Application of GaN-based ultraviolet-C light emitting diodes – UV LEDs – for water disinfection

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Application of GaN-based ultraviolet-C light emitting diodes e UV LEDs e for water disinfection M.A. Wu¨rtele a,*, T. Kolbe b, M. Lipsz c, A. Ku¨lberg c, M. Weyers c, M. Kneissl b,c, M. Jekel a a

TU Berlin, Fachgebiet Wasserreinhaltung, Sekr. KF4, Strasse des 17, Juni 135, 10623 Berlin, Germany TU Berlin, Institut fu¨r Festko¨rperphysik, Hardenbergstr. 36, 10623 Berlin, Germany c Ferdinand-Braun-Institut, Leibniz-Institut fu¨r Ho¨chstfrequenztechnik im Forschungsverbund Berlin e. V., Gustav-Kirchhoff-Straße 4, 12489 Berlin, Germany b

article info

abstract

Article history:

GaN-based ultraviolet-C (UVeC) light emitting diodes (LEDs) are of great interest for water

Received 15 June 2010

disinfection. They offer significant advantages compared to conventional mercury lamps

Received in revised form

due to their compact form factor, low power requirements, high efficiency, non-toxicity,

4 November 2010

and overall robustness. However, despite the significant progress in the performance of

Accepted 9 November 2010

semiconductor based UV LEDs that has been achieved in recent years, these devices still

Available online 16 November 2010

suffer from low emission power and relatively short lifetimes. Even the best UV LEDs exhibit external quantum efficiencies of only 1e2%.

Keywords:

The objective of this study was to investigate the suitability of GaN-based UV LEDs for

Disinfection

water disinfection. The investigation included the evaluation of the performance charac-

UV light

teristics of UV LEDs at different operating conditions as well as the design of a UV LED

Light emitting diodes (LED)

module in view of the requirements for water treatment applications. Bioanalytical testing

Water treatment

was conducted using Bacillus subtilis spores as test organism and UV LED modules with emission wavelengths of 269 nm and 282 nm. The results demonstrate the functionality of the developed UV LED disinfection modules. GaN-based UV LEDs effectively inactivated B. subtilis spores during static and flow-through tests applying varying water qualities. The 269 nm LEDs reached a higher level of inactivation than the 282 nm LEDs for the same applied fluence. The lower inactivation achieved by the 282 nm LEDs was compensated by their higher photon flux. First flow-through tests indicate a linear correlation between inactivation and fluence, demonstrating a well designed flow-through reactor. With improved light output and reduced costs, GaN-based UV LEDs can provide a promising alternative for decentralised and mobile water disinfection systems. ª 2010 Elsevier Ltd. All rights reserved.

1.

Introduction

Ultraviolet (UV) systems are of growing interest in many water treatment applications. Based on their ability to function as a broad-spectrum antimicrobial agent with minimal disinfection

by-product formation, they are an alternative to chemical disinfectants. UV light at the proper fluence and wavelength inactivates microorganisms by disrupting their DNA or RNA molecules, rendering them unable to reproduce (e.g. Blatchley et al., 2008; Bolton and Cotton, 2008; Kolch, 2007).

* Corresponding author. Tel.: þ49 30 314 25368; fax: þ49 30 314 79621. E-mail address: [email protected] (M.A. Wu¨rtele). 0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2010.11.015

1482

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

DNA absorbs UV light with wavelengths between 200 and 300 nm. Maximum UV light absorption through DNA is typically reached at a wavelength of around 260 nm, but the peak wavelength distribution is dependent on the target organism (Bolton and Cotton, 2008; Chen et al., 2009). Chen et al. (2009) presented that Bacillus subtilis spores have two absorption maximums, one below 240 nm and one at around 270 nm. The fluence-inactivation response curves for 254 nm and 279 nm wavelengths were shown to be similar (Chen et al.,2009). Different wavelengths are emitted depending on the UV source: Conventionally, UV light is generated from mercury lamps. Low pressure mercury lamps emit nearly monochromatic UV light at a wavelength of 254 nm, whereas medium pressure lamps emit a polychromatic spectrum with various wavelengths (Bolton and Cotton, 2008). A relatively new method to generate UV light is the use of LEDs (Vilhunen et al., 2009). UV LEDs offer the possibility to use the optimum wavelength for disinfection instead of the 254 nm emitted by low pressure mercury lamps (Vilhunen et al., 2009). Further benefits of UV LEDs in applications for water purification e particularly in discontinuously operated systems e include: no disposal problem (LEDs do not contain mercury) compact and robust design: more durable in transit and during handling (no glass or filaments) - faster start-up time (no warm-up time) - ability to turn on and off with high frequencies (for just in time applications) - lower voltages, low power requirements - potential for higher energy efficiency - potential for longer lifetimes and reduced frequency of replacement. -

