Metal Speciation by Supercritical Fluid Extraction with Online

Metal Speciation by Supercritical Fluid Extraction with Online Detection by Atomic Absorption Spectrometry. Jin. Wang, and William D. Marshall. Anal. ...
0 downloads 0 Views 874KB Size
Anal. Chem. 1994, 66,3900-3907

Metal Speciation by Supercritical Fluid Extraction with On-Line Detection by Atomic Absorption Spectrometry Jin Wang and William D. Marshall' Department of Food Science and Agricultural Chemistty, Macdonald Campus of McGill University, 2 1 , 1 1 1 Lakeshore Road, Ste. Anne de Bellevue, Quebec, Canada HOX 3V9

A silica flame-in-tube interface is described for the sensitive detection, by AAS, of As, Cd, Cu, Mn, Pb, Se, or Zn in supercritical (SC) fluid extractor eluate. In operation, analyte metal in aqueous medium is derivatized by in situ complexation with tetrabutylammonium dibutyldithiocambamate (TBADBDTC) and the product complex is mobilized into S C - C O 2 . The extraction process is equally efficient whether the mobile phase is presaturated with complexing agent during a static preequilibration stage of the process or this reagent is solubilized dynamically as the mobile phase traverses a saturation vessel placed in series with the extraction vessel. S C - C O 2 extractor eluate is nebulized into a diffused flame maintained within the upper region of the flame tube and the optical tube of the interface. Optimal flame conditions maintained with separate flows of 0 2 and H 2 to the base of the flame tube are slightly reducing for aqueous and S C - C O 2 mobile phases but slightly oxidizing for a methanolic mobile phase. For each analyte element, limits of detection were subnanogram to low picogram if standard was flow injected into the mobile phase. These sensitivities permitted differences in the rates of mobilization of analyte metal from different matrices to be explored as a technique for probing the interactions of the analyte metal with the matrix. A portion of the total Zn burden of fresh bovine liver slurry was rapidly mobilized in the absence of complexing agent, and the remainder was solubilized more rapidly than the Zn in a freeze-dried standard reference material of this tissue. Supercritical fluid mass transfer properties include solute diffusivities that are 1 order of magnitude greater and viscosities that are 1 order of magnitude less than those of liquid solvents.' In consequence, extraction with supercritical carbon dioxide ( S C - C O 2 ) has become a widely used technique for the rapid recovery of a wide variety of relatively nonpolar analytes from environmental/biological sample^.^-^ The recovery of more polar analytes has been somewhat less successful due to the nonpolar character of those SC fluids, which are readily generated in the laboratory. The solubilities of polar analytes have been increased by adding a polar modifier to the mobile phase,6s7 by resorting to more polar (1) Wach, F. Anal. Chem. 1994, 66, 369A-373A. (2) McHugh, M. A.; Krukonis, V. J. Supercritical Fluid Extraction; Principles and Practice; Butterworths: Boston, MA, 1986. (3) Stahl, E.;Quirin, K. W.; Gerhard, F. DenseGasesforExtracrionandRefining; Springer-Verlag: New York, 1988. (4) Hawthorne, S . B. Anal. Chem. 1990, 62, 633A-642A. ( 5 ) Veuthey, J. L.; Caude, M.; Rosset, R. Analusis 1990, 18, 103-111. (6) Wheeler, J. R.; McNally, M. E. J. Chromatogr. Sci. 1989, 27, 534-539. (7) Hawthorne, S . B.; Miller, D. J.; Walker, D. D.; Whittington, D. E.; Moore, B. L. J. Chromatogr. 1991, 541, 185-194.

