Determination of an Iron Suspension in Water by Laser-Induced

Sep 1, 1996 - We have applied laser-induced breakdown spectroscopy to quantitative analysis of colloidal and particulate iron in water. A coaxial samp...
2 downloads 9 Views 200KB Size
Download PDF
Anal. Chem. 1996, 68, 2981-2986

Determination of an Iron Suspension in Water by Laser-Induced Breakdown Spectroscopy with Two Sequential Laser Pulses Susumu Nakamura,† Yoshiro Ito,* and Kazuhiro Sone

Department of Mechanical Engineering, Nagaoka University of Technology, 1603-1 Kamitomioka, Nagaoka, Niigata 940-21, Japan Hitoshi Hiraga

Applied Laser Engineering Research Institute, 2085-16 Hukasawa, Nagaoka, Niigata 940-21, Japan Ken-ichi Kaneko

Niigata Thermal Power Station, Tohoku Electric Power Co. Inc., 2-200 Momoyama, Niigata, Niigata 950, Japan

We have applied laser-induced breakdown spectroscopy to quantitative analysis of colloidal and particulate iron in water. A coaxial sample flow apparatus developed in our previous work, which allowed us to control the atmosphere of laser-induced plasma, was used. Using sequential laser pulses from two Q-switched Nd:YAG lasers as excitation sources, the FeO(OH) concentration in the tens of ppb range was determined with an optimum interval between two laser pulses and an optimum delay time of a detector gate from the second pulse. The detection limit of Fe decreased substantially using two sequential laser pulse excitations: the 0.6 ppm limit of single pulse excitation to 16 ppb with sequential pulse excitation. The effects of the second laser pulse on the plasma emission were studied. The concentration of iron in fine particles in boiler water sampled from a commercially operated thermal power plant has been determined successfully by this method. The results show the capability of laser-induced breakdown spectroscopy in determining suspended colloidal and particulate impurities in a simple and quick way. Detection and control of suspending fine particles or colloids in a solution is of practical important in modern manufacturing. Size distribution and number density of suspended particles are determined by optical methods such as scattering. Chemical analysis of such particulate substances, however, is a rather complicated, time-consuming procedure in general: one must dissolve them, adjust their charge states and, after these procedures, determine their quantity by one of the suitable analytical methods. Iron-containing fine particles suspending in water are important in the control of boiler water quality at an electric power plant. Even a small particle of scale in the water can damage turbine blades severely, and therefore, highly cleaned, deionized pure water is supplied for a boiler. Particles of scale from pipes and * E-mail: [email protected]. † Present address: Department of Electrical Engineering, Nagaoka College of Technology, 888 Nishikatakai, Nagaoka, Niigata 940, Japan. S0003-2700(96)00116-3 CCC: $12.00

