ANALYTICAL CHEMISTRY, VOL.
(2) H. L. Kahn, M. Bancrofl and R. H. Emmel, Res./Dev., 27, 30 (1976). (3) 6. R. Culver and T. R. Surles, Anal. Chem. 47, 920 (1975). (4) G. Lundgren, Talanta, 23, 309 (1976).
2375
(19) J. V. Chauvin, M. P. Newton, and D. G. Davis, Anal. Chim. Acta, 65, 291 (1973). (20) 6.Watne and R. Woodriff, Appl. Spectrosc.. 30, 7 1 (1976). (21)J. Agget and A. J. Sprott, Anal. Chim. Acta, 72, 49 (1974). (22)R. E. Thiers in "Methods for Biochemical Analysis" D. Glick, Ed., Vol. 5, Interscience. New York, 1957,pp 274-309. (23) R. W. Karin, J. A. Buono, and J. L. Fasching, Anal Chem., 47, 2296 (1976). (24)Adel Abv-Samra, J. Steven Morris, and S.R. Koirtyohann, Anal. Chern., 47, 1475 (1975). (25) "Analytical Methods for Flame Spectroscopy", Varian Instrument Division, 61 1 Hansen Way, Palo Alto, Calif. 94303. (26) J. Smeyers-Verbeke, Y. Michotte, P.Van den Winkel, and D. L. Massart, Anal. Chem., 46, 125 (1976). (27)R. E. Sturgeon, C. L. Chakrabarti, I. S. Maines, and f'. C. Bertels, Anal. Chem., 47, 1240 (1975). (28)M. R. Sensmeier, W. F. Wagner, and G. D. Christian, Fresenius' 2. Anal. Chern., 277, 19 (1975). (29) T. Maruta and T. Takeuchi, Anal. Chim. Acta, 66,5 (1973). (30) Eugene L. Meier, U S . Army Medical Research and Development Command, Fort Detrick, Frederick, Md. 21701,personal communication, 1976.
(5) T. Surles, J. R. Tuschall. and T. T. Collens, Environ. Sci. Techno/., 9,
1073 (1975). (6) T. S. West and X. K. William, Anal. Chim. Acta, 57,281 (1971). (7) R. G. Anderson, H. N. Johnson, and T. S. West, Anal. Chim. Acta, 57, 281 (1971). (8) R. D. Reeves, C. J. Molnar, M. T. Glenn, J. R. Ahlstrom, and J. D. Winefordner, Anal. Chem., 44, 2205 (1972). (9) M. D. Amos, P. A. Bennett, K. G. Brodie, P. W. Y. Lung, and J. P. Matousek, Anal. Chem., 43, 211 (1971). (IO) A. Montaser and S. R. Crouch, Anal. Chem., 47, 38 (1975). (11)J. Y. Hwang, C. J. Mokeler, and R. A. Uilucci, Anal. Chem., 44, 2018 (1972). (12) N. S. McIntyre, M. G. Look, and D. G. Boase, Anal. Chem.. 46, 1983 (1974). (13) M. P. Bratzei, R. M. Dagneil, and J. D. Winefordner, Appl. Spectrosc., 24, 518 (1970). (14) S.R. Goode. A. Montaser, and S. R. Crouch, Appl. Spectrosc., 27,355, (1973). (15) W. Lund and 8. V. Larsen. Anal. Chim. Acta, 61,319 (1976). (16)W. Lund, B. V. Larsen, and N. Gundersen. Anal. Chlm. Acta, 61, 319 (1976). (17) M. P. Newton and D. G. Davis, Anal. Chem., 47, 2003 (1975). (18) M. P. Newton, J. V. Chauvin, and D. G. Davis, Anal. Left., 6. 89 (1973).
51, NO. 14, DECEMBER 1979
RECEIVED for review May 31, 1977. Resubmitted September 8, 1978. Accepted August 6, 1979.
Sampling at Constant Temperature in Graphite Furnace Atomic Absorption Spectrometry D. C. Manning, Walter Slavin," and S. Myers The Perkin-Elmer Corporation, Main A venue, Norwalk, Connecticut 06856
When using graphite furnace sampling in atomic spectrometry, it is advantageous to introduce the sample into a hot environment that is at thermal equilibrium (constant temperature). One means of doing this is to dry an aliquot of the sample on a wire made of a high melting point material, and then l o place the wire and dried sample into the hot furnace. Using this procedure, matrix interferences are eliminated or greatly reduced compared to the conventional procedure of introducing the sample onto the cold wall of the furnace for the determination of Pb and TI. High concentrations of chloride matrix provide a vapor phase interference.
