Atomic Absorption Spectrophotometry

Chapter 6

ATOMIC ABSORPTION SPECTROPHOTOMETRY

INTRODUCTION

Atomic absorption spectrophotometry involves the conversion of compounds to their constituent atoms and then excitation of the atoms by absorption of radiant energy. The amount of energy absorbed is measurable and is proportional to the concentration of absorbing atoms. Although atomic absorption was described by Kirchoff and Bunsen in 1860, and then used to identify the elements present in the solar atmosphere, its potential applications to analytical chemistry, with the notable exception of the determination of Hg, went unfulfilled for almost one hundred years. Then, in 1955, Walsh described both the theoretical applications of AAS to analysis and the design of an atomic absorption spectrophotometer utilizing a hollow cathode light source (Walsh, 1955). Subsequently, AAS expanded rapidly into many fields of analysis, including exploration laboratories where it soon replaced colorimetric procedures as the principal method for many trace metal determinations (Table 1-V; Fig. 1-9). Although already being partly replaced by ICP-ES in some large laboratories, it seems likely that AAS will continue to be the principal analytical method in most exploration laboratories over the next decade.

THEORY

Atomic absorption and atomic emission are related phenomena arising from changes in the energy levels of an atom's outer electrons (Fig. 6-1). If these are excited, for example in a flame or electric arc, they are raised from their neutral, or ground state, to higher energy levels, the transitions involved depending on the atomic configuration of the particular element and the selection rules of quantum mechanics. In collapsing back from higher to lower energy levels electrons loose energy by emission of photons whose energy, and hence wavelength, corresponds exactly to the difference between the initial and final energy levels of the transition. Each transition therefore gives rise to a narrow emission line at a fixed wavelength and each

110 High energy level, E2

Ä absorption Low energy level, E-j

Fig. 6-1. Excitation of an electron from a lower (£1) to a higher (#2) energy level by absorption of energy (E^) where E^ = E2 — E\. On collapsing back from E2 to E\ the electron releases a photon with an energy equal to E± and hence, from Planck's Law, a wavelength in nanometres (nm) of 1.24/ü^, if E± is in kilovolts.

atom (or ion) has its own unique emission spectrum (Fig. 6-2). Conversely, excitation of an electron to a higher energy level can result from absorption of radiant energy of the same wavelength as the emission line for that particular transition. This is the basis of atomic absorption spectrophotometry. The ratio of excited (N{) to ground state (N0) atoms at different temperatures is given by: i

Pi

(-ΕΛ

N, where P{ and P0 are statistical weights for the ground and excited state, Ei is the energy difference between the states, k is the Boltzmann constant, and T is temperature in °K. As the value of Ei increases (i.e. as wavelength decreases according to Planck's Law) the smaller will be the fraction of excited atoms for a given temperature. As temperature increases the number of excited atoms increases. However, even at high flame temperatures the proportion of excited atoms remains relatively low (Table 6-1). Atomic absorption measurements are usually made in flames with temperatures between 2000 and 3000°C so that virtually all the atoms are in the ground state (or energy levels very close to the ground state) and N0 is relatively insensitive to small fluctuations of flame temperature. Also, because only absorption lines originating in the ground state (or close to the ground state e.g. Sn 235.48 nm and Zr 360.12 nm) will be sufficiently abundant to cause measurable absorption, the absorption spectrum is much simpler than the emission spectrum (Fig. 6-2). The relationship between the intensity of an incident (I0) and transmitted (It) light beam, of wavelength λ, passing through an absorbing cloud of atoms follows Beer's law: It = I0 exp — (Kxcl)

Ill

10,000

20,000

30,000

40 000 E υ

50,000

μβο,οοο

^70,000

Fig. 6-2. Grotrian energy level diagram for the Au atom. The wiggly arrows indicate a transition that is forbidden by the selection rules, a and b are metastable levels. Only those wavelengths associated with the ground state (i.e. 164.6 nm, 242.8 nm and 267.6 nm) are involved in atomic absorption but the line at 164.6 nm is too low in the UV to be useful and the 242.8-nm line is usually used for AAS determination of Au. (Slightly modified from Ahrens and Taylor, 1961, Spectrochemical Analysis, 2nd ed., with permission of Addison-Wesley, Advanced Book Program.)

where Κλ is the absorption coefficient for the particular wavelength (λ), c is the concentration of atoms, and / is the length of the absorbing path. Therefore: logioCo/Λ) = K\cl = absorbance (A) With J0, as percent transmission, set to 100% (J0 — h) gives percent absorption, and: 100 absorbance {A) = log 100-(/0-/t)

112

TABLE 6-1 The relationship between temperature, wavelength and the number of excited (Nj) to neutral (N 0 ) atoms Absorption line (nm)

Pi IPo

Na Ca Cu Zn

2 3 2 3

589.0 422.7 324.8 213.8

1

Ni/N 0 2000°K 9.89 1.21 4.78 7.25

X X X X

3000°K 10~6 10" 7 10" 1 0 10" 1 5

5.81 3.54 7.70 5.40

X X X X

5000°K 10" 4 10" 5 10" 7 10" 1 0

1.51 3.31 2.83 4.27

X X X X

10" 2 10" 3 10" 4 10" 6

I\ and P0 are statistical weights derived from the quantum number J such that P = 2 J + 1: e.g. Zn 213.8 arises in the transition 'S0 -+ 'Px so that P0 = 1 and P{ = 3.

so that 1% absorption = 0.0044 absorbance units. In practice, the atomic absorption instrument is calibrated with standards of known concentration and it is therefore not necessary to know the values of either Κλ or /. The sensitivity for the determination of an element by AAS is defined as the concentration required to give a 1% absorption signal. Typical values are summarized in Table 6-II. The natural width of an absorption line is about 10" 5 nm. However, as a result of the Doppler effect and pressure broadening, caused by collisions between atoms, this is increased to 0.001—0.003 nm. Additional line broadening occurs in strong magnetic (Zeeman effect) or electrical fields (Stark effect). In some specialized applications, described further on p. 135, advantage is taken of pressure broadening or Zeeman broadening to measure and correct absorbance results for unwanted, non-atomic background absorption. INSTRUMENTATION

Schematic diagrams of two atomic absorption spectrophotometers are shown in Fig. 6-3. In the simpler, single-beam design, the instrument consists of: (1) a light source emitting the sharp line spectrum of the element to be determined, (2) a means of generating a cloud of atomic vapour, (3) a monochromator and slit to select the desired absorption line, and (4) a detector, amplifier and readout system. Components of the double-beam system are identical but the light from the source is split into two beams, one of which bypasses the atomic cloud and provides a reference against which instability in the output of the source can be corrected.

