NICKEL-COBALT-IRON-MANGANESE. A typical A F wavelength scan obtained while nebulizing a solution containing 1 ppm nickel, 1 ppm cobalt, 1 ppm iron, and 0.5 ppm manganese, is shown in Figure 4. This scan was taken between 227 and 285 nm, and thus incorporates the major resonance line of each of these elements. CONCLUSIONS
A scanning technique for multi-element determinations by AF, in conjunction with dual-element EDL appears to be a feasible method of analysis for a number of combinations of elements. The procedure is simple and all of the instrumentation is commercially available. The sensitivity and selectivity appear to be as good as for conventional A F with the same type of source and flame.
The time taken for a determination of zinc, cadmium, nickel, and cobalt is ca. 100 seconds; that for selenium, nickel, tellurium, and cobalt is ca. 140 seconds; and that for nickel, cobalt, iron, and manganese is ca. 150 seconds. These times may be improved upon by using a faster scanning speed, without too great a loss in sensitivity or reproducibility. The main disadvantage of the technique, as with any multielement atomic spectrometric method, is that the solutions must be of a limited concentration with respect to each element, to avoid more than one dilution, and, hence, more than one analysis of each sample RECEIVED for review July 31, 1972. Accepted October 2, 1972. We thank the Science Research Council for the award of a research studentship to J.D.N.
Determination of Sub-NanogramQuantitiesof Silver in Snow by Furnace Atomic Absorption Spectrometry Ray Woodriff, Bruce R. Culver,’ Douglas Shrader, and Arlin B. Super Montana State Unicersity, Bozeman, Mont. 59715 An analytical method for determining microtrace concentrations of Ag in snow is discussed. The method involves preconcentration of the Ag by solvent extraction and its subsequent determination by furnace atomic absorption (FAA). The extractant is a dithizoneCC14 solution. Over two hundred twenty-five snow samples were analyzed by this method. A comparison among microsampling boat flame AA, neutron activation analysis, and furnace AA is presented. The results obtained by the boat technique are generally higher than those obtained by the other two methods. Where neutron activation analysis is &15-40% reproducible, FAA’s reproducibility is 15%. The concentration ranges involved are on the order of 5 X 10-1‘ g/ml. The sensitivity for the FAA method is 5 X 10-13 g/ml.
SILVERIODIDE is the most commonly used agent in weather modification programs. The detection of silver in snow from clouds seeded with silver iodide is of interest in evaluating the success of such programs. Detection of silver in above background levels does not prove that the silver iodide crystals caused nucleation of supercooled cloud droplets and subsequent ice crystal growth and fall out. The silver iodide particles might have been scavenged by natural snowflakes and/or simply have been deposited on the snow surface. However, the absence of above-background levels of silver in an assumed target area indicates that proper targeting was not accomplished. Hence, measurement of the spatial variation of silver concentration in snow from a seeded storm or storms can be a valuable tool in the evaluation of a cloud-seeding experiment. During the past five years there have been two accepted techniques for measuring Ag in snow. The first and most widely accepted is neutron activation analysis (1). This expensive technique requires large samples (1 to 2 liters of Present address, Varian, Los Altos, Calif. 94022. ~~
(1) J. A. Warburton and L. G. Young, J . Appl. Meteorol., 7, 433
(1968). 230
melted snow), a source of neutron flux, and has a fractional standard deviation of + 15-40 in the range of Ag concentrations in snow. Recently another method has come into existence. This less expensive method is a flame atomic absorption technique involving a microsampling boat ( 2 , 3). The claimed detection limit for this technique ranges from 10-lO to 5 x lo-” giml of silver. These reported limits incorporate a preconcentration step and a 1OX recorder scale expansion. AAS has been used for the silver determination in seeded snow, but the procedure lacks reliability since the values reported are dependent upon considerable extrapolation ( 4 ) . Use of the microsampling boat improves this somewhat. Due to the extremely low Ag concentrations in snow, a method must be very sensitive to be useful. Using neutron activation analysis, E. Bollay Associates (5) found that the background concentration of silver in the spring snow pack of the western United States ranged from 0-20 X g/ml. Samples collected from the target areas of cloud seeding programs yielded silver concentrations ranging from 20-200 x lo-’* g/ml (melted snow). The sensitivity of furnace atomic absorption (FAA) for g/ml with a lox preconsilver is on the order of 5 x centration step and no recorder scale expansion. The reproducibility of the method in the 5 X 10-l1 g/ml range is = 5 for repeated determinations of the extractant solution from a single sample. FAA should be an excellent method for microtrace silver determinations. There are, however, many prob(2) Denver, Colorado Meeting, May 18, 1971, on “The determination of silver in terrestrial and aquatic ecosystems.” (Details available from H. L. Teller, Colorado State University, Ft. Collins, Colo.) (3) H. L Kahn, G. E. Peterson, and J. E. Schallis, A?. Absorption Newslett., 7, 35 (1968). (4) F. P. Parungo and C. E. Robertson, J . Appl. Meteorol., 8, 315 (1969). ( 5 ) E. Bollay Associates, Final Report, Bureau of Reclamation Contract No. 14-06-D-5573,20 pp (1965).
ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973
lems involved in handling such small concentrations of silver. Untreated aqueous silver solutions are unstable, especially if the silver concentration is below 10-8 g/ml. Silver is adsorbed onto the surface of the container wall; therefore, the solution must be stabilized. Contamination is another problem. Most containers, solvents, and reagents are contaminated with silver a t this concentration level. The method developed and the procedure used for handling the samoles is discussed. EXPERIMEWAL Apparatus. One set of borosilicate glasswaLL specifically for the determination of silver. It consists of six 400-mI beakers, six 150-mlTeflon (Du Pont) stoppered separatory funnels, one 100-ml graduated cylinder, one 50-1111 graduated cylinder, one 10-ml graduated cylinder, and two 500microliter pipets. The glassware was purchased new, and not used for any other purposes. All glassware was cleaned with a solution which was 6N in HCI, 0.75 N in HF, and 1 % H202. When not in use, the separatory funnels were kept filled with 0.1N HNOa. The beakers and graduated cylinders were periodically soaked in 0.1N H N 0 3 , then rinsed several times with doubly-distilled water. The furnace used in this study was a third-generation furnace. It is essentially the same basic design as a secondgeneration furnace with two major exceptions-the outside jacket is double walled stainless steel for water cooling, and the shield tube and heat sink is constructed from one piece of graphite (6). These two features seem to have greatly increased the lifetime of the heater tubes. One particular pair lasted 10 months without burning out. Previous to that, the maximum lifetime was one or two months. The furnace temperature was measured through the side tube with an optical pyrometer. Furnace temperatures can be varied from 200 to 2800 "C. The temperature for the silver analysis was either 1650 or 1800 "C. The power dissipated to acquire these temperatures was about 2500 to 3000 watts. The furnace is powered with a dc arc welder. A picture of the instrumentation is shown in Figure 1. The major components of the system are: the furnace, a Spex 8/r-meter Czerny-Turner Spectrophotometer, a Princeton Applied Research (PAR) HR8 lock-in amplifier and light chopper, a Honeywell 6-inch recorder, and the associated electronics (Hewlett-Packard and Fluke power supplies, Tektronix oscilloscope, etc.). The wavelength of the monochromator was set on the most sensitive silver line (3280.7 A). The hollow cathode light was chopped a t 400 Hz. The oscilloscope was used to monitor the signal from the photomultiplier, an RCA IP28. A chart speed of 2 min per inch was used to record the signal. Figure 2 shows typical traces. Reagents. The reagents used were dithizone (diphenylthiocarbazone), carbon tetrachloride, and nitric acid. The reagent grade dithizone was recrystallized five times to ensure its purity. The carbon tetrachloride was distilled, as was the nitric acid. All reagents were independently checked for silver. The water used for rinsing the glassware and making dilutions was doubly-distilled in borosilicate glass. Procedure. Snow samples were collected by two methods. In the first, new plastic liners were placed in plastic garbage cans and were exposed to snowfall in several locations in the Bridget Mountain Range of southwestern Montana. This range serves as the target area for a winter orographic cloud(6) Ray Woodriff and Douglas Shrader, ANAL CH~M., 43, 1918
(1971).
Figure 1. Picture of the instrumentation 1. The hollow cathode 2. PAR Model BZ-1 light chopper 3. The Furnace--a third generation model 4. Spex '/, meter Czerny-Turner Spectrophotometer 5. Honeywell &inch recorder 6. Tektronix type RM 503 oscilloscope
I. PAR HR8 lock-in amplifier
LA 1200
1
2d-L. 00
2000
1
Figure 2. Temps
seeding experiment being conducrcu uy iviuriiaira ~JLLLLC U L W versity. The plastic liners were collected and new liners exposed every few days to one week. As the experiment was randomized and used a 24-hour experimental unit, any particular snowfall sample may have occurred during a completely-seeded, partially-seeded, or completely-nonseeded period. A second method of obtaining snow samples involved the digging of pits through the entire snow pack. This was done after the end of seasonal seeding operations in March, 1971. Pits were dug generally downwind (east) and crosswind (south) of the Bangtail Ridge target area as well as in the primary target area. Samples were also taken near the seeding generators. All pit sample locations are shown in Figure 3.
