Scanning electron microscopy studies on surfaces from electrothermal atomic absorption spectrometry—III. The lanthanum modifier and the determination of phosphorus

Spec~ro&nica Acrcr, Vol.4lB,No. I I, wntedia Great Britain.

jq

OS84-8347/86 33.00+0.00

1175~1201,1986.

(c3 1986 Pa8amon JOUIDEIB Ltd.

Scanning electron microscopy studies on surfaces from electrothermal atomic absorption spectrometry-III. The lanthanum modifier and the determination of phosphorus BERNHARD

WELZ*, ADUON J.

CURTIUS? and GERHARD S~HLEMMER

Departmentof Applied Research,Bodenseewerk Perkin-Elmer

& Co GmbH, D-7770 ~~rlinge~

F.R.G.

and HUGO M.

ORTNER

and

WILHELM BIRZER

MetallwerkPlanseeGmbH, A-6600 Reutte,Austria

(Received 24 March 1986; in revised form 3 June 1986) Abstract-Morphological studieson graphitesurfacesby scanningelectronmicroscopyarepresentedfor platforms made from pyrolyticgraphite,and for polycrystalline electrograph&tubeswithpyrolyticgraphitecoatingin which

phosphorus was determined without and with the addition of higher concen~ations of ~n~nurn as the modifier. ~nth~urn causes severe pitting and corrosion of the graphite surf= already after relatively few determinations, and definite indication was found for the formation of intercalation compounds between lanthanum and graphite. No sign was found, however, for the formation of a dense coating of lanthanum carbide as proposed by several authors. The mechanism for the increase of phosphorus sensitivity is most probably the formation of a thermally stable compound involving lanthanum and phosphorus which leads to vaporization of phosphorus at high enough temperatures to obtain sufficient atomization and useful analytical signals. This is supported by the morphological changes of the graphite surface observed after application of higher lanthanum concentrations, and the resulting increased number of active carbon sites. Phosphorus alone also causes substantial corrosion of graphite, but with a completely different pattern. A very pronounced secondary coating of tube and platform wall is observed in the absence of lanthanum which is most probably supported by the formation and decomposition of compounds between phosphorus and graphite.

1. INTRODUCTION LANTHANUM WAS proposed

by MACHAT~as a matrix modifier for the determination of lead Cl, 21 and other heavy metals in biological materials [2, 33 as early as 1973. This author reported that matrix effects could be avoided by dissolving biological materials after dry ashing in a 1% lanthanum solution. This was confirmed by ANDERSSON [4] who found that lanthanum may also increase the sensitivity for lead due to the formation of a protective coating of lanthanum carbide at the graphite tube surface. Lanthanum was also reported to enhance the absorption signal and to eliminate interferences in the determination of beryllium in biological samples [5]. Soaking graphite tubes with lanthanum solution was found to reduce the memory effect and to increase the sensitivity for silicon [6], aluminium, beryllium, chromium and manganese [7]. These authors propose that competitive processes of ato~tion and carbide formation take place during the atomization step in a graphite tube. For elements such as lanth~~, carbide formation is dominant. Coating the inner surface with a thermally stable carbide precludes subsequent carbide formation of the analyte *Author for correspondence. TPresent address: Departamento de Quimica, Pontificia Universidade Cat6lica do Rio de Janeiro, Rio de Janeiro, Brasil. [l] G. MACHATA,Wiener K/in. Wochenschr. 85216 (1973). G. MACHATAand R. BINDER,Z, Rechtsmed. 73,219 (1973). G. MACHATA,WienerKlin. Wochenschr. 87,484 (1975). A. ANDERSSON, At. Absorpt. New&t. 15, 71 (1976). J. A. HURLBUT,At. Absorpt. Newsiett. 17, 121 (1978). D. B. Lo and G. 13. CHRISTIAN, Can. J. Spectrox. 2% 45 (1977).

(21 [33 [4] [S] [6] [7l

J. H. RUNNELS, R. ME~YFIELD

and H. B. FISHER,Anal. Chem. 47, 1258 (1975). 1175

1176

BERNHARD WELZet al.

element and thus favors atomization. An alternative explanation for the action of lanthanum is given by STURGEON and CHAKRABARTI [8] who point to the fact that this and other elements are active catalysts for graphitization. Graphitization can decrease the reactivity of the graphite surface. During the pretreatment stage of the graphite tube a carbide is formed which penetrates the pores of the graphite surface, dissolving the disordered carbon and precipitating it as nonreactive, small graphite crystallites. The improved sensitivities obtained for difficult-to-volatilize elements using carbide-treated tubes may therefore be the result of a more highly ordered graphite surface. Yet another explanation is given by EKLUNDand HOLCOMBE [9] who propose a preferential reaction of the added lanthanum with oxygen in the gas phase as the reason for the increased sensitivity found for lead and chromium. KOREEKOVA et al. [lo], finally, in their investigations of the reactions involved in the determination of arsenic assume that lanthanum stabilizes the oxidizing centres on graphite and in consequence also the created interlamellar arsenic compounds. The most widespread application of lanthanum as a modifier for graphite furnace AAS is in the determination of phosphorus [ 1l-151. The addition of lanthanum is reported to enhance the sensitivity for phosphorus up to six times over that obtained with aqueous solutions and eliminates most interferences observed. PERSSONand FRECH[16] point out that useful analytical signals for phosphorus can be obtained only if the vaporization of phosphorus compounds proceeds at high enough temperature. The first successful determination of phosphorus in a graphite furnace was reported by L’vov and KHART~YZOV [17]. These authors volatilized the analyte in a graphite tube heated to a constant temperature so that they did not need a modifier. In a Massmann-type furnace, however, this would lead to substantial pre-atomization losses, and one way to avoid this is to form thermally stable phosphorus salts. L’vov and RYABCHUK [ 181, similar to EKLUNDand HOLCOMBE [9], suggested that the enhancement of the phosphorus sensitivity in the presence of excess lanthanum can be attributed to purification of the purge gas in the way that lanthanum binds free oxygen in stable gaseous compounds. In essence, there is a wide variety of proposals on how the lanthanum modifier might act. Morphological studies should at least give information if changes of the graphite tube or platform surface, like coating with a carbide, are involved or not. In earlier work [ 191 we had already investigated the influence of lanthanum nitrate on graphite tube lifetime and surface morphology, but only a fairly low concentration of 100 mg 1-i lanthanum was used. After 200 determinations the platform cavity was fairly well coated with a layer of pyrolytic graphite, but no indication could be found for a coating with lanthanum carbide. In most applications, however, much higher lanthanum concentrations of up to 1% are applied or tubes are even soaked with lanthanum solution. The scope of this work was to investigate with scanning electron microscopy (SEM) and energy dispersive X-ray fluorescence spectrometry (ED XRF) if higher lanthanum concentrations cause any morphological changes at the graphite surface. The conditions chosen were those typically applied for the determination of phosphorus [20]. Visible alteration or corrosion of the surface is always related to eventual changes in the analytical performance of phosphorus in graphite furnace atomic absorption spectrometry (GF AAS).

