Table 11. Standard Additions

Analysis of EPA Reference Samples by

Concentration, ppb

Dilution Re1 Av Sample factor Obsd Accepted error, % recovery, %

EPA 1 1 : l O 2.43 2.60 -6.5 90.7 i 3.3 EPA 2 3:50 4.25 4.38 -3.0 93.9 i: 7.2 EPA 3 1:50 5.75 5.56 +3.4 97.4 i 3.3

increase in absorbance which is twice as large as the uncer- tainty associated with a blank analyzed in identical fashion. This 0.2-ppb detection limit can be reduced even further by using a larger arsine generator which can accommodate larger volumes of water sample. Alternatively, the SDDC solution after arsine trapping can be further concentrated by controlled evaporation of the solvent, the As absorbance increase being proportional to the volume contraction. Attempts to reduce the volume of the As-SDDC solution by more than a factor of 5 resulted in a loss of As due to precipitation of SDDC and/or complex. Using this latter approach, we have extended the detection limit to below 0.05 ppb for a 50-mL sample. Absorbance measurements employing pyrolytically coated graphite furnace tubes (9) did not differ significantly from those utilizing uncoated tubes.

To test the accuracy of our method, three EPA reference samples were analyzed by standard additions. The results of these analyses are summarized in Table 11. The reference samples were diluted so as to bring them into the range of arsenic concentration generally observed in natural waters. The average relative error of the method is typically less than *5%. Using the sensitivity computed from the data of Figure 5, the average percent recovery of the arsenic added to each of the EPA samples during the standard addition analysis can be calculated, and these results are also summarized in Table 11. The somewhat less than quantitative recovery may be due to a slight suppression of arsine generation by heavy metals present in the EPA reference samples (10, 11). In each case, better than 90% recovery was obtained.

Selected river water samples from various parts of North Dakota were analyzed by the method described in this report. In every case the standard addition curves exhibited good

linearity with As recoveries comparable to those achieved with the EPA samples. To illustrate a typical result, an analysis of a sample taken from Cedar Creek, North Dakota, yielded an As content of 8.9 ppb for an undigested aliquot and an As content of 17.8 ppb for an aliquot digested in HzSOd/HN03 prior to analysis. These results indicate that a substantial portion of the total arsenic in this particular sample exists in one or more organic forms not detected by our method without prior digestion of the sample. Under the acidic conditions of our arsine generation reaction, many of the organo-arsenic forms found in natural waters should be re- duced to the corresponding organo-substituted arsines (12). Certain of these organic forms, however, may not be trapped by the SDDC solution, possibly because of weak complex formation with SDDC or low vapor pressure of the higher molecular weight species. In any event, to obtain the total organic plus inorganic arsenic by this method, digestion of the sample is essential. We are currently exploring the nature and extent of interaction of various arsenic forms with SDDC as well as a method involving injection of the various arsenic hydrides directly into the graphite tube furnace. The results from these investigations will be reported a t a later date.

LITERATURE CITED (1) “Standard Methods for the Examination of Water and Waste Water,”

Michael J. Taras, Ed., American Public Health Association, Washington, D.C., 1971.

(2) F. J. Fernandez, At. Absorp. News/., 12, 93 (1973). (3) R. C. Chu, G. P. Barron, and P. A. W. Baumberger, Anal. Chem., 44,

1476 (1972). (4) P. D. Goulden and P. Brooksbank, Anal. Chem., 46, 1431 (1974). (5) F. D. Pierce, T. C. Lamoreux, H. R. Brown, and R. S. Fraser, Appl.

Spectrosc., 30, 38 (1976). (6) P. F. Wyatt, Analyst (London), 80, 368 (1955). (7) J. F. Kopp, Anal. Chem., 45, 1786 (1973). (8) G. C. Clarke in “Isolation and Identification of Drugs”, Pharmaceutical

Press, London, 1969, p 327. (9) R. E. Sturgeon and C. L. Chakrabarti, Anal. Chem., 4g, 90 (1977).