ANALYTICAL CHEMISTRY, VOL. 45, NO, 2, FEBRUARY 1973
231
Figure 3. Map of snow sample sites
SNOW SAMPLE SITE
After each snowpit was excavated to ground level, one wall was dug back with a clean plastic shovel. Three side-by-side samples of the pit wall were then placed in plastic bags for three separate determinations of silver concentration. Care was taken to obtain equal volumes of snow from all levels of the snow pack. Considerable attention was also given to avoiding any physical contact with the snow except for the plastic shovel and interior of the plastic sample bags. All samples were kept frozen until just prior to the extraction step. The sample size varied from 0.5-5 liters of melted snow. The sample was broken up in the bag and mixed thoroughly. A 400-ml beaker was filled with snow and the sample was melted on a hot plate. Immediately after the snow was melted, 1 ml of 0.1N HNOp was added to a 100-ml portion of the melted snow sample, decreasing the pH to 3. The sample was poured into a separatory funnel and 5 ml of lO-SM dithizone-CCL solution was added. The funnel was shaken vigorously for about 45 seconds. After the layers separated, a portion of the organic phase was transferred to a 5-ml holding bottle. The organic phase could be stored for one day under refrigeration with no appreciable losses of Ag. Using a 100-p1 Eppendorf pipet, 500 pI of the organic phase was placed onto a graphite cup and evaporated under an infrared lamp. The volume of a cup was about 120 p1. Duplicate cups for each sample were run. After the sample was on the cup, it could be stored for several days in a desiccator. For determination, the cups were screwed onto a '/*-inch graphite rod and inserted into the furnace. The results are reported as grams of Ag/ml of melted snow. The normal concentration range for silver in snow is on the order of lo-" g/ml(lOV g/ml = 1 ppm). RESULTS AND DISCUSSION
Several possible methods of sample preparation for silver concentrations in the picogram to nanogram range have been studied by the authors (7) as well as by others (1, 2 , 4 ) . Some (7) Ray Woodriff and Duane Siemer, Appl. Spectrosc., 23, 38 (1969).
232
of the methods were: coprecipitation of the silver with a carrier and subsequent centrifugation into a cup for analysis, preconcentration of the silver using an anion-exchange column, electroplating the silver onto a Pt or Ir filament for emission spectroanalysis in a R F induced discharge, and extraction of the silver into an organic solvent with some chelating agent for preconcentration. After carefully studying these methods, the last was chosen as the simplest, most reliable, and practical method. Dithizone (diphenylthiocarbazone) was chosen as the best chelating agent for silver because it has a high degree of specificity for silver over a wide pH range. The primary silver dithizonate, which is formed at pH's below 7, is soluble in organic solvents and practically insoluble in water. Chloroform is the usual solvent for dithizone because metal dithizonates are most soluble in chloroform. The silver concentration in chloroform, however, was too high (ca. g/ml) for it to be used. Several attempts were made to remove the Ag from the chloroform, but none were successful. Carbon tetrachloride was then selected as a solvent, and proved to be quite free from silver after being distilled once. The first attempts to extract the silver involved extracting 50 ml of melted snow into 5 ml of lO-5M dithizone in carbon tetrachloride and evaporating all 5 ml on a graphite cup. The background peaks were off scale using this technique. The sample size was cut in half, but the background peaks were still too high. Another problem encountered was high blanks due to the contamination of the glassware. Blank peaks ran between 20 and 50z scale deflection and were inconsistent. The glassware could not be completely cleaned of silver. Silver seems to slowly leach out of the glass over a period of weeks or months. This problem pertained mostly to the separatory funnels. Keeping the funnels filled with 0.1N HNOl between uses seemed to decrease the problem over a period of time, but did not entirely eliminate it. The two methods mentioned above were in essence putting 50 and 25 ml of melted snow on the cup. Since the Ag concentration was too high, the method presented was developed.
ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973
Table I. Comparison of FAA with Other Analytical Methods (Results in g/ml X 10'0 of Ag) Snow Boat AA NAA FAA sample (Salman) (Warburton) (Culver)
1.2'
1.0-
27.9 20.0 24.8 11.0 1.7 3.8 0.2 lost 5 lost 0.3 0.9 6 1.5 0.3 7 13.0 0.8 8 1 .o 0.6 9 3.7 0.4 10 2.4 11 0.6 0.2 lost 12ab 1.4 12b" 0.6 0.1 13 0.7 0.2 14 1.4 0.5 15 0.6 0.1 16 lost 0.2 a Within 20 m of seeding generator. b Unseeded area. c Fresh snow unseeded area.
