Why the filter is the whole product
A krypton-chloride lamp does not emit a single clean wavelength. Its main output is at 222 nm, but it also produces a small tail above 230 nm — and that tail behaves like conventional UV: it penetrates, and it damages. UV222® is built around a filtered krypton-chloride module. The filter is a hard multilayer dielectric coating on quartz — not an organic film that fades — and it cuts everything above 230 nm to 0.43% of the in-band output.
The invention behind it →
Schematic — wavelengths exact, intensities illustrative. The 0.43% figure is measured: Narita K et al., PLOS ONE 2022;17(5):e0267957.
What counts as a safe dose here
In the EU the limit that matters is ICNIRP, because Directive 2006/25/EC is built on it. The American ACGIH threshold is about twenty-one times more permissive on skin and seven times on the eye, and it has no legal standing in Cyprus or Greece.
We design every installation to 23 mJ/cm² over an eight-hour day.
| Limit over 8 hours | Skin | Eye |
|---|---|---|
| ICNIRP — the basis of EU law | 23 mJ/cm² | 23 mJ/cm² |
| ACGIH — United States | 479 mJ/cm² | 161 mJ/cm² |
Dose is distance and time
Dose = intensity × time. Output is fixed at the factory and cannot be turned up. The dose for a 90% reduction — D90 — is not one number: it runs from about 0.4 mJ/cm² for the coronaviruses to roughly 107 mJ/cm² for Aspergillus niger spores. Viruses are the easiest, then Gram-negative bacteria, then Gram-positive, then bacterial spores, then yeasts, then mould spores.
Why duty cycle exists
At one metre under a 60° beam the ICNIRP limit is reached after about 28 minutes — so across an eight-hour shift the lamp may run about 5.8% of the time.
60° beam, 115 mW
| Distance | Intensity | 90% of SARS-CoV-2, on a surface |
|---|---|---|
| 1 m | 13.8 µW/cm² | 43 s |
| 2 m | 3.42 µW/cm² | 2 min 54 s |
| 3 m | 1.52 µW/cm² | 6 min 31 s |
| 4 m | 0.86 µW/cm² | 11 min 36 s |
| 5 m | 0.55 µW/cm² | 18 min 7 s |
Fluence rate — what actually governs the air
The number in the table above is irradiance: what a surface facing the lamp receives. It is the right quantity for a surface dose and for eye and skin safety. It is the wrong one for the air.
A pathogen suspended in the air is not a flat surface pointing at anything. It is hit from every direction at once — straight from the lamp, and again off the walls and ceiling. The quantity that describes that is fluence rate, and it is what drives both air disinfection and the photochemistry that makes ozone.
The two have the same unit and are constantly confused, including in datasheets. The distinction is the whole basis of a competent installation.
| Fluence rate Ē₀ | Irradiance Ee | |
|---|---|---|
| Geometry | A sphere — photons from all directions | A flat plane with a direction |
| Angle | No dependence | Falls with cos φ — at 60°, half |
| Governs | Air disinfection · ozone | Surface dose · eye and skin safety |
| Measured by | Spherical radiometer — rare and costly; in practice simulated | An ordinary radiometer |
A hand-held radiometer reads irradiance. It equals fluence rate only when it is aimed straight at a single lamp, square on, with no reflections and no second source. In a real room with several lamps and reflective surfaces, an irradiance reading understates the fluence rate — which is why a survey with one meter and no model is not a survey.
How we arrive at it for your room
Three steps, in this order. The first takes a minute; the last takes a visit.
01 · ESTIMATE
The path-length model. If ceiling area A radiates power P downwards, the mean fluence rate is Ē₀ = P/A = P·l/V. It gives the theoretical ceiling — the real figure is always lower — and tells us roughly how many lamps a room needs.
02 · SIMULATE
Radiosity modelling in Illuminate, which is built for germicidal UV and accounts for wall reflections. The same two lamps moved to a diagonal raised the mean fluence rate by 50% in the published example. Position is not a detail.
03 · MEASURE
Point readings on site to check the model — taken at 1–2 m and directly under the lamp, where the result depends least on exactly where the sensor sits.
One room, both answers
Because pathogen kill and ozone are driven by the same quantity, they can be predicted together — from one figure, with one equation each.
