JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 419 High Current Pulsing of a Xenon Arc Lamp for Electrothermal Atomic Absorption Spectrometry Using a Linear Photodiode Array* Clare M. M. Smith and James M. Harnlyt USDA ARS BHNRC NCL Bldg. 161 BARC-East Beltsville MD 20705 USA Gary P. Moulton* and Thomas C. O’Haver Department of Chemistry University of Maryland College Park MD 20743 USA A series of 300 and 500 W xenon arc lamps normally operated at 20 and 35 A respectively have been pulsed as high as 300 A to achieve higher intensities in combination with a linear photodiode array detector. Initial tests without pulsing showed that the 500 W lamps are generally more intense but the 300 W lamps were more intense at 200 nm. With 300 A pulses both lamps showed a factor of 500 increase in the pulse intensity over the simmer intensity.With a 0.5 ms pulse and a 3.75% duty cycle pulsing at 300 A provided a factor of 18 increase in the integrated intensity over normal d.c. operation. The increase in integrated intensity can result in a comparable improvement in detection limits since the instrument is detector noise limited. Both the 300 and 500 W lamps exhibited failure after the equivalent of 200 atomizations at 200 A. With 100 A pulses the 300 W lamp was still operating after the equivalent of 800 atomizations. At both pulse levels the decrease in intensity with time was accelerated as compared with d.c. operation. It was concluded that an improved lamp design is necessary to make pulsed operation economically attractive.Keywords Continuum source atomic absorption spectrometry; linear photodiode array detection; xenon arc lamp; pulsing. In recent literature continuum source atomic absorption spec- trometry (CSAAS) has been described where a linear photodi- ode array (LPDA) detector’-’ is employed with pulsing of the xenon arc continuum source.1’2 The most significant results to date have arisen from the use of the LPDA.3*5 With d.c. operation of the continuum source and LPDA detection detection limits have been achieved which are comparable to conventional line source atomic absorption spectrometry (AAS) even in the far ultraviolet (UV) region for arsenic (193.7 nm) and selenium (196.0 nm). These improved detection limits are a result of the greater quantum efficiency and the multiplex advantage of the LPDA as compared with detection using a photomultiplier tube (PMT). Initial results suggested that further improvements in the detection limits can be achieved by pulsing the xenon arc continuum source.’?’ Improvements in the detection limits should be proportional to the increase in intensity since the limiting noise is the LPDA read noise.The use of an LPDA detector makes pulsing of the xenon arc source feasible. With a continuum source the optimum signal-to-noise ratio is obtained by ratioing intensities on and off the analytical line. With PMT detection this is accomplished by mechanical wavelength modulation using a quartz refractor plate on a galvanometer.6 The fastest wave- length modulation frequencies (200 Hz) are not compatible with a lamp pulse of 1 ms or less.In the research instrument used in this study,’-’ a short LPDA (256 pixels) is mounted in the focal plane of a high-resolution echelle spectrometer and dedicated to the measurement of the spectral intensities around a single absorption line typically a spectral range of less than 1 nm. At periodic intervals (> 50 Hz) the LPDA is read (the accumulated charge of each pixel is read giving an intensity for each pixel) and an absorbance is computed by ratioing intensities on and off the analytical line.3 Pulses of the lamp can be readily inserted between the LPDA reads. Because intensities from a single read are used to compute absorbance variations between lamp pulse intensities have no effect on the * Presented at the XXVIII Colloquium Spectroscopicum Inter- nationale (CSI) Post-Symposium on Graphite Atomizer Techniques in Analytical Spectroscopy Durham UK July 4-7 1993.t To whom correspondence should be addressed. 