JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 63 1 Tandem Sources Using Electrothermal Atomizers Analytical Capabilities and Limitations Heinz Falk Spectro Analytical Instruments Tiergartenstrasse 27 4 190 Kleve Germany Tandem emission sources with electrothermal volatilization and atomization of the sample such as furnace atomic non-thermal excitation spectrometry (FANES) or furnace atomization plasma emission spectrometry are compared with other excitation sources e.& an inductively coupled plasma. The interdependence between the excitation and atomization processes which causes restrictions for analytical procedures is considered. Of particular interest for the analytical applications of tandem sources is the influence of matrix constituents on the excitation conditions.Furnace atomizers are able to form relatively high matrix concentrations in the excitation plasma which can lead to a breakdown of the initial electron energy distribution. Correspondingly depression of the analytical signals by the matrix has been found for FANES which leads to some general conclusions on tandem sources. Keywords Atomic emission spectrometry electrothermal atomizer; tandem emission source; glow discharge matrix effect The glow discharge mode of excitation is well established in atomic spectroscopy. It is characterized by narrow spectral lines and a low background.' In addition the simple operation makes it especially suited for use as a spectros- copic light source. Of the glow discharges hollow cathode lamps have become broadly applicable in spectroscopy not only in atomic absorption but also as emission In the latter the glow discharge was originally used for both atomization of the sample by sputtering action and subsequent excitation usually in the negative glow plasma.A prerequisite for this procedure is that the sample consists of an electrically conducting material or has to be ground and mixed with a conducting powder such as c ~ p p e r . ~ The coupling of atomization and excitation leads to a relatively simple construction of such an emission source; it does however force the user to find an acceptable compromise for the operating parameters. For example changing the carrier gas from He to Ar will increase the density of the sputtered material within the plasma dramatically3 but will also shift the upper limit of the excited levels towards lower energies. The excellent signal-to-noise ratio achievable when using a glow discharge as the excitation source due to the non- thermal character led to a separation of the atomization step by using a separate device forming a 'tandem source' (see Fig.I). Transportation losses that occur while trans- ferring the sample vapour from the atomizer to the excitation volume can be avoided when the tube of the electrothermal atomizer forms the cathode of a hollow cathode discharge; this technique has been named furnace atomization non-thermal excitation spectrometry (FANES).6-8 Sample introduction drying and ashing are similar to those used in electrothermal atomic absorption spectrometry (ETAAS) but atomization is carried out at reduced pressure while the negative glow plasma inside the tube is maintained by the anode voltage applied to the electrode which is introduced into the vacuum vessel.By using FANES limits of detection typical for ETAAS are reached but in a simultaneous multi-element mode covering a high dynamic range of analyte concentrations. The original FANES instruments with the tube furnace as the cathode have been modified by the introduction of a small rod cathode into the furnace forming a hollow anode discharge (HA-FANES).9-11 This approach despite some experimental peculiarities is based on essentially the same physical principles viz. thermal atomization of a dry / Plasma - 7 ~ Tubing / f Sample Furnace 9===+ Photons / Plasma Anode T Fig.1 Schematic diagram of the two basic approaches for tandem sources consisting of an atomizer coupled to an excitation plasma. (a) Spatial separation of volatilization and excitation; and (b) volatilization and excitation in the same volume residue of the sample at reduced pressure and excitation of the vapours in a glow discharge. Similar to FANES from a constructional point of view is furnace atomization plasma emission spectrometry (FAPES).l2 Here a plasma is formed inside a tube furnace at atmospheric pressure with a high frequency antenna positioned on the furnace axis. The physical nature of a high-frequency plasma at atmospheric pressure differs from a glow discharge; however because of a number of similarities FAPES will be included in the following considerations.Glow discharge tandem sources are definitely suited for use in improving the analytical capabilities of conventional one-step emission sources. It is the aim of this paper to consider the analytical implications of the physical limita- tions of such tandem sources. Separation of Atomization and Excitation Tandem sources should allow a fairly complete separation of the processes of volatilization and atomization from the632 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 h 400 I I200 g 2 350 150 g 100 n 50 0 500 1000 1500 2000 250i Temperature/K Fig. 2 Temperature influence on discharge voltage and line emission in a FANES discharge.