1" 25 3 4
-
.8 0
z
a
m
cc 0
m
m
.6-
a .4
-
.2-
2
4 grams/,
6
8
10
12
II x IO of A g
Figure 4. Silver standard curves This new method not only decreased the Ag concentration to within the right range but also eliminated the blank problem from the glassware and saved time. The method involved taking a larger sample (100 ml), but only evaporating l / , ~of the extractant onto a cup (0.5 ml). This technique is equivalent to evaporating 10 ml of melted snow on a cup while at the same time cutting the blank one order of magnitude. The working range was 3 X g/ml up to about 3 x 10-1O g/ml. Some highly-seeded samples still went off scale, but in these cases a smaller portion of the extractant (200 pl) was evaporated on another pair of cups and rerun immediately. This step assumes the extraction coefficient is constant over the Ag concentration range in question. If this is not true, it may account for the lower values for samples 1 and 2 in Table I. Another problem encountered in handling samples containing such minute quantities of silver is that of surface adsorption. Silver has a tendency to adsorb onto container walls if the solution is not properly treated (8,9). This is why it is necessary to acidify the sample as quickly as possible after it has melted. Once the sample is acidified, the silver loss is minimized. The silver was always extracted immediately to avoid any losses. The optimum furnace temperature to analyze silver was determined. Figure 2 shows that maximum sensitivity is achieved somewhere between 1400 and 1800 "C. Two Ag standard curves were run, one at 1650 "C and the other at 1800 "C (Figure 4). Although the 1650 "C curve is obviously more sensitive, the 1800°C curve is more linear over a wider concentration range. Both curves were made by running standard silver solutions through the extraction procedure and plotting grams of Ag/ml of aqueous solution us. absorbance. (8) F.K.West, P. W. West, and F. A. Iddings, ANAL.CHEM.,38, 1566 (1966). (9) F. K.West, P. W. West, and T. V. Ramakrishna, Enciron. Sci. Terlinol., 1, 717 (1967).
4.2 2.6 0.8 0.8 0.7 0.7 0.3 0.6 0.5 0.3 0.5 0.1 0.1 0.5 0.6 0.5 0.5
Table 11. Comparison of Some Samples Determined on Different Days (Results in g/ml X 1O1Oof Ag) Analyses Sample No. 1 Date No. 2 Date No. 3 Date 105 0.17 2/19/71 0.53 6/24/71 0.44a 106 0.46 2/17/71 0.70 6/29/71 0.635 108 0.81 5/22/71 0.46 6/23/71 0.56 109 0.23 2/19/71 0.61 6/24/71 0.20 111 0.29 2120171 0.57 6/24/71 0.52 116 1.10 6/24/71 0.90 1 20 0.32 6/24/71 0.45 a Results are the average of triplicate determinations.
7/13/71 7/13/71 7/13/71 7/13/71 7/13/71 7/13/71 7/13/71
Over 225 snow samples were determined for Ag using the above mentioned method. The 17 snowpit samples were triplicated and sent to two other laboratories, Bureau of Reclamation Laboratory in Denver, Colo., and Desert Research Institute near Reno, Nev. In Denver, the samples were determined using a flame AA microsampling boat technique. The samples sent to DRI were done by Warburton using neutron activation analysis. Table I shows some results of these methods as compared to the furnace technique. In most cases the samples determined a t D R I are in good agreement with the furnace method. The results obtained in Denver are generally higher than the other methods. Using the furnace, some samples were found to have a 35% silver gradient from one end of the sample to the other. Four portions from one end of a sample and three from the opposite end were determined. The coefficient of variation of the quadruplicate and triplicate samples was less than + 5 %. This indicates a necessity for thorough sample mixing. The triplicate samples sent to the various laboratories mentioned above were carefully taken adjacent to each other and should have contained the same amount of silver. Of the 225 samples analyzed, 22 were unsatisfactory (i.e., the duplicates were quite different). These samples were redetermined with the tags coded to prevent prejudice on the part of the chemist. Some of the results are shown in Table 11. Every one of the redetermined samples was either almost equal to one of the other two earlier results or fell between the
ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973
233
two earlier results. The variation is probably due to the silver gradient mentioned earlier or to the increased handling, thus increasing the possibility of contamination in some of the determinations.
searchers including L’vov (IO), Massman ( I I ) , and West (12) have also worked on nonflame methods for atomic absorption. Of the four nonflame methods mentioned, only that of L’vov is not commercially available in this country.
