Part Z: Fundamentals and Znstrumentation 19 1.4 FLAMES 1.4.1 Fundamental Studies Theoretical studies of flame systems continue to be reported and in recent times the emphasis has shifted from the flame itself to the processes involved in atomization. New techniques and instrumentation often enable workers to consider new approaches to old problems and in some cases the flame provides a convenient test cell for examining the20 Analytical Atomic Spectroscopy validity of theoretical predictions.L’Vov er al. (171) have continued their serics of discus- sions on the theory of AAS with a paper concerned with the effect of incomplete sample vaporization. It is shown that curvature of the log A versus log C curve can be charac- terised by the ratio of the particle radius to the mean free path.When the particle radius is less than the mean free path the slope may decrease to 0.67 and when the radius is greater than the mean free path the slope may decrease further to 0.33. Experimentally found slopes for Ni and Co sulphates were 1 in an oxidising flame, and 0.57 in a reducing flame where reduction to involatile carbides was assumed. The difference between 0.57 and the predicted 0.67 was explained by poly-dispersity of the aerosol. A stochastic approach to the theoret- ical prediction of the processes of desolvation and atom production has been given by Li (922, 11 85).The expressions developed describe the atomic concentrations for a certain position in the flame as a function of solution uptake rate, droplet distribution, flame gas velocity, atom diffusion coefficient, atomization efficiency, mass of the analyte atoms, desolvation, solute vaporization and size of burner, The atomic concentrations calculated can be used to determine absorbances for AAS or radiances for AFS and AES.The dynamics of desolvating aerosol particles in high-temperature sources has been the sub ject of work by Hieftje and Boss (790, 11 87, 1242).Halls has shown (796, 1686) that the dissociation equilibrium theory of atom production fails in a number of cases to predict atomization efficiencies that agree with experience. In such cases atomization is achieved by fast binary reactions involving reduction of analytc oxides by flame species. The free energy available from the oxidation of C, or H radicals is just sufficient to permit the reduction of the oxides of elements that are known to be atomized well.Calculations of atomization efficiency for Na and Mg assuming reduction by H radicals gave good agreement with experimentally determined values. Alternatively, Hayashi et at. (849) have performed modelling of flame species in the N,O/C,H, flame based on equilibrium calculations and generally found satisfactory agreement between theory and experiment.Alkemade et al (812) have studied the emission and absorption line profiles of Sr and Ba in an air/C,H, flame using both a pressure scanned interferometer and Zeemaii scanning and use the results to show that the free atoms are in thermal equilibrium with the flame gases. In further work on Sr (68, 69) in O,/H,/Ar and N,O/CO flames they have deter- mined adiabatic collision cross-sections for Ar and H,O and quenching cross-sections for 0,, CO,, CO and N,.A further paper by the same authors (803) discusses the effect of line profile on the curve-of-growth, and gives numerically computed curves for various shapes, including the Voigt profile. The study of flame species provides continuing interest since they form the chemical environment in which atomization must occur and also are responsible for the background radiation against which the analytical signal is measured, Fowler and Winefordner (91 7) have studied scatter and fluorescence spectra of the air/C,H, flame irradiated under pulsed and d.c.conditions by a high pressure Xe lamp. The principal band spectra observed were those due to PO (220-275nm; 323-330nm), OH (280-295nm; 305-320nm) and CN (380-390 nm).A commercially available PH, gas filter when placed in the C,H, line was found to be effective in removing fluorescence contributions from PO. A report from Steiner (1381) suggested that impurities in C,H, can affect the accuracy of analyses if the variable composition of the cylinder outflow i s not taken into account.The use of molecular absorption for flame photometric determination has been limited because of the complex background spectra that occur. A twin magnetic-tape analogue signal-processing system capable of analyzing flame absorption spectra has been described by Haraguchi et al. (925). By recording signal and background spectra separately on different tracks and then playing back simultanewsly through a differential amplifier, background correction was readilyPart I: Fundamentals and Instrumentation 21 achieved and with the addition of time displacement, operations such as differentiation could be accomplished.The interferences caused by flame background absorption spectra have been studied by Daidoji (1 501).A CO, laser has been used to irradiate Ca and Mg