Sensitivities of inductively coupled plasma optical emission

Mar 15, 1992 - Alfred P. Weber, Rolf. Keil, Leo. .... Dr. Gregory L. Beutner, a principal scientist at Bristol-Myers Squibb, is the recipient of ... 1...
1 downloads 0 Views 806KB Size
672

Anal. Chem. 1992, 64, 672-677

Nemeth, of General Motors, St. Catharines, is thanked for providing samples of standard reference materials. Funds for this research were provided by the Research Advisory Committee of the Ontario Ministry of the Environment (Project 434 G). REFERENCES (1) (2) (3) (4)

Holak. W. Anal. Chem. 1989, 4 7 , 1712-1713. Welz, B. Chem. &. 1988, 2 2 , 130-133. Nakahara, T. Rag. Anal. At. Spectrosc. 1983, 6 , 163-223. Qodden, R.; Thomerson, D. K. Analyst (London) 1980, 705, 1137-1257. (5) Dedina, J. Rog. Anal. At. Spec17osc. 1988, 7 1 , 251-260. (6) Rlby, P. Q.; Haswell, S. J.; Olzeskowlak, R. J . Anal. At. Spectrom. 1989, 4 , 181-184. (7) Huang. B.; Zeng, X.; Zhang, 2.; Llu, J. Spectrochim. Acta, Part 6 1988, 438, 381-389. (8) Ebdon, L.; Sparkes, S. Mlcfochem. J . 1987, 36, 198-206. (9) Ek, P.; Hulden, S. U. Talenta 1987, 34. 495-502. (IO) Boempong, C.; Brindle, I. D.; Ceccarelli-Ponzoni, C. M. J . Anal. At. Spectnnn. 1987, 2 , 197-200. (11) Agsett, J.; Aspell, A. C. Analyst(London) 1978, 707. 341-347. (12) Braman, R. S.; Johnson, D. L.; Fweback, C. C.; Ammons, J. M.; Bricker, J. L. Anal. Chem. 1977, 49, 621-625. (13) Anderson, R.; Thompson, M.; Culbard, E. Analyst (London) 1986, 7 7 7 , 1143-1152, 1153-1158. (14) Van Clewenbergen, R. J. A,; Van Mol, W. E.; Adams, F. C. J . Anal. At. Spectrom. 1988, 3 , 169-176. (15) Welz, B.; Melcher, M. Analyst (London) 1984, 709, 573-575. (16) Sinemus, H. W.; Melcher, M.; Welz, B. At. Spectrosc. 1981, 2 , A - .1-Re --. (17) Haring, B. J. A.; Van Delft, W.; Bom, C. M. Fresenius' Z . Anal. Chem. 1982. 370,217-223.

(18) Stephens, P. At. Absorp. Ne&. 1979. No. 18, 118-120. (19) Terada, K.; Matsumoto, K.; Inaba, T. Anal. Chlm. Acta 1984, 758, 207-215. (20) Sun, S.; Xue, J. Youkuengye 1986, 5, 31-34. (21) Goulden, P. D.; Brooksbank, P. Anal. Chem. 1974, 46, 1431-1436. (22) . . Takahashi. Y.: Ono. T.: Yokovama. T.: Taruntaml. T. Chhtsu 1987. 24, 383-389. 131-142 K. S.; Meranger, J. C. Anal. Chlm. Acta 1981, 724, (23) Subramanlan, .- . . .-. (24) Chakrabortl, D.; De Jonghe, W.; Adam, F. Anal. Chlm. Acta 1980, 720, 121-127. (25) Chung, C.; Iwamoto, E.; Yamamoto, M.; Yamamoto, Y. Spectrochlm. Acta, Part 6 1984. 398, 459-466. (26) Hamilton, T. W.; Ellis, J.; Florence. T. M. Anal. Chhn. Acta 1980, 779, 125-233. (27) Boampong, C.; Brindle. 1. D.; Le, X-c.; PMwerbersky, L.; Ceccarelli Ponzonl, C. M. Anal. Chem. 1988. 6 0 , 1185-1188. (28) Smlth, A. E. Analyst (London) 1975, 700, 300-308. (29) Pierce, F. D.; Brown, H. R. Anal. Chem. 1978, 48, 893-695; 1977, 49, 1417-1422. (30) Welz, B.; SchubertJacobs, M. J . Anal. At. Spectrom. 1988, 7 , 23-27. (31) Hershey, J. W.; Kellher. P. N. Spectrochim. Acta, Part8 1988. 4 7 8 , 7 13-723. (32) Welz. B.; Melcher, M. Analyst (London) 1984, 709. 589-572. (33) Brindle, I. D.; Le, X-c.; Ll, X-f. Analyst (London) 1988, 773, 1377-1381. (34) Brindle, I . D.; Le, X-c. J . Anal. At. Spectmm. 1989, 4 , 227-232. (35) Brindle, I. D.; Le, X-c. Anal. Chlm. Acta 1990, 229, 239-247. (36) Andreae, M. 0.;Asmode, J. F.; Foster, P.; Van t'dack, L. Anal. Chem. 1981, 53, 1766-1771.