3900

Analytical Chemistry, Vol. 88, No. 22, November 15, 1994

supercritical such as N 2 O or ammonia, or by conversion of the analyte(s) to a less polar derivative either prior to or during the extraction process.I0 The recovery of metal ions by SC extraction has been little studied and has often been considered unfeasible. However, a recent study" has demonstrated the feasibility of mobilizing Zn2+, Cu2+, and Pb2+ from aqueous solution by in situ chelation using a SC-CO2mobile phase that had been saturated with a nonpolar complexing agent. The nonpolar product complexes were rapidly purged from the extraction vessel and were efficiently recovered in a methanolic trapping solution. Related studies have focused on the relative rates of purging of a mixture of preformed metal-acetylacetonate complexes from the extractor,I2 the separation of several metal-diisobutyldithiocarbamates, bis(trifluoroethyl)dithiocarbamates, or P-diketonates by S C - C O 2 c h r o m a t ~ g r a p h y , l ~and - ~ ~the partitioning of product metal complexes from aqueous solution or solids into S C - C O 2 by adding a complexing ligand to the SC phase.16 The availability of an on-line metal selective detection system for the SC-C02 extraction would (i) appreciably simplify and accelerate the analysis procedures and (ii) permit process optimization during the course of the extraction. de Galan17 has demonstrated that, for volatile analytes, if the temperature of the nebulizer assembly is sufficiently high to ensure complete analyte vaporization there is no need for an aerosol to physically transport the analyte to the detector. By venting the extractor effluent to atmospheric pressure, the SC-CO2 spontaneously reverts to a gas, which facilitates detector interfacing; however, the attendant adiabatic cooling can cause problems. An inherent difficulty with linear restrictors used to achieve a controlled depressurization is the possibility of reduced or interrupted flow due to plugging, which can occur when the mobile phase contains a high concentration of solute and/or water. Typically, these problems are circumvented by heating the restrictor or by maintaining the collection solvent at a constant temperature. (8) Hawthorne, S. B.; Langenfeld, J. J.; Miller, D. J.; Burford, M. D. Anal. Chem. 1992, 64, 1614-1622. (9) King, J. W.; France, J. E. In Analysis with Supercritical Fluids: Extraction and Chromatography; Wenclawiak, B., Ed.; Springer-Verlag: New York, 1992; pp 32-60. (10) Hawthorne,S. B.; Miller,D. J.;Nivens, D. E.; White,D.C. Anal. Chem. 1992, 64, 405-412. (11) Wang, J.; Marshall, W. D. Anal. Chem. 1994, 66, 1658-1663. (12) Saito,N.;Ikushima,Y.;Goto,T. Bull. Chem. Soc. Jpn. 1990,63,1532-1534. (13) Manninen, P.; Riekkola, M. L. J. High Resolut. Chromotogr. 1991,41,210211. (14) Laintz, K. E.; Yu, J.-J.; Wai, C.M. Anal. Chem. 1992, 64, 311-315. (15) Ashraf-Khorassani, M.; Hellgeth, J. W.; Taylor, L. T. Anal Chem. 1987.59, 2077-2081. (16) Laintz, K. E.; Wai, C. M.; Yonkcr, C. R.; Smith, R. D. Anal. Chem. 1992, 64, 2875-2878. (17) de Galan, L. Spectrochim. Acta 1981, 368, 71-76.

0003-2700/94/0368-3900$04.50/0

0 1994 American Chemical Society

sc-co2

@

9

Silica

from compressor

(

W

1

hl d. n

AAS Detector

Figure 1. Supercritical fluid extractor consisting of (a) a temperature and pressure equilibration device (TPED), (b) a saturation vessel (SV), and (c) an extraction vessel (EV) immersed in an insulated water bath (IWB). Pressure was monitoredwith (d) a pressure transducer mounted in series between the TPED and the SV and displayed on (e) a digital pressure indicator powered via (f) a power supply. Two-way flowthrough valves (9) positioned after the EV and after a six-port rotary injection valve (h) in combination with a dual-stem three-way selectlon valve (j) permitted the mobile phase to be directed to the EV or to the injection valve. Extractor eluate was channeled via a 30 cm X 0.05 mm i.d. silica transfer line to the interface mounted within the optical beam of the AA spectrometer.

Inductively coupled plasma optical emission and mass spectrometries have been used to monitor ferrocenes,18 tetraalkyltin~,'~ and alkyllead compoundsZoin the SC-CO2 effluent from capillary columns. In addition to cluster formation and wall condensation, which can result if insufficient heat is provided to the interface assembly,21a further difficulty with the use of atomic emission detectors is the C02-induced decrease in response, which can be offset only partially by increasing the power to the plasma. In addition, plasmas (ICP and MIP) display only a limited tolerance to the presence of CO2 in the inlet feed stock. SC-C02 has also been employed as a sample transport medium for the determination, by flow injection (FI) flame AAS, of the copper content of a Cu-pyrollidinedithiocarbamate complex. The authorsZ2reported that the use of a heated restrictor at the end of the transfer line of the FI system resulted in sensitivities which were approximately equivalent to (1.2-fold greater) conventional water-based aspiration/nebulization of this analyte. The objectives of the current study were to develop an interface for coupling SCF extraction with on-line detection by AAS. The required system was to be capable of operation at conventional extractor flow rates and to provide sensitivities that were at least equivalent to conventional flame AAS. In addition, the possibility of performing chemical speciation based on differences in the kinetic rates of purging was to be explored.

MATERIALS AND METHODS The analytical-scale extractor (Figure 1 ) was assembled in-house and consisted of a compressor which delivered liquid (18) Fujimoto, C.; Yoshida, H.; Jinno, K. J. Chromarogr. 1987, 411, 213-220. (19) Shen, W.-L.; Vela, N. P.; Sheppard, B. S.;Caruso, J. A. Anal. Chem. 1991, 63, 1491-1496. (20) Carey, J. M.;Vela, N. P.; Caruso, J. A. J . Anal. Afom. Spectrom. 1992, 7, 1173-1 181. (21) Vela, N. P.; Caruso, J. A. J . Anal. Afom. Specrrom. 1992, 7, 971-977. (22) Bysouth, S.R.; Tyson, J. F. Anal. Chim. Acra. 1992, 258, 55-60.