[8] [9] [lo] [ll] [12] [13]

[14] [lS] [16]

[17] fl8] [19] [ZO]

R. E. STURGEONand C. L. CHAKRABARTI,Prog.

Analyt. Atom Spectrosc. 1, 5 (1978). Chim. Acta 108,53 (1979). J. KORE~KOVA,W. FRECH,E. LUNDBERG,J. A. PER~~ONand A. CEDERGREN,Anal. Chim. Acta 130,267(1981). R. D. EDIGER, At. Absorpt. Newslett. 15, 145 (1976). R. D. EDIGER, A. R. KNOTT, G. E. PETERSONand R. D. BEATY, At. Absorpt. Newslett. 17,28 (1978). A. PRBV~T and M. GENTE-JANNIAUX,At. Absorpt. Newslett. 17, 1 (1978). F. J. SLIKKERVEER, A. A. BRAAD and P. W. HENDRIKSE,At. Spectrosc. 1, 30 (1980). B. WEU, U. VOELLKOPFand Z. GROBENSKI,And. Chim. Acta 136,201 (1982). J. A. PERSSONand W. FRECH, Anal. Chim. Acta 119, 75 (1980). B. V. L’vov and A. D. KHARTSYZOV,Zh. Prikl. Spektr. 11, 9 (1969). B. V. L’vov and G. N. RYABCHUK,Spectrochim. Acta 37B, 673 (1982). B. WELZ, G. SCHLEMMERand H. M. ORTNER, Spectrochim. Acta 41B, 567 (1986). A. J. CURTIUS,G. SCHLEMMERand B. WELZ, JAAS, submitted for publication. R. H. EKLUND and J. A. HOLCOMBE,Anal.

Lanthanum modifier

1177

2. EXPERIMENTAL SEM micrographs were obtained with a Jeol JSM-35 CF scanning electron microscope capable of actual magnifications from 10 to 50,000 x. In the dark region below the pictures, the following information is displayed from left to right: accelerating voltage in kV, magnification, SEM micrograph number, scale of picture in m, month and year when the picture was taken. Topochemical analyses were carried out with the associated energy dispersive X-ray fluorescence analysis system Model 3000 (Princeton Gamma Tech) with Si(Li) detector and a resolution of 150 eV. They are referred to as ED XRP analyses. Phosphorus determinations were carried out with a Perkin-Elmer ZeemanJ3030 Atomic Absorption Spectrometer with HGA 600 Zeeman graphite furnace and an electrodeless discharge lamp for phosphorus operated at 8 W. The wavelength was set to 213.6 nm, and all instrumental parameters were set according to the manufacturer’s recommendations. The temperature program used for graphite furnace analysis is given in Table 1. All data, including the time resolved atomization signals, were plotted on a Perkin-Elmer PR-100 printer. Pyrolytic graphite coated tubes, Perkin-Elmer Part. No. BOlO-9322, and pyrolytic graphite platforms (“L’vov Platform”), Perkin-Elmer Part No. BOlO-9324, were used and also experimental platforms with two separate sampling cavities, described in detail elsewhere [21]. A phosphorus stock solution containing 5000 mg l- 1 P was prepared by dissolving the appropriate amount of diammonium hydrogen phosphate (analytical reagent grade) in deionized water. Lanthanum stock solution containing 50 g l- ’ La was prepared by dissolving the appropriate amount of lanthanum nitrate hexahydrate (analytical reagent grade) in deionized water. Sufficient nitric acid was added so that the final solution contained 0.2 y0 (v/v) nitric acid. All further dilutions were made with nitric acid of the same concentration.

3.

RESULTS AND DISCUSSION

3.1. Tube conditioning SEM micrographs of pyrolytic graphite coated tubes and pyrolytic graphite platforms have been shown on several occasions so that reference can be made to previous publications C19,22]. Before tubes and platforms are used analytically it was found very useful to apply a

“conditioning program” in the graphite furnace. This consists of four “dry” heating cycles to 265o”C, the first with a ramp time of 60 s, the other three with a ramp time of 10 s each. The hold time at 2650°C was 2 s, in the first, and 10 s each in the other three cycles. One tube and platform was inspected with SEM after this conditioning to have a kind of a ‘blank” for comparison with the planned experiments. Like in earlier dry heating experiments [ 191, which, however, comprised many more heating cycles, tube and platform showed a thin secondary coating with pyrolytic graphite (Fig. 1).This secondary coating is most apparent in Fig. l(a) which shows the rim of the injection hole where the thin layer is partly removed. From this it appears as if the conditioning process is nothing but a “healing” of imperfections on the graphite surface by the secondary coating. It can be anticipated that this deposition of carbon from the vapor phase starts preferentially at lattice imperfections and active carbon

Table 1. Time-temperature program for the determination phorus. HGA-600 Z.eeman graphite furnace Step No.

Furnace temperature (“C)

1 2 3 4 5 6

90 120 1350 2650 2700 20

Time ramp hold (s) (s) 1 15 1 0 1 1

10 20 30 5 4 8

Internal gas flow (ml min- ‘) 300 300 300 0 300 300

of phos-

Read *

[21] B. WELZ, S. AKMAN and G. SCHLEMMER,Analyst 110,459 (1985). [22] H. M. ORTNER, G. SCHLEMMER,B. WELZ and W. WEGSCHEIDER,Spectrochim. Acta 4OB, 959 (1985).

1178

BERNHARD WELZet al.

