Evaluating Disinfection Performance and Energy Efficiency of a Dual-Wavelength UV-LED Flow-Through Device for Point-of-Use Water Treatment

Abstract

Ultraviolet-light emitting diodes (UV-LEDs) offer several advantages over conventional mercury-based UV lamps, including wavelength selectivity, compact size, design flexibility, instant on/off, power output adjustment, and mercury-free operation. These features position UV-LEDs as ideal candidates for point-of-use (POU) water disinfection systems, particularly in decentralized or resource-limited environments. In this study, we evaluated the microbial inactivation performance and energy efficiency of a bench-scale flow-through UV-LED POU system using indigenous heterotrophic plate count (HPC) bacteria, E. coli, and MS2 bacteriophage. The system was tested under various flow rates (1–4 L/min) and wavelength configurations (265 nm, 278 nm, and dual-wavelength combinations). MS2 bacteriophage was further used in collimated beam testing to validate UV-fluence-response curves and to estimate delivered doses in the flow-through POU device. HPC inactivation was enhanced under dual-wavelength conditions, suggesting wavelength-specific synergy, while E. coli showed high susceptibility across all wavelength configurations, achieving >2-log inactivation at significantly reduced UV-LED power (1/6 of that required for HPC) even at 4 L/min. Specific energy consumption analysis showed energy demands as low as 0.032–0.053 kWh/m³ for achieving 4-log inactivation of E. coli, with an estimated annual operating cost for UV-LED irradiation below $1.70. These findings demonstrate the potential of UV-LED-based POU devices as safe, energy-efficient, and cost-effective technologies for decentralized water treatment.

1. Introduction

Centralized municipal water treatment systems in the United States (U.S.) typically provide disinfected tap water that complies with regulatory microbial standards. Chlorine is widely used for residual disinfection throughout the distribution system, effectively suppressing microbial regrowth during transport. However, when tap water is stored or heated within the domestic hot water system (DHS), which can include water heaters, hot water piping network, and point of use (POU) devices such as faucets, the chlorine residual can rapidly dissipate due to thermal and chemical degradation [1,2]. This loss of residual disinfectant creates favorable conditions for microbial growth in plumbing systems, especially within the DHS. Factors such as plumbing materials, stagnation, and biofilms have been shown to influence microbial colonization and subsequent release into the water supply. Additionally, sporadic microbial contamination events may occur due to cross-connections, backflow, aging infrastructure, or seasonal changes in water chemistry, which further underscore the need for supplemental treatment barriers [3,4].

POU water treatment technologies offer a promising strategy to address water quality deterioration within household plumbing [5]. Among available approaches, ultraviolet (UV) disinfection is a well-established, chemical-free method that effectively inactivates a wide range of microorganisms, including bacteria, viruses, and protozoa [6,7,8]. Conventional UV disinfection systems commonly utilize low-pressure emitting at 254 nm or medium-pressure mercury lamps emitting a broad spectrum of UV. Although these systems are effective for microbial disinfection, they face several limitations in water treatment applications, such as fragility, extended warm-up time, mercury-related safety concerns, and a fixed spectral output. The mercury issue, in particular, is facing a trend of strengthening regulations internationally (i.e., the Minamata Convention on Mercury), which requires countries to phase-out or reduce the use of mercury in certain products. The U.S. also signed the Minamata Convention on 6 November 2013.

Ultraviolet light-emitting diodes (UV-LEDs) have emerged as a promising alternative to mercury-based UV systems. UV-LEDs offer advantages such as wavelength selectivity, compactness, instantaneous operation, long lifespan, and mercury-free operation [9,10]. These features have enabled UV-LEDs to be incorporated into small-scale POU water treatment devices [11]. Previous studies have shown that UV-LEDs can effectively inactivate a broad range of microorganisms [12,13,14,15,16,17]. Furthermore, dual-wavelength UV-LED configurations (i.e., combining emissions at two germicidal wavelengths) have gained attention as a strategy to expand molecular damage mechanisms [18,19,20,21,22,23,24]. Studies suggest that paring wavelengths may enhance disinfection by targeting both nucleic acids and proteins, although results generally show additive rather than synergistic effects within the UV-C band. However, most prior UV-LED studies have focused on batch disinfection using collimated beam conditions.

