Improved Drinking Water Disinfection with Uvc-leds for Escherichia Coli and Bacillus Subtilis Utilizing Quartz Tubes as Light Guide

A new approach is investigated utilizing light guidance capabilities of optical pure quartz glass in order to maximize drinking water disinfection efficiency with UVC-light-emitting diodes (LEDs). Two experimental setups consisting of soda-lime AR ® glass (VWR, Darmstadt, Germany) or HSQ ® 100 quartz glass (Heraeus, Wasserburg, Germany) reactors were designed to compare disinfection rates with and without total reflection of UVC radiation along the reactor walls. Each reactor was filled with 9 mL bacteria samples containing either E. coli DSM (Deutsche Sammlung von Mikroorganismen) 498 or B. subtilis DSM 402 strains (concentration 1–3 × 10 6 colony forming units (CFU)/mL) with and without additional mixing and irradiation periods of 10, 40, and 90 s. Disinfection rates were increased up to 0.95 log10 (E. coli) and 0.75 log10 (B. subtilis) by the light guide approach in stagnant samples. The same experiments with mixing of the samples resulted in an increased disinfection efficiency of 3.07 log10 (E. coli) and 1.59 log10 (B. subtilis). Optical calculations determine that total reflection is achieved with the applied UVC-LED's viewing angle of 15°. Furthermore measurements show that HSQ ® 100 quartz has a transmittance of 92% at 280 nm UVC irradiation compared to the transmittance of soda-lime glass of 2% (1 mm wall thickness).


Introduction
Contaminated drinking water is an often underestimated problem in First World countries because of the advanced disinfection methods in modern sewage plants and specialized facilities.However, nearly 2.5 billion people worldwide have no access to improved sanitation and 748 million of them still lack sufficient access to improved drinking water sources [1], accounting for over 1.7 million deaths a year, especially in developing countries [2,3].The upcoming technology of UVC-LEDs promises new inexpensive, environmentally friendly, and long-life disinfection systems for drinking water that can be used decentralized in combination with photovoltaic systems [3].UVC radiation (200-280 nm) in general is a very efficient way of inactivating pathogen bacteria in drinking water by directly damaging the DNA [4].Radiation of this wavelength range is able to enter the cell walls of microorganisms and is partly absorbed by the DNA.By this process, thymine dimers are formed, which prevent further replication of the DNA strains [5].These thymine dimers arise when an UVC photon gets absorbed by adjacent thymine nucleotides, merging them as shown in Figure 1b.A preferable wavelength for disinfection would therefore be expected at about 270 nm, which is the absorption peak of thymine [6,7].Today, commercial UV disinfection systems usually apply highly toxic and environmentally harmful low or medium pressure mercury vapor lamps, which have high optical output powers of several watts and a wavelength peak commonly in the UVC range of 254 nm [5,6].Normally, much UVC power of such a system is wasted due to inadequate optical coupling.Because of the high power output, these systems still guarantee sufficient fluence rates of over 40 mJ/cm 2 [3,5].Figure 1a shows a buildup concept for a traditional cylindrical UV disinfection reactor [9].The water flows alongside a mercury vapor lamp and is exposed to UVC-radiation on its way through the reactor.Because of the elongated construction of mercury lamps and the small dimensions of LEDs, this reactor design would be insufficient when applying only a single UVC-LED.However, there are approaches to take over this reactor concept with several UVC-LEDs placed alongside the reactor [10,11].The questionable environmental and toxic properties of mercury require adequate replacement technologies.In this context, deep UV-LEDs are a promising development and may be able to function as a substitute for mercury lamps in the near future.However, commercially available UVC-LEDs usually produce severely lower radiant output powers of mostly less than 3 mW [12], thus new and more efficient reactor concepts should be considered.
In order to create sufficient inactivation rates of at least three log10 levels, all of the generated UVC-LED radiation has to be harvested in a novel reactor design.Aiming at very high efficiency, a new concept was created by coupling an UVC-LED in parallel to a quartz tube and keeping UVC radiation inside the tube due to total reflection at the walls.With this approach, most of the optical power is available for disinfection along the whole reactor if certain boundary conditions are met [13].If normal soda-lime or silicate-based glass types, instead of quartz glass, are used in such a reactor design, the incoming UVC-rays will be absorbed almost completely at the walls.As a result, optical power is wasted and will no longer be available for inactivating microorganisms.A round pipe geometry is analyzed and applied for experimental setup in this work in order to verify the basic capabilities and benefits of this novel reactor design.If this concept proves to be effective, further optimized reactor designs according to this concept, especially for flow reactors, can be considered in future studies.