Fig. 1. Secondary coating of tube and platform with a thin layer of pyrolytic graphite after tube conditioning. (a) rim of injection hole with partly removed secondary coating, (b) bottom of the platform cavity.

sitesI at the tube and platform surface [19]. Conditioning consequently simply makes the graEwhitesurface less reactive. Another phenomenon which has also been observed in earlier investigations after pro1onged use of graphite tubes is graphite nodule formation [19,X&23]. Such nodules can befi ound at the graphite tube surface already after the conditioning step, as can be seen in Fig. l(a), and even in an unused tube, such “growth cone” structures, which are actually [23] H. M. ORTNER,W. BIRZER,B. WELZ,G. SCHLEMMER, A. J. CURTIUS,W. WEG~CHEIDER and V. SYCHRA, Fresenius Z. Anal. Chem. 323,681 (1986).

Lanthanum modifier

1179

defects in the pyrolytic graphite lattice, originate during the deposition process of pyrolytic graphite [24]. 3.2. Determination of phosphorus from a dual cavity platform When phosphorus is determined from a pyrolytic graphite platform in a pyrolytic graphite coated tube without the addition of a modifier, the sensitivity for this element is very poor with a characteristic mass of about 700 ng (0.0044 As)- ‘. When 0.5 % (m/v) lanthanum solution is added as the modifier the characteristic mass in this experiment is improved by more than two orders of magnitude to values around 5 ng *(0.0044 As)- ’ under otherwise identical conditions [20]. A very interesting phenomenon was observed when phosphorus was determined on a dual cavity platform [21] where the phosphorus solution was pipetted in one and the lanthanum modifier solution in the other cavity. As can be seen in Fig. 2 lanthanum clearly has an influence on the phosphorus signal even under these conditions where they do not come into contact in the condensed phase. Lanthanum increases the phosphorus signal already when it is applied for the first time. One explanation for this could be the mechanism proposed by L’vov and RYABCHUK [18] who assumed a preferential binding of oxygen in the gas phase by the lanthanum. Another explanation is that phosphorus is volatilized in some molecular form (e.g. an oxide) during the thermal pretreatment step but reacts in part with lanthanum in the other cavity to form a thermally stable compound instead of being carried out of the absorption volume without being atomized. Such reaction mechanisms have already been proposed for other analyte-modifier combinations when the dual cavity platform is used [21]. While lanthanum in the other cavity increases the phosphorus signal over that obtained in the absence of a modifier, the sensitivity is only about one fifth of that when the solutions are mixed (Fig. 2). The phosphorus signal increases, however, with the repeated application of lanthanum and reaches finally, after some ten determinations, about the same level which is obtained for the mixed solutions. A possible explanation for this slow increase in sensitivity would be that lanthanum is transported into the phosphorus cavity via the gas phase and alters the surface of this cavity for example by vapor deposition and formation of a lanthanum carbide coating. Inspection by SEM and ED XRF was therefore thought to be a useful diagnostic tool to learn more about the reactions and mechanisms involved. 0.3-

o- : ’ 1

5

iii No. of Determination

15

Fig. 2. Integrated absorbana signal for 0.2 pg P in the presence of 25 pg La (5 ~1 of a 0.5 % La solution as the nitrate)from the dual cavity platform. Determination No. 1:phosphorus only, without lanthanum; determination No. 2: blank (0.2% v/v HNOs); determinations 3-15: phosphorus and lanthanum. A: phosphorus and lanthanum solutions mixed in the same cavity. B: separated injection of phosphorus in one and lanthanum in the other cavity. [24] W. HUEITNERand C. BUSCHE,Fresenius Z. Anal. Chem. 323,674 (1986).

1180

BERNHARD WELZet al.

The dual cavity platform in which the phosphorus and the lanthanum solutions were pipetted into separate cavities was inspected after 15 applications of lanthanum solution and a total of 26 atomization cycles (including a few additional blank firings and applications of phosphorus only). This is shortly after the sensitivity for phosphorus has reached a stable value as is shown in curve B of Fig. 2. The bottom and the inner wall of the cavity in which the lanthanum solution was pipetted are shown in Fig. 3. In this cavity a lanthanum signal could be obtained upon ED XRF analysis, but the signal was much weaker than could be expected for a coating with a layer of lanthanum carbide. The change in morphology, if for example Fig. 3(a) is compared with Fig. l(b), is therefore predominantly due to pitting and corrosion and not to a carbide

Fig. 3. Morphology ofplatform cavity after 15 applications ofa 0.5 % lanthanum solution and a total of 26 atomization cycles. (a) platform cavity bottom, (b) inner wall of cavity.

Lanthanum modifier

1181

coating. The change in the morphology is most dramatic at the inner wall of the cavity where no indication is left for the originally lamellar structure (compare [19]). The bottom of the cavity in which the phosphorus solution was pipetted is shown in Fig. 4. There is no visible indication for any kind of a coating, and no lanthanum signal could be detected upon ED XRF analysis. There is, however, substantial pitting and corrosion in this cavity, but with a clearly different morphology compared to the lanthanum cavity. It cannot be decided from this micrograph if the corrosion in the phosphorus cavity is due to the attack of the phosphorus solution or due to an attack by lanthanum via the gas phase. The SEM inspection cannot give an answer to the question which way lanthanum increases the phosphorus signal when the two solutions are pipetted into separate cavities of the dual cavity platform. It could, however, clearly be shown that the sensitivity increase is not caused by a lanthanum carbide coating of the platform surface. Such a coating of the graphite surface with lanthanum carbide, while postulated by several authors [4,6,7], is not very likely anyway since the latter decomposes in water. The repeated application of aqueous solutions will most probably prevent the formation of a dense layer of lanthanum carbide. A reaction of lanthanum with carbon may well be responsible, however, for the severe pitting because lanthanum carbide, which undoubtedly forms at the graphite surface, is dissolved leaving reactive carbon sites at the surface. One possible explanation for the action of lanthanum as a modifier is that it simply creates active carbon sites which, in turn, stabilize the phosphorus [20]. A similar mechanism has already been proposed by KORE~KOVA et al. [lo] for the effect of lanthanum on the determination of arsenic. Both elements form very stable metallic carbides of the general form MezCg[25]. It is also in agreement with the observation that phosphorus, without the addition of a modifier, can be determined with a better sensitivity in an uncoated polycrystalline electrographite tube compared to a pyrolytic graphite coated tube. The stabilization of phosphorus at active carbon sites, however, appears not to be very effective because the

Fig. 4. Morphology of the phosphorus cavity bottom of the dual cavity platform after 26 atomization cycles.

