- Email: [email protected]
A fluidized-bed sampling system for the direct introduction of solids into an inductively coupled plasma—I. Performance characteristics
0584-8547/8613 oO+O.OO
Specrrochimrca Aera.Vol. 4lB, No. 9. pi. 865-674. 1986. Printed in Cire~tBritalo.
Pcrgamon Journals Ltd
A fluidized-bed sampling system for the direct introduction of solids into an inductively coupled plasma-I. Performance characteristics* K. NIMALASIRIDE Su_vAt and ROGER GUEVREMONT Atlantic Research Laboratory, National Research Council of Canada, 1411 Oxford Street, Halifax, Nova Scotia, Canada B3H 321 (Receiwd 19 August 1985; in revised form10 Jrrnuary1986) Abstract-A simple fluid&d-bed sample introduction system for the direct ICP spectrochemical analysis of solid samples has been evaluated. Studies requiring continuous introduction of powders are feasible. Several hardware design parameters,and experimental parameterswere optimized. Some characteristicsof atomization and emission of solid samples in a plasma have been considered. Emission of Si was not linearly related to the rate at which silica was delivered to the plasma at flows above 20mgjmin, while Cu emission remained linear above 80mg/min. Emission of Si and Zn were strongly dependent on parameters such as applied rf powerand carriergas flow, while emission of Cu was only slightly affected by changes in these parameters.
1. INTRODUCTION A NUMBERof problems have impeded the development of practical methods for flame or plasma analysis of solid samples. The sample must be introduced uniformly to the plasma to minimize variations in plasma characteristics (temperature, available energy, etc.). Uniform sample delivery (for most commercially available instrumentation) is prerequisite for wavelength scans, background corrections, and integration times of sufficient duration to maximize sensitivity. A minimum sample mass must be analysed to avoid problems related to sample inhomogeneity. Particles must be completely decomposed by the flame or plasma. Instrumental and measurement parameters must be understood and optimized. In addition, methods of standardization are required to simplify measurements of samples for which no reference materials presently exist. Several approaches for direct quantitative analysis of powders by flame or plasma have been described. Samples have been delivered in a slurry [l-3], sampled by arc and spark [4, 51, ablated by laser [6-83, introduced directly into the plasma [9-l 11, and heated by various electrothermal devices [12-141. Direct trace analysis of solids by atomic spcctrometry was reviewed by VANLOON [ 151 and by BROWNER [ 163. A fluidized-bed, or a gas/solid mixture having the properties of a liquid, can be formed and maintained by continuous input of mechanical energy either in the form of gas flows or *NRCCNo. 25900. +Present address:Mineral Resources Division, Geological Survey of Canada, 601 Booth Street, Ottawa, Ontario, Canada, KlA 0E8. [l]
J. B. WILLIS, Anal. C&m.
[2] N. MOHAMED and
47, 1752(1975). 53,450 (1981).
R. C. FRY, Anal. Chem.
[3] C. W. FULLER, R. C. HUTTONand B. PRBSTON,Analyst 106,913 (1981). [4] P. B. FARNSWORTHand G. M. HIEFTJE,Ad C/tern. 55, 1414(1983). [S] R. H. Scorr, Spectrochim.Acto 33B, 123 (1978). [6] T. ISHIZUKAand Y. UWAMINO, And. Chem. 52, 125 (1980). [‘I] M. THOMPSON,J. E. GOIJLTERand F. SIEPER,Awlyst 106, 32 (1981). [8] J. W. CARRand G. HORLICK, Spectrochim. Acto 37B, 1 (1982). [9] E. D. SALINand G. HORLICK, Anal. Ck-m. 51,2284 (1979). [lo] A. LORBER and Z. GOLDBART, Analyst 110. 155 (1985). [ll] A. G. PAGE, S. V. GODBOLE, K. H. MADRASWALA, M. J. KULKARNI, V. S. MALWPURKAR and Spectrochim. Acra 39B, 551 (1984). [12] A. M. GIJNN, D. L. MILLARD and G. F. KIRKBRIGHT,Adyst 103, 1066 (1978). [13] S. HANAMURA,B. W. SMITH and J. D. WINEFORDNER,And. Chem. 55,2026 (1983). [14] D. E. NIXON, V. A. FASSELand R. N. KNISELEY, AMY. Chem. 46,210 (1974). [15] J. C. VANLOON, Anal. Chem. 52,955A (1980). [16]
R. F. BROWNER,Tr. Anal. Chem. 2, 121 (1983). 865
B. D. JOSHI,
866
K. NIMALASIRI DE SILVAand ROGERGUEVREMONT
vibrations. The direct introduction of powders into a flame or inductively coupled plasma via a fluidized-bed approach has been described by DAGNALLet al. [ 171, HoAREand MOSTYN [ 183 and NG et al. [ 191. DAGNALL[ 171 compared fluidized-bed devices including a cyclone swirl cup and a cell with samples supported on a sintered glass disc. Samples including CaCO,, SiOZ, A&O3 and MgO were introduced, and the determination of Be and B in MgO described. Widespread use of uniform, finely divided substrates for chromatography and for preconcentration of trace components from complex matrices has suggested to us that further investigation of the fluidized-bed approach of sample introduction to the inductively coupled plasma for the analysis of these materials is warranted. We are interested, for example, in pellicular anion and cation exchangers (Whatman Inc., Clifton, NJ), reverse phase LC supports including Perisorb RP-18 (Merck, Darmstadt, Germany), organic polymers such as XAD resins (Rohm & Haas Co., Philadelphia, PA), chelating materials such as Chelex-100 (Bio-Rad, Richmond, CA), 8-hydroxyquinoline bonded to controlled pore glass (Pierce Chemical Co., IL) and to silica powders [20, 213. Automatic sampling hardware designed to preconcentrate trace components from seawater onto chelating substrates is being developed [22, 231. A fluidized-bed powder sampler for continuous and uniform delivery of particles to an inductively coupled plasma is described here. The methods which we have used to optimize a number of instrumental and measurement parameters for the analysis of powders are summarized. Our experience with this device suggests that analysis with a sequential spectrometer is possible (though not described in this paper). Questions related to efficiency of atomization and excitation of powder samples remain unanswered and are the subject of ongoing work. Aspects of data processing for quantitative analysis is considered in a following paper. Complex samples including marine sediments, coal and zirconium refractories have been introduced into the plasma; however, the present discussion will be restricted to only very simple matrices. 2. INSTRUMENTATIONAND MATERIALS 2.1. Spectrometers
Emission measurements were made with a dual spectrometer system consisting of a JY48P (multichannel spectrometer) and a JY37P (sequential spectrometer) (Instruments SA),both viewing a single plasma source. A summary of instrumentation and operating conditions appears in Table 1. Table 1. ICP instrumentation and operating conditions ICP system JY48P (Instruments SA, Metuchen, NJ) with 24 channels, 1 meter focal length, 2500 g/mm grating, 0.39 mn/mm dispersion, all elements measured in first order. JY37P (Instruments SA, Metuchen, NJ) sequential spectrometer,640 mm focal length, 1800 g/mm, 0.83 nm/mm dispersion, scan range 170-770 nm, complete computer control Plasma-Therm HFP-2500 rf generator, 27.12 MI-Ix. Digital Equipment PDPl l-23, RLO2 10M byte storage, TSX-Plus VS.0 time sharing extension of RT-11 (S & H Computers, Nashville, TN), time shared “simultaneous” operation of JY48P and JY37P spectrometers. Operating conditions rf forward power from 0.6 to 2.4 kW, reflected power 4 SW gas flows (a) coolant 16 l/mm argon (b) auxiliary 1.0 l/min argon (c) carrier 0.2 to 0.8 l/min argon. [17] R. M. DAGNALL, D. J. SMITH,T. S. WESTand S. GREENFIELD, Anal. Chim. Acta 54, 397 (1971). [18] H. C. HOAREand R. A. MOSTYN,Anal. Gem. 39, 1153(1967). [19] K. C. NG, M. ZEREZGHI and J. A. CARUSO,And. Chem. 56,417 (1984). [20] R. E. STURGEON, S. S. BERMAN, S. N. WILLIEand J. A. H. DESAULNIERS, Ad Chem. 53, 2337(1981). [21] M. A. MARSHALL and H. A. Mmou, And. Gem. 55,2089 (1983). [22] P. J. WANGERSKY, Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada (Personal communication). [23] Seakem Oceanography Ltd, Sidney, British Columbia, Canada, (Personal communication).