Previous studies have investigated microbial inactivation using UV-LEDs in flow-through reactor systems focusing on the effects of hydrodynamics, UV dose distribution, and residence time on treatment efficacy [25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40]. Among these, the use of dual-wavelength configurations using flow-through model systems remains largely unexplored, representing a critical gap in current knowledge. Understanding how multi-wavelength UV-LEDs perform under realistic flow conditions and water matrices are essential for developing novel POU devices that can be deployed effectively in households.

This study evaluates the disinfection performance and energy efficiency of a dual-wavelength UV-LED flow-through POU device treating domestic hot water. Indigenous heterotrophic plate count (HPC) bacteria and laboratory-spiked E. coli were used to assess microbial inactivation under multiple UV configurations and flow rates. Collimated beam test using MS2 bacteriophage validated fluence-response relationships. Specific energy consumption (SEC) was calculated to estimate the feasibility of deployment in residential settings. By investigating realistic water matrices and flow conditions, this study contributes to a better understanding of UV-LED disinfection for household tap water safety.

2. Materials and Methods

2.1. Microorganisms Preparation and Enumeration

Indigenous HPC bacteria were prepared by collecting domestic hot water from a laboratory faucet at the U.S. EPA Andrew W. Breidenbach Research Center in Cincinnati, Ohio. The water was collected in pre-sterilized 2 L glass bottles. At the time of collection, the water temperature was approximately 56 °C with the negligible chlorine residue. The water was cooled to room temperature prior to use.

E. coli (ATCC 25922) was cultured by inoculating two sterile 50 mL conical centrifuge tubes, each containing 30 mL of sterile tryptic soy broth (TSB; Fisher Scientific, Waltham, MA, USA). The tubes were briefly vortex mixed and incubated overnight in a shaker at 120 rpm 36 °C. After incubation, cultures were washed three times by centrifugation at 3000× g for 10 min using 30 mL of sterile phosphate buffer (pH 7.2). After removing the supernatant from the tubes, one of the final centrifuged pellets was resuspended with 10 mL of Butterfields buffer and merged into the other tube and stored at 4 °C until use.

MS2 bacteriophage (ATCC 15597-B1) stock cultures were prepared by mixing 30 mL of sterile TSB with MS2 and inoculating with 6 mL of mid-log phase E. coli F_amp host (ATCC 700891). This suspension was added to 90 mL of 0.7% tryptic soy agar (TSA; Fisher Scientific, Waltham, MA, USA) containing antibiotics (0.015 mg/mL each of streptomycin and ampicillin). The mixture was overlaid (6 mL per plate) onto 100 mm plates of 1.5% TSA also supplemented with the same antibiotics. After solidification, the plates were incubated at 36 °C for 24 h. Bacteriophages were harvested by adding 5 mL of TSB to each plate and incubating at room temperature for 2 h. The plates were swirled 20 times, and the broth was pooled, centrifuged at 3000× g for 5 min, and filtered through a 0.2 μm filter. The clarified supernatant was transferred into sterile containers and stored at 4 °C until use.

After each UV disinfection test, water samples were serially diluted in sterile Butterfield’s buffer. HPC bacteria were enumerated using the spread plate culture method on R2A agar (BD Biosciences, Franklin Lakes, NJ, USA) and incubated for seven days at 28 °C. E. coli was enumerated using the spread plate culture method on nutrient agar (BD Biosciences) and incubated at 36 °C for 18–24 h. MS2 bacteriophage was enumerated using the double agar layer method [41]. Briefly, 1 mL of each dilution was added to a soft top agar (3 mL) containing the E. coli host in mid-log phase, poured onto a TSA base, allowed to solidify, and incubated at 36 °C for 18–24 h.