4.H2 Figure 2. Silica T-tube Interface consisting of (a) an upper optical tube (13 X 1.1 cm i.d.1 mounted within the optical beam of the spectrometer and heated with a 12-turn coil of hlgh-resistance heating wlre, (b) a flame tube (3 X 0.8 cm i.d.) fitted with 02 and H2 gas Met ports [(d) 6 X 0.64 cm 0.d. (0.4 cm i-d.)], and (c) a sample Introductiontube [ 10 X 0.64cmo.d. (0.4cmi.d.)]. Quartzinserttubes(e, 8x0.32cmo.d.), constricted to a l-mm orifice at the exit end, were positioned within the gas entry ports and the sample introduction tube with modified Swageiok 0.64-0.32 reducing unions and made gas-tight with Vespel ferrules.

from the outlet siphon of a K-type cylinder of C02 via an in-line filter to a temperature/pressure equilibration device [TPED, an HPLC column assembly, 25 X 0.9 cm inner diameter (i.d.) filled with 4 mm diameter stainless steel balls]. Mobile phase was transferred from the TPED via a pressure transducer (PT) to a saturation vessel [SV, 25 X 0.9 cm (i.d.) HPLC column assembly] and then to an extraction [EV, 6.4 X 0.9 cm (i.d.)]. The TPEV, PT, SV, and EV were immersed in an insulated water bath (IWB). Standard stainless steel HPLC grade fittings and 0.16 mm (i.d.) transfer lines (rated to 6000 psi) were used throughout. A three-way flow-through valve, connected in series between the SV and the EV, permitted mobile phase to be directed to the extraction vessel or to a six-port rotary injection valve. Two-way flowthrough valves positioned immediately after the EV and after the rotary injectionvalve permitted the extractor to be operated in several different modes including (i) a static (no-flow) mode used to initially saturate the mobile phase with the ion pair complexing reagent, (ii) a static mode to equilibrate/saturate the aqueous sample with the ion pair reagent and mobile phase, (iii) a dynamic mode to purge analyte metals [as their dithiocarbamate (DTC) complexes] from the sample, or (iv) a mode to introduce metal-DTC standard into the mobile phase via the rotary injection valve. A 0.5 m X 0.16 mm (i-d.) stainless steel heated transfer line transported effluent from the extractor, via a capillary fused silica restrictor (23 cm X 50 pm i.d., SGE, Houston, TX), to the interface of the atomic absorption spectrometer (AAS, Philips Model PU 9100, equipped with a deuterium background correction system). Whereas pressure within the extractor was controlled by the compressor, the mobile phase flow rate was controlled by an appropriate choice of the length and inner diameter of the restrictor. The analog signal from the AAS was captured with a chromatographicdata reduction software package (Turbochrome 3, Perkin-Elmer Corp., Wilton CT). Interface. The three-component all silica interface (Figure 2) consisted of (a) an optical tube [ 13 X 1.1 cm inner diameter (23) Hendrick, J.; Taylor, L.T. Anal. Chem. 1989, 61,

1986-1988.

Analflical Chemistry. Vol. 66,No. 22, November 15, 1994

3901

(i.d.)], (b) a flame tube (3 X 0.8 cm i.d.) fitted with (d) gas entry ports (6 X 0.64 cm o.d., 0.4 cm i.d.) for 0 2 and H2, and (c) a sample introduction tube (10 X 0.64 cm 0.d. 0.4 cm id.). The optical tube was positioned within the optical beam of the AA spectrometer and heated with a 12-turn coil of 22-gauge Kanthal A-1 heating wire, 4-!2/m. Oxygen or hydrogen was delivered to the appropriate gas inlet port of the flame tube via a flow controller (0-500-mL or 0-2-L capacity, respectively, Matheson, Whitby, ON) connected to (e) quartz insert tubes (8 X 0.16 cm 0.d.) which extended concentrically inside the gas entry ports to within 1 cm of the center of the Y-shaped flame tube. The insert tube (e) was constricted to a 2-mm orifice at the exit end, positioned within the entry port with a modified Swagelok 0.64-0.32 reducing union and made gas-tight with Vespel ferrules. The fitting had been drilled out to permit the insert tube to pass through this fitting. Teflon tubing, which connected the flow controller with the appropriate gas inlet, was heat shrunk onto the entry of the insert tube to produce a gas-tight seal. The sample introduction tube, which met the optical tube at an angle of 30' relative to the plane formed by the optical tube and flame tube assemblies, was gently heated with a 15-turn coil of the same heating wire. The heating coils (positioned around theoptical and the sample introduction tubes) were separately energized by alternating currents rectified by variable transformers (Variacs). Extractor eluate contained in a 23 cm X 50 pm i.d. fused silica transfer line was vaporized into a diffused flame maintained within the upper region of the flame tube (b). The gaseous combustion products were entrained through the optical tube of the assembly. To initiate operation of the interface, the optical tube and the sample introduction tubes were heated to their normal operating temperatures ( 5 and 1.5 A, respectively). The flow of H2 to the interface was increased to 700 mL/min, during which the gas autoignited within the optical tube. The flow of 0 2 was increased slowly until it autoignited (typically at 50 mL/min) and then increased further to attain the normal operating flow rate. Visible flames emanating from both ends of the optical tube resulted in maximal response from the detection system. The capillary transfer line was extended into the sample introduction tube to a point just upstream from the optical tube, and the two-way flow-through valve from the EV was opened to permit the flow of eluate to the interface. To extinguish the diffused flame, the two-way flowthrough valve from the extractor was closed, the capillary transfer line was withdrawn from the sample introduction tube, and the flow of 0 2 to the interface was closed. Finally, the H2 flow was interrupted and the interface was permitted to cool to room temperature. Caution. The operator should be protected from possible injury caused by an explosion-induced shattering of the interface. The interface should be positioned directly below an efficient hood capable of venting the exit gases directly to the outside. The order of procedural steps for igniting and extinguishing the diffused flame must be followed. Complexing Agent and Metal Chelates. Tetrabutylammonium dibutyldithiocarbamate (TBADBDTC) and metalDBDTC complexes were prepared as described" previously. Repeated crystallization from isooctane, benzene, or hexane diethyl ether furnished analytical standards. 3902