However, due to the low output power and high costs of LEDs at this stage of development, UV LEDs have until now only been tested for water disinfection using UV sensitive microorganisms such as E. coli (Bak et al., 2010; Crawford et al., 2005; Vilhunen et al., 2009). Since in the future an increase of the output power of UVeC LEDs is clearly expected because of their high physical improvement potential (Khan et al., 2008; Kneissl, 2008; Adivarahan et al., 2009), tests with less UV sensitive organisms as e.g. B. subtilis spores e used as surrogate organism for protozoan (oo)cysts (Hijnen et al. (2006)) and in UV unit certification (DVGW (2006a), USEPA (2006)) are needed. After this development step, UVeC LEDs will be a good candidate for practical water disinfection modules. Currently, no standardised testing system or protocol exists for UV LEDs, whereas in the case of conventional mercury lamps researchers have made considerable headway in standardising protocols to investigate UV disinfection on bench-scale. They have found that factors such as water quality, especially the UV transmittance (UVT) of unfiltered water, and contact time influence the delivered fluence (e.g. Cantwell and Hofmann, 2008; Caron et al., 2007; Templeton et al., 2005). In contrast to chemical disinfection systems, the contact time in UV disinfection cannot be monitored directly. It is influenced by the flow rate and hydrodynamics, determining the specific path of the organism through the reactor (Blatchley et al., 2008, MWH, 2005). This complex

interplay of different factors that influence the UV disinfection performance in flow-through reactors led to the development of a method to calibrate the expected performance of full-scale units. This method called biodosimetry was originally proposed by Qualls and Johnson (1983). Today there are several standards in Europe and the USA that specify experimental protocols for performing microorganism inactivation versus fluence for mercury lamp based measurements using e.g. B. subtilis spores. These protocols include a “collimated beam device” (CBD) to deliver a highly uniform beam of UV light to a water sample in a Petri dish. According to Bolton and Linden (2003), it is not necessary to completely standardise a bench-scale apparatus, but basic guidelines should be considered when designing a modified apparatus for a specific application. Amongst other aspects, the design has to ensure that the beam irradiating the water sample is reasonably uniform and the divergence is small enough to ensure accurate sensor readings. The average fluence for all microorganisms in the suspension has to be kept equal by carefully controlled stirring (without vortex) (Bolton and Linden, 2003). The conventional design, whereby the water sample is irradiated from the top down, limits the use of UV LEDs because these methods need high power output sources to compensate for losses between the UV light source and the water sample. The challenge of this project was to develop a UV module that takes the relatively low output power of current UV LEDs into account. The design had to enhance the uniformity of the inhomogeneous light output from UVeC LED arrays while reducing UV light losses. Within the scope of this research, the applicability of UV LEDs for the disinfection of water was investigated following three steps: First, a UV module for static tests was constructed and tested with various water qualities. The findings were then compared to the results obtained with a conventional standardised mercury lamp system. After validating the constructed module, the influence of two different LED wavelengths on the disinfection of B. subtilis spores was compared based on their disinfection capacity and power consumption. In a last step of the investigation, real water disinfection applications were simulated with a bench-scale flow-through reactor.

2.

Material and methods

2.1.

Experimental set-up

Two UV LED disinfection modules were constructed in order to perform biodosimetry trials with GaN-based LEDs emitting at 269 nm and 282 nm. The LEDs were placed on the base of the water disinfection module at a distance of one cm in order to obtain a sufficiently high power density and a nearly homogeneous UV light distribution. In test module 1 (not shown) an array of 33 LEDs (269 nm) with an emission power of 0.33 mW at 20 mA per LED was arranged in a grid of 1 LED/cm2. Due to the LED array design only 28.5 of the LEDs were overlapping with the footprint of the Petri dish. For test module 2 (Fig. 1a and b) 35 LEDs (with an emission wavelength of 282 nm) were positioned in three concentric circles with diameters of 1.8 cm, 3.5 cm and 5.2 cm

