Anal. Chem. 1991, 63, 2352-2357
2952
like iron(II), cobalt(II), and copper(I1) have been reported to cause CL from luminol even in the absence of hydrogen peroxide (31). Thus interference from these metal ions is reckoned in the determination of gold using the luminol CL reaction in conventional aqueous solutions. At the moment, work is in progress on the solvent extraction of gold and its subsequent detection via CL reaction. The effects, if any, of foreign ions will also be included in our future investigation. In conclusion, this work suggests the possibility of developing a potentially effective hybrid method through coupling of a CL detection system with solvent extraction for the selective determination of gold a t trace levels.
LITERATURE CITED Fujiwara, T.; Kumamaru, T. Spectrochlm. Acta Rev. 1990, 13, 399. TOwnS6f1d, A. AMlyst 1090, 115, 495. Niemann, T. A. In Chemllumlnescence and photochemical Reaction Defection h Chromatography; Birks, J. W., Ed.; VCH: Weinheim, 1989; pp 99-123. Delumyea, R.; Hartkopf, A. V. Anal. Chem. 1978, 48, 1402. Montano, L. A.; Ingle, J. D., Jr. Anal. Chem. 1979, 51, 928. Boyle, E. A.; Handy, B.; van Geen, A. Anal. Chem. 1987, 5 9 , 1499. Fuliwara. T.: Tanimoto. N.; Huana. J.J.; Kumamaru. T. Anal. Cbem. 1080, 61, 2800. FuJiwara,T.; Kumamaru, T. Proceedings of International Trace Analysis Symposium 90 held at Sendai and Kiryu, Japan, July, 1990; pp 187-190. -. k g e s , J. Spectrochlm. Acta Rev. 1990, 13, 27. Kumamaru, T.; Okamoto, Y.; Yamamoto, M.; Obata, Y.; Onizuka, K. Anal. Chlm. Acta 1990, 232, 389. Okamoto, Y.; Takahashi, T.; Isobe, K.; Kumamaru, T. Anal. Scl. 1990, 6 , 401. Hoshino, H.; Hinze, W. L. Anal. Chem. 1987, 59, 496. Igarashi, S.; Hinze, W. L. AM/. Chem. 1988, 60, 446. Igareshi, S.; Hinze, W. L. Anal. Chlm. Acta 1989, 225, 147.
(15) See, e.g.: Pileni, M. P. In Structure and ReacHvity In Reverse Micelles; Pileni. M. P., Ed.; Elsevler: Amsterdam, 1989 (see also references cited therein). (16) FuJiwara,T.; Tanimoto. N.; Nakahara, K.; Kumamaru, T. Chem. Lett. 1991. 1137. (17) Lukovskaya, N. M.; Terietskaya, A. V.; Bogosiovskaya, T. A. Zh. Anal. Kblm. 1974, 29, 2268. (18) Nemcova, I.; Rychlovsky, P.; Kleszczewska, E. Talanta 1990, 37, 855
(19) &ri, A. R.; Huerta, E.; de ia Guardia, M. Fresenlus J . Anal. Chem. 1990, 338, 699. (20) Kamel. H.; Brown, D. H.; Ottaway, J. M.;Smith, W. E. Analyst 1978, 70 7 . 790. (21) Kalimann, S. Talanta 1987, 34, 677. (22) Wood, S. A.; Viassopouios, D.; Muccl, A. Anal. Cblm. ActB 1990, 229, 227. (23) Rockiin, R. D. Anal. Cbem. 1984, 56, 1959. (24) Marcus, Y.; Kertes, A. S. Ion Exchange and Solvent Exfraction of Metal Complexes; Wiley & Sons: New York. 1969; pp 694-736. (25) Mason, W. R., 111; Gray, H. B. I m g . Chem. 1988, 7 , 55. (26) Fernandez-Gutienez, A.; Munor de la Pena, A. In Mo/ecubr Lumlnescence Spectroscopy. Methods and Applications: Part 1 ; Schuiman, S. G. Ed.; Wiiey & Sons: New York, 1985; pp 463-546 (see also references cited therein). (27) Day, R. A.; Robinson, B. H.; Clarke, J. H. R.; Doherty, J. V. J . Chem. Soc., Faraday Trans. 11979, 75, 132. (28) Lang, J.; Mascolo, G.; Zana. R.; Luisi, P. L. J . Phys. Chem. 1900, 94. 3069. (29) Sunamoto, J.; Kondo, H.; Hamada, T.; Yamamoto, S.; Matsuda, Y.; Murakami, Y. Inwg. Chem. 1980, 19, 3668. (30) Robinson, B. H.; Steytler, D. C. J . Cbem. SOC.,Faraday Trans. 1 1979, 75, 481. (31) Kiopf, L. L.; Nieman, T. A. Anal. Chem. 1983, 55, 1080.