Sampling system for an ICP-I
867
Software supplied with the instrument was modified to operate with the TSX-Plus (S & H Computer Systems Inc., TN) operating system, a timesharing extension of RT-11 (Digital Equipment Corp., MA). The JY48P and JY37P run together in this environment. Errors in the JY48P software were corrected, notably a problem which led to incorrect internal standard ratio calculations with background corrected data The JY48P software was extended to acquire data from the JY37P (without scanning) as the N + 1 channel. New software was written for real-time graphic representation of emission detected at any JY48P channel and the JY37P spectrometer. Simultaneous real-time plots of intensity of emission of several elements were useful to establish the relationship among lines as a function of time, and for optimization of instrumental parameters. 2.2. Powder sample delivery A device for continuous delivery of a powder to an inductively coupled plasma (ICP) is shown schematically in Fig. 1. The powder sample was held in a 13 mm o.d. tube (2) threaded into an arm driven by an electric solenoid (1). The arm and sample tube travelkd 8 mm vertically. A glass cover (3) was threaded into the enclosure wall over the sample tube. Sample carrier gas entered at point (A) and exited through capillary (4). The solenoid arm fitted loosely around capillary (4) so as not to compress gas in the sample tube. Variable gas flow up the capillary (4) resulted in pulses in the emission signal. Activation of the solenoid suspended the sample particles in the gas and the resultant mixture flowed up the capillary (4). A second gas flow (B) was added in the closed cell (5). The sample passed to the base of the torch through capillary (6). Mass flow controllers (Tylan Corp., Carson, CA) with maximum flow limits of 300 and 1500 ml/min provided accurate gas delivery to points (A) and (B) respectively. Since flows (A) and (B) were controlled independently, the formation of the fluid&d-bed, and the transport of sample into the base of the plasma were separate and distinct. A microprocessor (SYS-2, Octagon Systems, Westminster, CO) controlled the solenoid and sensed the start of the spectrometer integration cycle. The frequency of solenoid motion, the relative durations of the activated and deactivated periods, the number of pulses and the stationary state (“up” or “down”) were adjustable input parameters to a simple “CONTROL BASIC” program stored on EPROM. Three operational modes of sampling were utilized: (1) continuous solenoid cycles, (2) fixed sequences of solenoid motion (e.g. three “pulses”) synchronized to integration, and (3) synchronized start of solenoid motion and integration, with total sample introduction during the integration period. The powder delivery unit was mounted below the torch box on a platform fixed to the bottom of the torch box support. The relative positions of the torch and delivery unit were therefore fixed. Observation height adjustments were made in the usual manner without added concern for the position to plasm0
t
I
1
Fig. 1. Schematic of a fluidized-bed system for the introduction of solids to an inductively coupled plasma. Components l-6 and gas flows A and B are described in Section 2.2.
K. NIMALASIRIDE S~LVA and ROGER GUEVREMONT
868
of the sample delivery unit. The platform position was adjustable to accommodate equipment up to 70 cm tall. Lowering of the platform disengaged the delivery capillary from the bottom of the torch, usually a 5 min operation. Precautions were taken to avoid air introduction into the system. The unit was thoroughly purged before use. During sample change the gas flows (A) and part of (B) exited through the sample introduction tube port. It was necessary to temporarily divert flow (B) while inserting the sample tube to avoid loss of sample. With the tube (2) and cover (3) in place the flow (B) was returned gradually. The pressure in the sampler increased until the gas flow to the plasma became constant. This delay was a function of the flow (A) and the volume of the sampler device. Certain powders accumulated severe electrostatic charges during vigorous shaking. Particularly troublesome were mixtures of particle sizes, where the fines coated the glassware and plugged the capillary tubes. It was found that argon presaturated with water resulted in less tendency for such charge build-up. 2.3. Silica powder Silica gel HF254 (Merck, Darmstadt, Germany) was separated into particle size fractions with six vertical columns of diameter from 2.5 cm to 7.5 cm (each 1 m in length). A slurry of the powder in distilled water was added to the column of narrowest diameter and a peristaltic pump used to maintain a constant flow of 15 ml/min through the series of columns. Approximately two days were required to reach equilibrium size distribution in the tubes. The particle diameters and the range of sizes of particles from each tube was determined by electron microscopy. The measurements described in this communication were all made with silica powders not subjected to prior chemical treatment or addition of spikes. The silica contained impurities of several elements at concentrations suitably high for the present investigation. It has been assumed that the impurities are distributed throughout the mass of the particulates and are not surface contaminants. 3. SAMPLEINTRODUCTION AND MEASUREMENT PARAMETERS 3.1. Sample delivery For the purpose of this study many experimental parameters were tixed. A summary is shown in Table 2. Measurements made with matrices other than silica powders, and of elements other than Si, Mn, Cu and Zn, are the subject of ongoing work. Most emission noise observed in this work was attributed to fluctuations in sample introduction rate. The uniformity of sample delivery was a function of rate of solenoid oscillation, particle size, sample delivery rate, gas flows (A) and (B), diameter of capillary tubes (4) and (6), length of capillary (6), shape of
