Spectral Distribution of UV Light During Aging of Medium Pressure UV Lamps

Transcription

1 Spectral Distribution of UV Light During Aging of Medium Pressure UV Lamps Charles M. Sharpless, Karl G. Linden Duke University, Department of Civil and Environmental Engineering 120 Hudson Hall Durham, NC Paris Neofotistos ONDEO Degremont, Inc Emerywood Parkway Richmond, VA ABSTRACT As part of a project to validate a UV reactor for disinfection of drinking water, the germicidal output of four medium pressure lamps housed in a continuously operating reactor was followed as a function of time. Spectra were measured monthly and were used in conjunction with radiometer readings to calculate germicidal fluence rates (using DNA absorbance weighting) at the radiometer sensor position with the lamps operating at full power (4 kw) during each measurement. Based on the measured fluence rates, the power was adjusted so that the UV lamps operated continuously to deliver a germicidal output of approximately 70% of that obtained at full power at the end of an initial 100 h burn-in period when the lamps were new. The experimental protocol was designed to ensure reproducibility of spectrometer and radiometer positioning and response. Under the conditions of this test, lamp aging has been relatively slight. The average depreciation in germicidal fluence rates for all four lamps was 17% after 3980 h of operation. Slightly more depreciation was observed in the UVC than in the UVB region (18% versus 15% for the average of all lamps). Consistent with this, the germicidal fluence rate depreciated slightly more rapidly than the total (200 to 350 nm) fluence rate. These changes indicate a more rapid depreciation of the shorter wavelengths, but the extent to which this is due to inherent changes in the lamp spectra compared to solarization of the quartz sleeves cannot be determined from the present data. Efforts will be made to separate these two effects by examining the quartz sleeves at the end of the tests. INTRODUCTION The way in which UV lamp irradiance depreciates over time has been a topic of interest ever since UV equipment was applied to pathogen inactivation (disinfection). Unfortunately, there is very little information available on the topic and even less has actually been published. In wastewater disinfection, many plants will operate the UV lamps past their expected life, because even with lower germicidal output than originally assumed, most plants will still meet their discharge permits. However, over the past several years more emphasis has been put on accurately assessing dose delivery, especially with the wide acceptance of UV disinfection for drinking water and wastewater reuse applications. This requires more careful monitoring of the UV lamps in the system and precise information regarding the depreciation of the UV lamps over time under worst-case conditions.

2 Currently, there is no industry standard protocol for aging and measuring the depreciation of UV lamps. Several approaches are possible: the UV lamps could be aged and measured in the air; they could be aged in water and measured in air; or they could be aged and measured in water. Lamp operation may also differ during aging; the lamps may or may not be cycled on and off, and the power levels may be set to provide a range of irradiance levels during the aging test. Measurement equipment has not been standardized either. For example, some equipment allows only a narrow band of wavelengths around 254 nm to be measured while others allow measurement of a broad spectrum of wavelengths in the UV-C and UV-B range. In December 2000, a protocol for lamp aging was introduced in the NWRI/AWWARF UV Disinfection guidelines [AWWARF, 2000]. The guidelines specified a narrowly defined method for determining lamp output depreciation over time. The guidelines specified aging in water, addressed dimming and cycling and minimum measuring standards. The NWRI/AWWARF guidelines were the basis for the aging and measurement of the UV lamps outlined in this paper. MATERIALS AND METHODS Overview: The tests were designed to determine the germicidal output of the lamps over time relative to their output after a 100 h burn-in period during which the lamps were operated at full power (4 kw). In the first test after this period, the germicidal output was measured at full power and at 75% power (3 kw), and the data were used to determine the operating power that would provide approximately 70% of the full power germicidal output. Between tests, the lamps were operated at sufficient power to maintain 70% germicidal output (i.e., after every test the operating power was increased slightly to account for lamp depreciation as determined during the test). Tests were conducted at two places. For approximately the first 4000 h of testing, the reactor was installed at the Appomatox River Water Authority drinking water treatment plant (ARWA, Chesterfield, VA) where water was taken post-filtration. A typical absorbance spectrum of the water delivered to the reactor is shown in Figure 1. Problems with maintaining adequate flow led to relocating the reactor to Degremont North American Research & Development (DENARD) where water was taken from the tap and re-circulated through the reactor and a chiller with the addition of approximately 10% fresh tap water to ensure adequate cooling of the lamps. All the data presented here were obtained after operation at ARWA, and tests are presently continuing at DENARD. During tests and daily operation of the reactor, the water flow rate was approximately 100 gal/min, which was sufficient to keep the temperature of the reactor at that of the water (at or below 78 o F). During daily operation, the lamps were subjected to one on/off cycle per day, and one automatic wiper cycle per day. UV Reactor: The UV reactor used for aging and monitoring the UV lamps was a modified Aquaray H 2 O 20 reactor with four medium pressure UV lamps. The UV lamps can be operated up to 4KW power and they have a 50 cm arc length. The UV lamp is designed so both electrical connections are on one side so they are accessed and maintained from one side of the reactor. The Aquaray H 2 O uses two state of the art electronic ballasts to power the UV lamps. The ballasts operate from 480 V, 3 phase, 50/60 Hz power and are designed to operate medium pressure discharge lamps. Each ballast has outputs for two UV lamps. In contrast to conventional ballasts (inductive lamp ballast or transformer with transductor), the lamp with an electronic ballast is operated at high frequency (approx. 250 khz). The lamp does not flicker and dimming is infinitely adjustable to a range between 20% and 100% of the electric power. Each UV lamp in the reactor is monitored by a UV sensor. Due to the modifications