RECEIVED for review May 8, 1991. Accepted July 19, 1991. This work was partially supported by a Grant-in-Aid for Scientific Research, No. 02453034,from the Ministry of Education, Science, and Culture of Japan.
Counting Metal Oxide Monolayers on a Metal Surface Using Plasma Desorption Mass Spectrometry Richard S. Juvet, Jr,,*JGunter Allmaier, and E. R. Schmid Znstitut fur Analytische Chemie der Universitat, Wahringer Strasse 38, A-1090 Vienna, Austria
Several serles of alumlnum oxMe clwter Ions produced from a thin fllm of alumlnum on polyester have been Identlfkd by plasma derorptlon mass spectrometry. Major porltlve-Ion serles Include (AI,O,),-AIOH+, (AI,O,), -H+, and (AI,O,), A1202+ wHh shorter serles of (AI,O,), *H,O+ and (AI,O,), -AI(OH),’; while major negative-ion serles Include (A120a), AIO( OH)-, (AI,O,),~AIO,~AI( OH),-, and (AI,O,), *AIO(OH),wHh shorter serks of (AI,O,),.AK)and (A120s),-. Values of n as large as 22 were observed for some of the alumlnum fllms studled. A graphlcal method Involving quantltatlve measurements allows an accurate evaluation of the number of oxldo “layers on the metal surface. Oxlde monolayers ranged from 15 to 24 depedng on the age of the wurlace and on L chemkal treatment. A reproduclblllty of flmonolayer Is obtained wHh slmllar results In both the poslve- and negative-Ion modes. The values measured agree closely wlth estlmates of oxlde thlckness determlned using X-ray photoelectron spectroscopy. A derorptlon mechanism based on quantltatlve experlmental measurements Is proposed.
.
-
‘Present address: Arizona State University, Department of Chemistry, Tempe, A2 85287-1604. 0003-2700/9110363-2352$02.50/0
Since the original research on 262Cf plasma desorption mass spectrometry (PDMS) by Macfarlane and Torgerson ( 1 , 21, the method of sample introduction has remained basically the same. Ionization and desorption take place on the surface of a thin foil by the passage of nuclear fission fragments from a 252Cfsource through the foil. The desorbed ions are accelerated by an electric field and separated according to their masses in a time-of-flight mass spectrometer. The foil most frequently used consists of an aluminized film of polyester. Nitrocellulose electrosprayed on the aluminum target was found by Jonsson et al. (3)to greatly enhance the intensity of the molecular ion, insulin, presumably because of desirable adsorption/desorption properties for proteins, and this modified target has since been used extensively in the study of peptides by PDMS. Since most studies to date using PDMS have concerned the determination of the molecular ions of high molecular weight peptides and sugars, background peaks found below m / z 10oO have been largely ignored. The origin of these peaks and the possibility of their elimination becomes important, however, if PDMS is to be expanded to the study of lower molecular weight compounds. Background interference has been traced to three sources: (1)to several series of aluminum oxide cluster 0 1991 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 63,NO. 20, OCTOBER 15, 1991 2353
ion peaks from the oxidized surface of the aluminum target (observed whenever the target is not completely covered with nitrocellulose or the sample); (2) to the vacuum pump oil molecular ion, M+, and its fragmentation peaks; (3) to fragmentation peaks from the nitrocellulose matrix. A detailed study of these interferences and a suggested new target material is given elsewhere (4). Others ( 5 , 6 )have noted interferences from aluminum oxide fragment peaks, and Voit and co-workers (7)have proposed structures for some of the series. In this paper a detailed study of several aluminum oxide cluster species is made and the application of 252Cf-PDMSfor estimating the number of monolayers of metal oxide on a metal surface is reported. These results have been confirmed by X-ray photoelectron spectroscopy. Accurate quantitative measurements by PDMS and the proposal of a desorption mechanism are also important parts of this publication. EXPERIMENTAL SECTION The plasma desorption mass spectrometer used in these studies was a BioqIon Model 20K PDMS with a 10-pCi 252Cfsource (BicrIon Nordic AB, Uppsala, Sweden). The accelerating potential was 17 kV in the positive-ion mode and -15 kV in the negative-ion mode. The length of the flight tube is 15 cm, and the time resolution selected was 1 ns/channel. Mass calibration was generally based on H+and Na+ in the positive-ion mode and on H- and CN- in the negative-ion mode in the recommended manner. For two cases, however, no trace amounts of Na+ could be detected, in which