1640
AnaL Chem. 1983, 55, 1640-1642
High Sample Convection Donnan Dialysis J. A. Cox" and G. R. Litwinski Department of Chemistry and Biochemistry, Southern Illinois University, Carbondale, Illinois 6290 I
Donnan dialysis has been shown to be a useful, quantitative method for preconcentration and matrix normalization in the determination of ionic species (1-5). In the method a reproducible fraction of the analyte ions is transported from a sample solution through an ion exchange membrane into a small volume of a relatively high ionic strength receiver solution. Quantification is based on linear calibration curves that are generated with a fixed-time kinetic model. The receiver electrolyte is an ideal matrix for electrochemical methods and is also suitable for many spectroscopic procedures. Both sheet and tubular (6, 7) membranes have been used. The latter is a particularly attractive form because of the high membrane area to receiver volume ratio that is possible. In turn, greater preconcentrations can be obtained. The preconcentration capability of Donnan dialysis also depends upon the rate of ion transport through the membrane. Several limiting steps for the overall flux in Donnan dialysis have been postulated including concentration polarization a t the solution-membrane interfaces and mass transport resistance within the membrane (4,8). The influences of the membrane structure and the receiver solution composition on the dialysis rate have also been examined (9-13). Donnan dialysis with a static receiver solution, a magnetic stirrer for sample convection, and sheet membranes yields only low preconcentrations in reasonable times (up to 1 h) when the receiver volume is compatible with the requirements of routine analytical methods, i.e. a few milliliters. Enrichment factors (the ratio of the analyte concentration in the receiver after dialysis to the initial sample concentration) are typically less than 10. High enrichments have been obtained with sheet membranes by using long dialysis times with a continuous flow system (14) and by using receiver volumes of less than 1mL (15). With tubular membranes, receiver recirculation, and sample convection with a magnetic stirrer enrichment factors of 40-50 have been obtained in 20 min (6). The receiver recirculation method has two drawbacks. The relatively large dead volume of common pumps limits the attainable enrichment factors, and the cost of pumps with the capability t o provide a constant flow rate is relatively high. In this paper, two simple and inexpensive methods are described for obtaining high enrichments in short times. EXPERIMENTAL SECTION All chemicals used were reagent grade. The receiver solution was generally a 0.2 M MgS04,0.5 mM A12(S04)3mixture that was prepared from a concentrated stock solution which had been purified by electrolysis for at least 1 week at a large mercury pool cathode at -1.3 V vs. SCE. This receiver had been previously shown to be well-suited for the Donnan dialysis of divalent cations (5, 10). All water used with distilled and doubly deionized. The membrane used was Nafion 811 cation exchange tubing (Du Pont Polymer Products, Wilmington, DE) with dry dimensions of 0.64 mm i.d. and 0.89 mm 0.d. Prior to initial use, each new section of membrane was soaked in 1 M NaOH and then in 1 M HCL for 30 min each. Immediately prior to each dialysis, the membrane was conditioned by pumping 2 M MgS04 through the tubing for 15 min at about 2 mL min-l. The membrane was then put into an initial state by pumping receiver electrolyte likewise. During the conditioning, the outer membrane surface was bathed in several changes of distilled water. Two types of dialysis cells were used. The f i s t design consisted of a square cross-section Plexiglas frame (Figure 1) sized to fit into the polystyrene copolymer container of a household blender (Osterizer 848-34). The Nafion tubing, either 7.0 or 12.0 m, was wrapped around the frame. The blender container held the 1000 0003-2700/83/0355-1640$01.50/0