113

TABLE 6-II Crustal abundance of some trace elements compared to their analytical sensitivity by atomic absorption spectrophotometry Element *

Ag As Au B Ba Be Bi Cd Co Cr Cu Hg Mn Mo Nb Ni Pb Rb Sb Se Sn Sr Ta Te U V

w

Zn Zr 1 2 3

4 5

Crustal 2 abundance (ppm) 0.07 1.80 0.004 10 425 2.8 0.17 0.20 25 100 55 0.08 950 1.5 20 75 12.5 90 0.2 0.05 2 375 2 2.7 135 1.5 70 165

Sensitivity concentration for 1% absorption flame (Mg/ml)

3

0.029 0.92 0.11 8.4 (a) 0.20 (a) 0.016(a) 0.20 0.011 0.053 0.055(b) 0.040 2.2 0.021 0.28 (a) 19 (a) 0.06 0.11 0.03 0.29 0.38 0.39 (a) 0.041 (a) 11 (a) 0.026 113 (a) 0.75 (a) 5.8 (a) 0.009 9.1 (a)

furnace (Mg/ml per 5μ1) 4

other methods

0.00004 0.02 0.002

0.002

0.00018 0.0014 0.00002 0.0012 0.001 0.0014 0.02 0.0001 0.008 0.002 0.001 0.0012 0.006 0.02 0.012 0.001

5

(c)

0.0001 (d)

0.08 0.002

(c) (c)

0.02 0.000016

Elements in italics detectable by flame AAS at normal crustal abundances with a dilution factor of 50. From Levinson (1974). Lean air/acetylene flame unless otherwise stated: data based on manufacturer's literature for Techtron VI: a = nitrous oxide/acetylene flame, b = reducing air/acetylene flame. Manufacturer's data for Model 63 (Varian Techtron) carbon rod atomizer with 5-μ1 injections. Other methods: c = hydride generation; d = flameless cold vapour method.

114 electronic

modulation "o-'t nm i

· i

I i'

[ Ί

n ·m

*γ«

^

\

>t lo

LIGHT SOURCE

ABSORPTION

DETECTION

B. «o""t

~γ*

i m

Rotating mirror sectors

Half-silvered mirror

Fig. 6-3. Schematic diagram of atomic absorption spectrophotometers: (A) single-beam, and (B) a double-beam instrument, h = hollow cathode light source; m = monochromator; and p = photomultiplier. IQ is the intensity of the signal from the light source and It its intensity after absorption by analyte atoms in the flame.

The light source With a width of 0.001—0.003 nm, atomic absorption lines are too narrow to be isolated by small monochromators, but if an attainable bandpass of 0.1 nm is used complete absorption of a continuum source by a line 0.002 nm wide only results in a 2% change in signal strength — too small a difference to be measured accurately. Walsh (1955) realized that this dilemma could be overcome by having the light source emit only the spectrum of the element of interest. Ideally the emitted lines should be sharper (narrower) than the absorption line of the ground state atoms. For a few volatile elements, such as Hg, As and Se, electrodeless discharge lamps are ideal sources of an intense, stable emission. With most elements, however, the best source is the sealed hollow cathode lamp. This consists of

115

a tubular cathode, containing the element(s) of interest, and an anode mounted in a sealed envelope containing an inert gas, usually Ar or Ne, at pressures around 2 mm Hg. For work in the ultraviolet, where the strongest absorption lines of many metals occur, the lamps have a silica window. A stabilized DC power supply, providing up to 800 V and a variable current from 2 to 50 mA, is used to operate the lamp. Inert gas ions, produced at the anode, are accelerated to the cathode where they sputter off and excite atoms of the cathode causing them to emit the desired sharp line spectrum together with that of the inert gas. The lamp is designed to enhance emission of ground state transition lines relative to non-absorption lines, and, because of the low pressure in the lamp, line widths of the emitted spectrum are less than widths of absorption lines, at atmospheric pressure, in the atomic absorption flame. Apart from aligning the beam from the lamp through the atomic cloud, the only variable in operating the hollow cathode lamp is the current. Above and optimum current, which varies for each element, absorption decreases with increasing current and the calibration line shows increasing curvature (Fig. 6-4). This arises from several causes: (1) as current increases the number of atoms sputtered from the cathode increases, causing resonance broadening — a special case of pressure broadening — so that the wings of the broadened emission line may exceed the width of the absorption line in the flame and therefore not be absorbable; (2) all the atoms sputtered from the cathode are not excited; as the number of ground state atoms increases, selfabsorption of the central portion of the absorption lines becomes significant; and (3) as cathode temperature increases, the number and intensity of non-ground state lines in the emitted spectrum increases; if these lines are not resolved from the absorption line, curvature of the calibration line results. Despite the decrease in absorption associated with increasing lamp current above the optimum value, this may still be a justifiable procedure when mea-

Ni, j j g / m l

Fig. 6-4. Influence of varying lamp current (mA) and slit width (nm) on the sensitivity for Niat 232.0 nm.

116

suring very small absorption signals. At low concentrations calibration curvature is not a problem and the reduced signal amplification (gain) required if lamp current is increased may result in a worthwhile improvement of the signal to noise ratio and hence of the detection limit (Table 6III). Increasing the lamp current will also sometimes improve the signal to noise ratio if the analyte must be measured in a strongly luminous flame. Production of atomic vapour With the exception of Hg, which is sufficiently volatile to produce ground state atomic vapour at room temperature, some heat source is needed to volatilize the sample, dissociate chemical compounds and produce a cloud of absorbing atoms. The temperature must not, however, be too high or too many atoms will be thermally excited to energy levels above their ground state. Furthermore, it is in the establishment of complex chemical and ionization equilibria during production of the atomic vapour, that many of the interference problems associated with AAS arise. Many methods of producing ground state atoms have been investigated but only flames and, to a lesser extent, electrically heated furnaces have found general application. Flames Usually the sample solution is aspirated, by a venturi driven by the pressure of the oxidant gas, into a chamber where the resulting aerosol mixes with fuel gas before entering the flame through a long, narrow slit. Burners of this type, known as laminar-flow or pre-mix burners, produce very steady, stable flames. Aspiration rates with aqueous solutions are usually 3—5 ml/ minute. However, only a small fraction of the solution reaches the flame; the remainder, consisting of the larger droplets, is trapped by baffles in the

TABLE 6-III Modifications to routine operating conditions Situation

Action

A.

High analyte concentration — to decrease sensitivity

(a) use less sensitive absorption line (b) rotate burner

B.

Very low analyte concentrations — to improve signal to noise ratio

C.

Strongly luminous flames — to improve signal to noise ratio

(a) increase slit (b) increase lamp current (c) both (a) and (b) (a) decrease slit (b) increase lamp current (c) both (a) and (b)