1 Present address Abbot Laboratories Chicago IL USA. variance of the signals. Since the wavelength region is moni- tored simultaneously the lamp pulse widths can be extremely short i.e. fractions of a millisecond. Pulsing is most efficiently accomplished using two power supplies connected in parallel to the lamp. The d.c. or simmer supply furnishes the simmer current while the pulse supply furnishes short high-current pulses and operates with a low (< 5%) duty cycle. The advantage of pulsing over d.c. operation is manifested in two ways. Firstly the maximum power level of the lamp can be exceeded during pulsing using the ‘burst’ mode of operation.Secondly the emitted intensity of the lamp in the UV increases exponentially with the pulse current. The ‘burst’ mode of operation is defined as the use of a short period of pulse-mode operation followed by a lengthy period of d.c. operation during which the lamp is allowed to ~ 0 0 1 . ~ Electrothermal atomization is ideally suited to the burst mode of operation. During this short burst or period of pulsed operation the power level of the lamp can be exceeded as pulsed operation need only last for the duration of the atomiz- ation cycle 1-5 s depending on the element being determined. Between atomizations 1-3 min of simmer are possible while the furnace cools the next sample is deposited dried and charred.With a 1 min simmer period and a 5 s burst the effective power rating of the lamp can almost be d o ~ b l e d . ~ ICL Technology the manufacturers of the Cermax xenon arc lamps (300 and 500 W) used in this study has shown that the relationship between the lamp intensity and the current is given by Ip/Is = (i,/i,)”.” (1) where I is the pulse intensity I is the simmer intensity i is the pulse current and is is the simmer current.8 It can be seen that with respect to the emitted intensity there is an exponential advantage to directing the lamp’s power into a high-current pulse rather than into d.c. (constant current) operation. For a fair comparison of d.c. and pulsed operation however the gain in intensity must be considered over the entire pulse cycle (pulse plus simmer).Thus the effective increase in integrated intensity viewed by the LPDA over the pulse cycle will depend on the pulse current and width and the duty cycle. These three factors are interactive and are limited by the effective power rating of the lamp. All the power cannot be directed into the pulse since the lamp cannot be420 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 turned off between pulses. A minimum simmer current of 14 A is required to prevent damage to the cathode from arc wander. High-current pulsing of the lamp is deleterious to lifetime of the lamp life and to inten~ity.~ The extent of the effect is dependent primarily on the pulse current. Repeated pulsing of a 300 W xenon arc lamp to 100 A for 0.5 ms produced a 65% decrease in intensity after 20 000 These data were obtained with a 0.4% duty cycle one pulse every 2min.Normal operation of the lamp is of course accompanied by a decrease in intensity. A loss of 35% in intensity is expected for 1000 h of ~peration.~ The many possible variations of the pulsing operation make it difficult to compare the number of lamp pulses to hours of lamp lifetime. In terms of the number of analytical experiments for which the lamp can be used however pulsing can be expected to have a deleterious effect. Only limited pulsing data have been reported previously. ‘y2 Initial pulsing results suffered from the lack of a separate power supply’ and true pulsing *as not possible. Instead a square-wave modulated peak was superimposed on the simmer intensity by supplying a reference frequency to the power supply.In this manner the lamp current was modulated between 20 A and 30-40 A at 60 Hz with a 50% duty cycle resulting in a doubling of the integrated intensity. The upper current limit was determined by the limit of the power supply. In the second study only preliminary data were available for the custom built power supply.2 In this study the intensities of a series of 300 and 500 W xenon arc lamps were compared for dx. and pulsed operation. Peak and integrated intensities were measured as a function of the pulse current. The detection limits obtainable with lamp pulsing were projected for arsenic as a function of the pulse intensity. Integrated intensities were measured as a function of the total number of pulses.The lifetimes of the lamps pulsed at 100 and 200A were compared with normal d.c. operation in terms of intensity and the length of lamp lifetime. Experimental Instrumentation The laboratory constructed CSAAS has been previously de~cribed.3.