Table 2. ExperimentalParametersfor this study Parametersfixed: (1) sample matrix,silica powder, no elemental spikes added (2) (3) (4) (5) (6) (7) (8) (9)
elements, Si, matrix impurities Mn, Cu, Zn plasma gas, argon gas flows, constant coolant and auxiliary flow quartz torch type Plasma-Therm Model T1.O quartz bonnet Plasma-Therm Model QB-1 torch extension was not used automatic rf tuning off-peak background correction, scanning entrance slit
Parameters variable: Sample introduction parameters: (1) sample particle size (2) sample introduction gas flow (A) (3) solenoid pulse timing and rate (4) shape and length of sample tube (5) diameter of capillary tubes (4) and (6) (6) length of capillary tube (6) Measurement parameters: (1) rf power (2) viewing height above torch (3) total sample carrier gas flow (A + B) (4) integration period
Sampling system for an ICP-I
869
sample holding tube and susceptibility of the sample to electrostatic charge. Several of these have been considered. 3.1.1. Sample introduction rate. Rate ofdelivery of sample to the plasma was determined by weighing the sample tube before and after the measurement period. Gas flow (A) was used as a practical control of sample introduction rate; typically from 10 to 100 ml/min was appropriate for silica powder delivery between 5 and 100 mg/min. The absolute rate was reproducible (at a given flow A) within 10 % at delivery rates over 50 mg/min. Adjustment offlow at low delivery rate was based on signal measurement and visual appearance of the plasma. Figure 2 illustrates the relationship between background corrected signals of Si and Cu and rate of delivery of silica (30 micron silica, carrier of 400 ml/min). Silicon emission is not linear with delivered mass at any of the wavelengths tested (251.611,220.798 and 205.813 nm). The Si emission at 251.6 nm remained non-linear with rate of delivery of sample, independent of photomultiplier voltage, and with a blind with about 20% transmission in front of the Si photomultiplier. 3.1.2. Sample introduction unifirmity. The total gas flow and the length of capillary (6) determined the delay between initiation of solenoid action and attainment of constant analyte emission signal, and similarly the decay time after sample introduction. The rise of emission to equilibrium, and the reverse process once solenoid motion is stopped is shown in Fig. 3. Delays between 2 s at flows above 600 ml/min and 10 s at flows below 400 ml/min were observed. This delay period was a measure of the effectiveness of capillary (6) in damping fluctuations in sample delivery rate; this is analogous to the electronic RC time constant for charging a capacitor. For comparison, data for introduction of liquid using a concentric nebulizer and conventional spray chamber is shown in the same figure. The powder introduction system provided analyte emission stable for periods of time typical of normal analytical measurements but showed drift resulting from depletion of sample at high rates of delivery and if measurements lasted several minutes. Figure 4 illustrates signal noise (attributed to fluctuations of sample delivery rate) over a short period of time. These noisy signals were chosen to show the high degree of correlation between emission of several elements, and they also show that the degree of correlation can signitkantly degrade for periods of time. Such departures from correlation may result from inhomogeneity of the sample. Figures 3 and 5-7 are based on emission signals with much less noise than that of Fig. 4, and these figures reflect fairly well the degree of noise observed without additional data processing (e.g. internal standards calculations). Figure 4 also illustrates one practical result of the nonlinearity of Si emission vs sample mass delivery rate. It is clear from Fig. 4 that the relative magnitude of “noise” of the Si emission is significantly less than that of the other elements. This is a consequence of the decrease in sensitivity of the Si emission to the changes in mass flow of sample at delivery rates over 20 mg/min.
+..-
,.*____* .__..;__............ .:: .-; _/’ ___.-*-,_1’
._,_
:,,
i
..d’ ,’
9
6
;
. . . ..__.
. ..6
.d’
I
I
I
I
20
40
60
80
FLOW
RATE
.
OF
POWDER,
mg /min
Fig. 2. Emission intensity of Cu (0) and Si (0) as a function of flow rate of silica powder to the Plasma.
870
K. NIMALASIRI DE SILVAand ROGERGUEVREMONI
TIME,
s
Fig. 3. Comparison of equilibration time,and noise levels for introduction of solid (silica, 20 mg/min) and liquid (Si 100 ppm, concentric nebulizer, Plasma-Therm) samples into the plasma: Si emission for silica, (-), and for liquid ( .. . .).
- 20 % _I I 0
1
1 8
I TIME,
I 16
I
I 24
s
Fig. 4. Normalized emission intensity of several elements during the introduction of silica powder to the plasma. An unusually noisy signal is illustrated.
3.1.3. Solenoid pulse timing. The frequency of activation of the solenoid, hence the amount of shaking of the powder, significantly affected the uniformity of Bow up to the plasma. Problems associated with static electric charges were severe with small particles and with vigorous mechanical agitation of the sample. For our system the relative standard deviation of emission signal was below 10% (tests based on 10 measurements using 1 s integration) at shaking frequencies above 5 Hz for silica particles of 20 p. This increased to about 12 % at 4 Hz and to over 60 % at 2 Hz. A more vigorous shake of 50 q particles was necessary, and the relative standard deviation of measured signal was below lOxat lOHz,about lS%at7Hzand80%at5Hz.
