Anal. Chem. 1988, 60,761-766 (13) Shoukry, A.
F.; Badawy,
1078. -.
S. S.; Ism, Y. M. Anal. Chem. 1987, 59,
(14) Chatten, L. G.; Pernarowski, M.; Levl, L. J . Am. fharm. Assoc., Scl. Ed. 1059, 48, 276. (15) Badawy, S. S.; Shoukry, A. F.; h a , y. M. Analyst ( b n d o n ) 1986, 7 1 1 , 1363.
761
(16) Antropov, L. I. TheafetlcalElectrochemisby; Mir Publishers: Moscow, 1972.
RECEIVED
for review March 19, 1987. Accepted December 9,
1987.
Metal Oxide Cloths as Charring/Atomization Surfaces in Furnace Atomic Absorption Spectrometry Thomas W. Brueggemeyer* Elemental Analysis Research Center, U.S. Food a n d Drug Administration, 1141 Central Parkway, Cincinnati, Ohio 45202
Carlos A. Bonnin U.S. Food a n d Drug Administration, P.O. Box 5719, Puerta de Tierra Station, S a n Juan, Puerto Rico 00906
The feaslblllty of uslng metal oxlde cloths as atomlzatlon surfaces In furnace atomlc absorptlon spectrometry was studied. Y,O, cloths were Introduced Into a preheated graphlte furnace on a probe attached to the autosampler arm. Thls permltted evaluatlon of the new surfaces wlth little s y e tern modlflcatlon. Data on temperature dependence and sensltlvlty for Ag, Se, Cd, and Pb were obtalned. With Pb as the analyte, studles on preclslon and Interferences were performed and compared to data obtained by uslng a conventional furnace program and graphite platform atomlzatlon. The potentlai for uslng the cloths as dlsposable charring/ atomlzatlon surfaces In real sample analysls was examined by uslng undlluted whole blood and total parenteral nutrition solutlons as matrlces.
Most work reported in the field of furnace atomic absorption spectrometry has involved a limited range of atomization surfaces. Resistively heated grahite has been used in the majority of cases. In recent years a radiatively heated graphite platform proposed by L’vov (1)has received wide-spread use in interference control. The chemical conversion of a graphite surface to that of a metal carbide has been frequently employed, particularly for the determination of analyte metals tending to form refractory carbides with an untreated graphite surface. Some work has been done with various metal foils or wires as atomization surfaces. Slavin lists many of these applications in a review monograph (2). Due to their relatively good electrical conductivity, graphite and certain metals can serve not only as atomization surfaces but as current-to-heat tranducers as well. A feature common to all these surfaces is the fact that they are somewhat reducing in nature. They can, at least in theory, donate electrons to analyte-containing molecules as part of the atomization mechanism-an idea proposed by Aggett and Sprott (3). They may also serve to regulate the gaseous furnace environment through reactions in which they combine with molecular oxygen a t elevated temperatures. Holcombe and Droessler (4)and Frech et al. (5)have recently addressed the interactions between graphite and oxygen. Little work has been reported dealing with oxide atomization surfaces. Silica surfaces were used for the trapping and
subsequent reatomization of relatively volatile elements in flame atomic absorption spectrometry (6). Karwowska and Jackson recently described the graphite furnace atomization of lead from alumina particles. Undertaken as a model for atomization from soil slwies, the study showed that the form of the alumina greatly affected atomization behavior (7). A number of refractory metal oxides can withstand a temperature in excess of 2000 “C, which is sufficiently high for the atomization of the more volatile analyte elements. Unlike graphite and refractory metals, which generally react with oxygen at elevated temperatures, the metal oxides are normally stable in an oxidizing environment but are often susceptible to reduction or dissociation (8). The commercial availability of certain metal oxide materials in the form of woven or knitted cloths offered a convenient starting point for the examination of new atomization surfaces. Their large surface area (about 250 cm2 total surface per 1 cm2 of cloth) and thinness (approximately 0.4 mm) were expected to prove beneficial. In this study, it was decided to insert the cloth surface on a probe into a preheated graphite furnace tube. Atomization probes