LED/OLED EQE Curves: Efficiency Droop & Roll-Off

How do you read an external quantum efficiency versus current density curve for an LED or OLED? Find the actual EQE denominator—electrons supplied to the device—then find the photon-output measurement used for its numerator. Read the current-density scale, locate the maximum or plateau, and calculate any efficiency roll-off against an explicitly defined reference. A falling EQE curve means fewer emitted photons per injected electron under the specified conditions; it does not uniquely identify the physical loss mechanism.

People searching for OLED EQE roll-off, LED efficiency droop, EQE versus current density, maximum external quantum efficiency, or why light-emitting diodes lose efficiency at high brightness need a method for separating measurements from mechanisms. The apparent peak can shift with voltage sweep direction, optical calibration, pixel area, pulse duration, temperature and how ‘roll-off’ is reported. This guide shows what the graph can establish before examining why the physics might change.

The central distinction is between a ratio and an absolute output. A device may continue emitting more light as current increases while its quantum efficiency falls. Reading that correctly is the beginning of an accurate comparison. The chapters below cover axes, denominators, device physics, worked arithmetic, diagnostic tests, cross-paper comparability and precise technical wording.

Define external quantum efficiency before interpreting it

External quantum efficiency in an electrically driven light-emitting device is commonly expressed as photons emitted from the device into the designated external region per electron delivered through the electrical circuit. The phrase external is important: it distinguishes photons that escape the structure from excitations that form internally or photons that never reach the outside world. Device architecture and measurement geometry may influence what is counted. Authors may estimate total external photons through integrating-sphere measurements, angle-resolved spectra or calibrated optical models. A photodiode pointed at one pixel does not automatically count every emitted photon.

Write the numerator and denominator as words before using a percentage. If an experiment counts 15 photons emitted to the exterior for every 100 injected electrons under a well-defined regime, its reported EQE is 15 percent. This simple interpretation does not say that only 15 percent of electrical energy became light; photons and electrons are counted, while energy efficiency also depends on voltage and photon energy. It also does not say how many excitations were generated inside the film. Those are separate quantities. Without this definition, discussion of efficiency droop easily becomes confused with brightness, power efficiency or luminous efficacy.

Distinguish EQE from internal quantum efficiency

Internal quantum efficiency is concerned with photons generated inside the active material or device per relevant input carrier or excitation event, depending on the system and definition. External quantum efficiency additionally reflects photon extraction from that structure. Light may be trapped in substrate modes, waveguide modes or plasmonic losses, and escape probability can depend on emitter position, dipole orientation, layer thickness and viewing direction. A device may therefore have excellent internal emission yield but considerably lower external photon output.

An EQE reduction with current could reflect nonradiative recombination, reduced carrier balance or a change in optical outcoupling, among other possibilities. It cannot simply be equated to a loss in intrinsic emitter photoluminescence quantum yield. If the paper reports both PLQY and EQE, read them as complementary experiments performed under different excitations and environmental conditions. Optical pumping can produce a different population of charges and excited states from electrically driven operation. Comparing the numbers may help identify hypotheses, but a mechanistic conclusion still needs an appropriate model and controls.

The meaning of efficiency droop and roll-off

Efficiency droop is the decline in conversion efficiency as drive current or current density increases, a phrase widely used in LEDs. Efficiency roll-off is common in OLED literature and often describes a fall in EQE or a related metric at higher drive or brightness. The words may overlap in practice but should not erase the plotted definition. A published ‘roll-off of ten percent’ might refer to a ten-percent relative reduction from maximum EQE, not a drop by ten percentage points. For example, a decline from 20 percent EQE to 18 percent EQE is a two-percentage-point absolute decline but a ten-percent relative decrease.

Ask what baseline the author used. The peak efficiency, a low-current plateau, an efficiency at a specified luminance, and an extrapolated zero-current value are not interchangeable. If a graph has no clear maximum because the lowest measured current is already past the peak, a quoted roll-off relative to that truncated range may underestimate the true decline. Report the chosen reference and the evaluated current or luminance along with the percentage. This habit prevents the same word from supporting incompatible numerical claims.