RECEIVED for review August 7,1991. Accepted December 17, 1991.

Sensitivities of Inductively Coupled Plasma Optical Emission Spectrometry for Dry and Wet Aerosols Alfred P.Weber, Rolf Keil, Leo Tobler, and U r s Baltensperger*

Paul Schemer Institute, CH-5232 Villigen PSI, Switzerland

The sensitlvlty of the lnducilvely coupled plasma optical "donspectrometry (ICP-OES) dgnal was Investigated for dry aerosols generated by spark dkcharges, as well as aerorok generated by a conventbnal ultrasonic nebuker (USN). Particle dzes of the generated aerosols were C300 nm and the ICP sendthrlty was not Influenced by partlcle size. The w a l l amount of resldual water after the desolvator (1.5 mg/mln) dld not change the plasma condltlons, dnce the exdtakn temperature In the plasma r " d constant wlthln 81 K for the dry and wet aerosols. Nevertheless, mlxlng of the &y aerosol wtth H,O/HNOs bopkt0 resutted In a reductkn of the ICP rlgnal by 43.1% which could be attributed to scavenging of analyte partlcks by the droplets generated In the USN. An absolute callbratlon of the ICP-OES signal by neutron acthratlam analyrl. was performed wing silver aers sols. Wlth them calibration data the nebullzlng efflclency of tho USN was detemlned, and a value of (5.2 f O S ) % was found.

INTRODUCTION Inductively coupled plasma optical emission spectrometry (ICP-OES)has become one of the most important techniques for rapid multielement analysis.' Usually, the sample is introduced by nebulizing a solution and transporting the formed

* To whom correspondence should be addressed. 0003-2700/92/0364-0672$03.00/0

droplets to the plasma. Quantification of unknown samples is performed by establishing identical conditions for standards and samples. Most commonly, pneumatic nebulizers and ultrasonic nebulizers are in use. Ultrasonic nebulizers (USN) typically have higher nebulization efficiencies than pneumatic nebulizers. In order to reduce the amount of excess solvent vapor, they are normally combined with a desolvator to produce a relatively dry aerosol to enter the plasma.2 (In the following, the term aerosol is used for suspensions of both liquid and solid particles in a gas.) The influence of the aeroaol generation on the instrument signal has been investigated by various Usually, sensitivity is defiied as the slope of the calibration curve. This means that the term sensitivity includes the production and transport of the aerosol as well as the photon yield in the plasma. As an example, for a higher nebulization efficiency a higher sensitivity should be expected. However, changes of nebulizer efficiency also change plasma conditions, e.g. by introducing variable amounts of water into the plasma. Thus it is not easily possible to investigate nebulizer efficiency and plasma conditions separately. On the other hand, methods for direct introduction of a solid material,avoiding dissolution of the sample, are available. The evaporationof solid samples is usually performed by laser ablation8or spark erosion? Both techniques are comparable with respect to analytical precision and detection limits, with the laser offering the advantage that both conducting and nonconducting samples can be used. Again, both techniques do not suffer from variable inputs of water, but they do not generate a constant aerosol for periods of hours. This would 0 1992 American Chemical Society