Several papers dealing with graphite furnace atomic spectrometry have pointed out the advantages of introducing an analytical sample into a hot environment, isothermal in time and space (1-7). Such an environment reduces or eliminates the analytical interferences associated with the variations in atomization conditions due to differences in sample matrix. In his early work, L'vov dried the sample on the end of a graphite rod which was then introduced into a preheated graphite tube (6). Woodriff and Ramelow ( 7 ) introduced the sample into a preheated tube using a small graphite cup. Littlejohn and Ottaway (8)used a furnace tube for emission. Their tube was designed to become hotter on the ends than in the center. When the sample was vaporized from the tube center, the vapor diffused toward the ends, encountering a hotter region that dissociated molecular bonds and excited the atoms more efficiently. In the furnace design described by Massmann, the sample is introduced to the inside wall of a cold furnace tube (9). The tube is then heated, and the analyte is volatilized a t a tem0003-2700/79/0351-2375$01 .OO/O
perature that is a t least partially dependent on the matrix composition of the sample. L'vov (2) points out that some of the potential difficulties inherent in heating the tube and sample simultaneously can be avoided if the sample is placed on a platform within the graphite tube. Slavin and Manning ( 4 ) have demonstrated the advantage of this idea for the determination of P b , using a commercial graphite furnace based on the Massmann design. However, use of the L'vov Platform delays the vaporization of the sample for only a limited time, which is not under the control of the analyst. The atmosphere within the graphite tube may not have reached constant temperature by the time the platform surface reaches the temperature a t which the sample vaporizes. This will depend, among other things, on the analyte, the matrix and the heating rate. I t would be preferable to introduce the sample after constant temperature conditions have definitely been established. L'vov (10) has suggested this might be done hy drying the sample on a wire, and then plunging the wire and dried sample into the preheated graphite tube. Since the time when the sample is introduced can be controlled, introduction into an environment a t constant temperature is assured. Drying the sample on a wire was used in early quantitative analytical spectroscopy. J. Janssen (11) used a platinum wire to introduce the sample into the flame of a Bunsen burner in 1870, attempting to determine sodium quantitatively. Since that time, many other workers have used the method (12), although not associated with the graphite furnace. Very recently Garnys and Smythe (13)have described a tungsten wire for introducing the sample into the graphite furnace. Although their arrangement is very similar to the one described in this paper, their goal appears to relate to sampling from a tungsten surface and to increasing sample throughput while this paper t2 1979 American Chemical Society
2376
ANALYTICAL CHEMISTRY, VOL. 51, NO. 14, DECEMBER 1979
-
__-__
Table I. Furnace Program Wire Sampling --__ 1 2
4
temp, "C ramp, s hold, s recorder, s read, s int. gas, mL/min Wall Sampling
650
EXPERIMENTAL APPARATUS AND PROCEDURE We used the Perkin-Elmer Model 5000 atomic absorption spectrophotometer equipped with background correction, a Model PRS-10 printer, a Model 56 recorder, an EDL power supply, and lead and thallium electrodeless discharge lamps. The Model HGA-600 graphite furnace, equipped with a Model AS-1 Auto Sampler, was used. The analytical wavelengths were 217.0 nm for P b and 276.8 nm for T1 and the spectral slit width was 0.7 nm for both elements. Pyrolytically-coated graphite tubes were used. During atomization when sampling from the wire, the internal argon gas flow was stopped. When sampling from the tube wall, the internal gas flow was 20 mL/min. We used this flow rate for wall sampling to remain consistent with our previous experience (14). The purpose of the very low flow of argon is to protect the windows of the cell from the deposition of matrix materials. Our previous paper described the determination of P b in a chloride matrix and showed that the chloride interference was the same when there was no flow and when the flow rate was 20 mL/min. To introduce the wire into the graphite tube conveniently and reproducibly, we mounted the wire in a holder that was placed in the sampling arm of the AS-1 Auto Sampler (Figure 1). The wire mount was designed to replace the standard pipet tip directly, which was removed and taped to the sampling arm. The sampling arm was manually rotated using the handwheel. The sample was deposited onto the coil on the end of the wire using a micro-dispensing pipet. Surface tension was sufficient to keep the droplet on the wire coil while it was transported