Anal. Chem. 1983, 55, 99-104
99
Characterization of Two Modified Carbon Rod Atomizers for Atomic Absorption Spectrometry Darryl D. Siemer” and Leroy C. Lewis Exxon Nuclear Idaho Go., Inc., Box 2800, Idaho Falls, Idaho 83402
A number of novel carbon rod atomizer conflguratlons were constructed and tested wlth the goal of Increasing the effective gas-phase temperatures experlenced by volatlle analyte metals. Both a “top-clamped” cup and a “tube-cup” atomlzer permitted Increases In the gas phase atomlzatlon temperature for lead and cadmium of more than a thousand degrees over those achlevable wlthi the two conventional CRA atomizer conflguratlons. Additionally, from 2 to 3 orders of magnitude more concomitant transltlon metal chloride salt can be accommodated wlth the Improved atomizer deslgns before matrlx effects are observed.
The determination of lead and cadmium by graphite furnace atomic absorption spectrometry (GFAAS) is complicated by a host of “matrix effects” not usually seen in flame AAS because their volati1i:zation temperatures overlap those of many of the compounds forming the bulk of many common sample matrices (e.g., transition or alkaline earth metal chloride salts). At thie relative low gas-phase temperatures actually present within furnaces of conventional construction during the time that the analyte vapor i s present, compound formation between covolatilized concomitants and the analyte is favored (1). Furnaces into which the sample is introduced after the optical path has been preheated to a relatively high temperature (e.g., the L‘VOVor Woodriff furnaces) are far less sensitive to matrix problems (2-4). The use of a “L’vov platform” in furnaces based on the Massmann design (e.g., the Perkin-Elmer HGA 2200, Varian Techtron GTA95, or the IL455/555/655 series) has been shown to retard evaporation of the analyte until tho tube wall is several hundred degrees higher than would otherwise be the case and, consequently, significantly reduces matrix problems (5, 6). However, owners of the Varian Techtron carbon rod atomizer (CRA) cannot use “platforms” because the atomizer’s tube diameter is too small (3 mm) to accommodate them. Two approaches to increasing the effective gas-phase temperatures in these furnaces have previously been investigated. The first involves heating a longer-than-standard atomization tube from the ends, not at the center, either with a pair of “Y”-shaped rods (7)or with four separate rods (8). The second approach, recently investigated in this laboratory, involves heating the “cup” configuration CRA atomizer with rods clamped across its top, not across the bottom as is usually the case (9). Although these approaches raised the gas-phase temperatures a few hundred degrees and significantly reduced the matrix problems in lead determinations, this writer felt that better results could be achieved with the basic CRA furnace design if the system were redesigned with the achievement of high gas-phase temperatures as a primary goal. Accordingly, this paper describes some of the experimentation done and characterizes both “cupl” and “tube-cup” designs featuring gas phase temperatures for lead and cadmium approximately loo0 degrees higher than are usually observed in thermally pulsed GFAAS furnaces. EXPERIMENTAL SECTION The modifications made to the Varian Model AA6 spectrometer and to the Model 63 CRA furnace power supply have been dis0003-2700/83/0355-0099$0 1.50/0
cussed in previous papers (10, 11). The important differences between them and the standard equipment include considerably “faster” analog electronics in the signal processing circuitry and the addition of an optional temperature feedback controller to the furnace power supply. A Thermodot optical pyrometer interfaced with the same Hewlett-Packard computer system used to collect the atomic absorption transient signal data permitted simultaneous recording of the temperatures of a selected point on the atomizer’s surface (IO). Modifications to the Model 90 CRA workhead were kept to a practical minimum in order to minimize duplication costs. First, the stack of alternating flat and corrugated stainless steel plates which are normally used to provide a laminar flow of inert gas to protect the hot graphite was replaced; instead a 4 mm thick aluminum plate pierced with 48 evenly spaced, 1.4 mm diameter (no. 54 drill) holes was used. This was done to give clearance to atomizer cups when they were lowered with respect to the center line of the rods, Second, a shield was fabricated from 2 mm thick aluminum sheet metal to loosely enclose the hot graphite working parts of the atomizer (Figure 1). This shield is very effective in retarding the back flow of air which invariably burns away the uppermost surface of both the atomizer and the rods in the standard, open-configuration workhead. Holes were drilled in the shield to accommodate the light beam and to provide access to the atomizer by the micropipet used to deposit the sample aliquots. Finally, the HEP 312 phototransistor used to provide the temperature feedback signal to the CRA power supply was mounted in a small aluminum block; this block was in turn screwed onto the rear electrode block of the workhead. A 25-mm long piece of ceramic tubing of the type usually used to shield thermocouple wires was used as a light pipe-collimator and heat shield for the transistor. The upper end of the ceramic tubing was inserted into a short piece of opaque, black, plastic