Photolysis at the Speed of Light: Chemical-Free Degradation of Trace Organic Contaminants by Bespoke Photolysis Using High-Intensity Ultraviolet C Light-Emitting Diodes

Abstract

Ultraviolet C light-emitting diodes (UV C LEDs) have demonstrated effectiveness in disinfection applications and proven suitability at scale for disinfection of municipal wastewater and drinking water. Technological advances in materials design and electrical efficiency have made high-intensity light delivery by UV C LEDs a reality and now poise these traditionally disinfection systems to serve a dual purpose for targeted remediation of trace organic contaminants (TrOCs). This work investigated the effectiveness of UV C light emission tailoring on the photodegradation dynamics of select TrOCs. Degradation kinetics and quantum yields of target compounds under 275 nm irradiation were governed by molar absorbance and chemical structure, and kinetics followed estrone (E1) > tryptophan > caffeine ≈ pCBA > urea. Secondary experiments compared the efficacy of a 275 nm UV LED and a medium-pressure mercury vapor (MP UV) system for photodegradation of two steroid estrogens, E1 and 17β-estradiol (17β-E2). Use of the 275 nm UV LED system substantially reduced fluence requirements and, in the case of 17β-E2, energy requirements, to achieve 90% degradation of the target compounds. Liquid chromatography–tandem mass spectrometry analysis of an E1 photodegradation product showed that the UV C LED system was more effective in eliminating both E1 and its associated photoproduct as compared to the MP UV system. This work demonstrates the effective use of UV LEDs for tailored photolysis of TrOCs and provides evidence for their use potential in applications outside of water disinfection.

1. Introduction

The widespread use of UV light in water treatment has long been recognized as a safe and effective method to inactivate pathogenic organisms and ensure water safety. UV water treatment systems were first produced commercially in the 1930s and are now commonplace in municipal and industrial applications. Commercial UV treatment systems currently rely on the use of mercury-based UV light-emitting lamps, which have historically been the only industry-ready UV treatment devices commercially available. The use of mercury in commercial and consumer products is being sequentially phased out in the European Union via Regulation-2024/1849, and the United Nations Minamata Convention, currently supported by 128 signatories, is set to ban mercury mining by 2032. Mercury has long been known for its neurotoxicity effects, and the realities of mercury mining can result in long-lasting environmental and human health effects to both miners and those living in affected regions. , It is therefore necessary not only to evaluate and validate technological water treatment alternatives to comply with current and future regulations but also to ensure safe development and sourcing of materials used in consumer and industrial products.

The emergence of solid-state light emitters (SSLEs), such as light-emitting diodes (LEDs), revolutionized consumer technology industries in the mid-20th century. , Today, SSLE devices can be found in almost all consumer and industrial devices requiring illumination. The versatility, adaptability, and efficiency of SSLEs have opened the door to illumination technologies that would not be possible with historical lighting alternatives. While visible- and infrared-light LEDs have been commercially available since the 1970s, UV-emitting solid-state light sources, or UV LEDs, have only become available to consumer and industrial markets in the past two decades and are now poised to disrupt the water treatment industry analogously to SSLEs in technological applications.

UV LEDs have many inherent properties that provide benefits over commercially available mercury vapor systems. ,, UV LED devices are made of semiconductor chips, which can be densely packed into UV reactors to provide effective and uniform treatment. Their design versatility and ability to run on battery power also make them an adaptable measure for point-of-use applications such as at-the-tap water disinfection for isolated areas (e.g., campgrounds and portable water disinfection devices). The dimmable nature of UV LEDs allows ramping up or down of UV light intensity in cases where water quality is affected by seasonal dynamics, whereas mercury lamps are designed to operate at a stable, consistent brightness, as frequent lamp cycling truncates the lifecycle of conventional systems. UV LEDs have proven effective for disinfection in small-scale and decentralized drinking water systems, biofilm-based Pseudomonas aeruginosa, and most recently for a full-scale 2-MLD wastewater treatment system.

UV LED wavelength emission can be tailored, depending on the treatment application. Unlike commercially available mercury vapor lamps, which emit monochromatically at 253.7 or 365 nm, or polychromatically between 200 and 600 nm, UV LEDs emit light within narrow wavelength bands (i.e., full width at half-maximum (fwhm) ∼ 10–15 nm), which are dependent on the band gap energy of the employed semiconductor materials. Most current-generation UV LEDs are composed of Al _x Ga_1–x N semiconductor materials, which are nontoxic, UV-transparent, and conductive. The composition of the Al _x Ga_1–x N material can be theoretically adjusted to achieve wavelength emission regions ranging from 210 to 365 nm. The versatility in emittance wavelength of Al _x Ga_1–x N-based UV LEDs makes them adaptable to treatment applications requiring light in wide ranges of the electromagnetic spectrum (i.e., UV A–UV C). The potential for bespoke UV LED treatment through specific wavelengths has demonstrated flexibility to target-specific biomarkers and thereby lower the required UV fluence to achieve pathogen inactivation objectives. −

Photolytic degradation of chemical contaminants can be optimized by tailoring UV LED wavelength emission to the absorbance spectrum of a compound of interest. Photons emitted within the typical UV treatment region of the electromagnetic spectrum (i.e., 300–200 nm) have energies ranging from 4.13 to 6.2 eV or 95.2–143 kcal mol^–1, which is sufficient to break most chemical bonds (e.g., C–H bonds in alkanes have bond dissociation energies (BDEs) ranging from 94 to 99 kcal mol^–1 while BDEs in C–C bonds range from 77 to 83 kcal mol^–1 in aliphatic compounds). Wavelength-specific photolysis of chemical contaminants has been limitedly investigated, ,− but technological advancements and significant improvements in power output have made delivery of high-intensity UV C light from LED sources a reality. This advancement along with wavelength tunability and development of high-density reactors positions UV LEDs as disruptive technology capable of applications beyond disinfection.

Here, we evaluate the efficacy of UV C LEDs to degrade TrOCs through photolysis-initiated reactions. Many TrOCs (organic compounds that generally occur in the environment at low (e.g., ng L^–1 and μg L^–1) concentrations and cause environmental harm) are difficult to remove and abate using traditional water treatment technologies and can cause unintended issues when interacting with conventional chemical disinfectants. Evaluation of alternative chemical-free degradation technologies to target TrOCs is of global interest.

We also introduce a novel, high-intensity flow-through UV C LED reactor that provides high fluence delivery (i.e., >1000 mJ cm^–2) in a matter of seconds and compare results from this novel device to a parallel experiment using the typical collimated beam apparatus, which is ubiquitous in bench-scale UV research. Finally, we evaluate and compare UV C LED treatment effectiveness and electrical efficiency to a traditional MP UV system under collimated beam setups and discuss future potential for UV C LEDs as a chemical-free water treatment tool and replacement for mercury vapor lamps. Wavelength-dependent interactions between many TrOCs and light are not well-understood across the UV C spectrum, but there are indications that, in some cases, photolysis alone could help mitigate unintended byproduct formation. The future use and design of tailored UV LED reactors could help reduce reliance on added chemicals such as ozone or peroxide in water treatment applications while also removing mercury from the treatment process.

