2050
Anal. Chem. 1982, 5 4 , 2050-2053
Double-Walled Furnace for Reduction of Matrix Interferences in Graphite Furnace Atomic Absorption Spectrometry Stephen R. Lawson’ and Ray Woodriff * Department of Chemistry, Montana State University, Bozeman, Montana 59717
A double-walled furnace was constructed which consists of an 18 mm long, 3 mm id. furnace tube held inside an 18 mm outer sleeve. The furnace tube Is lathed down 0.5 mm over the central 14 mm of length such that only 2 mm on elther end Is In dlrect physical contact wlth the outer sleeve. When held wlth the standard CRA supports, electrlcal current Is forced down the length of the outer sleeve and crosses only at the last 2 mm of elther end. This “double-walled”furnace was found to be spatially Isothermal during the atomlratlon of Pb or Co. Up to 10-fold Increases in the recovery of Pb and Co In chloride and sulfate salts compared to the CRA-63 furnace were obtalned. Comparison was made tq an endheated furnace uslng two-pronged supports developed earller.
Matrix interferences in electrothermal atomic absorption spectrometry (ETAAS) arise from changes in appearance temperature, rate of analyte evolution, and/or gas phase reactions which bind the analyte in molecular form thereby reducing the atomic absorption signal. Variations in the first two parameters do not affect the integrated absorbance signal if the sample is vaporized into a constant temperature environment (1). Gas-phase interferences can be substantially reduced or eliminated (depending on the matrix concentration) provided the temperature and residence time are both sufficiently large (2). Some recent work with pulse-type furnaces has therefore been aimed at delaying sample vaporization until the entire furnace is at least equal to or greater than the appearance temperature of the element of interest (3-6). The L’vov platform has accomplished this for larger diameter furnaces (-6 mm i.d.) like the Perkin-Elmer HGA-500 where the sample-containing platform is heated primarily by radiation from the walls (3-5). A different approach was required for the smaller diameter (3 mm i.d.) Varian CRA-63 furnace due to space limitations. Two-pronged supports, developed a t this laboratory and described in detail previously, have been utilized to restrict electrical current to cross only at the furnace ends (6). Analyte atoms and molecules always encounter a region hotter than the surface from which it vaporized before leaving the optical path. Longer f u r e c e tubes gave better recoveries of lead in chloride matrices since larger temperature differences between the tube center and ends coyld be achieved. The longest tubes that could be used were 18 mm since the atomizer head can only provide an inert gas sheath up to this length. Chakrabarti et al., using the extremely rapid heating rates (-60 K/ms) from a capacitive discharge across pyrolytic graphite, obtained a furnace which was isothermal in space and time (7). Recent studies have shown a marked reduction in matrix interferences with this system as long as the temperature was sufficiently high at the moment of vaporization (8). This technique represents a compromise between the earlier constant temperature furnaces of Woodriff and L’vov and the more popular small, pulse-type furnaces produced commercially. Present address: Department of Chemistry, Carleton University, Ottawa, Ontario, Canada K1S 5B6.