tubing and the phototransistor was glued into the other end. The rest of the temperature feedback controller was the same as described previously (11). To conserve argon, a homemade gas control box featuring an electronic timer and a solenoid gas valve was put in series with the inert gas line from the standard Model 63 gas control system. It is triggered “ON”at the initiation of the “ASH”cycle by a signal from the CRA power supply. An argon flow rate of 8 L/min was used. A low-pressure propane line was connected to the inlet side of the automatic gas solenoid of the standard gas control system. (The soldnoid valve was originally designed to permit the addition of hydrogen to the inert gas flow during the “ATOMIZE”cycle.) The occasional addition of propane (at a flow rate of 0.5 L/min) during an atomization cycle renews the surface of the hot graphite working parts of the atomizer with a layer of pyrolytic graphite and (in combination with the aluminum shield) extends their useful life to at least several weeks of continual use. All of the data presented in this paper were obtained with atomizer components coated “in situ” with pyrolytic graphite before the experiments were performed. The standard pyrolytic graphite Varian Techtron cups were subjected to the same pretreatment for the sake of consistency of temperature measurements made with the optical pyrometer. The appearance of the commercial graphite coating is different than that observed with the ”in situ” coating process and, presumably, the emissivities of the two surfaces are different. An inexpensive optical feedback control module compatible with Varian M 63 power supplies is commercially available (L. L. Elektronik, Department of Analytical Chemistry, University of Umel, S-90187, Umel, Sweden). Its sensitive phototransducer permits accurate control of “ASHING” as well as “ATOMIZE” Q 1982 American Chemical Soclety
100
ANALYTICAL CHEMISTRY, VOL. 55, NO. 1, JANUARY 1983
E
Flgure 1. Drawing of the modified CRA 90 workhead depicting the aluminum gas shield and the mounting for the temperature transducer.
Figure 2. Some of the atomizer configurations tested. The arrows
point to where the rods are clamped.
temperatures. The same supplier can also provide a gas control box similar to the homemade version outlined above. In most of this work, standard 4.6 mm diameter Varian Techtron CRA rods were used. All nonstandard graphite components were machined from a dense, nonporous 6.2-mm diameter spectrographic quality graphite rod (FXSI from POCO Graphite Inc., Decatur, TX). This material is quite strong without being brittle and can be readily machined to very close tolerances. RESULTS AND DISCUSSION The mean effective gas phase temperature within the atomizer while a typical volatile analyte metal is present was conveniently measured by comparing the relative integrated atomic absorbance signals observed at two prominent lead nonresonance lines. The first line, 368.3 nm (gf = 0.64), originates from a state 7819 cm-l above ground; the second at 280.2 nm (gf = 5.1) originates from an energy level 10650 cm-’ above ground. The spectroscopic constants for these lines are taken from ref 12. These lines are a good choice for these measurements because the relative magnitudes of the two gf values implies that at the temperatures of interest (1200-2800 K) to this study an identical aliquot of the same lead standard solution will give measurable signal peaks at both wavelengths. However, in this study sequential signal measurements at each wavelength were used for the spectroscopic temperature calculation. The 261.4 nm (nonresonance) and 283.3 nm (resonance) line pair that Ide et al. (13) used for a similar purpose is less satisfactory not only because the tremendous difference in sensitivity between those two lines requires two standard solutions but also because the 261.4-nm ‘‘line’’ is actually a doublet which renders its absorbance signal vs. mass of analyte response curve nonlinear when measurements are made with normal AAS equipment. The optical system used measured light passing through a round 2.4-mm hole in a baffle situated as closely as possible to the optical access hole through the atomizer on the side facing the monochromator. This optical beam is only slightly smaller than the access hole itself so all of the absorbance measurements (as well as any spectroscopic temperatures derived from them) reflect mean values across the access hole’s diameter. No attempt was made to spatially resolve these values either ucross the access hole or along the length of the atomization zone. The following equation was used to calculate the spectroscopic temperatures in the furnace:
where 4077 and 0.217 are constants which combine the relative
Seconds