* Conditions A 9 x lo2 Pa He 50 mA; B 5 x lo2 Pa Ar 40 mA; C He (319 nm) 9 x lo2 Pa He 40 rnA; and D Ar (45 1 nm) 10 x lo2 Pa Ar 40 mA excitation step.In other words while atomizing the excitation conditions have to be kept constant. This can be achieved by spatial separation of the atomizer and plasma which is however not very efficient owing to the transpor- tation losses (see Fig. 1). On the other hand all tandem sources that accomplish the more efficient approach in which atomization and excitation take place in the same volume such as in FANES and FAPES show an interaction between the atomization and excitation processes. Depending on the elements to be determined the furnace temperature varies between ambient and 3000 K. While keeping the pressure constant during atomization there is a decrease of carrier gas density in the atomizer which is proportional to the furnace temperature.As a result the discharge characteristics and consequently the excitation conditions are changed. More marked is that the thermo- ionic emission exhibits an exponential dependence on the temperature causing a sharp increase of thermal electrons above 1 500 K. This completely changes the character of the discharge which is then no longer self-sustained. As an example the discharge voltage for FANES is given in Fig. 2.8 The temperature dependence of the carrier gas emission lines is also shown in this figure reflecting the change in excitation conditions as a result of variations in the plasma parameters. Whereas the He line intensity decreases 5-fold during the transition from typical glow discharge conditions to a regime governed by thermal electrons the Ar intensity change is insignificant.These intensity variations are attributed to changes in the excitation conditions as the gas density is diminished by less than 15% within the critical temperature range. The He line at 319 nm with an excitation energy (Eex) of 23.7 eV is much more influenced by the temperature of the cathode than is the Ar line at 45 1 nm with an excitation energy of 14.6 eV. The high energy ‘tail’ typical for glow discharge^,^ disappears when thermo- ionic emission is dominant. Obviously the density of electrons at the low energy end remains almost constant despite a 1 0-fold decrease in electrical power consumption of the discharge when the transition from the low to the high temperature region takes place.The same effect has been reported for HA-FANES the Ar I line (420.1 nm Eex= 14.5 eV) emission is almost unaf- fected while reaching a temperature of 1500 “C but the Ar I1 line (434.8 nm Eex=19.5 eV ionization energy Ei,,=15.76 eV) shows a dramatic decrease when the thermo-ionic emission is governing the discharge.’ This behaviour of the FANES system is very important for its analytical application as most analytes have relatively low excitation energies. Consequently elements with low and those with high volatilization temperatures show acceptable degrees of excitation.8 Matrix Influences on Excitation Conditions Electron Density and Temperature In electrical discharges the density and energy distribution of the free electrons determine the excitation conditions. All of the other excitation processes such as ion recombina- tion or secondary collisions are ultimately powered by electrons.Therefore the changes in the density and energy distribution of the electrons in a tandem source will be considered as they directly affect the analytical applications. A collection of data on electron density and temperature for a variety of spectroscopic excitation sources is given in Tables 1 and 2. These data show the typical variation in electron density in low-pressure discharges of between 1 x lo1* and 1 x 1014 ~ m - ~ whereas atmospheric plasmas e.g. the inductively coupled plasma (ICP) have density variations of between 1 x lOI4 and 1 x 10l6 ~ m - ~ . However the electron temperatures in both source types are compar- able ranging between 5 000 and 10 000 K.It is worth noting that in a low-pressure discharge the plasma is not in local thermodynamic equilibrium (LTE) as can be seen from the difference between the gas and electron tempera- tures in Table 2. Usually the electron velocity distribution is not Maxwellian therefore the term ‘temperature’ has to be used with