~ A schematic diagram of the system used is given in Fig. 1. Either a 300 or 500 W Cermax lamp (Models LX300-UV and LX5OO-UV respectively ILC Technology Sunnyvale CA USA) was used. A pulse power supply was custom built for this project by the Electronic Development Group of the Department of Physics of the University of Maryland. This power supply was designed to deliver 200 A for 1.5 ms across 8 i2 with a 20% duty cycle but was found to be capable of delivering 280 A under these conditions. The current level and peak width of the pulse power supply were set manually.The start of each pulse was initiated by a digital signal from the computer. The pulse power supply was connec- ted to the lamps in parallel with the 300 and 500 W Cermax power supplies (Models PS300-1 and PS5OOSW-1 respectively ILC Technology). The 300 W supply was modified by placing a blocking diode in the cathode line to prevent the high current pulse from causing damage. A blocking diode was already in place in the 500 W power supply. During lamp ignition the pulse supply was disconnected from the lamp by means of a mechanical switch to prevent the r.f. ignition pulse from damaging the pulse power supply. An electronic shutter and driver/timer [Uniblitz Models LS6Z (normally closed) and T132 respectively Vincent Associates Rochester NY USA] was used to gate the exposure of the LPDA.For these studies a commercial furnace and power supply (HGA-500 Perkin- Elmer Norwalk CT USA) were used. Lamp Pulsing The 300 W lamps were normally operated at 20 A in the d.c. mode while the 500 W lamps were normally operated at 35 A. With pulsing both lamps were operated at a 20A simmer current. Current pulses of 20-300 A were superimposed on the simmer current. Data were acquired using two pulsing regimes. tn both cases the lamps were operated in the pulsed mode for 5 s with 120 s of simmer between bursts a timing scheme compatible with a normal electrothermal atomization (ETA) cycle. In the first pulsing regime a 3.75% duty cycle was used with a 0.5 ms wide pulse at 75 Hz (13.3 ms cycle).The majority of the results were obtained with the second pulsing regime of a 5% duty cycle with a 1 ms pulse width at 50 Hz (20 ms cycle time). In both cases with LPDA detection a 500 x 500 pm entrance slit was used. Intensity Measurements D.c. operation A series of 300 and 500 W lamps were compared in the d.c. mode of operation at eight different wavelengths between 201.1 and 402.2nm. These measurements were made at the centre of each order using PMT detection. The signal from the PMT was connected directly to an oscilloscope. Measurements were made with matching slit settings of 50 pm wide and 500 pm high and a PMT setting of 550 V. The 300 W lamps were operated at 20 A while the 500 W lamps were operated at 35 A. At all wavelengths intensity measurements were made on and between orders.The between-order measurements provided far stray light intensity which was used to correct the on-order intensities. The light measured between the orders of the echelle spectrometer is reflected light from the visible region and is a measure of the unabsorbable light on the order. The stray light was as large as 40% at 201.1 nm. Pulsing operation One set of experiments employed the PMT detector for pulsed operation. In this experiment the LPDA was removed from the spectrometer but was connected to the computer in the normal manner except for the ‘data out’ line. The PMT was installed in its usual position in the spectrometer and the analogue data line was hooked to the computer in place of the LPDA ‘data out’ line.Timing of the ‘read trigger’ was shifted so that the reading of 256 pixels corresponded to 256 analogue PMT measurements that started with the start of the lamp pulse. In this manner the intensity of the lamp and the following simmer intensity levels were measured at 30 ps intervals. The PMT measurements were made at 240.7 nm the cobalt wavelength. Intensity measurements made with the LPDA were made on and off the order to correct for far stray light. Measurements were also made with and without gating. For studies of the lamp intensity as a function of pulse current d.c. lamp measure- ments on and off the order were made between sets of pulsed lamp measurements. For lamp lifetime studies measurements were made continuously on-order with off-order measure- ments being made every 5000 pulses. Measurements were made in the d.c.mode at the start and end of each day. Between 10 000 and 15 000 pulses were made per day. Results and Discussion Comparison of Lamp Intensities for D.c. Operation The 300 W xenon arc Cermax lamps have been used almost exclusively for CSAAS. In this study both the 300 and 500 W lamp were used. The higher power of the 500 W lamp was attractive since higher pulse currents could be used and a longer lifetime was anticipated.” The 500 W lamps however are more expensive than the lower power lamps ($800 uersus $500). Initially a series of new 300 and 500 W lamps were compared for operation in the d.c. mode (see Experimental) at eightJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Simmer power supply 42 1 -- - _ _ _ _ _ _ _ _ _ _ _ - - - - ‘2 D Pulse power A supply I -L Video signal L\ I Trigger ADC I Master start Lamp pulse trigger Shutter trigger Clock DAC Computer Experiment start Int Fig.1 Schematic diagram of the CSAAS system wavelengths between 200 and 400 nm at operating currents of 20 and 35 A respectively. The intensities in Table 1 reflect the change in lamp intensity as well as the Cchelle spectrometer transmission efficiency as a function of wavelength. Zander et d.ll have reported almost a factor of ten decrease in the transmission efficiency of the Cchelle between 250 and 200 nm. Thus the majority of the loss below 250 nm can be attributed to the spectrometer. It can be seen that on average the 500 W lamps are more intense at most wavelengths.At 215 nm however the sources are comparable and at 201.1 nm the 300 W lamps are actually more intense. It is generally expected that the intensity ratio of the lamps is roughly proportional to the power ratio. The greater intensity of the 300 W lamp at 200 nm however is not all that surprising. Although the larger lamps generate more power the arc gap is larger and the ‘brightness’ (lumens per steradian) may not be as high as it is for lower power lamps. This ‘brightness’ is the most critical parameter of the lamp since the effective lamp intensity is based on the radiation that can be focused through a 500 x 500 pm entrance aperture. Intensity Versus Pulse Amplitude The intensity increases obtained by pulsing the lamps are shown in Figs. 2 and 3.A new 300 and a 500 W lamp (both number 1 lamps from Table 1) were pulsed at 75 Hz with a 0.5ms pulse width (3.75% duty cycle) for a series of pulse currents up to 300 A. The simmer current for both lamps was 20 A and each lamp was operated in the pulse mode for 5 s. All measurements were made at 240 nm. The current-voltage characteristics of the 300 and 500 W lamps are almost the same. The maximum current levels for both lamps (20 and 35 A respectively) assumes a 14 V drop across the electrodes. The power levels shown in Table2 were computed from current-voltage data from Chinnock’ for the 300 W lamp. In Fig. 2 the predicted and experimentally determined peak- to-simmer intensity ratios of the 500 W lamp are compared. Experimental results were obtained using PMT measurements (see Experimental) and the predicted values were computed from the pulse and simmer currents using eqn.(1). Fig. 2 shows Table 1 Lamp intensities (in arbitrary units) as a function of wavelength Wavelength/ nm 201.1 214.5 225.2 250.2 274.7 300.3 352.0 402.2 500 W lamps* 300 W lamps? 1 0.066 0.15 0.38 1.1 2.7 6.2 17.0 22.5 2 0.062 0.13 0.44 1.4 3.2 7.2 18.5 22.5 3 0.060 0.12 0.34 1.1 3.0 8.4 20.0 23.0 1 0.075 0.16 0.30 0.94 2.3 5.4 15.0 21.5 2 0.062 0.13 0.25 0.75 2.0 5.5 15.8 28.0 3 0.070 0.15 0.28 0.8 1 2.0 5.0 14.0 21.0 r 600 2 500 400 4d .- cn .- L E .- 300 c .- v1 5 200 r .- Q) 7 Q 2 100 0 Pulse current/A Fig.2 Peak intensity ratios as a function of pulse current for A 500 W lamp with PMT detection; and B predicted ratios from eqn. (1) * Operated at 35 A.t Operated at 20 A.422 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 I 1 150 i- PC Pulse current/A Fig.3 Integrated intensity ratios as a function of pulse current for A a 300 W lamp and LPDA detection; B a 500 W lamp with LPDA detection; and C predicted ratios Table 2 Lamp operating power Power/W Simmer Pulse current/A current/A 20 0 50 - 100 150 - 200 - 250 - 300 - - 0.5 ms* 75 Hz 280 328 378 434 495 558 619 1.0 ms* 50 Hz 280 343 41 1 48 5 567 651 732 * For 0.5 and 1.0 ms pulse widths the duty cycles were 3.75 and 5% respectively. The total cycle times were 13.3 and 20 ms respectively. that the predicted and experimental intensity ratios compare very well. Substituting the experimental data yields an average exponent of 2.2.Thus the intensity enhancement predicted by eqn. ( 1 ) was verified. The pulse power supply was only specified and calibrated for pulse currents as high as 200A. The agreement between the predicted values and the experimental results in Fig. 2 suggests that under the specific pulse parameters of the experi- ment the pulse power supply is still calibrated and capable of delivering current at 300 A. In Fig. 3 the ratios of the integrated intensities for pulsed and d.c. operation for the 300 W and 500 W lamp and the predicted values using an LPDA detector are compared. Pulsing conditions were identical to those for Fig. 2 except that the pulse currents were raised to 300 A in a different series of steps. Integration of the signal was accomplished with an LPDA (see Experimental). The predicted values were com- puted from eqn.