Worked example from Claus (2026). A 4 × 3 × 3 m room, 36 m³, two filtered KrCl lamps, mean Ē₀ = 1.8 µW/cm² — inside the ordinary commercial range of 0.5–2.5.
AIRBORNE PATHOGEN
94.5%
reduction at steady state, equivalent to 25.9 extra air changes per hour
OZONE, SAME ROOM
5.4 ppb
against a 50 ppb guideline — roughly a tenth of the limit
Which is why "near-total pathogen elimination with zero ozone" is not a marketing exaggeration — it is physically impossible. The two come from the same photons. Anyone claiming one without the other has either not measured, or is counting on you not asking.
And this is what the filter buys you
Ozone production per unit of fluence rate, measured. A filtered krypton-chloride lamp — which is what we install — makes less than half the ozone of an unfiltered one.
σNC in ppb·hr⁻¹ per µW/cm². The 10.5 figure was measured on a filtered Ushio Care222 lamp by Peng et al. (2023).
6.9 theoretical, perfectly monochromatic
10.5 a filtered Ushio Care222 lamp, measured
7–13 filtered krypton-chloride, typical range
> 20 unfiltered krypton-chloride
Claus, H. Fluence rate and ozone production from far-UVC lamps — a perspective. Photochemistry and Photobiology (2026). doi:10.1111/php.70139
How much air it is worth
In a 32 m³ chamber with 3 ACH baseline ventilation, far-UVC within the ICNIRP limit gave a 92% reduction in under 15 minutes — equivalent to 35 air changes per hour.
Eadie et al., Scientific Reports 12:4373 (2022)
35 UV222 at the ICNIRP limit
20 operating theatre ventilation
6–15 portable HEPA unit
Equivalent air changes per hour
Ozone: the honest number
Far-UVC lamps do produce ozone. Anyone who tells you otherwise is either being careless or hoping you will not check.
What it settles at is not a property of the lamp. It is fluence rate multiplied by the ozone coefficient, divided by how fast the room removes ozone — surface deposition, ventilation, the people in it. In a furnished room, deposition alone is usually the largest of those terms.
Ozone at steady state = σNC × Ē₀ ÷ removal rate
example above: 10.5 × 1.8 ÷ 3.5 hr⁻¹ = 5.4 ppb
example above: 10.5 × 1.8 ÷ 3.5 hr⁻¹ = 5.4 ppb
At the fluence rates ordinary installations run at, the rise is single-digit parts per billion against the WHO eight-hour guideline of about 50 ppb. In a cleanroom at 20–60 air changes an hour it is not a practical concern at all. In a small, poorly ventilated room it has to be designed for — which is exactly what the site study is for.
One thing we watch. There is an active proposal to tighten the guideline from 50 ppb toward 5. If it lands, it changes how aggressively any far-UVC installation can be sized — ours included. We would rather tell you that now than be surprised by it with you.
Peng et al., Environmental Science & Technology Letters (2023) · NIST IR 8550 (2024) · Claus, Photochemistry and Photobiology (2026)
The evidence on safety
Eyes
Four ophthalmologists worked daily for 36 months at up to 22 mJ/cm² per eight-hour day — effectively at the ICNIRP limit. No change in visual acuity, refraction or corneal endothelial cell density.
Sugihara K et al., Photochemistry and Photobiology (2025), PMID 39659140
Skin
Twenty volunteers exposed at 50–500 mJ/cm² showed no erythema at any dose. Sixty-six weeks of chronic exposure — five days a week, eight hours a day — produced no tumours in UV-sensitive hairless mice.
Fukui T et al., PLOS ONE 2020;15(8):e0235948 · Welch D et al. (2022)
For perspective
Ten minutes of sunshine at UV index 4 causes as much DNA damage in the basal layer of the skin as more than 30,000 hours under filtered far-UVC.
Modelled comparison, not a measurement — offered for scale.
What we don't claim
The five sentences a competitor will not put on their website.
Not zero ozone. We publish the range and design the ventilation around it.
Not a replacement for cleaning or HVAC. An additional control that sits between the two.
Not effective in shadow. Light disinfects what it reaches, and nothing else.
Not a medical device. Disinfection equipment, not indicated to treat patients.
Not a single universal dose. Every organism and every room differs — which is why we model yours.