lnterelement Effects and a Possible Excitation Mechanism in a LowPressure Microwave Induced Helium Plasma Hiroshi Kawaguchi,' lkuo Atsuya,* and Bert L. Vallee* Biophysics Research Laboratory, Department of Biological Chemistry, Harvard Medical School, and the Division of Medical Biology, Peter Bent Brigham Hospital, Boston, Mass. 021 15
When a low-pressure, microwave Induced hellum plasma Is used as an excitation source for metals vaporized from a filament, for many elements log-log callbration curves with other than a unit slope are obtained. Addition of potassium chloride or other salts to the sample not only enhances the spectral Intensities but also changes the slope? of the callbration curves to unity for these elements. Several techniques have been employed to explore the excltatlon mechanism In the plasma. A possible excltatlon mechanism is proposed in which both dissociation of the molecule and ionizationof the metal occur slmultaneously rather than stepwise.
Low-pressure, microwave induced emission spectrometry combined with thermal atomization of micro samples provides one of the most sensitive analytical techiques for trace elements in solutions ( 1 ) .This technique has been applied successfully for the determination of the metal content of biological materials such as metalloenzymes ( 2 , 3 ) .In the course of these experiments, emission intensity enhancements of 6to 1000-fold have been observed when KC1 is added to the samples, although the underlying mechanism has remained obscure. Busch and Vickers ( 4 ) ,Brassem and Maessen (5), and Avni and Winefordner (6) have recently reported on the fundamental properties of low-pressure microwave induced plasmas, such as electron temperature, excitation temperature, electron concentration, etc. However, the excitation mechanism which they proposed for low-pressure plasmas neither accounts for the various matrix effects reported previously (1) nor for those presented here.
EXPERIMENTAL Apparatus. Experimental equipment and conditions have been described ( I ) . A block diagram of the overall apparatus is shown in Figure 1and the important components and conditions are given in Table I. Procedure. Sample solutions of 5-pl volume are applied to the filament, and the solvent is vaporized slowly by passing a current of 2 to 3 A through the filament. When desolvation is complete, the discharge tube is evacuated to a pressure of 3.8 Torr while helium is flowing continuously. The microwave discharge is initiated, and the sample is vaporized by pulse heating of the filament to a temperature of 1800-2000 "C and is then swept into the discharge plasma by the gas flow. The heating pulse is obtained by discharging a 0.216-F capacitor charged to 7.5 V through the filament. The peak intensities of spectral lines are measured by an electronic peak detector and printed out digitally. Signals are also monitored on a dual-trace storage oscilloscope. The relative population of neutral atoms produced in the discharge tube by the filament is measured by atomic absorption by removing the microwave cavity from its position and aligning a hollow cathode source and a quartz condensing lens behind the discharge tube. Radiation from the hollow cathode traverses the discharge tube and is
Present address, The Faculty of Engineering, Nagoya University, Chikusa-ku, Nagoya, Japan. Present address, Kitami Institute of Technology, Kitami, Japan. 266
focused onto the entrance slit of the monochromator. Under operating conditions identical to those shown in Table 1(except for application of the microwave power), sample deposited on the filament is vaporized and absorption of radiation is measured by means of the storage oscilloscope. Preparation of Solutions. All analyte metal solutions were prepared either from reagent grade, soluble salts or metals dissolved in minimal amounts of acid and diluted to the desired concentration with distilled, deionized water. All analyte solutions of the metals were chlorides, except for cadmium, cobalt, copper, lead, magnesium, manganese, and zinc for which the nitrates were used, and molybdenum which was the molybdate.
