1. Introduction

Indoor air quality (IAQ) is a critical determinant of human health, as modern populations spend nearly 90% of their time indoors^[1]. In healthcare and other high-occupancy environments, ventilation performance is directly linked to the risk of airborne infection^[2,3]. The COVID-19 pandemic demonstrated that respiratory pathogens such as severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) can spread efficiently through aerosols suspended in indoor air^[4-6], making the control of airborne transmission an important objective in building environmental design^[7,8]. Biomedical interventions remain essential, but the persistence of airborne transmission has reinforced the need for complementary engineering controls^[9,10]. Among these controls, ultraviolet germicidal irradiation (UVGI) has received renewed attention as a means of continuously treating airborne microorganisms within occupied spaces and ventilation systems^[11,12]. Although ultraviolet radiation is well established for air disinfection^[13], its implementation in modern heating, ventilation, and air-conditioning (HVAC) systems requires careful consideration of lamp output, airflow condition, and exposure duration.

Ultraviolet radiation in the ultraviolet-C (UV-C) band (200-280 nm) exhibits germicidal activity through absorption by microbial nucleic acids, producing photochemical lesions that inhibit replication^[13,14]. Low-pressure mercury (LP-Hg) lamps emit predominantly at approximately 254 nm, close to the strongly germicidal wavelength region^[15]. Required UV-C fluence varies substantially among microorganisms and depends on organism type, aerosol state, environmental condition, experimental method, and the selected inactivation endpoint. For aerosolised SARS-CoV-2, doses of approximately 0.42-0.51 mJ/cm² have been reported under controlled experimental conditions^[16]. Because ultraviolet (UV) dose is the product of irradiance and exposure time, both radiation intensity and airflow condition govern in-duct disinfection performance, consistent with standardised fluence definitions^[17,18].

Within buildings, UVGI is implemented primarily through upper-room and in-duct systems^[15,19]. Upper-room UVGI irradiates the upper portion of occupied spaces and relies on room-air mixing, with substantial bacterial inactivation reported under controlled conditions^[20-22]. In contrast, in-duct UVGI places UV-C lamps within HVAC ducts to treat moving air and internal surfaces outside occupied spaces^[23,24]. This arrangement permits centralised treatment and continuous recirculation, but its effectiveness depends strongly on lamp output and air-residence time under actual duct operating conditions.

LP-Hg lamp output is strongly temperature-sensitive because mercury vapour pressure depends on the local lamp-wall and cold-spot temperatures^[25-27]. Lamp output is therefore governed by coupled convective and radiative heat transfer processes between the lamp surface and the surrounding airflow. In ventilation ducts, airflow can cool the lamp below its favourable operating range, reducing mercury vapour pressure and UV-C output^[25,28]. Higher air velocities and lower inlet temperatures intensify this effect^[26]. Consequently, prescribing constant lamp output can introduce uncertainty into in-duct UVGI assessments when airflow velocity and inlet temperature vary. Recent studies have examined radiation, airflow, and environmental effects in in-duct UVGI systems, but lamp output is still frequently prescribed rather than evaluated together with measured surface-temperature and irradiance behaviour^[11,28,29].

Electrical-to-radiative conversion further constrains UVGI performance because only part of the electrical input is converted into germicidal radiation, while the remainder is dissipated through heat and other losses^[15,30]. The 254 nm output of an LP-Hg lamp varies with its thermal condition and generally decreases when the lamp is substantially over-cooled or operated above its favourable temperature range^{[25-28,31,32]}. In ducted operation, changes in airflow velocity or inlet temperature can therefore produce measurable variation in lamp output. Lamp ageing, sleeve fouling, ballast characteristics, and installation geometry introduce additional variation during long-term operation^[26-28].

Ultimately, UVGI effectiveness is governed by the dose delivered to microorganisms, which depends jointly on irradiance and exposure time^[17,19]. Air velocity therefore plays a dual role by controlling both exposure duration and lamp thermal condition. Higher velocities shorten residence time and enhance convective cooling, potentially reducing dose through both mechanisms^[25,28]. Although considerable research has examined UVGI systems, thermal, optical, and microbial performance are frequently addressed separately. Photometric studies characterise radiation fields without directly resolving lamp thermal response, while microbial studies often report inactivation without quantifying lamp-temperature and spatial-irradiance variation^[28,29]. This separation limits the ability to compare lamp ratings and geometric arrangements under changing HVAC operating conditions.

To address this gap, the present study combines external lamp-surface temperature measurements, spectro-radiometric output characterisation, multi-lamp irradiance mapping, and Lagrangian UV-dose assessment for 95 W and 60 W LP-Hg lamps. In contrast to recent studies that primarily model radiation and flow fields or review individual in-duct design factors^[11,28,29], the present work evaluates how measured lamp thermal behaviour affects the irradiance inputs used in the downstream dose calculations. Specifically, the study: (i) quantifies convective and radiative heat transfer and external lamp-surface temperature distributions under controlled duct-flow conditions; (ii) relates measured external surface temperature to 254 nm output and intrinsic radiant efficiency; and (iii) compares multi-lamp UV dose and species-specific microbial response using measured irradiance fields and published susceptibility constants. The temperature and irradiance results were obtained experimentally, whereas the microbial inactivation results were derived from comparative calculations. Together, the results provide a measurement-informed basis for comparing lamp rating, airflow condition, and lamp arrangement within the tested in-duct UVGI system.