Sotera et al. encountered fewer interferences when the IL254 FASTAC autosampler was used to deposit samples in an IL (Instrumentation Labs) furnace compared to manual pipetting (9). The FASTAC autosampler deposites the sample solution as an aerosol onto the hot (420 K) furnace walls. The solution dries on contact leaving a uniform layer of fine crystals. The authors propose that this method of sample deposition overcomes sources of interference due to (1)matrix-dependent spreading of the sample droplet, (2) variable seepage of the solution into the graphite layers, and (3) formation of large crystals which can trap the analyte and be expelled from the furnace before decomposing. Definitely, more work needs to be done on this system to clearly elucidate the mechanism of the improvements. The purpose of this study was to test an alternate design for heating the ends of a graphite furnace first. The twopronged arrangement was limited to medium- and high-volatility elements because the center was only heated by end conduction. The center lagged behind the ends by 700 K when the ends were at 2500 K. A more direct source of heat input to the center was needed. The two-pronged supports were also less convenient to install on the CRA-63 atomizer head than the standard supports because one of the support holders had to be removed. This necessitated partially dismanteling the atomizer. Good electrical contact between the furnance and supports requires rotating the furnace tube by hand while pressure is applied from the forked holders. A double-walled furnace in which electrical current can cross only at the edges of an outer sleeve has been constructed and characterized. The above mechanical problems encountered with the two-pronged supports were overcome with this design while maintaining comparable recoveries in several matrices. The standard CRA supports are used, therefore no changes in operating procedures are required other than slightly higher voltage settings for a given temperature due to the increased mass and resistance. Studies were conducted on Co as well as P b in the presence of NaC1, MgClz, CaClZ,CuClZ,FeCl,, ZnC12, and ZnS04. Recoveries obtained for the CRA-63 and both design modifications are presented and compared. We have found the most informative approach to characterizing the double-walled furnace to be a comparison between furnaces. Primarily, the end-heated and double-walled furnaces are extensively compared in the following discussion. We have therefore included any data pn the end-heated furnace necessary for this discussion. ‘
EXPERIMENTAL SECTION Double-Wall Furnace. Interference studies were performed on the Varian CRA-63 atomizer with the standard pyrolytically coated 9-mm tube heated frpm the center using the standard supports. For comparison, an identical 9 mm tube (3 mm i.d., 5 mm 0.d.) was machined from Poco (Decatur,TX) AXF-5Q grade graphite. Eleven millimeter and 18 mm furnaces were therefore machined from the AXF-5Q graphite. Details of the 18-mm double-wall furnace are provided in Figure 1. The resistance of the double-walled furnace (0.14 3 ) was greater than both the end-heated (0.10 3) and CFL4 (0.07 3)furnaces due to the pressure contact between the tube and outer sleeve. This region did not, however, cause any problems with arcing.
0003-2700/82/0354-2050$01.25/00 1982 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982
Table I. Comparison of Recoveries Obtained for 480 pg of Pb Measured in Three Different Furnacesa 0.48 pg of 0.48 ug of 2.4 p g of 2.4 pg of 24 ug bf ZnCl, MgC1, ZnC1, NaC1 MgCl, 88 4 54+ 5 78f 5 62 f 6 12 11 mm (double walled) 98f 4 803: 5 76k 5 l o o k 4 94f 6 18 mm (double walled) 68+ 5 47 k 6 0 5 68 f 6 CRA-63 103 f 6 100 f 5 go+5 102f 5 871t 6 18 mm (end heated)
11 mm (doublle walled) 18 mm (doublle walled) CRA-63 18 mm (end heated)* a
2.4 pg of CUCl,
0.48 pg of CUCl,
2.4 pg of CaCl,
2 78 f 6 2 67+ 8
75+ 6 100 f 4 17+5 981 5
77 f 5 100 + 4 101 8 113 f 4
1.41 pg of FeCl, 14 94+ 6 0 70+ 8
0.48 pg of CaCl,
2.4 pg of ZnSO,
78f 5
66+ 6 67 f 5 27 k 8 go* 7
loo* 3 69 f 9
100 f 4
2051
Best recoveries obtained for end-heated furnace: 1.1K/ms at center; 1.9 K/ms at the ends.
Peak area measurements.
SLEEVE
AI1 meawrcmonts are
*
mm
EXPANDED SIZES FOR SLEEVE
TIME (S)
Flgure 2. Relative temperatures profiles of double-walled furnace operated at 0.4 K/ms: (a)output of pyrometer focused on the inside wall at one end of the furnace; (b) output of pyrometer focused through sample port: (c) 480 pg of Pb.
TOP
c
SLEEVE
CORE
-a=
1 TOM
t
Flgure 1. Cross-sectional views of double-walled furnace.