Sampling system for an ICP-I
871
3.1.4. Sample tube. Several sample tube shapes were tested. A tube with a conical bottom was found suitable for rapid and complete introduction of small samples, but the rate of introduction was not uniform. A tube with a spherical bottom was used in this work since the continuous, uniform introduction of part of a large sample (200-5OOmg) was a prerequisite for off-peak background correction. Tubes with flat bottoms were tested but no advantage was observed. Tubes of several lengths were tested. The distance between the end of capillary (4) and the bottom of the tube increased as the tube was lengthened, and more gas flow was necessary to carry equivalent mass of sample. Fluctuations in delivery rate became more severe. A short sample tube placed an upper limit on the mass of sample which could be held, consequently the duration of continuous flow before depletion of sample affected delivery rate. 3.1.5. Capillary diameter and length, and gasBow rate. Capillary tubes (4) of diameter 1.5 and 2.0 mm i.d. were compared. Sample introduction was more uniform with the narrower diameter capillary. At reduced sample delivery rates the low linear velocity ofcarrier gas resulted in the temporary suspension of heavier particles in the wider diameter capillary. Capillary tubes (6) of diameter 2 and 3 mm i.d. were tested and the wider was found unsatisfactory because of particle accumulation at low flow rates. The combination of diameter and length of tube (6) is an important factor in damping sample delivery variations. The maximum length of straight tube in the present design was 50 cm, and a coiled tube 2 m in length was tested. The increased length reduced fluctuations of sample introduction rate at low gas flows, but was ineffective above 500 ml/min. 3.1.6. Particle size. In preliminary qualitative experiments silica powders from below 10 p to above 80 pm in diameter were introduced into the plasma. It was expected that emission maxima would migrate higher into the plasma as larger particles were introduced, and increased time would be required to break down the particles. It was also expected that emission of Si could be used as a measure of mass of silica introduced to the plasma, independent of particle size. Preliminary experiments indicated that this was not the case, and emission from small particles was disproportionately high. These observations were consistent with the effects of total carrier gas flows described in Section (3.2.2). 3.2. Measurement parameters 3.2.1. Rf power. In an effort to assess the effectiveness of the plasma for destruction of particles and excitation of the resulting atoms, the emission of Si, Mn, Cu and Zn (silica powder samples) was measured from 0.6 to 2.4 kW forward rf power as a function of height. 30 m silica powder was introduced at approximately 20 mg/min, at a total carrier of 400 ml/min. The results are shown in Fig. 5. Wavelength profiles for each element indicated that background emission was insignificant for the data shown in this figure, as described in Section 3.2.4. The Si emission increased rapidly as the energy supplied to the plasma increased, and showed no evidence of reaching a plateau, even at 2.4 kW. 3.2.2. Sample carrier gaspow. Emission data was acquired at 5 mm increments above the load coil for a series of total carrier gas flow ranging from 200 and 800 ml argon/min. Rf power was lixed at 1.5 kW for these measurements. The Si, Zn, Mn and Cu emission intensity is illustrated as a function of carrier gas flow in Fig. 6. The maximum observed emission was just above the torch at the lowest argon flow rate capable ofpushing a channel into the plasma, in our case about 150 ml/min. As flow rate increased the maximum intensity decreased and the region of highest intensity showed an upward trend, moving approximately 10 mm higher above the load coil. Further work was carried out with total gas flow 400 ml/min, sacrificing some signal intensity for decrease in background intensity and structure at some wavelengths. 3.2.3. Observation height. Wavelength profiles of Si, Cu, Mn and Zn were recorded as a function of height above the load coil, and the results are shown in Fig. 7. Thirty micron silica powder was introduced at approximately 20 mg/min at a total carrier gas flow of 400 ml/min. Si emission was maximum from 8 to 16 mm above the load coil, the apparent decrease below 8 mm caused by attenuation of light by the quartz torch. Cu emission decreased only gradually with height above the load coil. Emission of many elements was recorded simultaneously with that of Si, Cu, Zn and Mn, and several show much more complex and interesting behaviour. In some regions of the spectrum the background emission low in the plasma at gas flows less than 200 ml/min, and high in the plasma (over 35 mm above the load coil) is intense and has complex structure. The Sn (189.9), As (193.6) and Se (196.0) lines were found to be susceptible to such emissions. 3.2.4. Off-peak background correction. The uniform introduction of samples was considered essential for valid off-peak background corrections with the JY48P and JY37P spectrometers as the background measurements are not made simultaneously in time with the analytical line measurement. Wavelength profiles across the emission lines shown in Fig. 7 were obtained by the JY48P which has a scanning entrance slit controlled by computer for background corrections. It is clear from these profiles that the background contribution to the emission intensities shown in previous figures is quite insignificant as compared to the line intensities.
872
K. NIMALASIRIDE
SILVA
and ROGER GUEVREMONT
Fig. 5. Emission of several elements as a function of applied forward rf power. Data was acquired over a range of viewing heights above the load
coil.
We consider errors in non-simultaneous background measurements as one of the factors most likely to degrade precision and detection limits when the system is used in the fashion described in this communication. Methods to assess background simultaneously with line measurement are being investigated. 4. DISCUSSION 4.1. Sample introduction
The design described here offers a unique separation of the sample stream formation process (e.g. nebulization process for liquids) from the transport process. As noted above this has made available the entire range of sample introduction rates (ranging over more than two orders of magnitude), in combination with the entire range of total gas flows. Changes in rate of sample introduction is possible while maintaining a constant total carrier gas flow. The experiment is more difficult to perform with liquid introduction systems, particularly since the actual liquid delivery rate to the plasma is usually unknown. A fluidized-bed approach for the formation of a stream of suspended particles suffers from the inherent problem that it may preferentially introduce particles of small size and lead to data biased by sample size segregation. The system described here uses mechanical motion to attempt to reduce the magnitude of such effects; however, since the problem has not been specifically addressed experimentally with this device, no estimation of the severity of the
873
Sampling system for an ICP-I
Mn
Si
0
20
10
HEIGHT,
: .’ ‘, ‘.
30
mm
Zn
‘.
: ,,.._.‘“‘.t ‘:.
z
I~~_
!y
I 0
I IO
20 HEIGHT,
mm
30
0
10
20 HEIGHT,
10
mm
Fig. 6. Emission of severalelements as a function of total carrierflow and of the viewing height above the load coil. Gas flows were 200 ml/min (. . . . .k 400ml/min (.--.-.); 600 ml/min (----); 800 ml/min (-).
problem is possible now. Nevertheless it will be noted again at this time that a wide variety of finely divided, and uniform particulates are being used in chromatography and for preconcentration from complex matrices. The present, admittedly restricted, objective is the analysis of such simple powders. 4.2. Atomic emission from powder samples The emission of Si was observed to be non-linear with rate of delivery of silica to the plasma. Among possible explanations was that the intensity of Si emission was strong enough to saturate the photomultiplier (PM) detector. Tests over a range of PM voltages, measurements at several Si wavelengths and measurements with a blind over the Si PM showed that this was unlikely. Other possible explanations are based on physical and spectroscopic phenomena in the plasma. The Si emission line may be self-reversed. It was observed that other, less strong, emission lines of Si are equally affected, and the problem appeared to be independent of observation position in the plasma. Since only Si (I) lines were investigated, further measurements with Si (II) lines should be considered before the possibility of self-reversal is ruled out. The silica particles may not be totally atomized by the plasma. This is in contradiction with the observation that some elements (e.g. Cu) showed nearly ideal behavior. Although the silica was used without chemical pretreatment or spiking it might be argued that Cu was only a surface contaminant. It might also be argued that Cu (and many other elements) could effectively be distilled out of hot silica particles (or droplets). Finally, it is possible that conditions in the plasma change as a result of increasing sample loads. This would result in decreasing energy available for excitation, and a concomitant decrease in observed emission. Some evidence shows that Cu is less affected by changes in
874
K. NIMALASIRI DE SILVAand ROGERGUEVREMONT
cu
amY
Avy dfw?zmm/nr
afw
7
.S‘Wfl
-iU
SW&7
LseFzmem/nr
Fig. 7. Emission and background intensity near analytical lines of several elements. Measurements were made at a series of viewing heights above the load coil.
available energy than is Si, and therefore exhibits less dependence on sample introduction rate. The Cu emission shown in Fig. 5 shows less dependence on forward rf power than does Si. Figure 7 similarly shows that Cu emission does not decrease with height above the load coil as rapidly as does the Si emission. 5. CONCLUSION The fluid&d-bed approach was found to be well suited for continuous and uniform introduction of silica powders of < 100 pm diameter. The device reported here was used to deliver a uniform flow of silica to the plasma for several minutes, and many experiments which are considered routine for liquid introduction systems have been shown to be possible. It is expected that the system will also be easily applied to studies related to particle size, and for measurements made on matrices other than silica.