,(lo) F. D. Pierce and H. R. Brown, Anal. Chem., 48, 693 (1976). (11) A. E. Smith, Analyst(London), 100, 300 (1975). (12) D. L. Johnson and R. S. Braman. Deep Sea Res., 22, 503 (1975).

RECEIVED for review January 24, 1977. Accepted April 15, 1,977. Financial support from the U.S. Department of the Interior, Office of Water Resources Research (C-6307) and from the Environmental Protection Agency (R803727-01-1) is gratefully acknowledged.

Comparisons of Methods of Hydride Generation Atomic Absorption Spectrometric Arsenic and Selenium Determination

Darryl D. Siemer” and Prabhakaran Koteel Department of Chemistry, Marquette University, Milwaukee, Wisconsin 53233

Experiments performed with simple apparatus to optimize hydride generator atomlc absorption arsenic and selenium determinations are described. Absolute sensitivities of 5.5 X lo7 and 1.5 X lo7 absorbance unlts/g are obtained by freezing out arsenic and selenlum hydrides in a “U” tube prlor to introductlon into a quartz tube burner. These sensitlvities are compared both with published values obtalned with other hydride-AA systems and wlth graphite furnace sensitivites to obtain relative system efficlencies. Many hydrlde-AA techniques in use are not very efficient In generating analytical signal per gram of analyte metal taken.

Atomic absorption (AA) has an advantage over other atomic

spectroscopic techniques in that the signal measured (ab- sorbance) is a direct indication of the number of atoms present per unit atomizer cross section per unit time which is to a large degree independent of the stability and intrinsic brightness of the light source. It is therefore possible to directly compare the efficiencies of various atomizers used in atomic absorption by looking a t values of published sensitivities. The term “Sensitivity” will be used throughout this paper to mean “the absorbance signal observed per gram of metal”. This is not the traditional AA definition of the term but it is in line with those used by practitioners of other disciplines who publish in this journal. Stone ( 1 ) has made this comparison for a variety of nonflame graphite furnaces and has shown that atomizers, which are large enough to contain all of the sample

1096 ANALYTICAL CHEMISTRY, VOL. 49, NO. 8, JULY 1977

atoms for a time sufficient to prevent diffusional loss until all of the sample has been atomized, produce sensitivities approaching the maximum values theoretically possible. This paper will compare various representative published hydride generation atomic absorption arsenic and selenium analytical techniques as to how efficiently the processes generate atomic absorbance signals per unit mass of trace metal.

Hydride generation atomic absorption (HGAA) techniques are much more sensitive than direct flame atomization largely because the analyte metal in from 2 to 100 cm3 of sample solution can be introduced into the atomizer in a far shorter time than possible if the solution itself were nebulized. Additional advantages accrue because conventional nebulizer efficiencies are rarely better than 10% and the hydride generation step separates the analyte from matrix concom- itants which often severely interfere in direct analysis. In comparison with furnace atomization, most published hydride generation AA approaches are from one to three orders of magnitude less sensitive on a signal per gram analyte basis but are usually more sensitive on a signal per unit concen- tration basis. The furnace techniques, however, are often unreliable because of high and variable matrix absorption or scattering signals seen in many “real” sample solutions. For these reasons the hydride based analytical techniques are more popular than either of the other AA procedures in laboratories charged with trace metal analyses of samples of widely varying composition.

There are two fundamentally different ways of imple- menting HGAA analyses. The simpler is to convey the metallic hydride as soon as it is produced directly into the atomizer (2-5). This type of procedure will be referred to as the “continuous flow” method in this paper. Since the dead volume of the generator, connecting tubes, and so forth is generally much larger than the actual atomizer volume, only a small fraction of the hydride can actually be atomized at any given instant even if the generator is efficiently designed so that the hydride is immediately separated from the sample liquid. In fact, the rate limiting steps are often either the stripping of the hydride from the sample solution, or the rate of reduction of the analyte to the hydride. These facts render absorbance peak height measurements per unit mass of analyte subject to variation if solution stripping gas flow rates, solution viscosities or volumes, matrix elements, solution temperatures, degree of subdivision of solid reductant, etc., are changed. Absorption signal integration is often of very great utility in reducing the effects of changes in experimental parameters upon the apparent instrument response (3, 4).

