsenide. Across the typical display working range up to 80 °C, InGaN loses only 10% of normalised light intensity, versus 30% for GaAs, which is not reliable by 100 °C. Displays capped at 50 °C could live with AlGaN; 80 °C and above favour InGaN.
In this short video, you can learn:
* InGaN red micro-LEDs hit 20% external quantum efficiency only at very low current densities, so droop remains the barrier at 500 A/cm².
* Green and red, not blue, are the problem at projection-drive current densities, while direct-view displays tolerate the droop.
* At 80 °C InGaN red loses just 10% of normalised light intensity against 30% for GaAs, which is unreliable by 100 °C.
📋 **Clip Abstract** InGaN red micro-LEDs reach 20% external quantum efficiency at low current density and 15% wall-plug efficiency across 5–50 micron sizes, but droop makes green and red problematic at the 500 A/cm² projection displays need, unlike blue. Temperature performance is the win: only 10% normalised intensity loss at 80 °C versus 30% for GaAs, and monolithic RGB on one chip removes dicing and placement.
About the speaker:
* Speaker: Steve Denbaars
* Company: University of California Santa Barbara
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#InGaNRed, #MicroLEDDroop, #MonolithicRGB, #WallPlugEfficiency, #MicroLEDDisplays, #ProjectionDisplays
This is a highlight of the presentation:
Development of GaN Based MicroLED Devices for Full Color Displays and High Speed Optical Communication
More Highlights from the same talk.
02:32 - 04:12
Can a sidewall DBR recover the light escaping the sides of a 5-micron micro-LED?
Can a sidewall DBR recover the light escaping the sides of a 5-micron micro-LED?
The sidewall-DBR micro-LED coats distributed Bragg reflectors around the sides to redirect light that would otherwise escape laterally. An eight-and-a-half-period DBR using silicon dioxide and titanium oxide sends most side-directed light back down, where it is reflected back up, so emission comes from the top surface. This eliminates side reflection light and increases forward light. The structure is not a resonant cavity, but it improves forward directionality.
The approach has been tested from 100 microns down to five microns. At the five-micron size, the effect is even bigger: compared with just titanium oxide around the sides, the DBR sidewall coating delivers about 25% more light out in the forward direction. It is not perfect and not a resonant cavity at this point, but it gets more light forward.
Far-field or airside radiation for both S polarized and P polarized light changes from a more Lambertian distribution to roughly plus or minus 30 degrees in angular dispersion. That is about 60-degree full width at half maximum. The DBR sidewall coating narrows emission, which will help fiber coupling, but mainly helps viewability.
In this short video, you can learn:
* An eight-and-a-half-period silicon dioxide/titanium oxide DBR around the sidewall redirects side-emitted light back up for top-surface emission.
* From 100 microns to five microns, the five-micron device shows about 25% more forward light with the DBR than with titanium oxide sidewalls.
* The airside radiation for S and P polarizations becomes roughly +/-30 degrees, about 60-degree full width at half maximum.
📋 **Clip Abstract** A sidewall distributed Bragg reflector made from eight-and-a-half periods of silicon dioxide and titanium oxide redirects laterally escaping light back up, increasing top-surface emission. Tested from 100 microns to five microns, the five-micron device gains about 25% more forward light versus titanium oxide sidewalls, while its far field narrows to roughly 60-degree full width at half maximum for both S and P polarizations.
About the speaker:
* Speaker: Steve Denbaars
* Company: University of California Santa Barbara
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#SidewallDBR, #SiO2TiO2DBR, #FarFieldPattern, #SPolarizedPPolarized, #MicroLEDDisplays, #OpticalInterconnect
12:13 - 14:17
Why can a simple on-off micro-LED beat co-packaged optics below one picojoule per bit?
Why can a simple on-off micro-LED beat co-packaged optics below one picojoule per bit?
Micro-LED links promise below one picojoule per bit. The data shown achieved 0.85 picojoule per bit at reasonable data rates. Co-packaged optics, now leading optical interconnects, is stuck around three to five picojoule per bit. Much of that penalty comes from the SerDes chip and driver. Micro-LEDs instead use non-return-to-zero transmission, avoiding PAM-4, forward error correction, and digital signal processing chips. Their simplicity is turning on and off with little error correction.