Figure 3. Absorbance signal transients seen with the standard, bottom-clamped, cup atomizer configuration. wavelengths, gf values, and lower state energies of the two lines, A2m.2 is the absorbance signal/g of lead at 280.2 nm, and A368.3 is the absorbance signal/g of lead a t 368.3 nm. Similar equations may be derived for other line pairs (e.g., the 283.3-nm resonance line and 368.3-nm nonresonance line pair) but their use is somewhat less convenient because different solutions must be pipetted for the two atomic absorption response measurements. The approximately 5% coefficient of variation of integrated atomic absorbance signals measured at the nonresonance lead lines results in an imprecision in the calculated spectroscopic temperatures of from 30 to 100 K over the range of temperatures observed. Of course, the accuracy of the measurements is determined by the accuracy of the gf values used in the calculation. For the purpose of comparing atomizers getting exact temperatures is not as important as is consistency of approach. In the course of this project approximately 20 different atomizer designs were built and tested. Length, diameter (inside and out), total mass, wall thickness, etc. were all varied. Figure 2 depicts a few of the various atomizers examined drawn to a common scale. In this figure the arrows point to where the center line of the rods were clamped with respect to the tube. In order to save space, only those experiments that definitely led to a clearer understanding of how these atomizers should be designed will be described in the rest of this paper. First, a complete description of the experiments done to characterize these atomizers will be presented using the standard cup-configuration CRA atomizer as an example (Figure 3). The rods were clamped across the bottom of the cup as is recommended in the manufacturer’s literature. In this figure, the atomic absorbance signals seen a t the two
ANALYTICAL CHEMISTRY, VOL. 55, NO. 1, JANUARY 1983
nonresonance lines (368.3 nm larger peak and 280.2 nm smaller peak) when 2000 ng of lead was atomized are depicted with solid lines. The dashed line is the signal seen from 2.5 ng of lead observed at the 283.3-nm resonance line. All of the figures (numbers 3-6) depicting actual instrument responses seen with the various atomizer configurations use the same format to portray the atomic absorption signals seen at the three lines. The temperature of the inner lip of the cup (recorded at the same level as the optical access hole) as measured by the optical pyrometer is alsio shown. A constant emissivity value of 0.7 was assumed for all measurements so recorded throughout the entire study. This value may not be accurate but since the object of the project was to compare different atomizers, the relative temperatures observed satisfy the intended purpose. In this experiment, the power supply was used in the standard mode (Le., no temperature feedback control) at its maximum heating rate (i.e., "10" on the front panel dial) for 1.5 s. This very short heating period was used for two reasons. First, all of the lead has left the atomizer within that time and, second, by the time thiat 2 s has elapsed, the bottom of the cup reaches a temperature sufficiently high to rapidly destroy the cup (by sublimation of the graphite). Also the use of the maximum possible heating rate ensures that the spectroscopic temperature measured represents the highest achievable with that atomizer and power supply combination. Several important conclusions can be drawn from an examination of Figure 3. First, the temporal overlap of the three atomic absorption transients (i.e., the shapes of the peaks) while not identical with one another is quite similar. This means that temperature measurements based on the relative size of the two nonresonance line transients are valid to use in describing the temperatures experienced by the groundstate atoms of interest in conventional GFAAS analytical work. This is not unexpected because the shapes of these transient signals should be largely determined by the supply function (71 or the rate at which the lead is volatilized) and not by the residence time (72) of individual atoms within the cup in atomizers of this design (14). In Massmann furnaces possessing large and varying (with respect to time) thermal gradients along the optical axis, the nonresonance line signals tend to be skewed to the right (i.e,, to higher temperatures) with respect to the reigonance line responses (15). Next, the apparent spectroscopic temperature calculated as explained previously is 1445 K-a couple of hundred degrees higher than is the temperature of the graphite at the same distance from the bottom of the cup at the time that the maximum signals appear. The reason for this is that the cross section of the cup is so large that the rapidly expanding gas does not achieve thermal equilibrium with the walls of the cup across the entire diameter of the cup. Since, in this case, the bottom of the cup is at a higher temperature than is the top, the gas passing upward within the cup is at a somewhat higher average temperature than are the walls of the cup at the same level. This has practical relevance in the design of the cup-type atomizers because it indicates that in order to achieve the highest possible gas-phase temperatures within a given region along a tube (i.e., at the optical access hole), the power should be applied at a point along the tube that the gas must first pass to reach the critical region. Figure 4 gives the results of a similar experiment performed with the same cup-rod combination but with the rods now clamped across the top, not the bottom, of the cup. The ends of the rods were machined to a radius of 2.8 mm instead of the usual 2.5 mm in order to accurately fit the larger upper diameter of the cup. The homemade temperature controller was used with the light collimator of the transducer directed at the upper lip of the (cup. The power supply heated the top
t
0.6{
1
2
3
101
2900
900
Seconds
Figure 4. Absorbance signal transients observed with the rods clamped across the top of a standard CRA atomizer cup.