The other way of implementing HGAA analyses incor- porates a collection of the hydride as it is generated prior to its introduction into the atomizer. This has been done by either freezing it out in a “U” tube immersed in a liquid nitrogen bath (6-8) or by collecting the hydride as well as the hydrogen gas generated in a balloon (9-10). Both of these experiments, if correctly performed, should free the overall analytical method from error caused by variations in ex- perimental parameters affecting reduction or gas stripping kinetics. The freezing-out technique should also give much better sensitivity because i t allows all of the hydride to be introduced into the atomizer with far less dilution with carrier gas. However, because of the use of inefficient gas transfer systems or ineffective atomizers, the actual results achieved have been much poorer then theoretically achievable.

EXPERIMENTAL The AA spectrometer, hydride generator, methods of solution

preparation, and quartz tube burner system have been described previously (3 ,4 ) . In addition, a “U” tube was constructed of 5-mm 0.d. borosilicate glass tubing indented in the same manner as is a Vigreux fractionating column. This method of increasing surface area does so by increasing the area of glass in immediate contact

,, -7 ! $

G d

D Figure 1. Freezing-out hydride analytical apparatus. (A) burner, (6) tube from oxygen source, (C) capillary, (D) bypass stopcock, (E) drying tube, (F) hydride generator, (G) tube from hydrogen regulator

with the coolant and is more efficient than the usual expedient of filling the “U” tube with glass beads. In some of the exper- iments, the hydride was fed directly into the quartz tube atomizer and, in others, the hydride was first frozen out in “U” tube and then introduced into the atomizer as a “slug” after complete reduction and quantitative stripping from the solution.

Figure 1 depicts the hydrogen gas control system used when metallic hydride freezing-out experiments were implemented. Most of the tubing used was 4.5-mm i.d. Tygon and lengths were reduced to practical minimums to reduce dead volume. The hydrogen regulator was set to about 20 #/inch2 and the needle valve was adjusted to keep the total HZ flow to about 2500 cm3/min (corrected for viscosity) as measured by a Gilmont flowmeter. This system keeps the total hydrogen flow entering the burner practically constant irregardless of the bypass stopcock position. A short length of 0.3-mm i.d. capillary tubing served to create enough back pressure in the line to ensure that when the bypass stopcock was fully opened, about 75% of the hydrogen gas flow went through the generator-drier-“U” tube circuit. The oxygen flow was kept constant at about 25 cm3/min by means of a high impedance sintered metal frit inserted between the tank regulator and the tubing connecting the oxygen supply to the burner.

The drying tube depicted in the figure was filled with 4 mesh calcium chloride because we found as did McDaniel et al. (6) that “drierite” absorbs H2Se (but not arsine).

In freezing-out experiments ap- proximately 4 mL of solution (6 M in HC1) were injected into the generator. The “U” tube was immersed in a dewar flask full of liquid nitrogen. The bypass stopcock was adjusted to allow approximately 100 mL/min of hydrogen to flow through the generator. Then 1 mL of 1% sodium borohydride solution was added through the septum. After approximately 1 min, the bypass stopcock was closed and the “U” tube placed into a beaker of room temperature water. After about 10 s, the stopcock was fully opened driving the metal hydride into the burner. The spec- trometer recorded the maximum atomic absorbance with an electronic peak retrieval system.