A light baffle containing a 1.5 mm diameter hole was used in all measurements to prevent excessive amounts of furnace light
from reaching the photomidtiplier. Another baffle (4 mm diameter hole) was placed on the opposite side of the atomizer head. This formed a box around the furnace which significantlyreduced the amount of oxidation experienced by the ends of the longer 18-mm furnaces. Power Supply. A 5-kW transformer was connected in parallel with a Varian CRA-63 power supply for Co recovery measurements and for testing the effect of rapid heating rates on Pb recoveries. Only the CRA-63 power supply was utilized for the lead recovery data obtained from the double-walled furnace presented in Table I. The parallel connection allowed the dry #andash steps to still be performed with the CRA supply while the atomization was initiated with a manually thrown double pole switch connected to a high voltage (208 V) variable transformer. The variable transformer controlled the voltage on the primary windings of the 5-kW step-down transformer. Atomization voltages up to 18 V and maximum currents of 180 and 130 A were recorded on the end-heated and double-walled furances, rerpectively. Optics and Electronics. The light source was a Westinghouse lead hollow-cathode lamp operated at 6 mA. A dual channel monochromator constructed in this laboratory (10) was operated at the 283.3-nm lead resonance line and a band-pass of 0.45 nm for all lead measurements. The cobalt 240.7-nm line at a bandpass of 0.45 nm was used for all cobalt measurements. The detector was a Hamwnatsu R212 photomultiplier operated at 890 V. The PMT signal was fed into an Ithaco (Ithaca, NY) lock-in amplifier with a tirne constant of 10 111s. This output was connected in parallel to an Omniscribe (Austin, TX) strip chart
recorder (for peak height measurements), a logarithmic amplifier followed by a V-F converter and counter (for peak area recording), and a Tektronik 564 dual trace storage oscilloscope. The oscilloscope trace was triggered by the power supply voltage level produced at the onset of the atomization stage. No background corrector was used. Separate measurements of background absorbance were made with aliquots of aqueous solutions of each matrix studied. The 283.3-nm line from the lead hollow cathode lamp was used for all background measurements. Temperature Measurements. The temperature of the inner surface of the atomizer was monitored with a phototransistor. Measurements at the furnace center were made by focusing through the sample port. Temperatures at the furnace ends were measured over an equal area of the inside wall. Details of the electronics and focusing optics have been described in an earlier paper (6). The pyrometer system was calibrated by focusing on the tungsten filament of a General Electric ribbon filament pyrometer (No. 431-P-732) calibrated between 1600 and 2400 K at 0.664 pm. Additional measurements below 1600 K were obtained from optical pyrometer (Leeds & Northrup, No. 1858692, Philadelphia PA) readings on the tungsten filament. The temperature vs. voltage plot obtained by this procedure was useful for heating rates and center-end temperature differences in a furnace. Dry and ash temperatures were measured with a Chromel-Alumel (type K) thermocouple. Reagents and Operating Conditions. Standard matrix solutions were prepared from reagent grade salts and doubly distilled water. The lo00 kg/mL lead and cobalt reference standards were prepared by dissolving the metal in distilled reagent grade nitric acid. Aliquots (2.4 pL) of either a 200 ng/mL lead solution or a 400 ng/mL cobalt solution were used for all measurements. Solutions were dried for 30 s at 120 " C and ashed for 15 s at 380 "C.