mL of sample solution. During the dialyses, the blender speed control was set at its lowest position. The second type of dialysis cell consisted of a 15 x 2.2 cm Plexiglas spool around which 10.0 m of Nafion was wrapped; the spool was fitted into a 4.5 cm i.d. cylindrical jacket that contained solution ports (Figure 2). The sample solution was circulated through the cylindricalcell at 4.4 L m i d using a Masterflex pump (Cole-Parmer;Chicago IL). For most applications,a second route for circulation of the sample was simultaneously used. The additional turbulence that is created by using two flow pathways inside of the sample cell yields higher enrichment factors with a minor increase in complexity of the system. The Nafion tubing exits the cell through small holes drilled in each end of the spool and extends about 1-cm beyond each end. The holes were just large enough to allow passage of the dry tubing; on wetting, the tubing swelled t o form a tight seal. The Nafion tubing contained a static volume of receiver electrolyte during the dialyses. Inlet and outlet lines to the Nafion tubing were made of Teflon (0.8 mm i.dJ; these were joined to the ends of the Ndion tubing with 1-cm lengths of 0.5 to 0.8 mm i.d. silicone rubber tubing. Sample inlet-outlet lines to the Figure 2 cell were of Tygon tubing; they were joined to the cell with barbed polypropylene fittings (Value Plastics, Loveland, CO). An Erlenmeyer flask served as sample reservoir. Metal ion determinations were by Cu(I1) ion selective electrode (HNU Systems, Inc., Newton, MA), flame atomic absorption (Varian 475), and electrothermal atomic absorption (Varian CRA 63 graphite furnace atomizer) methods. All standard solutions were prepared in the presence of the electrolyte used as the Donnan dialysis receiver solution. The concentration was matched to the value after dialysis, which differs by 10-20% from that of the initial receiver electrolyte because of osmotic dilution and electrolyte loss because of the nonideal membrane selectivity. The change in the electrolyte concentration is constant for a given dialysis time, membrane configuration, and electrolyte composition. In the experiments that did not use a flowing receiver electrolyte, the conditioned membrane tube was filled with receiver solution, and its inlet was closed with a stopcock. The outlet was placed in a 10-mL collection beaker. The membrane assembly was contacted with the flowed or mixed sample solution for a prescribed time. The receiver inlet stopcock was then opened and the receiver was blown into the beaker with a plug of air. The recirculated receiver experiments were similar except that a measured volume of receiver was cycled from the collection beaker through the Nafion tubing by a peristaltic pump (Model 375A, Sage Instruments). All dialyses were performed at room temperature, 24 & 2 "C. During a dialysis, the solution temperature rose by 2-3 "C depending on the dialysis time, sample volume, and rate of sample stirring. R E S U L T S A N D DISCUSSION Initial experiments were performed with the blender dialysis cell. The test ion was Cu(I1) in the range 1 X lo-* to 1X M (8 points). A linear least-squares fit of the data for samples above 5 x M in Cu(1I) yielded the following: slope (enrichment factor), 44; correlation coefficient 0.9998. The relative standard deviation for six replicate dialyses of a 2.5 X lo4 M sample was 7%. At Cu(I1) levels below 5 X M higher apparent enrichment factors were observed. The cause was traced to corrosion of the blender's metal stirring assembly. When 1000 mL of distilled water was stirred for 15 min in the blender, the Cu(I1) in the water reached a 4 X lo-'' M level. Thus, although the blender gives rapid and high enrichments, it cannot be used a t ultratrace levels for Cu(I1). T o determine the extent of the above impurity problem, 1000 mL of M HCl was stirred in the blender for 15 min, and the leachate was examined by argon inductively coupled 0 1983 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 55, NO. 9, AUGUST 1983
1641
Table I. Donnan Dialysis of Cd(I1) in the Flow Cella enrichment sample flaw rate, dialysis factor time, min L min-’ 40 15 1.7 52 15 2.8 5.3
4.4 4.4
15 15 15 15 30
68 53c 56 47 88
4.4 b 4.4 a Sample volume, 2 L except where otherwise noted; receiver volume, 4.6 mL (static) except where otherwise M. Sample noted; Cd(I1) concentration, 5.0 X volume, 1 L. Receiver volume, 6 mL, with recirculation at 2 mL min”.
s i d e vilew Figure 1. Frame for the blender dialyzer: (A) 12 mm diameter rod; (B) 2-mm holes for Nafion entry/exit; (C) 6 mm diameter rod. The membrane, which is not shown, is wrapped around the four 6inm rods.