© 1996 American Chemical Society

valves inevitably contaminate the water, especially when the plant starts up operation. A flushing procedure is, therefore, carried out from section to section of a plant and the water quality is checked at each step. At present, the water quality is monitored by off-line chemical analysis techniques in the plant and total amount of Fe and silica are checked: Fe concentration should be less than 20 ppb and silica less than 5 ppb. The analysis takes some tens of minutes and a huge amount of flushing water is consumed during this period. An on-line measuring technique for total amount of Fe, most of which is in the form of suspended solids and/or colloids, is necessary to shorten the time required to monitor the water condition during the flushing process so that one can economize on flushing water. It is also desirable that the method is able to determine silica simultaneously. In the course of developing an on-line monitoring method for boiler water quality at a thermal power plant, we have applied laser-induced breakdown spectroscopy (LIBS) to quantitative analysis of FeO(OH) in water and, with a coaxial sample flow apparatus, showed that the concentration of the turbid solution of iron colloids in a few ppm range was determinable.1 LIBS and laser-induced plasma spectroscopy have been applied to analytical measurement of gases,2 aerosols,3 and metal ions in solution4 and solids.5-8 Detection limits depended on target elements and ranged from some ppb to some percent. For the determination of a suspension, however, only a qualitative measurement for a turbid solution of CaCO3 has been reported9 so far. It was evident, however, that the test sensitivity should be improved significantly to apply to practical boiler water analysis. (1) Ito, Y.; Ueki, O.; Nakamura, S. Anal. Chim. Acta 1995, 299, 401-405. (2) Cremers, D. A.; Radziemski, L. J. Anal. Chem. 1983, 55, 1252-1256. (3) Radziemski, L. J.; Loree, T. R.; Cremers, D. A.; Hoffman, N. M. Anal. Chem. 1983, 55, 1246-1252. (4) Cremers, D. A.; Radziemski, L. J.; Loree, T. R. Appl. Spectrosc. 1984, 38, 721-729. (5) Kagawa, K.; Yokoi, S. Spectrochim. Acta, Part B 1982, 37, 789-795. (6) Iida, Y. Spectrochim. Acta, Part B 1990, 45, 1353-1367. (7) Kuzuya, M.; Matsumoto, H.; Takeuchi, H.; Mikami, O. J. Spectrosc. Soc. Jpn. 1992, 41, 327-331. (8) Kagawa, K.; Hattori, H.; Ishikane, M.; Ueda, M.; Kurniawan, H. Anal. Chim. Acta 1995, 299, 393-399. (9) Kitamori, T.; Matsui, T.; Sakagami, M.; Sawada, T. Chem. Lett. 1989, 22052208.

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996 2981

Figure 1. Schematic drawings of the experimental setup (a) and the coaxial nozzle (b).

One possible method was suggested from the so-called doubleexcitation method applied in spark and discharge emission analysis.10 In this method, a laser is used to evaporate or atomize small amounts of samples and then atomic emissions are excited by a discharge or a spark triggered by electrodes which are placed close to the irradiated surface. In LIBS methods, the second excitation source would be another laser pulse. Cremers et al.4 employed repetitive laser excitation in laser breakdown spectroscopy and observed a significant decrease in detection limit of boron ion in solution together with detection of the atomic emission of hydrogen and oxygen which were formed from water molecules. In this paper, we report LIBS detection of Fe concentrations less than 20 ppb. Using sequential laser pulses of two Q-switched Nd:YAG lasers as excitation sources, we have determined Fe concentration in the tens of ppb range in FeO(OH) suspensions with an optimum interval between two laser pulses and an optimum delay time of a detector gate from the second pulse. Effects of the interval and properties of two laser pulses were examined. The concentration of iron in boiler water sampled from a commercially operated thermal power plant has been determined successfully by this method. The results show the capability of LIBS in determining suspended colloidal and particulate impurities in liquids in a simple and quick way. They also show the merit of our coaxial flow system in controlling the nature of breakdown plasma. EXPERIMENTAL SECTION Two Q-switched Nd:YAG lasers were used: One delivered 8 ns pulses of up to 600 mJ (Continuum Co., NY82-20, hereafter L-1) and the other 6 ns and 160 mJ (Continuum Co., Surelite I-20, L-2) at 532 nm. Measured temporal pulse profiles, however, revealed that the second laser (L-2) pulse duration was shorter than the first one (L-1): 9-12 ns for L-1 and 2-4 ns for L-2. They were operated in the second harmonics at 20 Hz. Their beam divergences were different. Thus that of one laser was adjusted (10) Kubota, M. In Purazuma Hakkouhou (Plasma Emission Method for Analysis of Solid Samples); Murayama, S., Takahashi, T., Eds.; Gakkai Shuppan Center: Tokyo, 1982; pp 39-47 (in Japanese).

