UV disinfection: working principle, UV dose and process limits

UV disinfection inactivates microorganisms through photochemical damage to their nucleic acids. The effect achieved depends on wavelength, fluence, the sensitivity of the target organism and the optical properties of the treated medium. This page explains the working principle, dose-response relationships, suitable UVC technologies and the main process limits. The normative validation of drinking water reactors and safe plant operation are covered on dedicated specialist pages.

The central technical quantity is fluence – the amount of radiation that actually reaches the microorganism, not the lamp power and not the exposure time. It follows from wavelength, irradiation geometry, the UV transmittance of the medium (UVT) and the path the organism travels. There is no universal “dose against germs”; what governs is always the organism-specific dose-response curve.

Materials inside the irradiation chamber must be UV-C resistant: seals, plastics and coatings age considerably faster under 254 nm than in daylight.

The measurement side – reference measurement, plant monitoring and dose assessment at 254 nm – is covered on the page measuring and monitoring UVC radiation at 254 nm.

UVC radiation in the range of roughly 200–280 nm is absorbed by nucleic acids (DNA, RNA) and by other cellular and viral components. Absorption is wavelength-dependent and peaks near 260–265 nm, the absorption maximum of the nucleic acid bases. The absorbed photons trigger photochemical reactions, in particular the formation of pyrimidine dimers (covalent links between adjacent thymine or cytosine bases) within the DNA or RNA strand. These lesions alter the molecular structure so that replication and transcription are disrupted; the micro-organism loses its ability to reproduce or to cause an infection, without the cell necessarily being destroyed immediately. The strength of this effect is not a constant – it depends on the organism, strain, physiological state, surrounding medium, wavelength and applied fluence. In addition, cellular repair mechanisms (photoreactivation, dark repair) can reverse part of the damage after treatment. This interplay is why a single “standard dose” is not valid for every application and target organism.

UV disinfection systems differ primarily in the radiation source employed. The source determines spectrum, power density, electro-optical efficiency and therefore also the suitable application.

An important point for the technical assessment: microbial efficacy, nucleic-acid absorption and electrical efficiency of a source do not share the same spectral optimum. This is both the opportunity and the metrological challenge of LED and excimer technology – their dose-response data cannot simply be carried over from conventional low-pressure or amalgam lamps.

• Wavelength: determines the absorption behaviour of nucleic acids and thus the biological efficacy per absorbed photon; different source classes need different dose-response references.
• Irradiance (fluence rate): the radiant power per area at a given moment; it determines how quickly a dose builds up, and – within the reciprocity law and its limits – can also influence whether the same integrated dose produces the same inactivation.
• Fluence/dose: the time integral of irradiance along an organism's path; this, not exposure time alone, is the true reference quantity of the dose-response curve.
• UV transmittance (UVT) of the medium: governs, per the Beer–Lambert law, how strongly radiation attenuates with increasing path length in the medium; a falling UVT reduces the fluence at the target despite unchanged lamp power.
• Residence-time and flow profile: in reactors and air ducts, organisms cross a three-dimensional radiation field on different paths; short paths receive less fluence than long ones, regardless of the mean sensor reading.
• Geometry (distance, reflection, shadowing): determines what fraction of emitted radiation actually reaches the organism, particularly for surface and air disinfection with complex object geometry.
• Initial bioburden and turbidity: high particle or cell concentrations can shield organisms from one another and shift the effective dose-response relationship.

The effect of UV disinfection is given in log reductions – not in per cent, because inactivation runs exponentially. One log means that of a thousand organisms a hundred remain; four logs mean that of ten thousand, one remains.

• 1 log – 90 % inactivated
• 2 log – 99 %
• 3 log – 99.9 %
• 4 log – 99.99 %

The fluence needed for that depends on the organism – far more strongly than most designs assume. At 254 nm, four log reductions take:

• E. coli – about 8 mJ/cm²
• Legionella pneumophila – about 9 mJ/cm²
• Cryptosporidium parvum and Giardia – about 22 mJ/cm²
• Bacillus subtilis spores – about 81 mJ/cm²
• Adenoviruses – about 167 mJ/cm²

Between E. coli and adenovirus there is a factor of twenty. From that follows the sentence most often missing from a design: a system designed against bacteria is not therefore safe against viruses. Covering adenoviruses takes twenty times the fluence – so either twenty times the irradiance or twenty times the residence time.