RESULTS AND DISCUSSION The results of arsenic and selenium analyses performed with

our apparatus both with continuous hydride introduction and the freezing-out technique are listed in Table I. Freezing out the hydrides gave approximately twice as high peak absorbance values but, as expected, did not affect integrated atomic absorbance signals. Our HGAA values reduced to units of absorbance per gram of arsenic are listed along with some representative values for other published procedures in Table 11. The nonflame graphite furnace data are included to represent a standard to which the hydride techniques can be compared. Note that the analytical responses of the Woodriff, Perkin-Elmer 2100, and IL furnaces normalized to a uniform atomizer cross section are quite similar. Therefore it is reasonable to expect that the sensitivity per unit cross-section value achieved by these atomizers represents the maximum achievable by atomizers operated at atmospheric pressure. That is to say that the three atomizers all succeed in both conversion of analyte solutions to atoms with high efficiency

ANALYTICAL CHEMISTRY, VOL. 49, NO. 8, JULY 1977 1097

Ana ly t i ca l Procedure.

Table I. 50 ng SeIV in 4 mL

6M HC1, 1960 nm

Selenium Analysis and Arsenic Analysis

Continuous introduction Freezingout‘

Peak height, absorbance 0.346 = 0.707 n = l l RSD, % 2.1 3.0

RSD, % 2.1 5.2 Peak area 250 n = 6 249 n = 5

12.5 ng As”’ in 4 mL 6M HC1, 193.7 nm Peak height, absorbance 0.237 = 0.681 = 8 RSD, % 3.8 1.4 Peak area RSD, %

Three times the standard deviation of the blank for the freezingout-peak height measurements corresponded to 3.5 X l o - ’ ’ and 5.1 X g for arsenic and selenium, respectively. Analytical curves are linear from 0 to ap- proximately 0.7 absorbance unit.

and also containment of those atoms so that they all reside in the optical path simultaneously for a time sufficient for the spectrometer to make an absorbance measurement.

The data in the table also includes sensitivities obtained by HGAA techniques. These sensitivities have also been normalized whenever possible to a common atomizer cross

section. The last column in the table lists a figure for “efficiency” of the systems calculated by dividing the observed sensitivity per unit cross-section figure by the mean of the same parameter observed with the three graphite furnace atomizers. An effective cross section of 1 cm2 was assumed for the open flame atomizers. The “efficiency” figure indicates how close a hydride based analytical technique comes to generating the maximum possible analytical response (ab- sorbance) for a given mass of analyte. It is a measure of how well the generation system separates the arsine into a min- imum of carrier gas; how efficiently the gas is conveyed to the atomizer; and of how effectively the atomizer both converts the arsine to arsenic atoms and contains those atoms.

Table I11 gives a compilation of similar data taken from various sources pertaining to selenium determinations. Calculations similar to those in Table I1 have been made. It is apparent that many of the published HGAA techniques are not very “efficient”. There are many reasons for this. In many procedures an inert carrier gas (argon or nitrogen) is used to carry the hydride into the atomizer. This dilutes the hydride with so much extraneous gas that the atomizer cannot possibly contain all of the arsenic a t one time. In other procedures, the apparatus has a very large dead volume which again causes the hydride to be diluted. In techniques involving balloon storage of the hydride, large volumes of hydrogen gas are generated along with the metal hydride and cause problems. For example, in Chu’s (9) procedure (one of the more efficient

Table 11.

introduction4 Reference Atomizer cm Abslg X area Efficiency, %c

Arsenic Sensitivities 193.7 nm Line Sample Diameter, Sensitivity, Sensitivity

A Stone ( I ) Woodriff 0.8 1.7 X 10’ 8.8 x 1 0 7

A Perkin-Elmerb Model 2100 0.61 2.2 x 108 6.4 x 107

furnace

furnace

furnace

tube

A ILb Model 555 0.5 7.3 x l o 8 1.4 X 10*

B This work Flame in 1.0 1.9 x 107 1.5 x 107 15

B Duncan ( 2 ) Open flame ( ? I 2.2 x 106 2.3 B Smith (5) Open flame ( ? I 4.4 x 10’ 0.45 C This work Flame in 1.0 5.5 x 107 4.3 x 107 44

C Griffin (8) Open flame (?) 4.4 x 104 0.045 C Knudson ( 7 ) Open flame ( ? I 1.8 X lo6 1.8 C Knudson ( 7 ) HGA 2000 0.8 4 . 4 x l o 6 2.2 x l o 6 2.3

D Chu (9) Open flame (?I 2.6 x 105 0.27 D Chu ( 9 ) Heated quartz 2.3 8.8 X 10’ 3.7 x l o 6 3.8

tube

furnace

tube A, Graphite furnace atomization of discrete aliquots of solutions.