fabricated of graphite or metal have been described as ways to reduce interferences by utilizing the constant-temperature nature of the furnace tube (9-12). A convenient manner for introducing the probe is by attachment to the arm of an autosampler (9). When such a probe is configured to allow the rapid interchange of atomization surfaces, then the evaluation of these surfaces does not require major structural changes to the system. A system using interchangeable probes with external pretreatment has been suggested as a possible way to speed analyses (13). The work to be described here involves zirconia (ZrOP)and particularly yttria (Y20,)cloths for the charring/atomization of several relatively volatile metal analytes with particular emphasis on lead. EXPERIMENTAL SECTION Apparatus. A Model 5000 spectrometer, a Model 500 furnace system equipped with Zeeman effect background correction, and a Model AS-40 autosampler (all Perkin-Elmer Corp., Norwalk, CT) were employed. Data were acquired with a Perkin-Elmer Model 3600 data station. The commercial peak display software (DATA, version 4.0) was modified in this laboratory to generate peak heights and integrated areas, although areas only were used in this study. When samples were run in the conventional graphite furnace mode for comparison, a solid pyrolytic graphite L’vov
This article not subject to US. Copyright. Published 1988 by the American Chemical Society
762
ANALYTICAL CHEMISTRY, VOL. 60, NO. 8, APRIL 15, 1988
Table I. Spectrometric Conditions
element
wavelength, nm
slit, nm
Ag
328.1
0.7
Se
196.0 228.8
2.0 0.7
283.3
0.7
Cd Pb
integralamp power/cur- tion time, lamp" rent S HC1 EDL EDL EDL
10 mA 5.1 W 4.4 W 8.8 W
4 4 4
4
'HCl, hollow cathode lamp; EDL, electrodeless discharge lamp. Table 11. Platform Atomization Program
1
I
I. _ _ _4_ _ . _ .
- -1
,
rr
Flgure 1. Construction of probe from tungsten foil: (a) tungsten foil strip showing cuts and location of folds (dotted lines); (b) tungsten foil bent around silica tube with brass tube in posttion to be sild down; (c) finished probe as mounted in autosampler arm showing cloth in position.
platform was used inside a pyrolytically coated graphite furnace tube (Perkin-Elmer parts 121091 and 0290-1811, respectively). To accommodate the probe (which was mounted on the autosampler arm), the opening for sample injection in the left graphite contact ring was enlarged by using a small metal file. A rectangular slot 7.8 mm in the direction parallel to the light beam by 3.0 mm was cut from the outside of the ring through to the center. Because this slot was simply an enlargement of the injection opening at its center, it did not adversely affect the system when used in the conventional manner. The graphite tube itself (Perkin-Elmer part 0290-1821) needed modification in order to accept entry of the probe. Starting at the original injection hole, a small square metal file was used to cut a 7.2 mm X 1.9 mm rectangular slot in the tube. The 7.2 mm dimension of the slot was parallel to the central axis of the tube, and the slot was symmetrical about the original injection hole. The probe upon which the atomization cloths were placed was fabricated from 0.025 mm thickness tungsten foil (Alfa Products, Danvers, MA). The foil was cut to 6.8 mm X 25.4 mm with sciswrs. A 0.7 mm X 0.7 mm square was removed from one comer, and two diagonal cuts were made as shown in Figure la. The two folds shown were best made with the aid of a block of metal having a sharp 90O edge. This block was pressed atop the tungsten foil, leaving only the bottom 0.7-mm flap extending. This flap was then bent upward around the sharp edge of the block, producing a bottom ledge for holding the cloth. Repetition of the procedure produced the side edge. The tapered end of the foil (top) was bent around a 45 mm length of 2.29 mm 0.d. silica tube as seen in Figure lb. A 10 mm length of 0.125 in. 0.d. brass tubing (bored out slightly to slide over the silica tube) was then slid down as a sleeve holding the foil snugly in place (Figure IC).The entire probe was then snapped into the autosampler arm, replacing the sample delivery tubing. Because the tungsten foil is fragile,contact with the graphite ring or furnace tube can damage or destroy it. Thus, it was necessary