Read the current-density axis and its units

Typical plots use mA/cm², A/cm² or A/m², sometimes on a logarithmic axis. One mA/cm² equals 10 A/m², while one A/cm² equals ten thousand A/m². A fivefold separation on a logarithmic scale may occupy little horizontal distance. When comparing devices, pay attention to whether the axis represents total current, nominal current density through a pixel, or a different normalization such as current per active emitter area. The denominator matters for claims about local carrier density and bimolecular losses.

Current density alone does not guarantee the same exciton density in two stacks. Charge balance, recombination-zone width, layer thickness, transport and confinement all change where excited states form. A device producing the same nominal current density may concentrate recombination in a narrow region, making local interactions stronger. Conversely, a broader recombination zone can reduce local excited-state density even at identical J. This is one reason slope and peak position are clues rather than unique signatures of a particular annihilation pathway.

Read the vertical EQE axis without visual distortion

EQE can be plotted as a percentage, a fraction, or occasionally normalized to its own maximum. A curve labelled ‘normalized EQE’ does not reveal the absolute conversion yield unless the peak value is separately supplied. A curve reaching unity may simply mean 100 percent of the plotted reference, not 100 percent of electrons producing external photons. Check for linear versus logarithmic scaling and where the plotted range begins. A truncated y axis can make a small relative loss look dramatic, while a wide axis can conceal a meaningful device-level drop.

Some graphs overlay EQE with luminance, current efficiency or power efficiency on different vertical scales. Their curves can rise and fall at different conditions. Never assign a point from the efficiency line to the luminance axis just because they cross visually. Extract numbers from the correct labels, then note the current density and brightness corresponding to the same point. When a paper summarizes performance in a table, confirm whether the listed EQE is the maximum observed, the value at a practical luminance or the result at a specific current. Those are different performance statements.

Establish the optical photon count

A photon-count-based efficiency requires light measurement that can be translated into numbers of photons. Spectral radiance or radiant flux combined with wavelength-dependent photon energy is relevant, because each photon carries energy hc divided by wavelength. A green photon and a red photon carry different energies; counting watts of light without spectral correction is not the same as counting photons. The method must also account for the portion of light collected from the device, whether through angular integration or a justified emission model.

When a paper reports EQE derived from a photometer and an assumed Lambertian angular pattern, that assumption deserves mention. Top-emitting microcavity OLEDs may emit directionally enough that the assumption influences the value. A calibrated integrating sphere may estimate total emitted flux more directly, but its geometry, self-absorption, detector response and port effects still create uncertainty. Treat ‘EQE measured’ as shorthand for a documented chain of measurements and calculations. The quality of that chain determines whether small differences between samples can be trusted.

Electrical charge counting is another measurement

The electron-input count normally comes from measured electrical current integrated over time and divided by the elementary charge. This sounds straightforward until the sample exhibits leakage, capacitive transients, intermittent shorts or an unstable active area. In a rapid pulse, measured current may include a displacement component that does not correspond directly to steady carrier injection leading to emission. Comparing instantaneous optical response with total measured current without timing care can distort a calculated ratio.

Under steady forward operation, the current interpretation is often less ambiguous, but parasitic paths remain possible. Check whether the study uses measured device current rather than merely an instrument setpoint. If two devices have different leakage fractions, their external photons per total supplied electron can differ even with similar emission-layer physics. A complete mechanism analysis should distinguish a loss in electrical charge utilization from intrinsic quenching of excited states. The EQE curve records the overall outcome, not a direct count of each microscopic pathway.

Identify the peak, plateau and high-current region

A well-resolved EQE curve may rise at low current, reach a maximum and then decline at higher J. The rising section can reflect improved charge balance or changes in recombination probability as injection develops. The central plateau suggests efficiency is relatively stable over that measured interval. The descending section defines an operating region where photon output becomes less proportional to supplied charge. The shape is informative, but the three regions do not map automatically onto three unique chemical or electronic mechanisms.

If a paper claims an unusually high maximum, inspect how many points support it. A single noisy low-current measurement can create an artificial peak. If the peak occurs at such low brightness that the optical signal approaches a noise threshold, an application-level comparison at useful luminance may be more revealing. Conversely, a device with modest peak EQE but a broad plateau can outperform a high-peak device when operated brightly. Readers should preserve both the maximum and the operational efficiency profile rather than reducing the entire curve to one highlighted number.