2. Materials and Methods

2.1. Chemicals and Reagents

ACS-grade 4-chlorobenzoic acid (pCBA), caffeine, tryptophan, and urea were sourced from Sigma-Aldrich. 1 g L^–1 stocks of E1 and 17β-E2 (dissolved in methanol or acetonitrile) were obtained from Cerilliant Co. (Texas, USA). Working solutions of each of the target compounds were prepared from the neat chemical standards at a concentration of 1 g L^–1 in water, apart from E1 and 17β-E2, which were spiked directly from their purchased 1 g L^–1 stock in methanol and acetonitrile (respectively). LC-MS-grade methanol, LC-MS-grade formic acid, and Optima acetonitrile used in LC-MS/MS and HPLC-UV/vis analyses were purchased from Fisher Scientific. Laboratory-grade water with a resistivity of 18.2 mΩ cm^–1 and a total organic carbon (TOC) of <5 μg L^–1 was sourced from a Milli-Q purification system (Reference A+, Millipore) and used for the preparation of all chemical solutions and reagents.

2.2. Model Compound Selection

Primary experiments (Part 1) assessed the photodegradation dynamics of a range of organic compounds including pCBA, commonly used as a probe compound for estimation of ^•OH radical exposure in AOP systems; , urea, a waste product in municipal wastewater, industrial, and agricultural effluent streams; tryptophan, an amino acid indicative of microbial activity in biological systems; , caffeine, an alkaloid and a probe compound for ^•OH and reactive chlorine species (RCS) and a tracer for anthropogenic contamination; and E1, a steroid estrogen and relevant contaminant of concern in municipal, industrial, and agricultural waste streams. The range of compounds assed here was chosen based on differences in structure, molar absorbance in the 275 nm emittance range, photosensitivity to UV C irradiation, and environmental relevance. These experiments were conducted using a novel, 275 nm UV C LED flow-through reactor (described further in Section ).

Secondary experiments (Part 2) were conducted using two steroid estrogens (E1 and 17β-E2) as target compounds to evaluate and compare their photodegradation under 275 nm UV C LED and MP UV irradiation within collimated beam setups. Collimated beam testing is the current standard approach for bench-scale UV work, − and this approach was selected to provide appropriate context with industry-standard contaminant degradation work using UV light. This testing provided an opportunity to compare the conventional bench-scale UV exposure device to the novel flow-through system, which is optimized for studying high-intensity photolysis or photooxidation processes that require higher fluences. E1 and 17β-E2 were chosen for secondary experiments as they are well-known TrOCs that are of environmental concern and exhibit differing sensitivity to UV C irradiation. − The chemical structure, CAS number, and molar mass of each of the described compounds are given in Figure SI-1, while their molar absorbance spectra can be found in Figure .

1.

Molar absorbance of target compounds and relative emittance of the 275 nm UV C LED reactor.

Molar absorbance across wavelengths was calculated as per the Beer–Lambert law (eq ):

$$\epsilon =\frac{A}{lC}$$ 1

where A is the absorbance of a compound (unitless), ε is the molar absorptivity (M^–1 cm^–1), C is the concentration (M), and l is the cell path length (cm).

2.3. Reactor Characteristics and Design

2.3.1. Flow-Through UV C LED Reactor

Primary experiments were conducted using a bench-scale DC UV C LED reactor, the PearlAqua ThinFilm 534A2 (AquiSense Technologies), which consists of four main components: a light source assembly, a safety shield, a flow cell, and a driver assembly. The reactor contains an array of 192 UV C LED chips, which emit light at a 275 nm peak emission wavelength. The reactor was equipped with a 4 cm × 15 cm × 2.8 cm quartz laminar-flow cell with a volume of 4.23 cm³. The emittance spectrum of the reactor can be seen in Figure , while the laminar-flow cell is shown in Figure SI-3.

The driver assembly houses the power supply, which converts AC electrical input to DC electrical output for the lamp assembly and transmits data via MODBUS. An intermediate coolant loop is used to cool the driver assembly, and it routes the coolant (tap water) from the cooling block through the lamp assembly. A laptop is connected to the driver assembly and controlled the lamp output via the user interface. The interface enables control of the light source assembly and the power level and senses the temperature of the internal electrical components. Coolant water was supplied via connecting the cooling block to a barbed faucet tap using rubber hosing. The influent sample water was controlled using a peristaltic pump, which was calibrated to the range of flow rates (Table SI-2).

2.3.2. Collimated Beam Reactors

Secondary experiments were carried out using an AquiSense PearlBeam unit (UV LED experiments) with an irradiance of 3.62 mW cm^–2 (3.5 cm from the edge of the collimator) and 275 nm peak emittance and a Calgon Carbon collimated beam unit equipped with a 1000 W polychromatic MP UV lamp with an irradiance of approximately 2.80 mW cm^–2 measured 3.5 cm from the edge of the collimator and 200–600 nm spectral emittance (see Figure for the emittance spectrum).

3.

Degradation (A, C) and first-order kinetic (B, D) plots of E1 and 17β-E2 under 275 nm UV LED irradiation from bench-scale 275 nm LED or MP UV collimated beam reactors. Error bars represent one standard deviation from the mean value. Note the differences in x-axis values between compounds. Experimental parameters: pH = 7, 2 mM PO₄ ^2– buffer; [E1]₀ = 10 μg L^–1; [17β-E2]₀ = 10 μg L^–1. n = 3. Molar absorbance and relative emittance (E).

2.4. Experimental Procedures

2.4.1. UV C LED Photolysis of Target Compounds in a 275 nm High-Throughput Flow-Through Reactor

The starting concentration of the selected target compounds in primary experiments varied due to the range of analytical methods required for quantitation and solubility limitations (Table SI-4). The aqueous solution of the target compound was connected via rubber tubing to a water pump and the reactor. The starting sample volume was 2 L, which was spiked with a concentrated working solution of target analytes to achieve required concentrations. The samples were collected from the outlet of the reactor and circulated multiple times (Table SI-2) to achieve a range of UV fluences (as determined by uridine actinometry; see methods below). Samples were taken during each recirculation round. Experiments were carried out in triplicate. Between each sample run, an ultrapure water rinse of more than 90 s (approximately 150 mL) was completed at a flow rate of 100 mL min^–1. The sample volume collected for analysis was 10 mL. All experiments were conducted at room temperature (20–21 °C).