Hydride first frozen in “U” tube and then introduced into atomizer. duced into atomizer.

late efficiencies of HGAA svstems.

B, Hydride swept into atomizer as it is generated. C, D, Hydride first collected in a balloon and then intro-

Sensitivity data obtained from sales representatives of the instrument company selling the atomizer. A mean value of (sensitivity X area) of 9.7 X l o ’ Abs.cm2/g based on the reported graphite furnace data was used to calcu-

Table 111.

introduction4

Selenium Sensitivities, 196.0 nm Line Sample Diameter, Sensitivity, Sensitivity

Reference Atomizer cm Abslg X area Efficiency, %c

A Stone (1 ) Woodriff furnace 0.8 7.3 x 107 3.7 x 1 0 7 A ILb Model 555 . 0.5 4.4 x l o 8 8.6 X l o 7 A L’vov (1 1 ) L’vov furnace 0.25 4 .8 X 10’ 2.4 x 107 B This work Flame in tube 1.0 7.3 x lo6 5.7 x l o 6 12 B Duncan (2) Open flame (?I 9.9 x 105 2.0 B Smith (5) Open flame (?) 2.2 x 105 0.45 C This work Flame in tube 1.0 1.5 x 107 1 .2 x 107 24 C Knudson ( 7 ) PE Model 2000 0.8 1.1 x 105 5.5 x l o 4 0.11 C McDaniel (6) PE Model 2000 0.8 4.0 x 1 0 5 2 .0 x 1 0 5 0.41

See Table I1 footnotes. See Table I1 footnotes. The mean sensitivity X area value ( 4 . 9 X l o 7 Abs.cmz /g) for the 3 graphite furnaces was used for calculation of efficiencies,

1098 ANALYTICAL CHEMISTRY, VOL. 49, NO. 8, JULY 1977

systems) the 1 g of zinc used’can be predicted to produce about 350 cm3 of hydrogen. When this is swept into his very large quartz tube atomizer (about 60 cm3 volume) and heated to a quoted temperature of about 700 “C, the hydrogen will expand to about 1200 cm3. Division of the atomizer volume by that of the heated gas bearing the hydride indicates that it would be impossible to achieve more than 5% of the maximum possible absorbance signal. His observed efficiency is about 4% indicating that the atomizer used is fairly efficient in converting the hydride to atoms.

In theory it should be possible to achieve the best HGAA absolute sensitivities with the freezing-out, slug introduction techniques because they allow the metal hydride to be sep- arated from the relatively huge amounts of hydrogen gas created during the reduction step. These techniques should also allow the best sensitivities on an absorbance/concen- tration basis too because the kinetics of solution stripping and reduction of large volumes of sample solution no longer affect the rate of hydride introduction into the burner. However, it is apparent from Tables I1 and I11 that, in practice, freezing-out steps do not always result in improved absolute sensitivities. McDaniel (6) achieved far poorer analytical responses per unit mass of selenium than did either Duncan (2 ) or Siemer (3) who used simpler continuous flow through systems. This is probably due to poor atomization efficiency in the atomizer used or due to inefficient transfer of the hydride to the atomizer. We have obtained “efficiency” figures of around 14% and 33% for continuous flow and freezing-out techniques, respectively, for both arsenic and selenium. As expected, the “efficiency” figure for our continuous flow apparatus is sensitive to analyte solution volumes (3) . As the sample volume increases, a fixed hydrogen flow has to strip the hydride evolved from a greater amount of liquid. This takes longer and reduces peak absorbance though it does not effect integrated absorbance signals ( 3 , 4 ) . However, freez- ing-out the hydride eliminates reduction/stripping kinetics as experimental parameters and also frees the method from many of the bothersome matrix effects which have been apparent in many of the published reports. As has been pointed out by McDaniel, it is necessary to use much more borohydride in the presence of some of these interferents than in their absence. In a continuous flow system, using a huge excess of the borohydride creates so much hydrogen that the flame is strongly perturbed or even blown out. Incidentally, with a freezing-out procedure (providing enough borohydride is used) the often observed difference in response seen for arsenic I11 and arsenic V disappears, indicating that the interference is due to reduction kinetics of the arsenic species ( 4 ) .