RESULTS AND DISCUSSION The spatial temperature distribution throughout the double-walled furnance was far more uniform than that observed with the two-pronged support configuration. The temperature curves in Figure 2 illustrate the small temperature difference
2052
ANALYTICAL CHEMISTRY, VOL. 54, NO. 12, OCTOBER 1982
Table I1
-% recovery
doublewalled furnace
18 mm end-
18 mm end heated
heated furnace
94
70
81+ 7 302 8 82i 6 100 i 5
94
55
78
67
75
21
18 mm
480 pg of
Pb
+
1.41 p g of FeCI, (graphite) 480 pg of Pb -t 1 . 4 1 pg of FeCl, (pyrolytic coating) 480 pg of Pb + 2.4 p g of CuCl, (graphite) 480 pg of Pb t 2 . 4 p g o f CuCl, (pyrolytic coating)
Table 111. Comparison of Recoveries Obtained for 960 pg of Co Measured in Three Different Furnaces
between the furnace center and the ends 600 ms after power was applied (0.4 K/ms). This difference did not appreciably increase at an applied power of 2.4 kW (1.7 K/ms). In either case, the furnace becomes spatially isothermal before a volatile element such as lead enters the light path. The center 14 mm of the 18 mm inner tube is heated primarily by radiation from the outside sleeve. This was confirmed from temperature measurements taken a t several positions along this central portion. A 6-mm region to each side of the sample port was heated with temporal and spatial uniformity a t all power levels studied. The mode heating of this sample-containing central portion resembles that of a L’vov platform. However, upon leaving the atomizer surface, atoms encounter a region which is isothermal with the surface it left rather than hotter. Despite the absence of a higher temperature region through which atoms must diffuse before leaving the optical path, the double-walled furnace reduces chloride interferences to a slightly greater degree, in some cases, than the end-heated furnace. The only sizable differences in recovery between the two furnaces (see Table I) occurred when FeC1, or ZnSO, were added to lead nitrate. In the presence of 0.06% FeCl,, 94% recovery was obtained by using the double-walled furnace compared to 70% with the end-heated furnace. Although not a large improvement, such differences can provide information on furnace design features which are most beneficial. Two main features of the double-walled furnace distinguish it from the end-heated furnace: (1)near spatial isothermal heating and (2) a central zone where the sample is deposited which heats uniformly inward from the outside surface. The latter effect may influence solid phase reactions and/or the temporal concentrations of gas-phase matrix products. Unstable transition metal chlorides like FeC1, and CuCl, decompose with loss of chlorine at low temperatures: 688 K for FeC1, and 480 K for CuClz (11). Since the sample solution can soak into the porous ordinary graphite used in this study, the liberated chlorine may become trapped or intercalated within the graphite layers. This could delay or prolong the time during which interfering species are present in the furnace. To test the effect of ordinary graphite on the signals observed, we pyrolytically coated both furnaces as described by Siemer et al. (12). Only the inner tube was coated on the doublewalled furnace. The results listed in Table I1 reveal no change in recoveries from either CuC12or FeC1, when lead is atomized from a pyrolytically coated double-walled furnace. However, in the end-heated furnace, recovery of lead in CuClz decreases by a factor of 3. The decreased soaking of solution into the furnace wall apparently alters analyte-matrix surface reactions
18 mm CRA-63 double walled interferent
50i 7 25i 7
32i 7 88+ 6
101 t 5 68i 5 85t 5 100 4
*
FeCl, CUCI, MgCl,
CaCl,