For drinking water, DIN 19294 and ÖNORM M 5873 therefore set a fixed figure: a reduction-equivalent fluence of 40 mJ/cm² (400 J/m²). That value is a convention for the typical target organisms of drinking water; for resistant pathogens such as adenoviruses it is not sufficient (see table).

That single figure is a German and Austrian convention, not a world standard. The US EPA sets its required dose per pathogen and per log credit instead, and NSF/ANSI 55 sets two classes for domestic devices. How the four validation regimes compare, and why their numbers must not be read as equivalents, is set out under UV water treatment.

The figures above come from collimated beam tests: defined fluence, thin layer, known transmittance, uniform irradiation. A reactor meets none of these conditions. There, every water particle takes a different path and collects a different fluence; what governs is the reduction-equivalent fluence determined by biodosimetry – that is, by the inactivation actually achieved on a test organism, not by a calculation. How that proof is to be furnished under the applicable standards is covered under validating UV reactors.

Three quantities pull the fluence down in operation, and none of them appears in the lamp datasheet:

• Lamp ageing. The UV-C output falls to 70 to 80 % of the initial value over the service life – and it falls faster than the visible output. A lamp that still lights does not necessarily still disinfect.
• Transmittance of the medium. The transmittance at 254 nm (UVT; stated as the spectral absorption coefficient SAC₂₅₄) decides how deep the radiation penetrates at all. For surface water or process water it is well below that of drinking water.
• Geometry and shadowing. With surface disinfection this is the most common error: what lies in shadow receives no dose. Edges, bores, gaps and the underside of a part receive only a fraction of the nominal fluence even with generous design.

That is why fluence is demonstrated, not merely calculated. The calculation gives the design value; the irradiance is measured at the sensor and the reduction-equivalent fluence is determined by biodosimetry – that is the value that holds up in an audit.

Three tasks, three instruments:

• Determining the fluence in a trial. The BSH-03 CBD is a collimated beam device: it produces a parallel, uniform UV-C field above a sample dish and so reproduces exactly the conditions under which the table values were obtained. It allows the dose-response curve to be recorded for your own medium and your own organism – the basis of any design. For the collimated beam test with a medium-pressure lamp, the BSM-03 CBD stands alongside it.
• Monitoring the plant. The UVC-SE is a pressure-resistant sensor for installation in the reactor: it measures the irradiance where it arrives and signals the controller as soon as it falls below the required value. That turns a design into a monitored plant.
• Tracing the sensor. In drinking water disinfection the plant sensor must be compared against a traceable reference instrument – the DVGW reference radiometer is built for that. Which standard demands which proof is set out under validating UV reactors.

That closes the chain: collimated beam test for the curve, biodosimetry for the reduction-equivalent fluence of the plant, inline sensor for ongoing operation, reference radiometer for the sensor. Any link without the next one leaves a gap.

A common fallacy is equating electrical lamp power with optical UVC output and the fluence effective at the organism. Between these three quantities lie several stages of efficiency loss: electro-optical efficiency of the source, spectral composition, geometric distribution in the reactor and absorption in the medium. A nominally powerful installation can therefore still deliver insufficient fluence at the target.

A time value alone (“X seconds of irradiation”) is technically meaningless without reference to the actual irradiance at the organism's location. The same exposure time can lead to very different doses depending on distance to the source, UVT of the medium and the reflectivity of the surroundings.

A measurement taken outside the actual process location – for example at the reactor inlet rather than along the actual flow paths – only captures the optical boundary condition, not the actual fluence distribution. Two plants can show the same sensor reading and still produce different microbial doses because residence times and organism paths differ. This effect is especially pronounced in air ducts, where irradiance at the duct-wall sensor and the actual fluence along the airflow can diverge the most (Luo & Zhong, 2022).

The reciprocity law (Bunsen–Roscoe law) assumes that only the product of irradiance and time determines inactivation, regardless of how that product is composed. Current research shows, however, that this assumption does not always hold: for certain organisms and intensity ranges, observed inactivation rates deviate from pure time-dose reciprocity, partly due to repair processes and non-linear damage accumulation (Nature Scientific Reports, 2025; MDPI Water, 2024). For system design, this means that a target dose calculated purely from irradiance and time need not automatically match the actual biological effect.