Experimental Setup
An experimental setup allowing precise comparison of the disinfection with and without the concept of total reflection was designed.A schematic view can be seen in Figure 2, with both setups following the same buildup concept.The experiment consists of a glass tube each (either HSQ ® 100 quartz from Heraeus (Wasserburg, Germany) or AR ® soda-lime glass from VWR (Darmstadt, Germany)), 9 mL distilled water with inserted microorganisms (E. coli DSM 498 or B. subtilis DSM 402, concentration 1-3 × 10 6 colony forming units (CFU)/mL), and an UVC-LED type OPTAN280J from Crystal IS (Green Island, NY, US) delivering a measured radiant output of 0.97 mW (forward voltage: ca. 9 V; driving current: 100 mA; power dissipation: 0.9-1 watt) at a measured emission wavelength peak of 281.8 nm.The power supply TOE 8733 from Toellner (Herdecke, Germany) ensured correct voltage and current ratings during all experiments.Both glass tubes were identical except for their material with the dimensions of 12 mm (inner diameter), 14 mm (outer diameter), and a length of 100 mm.
For disinfection, the bacteria solution was exposed to UVC radiation in defined time intervals of 10, 40 and 90 s or doses of 8.64 mJ/cm 2 , 34.59 mJ/cm 2 , and 77.82 mJ/cm 2 , respectively.The bacteria solution inside the tube was mixed continuously with a MR 1000 magnetic stirrer from Heidolph (Schwabach, Germany).Additionally, stagnant experiments without mixing were performed.Because of the attenuation of UVC radiation along the glass tube, the bacteria solution close to the LED was exposed to much higher UVC doses than at the bottom.

Bacterial Strains and Growth Conditions
Two different non-pathogenic bacterial strains were applied for the experimental setup, Escherichia coli DSM 498 and Bacillus subtilis DSM 402.All bacterial strains were from DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (Braunschweig, Germany)) and were provided in freeze-dried condition stored at −75 °C.E. coli is a standard Gram-negative laboratory microorganism, which is responsible for many diarrheal diseases, and is therefore well suited as reference bacterium for disinfecting water [2,14].Gram-positive B. subtilis, on the other hand, is used as an additional reference microorganism, which delivers further information about the disinfection process with and without total reflection in comparison to E. coli [4,15].Due to structural differences in the construction and composition of their membrane, Gram-negative microorganisms tend to show increased susceptibility to UVC radiation [16,17].B. subtilis strains were applied in their vegetative state and not as spores for all experiments in this study.All bacterial strains were cultivated aerobically in Luria-Bertani (LB) broth consisting of distilled water, 10 g/L NaCl, 5 g/L yeast extract, and 10 g/L tryptone.Growth phase was realized using the incubator TH30 from Edmund Buehler GmbH (Hechingen, Germany) at 120 rpm and 37 °C for E. coli and 30 °C for B. subtilis over a period of 24 h, until stationary phase was reached.Initial concentration of this culture after 24 h was determined at 10 9 CFU/mL (E.coli, OD600nm = 0.76) and 10 8 CFU/mL (B.subtilis, OD600nm = 0.53) by spectrometry referenced on LB broth at 600 nm.Finally, bacterial cultures were inoculated into sterile 0.9% saline solution to a final concentration of approximately 1-3 × 10 6 CFU/mL.For B. subtilis 1 mL of the initial solution was diluted with 100 mL of 0.9% saline solution.