RESULTS AND DISCUSSION Effect of Potassium Chloride on the Slope of the Calibration Curves. The presence of 2 to 10 mM potassium chloride in sample solutions not only increases the spectral intensity of many elements by as much as 6- to 1000-fold,but also changes the slope of the log-log calibration curves for iron and cobalt to unity ( 1 ) . Subsequent experiments have shown that in the absence of this additive the slope of the log-log calibration curves is frequently much greater than unity. Table I1 qualitatively summarizes the characteristics of the slope of the log-log calibration curves of the elements listed obtained within the range of concentration from 0.001 to 2 pg/ml. For only 6 out of the 19 elements studied is the slope equal to unity in the absence of additives (Table 11, col. 3). In general for elements with steep calibration curves, poor detection limits and recoveries in intensity are observed. Although the addition of potassium chloride changes the slope of calibration curves for lithium, sodium, iron, cobalt, and chromium to unity, for other elements the slope is affected only slightly (Table 11, col. 4). Optimal concentrations of some additives, other than potassium chloride, can also change the slope to unity (Table 11, col. 5 ) . The calibration curves for calcium in the presence and absence of 12 pg/ml barium exemplify these matrix effects (Figure 2). In the absence of barium, the detection limit of calcium is about 0.5 pglml; in its presence, the calibration curve extends to 0.01 pg/ml. The flattening a t the low concentration end of the calibration curves in the presence of added barium may be due to contamination and increase in background emission. The addition of barium increases the intensity of the Ca 4227-A line by about 300-fold when employing 0.5 rcg/ml calcium. The results are analogous when calcium is the additive for either barium or strontium measurements. Thus, addition of 10pg/ml of calcium to barium standards or of 10 pg/ml of either calcium or barium to strontium standards enhances the intensities of the analyte and changes the slope of their calibration curves to unity. Interestingly, the addition of only 2 pg/ml of cobalt to iron standards enhances the intensity of the Fe 3719-A line to the same extent as does the addition of from 6 to 12 mM potassium chloride, resulting in a calibration curve of unity slope for iron. In the course of sample volatilization, the effect of additives on intensity variations of the He 2945-A line was measured by means of the storage oscilloscope to determine the influ-
ANALYTICAL CHEMISTRY, VOL. 49, NO. 2, FEBRUARY 1977
(Hel
Table I. Apparatus and Operating Conditions Microwave generator Cavity
Monsanto Research Corp.,Model 207,2450MHz, operated at 30 W Opthos Instrument Co., Evenson type quarterwave Discharge tube Quartz tubing of 4.3-mm o.d., 2.3-mm id., and 150 mm long Plasma gas Union Carbide Co., ultra high purity grade helium (99.999%), flow rate: 0.48 Umin at atmospheric pressure Plasma 3.8 Torr pressure Sample loading0.25-mm dia. tantalum wire, V-shaped, 15 mm filament high with a round bottom of 2-mm dia. Spectrometer McPherson Instrument Corp., Model 218,0.3-m modified Czerny-Turner grating: 1200linedmm, dispersion: 26.5 A/m, slit width: 0.05 mm Storage Telequipment, Model DM 64 oscilloscope
,-
'W LZ-1
1D.C.
Vacuum
Amplifier J I
Peak detector
Flgure 1. Block diagram of apparatus
Table 11. Behavior of the Slope of the Log-Log Calibration Curves in the Presence ar Additives Slope of calibration Wavecurve EleWithout Other adjuvants to l e n f h ' additive With KCl induce unit slope ment
Cd co Cr
13961.5 15535.6 I1 4554.0 14226.7 I1 3933.7 I 2288.0 13453.5 14254.3
cu Fe K La Li Mg Mn Mo Na Ni Pb Sr Zn
13247.5 13719.8 I 4044.1 I1 3949.1 16707.8 I 2852.1 I 4030.8 13798.3 I 5890.0 I 3414.8 I 4057.8 14607.3 12138.6
A1 Ba
Ca
0
'
a
3 3 2 2 1
4 2 1
4 3 2
4 1 1
20 3 4 1
5 1
6 3 3 2 2 1 1 1
1 1
.. 2 1 1 1 5 1 3 1 5 1
10 pg/ml of Ca 12 pg/ml of Ba
8 mM KCl HCl
+ 15 mM
2 pg/ml of Co 30 pg/ml of Li 20 wg/ml of Sr 1mM NaCl
c
L
loo 0.01
Ca concentration, 0.6 mM LiCl
uglml
Figure 2. Calibration curves for calcium in the presence (-)and absence (- - - -) of barium. Ba: 12 pglml; (0)Ca I 4227 A: ( 0 )Ca II 3934 A
10 pg/ml of Ba or Ca
100 pg/ml of A1 is not detected.