Material effects are frequently underestimated: fouling that accumulates over time on protective sleeves or quartz jackets reduces the effective UVT immediately around the lamp, without any change in electrical power draw. Likewise, cell aggregation and particles shield organisms inside; they receive a lower fluence than the sensor indicates. Dose-response data obtained under laboratory conditions with a reference organism in clear medium therefore apply only under comparable boundary conditions and must not be carried over uncritically to other wavelengths, organisms or water matrices.

The attenuation of UV radiation passing through an absorbing medium follows the Beer–Lambert law. The UV transmittance (UVT) of a water – usually measured at 254 nm – indicates what fraction of incident radiation remains after a defined path length. If UVT falls, for instance due to increased organic content, iron or turbidity, fluence decreases disproportionately with increasing distance from the lamp; a plant sized for a given UVT can undershoot its target dose under poorer water quality, even though lamp power and flow rate are unchanged. For polychromatic sources – medium-pressure lamps, but also LED arrays with several emission peaks – a second layer is added: absorption, the action spectrum of nucleic acids and sensor responsivity are all three wavelength-dependent and need not coincide. This is why polychromatic sources, or sources being compared across wavelengths, require spectral characterisation, while a broadband sensor can be sufficient for a known, unchanging monochromatic source. UVT itself is measured with a dual-beam UVC photometer at 254 nm (product FlowMissio).

Three physical relationships form the quantitative basis of UV disinfection:

Photon energy: E = h·c/λ, with h the Planck constant, c the speed of light and λ the wavelength. Shorter wavelengths carry more energetic photons; this explains why far-UVC (222 nm) delivers more energy per photon than 254 nm, while at the same time showing a shallower penetration depth into tissue or turbid media because of stronger absorption in near-surface material layers.

Time-dose reciprocity (Bunsen–Roscoe): log₁₀(N/N₀) = −k·D, with D = E_e·t (irradiance E_e multiplied by exposure time t) and k an organism- and wavelength-dependent inactivation rate constant in cm²/mJ. The model assumes a homogeneous, log-linear inactivation kinetic without significant threshold or tailing effects. As shown in the previous section, this assumption is not met across every intensity and time range.

Beer–Lambert law: E_e(z) = E_e,0 · 10^(−a·z), with a the decadic absorption coefficient of the medium and z the path length. UVT then follows as UVT = 10^(−a·L) · 100 % for a reference path length L. The model assumes a homogeneous, non-scattering medium; turbid or particle-laden media introduce additional scattering losses that the pure absorption law does not capture.

Practical consequence: a robust system design cannot rely on a single averaged dose. It must consider fluence-rate distribution (spatial), residence-time distribution (fluid-mechanical) and the dose-response curve (biological, wavelength-specific) together.

1. Assumptions. A study on the inactivation of antibiotic-resistant bacteria determined, for one test organism, a 1-log-reduction dose (D₁) of 4.11 mJ/cm² at 222 nm (KrCl* excimer lamp) and 8.99 mJ/cm² at 254 nm (low-pressure lamp) (RSC Environmental Science: Water Research & Technology, 2024). As a simplifying model assumption, this is extrapolated log-linearly.

2. Model. D(n-log) ≈ n · D₁ (valid only in the log-linear range, ignoring shoulder or tailing effects).

3. Calculation for an assumed 3-log reduction (99.9 %):

• 222 nm: D₃ ≈ 3 × 4.11 mJ/cm² = 12.3 mJ/cm²
• 254 nm: D₃ ≈ 3 × 8.99 mJ/cm² = 27.0 mJ/cm²

4. Result. In this particular experiment, the 254 nm source requires about 2.2 times the fluence of the 222 nm source for the same reduction.

5. Technical interpretation. The difference shows that the choice of radiation source directly affects the energetic and geometric design of a reactor. However, the values apply strictly to the organism and water matrix studied; carrying them over directly to other target organisms, real waters or technical plants without an own dose-response determination – for example via a collimated-beam test – is not technically justified.