2. Methodology

2.1 Experimental setup and data analysis

The in-duct UVGI system comprised an approximately 2.7 m long square airflow duct, with a cross-section of 0.61 m × 0.61 m and an epoxy-coated internal surface; the overall arrangement is shown in Figure 1. Airflow was supplied by a variable-speed axial fan, while inlet air temperature was regulated using an industrial chiller. Two LP-Hg UV-C lamps were evaluated: a 95 W OSRAM PURITEC 2G11 lamp and a 60 W OSRAM PURITEC 2G11 lamp. Their overall lengths were 0.533 and 0.408 m, with active emitting lengths of 0.50 and 0.38 m, respectively. Both lamps had an overall folded twin-tube width of approximately 0.040 m. The outer diameter of an individual cylindrical lamp limb was 0.0176 m and was used as the characteristic diameter in the heat transfer calculations. To minimise early-life output drift, the lamps underwent a 100-h burn-in period before testing^[26]. Lamps were mounted at designated installation ports within the test section. UV-C irradiance was measured using an Ocean Insight Maya 2000 Pro spectrometer connected to a 115 µm solarisation-resistant fibre-optic cable and a polytetrafluoroethylene (PTFE) cosine corrector, while external lamp-surface temperatures were recorded using HOBO contact sensors.

Display Full Size

Download

Figure 1. Schematic of the experimental in-duct UVGI system showing the airflow duct, cooling section, test section, lamp-array region, multiple downstream irradiance-measurement planes, measurement instrumentation, and the duct-exit plane used to report cumulative particle dose. UVGI: ultraviolet germicidal irradiation.

The apparatus consisted of a cooling section for inlet-air temperature control, a modular test section containing the lamps and measurement equipment, and a downstream tail section extending to the duct exit (Figure 1). The cooling section was connected to the chiller and conditioned the incoming air to the prescribed set point, enabling the effects of ambient temperature on LP-Hg lamp performance to be evaluated^[26,33]. The modular test section enabled different lamp arrangements to be examined, while the downstream section accommodated irradiance measurements at multiple axial planes and cumulative particle-dose evaluation at the duct exit. Specifications of the measurement equipment are summarised in Table 1.

Experiments were conducted at nominal airflow velocities of 1.1, 1.5, 2.0, and 2.5 m/s and inlet-air temperatures of 14-22 °C. The nominal low-flow condition of 1.1 m/s corresponded to a measured mean velocity of approximately 1.15 m/s. The measured value was used in the heat transfer and dose calculations, while the nominal value is retained in the figure labels and discussion. For each condition, the fan and chiller were adjusted to the target settings, after which the lamps were energised and allowed to reach a steady thermal and optical condition for approximately 60 min. The spectrometer wavelength scale was checked using the mercury-emission peak.

Two experimental groups were investigated. In the first group, single-lamp thermal-radiative characterisation was performed by masking one limb of each twin-tube lamp, thereby producing a single-emitting-limb configuration. Eight HOBO contact sensors were distributed among the root, mid, and tip axial stations and the upstream, downstream, and top circumferential positions shown in Figure 2. External lamp-surface temperature was recorded continuously for 60 min, and the reported values were time-averaged over the steady measurement period. Near-field UV-C irradiance was measured along the active lamp length with the probe positioned 5 mm from the external surface of the exposed lamp limb. In the second group, arrays containing two to six unmasked lamps were evaluated using the configurations shown in Figure 3. UV-C irradiance was recorded at 25 positions on a 5 × 5 cross-sectional grid (X = 10, 110, 210, 310, 410 mm; Y = 25, 126.25, 227, 328.75, 430 mm) at multiple axial planes between the lamp array and the duct exit. These multiple-plane measurements were used to reconstruct the downstream three-dimensional irradiance field for the Lagrangian dose calculations.

Display Full Size

Download

Figure 2. Experimental arrangement for thermal characterisation of the UV-C lamp, showing the root, mid, and tip axial stations and the upstream, downstream, and top circumferential measurement positions. UV-C: ultraviolet-C.

Display Full Size

Download

Figure 3. Multi-lamp configurations investigated using the 95 W and 60 W LP-Hg UV-C lamps. Circled locations indicate the active lamp positions for configurations 2-1, 2-2, 3-1, 3-2, 4-1, 5-1, 6-1, and 6-2. UV-C: ultraviolet-C.

Measured spectral irradiance was converted to 254 nm band irradiance by integrating the baseline-corrected mercury-emission peak over a ±0.5 nm wavelength interval centred on the measured peak^[31,34-37]. During the single-limb axial scans, the irradiance probe was positioned 5 mm from the external tube surface. The corresponding cylindrical measurement radius was therefore R_m = 0.0318 m, comprising the physical tube radius of 0.0088 m and the 0.005 m probe offset. For the multi-lamp measurements, the 5 × 5 data obtained at each axial plane were assigned to their corresponding cross-sectional and axial locations. Interpolation within the measured cross-sections and between successive axial planes was used to reconstruct the downstream three-dimensional irradiance field. Electrical power was represented using the manufacturer-rated lamp powers of 95 and 60 W; actual ballast input and ballast losses were not measured separately. Transient outliers were screened before averaging. The stated measurement specifications were ±0.1 m/s for airflow velocity, ±0.2 °C for temperature, and 0.1-0.2 nm for spectrometer wavelength resolution. The measured temperature and irradiance data were used to characterise lamp behaviour and provide inputs to the Lagrangian dose model.