Analytical Chemistry, Val. 66, No. 22, November 15, 1994

FIA-AAS. Residual analyte metal in the aqueous sample solution was quantified by flow injection analysis24of 10-50 p L aliquots into aqueous carrier solvent with detection by AAS using the method of external standards and/or standard additions. Samples. (i) NIST SRM 1557a, freeze dried bovine liver, or (ii) fresh bovine liver purchased from a local supermarket was solubilized (i) in 10% (w/v) KOH in 20% (v/v) methanol water or (ii) by sonicating 5 g of chopped wet tissue with 50 mL of distilled water (maximum speed, Model W- 185D, Ultrasonics Inc., Plainview, NY) once for 10 min with external cooling in an ice bath.

RESULTS AND DISCUSSION The design of the SC-CO2 extractor was essentially as described previously. In operation, the compressor delivered liquid phase C02, at constant pressure, from the siphon of a cylinder of the gas to three HPLC column assemblies mounted in series within an insulated water bath. The first column was filled with stainless steel balls and served as a temperature and pressure equilibration device. The second column was filled with complexing agent, TBADBDTC, contained in a snugly fitting cardboard thimble and served to saturate the SC-C02 mobile phase with this reagent. Finally, the third column assembly served as the extraction vessel. Design modifications included the addition of a three-way solvent switching valve (THV) positioned in series between the SV and the EV, which permitted the mobile phase to be directed to either the EV or to a six-port rotary injection valve. When combined with two two-way flow-through valves positioned immediately after the EV and the rotary injection valve, several modes of extractor operation were possible. Under static conditions (no flow), (i) the mobile phase or (ii) the mobile phase plus the aqueous sample could be saturated with the complexing agent, and under dynamic conditions, (iii) mobile phase saturated with complexing agent could be spiked with standard metal-dibutyldithiocarbamate (M-DBDTC) complex to calibrate the detector or (iv) analyte metal could be purged from the sample by in situ complexation and mobilized by dissolution of the resulting complex in the mobile phase. Initial feasibility trials were conducted with a silica T-tube interface24325 (prototype 1) composed of an upper optical tube and a lower sampleintroduction tube. The downstream portion of the sample introduction tube had been expanded to form a combustion chamber that housed a diffused flame supported by 0 2 and H2 which were added, via separate gas entry ports, to the base of the chamber. In operation, mobile phase was nebulized by a thermospray effect into the third entry port also located at the base of the chamber. For these studies, SC-CO2 at 50 "C and 24.05 MPa was delivered from the TPED directly to the interface via the rotary injection valve. Despite the fact that the strategy of thermospraying analyte metal, in either an aqueous or methanolic mobile phase, into the diffused hydrogen/oxygen flame had resulted in surprisingly sensitive responses to six analyte elementsz4 (AS, Cd, Cu, Hg, Pb, Se), this design had only a limited tolerance for C 0 2 . For flow rates of decompressed C02 corresponding to (24) Tan, Y.;Momplaisir, G.-M.; Wang, J.; Marshall, W. D. J. Anal. Atom. Specrrom., in press. (25) Momplaisir, G.-M.; Lei, T.: Marshall, W. D. Anal. Chem. 1994, 66, 3533.