Specrrochimrca Aera.Vol. 4lB, No. 9. pi. 865-674. 1986. Printed in Cire~tBritalo.
Pcrgamon Journals Ltd
A fluidized-bed sampling system for the direct introduction of solids into an inductively coupled plasma-I. Performance characteristics* K. NIMALASIRIDE Su_vAt and ROGER GUEVREMONT Atlantic Research Laboratory, National Research Council of Canada, 1411 Oxford Street, Halifax, Nova Scotia, Canada B3H 321 (Receiwd 19 August 1985; in revised form10 Jrrnuary1986) Abstract-A simple fluid&d-bed sample introduction system for the direct ICP spectrochemical analysis of solid samples has been evaluated. Studies requiring continuous introduction of powders are feasible. Several hardware design parameters,and experimental parameterswere optimized. Some characteristicsof atomization and emission of solid samples in a plasma have been considered. Emission of Si was not linearly related to the rate at which silica was delivered to the plasma at flows above 20mgjmin, while Cu emission remained linear above 80mg/min. Emission of Si and Zn were strongly dependent on parameters such as applied rf powerand carriergas flow, while emission of Cu was only slightly affected by changes in these parameters.
1. INTRODUCTION A NUMBERof problems have impeded the development of practical methods for flame or plasma analysis of solid samples. The sample must be introduced uniformly to the plasma to minimize variations in plasma characteristics (temperature, available energy, etc.). Uniform sample delivery (for most commercially available instrumentation) is prerequisite for wavelength scans, background corrections, and integration times of sufficient duration to maximize sensitivity. A minimum sample mass must be analysed to avoid problems related to sample inhomogeneity. Particles must be completely decomposed by the flame or plasma. Instrumental and measurement parameters must be understood and optimized. In addition, methods of standardization are required to simplify measurements of samples for which no reference materials presently exist. Several approaches for direct quantitative analysis of powders by flame or plasma have been described. Samples have been delivered in a slurry [l-3], sampled by arc and spark [4, 51, ablated by laser [6-83, introduced directly into the plasma [9-l 11, and heated by various electrothermal devices [12-141. Direct trace analysis of solids by atomic spcctrometry was reviewed by VANLOON [ 151 and by BROWNER [ 163. A fluidized-bed, or a gas/solid mixture having the properties of a liquid, can be formed and maintained by continuous input of mechanical energy either in the form of gas flows or *NRCCNo. 25900. +Present address:Mineral Resources Division, Geological Survey of Canada, 601 Booth Street, Ottawa, Ontario, Canada, KlA 0E8. [l]
J. B. WILLIS, Anal. C&m.
[2] N. MOHAMED and
47, 1752(1975). 53,450 (1981).
R. C. FRY, Anal. Chem.
[3] C. W. FULLER, R. C. HUTTONand B. PRBSTON,Analyst 106,913 (1981). [4] P. B. FARNSWORTHand G. M. HIEFTJE,Ad C/tern. 55, 1414(1983). [S] R. H. Scorr, Spectrochim.Acto 33B, 123 (1978). [6] T. ISHIZUKAand Y. UWAMINO, And. Chem. 52, 125 (1980). [‘I] M. THOMPSON,J. E. GOIJLTERand F. SIEPER,Awlyst 106, 32 (1981). [8] J. W. CARRand G. HORLICK, Spectrochim. Acto 37B, 1 (1982). [9] E. D. SALINand G. HORLICK, Anal. Ck-m. 51,2284 (1979). [lo] A. LORBER and Z. GOLDBART, Analyst 110. 155 (1985). [ll] A. G. PAGE, S. V. GODBOLE, K. H. MADRASWALA, M. J. KULKARNI, V. S. MALWPURKAR and Spectrochim. Acra 39B, 551 (1984). [12] A. M. GIJNN, D. L. MILLARD and G. F. KIRKBRIGHT,Adyst 103, 1066 (1978). [13] S. HANAMURA,B. W. SMITH and J. D. WINEFORDNER,And. Chem. 55,2026 (1983). [14] D. E. NIXON, V. A. FASSELand R. N. KNISELEY, AMY. Chem. 46,210 (1974). [15] J. C. VANLOON, Anal. Chem. 52,955A (1980). [16]
R. F. BROWNER,Tr. Anal. Chem. 2, 121 (1983). 865
B. D. JOSHI,
866
K. NIMALASIRI DE SILVAand ROGERGUEVREMONT
vibrations. The direct introduction of powders into a flame or inductively coupled plasma via a fluidized-bed approach has been described by DAGNALLet al. [ 171, HoAREand MOSTYN [ 183 and NG et al. [ 191. DAGNALL[ 171 compared fluidized-bed devices including a cyclone swirl cup and a cell with samples supported on a sintered glass disc. Samples including CaCO,, SiOZ, A&O3 and MgO were introduced, and the determination of Be and B in MgO described. Widespread use of uniform, finely divided substrates for chromatography and for preconcentration of trace components from complex matrices has suggested to us that further investigation of the fluidized-bed approach of sample introduction to the inductively coupled plasma for the analysis of these materials is warranted. We are interested, for example, in pellicular anion and cation exchangers (Whatman Inc., Clifton, NJ), reverse phase LC supports including Perisorb RP-18 (Merck, Darmstadt, Germany), organic polymers such as XAD resins (Rohm & Haas Co., Philadelphia, PA), chelating materials such as Chelex-100 (Bio-Rad, Richmond, CA), 8-hydroxyquinoline bonded to controlled pore glass (Pierce Chemical Co., IL) and to silica powders [20, 213. Automatic sampling hardware designed to preconcentrate trace components from seawater onto chelating substrates is being developed [22, 231. A fluidized-bed powder sampler for continuous and uniform delivery of particles to an inductively coupled plasma is described here. The methods which we have used to optimize a number of instrumental and measurement parameters for the analysis of powders are summarized. Our experience with this device suggests that analysis with a sequential spectrometer is possible (though not described in this paper). Questions related to efficiency of atomization and excitation of powder samples remain unanswered and are the subject of ongoing work. Aspects of data processing for quantitative analysis is considered in a following paper. Complex samples including marine sediments, coal and zirconium refractories have been introduced into the plasma; however, the present discussion will be restricted to only very simple matrices. 2. INSTRUMENTATIONAND MATERIALS 2.1. Spectrometers
Emission measurements were made with a dual spectrometer system consisting of a JY48P (multichannel spectrometer) and a JY37P (sequential spectrometer) (Instruments SA),both viewing a single plasma source. A summary of instrumentation and operating conditions appears in Table 1. Table 1. ICP instrumentation and operating conditions ICP system JY48P (Instruments SA, Metuchen, NJ) with 24 channels, 1 meter focal length, 2500 g/mm grating, 0.39 mn/mm dispersion, all elements measured in first order. JY37P (Instruments SA, Metuchen, NJ) sequential spectrometer,640 mm focal length, 1800 g/mm, 0.83 nm/mm dispersion, scan range 170-770 nm, complete computer control Plasma-Therm HFP-2500 rf generator, 27.12 MI-Ix. Digital Equipment PDPl l-23, RLO2 10M byte storage, TSX-Plus VS.0 time sharing extension of RT-11 (S & H Computers, Nashville, TN), time shared “simultaneous” operation of JY48P and JY37P spectrometers. Operating conditions rf forward power from 0.6 to 2.4 kW, reflected power 4 SW gas flows (a) coolant 16 l/mm argon (b) auxiliary 1.0 l/min argon (c) carrier 0.2 to 0.8 l/min argon. [17] R. M. DAGNALL, D. J. SMITH,T. S. WESTand S. GREENFIELD, Anal. Chim. Acta 54, 397 (1971). [18] H. C. HOAREand R. A. MOSTYN,Anal. Gem. 39, 1153(1967). [19] K. C. NG, M. ZEREZGHI and J. A. CARUSO,And. Chem. 56,417 (1984). [20] R. E. STURGEON, S. S. BERMAN, S. N. WILLIEand J. A. H. DESAULNIERS, Ad Chem. 53, 2337(1981). [21] M. A. MARSHALL and H. A. Mmou, And. Gem. 55,2089 (1983). [22] P. J. WANGERSKY, Department of Oceanography, Dalhousie University, Halifax, Nova Scotia, Canada (Personal communication). [23] Seakem Oceanography Ltd, Sidney, British Columbia, Canada, (Personal communication).