CONCLUSIONS
RECEIVED for review June 30, 1972. Accepted October 5, 1972. We wish to thank the National Science Foundation for their support of this research under Grant No. GP-28055. The U. S. Office of Education provided support for one of the investigators i.1 the form of a n NDEA fellowship. Also, this research was partially supported by the Division of Atmospheric Water Resources Management, Bureau of Reclamation, U. S. Diipartment of the Interior, under Contract No. 14-06-D-6798.
The method developed and the procedure used for determining microtrace concentrations of Ag in snow by furnace atomic absorption have been presented. Comparison of results with other methods of determination show that the FAA method gives results comparable with those obtained by neutron activation analysis. FAA’s reproducibility is much better than neutron activation analysis in the concentration ranges involved and determinations can be performed for a fraction of the cost. Thus, it is a very practical method of determination. FAA has been developed in our labs over the past several years. The first paper was presented on the method a t the National SAS Meeting in 1966. Other re-
(10) B. V . L’vov, Spectrochim. Acta, Part B, 24, 53 (1969). (11) H. Massman, ibid., 23, 215 (1968). (12) T. S . West and X. K. Williams, Anal. Chim. Acta, 45, 27 (1969).
Atomic Absorption Determination of Nanogram Quantities of Tellurium Using the Sampling Boat Technique Richard D. Beaty Department of Chemistry, Unicersity of Missouri-Rolla,
Rolla, M o . 65401
A method for the determination of ultra-trace quantities of tellurium has been developed, utilizing the sampling boat technique of atomic absorption. Two procedures were developed for the chemical separation of tellurium. In some types of samples, the tellurium can be directly extracted from 4M HCI solution into methyl isobutyl ketone. The relative standard deviation obtained for a typical sample treated in this manner and analyzed by atomic absorption was 5.0%. In samples where other constituents cause chemical interference with the extraction, a preliminary separation of tellurium by coprecipitation with selenium was employed. The relative standard deviation using this procedure increases to 6.6%, but few interferences are observed. Linear response occurs for a range of 5-100 ng tellurium.
THEABILITY to accurately analyze very small amounts of tellurium has become important with the arrival of a n increasing awareness of trace constituents in the environment. While tellurium is regarded as toxic, the toxicological properties of trace quantities of this element are generally unknown. A sister element, selenium, has been found to have a wide distribution in plants ( I ) , and this occurrence of selenium is attributed with having caused many livestock poisonings in western United States. While a similar study of tellurium has never been undertaken, it is logical to assume that this element could, at least, be a cocontributor in such incidents. Selenium has, also, been shown to exhibit interesting effects as an antidote for mercury poisoning (2), which prompts more (1) I. Rosenfeld and 0. A. Beath, “Selenium Geobotany, Biochem-
istry, Toxicity, and Nutrition,” Academic Press, New York, N.Y., 1964. (2) H. Ganther, C. Goudie, M. L. Sunde, M. J. Kapecky, P. Wagner, S. Oh, and W. G. Hoekstra, Science, 175, 1122 (1972). 234
interest in the unexplored toxic and nutritional role of trace amounts of tellurium. Another area in which reliable trace tellurium analyses would be of value is in the field of geochemistry. Data on the distribution of tellurium in the Earth’s crust are virtually nonexistent, except in areas of tellurium mineral deposits. More information on tellurium distribution would permit a better understanding of the geochemistry of this element. Existing methods for low-level tellurium determinations leave room for improvement. An analytical method for determining the general distribution of this relatively rare element in the environment would require the utmost sensitivity in the lower parts per billion range. The difficulty in sample decomposition of geological materials would limit the sample size to no more than one gram, and thus the method would have to be sensitive to a few nanograms of tellurium. Photometric methods are generally not sensitive enough and are subject to too many interferences to be seriously considered. Fluorescence has been used for tellurium at the nanogram level (3, 4), but with the inconvenience of working at liquid nitrogen temperatures. Tiptsova-Yakovleva et al. ( 5 ) report using ac polarography to achieve a tellurium sensitivity of about 50 ng/ml in the determination of cadmium and zinc tellurides, but electrolyte conditions are critical and would present a formidable separation problem in geological samples. Several papers have dealt with neutron activation (3) G. F. Kirkbright, C. G. Saw, and T. S . West, Analyst (Lortdon), 50, 457 (1969). (4) G. F. Kirkbright, C. G. Saw, and T. S. West, Tala/zra, 16, 65 ( 1969). ( 5 ) V. G. Tiptsova-Yakovleva and Yu. A. Figel’son, Zh. Anal. Khim., 23, 1415 (1968).
ANALYTICAL CHEMISTRY, VOL. 45, NO. 2, FEBRUARY 1973