2.2 Convective-radiative heat transfer

LP-Hg lamp output is influenced by external lamp-wall thermal condition because mercury vapour pressure depends on local lamp temperature^[25-28]. In duct-mounted operation, convective cooling and long-wave radiative exchange jointly determine the external lamp-surface temperature and consequently influence UV-C output. To compare the thermal response of the 95 W and 60 W lamps, the ratio of convective to radiative external surface heat loss, r, was evaluated as

(1)$r=\frac{Q_{\text{conv}}}{Q_{\text{rad}}}$

where Q_conv and Q_rad denote the convective and radiative heat losses, respectively, given by:

(2)$Q_{\text{conv}}=hA(T_s-T_{\mathrm{amb}})$

(3)$Q_{\text{rad}}=\epsilon \sigma A(T_s^4-T_{amb}^4)$

Here, h is the convective heat transfer coefficient, A is the surface area of the cylindrical lamp, ε is the emissivity of the quartz envelope, σ is the Stefan-Boltzmann constant, and T_s and T_amb are the lamp surface and ambient air temperatures in Kelvin, respectively. The convective heat transfer coefficient h was evaluated using the Nusselt number formulation:

(4)$h=\frac{Nu\cdot k}{D}$

where k is the thermal conductivity of air and D = 0.0176 m is the outer diameter of an individual cylindrical lamp limb. For a single cylindrical tube in crossflow, natural convection was modelled using the Churchill and Chu correlation (1975), while forced convection was represented using the Churchill and Bernstein formulation (1977), consistent with previous UVGI heat transfer studies^[26,28]:

(4a)$N{u}_N={[0.60+\frac{0.387R{a}^{\frac{1}{6}}}{(1+{\left(\frac{0.559}{Pr}\right)}^{\frac{9}{16}})\frac{8}{27}}]}^2$

(4b)$N{u}_F=0.3+\frac{0.62R{e}^{\frac{1}{2}}{\mathrm{Pr}}^{\frac{1}{3}}}{[1+{\left(\frac{0.4}{Pr}\right)}^{\frac{2}{3}}]\frac{1}{4}}{[1+{\left(\frac{\mathrm{Re}}{282000}\right)}^{\frac{5}{8}}]}^{\frac{4}{5}}$

To provide a continuous combination of the natural- and forced-convection contributions, the two correlations were combined using the generalized p-norm formulation in Eq. (4c), following the asymptotic blending framework introduced by Churchill and Usagi and subsequently applied in mixed-convection studies^[38-40].

(4c)$Nu={({\mathrm{Nu}}_N^n+{\mathrm{Nu}}_F^n)}^{\frac{1}{n}}$

An exponent of n = 3 was adopted as a fixed blending parameter for the present comparative calculations rather than as a universally established value for cylinder crossflow. This formulation provides a continuous combination of the natural- and forced-convection contributions used in the external surface heat-transfer estimate. Substitution of Eq. (4) into Eq. (1) gives the heat-loss ratio in terms of the fluid properties and lamp parameters:

(5)$r=\frac{Nu\cdot k(T_s-T_{amb})}{D\cdot \epsilon \sigma (T_s^4-T_{amb}^4)}$

Lamp-surface temperatures were measured at three axial stations. For profile plotting and axial heat-loss integration, the measured directional temperature profiles were interpolated or extrapolated to the active-length endpoints, x = 0 and x = L. The active lamp length:

(6a)$Q_{\text{conv},\text{tot}}={\int }_0^{L}h\left(x\right)A^{\prime }\left(x\right)\left[T_s\right(x)-T_{\mathrm{amb}}]dx$

(6b)$Q_{\text{rad,tot}}={\int }_0^L\epsilon \sigma A^{\prime }\left(x\right)\left[T_s\right(x{)}^4-T_{amb}^4]dx$

where A'(x) = πD d is the differential cylindrical surface area. The resulting Q_conv,tot and Q_rad,tot represent estimated convection and long-wave radiation from the evaluated external lamp surface. They do not include ballast losses, electrical losses, or heat transfer through the lamp base and electrical connection. The dimensionless ratio r = Q_conv,tot/Q_rad,tot was used to compare the relative external surface heat-loss components under different airflow velocities and ambient temperatures.

Circumferential surface-temperature non-uniformity was quantified using the angular thermal non-uniformity index:

(7)$ATNI=\frac{{\overline{T}}_{\text{downstream}}-{\overline{T}}_{\text{upstream}}}{{\overline{T}}_s}$

where ${\overline{T}}_{\text{upstream}}$ and ${\overline{T}}_{\text{downstream}}$ are the mean upstream and downstream external surface temperatures, respectively, and ${\overline{T}}_{s,K}$ is the mean external lamp-surface temperature expressed in Kelvin. Because the numerator is a temperature difference, it has the same numerical magnitude when expressed in degrees Celsius or Kelvin. ATNI therefore provides a dimensionless measure of circumferential external surface-temperature non-uniformity.