Drinking water supply: UV disinfection is the only approved physical disinfection method for public water supply in Germany. The critical process quantity is the reduction equivalent fluence (REF) of the complete plant, which must be demonstrated by biodosimetry per DIN 19294-1 or -2. UV disinfection is attractive here because it works without chemical additives and without altering taste.

Municipal and industrial wastewater and process-water treatment: here, UV disinfection and UV/H₂O₂ oxidation increasingly appear as a fourth treatment stage for trace-substance removal. Reaction mechanisms and the design of the oxidation processes are covered under photolysis and AOP in water technology. The critical factor is the often lower and more variable UVT compared with drinking water, which calls for dynamic, UVT-dependent dose control.

Ultrapure water in semiconductor manufacturing and pharmaceuticals: process water with very high UVT places different demands on reactor design and material compatibility; here the photo-oxidative effect of UV on TOC reduction is often as relevant as disinfection itself.

Food and beverage industry: UVC is used for surface decontamination of packaging materials in aseptic filling and for treating clear liquids, an application area with points of contact to the Plants, Agriculture & Food Technology application page. What matters here is the actual irradiation of the relevant surface, taking geometry and conveyor speed into account, not lamp power.

HVAC and air disinfection: in-duct UVGI systems inactivate bioaerosols as air flows through the duct. What matters is residence time within the irradiation field as a function of air velocity and duct geometry; simple wall-sensor readings represent the actual fluence distribution along the flow only imperfectly (Luo & Zhong, 2022).

Medical technology and hospital hygiene: UVC robot systems and irradiation cabinets are used for terminal disinfection of rooms, instruments and surfaces – an application field closely related to the Medicine & Phototherapy application page. A critical factor is shadowing by furniture, equipment and complex object geometries, which does not occur in laboratory testing; studies show systematic deviations between controlled test conditions and real clinical operation (Casini et al., 2024).

Research, photobiology and biotechnology: bacteriophages are frequently used as surrogate organisms to characterise dose-response curves without having to work with pathogenic target organisms – see also the Photobiology & Biotechnology application page; the observed UVC sensitivity cannot be predicted from taxonomic relatedness of the phages alone (Schubert et al., 2023).

Far-UVC at around 222 nm is under investigation for rooms with people continuously present. Exposure limits, ozone formation, shielding and occupational safety are covered on the safety of UV disinfection page; for process design, all that needs stating here is that these sources require a safety and measurement concept of their own.

For highly turbid or strongly absorbing media, where conventional UV reactors reach their design limits, UV processes are frequently combined with upstream filtration or membrane technology to secure sufficient UVT. For surfaces with complex three-dimensional geometry, where fixed lamp arrangements cause shadowing, mobile systems with several irradiation positions or robotic systems are used; their effectiveness has to be demonstrated in the field separately, not only in the laboratory.

Measurement requirements follow from the specific application. The quantities that must generally be captured are irradiance at the relevant process location, the UV transmittance of the medium (for water applications) and – for changing or polychromatic sources – the spectral distribution of the radiation.

A broadband measurement is sufficient when the source is known, spectrally stable and largely monochromatic, as with established low-pressure lamp installations with a defined ageing behaviour – a calibrated UVC sensor is suitable here. As soon as several source technologies are compared, a polychromatic source is involved, or spectral shift due to ageing needs to be documented, a spectral measurement with a spectroradiometer such as the SR900 becomes necessary, since a broadband sensor cannot resolve different wavelength components separately. For guidance on choosing the right sensor per source technology, see Selecting UV Sensors and How UV Sensors Work.

Spatially resolved measurements become necessary when the fluence distribution within a reactor, air duct or irradiation cabinet is itself the subject of investigation – for example when validating new reactor geometries. Time-resolved measurements (long-term monitoring) are relevant for detecting lamp ageing and fouling of protective sleeves early, before the target dose is undershot.

Permanently installed, digital sensors such as the PLC.D series handle this ongoing monitoring in technical plants and transmit not only the current reading, but also sensor type and calibration status to the control system. This information enables process control that goes beyond simple threshold monitoring: instead of a fixed exposure time, the controller can end exposure once a defined target dose is reached – a principle implemented, for example, by the UV-MAT dose control in irradiation chambers – so that lamp ageing does not directly affect the applied dose. Combined with flow and UVT measurement, dynamic dose control (dose pacing) can be realised, in which the plant adapts lamp power or the number of active modules to fluctuating water quality and flow rates instead of running a fixed, worst-case-oversized output. For documenting biological test series and regulatory evidence, reading, calibration status, sensor position, distance, sample geometry and exposure time must all be recorded together, since only this combination allows a traceable reconstruction of the applied dose.