3 required for the lamp measurement, special sensor ports were used in order to bring the sensor port closer to the UV lamp reducing the water gap to 3.8 cm (Figure 2). Figure 1: Absorbance spectrum of post-filter AWRA water (October, 2002) Absorbance (cm -1 ) Figure 2: Schematic of the sensor port arrangement. Radiometer/sensor in sensor port Sensor port Reactor wall Quartz sleeve UV lamp Test Equipment: Transmittance spectra of optical components used in the tests (filters and screen diffusers) were measured in the laboratory at Duke University on a Cary Bio-100 double-beam scanning spectrometer (Varian, Inc.). The same instrument was used to record the absorbance spectra of water from the reactor. An IL1700 radiometer system was used (International Light, Inc., Newburyport, MA). The detector was an SED240 equipped with a W-diffuser and a QNDS1 filter. The detector response function as determined by the manufacturer during its most recent calibration (July, 2002) is shown in Figure 3 along with the transmittance spectrum of the QNDS1 filter.

4 Figure 3: Transmittance spectrum of the QNDS1 filter and SED240 detector response function including W-diffuser %T SED 240 Response (ma cm 2 mw -1 ) A hand-held fiber-optic spectrometer (S2000, Ocean Optics, Inc., Dunedin, FL) was used without the fiber-optic cable to obtain lamp emission spectra. The spectral response of the spectrometer was calibrated in the laboratory at Duke against a NIST traceable D 2 lamp (Thermo Oriel, Stratford, CT). The spectrometer was used in scope mode, which directly records counts from the diode array detector. Counts were converted into spectral irradiance (mw cm -2 nm -1 ) using the conversion factors determined with the calibrated D 2 lamp. The radiation intensity was too high to connect the spectrometer directly to the sensor port, so an adapter was made that housed two diffuser screens with meshes crossed at 45 o. While the individual screens have an essentially flat spectral response, the crossed screens show a slight wavelength dependence of the transmittance. The transmittance spectrum of the crossed screens is shown in Figure 4. Figure 4: Transmittance spectrum of the crossed screens in the spectrometer adapter %T