(1) based on the simmer and pulse current and the duty cycle. It can be seen that the effective intensity increase at pulse currents below 200A is almost identical for both lamps and agrees well with the predicted values. With a 200A pulse the increase in integrated intensity is a factor of 70 and 64 respectively for the 500 and 300 W lamps as compared with the predicted ratio of 70. At pulse currents above 200 A both lamps show a reduced response. It was noticed when using the 300 W power supply that the simmer current did not return to its original level after a pulse but fell to a lower level. This effect was accompanied by a visible drop in the current meter of the simmer supply. The result was a decrease in the integrated intensities.This most probably accounted for the poor performance of the 300 W lamp above 200A. A subsequent study showed that this decrease was produced by the blocking diode which had been inserted in the line to the 300 W supply and was not owing to the lamp. No such effect was observed with the 500 W lamp and power supply or when running the 300 W lamp on the 500 W supply. The reduced response of the 500 W lamp above 200A could be due to the power limit of the pulse power All the results presented to this point were obtained without gating the LPDA exposure with the electronic shutter. Gating has been shown to be essential to accurate background correc- tion and to eliminate flicker noise when the array is read in a sequential manner.3 In addition gating minimizes the contri- bution of the simmer intensity and provides better signal-to- noise ratios when measuring integrated intensity at low pulse currents. Consequently the rest of the data were obtained with gated exposure of the LPDA.supply Detection Limits Versus Pulse Amplitude Integrated intensity measurements as a function of the pulse amplitude were repeated at the arsenic wavelength of 193.7 nm using a 300 W lamp a 50 Hz pulse frequency a 1 ms pulse width (5% duty cycle) and gating which permitted trans- mission of approximately 18% of the total integrated intensity. Thus the gate was open for approximately 3.6 ms of the 20 ms cycle time. At 193.7 nm it was necessary to correct for the far stray light of the 6chelle. This was achieved by making intensity measurements on and off the order.The increase in the integrated intensity produces a pro- portional decrease in the absorbance noise with the LPDA d e t e ~ t o r . ~ This is a result of the detector noise being limiting. The decrease of the experimentally measured absorbance noise with increasing pulse current and increasing integrated inten- sity is clearly shown in Table 3. As a consequence of the reduced absorbance noise the detection limits will improve proportionally since the characteristic mass is unaffected by intensity and remains constant. The predicted detection limits for arsenic as a function of pulse current are presented in Table 3. The detection limit with no pulsing was experimentally determined.5 Gating of the LPDA exposure decreased the integrated intensity by a factor of five and caused a decrease in the detection limits by a comparable factor.With increasing pulse current the predicted detection limit decreased as expected. The improvement in the detection limit with a 300 A pulse was a factor of 20 compared with d.c. operation with 18% gating. The use of an 18% gate is extreme. In general a 75% gate is used in routine operation; the shutter is closed for approxi- mately 5 ms the time necessary to read 64 pixels at a 26 kHz pixel read frequency. A 75% gate produces a detection limit of 33 pg for arsenic and pulsing of the lamp to 300 A results in a detection limit of approximately 0.3. A faster analogue- to-digital converter can be used to increase the gate trans- mission further and reduce the detection limit. Ideally a solid- state detector capable of simultaneously shifting all pixels into Table 3 Detection limits for arsenic as a function of pulse current Detection limit/pg Pulse current/A 0 0 40 80 120 160 200 240 270 300 Gated* No Yes Yes Yes Yes Yes Yes Yes Yes Yes Noise? 0.01 30 0.0662 0.0120 0.0095 0.0032 0.0021 0.001 1 0.0009 0.0008 0.0007 18% Gatet 25 125 22 18 6 4 2 2 2 1.3 75% Gate$ 25 33 6 5 2 1 0.5 0.4 0.3 0.3 * 18Y0 Transmission gate.t Experimentally determined. $ Computed from duty cycle.