2982

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

by additional lenses and then two beams were focused onto the stream of sample solution by a quartz lens of 200-mm focal length. The pulse energy of L-1 was changed from 100 to 500 mJ/pulse using an attenuator but in most cases it was 150 mJ; that of L-2 was kept to 150 mJ. Commercially available FeO(OH), special reagent grade, was mixed in deionized and doubly distilled water to form a colloidal suspension. We used a coaxial configuration nozzle in which the sample solution flowed down the central nozzle at a constant flow rate and purge gas flowed through the outer shroud.1 Schematic drawings of the experimental setup and details of the nozzle are shown in Figure 1. The sample solution was filled in a bottle placed in a ultrasonic bath and pumped up by a small pump to the nozzle. The sample water was corrected under the irradiating point and was fed again into the bottle. The ultrasonic bath operated during measurements to prevent precipitation and adhesion of colloids in the bottle or tubing. The stream of water was irradiated by laser pulses which hit it in the rectangle, and luminescence spectra were taken from the direction perpendicular to both the water flow and the laser beam. The luminescence was collected by a telescope and transferred with an optical fiber to a multichannel photodetector (spectral multichannel analyzer, SMA, Princeton Instruments, Inc., IRY-1024G) mounted on a 32-cm polychrometer with a grating of 1200 grooves/mm. The nominal resolution of the system was 0.24 nm. The SMA was operated in a gate mode, and data were accumulated for 50-2000 laser shots. The timing of the two lasers and SMA gate was controlled by a digital pulse generator (Stanford Research Systems, Inc., DG535) and was monitored by a digital oscilloscope. The measured spectra were transferred to a personal computer and analyzed using a data-processing program we developed. Our program was capable of determining the emission peak intensity by both height and area above background signal. First, the data were smoothed by weighted averaging of nine data points. A region for noise level determination was selected on a display, it was from 225 to 230 nm in most cases, and its noise level (N) was calculated as the root mean square (σ) of the signal intensity

Figure 2. Luminescence spectrum obtained for a 10 ppm FeO(OH) suspension with He gas flow obtained by single pulse excitation (a) and sequential pulse excitation (b).

in this region and was set to be 2σ. The background signal was approximated by a second-order function deduced from the data in which peaks larger than 3σ of the noise level were omitted. A target peak was selected on the display and the peak above the background was approximated by a Gauss function; then its area and height were calculated. The intensity of the selected peak was indicated by two values: (1) the signal to noise (S/N) ratio, where S was the height above the background signal and N was the noise level, and (2) the area above the background level. The intensity of luminescence and thus the value of the peak area and the peak height vary with experimental conditions such as laser pulse energy. The value of S/N is, however, expected to be insensitive to these conditions. We used it in optimizing the measuring conditions, and the peak area was used in quantitative measurements under the optimum conditions thus determined. RESULTS Spectra Obtained for Single and Two Sequential 532-nm Irradiations. Luminescence spectrum in the near-UV region recorded for a 10 ppm FeO(OH) suspension by single laser (L-1) irradiation is shown in Figure 2a. It was taken between 500 and 1500 ns after a laser pulse in He and accumulated for 500 pulses. The laser pulse energy was 100 mJ. In our previous study, we operated a laser at its fundamental wavelength, 1064 nm.1 We doubled the frequency this time because it was easier to coincide two laser beams when the laser was in the visible region. A change of the emission spectra due to the change of excitation wavelength was not observed: we could identify emissions of the Fe atom at 260, 261, 268, 299, and 301 nm and those of the Fe ion at ∼274 nm in addition to a broad and continuous emission, which would be due to the blackbody radiation and/or the radiative recombination of the breakdown plasma. These lines coincide with the values in wavelength tables of Fe emissions in arc and/ or discharge analysis11 within our spectral resolution. The effects of the purge gas were also the same: the intensity ratio of Fe emission lines to the background emission (S/B ratio) was the largest in He, next in air, and the lowest in Ar. The luminescence spectrum obtained for the same suspension by two sequential laser pulse irradiation is shown in Figure 2b. It was taken between 500 and 1500 ns after the second laser pulse (11) MIT Wavelength Tables, 1969 ed.; MIT Press: Cambridge, MA, 1969. Also in tables in ref 10.