117

mixing chamber and drains to waste. This low efficiency in transfering sample to the flame is one of the principal limitations on analytical sensitivity. A variety of gas mixtures have been used but the lean, oxidizing air/ acetylene flame has the most general application. Its temperature (^2200°C) is sufficiently high to ensure breakdown of most compounds but is not so high as to cause significant ionization for elements other than the alkalies. As would be expected the cooler (1800—1900° C) air/propane and air/coal gas flames are more prone to chemical interferences and are therefore seldom used. The hotter (2955°C), fuel-rich nitrous oxide/acetylene flame provides a reducing environment for determination of those elements — especially Mo in exploration samples — which tend to form refractory oxides (Fig. 1-9). Its relatively high temperature causes significant ionization of both alkali and alkali earth elements, and for Ba the ion absorption line at 455.4 nm provides better sensitivity than the atomic line (553.6 nm) unless ionization is suppressed (p. 128, Fig. 6-10). Wavelengths below 200.0 nm are strongly absorbed by the air/acetylene flame. Hydrogen/argon and hydrogen/nitrogen flames are more transparent in this region and are therefore recommended for measurements on the As and Se absorption lines at 193.7 nm and 196.0 nm, respectively. Whichever gas mixture is chosen, the heat of the flame is used to complete the process of evaporation which begins in the mixing chamber, and then to volatilize the solids and dissociate the chemical compounds producing a reservoir of atoms. The number of ground-state atoms produced depends on the aspiration efficiency and on the equilibrium established in the flame between the dissociation of compounds, formation of new compounds and the extent of ionization. The equilibrium condition is determined by flame temperature and composition, the solvent used, and on the other elements introduced into the flame by the sample. Consequently, production of ground state atoms will vary from sample to sample, causing chemical and ionization interferences, and, since flames are strongly zoned, in different parts of the flame. The distribution of absorbing atoms in a laminar air/acetylene flame is shown in Fig. 6-5 from a study by Rann and Hambly (1965). In the lean flame the maximum density of ground-state atoms, and hence greatest analytical sensitivity, is found about 0.5 cm above the burner where the flame is hottest. Higher in the flame the number of atoms decreases due to dilution by flame gases and recombination of atoms. Molybdenum and other refractory elements show a particularly rapid recombination rate, which can be somewhat reduced by use of fuel-rich air/acetylene flame. Better sensitivity for these elements can, however, be obtained in nitrous oxide/acetylene flames (Fig. 1-9). For optimum sensitivity the burner height and fuel to oxidant ratios and pressures should be determined separately for each element. In practice,

118

Rich

Lean COPPER

Rich

Lean MOLYBDENUM

Fig. 6-5. Distribution of Cu and Mo atoms in lean and fuel rich air-acetylene flames. Contours are drawn at intervals of 0.1 absorbance units with maximum absorbance in the centre: height is measured above the burner slot. (Reprinted with permission from Rann and Hambly (1965), Distribution of atoms in an atomic absorption flame. Analytical Chemistry, 37: 879—884. © 1965 American Chemical Society.)

however, these parameters are not too critical in the lean air/acetylene flame and the best conditions for any one of the elements which are probably most frequently requested in the exploration laboratory (Cu, Co, Ni, Pb, and Zn) will usually be satisfactory for the remainder of the group. Considerably greater care is needed to optimize flame conditions and the position of the light beam in the red cone or feather of the fuel-rich nitrous oxide/acetylene flame. Furnaces Resistance heated furnaces offer several theoretical advantages over flames as a means of producing ground-state atoms, notably more efficient utilization of the sample solution and a longer residence time for atoms in the furnace than in the flame. This gives sensitivities appreciably better than those attainable in flames (Table 6-II) and enables very small quantities of material to be analyzed. Major disadvantages of furnaces, compared to flames, are more severe interferences and their much lower analytical productivity. The simplest furnaces are the silica tubes, heated either by resistance wire (Chu et al., 1972) or by mounting them above an air/acetylene flame (Thompson and Thomerson, 1974), used for atomization of elements, such as As, that form gaseous hydrides. The hydride generating system and furnace used by Aslin (1976) to determine As and Sb in geochemical samples is shown in Fig. 6-6. Less specialized furnaces for general analytical applications are usually resistance-heated graphite tubes (Fuller, 1977). Alternatively, a strip (filament) of an unreactive metal, for example tantalum, enclosed in a chamber

119

Nitrogen

Dry Nitrogen B Tygon tubing

Ash

Atomize

HOLLOW CATHODE LAMP

Plastic stopper (tight fit)

ΊΓ

CONTINUUM LAMP Eppendorf tip Sodium borohydnde solution

NET SIGNAL time -

Fig. 6-6. Apparatus for AAS determination of elements forming gaseous hydrides. Cell A is a 18 mm X 16 mm plastic test tube with a hole, 30 mm from the bottom, to take the tip of a micropipet. A l-μΐ volume of sample is injected into the borohydride solution in the cell. Gaseous hydrides are then swept by a flow of nitrogen into the silica absorption cell (D) heated by the air/acetylene flame of a triple-slot burner. (From Aslin, 1976.) Fig. 6-7. Typical absorption curves for the dry, charr and atomize stages of electrothermal atomizers. Subtraction of non-atomic absorption, measured with a continuum source, from total absorption, measured with the hollow cathode lamp, gives the signal due to the analyte. During drying and charring, smoke and vapour can give large nonatomic absorption signals.

can be used. The power supply to such furnaces must provide three cycles, each with variable time and current settings, for drying, charring, and finally volatilization and atomization of the analyte. At the high temperatures (up to 3000°C) used for atomization, the furnace material would be rapidly consumed by oxidation in air. It is, therefore, arranged for the furnace to be sheathed with an inert gas (argon or nitrogen) or, to provide a more reducing environment, hydrogen. Sample solutions are usually introduced into the furnace with a micropipet (1—100 μΐ) and the solvent evaporated during the drying cycle. The current is then increased sufficiently to char and burn off organic matter and, if possible, other matrix constituents. Ideally this step, which produces smoke, should be completed with no volatilization of the analyte before the atomization cycle begins. If charring overlaps with atomization, light losses caused by the smoke must be subtracted from the true absorption signal by simultaneous background correction. Some typical drying, charring and atomizing, cycles are shown in Fig. 6-7.

120

Wavelength selection Hollow cathode lamps or electrodeless discharge lamps emit a virtually pure spectrum of the element to be determined. Consequently, requirements for wavelength selection are relatively simple, it only being necessary to isolate the atomic absorption line from any neighbouring non-absorbing lines emitted by the light source. For many elements, with simple spectra, a 1-nm or wider bandpass is adequate. However, for elements with more complex spectra, for example Mn and Ni, resolution of at least 0.1 nm is necessary. These requirements can be met by small, relatively inexpensive, grating monochromators. When low absorbance values are to be measured curvature of the calibration line, due to inclusion of non-absorbing lines in the bandpass, is not usually significant and it may be worthwhile to increase the slit width, thereby letting more light into the monochromator and allowing the gain, and hence noise, to be reduced (Table 6-III). It should be noted that doubling the slit width will increase the light intensity for a sharp line source twofold, whereas the intensity of a continuum source will be squared. Detection and readout systems Light from the monochromator enters a photomultiplier where it is converted to an electrical signal and fed to an amplifier whose output goes to the readout system. Photomultipliers supplied with atomic absorption spectrophotometers usually have their optimum sensitivity in the 200.0- to 700.0nm region. Photomultipliers giving better response in the UV below 200.0 nm are available to improve sensitivity for determination of As and Se. In order to detect only the absorption signal and to reject flame emission, particularly in the visible light region, the output of the light source must be modulated and the detector tuned to accept only the modulated signal. This can be achieved by either mechanically chopping the light between the source and flame with a revolving mirror (Fig. 6-3), or by using alternating current or modulated direct current as the lamp power supply. With strongly emitting flames, as when large amounts of Ca are present, increased noise and apparent interferences can result from the inability of the electronics to fully reject the DC signal from the flame (Rooney and Woolley, 1978). Increasing the brightness of the light source reduces the problem. Readout systems with early atomic absorption spectrophotometers were simple galvanometers calibrated in percent absorption or absorbance units. Digital readouts, providing calibration in concentration units and features such as curve correction, peak signal retrieval and signal integration, are now standard with many instruments. Rapid detector response and peak signal retrieval are particularly valuable with furnace methods that give very fast, transient signals.