containing flames (338) and has been found to produce unattributed emission peaks. It is suggested that the emission may be from metal oxide particles undergoing candoluminescence, which is in some way stimulated by the i.r. radiation. Baronavski and McDonald (877) used a flash lamp pumped dye laser to excite fluorescence of the C , Swan band system in an O,/C,H, flame, From these measurements they determined the total lower electronic state population of C, to be 4.5X 1015 molecules cm-3.The concentrations of NO and free 0 atoms in the air/C,H, flame have been studied using gas probe extraction methods (240). The maximum concentration of NO was 1600 ppm, which occurred 60mm from the burner, while maximum 0 concen- tration was 3000 ppm and occurred 30mm from the burner; peak temperature was attained 40mm above the burner.Halls (1400, 1693) has measured the atomizaton efficiencies of Na, K and Li in the air/C,H, flame and obtained values of 0.98, 0.85 and 0.26, respec- tively; the value for K is somewhat higher than that normally expected. Atomization efficiency was found to be controlled by the equilibrium of the frec metal with the hydrox- ide and no evidence was found supporting the formation of oxides or carbides.Temperature-measurement techniques are of considerable importance and new methods are continually being sought. Winefordner et al. have published further papers (987, 993, 997, 1421) describing use of the “two-line fluorescence method” (see ARAAS, 1972, 2, Ref. 616 and ARAAS, 1973, 3, Ref. 711) for the measurement of localised temperatures in air/C,H,, N,O/C,H, and Ar/H, flames. The fluorescence technique has the advantage over other methods that it does not integrate the temperature and allows measurement of the temperature in localised volumes without the need for Abel transformation. Fassel and co-workers (1 537) have compared flame-temperature measurements using the reversal method and Fe two-line or slope technique and show that serious errors may arise in the latter case.These were attributed t o the uncertainties in the transition probabilities. 1.4.2 Interference Studies A large proportion of flame literature is devoted to the study of interferences. Much of the information is, however, of limited general use since in relatively few cases do authors give the mechanism of the process behind the interference.Often success amounts to circumven- tion of the problem, entirely satisfactory when answers are required, but this pragmatic approach leads to much duplication of effort and the recent literature suggests that progress towards the systematic treatment of interferences is relatively slow.This view is supported by Cresser (534) who emphasized the inadequacy of many interference studies. Fortunately the above comments do not apply universally and valuable work on inter- ferences has been described by several authors. Varju (715) has studied the effect of HCI and HNO, in the range 04.5 M on solutions containing Zn, Cu or Mn. In each case, plots of the percentage interference against acid molarity yielded straight lines that were mutually parallel.This was taken to indicate that the apparent interference was non-specific and in this case a correlation was indicated with the density and viscosity of the solution. Rubeska (688, 1142; see also ARAAS, 1976, 6, Refs. 861, 868, 924) has continued his excellent work on interference mechanisms associated with refractory species in the N,O /C,H, flame.By considering the parallel processes of volatilization and reduction, it has been found that interference behaviour can be classified according to the relative volatility of the metal, oxide and carbide species. Elements for which volatility increases from oxide to carbide, e.g., Al, Be, La and Y, are least susceptible to interference, elements for which volatility decreascs from oxide to carbide, e.g., Mo, B, Si, Ti, V, W, are more22 Analytical A tomic Spectroscopy susceptible, and the interference pattern for elements that are involatile in all three forms, e.g., Hf, Nb, Ta, Th and Zr, depends on the nature of their crystalline structure.Chakrabarti and Naranjit (1443) have studied the interference of phosphoric acid on Ca in the N,O/C,H, flame.It was suggested that the high boiling-point H,PO, produced larger droplets from the nebulization chamber, which therefore limited the occurrence of lateral diffusion in the flame. Hence, absorbances were enhanced at the flame centre as compared with CaCI, and reduced at the flame edge. The interference was overcome by determination of the Ca by ETA.The addition of carbon black (1516) to the air/C,H, flame has been found to produce enhanced atomization for several elements, particularly those having a strong affinity for 0, e.g., A1 and Ti. Further, the interferences of phosphate and fluoride on Ca and Fe determinations were eliminated. Pszonicki and Krupinski (690) have observed that inter- ference effects in the N,O/C,H, flame are consistent and that single species interferences can be predicted