ence of the additives on excitation conditions in the plasma. The lower energy state of this line is a metastable state of helium. Five pl of 10 mM potassium chloride maximally decreases the intensity of the helium line by as much as 95%. In contrast, 5 pl of 10 pg/ml of calcium chloride does not affect the intensity of this line significantly, and even 100 pglml of calcium chloride decreases it by only 3%. Clearly, while potassium chloride drastically affects the excitation conditions of the helium plasma, comparable concentrations of calcium do not. Thus, enhanced intensity of spectral lines of metals cannot be attributed solely to alterations in excitation conditions of the plasma, but must also relate to the conditions which govern the introduction of samples, such as solutevolatilization phenomena. Effect of Phosphate on the Intensity of Copper. In low temperature flames and low power microwave discharges at
atmospheric pressure, phosphate interferes with calcium and strontium intensities and that of other elements (7), an effect observed also in the present low pressure system. In some instances, however, it can be eliminated by the addition of larger amounts of alkaline-earth elements. For example, 0.1 mM potassium dihydrogenphosphate suppresses the Cu 3247-w emission line, even in the presence of 8 mM potassium chloride, but it is restored progressively by the addition of strontium until, at 40 pg/ml of strontium, the effect of 0.1 mM phosphate is eliminated completely. This circumstance frequently serves to eliminate the interference of phosphate both in flame emission and absorption methods, and the presumable mechanisms have been discussed (8). Volatilization of Samples from the Filament. When filaments are heated to high temperatures, a fraction of the sample elements vaporize in the form of neutral atoms and are the basis for some recent flameless atomic absorption systems. This technique was applied to the present system to examine the volatilization process of samples in the presence and abANALYTICAL CHEMISTRY, VOL. 49, NO. 2, FEBRUARY 1977 * 267
Atomic absorption
0.6-
77
0.2
V
Emission
t 0.2
i
I
Voltage ' 1 filamenton i
;
I
I
'
' Delay time
c0
Figure 4. Schematic diagram showing the measurement of delay time
Table 111. Delay Time in the Appearance of Emission and Atomic Absorption Signals with Reference to the Initiation of Heating of the Sample Filament
cu
Element Cd
co
cu
Distance from filament, mm
Fe Mg Mn Zn
Figure 3. Variation of the peak absorbance as a function of the distance from the filament in the presence (0)and absence (0) of 10 mM potassium chloride.
Emission, ms Atomic absorption, ms Wave Wavelen th, Without KC1 len th, Without KC1 matrix matrix matrix matrix
X
2288.0 3453.5 3247.5 3719.9 2852.1 4030.8 2138.6
X
25 18 72 18 80 45 25
35 40 32 25 36 30 32
'
2288.0 2407.3 3247.5 2483.3 2852.1
20 100 70 90 50
...
*..
2138.6
20
20
...
20 100
70 70 50
Concentration (wg/ml): Cd, 0.6; Zn, 0.2; Mg, 0.6; Co, 20, Cu, 10; Fe, 10
sence of potassium chloride. The absorption of a spectral line of each of the elements was examined a t different positions along the discharge tube. The variation of the peak absorbance for 6 elements as a function of the distance from the filament is shown in Figure 3. These elements were selected owing to their relative sensitivity of detection by atomic absorption spectrometry. Measurements could not be performed closer than 40 mm from the filament because of the system design. Background absorption measured with a deuterium source was negligible in all instances. The decrease in absorbance as a function of increasing distance from the filament may reflect the decrease in the population of neutral atoms. The rate of decrease in absorbance differs for each element, and potassium chloride affects it but slightly. From the linear velocity of carrier in the discharge tube of 20 cm/ms a t 3.8 Torr and 23 "C, the half-lives of the neutral cadmium and iron atoms, for example, can be calculated to be 0.6 and 0.05 ms, respectively. The half-lives of neutral atoms decrease in the order Cd > Zn > Mg > Co > Cu > Fe, quite similar to those observed in the long pathlength cell employed for flame atomic absorption (9). The presence of 10 mM potassium chloride negligibly affects the population of neutral magnesium and cobalt atoms; for cadmium the effect is positive and for zinc, iron and copper, it is negative. Considering the large effect of potassium chloride in enhancing the emission intensity of these elements when excited by the microwave discharge ( I ) , these results show that there is at best only slight correlation between the population of neutral atoms entering the discharge and spectral line intensity. Further, the absorbance decreases as the distance from the filament increases, but the intensity of sample emission lines does not decrease when both the area of the discharge under observation and the microwave cavity 268
itself are simultaneously moved further away from the filament. Delay Times When Emission a n d Absorption Peaks Appear. Using initiation of heating of the sample filament as the reference point, there is a finite delay prior to the appearance of the emission signal ( I 1, and an even longer delay in the appearance of neutral atoms (plasma off) as evidenced by the atomic absorption signal. The appearance of signals on the storage oscilloscope is shown schematically in Figure 4 and the delay times measured both in the presence and absence of potassium chloride are summarized in Table 111. Since the delay times have about 20% uncertainty and change gradually as the number of heating cycles of the sample filament increases, the data in Table I11 cannot be considered absolute. However, the results demonstrate three important facts. First, when potassium chloride is added to the sample, the delay times of the emission signal approach about 33 ms from both sides, a value approximately coincident with the delay time of the potassium emission line. In contrast, potassium chloride does not vary the delay times of the atomic absorption signals, except for an apparent effect in the case of iron, which is perhaps within experimental error. Second, the delay times of the emission and atomic absorption signals are unrelated. This observation requires the postulate of another excitation mechanism which assumes direct formation of excited atoms from other than neutral atoms in the helium plasma. Last, the order of "delay times" of the atomic absorption signal of the elements coincides approximately with that of the half-life of the population of their neutral atoms, described earlier. Since the half-life correlates with the vapor pressure of metals ( 9 ) ,the above facts can be understood readily in terms of solute-vaporization effects. On the other hand, in the absence of potassium chloride, the values of the delay times of the