121 APPLICATIONS

Furnace techniques are well suited to direct analysis of waters but cannot compete, either in productivity or freedom from interferences, with flame AAS for the rapid routine determination of most of the elements of interest in soils and sediments. Consequently, even although concentration by solvent extraction may be required (Table 6-IV), most AAS analyses in the prospecting laboratory involve aspiration of solutions into laminar flow flames. Notable exceptions are the flameless determination of Hg and the electrothermal decomposition of gaseous hydrides. Willis (1975) has described an interesting technique for avoiding the sample decomposition stage, whereby a 5% suspension of the powdered sample in water is aspirated directly into the flame: Cu, Ni, Co, Zn and Pb were determined in stream sediments.

TABLE 6-IV Application of some solvent extraction schemes to geochemical samples l

Element

Extraction scheme

Ag

TOTP-MIBK

C h a o e t a l . (1971) Bratzel et al. (1972)

Au

Cl-MIBK Br-MIBK DBS-toluene

Bratzel et al. (1972) Tindall (1965) Thompson et al. (1968) Rubeskaet al. (1977)

Reference

Be

acetylacetone

Terashima (1973)

Bi

APDC-MIBK

Ficklin and Ward (1976)

Mo

SCN-MIBK Aliquat 336-MIBK

Kim et al. (1974) Rao (1971)

Pd

DBS-toluene

Rubeskaet al. (1977)

Pt

Sn n -MIBK

Stanton and Ramankutti (1977)

Sb Sn

Cl-MIBK TOPO-MIBK TOPO-MIBK

McHugh and Welsch (1975) Welsch and Chao (1976)

Te

Br-MIBK

Nakagawa and Thompson (1968) C h a o e t a l . (1978)

Tl and In

Br-MIBK

Hubert and Lakin (1973)

w

SCN-Alamine 336-MIBK

K i m e t a l . (1976)

1

Welsch and Chao (1976)

Abbreviations: TOTP = triisoctyl thiophosphate; MIBK = methyl isobutylketone, 4methyl-2-pentanone; Cl = chloro; Br = bromo; DBS = dibutyl sulphide; APDC = ammonium pyrrolidinedithiocarbamate;SCN = thiocyanate; TOPO = trioctylphosphine oxide.

122

Flame atomic absorption When solids are to be analyzed, the sample solution is generally either a dilute acid or some other aqueous extract prepared as described in Chapter 4. If analytical sensitivity is too low for direct determination of the analyte in the original solution it may be concentrated into an organic phase by solvent extraction. Operation and calibration Operation of atomic absorption spectrophotometers is usually straightforward, routine operating conditions being adequately described in most manufacturer's manuals. At low concentrations, requiring scale expansion, or when working with intensely luminous flames some deviation from these conditions may improve signal to noise ratios (Table 6-III). High-precision methods, for example standard additions or closely bracketing unknowns with standards, are too time-consuming for routine use in the exploration laboratory and the calibration curve is therefore usually based on a reagent blank and two or three standard solutions. Also, to avoid further calculations, the sample dilution factor (i.e. volume (ml)/weight (g)) is taken into account and the instrument calibrated directly in concentration units. On instruments not equipped with digital readout this can be done by inserting a calibration scale (one for each element) over the face of the meter readout and using scale expansion to make any necessary small adjustments for daily changes of sensitivity, so that concentrations and the corresponding readings for standard solutions coincide. Ideally, standard solutions should have the same composition as sample solutions. However, differences in absorption between dilute aqueous solutions are often not too great and standards in dilute hydrochloric or nitric acid will often suffice for analysis of samples in other dilute aqueous media. This must, of course, be established for the particular solutions to be analyzed. Preparation of standards is described further in Appendix 2. Sample solutions with element concentrations above the concentration range of the instrument can be brought on-scale by use of a less sensitive absorption line, by shortening the absorption light path by rotating the burner, or by further dilution. When a constituent with a wide concentration range, or several constituents at very different concentrations Eire to be determined from the same original solution, several dilutions may be needed. Under these circumstances an automatic dilutor, of the type used in clinical laboratories, providing a ten- to twenty-fold dilution saves considerable time. Sensitivity Of the trace elements in Table 6-II, crustal abundance of twelve (Ba, Be, Co, Cr, Cu, Mn, Ni, Pb, Rb, Sr, V and Zn) is such that they should be detectable with routine operating conditions and dilution factors in the

123

range 20—50 after complete, or almost complete, sample dissolution. Fortunately, this group includes many of the elements most often sought in the exploration laboratory. Sensitivity for several other important elements, notably Ag, Cd and Mo, is borderline and these will be reported as not detected in many background samples. The remaining elements will usually only be detectable after their pre-concentration or by using special techniques to enhance sensitivity. Because of its simplicity and rapidity, solvent extraction, whereby an uncharged ion association or metal chelate is extracted from an aqueous phase into an immiscible organic solvent, has found many analytical applications in the concentration and separation of trace elements from interferences (Morrison and Freiser, 1957; Stary, 1964). When used in conjunction with AAS the organic solvent, in addition to having a low solubility in, and providing a clean separation from, the aqueous phase, must have good burning characteristics in the air/acetylene or nitrous oxide/acetylene flames. Despite a rather high solubility in water (22 ml/1) methyl isobutylketone (MIBK: 4-methyl-2pentanone) fulfils most of these requirements and has become the preferred solvent in many extraction schemes. Compared to aqueous solutions, use of MIBK also results in a two- to five-fold increase in sensitivity, apparently due to an increase in the aspiration rate and more efficient transfer of the analyte to the flame (Allan, 1961; Chakrabarti and Singhal, 1969). Although many solvent extraction schemes are described in the analytical literature only a few have been applied to routine analysis of exploration samples (Table 6-IV). Extraction of the chloro- or bromo-aurate ion into MIBK in the determination of Au is probably the most frequently used. Prior to solvent extraction Au can be solubilized with either aqua regia digestion (Tindall, 1965) or, in the rapid method described by Thompson et al. (1968), a hydrobromic acid/bromine mixture. Campbell (1980) found that interferences from Mn at the 242.8-nm Au line could be overcome by boiling the solution to remove unreacted bromine. With little modification solvent extraction from hydrobromic acid/bromine can also be used for pre-concentration of In and Tl (Fratta, 1974; Hubert and Lakin, 1973), Te (Nakagawa and Thompson, 1968) and Pt (Stanton and Ramankutti, 1977). Some separation schemes, for example extraction of the red Mo-SCN complex or the Pt IV compound formed in the presence of SnCl2, have evolved directly from the colorimetric methods employed for determination of these elements. In addition to solvent extraction schemes for individual elements, multielement extraction schemes have also been described for determination of trace elements in geological material. Hannaker and Hughes (1977, 1978), for example, developed an extraction procedure to reduce interferences and increase sensitivity in the determination of Cu, Ni, Co, Cr, Ag, Pb, Bi, Cd, Zn, Mn, Au, Tl, Sb, Ga and Mo. Fortunately, providing background absorption is corrected for (p. 128), interferences in flame AAS are seldom suffi-