1--i
5 cm
-I
Figure 2. Flow cell for Dorrnan dialysis: (A) Nafion 81 1 membrane; (B) end of spool with groove for O-ring; (C) 2.2 cm rod; (D) 3-mm rod to prevent collapse of membrane against spool (two of the four whlch are present are shown); (E) O-ring; (F) 7/,8-27 threaded sample outlet: (G) cell jacket; (H) 7/,,-27 threaded sample inlet. The cell construction is all Plexiglas.
plasma spectrometry. The following metals were not detededl: Ni, Fe, Co, Cr, Pb, Al, and Cd. Only Cu and Zn were found to exceed 0.01 ppb; their levels were Cu (27 ppb) and Zn (12 ppb). Hence, the blender dialysis cell can be used for most routine purposes. The study was extended to lower conceintration levels with Cd(I1) as the test ion. In this case, enrichment factors with 15 min dialyses were determined for six samples in the 5 X lo4 to 1 X lo+ M range. The average enrichment factor was 46 f 2. The constant value demonstrated the absence of a Cd impurity a t the 5 X M level. Since the blender motor could not be expected to precisely maintain a constant stirring speed, the effect of stirring speed in the sample solution was examined. A variac was used to change the power while the blender was! set at its highest setting. Changing the power from 42 to 72 V resulted in a significant stirring rate increase (comparablle to using different settings on the panel), but the enrichment, factor changed by only 10%. In cases where contamination would limit the use of the blender dialyzer, an alteirnative all-plastic cell (Figure 2) can be used. Mixtures containing Cu(I1) and ICd(I1) in the range of 5 X to 5 X lo4 M were dialyzed for 15 min using thiB cell. Enrichment factors for Cd(I1) and Cu(I1) were 55 f 4 and 54 f 4,respectively, with a static receiver. The similarity of the enrichment factors for Cd and Cu agrees with previous reports using the Mg(II),IAl(III) receiver solution (6,lO)and suggests that Donnan dialysis may be used to obtain simul-
taneous enrichments for a variety of analyte ions without significantly changing their relative concentrations. The relative standard deviations (8%)agree with a previous report (10) for nonthermostated Donnan dialysis with nominal stirring speed control using sheet membranes and the Mg(11)/Al(III) receiver. The static receiver mode of dialysis was compared with the recirculated receiver mode using a 6-mL receiver volume and 2.0 mL min-’ flow rate (Table I). The static receiver mode yielded slightly higher enrichments which indicates that there is little benefit in recirculating the receiver. A second experiment at 4.0 mL min-l receiver flow rate gave similar results. It was previously reported that with Nafion 811 the Donnan dialysis enrichment efficiency reached its maximum thus, higher flow rates at a receiver flow rate of 3 mL m i d (6); were not employed. Although recirculation reduces concentration polarization at the membrane-receiver interface, this effect is so small that it can be easily out-weighed by the dead volume of the pump and associated tubing. As a practical matter, if a continuous on-stream application is not required, there is little or no advantage to flowing the receiver when using reasonably short lengths (up to about 10 m) of Nafion 811. Since the back pressure of the lengths of Nafion 811 used in this work is low, gravity flow, gas pressure, or syringes may be used to fill, empty, and condition the membrane. A pump would therefore not be needed. Table I also shows the effect of sample volume, dialysis time, and sample flow rate on the static receiver experiment. For dialysis times well-below the Donnan limit (EF for a 2-L sample of about 400),the quantity of the analyte in the receiver should increase in direct proportion to dialysis time. However, a small amount of osmosis occurs (5% and 10% receiver volume changes for 15 and 30 min dialyses, respectively); the cation transport number is less than unity and the sample becomes significantly depleted of analyte. Direct proportionality is therefore not