Why EQE may rise at the lowest current

Low-current behaviour can be limited by unbalanced carrier injection, trap-assisted processes, leakage, recombination-zone placement or optical signal-to-noise. As drive increases, charge supply can become more balanced and a larger fraction of injected carriers can reach radiative recombination. In such a device, EQE rises to its maximum. Yet the initial increase cannot be universally assigned to improved injection because the spectrum, contact conditions and traps might also change. Some devices show a nearly flat curve at low current; others show noisy values or a monotonic decline.

The reading rule is to distinguish the raw trace from the proposed model. Check whether the authors provide J–V–L curves, current-voltage fits, transient electroluminescence, temperature dependence or simulations constrained by independent measurements. A mechanism that explains several separate observations is stronger than one chosen only because it resembles a textbook sketch. When summarizing, say the efficiency increased over the specified low-current range and name any tested explanation as a proposal unless the evidence establishes it more firmly.

Triplet–triplet annihilation is not an automatic diagnosis

In some phosphorescent OLEDs, long-lived triplet excitons can interact at high density. Triplet–triplet annihilation can reduce the number of excitations available for radiative emission, depending on the microscopic system and reaction products. Such interactions become more likely when excited-state populations build up in a confined region. A descending EQE curve is consistent with density-dependent loss of this kind, but it is not uniquely diagnostic. The same curve can emerge from different processes or mixtures of processes.

Evidence for triplet–triplet annihilation may come from lifetime changes, time-resolved emission, excitation-density dependence and appropriately tested kinetic models. The recombination-zone width matters because the local concentration of triplets can differ even at identical electrical current. Literature comparing conventional and graded emitting layers has shown that spatial distribution can alter roll-off. That physical lesson is more general than any single model fit. Avoid calling every high-current loss ‘TTA’ when a competing charge-balance or triplet-polaron explanation has not been examined.

Triplet–polaron quenching provides another possibility

Polaronic charge carriers and excited triplet states can interact, allowing nonradiative loss channels that reduce emission yield. Triplet–polaron quenching is frequently considered alongside triplet–triplet annihilation in OLED studies. Distinguishing them can be difficult because their contributions depend on charge and exciton densities that both vary with drive. A model may fit a set of steady-state EQE data with more than one plausible parameter combination.

Time-resolved electroluminescence, photoluminescence under bias, magnetic-field measurements and comparison of stacks with varied recombination profiles can help constrain the mechanism, though each technique introduces its own assumptions. Published studies have reached different dominant-mechanism conclusions for different device architectures. That is not necessarily a contradiction: materials and spatial profiles genuinely differ. A careful reader should identify which device class the mechanism claim concerns and what independent observables support it. Report ‘consistent with triplet–polaron quenching’ where that is the level of evidence, not ‘proved by the EQE curve’.

Singlet-related loss and other excited-state interactions

Fluorescent systems can involve singlet excitons, triplets and conversion between excited-state populations. Depending on the emitter and architecture, singlet–singlet annihilation, singlet–triplet annihilation, delayed fluorescence pathways or other nonradiative processes may alter the brightness-to-current relation. A time-delayed emission component is particularly relevant when excitations persist long enough for density-dependent interactions. But the presence of delayed light does not identify a specific annihilation rate from a single steady-state EQE plot.

To compare mechanisms, look for independent temporal signatures, concentration dependence and controls that change one interaction more strongly than another. Models should state whether they assume a spatially uniform exciton population or allow a distributed recombination zone. If a fitted constant changes dramatically when the model becomes more realistic, its earlier value may not represent a stable physical property. The goal of graph interpretation is not to memorize a single mechanism name; it is to understand which observables would distinguish plausible explanations.

Auger recombination and inorganic LED droop

In many inorganic LEDs, efficiency droop has been studied through mechanisms including Auger recombination, carrier leakage, localization effects and other loss processes. An Auger process transfers recombination energy to a carrier rather than emitting a photon. In suitable conditions, its contribution can increase as carrier density rises. Yet observing an EQE decline alone cannot separate Auger loss from carrier overflow, heating or changes in injection efficiency. Device design, alloy composition and confinement conditions matter.

Do not transfer a mechanism from one LED technology to another purely because both plots bend downward. A phosphorescent OLED and a III-nitride LED can have similar-looking efficiency traces for fundamentally different reasons. Responsible scientific translation preserves the material system and the evidence route. When authors use ‘droop’ as an umbrella term, interpret it as the observed reduction in efficiency with increasing drive until specific physical measurements are introduced. A high-quality explanation keeps the experimental result separate from the ongoing debate over which losses dominate in a given device.