124

ciently troublesome to warrant use of solvent extraction for those elements with adequate sensitivity for their direct determination in acid extracts. Viets (1978) has described a solvent extraction scheme using tricaprylylmethylammonium chloride (Aliquat-336) for rapid determination of Ag, Bi, Cd, Cu, Pb, and Zn after their extraction with the sulphide selective potassium chlorate/hydrochloric acid leach. Analysis of waters by AAS has been reviewed by Ediger (1973). Trace metals at parts per billion (ppb) concentrations are not usually directly detectable in the flame and must therefore be concentrated. The simplest methods of concentration are evaporation or solvent extraction. Extraction with APDC*-MIBK is suitable for many elements (Mulford, 1966) and has been used by Brown et al. (1970) to concentrate Ag, Cd, Co, Cr, Cu, Ni, and Pb from samples acidified to pH 2.8. However, Kinrade and Van Loon (1974) found extraction with APDC-DDDC **-MIBK to be more effective over a wider pH range (Fig. 6-8). With a citrate buffer at pH 5.0 extraction of Ag, Cd, Co, Fe, Ni, Pb and Zn is quantitative. References to many other solvent extraction schemes are to be found in Ediger (1973) and the Annual Reports of Analytical Atomic Spectroscopy. Careful cleaning of storage bottles and reagents is necessary to free them from trace metal contaminants at ppb concentrations. Without resorting to solvent extraction, sensitivity for some elements can be substantially increased by adjusting bulk composition of the solution to ensure that formation of non-absorbing species, for example ions and refractory oxides, is minimized. At the same time the amount of solution reaching the flame can be increased. For example, Sen Gupta (1976) and Ooghe and Verbeek (1974) increased sensitivity for rare earth elements by suppressing their ionization in the nitrous oxide/acetylene flame with an alkali metal buffer. A further increase in sensitivity was obtained by aspiration of an ethanolic, rather than an aqueous solution, to improve overall atomization efficiency. A number of elements, notably Zr, Hf, Ta and Ti, have very poor sensitivities even in the nitrous oxide/acetylene flame because of the formation of their refractory oxides. All these elements also form strong fluoride complexes, with much lower melting points than the oxides, and Bond (1970) reported up to a tenfold increase of their sensitivity in the presence of ammonium fluoride. This is presumably due to the formation, volatilization and direct atomization of the fluoride complexes without formation of the refractory oxide. Similarly, Husler (1972) obtained a marked increase in sensitivity for Nb in the presence of hydrofluoric acid and an alkali-metal buffer. * Ammonium pyrrolidine dithiocarbamate ** Diethylammonium diethyldithiocarbamate

125 Cd scale = 2.5 Α

—Α

2.0 η

1

9

· — · Cd

I

D—D

1.6 A j

0

Α

o

2

4

6

CO

o Cu

8

10

12

14

12

14

pH Zn scale = 2.5 ♦ — ♦ Ni

1.6 H

0—0 Pb

LU

Δ—Δ

0

2

4

6

8

Zn

10

PH

Fig. 6-8. Influence of pH on the extraction of metals into MIBK with 1% w/v APDC + DDDC. (Reprinted with permission from Kinrade and Van Loon, 1974, Solvent extraction for use with flame atomic absorption. Analytical Chemistry, 46: 1894—1898. © 1974 American Chemical Society.)

Interferences Freedom from interferences is often cited as one of the principal advantages of AAS. Nevertheless, if samples and standard solutions differ in bulk composition a variety of interferences can cause enhancement or depression of the analyte signal. Principal sources of interferences and some remedial methods are summarized in Table 6-V. Most interferences could be overcome either by careful matching of sample and standard solutions, or by use of the method of additions. Alternatively, the analyte could be isolated from interferents by solvent extraction or some other separatory technique. For example, prior to analysis of

126 TABLE 6-V Some interferences in the determination of trace elements in geological matrices by flame atomic absorption spectrophotometry Element

Interference

Comments

Reference

Ag Ba*

B

background correction

Minkkinen (1975)

C, I

suppression by Si, Al and P — add oxine or NH4CI; enhancement by alkalies and alkali earths — add K radiation buffer

Marutaetal. (1972)

Miller and Cazalet (1979)

Cioni et al. (1976a)

Be*

C

suppression by Al

Terashima (1973)

Cd, Co

B

background correction

Fletcher (1970)

Cr

C

suppression by Fe, Na and K; enhancement by AI, Mg, Ca — interference and sensitivity reduced in Ν 2 0 / ϋ 2 Η 2 flame

Mo*

C

suppression by alkalies, Ca, Fe — add up to 1000 Mg/ml Al

Van Loon (1972) Sutcliffe (1976)

Ni, Pb

B

background correction

Fletcher (1970)

Rb

I

add K radiation buffer

Sr*

C, I

add K radiation buffer and La releasing agent

Carter et al. (1975)

v*

C

suppression by Fe and Ti — add 1000Mg/ml Al

Terashima (1973)

Zr*

C

add NH 4 F

Bond (1970)

1 2

Determined in lean air/acetylene flame except (*) in nitrous oxide/acetylene flame. B = background absorption; C = chemical (suppression); I = ionization (enhancement).

Fe-rich solutions, produced by acid decomposition of magnetite or pyrite, ferric iron can be extracted into MIBK from 6 M HCl leaving many trace elements in the aqueous phase (Nakagawa, 1975; Sinclair et al., 1977). However, such procedures are generally too time-consuming for routine analysis and it is, therefore, fortunate that for prospecting purposes analytical interferences are normally within tolerable limits. Spectral interferences. Spectral interferences arise when absorbing lines of other elements coincide or overlap with the absorption line of the analyte. Because of the simplicity of absorption spectra and the purity of the spectrum emitted by a hollow cathode lamp, such interferences are seldom encountered. One of the few cases that might be encountered in routine

127

analysis of geological matrices is the overlap of Fe 213.859 nm on Zn 213.856 nm causing high absorption readings for Zn in Fe-rich materials. Kelly and Moore (1973) found that 0.2-2.0% Fe in solution gave apparent Zn concentrations ranging from 0.031 to 0.26 Mg/g. Chemical interferences. As the solvent evaporates in passing from the spray chamber into the lower part of the flame a condensed phase consisting of solid clotlets is formed. If the composition of the sample results in formation of refractory compounds which do not dissociate (or only dissociate above the zone illuminated by the light source) the population of absorbing atoms and hence the absorption signal will be correspondingly reduced. This is particularly significant when silicates are analyzed for Ca or Mg in the air/ acetylene flame (Fig. 7-10) and is also well illustrated by the severe depression of Mo readings in the presence of Ca in the nitrous oxide/acetylene flame, probably due to formation of calcium molybdate (Fig. 6-9). Addition of Al which preferentially combines with the Ca and acts as a releasing agent for the Mo, minimizes the interference. Van Loon (1972) recommends addition of sufficient Al to give a final concentration in solution of 1000 Mg/ml for a 0.3-g sample. lonization interferences. lonization interferences are associated with the determination of the alkali metals and, in the hotter nitrous oxide/acetylene flame, the alkali earths. The alkalies have relatively low ionization potentials and are appreciably ionized even in cool flames. If another alkali is added to a solution already containing an alkali, its ionization contributes to the total 0.5Ί

g

< er °· 3 Ί

°

/

CO

CD <

7

/

/

02J / 1 / /

'

,

10 uq/ml

Mo

10^ig/ml Mo + 1 0 0 p g / m l Ca

/

1

200

1

1

1

400 600 800 ALUMINIUM, ^g/ml

1

1000

Fig. 6-9. Interference of Ca in the determination of Mo in the nitrous oxide/acetylene flame, and the influence of varying concentrations of Al. At approximately 1000 Mg/ml Al the suppression due to Ca is almost eliminated.