Another potential advantage of freezing-out procedures is that the hydride generation/stripping steps can be isolated from the actual atomic absorption procedure. There is no fundamental reason that it should not be possible to generate and transfer the metallic hydrides from a group of samples

into a series of labelled “U” tubes stored in a liquid nitrogen bath at one time and run the contents of the tubes into a suitable atomizer a t another time. This would save valuable instrument time. There is also potential for speciation of the various chemical forms that a metal may be found in real samples. For example, various forms of methylated arsenic form volatile hydrides with differing boiling points (12). It may prove possible to determine these hydrides by gradual heating of a “U” tube containing a frozen mixture of all of them and observing the arsenic peak integral as a function of time. A response appearing like an integral of a multipeak NMR spectrum would be expected with the height of each step proportional to the amount of each arsenic species.

For general analytical work we have found the simple continuous flow system adequate ( 3 , 4 ) . Signal integration frees the technique from most kinetic interferences and the procedure is very fast. The addition of a second aliquot of borohydride reveals the presence of persistent kinetic in- terferences by giving a second nonzero signal integral. However, we have found in practice, that ashed samples of foods, vegetation, and animal feeds do not possess these interferences ( 3 , 4 ) . Our hydride generator-atomizer system gives absolute sensitivities approaching the theoretical maximums and only little more improvement can be expected. Detection levels are presently largely determined by in- strumental noise because we are forced to use high photo- multiplier voltages and amplifier gains to process the signal. Substitution of a R106 for the R372 photomultiplier tube presently used in our instrument should alleviate this problem to a large degree.

ACKNOWLEDGMENT The authors express their thanks to Bob Emmel and John

Sotera of Instrumentation Laboratory, Inc., for their continued support of our research efforts.

LITERATURE CITED (1) R. Stone, W.D. Thesis, Montana State University, Bozeman, Mont., 1974. (2) L. Duncan and C. Parker, “Applications of Sodium Borohydride for Atomic

Absorption Determination of Volatile Hydriies”, Varian Techtron Technical Topic, Varian Techtron Pty, Ltd., Springvaie, Victoria, Australia.

(3) D Siemer and L. Hageman, Anal. Lett., 8. 323 (1975). (4) D. Siemer, P. Koteel, and V. Jariwala, Anal. Chem., 48, 836 (1976). (5) A. Smith, Ana/yst(London), 100, 300 (1975). (6) M. McDaniel. A. Shendrikar. K. Reismer. and P. West, Anal. Chem.. 48,

2240 (1976). E. Knudson and G. Christian, Anal. Lett., 6 , 1039 (1973). H. Griffin, M. Hocking, and D. Lowery, Anal. Chem., 47, 229 (1975). R. Chu, G. Brown, and P. Baumgarner, Anal. Chem., 44, 1496 (1972). D. Manning, At. Absorption. News/., 10, 123 (1971). B. V. L’vov, “Atomic Absorption Spectrochemical Analysis”, American Elsevier Publishing Company, Inc., 52 Vanderbilt Ave, New York, N.Y., 1970 p 228. R. Braman and C. Foreback, Science, 182, 1247 (1973).

RECEIVED for review March 10,1977. Accepted April 11,1977. Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for support of this research.

ANALYTICAL CHEMISTRY, VOL. 49, NO. 8, JULY 1977 1099