Sampling system for an ICP-I
867
Software supplied with the instrument was modified to operate with the TSX-Plus (S & H Computer Systems Inc., TN) operating system, a timesharing extension of RT-11 (Digital Equipment Corp., MA). The JY48P and JY37P run together in this environment. Errors in the JY48P software were corrected, notably a problem which led to incorrect internal standard ratio calculations with background corrected data The JY48P software was extended to acquire data from the JY37P (without scanning) as the N + 1 channel. New software was written for real-time graphic representation of emission detected at any JY48P channel and the JY37P spectrometer. Simultaneous real-time plots of intensity of emission of several elements were useful to establish the relationship among lines as a function of time, and for optimization of instrumental parameters. 2.2. Powder sample delivery A device for continuous delivery of a powder to an inductively coupled plasma (ICP) is shown schematically in Fig. 1. The powder sample was held in a 13 mm o.d. tube (2) threaded into an arm driven by an electric solenoid (1). The arm and sample tube travelkd 8 mm vertically. A glass cover (3) was threaded into the enclosure wall over the sample tube. Sample carrier gas entered at point (A) and exited through capillary (4). The solenoid arm fitted loosely around capillary (4) so as not to compress gas in the sample tube. Variable gas flow up the capillary (4) resulted in pulses in the emission signal. Activation of the solenoid suspended the sample particles in the gas and the resultant mixture flowed up the capillary (4). A second gas flow (B) was added in the closed cell (5). The sample passed to the base of the torch through capillary (6). Mass flow controllers (Tylan Corp., Carson, CA) with maximum flow limits of 300 and 1500 ml/min provided accurate gas delivery to points (A) and (B) respectively. Since flows (A) and (B) were controlled independently, the formation of the fluid&d-bed, and the transport of sample into the base of the plasma were separate and distinct. A microprocessor (SYS-2, Octagon Systems, Westminster, CO) controlled the solenoid and sensed the start of the spectrometer integration cycle. The frequency of solenoid motion, the relative durations of the activated and deactivated periods, the number of pulses and the stationary state (“up” or “down”) were adjustable input parameters to a simple “CONTROL BASIC” program stored on EPROM. Three operational modes of sampling were utilized: (1) continuous solenoid cycles, (2) fixed sequences of solenoid motion (e.g. three “pulses”) synchronized to integration, and (3) synchronized start of solenoid motion and integration, with total sample introduction during the integration period. The powder delivery unit was mounted below the torch box on a platform fixed to the bottom of the torch box support. The relative positions of the torch and delivery unit were therefore fixed. Observation height adjustments were made in the usual manner without added concern for the position to plasm0
t
I
1
Fig. 1. Schematic of a fluidized-bed system for the introduction of solids to an inductively coupled plasma. Components l-6 and gas flows A and B are described in Section 2.2.
K. NIMALASIRIDE S~LVA and ROGER GUEVREMONT
868
of the sample delivery unit. The platform position was adjustable to accommodate equipment up to 70 cm tall. Lowering of the platform disengaged the delivery capillary from the bottom of the torch, usually a 5 min operation. Precautions were taken to avoid air introduction into the system. The unit was thoroughly purged before use. During sample change the gas flows (A) and part of (B) exited through the sample introduction tube port. It was necessary to temporarily divert flow (B) while inserting the sample tube to avoid loss of sample. With the tube (2) and cover (3) in place the flow (B) was returned gradually. The pressure in the sampler increased until the gas flow to the plasma became constant. This delay was a function of the flow (A) and the volume of the sampler device. Certain powders accumulated severe electrostatic charges during vigorous shaking. Particularly troublesome were mixtures of particle sizes, where the fines coated the glassware and plugged the capillary tubes. It was found that argon presaturated with water resulted in less tendency for such charge build-up. 2.3. Silica powder Silica gel HF254 (Merck, Darmstadt, Germany) was separated into particle size fractions with six vertical columns of diameter from 2.5 cm to 7.5 cm (each 1 m in length). A slurry of the powder in distilled water was added to the column of narrowest diameter and a peristaltic pump used to maintain a constant flow of 15 ml/min through the series of columns. Approximately two days were required to reach equilibrium size distribution in the tubes. The particle diameters and the range of sizes of particles from each tube was determined by electron microscopy. The measurements described in this communication were all made with silica powders not subjected to prior chemical treatment or addition of spikes. The silica contained impurities of several elements at concentrations suitably high for the present investigation. It has been assumed that the impurities are distributed throughout the mass of the particulates and are not surface contaminants. 3. SAMPLEINTRODUCTION AND MEASUREMENT PARAMETERS 3.1. Sample delivery For the purpose of this study many experimental parameters were tixed. A summary is shown in Table 2. Measurements made with matrices other than silica powders, and of elements other than Si, Mn, Cu and Zn, are the subject of ongoing work. Most emission noise observed in this work was attributed to fluctuations in sample introduction rate. The uniformity of sample delivery was a function of rate of solenoid oscillation, particle size, sample delivery rate, gas flows (A) and (B), diameter of capillary tubes (4) and (6), length of capillary (6), shape of