128

electron population in the flame. This results in partial suppression of ionization for the first alkali and an increase in the number of its ground-state atoms causing an increase in the absorption signal. The effect of varying concentrations of alkalies in sample solutions can be overcome and sensitivity increased by addition of an alkali metal ionization-suppression buffer, for example, a large excess of Cs in the determination of Rb or Sr. Enhanced absorption by the Ba 553.6-nm atom line due to ionization suppression by increasing concentrations of K and Ca is shown in Fig. 6-10: there is a corresponding reduction of absorption at the 455.4-nm ion line. However, in the case of Ba the interactions between chemical and ionization interferences are so complex that effective correction with geological matrices may be impossible (Cioni et al., 1976a). Despite this, Miller and Cazalet (1979) reported obtaining satisfactory results for prospecting purposes using the 455.4-nm line and addition of K (1000 Mg/ml) to standards. Use of this line also avoids increased signal noise associated with the coincidence of a Ca bandhead with the atom line (Rooney and Woolley, 1978). Background interference. Background absorption is a special case of spectral interference resulting from absorption by molecular species in the flame and possibly also by light scattering by solid particles. It has been described in detail by Billings (1965) and Koirtyohann and Pickett (1966). Unlike the sharp absorption lines resulting from atomic absorption, background absorp0.15

0.15

A 40ppm Ba + 10,000 ppm K • 40 ppm Ba + 2,000 ppm Ca

o z < CO

o

c/> CO

<

0.05 H

0.05 -4

1000

2000

3000

CONCENTRATION OF INTERFERENT, }jg/ml

CONCENTRATION OF ALUMINIUM, jig/ml x103

Fig. 6-10. Determination of Ba in the nitrous oxide/acetylene flame. A. Effects of various elements on the absorbance of a solution containing 40 ppm Ba at the 553.4-nm atom line: the dashed line shows the effect due to addition of Ca at the 455.4 nm ion line. B. Effect of Al additions on absorption by Ba-K and Ba-Ca mixtures at the 553.4-nm line (From Cioni et al., 1976a.)

129

tion is a broad band phenomenon the absorption being additive on any atomic absorption due to the analyte. Fletcher (1970) and Foster (1971) have shown that background absorption can give large errors when low concentrations of an element are to be measured in geochemical samples. For example, Fletcher (1970) found that a Pb-free synthetic granodiorite gave an absorption equivalent to 25 ppm Pb (Table 6-VI). On real samples background interference was particularly severe for Pb, Co and Ni, but not Cu and Zn. Ag is also affected and, since Ag values of only a few parts per million may be very significant, spurious geochemical patterns and anomalies can easily be created. Minkkinen (1975) found that Ag-free solutions containing 20,000 ppm Ca gave readings equivalent to 4 ppm Ag. Background absorption is best corrected by its measurement with a continuum light source, either a hydrogen or deuterium lamp in the UV region, and then subtracting it from total absorption, measured with the hollow cathode lamp, to give a corrected value (Fig. 6-11). This can be done automatically on many instruments and is recommended for routine determination of Ag and Pb. If a continuum lamp is unavailable background absorption can still be measured if the light source emits a suitable nonabsorbing line close to the absorption line. For example, background absorption at Pb 216.99 nm can be estimated using Pb 220.35 nm. Background absorption is additive and the error, on a percentage basis, is therefore greatest at low concentrations. At high concentrations it may be insignificant. Furthermore, if background absorption and a suppressive interference are both present, background correction can give the best results at low concentrations but cause over correction at high values (Fig. 6-12). Fortunately, in exploration geochemistry negative errors at high concentrations are usually less serious than large positive errors at background levels.

TABLE 6-VI Correction for background absorption in lead analysis of a synthetic rock solution (from Fletcher, 1970) Pb added (ppm) * 0.0

20.0 50.0 100.0 200.0

Absorbance

Pb found (ppm)

1

cathode lamp

H 2 lamp

corrected

uncorrected

corrected

0.015 0.025 0.041 0.063 0.111

0.015 0.015 0.015 0.016 0.015

0.000 0.010 0.026 0.047 0.096

24.8 47.7 82.1 122.2 217.7

21.0 53.5 95.5 190.9

Dilution factor = 50

0.0

130 SOURCE

FLAME

DETECTOR

READOUT

L.JbJ 100

100

A%

l%

100

A%

Fig. 6-11. Background correction using a continuum source. The profile of the atomic absorption line is shown in solid black and the continuum is represented, over a slit width of Δλ, as the area within the dashed line. At the light source the emission intensity of the atomic line and continuum are balanced at 100%. In the flame atomic absorption is superimposed on background absorption so that the intensities reaching the detector are Ia and Ib) respectively. The readout displays these intensities as a net signal equal to (100 — Ia) — (100 -Ib).

100

—Φ—

90

U n c o r r e c t e d , Pb line 2170.5 Ä

--O—

Corrected on H-famp

---a—-

Corrected on Pb non-absorbing line 2 2 0 4 Ä

80 70 60 50 40 30 20 10 0\-

-10

O^

■r-ψ-^-ψ: 12

4 6 Θ 10 ΡΡΦ Pb in solution

Fig. 6-12. Errors in uncorrected and background-corrected measurement of Pb in the presence of 4000 ppm Fe. (From Govett and Whitehead, 1973.)

131

Electrothermal

atomization

For those elements readily determined in the flame use of electrothermal furnaces offers no advantages. Potential applications of electrothermal methods in the exploration laboratory are therefore largely confined to determination of elements with inadequate sensitivity in the flame and to determination of ppb concentrations in weak partial extracts of soils and sediments or in natural waters. Use of electrothermal atomization in these situations should be carefully weighed against the merits of solvent extraction and conventional flame atomization — particularly if sampling criteria require decomposition of relatively large sub-samples anyway. Geochemical matrices are usually analyzed by injection of an aliquot of sample solution, although direct atomization from a few milligrams of finely ground powder is also possible (Langmhyr et al., 1974; Langmhyr, 1977) and Friedrich et al. (1973) injected an ethanolic suspension to determine Cu in soils and sediments. When acid decomposition is employed, nitric acid is preferable to a halogen acid to avoid losses by direct volatilization of the undissociated metal halide: with oxy-acids atomization proceeds via the relatively less volatile metal oxide. Loss of analyte can also occur, especially with the more volatile elements, if temperatures are too high during sample charring (Fig. 6-7). Ediger (1975) reduced such losses by matrix modification. For example, addition of Ni retards loss of Se and Te up to 1200° C (probably due to formation of nickel selenides and tellurides) and has been applied to determination of Se in sediments (Martin et al., 1975). Compared to flames, non-specific absorption, due to smoke and high atom densities, is more severe with electrothermal devices and requirements for correction of background absorption are correspondingly more stringent. Other types of interference also occur in analysis of geochemical matrices. However, the processes and factors involved are not well understood. In a study of determination of Pb in carbonates, Campbell and Ottaway (1974, 1975) found that Ca depressed the analyte signal in hydrochloric but not in nitric acid. A similar phenomenon has been reported by Sighinolfi (1972) in the determination of Be, and Cruz and Van Loon (1974) attributed the depressant effect of Ca on Cd, Co, Cu, Ni, Pb and Zn to their occlusion within a refractory matrix formed in the furnace. Hutton et al. (1977) suppressed ionization interferences in the determination of Ba in limestones by addition of Cs. Interference problems could also be avoided by solvent extraction. This, however, would seem to negate the advantages of the great sensitivity of electrothermal atomization except for a few elements such as Au (Bratzel et al., 1972; Meier, 1980), with such low concentrations that adequate sensitivity can sometimes only be achieved by combining furnace AAS with solvent extraction. Examples of the applications of flameless AAS to geochemical analysis, in addition to those already mentioned, include: Ba (Cioni et al., 1976b): Bi