case calibration was based on H+and the pump oil M+ peak at 408.72 Da. Additional details of the instrument and handling procedures have been published earlier (8). Background spectra were obtained on aluminized polyester targets that were either the commercially available ones supplied through BicrIon Nordic, stored covered in air at room temperature and used ca. 10 months after delivery, or were prepared in our laboratory from a sheet of polyester, 2 pm in thickness, coated on both sides with aluminum, 1 pm in thickness (Type B(14), Alexander Vacuum Research, Greenfield, MA), and affixed to brass ring supports, 13 mm in diameter (Bio-Ion Nordic), with epoxy glue. A total of 1OOO OW50 OOOOOO primary events were used in obtaining the spectra reported, and both positive and negative charged ions were examined. Quantitative determination involved computer evaluation of peak a r e a in total ion counts with the peak base measured from the mean of baseline noise fluctuations. Since peak area is directly proportional to primary event start counts, for comparison purposes areas have been normalized to 1OOOOOO primary events in most data reported. X-ray photoelectron spectroscopicestimates of the aluminum oxide thickness were performed by using a Kratos (Manchester, England) Model XSAM 800 unit with Mg Ka radiation. RESULTS AND DISCUSSION Spectra obtained by using both commercial aluminum PDMS targets, exposed to air for at least a 10-month period, and aluminum targets more recently prepared in our laboratory showed, in addition to vacuum pump oil peaks with m/z generally less than ca. 200 Da, peaks with masses as large as 2295 Da in the positive-ion mode and 2303 Da in the negative-ion mode. The spectra for both the positive- and negative-charged species (Figure 1)show an obvious regularity with an average difference in corresponding m/z values in each series equal to 101.92 for all positive-charged ions and 101.83 for all negative-charged ions, in excellent agreement with the value 101.96 for the molecular weight of A1203. For the positive-charged peaks three major series may be discerned corresponding to the series (Al2O3),,.A1OH+, n = 0-22, n = 0-21. The (A1203),.H+,n = 0-22, and (A1203),,.A1202+, upper limit for each series increases with the age of the exposed aluminum surface. The upper limits stated were observed by using a large number of start counts, 50000000, and the older, commercially made aluminum target. The upper limits detectable with 2000000 start counts on the same
A 300
'
'
600'
'
900
'
'
1200
'
'
15007-
'
16W
'
16W '
I
MIZ 245481
B '
'sbo' ,960'
'
1200
'
-
15W '
MI2
oxide cluster peaks from an aluminum surface exposed to air for at least a 10-month period. 252Cf-PDMSspectra with 50 000 000 start counts are shown: (A) positive-ion mode; (B) negative-ion mode. Flgure 1. Aluminum
sample was approximately three A1203units lower for each series owing to the decreased signal-to-noise level. The aluminum target prepared in our laboratory and exposed to air for a shorter period of time showed upper limits a significant seven to nine A1203 units less than the maximum limits given above with both 1000 000 and 4 000 000 start counts. When the laboratory-prepared target was acid-washed with a 0.1% trifluoroacetic acid solution, rinsed thoroughly with water, and spin-dried, the upper limit of each series detectable was reduced an additional two to four units in both the positiveand negative-ion modes. In addition to these three major aluminum oxide series, two other shorter series were identified in the positive-ion mode. The relatively smaller peak areas for the members of these series suggest that they are less stable or are less abundant than the major series. Found were (A1203),.H@+, n = 0-3, and (A1203),.A1(OH)2+,n = 0-4. For the negative-charged species the three major series (A1203),.A10(OH)-, n = 0-22, (A1203),~A102~A1(OH)2(or (A1203)n+l~H20-), n = 0-17, and (Al,O,),AlO(OH),-, n = 0-11, were identified, again by using the older commercially made aluminum target and 50 000 000 start counts. I t should be noted that the maximum value of n found is identical in both the positive- and negative-ion modes for this target. In addition to these three major aluminum oxide anion series, two additional series, in some cases barely discernible above baseline, were found: (A1203),-, n = 1-9, and (A1203),.A10-, n = 0-9. The identity of the various species listed is based solely on the apparent m/z value of each member of the series. A summary of calculated and observed peak maases is given in Table I, and deviations are discussed below in detail.