experimentally observed. The 20% drop in E F when going from a 2 L to a 1 L sample points to the advantage of using large samples to minimize depletion effects. The rapid increase in EF with sample stirring rate illustrates the importance of concentration polarization at the interface of the sample and the membrane. At sufficiently high sample flow rates, the enrichment factor would become nearly independent of this parameter (8,9). With the blender system, the enrichment factor was nearly independent of stirring rate which indicates that the turbulence was so high that concentration polarization at the sample-membrane interface was not the major rate-determining transport step. In Donnan dialysis experiments using sheet membranes, conditioning is generally performed by placing the entire membrane in the conditioning solution. With tubular membranes it would be desirable to condition the membrane from the receiver side only in order to reduce the consumption of the purified conditioning solution. The practicality of such
1642
Anal. Chem. 1983, 55, 1642-1645
a conditioning scheme was examined by first dialyzing a mixture that contained lom4 M CuSOl and CdSOl for 15 min using the Figure 2 cell with a static receiver. The receiver was then eluted with 2 M MgSO, a t 2 mL min-l while distilled water was passed through the sample compartment at 100 mL mi&. Effluent fractions of the MgSO, solution were collected at timed intervals, and the Cu(I1) and Cd(I1) therein were determined. The high ionic strength conditioning solution rapidly strips the residual analyte from the membrane. The Cu and Cd concentrations drop from 3 X M in the fraction that was collected after 1 min to 1 X lo4 M in the 30-min fraction. The drop of 3 orders-of-magnitude in 30 min is observed regardless of the initial concentration. Hence, a rinse of the tubing with 30 mL of 2 M MgSO,, followed by 30 mL of the receiver solution, each at about 2 mL m i d , is a sufficient conditioning step when the sample concentration is not lowered by more than 2 orders-of-magnitude in consecutive experiments. In the most extreme case studied, when a 5 X lo4 M sample followed a 1 x M Cd(I1) solution, 45 min was required to condition the membrane. In conclusion, simple Donnan dialysis systems that provide enrichment factors of over 50 in 15 min have been described. The systems are compatible with such convenient analytical methods as ion selective electrode potentiometry, pulse polarography, and flame atomic absorption spectrometry; these methods can, therefore, be routinely applied to the determination of metals at the nanomolar level. Our earlier studies have demonstrated that sample matrix effects on Donnan
dialysis are not significant over a wide range of conditions (1-3).
ACKNOWLEDGMENT
Y. T. Kim, SIU-C, performed certain of the dialysis experiments. M. Kaiser, Environmental Trace Substances Center, University of Missouri, performed the ICP measurements. LITERATURE CITED (1) Lundquist, G. L.; Washinger, G.;Cox, J. A. Anal. Chem. 1975, 4 7 , 3 19-322. (2) Cox, J. A.; Cheng, K. H. Anal. Left. 1978, IY (A8), 653-660. (3) Cox, J. A.; Twardowski, Z.Anal. Chlm. Acta 1980, 119, 39-45. (4) Blaedel. W. J.; Haupert, T. J.; Evenson, M. A. Anal. Chem. 1969, 4 1 , 583-590. (5) Wilson, R. L.; DiNunzio, J. E. Anal. Chem. 1981, 53, 692-695. (6) Cox, J. A.; Twardowskl, 2 . Anal. Chem. 1980, 52,1503-1505. (7) Cox, J. A.; Carnahan, J. A@/. Specfrosc. 1981, 35,447-448. (8) Lake, M. A.; Melsheimer, S. S.AIChE J . 1978, 24, 130-137. (9) Wen, C. P.; Hamii, H. F. J . Membr. Sci. 1981, 8, 51-68. (IO) Cox, J. A.; DINunzio, J. E. Anal. Chem. 1977, 4 9 , 1272-1275. (11) Cox, J. A.; Gajek, R.; Litwinski, G. R.; Carnahan, J. Anal. Chem. 1982, 54, 1153-1157. (12) Cox, J. A.; Olbrych, E.; Erajter, K. Anal. Chem. 1981, 53, 1308-1309. (13) Cox, J. A.; Cheng, K. H. Anal. Chem. 1978, 50,601-602. (14) Davls, T. A.; Wu, J. S.;Baker, 8. L. AIChEJ. 1971, 17, 1006-1008. (15) Blaedel, W. J.; Kissel, T. R. Anal. Chem. 1972, 4 4 , 2109-2111.