Charge imbalance can make efficiency fall

Light emission usually requires appropriate electron and hole supply into an effective recombination region. If increasing voltage or current changes one carrier’s injection much more than the other, more injected charge can be lost through leakage, interface accumulation or nonradiative pathways. The resulting external photon yield per electron may fall even if intrinsic radiative probabilities inside the emitter stay comparatively stable. This is a charge-utilization issue as much as an excited-state issue.

To test charge balance, researchers may vary transport layers, blocking layers, thicknesses or injection contacts. However, those changes often also alter electric fields, morphology and optical cavity conditions, so causal attribution requires careful controls. A lower drive voltage and a lower EQE roll-off do not necessarily arise from the same improvement. Compare relevant electron-only and hole-only experiments, recombination-zone probes and optical data where available. Avoid using ‘improved balance’ as a universal explanation for every smoother efficiency curve unless that balance was actually measured or well constrained.

Recombination-zone width changes local exciton density

A physically important distinction is nominal drive density versus the density of excitations in the places they form. If recombination occurs in a very thin region, the local density can be high even at moderate overall current. Broader spatial recombination can reduce collision probabilities and some density-dependent loss mechanisms. Layer grading or tailored transport can change this distribution. Therefore two OLEDs measured at the same J can have different efficiency roll-off despite having similar total current.

An EQE curve alone does not show a spatial recombination profile. To make a statement about zone broadening, researchers need spatial probes, carefully designed sensing layers, device simulations constrained by data or other diagnostic experiments. The curve may motivate the hypothesis, but not verify the geometry. When teaching the interpretation, distinguish current per macroscopic area, excitons per active volume and lifetime of those excitons. These variables influence each other without becoming the same measurable quantity.

Thermal droop and device temperature

Electrical power that does not leave as useful light contributes to heating or other energy-transfer pathways. The sample temperature can then influence carrier transport, emission probability, traps and material stability. In high-current experiments, EQE measured under short pulses may differ from EQE under continuous drive because average thermal loading differs. A device operating on a heat sink may also behave differently from an otherwise identical free-standing sample.

Look for temperature-controlled measurements, pulsed-versus-continuous comparisons and power-density reporting. An EQE decline that strengthens with temperature may support thermal involvement, but temperature also changes several mechanisms at once. Do not diagnose ‘Joule heating’ from a downward curve without temperature context. Equally, absence of a visible temperature measurement should prevent categorical claims that heating played no role. A good performance comparison states ambient and, if available, junction or active-layer temperature, duty cycle and measurement duration.

Optical outcoupling can depend on device state

EQE combines internal excitation processes with the probability that emitted photons escape to the measured external region. Optical cavity effects can shape the emission spectrum and angular distribution. Changes in recombination-zone position, emitter dipole distribution or layer refractive properties can therefore influence detected optical output. In most simple comparisons these effects are treated as part of the device’s optical design, but they should not be forgotten when interpreting a current-dependent EQE curve.

If the emission pattern changes with current, an EQE calculated from a fixed on-axis assumption may acquire an artificial trend. Angle-resolved electroluminescence can help test this possibility. Spectral changes may also affect detector calibration. A narrow on-axis collection aperture could record a decline even while total integrated photon output behaves differently. This does not mean every roll-off report is wrong. It means the measurement model and calibration must support the degree of mechanistic certainty claimed. External performance is intrinsically a system-level observable.

Read efficiency against luminance, not only J

An OLED EQE-versus-current curve tells how performance changes with electrical drive density. An EQE-versus-luminance curve tells how efficiently it operates at a visually defined output. Both can be valuable. In applications that require a target brightness, comparing EQE at 100 or 1000 cd/m² may be more meaningful than comparing maxima at very dim conditions. Yet luminous output itself depends on the spectrum and viewing geometry, so matched cd/m² is a photometric rather than equal-photon condition.

Before ranking device designs, identify the application. Displays prioritize visibility, colour stability, power and operating lifetime in addition to quantum efficiency; specialized photonic devices might prioritize photon flux or spectral properties. A material that shines at low current but droops rapidly may be excellent for low-intensity sensing yet poor for sustained high-brightness lighting. Keep the use case explicit and avoid announcing a universal winner from one curve. Scientific comparison becomes stronger when it names the quantity held constant.