to the graphite tube. We have routinely used 5-pL sample drops; however we were able to transport 10-pL drops to the furnace using this coil design, with careful manipulation of the sample arm. We have chosen tungsten wire (0.010-in. diameter) for our experiments, but other high-melting-point metals or graphite may be advantageous. The furnace controller program in Table I was used for Pb and T1. To avoid memory effects due to ineffective cleaning of the tube and wire after each firing, it was necessary to preheat the wire before depositing the sample and to fire off the residue after the analysis. After the sample droplet was pipetted onto the wire coil, the graphite furnace power controller start button was pushed. The sample arm was rotated into a position to bring the wire coil 1to 2 mm above the sample introduction hole during the 5-s ramp period of step 1. The arm was held in this position while heated argon from the furnace evaporated the solvent. The liquid outside the wire coil evaporated more quickly than the liquid within the coil. When only the liquid within the coil remained, the coil was inserted through the sample hole into the tube (still at 650 "C). The fact that the drop was completely dry was determined by holding a dental mirror on the left-hand side of the furnace and observing the wire illuminated by the light source. When the drop was dry, the wire was removed from the tube a distance of 3 to 5 cm and held. This was done before step
_____ 4 2400
3
5
2700 1
1800 1
35
4
16 t7
Flgure 1. Tungsten wire and holder mounted on the AS-1 auto sampler accessory
describes the reduction of interference effects. We have conducted a number of experiments using wire sampling. It was not a very convenient technique to use, but it did demonstrate the expected reduction of matrix interference.
program step
5
1
30 1
4
6
T-10
+5
0
0
program step temp, "C ramp, s hold, s recorder, s read, s int. gas, mL/min
1
2
3
4
120
550 15
20
15
2500 9 9 0 0
2700
1
1 5
20
2 started to avoid prematurely vaporizing the sample. Step 2 was a high temperature purge to clean the tube of any
residue from the drying step. The wire with the dried sample was introduced into the tube 10 s after the beginning of step 3, the atomization step. Previous experience (15) indicated that this was enough time for the tube to reach constant temperature. During this step the internal purge gas was stopped to minimize convective loss of the sample, and a recorder tracing of the absorption signal was made to show the peak shape. The recorder was started 7 s after the beginning of step 3, which was 3 s before the wire was introduced. At the end of atomization, the tube and wire were heated to 2400 O C (step 4) to clean any remaining matrix residue from the wire. In step 5, a reference or base-line reading was taken with the wire in the tube in the same position as in step 3. The sample signal was the net of these two readings. We used this procedure since even with background correction, it was not possible to adjust the wire so it would cause no deflection of the base line when the wire was inserted into the tube. As in our previous work (4, 14), we used integrated absorbance (abs-s) analytical signals, instead of peak height. At the conclusion of the atomization program, the tungsten wire remained inside the graphite tube in the argon atmosphere while both tube and wire were cooled to avoid oxidation.
RESULTS AND DISCUSSION This sampling arrangement is not particularly attractive for routine spectrochemical analysis. The drying of the sample droplet must be closely monitored. The wire must be removed from the tube after the sample is dried, but before the tube is heated t o its final atomization temperature. We have not investigated the possibility or desirability of ashing prior t o atomization. The objective of this work has been to determine whether matrix interferences are reduced or eliminated by introducing the sample into the graphite tube after it has reached constant temperature. Using the program summarized in Table I and the protocol previously described, consecutive integrated absorbance signals for 0.5 ng P b were reproduced with a relative standard deviation of 2% (k10 pg). Nine t o twelve consecutive sampling sequences were used t o calculate this precision. T h e same reproducibility was observed for P b in the presence of 0.2% MgC12. Figure 2 shows a recorder tracing of five consecutive absorption peaks produced by 0.5 ng of Pb. Using wire introduction of the sample, we have seen no instance of multiple absorption peaks as is sometimes observed (Figure 3) when vaporizing the sample from the wall. If these multiple peaks from the wall were caused by condensation and re-evaporation of'the lead along the tube length, as has been proposed ( 2 , 4 ) ,one would not expect to see them when
ANALYTICAL CHEMISTRY, VOL. 51, NO. 14, DECEMBER 1979
2377
IO-
W 060
'?