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

in order to perform additional biodosimetry trials with a simple flow-through reactor. At the same current of 20 mA the emission power of the 282 nm LEDs was nearly two times higher than that of the 269 nm LEDs. The module characteristics are summarised in Table 1. A schematic diagram of the static and flow-through experimental set-up is shown in Fig. 2 (a and b). For static tests (Fig. 2a) a Petri dish with a diameter of 6 cm was placed on the top of the UV LED array with a 2 mm thick Suprasil base that allows over 90% of the UVeC light to be transmitted. The output power of the LEDs was measured with a UV sensitive silicon photodiode, guaranteeing the same irradiation fluence during all microbiological tests. In order to obtain a homogeneous irradiation of the entire water volume, the water was stirred during the tests from the top of the Petri dish with an electrically driven stirrer. For the 282 nm UV LED module also a simple flow-through reactor was designed with a maximum flow rate of w12 ml/min (inset of Fig. 1a). The reactor body was made of aluminium (for UV reflection) into which 6 mm wide and 5 mm deep water channels were milled (see inset of Fig. 1a). The reactor body was then covered with a 2 mm thick Suprasil window to enable UV exposure of the water channels. For the flow-through tests the Petri dish of the disinfection module is replaced by the flow-through reactor. The test water flows from the feed reservoir through the flowthrough reactor and is catched in the catchment tank (Fig. 2b).

2.2.

Characterisation of LEDs

A series of GaN-based UV LEDs with emission wavelengths of 269 nm and 282 nm were then characterised in order to investigate the optimum operating conditions for the water disinfection modules. A typical emission spectrum of a 269 nm and a 282 nm LED is shown in Fig. 3. The spectra were measured under continuous-wave (cw) conditions at 20 mA with an optical fiber spectrometer. No variation in the peak wavelength can be observed for the 269 nm LEDs. UV LEDs emitting at 282 nm showed a slightly larger distribution of emission wavelength, but the difference was still less than 1 nm between the different LEDs. The measured full-width at half-maximum (FWHM) of the emission spectra was typically between 10 and 11 nm. The current-voltage characteristics for the UV LEDs were measured with an Agilent power supply and a Keithley 2000

1483

multimeter under cw conditions. All LEDs had very similar current-voltage characteristics with typical operating voltages (at 20 mA) of around 5.8 V for the 269 nm LEDs and operating voltages of around 6.3 V for the 282 nm LEDs. For the measurement of the emission power a calibrated silicon photodiode with a detector area of 100 mm2 was used. In Fig. 4 a typical light output power vs. current (LI) characteristic for a 269 and 282 nm LED is plotted. For the 269 nm LEDs a typical emission power of 0.33 mW was observed at a current of 20 mA. For the 282 nm LEDs the emission power of 0.65 mW measured at 20 mA was almost double that of the 269 nm LEDs. Maintaining a constant output power is a critical parameter for a water disinfection module. Fig. 5 shows the change in LED light output for a 269 nm LED measured at a fixed current of 20 mA. After 100 h the emission power had decreased by 40%, while the series resistance had increased from 11.8 Ohm (0 h) to 12.9 Ohm (after 100 h). However, no change in the emission wavelength was observed. The reasons for the strong degradation in the first operating hours are not completely understood but most likely related to the high defect density in the AlGaN materials. However, after the initial drop the emission power nearly stabilizes at operating times longer than 100 h. The strong degradation of the emission power requires active monitoring of the light output during testing and adjustment of the drive current in order to maintain a constant optical power density. It can be expected that newer generations of UV LEDs with reduced defects densities will exhibit significantly reduced degradation rates and improved lifetimes. For example, recent reports by Adivarahan et al. (2009) (Adivarahan et al., 2009) show lifetimes well in excess of 1500 h with an output power of 42 mW for a LED lamp emitting at 280 nm. Further improvements in UV LED output power and lifetimes can be expected.

2.3.

Disinfection tests

A UV fluence-inactivation response curve of B. subtilis ATCC 6633 spores was generated with a laboratory apparatus especially designed for UV LEDs. The test organism was obtained from the Institute of Hygiene and Public Health (University of Bonn), where it was cultivated and characterised according to the German standard DVGW 2006c with monochromatic low pressure UV lamps (DVGW 2006b). In the course of the cultivation process B. subtilis spores (obtained from the DSMZ,

Fig. 1 e Test module 2; a) basic unit and flow-through reactor; b) LED array with stirrer unit.

1484

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

Table 1 e LED parameters during tests (mean values) of the disinfection modules (stationary tests) Module 2 was constructed to perform flow-through tests. Module

1 2

LED wavelength (nm) 269 282

LED output power during tests (mW) 0.16 0.19

Number of LEDs

Number of active LEDs

LED configuration

Output power disinfection module (mW)