Figure 3. Effect of interval between two laser pulses on S/N of the 260-nm peak of a 10 ppm FeO(OH) suspension. The SMA gate was opened from 500 to 1500 ns after the L-2 pulse while the interval of the L-1 and L-2 laser pulse changed from +4600 to -500 ns. Minus sign means the L-1 pulse irradiated after the L-2 pulse. The line is only indicative.

in He and accumulated for 500 pulses. The pulse energy of both lasers was ∼150 mJ, and the interval between them was 1000 ns. The essential features of the spectrum are not changed from that obtained by single pulse excitation, but the S/B ratio evidently increases. The Fe concentration was estimated using the emission peak observed at 260 nm. This peak was relatively separated from other lines and thus we could easily determined its intensity. The peak might consist of an atomic line at 259.96 nm and ionic lines of 259.84 and 259.94 nm, both of which were known to be strong lines in ICP analysis.12 These lines were not separated by our detection system partly due to broadening of the spectral lines by the Stark effect in the breakdown plasma.9,13 Iron has stronger emission lines in at longer wavelengths, but they overlapped with much stronger background continuous emission in our measurements. In addition, we kept in mind the possibility of simultaneous determination of Fe with silicon, which has a characteristic emission line at 251.6 nm coming very close to, but spaced from, the Fe 260-nm line. Effects of the Interval and the Energy of Laser Pulses and Optimization of the Measuring Conditions. The effect of laser pulse interval on the S/N of the 260-nm emission peak was examined, and the results are shown in Figure 3. In the measurement, the SMA gate was opened from 600 to 1600 ns after an L-2 pulse while the interval of L-1 and L-2 laser pulses changed from +4600 to -500 ns. Here, the minus sign means the L-1 pulse irradiated after the L-2 pulse. The figure indicates that the intervals between +800 and +2500 ns give high S/N values. Even though the absolute intensity of emission decreased as the delay increased, the S/N did not correlate with the intensity. The effect of purge gas was checked for He and air, and the optimum pulse interval was the same for both cases. The S/N and S/B values were greater in He than in air. The results for an inverted sequence of lasers (i.e., the SMA gate was opened after L-1 with delays from 600 to 1600 ns) showed poorer S/N values than the noninverted sequence. This would be due to the difference of pulse duration of L-1 and L-2. The (12) See tables in: Ekitai Shiryouno Hakkoubunseki (Emission Spectroscopic Analysis of Liquid Samples); Takahashi, T., Murayama, S., Eds.; Gakkai Shuppan Center: Tokyo, 1983 (in Japanese). (13) Smith, D. C. J. Appl. Phys. 1977, 48, 2217-2225.

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2983

Figure 4. Variations of S/N with delay time of the SMA gate relative to the second pulse. The gate width was 1000 ns, and the pulse interval was 1000 ns. The pulse energy of L-1 was changed to 150, 350, and 500 mJ while that of the L-2 was 150 mJ. Fe concentration of the sample is 6.3 ppm. The line is only indicative.

measured L-2 pulse had almost 3 times shorter duration than L-1, and if pulse energies were equal, the peak power of L-2 was ∼3 times higher than that of L-1. The effect of the pulse energy of L-1, the preceding laser pulse, was studied from 150 to 500 mJ. In He, we found that a higher energy pulse could ignite brighter sparks but the quality of the measured spectra was rather insensitive to the energy. In air, higher energy pulses resulted in better S/N spectra and a 500mJ pulse could result in a spectrum with a quality similar to that obtained in He. A large increase in the S/N was observed at delay times longer than ∼500 ns when the energy of L-1 was 150 mJ, and this interval seemed to become shorter with increasing pulse energy. The interval of two laser pulses were chosen to be 1000 ns to get a good S/N spectrum with high intensity. Then we examined the effects of delay of the SMA gate and of its width. Figure 4 shows variations of S/N with delay time of the SMA gate relative to the second pulse. The gate width was kept to 1000 ns, the pulse interval was 1000 ns, and the pulse energy of L-1 was changed to 150, 350, and 500 mJ while that of L-2 was kept at 150 mJ. The Fe concentration of the sample was 6.3 ppm [i.e., 10 ppm FeO(OH)]. The results are rather scattered, but roughly speaking, delays from 400 to 1500 ns give high-S/N spectra. The pulse energy of the first laser has little effect on S/N values and their variations. From these and other results, we determined the optimum measuring conditions as follows: (1) He gas environment; (2) interval of successive laser pulses, 1000 ns; (3) delay of SMA gate to the second pulse, 500 ns; (4) gate width, 3300 ns; and (5) laser pulse energy, 150 mJ for both lasers. The data were accumulated for 1000-2000 pulses, which required only 50-100 s, to get better spectra. Detection Limit and Linearity of the Measurement and Determination of Fe in Boiler Water. In quantitative measurements, the peak intensity was estimated by a peak area above the background level as mentioned before. Figure 5 shows the correlation of the peak intensity with the Fe concentration of an FeO(OH) suspension. The Fe concentration in the abscissa was determined by the chemical analysis procedure, TPTZ method,14 usually carried out at the Niigata Thermal Power Station in (14) JIS (Japan Industrial Standards) B-8224-32, 2.