2.2. Experimental Setup

A bench-scale flow-through UV-LED POU system (HEPS Inc, Seongnam, Republic of Korea) consisted of two cylindrical UV reactors deployed in series, two UV-LED plates, and a controller integrated with a power supply unit (Figure 1A). The UV reactors were carefully designed to maximize UV-LED output. Each reactor was made of polytetrafluoroethylene (PTFE), a physically and chemically robust material that allows treatment of a wide variety of waters and provides high UV reflectivity. The reactor had an inner diameter of 25 mm and a working volume of 59 mL. A UV-LED plate was mounted at the end of each reactor to provide UV irradiation in the opposite direction (counter-current) as the water flows. The LED chips were separated from the reactor interior by a 6 mm thick synthetic fused silica window (Schott North America, Inc., Rye Brook, NY, USA) to protect them while maximizing UV transmittance (98.6 ± 0.8% in the wavelength range between 250 nm and 300 nm). The system was connected to a reservoir containing feed water, which was pumped through the reactors (AC-5C-MD; March Mfg., Inc., Glenview, IL, USA). Pumping speed was controlled manually by measuring the flow rate using a flow meter (F440; Blue-White Industries, Huntington Beach, CA, USA). As water flowed through the reactors, it was irradiated by the UV-LEDs (Figure 1B). Samples were collected before and after treatment to assess microbial inactivation.

2.3. UV Irradiation

Two types of UV-LEDs (Photon Wave Co., Ltd., Yongin, Republic of Korea) with peak emission wavelengths of 265 nm and 278 nm were used. Emission spectra of each UV-LED were measured using a spectroradiometer (ILT 950, International Light Technologies, Peabody, MA, USA) (Figure S1). According to the manufacturer’s specifications, the nominal radiant flux of each UV-LED is 37.5 ± 3.7 mW at 265 ± 3 nm and 42 ± 4.2 mW at 278 ± 3 nm. Six LED chips arranged in a 2 by 3 array were mounted on each UV-LED plate (Figure 1C). Three types of UV-LED plates were employed for the study, which were two single-wavelength plates (265 nm or 278 nm only) and one dual-wavelength plate with alternating rows of 265 nm and 278 nm LEDs. Each UV-LED plate was operated at a current load of 350 mA. Using these UV-LED plates, four distinct wavelength combinations were tested (

Table 1. UV-LED wavelength combinations.

Combination	Reactor #1	Reactor #2	Configuration Type
1	265 nm	265 nm	Sequential single-wavelength
2	278 nm	278 nm	Sequential single-wavelength
3	265 nm/278 nm	265 nm/278 nm	Simultaneous dual-wavelength
4-1	265 nm	278 nm	Sequential dual-wavelength
4-2	278 nm	265 nm	Sequential dual-wavelength (reversed)

). UV intensity was adjusted via the controller by selectively activating specific LED chips. Flow rate was varied to control hydraulic residence time (HRT), which directly influenced UV fluence received by the microorganisms.

2.4. UV-LED Flow-Through System Inactivation Tests

For all inactivation tests, 15 L of domestic hot water was used as the feed water. For HPC experiments the domestic hot water was allowed to cool to room temperature, and the indigenous bacteria present in the feed water were used in all experiments. For E. coli and MS2 tests, the domestic hot water was autoclaved, cooled, and spiked with respective stock cultures to achieve approximately 10⁶ CFU/mL (E. coli) or 10⁶ PFU/mL (MS2). Microbial concentrations in influent (before UV) and effluent (after UV) samples were determined as described in Section 2.1. Log inactivation (I) was calculated using the following equation:

$$I= -{log}_{10}{\displaystyle \frac{N}{N_0}}$$

(1)

where I is the log inactivation, N₀ is the concentration before UV treatment (CFU/mL or PFU/mL), and N is the concentration after UV treatment. Each wavelength combination was tested at three flow rates: 1, 2, and 4 L/min. It is worth noting that although flow rates above 4 L/min were not tested directly due to benchtop setup limitations, the results obtained provide a relevant performance basis for estimating reactor behavior under higher-flow scenarios. All tests were performed in triplicate with duplicate plating for each sample.

