Efficiency droop is the single most consequential unsolved problem in LED semiconductor physics. Every LED manufacturer, every photonics researcher, and every lighting engineer who has pushed a device to high drive currents has encountered it: the puzzling, reproducible, and commercially devastating decline in efficiency as current increases. Understanding precisely why it happens—and what can be done about it—determines whether UV LED technology becomes commercially viable at scale.
The phenomenon is straightforward to observe but took decades to explain convincingly. Apply increasing current to an InGaN blue LED. At low current densities (below roughly 1–5 A/cm²), efficiency climbs as expected. Then it peaks. Then, as current continues to rise toward the 20–50 A/cm² range typical of real applications, efficiency falls. Not catastrophically, not linearly, but in a smooth, seemingly unavoidable curve downward. The LED still emits light—sometimes more light in absolute terms—but the fraction of injected electrons converted to photons collapses.
The ABC Rate Equation Framework
To understand droop, you need the ABC model—the standard framework for describing recombination processes in LED active regions. The model categorizes carrier recombination into three competing pathways, each described by a rate coefficient:
- A coefficient (Shockley-Read-Hall, or SRH recombination): Non-radiative recombination at crystal defects and impurities. This pathway dominates at very low carrier densities and limits efficiency in defect-rich materials like high-aluminum AlGaN. Rate scales linearly with carrier density n:
An. - B coefficient (Radiative recombination): The desired process—an electron and hole recombine and emit a photon. Rate scales as
Bn², meaning efficiency climbs with carrier density as this process increasingly dominates over SRH. - C coefficient (Auger recombination): A three-body non-radiative process that dominates at high carrier densities. Rate scales as
Cn³. This term becomes dominant at the carrier densities achieved under high drive current, and it is the primary driver of droop.
The internal quantum efficiency (IQE) of the active region is thus:
The IQE Expression
IQE = Bn² / (An + Bn² + Cn³). At low carrier density, the An term limits efficiency. At intermediate density, Bn² dominates and efficiency peaks. At high density, Cn³ takes over and efficiency drops. This cubic dependence on carrier density is why droop becomes catastrophic rather than gradual at high currents—doubling current density more than doubles the Auger loss rate.
Auger Recombination: The Primary Culprit
The Auger process transfers the energy of an electron-hole recombination event to a third carrier (either another electron or another hole) rather than emitting a photon. That third carrier then loses its energy as heat through phonon emission. The net result: one electron and one hole consume one photon's worth of energy, produce zero photons, and heat the junction instead.
In Auger recombination, three carriers must simultaneously occupy the same spatial region for the process to occur—hence the cubic dependence on carrier density. At the carrier densities required for 1-watt-class LED chips (roughly 10¹⁸–10¹⁹ cm⁻³), Auger events become frequent enough to significantly deplete the radiative recombination budget.
The Auger coefficient C in InGaN is surprisingly large—roughly 10⁻³⁰ to 10⁻²⁸ cm⁶/s, several orders of magnitude higher than what classical Auger theory predicts for a material with this bandgap. The discrepancy was a source of significant scientific controversy for years. Current understanding attributes the enhanced Auger rate to phonon-assisted indirect Auger processes enabled by alloy disorder in InGaN, and to carrier delocalization effects that increase the effective carrier density in regions where recombination occurs.
Carrier Overflow: The Second Mechanism
Auger recombination does not tell the complete droop story. A second significant mechanism is carrier overflow—high-energy carriers that escape the quantum well active region before recombining.
LED active regions typically consist of multiple quantum wells (MQWs), thin layers (2–5 nm) of lower-bandgap material sandwiched between barrier layers. At low current, carriers thermalize into the wells and recombine. At high current, the injection rate exceeds the capture rate, and carriers spill over the wells into the p-cladding layer. Once in the p-cladding, they recombine non-radiatively or recombine far from the intended active region, contributing nothing to photon output.
An electron blocking layer (EBL)—a thin, high-aluminum-content AlGaN layer between the last quantum well and the p-cladding—is the standard mitigation strategy. The EBL creates an additional potential barrier that reflects overflowing electrons back toward the active region. However, EBL design involves tradeoffs: too high a barrier impedes hole injection into the wells, reducing efficiency at all current densities. The optimization is genuinely non-trivial and remains an active area of device engineering.
Blue InGaN vs. Deep UV AlGaN: Droop Comparison
| Parameter | Blue InGaN LED (450 nm) | Deep UV AlGaN LED (265 nm) |
|---|---|---|
| Peak EQE | 75–85% | 3–8% |
| Droop onset current | ~5–10 A/cm² | ~1–3 A/cm² |
| EQE at 35 A/cm² | 50–60% of peak | 30–50% of peak |
| Primary droop driver | Auger recombination | Auger + defect-mediated SRH + TM polarization loss |
| Defect density (typical) | 10⁸–10⁹ cm⁻² | 10⁹–10¹⁰ cm⁻² |
| Substrate | Sapphire or GaN-on-Si | Sapphire or AlN |
Data represents typical commercial device performance ranges; research devices may achieve better results
Why Droop Hits UV LEDs Harder
Blue InGaN LEDs achieved 80%+ EQE partly because the semiconductor community spent thirty years—and Nakamura, Amano, and Akasaki's Nobel Prize-winning work—solving the underlying material and device challenges. Modern LEDs are mature devices with well-understood defect chemistry and deposition processes. Deep UV AlGaN is decades behind.