2.3 UV output and intrinsic efficiency

To characterise intrinsic UV-C output, axial irradiance profiles were acquired from the 95 W and 60 W twin-tube LP-Hg lamps with one lamp limb masked^[31-35]. The irradiance probe was positioned 5 mm from the external surface of the exposed limb. The measured 254 nm band irradiance was integrated over the corresponding cylindrical measurement surface:

(8)$P_{254}=2\pi R_m{\int }_0^L{E}_{254}\left(x\right)dx$

where R_m = 0.0138 m is the radial distance from the lamp centreline to the irradiance probe, and L is the active emitting length. The active lengths were 0.50 m for the 95 W lamp and 0.38 m for the 60 W lamp. The existing geometric end correction was then applied:

(9)$f_{\text{endcap}}=1+\frac{R_m}{L}-\sqrt{1+{\left(\frac{R_m}{L}\right)}^2}$

The corrected total UV-C output was then obtained as:

(10)$P_{\text{total}}=\frac{P_{254}}{1-f_{\text{endcap}}}$

Because only one lamp limb was exposed during each axial scan, the nominal electrical power allocated to the emitting limb was 47.5 W for the 95 W lamp and 30 W for the 60 W lamp, assuming equal division of the nominal whole-lamp rating between the two symmetric limbs. These values represent allocated nominal power rather than directly measured electrical input. Intrinsic radiant efficiency was evaluated using Eq. (11). Relative UV-C output was normalised independently using the maximum measured value for each lamp type, as defined in Eq. (12):

(11)${\eta }_{UV-C}=\frac{P_{\text{total}}}{P_{\text{elec}}}\times 100\%$

(12)${\eta }_{\text{rel}}=\frac{P_{\text{total}}}{P_{\text{total,}max}}\times 100\%$

For the analysis of UV-C output as a function of lamp temperature, the lamp-temperature parameter was defined as the minimum of the eight external contact-sensor measurements, T_s,min. This parameter is used as an external cold-spot proxy and should not be interpreted as a direct measurement of the internal mercury cold spot.

2.4 Lagrangian UV dose and comparative microbial inactivation assessment

To quantify in-duct UV-C exposure, a Lagrangian particle-tracking framework was used to simulate axial transport of discrete particles seeded across the 0.6 m × 0.6 m duct inlet. Particle inlet locations were initialised using a stratified quasi-random approach with shuffled axes, functionally similar to Latin Hypercube sampling, to provide spatially representative cross-sectional coverage^[41]. The physical 5 × 5 irradiance-measurement locations at each axial plane were mapped to the computational cross-section. Measurements obtained at multiple axial planes between the lamp array and the duct exit were used to reconstruct the downstream irradiance field.

Each particle was transported at the mean axial airflow velocity, U_z, under a plug-flow assumption, with its cross-sectional coordinates held constant. At each axial position, the local band-integrated UV-C irradiance, E(x,y,z), was obtained from the reconstructed field, and cumulative dose was evaluated by integrating irradiance over the particle residence time:

(13)$D={\int }_0^L\frac{E\left(z\right)}{U_z}dz$

The accumulated dose was reported at the duct exit, approximately 1.3 m downstream of the final lamp array. The model neglects turbulent cross-stream dispersion, particle inertia, deposition, agglomeration, and transverse migration. It therefore provides a comparative exposure calculation rather than a complete representation of particle transport in turbulent duct flow^[34,42].

Species-specific microbial response was calculated using the first-order exponential inactivation relationship:

(14)$S=\exp (-kD)$

(14a)$\overline{S}=N^{-1}\sum _{i-1}^N{S}_i$

(15)$LRV=-{\log }_{10}\left(\overline{S}\right)$

(16)${\eta }_{\text{rem}}=(1-\overline{S})\times 100\%$

where k is the published species-specific UV susceptibility constant in m²/J and D is the cumulative dose received by each particle. This relationship is widely used for comparative UV-disinfection calculations^[34,42]. The susceptibility constants used in the present study were obtained from published sources, including the compilation by Kowalski^[25], and are summarised in Table 2. Representative inlet concentrations, C_in, were specified as DNA copies/m³ for human adenovirus, and colony-forming units (CFU)/m³ for culturable bacteria and fungi. The reporting basis of the selected concentrations is identified in Table 2. The mean survival fraction across the tracked particles was used to calculate the corresponding log-reduction value and removal efficiency, η_rem. The resulting microbial inactivation values are comparative calculations derived from measured irradiance inputs and published susceptibility constants. They were not experimentally validated using controlled bioaerosol measurements.