Relative roll-off: worked arithmetic

Consider an illustrative OLED with maximum EQE of 24 percent and EQE of 18 percent at a specified high current density. The absolute decrease is six percentage points. The relative roll-off from the measured peak is (24 minus 18) divided by 24, which equals 0.25, or 25 percent. Both numbers may be useful, but they answer different questions. If a summary simply says ‘efficiency dropped by six percent’, it could be interpreted incorrectly as a relative change.

Now imagine a second device with maximum EQE of 18 percent and a high-current value of 16 percent. Its absolute decline is two percentage points and its relative decline is about 11.1 percent. It has less relative roll-off but a lower high-current EQE than the first device, which still delivers 18 percent. Thus lower droop does not automatically mean higher operational photon conversion. Report both absolute EQE and the relative metric at the same operating point. This example also shows why a normalized-only curve can mislead readers about practical performance.

Define J90 and other retention metrics

Some studies express an efficiency retention point, such as J90, the current density where EQE reaches 90 percent of a reference value, often the peak. Its usefulness depends on a precise definition. If efficiency oscillates near its maximum or the measurement starts above the true peak, the extracted threshold may depend on smoothing and interpolation. A quoted J90 cannot be fairly compared with another paper’s J80 or with a luminance-based L90 without understanding the definitions.

When a threshold is extracted from discrete points, record whether it was interpolated and whether the curve is monotonic over that region. If the current axis is logarithmic, interpolation in log-current and linear current may produce different estimates. Report the method when precision matters. A retention metric compresses a curve for convenient comparison but discards other features, including low-current performance and the shape beyond the threshold. Use it as a summary coordinate, not as a replacement for the full plotted range.

Do not confuse roll-off with operational ageing

Efficiency roll-off describes how efficiency varies with drive condition, usually within a characterization sequence. Operational ageing describes how device performance evolves over time during use. A device can show pronounced instantaneous roll-off but good lifetime at moderate brightness, or limited roll-off yet poor chemical stability. Measuring current-dependent EQE during a voltage sweep is not the same as monitoring a device under constant-current operation for hundreds of hours.

Ageing may still contaminate a roll-off experiment if the device deteriorates during the scan. Repeat the scan in both directions, monitor spectra and consider rest or recovery protocols where scientifically justified. If high-current points are measured after prolonged stressing, a falling ratio might combine genuine current-density dependence with accumulated degradation. Time stamps and measurement ordering are therefore part of the curve’s meaning. To keep terminology precise, reserve ‘roll-off’ for drive-dependent efficiency behaviour and use ‘degradation’ or ‘lifetime decay’ for changes over elapsed operating time.

Correct for optical spectral response

Spectrometers and photodiodes have wavelength-dependent sensitivities. If an OLED spectrum shifts or changes width as current increases, a detector calibrated only at one wavelength can create a false efficiency trend. Proper photon accounting corrects spectral responsivity and integrates photon flux rather than assuming every optical count represents the same number of photons. When the spectrum is narrow and stable, the numerical impact may be modest; when colours differ, it may become substantial.

Check whether the authors calibrated across the relevant emission band, not merely at its centre. Spectral sidebands can contribute additional photons even when they look weak in a normalized plot. Angular dependence can change spectral shape as well. A convincing EQE method specifies detector and collection geometry, calibration standards and any optical model. Precision requires the chain from raw photons to published percentages to remain auditable. This is especially important for record claims or small improvements that might fall within calibration uncertainty.

Pixel area and current spreading

The current-density axis generally divides current by a defined geometric area. If current crowds near an electrode contact, local density can be higher than nominal J. Some pixels may emit nonuniformly, leaving part of the supposed area dark. A smaller truly active region raises local excitation density and can enhance density-dependent losses even if the reported nominal current density seems unchanged. Device size comparisons are therefore not automatically apples to apples.

Images of electroluminescence uniformity, alternative electrode layouts and area-scaling studies can reveal whether current spreading is a likely contributor. In some research architectures, contact resistance and transparent-electrode sheet resistance also influence distribution. A high-current EQE decline could be amplified by local hot spots. The careful reader uses the published geometric J as it is defined while avoiding unwarranted statements about microscopic uniformity. Data that report emission maps alongside EQE help close this interpretation gap.