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 423 read registers could be used which would result in no deterior- ation of the detection limit and would still provide the accurate background correction. The use of a 300 A pulse on a routine basis is not practical for a 300 W lamp.At 300 A the lamp is operated at a power level of 732 W. This is 20% higher than the factor of two increase that is tolerable with burst-mode operation7 The factor of two is obviously conservative or else lamp failure would have occurred. The 300 A pulse current is still below the factor of two for the 500 W lamp. Realistically however pulse currents in the 100-200A range should prove more useful. Pulsing in this range would result in improvement of the detection limit of arsenic from 25 to between 20 and 0.5 pg (Table 3). Lamp Lifetime It is clear that there is a substantial detection limit advantage to be obtained by operating the xenon arc lamp at high pulse currents. However there is a major trade-off involved in terms of lamp lifetime. The detection limit values given in Table 3 were obtained with a new lamp and the measurements were made within the first few hours of operation.With continual pulsing at a level which exceeds the power rating of the lamp a significant reduction in lamp lifetime can be expected. The longevity of a new 500 W lamp pulsed to 200 A at 50Hz with a 5% duty cycle and a simmer current of 20A was studied. Intensity measurements were again made at 193.7 nm and were corrected for far stray light (see Experimental). The plot of relative output versus number of pulses is shown in Fig. 4. It can be seen that initially the integrated intensity decreased extremely rapidly with the number of pulses. After 5000 pulses the integrated intensity decreased by approximately 95% of its original value.Between 5000 and 50000 pulses the pulsed intensity of the lamp remained fairly constant at 5% of the original value. During this interval the variation between pulses was approximately 2%. Although the integrated intensity with the 200A pulse decreased dramatically the ratio of the pulsed to d.c. operation was still approximately the same; that is pulsing is still providing the same improvement in lamp intensity as com- pared with d.c. operation as it did when the lamp was new. Thus the d.c. lamp intensity had also decreased by 95% of its original value. The 500 W lamp was still operational after 75000 pulses. After 50 000 pulses however the integrated intensity became extremely erratic. The LPDA was successful in removing flicker noise from the absorbance calculation but the variation between pulses was extreme.Integrated intensity uersus the number of 200 A pulses for a loo 6 1 b 20dOO 4odoo 60d00 80d00 / d o 0 0 No. of pulses Fig. 4 Relative intensity as a function of number of pulses for A 300 W lamp with 200 A pulses; B 300 W lamp with I00 A pulses; and C 500 W lamp with 200 A pulses new 300 W lamp is also shown in Fig. 4. This study employed conditions identical with those for the 500 W lamp. Initially the behaviour of the 300 W lamp was very similar to that of the 500 W lamp i.e. a rapid decrease in the integrated intensity was observed in the first 6000 pulses followed by stable operation for the next 50 000 pulses. Variation between pulses was again around 2%.The integrated intensity of the 300 W lamp decreased by only 90% of its original intensity compared with 95% for the 500 W lamp. After approximately 60000 pulses the lamp failed to ignite. This was preceded by frequent incidents when the lamp extinguished in the course of an experiment. Finally a new 300 W lamp was pulsed under conditions identical with those used previously using a pulse current of 100 A. The same dramatic decrease (to approximately 10%) in the integrated intensity is observed in the first 6000 pulses. As with the 300 W lamp with a 200 A pulse very stable operation was observed beyond 6000 pulses with little variation between pulses. At 100 A however the lamp is still operational after 240000 pulses with a variation between pulses of less than 2%.A 500 W lamp was not tested with 100 A pulses. It is reasonable to assume however that the 500 W lamp would exhibit the same extended lifetime (as compared with 200A pulsing) as the 300 W lamp. During the period when the output of the lamps was relatively stable (from around 20000 to 50000 pulses) the integrated intensities of the 300 and 500 W lamps operated at 200 A were approximately equal to the 300 W lamps operated at 100 A delivering approximately one third that intensity. The 90% loss in intensity observed with a 100 A pulsing of the 300 W lamp is considerably greater than that reported by the manufacturer