2U
TIME
-
051
0 O4 23 '-I
Figure 2. Five absorption peaks, each produced by 0.5 ng Pb sampled from a tungsten wire. The time axis is shown on the chart
1
0
25
50
45secJc
,
75
100
L-_
! 25
n g Pb
I
Figure 5. Pb calibration graph obtained by wire sampling
0,3t v)
m
4 0.1
TIME +
Figure 3. Example of multiple Pb peaks sometimes seen when sampling from the wall
5 7L
-
,
3 6t
n
7 1i e c
A
WIRE
' 9 1
0.001 0,002 0.004 0.01 0.02 0.04 0.1 0.2
0 1600°C
TIME
22OO.C
0.4
% MgCI2
Figure 6. Comparison of the interference effect of MgCI, on TI, sampling from a wire and from the furnace wall. One ng of TI was used on the wall and 0.3 ng TI on the wire
___________
:21 1400.C
WALL
__
Table 11. Recovery of Pb Added to Each of Five Matrix Compounds; Comparison of Wire Sampling to Wall Sampling NaFI,- Na,NaC1, CaCl,, MgCI2, PO,, SO,, 1% 0.2% 0.2% 0.5% 0.5% percent recovery wire sampling 92% 97% 9570 91% 67% 18% 89% 20% 74'70 wall sampling 73%
+
Figure 4. Pb absorption peaks at several temperatures. The time axis is expanded as shown on the chart
introducing the sample into an environment that had already reached constant temperature. Figure 4 shows tracings using 0.5 ng P b run a t three different tube temperatures, but with the time scale expanded to more clearly show absorption peak shape. T h e relationship of P b concentration to integrated absorbance is shown in Figure 5 . T h e sensitivity (element mass which corresponds to a given integrated absorbance signal) of this technique is about equal to t h a t observed with the conventional method of depositing the sample on the wall. We investigated matrix interferences for P b and T1. For Pb, five matrix compounds were chosen based on our previous observation ( 4 ) of their interference effects. A concentration level for each was used which produced a background-only signal of between 0.8 and 1.0 absorbance unit when 5 pL was sampled on the tungsten wire. This absorbance was chosen as the limit that the background corrector could handle. P b was added to solutions of the five compounds to produce a concentration of 0.1 pg/mL, and each resulting solution was sampled by the wire and also by vaporization from the tube
wall for comparison. A P b solution without the matrix was also sampled for reference. The recoveries were calculated and are summarized in Table 11. Percent recovery is calculated as the integrated absorbance signal for P b in the presence of the matrix divided by the signal in the absence of matrix. Recovery with wire sampling was close to 100% for all the matrices except Na2S04and was always greater than for wall sampling. The recovery in some cases was better than we had found with the L'vov Platform ( 4 ) . We made a similar study of T1 interference by MgCl,, comparing wire sampling with wall sampling. T h e results are summarized in Figure 6. Integrated absorbance is plotted against increasing MgC1, concentration. Onset of matrix interference when sampling on the wall is between 0.001 and 0.002% MgCl,, and onset when sampling by wire is between 0.01 and 0.02%. L'vov ( 2 ) proposed that a part of the chloride interference is a gas phase phenomenon. For T1 a hotter tube environment would cause the equilibrium: TlCl
-
T1 + C1
to be driven to the right. We have investigated the interference of MgClz on T1 for several furnace tube temperatures.
2378
ANALYTICAL CHEMISTRY, VOL. 51, NO. 14, DECEMBER 1979 O30r
an environment at constant temperature. This method can be used to reduce the various matrix effects often observed when sampling a solution deposited on the wall. 2200'
u 0
020t
2700'
ACKNOWLEDGMENT
0
We acknowledge numerous helpful discussions with A. Walsh and B. V. L'vov.
LITERATURE CITED
0001 0002
OW5
001
002
005
01
02
05
(1) B. V. L'vov, L. A. Pelieva, and A. I. Sharnopolskii, Zh. Prikl. Spectrosk., 27, 395 (1977). (2) B. V. L'vov, Specfrochim. Acta, Part 6, 33, 153 (1978). (3) R. E. Sturgeon, Anal. Chem., 49, 1255A (1977). (4) W. Slavin and D. C. Manning, Anal. Cbem., 51, 261 (1979). (5) D. C. Gregoire and C. L. Chakrabarti, Anal. Chem., 49, 2018 (1977). (6) 6. V. L'vov, Specfrochim. Acta, 17, 761 (1961). (7) R. Woodriff and G. Ramelow, Spectrochim. Acta, Part 6, 23, 665 (1968). (8) D. Littlejohn and J. M. Ottaway, Analyst (London), 103, 662 (1978). (9) H. Massmann, Specfrochim. Acta, Part B , 23, 215 (1968). (10) 6. V. L'vov and L. A. Pelieva, Zh. Anal. Chim. (Russian), 33, 1572 (1978). (11) J. Janssen, Compf. Rend., 71, 626 (1870). (12) R . Mavrodineanu and H.Boiteux, "Flame Spectroscopy", Wiiey & Sons, New York, 1965. (13) V. P. Garnys and L. E. Smythe, Anal. Chem., 51, 62 (1979). (14) D. C. Manning and W. Slavin, Anal. Chem.. 50, 1234 (1978). (15) W. Slavin, S. Myers, and D. C. Manning, unpublished results.