RECEIVEDfor review March 11,1983. Accepted April 18,1983. This work was supported by the U.S. Environmental Protection Agency under Cooperative Agreement USEPA-CR809397 with the Athens Environmental Research Laboratory.
Sample Preparation for Monitoring Asbestos in Air by Transmission Electron Microscopy Carry Burdett and Anthony Rood* Health & Safety Executive, 403 Edgware Road, London NW2 6LN, England
Measurements of environmental asbestos concentrations by transmission electron microscopy (TEM) have been reported regularly in recent years, and the results have been summarized by Spurny et al. ( I ) and Sebastian et al. (2). This work represents the first detailed survey to be undertaken in the United Kingdom since that of Rickards (3) in 1972. A novel centrifuge technique was used in the preparation of samples taken in the survey, which has reduced sample preparation times from many hours to a few minutes. The full results of the survey will be published in the Annals of Occupational Hygiene in the near future, although a few are quoted here for illustration. An example is also given of a fiber size distribution determined by means of the technique from an air sample taken outside a factory using chrysotile. This is relevant to recent work on size-selective fiber retention in the human lung ( 5 ) . The technique previously employed by many workers has been to collect a sample on a membrane filter, then to ash the filter in a plasma oven, and to resuspend the residue in distilled water prior to TEM preparation by the “Nuclepore Jaffe-wick” procedure of Sebastien et al. (4). Complete preparation by this method can take many hours (or days). Beamen and File (6) have emphasized the possibility of losses in preparation, we attempt to quantify fiber recovery for the centrifuge technique described. EXPERIMENTAL SECTION
Sampling. The samples were collected for the survey by using two Leap 3620 high-volume electrostatic precipitator samplers mounted side-by-sidein a Land Rover. These operate at about
8 m3/min. The precipitator plates are washed by a detergent solution, which concentratesthe particles from the air into about 200 mL of solution. According to the manufacturers’ data, the collection efficiency of the samplers falls with decreasing particle size to about 80% at sizes similar to chrysotile fibrils. Specimen Preparation. On return to the laboratory, the 200-mL samples are resuspended with an ultrasonic probe (operating at 70 W metered energy) for 20 min. This period of time has been found ideal for breaking down fiber aggregates which occur when asbestos suspensions are allowed to settle (7). A 400-pL sample from this suspension is placed in a centrifuge tube as shown in Figure 1 (a reusable or disposable tube is suitable); 20 pL of “AnalaR”acetone is added at this stage to reduce surface tension effects. A carbon-coated, 300-mesh TEM grid is introduced edge-on and placed flat on the bottom of the tube with a pair of croas-over tweezers (this procedure has been found to prevent the carbon film from being dislodged from the grid). The grid area is approximately one-tenth of the cross sectional area of the centrifuge tube. Samples are then centrifuged (in a Dupont-Sorvall RC-5B) for 10 rnin at 13000 rpm, after which the supernatant liquid is drawn off carefully with a Pasteur pipet or by capillary action with a tissue (see Figure 1).If a disposable tube is used, the base is simply punctured and the supernatant liquid is allowed e0 drain off. Grids are then removed and left t o dry in a dust-free atmosphere for a few minutes. ‘The design of the centrifuge rotor allows pairs of tubes to be run under the same conditions, and duplicate samples are always prepared. Fibers precipitate vertically, with respect to the tube axis during centrifugation, and the carbon film intercepts them directly. A small number of fibers beneath the plane of the carbon film are lost as they precipitate to the base of the tube. Inspection, in the TEM, of paired samples each of which were in a different relative position at the base of the
0003-2700/83/0~55-1642$O1.50/0 Published 1983 by the American Chemical Society