to make careful adjustments of the autosampler translation sledge, the insertion depth stop, and the graphite tube to ensure that the probe moved freely into the furance and stopped about 1 mm from the inside bottom of the tube. CAUTION The probe shaft must be made of a material that is electrically insulating such as silica or a ceramic in order to prevent any possible current path between the furance and the operator or the autosampler. The cloths used as atomization surfaces were obtained as 1 ft2 sheets from Zircar Products, Inc. (Florida, NY). The zirconia (ZYK-15) and yttria (YK-15) cloths used were nominally 0.015 in. thick and had a tricot knit. A sharp razor blade and straight edge were used to cut cloths into the desired rectangular configurations. Generally, the cloths were cut to be approximately 5.9 mm (dimension parallel to the light beam) by 4.8 mm (dimension parallel to the probe arm). When resting on the probe,
step dry 1 dry 2 atomize clean
hold time, s
argon flow, mL/min
10
20 30
300 300
a
0
5
0
2600
1
3
300
temp, "C ramp time, s 120 300
20
"Atomize temperatures: Ag, 1700 "C; Se, 2200 "C; Cd, 1700 "C; Pb, 1800 "C.
a cloth did not extend beyond the edges of the tungsten foil. To dry/char sample-containing cloths in the muffle furnace, rhodium cloth holders were used. These were cut on a shear from 0.25 mm thick Rh foil (Alfa Products) and were approximately 2 mm in width and 25 mm in length. The upper end was cut at approximately a 30° angle, producing a "knife" for piercing the center of the cloth and holding it in place, yet not seriously damaging it. Being made of Rh, the knives could be cleaned in nitric acid between uses and resisted degradation in the muffle furnace. Each knife, with a cloth attached, was placed into a 10 mm deep, 2.3 mm diameter hole drilled into an aluminum block. After application of samples, the entire apparatus was then placed in a muffle furnace and covered with a glass beaker to exclude dust during the thermal pretreatment stage. Reagents. Argon (99.995% purity, Wright Bros., Cincinnati, OH) was used throughout the study as the graphite furnace purge gas. Deionized, distilled water with a specific resistivity of 18 MQ cm was obtained from a commercially available system (Millipore Corp., Bedford, MA). Double distilled nitric acid (GFS Chemicals, Columbus, OH) was used in dilutions. Commercial standards, 1000 pg/mL, for Ag, Se, Cd, and Pb (Spex Industries, Metuchen, NJ) were diluted in 1% nitric acid to individual 1 wg/mL standards. These 1w/mL standards were diluted on a daily basis, again with 1%nitric acid, to the appropriate level for injection onto the cloths or onto the platform: 50 ng/mL for Se and Pb, 5 ng/mL for Ag, and 2.5 ng/mL for Cd. In the interference study, reagent grade chemicals were used to make salt matrix solutions of the appropriate concentration. These solutions were then used in place of the 1%nitric acid for preparation of 50 ng/mL P b standards. Procedure. Table I gives spectrometric conditions used for the analyte elements regardless of whether atomization was from the probe/cloth or the platform. Table I1 gives the program used for conventional platform atomization. The same program was used for all four analyte elements, with only the atomization temperature differing. Maximum-power heating and argon gas interruption were used for atomization. Each cloth was given a preliminary thermal cleaning by carefully placing it upon the probe, rotating it manually into the furnace, and then applying a 15-s 2000 OC program. The probe was not rotated out of the furnace (and its argon environment) until at least 5 s after heating ended in order to prevent oxidation of the hot tungsten foil. In the repetitive mode, where a cloth was used for several firings, 5 MLof solution was added to the cloth as it rested upon the probe. The solution was dried in the furnace tube for 60 s at 300" C. In the batch drying/charring mode, each thermally cleaned cloth was mounted on a Rh knife. After the addition of 5 or 10 pL of sample to each cloth, the entire batch was placed in a muffle furnace set to ramp from ambient temperature to 400 "C in 10
ANALYTICAL CHEMISTRY, VOL. 60, NO. 8, APRIL 15, 1988 0 763 0. 15T 0.151
I
0 1 0 0. 15
0
-j.-";.---------
t
1700 OC
0
01
fl 0.