Measurement dwell time and sweep direction

At every current-density setting, instruments need time to settle and to collect optical signals. Long dwell can cause temperature changes or ageing; short dwell can capture non-steady carrier or exciton dynamics. A rising-current scan and falling-current scan may therefore differ. When the curves are treated as steady-state performance, look for evidence that the dwell time was sufficient for the relevant transients, without allowing unacceptable drift.

Some authors use pulsed current with time-gated detection; others hold each point at constant current. Both can be appropriate for different questions, but they do not necessarily produce identical high-drive EQE values. A dataset obtained from the most favourable short pulse should not be silently compared with a continuous measurement under thermal load. Record pulse width, duty cycle, timing of optical integration and cooling conditions. This transforms an unexplained discrepancy between papers into a concrete difference in methods that readers can evaluate.

Carrier recombination and spectra are complementary

An efficiency curve alone gives a net photon-yield outcome. Electroluminescence spectra reveal which wavelengths were emitted and how the distribution changes with drive. Time-resolved emission can constrain population lifetimes and delayed components. Photoluminescence quantum-yield measurements probe emitter properties under optical excitation. Each method contributes different evidence about recombination and emission, but no one technique should be allowed to answer questions outside its scope.

For example, a red shift accompanying roll-off may suggest temperature or cavity changes but does not prove heating is the sole efficiency-loss cause. Shortening excited-state lifetime under bias may support quenching, yet charge balance still changes the number of excited states created. The strongest interpretations compare multiple quantities with a model explicit about its assumptions. In educational writing, it is useful to explain how the pieces interlock: J–V–L provides the electrical and photometric operating relationship, spectra describe emitted photons, and EQE summarizes external photon output per supplied charge.

Models should be tested against more than one curve

Many kinetic models fit EQE as a function of J using rate equations for excitation generation, radiative decay and competing loss channels. A good fit is necessary but not sufficient evidence of the assumed mechanism. Different combinations of carrier balance, annihilation constants and spatial distributions may produce similar steady-state curves. This is an identification problem, not merely a curve-fitting problem. Parameter uncertainty and correlation should be considered alongside visual agreement.

Look for independent measurements that constrain rates or densities, sensitivity tests and predictions at conditions outside the fitted dataset. If a model fitted at room temperature also predicts trends under pulsed driving or altered layer structures, the interpretation grows more credible. If its central constant has to change arbitrarily between nearly identical conditions, the physical story deserves scrutiny. Explain to learners that a smooth line through data is not the same as a demonstrated mechanism. Mathematical flexibility can fit observations without uniquely revealing the underlying process.

Device-to-device variation and statistical confidence

A curve selected from a single champion device may not represent the typical performance of a material system. Reported maxima can vary across fabricated pixels because of thickness, contact quality, contamination, local defects, encapsulation and instrument uncertainty. A meaningful comparison specifies how many devices were tested, whether data represent an average or best specimen, and the distribution of efficiency at practical operating points.

When differences are small, statistical uncertainty matters as much as the curve shape. Suppose an EQE peak differs by one percentage point between two batches but device variability is several points. A claim of systematic improvement cannot be established from the highest individual trace alone. Independent fabrication runs and matched measurement protocols strengthen evidence. A reader may still describe the published champion values accurately; the error would be to translate one exceptional trace into a claim of reliable yield. The distinction between observed performance and reproducible improvement is central to engineering judgment.

What high peak EQE cannot tell you

A very high peak EQE can demonstrate favourable photon conversion at a particular drive condition. It does not necessarily mean low operating voltage, high luminous efficacy, stable colour, uniform emission, resistance to environmental damage or long lifetime. Those properties require separate measurements. It is possible to improve optical outcoupling and obtain higher EQE without solving electrical losses or chemical degradation. It is also possible for a device with slightly lower peak EQE to consume less power at application brightness.

Read the whole performance envelope. Find EQE at a reference luminance, voltage and J at that point, spectral stability over current, ageing tests and measurement uncertainty. A publication may emphasize peak quantum efficiency because that was the scientific question; it need not answer a commercial product question. Problems arise when external summaries imply otherwise. Good translation keeps a result attached to its experimental scope and refuses to inflate it into a comprehensive performance verdict.