3. Results and Discussion

3.1 Convective-radiative heat transfer

Figure 4 shows substantial axial and circumferential variation in external lamp-surface temperature. The upstream surface generally remained cooler than the top and downstream surfaces because of its direct exposure to the incoming airflow. The axial distributions were not universally monotonic; depending on lamp rating, airflow velocity, ambient temperature, and circumferential position, local minima or maxima occurred near the measured mid station. Differences among the axial stations may reflect lamp-end effects, heat transfer through the base region, and local thermal boundary conditions. Increasing airflow velocity generally reduced external surface temperature, while increasing ambient temperature shifted the profiles upward. The 95 W lamp maintained higher and more stable surface temperatures over a broader range of conditions, whereas the 60 W lamp exhibited greater sensitivity to convective cooling. These observations demonstrate that lamp thermal behaviour should be represented using both axial and circumferential measurements rather than a single uniform surface temperature^[26,28].

Display Full Size

Download

Figure 4. Axial external lamp-surface temperature profiles at different nominal airflow velocities and ambient temperatures, where denotes axial position along the active lamp length. Panels (a), (c), (e), (g), and (i) show the 95 W lamp at T_amb = 14, 16, 18, 20, and 22 °C, respectively, while panels (b), (d), (f), (h), and (j) show the corresponding 60 W conditions. Colour denotes airflow velocity, and line style denotes the upstream, downstream, and top measurement positions.

The integrated convective and radiative external surface heat losses are presented in Figure 5. For the 95 W lamp, the total estimated surface heat loss was approximately 18-22 W and varied over a comparatively narrow range under most conditions. The 60 W lamp produced lower total surface heat loss, approximately 13-18 W, with greater relative variation across airflow velocity and ambient temperature. Convective heat loss formed the dominant component for both lamps, while the radiative contribution generally decreased when stronger convective cooling reduced external surface temperature. These quantities represent estimated heat transfer from the evaluated external lamp surface and do not include ballast losses or heat transfer through the lamp base. The narrower relative variation of the 95 W lamp is consistent with its more stable measured temperature and irradiance response^[26,28].

Display Full Size

Download

Figure 5. Integrated external surface heat loss for (a) the 95 W lamp and (b) the 60 W lamp, separated into radiative, Q_rad, and convective, Q_conv, components and grouped by nominal airflow velocity and ambient temperature. No measurement was obtained at U = 2.5 m/s and

Figure 6 presents the Kelvin-normalised angular thermal non-uniformity index as a function of Reynolds number. For both lamps, ATNI generally decreased with increasing Reynolds number, indicating that stronger airflow reduced the circumferential surface-temperature difference relative to the mean absolute surface temperature. The 95 W lamp exhibited ATNI values of approximately 0.054-0.083, while the 60 W lamp ranged from approximately 0.047 to 0.091. The relative ordering of the two lamps was condition-dependent, particularly at the lower velocities; therefore, the 95 W lamp did not exhibit consistently higher ATNI under every condition. Ambient temperature introduced secondary variation within each airflow group, but the principal trend was reduced circumferential non-uniformity with increasing airflow.

Display Full Size

Download

Figure 6. Angular thermal non-uniformity index as a function of Reynolds number for (a) the 95 W lamp and (b) the 60 W lamp, with ambient temperature indicated by colour. The mean absolute external surface temperature in the ATNI denominator was expressed in Kelvin.

The convection-to-radiation external surface heat-loss ratio, r, increased monotonically with airflow velocity for both lamps (Figure 7). At the nominal low-flow condition, r was approximately 4.6-5.0 and increased to approximately 7.1-7.6 at 2.5 m/s. This trend indicates an increasing contribution of airflow-driven convection relative to long-wave radiation. The 95 W lamp generally exhibited slightly lower r values because its higher external surface temperature increased the radiative contribution. Ambient temperature produced comparatively small variation within each velocity group. The ratio r is used here to compare the relative external surface heat-loss components and does not independently define natural-, mixed-, or forced-convection regimes.

Display Full Size

Download

Figure 7. Convection-to-radiation external surface heat-loss ratio, r = Q_conv/Q_rad, as a function of nominal airflow velocity for (a) the 95 W lamp and (b) the 60 W lamp, with ambient temperature indicated by colour. The nominal 1.1 m/s condition corresponds to a measured mean velocity of approximately 1.15 m/s used in the calculations.

Figure 8 presents the relationship between Reynolds number and Rayleigh number, with the convection-to-radiation heat-loss ratio indicated by colour. The calculations used the 0.0176 m outer diameter of an individual cylindrical lamp limb as the characteristic length. Across the investigated conditions, Re_D ranged from approximately 1.2 × 10³ to 2.9 × 10³, while Ra_D ranged from approximately 7.2 × 10³ to 1.30 × 10⁴. Increasing Reynolds number was generally accompanied by decreasing Rayleigh number and increasing r, reflecting the progressively greater contribution of airflow-driven convective cooling relative to long-wave radiation. The map is therefore presented as a comparison of operating conditions rather than as a formal classification of natural-, mixed-, and forced-convection regimes.

Display Full Size

Download

Figure 8. Reynolds-Rayleigh operating map for (a) the 95 W lamp and (b) the 60 W lamp, with colour representing the convection-to-radiation external surface heat-loss ratio, r = Q_conv/Q_rad.