Table 2. ExperimentalParametersfor this study Parametersfixed: (1) sample matrix,silica powder, no elemental spikes added (2) (3) (4) (5) (6) (7) (8) (9)
elements, Si, matrix impurities Mn, Cu, Zn plasma gas, argon gas flows, constant coolant and auxiliary flow quartz torch type Plasma-Therm Model T1.O quartz bonnet Plasma-Therm Model QB-1 torch extension was not used automatic rf tuning off-peak background correction, scanning entrance slit
Parameters variable: Sample introduction parameters: (1) sample particle size (2) sample introduction gas flow (A) (3) solenoid pulse timing and rate (4) shape and length of sample tube (5) diameter of capillary tubes (4) and (6) (6) length of capillary tube (6) Measurement parameters: (1) rf power (2) viewing height above torch (3) total sample carrier gas flow (A + B) (4) integration period
Sampling system for an ICP-I
869
sample holding tube and susceptibility of the sample to electrostatic charge. Several of these have been considered. 3.1.1. Sample introduction rate. Rate ofdelivery of sample to the plasma was determined by weighing the sample tube before and after the measurement period. Gas flow (A) was used as a practical control of sample introduction rate; typically from 10 to 100 ml/min was appropriate for silica powder delivery between 5 and 100 mg/min. The absolute rate was reproducible (at a given flow A) within 10 % at delivery rates over 50 mg/min. Adjustment offlow at low delivery rate was based on signal measurement and visual appearance of the plasma. Figure 2 illustrates the relationship between background corrected signals of Si and Cu and rate of delivery of silica (30 micron silica, carrier of 400 ml/min). Silicon emission is not linear with delivered mass at any of the wavelengths tested (251.611,220.798 and 205.813 nm). The Si emission at 251.6 nm remained non-linear with rate of delivery of sample, independent of photomultiplier voltage, and with a blind with about 20% transmission in front of the Si photomultiplier. 3.1.2. Sample introduction unifirmity. The total gas flow and the length of capillary (6) determined the delay between initiation of solenoid action and attainment of constant analyte emission signal, and similarly the decay time after sample introduction. The rise of emission to equilibrium, and the reverse process once solenoid motion is stopped is shown in Fig. 3. Delays between 2 s at flows above 600 ml/min and 10 s at flows below 400 ml/min were observed. This delay period was a measure of the effectiveness of capillary (6) in damping fluctuations in sample delivery rate; this is analogous to the electronic RC time constant for charging a capacitor. For comparison, data for introduction of liquid using a concentric nebulizer and conventional spray chamber is shown in the same figure. The powder introduction system provided analyte emission stable for periods of time typical of normal analytical measurements but showed drift resulting from depletion of sample at high rates of delivery and if measurements lasted several minutes. Figure 4 illustrates signal noise (attributed to fluctuations of sample delivery rate) over a short period of time. These noisy signals were chosen to show the high degree of correlation between emission of several elements, and they also show that the degree of correlation can signitkantly degrade for periods of time. Such departures from correlation may result from inhomogeneity of the sample. Figures 3 and 5-7 are based on emission signals with much less noise than that of Fig. 4, and these figures reflect fairly well the degree of noise observed without additional data processing (e.g. internal standards calculations). Figure 4 also illustrates one practical result of the nonlinearity of Si emission vs sample mass delivery rate. It is clear from Fig. 4 that the relative magnitude of “noise” of the Si emission is significantly less than that of the other elements. This is a consequence of the decrease in sensitivity of the Si emission to the changes in mass flow of sample at delivery rates over 20 mg/min.
+..-
,.*____* .__..;__............ .:: .-; _/’ ___.-*-,_1’
._,_
:,,
i
..d’ ,’
9
6
;
. . . ..__.
. ..6
.d’
I
I
I
I
20
40
60
80
FLOW
RATE
.
OF
POWDER,
mg /min
Fig. 2. Emission intensity of Cu (0) and Si (0) as a function of flow rate of silica powder to the Plasma.
870
K. NIMALASIRI DE SILVAand ROGERGUEVREMONI
TIME,
s
Fig. 3. Comparison of equilibration time,and noise levels for introduction of solid (silica, 20 mg/min) and liquid (Si 100 ppm, concentric nebulizer, Plasma-Therm) samples into the plasma: Si emission for silica, (-), and for liquid ( .. . .).
- 20 % _I I 0
1
1 8
I TIME,
I 16
I
I 24
s
Fig. 4. Normalized emission intensity of several elements during the introduction of silica powder to the plasma. An unusually noisy signal is illustrated.
3.1.3. Solenoid pulse timing. The frequency of activation of the solenoid, hence the amount of shaking of the powder, significantly affected the uniformity of Bow up to the plasma. Problems associated with static electric charges were severe with small particles and with vigorous mechanical agitation of the sample. For our system the relative standard deviation of emission signal was below 10% (tests based on 10 measurements using 1 s integration) at shaking frequencies above 5 Hz for silica particles of 20 p. This increased to about 12 % at 4 Hz and to over 60 % at 2 Hz. A more vigorous shake of 50 q particles was necessary, and the relative standard deviation of measured signal was below lOxat lOHz,about lS%at7Hzand80%at5Hz.