[25] F. BENESOVSKY,

Ullmanns Encykloptiie

Chemie, Weinheim (1975).

der technischen Chemie, 4th edition, Vol. 9, pp. 122-136. Verlag

BERNHARDWELZ et al.

1182

sensitivity without a matrix modifier is always substantially lower and decreases further with increasing thermal pretreatment temperature [20]. The most obvious mechanism of stabilization appears therefore to be the formation of a thermally stable compound between phosphorus and lanthanum with some participation of carbon. This compound can be formed via gas phase transport of a phosphorus oxide into the lanthanum cavity, and, after several atomization cycles, also by transport of lanthanum into the phosphorus cavity. The former explains the increased sensitivity in the first determination of phosphorus with lanthanum in the other cavity, and the latter explains the further increase in sensitivity during the following determinations after repeated application of lanthanum until essentially the same value is reached as with the lanthanum admixed to the phosphorus solution. 3.3. Endurance test 0.5 % lanthanum solution Usually, endurance tests include several hundred determinations before massive corrosion is observed, particularly if samples are atomized from a platform [ 191. However, in the case of phosphorus determination with lanthanum as the modifier, we observed a strange phenomenon already after some 20-30 determinations. Toward the end of the phosphorus absorption signal, very rapid spikes appear in the background channel which disturb the measurement and make a reasonable signal evaluation difficult or even impossible (Fig. 5). This spiking is very random and occurs earlier in the atomization cycle with increasing number of determinations, but may disappear again for several determinations. The endurance test was interrupted and tube and platform inspected after 38 applications of 0.5 y0 lanthanum and a total of 45 atomization cycles (including seven atomization cycles with blank solution or phosphorus only). As can be seen in Fig. 6, the platform is substantially bent by that time, a phenomenon which had not been observed previously with any of the other samples or matrix modifiers. Closer inspection of the cavity, in which lanthanum solution was applied, shows severe corrosion all around (Fig. 7). The inner wall of the cavity no longer exhibits any lamellar structure and the bottom shows severe pitting and disintegration of the graphite material. Figure 8 shows a fracture pattern of the deliberately broken lanthanum cavity. A light, flaky surface layer can be found at the cavity bottom on top of the lamellar pyrolytic graphite.

0.3 A

0

Fig. 5. Time resolved atomization signals for 0.2 pg P atomized from the dual cavity platform with 25 pg La in the other cavity. Severe “spiking” in determinations Nos 30 and 33, but not in No. 31.

1183

Lanthanum modifier

Fig. 6. Dual cavity platform after 38 applications of lanthanum (0.5 % solution) and a total of 45 atomization cycles.

This cavity gave a strong lanthanum signal upon ED XRF analysis. Due to the roughness of the inspected area and the consequent electron scatter, it was not possible, however, to relate this lanthanum signal to a particular portion of the platform or to a morphological structure, for example to the light flakes. But it is obvious from Fig. 8(a) that there is no dense layer of lanthanum carbide which could be considered as a kind of a coating and which would prevent contact of the analyte element with the underlying carbon material. This becomes even more apparent from Fig. 8(b) which shows the open structure of separated graphite layers. These layers are of course not the individual atomic graphite layers, the distance of which is typically in the order of a few tenths of a nm, but multiatomic layers that are formed during crystal growth. The original distance of these layers has been widened substantially, most probably by the formation of intercalation compounds with lanthanum. The formation of such intercalation (inclusion) compounds results in an increase in graphite volume until, finally, it “puffs up” [26]. This increase in the graphite volume at the surface causes a substantial shearing force which finally leads to the bending of the platform as shown in Fig. 6. Another phenomenon observed in the lanthanum cavity is the formation of bubbles or blisters, examples of which are depicted in Fig. 9. These blisters give only the background signal but no response for lanthanum upon ED XRF analysis. They all appear to be hollow shells which may have formed during thermal pretreatment and burst during atomization to release their contents. While we certainly don’t have enough evidence to propose a mechanism for the formation of these blisters, it should be mentioned that low hydrocarbons and hydrogen are formed upon hydrolysis of lanthanum carbide with water [25]. It is not unlikely that these blisters are the reason for the severe spiking which is depicted in Fig. 5. If particles sputter through the absorption volume because of such blisters bursting at high temperature, this may well lead to spurious signals like those observed in our experiments. The phosphorus cavity of the dual cavity platform also shows substantial pitting, as can be seen in Fig. 10. The morphology of the corrosion, however, is different from that of the lanthanum cavity. This becomes apparent when Fig. 10 is compared with Fig. 7(b) which was taken at the same magnification. As in the previous experiment, it is not possible to decide if

[26] W. R&ORFF, E. STUMPP,W. SPRIESSLER and F. W.

SIECKE,

Angew.

Chem. Int. Ed. EngL 2, 67 (1963).

1184

BERNHARDWELZ

et al.

Fig. 7. Severecorrosion of the lanthanum cavity of the dual cavity platform shown in Fig. 6. (a) inner wall of the cavity, (b) cavity bottom.

this corrosion is due to a condensed phase interaction with the phosphorus solution or due to an attack of lanthanum via the gas phase. Again, however, there is no sign for a lanthanum carbide coating in the phosphorus cavity which could be made responsible for the increase in phosphorus sensitivity observed experimentally. A totally new corrosion pattern, which is depicted in Fig. 11, was found upon inspection of the tube in which the endurance test was performed. The surface morphology, when compared to that of an unused or conditioned [Fig. l(a)] pyrographite coated tube, has undergone a dramatic change. Quite uniform pitting is found all over the tube surface, in and around the injection hole. This corrosion is most likely caused by an attack of lanthanum via the gas phase. When volatilized at high temperatures, lanthanum can react with active carbon

Lanthanum modifier

1185

Fig. 8. Fracture pattern of the lanthanum cavity at different magnification. The light surface layer gives a strong La signal upon ED XRF analysis. (a) fracture pattern overview, (b) macro layers separated by intercalation, detail of (a).

sites at the tube surface to form a lanthanum carbon species, for instance lanthanum carbide. This may then react with water vapor generated during the drying step, and decompose, creating in this way additional reactive carbon sites at the surface. Another interesting phenomenon can be seen in Fig. 11(b) which shows corroded and uncorroded graphite nodules. The corroded nodules can be considered to be “primary” growth cones which had already formed during the pyrolytic graphite coating process or in the conditioning phase. The uncorroded graphite nodules, however, must have been generated very recently, i.e. during the last few atomization cycles before tube inspection, This appears to confirm the previously made assumption that the graphite nodules may grow quite rapidly under favorable conditions.