33 35

German Resource Centre for Biological Material, Braunschweig, Germany) were cultivated on Columbia agar plates (24 h, 37  1  C). Individual colonies were then inoculated into nutrient solution. After incubation (72 h, 37  1  C) the liquid culture was dispersed in an ultrasonic bath (50 kHz, 2 min 10  C) and separated by centrifugation (ca. 500 g, 5e10  C, 15 min). Spores were washed three times with sterile, deionised water by collecting, centrifuging (5000 g, 15 min) and resuspending. The suspension was placed in a water bath at 80  C for 10 min to destroy remaining vegetative cells and refrigerated at 4  C. It was delivered in deionised water with a concentration of 109 cfu/ml. For the exposure tests the applied test organism was suspended in the test water according to DVGW 2006c to obtain a concentration of 106 to 107 cfu/ml. Tests were conducted at room temperature (23  2  C). The tests were performed with stationary samples of 30 ml, exposed successively to decreasing UV fluences. The fluence (J/m2) was calculated as the product of irradiance (W/m2), exposure time (s), number of LEDs and corrected for UVT. During tests with the 269 nm LEDs, samples of 1.5 ml were taken after 372, 248, 155, 62 and 0 s. Samples of the tests conducted with the 282 nm LEDs were taken after 255, 170, 106, 43 and 0 s to achieve comparable fluences (approx. 600, 400, 250, 100, 0 J/m2). Flow-through tests were performed in a single pass operation mode. Different fluences were obtained by varying the flow rate and optical power output of the LEDs. Flow rates were chosen based on limitations of the module design and the available light output power from the UV LEDs: 10.8  0.6 ml/min and 7.8  0.6 ml/min, resulting in a residence time of 45.8 s and 63.5 s respectively and laminar conditions. The LED (282 nm) power at 10.8 ml/min was set at 0.5; 0.7; 0.9 mW and at 0.35; 0.49; 0.59 mW for the flow rate of 7.8 ml/min.

28.5 35

2

1/cm raster 3 circles

4.56 6.65

The experiments were performed based on DVGW 2006b starting with the highest fluence. The following test protocol was applied (cp. Fig. 2b): 1. initialising the system (adjustment of flow rate and UV power), 2. sampling in feed reservoir, before UV exposure 3. turning on UV light, 4. starting flow, discarding five test cell volumes, 5. sampling after UV exposure (after 1, 2, 3 and 4 min), 6. sampling in feed reservoir, before UV exposure.

2.4.

Microbiological analysis and data analysis

Determination of spores was performed according to DVGW 2006b. The viable microorganisms were enumerated by diluting the samples, plating and incubating the dilutions and counting the colonies that arose as colony-forming units (cfu) on plates with 1e300 colonies. Samples were cultivated in triplicates. The log inactivation (RF ¼ log (N0/N )) was plotted against the fluence to derive a fluence-inactivation relation. The fluence-inactivation response curve of B. subtilis can be described by three phases (Hijnen et al., 2006): A shoulder phase, a log-linear phase and a tailing phase. When applying low fluences, log (N0/N ) changes only slightly with increasing fluence. Researchers attributed this phase to DNA repair or the requirement of several DNA damage sites. After an offset fluence, inactivation starts in a log-linear relationship, followed by a sometimes existing tailing phase, in which log (N0/N ) again changes slowly with fluence. Causes for the tailing phase are still under discussion, possible causes could be microorganism clumping or association with particles, experimental bias or hydraulic effects. This curve progression, not considering the tailing phase, was described by a shoulder model:

Fig. 2 e Schematic diagram of the experimental set-up for a) static tests and b) flow-through tests.

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

Fig. 5 e Emission power measured over 100 h for a 265 nm LED at a fixed drive current of 20 mA.

Fig. 3 e Typical emission spectrum of a) 269 nm and b) 282 nm LED.

RF ¼ k$Fluence  b;

(1)

where RF is the decimal reduction factor (¼ log(N0/N ), k is the inactivation rate constant (m2/J) and b is the offset value  a negative value  which crosses the fluence axis at the fluence, where log-linear relationship starts (Hijnen et al., 2006). The goodness of fit of the linear regression was analysed with the coefficient of determination (r2) and the residual standard deviation (RSD). The log inactivation results presented in Section 3.3 are geometric mean values. In the case of the flow-through tests, N0 was calculated by averaging the test results of the three samples taken before UV exposure.

2.5.

1485

Water quality

Tests were performed with different water qualities: deionised water (DI), tap water (TW), surface water (SW) and secondary effluent (SE). Water samples were taken in Berlin, Germany. Tap water was obtained from the local water supply of the city of Berlin. Surface water samples were taken at the Landwehrkanal and secondary effluent was provided by the waste water treatment plant Ruhleben. The absorption coefficient (A) and the UV transmittance (UVT) of the filtered and unfiltered test waters were measured with a two beam spectrometer (model Lambda 12, Perkin Elmer) in a five cm quartz cuvette. The test waters were filtered with 0.45 mm cellulose nitrate filters (Sartorius AG).

Fig. 4 e Typical light output vs. current characteristic (LI) of a) 269 nm and b) 282 nm LED.