Analysis Methods-Recovery and Enumeration of Bacteria
Disinfection rates were measured by membrane filtration as described by DIN EN ISO 8199:2007 [18,19].In this method, microorganisms are collected on a membrane filter and incubated on a specific nutrient pad (Endo for E. coli, Caso for B. subtilis, both from Sartorius (Goettingen, Germany)).In order to enumerate viable microorganisms, each sample was six-fold serially diluted in 0.9% saline solution and filtered.Colonies were counted after incubation for 24 h at 37 °C (E.coli) and 30 °C (B.subtilis).All experiments were performed in triplicate.The actual concentration of microorganisms inside a bacterial solution was then calculated by Equation ( 1), leaving the weighted average of the counting from each experiment expressed in CFU per mL as a result [19].The colony count range for valid results was set from 10 to 200 colonies, discarding filters with a higher or lower number of colonies.Before each disinfection process, the concentration of the initial stock solution N0 was measured by filtration as described above.The inactivation rate was furthermore determined by dividing the filtered concentration of the initial stock solution N0 by the evaluated bacterial concentration Ni after defined time intervals ti of 10, 40, and 90 s of UVC irradiation.The overall logarithmic inactivation rate was finally determined by calculating log10 [20].
Additionally, standard deviation was calculated between the three singly verified inactivation rates of each experiment with Excel ® (Microsoft, Redmond, WA, USA).

Output Power and Wavelength of UVC Light Source
Important features of the applied UVC light source are the actual output power and the emission wavelength peak, which set disinfection capabilities.The actual wavelength peak of the OPTAN280J UVC-LED [12] was determined by an Avaspec 2048 XL CCD spectrometer from Avantes (Apeldoorn, The Netherlands) at 281.8 nm and can be seen in Figure 3.The Avaspec 2048 spectrometer measures incoming photons in digitally recorded counts (ADC) according to an adjusted integration time (here: 50 μs) and, therefore, only delivers relative values visualizing the wavelength peak of this LED and not the overall output power.
Additional absolute values for the emitted optical radiant output power are delivered by measurements with an OPM150 power meter from Qioptics (Goettingen, Germany).Over a defined time interval of 90 s, a slight degradation (from 1.04 mW to 0.95 mW) of the LED output power was observed, leading to an average power output of 0.97 mW.In the datasheet, the overall radiant output of this LED type is defined between 1-2 mW and, therefore, corresponds to the measured value here [12].This value will also be the referenced as UVC power for the following experiments which leads to an overall UVC dose of 77.82 mJ/cm 2 for 90 s.

Investigation of Quartz as Light Guide for UVC radiation
Transmission spectra for HSQ ® 100 quartz and soda-lime AR ® glass are critical parameters for describing the actual light guidance effect.Therefore, the transmission curves for the applied quartz and soda-lime glass were measured with a Specord 250 Plus spectrophotometer from Analytik Jena (Jena, Germany).Figure 4 shows these measured transmissions in a wavelength range from 200 nm to 450 nm referenced on air.Soda-lime AR ® glass has a transmission rate of 2% at 280 nm.On the other hand, HSQ ® 100 quartz transmits 92% of the incoming UVC radiation in this wavelength.In practice, the strong absorbance of soda-lime glass would mean losing optical power at the reactor walls.Thus, 90% of UVC radiation impinging the glass walls could theoretically be saved by HSQ ® 100 due to these measured high transmittance characteristics.Further considerations are taken into account for describing reflection and transmission properties of UVC inside a quartz glass tube.The general optical equations (Snell's law (Equation ( 2)) and Fresnel laws) are valid here and allow the calculation of reflectance of UVC at different media interfaces [9].For getting a better understanding, Figure 5 represents a schematic view of the investigated problem description with interfaces from water to quartz and from quartz to water.Additionally, the refraction indexes of the different media of water, quartz, and air are depicted.The incident angle is described by Φ and the refraction angles by Φ , Φ , which can be calculated by Snell's law [9]: Reflection and transmission characteristics from water to quartz as well as from quartz to air are calculated by Fresnel's equations [9] and plotted in Figure 6.
According to these results, the total reflection at the quartz-air transition is given for incident angles Φi greater than 48°.Radiation impinging the quartz surface at a lower incident angle will be emitted almost completely on the outside and will be lost for disinfection.However, the applied UVC-LED has a viewing angle of 15° and, therefore, total reflection at the quartz walls can be ensured.