0
4
0
0 15
'I
4 T l m e
C s e c >
Figure 4. Atomization profiles resulting from 12.5 pg of Cd on yttria
cloth.
I O I
2000 O C
0
4
1
--:
I
T z m e
Time
(=eo>
Flgure 3. Atomization profiles resulting from 250 pg of Se on yttria
cloth. min and to hold at 400 "C for about 1h. After cooling, each cloth was transferred with plastic tweezers to the probe for its atomization step. The furnace program for atomization from the probe/cloth consisted of two steps. The first step was essentially a waiting period, 9 s in duration, allowing the furnace to achieve the selected steady-state temperature. Probe insertion took place during the second step, which lasted 11s at the same temperature. Argon flow was stoped during the second step to lengthen atom residence time and thus enhance sensitivity. Manual insertion of the probe was performed 5 s into step 2 as counted by the digital timer on the furnace controller and began with the probe poised about 30 mm above the furnace tube. Extremely rapid or jerky rotation of the autosampler arm had to be avoided to prevent the cloth from falling.
RESULTS AND DISCUSSION Atomization Peak Profiles. The atomization behavior of four volatile elements (Ag, Se, Cd, Pb) was investigated by use of zirconia (ZrOz)and yttria (Y203) cloths introduced into a preheated graphite furnace on a tungsten foil probe. Little difference was found between the performance of zirconia and yttria cloths for any of the parameters investigated. Because yttria was generally less contaminated with the analyte metals examined, and more readily cleaned by a preliminary firing, it was used for most of the work described. It can be seen in Figures 2-5 that an increase in the steady-state temperature of the graphite tube led to sharpening of the peak profiles. With the exception of selenium, the peak areas for the analytes did not vary significantly over the temperature ranges shown in Figures 2-5. At temperatures
-+
4
Figure 5. Atomization profiles resulting from 250 pg of Pb on yttria
cloth. lower than those shown, some peak area reduction was seen due to incomplete atomization of analyk-containing molecules. This can be seen for selenium a t 1800 "C. A base line distortion spike can be seen in each absorbance profile prior to the analyte peak. This resulted from the probe entering the furnace tube and thereby disrupting the light beam. This light blockage was treated by the Zeeman effect background correction system as a nonatomic background signal whose rapid rise time was too fast to allow total correction (14). The positive and negative components of this portion of the signal tended to cancel and made little net contribution to the integrated peak areas. The spike could be minimized and sometimes eliminated by adjustment of probe location in the tube and by lamp alignment. Because probe introduction was manual, i t was not well synchronized with the start of data acquisition. To compensate and thereby allow direct comparison of absorption profiles for various elements and different atomization temperatures, the data in Figures 2-5 were shifted left or right as necessary after plotting. This was done by aligning all the probe introduction spikes at the same point along the time axes, since each represents the time of probe entry into the furnace tube on that particular run. It can be seen in the figures that there was a progressively shorter delay between insertion and analyte appearance as higher temperatures were utilized. Because this delay includes the time necessary for the atomization surface to go from room temperature to the appearance temperature of the analyte, i t should depend upon a number of factors in addition to furnace temperature. Among these are the heat capacity of the probe/cloth in the furnace, the thermal conductivity of the
764
ANALYTICAL CHEMISTRY, VOL. 60, NO. 8, APRIL 15, 1988
Table 111. Sensitivity Comparison-Cloth vs Platform
element
atomization temp, O C
Ag
1700 2200
Se Cd
1700
Pb
1800
characteristic mass h). De platform cloth/probe 1.5 38
0.65 13
2.5 28 0.90 19
probe, and the absorptivity of the cloth surface toward black-body radiation emanating from the furnace tube walls. Sensitivity Comparison. Table I11 shows a comparison of sensitivities for atomization from the yttria cloth and conventional atomization from the solid pyrolytic graphite L'vov platform. The mo value, or characteristic mass, represents the amount of analyte, in picograms, which was found by extrapolation to give a peak area of 0.0044 A s. The actual analyte masses injected (in 5 p L of solution) were 250 pg of Se and Pb, 25 pg of Ag, and 12.5 pg of Cd. The smaller the characteristic mass, the greater the sensitivity. Both probe and platform mo values were obtained a t an atomization temperature giving optimal integrated area for the cloth. This was found to be within 100 "C of the ideal temperature