ANALYTICAL CHEMISTRY, VOL. 63, NO. 20, OCTOBER 15, 1991
2954 ~
~~
Table I. Agreement between Observed and Calculated Peak Masses for Various Aluminum Oxide Cluster Ions" species
av mass error, Da +I.%
+0.29 -0.17
-0.06
+0.12
species
av mass error, Da
(AI~O~),,~AlO(OH)- -0.67 (n = 0-17:-0.07) (n = 18-22-2.04) (A1203),.A102.N(OH)2- +0.81 (n = 0-10+0.06) (n = 11-17:+1.83) (A1203)n.A10(OH)~ +3.23 (n = 0-2:+0.22) (n = 3-6:+2.13) (n = 7-11:+5.84) (A1203),*A10-2.33 (n = 0-9) (Al2O3In-1.71 (n = 1-9)
2MO
-
E
1800
2 2a 9
1400
Q
t
1000
$ 600
Observed masses are the average of five determinations. 200
As was the case for the positive-charged ions, the maximum value of n observed was less in each series for the newer laboratory-made targets and was further reduced by washing with 0.1% trifluoroacetic acid solution. It appears from these preliminary experiments that the upper values observed are a function of the degree of oxidation of the aluminum surface and to some extent of the number of start counts used in obtaining the data, since this influences the signallnoise ratio. Moreover, the amount of oxidized surface can be reduced by acid washing. It would be tempting a t this point, though incorrect, to assume that one could approximate the number of monolayers of oxide coating from the mass of the highest observed peak. However, since peaks with the largest m / z are barely discernible above baseline noise and this estimate could therefore deviate from observer to observer, such a method for counting monolayers of surface oxide would be considered only approximate at best. Moreover, the mechanism of desorption should have an influence on the validity of the results. One might picture a desorption mechanism such as the energy of a fission particle from the 252Cfsource, bulletlike, knocking out a cylinder, or perhaps a cone or crater, of surface oxide as it penetrates the target, causing ionization. The charged cylinder or cone may fragment to the lower molecular weight particles observed, but some charged cylinders or cones remain intact and produce an upper limit in mass. If the charged cluster projectile of surface oxide is cylindrical in shape, then the upper limit in n should correspond exactly to the number of monolayers of oxide coating present on the sample surface, provided sufficient ions of that mass are present to detect above instrument noise level. If the cluster projectile is a crater and more conical in shape, then the value of n observed becomes only a maximum limit for the number of monolayers, the true value being smaller by an amount dependent upon the shape of the crater produced. Knowledge of the shape of the projectile therefore is essential if an accurate estimate of the number of monolayers in the oxide coating is to be determined. Fortunately, in the case of aluminum oxide monolayers it may be possible to reach a conclusion concerning the desorption mechanism on the basis of the experimental data obtained in the course of this study. Proposal for t h e Desorption Mechanism. For each of the several series of anionic and cationic clusters, if a plot is made of the particle ion peak area vs. n, a smooth curve is produced throughout most of the plot, with area decreasing as n increases. The smoothness suggests that no "magic numbers" exist, as has been found for cadmium iodide and several metal clusters by using the SIMS technique (9, IO), at least within the 22 units of the present sample observable. Moreover, the shape of these curves suggested that there might
02
0.4
06
0.8
1.L
06
08
11
l/n
2600
22770 \
2
5 9s 2
1.900
1400
8 %
600
200
02
04 1/n
Flgwe 2. 25%f-PDMSpeak area vs. reciprocal of the number of A1203 clusters in various fragment ions from an aluminum surface exposed to air at room temperature for at least 10 months (50 000 000 start counts, area counts normalized to 1 000 000 start counts): (A) positive-ion mode; (e) negatlvslon mode.
be a simple reciprocal relationship between peak area (proportional to the amount of ion fragment produced) and the value of n. This is confirmed in Figure 2A,B for the higher members of the series for one ion in both positive- and negative-ion modes. In both positive- and negative-ion modes it is the predominant species with the longest series that shows a linear relationship. The other, shorter series produce curvilinear relationships in both modes. Nonlinearity of the two predominant series also occurs for the first two or three members of both series. The deviation from nonlinearity may be an indication of preferred cleavage or of preferred charge stabilization-perhaps an indication of a change in projectile shape (e.g., a cylindrical projectile becoming more conical at the surface). The extrapolated intercept with the x axis, however, for all curves seems to coincide within experimental error. The intercept with the x axis corresponds to the value of n at which cluster peaks no longer exist (Le., a peak area of zero counts). The linear and nonlinear behavior of these curves is undoubtedly significant, and an interpretation of the desorption
ANALYTICAL CHEMISTRY, VOL. 63, NO. 20, OCTOBER 15, 1991
A d
E P.
