JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL. 6 25 Comparison of Atomization Mechanisms for Group IIA Elements in - ‘7 r; r-r; fir v IC ( i i f i ’ - ( _ ^ I - _ - -- rr-.r 7 - - i - - Laurie J. Prell* and David L. Styrist Pacific Northwest Laboratory,$ Richland WA 99352 USA David A. Redfield Northwest Nazarene College Nampa ID 83651 USA Atomic absorption and mass spectrometry were used simultaneously to elucidate mechanisms responsible for atomization of Group IIA elements (beryllium magnesium calcium strontium and barium) in pyrolytic graphite furnaces. Gaseous species of these elements deposited as the nitrates and vaporized in 1 atm of nitrogen and in vacuum were analysed in real-time by mass spectrometric sampling. The principal gas phase analyte species observed were carbides oxides and hydroxides.Excluding beryllium the data suggest that Group I IA atomization and carbide formation begin with the dissociative adsorption of the oxides and a perturbation of the surface state associated with the resulting adsorbed elements. Gas phase oxides are formed as a result of associative adsorption and the hydroxides are formed by homogeneous gas phase reactions of the carbides and oxides with water vapour. Keywords Atomization mechanisms; electrothermal atomic absorption spectrometry; mass spectrometry; Group IIA elements Careful investigation of the mechanisms that control the atom- ization of specific elements are providing a valuable database for achieving a more expanded view of fundamental electro- thermal atomization processes.However this approach does not indicate the scope of influence of the controlling mecha- nisms; the question of the validity of extending a given mecha- nism to include other elements is left unanswered. It is more valuable therefore to incorporate a group of elements in a modelling process in order to address this question. The Group IIA elements with the exception of radium are ideal for the purpose of investigating such generalizations within a group. Electrothermal atomic absorption spectromet- ric studies of this group involve primarily intra-group inter- ferences and Atomization mechanisms have been investigated for each of these elements but for the most part they remain to be determined. A systematic study of this group will at the very least provide experimental consistency in the vaporization data.Maessen and Posma4 suggested that beryllium vaporized from the solid metal after it was formed from the reduction of the oxide; however Maessen et al.s questioned these argu- ments. Frech et al.s performed high-temperature equilibrium calculations which indicated that beryllium should remain non-volatile up to 1400 K and analyte losses from beryllium hydroxide formation could occur as low as 1200 K if a rela- tively large volume of water was used. The X-ray diffraction analysis by Runnels et a1.6 found beryllium oxide on the inner furnace surface after heating to high temperatures. They attri- buted this to the beryllium carbide reacting with water while the sample was being transferred into the X-ray diffractometer.The last two research groups proposed that carbide formation competes with atomization. Styris and Redfield’ evaluated atomization mechanisms for beryllium by mass analysing gas phase molecular and atomic species evolved in the furnace. Their results imply atomization by thermal decomposition of adsorbed oxides. Several workers support the idea that magnesium vapour results from volatilization of the oxide followed by dissocia- tion in the vapour Khntor et al.H noted that the appearance temperature of magnesium was independent of the * Present address Lockheed Engineering and Science Company 1050 t To whom correspondence should be addressed. 3 Operated for the US Department of Energy by Battelle Memorial E. Flamingo Road Las Vegas NV 891 19 USA.Institute under contract DE-AC06-76RLO 1830. deposited species chloride or nitrate. They suggested that the free atom signal was due to oxide volatilization and dissocia- tion. Hutton et al.9 monitored oxide and hydroxide molecular emission spectra from the deposited nitrate in the furnace. The appearance of the oxide with the free atom suggested the above mechanism. However the large amounts of magnesium (500 pg) may have perturbed the system by producing changes in the dispersion of deposited analyte and thus contributed to changes in the vaporization scheme. Ohta and Su” atomized magnesium in tungsten tubes obtaining a detection limit 40 times less than that for graphite furnaces. They suggested that carbide formation may be interfering with the formation of the magnesium atom.Little effort has been made to understand calcium atomiza- tion. Hutton et a/.’ used 100 pg of calcium to provide molecu- lar emission spectra and noted the oxide appearing with the free atom and the hydroxide at a high temperature suggesting that free calcium results from volatilization and dissociation of the oxide. These results were disputed by L’vov’* who sug- gested the formation of calcium cyanide or calcium carbide instead of the oxide. Suzuki and Ohta13 used a molybdenum micro-tube to test their own hypothesis that carbide formation slowed atom formation. They observed peak-shape improve- ment when hydrogen gas was introduced. Barium probably the most extensively studied Group IIA element has provoked considerable disagreement concerning atomization mechanisms.StyrisI4 used vacuum vaporization and mass spectrometric analysis to help identify the atomiza- tion mechanisms of barium chloride in both pyrolytic graphite and tantalum furnaces. It was concluded that adsorbed barium oxide is reduced by the surface to give adsorbed barium which desorbs at higher temperatures. Jasim and Barbootiis proposed that barium is volatilized from the condensed phase after the oxide is reduced by the carbon surface. However Nagdaev and BukreevI6 suggested from appearance tempera- ture data that barium oxide dissociates after sublimation. Stur- geon et a[.” agreed with Nagdaev and Bukreev suggesting that the chlorides and oxides are thermally dissociated after va- porization. L’vov et al.ln used semi-empirical estimates of the heat of formation for the carbide” to surmise that the poor sen- sitivity of barium was due to the formation of low volatility carbides that are intercalated in the graphite.Intercalation de- creases the sensitivity for barium because the carbides are un- available for atomization.”’