The absorbance at 254 nm relates to a one m path length. UV transmittance is defined as the percent transmittance in the medium when the path length is one cm and the wavelength is 254 nm. Since tests were performed with UV LEDs emitting at 269 and 282 nm, A and UVT were also determined for these wavelengths. Fluence correction was performed with UVT (254). Table 2 summarises the measured water absorption parameters. The high turbidity of the B. subtilis suspension reduced UVT (254) of the unfiltered deionised water samples. Higher absorption of the test waters used in experiments with the 282 nm LEDs was caused by a higher initial spore concentration. The tap water of the city of Berlin contains a high amount of UV active dissolved organic matter, resulting in a high UV absorbance of the tap water tests samples.

3.

Results and discussion

3.1.

Module development and validation

The test modules were developed based on the low output power of the UV LEDs causing long irradiation times and low flow rates. The design consists of a LED array, irradiating the water sample from the bottom up, in contrast to the conventional collimated beam device (CBD), where the mercury lamp is located on top of the water sample. The inhomogeneous light emission was addressed by constant stirring during static tests. Validation tests were performed with arrays of UV LEDs emitting at 282 nm. B. subtilis spores, suspended in deionised water, were used as test organism and exposed successively to UV light. In a first step of the validation process, the reproducibility of the test results obtained with the newly designed module was evaluated by the standard deviation of repeated test runs and by applying various water qualities. In a second step of the validation, the test results were compared to disinfection tests conducted on a standardized mercury lamp CBD. The inactivation results to evaluate the reproducibility of the test module with different water qualities are presented in Fig. 6. Increasing UVT was considered in the fluence calculation causing lower applied fluences at the same exposure time, for waters containing higher amounts of UV absorbing compounds.

1486

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

Table 2 e Water quality parameters (mean values) of the applied test waters. Parameter

Unit

Deionised Water (DI)

Tap Water (TW)

Surface Water (SW)

Secondary Effluent (SE)

Test waters for tests conducted with 282 nm LEDs: Unfiltered A (254) [1/m] UVT (254) % A (282) [1/m] Filtered A (254) [1/m] A (282) [1/m]

2.7 94.1 2.3 0.69 0.43

10.8 78.0 8.2 7.9 5.4

18.4 65.5 13.4 15.9 11.0

28.7 51.7 22.1 23.6 17.6

Test waters for tests conducted with 269 nm LEDs: Unfiltered A (254) [1/m] UVT (254) % A (269) [1/m] Filtered A (254) [1/m] A (269) [1/m]

1.1 97.5 n.a. 0.8 n.a.

e e e e e

e e e e e

e e e e e

n.a.: not available.

The UV sensitivity of a microorganism is described by the inactivation rate constant k (m2/J) and the offset value b, derived from Eq. (1) (Section 2.4). Linear regression was performed for all data points between 100 and 600 J/m2 according to Eq. (1) with a high goodness-of-fit (RSD ¼ 0.44; r2 ¼ 0.94). Data points beneath 100 J/m2 were excluded because of an observed shoulder effect and therefore a non-log linear relationship. The maximum deviation between the 282 nm LED test runs in DI water (0.5 log reduction) was in the same order of magnitude as the standard deviation of the triplicate analysis (0.4 log reduction), indicating that the test set-up generates reproducible results. Reproducibility of the tests conducted with the UV LED test module is also underlined by a highest standard deviation of only 0.8 between all the conducted experiments. In a second validation step, the same organism was used under different experimental conditions: in the 282 nm LED module and the 254 nm mercury lamp CBD. According to

Fig. 6 e Fluence-inactivation response and the curve derived by linear regression of B. subtilis spores in different waters (DI: deionised water, TW: tap water, SW: surface water, SE: secondary effluent) with different qualities (UVT as representative parameter) for 282 nm LEDs; in parentheses are the numbers of test batches. Presented data are geometric mean values ± standard deviation of three test rows. Test rows were cultivated in triplicates.

literature data, the fluence-inactivation response curve of B. subtilis spores should be comparable at wavelengths of 254 nm and 279 nm (Chen et al., 2009) and therefore the sensitivity e described by the inactivation rate constant k (m2/J) e of the surrogate spores should be comparable in both experimental set-ups. Spore inactivation results in DI water obtained with 282 nm LEDs and a conventional mercury lamp are presented in Fig. 7. A shoulder effect was observed for the 282 nm LEDs, whereas the mercury lamp inactivation curve had no shoulder. Tailing was neither observed for the 282 nm LEDs nor for the mercury lamp at the investigated fluences. A linear fluence-inactivation relationship was observed in both experimental set-ups between 100 and 600 J/m2. Regression analysis lead to low RSD and r2 values for the 282 nm LEDs (RSD ¼ 0.281, r2 ¼ 0.99) and the mercury lamp CBD results (RSD 0.194, r2 ¼ 0.98). A comparison of the inactivation rate constants k of the 282 nm LEDs (0.0132 J/m2) and the mercury lamp (0.0056 J/m2) indicate a two times higher sensitivity of the B. subtilis spores in the LED module. The difference of spore reduction between the 282 nm LEDs and the mercury lamp therefore increases