of the cup at the maximum possible rate until that part of the cup reached approximately 2400 K and then held that temperature constant. Only 250 ng of lead was needed in this case to get the two nonresonance lead signal transients (depicted with solid lines-the peak at 368.3 nm being the larger) but the same 2.5 ng of lead was used for the resonance line signal (dashed line). Note that in this case, the sizes of the signals per gram of lead seen at the two nonresonance lines are much larger than they were in the last figure and also that the ratio of the size of the peak seen at the 280.2-nm line to that at the 368.3-nm line has increased. This, of course, reflects the much higher mean gas phase temperature experienced by the lead within the optical path-2145 K as compared to the 1445 K observed with the standard atomizer configuration. However, the magnitude (area) of the resonance line signal is only about 90% as large as that observed with the standard atomizer configuration. This reflects the higher diffusion coefficients that lead atoms possess at the greater gas temperature present with the second cup-rod configuration. However, the signal is not attenuated as much as might be expected by a consideration of the difference in the atomic diffusion coefficients alone; this indicates that the convective expulsion of volatile analyte atoms by the rapidly expanding argon gas strongly affects atomic residence times (7J in (at least) the first set of experimental conditions. Experiments performed with cup D in Figure 2 (quite long with uniform cross sections both inside and out) involving changing of the position of the rods with respect to the optical access hole, indicated that the highest possible useful gas-phase temperatures are achieved with rods clamped so that their center line was approximately 1mm below the center of the hole. The position of the rods determines the temperature differential between the top and the bottom of the cup, both during the initial rapid heating which occurs when the power is first applied and after the entire cup reaches thermal equilibrium (after a few seconds). With a sufficiently long cup clamped at the top, it is possible to achieve a stable gas-phase temperature in the light path of 3000 "C before the lead at the bottom starts to volatilize. However, a practical cup atomizer must be capable of being heated at the bottom to a sufficiently high temperature to prevent the accumulation of the less volatile common matrix concomitants as the cup is used for repeated analyses of real samples. For the purpose of this project, one of the desired design criteria for the atomizer was an eventual temperature at the point upon which the sample is initially placed of 2000 K. This temperature was chosen because it is at least a couple of hundred degrees higher than the "appearance temperatures" of most of the elements (e.g., Na, K, Fe, Mg, Cu, Sn, Ca, etc.) that make up
102
ANALYTICAL CHEMISTRY, VOL. 55, NO. 1, JANUARY 1983
the bulk of the common sample matrices in which lead is apt to be determined (2). There seems to be little to be gained by designing atomizer cups with two different diameters-inside or out. At first thought, a cup featuring a small diameter, low “dead gas volume” sample well at the bottom should give better analytical sensitivity (cup C, Figure 2). The reason for this is that the convective effect of the expanding argon gas carrying the volatilized analyte out of the light path should be lessened. However, the actual sensitivity experimentally achieved was not appreciably better than with a cup of uniform cross section (e.g., cup B, Figure 2). The reason for this is probably a greater loss of analyte vapor through the walls of the cup due to a higher incidence of wall impacts. A serious disadvantage of cups featuring small sample wells is that only tiny sample aliquots can be dried without forceful ejection of the droplet by the steam generated (“bumping”). Cups of extremely small internal diameters (less than 2 mm) offer slightly greater absolute sensitivities (assuming equal wall thicknesses and lengths) than do those close to the standard (3 mm) size. However, the improvement is very slight and sample capacity drops rapidly as the diameter of the cups is reduced. The standard internal dimension is a good compromise. An external cup diameter of 5 mm (0.20 in.) is convenient because standard CRA rods fit cups of that size without modification. Experiments performed with homemade rods with ends 6.2 mm in diameter (the size of standard spectrographic quality graphite rod stock) did not indicate that bigger rods are worth the additional trouble that it takes to machine and install them. Somewhat more current is drawn from the power supply because the cup-rod contact resistance is lower with the larger rods. However, the additional heating power is largely lost through increased thermal conduction down the rods themselves. The optical