High-aluminum-content AlGaN (above ~50% Al, needed for sub-280 nm emission) introduces problems that compound each other:
- Crystal quality: Growing AlGaN on sapphire or SiC generates threading dislocations at densities of 10⁹–10¹⁰ cm⁻², compared to 10⁸–10⁹ for InGaN. Each dislocation is a non-radiative recombination center that amplifies the A coefficient and accelerates efficiency loss at low current before droop even begins.
- TM polarization: High-Al AlGaN emits predominantly in the transverse magnetic (TM) polarization, meaning photons propagate parallel to the substrate surface rather than perpendicular to it. The result: a large fraction of generated photons undergo total internal reflection and never exit the chip, suppressing light extraction efficiency (LEE) to 5–10%.
- p-type doping: Efficient p-type doping of high-Al AlGaN is notoriously difficult. Magnesium dopants activate poorly, and the resulting high hole resistivity forces devices to operate at higher applied voltages, reducing wall-plug efficiency and increasing junction temperature—which then exacerbates droop through thermally assisted Auger processes.
The Compounding Effect
In a deep UV-C LED, you are fighting SRH recombination from defects, Auger recombination from high carrier density, carrier overflow from imperfect quantum well confinement, TM polarization from the crystal symmetry, and poor p-doping all simultaneously. Each of these, alone, would be manageable. Together, they limit UV-C EQE to a fraction of what blue LEDs achieve, and each one interacts with the others in ways that make optimization brutally non-linear.
Engineering Strategies to Reduce Droop
Pulsed Operation
Operating LEDs in pulsed mode—high peak current for microsecond bursts, with low duty cycles—exploits a simple fact: carrier density in the active region during a pulse is high enough to produce intense photon output, but the average junction temperature stays low because heat has time to dissipate between pulses. Peak IQE at a given instantaneous current is higher at lower junction temperatures, so pulsed operation can recover 10–20% relative efficiency compared to DC operation at the same average power.
For germicidal UV applications, there is an additional benefit: some research suggests pulsed UV doses may be more bactericidal per photon than continuous-wave doses of equivalent total fluence, possibly because bacterial DNA repair mechanisms require continuous light to activate. The quantum efficiency bottleneck in UV-C LEDs article explores these operational tradeoffs in detail.
Distributed LED Arrays
Rather than driving a single chip at 50 A/cm², designers can drive twenty chips at 2.5 A/cm² each and achieve the same total optical output while operating near the efficiency peak for each individual device. This "current spreading" architecture has obvious cost and assembly implications, but for high-value applications like pharmaceutical UV disinfection reactors or precision photolithography sources, the efficiency gain justifies the added complexity.
AlN Native Substrates
Replacing sapphire with native aluminum nitride (AlN) substrates reduces threading dislocation density in AlGaN films by one to two orders of magnitude. The lattice match between AlN substrate and AlGaN epilayers is far superior to the AlGaN-on-sapphire system. Research groups using AlN substrates have demonstrated UV-C LEDs with EQE exceeding 10%—roughly double the commercial standard. The barrier is cost: AlN wafers currently cost $500–$2,000 each versus $10–$50 for sapphire. Scale will bring this down, but it is not yet a production-line reality.
Nanophotonic Extraction Structures
Since TM polarization suppresses photon extraction in deep UV AlGaN, surface structuring can partially recover those trapped photons. Photonic crystal patterns etched into the LED surface create periodic refractive index variations that diffract otherwise-trapped photons into escape angles. Nanoscale roughening of the p-GaN or p-AlGaN contact layer serves a similar function. Demonstrated extraction improvements from these techniques range from 30% to 200% relative—a significant boost on top of the base EQE.
Droop Mitigation: Performance Benchmarks
Research Directions and Where the Field Is Heading
Three research threads show the most near-term promise for breaking the droop barrier.
Polarization engineering uses strain and composition gradients in the quantum well structure to shift emission from TM toward TE polarization without changing the emission wavelength. Several groups have demonstrated partial polarization conversion using graded AlGaN quantum wells. If successful at scale, this alone could double light extraction from deep UV-C devices.
Tunnel junction contacts replace the high-resistance p-AlGaN layer with a junction structure that allows electron injection without requiring efficient p-type doping. The DOE Solid-State Lighting program has funded multiple tunnel junction approaches for UV LEDs as part of its energy efficiency research portfolio. Early results show operating voltage reductions of 1–2 V, directly improving wall-plug efficiency.
Micro-LED architectures take the distributed array concept to its logical extreme: thousands of micro-scale (1–50 μm) LED pixels driven at very low individual current density. Each micro-LED operates near its efficiency peak. The aggregate output matches or exceeds a conventional large-chip device with dramatically less droop-induced loss. Micro-LED manufacturing at UV wavelengths is at an early research stage, but the concept is proven at visible wavelengths and the physics translates directly.
The AlGaN material system still has substantial room for improvement in crystal quality, doping control, and device architecture. Researchers at institutions from MIT and Stanford to RIKEN and EPFL are making incremental but steady gains. The trajectory of the last decade—roughly doubling of commercial UV-C LED efficiency every four to five years—suggests that the 20% EQE milestone for deep UV devices is achievable before 2030. At that point, the efficiency droop problem does not disappear, but its practical impact shrinks to the point where UV LED systems become cost-competitive with mercury across essentially all application categories.
The National Institute of Standards and Technology maintains photonics metrology facilities used to characterize next-generation LED devices with the precision necessary to validate these performance claims. Standardized measurement protocols are critical because droop characteristics depend heavily on measurement conditions: junction temperature, drive waveform, and aging state all affect the EQE curve significantly.