greater than 13.74 MPa extractor operating pressure, it was not possible to maintain a flame within the interface. To increase the pressure operating range of the detection system, several design modifications were evaluated. These modifications included increasing the inner diameters of both the optical tube and combustion chamber from 0.9 to 1.7 cm and moving the 0 2 and H2 gas entry ports from the base to new positions on opposite sides of the chamber and located approximately midway between the entrance and exit of the combustion chamber. With this new design (prototype 2), only the region of the combustion chamber between the thermospray tube and the gas entry ports was heated. Although the prototype 2 design supported the requisite flow rates of decompressed COz, the principal shortcoming proved to be the limited range of analyte metals that could be detected with sufficient sensitivity to be useful for characterizing the metal burdens of biological/environmental samples. Only cadmium- or zincDTC proved to be sufficiently volatile. Also, the prototype 2 interface was not compatible with either aqueous or methanolic liquid mobile phases. Aqueous solutions containing zinc or cadmium were extracted to demonstrate the feasibility of using SFEprototype 2-AAS to characterize samples with respect to their metal burdens. Following 20 min of static equilibration at 24.05 MPa, 3 mL of aqueous solution [20 pg of Cd/mL as Cd(NO&] contained in the 4.6-mL capacity EV was extracted for 20 min with a SC-C02 saturated with TBADBDTC (decompressed flow rate, 900 mL/min). The change in AAS response with time is presented in Figure 3. Earlier results had demonstrated that, under the extractor operating conditions, (i) the mobile phase rapidly became saturated with the complexing reagent and (ii) a large excess of this reagent (relative to the quantity of analyte metal) was present in the mobile phase during the entire extraction. It was assumed that this excess was sufficiently large that its concentration could be considered constant. As a working model for the process, it was assumed that metal analyte (M) present in aqueous solution or slurry in various physical and/ or chemical forms ( M i ) are in equilibrium with free metal cation (M"+). Free metal cation is in equilibrium with nonpolar complex {M(DTC), aqjin the aqueous phase which, in turn, is in equilibrium with complex in the SC-CO2 phase {M(DTC), SC-COJ. If there is an appreciable excess of complexing agent, and if the rates of M"+ complexation, transfer of the resulting complex to the SC-C02 phase, and the rate of exchange of the head space gas are also fast, then the rate-determining step(s) will be the rate(s) of conversion from Mi to a form (M"+) that can react with the complexing agent. Provided that these assumptions are valid, a pseudofirst-order model is predicted to adequately describe the kinetics of the extraction process. It might then be possible to discriminate between different chemical species (Mi) of the analyte element on the basis of the relative rates of their mobilizationlpurging from the sample. The decay portion of the AAS response with time was modeled using a singleexponential decay or the sums of either two- or threeexponential decay functions. Using xz as a parameter of the goodness of fit of the model to the data, the sum of twoexponential decay functions provided an appreciably better fit to the data than either the single or the combination of

Zinc

-2'

'

0

'

' 2

'

' 4

'

' 6

'

'

6

'

' ' ' ' ' ' ' ' ' 10 12 14 16 18

Time (min)

Cadmium

-5'

0

'

2'

4

6

8

10

Time (min) Figure 3. AAS response curve for the SC-C02 complexometrlc extraction of 12 pg Zn/mL or 20 pg Cd/mL performed at 24.05 MPa and 50 OC for 25 min. The modeled declination portion of the response curve (3) was decomposed mathematically into two component exponential decay curves (1 and 2).

three decay functions (Table 1). However, there was no significant lack of fit for any of the models to the data. Flow injection analysis of the levels of residual Cd in the extracted sample indicated greater than 94% removal after 30-min extraction. These results corroborated earlier observations that the extraction process can be made to be very efficient. Analogous results (Table 1) were obtained when 3 mL of Zn(NO3)z or Cu(N0& in distilled water (12 pg of Zn or 5 pg of Cu/mL) was extracted at 50 O C and 24.05 or 27.48 MPa, respectively. Detection of Cu in the extractor eluate was performed using a modified interface (prototype 3, see below) to nebulize/atomize this analyte. The apparent rate constants that result from this fitting process can be interpreted to represent operationally dependent baseline values and can serve as a basis of comparison with slower extractions of other analyte metal species. It is to be recognized that the ability to successfully distinguish between kinetically distinct processes (to mathematically decompose an overall detector response curve into its components) will depend on the differences in the magnitude of the rate constants for the processes and on the proportion of the total analyte that undergoes each process. That 97.5 and 95.3% of the purged Cd and Cu in these extractions were best fitted to a single kinetic process is gratifying. The zinc results are somewhat more difficult to rationalize in that only 85% of the substrate analyte was fitted by the dominant process. None the less, after 4 half-lives of the faster process (3.5 min, when Ana&tical Chemistry. Vol. 66, No. 22, November 15, 1994

3903

Table 1. Klnetlc Models, Apparent Percent Composttion, and Efflclency of the SC-C02 Complexometrlc Exiractlon of Cadmlum, Copper, or Zlnc from Aqueous Solutlon % components'

analyte

model

tl

0.63 0.04 0.51 4.49 0.49 1.29 1.92 1.26 1.05

12

t3

2.55 3.05

10.9

2.18 2.80

55.7

21.3 2.20

15.2

X2

1

3.018 0.644 1.631 1.349 0.039 0.050 0.201 0.57 0.064

100 2.5 23.5 100 4.7 30.5 100 15.0 15.9

2

3

97.5 52.0

24.5

95.3 48.6

20.9

85.0 10.7

73.4

% removalb

94.4 91.9 92.1

a The apparent fractional composition was estimated by extrapolating the declination portion of the AAS response curve back to a maximum and assumed an identical response for equivalent quantities of different analyte "chemical species". Efficiency was determined from the residual levels of analyte in the aqueous sample after a 30-min extraction. Whereas detection of Cd or Zn was performed with the prototype 2 interface, the detection of Cu was performed with the prototype 3 interface.