Samples undergoing LC-MS/MS quantitation (E1, 17β-E2, caffeine, and tryptophan) were fortified with their associated internal standards prior to analysis. All data at or above calculated method detection limits (MDLs) were reported as their quantitated value. Only data at or above calculated limits of quantitation (LOQs) were used in kinetic calculations. MDLs and LOQs of target compounds are given in Table SI-4.

2.4.2. Comparison of UV C LED and MP UV Photolysis of Steroid Estrogens under Collimated Beam Setups

Secondary experiments were carried out using 50 mL of water samples spiked with 10 μg L^–1 E1 or 17β-E2 from a 1 g L^–1 solution in methanol or acetonitrile, respectively. Samples were exposed to fluences of 25–10,000 mJ cm^–2 under UV C LED (275 nm) or MP UV irradiation in traditional collimated beam-type setups. Samples were adjusted to pH 7 using a combination of potassium phosphate monobasic (KH₂PO₄) and potassium phosphate dibasic (K₂HPO₄) at 2 mM concentrations.

Samples were stirred constantly during treatment and equilibrated for a period of 1 min before starting experiments. 240 μL aliquots were removed from the 50 mL sample throughout the experiment at time points corresponding to chosen UV doses. The smallest possible aliquot volume was used as to not significantly affect the path length of the water sample and subsequently the UV fluence received by the water matrix (<5% change in the sample volume). 240 μL aliquots were placed in 2 mL microcentrifuge tubes, fortified with a 40 μg L^–1 13C₆-17β-E2 internal standard, and transferred to 2 mL amber autosampler vials before being analyzed via LC-MS/MS. All experiments were conducted at room temperature (20–21 °C). All data at or above calculated method detection limits (MDLs) are reported. Only data at or above the calculated limits of quantitation (LOQs) were used in kinetic calculations.

2.4.3. Kinetic Rate Constant, Quantum Yield, and Electrical Efficiency Order (EEO) Calculations

Fluence-based first-order rate constants (k _obs) and coefficients of determination (r ²) for individual compounds were determined from the linear regression of the first-order degradation plots (eq ).

$$\ln \frac{[C_{\mathrm{A}}]}{[C_{0,\mathrm{A}}]}=k_{\mathrm{obs}}$$ 2

where [C _{0, A}] is the initial concentration of compound A, [C _A] is the concentration of compound A under a specific treatment condition, and F is fluence (mJ cm^–2). The polychromatic and monochromatic quantum yields of each of the above test compounds were calculated as described in eq and below:

$${\mathrm{\Phi }}_{\mathrm{A},\mathrm{poly}}=\left(\frac{10}{\ln (10)}\right)\frac{k_{\mathrm{obs},t}}{\mathrm{PF}{\int }_{{\lambda }_2}^{{\lambda }_1}{\epsilon }_{\mathrm{A}}(\lambda )E_{\mathrm{p},\lambda }^0\times \mathrm{RF}(\lambda )\mathrm{d}(\lambda )}$$ 3

$${\mathrm{\Phi }}_{\mathrm{A},\mathrm{mono}}=\left(\frac{10}{\ln (10)}\right)\frac{k_{\mathrm{obs},t}}{\mathrm{PF}\times {\epsilon }_{\mathrm{A}}(\lambda )E_{\mathrm{p},\lambda }^0\times \mathrm{RF}(\lambda )}$$ 4

where Φ_A,poly and Φ_A,mono are the polychromatic and monochromatic Φ of compound A, respectively. Φ_A,poly was calculated using the wavelength ranges of 258–315 and 200–350 nm under 275 nm and MP UV irradiation, respectively. Φ_A,mono was determined at 275 nm for UV C LED experiments. k _obs,t is the time-based first-order reaction rate constant of compound A (s^–1), PF is the Petri factor, RF is the reflectance factor, ε_A is the molar absorption coefficient of compound A (M^–1 cm^–1), and $E_{\mathrm{p}}^0$ is the photon irradiance (Einstein s^–1 m^–2 nm^–1).

In secondary experiments, the electrical efficiency order (EEO) was calculated to determine the amount of energy required to reduce the concentration of E1 and 17β-E2 by one order of magnitude under 275 nm UV LED or MP UV irradiation using eq : ,

$$EEO=\frac{1000Pt}{V\times \ln (C_0/C_t)}$$ 5

$$P=\frac{IA}{{10}^6\times \eta }$$ 6

$$EEO=\frac{1000t}{V\lg (C_0/C_t)}\times \frac{IA}{{10}^6\eta }=\frac{\ln (10)\times It}{\ln (C_0/C_t)}\times \frac{1000A}{{10}^6\eta V}=\frac{0.64A}{{10}^6\eta V{k}_{obs,t}}$$ 7

where P is the incident power on the surface of the water (kW), t is time for sample irradiation, V is the volume of the sample (L), C ₀ and C _t are sample concentrations before and after irradiation, respectively, I is the irradiance on the surface of the sample (mW cm^–2), A is the sample surface area (cm²), and η = WPE for each light source (3.5% for the 275 nm UV LEDs and 15% for the MP UV lamp). 95% confidence intervals were calculated to determine measurement uncertainty in these metrics (k _obs, Φ_A,poly, Φ_A,mono, and EEO). Differences were considered statistically meaningful when 95% confidence intervals did not overlap.

2.5. UV Fluence Calculations

2.5.1. Uridine Actinometry

The irradiance of the 275 nm ThinFilm Reactor was calculated using the methods outlined by Pousty et al. and the following assumptions: the Petri factor (PF) was assumed to be 1 since the irradiance of different parts of the cell should not differ within the ThinFilm Reactor, the molar absorption coefficients of uridine used in calculations were 10,185 M^–1 cm^–1 at 262 nm and 8000 M^–1 cm^–1 at 275 nm, and the calculated irradiance via uridine actinometry was 249 mW cm^–2. Table SI-2 outlines fluences for each target analyte throughout experimental runs, which were similar to the modeled fluences provided by the manufacturer (within 15.3%).

2.5.2. Collimated Beam Fluence Calculations

UV irradiance measurements were performed by using an OceanOptics USB2000 spectroradiometer for collimated beam experiments. The average irradiance ( $E_{\mathrm{ave}}^{\prime }$ , reported in mW cm^–2) of each system was calculated by multiplying the measured irradiance (E ₀) at the center of the sample surface by correction factors described by Bolton and Linden:

$$E_{\mathrm{ave}}^{\prime }=E_0\times \mathrm{Petri}\mathrm{factor}\times \mathrm{reflection}\mathrm{factor}\times \mathrm{water}\mathrm{factor}\times \mathrm{divergence}\mathrm{factor}$$ 8

where the Petri factor is the ratio of the average irradiance over the area of the sample, the reflection factor accounts for the change in the refractive index as the light beam enters water, the water factor accounts for light absorbed by the water matrix, and the divergence factor accounts for the divergence of the light beam as it exits the lamp. Sampling times were then determined by dividing $E_{\mathrm{ave}}^{\prime }$ by the desired fluence. The integration ranges used to measure the intensity of each of the light sources were 200–350 nm (MP UV) and 258–315 nm (275 nm UV LED).