.” StyrisI4 observed gas phase barium carbide in mass spectrometric data. Equilibrium calcu- lations by Frech et al.? show that condensed barium carbide is26 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL. 6 only stable over a small temperature range between 1500 and ZOO0 K and that its formation is highly dependent on oxygen partial pressures. The Group IIA element whose atomization mechanism was studied most recently is strontium.22 The coincidence of the strontium atom signal with the molecular emission spectra of the oxide and hydroxide has been attributed to direct volatiliza- tion of strontium oxide (50pg of strontium).Y Nagdaev and Bukreevl‘j used the appearance temperature to conclude that free strontium formed from the sublimation of strontium oxide.L’vov et a1.23 atomized strontium from both tantalum and graphite platforms noting that the increase in the atomic ab- sorption (AA) signal was enhanced by the tantalum platforms. They concluded that the shape of the absorption signal was governed by atom supply and that pulse tailing was due to the formation of carbon shells on analyte micro-particles. Ohta and Su” used tungsten furnaces to test their hypothesis that carbide formation interferes with atom formation; they obtained detec- tion limits 250 times lower than those with graphite furnaces.When atomizing from a molybdenum furnace Suzuki and Ohta13 obtained detection limits of 5 pg and noted that 100-fold amounts of magnesium calcium and sodium did not interfere. Mass analysis of the gas phase molecular and atomic species in the furnace implies that the vaporization of strontium species involves three different types of adsorption sites.22 The ‘type 1 ’ sites involve dissociative adsorption and carbide formation. Associative adsorption occurs at ‘type 2’ sites and ‘type 3’ sites involve dissociative adsorption and free atom release. The above discussions indicate confusion concerning atomi- zation mechanisms for Group IIA elements. Experimental clarification must include information on real-time formations of intermediates that are produced in the furnace.Such infor- mation on atmospheric pressure vaporization is currently only available for beryllium’ and strontium.22 The experiments de- scribed here elucidate the remaining Group IIA element vapor- ization mechanisms by using real-time mass spectrometric analysis of the gaseous species associated with the atomization of these elements from pyrolytic graphite coated graphite fur- naces. The validity of extending to other Group IIA elements the mechanisms involving three types of adsorption sites pre- viously proposed for is therefore evaluated. Experimental Apparatus A mass spectrometry provided the capability for real- time molecular beam sampling and identification of gaseous species generated within pyrolytic graphite coated graphite furnaces (Thermo Jarrell-Ash Franklin MA USA) and heated in 1 atm of nitrogen (99.997%).In a few instances as noted in the text a second mass spectrometry apparatusi4 was employed to provide vacuum-vaporization data for the purpose of deter- mining carbide formation mechanisms. Procedures The dissolved nitrate of a particular Group IIA element was used for each sample Beryllium samples were prepared by dissolving beryllium nitrate (K & K Laboratories Painsville NY USA) in de-ionized water and then diluting to the appro- priate concentrations. Calcium and strontium samples were prepared from lo00 pg 1-I atomic absorption standard nitrate solutions (Spex Edison NJ USA). Magnesium and barium samples were prepared similarly with Specpure (AESAR Sea- brook NH USA) atomic absorption standard nitrate solutions.Solution volumes of either 2 or 42 pl were deposited into the furnace. The larger volume was used in order to decrease surface concentrations of the analytes while keeping the mass of the given analyte constant. Analyte masses are given in Table 1 column 3. The temperature programme for the atomization step for be- ryllium was set to ramp to 2700 K in 1 s and hold for 8 s. A 1 s ramp to 3000 K with a 5 s hold was used for the four other ele- ments. An Ircon 1100 series optical pyrometer (Ircon Niles IL USA) was used to monitor furnace temperatures by observ- ing the central outer portion of the surface of the furnace. The pyrometer output was calibrated against the inner-surface tube temperature to 2600 K with a 0.08 mm diameter W-5% Re W-26% Re thermocouple (Omega Engineering Stamford CT USA).The Extrel quadrupole mass analyser (Extrel Pittsburgh PA USA) was calibrated with water nitrogen oxygen argon carbon dioxide and trichlorotrifluoroethane. Electron energies for ionization were beryllium 30; magnesium 40; calcium 40; strontium 40; and barium 35 eV for the atmospheric pressure vaporization experiments; electron energies of 16 eV were used in the vacuum vaporization experiments. Three beams of a dual beam dual trace and single-sweep oscilloscope provid- ed temporal responses of the pyrometer atomic absorption spectrometer and mass analyser; the mass analyser was set to monitor a single mass during this sweep. The Tektronix 5 110 (Tektronix Beaverton OR USA) oscilloscope was triggered from the ‘read’ output of the furnace power supply and record- ed with a Tektronix C5C oscilloscope camera.Results