access hole can advantageously be made somewhat larger than the 2.4 mm diameter hole used in the standard Varian Techtron CRA cups. A variation in these hole sizes from 2.0 to 2.8 mm in diameter in a single cup (cup B, Figure 2) caused no significant change either in gas-phase temperatures or in the analytical sensitivity achieved with the atomizer. The larger hole sizes, however, cause less trouble with vignetting of the light beam by the edge of the hole when the atomizer expands upon heating-a practical advantage. The length of the cup should be sufficient to give an adequate temperature differential between the top and the bottom a t the time that the analyte starts to volatilize-but not longer. Excess length reduces the final temperature achieved a t the bottom of the cup (causing matrix buildup) and increases the dead volume of the atomizer causing dilution of the volatilized analyte and therefore reduced analytical sensitivity. Typical results obtainable with a good overall compromise cup atomizer design are depicted in Figure 5. The CUP used (cup D, Figure 2) has the following dimensions: 11.0 mm overall length; 5.02 mm outside diameter; 3.1 mm internal diameter (no. 31 drill); 9.91 mm depth; 2.64 mm diameter optical access hole (no. 37 drill) centered 2.34 mm below the rim of the cup. A narrow ridge 5.3 mm in diameter and 0.80 mm wide was left at the top of the cup to give a reproducible rod-cup geometry when the cup is pressed down into place. The signal responses depicted are those from 100 ng of lead for the nonresonance lines (solid traces), 2.5 ng for the resonance line (dashed trace), and 5 x g of cadmium (dotted line). Note that the signals for both lead and cadmium appear after the top of the cup has reached the temperature setpoint of the power supply (in this instance 2400 K). The spectroscopic temperature based on the relative size of the nonre-
0.8
,
I
I
Czsoo
Seconds
Figure 5.
Absorbance signal transients observed with an optimized
CRA cup atomizer.
sonance line lead signals was 2339 K. The temperature of the inside bottom of the cup as measured by the optical pyrometer (lower temperature trace) reached the desired 2000 K in approximately 4 s. In this one measurement the emissivity setting used with the pyrometer was 1.0 because the geometry approximated that of a “black body” radiator (which fact makes the nature of the graphite surface less important in determining its radiative properties). The cup atomizer used to generate the signals depicted in Figure 5 achieved the main goals of this project, i.e., effective gas-phase temperatures for volatile metals similar to those observed in an air-acetylene flame while simultaneously preventing matrix accumulation in the cup. However, the absolute sensitivity of the atomizer (based on the relative sizes of the integrated atomic absorbance signals per gram of analyte) is approximately five times worse than can be achieved with the standard tube atomizer (using matrix-free standards). In order to combine the relatively high gas-phase temperatures achievable with the optimized cup design with the superior sensitivity of the tube-type atomizer, the “tube-cup” atomizer depicted as E in Figure 2 was constructed. Its design philosophy was such that when the analyte volatilized from the cup it would have to pass through a preheated optical path (the tube) far longer than the internal diameter of the cup itself in order to escape from the light beam. Of course, this raises the total fraction of analyte atoms within the light path a t any one time and consequently increases the analytical response. Its physical dimensions were as follows: tube length, 13.6 mm; outer tube diameter, 5.08 mm; tube inner diameter, 2.81 mm (no. 34 drill); sample access hole, 1.40 mm (no. 54 drill); hole for the cup opposite the pipet access hole, 3.66 mm (no. 27 drill); cup length, 7.4 mm; cup inner diameter, 2.81 mm (no. 34 drill); cup depth, 7.0 mm; outer diameter of cup, 3.66 mm (an interference fit into the hole drilled into the tube). After the tube-cup atomizer was constructed, it was clamped into the workhead exactly as a standard CRA tube atomizer would be. It was heated to 2400 K in a 16:l argon/propane sheath gas atmosphere for 6 s. This was repeated a total of five times. The resulting layer of pyrolytic graphite coated the entire surface of the atomizer and effectively sealed the cup to the tube. Figure 6 shows the results obtained when the same series of experiments performed with the cup in Figure 5 was repeated with the tube-cup atomizer using an identical power supply setting. The tracings on this figure indicate the same experiments as in Figure 5 except that the resonance line signal for lead (dashed line) represents that seen with 1 ng (not 2.5 ng) of lead. The overall effect of the change in design is a &fold increase in analytical sensitivity while retaining a high effective gas-phase temperature (2436 K in this case).