132

(Ficklin and Ward, 1976); Cd (Gong and Suhr, 1976); Te (Corbett and Godbeer, 1977; Chao et al., 1978); Mo (Kontas, 1976); and precious metals (Fryer and Kerrich, 1978). Other references and a discussion of the determination of individual elements are to be found in Fuller (1977). In marked contrast to analysis of solids, a rapidly expanding literature on application of flameless AAS to water samples has developed in response to environmental problems. Natural fresh waters contain a wide range of dissolved ions and constitute a complex, non-uniform analytical matrix. Consequently, background absorption and inter-element interferences, although less severe than with solids, are still present. Nevertheless, multi-element studies by Edmunds et al. (1973) and Barnard and Fishman (1973) have shown that for rapid screening and anomaly detection, direct injection and comparison with standard calibration curves could be adequate if accompanied by simultaneous background correction. Fordham (1978) obtained more accurate results by the method of additions: however, only sixty determinations per day were possible. Hydride

generation

Hydride generation combined with AAS (or ICP-ES, p. 164) provides a rapid and very sensitive method for determination of elements forming gaseous hydrides. It has been applied to analysis of geochemical samples, including waters, for As, Sb, Se, Sn, and Te and may also be applicable to estimation of Ge and Bi — all elements difficult to determine by conventional flame methods. Initial studies of hydride generation-AAS involved reduction of As 3+ to arsine (AsH 3 ) in zinc-stannous chloride/hydrochloric acid reduction cells (Fernandez and Manning, 1 9 7 1 ; Dalton and Malanoski, 1971). The arsine was then swept, either directly or after collection in a reservoir, into the transparent argon/hydrogen or nitrogen/hydrogen flames. Subsequently, hydrides of As, Bi, Sb, Se, Sn and Te were all generated more conveniently from acidic sample solutions using sodium borohydride, either as pellets or a stable alkaline solution, as a reductant (Fernandez, 1973; Thompson and Thomerson, 1974). Precise control of the acidity of the sample solution is not critical except in the determination of Sn. Rapid quantitative evolution of the hydrides only proceeds from the lowest valency state of the elements. Se 6+ and Te 4 + can be reduced to their trivalent states by warming in hydrochloric acid or directly by the action of the borohydride. If, however, arsine and stibine are to be evolved from weakly acidic solutions, sodium or potassium iodide must be added as prereductants to ensure quantitative reduction. Sensitivities attainable by decomposing the hydrides, both in flames and in resistance or flame heated cells, are summarized in Table 6-VII. These values are very much dependent on generator design, reagent strengths, solu-

Aqua regia HF/HCIO4/HNO3 6MHC1

HF/HCIO4/HNO3

Sb

Te 5% NaBH 4

2% NaBH 4

2% NaBH 4

KI/SnCl 2 /Zn

resistance heated quartz cell

flame heated quartz cell

resistance heated silica tube

Ar/H 2 flame

resistance heated quartz cell

N2/H2/air flame

resistance heated quartz cell

flame heated quartz cell

Atomizer

:

no serious interferences for most samples: up to 2 0 0 ppm Cu can be tolerated

interference from Ni and Ag reduced by addition of 0.01 M EDTA

no interferences from Cd, Cu, Pb, Zn, Mg, Ca, Fe and Al at concentrations found in soil samples

K M n 0 4 added to avoid volatilization losses; interference from Pb minimized by addition Fe3+

KI added as a pre-reductant; rapid technique for analysis of soils, waters and vegetation

modification for rapid analysis of exploration samples

semi-automated Technicon system for analysis of soil and vegetation

interference from Ni prevented by addition of 0.01 M EDTA

Comments

6

1

7

5

4

3

2

1

Reference

Wauchope (1976); 5 = Terashima (1976); 6 = Green-

5 ppb on 0.25-g sample

0.08 ppm on 0.05-g sample

0.001 ^g/ml

0.04 ppm for 1 g sample

10 ng/ml in solution

0.25 ppm on 0.1 g sample

0.16 ppm on 0.5-g sample

Detection limit

References: 1 = Aslin (1976); 2 = Vijan et al. (1976); 3 = Kokot (1976); 4 land and Campbell (1976); 7 : = Thompson and Thoresby (1977).

HN03/H2S04

As

1

HF/HCIO4/ HN03/KMn04

As

HNO3/HCIO4

1% KI/ 1% NaBH 4

HC1

1% NaBH 4 solution

As

HNO3/HCIO4

As

2% NaBH 4 solution

5% NaBH 4

aqua regia HF/HCIO4/HNO3 6 M HC1

As

Generator

As

Decomposition

Element

Application of hydride generation to analysis of geochemical samples by atomic absorption spectrophotometry 1

00 GO

134

tion volumes and gas flow rates. Greenland and Campbell (1976), determining Te in rocks, found it difficult to maintain maximum sensitivity without constant attention to the equipment. They, therefore, opted to use less sensitive conditions for routine analysis: even so, as little as 5 ppb Te could be detected with a 0.25-g sample. As usual, early reports on the absence of interferences proved illusory. Interferences reported by various workers do, however, differ somewhat, apparently depending on reagent strengths and volumes and order of mixing. In a detailed study, Smith (1975) generated the hydrides of As, Bi, Ge, Sb, Se and Te from dilute hydrochloric acid with sodium borohydride. Using an argon/hydrogen flame he found: Cu, Ag, Au, Pt, Pd, Rh, Ru, Ni and Co always suppressed hydride generation — probably due to their preferential reduction and precipitation, resulting in coprecipitation or adsorption of the analyte, and consumption of reductant; and the hydride-forming elements mutually interfere. Belcher et al. (1975) found that suppression of the evolution of arsine and stibine, from 0.1 M hydrochloric acid, by Co, Ni, Zn, Fe, Bi, Cd and Cu could be almost eliminated by addition of EDTA. In more strongly acidic media (5 M HC1), Aggett and Aspell (1976) found no interference from up to a 10,000-fold excess of Cu on the evolution of arsine. Problems with high reagent blanks for As, Sn and Sb in sodium borohydride have been reported. Rapid methods of analysis of geochemical samples for As, Sb, Se and Te have been described (Table 6-VII). With hot acid decompositions it is necessary to prevent volatilization of As and Se by maintaining them in their highest valency state, which must then be reduced to the trivalent state for quantitative hydride generation. Large amounts of nitric acid remaining after decomposition cannot be tolerated and should be removed by evaporation with perchloric acid. Interference studies suggest that Fe, Cu, Ni, Mn and Co are most likely to be troublesome in geological materials. Aslin (1976) found that Ni suppressed evolution of arsine from 1.5 M hydrochloric acid and both Ni and Ag suppressed stibine; the interferences could, however, be controlled by addition of EDTA and it then became possible to determine 2 ppm Sb in the sulphide standard SU-1. Thompson and Thoresby (1977) and Thompson et al. (1978b) found no significant interferences in the evolution of arsine from 5 M hydrochloric acid. However, the latter authors did find it necessary to avoid interference from Cu in the hydride generation of Bi, Se and Te by their coprecipitation and separation on lanthanum hydroxide. Greenland and Campbell (1976) also reported interference from Cu, in excess of 200 ppm, in the determination of Te. Determination of mercury by flameless AAS Estimation of Hg by flameless AAS predates conventional AAS by many years and concern over environmental concentrations of Hg and its use as a