Table 2. Estimated LlmHs of Detectlon. (LODe) for AAS Detection of Analyte Element In Different Moblie Phases

H20

CH30H

opt flow rate (mL/min) analyteC X (nm) As Cd Cu Mn Pb Se Zn

193.7 228.8 324.7 279.5 217.0 196.0 213.9

SC-CO2b opt flow rate (mL/min)

opt flow rate (mL/min)

H2

02

LOD(pmo1)

1050 1450 2200 2200 1350 1050 1250

460 325 645 685 485 460 325

5.42 0.14 0.96 2.00 0.27 9.31 0.17

H2

0 2

LOD(pmo1)

H2

02

LOD (pmol)

210 220 150 540

245 525 585 325

0.04 0.73 1.91 0.28

1010 1460

455 525

0.13 2.83

150

285

0.13

1105

455

0.46

Limit of detection for FIA-AAS were estimated as 3 (peak-to- eak baseline noise)/slope of the calibration plot for standards. Operating conditions were 50 OC and 24.05 MPa with a decompressed C02 flow rate of500 mLdmin. Analytical standards were nitrate salts, HsAs04, or H2Se04 in H20 mobile phase or M-DBDTC standard in methanol (methanolic or SC-C 2 mobile phase).

the contribution to zinc mobilization from this faster process will be less than 1% (6.25%X 15%), the molar fraction of analyte remaining in the aqueous sample (m/rno)is predicted to be 0.75 for the slower process. A maximum of less than 40% will have been mobilized by this time. Interestingly, a somewhat analogous mathematical expression (the sum of exponential decay curves) results from a (hot ball model) for S C extractions which predicts the changes in rate of solute recovery in terms of the size of the host matrix particles and the diffusivity of the solute or solvent through the host particles (unspecified process which can be approximated by a diffusion controlled process). For a single solute, this model predicts that a plot of In (m/mo) vs time will be curvilinear initially, becoming linear during the latter stages of the extraction. To improve the detector response to other analyte metals, an alternate strategy for introducing mobile phase to the interface was evaluated. The major interface modification (prototype 3, Figure 2) involved the introduction of the superheated SC-C02 extractor eluate into the upper region of the diffused flame rather than thermospraying liquid eluate into the base of the flame. In operation, extractor eluate contained in a 30-50pm i.d. silica transfer line was superheated and sprayed into the tip of a diffused flame (no visible flame front) maintained within the upper region of the flame tube and in the optical tube. Whereas the operating pressure of (26) Bartle, K. D.; Clifford, A. A,; Hawthorne, S. B.; Langenfeld, J . J.; Miller, D. J.; Robinson, R. J . Supercrit. Fluids 1990, 3, 143-149. (27) Bartle, K. D.; Boddington, T.; Clifford, A. A,; Hawthorne, S. B. J . Supercrit. Nuids 1992, 5, 207-212.

3904

Analytical Chemistry, Vol. 66, No. 22, November 75, 1994

the extractor was maintained by the compressor, the flow rate of mobile phase was controlled by the length and inner diameter of the silica transfer line. Typical flow rates of decompressed mobile phase at the exit of the extractor were in the range of 600-900 mL/min (corresponding to 0.5-0.8 mL/min compressed fluid). The displacement of the sample introduction tube from the base to the tip of the combustion chamber appreciably improved the sensitivity of the detection system to zinc or cadmium as well as increasing the number of analyte metals that could be detected. The system response (peak area) to seven elements entrained by flow injection into an aqueous, methanolic, or SC-C02 mobile phase was maximized by optimizing the flow rates of H2 and 0 2 to the flame tube by use of a univariate procedure. Whereas soluble nitrate salts or acidified solutions of selenate or arsenite served as calibration standards for the aqueous mobile phase, metaldibutyldithiocarbamate (M-DBDTCz) complexes were used for either the methanolic or SC-C02 mobile phase. Optimized interface operating parameters and the estimated limits of detection (LODs) for these analyte elements are presented in Table 2. For the aqueous mobile phase, a maximum response to the analyte metal was obtained with slightly reducing atmospheres (mean ratio H2/02, 3.2 f 25%). However, the optimized ratios of gas flow rates for this interface design were considerably less than the optimum ratio for As or Se (12 and 32, respectively) or the mean ratio for the maximum response to Cd, Cu, Pb, or Zn (4.9 f 15%)observed for the previous interface de~ign.2~~25 By contrast, optimized gas flow rate

5.8

-

5.8

-

5.4

-

5.2

-

7no

sJ

S E

5.5

0.8285.8

7.4

-

w

a c

cn

0

Q

3

methanol

o.saeng

7.0 5.8 -

5.4

-

5.2

-

cn

a L

1

6.0

-

8.2

-

5.8

-

5.4

-

Re

0.9939

5 '

8

1o.ong

I l l

6

water

1 .On0

5.8

5.8

5.8

5.8

5.4

5.4

5.2

5.2

5

5

water R2= 0.9950

Flgure 4. Typical AAS responses for the flow injection of various quantities (as metal analyte) of Cu (A) or Zn (6)standard into an SCC02, a methanolic, or an aqueous mobile phase.