ANALYTICAL CHEMISTRY, VOL. 55, NO. 1, JANUARY 1983
t
3900
Seconds
Flgure 6. Absorbance signal transients observed with the tube-cup atomizer.
f
-$ C
-3
4!
-1
0
1
2
3
I . o g l o p g Added CuClz
Flgure 7. Integrated atomic absorbance signals per nanogram of lead as a function of the amount of copper chloride added. In order to assess the practical value of these new atomizer designs in ameliorating gas phase related matrix problems, we prepared a series of Cu(I1) chloride solutions containing, 0.1, 1,10, and 100 pg of the salt per 5-pL aliquot. This salt was chosen for illustrative purposes because its severe depressive effect on lead (GFAASsignals is known to be caused by gas phase reactions between the lead and the chlorine released when the readily dissociated copper salt is covolatilized with the lead (1). Figure 7 shows the integrated signals per nanogram of lead obtained with four different atomizer configurations as a function of the amount of copper(I1) chloride added. No “Ashing” heating stage was used and the lead standard solution used contained no added acid (which would have acted as a “Matrix Modifier” during the drying step). Curve A is that obtained with the standard CRA cup configuration (rods across the bottom) atomizer with power supply settings of “8” for 3 s (no temperature feedback control). The second curve (”B”) was generated with a standard tube configuration CRA used with a power supply setting of “7” for 4 s. Curve “C” represents the signals obtained with the top-clamped homemade cup used as described in the discussion of Figure 5. The last curve (ID) shows the results seen with the “tube-cup” atomizer previously described. Both the tube-cup alnmizer and the homemade cup evince little matrix effect until the amount of copper chloride covolatilized with the lead exceeds 40 pg-approximately 3 orders of magnitude more than can be accommodated by the popular standard tube configuration CRA atomizer ( I ) . Even at the 100-pg concomitant level enough of the lead signal remains (60-80%) that analyses by the standard addition technique are still possible. Similar experiments revealed that neither iron(II1) nor magnesium chloride salts gave a serious signal perturbation until the amount covolatilized exceeded 40-100 pg. A study was performled of the effect of magnesium chloride on thallium signals; this study was similar to that done by Manning et al. (16) in their evaluation of a Massman-type
103
furnace modified so that samples were introduced on a tungsten wire after the tube itself had achieved thermal equilibrium. The chlorine-metal bond is considerably stronger between thallium and chlorine than between chlorine and lead. This, of course, implies that even higher effective gas-phase temperatures are needed to eliminate the matrix effect. The “tube-cup” atomizer described above gave results similar to those achieved by Manning et al., i.e., a 50% suppression of analytical response occurring at a MgClz concentration of approximately 0.02 % . To further increase the effective gas-phase temperature obtainable with the “tube-cup” atomizer, we shortened the tube portion of the atomizer to 9.5 mm (from 13.6 mm) and tapered the outer 2 mm of each end to a diameter just slightly larger than that of the hole itself. These modifications reduced the mass of the atomizer to 0.225 g (from 0.360 g) and consequently increased the rate of heating achieved with the power supply used. With this atomizer a stable gas-phase temperature of 2700 K is readily achieved during the entire time that thallium is in the optical path. The magnesium chloride concentration necessary to reduce thallium signals by 50% was greater than 4% with this atomizer. This concentration is much greater than that which the furnace used by Manning et al. could accommodate even when it was used at a nominal temperature of 2700 OC (16). The reason for this “tube-cup’’atomizer’s superior performance is probably that it does not have relatively cool tube ends through which the analyte and covolatilized matrix must pass in order to escape from the optical path. Of course in actual analytical practice, an analyst using either of these atomizers should take advantage of