and Discussion Temperature-dependent profiles of mass spectra (MS) obtained during the atomization step at atmospheric pressure are shown in Fig. 1 (oxides) Fig. 2 (carbides) and Fig. 3 Be 0 loo0 2000 3Ooo TIK Fig. 1 Composite oxide MS temperature profiles for Group IIA elements vaporized at atmospheric pressure. Intensities are plotted relative to the free atom MS signal. Ordinate scales differ from element to element.The mk values are shown in Table 1. Sample volume is 2 p1 0 lo00 2000 3Ooo TiK Fig. 2 Composite carbide MS temperature profiles for Group HA ele- ments vaporized at atmospheric pressure.Intensities are plotted relative to the free atom MS signal. Ordinate scales differ from element to element. The mk values are shown in Table I . Sample volume is 2 plJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 199 I VOL. 6 27 Table 1 Experimental summary and implied reactions Species Be0 (Beoh (BeO) MgO CaO SrO BaO SrO BaO Bell MA Can** sdl Ball Be,C BeC Be& Be$ Mg2C3 LTtt MgC2 LT CaC LT CaC2 LT SrC LT BaC,. LT$$ CaC HT CaC HT BaC HT BaC HT SrC2 HT SK2 LT MgCZ HT§§ Be(02H),SS Mg(OH)2 Ca(OH) Sr(OH) Ba(OH) BeNCN BeCN MgCN CaCN SrCN BaCN mlz 25 50 100 42 56 104 154 I 92 170 9 24 88 138 30 33 42 66 84 50 52 64 100 112 1 62 50 52 64 150 162 112 45 60 74 122 172 49 35 50 66 114 164 - Appearance Large-volume Analyte mass/ng temperature*/K load effect Implied reaction (reaction or reference No.) 20 20 20 50 50 50 90 200 200 10 25 50 50 90 20 20 20 20 25 100 200 50 50 50 90 100 100 50 90 90 50 20 50 50 50 90 20 20 25 50 50 90 2465 2590 2475 430 1650 1040 450 1700 2000 1990 1580 1940 1640 2080 2480 2355 2410 2465 440 410 600 590 420 430 500 1560 2020 2020 2215 2200 1995 1220 400 610 400 400 1750 2575 1725 2080 1730 2280 - NC t - - 0.1 x decrease c<O.1 x decrease Decrease$ - - NC 0 . 3 ~ decrease II 0 . 6 ~ decrease II 0 . 7 ~ decrease II 0 . 8 ~ decrease II - - NC - 5x increase ceO.1 x decrease 15x increase 40x increase Increase 0 . 2 0 ~ decrease 0 . 2 5 ~ decrease 0 . 2 5 ~ decrease 0 . 0 7 ~ decrease Increase 25x increase 20x increase 30x increase 12x increase - - - - - - - - - - (BeO) (ad)+(BeO),(g) + .. . + BeO(g); (reference 7) MO(ad type 2)+MO(g); (reaction 1) SrO(g) + Sr(g) +Sr,O(g); (reference 22) Ba,O,(g) + e- +BaO '+ (g) + O(g) + 2e- BeO(ad)+Be(g) + O(g); (reference 7) MO(ad type 3) + C :M(ad. type 3) + CO(g); T < T2 (reaction 3) M(ad type3) %M(g); T c T (reaction 2) MO(ad type 1) + C(s)+MCz(g) + CO(g); (reaction 4) * Appearnace temperature for a 2 pl volume. t NC = no measurable change. $ Thermal MS profile changes from multiple to a single peak which appears at 1720 K . 0 Ionizer-induced fragment of possibly Ba,02. 1 Direct comparison of the atomic absorption and mass spectrometric signals is not possible. Instrument constants that relate the relative signals to the total amount of particu- I1 Large-volume load effect for atomic metal species based on AA data only.**The 40 u stable isotope is not measurable because of the presence of "Ar. tt LT = low temperature. $3 Appeared only in 42 pI instance. $8 HT = high temperature. lar species in the furnace are not known and all species for a given element have not necessarily been analysed in the 42 WI instance. (hydroxides). The beryllium and strontium spectra which have been reported in references 7 and 22 respectively are includ- ed for completeness. The magnitude of the arbitrary ordinate unit remains constant only for species involving a given element. Molecular spectra are normalized to the full-scale free atom signal. The m/z values used in these experiments and their respective appearance temperatures are shown in Table 1 columns 2 and 4 respectively.For clarity no attempt has been made to show the full complement of species observed for each element; these species are listed in Table 1 column 1. In summary all Group IIA elements exhibit high- temperature (appearance temperature greater than or equal to that of the free atom) gas phase carbides. Magnesium calcium strontium and barium (42 pI volume only) exhibit low-temperature (appearance prior to atomization) gas phase carbides. Low-temperature hydroxides appear in all instances. Oxides appear at temperatures lower than those of the corre- sponding free atoms for all elements except beryllium.28 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL. 6 UlgCOH) A t r - Ca(OH) 0 lo00 2000 3000 TK Fig. 3 Composite hydroxide MS temperature profiles for Group IIA ele- ments vaporized at atmospheric pressure.Intensities are plotted relative to the free atom MS signal. Ordinate scales differ from element to element. The mlz values are shown in Table 1 . Sample volume is 2 pl. Since there was no observed beryllium hydroxide in the 2 1.11 instance no spectrum is shown here for Be The following discussions treat separately the oxide carbide and hydroxide species shown in Figs. 1-3; each group is dis- cussed under its respective sub-heading. Mechanisms are dis- cussed in detail and when sufficient data exist thermodynamic evaluations are provided. Thermodynamic data on adsorbed phases of these species are not available but arguments based on knowledge of general adsorption process- es are presented.Among the mechanisms that are implied by the data there exists a set that is consistent with all the data for any one element. If inter-element consistencies between these sets of mechanisms can then be found it may be possible to establish a general mechanism for the entire group. Oxides and Atomization Temperature profiles of the oxides observed for each element are shown in Fig. 1. The Sr,O+ discussed in reference 2 1 and BaO,’ discussed briefly below were also observed but are not shown in Fig. 1 because they were minor species and unique to strontium and barium. Beryllium is unique in Group IIA because it is the only element of the group having an oxide appearing at a tempera- ture