Fig. 7 e Fluence-inactivation response and curves derived by linear regression of B. subtilis spores in deionised water for 282 nm LEDs and a mercury lamp. Presented data are geometric mean values ± standard deviation of three test rows. Test rows were cultivated in triplicates.

1487

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

Fig. 8 e Comparison of fluence-inactivation response and curves derived by linear regression of B. subtilis spores in deionised water, obtained with UV LEDs of 269 and 282 nm in the static apparatus. Presented data are geometric mean values ± standard deviation.

with higher fluences. Based on the presumption that the different wavelengths should result in the same sensitivity (according to Chen et al., 2009), the difference of the disinfection kinetics has to be attributed to the different experimental set-ups. Various factors, such as the condition of the test suspension, the fluence calculation and/or the different constructions of the LED apparatus and the CBD, might influence the test results. At this stage, a clear explanation for this slightly higher disinfection capacity at higher fluences compared to the conventional UV source still has to be investigated. However, since the LED system generated reproducible results, the inactivation results obtained for different wavelengths on the same experimental set-up are comparable. The influence of different LED wavelengths on the disinfection of B. subtilis spores is discussed in the following section.

3.2. Comparison of LEDs emitting at 282 nm and 269 nm The disinfection capacity of 269 and 282 nm emitting LEDs were investigated by running disinfection tests with deionised water. According to former research, the 269 nm wavelength

is the absorption maximum of the B. subtilis spores (Chen et al., 2009) and should therefore show a greater inactivation than the 282 nm LEDs. On the other hand, the 282 nm LEDs have a higher photon output. The consequences are discussed in the following. Fig. 8 presents the inactivation curves derived from the results obtained with different UV LED wavelengths in the static apparatus configuration. Deduced from linear regression B. subtilis is equally sensitive to both wavelengths above 100 J/m2 (k282 ¼ 0.0132; k269 ¼ 0.0133). The offset value for the 282 nm LEDs is negative, indicating a shoulder effect, as discussed above. The offset value for the 269 nm LEDs is positive and therefore no shoulder effect is existent. Since the 269 nm LEDs show no shoulder effect, the absolute difference of more than 1 log in the investigated fluence range demonstrates a higher absolute disinfection using 269 nm LEDs. This enhanced disinfection capacity could be attributed to the higher germicidal effectiveness at the wavelength of 269 nm. In a next step, the 269 and 282 nm LEDs were compared based on the same input power and time, resulting in different fluences. A model calculation was conducted based on a nominal drive current of 20 mA, at which the 269 nm LEDs have an optical power output of 0.33 mW, whereas the 282 nm LEDs have an optical power output of 0.65 mW. Various resulting fluences were calculated as shown in Section 2.3. The resulting inactivation was calculated from the derived inactivation curves in Fig. 8. Results of the model calculation are summarised in Table 3. Although the 269 nm LEDs exhibit a better germicidal effectiveness, the spore inactivation caused by the 282 nm LEDs is significantly better than for the 269 nm LEDs during the same time span and at the same input power. Irradiation for a period of 300 s, for example, leads to an applied fluence of 175 J/m2 for the 269 nm LEDs and an applied fluence of 345 J/m2 for the 282 nm LEDs. At this fluence the reduction factor for the 269 nm LEDs is 0.9 logs lower than for the 282 nm LEDs. This is due to the higher photon output (at the same current) of the 282 nm LEDs. The higher disinfection capacity of the 269 nm LEDs, which is due to an output wavelength close to the absorption maximum of the B. subtilis spores, is compensated by a higher photon efficiency of the 282 nm LEDs. As a consequence, the use of 282 nm LEDs is preferable for the overall performance of the UV purification module as long as the performance of the LEDs at shorter

Table 3 e Summary of model calculation for the comparison of power consumption and inactivation performance of the 269 nm and 282 nm LEDs; boundary conditions: input current of 20 mA and a total of five LEDs. Reduction factor: RF269 [ 0.01133 3 D 0.5547; RF282 [ 0.0132 3 L 0.720.