who reports a 35% loss in intensity after 20 000 pulses. In this study however longer pulses ( 1 ms versus 0.5 ms) and a higher duty cycle (5% versus 0.4% or less) were used.It must be stressed that the original reference intensities in this study were determined as the average intensity for the first 360 pulses. A delay of several thousand pulses in determin- ing the reference intensity would make a considerable difference in the shape of the curves in Fig. 4. Pulsing in Perspective It is not possible to consider the merits of pulsing of the Cermax lamps without considering the behaviour of the lamps in the d.c. mode of operation. In Fig. 5 integrated intensity is plotted as a function of d.c. hours of operation for five 300 W lamps one 500 W lamp (none of which were pulsed) and the manufacturer’s predicted behaviour. The broken lines represent the mean and the extremes of lamp intensity specified by the manufacturer.The termination of an experimental lamp trace 120 1 0 ’ I 1 1 10 100 Period of operatiodh Fig.5 Corrected intensity as a function of hours of operation for 300 W (A-E) and 500 W (F) lamps operated in d.c. mode. The dotted lines represent the mean and extremes of lamp intensity specified by the manufacturer424 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 in Fig. 5 does not mean that the lamp failed it represents the point at which the lamp was removed from the study. It can be seen that the experimental plots of the lamps all fall close to or below the lower extreme predicted by the manufacturer. After 10-40 h of operation the intensity of three of the lamps had fallen to 50% of their original intensity. The 500 W lamp lost intensity at a faster rate than the 300 W lamps but it is difficult to reach any conclusions with such limited data.During the course of uninterrupted lamp oper- ation there was some decrease in intensity depending on the length of operation. The largest drops in intensity were observed between the time the lamps were shut off for the night and re-ignited the next day. Attempts to re-optimize the position of the lamp to achieve the intensity of the previous day were always futile. A main source of error in comparing the experimental data and the manufacturers data is the determination of the initial intensity. In this study the initial intensity of the lamps were made within minutes of the first ignition. As mentioned pre- viously a delay in acquiring the intensity to be used as the reference would considerably change the shape of the curves in Fig.5. The observation that most of the lamps tested in this study fall close to the lower limit suggests that the manufac- turer's reference point represents the intensity averaged over a longer period of time than employed in this study. Using arsenic as a reference element detection limits obtain- able at various stages in the lifetime of the 300 W lamp in d.c. and pulsed mode (using 100 and 200 A pulses) were computed (Table 4). For these computations it was assumed that pre- atomization and atomization cycle data acquisition required 5 s and the entire ETA cycle required 3 min. It was further assumed that a 75% gate was used and the pulsing frequency was 60 Hz thus 300 pulses were required per atomization. The detection limits for the pulsing of the 300 W lamp are based on the experimental results in this study.The detection limits for d.c. operation were based on the manufacturers predicted behaviour and the data in Fig. 5. As expected pulsing at either current and with either a 18 or 75% gate offers significant improvement in the detection limit for the first 20 atomizations (the first hour of operation or 6000 pulses). With the 18% gate after the first 20 atomiza- tions d.c. operation and pulsing at 200A are comparable. After 200 atomizations at 200 A the 300 W lamp fails and d.c. operation offers detection limits a factor of two better than those achieved with pulsing at 100A with an 18% gating. After 1000 atomizations d.c.operation and pulsing the lamp at 100A with 75% gating provides the same detection limit. This comparability should last until the lamp fails since the intensity of the lamp is stable in both modes of operation. An alternative means of evaluating the lamp lifetime is to consider the number of experiments that can be run with the 300 and 500 W lamps with 100 and 200 A pulses. Both lamps proved to have pulsed lifetimes of 60000 pulses for a 200 A pulse. Based on a 60 Hz pulsing frequency 5 s of data acqui- sition per atomization and 50 atomizations per experiment (five standards in triplicate five samples