-4
d
/K
Possible
Units No. -
m o m
6 5 4 3 2 1
1 2 3 4 5 6
C Flgure 3. (A) Schematic diagram of the proposed desorption mechanism showing cylindrical projectile and crater formation, fractionation and protonation, and the inverse relationship between the peak area and nwnber of cluster units. Reasonable aluminum oxide configurations leading to (8) cylindrical cluster species and (C) noncyiindrical crater species.
mechanism for an aluminum oxide coating should explain this behavior. Although the sensitivity of the PDMS detector tends to decrease with an increase in mass, for an instrument that routinely detects masses as large as 25 000 Da, instrumental variations in sensitivity will not explain the various curve shapes of Figure 2, covering the relatively narrow mass range of ca. 300-2000 Da. A linear relationship is an indication that cluster cleavage during fragmentation occurs with equal probability at any position to produce cluster ions of any size without preference. Variation in peak area for the various members of the series is then solely a function of the number of such clusters possible with random cleavage. It would appear that only a cylindrical projectile can meet the requirement of equal probability in bond cleavage. The schematic diagram of Figure 3A is useful in following this reasoning. Consider the hypothetical case in which there are six monolayers of oxide on the surface and a cylindrical projectile of aluminum oxide clusters is ejected. From this projectile only one cluster of six units can be produced if ionization occurs without fragmentation, while fragmentation can lead to two possible clusters of five units, three clusters of four units, four clusters of three units, five clusters of two units, or six single units of A1203. That is, the number of particles is inversely proportional to the number of aluminum oxide units in the fragment, as was found for the two predominant species in the positive- and negative-ion modes, (A1203),*A10H+and (A1203),.A10(OH)-. The series (A120~)..AIO~~Al(OH)2-, or alternatively written, (A1203)n+l. H20-, also produces close to a linear relationship (Figure 2B), but all others are clearly nonlinear. Close observation of the structure in Figure 3B shows that there are no bonds between adjacent metal oxide molecules but only between oxides projecting toward the surface. If a structure similar to this exists, then a cylinder of oxide clusters is completely reasonable and, in fact, would be expected. Such a model would explain all the experimental results. Protons are in great abundance, as evidenced by the large peak area at m l t = 1,
2355
at least 10 times larger in area than even the largest aluminum oxide cluster peak. The great abundance and high mobility of protons should allow them to collide with some of the fragments early in the acceleration process, thus eliminating the requirement for hydrogen atoms being present in the lower aluminum oxide layers prior to being ejected. The nonlinear curves of Figure 2 indicate that fragmentation with the formation of larger clusters (i.e., those with smaller values of l / n ) occurs less frequently than would be dictated by random cleavage produced through equal probability of cleavage at any position since the integrated peak area counts are considerably less for the larger fragments than observed for the linear random distribution examples discussed above. On the other hand, cleavage of small fragments occurs readily, as seen by the rapid increase in peak area for smaller cluster ions (e.g., the (A1203),-H+curve in Figure 2A). Such a situation might be expected to occur if a more conical-shaped fragment with a two- or three-dimensionalnetwork of bonds were formed with increased intermolecular bonding, perhaps as shown schematically in Figure 3A,C. Thus fragments of different cluster series may come from different locations with respect to the point of impact of the =2cf fission product or from different positions on the metal surface. One other experimental observation may be added to substantiate this conclusion. In each case peak and series identification was made on the basis of molecular weight and consideration of all atomic species possible. Table I gives the agreement found for the several series for both positive- and negative-ion modes. It should be noted that excellent agreement between apparent and calculated masses is obtained for the first 11 members of the major series (Al2O3),.A1OH+,the first 18 members of the major series (Al2O3),.Al0(OH)-,and the first 11 members of the series (A1203),AI02~Al(OH)~. For higher members of the series the apparent mass exhibits a sudden increase in error and the peaks become broader and less symmetrical (see insert, Figure 2A, m / z > 1000 Da), although no discontinuity in peak area occurs. Agreement in calculated and observed masses is also excellent for all other positive-ion series but is poorer for the short, negative-charged series. Although these peaks were smaller, broader, and generally of poorer symmetry than the major ion fragments, and therefore somewhat harder to achieve reproducible results, there can be no question that the difference between apparent mass and true mass is real. In all cases there is no reasonable alternative structure to those proposed. There may be some significance to the observation that the value of n at which