the benefits to be gained by using any convenient “matrix modification” and/or “Ashing” sample pretreatment steps known to be beneficial prior to performing the actual atomization step. However, what the improved atomizer designs accomplish is to make those sample preparation steps far less critical. Therefore, these atomizers reduce the difficulty in getting reliable analytical values with samples of unknown or “difficult” composition. An additional power supply modification incorporating a second temperature sensing transducer directed at the bottom of the atomizer cup in order to permit positive control of the temperature during the “Ashing” heating stage would be very useful for practical analytical work. Experiments with all of the atomizer configurations used indicated that use of the fastest possible heating rate is essential if the desired large temperature differential between the optical path and the sample deposition point is to be achieved. This implies that some form of feedback control of the maximum temperature should be employed to prevent over-heating of the graphite surrounding the point at which the rods are clamped. The potential heating rate achievable with the standard CRA M90 power supply is considerably greater than that of the CRA 63 used for this project (because the main power transformer has a rating of 3 kW as opposed to only 1.5 kW); however, the actual maximum heating rate (with standard CRA tubes) achieved with the M90 system is only about half as great as that gotten with the earlier model. The M90 power supply can be adjusted to permit more rapid heating ramp rates (see the CRA M90 service manual, adjust RV 104, to accomplish this). However, in any case, the “temperature” meter on the front of the power supply will give only a rough approximation of the actual graphite temperature because the resistances of the unconventional atomizers are not the same as those of the standard components. The modified atomizers described in this paper were designed, first, to minimize the serious gas-phase matrix effects encountered in the GFAAS determination of volatile analyte
104
Anal. Chem. 1983, 5 5 , 104-110
metals and, second, to be compatible with commercially available CRA workheads and power supplies with a minimum of hardware changes. The maximum volatilization temperatures achievable with them are too low for the determination of refractory analytes so the standard atomizer configurations must be used for metals less volatile than copper. The principle followed in the design of these CRA atomizers is similar to that used by Littlejohn and Ottaway in their paper describing a Massmann tube furnace optimized for atomic emission analyses (17). Presumably then their furnace would evince lessened gas-phase matrix effects when used for absorbance measurements and, conversely, the modified CRA atomizers should be superior to the standard components for the atomic emission analysis of relatively volatile metals. LITERATURE CITED (1) Czobik, E. J.; Matousek, J. P. Anal. Chem. 1977, 50, 2-10. (2) L’vov, B. V. Spectrochlm. Acta, Part 6 1978, 338, 153-193.
(3) Hageman, L.; Mubarak, A.; Woodriff, R. Appl. Spectrosc. 1979, 33, 226-230. (4) Hageman, L.; Nichols, H. A.; Vlswandham, P.; Woodriff. R. Anal. Chem. 1979, 51, 1406. (5) L’vov, B. V. Spectfochlm. Acta, Part 8 1978, 336, 153. (6) Savin, W.; Manning, D. C. Anal. Chem. 1979, 51,261-265. (7) Lawson, S. R.; Woodriff, R. Spectrochim. Acta, Part 6 1980, 356, 753. (8) Siemer, D. D. Anal. Chem. 1982, 54, 1659-1663. (9) Siemer, D. D. Appl. Spectrosc., in press. (10) Siemer, D. D.; Baidwin, J. M. Anal. Cbem. 1980, 52 295. (11) Siemer, D. D. Appl. Spectrosc. 1979, 33, 613. (12) Corliss, C. H.;Bozman, W. R. N6S Monogr. ( U S . ) No. 53, 289. (13) Ide, Y.; Yanagisawa, M.; Kitagawa, K.; Takeuchi, T. J. Spectrosc. SOC. Jpn. 1975, 24, 1435-1437. (14) Van der Broek, M. J. T.; de Gaian, L. Anal. Chem. 1977, 49, 2186-2186. (15) Sturgeon, R. E.; Berman, S.S.Anal. Chem. 1981, 53, 632-639. (16) Manning, D. C.; Slavin, W.; Myers S. Anal. Chem. 1979, 51, 1375-2378. (17) Littiejohn, D.; Ottaway, J. M. Anal. Chim. Acta 1979, 107, 139-158.
RECEIVED for review July 8, 1982. Accepted October 7, 1982.