in triplicate and 20 blanks) one experiment requires 15 000 pulses. This means Table 4 Detection limits (pg) for arsenic as a function of lamp lifetime that with this pulsing regime both Cermax lamps pulsed at 200A could only be used for four experiments.Pulsed at 100 A the 300 W lamp and probably the 500 W lamp could be used for 16 experiments and most probably more. Physical Changes in the Lamps In both d.c. and pulsed operation it appears that physical changes in the position of the cathode relative to the anode dictate the effective lifetime of the lamps. It is well documented that in d.c. operation the intensity of the xenon arc lamps decreases asymptotically to a level that remains steady for the duration of the lifetime of the lamp. This corresponds to an initial melting-back of the pointed tip of the cathode until it reaches a blunt shape that is relatively impervious to further change in shape. Eventually the continued heating of the three struts that support the cathode produces metal fatigue and results in a shift in the position of the cathode such that the arc gap is too large to permit ignition of the lamp. In the d.c.mode this event usually comes after more than 1000 h of operation. When pulsed at 200 A both the 300 and 500 W lamps experience an acceleration of both processes the melting of the tip and the eventual extreme shift of the cathode away from the anode. The melting of the cathode tip increases the arc gap and results in a larger more diffuse arc. The more diffuse arc is less bright in the UV and is less than optimum for focusing through a 500pm square aperture. Consequently a loss in intensity in the UV accompanies the ageing of the lamp and this loss is more pronounced than at longer wavelengths.It appears that pulsing at currents as high as 200A not only accelerates the cathode melting process but exceeds the limits reached with d.c. operation. After 6000 pulses at 200A the d.c. intensities of the 300 and 500 W lamps were far less than predicted for normal deterioration. Examination of the failed 300 W lamp revealed a separation of the cathode structure of the lamp. This was not true of the 500 W lamp supporting the manufacturers statement that the 500 W lamp was more suited for the high current pulses. Conclusions For d.c. operation the 500 W xenon arc lamp is in general a factor of 1.5 more intense except below 210nm where the :300 W lamp is more intense. Pulsing at high currents provides a significant increase in the integrated intensity of the LPDA and can result in a factor of up to 80 improvement in the detection limit for a new lamp.Pulsing at 200A induces premature failure of both the 300 and 500 W lamps after 60 000 pulses. The cost of the lamps make such limited lamp lifetime intolerable. Pulsing at 100 A is less deleterious but also provides less significant improvements in detection limits. After the first .SO h a 300 W lamp pulsed at 100 A is at best comparable to d.c. operation. While the intensity increase obtained with high current pulsing of new lamps will continue to be attractive the practical use is very limited. An improved lamp design is Number of atomizations 1 50 100 150 200 500 1000 Time/h 0.05 2.5 5.0 7.5 10 25 50 Number of pulses 300 15 000 30 000 45 OOO 60 000 150 000 300 000 D.c. operation Pulsed operation ( 18% transmission gate) Pulsed operation (75% transmission gate) ILC This study 25 25 26 26 27 29 28 30 30 32 32 37 33 52 100 A 200 A 9 3 70 22 75 25 75 25 80 30 80 - 80 - 100 A 200 A 3 0.5 22 4.7 25 4.5 25 4 30 5 30 30 - -JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 425 necessary before pulsing of xenon short arc lamps will be feasible for CSAAS. References 1 Moulton G. P. O’Haver T. C. and Harnly J. M. J. Anal. At. Spectrom. 1989 4 673. 2 Moulton G. P. O’Haver T. C. and Harnly J. M. J. Anal. At. Spectrom. 1990 5 145. 3 Harnly J. M. J. Anal. At. Spectrom. 1993 8 317. 4 Smith C. M. M. Nichol R. and Littlejohn D. J . Anal. At,. Spectrom. 1993,8 989. 5 Smith C. M. M. and Harnly J. M. Spectrochim. Acta Part B in the press. 6 Harnly J. M. Anal. Chem. 1986 58 933A. 7 Chinnock R. ICL Engineering Note No. 152 Use of X e Short Arcs as Pulsed Light Source ILC Technology Sunnyvale CA 1982. 8 Naval Research Laboratory Ultraviolet Output from Pulsed Short Arcs Washington D.C. Memorandum Report 2427 April 1971. 9 ILC Technology Sunnyvale CA product literature. 10 ILC Technology Sunnyvale CA personal communication. 11 Zander A. T. Miller M. H. Hendrick M. S. and Eastwood D. Appl. Spectrosc. 1985 39 1. Paper 3/04066F Received July 12 1993 Accepted October 1 I 1993