2.5. Specific Energy Consumption

Specific energy consumption (SEC) was calculated as the energy used per volume of treated water to achieve a target log inactivation, at each UV wavelength combination and flow rate.

$$\mathrm{S}\mathrm{E}\mathrm{C}={\displaystyle \frac{I_t}{I_d}}\times {\displaystyle \frac{P_{265}+P_{278}}{Q}}$$

(2)

where I_t is the target log inactivation, I_d is the actual log inactivation, P₂₆₅ and P₂₇₈ is the energy consumption (based on manufacturer information) for UV-LED emitting at 265 nm (2.1 × 10⁻³ kW/chip) and 278 nm (2.2 × 10⁻³ kW/chip), respectively, and Q is the flow rate (m³/hr). This calculation allowed comparison of energy efficiency across different UV wavelength combinations and flow conditions. It should be noted that the calculation assumes proportionality between the UV-LED energy consumption and delivered fluence, which may not strictly hold under various operating conditions.

3. Results and Discussion

3.1. Biodosimetry and Fluence Estimation

To develop a fluence-response relationship, a custom-built collimated beam apparatus equipped with a low-pressure mercury lamp emitting at 254 nm (XX-15S, UVP, Upland, CA, USA) was used (Figure S2). The UV fluence was calculated following standard protocols [42], incorporating corrections for sample depth, incident irradiance (measured by a calibrated radiometer, ILT 2400, International Light Technologies, Peabody, MA, USA), Petri factor, reflection factor, divergence factor, and water absorbance at 254 nm measured by a spectrophotometer (UV 5; Mettler Toledo, Columbus, OH, USA).

MS2 bacteriophage, suspended in autoclaved and tempered domestic hot water, served as a biodosimetric surrogate due to its well-characterized UV sensitivity. For each exposure, 10 mL of the MS2 suspension was placed in a sterile Petri dish (52 mm inner diameter) and irradiated for predetermined durations to deliver fluences of 20, 40, and 60 mJ/cm². Each fluence condition was tested in triplicate. Incident irradiance was measured before and after each test to confirm stability, and sample temperature was monitored to ensure that it remained constant during irradiation. MS2 inactivation was then measured under different flow rates and wavelength combinations using the UV-LED POU system.

The resulting fluence-response curve obtained from the collimated beam test was used to determine the reduction equivalent dose at 254 nm (RED₂₅₄) for the flow-through UV-LED POU system (Figure 2). By matching the log inactivation achieved by the flow-through UV-LED reactor system to the MS2 fluence-response curve, equivalent fluences (RED₂₅₄ values) were inferred for each flow rate and wavelength configuration tested. As expected, higher flow rates reduced HRT, which in turn lowered UV fluence and resulted in decreased inactivation. The inactivation data from the flow-through system were consistent with the fluence-response trend observed under collimated beam conditions, confirming that the estimated UV fluences in the reactor aligned with theoretical expectations. It is worth noting that while HRT decreases inversely as the flow rate increases by multiples, RED₂₅₄ does not decrease at the same rate as HRT. This will be further discussed in the next section.

3.2. Inactivation of HPC Bacteria and E. coli

The initial concentration of HPC bacteria in the collected domestic hot water ranged from 10⁵ to 10⁶ CFU/mL, depending on collection date. UV transmittance (UVT) of the tap water at 254, 265, and 278 nm were consistently high (95–97%), indicating minimal light attenuation and suitable conditions for UV-based disinfection (Figure S3).