1186

BERNHARDWEU et ai.

Fig. 9. Blister formation found in the lanthanum cavity of the dual cavity platform. (a) inner wall of the cavity, (b) cavity bottom.

A comparison should also be made between Fig. ii and the bottom of the phosphorus cavity shown in Fig. 10. The corrosion patterns of the two micrographs exhibit similarities which propose that the source and the mechanism of the attack could be similar as well. This means that the pitting in the phosphorus cavity could well be caused by a gas phase attack of lanthanum which is volatilized from the other cavity. The lanthanum-carbon species formed in this reaction, of course, dissolves and decomposes again upon the injection of the next analyte solution, leaving reactive carbon sites on the platform surface. 3.4. Corrosion test with 2% lanthanum solution and “mixed” injection The integrated absorbance signal for phosphorus increases with increasing lanthanum concentration up to about 2 y0 lanthanum. Above 1% lanthanum, however, the signals

Lanthanum modifier

1187

Fig. 10. Pitting at the bottom of the phosphorus cavity of the dual cavity platform shown in Fig. 6.

which proposes that phosphorus is retained one way or another. An additional experiment was therefore carried out using 5 ,ul of a 2 % lanthanum solution (100 pg La) as the modifier. In addition, the analyte solution containing 0.2 pg phosphorus was mixed with the lanthanum solution in one cavity of the dual cavity platform. The other cavity was left empty throughout this experiment to gain more insight into the gas phase attack of lanthanum on the graphite surface. With the 2 % lanthanum solution, the spiking, shown in Fig. 5, started to occur much earlier. The experiment was therefore interrupted after only 11 applications of lanthanum and a total of 17 atomization cycles, including six atomizations with blank solution or phosphorus only. The platform showed about the same degree of bending as that of the previous experiment, depicted in Fig. 6. A perpendicular view into the two cavities,, at the low magnification of Fig. 12(a), shows very pronounced corrosion in the cavity which contained the lanthanum and phosphorus solutions. The other cavity, into which no solution was injected at any time, however, appears to be’corroded to some extent as well. A closer look into the (deliberately broken) sampling cavity [Fig. 12(b)] reveals the extent of pitting and disintegration that only 11 applications of the 2 % lanthanum solution have caused. The light, flaky layer at the bottom of the cavity, like in the previous experiment,-gave a strong lanthanum signal upon ED XRF analysis. Figure 13 depicts revealing details of the surface morphology on the platform cavity bottom. It is obvious that the surface is nowhere sealed or covered by a layer of a carbide. Even if the light, flaky material is considered to be lanthanum containing carbon species, it is more or less loosely spread over the surface and can by no way prevent “physical contact between the carbon of the furnace and the subsequent sample” as proposed for example by RUNNELS et al. [7]. Clearly visible, however, is the pronounced separation of the pyrolytic graphite layers which is almost certainly caused by the intercalation of lanthanum. It is not unlikely that the strong lanthanum signal found upon ED XRF inspection originates at least in part from these areas, i.e. from intercalation compounds of lanthanum. The light flakes are most probably disintegrated pyrolytic graphite which may, of course, have lanthanum bound to its surface. The disintegration of the graphite at its surface may at least in part be due to the gasification reaction between oxygen and graphite which is known to be catalyzed by many metals. Figure 14 shows an additional view of the fracture pattern of the sampling cavity, exhibiting the thickness of the intercalation layer which is between 15 pm and 35 pm, The

exhibit

substantial

tailing

1188

BERNHARDWEU et

al.

Fig. 11. Injection hole area of graphite tube used for endurance test with lanthanum. (a) tube inside and injection hole area, (b) corroded and uncorroded graphite nodules in the injection hole area.

micrograph also shows the distinct difference in the morphology between the corroded intercalation layer in the upper third of Fig. 14 and the unaffected base material that appears at the bottom of the micrograph. Figure 15 finally shows the bottom of the “empty” cavity of the dual cavity platform into which no solution at all was pipetted throughout the experiment. The cavity nevertheless exhibits a corrosion having some similarity with that shown in Fig. 3(a) which could be defined as “moderate lanthanum corrosion”. This proposes that any corrosion found in the phosphorus cavity, in the experiments where phosphorus and lanthanum solutions were pipetted into separate cavities, is not due to the phosphorus solution but caused by an attack of lanthanum via the gas phase. If, however, lanthanum is deposited in the other cavity, and

Lanthanum modifier

1189

Fig. 12. Dual cavity platform after 11 applications of 2 % lanthanum solution and a total of 17 atomization cycles. (a) “sampling” cavity (right) and “empty” cavity (left), (b) deliberately broken sampling cavity.

accumulated with time, formation of a thermally stable compound between phosphorus and lanthanum is the most likely mechanism of stabilization. In addition to corrosion and pitting, the micrograph of Fig. 15 also shows some blisters similar to those in Fig. 9. As any reaction and attack in this cavity must proceed via the gas phase, the most likely explanation for their formation is by decomposition of a lanthanumcarbon species upon reaction with water. 3.5 Lanthanum memory efect The assumption that a dense coating with a layer of lanthanum carbide may be the reason for the enhanced sensitivity was supported by the fact that already a single application of high

1190

BERNHARDWEU et al.

Fig. 13. Bottom of sampling cavity. (a) disintegrated material and separation of layers, (b) separated pyrolytic graphite layers, detail of (a).

concentration of lanthanum or soaking the tube with a lanthanum solution has a long lasting effect [4,6,7]. We have also found for lanthanumconcentrations higher than 0.2 y0 (m/v) that after a few additions of this matrix modifier solution the phosphorus signal was altered permanently in the subsequent determination even when no more lanthanum was added [ZO]. It was shown in the previous experiments that there is no indication ofa lanthanum carbide coating in the form of a dense layer, but that there is nevertheless some lanthanum left on the graphite surface after its repeated application. The memory effect can therefore be caused by the residual lanthanum and/or by the changes in graphite morphology, i.e. the increased number of active carbon sites. In order to discriminate between these two possibilities, 20 determinations of phosphorus were carried out on a platform with the addition of 20 pg

Lanthanummodifier

1191

Fig. 14. Fracture pattern of sampling cavity. Thickness of intercalation layer cu. 20 pm.