care. Consequently for a glow discharge the term ‘electron temperature’ refers to an atomic ensemble in which the population of energy levels follows a Boltzmann distribution corresponding to that temperature and are essentially excited by the aforesaid electrons. In this sense the electron temperature of non-LTE plasmas yields infor- mation about the excitation of atoms and ions in such a system. Matrix Concentration in Emission Sources The interaction of the source plasma with the matrix constituents is a major reason for changes in the excitation conditions.The matrix concentration in the excitation region is considered below. For comparison purposes the ICP will be discussed first. The concentration of the sample vapour relative to the carrier gas in the ICP is given by where Qa is the aerosol gas (+plasma gas) flow rate in mol min-I; Q is the sample flow rate in mol min-l; andf is the nebulizer efficiency. When the sample contains the matrix concentration c (g ml-l) then the matrix contribution relative to the camer gas (crel ,,,) becomes As an example the following assumptions are made Qa= 2 1 min-l of Ar; Qs= 1 ml min-l of H20 ;f (pneumatic)= 3%; f (ultrasonic)=30°/o;16 and cm= 1%.Some results are also shown in Table 3. For an ultrasonic nebulizer with a cooler it is assumed that the aerosol reaches the saturation vapour pressure of water at the cooling temperature. As the example shows the matrix concentration in the ICP is fairly low under typical operating conditions whereas the concentration of water vapour is such that it will always be an influencing factor. The solvent concentra- tion has to be reduced when using ultrasonic nebulization. If sample introduction is carried out by sample sputter- ing e.g. by using a hollow cathode discharge with a discharge current of 50 mA a typical sample introductionJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 633 Table 1 Electron density of various spectroscopic radiation sources Gas Discharge Pressure/ temperature/ current Electron Source Gas kPa K density/mA cm-* den~ity/cm-~ GD* GD HCDt MIPS MIP FANES FANES FANES ICPC'l ICPh 11 Ar Ar Ar Ar Ar Ar Ar He Ar Ar 1.4 0.53 1.2 0.26 101.3 2.0 2.0 1.3 101.3 101.3 *GD glow discharge.t HCD hollow cathode discharge. $ MIP microwave-induced plasma. 5 NR not recorded. 1 ICPc 'cool' region of ICP above the coil. 11 ICPh 'hot' region of ICP inside the coil. 500 500 500 NR 500 2 500 500 5 000 8 000 150 98 100 NR§ NR 35 35 35 - 8 x lOI4 3.5 x 10" 1 x 1013 I x 1013 I x 1014-1 x 1016 3 x 1014 4~ 1014 2 x 1013 I x 1014-1 x 1015 I x 1015-1 x 1016 Method Spectrometric Probe Probe Probe Stark Spectrometric Spectrometric Saha Spectrometric Spectrometric Reference 13 14 1 15 16 8 8 8 17 17 Table 2 Electron temperatures measured in various spectroscopic sources Discharge Gas current Pressure/ temperature1 density/ Source Gas Wa K mA cm-* GD GD HCD HCD MIP MIP FANES FANES FANES FANES ICPC ICPh Ar 1.4 Ar 0.53 Ar 0.5 Ar 2.0 Ar 3.0 He 0.7 Ar 2.0 Ar 2.0 He 1.3 He 1.3 Ar 101.3 Ar 101.3 500 500 500 1500 500 500 500 2 500 500 2 500 5 000 7 500 150 98 13 80 NR NR 35 35 35 35 Electron temperature/ K 5 000 12 000 3 200 4 000 4 500 3 400 8 000 8 300 10 000 10 300 5 500 8 000 Method Spectroscopic Probe Spectroscopic Spectroscopic Spectroscopic Spectroscopic Spectroscopic Spectroscopic Spectroscopic Spectroscopic Spectroscopic Spectroscopic Reference 13 14 18 19 16 16 8 8 8 8 17 17 Table 3 Matrix concentrations in various emission sources relative to carrier gas (calculated) Source type Expected Maximum concentration (%) tolerable sodium Solvent Matrix concentration* (%) ICP pneumatic nebulizert,$ 1.9 0.0 19 0.05 ICP ultrasonic nebulizert,$ 19 0.16 0.05 ICP ultrasonic nebulizer + coolee 1 0.16 0.05 FANES (Ar)t,§"l 0 8 0.0005 0.002 Hollow cathode lamp§ 0 0.00 1-00 1 FAPESt,$,I II 0 8 0.0 1 * 10% loss of discharge power via matrix excitation.t Sample 1 % aqueous solution. §Discharge current 50 mA; and rdp= 1. fi 10 pl sample T= IOQO K. 11 T,=5000 K n,= 1 x lOI4 ~ m - ~ discharge power= 100 W. $rdp =0.1. Ratio of expected concentration to tolerable 0.4 J 3 16 000 800 0.5-5 rate of 0.2-1 pg s-l for metals is ~btained.