All UV wavelength configurations demonstrated comparable inactivation performance for HPC bacteria, except the 278–278 nm configuration, which yielded consistently lower log inactivation (Figure 3A). This reduced efficacy can be explained by two main factors. First, compared to 265 nm UV irradiation, 278 nm is less effective at damaging microbial DNA due to reduced photon energy and lower absorption by nucleic acids, whose peak absorbance lies around 260–265 nm [14,18,30]. As a result, 265 nm generally exhibits higher microbial inactivation rates. Second, the heterogeneous composition of the HPC community likely amplifies wavelength-specific effect, as some species or strains may be more resistant to damage induced by longer wavelengths. Interestingly, dual wavelength configurations, both simultaneous (265/278 nm per plate) and sequential (265→278 nm), led to slightly higher inactivation of HPC bacteria compared to single-wavelength configurations. This suggests potential synergistic effects, possibly due to multi-targeted damage across different biomolecules such as DNA, proteins, and membranes. These components may each absorb different UV wavelengths more efficiently, thereby enhancing overall microbial damage. Prior research supports such multi-wavelength synergy, particularly in complex microbial communities with varied UV sensitivities [6].

Compared to HPC bacteria, E. coli exhibited markedly higher sensitivity to all UV configurations (Figure 3B). Inactivation testing was conducted using only one LED chip per plate (1/6 the maximum power) due to the high susceptibility of E. coli; full power settings resulted in complete inactivation beyond the detection limit. Despite this reduced intensity, all wavelength configurations achieved >5-log inactivation at 1 L/min and >2-log inactivation at 4 L/min. This confirms that E. coli is highly susceptible to UV irradiation, even under lower power conditions. Because only one UV-LED chip was used, simultaneous dual-wavelength exposure was not applicable. However, sequential dual-wavelength testing was performed, including trials where the order of UV wavelengths was reversed (i.e., 265 → 278 and 278 → 265 nm). Regardless of the order, the inactivation results were similar. Notably, unlike HPC bacteria, E. coli did not exhibit any clear synergistic effects from multi-wavelength exposure, indicating that wavelength-specific synergy may be organism-dependent. Moreover, 278 nm irradiation showed comparable inactivation performance to other wavelength configurations, which was also observed for MS2. For these clonal test organisms (also known as pure cultures), wavelength-dependent differences were less pronounced, likely because the higher intensity of the 278 nm LEDs (Section 2.3) partially compensated for their lower germicidal efficiency, resulting in only modest differences compared to 265 nm.

Increasing flow rate from 1 to 4 L/min reduced HRT per reactor from 3.5 to 0.9 s. As expected, this reduction in exposure time diminished inactivation efficiency due to a lower delivered UV fluence. Nevertheless, the system still achieved 2.5–3 log inactivation of HPC bacteria even at the highest flow rate, demonstrating substantial disinfection capacity under high-throughput conditions. However, for all microorganisms tested, the increase in inactivation with longer residence time was not linearly proportional, with HPC exhibiting a more apparent deviation from linearity. While theoretical fluence increased approximately four-fold from 4 L/min to 1 L/min, log inactivation improved by only 1.3–1.5 times (up to 2.2 times for 278 nm). This nonlinearity may result from hydrodynamic conditions within the reactor. Specifically, laminar flow, short-circuiting, poor mixing, and shadowing effects can lead to uneven UV dose distribution, where some bacteria receive less exposure than others. The estimated Reynolds numbers were approximately 845 and 3390 at 1 L/min and 4 L/min, respectively, indicating a transition from laminar to weakly turbulent flow, which may influence UV exposure uniformity.

Additionally, the heterogeneous composition of the HPC bacterial community likely contributes to the observed dose–response behavior. HPC bacteria include a variety of species with diverse UV tolerances. Some may aggregate, shield each other, or possess efficient repair mechanisms, which can reduce the net effect of increased fluence. In contrast, E. coli represents a uniform, UV-sensitive population that conforms more closely to expected dose–response kinetics.