135

pathfinder for sulphide deposits have produced an extensive literature on its determination in natural media. Ure (1975), for example, cites more than 400 papers on determination of Hg by flameless AAS after its release from samples either by pyrolysis or acid digestion followed by reduction in solution to elemental Hg. Pyrolytic methods can be used to determine either total Hg or, by controlling the heating rate, to determine different forms of Hg and its partitioning between different minerals (Koksoy et al., 1967; Watling et al., 1973). Pyrolysis, however, has the disadvantage that geochemical samples, especially those rich in organic matter or sulphides, produce smoke or S0 2 causing non-specific absorption at the 253.7-nm Hg line. With conventional atomic absorption instruments, background correction with a continuum source can be used as already described. Alternative correction methods include: (1) Measurement of non-specific absorption on the edges of the 253.7-nm line using either (a) the wings of a pressure broadened, self-absorbed line produced in a Hg-saturated cell as shown in Fig. 6-13 (Barringer, 1966); or (b) the Zeeman effect to split the absorption line (Hadeishi and McLaughlin, 1966;Robbins, 1973). (2) Dividing the gas flow from the heated sample into two streams (Fig. 614), one of which passes through a PdCl2 filter to remove Hg, before entering the reference cell of a double-beam instrument (James and Webb, 1964). A similar method has been applied to recirculating single-beam systems (Windham, 1972). (3) Collection of Hg liberated during pyrolysis by amalgamation on Au or Ag, from which it is subsequently re-released by heating (Vaughn and McCarthy, 1964; Vaughn, 1976; Marinenko et al., 1972). Organic compounds condensing on the Au trap can be troublesome, either due to their occlusion of Hg or their re-release on heating causing background absorption. Azzaria and Webber (1969) and Weissberg (1971) prevented condensation of organics by maintaining the trap at 150—170° C. As an alternative to amalgamation, Hg released by pyrolysis can be collected in a potassium permanganate solution and subsequently determined by the cold vapour method described next. The cold vapour method is readily adaptable for use on any conventional atomic absorption spectrometer (Fig. 6-15). In the original procedure, described by Hatch and Ott (1968), Hg was liberated by the action of sulphuric acid, hydrogen peroxide and potassium permanganate on powdered rock samples. After cooling, excess permanganate and precipitated manganese oxides were reduced with hydroxylamine sulphate. Hg is then reduced to the elemental form with stannous sulphate and flushed into an absorption cell by a stream of air. Subsequently, many variations of the cold vapour method and a wide variety of decomposition mixtures have been described (Ure, 1975). Some workers have combined cold vapour and amalgamation meth-

136

A

it

I

Differential I Amplifier

Fig. 6-13. Design of a spectrometer using pressure broadening of the 253.7-nm Hg line to correct for non-atomic absorption in the determination of Hg. A pressure-broadened Hg spectrum, emitted by a low-pressure mercury lamp, passes through an absorption chamber and is then split into two beams, one of which falls directly on a photomultiplier (PMA). The other beam passes through a cell saturated with mercury vapour so that the central portion of the 253.7-nm Hg line is completely absorbed leaving only the wings of the line to reach the reference photomultiplier (PMR). When a vapour causing non-atomic absorption is pumped into the absorption chamber, output from both PMA and PMR is proportionally reduced: atomic absorption due to Hg only reduces the signal fromPM A . (Modified from Barringer, 1966.)

ods (Head and Nicholson, 1973; Huffman et al., 1972). Providing potassium permanganate is not used in the digestion solution Hg can be reduced directly with Sn11 without addition of the hydroxylamine. Loss of Hg in the absence of permanganate has, however, been reported (Iskander et al., 1972). Interferences arise, particularly if the Hg is recycled through the absorption cell and aeration vessel, from elements which are also reduced to elemental form (Au, Pt, Pd, Te) and then amalgamate with Hg (Jonasson et al., 1973). Interferences from Cu and Ag, reported by Band and Wilkinson (1972) in a chloride-rich system using hydroxylamine hydrochloride, are avoided by reduction with stannous sulphate (Jonasson, 1974). Indirect determinations of elements by AAS A large number of elements are either not determinate by AAS or are only determinable at sensitivities too low to be of practical value. In some

137 (?fi

Ultra-violet

Lamp

/\

Photocell

hm.

^m.

mm Glass w o o l '

Photocell

E&.

4

Suction pump

m

W^A Palladium chloride on glass wool

Silica gel

tm Fig. 6-14. A double-beam mercury vapour meter. The sample stream is split into two halves, one of which passes through glass wool loaded with palladium chloride, to remove Hg but not constituents causing non-atomic absorption, before entering the reference absorption cell on the right side of the instrument. Output from the two photomultipliers is compared. (From James and Webb, 1964.) Infra-Red Heater " I I I

Lamp

i \ | | | I i I

\ \ V

Hg° (U Absorption

Cell

Quartz Window Air

Gas Washing Bottle

: Ä ■

ym Fig. 6-15. Cold vapour generation and determination of Hg by AAS. In the gas washing bottle Hg 2+ is reduced to Hg° with Sn 2+ and then swept into the absorption cell by a stream of air. The IR heater lamp prevents condensation of water vapour in the cell.

138

cases these elements can be determined by indirect methods based on either changes in the absorption signal of an element after its reaction with the analyte; or the formation of complexes containing the analyte and an element readily determined by atomic absorption. Kirkbright and Johnson (1973) have reviewed indirect methods for some 16 elements. One of the few geochemical applications is the determination of sulphate in soils, by measurement of excess Ba after precipitation of barium sulphate (Varley and Chin, 1970). The author, following Kirkbright et al. (1967), has determined phosphate by extraction of the phosphomolybdovanadic acid complex, used for colorimetric estimation of phosphate (Peachey et al., 1973), into MIBK and measurement of the Mo content of the organic phase by AAS. The 11 : 1 ratio of Mo to P in the acid complex makes the method extremely sensitive. Similar amplification schemes, based on formation of molybdenum-heteropoly compounds for As, Ge, Si, V, Nb, Th, and Ti, are summarized by Kirkbright and Johnson (1973).