2.6. Analytical Methods

Caffeine, tryptophan, E1, and 17β-E2 were analyzed via LC-MS/MS using an Agilent 1260 liquid chromatography system and an Agilent 6460 triple quadrupole mass spectrometer as previously described. , Caffeine and tryptophan were quantified under the same analysis method, while a separate method was used for measurement of E1 and 17β-E2. Mass spectrometry acquisition of all compounds was conducted using multiple reaction monitoring (MRM) in either electrospray ionization (ESI) positive (caffeine and tryptophan) or negative (E1 and 17β-E2) mode. Data acquisition and analysis were carried out using Agilent MassHunter software (version rev B.10.00). Detailed methods can be found in SI-Texts 1 and 2, and analyte-specific parameters can be found in Table SI-3. pCBA and urea were quantified using HPLC-UV/vis spectroscopy and UV/vis spectroscopy, respectively, and are detailed in SI-Texts 3 and 4. MDLs and LOQs for all analytes were determined as per the US EPA Definition and Procedure for the Determination of the Method Detection Limit, Revision 2. Target concentrations, analysis methods, MDLs, LOQs, and associated internal standards (where applicable) can be found in Table SI-4.

3. Results and Discussion

3.1. Part 1: Degradation of TrOCs in a High-Intensity Flow-Through UV C LED Reactor

3.1.1. Degradation and Kinetics

The PearlAqua ThinFilm reactor (AquiSense, Erlanger, KY, USA) is a high-intensity flow-through UV C LED instrument, which was used to assess the photodegradation of target compounds under high-intensity UV C light (peak 275 nm emittance; Figure ). The ThinFilm instrument is a tailored photolytic UV LED reactor whose UV emittance profile can be manufactured to the absorbance spectrum of a given target. Target compounds (pCBA, caffeine, urea, tryptophan, and E1) were exposed to a range of delivered fluences (0–3200 mJ cm^–2; Table SI-2) to assess photodegradation dynamics. Figure A–C illustrates the process flow of the tailored photolytic UV LED reactor where a working solution of a target organic compound is pumped through a flow cell and exposed to UV C LED-generated light. Figure i–iii shows the degradation kinetics of the five targets grouped by (i) 275 nm UV-susceptible compounds and (ii) 275 nm UV-resistant compounds.

2.

Degradation and first-order kinetic plots of target compounds under 275 nm irradiation from a flow-through UV C LED reactor. (A) LED panel, (B) flow cell, and (C) cooling block. (i) 275 nm susceptible compounds, (ii) 275 nm resistant compounds, and (iii) degradation kinetics of all target compounds. Experimental parameters: pH = 7, 2 mM PO₄ ^2– buffer; [urea]₀ = 50 mg L^–1; [pCBA]₀ = 10 μM; [caffeine]₀ = 10 μg L^–1; [tryptophan]₀ = 100 μg L^–1; [E1]₀ = 10 μg L^–1. n = 3.

A photodegradation reaction will only occur if (1) a photon is absorbed by the target compound, (2) the energy of the absorbed photon is higher than the bond dissociation energy of a given chemical bond, and (3) the chemical structure of a target compound favors bond cleavage rather than energy dissipation through nondissociative pathways. As such, the efficacy of 275 nm light in degrading each of the compounds was governed by both the spectral absorbance and the chemical structure of each compound. All compounds, with the exception of urea, exhibited moderate to high molar absorbance in the emittance range of the UV LED reactor (Figure ). The compounds that exhibited minimal degradation under 275 nm irradiation (pCBA and caffeine) followed a similar degradation profile in resisting photodissociation from 275 nm light (Figure ii).

3.1.1.1. Degradation of Compounds Sensitive to 275 nm UV Light

Tryptophan was degraded at the second highest rate of all target compounds evaluated and exhibited >95% degradation (below detectable limits) at the highest fluence condition (k _obs = 1.62 × 10^–3 cm² mJ^–1; Table SI-1). Tryptophan exhibits high absorbance in the 275 nm wavelength range (ε_275 nm = 6543 M^–1 cm^–1) which is attributed to its highly conjugated indole ring and carbonyl group on C1 (Figure SI-1). Previous work has similarly observed that tryptophan will readily degrade under UV irradiation and can both undergo direct photolysis and participate in auto-oxidation reactions to form reactive oxygen species (ROS), which can further contribute to degradation. ,

E1 was degraded more rapidly than all other analytes during treatment (9.22 × 10^–3 cm² mJ^–1; Table SI-1) and demonstrated >95% removal (below detectable limits) at a fluence of 616 mJ cm^–2 (Figure ). The photochemistry of E1 is well-documented. ,− E1 is known to be highly susceptible to photolysis under LP UV, MP UV, and even solar irradiation ,,,− and has an absorbance band around 280 nm with peak λ = 281 (Figure ) and ε_{275 nm–281 nm} = 2027–2568 M^–1 cm^–1. The presence of the phenolic ring on E1 is responsible for the observed 280 nm absorbance band, while the C17 carbonyl group provides weak absorption at ∼290 nm, ,, both of which overlap with the emittance of the 275 nm UV C LED reactor used in this study.

During E1 photolysis, we observed the evolution and disappearance of a photoproduct throughout treatment, which had the same quantifier and qualifier ion transitions as E1 (m/z 269.1 → 145 and 269.1 → 143.2, respectively) and a slightly later retention time (6.62 min for E1 vs 7.16 min for the photoproduct; Figure SI-4). The ion transitions and nominal mass of the photoproduct are consistent with the 13α-epimer of E1, lumiestrone, which is structurally identical with E1 with the exception of a rotation of the C13 methyl group.

Lumiestrone has been reported in the literature to be the primary photoproduct of E1 ,, and demonstrates mild estrogenic activity as compared to its parent compound. The primary mode of E1 degradation at ∼280 nm emittance is dependent on photon absorption by the phenol group and transfer of excitation to the carbonyl group, leading to epimerization and subsequent formation of lumiestrone. This reaction is both solvent- and wavelength-dependent. Formation of lumiestrone is conserved across 255–320 nm wavelengths; however, a higher yield of the epimer has been reported via photolysis at shorter (e.g., 255–280) wavelengths. We observed rapid evolution of the photoproduct, with peak formation occurring at a fluence of 231 mJ cm^–2 (Figures SI-4 and SI-5A). Photoproduct response then decreased with increasing fluence to a maximum of 2464 mJ cm^–2, where the photoproduct was reduced by >98% of maximum peak response observed at the 231 mJ cm^–2 condition.