greater than that of the free atom. A detailed discussion of the beryllium vaporization mechanisms is given by Styris and Redfield.7 These workers concluded that Be(g) is the result of the thermal dissociation of the adsorbed monoxide and that gaseous polymeric oxides are produced during desorption of the higher polymers.Appearance temperatures of the oxides of magnesium calcium strontium and barium (Fig. 1) precede those of their respective free atoms. Either bulk vaporization desorption or oxide release from direct dissociation of the nitrate are poten- tial mechanisms. The last mechanism does not evidently apply to the beryllium calcium strontium or barium because the appearance temperature of beryllium oxide is greater than that of the free metal; the appearance temperatures of hydroxides and carbides of calcium and strontium are greater than those of the respective oxides; and the barium oxide appearance tem- perature depends on the sample volume.Vaporization does not appear to be the release mechanism for the oxides because melting points*‘ range from 1300 to 2700 K higher than their appearance temperatures. Sublimation is not the mechanism either; all known sublimation pointsZ6 are at least 800 K higher than the appearance temperatures. The results can be ex- plained however by the desorption reaction where M implies Mg Ca Sr and Ba and ‘type 2’ refers to one of the three types of active sites available for oxide adsorption. For a physical description of each type of site in terms of the MO(ad type 2) -+MO(g) ( 1 ) related potential energies the reader is referred to reference 22.However a brief description of the sites is presented below. Two of these sites (‘types 1 and 3’) are postulated to explain how the metal oxide forms free metal atoms at one tempera- ture and metal carbides at a lower temperature (Fig. 2). It is thought that some of the available oxide dissociatively adsorbs on these two sites and the resulting oxygen atom reacts exo- thermically with carbon. Excitation of the lattice from the heat of this reaction will result in the release of the carbide at low temperatures should the lattice vibration modes be appro- priately coupled to the adsorbed species. This evidently occurs at ‘type 1 ’ sites (low-temperature carbides are observed) but not at the ‘type 3’ sites where the metal atom remains until desorbed at higher temperatures.Metal oxide desorption at more elevated temperatures requires the oxide to be associa- tively adsorbed as a molecule on the ‘type 2’ sites where it remains until the thermal energy required for molecular de- sorption is provided. It will be shown later that the above picture of ‘type 3’ sites is too simplistic to account for the ob- served high-temperature carbides and that a quantum descrip- tion of this site is necessary. In summary the involvement of ‘type 1 ’ sites is being invoked for dissociative adsorption and low-temperature carbide formation ‘type 2’ sites for associa- tive adsorption and oxide release and ‘type 3’ sites for dissoci- ative adsorption and free atom release. The broad thermal profile of BaO+ shown in Fig. 1 for the 2 pl samples implies a wide range of desorption energies for barium In this instance dissociative adsorption of the oxide at ‘type 1’ sites may be inhibited by the relatively high energies required for the dissociation of this oxide.Larger amounts of the oxide must then be adsorbed associatively. The ‘type 2’ sites therefore become saturated and the excess of oxide adsorbs on a range of available sites characterized by lower desorption energies. The use of larger volume (42 pl) samples resulted in the appearance of a single-peak barium oxide spectrum at 1720 K and a carbide profile at 500 K (Table 1). This should be com- pared with Figs. 1 and 2 for Iow-volume samples. The change can be accounted for by water-induced surface topology changes providing additional adsorption sites or an increase in the availability of ‘type 1’ and ‘type 2’ sites due to the more dispersed oxide; the ‘type 2’ sites are unsaturated because of this availability. Molecular desorption from ‘type 2’ sites would then be solely responsible for the observed single-peak oxide spectrum.To test the hypothesis of surface-induced topology changes 40p1 of water were desposited on the furnace surface and allowed to dry. A 2 pl volume of barium nitrate solution was then deposited at the same location. The resulting barium oxide spectrum was identical with that of Fig. 1 for the 2 pl volume hence water is not inducing the observed spectrum change. The important implication is that the dispersed low surface concentration of barium associated with large-volume samples increases the probability of barium oxide being disso- ciatively adsorbed at the high activation energy desorption ‘type 1’ sites.There is the possibility that the broad BaO’ spectrum is produced by the barium oxide slowly percolating from the bulk carbon or from near-surface regions. The fact that the barium oxide signal is weak compared with the other Group IIA oxide signals supports this; however the large volume load data are not explained by this possibility. A 170 u mass peak which may correspond to BaO,’ ap- peared at 2000 K (Table 1). It is questionable whether this species originates in the furnace. Certainly BaOz forms readily in air at 770 K from barium?’ or from BaO,*X but de- composes at 1070 K under equilibrium conditions.zH.z9 Further- more the peroxide has not been observed in Knudsen cells with the oxygen partial pressure as high as 101.3 Pa.30 Unfor- tunately furnace-related oxygen partial pressures reported by different workers range from 1 x lo-” to 1 at 2000 K.