by exponential equations employing two coefficients, one describing the maximum degree of interference and its sign and one defining the shape of the interference function.It is stressed that the effects of individual interferents are not additive and there- fore the method cannot be applied when several interferents are present.A potential source of error in the determination of Si in natural waters has been identified by DeVine and Suhr (136). They showed that the absorption of Si in the N,O/C,H, flame could be enhanced by as much as 50% by the presence of Na. Although some background absorption was observed in the presence of Na, background correction had only a minimal effect in correcting the interference.The authors pointed out that as most Si standards were made from commercially available sodium silicates with unspecified Na content, existing analytical results had to be treated with some caution. The detennina- tion of Ba in silicates has been shown (293) to be affected by interferences from Ca and K and by synergistic interference effects, e.g., A1 in the presence of Ca and K, and Na and K in the presence of Ca.Dithiocarbamate has been found to be an effective releasing agent (846) in controlling the mutual interferences of the noble metals when determined directly in organic extraction systems, Kaszerman and Theurer (75) have reported increased sensitivity for the As(II1) state compared with As(V) in an Ar/H, diffusion flame; similar but less pronounced variations were found in the N,O/C,H, flame.Further, the presence of acids produced marked interferences on As(V) but had less effect on As(1II); KT was therefore used to reduce all As to As(I1I) before analysis. Oguro (92, 820, 544, 827, 340, 226) has reported an enhancement effect of NH,C10, in the determination of Eu and Yb by both AAS and AES in air/C,H, and air/H, flames.A twin-nebulizer technique has been employed by Kato et al. (318) to study the inter- ference of Ti and Nb peroxo-complexes on the determination of Cu, Mn and Co in Ni-based alloys in the air/C,H, flame. The interference was attributed to the formation of refractory compounds that occlude the Cu, Mn and Co leading to incomplete dissociation. Other rcference of interest - Removal of A1 interferences on Be by HF: 140. 1.4.3 Devices for Sample Introduction The sampling and atomization of solid materials presents particular difficulties for flame- spectrometric methods, which have been developed almost exclusively for solution analysis. Some solid materials are difficult to dissolve, contamination by the reagents used can prcsent problems and the dilution factor can render the analyte species unmeasurable. A common approach to overcoming these problems is to sample the solid directly using a higher temperature source, such as an arc, to vaporize the matrix material.Thus, Pungor et al. (235, 704) have described a d.c. arc nebulizer for use on small powder, metal or liquidPart I : Fundamentals and Instrumentation 23 samples and have used the device for the determination of Na and K in industrial Al,O, and electrocarborundum.Using an air/C,H, flame, detection limits were also established for Ca, Cd, Cr, Cu, Fey K, Li, Mg, Na, Pb and Zn in synthetic materials. Similarly, Posta and Papp (906) have designed an arc cell with an electrode specially configured for vapor- ization purposes and claimed increases of sensitivity in AAS of 2 orders of magnitude, compared with the equivalent solution technique.Pungor et al. (235) have proposed an apparatus for the nebulization of samples with a d.c. arc and subsequent analysis of the fine particles by AES. Operational parameters have been studied and methods presented for the determination of Na and K in aluminium and Pb in copper. A somewhat unusual device has been described by Razumov (466), which consists of an arc cell with the sample elec- trode at one end, and the whole is mounted in the flame produced by a standard long-path absorption burner.Other arc atomizaton devices have been reported by Chupakhin et af. (1362) and Rasumov (1512), and a review of the determination of powdered samples by AAS has been given by L'Vov (1667).The simplest approach to direct solid sampling is the dispersion method, in which the sample is finely ground, dispersed in a solvent and then sprayed in the usual manner. Labrecque (728) has reported results by this technique for the determination of Mo and Co in desulphurization catalysts and for Al, Si and Ti in laterites using both air/C,H, and N,0/C2H, flames. Fuller (294) has used the dispersion method for the determination of tracc elements in TiO, pigments.Chemical transforrnatiort has become more common in recent years as a mcans of improving the volatility of materials or of avoiding interferences. An example is flame gas-solid interaction, which has been employed by Fike and Frank (1094) for the determina- tion of I and Br at the ppb level.A standard O,/H, Beckmann burner was fitted with a stainless-steel chimney drilled with 6 holes