ANALYTICAL CHEMISTRY, VOL. 64, NO. 6, MARCH 15, 1992

be necessary to study, e.g., the influence of the water loading on the ICP signal. This disadvantage is eliminated when spark discharges are produced between two electrodes consisting of the material of interest. Such spark discharge aerosol generators easily deliver an aerosol of constant concentration during periods of hours.1° In our research, we study methods for on-line maea analysis of aeroeol particles. ICP-OES represents a valuable possibility for this purpose." It offers the advantage of an on-line measurement with multielement analysis capability and low detection limits.However, in order to use it for absolute mass determinations, it had to be calibrated by an independent technique, for which neutron activation analysis (NAA) was the method of choice. For several reasons this calibration was not performed with the conventional aerosol generated by a nebulizer but rather with a dry aerosol generated with a spark discharge generator. First, the problem with variable water loading as mentioned above could be investigated. Second, the pressure increase when introducing a filter behind the USN was too high to ensure a correct operation of the USN. To study the influence of dry and wet aerosols on the plasma conditions the spark discharge generated aerosol was passed through an ultrasonic nebulizer with optional mixing with droplets of H20/HNO3 Alternatively, the ultrasonic nebulizer was used to produce an analyte aerosol conventionally. This approach offered the possibility of separating nebulizer and plasma effects and was used to study the influence of the water loading on the ICP sensitivity. Thus,this paper presents a method to study one parameter while maintaining the remaining parameters constant, which is highly d e ~ i r a b l e .Silver ~ was chosen for most experiments because of low detection limits in ICP-OES and NAA. Iron was chosen for additional investigations on plasma conditions. Iron and silver can easily be generated as an aerosol by spark discharges. EXPERIMENTAL SECTION An Applied Research Laboratory (ARL) ICP spectrometer (Model 3410) with a minitorch was used. The forward power of the radio frequency (27.12 MHz) generator was 650 W. A 1-m monochromator in Czerny-Turner arrangement with a 2400 grmves/mm grating and 20-pm exit slits was used. The resolution of the monochromator was 0.013 nm, and the observation zone was 10 mm above the load coil. The spatial window had a diameter of 8 mm. Silver aerosols were produced by two methods. In the conventional method a solution of Ag (Baker, lo00 ppm standard solutionprepared from Ag metal and 1M HNOJ in 0.14 M HN03 (Merck, suprapur) was nebulized using an ultrasonic nebulizer (ARL). The flow rate for the Ag solution was held constant at 2.24 mL/min by a peristaltic pump (Gilson, Miniplua 2). Freshly deionized water (Millipore, Milli-Q) was prepared for each experiment. The USN was followed by a heater and a condenser to eliminate most of the solvent in the carrier gas. The heater and the condenser were set to 121 "C and 1OC, respectively. The desolvator was activated for all experiments where the USN was turned on. The portion of the remaining solvent reaching the ICP was measured gravimetrically by passing the dried aerosol through a tube filled with magnesium perchlorate. Alternatively, the silver aerosol was produced by spark discharges. The spark discharge generator consisted of a capacitor and of two cylindricalAg electrodes of 3-mm diameter.12 Argon was used as carrier gas, with a flow rate of 1.4 L/min. The capacitor was charged continuously by the current source until the breakdown voltage was reached at the electrodes. The following spark produced primary spherules, which formed agglomerates by diffusional coagulation according to size and concentration. Aerosol mass concentration was varied by changing the current that loaded the capacitor. The breakdown voltage was about 3 kV (in argon),and currents up to 1.0 mA were chosen. For a given current, number and mass concentrations were constant within 10% during 8 h. Operation was possible for about

* lNFl

673

I

I

USN

S

Flgurr 1. Schematic diagram of the experiment: MFC, mass flow controller; SDG, spark discharge generator; P, pressure sensor: V, three-way valve; NF, Nuclepore filter for NAA BP, bypass; R, rotameter; S, solvent;PP, peristattic pump; D, drain; CW, condenser waste; H, heater; C, condenser.

