Quick Answer
UV-LED lifetime is not "hours until off", but "hours until 70 % of initial output" — the L70 value. Industry specification: high-quality UV-A LED modules reach 20,000–30,000 h L70, a typical medium-pressure mercury lamp 1,000–2,000 h. UV-C LEDs are not yet in this league — current research results for AlGaN far-UVC around 230 nm show ~1,500 h L70 at 100 mA / 25 °C with a 20-quantum-well design.
The single most important physical lever is junction temperature. For every +10 °C at the junction, the L70 lifetime roughly halves (Arrhenius effect). A UV-A LED with a datasheet L70 = 30,000 h at 25 °C, run in passive cooling at a 65 °C junction temperature, in reality only lasts ~7,500 h to the L70 point — four halvings.
Key practical consequences:
- Active cooling is mandatory for high-power modules — otherwise the datasheet lifetime becomes marketing fiction
- Periodic radiometer measurement (monthly to quarterly) is the only reliable maintenance strategy — "the lamp still glows" says nothing about UV output
- UV-C LEDs have a burn-in phase: in the first ~100 h output drops to 50–70 % of the initial value — this is material diffusion, not a defect. Datasheets usually quote the post-burn-in output; some quote the initial peak
1. What L70, L80, L90 Mean — and Why L70 Is the Standard
LED lifetime is not defined as "until fully dark" (LEDs rarely fail that way), but as a lumen-maintenance threshold:
| Marker | Definition | Use |
|---|---|---|
| L90 | Hours until output has fallen to 90 % | Display light sources, semiconductor lithography |
| L80 | Hours until output has fallen to 80 % | High-end lighting, critical applications |
| L70 | Hours until output has fallen to 70 % | Industry standard for UV-LED curing modules |
| L50 | Hours until output has fallen to 50 % | Rarely used; "half-life" marker |
The B extension (e.g. L70B10) is a statistical marker: B10 means "10 % of the LEDs are below the L70 point after this time". L70B50 would be "50 % are below it" — a less strict test than L70B10.
Industry practice when reading datasheets: if a manufacturer only states "lifetime" without an L marker and without a junction temperature, two essential parameters are missing. That is a legitimate reason to ask follow-up questions.
2. The Two Main Mechanisms of UV-LED Ageing
UV-LEDs age primarily through two physically distinct processes. Both run in parallel, but are thermally accelerated.
2.1 Thermal Ageing (Arrhenius Mechanism)
The junction temperature (T_j) drives several parallel degradation processes:
- Yellowing of the silicone/epoxy encapsulant (in UV-A modules)
- Intermetallic growth at solder joints
- Thermo-mechanical stress between chip and substrate (CTE mismatch)
- Phosphor quenching (in the few UV-A LEDs with conversion phosphors)
Arrhenius model: the rate of these processes roughly doubles per +10 °C T_j. In quantitative terms: L70 halves per +10 °C T_j.
In lifetime studies, the measured Arrhenius activation energies for UV-A LEDs at 365 nm are:
- 0.13 eV in continuous operation
- 0.20 eV in cycled (switched) operation
→ The cycled mode ages faster per effective radiation hour, because additional thermo-mechanical stress from the warm-up/cool-down cycles applies. Relevant for curing applications with high duty cycling.
2.2 Electrically Driven Defect Generation (AlGaN-Specific)
In UV-C LEDs (AlGaN with high aluminium content, wavelengths <280 nm) there is an additional ageing mechanism that does not play a dominant role in UV-A:
- Point-defect generation through electrical stress in the quantum-well region — defects capture charge carriers and remove them from the radiative process
- Hydrogen out-diffusion from the active region — hydrogen stabilises the acceptor doping; its diffusion away reduces the hole concentration
- Trap-assisted tunnelling — increased leakage current, though it is not the primary effect under standard driving conditions
These mechanisms explain why UV-C LEDs do not simply last as long as UV-A despite better chip quality: the higher Al content in the AlGaN makes the material more defect-prone. Only through substrate changes (AlN single-crystal instead of sapphire), quantum-well multiplication (20+ wells instead of 5–6) and clean epitaxy does UV-C reach the >1,000 h league.