3.1.1.2. Degradation of Compounds Resistant to 275 nm UV Light

Urea was negligibly degraded under 275 nm irradiation and exhibited a maximum of mean <1% removal even under the highest fluence condition (3081 mJ cm^–2; k _obs = 7.92 × 10^–6 cm² mJ^–1; Figure and Table SI-1). Urea exhibits very low or no absorbance within the emittance range of the 275 nm LED (0–1.2 M^–1 cm^–1 between 258 and 315; Figure ). Photons must be absorbed by a compound to undergo photochemical reactions (Grauss–Draper law). Degradation data for urea were a poor fit for first-order kinetics (r ² = 0.641; Table SI-1), which is unsurprising due to the low removal exhibited throughout experiments.

pCBA was minimally degraded during treatment and exhibited a maximum of 19% mean removal under the highest fluence conditions (k _obs = 7.23 × 10^–5 cm² mJ^–1; Figure and Table SI-1). pCBA exhibits moderate absorbance within the UV C LED emittance range (Figure ) with ε_275 nm = 600 M^–1 cm^–1 and thus would be capable of photon absorption under this scenario. It is well-known that pCBA does not readily undergo photolysis under both low-pressure (LP UV) and MP UV irradiation and its photostability has made it widely used as a ^•OH probe in advanced oxidative process (AOP) studies. ,, pCBA contains a conjugated benzene ring, carbonyl group on C1, and lone pair electrons on the Cl atom, which contribute to its UV absorbance (Figure SI-1). The electronegativity of the Cl atom on C4 creates an electron-withdrawing effect on the benzene ring. This can reduce photodegradation efficiency as has been described for other chlorophenols undergoing UV irradiation and explains the minimal photodegradation observed here and by others.

Caffeine was similarly degraded to pCBA exhibiting a maximum of 17% (k _obs = 6.61 × 10^–5 cm^–2 mJ^–1) mean removal at the highest fluence condition, and 95% confidence intervals indicate no statistically significant difference in degradation kinetics between pCBA and caffeine (Table SI-1). Though caffeine exhibits very high molar absorbance within the relevant emittance range (ε_275 nm = 9884 M^–1 cm^–1), minimal degradation of caffeine under UV irradiation has been similarly observed in previous work. − Caffeine absorbs readily within the typical UV treatment range (i.e., 200–300 nm) due to its extensive electron delocalization across the purine ring system from lone pair electrons on the N atoms, double bonds in the imidazole ring, and carbonyl groups on C2 and C6 (Figure SI-1). Caffeine has been observed to have very short excited-state lifetimes (e.g., <1 ps), favoring rapid, nondissociative conversion of absorbed energy over photodissociation, which explains its relative photostability.

3.1.2. Quantum Yield Determination

UV LEDs have historically been treated as monochromatic light sources, but by nature, UV LEDs emit polychromatically over wavelength bands (e.g.., Figure ), which are much broader than the emittance range of traditional monochromatic (e.g., LP UV) light sources. As such, we calculated and compared both the polychromatic and monochromatic quantum yields of each of the above test compounds (eq and eq ; see methods). Calculated quantum yields can be found in Table below. Quantum yields for urea were not calculated due to negligible degradation and poor fit for first-order kinetics (described above; Table SI-1).

1. Polychromatic (Φ_poly,275nm) and Monochromatic (Φ_mono,275nm) Quantum Yields for the Target Compounds .

analyte	Φpoly,275nm (mol E s–1)	Φmono,275nm (mol E s–1)	percent difference (%)
pCBA	0.0271 [0.0230, 0.0312]	0.0225 [0.0191, 0.0259]	18.6
caffeine	0.00147 [0.00129, 0.00166]	0.00128 [0.00112, 0.00144]	13.8
tryptophan	0.0486 [0.0473, 0.0499]	0.0466 [0.0454, 0.0477]	4.2
E1	0.843 [0.786, 0.900]	0.857 [0.799, 0.915]	1.7

^aBrackets indicate 95% confidence intervals.

The calculated 95% confidence intervals indicate that differences in monochromatic (Φ_mono,275nm) and polychromatic (Φ_poly,275nm) quantum yields were not statistically significant within analytes (Table ) but differences in mean values were >10% for UV-resistant compounds (pCBA and caffeine). Φ_poly,275nm for pCBA and caffeine were mean 18.6 and 13.8% higher than Φ_mono,275nm, respectively, indicating that, in these cases, monochromatic quantum yield calculations may underestimate actual quantum yield. In contrast, differences in monochromatic and polychromatic values were less pronounced for UV-sensitive analytes. Φ_poly,275nm was <5% higher and <2% lower than Φ_mono,275nm for tryptophan and E1, respectively, indicating minimal differences in monochromatic and polychromatic quantum yields for these compounds. Li et al. found that monochromatic quantum yield calculations underestimated true quantum yield values of chlorine species (HOCl and OCl^–) by 5.66–14.63% under 265 and 280 nm UV LED irradiation. In contrast, Pousty et al. determined that differences between polychromatic and monochromatic quantum yields for uridine were negligible under 279 nm UV LED irradiation and that the studied 279 nm UV LED could be treated as a quasi-monochromatic light source. Accordingly, there may be situations where UV LEDs can be treated as monochromatic light sources without significant error, but this is almost certainly dependent on the emittance wavelength and fwhm of the UV LED light source, as well as the chemical structure, molar absorbance, and subsequent UV sensitivity of the compound of interest. In the majority of cases, it may be most appropriate to treat these light sources as polychromatic unless the use case has been proven otherwise for a particular light source and compound combination. In the section below, Φ_poly,275nm for all analytes was chosen to provide comparison to the literature.

3.1.2.1. Discussion of Quantum Yields for pCBA and Caffeine

Polychromatic and monochromatic quantum yields at 275 nm (Φ_poly,275nm and Φ_mono,275nm) have not previously been reported in the literature for the target compounds evaluated in this paper. Calculated mean Φ_poly,275nm for pCBA was 0.0271 mol E s^–1 in this work (Table ). Rosenfeldt and Linden calculated monochromatic (LP UV; 254 nm) and polychromatic (MP UV; 200–300 nm) quantum yields of pCBA as 0.013 ± 0.002 and 0.018 ± 0.004 mol E s^–1, respectively, which are mean 70.3 and 40.4% lower (but within the same order of magnitude) than is reported in this work. Similarly, Shu et al. reported Φ_Caffeine under MP UV irradiation as 0.0003 ± 0.001 mol E s^–1, which is nearly an order of magnitude lower than the calculated Φ_poly,275nm for caffeine in this study (0.00147 mol E s^–1).