30 h without replacing the electrodes. After replacing, concentrations started at lower values and increased to constant values within 2 h. The size distributions of the silver particles were measured by differential mobility analyzers (DMA),either by a TSI Model 3071 for particles with mobility equivalent diameters d, 2 11 nm or by a Hauke DMA for d, 2 5.5 nm. Number concentrationsof the selected size intervals were determinedeither by a condensation particle counter (CPC, TSI Model 3022) or an electrometer. The carrier gas flowing into the ICP was argon at a flow rate of 0.8 L/min. Ar was also used for the auxiliary gas (1.0 L/min) and for the coolant gas (7.5 L/min). All gases had a purity of 99.998% (H20 I5 ppm, O2 5 3 ppm). For the calibration experiment, the aerosol was split, one fraction (0.8 L/min) flowing through the USN into the ICP, the other one (0.6 L/min) through a Nuclepore filter (pore size 0.4 pm, thickness 10 pm) for NAA (Figure 1). This filter also samples aerosol particles smaller than the filter pore size with high efficiency, since besides impaction, interception and diffusion also contribute to the fiter efficiency.13 The efficiency of the Nuclepore filter was determined experimentally by feeding the outlet of the filter to the ICP-OES apparatus and was found to be >99.94%. The total gas flow and the flow through the filter were measured by mass flow controllers (Brooks 58503 and Brooks 5850TR). In addition, the gas flow through the USN was monitored by a rotameter. The pressure regime was controlled manually by means of a U-tube manometer downstream of the spark discharge generator. For changing the filter a bypass was available. The sampling time for the NAA varied from 5 to 15 min according to the mass concentration. For the determination of Ag by NAA, every filter was irradiated separately in a polyethylene container for 4 min at a neutron flux of 2.2 X 1013n/cm2.s in the research reactor "Saphir". Transport to and from the reactor was done by a pneumatic transport system (Rabbit). The nuclear reaction lo7Ag(n,y) lWAg(cross section u = 35 b) was used to activate the silver. From the subsequent decay of lWAg(Ti,2 = 2.42 min, E, = 632.9 keV), y-ray spectra were collected for 6 min after a decay time of 2 min with a Ge(Li) detector (FWHM (1332) = 2.7 keV) coupled to a Nuclear Data 9900 NVAX-based multiuser data acquisition and processing system. Results were obtained by comparison with separately irradiated Ag standards Counting statistics were always better than 2%, and estimated overall accuracy was better than 5%. Nuclepore filter blanks were 500 pg/L the mode of the size distribution increases with increasing analyte concentration. For concentrations < 500 pg/L, however, the mode is virtually independent of the analyte concentration, due to the presence of residual particles in HzO as well as the added HNO,. It is known that even when nebulizing ultrapure water, high number concentrations of residual particles are found.14J5 Thus, these residual particles generate a lower limit for the analyte particle size. Without these residual particles, a t low concentrations the analyte aerosol would consist of very small particles and would suffer from substantial transport losses due to the high diffusion coefficient of these small particles, As an example, this loss would amount to 12% for 10-nm particles for a transport distance of 1 m at a flow rate of 0.8 L/min.16 From the number concentrations, mass concentrations were calculated assuming spherical particles. Correcting for the residual particles, these mass concentrations were proportional to the analyte concentrationsin the silver solutions. In the size range covered with this type of aerosols (i.e. up to 300 nm), the ICP sensitivity did not show any dependence on the particle size distribution. This is reasonable, since Bochert found for single manganese(I1) sulfate particles produced by a vibrating orifice aerosol generator that the ICP signal was proportional to the particle mass for particles up to 8 pm (experimentalconditions: plasmatherm torch with a power of 1700 W, coolant gas 30 L/min Ar, and carrier gas 1 L/min Ar)."J8 Then, the same measurements for the aerosols generated by spark discharges were performed. Figure 3 shows the size distributions for the Ag aerosol particles at various loading currents. These particles were formed by diffusional coagulation of primary spherical particles. Thus,in contrast to the particles produced by the nebulizer, these particles are agglomerates and exhibit a complicated shape resulting in a different relationship between particle diameter and particle volume than in the case of spheres.19 It has been shown in simulations20~21 and experimentszzthat diffusion-grown agglomerates can be described by the concept of fractals and that over a certain size range the following equation is valid:

M

-

dpDf

IO

Diameter [nm]

(1)

Figure 3. Mass size distributions of silver agglomerates formed by spark discharges as a function of the loading current (0.4-1 .O mA). Mass concentrations were calculated with a fractal dimension D, = 2.2.'' The megq distributkn of kon aggkmerates produced h the same way is included, assuming the same fractal dimension as for the silver agglomerates.

where M is the mass and d, a diameter of the particle, e.g. the mobility equivalent diameter, and Df a structural parameter of the agglomerate called the fractal dimen~i0n.l~ As an example, slender rods show Df = 1 and the particle mass scales with the particle length. For Df = 2 the mass is proportional to the surface as in the case of a disk, and for Df = 3 the mass is proportional to the volume according to the behavior of a sphere. For particles formed by agglomeration processes Df is typically