2.3 UV-C-Specific Burn-In Phase
A peculiarity of UV-C LEDs: within the first ~100 hours the optical output drops abruptly to 50–70 % of the initial value. After that the output stabilises and the "normal" degradation path begins.
This is not a defect — it is material diffusion in the fresh epitaxy, primarily hydrogen migration. High-quality manufacturers therefore sell post-burn-in-stabilised modules and state this in the datasheet. Lower quality quotes the unstabilised initial peak as "irradiance" — a deliberate piece of marketing window-dressing that no longer holds after a week of operation.
Practical heuristic: with a new UV-C LED system, run it in for 100–200 h first, then take a baseline measurement with a radiometer for maintenance planning.
3. UV-A vs. UV-C — the Lifetime Asymmetry
| Wavelength | Typical L70 (industry module) | Main limit | Practical replacement cycle |
|---|---|---|---|
| 365 nm (UV-A) | 20,000–30,000 h | encapsulant yellowing + thermal | 2–5 years continuous operation |
| 395/405 nm (UV-A near-VIS) | 30,000–50,000 h | thermally dominated | 3–7 years continuous operation |
| 275 nm (UV-C E. coli) | 5,000–15,000 h | AlGaN defect generation | 1–2 years continuous operation |
| 265 nm (UV-C DNA peak) | 3,000–10,000 h | AlGaN + stronger burn-in | 0.5–1.5 years continuous operation |
| ~230 nm (far-UVC) | ~1,500 h (2024 research record) | massive defect rate | research / specialist use |
→ The lifetime gap between UV-A and UV-C is 5–10× at the current state of the art and is closing, but more slowly than the efficiency gap. Anyone planning a UV-C system should expect at least two module replacements over the operating life cycle — something that does not arise with UV-A.
4. When Is an LED "at the End"?
Three maintenance thresholds, depending on application tolerance:
4.1 Curing (Adhesives, Coatings)
Curing applications are dose-controlled. As output falls, the exposure time must be increased to reach the same dose (mJ/cm²). There are two failure modes:
- Output falls below the dose-achievable minimum → curing process fails
- Output still sufficient, but dwell time too long → cycle-time loss, production bottleneck
In both cases L70 is the right maintenance limit — beyond it the cycle-time penalty becomes untenable.
4.2 Disinfection (Water, Air, Surface)
Disinfection applications are also dose-controlled, but often governed by a validation constant: the system must deliver a validated mJ/cm² dose, otherwise hygiene compliance is breached.
Here L80 is the more common maintenance limit — the safety buffer to the validation threshold is tighter.
4.3 Exposure Systems (Photolithography, 3D Printing)
In 3D resin printing LCD/MSLA systems there are two wear parts:
- LED array (UV-A at 405 nm typical) — L70 after 5,000–15,000 h
- LCD exposure mask — lifetime ~400–1,000 print hours (per Prusa3D community discussion)
The LCD mask is usually the weaker link, not the LED. This is an important practical point for hobbyists: if the printer shows weak layers, it is usually the LCD mask, not the LED array.