for the platform. In some cases, the use of matrix modifiers would have led to improved platform sensitivity and quite possibly to improved sensitivity from the cloths as well. It can be observed that there is not a large change in analytical sensitivity with the probe/cloth system. When platform atomization signals for 250 pg of P b were examined, a reduction of peak area by about 35% was found for tubes in which the injection holes had been enlarged. This accounts for most of the sensitivity difference between probe/cloth and platform techniques. A shortened residence time and perhaps increased oxygen entry may be responsible. The fact that Se exhibited greater sensitivity from the cloth despite the tube modification is not readily explained. Pb Atomization. Beginning at this point, experiments involved only P b as the analyte metal. Unless otherwise indicated, solutions were added via micropipet to a cloth already resting upon the tungsten foil probe and then dried by manual rotation into the graphite furnace maintained at 300 "C for 60 s. In order to verify that P b atomization was taking place from the cloth and not from the tungsten foil beneath it, a yttria cloth to which 250 pg of P b had been added was removed from the probe after drying. Insertion of the bare probe into the 1800 OC furnace gave little P b atomization signal. The signal was recovered when the previously dried cloth was replaced on the probe and reinserted into the heated tube for atomization. In the repetitive mode where a cloth was reused for many firings, sample volume was limited to about 5 pL because larger volumes visibly spread from the cloth to the underlying probe surface. However, when the cloths were spiked with solution and dried externally before being placed on the probe, 10 p L of solution could be used. The repetitive mode was more convenient for most preliminary studies, while the batch mode-externally drying and charring a large number of cloths in parallel fashion-was more practical for real sample analysis. Figure 6 shows a series of peak profiles resulting from the atomization of varying amounts of P b from a yttria cloth using a furnace temperature of 1600 "C. Solution injection volumes of 5 pL were used. A correlation coefficient of 0.9991 was obtained up to an analyte mass of 2.5 ng using peak areas. Precision Comparison. Table IV compares precisions obtained by using a yttria cloth on the probe vs the L'vov platform with conventional atomization. It can be seen that at both the base-line and 250-pg levels the precisions were not markedly different for the platform and the cloth/probe
i'
-, = - n e
Flgure 6. Peak profiles from F% atomization off p i a cloth using 1600
"C furnace temperature. Five microliter solution volume. Table IV. Precision Comparison for Pb
yttria cloth
L'vov platform
N
base line furnace firing 5rL 50 ng/mL Pb
10
av area
std dev 5
av area std dev % RSD
insertion
0.0005 A s 0.0016
0.0033 A
0.0830
0.0026
0.0543 0.0024
3.1%
4.4%
s
0.0014
Table V. Relative Pb Peak Areas in Various Matrices YO
matrixa 1% "03 1%NaCl
0.2% CaCl, 0.25% MgClz 0.5% NaH2P04 0.5% KZSO4 0.5 Na2S04
wall
platform
cloth/probe
100 95 58 12 150 41
100
110
100 116 115
43
104 90
118
108
119
91 95
58 58
"Five microliters of 50 ng/mL Pb in indicated matrix; 1700 " C atomization. technique. Although the furnace system was equipped with an autosampler, injections onto the platform were performed manually in order to match the sample introduction used with the cloths. This ensured a more valid precision comparison. When cloth-to-cloth signal variation was compared to precision from a single cloth, no significant difference was seen. A 5.9 mm X 4.8 mm cloth weighing about 6.5 mg was the largest size compatible with the chosen probe and graphite tube configurations. Interference Study. Table V compares three atomization methods in terms of matrix interferences observed when 250 pg of P b was atomized at 1700 "C. The signals obtained in the indicated matrix have been expressed as percentages of the integrated area found for P b in 1%nitric acid using the particular method. No effort was made to remove matrix through a charring treatment-only a 300 OC drying stage was employed. The duplicate determinations generally agreed within 5%, and no P b contamination was seen in the unspiked matrices. The interferences were clearly worst with wall atomization and the least severe with platform atomization. It is not clear at this time why significant signal enhancements were seen in certain matrices when the probe/cloth method was used. The suppression observed in the sulfate matrices
ANALYTICAL CHEMISTRY, VOL. 60, NO. 8, APRIL 15, 1988
765
Flgure 7. Pb atomization profiles from undiluted whole blood on yltria cloths with muffle furnace charring and 1700 "C atomization temperature: solid line, blood: dotted line, blood 500 pg of Pb.