3.3. Specific Energy Consumption and Implication

Based on experimental results, SEC of the UV-LED POU device was estimated for achieving 4-log inactivation (i.e., 99.99% inactivation) of E. coli under different flow rates and UV wavelength combinations. SEC was calculated using the actual log inactivation achieved, power consumption of UV-LED chips, and flow rate. As shown in Figure 4, higher flow rates resulted in lower SEC, primarily because the inactivation efficacy remained relatively high despite shorter residence times at increased flow rates. For example, at 4 L/min, although HRT was significantly reduced, the device still achieved >2-log inactivation of E. coli using only one UV-LED chip per reactor (1/6 of maximum power). This performance translates to lower energy cost per cubic meter of treated water. This trend suggests that the inactivation efficiency of the UV-LED POU device does not decrease proportionally with shorter residence time, likely due to effective UV fluence delivery even at high flow rates. Improved mixing and more favorable hydrodynamic conditions at higher flow may contribute to more uniform fluence distribution within the reactor. However, further investigation is required to precisely quantify fluence profiles and identify any flow-dependent enhancement mechanisms. In addition, it is worth noting that the SEC analysis is based on simplifying assumptions that may vary under different operating conditions, such as flow regimes, reactor reflectivity, and non-uniform fluence distributions. Therefore, the reported SEC values should be interpreted as indicative benchmarks for comparing wavelength configurations and flow rates, rather than as absolute measures of energy efficiency.

Annual energy consumption was estimated assuming a household water usage of 190 m³/year (approximately 520 L/day) and the 2023 U.S. national average electricity price of $0.16 per kWh [43,44]. The current UV-LED POU device only consumes 6 to 10 kWh a year and removes 99.99% E. coli, therefore the annual operating cost associated with UV-LED power consumption ranged from $0.96 to $1.62 depending on the flow rate. These values reflect only the energy used by the UV-LED chips and do not include power for other components (e.g., pump, controller) or maintenance-related costs, such as light source replacement and. Nevertheless, the low energy demand highlights the potential for energy-efficient and decentralized water treatment. Overall, these findings indicate that the UV-LED POU system offers a highly energy-efficient solution for microbial disinfection, particularly for E. coli. The low power demand and minimal operating costs further enhance its potential for practical implementation in residential and decentralized water treatment applications. In particular, assuming the commercialization of UV-LED POU devices for residential use (e.g., showerheads and faucets), the pump used to evaluate the applicability of UV-LED POU devices can be replaced with a resistor that allows the desired flow rate, and the complex controller can be replaced with a simple DC adapter. The LED module with integrated heatsinks and UV reactor will also be compact through innovative designs. With these auxiliary components reduced, maintenance costs will be negligible.

4. Conclusions

This study demonstrated the disinfection performance of a flow-through UV-LED POU system using both indigenous HPC bacteria and spiked E. coli under various operating conditions. The results highlight the system’s strong disinfection capabilities, even at high flow rates and reduced UV power. Combined UV-LED wavelength configurations exhibited a slight synergistic effect in the inactivation of HPC bacteria, likely due to varied action spectra across a diverse microbial community. In contrast, E. coli, representing a more uniform and UV-sensitive organism, showed no measurable synergistic benefit from dual-wavelength exposure. Despite using only one-sixth of the maximum UV power for E. coli tests, the system achieved >2-log inactivation even at 4 L/min, underscoring its high efficacy and energy efficiency. Specific energy consumption analysis showed that operating the UV-LED system at higher flow rates reduced energy usage per unit volume of water treated, making it more practical for real-world applications. The estimated annual energy consumption ranged from 6 to 10 kWh, corresponding to an operational cost of $0.96–$1.62, assuming typical residential water usage and electricity rates. These values reflect only the power required by UV-LEDs and exclude auxiliary components or maintenance. Overall, the findings support the feasibility of UV-LED-based POU devices for household water disinfection. The demonstrated performance, energy efficiency, and compact design suggest that UV-LED systems hold strong potential for decentralized water treatment, especially in settings lacking access to reliable centralized infrastructure. Future research should focus on fluence distribution modeling, microbial regrowth analysis, and long-term system reliability, including maintenance costs associated with biofilm control to further optimize device performance.