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL.6 29 Semi-empirical evaluations indicate that oxygen partial pres- sures associated with the gas phase decomposition of MO type compounds range from 0.001 to 100 Pa depending on the amount of It is therefore unlikely that BaO is present in the furnace. Ionizer-induced dissociation of Ba202(g) is probably responsib1e.N It has already been established that strontium is not vapor- ized from the condensed phase but its presence is due to de- sorption from ‘type 3’ sites.Furthermore previous barium vacuum-vaporization datai4 suggest that a desorption mecha- nism is also responsible for free barium. A 100 K decrease in the calcium appearance temperature when the analyte mass increases from 10 to 200 ng is observed. This results from either an nth order of desorption reaction (n > 0)36 or from increased occupancy of lower activa- tion energy desorption sites which increase the rates of de- sorption at a given temperature.’’ It is concluded therefore that magnesium calcium strontium and barium are desorbed and that their desorption can be described by the reaction which proceeds after the oxide molecule dissociatively adsorbs at a ‘type 3’ site. As discussed earlier if the dissociative ad- sorption products react with the surface to form CO(g) as was the situation for ‘type 1’ sites the associated exothermic energy must be poorly coupled to the lattice at the ‘type 3’ location since the Group IIA metal remains adsorbed to rela- tively high temperatures.The adsorbed free analyte species in reaction (2) is one of the products of the dissociative adsorption described by the reaction (3) Assuming equilibrium upper bound values for the Gibbs free energies for reaction (3) were derived using statistical thermo- dynamic procedures described in reference 38. Published rota- tional constants were used to calculate moments of inertia for the metal oxide.26 The maximum centre-of-mass to atom separ- ation was used as the width of the potential well and the atomic radii of the metals were used to approximate the surface to atomic metal separation.The resulting upper bounds for the Gibbs free energies for reaction (3) are -147 kJ mol-I at 1000 K for Mg; -148 kJ mol-’ at 1650 K for Ca; -137 kJ mol-I at lo00 K for Sr; and -138 kJ mol-I at 2000 K for Ba all of which become more negative with increasing temperature. MO(ad type 3) + C(s) -+M(ad type 3) + CO(g) Carbides Magnesium calcium and strontium carbide spectra (Fig. 2) appear at low and high temperatures. It will be shown that the high-temperature carbide formation mechanism proposed in reference 22 to explain the strontium results applies only to strontium. Surface state perturbation arguments will be pre- sented to explain the high-temperature carbides associated with the other elements. Condensed phase reactions are re- sponsible for the low-temperature carbides because molecular gas phase species necessary to form the carbide are not present at 450-600 K.Magnesium carbide exists in two forms at low temperature Mg,C and MgC,. The MgC is not an ionizer-induced dissoci- ation product of Mg,C,. If it were MgC should appear simul- taneously with the MgC,; it does not. Now Mg,C is known to form when an excess of carbon is available;a) hence some of the ‘type 1 ’ sites may be contributing to the excess of carbon. Large-volume (42 pl) solution loads greatly increased the magnitude of low-temperature carbide spectra for magnesium calcium strontium and barium (see Table 1 column 5 ) . In fact large-volume loading was a necessary condition for low- temperature barium carbide formation; this large-volume sample load result is not included in the spectra of Fig.2. These low-temperature carbide enhancements are attributed to increased surface to volume ratios of the oxide crystallites a result of the increased surface dispersion associated with large volumes. The probability of interaction of the oxide molecules with available active sites on the surface therefore increases. As suggested previously for strontium,22 the formation of low-temperature carbides is explainable through the reaction (4) The ‘type 1 ’ sites react with the adatoms from dissociative ad- sorption of the oxide a mechanism that requires participation of two surface defect^;^' the reaction products are MC,(g) and CO(g). This adsorption site type explains the following seem- ingly contradictory results; increases in low-temperature carbide signals with increased solution volume suggest that ‘type 1 ’ sites [reaction (4)] are preferred over other oxide ad- sorption sites i.e.‘type 1 ’ sites involve the greater heat of ad- sorption @,); and the low-temperature carbide appearance suggests a low heat of desorption. This is possible if the adatom products (M,O) of dissociative adsorption chemically bond with ‘type 1 ’ site carbon atoms. The resulting exothermic C-0 reaction supplies the energy necessary to desorb the prod- ucts at low temperatures hence the heat of desorption is effec- tively decreased. The site is envisaged as a surface defect (e.g. an edge dislocation) containing a ‘dangling’ carbon chain and a neighbouring defect with at least one ‘dangling’ carbon.This defect geometry is particularly reasonable for the pyrolytic graphite coated highly textured graphite substrates. A more detailed physical description of these sites is given in Appen- dix 1 of reference 22. When large-volume solution loads were used the low- temperature calcium carbide signal increased significantly (see Table 1 column 5) and appeared at the lower 450 K tempera- ture; the large-volume load disperses the sample and thus de- creases surface concentrations. This increases the probability of calcium oxides finding ‘type 1 ’ sites and adsorbing dissocia- tively. An increased probability for carbide formation follows. It is not clear how the activation energy for desorption dimin- ishes but the decreased appearance temperature implies that it does.As discussed above low-temperature barium carbides were observed only when barium was loaded as a 42pl solution. This