Absorbance Characteristics of UVC radiation in Different Media
One concern with UVC radiation is its strong attenuation, especially in polluted water.This effect has a direct influence on the evaluated reactor design with quartz tubes and is therefore very important in this scientific study.The attenuation of UVC inside a medium can be explained theoretically by Equation ( 5) [21]: (5) I(z): Beam intensity at distance z from radiating source (W/cm 2 ); I0: Initial intensity at z = 0 (W/cm 2 ); c: Concentration of absorbing molecule (mol/L); z: Layer thickness (cm); ε: Molar extinction coefficient (L/(mol•cm)).which leads to the formulation of Beer's law [9,22] log ( ) = ε = (6) A: Absorption (L/cm).
If monochromatic and collimated light passes through a homogeneous medium, the incident light is absorbed in proportion to the distance z from the radiating light source.Attenuation of I0 then follows Equation ( 6) and is dependent on the layer thickness z and a material-specific extinction coefficient ε [21].Extinction ratio ε is also referred to as absorption A or optical density OD.The absorption of the investigated bacteria media determines how deep UVC radiation is able to penetrate the cultivated bacteria solution.The measured absorbance of the applied B. subtilis, E. coli, 0.9% NaCl solution, and tap water referenced on distilled water from 250 nm to 300 nm can be seen in Figure 7.In Figure 8, resulting fluence rates of UVC through a defined medium are calculated by applying extinction coefficients ( ε) of 0.023 (~E. coli solution (10 6 CFU/mL)), 0.037 (~tap water from Ulm, Germany), and 0.081 (~B.subtilis solution (10 6 CFU/mL)) at a varying layer thickness z from 0 mm to 100 mm, neglecting scattering effects.Equation ( 5) delivers attenuation as described by Beer's law.The resulting fluence rate can be derived by taking the irradiated area into account, which is determined by the LED's viewing angle of 15° and the layer thickness z.Effects of total reflection (quartz) and absorbance (soda-lime) are recognizable as soon as the irradiated area reaches the inner area of the tubes.
As consequence of UVC attenuation, according to Beer's law, a high extinction coefficient prevents the radiation from entering deep into the media and the investigated light guide effect of quartz glass is severely restricted.An extinction coefficient below 0.1 would theoretically be a realistic absorbance range for average tap water quality, which also corresponds to the findings here [23,24].
According to Li et al., a high reflection coefficient at the inner wall of an UVC reactor can lead to an increased weighted average in fluence rate [13].The calculations from Figure 8 correspond to these findings.
Another issue is the interaction of UVC radiation with solid particles [25], causing scattering effects or delivering shaded areas for microorganisms, which reduces disinfection efficiency and disturbs the light guide effect as well.Figure 9 gives a summary of the possible interactions with solid particles concerning UVC disinfection according to Blume et al. [26].Thus, the theoretical basic conditions disclose that the highest inactivation rates can be achieved in a medium that has low absorbance and contains only little solid particles.Both of these demands could be met by a pre-switched filtering system, for example.