Time (s) 200 250 300 350 400

269 nm

282 nm

Output: 0.33 mW

Output: 0.65 mW

2

Difference

Fluence (J/m )

RF269 (log N0/N)

Fluence (J/m2)

RF282 (log N0/N )

RF282 e RF269 (log N0/N )

117 146 175 204 233

1.9 2.3 2.7 3.1 3.6

230 287 345 402 460

2.4 3.0 3.7 4.4 5.1

0.5 0.7 1.0 1.3 1.5

1488

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

Nevertheless, an increase in the applied fluence leads to a higher inactivation, indicating a promising design of the UV LED configuration in the flow-through reactor. In the German UV reactor certification for potable water production a target fluence of 400 J/m2 is given (DVGW 2005). Based on the sensitivity of the applied test organism in CBD tests, 400 J/m2 are reached, when the spores are reduced by 3 log (cp. 3.1). However, even with a lower sensitivity of the applied test spores in flow-through tests we were able to demonstrate a 3 log reduction in B. subtilis spore count applying only a minimally higher fluence of approx. 450 J/m2 than needed in CBD tests. Fig. 9 e Fluence-inactivation response and the curves derived by linear regression of B. subtilis spores in deionised water, obtained with the apparatus designed for 282 nm UV LEDs in static tests (regression analysis of static tests in DI water) compared to results obtained during flow-through tests. Presented data are geometric mean values ± standard deviation.

wavelength is not improved. In the future, with an increasing output power of the LEDs available for both wavelengths, the trend may be different. The optimum wavelength will have to be chosen based on a comparison between the UV output power and the reduction factor applying the same fluence.

3.3.

Flowethrough tests

During this study flow-through tests were conducted with the higher output 282 nm LEDs. Flow-through tests were performed with a flow-through reactor to obtain the first results for the applicability of UV LEDs under real conditions. Differing fluences were obtained by varying the flow rate and optical power output of the 282 nm LEDs (see 2.3). The flow rate was adjusted to 10.8  0.6 ml/min and 7.8  0.6 ml/min, resulting in laminar conditions. The results are presented in Fig. 9. Linear regression was performed, including all data obtained with a flow rate of 10.8  0.6 ml/min and 7.8  0.6 ml/min, to investigate, if the flow rates have an influence on the spore inactivation. The low RSD of 0.20 and a goodness-of-fit of r2 ¼ 0.91 indicate that e applying laminar conditions e the flow rate has no significant influence on the spore reduction. Therefore the fluence-inactivation response curve e including data of differing flow rates e was used for further evaluation. With regard to this curve, the sensitivity (inactivation rate constant k) of the B. subtilis spores was more than halved in flow-through tests compared to static test results (represented in Fig. 9 as regression analysis of static tests). Although nominally the same fluence was applied, the inactivation of B. subtilis in the flow-through test reactor was reduced compared to static tests. This is a common phenomenon when up-scaling UV reactors and constructing flow-through reactors instead of static reactors. These results indicate that the flow conditions lead to areas of lower UV irradiation by incomplete illumination and short-circuiting by shadowing effects within the flow-through test reactor, reducing the overall disinfection efficiency.

4.

Conclusions

The first test results indicate an effective inactivation of B. subtilis spores through UV LEDs emitting at wavelengths of 269 and 282 nm. However, water purification applications where water has to be disinfected within a few seconds are still limited by long exposure times of up to 3e4 minutes, which are caused by the low output power of the UVeC LEDs. The following conclusions can be drawn: Test UV LED water purification modules generate reproducible results for different water sources. - The germicidal effectiveness of 269 nm LEDs is higher than that of 282 nm LEDs. - The lower inactivation of 282 nm LEDs is currently more than compensated by a higher light output and therefore the overall performance of the 282 nm LED module is preferable. - Flow-through test conditions reduce the disinfection capacity but the inactivation is still linearly dependent on the fluence, when applying laminar flow conditions. - With improved power output, life span and reduced costs through production scale-up and market competition e in future e LEDs are a promising means for decentralised and mobile water disinfection systems. -

Acknowledgements This work was performed under contract with the Berlin Centre of Competence for Water in the frame of the FP6 project TECHNEAU, and financed by the European Commission and Veolia Water. The authors thank Boris Lesjean and Eric Hoa from the Berlin Centre of Competence for Water and Florencio Martin from Veolia Water, Anjou Recherche, for their helpful expertise and the Berliner Wasserbetriebe for providing the samples of secondary effluents. Katharina Kutz is gratefully acknowledged for her laboratory work.

references

Adivarahan, V., Heidari, A., Zhang, B., Fareed, Q., Hwang, S., Islam, M., Khan, A., 2009. 280nm Deep Ultraviolet light