Relevant measurement uncertainties concern traceability of calibration to national standards, the angular response of the sensor (cosine error at non-perpendicular incidence) and the temperature dependence of sensor responsivity. Qualified testing of drinking-water plants requires standard-compliant reference sensors whose requirements are set out in DIN 19294 and ÖNORM M 5873 respectively, and which must be designed differently depending on lamp type (low-pressure or medium-pressure). For the initial determination of a dose-response relationship for a target organism, a collimated-beam test with defined sample geometry and known irradiated area provides the controlled reference conditions that a pure system measurement cannot offer.

Under the theme UV disinfection, the publication finder currently lists 16 works examining UV and UVC radiation in a disinfection context. A recent study tested the effectiveness of UVC irradiation against the Mpox virus on contaminated surfaces and thereby provides pathogen-specific data for a virus for which few UV dose-response data existed so far.

UV-C Irradiation Effectiveness on Mpox-Virus-Contaminated Surfaces. Gidari, Anna, et al. Pathogens, 15(1), 2026, Art. 78. DOI

Another study compared the same UVC robot system under controlled laboratory conditions and in real hospital operation, illustrating how far laboratory figures and field conditions can diverge.

Antimicrobial efficacy of an experimental UV-C robot in controlled conditions and in real hospital scenario. Casini, Beatrice, et al. Journal of Hospital Infection, 2024. PubMed

A study testing germicidal treatment cabinets and carousels used a bacteriophage as a surrogate organism and visualised the spatial dose distribution inside the cabinet – relevant for assessing shadowing effects.

An Assessment of Germicidal Ultraviolet Treatment Cabinets and Carousels Using a Bacteriophage Surface Challenge. Brookes, Jodi, et al. Applied Biosafety, 28(4), 2023, pp. 242–255. PubMed

A study on the genetic diversity of lactococcal bacteriophages investigated whether their relatedness allows predictions about UVC susceptibility, providing direct evidence for the organism-specific nature of the dose-response curve.

Does the high biodiversity of lactococcal bacteriophages allow predictions about their different UV-C susceptibilities? Schubert, Christina, et al. International Journal of Food Microbiology, 401, 2023, Art. 110274. PubMed

A modelling study developed and validated an improved mathematical irradiance model for in-duct UVGI applications, illustrating how far fluence and surface readings can diverge in air ducts.

Development and experimental validation of an improved mathematical irradiance model for in-duct ultraviolet germicidal irradiation applications. Luo, Hao, and Lexuan Zhong. Building and Environment, 226, 2022, Art. 109699. ScienceDirect

A study on the decontamination performance of a dielectric-barrier discharge (DBD) with a UVC-emitting phosphor describes an alternative, low-mercury source technology alongside conventional lamps and LEDs.

Decontamination efficiency of a DBD lamp containing an UV-C emitting phosphor. Caillier, Bruno, et al. Photochemistry and Photobiology, 91(3), 2015, pp. 526–532. PubMed

Further use cases by topic are listed in the overview Customer Publications – by Topic.

The publication finder lists all publications on UV disinfection we know of – searchable by device, field and keyword.

• DVGW worksheet W 294-1 (December 2023): Planning, operation and monitoring of UV disinfection plants in water supply. DVGW
• DIN 19294-1 to -4 (2020/2026): Devices for disinfecting water using ultraviolet radiation – requirements and testing. DVGW-Regelwerk
• US EPA: Ultraviolet Disinfection Guidance Manual (UVDGM) and UV Treatment Toolkit, 2006/2022. US EPA
• ISO 15858:2016, UV-C Devices – Safety information – Permissible human exposure. ISO
• IEC 62471-6:2022, Photobiological safety of lamps and lamp systems – Part 6: Ultraviolet lamp products. IEC
• IUVA: Far UVC Radiation for Disinfection of Air and Surfaces – scientific statement on 222 nm applications. IUVA
• A complete overview of the applicable UV standards by country and lamp type is maintained at Guidelines, Standards and Norms in UV.