is because the BaO dissociation energy is 540 kJ m01-’,~~ which is 3045% greater than the dissociation energies for the oxides of the other elements in the group. It is therefore ener- getically more difficult to adsorb barium oxide dissociatively. Such adsorption can be enhanced however. The more dis- persed barium oxides in the large-volume loadings have higher probabilities of being located in the proximity of sites with high heat of adsorption that exhibit lower crossover energies and hence lower activation energies of adsorption. Dissocia- tive adsorption is enhanced and this enhances the probability of carbide formation.It is perceived that barium oxide mole- cules from low-volume loadings desorb before they receive sufficient energy from the lattice to overcome the dissociative adsorption activation energy barrier. Fig. 2 shows high-temperature strontium carbide appearing several hundred degrees after the free atom. When a calcium interferent was added to strontium both the free atom and carbide appeared at lower temperatures.22 This mutual shifting implies that carbide formation is associated with the free atom. Furthermore high-vacuum vaporization data from both the outer and inner furnace surfaces’* indicate that most of the high-temperature strontium carbide is formed by the heteroge- neous reaction between the free atoms and the graphite surface (i.e.second wall) in the furnace as shown in the reaction MO(ad type 1) + 3C(s) +MC,(g) + CO(g) There is no indication from the vacuum vaporization data that other Group IIA elements participate in similar second-wall re- actions. The high-temperature gas phase carbides therefore evolve directly from the condensed phase a desorption30 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL. 6 process which is inconsistent with the low heat of desorption implied by the low-temperature appearance of these same car- bides. This paradox a manifestation of assuming that desorp- tion energies alone are controlling the rate of release of both carbides vanishes should the high-temperature carbide spectra be controlled instead by the carbide formation mechanism de- scribed below.The following mechanism based on solid state band theory is proposed. The carbide formation is initiated through adsor- bate-adsorbent electronic energy configuration changes that occur when surface localized electronic levels (surface states) associated with the adsorbate are perturbed by electron trans- fer between the bulk electronic band structure and the surface For example two surface states are indicated by the two observed high-temperature species (free analyte and carbide) One of these states is the ground state of the ad- sorbed analyte atom at the ‘type 3’ site. The second state which is related to the carbide formation is induced by elec- tron transfer between the above mentioned ground state and the electronic band structure of the ‘type 3’ site.When this occurs the localized adsorbate-adsorbent pair might no longer compose a minimum surface state energy configuration; the ionized adsorbate rapidly relocates to a ‘type 3’ site for example in order to achieve a new ground state configuration. The present data indicate formation of the carbide at the relo- cation position. Note that the two ground state configurations depend on occupancy (with or without the transferred elec- tron). This Franck-Condon type splitting of surface states de- scribed by Morrison,43 provides the ground states from which the observed high-temperature desorption occurs. Evidently the electron exchange between the graphite and strontium surface states occurs at a temperature greater than the appear- ance temperature for free strontium; thus the strontium carbide must form to a large extent by reaction (5).The question of why only strontium and not calcium or barium should be involved in the second-wall reaction now becomes important. A similar question addresses why the strontium desorption energy should exhibit the lower magni- tude. An answer based on Gadzuk’P quantum treatment of surface-induced perturbations of energy level spectra of free adsorbate atoms is presented. The theory that evolved predicts the formation of resonance states (virtual electronic excited states) due to the coupling of free adsorbate atom excited states with the band structure of the adsorbent. The virtual states which control the heat of have been ob- served by Plummer and Young4 for calcium strontium and barium on tungsten by field emission techniques. Plummer and Young suggest from their treatment of the structure in the energy distribution of the field-emitted electrons that the barium adatom ground state is a mixture of the ground state of the free atom and the first two excited states and that these states are shifted and broadened to form three overlapping bands below the Fermi surface.Calcium however has only its free atom ground state and the first excited state that are broad- ened and shifted by the surface interactions hence two virtual states exist below the Fermi surface. Finally strontium field emission data suggest that only a broadened free atom ground state below the Fermi surface accounts for the adsorbed ground state configuration.Similar adsorption behaviour might be expected on pyroly- tic graphite (PG) as non-localized or averaged work functions for PG and tungsten are similar (4.647 and 4.5 eV,4s respective- ly) and the substrate manifests itself through the work function in the theory of Gadzuk4. The more open structure of graphite will of course be influential but the relative behaviours of the adsorbates on the PG and on tungsten should be similar. It is expected that the number of virtual states below the Fermi surface of graphite should be greatest for barium and least for strontium. The strontium would therefore have the lowest heat of adsorption of the three elements