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 1 4 8 1 e1 4 8 9

emitting diode lamp with an AlGaN multiple quantum well active region. Applied Physics Express 2, 102101. Bak, J., Ladefoged, S.D., Tvede, M., begovic, T., Gregersen, A., 2010. Disinfection of Pseudomonas aeruginosa biofilm contaminated tupe lumens with ultraviolet C light emitting diodes. Biofouling 26 (1), 31e38. Blatchley, E.R., Shen, C., Scheible, O.K., Robinson, J.P., Ragheb, K., Bergstrom, D.E., Rokjer, D., 2008. Validation of large-scale, monochromatic UV disinfection systems for drinking water using dyed microspheres. Water Research 42 (3), 677e688. Bolton, J.R., Cotton, C.A., 2008. The Ultraviolet Disinfection Handbook. American Water Works Association, Denver. Bolton, J.R., Linden, K.G., 2003. Standardization of methods for fluence (UV dose) determination in bench-scale UV experiments. Journal of Environmental Engineering-Asce 129 (3), 209e215. Cantwell, R.E., Hofmann, R., 2008. Inactivation of indigenous coliform bacteria in unfiltered surface water by ultraviolet light. Water Research 42 (10e11), 2729e2735. Caron, E., Gabriel Jr., C., Barbeau, B., Payment, P., Pre´vost, M., 2007. Impact of microparticles on UV disinfection of indigenous aerobic spores. Water Research 41, 4546e4556. Chen, R.Z., Craik, S.A., Bolton, J.R., 2009. Comparison of the action spectra and relative DNA absorbance spectra of microorganisms: information important for the determination of germicidal fluence (UV dose) in an ultraviolet disinfection of water. Water Research 43 (20), 5087e5096. Crawford, M.H., Banas, M.A., Ross, M.P., Ruby, D.S., Nelson, J.S., Boucher, R., Allerman, A.A., 2005. Final LDRD Report: Ultraviolet Water Purification Systems for Rural Environments and Mobile Applications. Sandria National Laboratories, Albuquerque, New Mexico, pp. 1e37. DVGW, 2005. Arbeitsblatt W 290: Trinkwasserdesinfektion Einsatz und Anforderungskriterien. Deutsche Vereinigung des Gas- und Wasserfaches, Bonn. DVGW, 2006a. Arbeitsblatt W 294-1: UVeGera¨te zur Desinfektion in der Wasserversorgung Teil 1: Anforderungen an

1489

Beschaffenheit, Funktion und Betrieb. Deutsche Vereinigung des Gas- und Wasserfaches, Bonn. DVGW, 2006b. Arbeitsblatt W 294-2: UVeGera¨te zur Desinfektion in der Wasserversorgung Teil 2: Pru¨fung von Beschaffenheit, Funktion und Desinfektionswirkung. Deutsche Vereinigung des Gas- und Wasserfaches, Bonn. DVGW, 2006c.ArbeitsblattW294-3: UVeGera¨te zurDesinfektionin derWasserversorgung Teil 3: MessfensterundSensoren zur ¨ berwachungvonUV-Desinfektionsgera¨tenradiometrischenU Anforderungen, Pru¨fungundKalibrierung. Deutsche Vereinigung desGas- undWasserfaches, Bonn. Hijnen, W.A.M., Beerendonk, E.F., Medema, G.J., 2006. Inactivation credit of UV radiation for viruses, bacteria and protozoan (oo)cysts in water: a review. Water Research 40, 3e22. Khan, A., Balakrishnan, K., Katona, T., 2008. Ultraviolet lightemitting diodes based on group three nitrides. Nature Photonics 2, 77. Kneissl, M., 2008. Ultraviolet light-emitting diodes promise new solutions for water purification. World Water & Environmental Engineering 31 (3), 35. Kolch, A., 2007. UV-Disinfection of drinking water e the new DVGW work sheet 94 part 1e3. IUVA News 9, 17e20. MWH, 2005. Water Treatment: Principles and Design. John Wiley & Sons, Inc, Hoboken, NJ. Qualls, R.G., Johnson, J.D., 1983. Bioassay and dose measurement in UV disinfection. Applied and Environmental Microbiology 45 (3), 872e877. Templeton, M.R., Andrews, R.C., Hofmann, R., 2005. Inactivation of particle-associated viral surrogates by ultraviolet light. Water Research 39, 3487e3500. USEPA, 2006. 815-R-06e007. In: EPA (Ed.), Ultraviolet Disinfection Guidance Manual for the Final Long Term 2 Enhanced Surface Water Treatment Rule (WashingtonDC, USA). Vilhunen, S., Sa¨rkka¨, H., Sillanpa¨a¨, M., 2009. Ultraviolet lightemitting diodes in water disinfection. Environmental Science and Pollution Research 16 (4), 439e442.