2984 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

Figure 5. Correlation of 260-nm peak intensity with Fe concentration of a FeO(OH) suspension. Fe concentrations on the abscissa were determined by the TPTZ method that is normally carried out at the plant.

Figure 6. Correlation of 260-nm peak intensity with Fe concentration in boiler water sampled from a plant at the Niigata Thermal Power Station. The thin line in this figure is that from Figure 5 and is shown for comparison.

monitoring boiler water. The TPTZ method can be summarized as follows: (1) Hydroxylamine hydrochloride and 2,4,6-tri-2pyridyl-1,3,5-triazine were added in dilute acid solution containing Fe. (2) Ammonium acetate was added in the solution and the pH value was prepared in the range of 4-5. (3) The Fe concentration was determined by measuring the absorbance of the Fe(II) coordination compound produced in the solution. As shown in this figure, we have a good linear correlation in the tens of ppb range. The detection limit where S is equal to 2σs is estimated to be 16 ppb. Here, the σs is the root mean square variation of the data points presented in the figure. Figure 6 shows the correlation of the peak intensity with Fe concentration of boiler water sampled from a plant at the Niigata Thermal Power Station. The results show a good correlation; however, the slope of the correlation line is smaller than that in Figure 5. Temporal Change of the Luminescence. The temporal change of the luminescence intensity was measured under the optimum measuring conditions. The SMA was operated with a 50-ns gate width and the gate delay times were varied by 50-ns intervals. The spectra were accumulated for 100 shots at each delay time and the emission intensity at 260 nm, which corresponds to the Fe emission lines, was plotted. The results are shown in Figure 7. For single pulse excitation, the temporal changes of the emission in He and in air consisted of two components: a fast

Table 1. Comparison of S/N Obtained for Single, Simultaneous, and Sequential Excitations S/N singlea simultaneousb sequentialc a

Figure 7. Temporal change of the luminescence intensity at 260 nm obtained for a 100 ppm FeO(OH) suspension with He gas flow. The abscissa is delay times from the first laser, i.e., at time 1000 ns the second laser is fired. The gate width of SMA was 50 ns.

Figure 8. Spectra of the fast component recorded at (a) 100 and (c) 1100 ns after L-1, i.e., immediately after the first and second laser pulses, and those of the slow component recorded at (b) 600 and (d) 1600 ns after L-1. (c) and (d) correspond at 100 and 600 ns after the second L-2 pulse, respectively. The gate width was 100 ns.

component and a slow component with lifetimes of 50 and 240 ns in He and 60 and 380 ns in air, respectively.1 For sequential excitation, we observed the repetition of the two components’ decay: the decay constant for the fast component in the second pulse excitation slightly increases from the first pulse (single pulse) excitation while that for the slow component is nearly the same. Figure 8 shows spectra of the fast component recorded at (a) 100 and (c) 1100 ns after L-1, i.e., immediately after the first and the second laser pulse respectively, and those of the slow component recorded at (b) 600 and (d) 1600 ns after L-1. (c) and (d) are the spectra at 100 and 600 ns after the second L-2 pulse, respectively. The gate width was 100 ns. The emission

10.2 17.5 55.2

L-1, 150 mJ. b L-1, L-2, 150 mJ. c Pulse interval 1600 ns.