3.2 UV output and intrinsic efficiency

Figure 9 shows mean 254 nm band irradiance as a function of nominal airflow velocity and ambient temperature. For the 95 W lamp, irradiance ranged from approximately 273 to 316 W/m² and remained comparatively stable over much of the investigated range. The 60 W lamp exhibited a substantially stronger dependence on operating condition, ranging from approximately 151 to 262 W/m², with output generally increasing as ambient temperature increased and decreasing as airflow velocity increased. These results indicate that the 95 W lamp maintained greater optical stability, while the 60 W lamp was more susceptible to external convective cooling, consistent with previous observations of temperature- and airflow-dependent LP-Hg lamp output^[26]. The blank cell at 2.5 m/s and 14 °C indicates that no measurement was obtained for that combination.

Display Full Size

Download

Figure 9. Mean 254 nm band irradiance as a function of nominal airflow velocity and ambient temperature for (a) the 95 W lamp and (b) the 60 W lamp. Numerical values are shown to one decimal place. No measurement was obtained at U = 2.5 m/s and T_amb = 14 °C.

Relative UV-C output, normalised independently using the maximum measured value for each lamp type, is shown in Figure 10. The 95 W lamp maintained approximately 97-100% output at 1.1 and 1.5 m/s over the investigated ambient-temperature range. Greater reductions occurred at 2.0 and 2.5 m/s under the lower-temperature conditions, but output recovered as ambient temperature increased. In comparison, the 60 W lamp exhibited a stronger and more systematic temperature dependence, increasing from approximately 58-74% at 14 °C, depending on airflow velocity, to approximately 86-100% at 22 °C. The results confirm that the lower-power lamp was more sensitive to combined airflow and ambient-temperature effects^[25-28]. Because output was normalised independently for each lamp, Figure 10 should be used to compare relative stability rather than absolute radiant output.

Display Full Size

Download

Figure 10. Relative UV-C output as a function of ambient temperature at different nominal airflow velocities for (a) the 95 W lamp and (b) the 60 W lamp. Output was normalised independently using the maximum measured value for each lamp type. UV-C: ultraviolet-C.

Figure 11 presents relative UV-C output as a function of the minimum measured external surface temperature, T_s,min. This parameter is the minimum of the eight external contact-sensor measurements and is used as an external cold-spot proxy. For the 95 W lamp, output remained comparatively stable over the investigated range, with the highest measured values occurring at T_s,min of approximately 30-32 °C. The 60 W lamp exhibited a substantially stronger increase in output with increasing T_s,min, indicating greater sensitivity to external cooling. Because of the measured axial and circumferential thermal non-uniformity, T_s,min was lower than the corresponding mean external surface temperature. The available measurements do not establish a complete high-temperature descending branch, and no overheating threshold is inferred from Figure 11.

Display Full Size

Download

Figure 11. Relative UV-C output as a function of the minimum measured external lamp-surface temperature, T_s,min, at nominal airflow velocities of 1.1-2.5 m/s. Panels (a), (c), (e), and (g) show the 95 W lamp, while panels (b), (d), (f), and (h) show the corresponding 60 W conditions. T_s,min is the minimum of the eight external contact-sensor measurements and is used as an external cold-spot proxy. Relative output was normalised independently for each lamp type. UV-C: ultraviolet-C.

3.3 Measured multi-lamp irradiance distributions

The multi-lamp arrangements were experimentally evaluated at nominal airflow velocities of 1.1 and 1.5 m/s. These conditions were selected for the subsequent dose calculations because they combined comparatively high measured lamp output with longer particle residence time than the higher-velocity conditions. They should not be interpreted as a universal optimum for all HVAC systems.

Figure 12 presents the measured 254 nm band-irradiance distributions across a representative 5 × 5 sampling plane located 800 mm downstream of the lamp arrangement. This sample plane is one of the multiple axial measurement planes used in the reconstruction of the downstream irradiance field. For both airflow conditions, the 95 W arrangements generated higher absolute irradiance than the corresponding 60 W arrangements.

Display Full Size

Download

Figure 12. Absolute 254 nm band-irradiance distributions measured at 25 positions on a representative 5 × 5 sampling plane located 800 mm downstream of the lamp arrangement: (a) nominal U = 1.1 m/s, 95 W; (b) Nominal U = 1.1 m/s, 60 W; (c) U = 1.5 m/s, 95 W; (d) U = 1.5 m/s, 60 W. Curves represent arrangements 2-1, 2-2, 3-1, 3-2, 4-1, 5-1, 6-1, and 6-2.

The measured profiles exhibited spatial peaks associated with the active lamp positions and the degree of optical overlap among the arrangements. Increasing airflow velocity from nominal 1.1 to 1.5 m/s produced a modest reduction in irradiance magnitude while retaining the principal spatial patterns. Increasing lamp count generally increased irradiance, whereas lamp position influenced peak location and the extent of lower-irradiance regions. These measured fields provided the inputs for the downstream Lagrangian dose calculations.

3.4 Lagrangian UV dose and comparative microbial inactivation assessment

Irradiance measurements obtained on 5 × 5 grids at multiple axial planes were used to reconstruct the downstream radiation field between the lamp array and the duct exit. Particle-specific UV dose was accumulated along each trajectory through this reconstructed field. Figure 13 and Figure 14 present cumulative dose and corresponding calculated SARS-CoV-2 log reduction at the duct exit for arrangement 3-2 at nominal airflow velocities of 1.1 and 1.5 m/s, respectively. The duct exit was located approximately 1.3 m downstream of the final lamp array. Results for the remaining lamp arrangements are provided in the Supplementary materials.