hibiting an enlarged surface area compared to a less porous charring surface. The TPN samples proceeded from colorless to brown to black to gray and finally to a white apprearance virtually indistinguishable from the original cloth. Whole blood samples progressed from red to brown to black and eventually faded to a faint tan-probably due to remaining iron oxide residue originating from hemoglobin. A 400 O C thermal pretreatment was not sufficiently hot to remove most inorganic matrix salts. It was not essential to continue charring until the cloths were nearly colorless. Equivalent results were obtained when cloths were gray in color. The completeness of charring on the cloths compared to the more conventional furnace aproach is probably a result of two factors. First, the spreading of sample over the many fibers of the cloth permitted more exposure to the atmospheric oxygen. Second, a much longer time was allotted to the thermal pretreatment stage (about 1 h or more, compared to perhaps 1 min in the conventional graphite furnace mode). This lengthy charring stage can be practical if a large number of sample-containing cloths are treated in a batch.
has been seen before. Manning et al. (9) reported a comparable degree of interference (33% signal lass) for Pb in a 0.5% sodium sulfate matrix atomized from a tungsten wire probe. Real Sample Analysis. The direct injection of certain undiluted fluids into the graphite furnace is troublesome because incomplete charring can lead to a carbonaceous buildup. In this laboratory, whole blood and total parenteral nutrition (TPN) solutions have caused this problem. T P N solutions are supplied intravenously to individuals unable to digest normal food and consist of high levels of several nutrients: up to 25% glucose and 4% amino acids, as well as various electrolytes, essential trace elements, and vitamins. When 10 p L of an undiluted T P N solution was pretreated conventionally on a L'vov platform a t 500 "C with oxygen during the charring, 45% of the total carbon injeded remained on the platform after charring. Even after atomization, 20% by weight remained as residue. Such buildup causes light beam blockage and inhibits reproducible sample introduction. The dilution of such samples-particularly with surfactant solutions-may prove effective in these cases. As an alternate approach, the idea of disposable charring/atomization surfaces for difficult fluids seemed compatible with the properties of the metal oxide cloths. An undiluted whole blood sample and a TPN solution were each analyzed by using the batch charring approach in a 400 OC muffle furnace. Standard additions were used to calculate the lead levels. Before being charred three of the six cloths used for each matrix were injected with 5 pL of sample alone, and the other three with sample containing an additional 100 ng/mL Pb. The integrated area vs concentration slope obtained in the blood matrix was 10% greater than with aqueous standards. The slope in the T P N matrix showed 7% suppression. Relative standard deviations of approximately 5% were obtained for the lead concentrations in each matrix. A lead level of 102ng/mL was obtained for the unspiked blood. For comparison, three 1-mL aliquots of the same blood were dry ashed using a typical muffle furnace procedure and the ash subsequently dissolved in 4% nitric acid. Graphite furnace atomic absorption spectrometry (GFAAS) determination of P b using L'vov platform atomization yielded a level of 107 ng/mL with a 10% relative standard deviation. Figure 7 shows the resultant peak profiles when whole blood samples were atomized from a yttria cloth at 1700 "C. Intermittent observation of the cloths during the charring procedure revealed that shortly after the drying of the samples, the charring of organic substances began. Examination of partially charred cloths under a microscope showed that the darkened carbon was spread evenly over the cloth fibers, thereby ex-