Worked example: photon conversion versus brightness

Imagine an illustrative emitter whose EQE falls from 20 percent to 15 percent as current density doubles. The emitted photon rate per active area is proportional, under a consistent definition, to current density multiplied by EQE. At the first point the relative value is 1 times 0.20, or 0.20; at the second it is 2 times 0.15, or 0.30. Thus external photon output increased by 50 percent even though efficiency dropped by 25 percent relative to its starting value.

This example addresses a common misreading: droop does not necessarily mean the device emits fewer photons than before. It means the conversion yield worsens as electrical input increases. The relationship between photon output and luminance depends additionally on wavelength and angular emission, so the same numbers do not prove a 50-percent increase in cd/m². State which output is being counted. If the experiment involves changing spectral colour, photometric and quantum metrics may diverge further. The careful interpretation survives both effects by keeping denominators visible.

Worked example: why percentage points matter

Suppose an OLED reports EQE values of 22, 20 and 16 percent at successive current densities. From the first to the third point the absolute loss is six percentage points, while the relative loss is six divided by 22, about 27.3 percent. If another device declines from 12 to 10 percent, its absolute loss is two points and relative loss about 16.7 percent. It is possible to describe the second device as having a smaller relative roll-off even though its performance stays below the first across all three points.

A useful comparison should present peak EQE, EQE at a named operating point and the reference used for relative droop. Adding current density and luminance removes still more ambiguity. If one author quotes a 27-percent roll-off at 1000 cd/m² and another quotes a 17-percent roll-off at 100 mA/cm², those statements are not directly ranked until their conditions are aligned. Good numeracy is not difficult arithmetic here; it is choosing consistent definitions before doing elementary arithmetic.

Worked example: matched current, unequal local density

Consider two idealized pixels each receiving 30 mA/cm² by nominal geometric area. Device A distributes radiative recombination through a broad effective emitting region, while Device B concentrates excitations in a narrower zone. Even if the same total number of charge carriers enters each pixel per second, the local exciton densities may differ significantly. Density-dependent reactions can then produce stronger loss in B, giving a lower EQE at the same nominal J. The curve alone would show the outcome but not reveal the spatial cause.

This thought experiment explains why designs that redistribute recombination can reduce efficiency roll-off. It should not be mistaken for a measurement of layer width or exciton concentration. Evidence must establish those quantities independently. If a paper attributes improved high-current EQE to broadening of the recombination zone, ask how the profile was measured or constrained. The most useful teaching question is: Which quantity is the same between samples, and which local quantity may still differ? That prevents a normalization from being confused with an equal physical state.

Interpret a curve with uncertainty bands

Imagine that a paper plots average EQE with a shaded region showing one standard deviation across devices. At high current, Device A averages 18 percent and B averages 19 percent, with uncertainty ranges that overlap. It would be accurate to say B has a slightly higher observed mean under the specified test. It would not automatically be accurate to say that B is demonstrably superior in every fabrication batch. The degree of statistical confidence depends on sample counts, distribution, independence and the meaning of the bands.

Calibration uncertainty may also move both curves in the same direction without appearing in device-to-device variability. This distinction matters when papers show very narrow replicate error bars but use the same potentially biased detector calibration for all samples. A strong comparison names systematic and random uncertainty where possible. If those details are absent, choose language proportional to the visible evidence. Precision in science is not achieved by asserting certainty; it is achieved by making the limits of the measurement legible.

Translation and technical vocabulary traps

Words such as droop, roll-off, quenching, degradation, efficiency, efficacy, radiance and luminance have overlapping everyday associations but different technical meanings. A learner or translator should not treat them as interchangeable. ‘Droop’ refers here to drive-dependent efficiency decline; ‘degradation’ usually describes change over elapsed operation; ‘quenching’ is a candidate physical loss process. ‘External quantum efficiency’ is a photon-per-electron ratio, not lm/W or cd/A. Maintaining these definitions makes a technical explanation shorter and more accurate.

Build a small two-column glossary in your notes: each term beside the quantity actually measured or modelled. Then test boundary cases. Could luminance increase while EQE decreases? Yes. Could EQE stay constant while voltage rises? Yes, at least in principle. Could an emission spectrum remain stable while current efficiency changes? Yes. Counterexamples teach meaning more robustly than memorizing synonyms. When communicating across languages, preserve the variables, units and stated conditions before choosing stylistically elegant wording. Technical accuracy begins with conceptual alignment rather than literal word replacement.