and/or the temporal overlap between gas-phase products. This change is only important for the end-heated furnace. For comparison, the recoveries of cobalt in FeCl,, CuCl,, MgCl,, and CaC1, were measured in both furnaces (Table 111). Recovery of cobalt in CuClz is lower by a factor of 2 in the end-heated furnace than in the double-walled furnace. The only other measurable difference in recovery occurs when FeC1, is used as the matrix. Thus a similar pattern is observed between furnaces for the interference of FeCl, and CuC12 on a high volatility (Pb) and medium volatility (Co) element. Whether or not heating the furnace wall from the outside (as in the double-walled furnace) is the key factor remains unknown. Further study is still required. A similar difference exists with the Pb-ZnSO, system. Vaporization of a solution of 200 ppb lead in 0.1% ZnSOd in the end-heated furnace yields a 90% recovery, while only 67% of the lead signal is recovered from the double-walled furnace. This can be attributed to the 700 K difference in temperature between the hot ends and cooler center at the moment the lead absorbance reaches a maximum. With 1040 K as the lead appearance temperature (13), this corresponds to -1700 K at the furnace ends when the absorbance maximizes. Hageman et al. (2) obtained a 96% recovery in a CTF at 1900 K for lead in 1% Na2S04. This indicates that higher temperatures should remove the remaining interference. Longer residence times probably are not required. Frech et al. (14) using a theoretical treatment on the Pb-Na2S0, system in a HGA-74 furnace under isothermal conditions concluded that the reactions which formed PbS were rapid enough to reach equilibrium (7 = 200-300 ms). The equilibrium partial pressure of oxygen was reached at 1800 K but not at 1500 K. Manning and Slavin (4)noticed the same persistence of interference effects from sulfates, phosphates, and perchlorates using the L’vov platform. The difference in temperature between the platform and the wall was a t most 300-400 K which explains the lack of full recovery. Achieving higher atomization temperatures is straightforward with a constant temperature furnace but can become a difficult problem in pulse-type furnaces especially with high volatility elements. Both the L’vov platform and end-heated furnace, in their present forms, are limited in the magnitude of the sample-furnace temperature difference obtainable by their heating rates. The delay in sample heating inherent in their design relaxes the required rates somewhat from those needed by Chakrabarti et al. (8) using capacitive discharge heating. Heating rates up to 63 K/ms were studied, but 14 K/ms eliminated the interference of 2.5 pg of MgClz on 1.0 ng of Pb. In contrast to the end-heated furnace, recoveries in the double-walled furnace did not improve with an increased heating rate over the range of rates available: 0.4-1.8 K/ms. This is because of the nearly isothermal condition developed over the length of the furnace. Thus, heating rates closer to those used in the capacitive discharge techniques may be required to eliminate some matrix interferences. For this to be done, the outer sleeve on the double-walled furnace would have to be increased from the 0.5 mm thickness currently used to 1.0 mm to withstand the additional stress from faster heating. At rates of 1.0-1.8 K/ms, the heat and pressure
2053
Anal. Chem. 1982, 5 4 , 2053-2056
LITERATURE CITED
exerted on this section during the atoinization cycle cause it to gradually bend inward. The curvature remains after cooling and eventually becomes large enough to allow arcing between the inner and outer walls at the furnace center. The sample heats first and interferences suddenly become severe. Increasing the wall thickmess to 1.0 mm increases the useful life of the sleeve from 5-43 cycles to about 20 a t 1.5 K/ms. The inner tube has a life of twice this due to the additional shielding from oxygein that the outer sleeve provides.
L’vov, B. V. Spectrochim. Acta 1961, 17,761. Hageman, L.; Nichols, J.; Viswanadham, P.; Woodriff, R. Anal. Chem. 1979, 5 1 , 1406. L’vov, B. V. Spectrochim. Acta, P a r t 6 1978, 336, 153. Slavin, W.; Mannlng, D. C. Spectrochim. Acta, Part 8 1980, 358, 701. Slavin, W.; Mannlng, D. C.: Myers, S. Anal. Chem. 1979, 51, 2375. Lawson, S. R.; Woodriff, R. Spectrochim. Acta, Part 6 1980, 356, 753. Chakrabartl, C. L.; Hamed, H. H.; Wan, C. C.; Li, W. C.; Bertels, P. C.; Gregoire, D. C.; Lee, S . Anal. Chem. 1980, 52, 167. Chakrabartl, C. L.; Wan, C. C.; Hamed, H. H.; Bertels, P. C. Anal. Chem. 1981, 53, 444. Kahn, H. L.; Conley, M. K.; Sotera, J. J. A m . Lab. (Fairfield, Conn.) 1980, Aug, 72. Woodriff, R.; Shrader, D. Appl. Spectrosc. 1973, 27, 181. Duval, C. “Inorganic Thermogravimetrlc Analysis”, 2nd ed.; translated by Oesper, R. E.; Elsevier: New York, 1963. Siemer, D. D.; Woodriff, R.; Watne, B. Appl. Spectrosc. 1974, 2 8 , ..