5 Testing Procedure: Each test followed a standard procedure beginning with turning off the reactor and draining the water. The reactor was then opened, and the quartz sleeves were cleaned with a citric acid solution followed by a water rinse. Sleeves and lamps were visually inspected for wear while the reactor was open. In the case of lamp 2, the mounts were positioned such that the return wire for the electrodes passed in front of the sensor window during normal operation so the lamp was rotated 180 o during tests to prevent this from interfering with the measurements. The regular sensor ports were then replaced with sensor ports identical to the ones used in daily operation but that were used only during tests to prevent them from solarizing or otherwise aging. At this point, the reactor was sealed and filled; tests were conducted at a flow rate of approximately 100 gal/min, which was also the regular operating flow rate. Each lamp was tested individually at 4 kw (full power) while the other lamps remained off. Several measurements were made for each lamp. The radiation intensity was quantified three different ways: by radiometer, by standard sensor included with the reactor, and by DVGW-type sensor. Other data collected include the emission spectra of the lamps and their operating current. The germicidal and 200 to 350 nm fluence rates were determined by combining radiometric and spectrometric measurements, and the method is described in detail below. Briefly, the radiometer detector was placed in the reactor sensor port(s) to measure the apparent fluence rate; the term apparent is used because the value measured by the radiometer is a function of the water transmittance, the optics and sensitivity of the detector, and the spectral distribution of the radiation. Corrections for these parameters were made mathematically as described below. The calibration plane of the detector defined the plane in which the fluence rate was measured. To ensure reproducibility of the detector placement from test-to-test, a PVC mount was constructed that could be screwed into the sensor ports of the reactor (Figure 5). A similar arrangement was adopted for mounting the spectrometer used to obtain emission spectra of the lamps (uncorrected for water transmittance). Figure 5: Arrangement of the radiometer detector during tests. Reactor Sensor Port Calibration Plane W-Diffuser Detector Mount SED240 Detector QNDS1 Filter Irradiance Calculation: The basic method for calculating the fluence rate using a radiometer when the lamp emission spectrum is known has been described previously (Linden and Darby, 1997). The method involves calculating the radiometer sensitivity factor from equation 1.

6 SF 1 = RRF(λ) P(λ) (1) λ Here, SF is the sensor factor, λ represents the wavelength, RRF is the relative response factor of the radiometer (relative to the value used on the IL1700 during measurement to convert current to irradiance), and P is the lamp probability function (i.e., the normalized emission spectrum over the range of wavelengths to which the radiometer detector responds). The fluence rate at the detector is obtained by multiplying this factor by the irradiance read off the radiometer. For these tests, additional calculations were necessary, because of the use of the QNDS1 filter, the screens in the spectrometer adapter, and alteration of the observed lamp spectra due to absorption of radiation in the water. To correct for the use of the QNDS1, the SF is altered according equation SF = (2) RRF(λ) TQNDS (λ) P(λ) λ Here, T QNDS (λ) is the percent transmittance of the QNDS1 filter at each wavelength. To correct for the screens in the spectrometer adapter, P(λ) is altered according to equation 3. P(λ) TAS(λ) P' (λ) = (3) P(λ) λ TAS(λ) Here, T AS (λ) is the percent transmittance of the adapter screens. Note that this equation defines normalized values of P (λ). The spectral distribution of the fluence rate uncorrected for the water absorbance was then calculated from equation 4. E (λ) = P' (λ) MR SF (4) Here, MR is the radiometer reading during the test. Finally, correction for the water absorbance was made using equations 5 and 6 (Morowitz, 1950). E(λ) E' (λ) = (5) S 3.8a(λ) (1 10 ) S = ln(10) 3.8 a(λ) Here, E (λ) is the fluence rate at the radiometer detector calibration plane in the absence of absorption by the water, and a(λ) is the absorbance coefficient of the water (cm -1 ). The fluence rate from 200 to 350 nm was calculated as the sum of the E (λ) values. Germicidal fluence rate, E G, was calculated by weighting the E (λ) values by the relative absorbance spectrum of DNA (with 254 nm given a weighting factor of unity) and summing according to equation 7. E G λ (6) = E'(λ)A' (λ) (5) DNA