and should appear at a lower temperature as observed. The over-all picture associated with high-temperature Group IIA carbides is simply that these carbides form as a result of Franck-Condon splitting of the ground surface states that evolve from the electronic resonance states associated with the adsorbate-bulk adsorbent pair.The weaker resonances expect- ed for strontium indicate an earlier release of strontium and thus a second wall reaction leading to carbide formation. The data suggest that the high-temperature magnesium carbide for- mation mechanism is similar to that of calcium and barium carbides. The beryllium carbides are the only Group IIA carbides whose formation is not explained by the above arguments. Earlier data from mass spectrometric investigations of berylli- um atomization indicate that beryllium carbide formation is closely related to the free oxide instead of the free metal and this formation can be described by the heterogeneous reaction’ xBeO(g) + (x +y )C(s)+Be,,C,(g) + xCO(g) Unlike the adsorption of the other Group IIA oxides only as- sociative adsorption occurs for beryllium oxide.This is evi- dently the reason for the absence of low-temperature beryllium carbides that require ‘type 1 ’ site dissociative adsorption. It is not clear however why beryllium is not adsorbed dissocia- tively. Perhaps the smaller size of the oxide allows such efficient trapping of the molecule that dissociative adsorption does not become energetically favourable. Changes in the magnitudes of products desorbing from each type of site indicate the relative probability of adsorbing a par- ticular species. The relatively large volume-induced increase in the intensity of the low-temperature carbide spectra (Table 1) and the corresponding decrease associated with the high- temperature atomic spectra suggest that ‘type I ’ sites are pre- ferred for oxide adsorption. This implies that the heats of adsorption associated with ‘type 1 ’ sites are of greater magni- tude than those at ‘type 3’ sites.Similarly by comparing rela- tive intensity changes for the gas phase oxide and free metal atom it is seen that the heats of adsorption at ‘type 3’ sites Q (3) are greater than those at ‘type 2’ sites Q(2). It is concluded that heats of metal oxide adsorption [Q(l)] for ‘type 1’ sites are such that Q( 1) > Q(3) > Q(2). This conclusion was made initially for strontium.’’ The present results extend this conclu- sion to magnesium calcium and barium.Hydroxides Temperature profiles associated with hydroxide formation are shown in Fig. 3. The formation of the hydroxides of magne- sium calcium and strontium are coincident with the low- temperature formation of carbides and exhibit signal intensity increases that are coincident with increases in the intensities of their respective carbides. The increases were observed when large-volume loads were used (see Table 1 column 5 ) and when calcium was used to enhance low-temperature strontium carbide formation.’? These observations support a reaction of the type (7) for Group IIA elements4() if this type of reaction is applicable to gas phase carbides. Calcium carbide and hydroxide appeared near 600 K when calcium was loaded as a small-volume sample.However when using large-volume sample loads the carbide and hydroxide appeared at 450 K. The decrease in the calcium hydroxide appearance temperature is perceived to be due to sample volume load promotion of dissociative adsorption as described previously under Carbides. Lowering of the temper- ature at which the carbide forms results in a lowering of the appearance temperature of the hydroxide formed by reaction (7) in the gas phase. The hydroxide of barium exists in spite of the paucity of a low-temperature carbide when low-volume loads are used.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL. 6 31 Reaction (7) does not apply but coincidence of the barium oxide (Fig. 1) and the hydroxide (Fig. 3) suggests the reaction Newbury30 observed a linear increase in the barium hydroxide generated in proportion to the amount of water vapour present in Knudsen cells.Beryllium hydroxide was observed by Styris and Redfield’ only when large-volume solution loads were used. In this in- stance a heterogeneous form of reaction (8) explains the for- mation of the hydroxide i.e. a reaction of water vapour with adsorbed instead of gas phase oxide. This is consistent with the nature of beryllium oxide to be adsorbed associatively but not dissociatively as discussed in the latter part of the previous section. The beryllium hydroxide was not formed when a 2 pl load solution was used because either the necessary amount of water vapour was then unavailable or the beryllium oxide was not sufficiently dispersed over the surface.Conclusion Analyses of real-time development of Group IIA gas phase species (except radium) in pyrolytic graphite coated graphite furnaces show that analyte losses occur through formation of gaseous carbides oxides and hydroxides. The data presented here are consistent with the conclusion of earlier workers that carbide formation is a competing mechanism in the atomiza- tion of the species.6.1’-’3.23 The data fail to support the sublima- tion processes proposed by Nagdaev and Bukreev,I6 to explain vaporization of barium and strontium. The mechanism involv- ing three types of adsorption sites previously proposed for strontium,21 was found to be applicable to other Group IIA elements. Atomization of oxides of the Group IJA elements excluding beryllium is the result of dissociative adsorption of the oxide.The free element and high-temperature carbides are a result of Franck-Condon splitting of the adsorbed elements surface states and associated equilibrium established between the re- sulting states. Strontium is desorbed early because it contrib- utes only a single virtual state below the Fermi level of