lines of Fe dominated in the slow-component spectra while a broad, featureless emission was observed in the fast-component spectra. In addition, we see that S/B ratio in (d) (0.81) is better than (b) (0.62). DISCUSSION The spectra obtained show that the wavelength of the laser pulse does not have a significant effect on the spectrum of the breakdown plasma so long as the peak power of a laser is high enough to induce the breakdown. Figure 3 clearly shows that the increase in S/N and S/B is due to the sequential pulse excitation with a certain time interval and not due to a simple increase in laser pulse energy resulting from using two lasers. Zero interval data, in which the two laser pulses hit the sample simultaneously, show a rather low S/N compared to those for the longer intervals. Total intensity was nearly the same when we used only the L-1 at equivalent pulse energy to the sum of L-1 and L-2, the S/N was similar also. Comparison of S/N obtained for single, simultaneous, and sequential excitations is shown in Table 1. The optimum interval between two laser pulses is shorter than that reported by Cremers et al.4 They reported that they got their best results with an interval of 18 µs, which is more than 10 times longer than our optimum condition. This discrepancy would be due to the difference in the constituents of the breakdown plasma. They induced the plasma in a closed cell so that the plasma was from species produced from water molecules. We employed a coaxial sample flow system, and the produced plasma involved purge gas molecules.1 Therefore we should expect different plasma lifetimes and thus the optimum interval would change. In the previous study, we reported that the fast components have lifetimes dependent on the atmospheric gases used which vary from 50 ns in He to 320 ns in Ar. The slow components have longer lifetimes in both He and air and are considered to be effective emission lifetimes of Fe in those environment. We proposed that the thermal properties and the ionization potential of purge gas control the temporal change of the density and the temperature of the plasma.1 A molecule with a lower ionization potential would be ionized more easily and result in more efficient evolution of breakdown plasma with higher density and temperature. Its temperature would be kept high for a longer period in a gas of lower thermal conductivity. The fast component had the shortest lifetime in He, which has the highest ionization potential and the largest thermal conductivity. It had the longest lifetime in Ar, which has the lowest ionization energy and the smallest thermal conductivity. Thus, both sequences of ionization potentials and thermal conductivities of gas examined so far show good correlation with our results. Similar effects of purge gases to laser-induced plasma emission were reported for solid metal samples.6,7 Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

2985

This is one of the advantages in our system in comparison to previous solution measurements using cells.4,9 Plasma formed in liquid contained in a cell involve radicals, atoms, and ions come from liquid molecules. The nature of plasma formed is determined by these constituents, except for parameters controllable by a laser itself. With our flow apparatus, laser-induced plasma initially formed in a liquid stream is expected to propagate quickly, within the pulse duration, to the direction of incoming laser pulse and extend into purge gas flow. High-speed growth of plasmas toward the laser pulse is frequently observed in laser-induced breakdown.15 The plasma formed contains fragments from both liquids and purge gas molecules. Thus, we can control the nature of laser breakdown plasma to some extent by adding other species in it as purge gases. We now consider the reason why successive laser pulse excitation is capable of decreasing the detection limit. The second laser pulse excites a new spark in the middle of expanding gas/ plasma of water, purge gas, and Fe formed by the first pulse. The shock front of laser-induced plasma in liquid might propagate at a velocity of the order of 1000 m/s.15,16 Thus, the front is expected to move more than 1 mm during a 1-µs interval of two laser pulses. The diameter of the sample flow at the intersection with the laser beam was ∼1 mm. A significant decrease in the detection limit was observed at intervals longer than 500 ns, when the shock front was expected to have reached the surface of the stream. The second laser pulse hits this expanding plasma cloud. If we assume local thermal equilibrium in the initial plasma, atoms and ions involved in the plasma have the same thermal kinetic energy. Therefore, we expect that light molecules and atoms expand faster than Fe atoms/ions. The second laser pulse ignites a spark at lower power density since some electrons already exist and the avalanche need not be initialized by a “lucky” electron.15 Even though the second laser must ignite a spark in a lower density gas phase, it transfers enough energy to excite it again and most of the energy would be used to excite atoms and ions in the cloud since there would be no need to atomize and vaporize the iron particles and the water. The background broad emission of the second spark decays faster than that of the first spark due to lower density of the plasma where the relative density of Fe atoms/ ions significantly increases by the velocity difference in expanding gas. Thus, we can get a spectrum of improved S/B ratio after the second pulse. The detection limit depended upon the quality of spectra, which we can improve by increasing a number of accumulation. Thus, we clearly observed the 260-nm peak in the spectra from samples of 10 ppb or lower Fe concentration when more than 2000 shots were accumulated. The increase in accumulating number, however, means an increase of measurement time. We imposed on ourselves a condition that it should be at least 1 order of magnitude shorter than the conventional chemical analysis. This means it should be less than 2-3 min. A total of 2000 shots took 100 s in our experiment, which met the condition. The discrepancy in the slopes of Figures 5 and 6 would be due to the selective sensitivity of iron colloids and/or particles in the breakdown spectrometry. The Fe concentration of boiler (15) Docchio, F.; Regondi, P.; Capon, M. R.; Mellerio, J. Appl. Opt. 1988, 27, 3661-3668. (16) Bell, C. E.; Landt, J. A. Appl. Phys. Lett. 1967, 10, 46-48. (17) Ito, Y.; Nakamura, S.; Jo, M.; Sone, K.; Hiraga, H., to be submitted.