Disinfection Results for E. coli DSM 498 and B. subtilis DSM 402 in Stagnant Bacteria Solution
Figure 10 shows the evaluated inactivation rates for B. subtilis DSM 402 with UVC irradiation periods of 10 s, 40 s, and 90 s, comparing quartz glass and soda-lime glass without additional mixing of the samples during the experiments.The initial B. subtilis concentration N0 of the samples was determined at 1-3 × 10 6 CFU/mL.Additionally, the corresponding regression line as well as the correlation coefficients of each disinfection process, evaluated with Excel ® , are given.Standard deviation is marked with error bars.The overall log10 inactivation of B. subtilis reached from 1.04 log10 (soda-lime glass) to 1.79 log10 (quartz glass) after 90 s of UVC irradiation.Quantifiable differences of the inactivation efficiency between soda-lime and quartz glass were determined at 0.06 log10 (10 s), 0.57 log10 (40 s), and 0.75 log10 (90 s).The numerical values and the corresponding standard deviation for the data shown in Figure 10, as well as the differences in disinfection efficiency between the two glass types, can be seen in Table 1.The results for the stagnant experiments applying E. coli DSM 498 can be seen in Figure 11 as well as Table 2.The initial E. coli concentration N0 of the samples was also determined at 1-3 × 10 6 CFU/mL for all experiments.
Quartz glass shows higher inactivation rates of 0.35 log10 (10 s), 0.53 log10 (40 s), and 0.95 log10 (90 s) in the experiments, with inactivation results up to 2.80 log10 (quartz glass) and 1.85 log10 (soda-lime glass) after 90 s of UVC irradiation.The same experiments were carried out with continuous mixing of the samples during UVC irradiation.The results from Figure 12 show higher disinfection rates compared to the previous results without mixing.An overall reduction for E. coli of 1.21 log10 (soda-lime glass) and 4.28 log10 (quartz glass) after 40 s was observed.The investigated difference between quartz and soda-lime glass is 0.74 log10 (10 s) and 3.07 log10 (40 s).After 90 s, all filters for the disinfection experiments with quartz glass contained zero colonies and were therefore under the detection limit of 10 1 CFU/mL.The highest disinfection rate for soda-lime glass was measured at 2.59 log10 after 90 s of UVC irradiation.
A summary of all evaluated log10 inactivation rates and the corresponding standard deviation during this experiment are given in Table 3.There are noticeable improvements in E. coli inactivation with HSQ ® 100 glass in comparison to AR ® glass up to 3.07 log10.
The corresponding results for B. subtilis from Figure 13 and Table 4 support these results.Here, the overall reduction was determined at 2.47 log10 (soda-lime glass) and 3.86 log10 (quartz glass) after 90 s UVC-treatment.Increased inactivation rates of 0.29 log10 (10 s), 1.59 log10 (40 s), and 1.39 log10 (90 s) for quartz glass were measured uniformly during the experiment.Thus, the overall inactivation rate in the mixed disinfection experiments with B. subtilis was increased by quartz glass as well.However, the investigated differences between inactivation efficiency with and without the light guide approach are lower compared to the results with E. coli.All numerical inactivation results are highlighted in Table 4.
Uniformly increased disinfection rates are visible for both bacterial strains by utilizing quartz instead of soda-lime glass.This outcome indicates that the light guide effects for UVC disinfection deliver strong improvements in disinfection efficiency with UVC-LEDs of several log10 levels, especially in mixed samples.