How does UV disinfection work technically?
UVC radiation is absorbed by nucleic acids in micro-organisms and produces photochemical damage, in particular pyrimidine dimers in DNA or RNA. This damage prevents replication and reproduction without necessarily destroying the cell immediately. The effect is not a chemical reaction but a physical-photochemical process, with no additives in the treated medium.

Which wavelength is best suited to UV disinfection?
The range around 260–265 nm, the absorption maximum of the bases, is the most effective against nucleic acids. Conventional low-pressure lamps emit almost monochromatically at 253.7 nm, close to this optimum. UV LEDs and far-UVC sources at 222 nm show partly different, organism-specific action profiles and require their own, wavelength-specific dose-response evidence before practical use.

What dose is required for a given log reduction?
There is no universal dose. The required fluence depends on the target organism, the water matrix, the wavelength and the desired reduction. Reliable values come from collimated-beam tests with the specific target organism under defined sample geometry, not from generic tables for other applications or organisms.

What is the difference between irradiance, dose and fluence?
Irradiance is radiant power per area at a given moment. Dose (exposure) is its time integral over a fixed area geometry. Fluence additionally refers to an object that can receive radiation from all directions – relevant for organisms in three-dimensional radiation fields such as reactors or air ducts. The formal definitions of irradiance, radiant exposure and fluence are given under radiometric quantities.

Why does a UV plant not work reliably despite adequate lamp power?
Common causes are falling UV transmittance of the medium, fouling of protective sleeves, unfavourable flow distribution with short residence times on some paths, or a measurement position that does not reflect the actual fluence at the process location. Electrical power alone is no proof of adequate disinfection performance – it is merely an operating parameter of the source. The underlying definitions and units are listed in the overview of radiometric quantities.

How is the UV dose measured and monitored in a plant?
For known, stable sources, calibrated broadband sensors at defined measurement positions are sufficient. For polychromatic sources or when comparing several source technologies, a spectral measurement with a spectroradiometer is required, since a broadband sensor cannot separate wavelength components. Drinking-water plants require standard-compliant reference radiometers per DIN 19294 or ÖNORM M 5873.

Which technology suits which application?
Low-pressure and amalgam lamps dominate large-volume water disinfection with established standards. Medium-pressure lamps suit high throughputs and oxidation processes. UV LEDs are advantageous for small flow rates, point-of-use applications and wherever mercury-free operation is required. Far-UVC is still in the research and validation phase for applications in occupied spaces.

Why is electrical lamp power not a suitable indicator of disinfection performance?
Several loss mechanisms lie between electrical power draw and the fluence effective at the organism: electro-optical efficiency of the source, spectral composition, geometric distribution within the reactor and absorption in the medium. Only the actually measured or validated fluence at the target is a reliable process quantity – not the datasheet value of the lamp.

How long does disinfection take?
The exposure time is the required fluence divided by the irradiance at the location of the organism. For four log reductions at 254 nm this means 8 mJ/cm² for E. coli, 22 mJ/cm² for Cryptosporidium and Giardia, 81 mJ/cm² for Bacillus subtilis spores, 167 mJ/cm² for adenoviruses and 40 mJ/cm² as the reduction-equivalent fluence under DIN 19294. In a closed irradiation chamber at 10 mW/cm² the 40 mJ/cm² are reached after four seconds, at 85 mW/cm² after half a second. In a reactor the residence time decides instead, and the absorption of the medium lowers the irradiance with depth. A time without the corresponding irradiance and target organism is therefore not transferable.

This application is closely related to adjacent fields of optical radiation metrology – for example Photobiology & Biotechnology (bacteriophages as surrogate organisms, general dose-response characterisation), Medicine & Phototherapy (UVC robot systems in hospital hygiene), and Plants, Agriculture & Food Technology (surface decontamination of packaging and food). Mobile UV-C systems and manipulators, where the fluence per surface element only comes about along a path, are covered under robotics. The full picture is in the UV applications overview.

Unsure whether the irradiance measured at the sensor actually corresponds to the fluence your target organisms receive in the reactor or air duct? A reliable answer requires a measurement and validation strategy matched to the application and source technology – from spectral characterisation with the SR900 to control measurement with the RMD Pro. Talk to us about your plant or testing task.