Differences in reported values could indicate wavelength dependence in the spectral ranges evaluated or could be related to variability in the calculation or experimental setup. Differences in light intensity could also potentially explain variation between calculated values in the literature; Linden and Darby showed that increasing light intensity in LP UV systems can result in higher apparent quantum yields of uridine, but only to a maximum value of 0.4 mW cm^–2, after which no further increase was observed up to a maximum of 3.2 mW cm^–2. More recently, Rajesh et al. observed an over two-fold increase in quantum yield of dibromoacetonitile under high-power-output vs low-power-output 265 nm UV LEDs.

In this work, the calculated fluence rate of the reactor under uridine actinometry was 249 mW cm^–2. There are no commercially available mercury-based systems and no other UV C LED systems outside of the reactor type employed in this study, which can produce such high-intensity UV light. A conventional system could theoretically be designed to deliver high-intensity light but would inevitably also deliver so much heat to a sample that it would be very challenging to mitigate the impact of heat on samples. The kinetic and photochemical behavior of analytes under this high-intensity UV light source is likely not directly comparable to lower-intensity light sources. Further, the reactor used in this experiment is a flow-through design and has not previously been described for degradation of organic compounds using UV LEDs. The dynamics of each experimental setup likely require individual considerations for the specific water chemistry that is being investigated in an experiment.

3.1.2.2. Discussion of Quantum Yields for Tryptophan and E1

Φ_Tryptophan varies from below 0.01 to 0.35 in the literature; Callis and Liu discussed that this may be related to differences in electron transfer caused by environmental variations during measurement. Calculated Φ_poly,275nm for tryptophan in this study is within this range of values (0.0486 mol E s^–1; Table ), but it is relevant to note that the above-reported values are based on fluorescent quantum yield determinations over a variety of wavelength ranges vs the time-based kinetic quantum yield calculation method employed in this work.

Reported quantum yield values for E1 in the literature also vary widely and have been reported from 0.065 mol E s^–1 (10 μg L^–1, LP UV irradiation) to 0.35 ± 0.14 mol E s^–1 (5 μg L^–1, LP UV irradiation) and as high as 5.14 mol E s^–1 (1 mg L^–1, LP UV irradiation). The cause of such large discrepancies is unclear in the literature as all three example references above used the same method to calculate quantum yield at the same irradiation wavelength, though Pereira et al. performed experiments using an analyte concentration that was two orders of magnitude or greater than the others described above. The calculated quantum yield for E1 in this study is within the range demonstrated in the literature (Φ_poly,275nm = 0.843 mol E s^–1), but further research would be needed to explain if variation from values previously reported is due to wavelength dependence, differences in calculation, variation in experimental variables, or other factors.

3.1.3. Summary of Findings

This work presents the first demonstration of k _obs, Φ_poly,275nm, and Φ_mono,275nm for the described compounds under 275 nm irradiation and demonstrates the importance of compound structure–wavelength interactions in target-specific photolytic water treatment processes. This work also provides the first use and application of a novel, flow-through UV C LED reactor, which allows for substantially higher sample throughput and lower fluence delivery time as compared to traditional collimated beam setups (e.g., Table SI-2). The UV absorbance profile and chemical structure of the target are the two primary governing factors when determining the efficacy of photolytic treatment, and compounds with higher quantum yields are more susceptible to photolytic degradation. The consideration of each of these features is key to developing compound-specific UV LED systems for robust water treatment approaches.

3.2. Part 2: Comparison of UV C LED and MP UV Photolysis for Degradation of Steroid Estrogens

In secondary experiments, the effectiveness of 275 nm LED and MP UV irradiation was assessed for driving the photodegradation of E1 and another steroid estrogen, 17β-E2. These experiments were conducted to provide additional context to the flow-through results described above that demonstrated the effectiveness of UV C LED exposure at 275 nm for driving the photodegradation of E1. 17β-E2 was chosen as an additional target compound because of its similar structure and absorbance profile to E1 but comparative resistance to photodegradation in comparison to its sister compound. , Further, both compounds are environmentally relevant TrOCs, which cause endocrine-disrupting effects in aquatic organisms, and evaluation of effective treatment techniques is relevant to both municipal and agricultural sectors, which are the primary producers of estrogenic waste released into aquatic environments. ,,

It is relevant to note that the semiconductor chips in the collimated beam reactor evaluated here are paired with those used in the flow-through reactor described in Part 1 experiments, meaning that the emittance spectrum is identical for both systems. Pairing of these LEDs was done intentionally to ensure that differences in manufacturing were eliminated from influencing the results of this work, as LEDs can vary by a few nanometers in emittance and by milliwatts in output if chips are taken from different batches. This attention to detail was considered necessary as the experimental detail on the flow-through reactor has not been published previously. The degradation data for these compounds under each light source and the absorbance/emittance spectra of the reactors and analytes can be found in Figure below.

E1 and 17β-E2 were degraded by both light sources, but the MP UV system required substantially higher fluence delivery to achieve comparable removal to the 275 nm UV LED system. It is relevant to note that the MP UV system demonstrates only 17% overlap (when accounting for the light emitted in the 200–600 nm range) in total light emission with the molar absorbance spectrum of each estrogen, whereas the 275 nm UV C LED emittance spectrum demonstrates complete overlap with the lower absorbance band of both compounds. The MP UV light source was unable to degrade 17β-E2 to below detectable limits even at a fluence of 10,000 mJ cm^–2, while complete (below detectable limits) removal was achieved at a fluence of 2000 mJ cm^–2 using the 275 nm UV LED light source.

3.2.1. UV C LED and MP UV Degradation of E1

The calculated 95% confidence intervals indicate statistically significant differences between Φ_poly,275nm and k _obs values for E1 in 275 nm flow-through reactor (Part 1) and collimated beam (Part 2) experiments (Table , Table SI-1, and Table ). The mean 15.8 and 56.4% differences observed between first-order rate constants and quantum yields (respectively) are likely due to differences in the experimental setup and/or light intensity (249 mW cm^–2 (flow-through reactor) vs 3.62 mW cm^–2 (collimated beam)). We discussed the implications of this in earlier experiments as it pertains to quantum yield, and the same discussion would apply here. A similar phenomenon has also been observed for degradation kinetics of other TrOCs; Rajesh et al. found that increasing the irradiance of 265 nm UV LEDs resulted in a 2.3-fold increase in the kinetic rate constant of dibromoacetonitrile even when controlling for fluence.