5. Maintenance Strategy — a Radiometer Is Mandatory
From the industry maintenance handbooks (UVNDT, Gigahertz-Optik, Dymax):
5.1 Baseline + Periodic Measurement
- Baseline measurement with a calibrated UV radiometer directly after burn-in (UV-C) or after installation (UV-A) — this initial irradiance is the reference for all later maintenance measurements
- Periodic re-measurement: monthly for critical applications, quarterly for tolerant ones
- Mapping across the area: not just a single point — area emitters often age unevenly (hot spots run hotter → faster; multi-LED failure in the centre position)
5.2 Which Values to Watch
| Measurement | Meaning |
|---|---|
| Peak irradiance (W/cm²) | falls with ageing — the most reliable value |
| Spectral distribution | should stay stable — a shift points to a phosphor-conversion problem (rare in pure UV pumps) |
| Spatial uniformity | standard deviation across the area — if it grows, individual LED failures are beginning |
| Junction temperature (via NTC or thermal imaging) | indirect maintenance indicator — detects cooling degradation |
5.3 Immediate Actions on an Ageing Signal
When a power drop is detected:
- First — clean the optics (clean a yellowed silicone window, dust on lenses). Often "ageing LEDs" are in reality just dirty
- Second — check the cooling (fan, thermal paste, water-cooling flow rate) — degraded cooling simulates LED ageing
- Third — check the driver current (constant-current drivers can develop drift themselves as they age)
- Only then — module replacement
6. Where LUVEX Practice Differs from the Manufacturer Doctrine
- L70 without T_j is useless. Anyone buying a UV-LED module should tie the L70 guarantee to T_j conditions, not to a "datasheet value". At 60–70 °C T_j (typical in passively cooled high-power applications) the real L70 lies well below the datasheet — an L70 multiplier per +10 °C belongs in the design diagram, not in the fine print
- L70B50 vs L70B10 is not a detail. If a manufacturer states only "L70" without a B marker, the milder B50 test is most likely meant. In a system with 100 LED chips, B50 means that statistically 50 LEDs are already below L70 when the system is considered "still within the L70 range"
- UV-C burn-in is often concealed in marketing. Initial peak irradiance values are tempting in comparison tables, but after 100 h of burn-in the picture looks different. Fair manufacturers quote post-burn-in stabilised values; ask for them in the datasheet
- Cycling stresses, but continuous-on stresses too. There is no stress-free operating mode for UV-LEDs. Anyone planning a system with high duty cycling (e.g. inline printing with pulsed exposure) must ask for the cycled lifetime separately — it is usually not the datasheet default
7. Cross-References
- LED Area Emitters — UV-LED Arrays for Homogeneous Area Exposure — construction context (junction temperature, COB vs SMD, AlN substrate)
- UV Lamp Anatomy — Construction & Components — top-level component overview (lifetime chapter as a part)
- UV Lamp Technology — Low-Pressure, Amalgam, Medium-Pressure, LED, Excimer — lifetime comparison of the lamp classes
- Ballasts and Drivers — Which Type for Which UV Application — driver drift as a co-ageing factor
- Forward refs (coming): UV-LED thermal design in detail · UV radiometer selection matrix · UV-LED maintenance schedule templates per application class
8. Sources + Trust Anchors
Trust anchor — regulatory:
- US DOE Operating Lifetime Study of UV LEDs (2022) — the primary independent measurement study, covering UV-A to far-UVC
Academic (gold standard):
- Lifetime Analysis of Commercial 3W UV-A LED (MDPI Crystals, 2020) — Arrhenius activation energies 0.13/0.20 eV
- Review — Reliability and Degradation Mechanisms of Deep UV AlGaN LEDs (IOPscience ECS, 2023)
- Modelling the electrical degradation of AlGaN UV-C LEDs (AIP APL, 2023)
- Diffusion mechanism as cause of optical degradation in AlGaN UV-C LEDs (Nature Sci Reports, 2025)
- Impact of operation parameters on the degradation of 233 nm AlGaN far-UVC LEDs (AIP JAP, 2022)
- Efficiency- and lifetime-limiting effects of commercially available UV-C LEDs: a review (IOPscience J. Phys.: Photonics)
- Reliability Analysis of AlGaN-Based Deep UV-LEDs (PMC NIH)
- High-Power UV-LED Degradation: Continuous and Cycled Working Condition Influence (ResearchGate, 2015)
Industry documentation:
- UVNDT — UV LED Curing System Maintenance Checklist
- Gigahertz-Optik — Irradiance Measurement Application Note
- Dymax — UV LED Curing Systems Measuring (whitepaper)
- BlazeAsia — UV profiling for preventive maintenance
- Violumas/BosElec — Understanding UV LED Lifetimes
Community / practice:
- Prusa3D Forum — LCD Panel & UV LED Lifetime Discussion
- Liqcreate — Reasons to reduce UV power for resin (practice note)
Note: specific L70 hour values are manufacturer-specific and time-dependent. The ranges given here are the 2024–2025 industry consensus; evaluating a specific module always requires a datasheet comparison and, where needed, a follow-up question to the manufacturer about T_j conditions.