Display Full Size

Download

Figure 13. Cumulative Lagrangian UV dose and corresponding calculated SARS-CoV-2 log reduction at the duct exit for arrangement 3-2 at nominal U = 1.1 m/s: (a) Total UV dose, 95 W; (b) Total UV dose, 60 W; (c) Calculated log reduction, 95 W; (d) Calculated log reduction, 60 W. The duct exit was located approximately 1.3 m downstream of the final lamp array. Markers indicate the cross-sectional positions of the tracked particles. UV: ultraviolet.

Display Full Size

Download

Figure 14. Cumulative Lagrangian UV dose and corresponding calculated SARS-CoV-2 log reduction at the duct exit for arrangement 3-2 at U = 1.5 m/s: (a) Total UV dose, 95 W; (b) Total UV dose, 60 W; (c) Calculated log reduction, 95 W; (d) Calculated log reduction, 60 W. The duct exit was located approximately 1.3 m downstream of the final lamp array. Markers indicate the cross-sectional positions of the tracked particles. SARS-CoV-2: severe acute respiratory syndrome coronavirus 2; UV: ultraviolet.

At the nominal airflow velocity of 1.1 m/s, the 95 W arrangement produced higher cumulative dose than the corresponding 60 W arrangement. At 1.5 m/s, the cumulative dose decreased for both lamp ratings because of the shorter residence time. The principal spatial patterns remained similar between the two airflow conditions because the Lagrangian model retained fixed particle cross-sectional coordinates. The model therefore does not represent enhanced transverse mixing at the higher velocity. Regions of lower cumulative dose correspond to trajectories passing through lower-irradiance portions of the reconstructed field, while higher-dose regions reflect sustained exposure through overlapping lamp-radiation fields.

Application of the published SARS-CoV-2 susceptibility constant to the calculated dose distributions produced theoretical log reductions substantially greater than 6. These values should not be interpreted as directly measurable or guaranteed practical performance. Using the upper end of the reported SARS-CoV-2 airborne concentration range, 3.38 × 10³ RNA copies/m³, as the representative inlet concentration in Table 2, a 6-log reduction would correspond mathematically to 3.38 × 10^-3 RNA copies/m³, already below one RNA copy per cubic metre. Additional calculated log reduction therefore indicates that the first-order model has exceeded a practical reporting threshold rather than demonstrating experimentally verifiable complete inactivation. The values are retained for relative comparison among lamp arrangements, lamp ratings, and airflow conditions. Figure 15 extends this comparative assessment to the eight microbial species listed in Table 2 using published UV susceptibility constants.

Display Full Size

Download

Figure 15. Mean cumulative Lagrangian UV dose at the duct exit and corresponding microbial log reductions calculated using published UV susceptibility constants at airflow velocities of (a) nominal 1.1 m/s and (b) 1.5 m/s for the 95 W and 60 W lamps. The duct exit was located approximately 1.3 m downstream of the final lamp array. The calculated microbial reductions were not experimentally validated using bioaerosols. UV: ultraviolet.

The microorganisms considered span several orders of magnitude in published UV susceptibility. The model produced calculated reductions exceeding 6 log for several susceptible viral and bacterial species, while the calculated fungal-spore response was substantially lower. Aspergillus niger remained below approximately 0.05 log, while A. flavus and A. glaucus generally remained below approximately 0.4 log. The lower UV-C susceptibility of fungal spores has been associated with protective conidial-wall structures and their stress-tolerant dormant physiological state^[48], as well as UV-protective pigments and other secondary metabolites^[49]. This interpretation is consistent with experimental UVGI results showing that relatively high doses are required to inactivate Aspergillus spores^[50]. Consequently, an in-duct UVGI system designed for a susceptible virus or vegetative bacterium may not provide equivalent control of resistant fungal spores.

Increasing airflow velocity from nominal 1.1 to 1.5 m/s reduced cumulative dose because of the shorter residence time. The 95 W arrangements generally delivered higher mean dose than the corresponding 60 W arrangements, while lamp position and optical overlap influenced spatial uniformity. These comparisons demonstrate that lamp rating, airflow condition, and geometric arrangement must be evaluated together.

The microbial results should be interpreted as comparative model predictions rather than experimentally validated performance. Their absolute values depend on the reconstructed irradiance field, the plug-flow transport assumption, the published susceptibility constants, and the experimental conditions under which those constants were obtained.

The supplementary arrangements presented in Figures S1-S14 exhibited the same overall trends. Increasing lamp count generally increased mean cumulative dose, while lamp position influenced the spatial distribution and extent of lower-dose regions. Six-lamp arrangements produced the highest overall dose levels, and arrangements with greater optical overlap generally reduced the extent of local low-irradiance trajectories. Increasing airflow velocity from nominal 1.1 to 1.5 m/s reduced modelled dose across the arrangements because of the shorter particle residence time. These supplementary results support the main comparison of lamp rating, airflow condition, and lamp arrangement.