The data presented suggest a guarded optimism concerning the use of yttria cloths as atomization surfaces in furnace atomic absorption spectrometry. Precision and linearity were acceptable in comparison to conventional operating procedures. In the case of Pb, there was a loss of sensitivity as evidenced by a somewhat larger characteristic mass value, m,,, There was also some increased base-line noise due to light beam blockage by the probe. Somewhat surprisingly, this visible increase in noise level did not manifest itself in a deterioration of run-to-run base-line precision (see Table IV). This noise was reduced through probe adjustment within the light beam. Complete elimination of beam blockage may entail replacing the convenient autosampler arm with a dedicated cloth insertion mechanism. The current procedure limits solution volume to about 10 pL for real samples. The modification of the system to accommodate larger cloths and subsequently larger sample volumes should improve solution detection limits. The overall throughput of the procedure is not currently acceptable for routine analysis. The method involves the preparation and handling of small, fragile cloths, which is a difficult series of tasks to automate. The prefiring of cloths to remove contamination is necessary in the case of Pb, but may not be for less ubiquitous metals. Attempts to acid-leach P b from cloths proved unsuccessful. The enlargement of the injection hole in the graphite tube into a rectangular slot damaged the pyrolytic graphite coating. This in turn led to more rapid graphite deterioration near the slot. Recoating of the tube prior to use with a pyrolysis gas like methane should be possible. These experiments have shown that while graphite is normally the surface of choice for trace metal atomization, other surfaces are worthy of consideration in certain cases. This probe technique that has been described permits the preliminary investigation of novel surfaces without demanding a complete redesign of the existing commercial system. The idea that an atomization surface can be disposable may remove certain constraints previously faced in an analysis. If the charring and atomization conditions yield precise, accurate signals, then the condition of the surface afterward is not important-a fresh surface is used for the next run. Thus, residual matrix will not be able to accumulate over a series of firings. Carbonized organic material or refractory inorganics can remain on the discarded surface as long as analyte atomization is complete. Registry No. Yz03, 1314-36-9; Ag, 7440-22-4; Se, 7782-49-2; Cd, 7440-43-9; Pb, 7439-92-1; ZrO,, 1314-23-4.
.:..
0
1,d L, ...
.................., ...... , .... ... . , . ...
.,
,
a
3 T i m e
C s e c >
+
CONCLUSIONS
Anal. Chem. 1988, 60,766-773
766
LITERATURE CITED (1) L'vov, B. V. Spectrochlm. Acta, Part B 1978, 338, 153. (2) Slavh, w. Qaph/& F~~~~~~AAS A sourcem k ; The perkinzlmer Corp.: RMgeHekl, CT, 1984; pp 33, 43. (3) Aggett, J.; Sprott. A. J. Anal. Chlm. Acta 1974, 72, 49. (4) Holcombe, J. A.; Droessier, M. S. Fresenlus' 2.Anal. Chem. 1988, 323,689. (5) Frech, W.; Lindberg, A. 0.;Lundberg, E.: Cedergren, A. Fresenius' 2. Anal. Chem. 1988, 323,716. (6) Lau, C.; Held, A; Stephens, R. Can. J . Spectrosc. 1976, 21, 100. (7) Karwowska, R.; Jackson, K. W. J . Anal. At. Spectrom. 1987, 2, 125. (8) Campbell, I. E.; Sherwood, E. M. Mgh Temperature Materials and Technology; Wiiey: New York, 1967; Chapter 8.
(9) Manning, D. C.: Slavin, W.; Myers, S. Anal. Chem. 1979, 51, 2375. (IO) Manning, D. C.; Slavln, W. Anal. Chim. Acta 1980, 118, 301. (1 1) Girl, S.K.; Littlejohn, D.; Ottaway, J. M. Anawst (London) 1982, 107, 1095. (12) Giri, S. K.; Shields, C. K.; LlttieJohn,D.: Ottaway, J. M. Anawst (London) 1983, 108, 244. (13) Holcombe J. A.; Kolrtyohann, S.R. Spectrochim. Acta, Parf B 1984, 398, 243. (14) Harniy, J. M.; Holcombe, J. A. Anal. Chem. 1985, 5 7 , 1983.
for review August l4, 1987. Accepted October l 2 9 1987.