the graphite; a smaller heat of adsorption relative to calcium and barium is indicated. The carbide of strontium is consequently formed by a heterogeneous (second wall) reaction. The oxide desorbs at lower temperatures after being associatively ad- sorbed at ‘type 2’ sites characterized by the lower heat of ad- sorption Q(2). The low-temperature carbides are the result of dissociative adsorption of oxide at the higher heat of adsorp- tion Q(1) (‘type l’) and reaction between the adsorption products and the carbon at the adsorption site.Exothermic energy from this reaction (CO formation) results in the low- temperature release of the carbide into the gas phase. The hy- droxide is formed by the reaction of water vapour with the low-temperature carbides. It may also be the result of the hy- dration of the adsorbed oxide but this is probably a minor path of formation. For those elements in which the oxide precedes free-atom formation the large-volume loads induce a 540-fold increase in carbide and hydroxide precursor losses. Noticeable decreas- es in high-temperature mass and atomic absorption spectro- metric intensities result from these volume-induced precursor losses.It would appear that by increasing the sample volume keeping the analyte mass constant the sample dispersion on the graphite surface is enhanced. This results in an increase of the probability of interaction with higher energy (‘type 1 ’) ad- sorption sites responsible for low-temperature carbide formation. Only the vaporization of beryllium appears to be controlled by different mechanisms. These mechanisms have been dis- cussed in reference 7. The authors express their gratitude to D. R. Ells for his invaluable assistance with the apparatus. This work was sponsored by the Director Office of Energy Research Office of Basic Energy Science Chemical Sciences Division of the US Department of Energy and performed under contract DE- AC06-76RLO 1830.Financial support for David A. Redfield was provided by the Northwest College and University Association for Science Washington State University under contract DE-AM06-76-RLO 2225 with the US Department of Energy. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 References Electrothermal Atomization for Atomic Absorption Spectrometry ed. Fuller C. W. The Chemical Society London 1977 pp. 67-69 74 and 80. Graphite Furnace AAS A Sourcehook ed. Slavin W. Perkin-Elmer Norwalk CT 1984 pp. 79-80 S 9 1 120 and 141. Frech W. Lundberg E. and Cedergren A. frog. Anal. At. Spectrosc. 1985,8,257. Maessen F. J. M. J and Posma F. D. Anal. Chem. 1974,46 1439. Maessen F. J. M. J. Balhe J. and Massee R. Spectrochim.Acta Part B 1978,33,3 1 I . Runnels J. H. Merryfield R. F. and Fischer H. B. Anal. Chem. 1975,47,1258. Styris D. L. and Redfield D. A. Anal. Chem. 1987,59,2897. Kintor T. Laszlo B. and Pungor E. Spectrochim. Acta Part B 1983,38,58 1 . Hutton R. C. Ottaway J. M.. Epstein M. S. and Rains T. C. Analyst 1977. 102,658. Gregoire D. C. and Chakrabarti C. L. Spertrochim. Acta Part B 1982,37,611. Ohta K. and Su S. Y.. Anal. Chem. 1987.59,539. L’vov B. V. Zh. Prikl. Spektrosk. 1979,31 16. Suzuki M. and Ohta K. Talanra 1981,28 177. Styris D. L. Anal. Chem. 1984,56 1070. Jasim F. and Barbooti M. M. Talanta 1981,28 353. Nagdaev V. K. and Bukreev Yu. F. J . Appl. Spectrosc. (USSR) 1980,33,6 18. Sturgeon R. E. Chakrabarti C. L. and Langford C. H.. Anal. Chem. 1976,48 1792. L’vov B.V. Kotshov D. A. Krugikova L. D.. and Polzik L. K. Spectrochim. Acta. Part B 1976,31,49. L’vov B. V. Spectrochim. Acta Part B . 1978,33 153. Fischer J . E. Comments Solid Stare Phys. 1978,8 153. Fischer J. E. Comments Solid State Phys. 1979,9,93. Prell L. J. Styris D. L. and Redfield D. A. J. Anal. At. Spectrom. 1990,5,23 1. L’vov B. V. Nikolaev A. V. and Polzik L. K. Specmxhim. Acta PartB 1988,43 1141. Styris D. L.. Freseniirs Z . Anal. Chem. 1986,323,7 10. Chemistry of the Elements eds. Greenwood N. N. and Eamshaw A. Pergamon Press Oxford. 1984 pp. I 17-1 54 and 29-379. JANAF Thermochemiiul Tables eds. Chase M. W. Davies C. A. Cowney J. R. Jr. Frurit P. J. McDonald R. A. and Syverud A. N. 3rd edn. (from J . Phys. Chem. Ref Data 1985,14 incl.) Schofield K. Chem Re\,. 1967,67,707. Kedrovskii 0. V. Kovtunenko I. V. Kiseleva E. V. and Bundel A. A. Russ. J . Phys. Chem. 1967,41,205. Hundbook ($ Chemistry and Physics ed. Weast. R. C. CRC Press. Cleveland OH 57th edn. p. B-93. Newbury R. S. US Atomic Energy Commission UCRL-12225-T. 1965. Cedergren A.. Frech W. and Lundberg E. Anul. Chem. 1984 56 1382. L’vov B. V. and Ryabchuk G. N. Spetwochirn. Actu. Part B 1982 37 673. Sturgeon R. E. Siu K. W. M. Gardner G. J. and Berrnan S. S. Anul. Chern. 1986,58,42. Sturgeon. R. E. and Falk H. Spectrochim. Acta Part B 1988 43 421. Sturgeon R. E. and Falk H. J . Anal. At. Spectrom. 1988,3,27. McNally J.. and Holcombe. J. A.. Anal. Chem. 1987,59 1 105. Chemisorption An E.vper.imental Appi-ouch ed. Welder G. Butter- worths Boston 1976 ch. 2. Styris D. L. and Redfield D. A. Anul. Chem. 1987,59,2891.32 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY FEBRUARY 1991 VOL. 6 39 Structural Inorganic Chemistry ed. Wells A. F. Oxford University Press Oxford 1984 p. 1382. 40 Advanced Inorganic Chemistrv eds. Cotton F. A. and Wilkinson G. Wiley New York 4th edn. 1980 pp. 271-288 and 351-373. 41 Chemisorption of Gases on Metals ed. Tompkins F. C. Academic Press New York 1978 pp. 4-6. 42 The Chemical Physics of Surfaces ed. Momson S . R. Plenum Press New York 1977 pp. 185-188. 43 Momson S. R. Surf. Sci. 197 1 24 1 10. 44 Gadzuk J. W. Phys. Re\!. 1970 1,2 1 10. 45 Schmidt L. D. and Gomer R. J . Cham. Phys.. 1966.45 1605. 46 Plummer E. W. and Young R. D. Phys. Rev. B 1970,1,2088. 47 Graphite and Its Crystal Compounds eds. Ubbelohde A. R. and Lewis F. A. Clarendon Press Oxford 1960 p. 106. ' Paper Ol016791 Received April I7th I990 Accepted August 31st 1990