ratios for the maximum response to these analyte-dithiocarbamate complexes in the methanolic mobile phase were distinctly oxidizing (H2/02, 0.74 f 75%) and only slightly reducing (2.5 f 12%) for the SC-C02 mobile phase. In all cases, the corresponding limits of detection of these analyte metals ranged from subnanograms to low picograms. Typical responses (prototype 3 interface) for the FI of zinc or copper in each of the mobile phases are presented in Figure 4. Whereas the injection of mobile phase resulted in no detectable instrumental response for either the water or methanol mobile phase, the injection of methanol into the SC-CO2 phase resulted in a minor blank signal equivalent to

approximately 20%of the lowest quantity of standard (Figure 4) injected into this carrier. The use of the deuterium background correction system appreciably degraded the performance of the detection system (increased background noise) and was not used for these studies. For the SC-CO2 mobile phase, the FI of methanolic standards of either MnDBDTC or Pb-DBDTC at levels corresponding to submicrogram quantities of analyte metal did not result in any detector response. By contrast, the mobilization of 30 pg of either Pb or Mn from distilled water ( 3 mL) by complexation with DBDTC and mobilization into SC-CO2 (28.86 MPa, 50 "C) did produce an appreciable detector response (Figure 5). Analyticai Chemistry, Vol. 66, No. 22, November 15, 1994

39Q5

s "O

1 e,8

t

~~

L

24 '-

ll-

I s

22 I-

-

S.6

B

-

3

5.0 -

f 5.4

't 1i--1 P 14

o

12

g io i-

4

,

l

0

4

I

/

0

#

l

12

,

,

18

,

,

,

20

___

Table 4. Purglng of Cadmium Catlon from Aqueous W i o n into SCCO, by Compiexometric Extractlon wlth TBADBDTC

soln [Cd2+] extctn conditions",* gases (mL no. (pg/mL) andduration 0 2 H2

0.10 0.20 0.30 0.40 1.0 2.0 3.0

1

2 3 4 5 6 7

3500/50/25 3500/50/25 3500/50/25 3500/50/25 3700/50/30 3700/50/30 3700/50/30

525 525 525 525 120 120 120

700 700 700 700 987 987 987

95 anaiyte

removalC R2d

98.9 97.1 95.3 94.4 93.1 92.8 92.1

0.9803 0.9893

a Operating pressure ( si), temperature ("C), and duration (min). Solutions were extractedPdynamicallywithout a prior presaturation of the mobile phase with the complexing a ent. CPercent removal was calculated from the residual levels of CdB+ in the aqueous sample as determined by FIA. For the linear regression of the area under the modeledAASresponsecurve{(Y- YO) A2e(-xlr~)on [CdZ+].

I

24

limo (min)

5

Figure 5. AAS response curves for the compiexometric extraction into SC-C02 of 3-mL aqueous solution containing 30 pg (as metal anaiyte) of either Pb(NO& or Mn(NO&. Extractions were performed at 28.86 MPa and 50 O C without presaturation of the mobile phase with compiexing agent. Curves 1 and 3 represent AAS signal in the absence of compiexing ligand.

F

v

Table 3. Figures of Merlt for the Dynamic Complexmtric Extraction into SC-C02 of Copper, Manganese, or Lead from Aqueous Solution Containing 5 or 10 pg/mL Analyte Metal Cation % RSD for the

analyte

conc (rg/mL)

cu

5

Mn Pb

10 10

cumulative peak area 10 min 20 min 40 min

f10.2 f10.9 flO.l

f11.3

f12.2 fll.5

f12.1 f15.8 f14.2

% removalo

92.0f0.008 93.6f 0.01 94.4f 0.01

Percent removal was calculated from the residual levels of analyte in the aqueous sample as determined by FIA. (1

The repeatability of extraction process was tested by performing three replicate extractions on 3-mL aliquots from solutions containing 5 pg of Cu/mL, 10 pg of Pb/mL or 10 pg of Mn/mL. For these trials, no static preequilibration of the mobile phase with the complexing reagent was performed-rather the analyte was purged from the sample using mobile phase that had dissolved complexing agent as it passed through the SV. Arbitrarily, these extractions were performed for 40 min at 27.48 (Cu) or 28.86 MPa (Mn, Pb) and 50 OC. At the termination of each trial, residual quantities of analyte metal in the aqueous aliquot were determined by FIA with quantitation by standard additions and by the method of external standards. The results were identical using both calibration techniques. As recorded in Table 3, the extractions were both very efficient for all three metals (>92% removal) and highly repeatable (-0.1% RSD). The cumulative area under the AAS response curve however was appreciably more variable (f12.1, f15.8, and f14.2% RSD for Cu, Mn, and Pb, respectively). A portion of this variation was apparently caused by longer term instrumental drift in that the variability in the AAS signal (in terms of cumulative area) increased modestly with longer extraction times. Additionally, the AAS signal typically did not return to values for mobile phase in the absence of pairing ion. For this reason, the mathematical models that were used throughout these studies included a 3906

Analyticel Chemistry, Vol. 66, No. 22, November 15, 1994

.-N

1e-

0

-

f

c8

3

141210-

&.