Fig. 15. Surface of empty cavity bottom with blister formation.

lanthanum (10 ~1 of a 0.2 % solution) in each determination. After this, the next 40 determinations of phosphorus were carried out without the addition of lanthanum to see the memory effect. In a second experiment the platform was taken out of the tube after 20 determinations of phosphorus in the presence 20 pg lanthanum and inserted in a new, unused tube for the next 40 determinations without the addition of lanthanum. Finally, in a third experiment, the platform was again taken out of the tube after 20 determinations of phosphorus in the presence of lanthanum, and immersed into 200 ml of deionized water. The

1192

BERNHARD WEU

et al.

platform was left in water at room temperature for three days, and finally heated to nearboiling for one hour to dissolve and decompose lanthanum carbide from its surface if it had formed. After this treatment the platform was dried and inserted into an unused tube for the additional 40 phosphorus determinations without lanthanum as in the previous experiments. The integrated absorbance values recorded in these three experiments are shown in Fig. 16, and, at first glance, there is surprisingly little difference between the three sets of data. In any case the phosphorus signal drops asymptotically over the next 30 determinations after the last application of lanthanum to an integrated absorbance value which is about one tenth of the original signal in the presence of lanthanum. The remaining signal is, however, clearly above the blank, and it is also higher than the signal for phosphorus in the absence of lanthanum under otherwise identical conditions, i.e. atomization from a platform in a pyrolytic graphite coated tube. In all these experiments the platforms were slightly bent after 20 applications of 20 pg lanthanum each. This bending seems to be essential for the observed memory effect because a much earlier return of the phosphorus signal to the baseline was found with only slightly lower lanthanum concentrations which did not cause visible bending of the platform. Upon more careful investigation of the results depicted in Fig. 16, it becomes obvious that the integrated absorbance values obtained in the first experiment, where the platform remained in the original tube, are clearly higher over some 30 determinations after the last application of lanthanum compared with those where the platform was removed from its tube and inserted into an unused one. In addition, this difference remains fairly constant within a factor of two over nearly 30 determinations. In the two experiments where the platform was taken out of the original tube and inserted into a new one, the integrated absorbance signals are essentially identical within experimental error after about 10 determinations without the addition of lanthanum. The platform which was left in water for three days to dissolve lanthanum carbide, in case it was formed, gives a somewhat lower integrated absorbance signal only over a few determinations. There is certainly no big difference between this platform and the one which was simply taken out of one tube and inserted into a new one. This could mean that either it is not the lanthanum but the change in surface morphology which causes the sensitivity increase for phosphorus or the lanthanum is not removed very effectively from the platform by the washing procedure applied in this experiment. It was expected that SEM inspection could help to distinguish between these two possible explanations. Figure 17 shows the bottom of the platform cavity from the third experiment where the platform was taken out of the tube after 20 applications of lanthanum, immersed into water, and used for an additional 40 determinations of phosphorus without the addition of lanthanum. The morphology is very much the same as that depicted in Fig. 3a, which was 31

-

i

lil

wthout La

ti

ti

M

60

No of Determmot~on

Fig. 16. Integrated absorbance signal for 0.2 yg P atomized from a L’vov Platform in a pyrolytic graphite coated tube. 20 determinations with the addition of 2Opg lanthanum, each and no lanthanum added over the next 40 determinations. (0) same tube and platform over entire experiment; ( + ) platform inserted into new tube after 20 determinations; (0) platform soaked in water for three days after 20 determinations, then inserted into new tube.

Lanthanum modifier

1193

Fig. 17. Surface of platform cavity bottom after 20 applications of 0.2% lanthanum, three days soaking in water, and 40 additional determinations of phosphorus without lanthanum addition.

taken after 15 applications of a 0.5% lanthanum solution. This means that neither the soaking with water nor the 40 additional atomization cycles had any major visible influence on the platform morphology. It should be noted in particular that there is no sign of a secondary coating of the platform cavity even after 40 atomization cycles in the absence of lanthanum. Changes in the morphology cannot be made responsible therefore for the drop in the integrated absorbance of phosphorus by one order of magnitude. There may be, however, an annealing effect during the 40 high temperature cycles which reduces the number of active carbon sites where a stable lanthanum-phosphorus-carbon compound could form. Such details at the atomic level can obviously not be detected by SEM inspection. Upon ED XRF inspection the platform cavity of Fig. 17 gave a clear signal for lanthanum. This means that there is still some lanthanum left on the platform, even after the soaking in water and the 40 determinations of phosphorus without the addition of lanthanum. This suggests that at least part of the lanthanum is bound very firmly to the graphite, most likely in the form of an intercalation compound, which is volatilized only very slowly. The severe pitting and corrosion, which creates a lot of active carbon sites and opens access to underlying graphite layers, makes such an intercalation quite probable. The memory effect shown in Fig. 16 is therefore most likely caused by residual lanthanum retained by the graphite in the form of intercalation compounds, and released only slowly and incompletely [27]. The observation that there is a direct correlation between the bending of the platform-aused by intercalation-and the memory effect, is another indication that the residual lanthanum retained in the graphite lattice plays an important role in the stabilization of phosphorus. The residual lanthanum is most probably also responsible for the absence of any secondary coating on the platform. The platforms of the other two experiments showed essentially the same morphology and are therefore not depicted here. This is another proof that the soaking procedure has no influence on the platform morphology. Figure 18, finally, shows the inner tube surface near the injection hole of the tube from which the platform was removed after 20 atomization cycles with 0.2 % lanthanum solution [Fig. 18(a)], and of the tube in which the water soaked platform was re-inserted and used for 40 additional determinations of phosphorus without lanthanum, respectively [Fig. 18(b)]. The former exhibits substantial corrosion showing some similarity to that depicted in Fig. 11, which is after about twice as many applications of 0.5 ‘A lanthanum solution. 1271 G. R. HENNIG,Progress in Inorganic Chemistry, Vol. 1, F. A. COITON, Ed., p. 125. Wiley-Interscience, New York (1959).

1194

BERNHARDWELZ

et al.