2. Polychromatic (Φ_poly,275nm for the 275 nm LED or Φ_poly,MPUV for the MP UV Lamp) and Monochromatic (Φ_mono,275nm) Quantum Yields for Target Compounds.

analyte	irradiation source	fluence-based rate constant k obs (/× 10–3; cm2 mJ–1)	r 2	Φpoly (mol E s–1)	Φmono,275nm (mol E s–1)	% difference
E1	275 nm LED	7.87 [7.63, 8.10]	0.9827	0.472 [0.458, 0.486]	0.480 [0.467, 0.494]	1.7
E1	MP UV	2.37 [2.15, 2.61]	0.9949	0.274 [0.258, 0.290]
17β-E2	275 nm LED	4.04 [3.01, 5.06]	0.9712	0.0794 [0.0592, 0.0997]	0.0794 [0.0592, 0.0997]	0
17β-E2	MP UV	0.215 [0.183, 0.247]	0.9500	0.0537 [0.0451, 0.0623]

E1 was degraded by >90% under 275 nm irradiation at 50% of the fluence required under MP UV (Figure ). Fluence-based rate constants were >3× higher (Table ), and polychromatic quantum yields were 53.1% higher under 275 nm irradiation as compared to MP UV (0.472 mol E s^–1 (Φ_poly,275nm) vs 0.274 mol E s^–1 (Φ_poly,MPUV)) indicating wavelength dependence within the 200–350 nm region. As described earlier, E1 is known to undergo rapid photodecomposition when exposed to light over wide ranges of the UV spectrum, ,,,− and so, effective degradation under both 275 nm and MP UV conditions would be expected. The more effective degradation observed under 275 nm light is due to the targeted nature of the 275 nm reactor emittance spectrum over the 281 nm absorbance band of E1 (Figure ). The peak emittance of the reactor and the 281 nm absorbance band of the analyte are overlapping (peak reactor emittance = 275 nm, peak E1 absorbance = 280–282 nm, and ε = 2569 M^–1 cm^–1) in comparison to the MP UV lamp, which emits sporadically in the 200–300 nm absorbance range of E1 (Figure ). E1 also has an absorbance shoulder at ∼220 nm (ε = 7841 M^–1 cm^–1) and a second absorbance peak at 203 nm (ε = 30,146 M^–1 cm^–1), which could be theoretically targeted by the MP UV lamp. The absorbance peak observed at 281 nm corresponds to E1’s lowest energy absorption band. If a photon absorbed by a compound has energy sufficient to promote a compound to its first excited state and result in a photochemical reaction, then the absorption of photons of higher energy will not increase the rate of a photochemical reaction and the additional energy will be lost as heat. Accordingly, the targeted emission of the 275 nm reactor is more efficient at driving photodegradation reactions for E1 than the MP UV lamp.

The same photoproduct of estrone as discussed in earlier experiments (consistent with lumiestrone) was detected under both 275 nm and MP UV irradiation. Photoproduct dynamics under 275 nm irradiation were similar to flow-through reactor experiments: peak photoproduct evolution was observed at 250 mJ cm^–2 as compared to 231 mJ cm^–2 for the 275 nm flow-through reactor. The photoproduct was nondetectable at a fluence of 5000 mJ cm^–2 (Figure and Figure SI-5B). In contrast, peak photoproduct evolution was not observed until 2000 mJ cm^–2 under MP UV and the photoproduct was still observed at ∼51% of maximum peak response at the highest fluence conditions (5000 mJ cm^–2; Figure and Figure SI-5C). Accordingly, this tailored wavelength approach to photolytic degradation demonstrated more rapid degradation of E1 and improved reduction in its associated photoproduct as compared with MP UV irradiation.

4.

E1 and associated photoproduct response as a function of fluence under 275 nm (A) and MP UV (B) irradiation.

3.2.2. UV C LED and MP UV Degradation of 17β-E2

17β-E2 was degraded by >80% under 275 nm irradiation at 5% of the fluence required (500 mJ cm^–2) under MP UV (10,000 mJ cm^–2) and >90% under 275 nm irradiation at 10% of the fluence required as compared to under MP UV irradiation (Figure ). Complete (below detectable limits) degradation was observed under 275 nm irradiation at a fluence of 2000 mJ cm^–2. In contrast, 17β-E2 was still detectable even at very high (i.e., 10,000 mJ cm^–2) fluence conditions under MP UV irradiation. Fluence-based rate constants were also more than an order of magnitude higher under 275 nm irradiation (4.04 × 10^–3 cm² mJ^–1 vs 2.15 × 10^–4 cm² mJ^–1 under 275 nm irradiation and MP UV, respectively). The calculated polychromatic quantum yield was also higher under 275 nm irradiation (0.0794 mol E s^–1 (Φ_poly,275nm) vs 0.0537 mol E s^–1 (Φ_poly,MPUV)), but 95% confidence intervals indicate that this difference was not statistically significant (Table ).

17β-E2 has a very similar structure and absorbance profile to E1 with an absorbance band at 281 nm (ε = 1498 M^–1 cm^–1), an absorbance shoulder at ∼220 nm (ε = 6129 M^–1 cm^–1), and a second absorbance band at ∼200 nm (ε = 31,598 M^–1 cm^–2; Figure ). It is likely that the same spectral dynamics described above for E1 are also applicable to 17β-E2, i.e., targeted photochemical activation of the lower energy band (∼280 nm) by the 275 nm UV C LED can more effectively evoke photochemical reactions than the polychromatic distribution of light by the MP UV light source. In this case, the reduction in required fluence over commercially available systems is significant (to be redundant, 10% of the required fluence to achieve 90% degradation), which has meaningful implications for stakeholders looking for treatment options to remediate estrogenic compounds (i.e., agriculture, pharmaceutical, and municipal, among others).

3.2.3. Electrical Efficiency Order (EEO)

The electrical energy required to achieve 90% degradation (electrical efficiency order; EEO) of E1 and 17β-E2 under both 275 nm and MP UV irradiation was calculated to determine direct energy cost differences between the two bench-scale systems (eq , see methods). Both analytes were degraded at a significantly higher rate under 275 nm irradiation (Table ) than that under the MP UV system. The waste heat energy is directed in the same direction as light energy for MP UV systems imparting practical upper boundary conditions for the MP UV system that would not exist for the UV LED system (in which heat and light energy move in opposing directions). High (≥2000 mJ cm^–2) fluence conditions would be impractical or impossible in a full-scale MP UV system. Nonetheless, for illustrative purposes, EEO values were 25% higher under 275 nm irradiation for E1 (mean 1.31 and 1.02 kWh m^–3 order^–1 for the 275 nm UV C LED and MP UV lamp, respectively; Table ) while energy usage was reduced by over four-fold for 17β-E2 under 275 nm conditions with the current 3.5% WPE UV C LED system (2.66 vs 11.34 kWh m^–3 order^–1 for the 275 nm UV LED and MP UV lamp, respectively). This represents significant potential for a direct UV LED photolysis system that is neither practically possible for mercury UV systems nor theoretically more efficient for 17β-E2 with the current UV C LED technology.

3. EEO Values for 17β-E2 and E1 under 275 nm UV LED and MP UV Irradiation .

analyte	irradiation source	EEO (kWh m–3 order–1)
E1	275 nm LED (3.5% WPE)	1.31 [1.27, 1.35]
	MP UV (15% WPE)	1.02 [0.916, 1.13]
17β-E2	275 nm LED (3.5% WPE)	2.66 [1.89, 3.43]
	MP UV (15% WPE)	11.34 [9.78, 12.91]

^aBrackets indicate 95% confidence intervals.