to allow the passage of combustion air. A silver tube was placed on top of the chimney. When aqueous iodide or bromide was sprayed, AgX was formed on the tube surface, spraying was maintained for 1-5 min to allow accumu- lation of AgX.The solvent was then changed from H,O to 95% ethanolic solution, the rise in flame temperature vaporized the AgX and the halide was determined by measurement of the emission intensity of the Ag line at 338.2nm. Mallett and Royal (818) have reported the determination of 0 s in the N,0/C2H, flame by the prior formation of volatile ammo- nium chloro-osmate in a heated quartz furnace.Several metals, e.g., Au, Ru, Ni, Fe, Se, La and Al, interfered, but this was overcome by distillation of the 0 s and collection in HC1 solution before analysis. Similarly, Chapman and Dale (524) have reported improved sensitivities for the determination of Si and B by introduction of the volatile fluoride into the flame. The fluoride is generated by reaction with CuOHF at 345 "C.The use of the Delves cup technique for the determination of Cd in blood has been hindered by the large background absorption /scatter signals produced by combustion of the sample. It has been found (305, 710) that addition of (NH,),HPO, to the sample retards the atomization of Cd allowing resolution of its absorption peak from that of thc sample matrix.A detection limit of 16X 10-12 g was reported with an RSD of 0.056 at the 10 ppb level. MetaE speciation studies have become increasingly important particularly in the environmental ficld where chemical form rather than total content is often the determining factor in toxicity. Flame emission and absorption spectroscopy have been used for some time as detectors for GC and now their use is being extended for LC and HPLC.Van Loon et al. (1083) have demonstrated the use of AAS for the determination of Cr(V1) and Cr(II1) in the eluent from ion-exchange columns and have reported the analysis of alkyl and aryl Zn compounds separated on HPLC columns. Van Loon and co-workers (1423, 833) have also employed non-dispersive AFS as a means of chromatographic detection.GC coupled24 Analytical A tomic Spectroscopy with a flame emission detector has been used by Jungers et al. (932) to determine total S content in petrol. Analysis time was less than 5min with a detection limit of 0.002% and RSD of 0.1. Similar equipment has been used by Pearson and Hines (931) for the analysis of H,S, COS, CS, and SO, in inert gases and flowing hydrocarbon streams.Koop et al. (1062) have demonstrated that metal speciation in solid samples can be investigated by ramp heating of an ETA furnace, followed by flame or ICP analysis of the evolved vapcrur. Improvements in nebulizer design and use continue to be reported and indeed there is still much scope for development. The Babington nebulizer (see ARAAS, 1975, 5, Ref. 164) has been found to be an effective device for nebulizing high solids content solutions or those having high viscosity. Fry and Denton (484) have described the design of such a device and reported its use for the determination of Cu and Zn in whole blood, urine, sea-water, evaporated milk and tomato sauce. Hieftje and Savage (1238) have continued development of the electric-field pneumatic nebulizer (see AKAAS, 1976, 6, Ref. 11 37) and have described operating conditions and parameters for a practical device that can be coupled to flame or plasma atomizer instruments. The discrete-droplet generator of Malmstadt et al. (Anal. Chem., 1968, 40, 1860) has been employed to produce a “source modulator flame” for AAS with a continuum source (147). The atomic clouds produced by the atomization of indivi- dual droplets pass through the source beam causing it to be modulated.Synchronous detection becomes possible and the authors demonstrate that such a system is capable of simultaneous multi-element analysis, with relative freedom from narrow spectral line inter- ference. Tominaga et al. (1152) have found that the addition of the surfactant Na dodecyl sulphate (SDS) can increase Cr absorption by up to 120% in the air/C,H, flame, provided that it is present above the critical micelle concentration. Several interferences, i.e., those due to Fe, Ni, Co, C1-, NO,- and SO,2- were eliminated by the presence of SDS and it is suggested that the surfactant both improves nebulizer performance and enhances the atomization of the Cr. Other references of interest - Arc atomizer and source for AAS: 170.Arc discharge atomizer for AAS: 1502. Automation of sample introduction in FAAS: 1066. Controlling sample flow in flame AAS: 1122. GC-AAS for detection of volatile metal chelates: 420. Gas dynamics studies of chamber-electrode atomizer: 1509. Review of direct atomization of geological samples: 487. Ta cup for determination of Pb in sea-water: 76. 