Conclusions
Quartz tubes applied as light guides for UVC disinfection allow harvesting almost all of the incoming UVC power and using it for disinfection due to total reflection.A clearly visible difference could be identified between stagnant and mixed samples during UVC irradiation here.This difference could be explained by the mentioned attenuation of UVC radiation along the tube and the absorption and total reflection at the reactor walls, respectively.With this method, only the upper part of the bacteria sample is exposed to these high doses of nearly 80 mJ/cm 2 if there is no mixing during the experiment.The fluence rate decreases by 65% in the investigated B. subtilis solution and by 55% in the E. coli solution after a layer thickness of just 20 mm and tends close to zero after 60 to 70 mm without total reflection (see Figure 8).In contrast, utilizing total reflection with quartz, the fluence rate drops only by 40% (B.subtilis solution) and 22% (E. coli solution) after a layer thickness of 20 mm.The higher disinfection rates for quartz glass are therefore most likely caused by the slower descent of fluence along the tube and the resulting increase of the UVC dose.However, the disinfection results with stagnant bacteria samples show only minor improvements in efficiency with the light guide approach.Furthermore, as a consequence of these similar disinfection capabilities, the error bars of the calculated standard deviation partly overlap.Therefore, no clear improvements between quartz glass and soda-lime glass can be observed for the stagnant experiments in practical use.All experiments were repeated with additional mixing, showing much higher disinfection rates for both bacterial strains.By mixing, all bacteria contained in the irradiated samples were driven close to the UVC-LED and were therefore exposed to a much higher dose than without mixing.This effect can clearly be seen in the experiments.B. subtilis strains were more resistant to UVC radiation throughout the experiments than E. coli, which is most likely caused by the mentioned different Gram stain of both microorganisms [16,17].The quartz and, thus, the light guide concept showed higher disinfection rates of up to 3.07 log10 (E.coli) after 40 s of UVC irradiation and 1.39 log10 (B.subtilis) after 90 s of UVC irradiation.A strong improvement in disinfection efficiency from applying quartz as a light guide can be verified here in comparison to soda-lime glass.These results suggest that this concept is best suited for flow reactor designs in practical use.Because of the limited radiant output power of today's UVC-LEDs, which is usually less than 3 mW, and the high cost of at least several hundred dollars, the concept of total reflection could improve the profitability of such a system.Furthermore, a geometry that is able to homogenize the incoming radiation inside the tube could offer even more advantages.A possible way of circumventing the strong absorbance would be to couple UVC with UVA-LEDs.This combination showed increased disinfection results according to Chevremont et al. [20].

Figure 2 .
Figure 2. Schematic illustration of the designed experimental setup: (left) novel approach utilizing HSQ ® 100 quartz glass for total reflection and (right) reference setup with AR ® glass for quantification of the disinfection enhancements of this concept.

)N:
Number of CFU/mL; ni: Number of filters per dilution i; Z: Sum of counted colonies on filters with different dilutions d1, d2, …; VS: Reference volume, here: 1 mL; Vi: Volume per dilution i; Vtot: Sum of all used reference volumes.

Figure 3 .
Figure 3. Measured spectrum of OPTAN280J UVC-LED from 250 to 350 nm in ADC (analogue to digital converted) counts.

Figure 5 .
Figure 5. Optical reflection characteristics at water-quartz and quartz-air interfaces.

Figure 6 .
Figure 6.(a) Reflection and transmission from water to quartz with incident angles at the water-quartz interface Φi of 0° to 90°; (b) resulting reflection and transmission from quartz to air observed for incident angles at the water-quartz interface Φi of 0° to 90°.

Figure 9 .
Figure 9. Particle interactions with UVC radiation inside a medium.

Table 1 .
Inactivation results for HSQ ® 100 and AR ® glass types for B. subtilis DSM 402 after different UVC irradiation periods and the corresponding differences of microbial inactivation in log10 in stagnant samples.

Table 2 .
Inactivation results for HSQ ® 100 and AR ® glass types for E. coli DSM 498 after different UVC irradiation periods and the corresponding differences of microbial inactivation in log10 in stagnant samples.

Table 3 .
Inactivation results for HSQ ® 100 and AR ® glass types for E. coli DSM 498 after different UVC irradiation periods and the corresponding differences of microbial inactivation in log10 in mixed samples.

Table 4 .
Inactivation results for HSQ ® 100 and AR ® glass types for B. subtilis DSM 402 after different UVC irradiation periods and the corresponding differences of microbial inactivation in log10 in mixed samples.UVC Irradiation/