deviations become large seems to coincide closely with the maximum size observed for one of the other series. An apparent mass larger than the true mass would be expected if there were a slight retardation in the ejection of the fragment, perhaps owing to multiple intermolecular bond cleavages near the periphery of the crater or to cleavages at greater depth within the oxide layer. An apparent mass less than the true mass would be expected if further fragmentation occurred near the end of the acceleration process, thus leading to enhanced cluster acceleration and earlier arrival at the detector. This subject has also been discussed by Demirev et al. (11). Those species showing significant negative errors in apparent mass are shorter series with values of n less than 10. The particular ion formed may thus be a function of the energy available for bond cleavage in relationship to the position at which the 252Cffission product strikes the target. Counting Metal Oxide Monolayers. If the above desorption mechanism for aluminum oxide clusters is correct and those ions producing linear plota in Figure 2 are produced from a column of aluminum oxide molecules ejected from the surface of the metal, then a plot of integrated peak area vs.
2356
ANALYTICAL CHEMISTRY, VOL. 63,NO. 20, OCTOBER 15,
l/n
P
““t
l/n
0
Flguro 4. Estlmatlon of ox& monolayers on aluminum surfaces of different age and chemlcal treatrmnt: (A) from (AI@3)n-AIOH+ serles; (e)from (AI@,),.AIO(OH)- wries. Surfaces: cvve A, IO-month old, alr-exposed alumlnum; curve B, newer aluminum surface; curve C, surface acid-washed with 0 . 1 % trlfluoroacetlc acM. Number of monolayers calculated from x axis Intercepts: (A) 24, 18, 16, respectively; (B) 23, 18, 15, respectively.
l / n for each member of the series, extrapolated to zero area, should provide an actual count of oxide monolayers. It should be noted that the number of start counts used in obtaining these data has no effect on the extrapolated result. Parts A and B of Figure 4 show the linear portion of curves for (Al2O3),.AlOH+ in the positive-ion mode and (Al2O3),.A1O(OH)in the negative-ion mode for three aluminum targets of different age and chemical treatment. It is interesting to point out that the intercepts with the x axis are almost identical for both positive- and negative-ion modes. The intercept for the aluminum target greater than 10 months old is the reciprocal of 25.0 in the positive-ion mode and 23.4 in the negative-ion mode. The newer aluminum target prepared in our laboratory gives intercept reciprocals of 18.9 in the positive-ion mode and 19.0 in the negative-ion mode. The laboratory-prepared aluminum target, when acid-washed with
1991
0.1 % trifluoroacetic acid as previously described, gives intercept reciprocals of 16.9 in the positive-ion mode and 16.0 in the negative-ion mode. These results therefore suggest that the older aluminum target has 23-24 monolayers of oxide coating, the newer aluminum target has 18monolayers of oxide coating, and the acid-washed aluminum target has 15-16 monolayers. Although a t least 10 different crystal structures for aluminum oxide formed under various conditions, generally at high temperature, have been characterized by X-ray methods, the crystal structure of A1203 formed at room temperature on the surface of aluminum metal has not been determined. We have attempted to determine this structure by using low-angle X-ray diffraction measurements performed over a 72-h period but were unsuccessful owing to the thinness of the oxide layer on the aluminum surface. Cotton and Wilkenson (12) state that the aluminum oxide which forms on the surface of metallic aluminum differs in structure from a-A1203,which is very hard and resistant to hydration and attack by acids, and yA1208,which readily takes up water and dissolves in acids. They state, without literature reference, that this form of A1203has a rock-salt defect structure which produces an arrangement of Al and 0 ions like rock-salt, but missing every third A1 ion. Our PDMS experiments would add that this form is also attacked by acid and probably does not form at thicknesses much greater than about 25 monolayers in air at room temperatures. XPS Confirmation of PDMS Results. An independent technique has been used to corroborate the number of monolayers determined by our PDMS method. Although there is no other known method for directly counting oxide monolayers, X-ray photoelectron spectroscopy (XPS) was used to approximate the thickness of the oxide layer. With this value and an estimate of the thickness of a single monolayer, an approximation to the number of monolayers may be calculated. Measurements were made on the sample of aluminized polyester employed in our laboratory-preparedtargets by using the XPS method of Ebel and Liebl(13). With the value X = 15.02 8, for the inelastic mean free path of 1.2536-keVphotoelectrons in aluminum oxide, calculated from the data determined by Reich et al. (14), an estimate of the oxide thickness of 41 f 5 8, was obtained, the range being calculated from maximum and minimum slopes in the experimental measurements. If the monolayer distance is assumed to be 2.1 A (15),then the number of monolayers range from 17 to 22, a value encompassing our value, 18, measured by PDMS. Thus the XPS data confirms convincingly that the interpretation of the PDMS data is correct.