Detection of Flowing Samples with a Selective Concentration Gradient Method Janusz Pawliszyn'
Department of Chemistry and Biochemistry, Utah State University, Logan, Utah 84322-0300
Refractive index gradients generated in the flowing medium are measured with the help of a single probing light beam (Schfleren optlcs). These gradients are associated with any solute present In the medium. Addmonaily, excitaUon energy is introduced to the detection volume. After the relaxation process, the temporal temperature gradlent relates closely to the concentration gradient of the absorbing solute only. Both universal (concentration gadlent) and selective (temperature gradient) information associated with the solute can be acquired by monitoring the deflection of the probe light beam in nanoiiter detection volume. These data can be utilized to differentiate between samples on the basis of their molar absorption coefficients. A one-dimenslonai mathematical model was developed to descrlbe the heat transfer assodated wlth the selective concentratlon gradient method in flowing streams. Theoretical results closely correlate with experimental data. This method has been appHed to capillary liquid chromatography detection, characterization of freshness of fruit Juices, and contamlnatlon of organic solvents (fuets)wlth water. The detection ilmlt of the selective concentratlon gradient method is about 6 X lod absorption unlt for 1-mJ excitation pulses. Thk corresponds to a fraction of fentomole of congo red detected In a few nanollter volume.
In many modern analytical methods, samples are passed through capillaries or other low-dispersion systems. This approach is necessary to reach the technique's highest efficiency. The solutes in these methods are transported as narrow bands, characterized by high concentration gradients. Therefore, low-volume detection schemes are required to characterize the distribution of samples in these flowing streams (1-3). Recently the Schlieren optics detection scheme was proposed for the high-efficiency applications mentioned above (4). In this method the signal magnitude is proportional to the concentration gradient providing high sensitivity. Only Present address: Department of Chemistry, University of Waterloo, Waterloo, ON, Canada N2L 3G1.
a single light beam is required to measure the gradient, allowing small volume detector designs. The powerful feature of this simple method is that it can simultaneously provide both universal and selective information about the sample in an inexpensive design. The interesting concept of simultaneous universal and absorption detection based on the single monitoring scheme was first explored by Woodruff and Yeung ( 5 ) . In their arrangement they used a Fabry-Perot interferometer (concentration method) to monitor changes of the refractive index associated with sample and temperature rise produced during photothermal experiment. However, the concentration gradient approach and Schlieren optics detection are much simpler and less expensive. The universal mode of this technique has already been investigated experimentally (6). Any solute with a refractive index (RI) different from the solvent can be detected in this method. RI gradient generates a deflection signal proportional to the concentration gradient and the sample concentration itself for over 4 orders of magnitude. The detection limit of this method can be correctly given in refractive index gradients units: 104/m for a 200-pm detector cell. However, for a given detector design and experimental conditions, the detection limit can be expressed in concentration units. The conditions used in ref 6 resulted in molar detection limits of this method about the same order of magnitude compared to the best refractive index detector (5). However, the detection volume can be as small as a few nanoliters, which gives mass limits of only a few picograms of such substances as sucrose. The concentration gradient method is able to effectively distinguish between sharp peaks and broad variations in the refractive index due to gradient elution conditions. Enhancement in sensitivity of this method can be accomplished by expanding the effluent from the capillary into a larger detector with the help of a sheath flow method. In this paper, the selective mode of the concentration gradient is investigated. Firstly, a one-dimensional theoretical model for the selective mode of the concentration gradient method is developed. It describes the expected transient deflection signals associated with the thermal relaxation of the system after the excitation pulse. These signals carry the selective information about solute. Secondly, the predictions are compared with experimental results, and a number of
0003-2700/88/0360-0766$01.50/00 1988 American Chemical Society