~ The density of sputtered material ranges from 1 x 10" to 1 x 10l2 ~ m - ~ . ~ ~ Therefore the concentration of the sample relative to the (3) carrier gas is approximately 0.00 1 -0.0 1%.In a graphite furnace the dry residue of the sample is volatilized in around 0.2 s. The maximum concentration of sample species inside the furnace is given by where fat is the atomizer efficiency; V is the sample volume; N is Avogadro's number; n is the carrier gas density; M is the molar mass; and V the atomizer volume. In a low-pressure situation such a system has a typical634 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 120 -5 - 0 20 40 60 80 100 U A Electron energyJeV Fig. 3 A Excitation cross-section of Na (589.6-589.9 nm).22 B Dissociation cross-section of H2.24 C Ionization cross-section of Ar I.23 D Ionization cross-section of Hg 123 efficiency of 1% and at atmospheric pressure about 50°/6.21 By using a 10 pl sample containing a matrix concentration of 1% (M= 50) at a pressure of 2 P a an atomizer volume of 1 cm3 and a temperature of 1000 K which applies to FANES crrl,,,=8% is achieved.For FAPES operated at atmospheric pressure the same value is obtained. Such a relatively high matrix concentration is present in the furnace for approximately 0.2 or 0.5 s for FANES or FAPES respectively at the volatilization temperature of the matrix. In Table 3 various emission sources are compared and show that a relatively high matrix concentration can occur in furnace-based tandem sources. Interaction of Plasma Electrons With Matrix Species Electrons in emission sources lose their energy mainly by excitation or ionization collisions with the carrier gas but also with the matrix species.The number of excitation impacts (k,,) per time and volume unit is where n is the atom density; n is the electron density; u is the electron velocity; E is the electron energy; A,(E,) is the excitation cross-section; and f,(E,) is the electron distribution function. The expression f,(E,) x dEe has to be dimensionless. There is a similar expression for the ionization collision rate where the excitation cross-section is replaced by the ionization cross-section. However the excitation rate has to be balanced with a corresponding collisional de-excitation rate before the emission losses can be calculated. In fact the relative population in a 2-level system can be written as n* CE - no CD+RD ( 5 ) where n* is the excited-state population; no is the ground- state population; and CE CD and RD are the rates for collisional excitation collisional de-excitation and radiative de-exci t at ion respectively .Collisional de-excitation is typically 10 times higher than radiative de-excitation in an ICP,22 but in low-pressure discharges radiative de-excitation is the dominating pro- c ~ s s . ~ ~ In eqn. (4) n can represent the carrier gas in addition to species introduced by the sample. The actual energy loss of the plasma electrons depends on the excitation and ioniza- tion functions. Fig. 3 shows some typical examples with a very steep rise at the excitation or ionization energy.24*2s For comparison a dissociation cross-section function is in- cluded in Fig. 3 as this might also contribute to the energy losses of the plasma electron^.^^ Only energies of less than about 20 eV have to be taken into account as only electron temperatures of less than 10 000 K (see Table 2) corresponding to an average electron energy of 2 eV occur.In this instance the functions shown in Fig. 3 can be replaced by a rectangle starting at the corresponding energy for excitation ionization or dissocia- tion respectively. Under this pre-supposition and the assumption that the plasma electrons follow a Maxwellian distribution eqn. (4) can be integrated as follows ke,= 4 x n x n x A, x r I exp( a) k x T x (*+ k x T I ) ] where A, is the amplitude of the steplike excitation function; k is the Boltzmann constant; T is the electron temperature; and Eo and Em are the integration limits.Assuming that the photons emitted from the excited atoms leave the plasma the power loss (Pex) by this effect becomes pex = kex rdp Eph Vso (7) where Eph is the photon energy; V is the source volume; and rdp = RD/(CD+RD). By using eqns. (5-7) the concentration of any species in the plasma which would dissipate a given fraction of the electric power applied to the plasma via excitation losses can be calculated. This has been done for Na as the matrix component where the tolerable excitation loss is assumed to be 10%. The result is shown in Table 3 which gives only an upper limit for that matrix concentration causing a notice- able influence on the plasma. The calculations for the ICP apply to the 'normal' observation height about 20 mm above the coil where depression effects by the matrix are dominant.26 The power loss given by eqn.