Sampling system for an ICP-I
871
3.1.4. Sample tube. Several sample tube shapes were tested. A tube with a conical bottom was found suitable for rapid and complete introduction of small samples, but the rate of introduction was not uniform. A tube with a spherical bottom was used in this work since the continuous, uniform introduction of part of a large sample (200-5OOmg) was a prerequisite for off-peak background correction. Tubes with flat bottoms were tested but no advantage was observed. Tubes of several lengths were tested. The distance between the end of capillary (4) and the bottom of the tube increased as the tube was lengthened, and more gas flow was necessary to carry equivalent mass of sample. Fluctuations in delivery rate became more severe. A short sample tube placed an upper limit on the mass of sample which could be held, consequently the duration of continuous flow before depletion of sample affected delivery rate. 3.1.5. Capillary diameter and length, and gasBow rate. Capillary tubes (4) of diameter 1.5 and 2.0 mm i.d. were compared. Sample introduction was more uniform with the narrower diameter capillary. At reduced sample delivery rates the low linear velocity ofcarrier gas resulted in the temporary suspension of heavier particles in the wider diameter capillary. Capillary tubes (6) of diameter 2 and 3 mm i.d. were tested and the wider was found unsatisfactory because of particle accumulation at low flow rates. The combination of diameter and length of tube (6) is an important factor in damping sample delivery variations. The maximum length of straight tube in the present design was 50 cm, and a coiled tube 2 m in length was tested. The increased length reduced fluctuations of sample introduction rate at low gas flows, but was ineffective above 500 ml/min. 3.1.6. Particle size. In preliminary qualitative experiments silica powders from below 10 p to above 80 pm in diameter were introduced into the plasma. It was expected that emission maxima would migrate higher into the plasma as larger particles were introduced, and increased time would be required to break down the particles. It was also expected that emission of Si could be used as a measure of mass of silica introduced to the plasma, independent of particle size. Preliminary experiments indicated that this was not the case, and emission from small particles was disproportionately high. These observations were consistent with the effects of total carrier gas flows described in Section (3.2.2). 3.2. Measurement parameters 3.2.1. Rf power. In an effort to assess the effectiveness of the plasma for destruction of particles and excitation of the resulting atoms, the emission of Si, Mn, Cu and Zn (silica powder samples) was measured from 0.6 to 2.4 kW forward rf power as a function of height. 30 m silica powder was introduced at approximately 20 mg/min, at a total carrier of 400 ml/min. The results are shown in Fig. 5. Wavelength profiles for each element indicated that background emission was insignificant for the data shown in this figure, as described in Section 3.2.4. The Si emission increased rapidly as the energy supplied to the plasma increased, and showed no evidence of reaching a plateau, even at 2.4 kW. 3.2.2. Sample carrier gaspow. Emission data was acquired at 5 mm increments above the load coil for a series of total carrier gas flow ranging from 200 and 800 ml argon/min. Rf power was lixed at 1.5 kW for these measurements. The Si, Zn, Mn and Cu emission intensity is illustrated as a function of carrier gas flow in Fig. 6. The maximum observed emission was just above the torch at the lowest argon flow rate capable ofpushing a channel into the plasma, in our case about 150 ml/min. As flow rate increased the maximum intensity decreased and the region of highest intensity showed an upward trend, moving approximately 10 mm higher above the load coil. Further work was carried out with total gas flow 400 ml/min, sacrificing some signal intensity for decrease in background intensity and structure at some wavelengths. 3.2.3. Observation height. Wavelength profiles of Si, Cu, Mn and Zn were recorded as a function of height above the load coil, and the results are shown in Fig. 7. Thirty micron silica powder was introduced at approximately 20 mg/min at a total carrier gas flow of 400 ml/min. Si emission was maximum from 8 to 16 mm above the load coil, the apparent decrease below 8 mm caused by attenuation of light by the quartz torch. Cu emission decreased only gradually with height above the load coil. Emission of many elements was recorded simultaneously with that of Si, Cu, Zn and Mn, and several show much more complex and interesting behaviour. In some regions of the spectrum the background emission low in the plasma at gas flows less than 200 ml/min, and high in the plasma (over 35 mm above the load coil) is intense and has complex structure. The Sn (189.9), As (193.6) and Se (196.0) lines were found to be susceptible to such emissions. 3.2.4. Off-peak background correction. The uniform introduction of samples was considered essential for valid off-peak background corrections with the JY48P and JY37P spectrometers as the background measurements are not made simultaneously in time with the analytical line measurement. Wavelength profiles across the emission lines shown in Fig. 7 were obtained by the JY48P which has a scanning entrance slit controlled by computer for background corrections. It is clear from these profiles that the background contribution to the emission intensities shown in previous figures is quite insignificant as compared to the line intensities.
872
K. NIMALASIRIDE
SILVA
and ROGER GUEVREMONT
Fig. 5. Emission of several elements as a function of applied forward rf power. Data was acquired over a range of viewing heights above the load
coil.
We consider errors in non-simultaneous background measurements as one of the factors most likely to degrade precision and detection limits when the system is used in the fashion described in this communication. Methods to assess background simultaneously with line measurement are being investigated. 4. DISCUSSION 4.1. Sample introduction
The design described here offers a unique separation of the sample stream formation process (e.g. nebulization process for liquids) from the transport process. As noted above this has made available the entire range of sample introduction rates (ranging over more than two orders of magnitude), in combination with the entire range of total gas flows. Changes in rate of sample introduction is possible while maintaining a constant total carrier gas flow. The experiment is more difficult to perform with liquid introduction systems, particularly since the actual liquid delivery rate to the plasma is usually unknown. A fluidized-bed approach for the formation of a stream of suspended particles suffers from the inherent problem that it may preferentially introduce particles of small size and lead to data biased by sample size segregation. The system described here uses mechanical motion to attempt to reduce the magnitude of such effects; however, since the problem has not been specifically addressed experimentally with this device, no estimation of the severity of the
873
Sampling system for an ICP-I
Mn
Si
0
20
10
HEIGHT,
: .’ ‘, ‘.
30
mm
Zn
‘.
: ,,.._.‘“‘.t ‘:.
z
I~~_
!y
I 0
I IO
20 HEIGHT,
mm
30
0
10
20 HEIGHT,
10
mm
Fig. 6. Emission of severalelements as a function of total carrierflow and of the viewing height above the load coil. Gas flows were 200 ml/min (. . . . .k 400ml/min (.--.-.); 600 ml/min (----); 800 ml/min (-).
problem is possible now. Nevertheless it will be noted again at this time that a wide variety of finely divided, and uniform particulates are being used in chromatography and for preconcentration from complex matrices. The present, admittedly restricted, objective is the analysis of such simple powders. 4.2. Atomic emission from powder samples The emission of Si was observed to be non-linear with rate of delivery of silica to the plasma. Among possible explanations was that the intensity of Si emission was strong enough to saturate the photomultiplier (PM) detector. Tests over a range of PM voltages, measurements at several Si wavelengths and measurements with a blind over the Si PM showed that this was unlikely. Other possible explanations are based on physical and spectroscopic phenomena in the plasma. The Si emission line may be self-reversed. It was observed that other, less strong, emission lines of Si are equally affected, and the problem appeared to be independent of observation position in the plasma. Since only Si (I) lines were investigated, further measurements with Si (II) lines should be considered before the possibility of self-reversal is ruled out. The silica particles may not be totally atomized by the plasma. This is in contradiction with the observation that some elements (e.g. Cu) showed nearly ideal behavior. Although the silica was used without chemical pretreatment or spiking it might be argued that Cu was only a surface contaminant. It might also be argued that Cu (and many other elements) could effectively be distilled out of hot silica particles (or droplets). Finally, it is possible that conditions in the plasma change as a result of increasing sample loads. This would result in decreasing energy available for excitation, and a concomitant decrease in observed emission. Some evidence shows that Cu is less affected by changes in
874
K. NIMALASIRI DE SILVAand ROGERGUEVREMONT
cu
amY
Avy dfw?zmm/nr
afw
7
.S‘Wfl
-iU
SW&7
LseFzmem/nr
Fig. 7. Emission and background intensity near analytical lines of several elements. Measurements were made at a series of viewing heights above the load coil.
available energy than is Si, and therefore exhibits less dependence on sample introduction rate. The Cu emission shown in Fig. 5 shows less dependence on forward rf power than does Si. Figure 7 similarly shows that Cu emission does not decrease with height above the load coil as rapidly as does the Si emission. 5. CONCLUSION The fluid&d-bed approach was found to be well suited for continuous and uniform introduction of silica powders of < 100 pm diameter. The device reported here was used to deliver a uniform flow of silica to the plasma for several minutes, and many experiments which are considered routine for liquid introduction systems have been shown to be possible. It is expected that the system will also be easily applied to studies related to particle size, and for measurements made on matrices other than silica.