The measured results demonstrate that airflow influences in-duct UVGI performance through its effects on external lamp-surface temperature and UV-C output, while cumulative UV dose is additionally governed by particle residence time. The 95 W lamp maintained a more stable measured optical response across the investigated conditions, whereas the 60 W lamp showed greater sensitivity to changes in ambient temperature and airflow. These observations are consistent with previous studies and reviews identifying lamp temperature, airflow velocity, residence time, and lamp arrangement as important design parameters for in-duct UVGI systems^[15,28].

The present study builds on related in-duct UVGI and system-efficiency studies^[11,30,51] by combining external lamp-surface temperature measurements and spectro-radiometric output characterisation with irradiance measurements obtained at multiple axial planes and subsequent Lagrangian dose calculations. Unlike approaches that prescribe constant lamp output, the present measurement-informed framework accounts for the different measured thermal responses of the two lamp ratings. It therefore provides a consistent basis for comparing lamp rating, airflow condition, and lamp arrangement.

For the present apparatus, nominal airflow velocities of 1.1-1.5 m/s provided favourable combined irradiance and residence-time conditions. This range represents the operating envelope identified for the tested duct geometry and lamp configurations rather than a universal HVAC optimum. In practical applications, the selected duct velocity should also satisfy the required volumetric airflow, available duct area, treatment-section length, fan-energy and pressure-loss constraints, acoustic requirements, and intended microbial-control target^[15,28].

The microbial reductions were calculated using a simplified Lagrangian framework and published UV-susceptibility constants. The framework assumed plug flow with fixed particle cross-sectional coordinates and neglected turbulent cross-stream dispersion, particle inertia, deposition, agglomeration, and transverse migration. The irradiance field was reconstructed from 5 × 5 cross-sectional measurements obtained at multiple axial planes between the lamp array and the duct exit, using spatial and axial interpolation. These assumptions and reconstruction procedures introduce uncertainty into the absolute cumulative-dose and microbial-inactivation values. In addition, the published susceptibility constants were obtained under microbiological and environmental conditions that may differ from those in the present duct. Nevertheless, all lamp ratings, airflow conditions, and geometric arrangements were evaluated using the same measured inputs and calculation procedures, providing a consistent basis for comparative assessment. Higher calculated log-reduction values should therefore be interpreted as theoretical indicators of very low relative survival rather than as direct experimental measurements or guaranteed practical performance. Future work may further assess the framework through controlled microbial measurements and more detailed particle-transport analysis.

4. Conclusions

This study combined external lamp-surface temperature measurements, spectroradiometric characterisation, multi-plane irradiance measurements, and Lagrangian UV-dose calculations to assess 95 W and 60 W LP-Hg lamps in an in-duct UVGI system. The principal findings are as follows:

● Increasing airflow velocity enhanced convective cooling and generally reduced the external lamp-surface temperature. The 95 W lamp maintained a more stable 254 nm output, whereas the 60 W lamp was more sensitive to changes in airflow velocity and ambient temperature.

● The convection-to-radiation external surface heat-loss ratio increased with airflow velocity. The Reynolds number ranged from approximately 1.2 × 10³ to 2.9 × 10³, while the Rayleigh number ranged from approximately 7.2 × 10³ to 1.30 × 10⁴.

● Increasing lamp count and optical overlap generally increased the measured irradiance and calculated cumulative dose, whereas increasing airflow velocity reduced the cumulative dose because of the shorter particle residence time.

● Application of published UV-susceptibility constants produced calculated reductions exceeding 6 log for susceptible viral and bacterial species, whereas the calculated reductions for Aspergillus species generally remained below approximately 0.4 log.

Cumulative particle dose was calculated using irradiance measurements obtained on 5 × 5 grids at multiple axial planes and evaluated at the duct exit, approximately 1.3 m downstream of the final lamp array. Although the microbial reductions were calculated using published susceptibility constants rather than measured directly using bioaerosols, all lamp ratings, airflow conditions, and arrangements were assessed using the same measured inputs and calculation procedures. The framework therefore provides a consistent basis for comparing lamp rating, airflow condition, and lamp arrangement in in-duct UVGI systems.

Supplementary materials

The supplementary material for this article is available at: Supplementary materials.

Acknowledgements

ChatGPT by OpenAI was used during manuscript revision solely to assist with language editing, grammatical correction, and terminology consistency. It was not used to generate the study design, experimental data, numerical calculations, figures, or tables. All scientific interpretations, revised text, and conclusions were independently reviewed and approved by the authors. The authors are responsible for the accuracy and scientific content of the article.

Authors Contribution

Kanagalingam S: Conceptualization, data curation, formal analysis, writing-original draft.

An H: Conceptualization, writing-review & editing.

Fong YM: Data curation, writing-original draft.

Soh CB, Chien SC, Wang PC, Yu SCM: Writing-review & editing.

Conflicts of interest

The authors declare no conflicts of interest.

Ethical approval

Not Applicable.

Consent to participate

Not Applicable.

Consent for publication

Not Applicable.

Availability of data and materials

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Funding

The authors would like to thank the Ministry of Education (MOE) of Singapore (Grant No. T2EP50221-0042).

Copyright

© The Authors 2026.