Fig. 18. Inner tube surface near injection hole. (a) after 20 applications of 0.2 % lanthanum, (b) after 40 determinations of phosphorus on water soaked platform from Fig. 17 without addition of lanthanum.

As this corrosion is almost certainly caused by a vapor phase attack of lanthanum it can be anticipated that some lanthanum is also retained on the graphite tube and not only on the platform. The difference in the integrated absorbance signal for phosphorus depicted in Fig. 16 for the platform in the original tube and the platform inserted into an unused tube may well be due to the residual lanthanum at the tube surface, which contributes to the total amount of lanthanum left in the furnace. The new tube, in which the platform soaked in water was inserted [Fig. 18(b)], also shows moderate corrosion after the 40 determinations of phosphorus, which is certainly not simply due to the thermal charge of the 40 atomization cycles (compare [19]). The most likely explanation is that lanthanum, which is volatilized from the platform in every atomization step, is readily adsorbed by the yet lanthanum-free tube until a certain surface coverage is

Lanthanum modifier

1195

obtained. This adsorbed lanthanum then causes the corrosion seen in Fig. 18(b), which is obviously less severe because of the smaller, amount of lanthanum involved. 3.6. Corrosion test with phosphorus On several occasions during this work it was concluded from observations that corrosion on spots that have not been in contact with lanthanum in the condensed phase was due to an attack of the modifier via the gas phase. It was also assumed that the absence of a secondary coating on the platform and essentially also on the tube could primarily be ascribed to reactions of lanthanum. To substantiate these conclusions and assumptions, a few additional experiments were carried out to investigate the influence of phosphorus alone on graphite surfaces. In a first experiment 0.2 pg phosphorus as a solution of the dibasic ammonium phosphate in dilute nitric acid were pipetted onto the platform in a pyrolytic graphite coated tube. After 410 determinations using the temperature program given in Table 1, tube and platform were inspected with scanning electron microscopy. The visible appearance of the tube was still good by that time and the phosphorus signal had not changed significantly. Figure 19 shows morphological details of the platform and the inner tube wall. The bottom of the platform cavity [Fig. 19(a)] exhibits only corrosion and no sign of any secondary coating. The corrosion of the platform surface is, however, not as severe as that shown in Figs 10 and 17. This is particularly true when it is taken into account that in these latter experiments the platform was inspected after about 40 heating cycles which is only one tenth of those in the current experiment. Figure 19(b) shows that there is substantial corrosion at the inner tube wall where the pyrolytic graphite coating is partly removed so that the underlaying polycrystalline graphite becomes visible. At the same time, however, a substantial secondary pyrolytic graphite coating with the formation of graphite nodules becomes apparent at the upper and outer part of the platform. This phenomenon has not been found in the experiments with lanthanum added to the phosphorus solution. A pronounced formation of graphite nodules was also found around the injection hole area (now shown here) which confirms that lanthanum is actually the inhibitor of the nodule formation. To further investigate the corrosion pattern caused by phosphorus compounds we increased the analyte mass to 1 pg phosphorus. This has the additional advantage of compensating for the substantially lower sensitivity for phosphorus obtained in the absence of a modifier [20]. The tube failed after 349 determinations. In addition, the sensitivity for phosphorus started to increase slowly after about 200 determinations and reached almost a ten times higher value towards the end of the tube life. The highest values obtained were close to those which are typically found for uncoated tubes only, which give the best sensitivity for phosphorus when no modifier is used [ZO]. The tube from this endurance experiment is depicted from its outside in Fig. 20, which shows deformation and substantial secondary coating all around its central part. Particularly revealing is the corrosion pattern which becomes apparent when the tube is cut (Fig. 21). Around its central part the graphite of the tube is severely or totally disintegrated, most probably by volatilization of predominantly binder graphite [19]. The disintegration is, however, more severe than could be expected for the thermal charge only. Phosphorus suboxides can react with graphite to form compounds that decompose again at high temperatures, and thus contribute actively to graphite disintegration [20]. It is interesting to note that the shape of the tube is largely maintained inspite of the corrosion and that the actually broken tube is held together by disintegrated material. Figure 21 also shows a clear connection between graphite disintegration and secondary pyrolytic graphite coating. This secondary coating and nodule formation are most pronounced at the platform directly opposite to the most disintegrated tube area. No secondary coating but rather pronounced corrosion and pitting can be seen at the bottom of the platform cavity. This is in agreement with the previous experiment and also confirms the direct corrosive attack of phosphorus according to the above-mentioned mechanism. A substantially better sensitivity for phosphorus is obtained in an uncoated polycrystalline electrographite tube. In another experiment a solution containing 1 /rg phosphorus was

1196

BERNHARDWELZ et al.

Fig. 19. Morphological changes at tube and platform surface after 410 atomization cycles with 0.2 pg P as the dibasic ammonium phosphate. (a) corrosion at the platform cavity bottom, (b) corrosion of the inner tube wall (upper right part) and secondary deposition of pyrolytic graphite on platform rim (lower left part).

pipetted directly onto the wall of an uncoated tube, which eventually failed after 272 determinations. It is interesting that the sensitivity for phosphorus did not change by more than 20 % over the entire lifetime of the tube [20]. After its failure, the tube was very brittle and showed several cracks at its surface as depicted in Fig. 22(a). Upon closer inspection it becomes apparent that the tube is fairly well covered with a secondary layer of pyrolytic graphite, and the original polycrystalline morphology becomes vilible only in the depth of the cracks [Fig. 22(b)]. The similarity between the tubes from the two experiments becomes even more obvious from Fig. 23 which shows the areas around the injection hole of the two tubes. There is no

Lanthanum modifier

1197

Fig. 20. Pyrolytic graphite coated tube after 349 atomization cycles with 1 pg P. (a) total view, (b) injection hole area with secondary deposition of graphite and a crack.

apparent difference in the morphology of the originally uncoated tube [Fig. 23(a)] and the one which was coated with pyrolytic graphite [Fig. 23(b)]. Both tubes are equally disintegrated and have a very similar secondary coating of nodular pyrolytic graphite. It is not at all surprising that both tubes exhibit essentially the same sensitivity for phosphorus at that stage close to the end of their lifetime [20], which again demonstrates the close relation between tube morphology and the sensitivity obtained for the analyte element. These experiments have shown that phosphorus itself contributes to the corrosion of graphite tubes and platforms. This is the first time that we have found corrosion due to the analyte element itself. It has to be taken into account, however, that usually much smaller analyte masses are applied, and that the attack of phosphorus on the graphite is clearly massdependent. There is no doubt that the corrosion due to phosphorus is significantly different from that due to lanthanum with respect to its degree as well as to its morphological pattern.