1.4.4 Other Studies Flame methods are widely used for routine analysis in laboratories throughout the world and an inevitable consequence of this is the appearance of papers describing devices that improve the performance of the flame method. Most of the devices reported do have some advantages to offer, usually increasing the residence time of the analyte in the flame, but none so far has given universal improvement and the benefits are usually related to a specific group of elements in a certain matrix type. An example of such a device is that dcscribed by Watling (1281).A slotted quartz tube was placed in the flamc such that a portion of the flame gases entered the slot and burned inside the tube, while the remainder formed a flame sheath round the outside.The absorption was measured along the tube axis and 4-10 fold improvements in sensitivity for the volatile elements As, Sb, Se and Hg were obtained. The improved sensitivity was gained at the expense of enhanced interference effects, which might preclude use of the device for all but simple sample matrices.Part I: Fundamentals and Instrumentation 25 Solvent extraction provides the analytical chemist with a tool for overcoming some of the difficulties posed by complex sample matrices.These may be of the interference type, or those related to sensitivity limits often imposed by the dilution factor necessary to take solid samples into solution form. The high-background levels and flame instability caused by spraying organic solvents is well known and often necessitates inclusion of a back- extraction step in the analytical procedure to obtain the more favourable aqueous form.Kono (31 1) has described an 0, sheathed flame system which apparently overcomes some of the problems and allows satisfactory analyses to be carried out directly on the organic phase. Similar claims for the N,0/H2 flame have been made by Lukasiewicz and Buell (1212) where the presence of the organic solvent had a beneficial effect in removing strong NO absorption bands from the background spectrum.Hieftje and Saturday (1262, 1179) have continued their studies on inert gas diluted flames (see ARAAS, 1976, 6, Ref. 1179) and have determined free-atom fractions under various flame conditions. Improvements in technique continue to be reported, particularly for samples that have specific composition or present difficulties when analyzed by conventional methods.Henrion et al. (267) have determined Sb in Bi-Sb semiconductors by chemiluminescent excitation in an air/H, flame. Propanol was the active species and was added to the sample solution; Sb or added Sn was used as internal standard.Two methods have been reported for the indirect determination of fluoride, the first involved precipitation of the fluoride as CeF, followed by filtration and determination of Ce by flame emission (222), and the second, the direct measurement of the intensity of the InF band at 328.12 nm in a N,/H, diffusion flame (429). The emission determination of Ag at 328.1 nm at low resolution is hindered by OH back- ground emission. Improved precision has been obtained using the Cr line at 357.9 nm as an internal standard (103).The excitation energy of the two lines differs by less than 1 eV. Urbain and Cattenot (689) have found that the atomization efficiency and therefore sensitivity of determination of the noble metals Pt, Rh and Ru is greatly enhanced by the presencc of SO,,- ions, The authors claim that SO,,- acts as an oxidising agent thereby preventing the formation of stable inter-metallic aggregates.Preconcentration techniques present several problems in use; a loss of precision is almost inevitable and the ever present danger of contamination becomes critical. Neverthe- less such techniques are often essential when adequate sample is available but the concentra- tion is too low for direct analysis. Jackwerthe and Messerschmidt (502) have described the preconcentration of trace metals dissolved in HCIIHNO,, on to high purity Ga. Those elements which are nobler electrochemically than Ga, e.g., Ag, Au, Bi, Co, Cd, Fe, Hg, Ni, Pb, Pd and Sn, are spontaneously deposited and enrichment factors of 103 can be obtained. Certain elements, e.g., Cd, Co, Fe, T1 and Zn, can be directly extracted from the Ga without the need for dissolution using HI and I,. A similar procedure whereby trace elements are electrolytically deposited onto a platinum spiral has been described by Lund et al. (1625). The spiral is then placed into an air/C,H, flame to volatilize the deposited material and the resulting vapour is localised by placing a quartz tube just above the spiral position. The technique was applied to the determination of the volatile elements Ag, Bi, Cd, Hg, Pb, Se, Te, Th and Zn in sea-water, urine and biological samples. Other references of interest - Absorption-line modulation by Zeeman scanning: 864. Characterisation of signals for micro-sample AAS: 894. Comparison of AES and AAS for steel analyses: 1452. Computer assisted construction of AAS response surfaces for Ru and Rh: 1231. Determination of halides by GaX and InX band spectra: 1656. Gas control system for flame photometer: 1505. MECA: determination of C,H,Br,: 190; determination of P compounds: 1018.