1,

~

Time (min) Flgure 6. AAS zinc response profiles for the extraction, into SC-C02, of freezdried bovine liver standard in the (1) absence or (2) presence of compiexing agent or (3) an aqueous solution containing 12 pg of Zn in the presence of compiexing agent. The response curves 4 and 5 resultedfrom the SC-C02extractionof fresh bovine liver homogenate in the absence (4) or presence of compiexing agent (5). Extractions were performed at 20.04 MPa and 50 O C with a static equilibrationof the mobile phase with the complexingagent for 20 min prior to dynamic mobilization of anaiyte Zn.

term (YO) to compensate for the contribution of complexing reagent in the mobile phase to the AAS response. Variability was further explored by extracting aqueous solutions containing 0.1-3.0 pg of Cd as Cd(N03)~.The extractor and interface operating conditions and the resulting efficiencies of extraction are summarized in Table 4. Despite a modest change in the shape of the AAS response with increasing metal loading in the aqueous sample, as well as a modest decrease in the efficiency of the process, the cumulative area under the AAS signal was highly correlated with the concentrations of Cd in the sample solutions (for extraction performed with the same interface operating parameters). As reported in Table 4, R2 for the linear regression of the optimized model of the AAS response ( ( Y - YO)= Ale(-X/tI) + A2e(-X/t2)jon [Cd2+] for four or three standard solutions extracted under identical operating conditions was 0.9803 and 0.9893, respectively. To demonstrate the feasibility of working with biological materials, a standard reference material (NISTSRM 1557a)

Table 5. Mathematlcai Models for the Declination Portion of the Zinc or Cadmlum AAS Response to Eluate from SC-C02 Complexometrlc Extraction of Fresh and Defatted Freeze-Drled Bovlne Liver 3'% components

matrix

SRM Freeze-dried liver fresh liver hgtC

sc-co2 + Ld sc-co2

Zn(NO& in distilled water

model'

tl

sumof 2

2.50

sum of 3 one sum of 2

1.03 0.55 1.26

12

t3

22.3 4.51 27.3

78.8

XZ

0.025

92.0

8.0

0.036 0.208 0.057

17.5 100 15.0

17.3 85.0

S removal* 90.1

65.2

85.6 5.6 92.1

Responses were modeled as one, the sum of two, or the sum of three exponential decay curves. After 30-min extraction at 50 OC and 24.5 MPa as determined by residual level of analyte metal in the sample. hgt, homogenate. L, DBDTC ligand.

of freeze dried bovine liver was dissolved in methanolic KOH and subjected to SC-CO2 extraction at 20.03 MPa and 50 OC for 15 min. The resulting AAS Zn traces are presented in Figure 6 and the relevant data in Table 5 . Whereas the extraction of aliquots of the solution with SC-C02 alone did not result in a perceptible AAS Zn response (curve 1, Figure 6 ) ,saturation of the mobile phase with the complexing reagent resulted in an AAS response curve (curve 2) that was similar but not identical to the AAS response curve for the extraction of 36 pg of Zn(NO3)z from 3 mL of distilled water (curve 3 ) . Surprisingly, when fresh bovine liver, purchased from a local supermarket, was sonicated and then extracted under identical conditions, a more rapid rate of mobilization of the Zn content was evident (curve 5 ) and approximately 86% of the Zn burden was recovered. When a separate aliquot of the same slurry of the fresh liver was extracted with SC-CO2 in the absence of complexing ligand, a portion (5.6%) of the zinc was mobilized from this matrix (curve 4). The chemical identity of the nonpolar fraction of the total Zn content remains to be determined. Apparently, the preparation and/or slurrying of the freeze-dried material in KOH has appreciably altered the interactions of the Zn with this matrix. The efficiency of extraction as determined by the residual zinc content in the aliquots of samples after extraction are recorded in Table 5 .

CONCLUSIONS A novel interface for coupling aqueous, methanolic, or SCC02 solvent streams with on-line detection by AAS has been developed. The approach of SC-CO2 extraction coupled with on-line detection by AAS provides (i) a novel method for characterizing different physical and/or chemical forms of an analyte trace metal based on differences in their relative rates of mobilization. To the extent that successful mathematical models can be developed for the extraction process, this approach can be used to (ii) estimate analyte metal burdens in samples without having the complete the extraction, and (iii) on-line detection will be useful to optimize the extractor operating parameters so as to maximize rates and efficiencies of metal mobilization. ACKNOWLEDGMENT Financial support from the Natural Science and Engineering Research Council of Canada (NSERC) in the form of an operating and a strategic operating grant is gratefully acknowledged. Received for review June 20, 1994. Accepted August 18, 1994.' *Abstract published in Advance ACS Abstracts, October 1 , 1994.

Analytical Chemistw, Vol. 66, No. 22, November 15, 1994

3907