0
% MgCk2
Figure 7. Effect of atomizing temperature on MgCI, interference of TI
absorption
Plotted in Figure 7 is the T1 absorbance as the MgC1, is increased, a t three temperatures: 1800, 2200, and 2700 "C. T h e absorbance signal for T1 is larger a t lower temperatures because the residence time is longer in the furnace. However the chloride interference is more severe a t lower temperature, as predicted. Introducing a sample dried on a wire is a convenient method for ensuring that the sample is vaporized and atomized within
RECEIVED for review June 27,1979. Accepted September 4, 1979.
Reduction of Argon Consumption by a Water Cooled Torch in Inductively Coupled Plasma Emission Spectrometry Guy R. Kornblum, Wouter Van der Waa, and Leo de Galan" Laboratorium voor Analytische Scheikunde, Technische Hogeschool Delft, Jaffalaan 9, 2628 BX Delft, The Netherlands
By adding a water cooled jacket to the conventional three-tube quartz torch, the argon consumption In inductively coupled plasma (ICP) spectrometry has been reduced 10-fold. Two designs are described and evaluated. In comparison with the conventional high-flow ICP, the plasma generated in the novel torch design shows a reduced background Intensity and comparable excitation temperature, dynamic range, and freedom from interferences. Limits of detection decrease with the sample carrier gas flow and presently range from 0.01 to 1 mg/L for a carrier gas flow of 0.1 L/mln.
T h e present state of the art of the inductively coupled radiofrequency argon plasma has been reviewed (1, 2). A disadvantage of all commercial instruments and nearly all research units is the large consumption of argon gas. Whereas only 1 L/min is used to carry the nebulized sample solution into the plasma, typically between 10 and 20 L/min are required to shield the outer quartz tube of the plasma torch from the hot plasma gas. Occasionally, a third argon flow (plasma gas) is used, especially for nonaqueous solvents. This large consumption of argon not only adds to the running costs of the ICP, but it creates the practical problem of regular supply. Continuous operation requires one 50-L gas tank per day and has induced some laboratories to install large units with liquid argon. Several proposals have been made to decrease the need for argon. A rather drastic solution is to replace the main argon flow by nitrogen, but this requires a much higher rf power (>5 0003-2700/79/0351-2378$01,00/0
kW into the plasma), which is undesirable for obvious reasons. With the more popular medium power rf generator ( 1 2 kW) Dagnall, Smith, and West (3) succeeded in a partial replacement of the main argon flow, so that their ICP ran on 7 . 5 L/min Ar and 7.5 L/min N2in addition to 0.5 L/min of argon as the carrier gas. Scott et al. ( 4 ) optimized the design of the three-tube configuration and utilized a total argon flow of 10 L/min. A similar approach was followed by Allemand and Barnes ( 5 ) ,who in a later publication (6) reduced the main argon flow by applying an angular gas introduction system that creates a so-called swirl flow. In this way the main argon flow could be decreased from 18 to 10 L/min in a commercial instrument and from 12 to 5 L/min in their own research unit. In all these attempts the total argon consumption is reduced by no more than a factor of two with a lower limit of about 6 L/min. The torch design described in this work permits a reduction with another factor of three to 2 L/min. Torch Design. Apart from sustaining the rf plasma, the main function of the large argon flow used in present ICP torches is to prevent the outer quartz tube from melting. I t is, therefore, commonly referred to as the coolant gas. Consequently, a substantial reduction in argon consumption can be realized, if the cooling is performed by other means. In the proposed torch design, this is realized by fitting a water jacket around the torch. For obvious reasons the inlet point of the water supply cannot be located within or on top of the rf coil. Hence both the inlet and the outlet point of the cooling water must be situated a t the bottom end of the jacket. In order to force the water toward the hot plasma region inside the rf coil, where 1979 American Chemical Society