(7) can be observed as a signal depression if this effect is not compensated for by other effects which act in the opposite direction e.g. the increase in the electron density by matrix ionization. This takes place at low observation heights in the ICP. Table 3 shows that the maximum matrix concentration to be expected under realistic analytical conditions is usually considerably higher than the tolerable amount. Only in sputter sources does the influence of the matrix on the excitation conditions in the plasma appear to be negligible. Despite the uncertainties involved in the estimate it can be seen that influences of the matrix in tandem sources such as FANES and FAPES will start at relatively low concentrations.Analytical Implications In contrast to a nebulizer-operated ICP analyte and matrix are not necessarily present in the furnace volume at the same time in electrothermal atomizers. The actual matrix concentration for the duration of the analyte peak depends on the volatilization temperatures and residence times. Residence times can vary between typically 0.1 and 0.5 s for both FANES and FAPES. As Table 3 shows the worst instance occurs when analyte and matrix peaks coincide. Here it would be expected that in FANES the matrix influences become noticeable at matrix concentrations as low as 0.001%. Indeed a depression of the Cu signal in FANES for NaJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 635 140 t 2 20 401 Y ~ 1 lo-' 1 0.01 0.1 Na concentration (%I Fig.4 Emission intensity of Cu (324.7 nm) from FANES as a function of Na concentration (20 pl sample) for analytes with different volatilities. A Al; B Fe Ni Co and Cr; and C Cu. Volatilization temperatures Na 800; Cu 800; Fe Ni Co and Cr 850-1000; and Al 1100 "C concentrations of above 0.001% has been observed8 as shown in Fig. 4. For Fe Ni Co and Cr the depression starts at one order of magnitude higher Na concentrations whereas for Al such an effect does not appear at all. Under low-pressure conditions Cu and Na atomization takes place at 700-900 "C but Al is atomized at 1200-1 500 "C. The other elements mentioned are also volatilized within that range. The matrix effect on this group of elements could have been further reduced by using a more adequate heating programme allowing more of the matrix to volatil- ize before the Fe Ni Co and Cr peaks appear.Conclusions Despite the rather crude estimates of matrix effects in emission sources carried out in this work the following conclusions can be drawn (i) the separation of the volatili- zation and atomization stage from the excitation process cannot be completely achieved for tandem sources in which both processes take place in the same volume; (ii) glow discharges are more prone to matrix influences than excitation sources when working at atmospheric pressure; (iiz) tandem sources including furnace atomizers are very efficient therefore analyte and matrix concentrations in the excitation source are relatively high; (iv) for emission systems such as those used in FANES and FAPES an adequate temperature programme is crucial to keep matrix effects at an acceptable level; and ( v ) tandem sources need elaborate procedures for analytical application which might restrict their multi-element capability for samples containing high matrix concentrations. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 References Falk H.Spectrochim. Acta Part B 1977 32 437. Grimm W. Spectrochim. Acta Part B 1968 23 443. Improved Hollow Cathode Lamps ed. Caroli S. Ellis Hor- wood Chichester 1985. Broekaert J. A. C. J. Anal. At. Spectrom. 1987 2 537. Brenner I. B. Laqua K. and Dvorachek M. J. Anal. At. Spectrum. 1987 2 623. Falk H. Hoffmann E. and Ludke Ch. Spectrochim. Acta Part B 1981,36 767.Falk H. Hoffmann E. and Ludke Ch. Spectrochim. Acta Part B 1984 39 283. Falk H. Hoffmann E. and Ludke Ch. Prog. Anal. Spectrosc. 1988 11 417. Ballou N. E. Styris D. L. and Harnly J. M. J. Anal. At. Spectrom. 1988 3 1141. Harnly J. M. Styris D. L. and Ballou N. E. J. Anal. At. Spectrom. 1990 5 139. Riby P. G. Harnly J. M. Styris D. L. and Ballou N. E. Spectrochim. Acta Part B 199 1 46 203. Sturgeon R. E. Willie S. N. Luong V. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. Patel B. M. and Winefordner J. D. Can. J. Spectrosc. 1987 32 138. Fang D. and Marcus R. K. Spectrochim. Acta Part B 1990 45 1053. Brassem P. and Maessen F. J. M. J. Spectrochim. Acta Part B 1974 29 203. Zander A. T. and Hieftje G. H. Appl. Spectrosc. 1981 35 357. Magyar B. Guide-lines to Planning Atomic Spectrometric Analysis AkadCmiai Kiiido Budapest 1982 p. 145. Mehs D. M. and Niemczyk T. M. Appl. Spectrosc. 1981,35 66. Dobrosavjevic J. S. and Marinkovic M. Spectrochim. Acta Part B 1974 29 87. Kowollik G. Investigations on Emission and Absorption in Hollow Cathode Lamps for Atomic Absorption Spectrometry at Low Currents Thesis Humboldt University Berlin 1979. Falk H. and Tilch J. J. Anal. At. Spectrum. 1987 2 527. Lovett R. J. Spectrochim. Acta Part B 1982 37 969. Griem H. R. Plasma Spectroscopy McGraw Hill New York 1964. Massey H. S. W. and Burhop E. H. S. Electronic and Ionic Impact Phenomena vol. 1 Clarendon Press Oxford 1969. Physikulisches Taschenbuch ed. Ebert H. F. Vieweg & Sohn Braunschweig 1967 p. 4 18. Sun D. Zhang Z. Qian H. and Cai M. Spectrochim. Acta Part B 1988 43 391. Paper I /01508G Received March 28th 1991 Accepted August 21st I991