7 In this equation, A DNA (λ) are the absorption coefficients of DNA relative to 254 nm. The DNA spectrum used in these calculations is shown in Figure 6; the spectrum from 200 to 300 nm was taken from Jagger, 1967 and above 300 nm it was fit with a gaussian curve. Figure 6: Germicidal weighting curve [from Jagger 1967] used to calculate E G. This is the DNA absorption spectrum relative to 254 nm Germicidal Weighting Factor RESULTS AND DISCUSSION Figure 7 shows how the observed spectrum of lamp 1 changed during the first 3980 h of operation. The spectra shown in the figure are corrected for all the effects described in the preceeding section and represent the spectral distribution of the fluence rate at the detector calibration plane (i.e, they are E (λ) values, the irradiance expected if the reactor were filled with a non-uv absorbing medium). There is a distinct drop in the fluence rate at all wavelengths that is most pronounced below 240 nm. Similar changes were also observed for the other three lamps. It is not possible from the present data to determine whether these changes are due solely to lamp aging or to a combination of aging and solarization of the quartz sleeves. For the purposes of these tests, such a distinction is not necessary. However, when the tests are completed the sleeve transmittance will be measured and compared to that of new sleeves to assess what fraction of the observed depreciation can be attributed to sleeve aging and what to lamp aging. Of particular interest in these tests are the rates at which germicidal, total (200 to 350 nm), UVB (280 to 315 nm), and UVC (200 to 280 nm for these lamps) fluence rates change in the reactor during full-time operation of the system. The change in UVA fluence rate cannot be adequately assessed from the data due to the limited response range of the SED240 detector (Figure 3). For lamp 1, the change in germicidal and total fluence rates is shown in Figure 8. From the figure, it is clear that the germicidal fluence rate depreciates more rapidly than the 200 to 350 nm fluence rate, consistent with the observed spectral changes (Figure 7).

8 Figure 7: Spectral changes for lamp 1 over the first 3980 hours of operation hours 3980 hours 1.4 Irradiance (mw cm -2 ) Figure 8: Change in germicidal and total fluence rate with time for lamp Total Germicidal 0.95 Relative Irradiance Hours The relative change in germicidal versus total irradiance is reflected in a more rapid decrease of the UVC fluence rate compared to the UVB fluence rate, which is shown in Figure 9. These changes are similar to previously reported trends (Schenck, 1987), but the depreciation rates observed here are much slower than expected based on the results discussed by Schenck. The reason for this discrepancy is not presently known, but the trends observed here are consistent with the manufacturer s life expectancy of 6000 h (or more) for these lamps to operate effectively for UV disinfection.

9 Figure 9: Change in UVB and UVC fluence rate with time for lamp Relative Irradiance UVC UVB Hours CONCLUSION Because few data are available in the published literature, the aging profile of UV lamps is not well understood. This ongoing study provides information on the manner in which MP UV lamps age with regard to their total output as well as changes in their spectral distribution. Over 4000 hours of investigation, the MP lamps tested decreased in irradiance by approximately 17%. Based on one previous report (Schenck, 1987), these lamps appear to be aging fairly slowly. Slightly more depreciation has been observed in the UVC than in the UVB region (18% versus 15% for the average of all lamps). Consistent with this, the germicidal fluence rate depreciated slightly more rapidly than the total (200 to 350 nm) fluence rate. In general, shorter wavelengths tended to age more rapidly, which may be due to a combination of lamp and sleeve aging. LITERATURE CITED AWWARF (American Water Works Research Foundation) (2000) Ultraviolet Disinfection Guidelines for Drinking Water and Water Reuse, Project Report #2674; American Water Works Association, Denver, CO. Linden, K.G.; Darby, J.L. (1997) J. Environ. Eng.-ASCE, 123(11): Morowitz, H. J. (1950) Science, 111: Jagger, J. H. (1967) Introduction to Research in UV Photobiology; Prentice-Hall, Inc., Englewood Cliffs, N.J. Schenck, G.O. (1987) "Ultraviolet Sterilization" in Handbook of Water Purification, 2nd ed.; Lorch, W. (Ed.); Wiley and Sons, NY.