All knowledge articles

UV-C Dose & Log Reduction — the number behind every disinfection claim

· May 30, 2026 · #uv-c#dose#fluence#log-reduction#d90#disinfection#fundamentals#microbiology

Quick answer

"Kills 99.9% of germs" is, on its own, an empty claim. It only becomes an engineering statement once you attach three things: which organism, in which medium (air, surface or water), and at what UV dose delivered to the target. The dose — properly called fluence, measured in mJ/cm² — is the single number that connects a UV-C lamp to a biological result.

Every organism has a characteristic dose-per-log: the fluence needed to cut the surviving population by a factor of ten (1 log = 90%). For many vegetative bacteria that is under ~12 mJ/cm² for the first log; for viruses it is typically higher; for spores and a few stubborn fungi it runs into the hundreds of mJ/cm². A "99.9%" headline tells you none of this — the same lamp can be brilliantly effective against one organism and nearly useless against another at the same exposure. The rest of this article gives you the framework to turn a claim into a specification.

1. The two numbers people constantly confuse: irradiance vs. dose

This is the most common and most expensive mix-up in UV-C.

  • Irradiance (also "fluence rate") — the instantaneous UV power hitting a unit area, in mW/cm². Think of it as how hard the light is shining right now.
  • Dose (also "fluence") — irradiance accumulated over time, in mJ/cm². Think of it as the total energy delivered.

The relationship is deliberately simple:

Dose (mJ/cm²) = Irradiance (mW/cm²) × Exposure time (s)

This is the basis of the IUVA reference method, the Protocol for the Determination of Fluence (Bolton & Linden, 2003) — the document most credible UV testing ultimately traces back to. The units mW/cm² and mJ/cm² look almost identical and are routinely swapped in datasheets and sales copy; whenever you read a UV number, check which one it is.

Practical consequence — reciprocity. Because dose is the product of irradiance and time, a weak lamp for a long time can deliver the same dose as a strong lamp for a short time. This intensity–time reciprocity (the Bunsen–Roscoe principle) holds well across a wide range in practice, but not without limits — at very low irradiance, biological repair (Section 4) can outrun the damage, and at very high irradiance some effects saturate. Treat reciprocity as a strong rule of thumb, not a law you can extrapolate forever.

2. Log reduction & the D90 value — first-order kinetics

UV-C inactivation, in its idealised form, follows first-order kinetics: each additional increment of dose removes the same fraction of the survivors, not the same absolute number. That is why disinfection is expressed in logs:

Log reduction Survivors removed Fraction surviving
1-log 90% 1 in 10
2-log 99% 1 in 100
3-log 99.9% 1 in 1,000
6-log 99.9999% 1 in 1,000,000

The key per-organism constant is the D90 value (a "D-value"): the dose required for one log of reduction. In the ideal first-order world, the dose for N logs is simply N × D90 — so a 6-log target needs roughly six times the 1-log dose. (Section 3 explains why real systems rarely behave that cleanly, so this should be used to size up, not to promise.)

Number anchors (from a meta-analysis of 413 studies and primary papers):

  • Vegetative bacteria: the first log (90%) is reached below ~12 mJ/cm² in the large majority of studies, and at ≤5 mJ/cm² in about 82% of cases. Salmonella Typhimurium, for example, shows a sensitivity near 1.9 mJ/cm² per log, with ~7.8 mJ/cm² giving a 3-log cut.
  • Viruses: generally more resistant — up to ~20 mJ/cm² for the first log in about 75% of studies, and up to ~80 mJ/cm² in roughly 22%.
  • Resistant organisms: bacterial spores and some fungi sit far higher. A 6-log inactivation of Candida auris on a carrier has been reported needing on the order of ~597 mJ/cm² — two orders of magnitude above an easy bacterium.

The span from ~2 mJ/cm² to ~600 mJ/cm² for "the same job" (one organism vs. another) is exactly why a single "kills 99.9%" figure is meaningless without the target named.

3. Why real-world curves aren't straight lines: shoulder & tailing

Plot log-survival against dose and the textbook says you get a straight line. Real data usually doesn't:

  • Shoulder — at low doses the curve is flatter than expected before steepening. Attributed to a damage threshold / built-in repair capacity that must be saturated before net kill accelerates.
  • Tailing — at high doses the curve flattens again, so the last logs cost disproportionately more dose. Common causes are cell or spore aggregation, shielding by suspended solids, flocs or organic matter, hydraulic short-circuiting in flow systems, and simply approaching the assay's detection limit.

Engineering takeaway: never linearly extrapolate a measured 2-log dose to a 6-log requirement. The tail is where projects fail. Size for the tail and the worst-case shielding, not for the clean collimated-beam line.

4. The part nobody markets: microbial repair (photoreactivation & dark repair)

UV-C works by damaging genomic DNA — chiefly by forming pyrimidine dimers that block replication. Crucially, that damage is, for some organisms, repairable:

  • Photoreactivation — under subsequent visible/UV-A light, the enzyme photolyase can reverse dimers, and a fraction of "inactivated" organisms recover.
  • Dark repair — other repair pathways operate without light.

The degree of reactivation depends on the delivered UV-C dose (higher dose suppresses it), and reactivation curves can be modelled and predicted. This is a real limitation of UV-C, not a footnote: a system validated immediately after exposure can show apparent regrowth hours later. It is one reason designers build in a dose margin above the bare lab D-value, and why downstream conditions (light exposure, hold time) matter.

5. The medium changes everything: air vs. surface vs. water

A D90 measured under a collimated beam in the lab is an idealised value — a uniform, fully-known dose to a thin, clean sample. Real installations almost never deliver that uniformly; the meaningful quantity becomes the dose distribution across the treated volume, and the minimum dose any pathogen receives governs performance.

Medium Dominant dose-limiting factor What actually has to be controlled
Water UV transmittance (UVT) / absorption, suspended solids clarity, lamp-to-target path length, flow/residence time
Air residence time at velocity, lamp geometry airflow rate, duct/chamber design, number & placement of lamps
Surface geometry, shadowing, angle of incidence line-of-sight to every surface, distance, exposure time per spot

A concrete surface example for calibration of intuition: surface-dried SARS-CoV-2 has been shown to reach a >6-log reduction at roughly 3.5 mJ/cm² at 254 nm — but that is the dose at the surface, under direct line of sight. Shadowed or angled areas in the same room may receive a small fraction of it. Modelling the dose distribution (rather than trusting a single nameplate figure) is what separates a validated system from a hopeful one.

6. From a claim to a specification (the practical loop)

To convert "99.9%" into something you can engineer and defend, answer three questions in order:

  1. Which organism? → fixes the target D90 / dose-per-log (Section 2). Design for the most resistant relevant organism, not the easiest.
  2. Which medium and geometry? → fixes the delivery losses and the minimum delivered dose (Section 5).
  3. What margin? → add headroom for tailing (Section 3) and repair (Section 4), plus a validation safety factor.

And then measure — validation means confirming the delivered dose with calibrated radiometry, not trusting a lamp's nameplate output, which degrades over life. (A dedicated article on UV-C validation, radiometry and the measurement obligation is coming.)

Cross-references

  • UV-C validation & dose measurement — (coming): how delivered dose is actually confirmed, radiometer types, and where measurement is legally required.
  • Air vs. surface vs. water UV-C in practice — (coming): per-medium system design and the dose-distribution problem in depth.
  • For the German practice context, the relevant industry guideline is AG LUV Guideline 100 (co-authored by Dr. Michael Calenberg) — worth consulting alongside the international sources below.

Sources