ACKNOWLEDGMENT We thank Professor William Glaunsinger and Dr. Mahesan Chelvayohan, Department of Chemistry, Arizona State University, for running the X-ray photoelectron spectrometric analysis of our sample. Registry No. Al, 7429-90-5; AlzOs, 1344-28-1. LITERATURE CITED (1) Torgsrm, D. F.; Skowronskl, R. P.; Macfarlane, R. D. &oc/”. Bophys. Res. Commun. 1974. 80, 616-621. D. F. Sdbnoe 1976, 101, 920-925. (2) Mockriano, R. D.; tor-, (3) Jonseon, 0. P.; Hedln, A. B.; Hakansson, P. L.; Sundqvlst, B. U. R.; Save, B. 0. S.; Nielsen, P. F.; Rospstorff, P.; Johanwon, K.4.; Kamnsky, I.; Llndberg, M. S. L. Anel. ’3”.1986, 58, 1084-1087. (4) Juvet, R. S.; Alknekr, G.; Schmld. E. R.. A M / . chkn.Act8 1990. 247, 241-248. (5) Macfarlane, R. D. In Son Ionlzatbn B b b g h l Mss Spsctromehy; Manls, H. R., Ed.: Heydm 6 Sons,LM.: London, 1981; pp 110-119. (6) Tuwynskl, W.; Hllf, E. hbor R8X/S, h b W 2000 1989, 99-103, (7) Noes, B.; Nkschkr, E.; Volt, H. Lwt. Notes phvs. 1987, 260, 72-78. (8) Sundqvlst, B. U. R.; Kamnsky, I.;Hakansson. P. L.; K)eMberg, J.; Saiehpour, M.; Wlddlyasokera, S.; FoMmnn, J.; Peterson, P. A,; RoepS t M , P. BOnWd. M S S SpOCLrom. 1984, 1 1 , 242-257. (9) Katakw, I.; Ichlhara, T.; Ito, H.; Matwo. T.; Sakural, 1.;Matsuda. H. Int. J . M s s Spsctrom. Ion phys. 1989, 01, 93-97.
Anal. Chem. 1901, 63, 2357-2365 (10) Katakuse, 1.; Ichlhara, T.; Fujlta, Y., Matsuo, T.; Sakurai, T.; Matsuda. H. Int. J . Mss Spectrom. Ion phys. 1986, 69, 109-114. (11) Demlrev, P.: Olthoff, J. K.; Fenselau. C.; Cotter, I?. J. Anal. Chem. 1987, 59, 1951-1954. (12) Cotton, F. A.; Wllkenson. G. In Advanced Inorgenic Chemistry; Wlley: New York, 1988; p 211. (13) Ebei. M. F.; Lkbl, W. J . €leclron Spectrosc. Relet. phenom. 1979, 16, 463-470. (14) Retch, 1.: Yarrhemski, V. G.; Nefedov, V. 1. J . €kctron Spectrosc. Rebt. phenom. 1988, 46, 255-267. (15) Hannawak files: average value for the varlous known alumlnum oxides. Joint Committee for Powder DiffractionlStandards. Powder Dif-
2357
fraction and RetrlevallDlsplay System, Versbn 2.10. PDF-2 Data Base, International Centre for Dtffractbn Data, Swarthmore. PA, 1989.
RECEIVED for review April 10, 1991. Accepted July 29,1991. This work was presented at a meeting of the Deutsche Physikalische Geseuschaft, Munich, Gemany, March 12-16, 19g0, and in part$ by Grant p6854 from the Austrian “Fonds zur Forderung der Wissenschaftlichen Forschung”.