2986

Analytical Chemistry, Vol. 68, No. 17, September 1, 1996

water determined by the TPTZ method is considered to be the total amount of iron including both solid/colloids and ions while that estimated by LIBS should be of solid and colloids. In the emission spectra for an FeO(OH) suspension, the emission line of Fe at 260 nm was clearly observed, while this line did not appear in the spectra for an FeCl3 aqueous solution under the same observation conditions. The same selectivity has been reported in the case of calcium: atomic and ionic emissions of Ca were observed for a CaCO3 turbid solution while no Ca emission was observed for a Ca2+ solution of equivalent Ca concentration.9 Hence, the LIBS is effective for analysis of microparticulate impurities in liquids. In practical application of the method, the good linearity of the correlation line in the figure would allow us to use a calibrated graph which is made using boiler water of different Fe concentrations. The method is proved to have potential applicability in the analysis of boiler water quality at a thermal power plant. Practical applicability depends on a cost estimation: the cost of constructing a measuring system easily handled and maintained on a practical site, its economical benefit, and marketing research of the equipment, etc.; and these have not estimated fully yet. The method should be usable for simultaneous multielements analysis so far as spectral overlapping does not occur. In fact, we can measure the concentration of silicon simultaneously with that of Fe.17 CONCLUSION The method described here allows us to determine iron concentration in water as low as 16 ppb both in FeO(OH) colloidal suspension and in boiler water sampled from a thermal power plant. It takes only 100 s even when we accumulate 2000 pulses. It allows us to skip pretreatment of the sample and to measure the concentration continuously. The detection limit depends on element measured, but it should be decreased by proper choice of purge gas, laser power, a measuring timing relative to the excitation and by adopting the sequential excitation. The results show the capability of LIBS in the quantitative determination of colloidal and particulate substances in liquids. It provides a simple and quick method without any solubilizing procedures needed. Our coaxial flow method have several advantages of LIBS of fine particle suspensions and liquids: (1) It is possible to increase the sensitivity by controlling plasma constituents with a proper purge gas. (2) Simple cell-less structure results in easy handling. (3) Fresh sample liquid is continuously supplied and thus the online measurement can easily be achieved. (4) There is no interference from bubbles formed by the preceding spark. ACKNOWLEDGMENT This work was supported by a joint research program of Tohoku Electric Power Co., Inc. and Applied Laser Engineering Research Institute (ALERI). We thank to Messrs. T. Kobayashi, M. Jo, and K. Kuno for their help in the experiment. Received for review February 5, 1996. Accepted June 10, 1996.X AC9601167 X

Abstract published in Advance ACS Abstracts, July 15, 1996.