Synthesis of the 38 Tetrachlorodibenzofuran Isomers and Identification by Capillary Column Gas Chromatography/Mass Spectrometry Thomas Mazer,” Fred D. Hileman, Roy W. Noble, and Joseph J. Brooks Monsanto Research Corporation, Dayton Laboratory, 1515 Nicholas Road, Dayton, Ohio 454 18
The 38 positional isomers of tetrachlorodibenzofuran have been synthesized by pyrolysis of specific polychlorinatedbiphenyl congeners, ultraviolet photolysis of pentachiorodibenzofurans, and chlorination of trichlorodibenzofurans by aromatic substitution. The specificity of these reactions in combination with caplilary column gas chromatography with mass spectrometric detection has allowed each of these isomers to be identified based on their relative elution order.
In recent years the polychlorinated dibenzofurans [PCDFs] have been the subject of an intense research effort owing to their structural similarity to the polychlorinated dibenzo-pdioxins [PCDDs] (1). Particular attention has been given to the tetrachlorodibenzofurans [TCDFs] for which there are 38 positional isomers, yet no analytical scheme allowing identification of specific TCDF isomers has yet been reported. In this paper we report on the synthesis of the 38 TCDF isomers by oxidative pyrolysis, under carefully controlled reaction conditions, of specific PCB congeners (2,3), ultraviolet [UV] photolysis of pentachlorodibenzofurans [PenCDF], and chlorination by electrophilic aromatic substitution of specific trichlorodibenzofuran [TrCDF] isomers. Characterization of the TCDF isomers was accomplished by high-resolution gas chromatography/mass spectrometry and UV photolysis ( 4 ) . The final result has been the development of an analytical scheme allowing for the analysis of 2378-TCDF (notation for symbols excludes the commas necessary in full names). EXPERIMENTAL SECTION Caution. Persons attempting to synthesize these compounds should first familiarize themselves with their safe handling and disposal.
PCB Congeners. All the PCB congeners used in this study were obtained from Ultra Scientific, Inc., Hope, RI, with the exception of 2,2‘,3,4‘- and 2,3,3’,4’-tetrachlorobiphenyl, 2,2’,3,3’,6-pentachlorobiphenyl, and 2,2’,3,3’,4,6’- and 2,2’,3,3’,5,6’-hexachlorobiphenyl which were received from C. A. Wachtmeister, Stockholms Universitet, Wallenberglaboratoriet and 2,3,3’,4,5S-10691 Stockholm, 2,3,4,4/-tetrachlorobiphenyl pentachlorobiphenyl which were received from J. Pyle, Miami which University, Oxford, OH, and 2,2/,3,4’,5,6-hexachlorobiphenyl was received from M. D. Mullins, EPA, Grosse Ile, MI. The purity of the PCB congeners was generally greater than 97% assuming an equivalent FID response for all PCBs. In those cases where a given PCB congener was contaminated with another PCB congener, pyrolysis often yielded small amounts of PCDF isomers other than those expected. In all cases these were readily discernible from the desired PCDF isomer. PCB Pyrolysis. Miniglass ampules 5-6 cm in length were prepared by sealing the large end of disposable borosilicate glass Pasteur pipets (Model 13-678-20C,Fisher Scientific Co., Cincinnati, OH). Ten microliters of a solution of the PCB in hexane [ 10 kg/mL] was placed in the ampule and the solvent allowed to evaporate. The tip of the ampule was then flame sealed and the ampule placed in a large vial along with a cold junction referenced chromel-alumel thermocouple which was connected to a digital voltmeter to allow accurate temperature measurements. The ampule and thermocouple were placed into a muffle furnace [Type 1500, Thermolyne Corp., Dubuque, IA] operated at 600 “C. When the temperature of the ampule reached 550 “C, pyrolysis was allowed to continue for an additional 5 s. The ampule was removed from the furnace and allowed to cool to ambient temperature at which time it was opened and the contents thoroughly rinsed out with 1 mL of hexane. In some cases unreacted PCBs were removed by subjecting the pyrolysate to chromatographic separation on a minicolumn of Woelm Basic Alumina (ICN Pharmaceuticals, Cleveland, OH) by eluting the PCBs with 10 mL of 2% methylene chloride in hexane, and then eluting the retained PCDFs with 15 mL of 50%
0003-2700/83/0355-0104$01.50/00 1982 American Chemical Society