Initial Characterization of a Plasma Gun Source for Atomic Spectroscopy Joel M.Goldberg* and David S . Robinson’
Department of Chemistry, University of Vermont, Burlington, Vermont 05405-0125
A plasma gun Is descrlbed and evaluated as an emission source wltabk for direct soM mmpllng methods of spectrochemkd analysk. Atomlzatlon occurs In the radially contlned currmt-carrylng portkn of the plasma, and emlsdon is measured In the expelled, cooling plasma plume; this allows both spatial and temporal segregation of these competing proceoes. Temporally and spatlally resolved spectroscoplc studles of the emlssion from the plasma gun source are presented. The lnltlal plasma front Is expelled from the discharge tube at about 3000 m/s and then slows to about 700 m/s at a hdght of 8-10 mm. Plasma fonnatbn and hnpkskn processes result in the expulslon of a sharp, rapidly propagating pulse of plasma, whlch appears to combine or coillde wlth plasma that Is b l n g contlnuously expelled due to the oscillatory heatlng and coding of the plasma In the dlscharge tube. Spatially resolved emlolon measurements show slgnlflcant Increases In both continuum and ilne emlsslon at heights greater than 10 mm. It is suggested that this behavior le due to both colllslonal and optical reexcitation processes. It k shown that masking of emhrlon from the lower r.gkn, of the pknw (0-10 mm) and gatlng omlsdon from the plum after the flrd curnnt half-cycb (after about 75 ps) can enhance V( I I ) llne-to-background ratlos by almost 10-fold. Applicatlon of the source to the direct qualltative analysls of a USGS reference rock sample (MAQ-1, Marlne Mud) Is presented.
The determination of elemental concentrations directly in refractory, nonconducting solids remains as one of the unsolved challenges in atomic emission spectrochemical analysis. Direct solid sampling (DSS) approaches based on segregation of the atomization and excitation steps have produced some of the most promising results (1-5); however, there are very few sources that are suitable atomizers for the analysis of very refractory materials (6, 7). We have been investigating the properties of an imploding thin-film plasma (ITFP) source that shows promise for development as an atomization source
* To whom correspondence should be addressed. ‘Current address: Sentex Sensing Technology, Inc., 553 Broad Ave., Ridgefield, NJ 07657.
capable of analyzing solid refractory materials. The ITFP is created by discharging a high-voltage capacitor bank through a cylindrically symmetric thin conductive film. The thin film is coated on the inner wall of a nonconductive substrate, and the plasma that is formed implodes and is then confined radially. Solid powders deposited on the surface of the thin film in the discharge tube come into intimate contact with this hot plasma and are rapidly converted to an atomic/ionic vapor. We have studied this plasma source in considerable detail (8-12)and found that: (1) its extremely high power densities enable the complete atomization of compounds as refractory as vanadium carbide, (2) axial observation of emission from the plasma is not suitable for analytical measurements due to severe self-reversalof even analyte ion linea, and (3) confinement of the plasma for in situ reexcitation is hampered by significant axial expansion of the plasma. Coupling this atom cell with an external reexcitation source, then, would require transport of the vapor produced by the ITFP. In this report, we describe an ITFP source that has been modified so that plasma may expand axially in only one direction, allowing measurement of emission from the expelled plasma plume as well as direction of the sampled vapor into an appropriate reexcitation source (such as an ICP). Due to the high velocity with which this device expels plasma, it is referred to as a plasma gun. We report here on the initial spectroscopiccharacterization of the plasma gun as an atomic emission source for direct solid sampling.
EXPERIMENTAL SECTION The dischargecircuit and chamber have been described in detail previously (8). Discharge Cassette. The plasma gun discharge cassette shown in Figure 1 is a modification of the original ITFP discharge cassette described previously (8). The cassette consists of a lower brass electrode support block (marked a), a brass side support block (marked d), an upper brass electrode support block (marked b), and two polycarbonate insulating blocks (marked c). Electrical contact with the capacitive discharge circuit is made via three hemispheric dimples milled into the bottoms of the lower and side brass support blocks (a and d, respectively),which mate with three raised electrodes inside the discharge chamber (not shown). Electrical contact with the discharge tube is made by using graphite electrodes that press fit into the lower and upper brass support blocks (a and b, respectively). Polycarbonate shields (marked e) provide electrical insulation as well as ensure good
0003-27001@1/0363-2357$02.50/0 0 1991 American Chemical Sockty