The field is at its infancy. Claims of one to five dollars per gigabit are highly speculative. Costs can be driven down tremendously. In LED lighting, a two hundred dollar bulb in 2000; a large lighting company said come back in fifty years. Ten years later that company was out of business. By 2010 the bulb was under ten dollars, and by 2014 it was a dollar something. The same cost trajectory can happen here.
Gallium arsenide and indium phosphide co-packaged optics have a big temperature problem. That is why indium phosphide lasers are pulled outside the chip. Micro-LEDs can be placed right onto the chips. They run at 100 and 150 degrees Celsius, and data was shown out to 200 degrees Celsius. This thermal tolerance reduces the need for external lasers and supports tighter integration with the driver electronics.
In this short video, you can learn:
* Micro-LEDs achieved 0.85 picojoule per bit at reasonable data rates, versus co-packaged optics stuck around three to five picojoule per bit.
* Non-return-to-zero transmission avoids PAM-4, forward error correction, digital signal processing, and the extra power they require.
* Gallium arsenide and indium phosphide co-packaged optics pull lasers off-chip due to temperature, while micro-LED data ran to 200 degrees Celsius.
📋 **Clip Abstract** Micro-LEDs reached 0.85 picojoule per bit using simple non-return-to-zero on-off transmission, while co-packaged optics remain at three to five picojoule per bit because of SerDes, driver, PAM-4, forward error correction, and DSP overhead. The approach is still at its infancy, with speculative one-to-five-dollar-per-gigabit cost estimates, but LED lighting's collapse from a two-hundred-dollar bulb to a dollar something suggests large cost reduction is possible, and micro-LEDs can sit on-chip and run to 200 degrees Celsius.
About the speaker:
* Speaker: Steve Denbaars
* Company: University of California Santa Barbara
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#PicojoulePerBit, #NonReturnToZero, #CoPackagedOptics, #ForwardErrorCorrection, #OpticalInterconnects, #MicroLEDDisplays
04:46 - 07:03
Why do micro-LED papers hide external quantum efficiency, and what does deleting ITO buy?
Why do micro-LED papers hide external quantum efficiency, and what does deleting ITO buy?
Many micro-LED papers omit external quantum efficiency. Calibrated integrating spheres show an omnidirectional reflector around the device yields a 25% improvement across sizes from five to 100 microns, compared with no DBR. The data also expose drive current: good efficiency persists to 500 amperes per centimeter, reaching about 28%. Further improvement requires reducing chip losses.
One loss reduction is removing the indium tin oxide layer, which is about 10% absorbing, and replacing it with n-type gallium nitride. Tunnel junctions are widely used in gallium arsenide VCSELs, providing the transparency needed for an inner cavity contact and good vertical cavity emission. The same approach was applied to a micro-LED: ITO was removed and replaced with an n-type gallium nitride tunnel junction.
Normally tunnel junctions do not work well in lighting-class LEDs, which are a millimeter or larger. In micro-LEDs, hydrogen passivation that passivates the p-type dopant can be driven out the sidewall, giving complete p-type activation. Hydrogen does not diffuse through n-type gallium nitride, so this works only for micro-LEDs. The result: external quantum efficiencies of 55% at low current densities and wall-plug efficiencies of 52%.
In this short video, you can learn:
* Calibrated integrating spheres reveal a 25% improvement from an omnidirectional reflector for micro-LEDs from five to 100 microns, versus no DBR.
* Removing about 10% absorbing ITO and replacing it with an n-type gallium nitride tunnel junction reduces chip loss, an approach used in gallium arsenide VCSELs.
* Hydrogen passivation can be driven out the sidewall in micro-LEDs to activate p-type dopant, enabling 55% EQE at low current densities and 52% wall-plug efficiency.
📋 **Clip Abstract** This clip shows calibrated micro-LED efficiency data, including a 25% omnidirectional reflector gain and about 28% at 500 amperes per centimeter. It then explains removing 10% absorbing ITO for an n-type gallium nitride tunnel junction, using micro-LED sidewall hydrogen out-diffusion to activate p-type GaN, reaching 55% EQE and 52% wall-plug efficiency.
About the speaker:
* Speaker: Steve Denbaars
* Company: University of California Santa Barbara
* Event: Eindhoven 2026
* Location: High Tech Campus, Eindhoven
#ExternalQuantumEfficiency, #IndiumTinOxide, #TunnelJunction, #WallPlugEfficiency, #MicroLEDDisplays, #OpticalCommunication