4. Prospects for UV C LEDs as a Climate-Robust Treatment Technology

UV C LEDs have confirmed effectiveness for full-scale disinfection in drinking water and wastewater , and have demonstrated the ability to replace mercury lamps in most water treatment applications. Limitations on their widespread use have largely been related to historical constraints in power output and energy efficiency. Power outputs of UV C LEDs have improved by over two orders of magnitude (139×) in the past 10 years, and current efficiency limitations, particularly for UV C LEDs, are not due to the fundamental quantum mechanics of the system but technological limitations. It is projected that technological advances could see the development of more WP efficient UV C LEDs within the next 10 years, which would only increase practical capability to disrupt water treatment design possibilities.

This work demonstrates the significant use potential of UV C LEDs for removing estrogens from water. Estrogenic compounds are used as feed additives or implants in aquaculture ,,, and agriculture systems , (respectively), as active ingredients in oral contraceptives and hormone replacement therapies, , and are also present naturally in the human body. These scenarios result in significant inputs of estrogens into municipal treatment plants and the global environment. Steroid estrogens are often not effectively removed or degraded by standard treatment practices employed at municipal wastewater treatment plants, − and these compounds require the addition of chemical agents (to facilitate AOPs), high-pressure membranes (e.g., reverse osmosis (RO) or nanofiltration (NF)), adsorbent materials (e.g., granular activated carbon (GAC)), or ion-exchange technologies to effectively remediate. In the latter cases, processes result in secondary waste streams (e.g., membrane concentrate, exhausted adsorbent, and spent ion-exchange resin), which need to be further treated, regenerated, or disposed of, resulting in significant cost, materials waste, and carbon footprint. The current work illustrates the substantial promise of UV LEDs to provide effective, chemical-free, bespoke treatment of steroid estrogens with no secondary waste streams, and tailoring the wavelength emittance could see similar applications for a wide range of concerning contaminants.

The majority of previous research evaluating the utility of photolytic degradation technologies for contaminant remediation focuses on mercury-based light sources (e.g., LP UV and MP UV), which have a limited range of spectral output. The use of tailorable LED-based light sources provides wider application of photolytic technologies; for example, future potential could see bespoke UV LED technology design for even recalcitrant compounds such as per- and polyfluoroalkyl substances (PFAS), which require large energy inputs to break C–F bonds (96.8–131.6 kcal mol^–1 ). This is no longer an impossibility with the current technology trajectory. It is possible for LEDs to approach 100% WPE under ideal scenarios, and current limitations are technological and not fundamental. Current efficiency constraints for UV C LEDs are primarily related to materials quality, doping, carrier injection, and light extraction, limitations that are driven by materials design and device engineering rather than any inherent physical barriers.

By the early 21st century, SSLE devices had largely replaced technological illumination alternatives in consumer products such as vacuum tubes and pressured gases due to their improved efficiency, reliability, longer lifetime, and cost effectiveness in comparison to their historical counterparts. It is inevitable that the same will be the case for UV LED replacement of mercury vapor lamps in water treatment. Ultraviolet diodes are on a similar trajectory as visible-light diodes were several decades ago. Cathode ray tubes (CRTs) were the primary display technology throughout the mid-20th and early 21st century. The onset of LED-based displays ushered in the emergence of 1080p as a standard resolution for televisions and monitors in the early 2010s, which resulted in the displacement of CRT televisions. 4k displays are now repeating this cycle with 1080p screens. UV LEDs are now at the tipping point of displacing an established technology for disinfection with the use of full-scale UV LED disinfection. This work makes it possible to imagine how advancement in UV LED technology could foster algorithmic-driven water treatment, dual-purpose disinfection/photolysis reactors, and dynamic adjustment of emitted wavelengths in large-scale reactors.

5. Conclusions

This characterization study using six different target compounds demonstrated the unique potential that tailored light output can provide for water treatment. Estrogenic and amino acid compounds were most susceptible to degradation at 275 nm, while targets such as urea, caffeine, and pCBA were unaffected or minimally degraded by exposure to 275 nm light. This result highlights that alignment between wavelength emittance and treatment objective is an important consideration for future design of UV LED-based photolysis systems, and evaluation of the chemical structure and molar absorbance is necessary to provide an adequate assessment of treatment capabilities for TrOCs. Additionally, our results suggest that quantum yield is a good proxy for potential treatment outcomes and can help to infer potential energy requirements.

Secondary experiments comparing 275 nm UV C LEDs and an MP UV system further demonstrated the utility of a tailored treatment platform. For 17β-E2, 90% degradation was achieved at fluences 1/10th of that required for similar degradation using the MP UV system. In the case of E1, both the analyte and its associated photoproduct were degraded rapidly under 275 nm irradiation, while E1 degradation rates were significantly lower under MP UV irradiation, and the associated photoproduct persisted at even high (5000 mJ cm^–2) fluences. Furthermore, the EEO for 17β-E2 using the 275 nm UV C LED system was about 4-fold lower than that of the MP UV system when using systems with WPE of 3.5% (UV C LED) and 15% (MP UV). The WPE of UV C LEDs is expected to increase substantially in the next 10 years, which would drastically reduce the energy needs of these systems when compared to conventional mercury lamps in this scenario. This result indicates that tailored use of UV light to treat target compounds can overcome some of the current technological limitations that LEDs have compared to conventional systems. This benefit is in addition to the environmental considerations that have been highlighted in this work.

This research provides the groundwork for scaling these technologies via the calculation of energy requirements needed for UV C LEDs compared to conventional alternatives. Future work should focus on the evaluation of variable parameters and water matrix effects on treatment efficacy, analyte kinetics, and compound-specific response, while testing of this technology at pilot-scale for this use case is required to demonstrate real-world utility and future potential. This work and proof of concept have implications across scientific and engineering disciplines given the economic and environmental drivers to eliminate mercury use and reduce the reliance on chemicals in the water treatment process.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsestwater.5c00526.

• Additional details on target compounds, experimental setup, and supplementary text, tables, and figures as mentioned in the manuscript (PDF)

CRediT: Jessica L. Bennett conceptualization, data curation, formal analysis, methodology, validation, visualization, writing - original draft, writing - review & editing; Sean A. MacIsaac conceptualization, data curation, methodology, visualization, writing - original draft, writing - review & editing; Jin Li conceptualization, investigation, methodology, writing - original draft; Metyn B. Rehman investigation, methodology, writing - original draft; Rae E. Fitzgerald investigation, methodology; Amina K. Stoddart conceptualization, funding acquisition, supervision, writing - review & editing; Graham A. Gagnon conceptualization, funding acquisition, supervision, writing - review & editing.

The authors declare no competing financial interest.