(CONCLUSIOM This study of a double-walled furnace demonstrated the reduction in gas-phase interferences which can be obtained by simply increasing the residence time and maintaining a spatially uniform temperature. Heating the end first should decrease any temperature lag of the gas within the furnace. Although this factor is considered small (-50-100 K) in the CRA due to the small internal diameter (15),it might contribute to the gas-phme reactions considered here. Use of the double-walled furnace in combination with matrix vaporization in the ash cycle and use of H, in the sheath gas during atomization can significantly extend the range of interference-free operation of a GFAAS system.
582. Sturgeon, R. E.; Chakrabarti, C. L.; Langford, C. H. Anal. Chem. 1978, 4 8 , 1792. Frech, W.; Persson, J.-A.; Cedergren, A. Prog. Anal. A t . Spectrosc. 1980, 3,279. van den Broek, W. M. G. T.; de Galan, L.; Matousek, J. P.; Czobik, E. J. Anal. Chim. Acta 1978, 100, 121.
RECEIVED for review August 24, 1981. Resubmitted and accepted June 17, 1982.
Treatment of Replicate Measurements in Kinetic Analysis Gregory R. Philllps, Joel M. Harris,, and Edward M. Eyring” Department of Chemistry, University of Utah, Salt Lake City, Utah 84 112
The common practice1 in kinetic studles of replacing replicate measurements by thelr average cause8 a loss of lnformatlon about the rellablllty of calculated rate constants as well as the agreement between experimental data and assumed reactlon mechanism. An approach to the use of the lnformatlon present in replicates as check the adequacy of an assumed mechanism and to prevent the unnecessary lengthenlng of Confidence intervals Is dlscussed, along wlth the clrcumstances under which such measures are applicable.
In recent years, kinetic methods have become a widely accepted analytical technique, the applications of which include the investigation of complexation, acid-base, and enzyme reactions. Guilbault ( I , 2) and Mark and Rechnitz (3)have reviewed the practical aspects of kinetic analysis, and more recent results have been surveyed by Mark and co-workers (4, 5 ) . The successful application of lkinetic techniques to analytical problems is facilitated by an understanding of the reaction mechanism involved. This knowledge helps establish optimal conditions for an analytical determination. In the development of kinetic methods, a simple but useful aid in the interpretation of kinetic data seems to have been overlooked. Appropriate treatment of replicate kinetic measurements provides information about the adequacy of an assumed reaction mechlanism that is not otherwise available. In addition, the confidence intervals for kinetic parameters derived from experimental data by this method are better estimated. A common practice in determining a kinetic mechanism is to make replicate measurements of reaction rates at each of
a series of concentrations, average the rates at each concentration, and then determine rate constants from these mean values. Such a procedure is both unnecessary and wasteful of information, information which is not replaced by measurements at a larger number of different concentrations. The repeat measurements are especially useful since they permit a quantitative and objective test of the postulated mechanism.
THEORY In general the reaction rate, Ri, measured at a series of reactant concentrations, ci, depends on rate constants, k , equilibrium constants, K , and the particular concentrations used
Ri = f(k,K,cj) (1) Since any measurement involves error, the observed relationship is
+
Ri = f(k,K,ci) ei (2) where f(k,K,ci)is the expected or “true” value and eiis an error term. It is generally assumed that the concentrations are known without error, and thus only the rates, Ri, contain error terms. In practice, this is the situation when the relative errors in the concentrations are much less than those in the rates. In studies where this is not the case, more complicated statistical techniques are necessary (6). This work concerns itself with relationships between the rates and concentrations which are linear in the rate constants. Such a relationship can be expressed in vector notation (7) R=Xp+e
(3)
where B = (R1,..., R,J’ is an n-dimensional vector of observed rates, B = (PI, ..., PPI’ is a p-dimensional vector of unknown
0003-2700/82/0354-2053$01.25/0 62 1982 American Chemical Society