An efficient reading route for unfamiliar papers

Start with the figure caption, device structure and definitions. Locate the y-axis EQE metric, the x-axis drive metric and the values at two or three operating points. Ask whether optical output was spectrally and angularly calibrated, whether the current is steady or pulsed, and whether the result represents one device or a distribution. Only then read the authors’ mechanism discussion. This order prevents the paper’s narrative from becoming a substitute for your independent observation.

Next compare the EQE graph with J–V–L, electroluminescence spectra, thermal information and time-resolved results. If the claimed mechanism predicts a trend that a complementary plot does not show, note the tension. If the evidence is incomplete, state what additional measurement would help. Finish with one sentence describing the measured trend and one sentence naming the best-supported explanation at its correct confidence level. That small discipline improves journal reading, technical writing, peer discussion and scientific translation.

Frequently asked question: is efficiency droop always harmful?

Efficiency droop reduces photons produced per supplied electron over the affected current range, so it can matter for energy use and thermal management. Whether it is decisive depends on the intended operating condition. A sensor or low-brightness display may work mainly near a high-efficiency plateau, where high-current droop rarely determines real use. A lighting source designed for sustained bright operation may be much more sensitive to the same decline. Device optimization is therefore an application-dependent tradeoff rather than a competition for one abstract curve shape.

Excessive current may also accelerate ageing, but instantaneous droop alone is not a lifetime measurement. Assess operating voltage, total power, spectral quality and longevity separately. A sharp efficiency decline should motivate design investigation rather than automatically disqualifying a material. Reporting both peak EQE and EQE at a relevant target output gives readers enough information to evaluate the practical significance.

Frequently asked question: can EQE exceed 100 percent?

For a conventional electrically pumped LED defined as one externally emitted photon per injected elementary charge carrier, values above unity require special care and may indicate a different counting convention, physical multiplication process, or measurement error. Some specialized devices can use gain or multiple-photon mechanisms under specific conditions; the correct interpretation depends on precisely what input event and output event are counted. A number above 100 percent is not automatically physically impossible across every conceivable photonic system, but it is not routine evidence of a conventional single-exciton LED operating normally.

Before accepting an extraordinary EQE, inspect charge accounting, device leakage, detector calibration, luminescence background and the definition of electron flux. Compare the result with spectra, energy balance and independent repetitions. Especially when a graph comes from a technology unfamiliar to the reader, the correct first question is what the authors put in the numerator and denominator. Understanding the accounting resolves many apparent paradoxes before speculative physics is necessary.

Frequently asked question: why does EQE differ between measurement laboratories?

Differences can arise from optical collection angles, assumptions about angular emission, spectral calibration, pixel area, drive mode, measurement duration, temperature and data processing. A laboratory integrating total optical flux may obtain a different value from one using a fixed forward photodiode with a Lambertian approximation. Neither measurement should be judged from its headline number alone; each should be understood through its traceable method and uncertainty.

Cross-laboratory comparisons are strongest when common reference devices, calibration procedures and operating points are used. In the absence of harmonized methods, describe findings as measured under their own stated protocols. Avoid decimal-level rankings that ignore systematic differences. This is particularly relevant when discussing a narrow performance record or an incremental improvement. A result is useful when readers understand the conditions under which it was obtained.

A concise reporting model

A defensible result summary could read: ‘The device reached a maximum external quantum efficiency of the reported value at the measured current density. EQE declined to the reported operating value as current density increased, representing a specified relative roll-off from that peak. The change indicates reduced external photon output per supplied electron under these conditions. The proposed quenching or transport explanation is supported by the additional measurements described separately.’ This form keeps observation, arithmetic and mechanism in separate clauses.

If the paper reports only normalized EQE, remove absolute efficiency claims. If measurements are pulsed, mention their time profile. If the curve was recorded during ageing, state the operating history. If different spectral detectors or angular assumptions were used, note the comparison limit. Concise technical writing is not vague writing; it is explicit about which assumptions support which conclusions. The reader should be able to reconstruct both what was measured and what remains uncertain.

How to Translate Easily | How to Read LED and OLED EQE–Current Density Curves, Efficiency Droop and Roll-Off