1198

BERNHARDWELZ et al.

Fig. 21. Same tube as in Fig. 20, cut into halves. (a) center part of tube with inserted platform, (b) cross-section through disintegrated tube wall and nodular pyrolytic graphite deposits at outer platform wall and platform rim.

4. CONCLUSIONS

In our investigations we have found no indication for the formation of a dense layer of lanthanum carbide which would preclude contact of the analyte element with graphite proposed by several authors [4,6,7]. Thisdoes not mean that lanthanum carbide may not E formed during thermal pretreatment, but this carbide is known to decompose upon cant act with water so that the formation of a dense layer is not observed. More likely than Ithe formation of “lanthanum carbide” is that of a surface-stabilized lanthanum-carbon species which may not have many of the chemical similarities to lanthanum carbide as a btulk compound.

Lanthanum modifier

1199

Fig. 22. Uncoated polycrystalline electrographite tube after 272 atomization cycle-s with 1 pg P. (a) tube outside with cracks, (b) detail of crack depicting secondary layer of nodular pyrolytic graphite on top of the polycrystalline graphite.

The mechanism of catalytic graphitization proposed by STURGEONand CHAKRABARTI [S] is not very likely either. They assume that during the thermal pretreatment step a carbide is formed which penetrates the pores of the graphite surface, “dissolving the disordered carbon and precipitating it as nonreactive, small graphite crystallites”. They then believe that the improved sensitivity obtained for a number of elements using carbide-treated tubes may be the result of a more highly ordered graphite surface. We agree with the formation of acarbide which penetrates into the pores and preferentially reacts with disordered carbon. The morphological studies of this investigation give no indication, however, that this leads to a more highly ordered graphite surface. The opposite appears to be the case, i.e. the carbide penetrating into the pores leads to massive corrosion and layer swelling and creates more and more reactive carbon sites when it decomposes again, in part, upon contact with water.

1200

BERNHARDWELZ~~

al.

Fig. 23. Injection hole areas of the tubes from Figs 21 and 22 from the outside with nodular pyrolytic graphite deposits. (a) previously uncoated polycrystalline electrographite tube, (b) originally pyrolytic graphite coated tube.

The effect of a single application of a high lanthanum concentration, or the memory effect shown in this work for a 0.2 % lanthanum solution, clearly demonstrates that lanthanum is retained on or in the graphite for a long time and over many atomization cycles. Morphological studies have shown that this is most probably in the form of intercalation compounds. Such compounds are much more resistant against chemical attack, as could be shown in the soaking experiment with water which had little effect on the lanthanum concentration on the platform. The lanthanum is most probably not only retained by the surface on which it is pipetted as the solution, i.e. the platform, but also by the surface which it comes into contact with after its volatilization, i.e. the entire graphite tube. While it appears to be obvious that it is the lanthanum itself which leads to the increased

Lanthanummodifier

1201

sensitivity, and not a more highly ordered graphite surface-or a surface sealing with a carbide layer, there are still several possible reaction mechanisms. EKLUNDand HOLCOMBE [9] and L’vov and RYABCHUK[NJ proposed a preferential reaction of lanthanum with oxygen in the

gas phase and thus a purification of the purge gas from the O2 impurity. This cannot be excluded as one of the reactions that may occur in the graphite tube furnace. It is unlikely, however, that this is the only or the predominant mechanism. This “purification” mechanism can only be effective at relatively high temperatures when lanthanum is atomized and can influence gas phase equilibria. The main reason for the low sensitivity of phosphorus in the absence of a modifier was, however, shown to be pre-atomization losses in the form of PO, [16], which cannot be avoided by the proposed mechanism. Stabilization of the oxidizing centers on graphite by lanthanum, and in consequence of the created interlamellar compounds, was proposed by KORECKOVA et al. [lo] for arsenic. This appears to be one possibility since we could show the influence of lanthanum on the graphite morphology which makes the formation of interlamellar (intercalation) compounds very likely. This mechanism cannot explain, however, the increase in sensitivity found for phosphorus in the first atomization from the dual cavity platform. We therefore believe that the predominant reaction of lanthanum is the formation of thermally stable phosphorus compounds which avoid low-temperature losses and lead to vaporization of phosphorus at high enough temperatures. This compound formation is apparently supported by the changes in graphite surface morphology that are caused by the attack of high lanthanum concentrations. These morphological changes contribute to retain lanthanum in the furnace tube by facilitating the formation of intercalation compounds. The larger number of active carbon sites created by the attack of lanthanum at the graphite surface may also contribute directly to the stabilization of the analyte element and hence to the sensitivity increase. This effect does not contribute very much to the sensitivity increase for phosphorus according to our experience, it may, however, be of greater importance for other elements. In the experiments with the dual cavity platform the stabilization of phosphorus is predominantly via heterogeneous reactions. In the first determination, volatilized phosphorus oxide reacts with lanthanum in the condensed phase and forms a stable compound. In the subsequent determinations, more and more lanthanum is volatilized and condenses on previously lanthanum-free surfaces. In this way, the previously lanthanum-free cavity, after a while, contains enough of this modifier to guarantee stabilization of phosphorus. In the absence of lanthanum as the modifier, phosphorus itself also causes corrosion at the graphite tube and platform surface. This corrosion is, however, different from that observed for lanthanum. Phosphorus prevents the formation of a secondary graphite coating on the surface which the analyte is in contact with in the condensed phase, for example on the platform cavity bottom. It apparently supports, however, the formation of nodular pyrolytic graphite on other parts of tube and platform by the formation of compounds with graphite which decompose at high temperatures. Lanthanum, in contrary, inhibits the formation of nodular pyrolytic graphite layers completely, so that only corrosion is observed in the presence of high lanthanum concentration. Acknowledgement-This research was supported in part by the Conselho National de Desenvolvimento Cientifico e Tecnol6gico (CNPq), Brasilia, and by the Forschungsfiirderungsfonds fiir die gewerbliche Wirtschaft, Vienna.