ANALYTICAL CHEMISTRY, VOL. 49, NO. 2, FEBRUARY 1977
L
I
4In
c C
-
3
f
2 = n
82-
.-cI2n,
c -
I
I
1 2 KCI concentration,
I
I
3
4
4d
mM
I
I
1 KCI
I
2 3 concentration, m M
I
4
Figure 5. Variation of the intensity of the Zn 2139-A line as a function of the concentration of potassium chloride in the zinc solution
Figure 6. Variation of the intensity of the Cu 3247-A line as a function of the concentration of potassium chloride in the copper solution
Distilled water (a)and potassium chloride solution (b)are deposlted on the loops of the filament (see text). Zn = 0.032 Wglml
Distilled water (a) and potassium chloride solution (b)are deposited on the loops of the filament (see text). Cu = 0.032 pg/ml
emission signal are independent of the vapor pressure of metals. Effect of Potassium Chloride Introduced Separately. In order to examine the effect of potassium chloride on the volatilization and excitation processes from the rest of the variables, a modified tantalum filament was devised, and the remainder of the conditions was left unaltered. The filament, normally “V”-shaped, is altered to assume the shape of a “W” so that two different solutions can be deposited singly in each of the depressions of the W. Five p1 of a 0.5 pglml solution of zinc, containing no other additives, is deposited in one of the depressions and 5 pl of potassium chloride, 0 to 4 mM, is deposited in the other. The sample solutions are dried and volatilized as usual, and the emission intensity of the Zn 2139-A line is measured for each pair. As the potassium chloride concentration increases from 0 to 1mM, the intensity of the zinc line increases by only 20% and remains unchanged in the concentration range from 1 to 4 mM. Analogous results are obtained when studying copper. Apparently, potassium chloride affects the excitation process in the microwave-induced helium plasma only slightly, as is confirmed by the following experiments using the W-shaped filament. Five p1 of metal solution containing 0.1 to 4 mM potassium chloride are deposited in one of the wells of the W filament, (I). In experiment (a), 5 pl of potassium chloride solution, from 3.9 to 0 mM, replace the water ip the second well, (11).These concentrations are chosen to hold the total concentration of potassium chloride in wells (I) and (11)equal to 4 mM. In experiment (b) only water is used in well (11).Figures 5 and 6 show the variation of intensity of Zn 2139 A and Cu 3247 as a function of the concentration of potassium chloride in the solutions in (I). Curves (a) and (b) denote experiments (a) and (b),respectively. The intensity differences between (a) and (b) indicate the effect of potassium chloride on the excitation process. These data show clearly that potassium chloride enhances the intensity largely by affecting the volatilization of metals from the filament. A Possible Excitation Mechanism. Busch and Vickers ( 4 ) have proposed a radiative recombination model as the mechanism responsible for excitation in low-pressure noble gas plasmas sustained by microwave power. They have discussed the importance of Penning ionization by metastable noble gas atoms. The ionization-radiative recombination equilibrium can be described by:
where M* represents an excited atom. The Penning ionization reaction can be written as:
M
+e
M+ + e
-
-+
+e +e M* + hvcont
M+
Armeta+ M
Ar
+ M+ + e
(3)
where Armetais a metastable argon atom. The excitation mechanism in a low-pressurehelium plasma is not readily explained by reactions 1-3 considering the unusual spectral emission and matrix effects described here. Since they dealt only with mercury, Busch and Vickers did not comment on the formation of the neutral atom, M, in reactions 1 and 3. The present experiments show that this process must be related directly to the excitation mechanism. Busch and Vickers examined the low-pressuredischarge under conditions substantially different from those here employed. Hence, their interpretations and mechanistic deductions could not encompass the results here obtained and conclusions which are now possible. We, therefore, would like to extend their interpretations to the present conditions and enlarge upon this theory. It is clear that our views may similarly be in need of amendment as the low-pressure discharge is studied further and should be considered a “working hypothesis.” In order to account for all of the experimental facts found, we propose that apart from reactions 1 and 3, the following ionization reactions have to be considered in addition: M C l f e * M + + C1+ e + e
+
Hemeta MC1- He
(1)
(2)
-
+ M+ + C1+ e
(4)
(5)
where M is a metal, MC1 a metal chloride and Hemetais a metastable helium atom. Other compounds of the metal with relatively high vapor pressure may replace MCl. These postulated reactions are possible since in the low-pressurehelium plasma, the electron temperature is as high as 5 X lo4 K ( 4 ) , and the metastable levels of He are at 19.8 and 20.6 eV. A fraction of the chloride molecules may be dissociated even while on the filament during volatilization, and the resultant neutral atoms then enter the plasma, as observed by atomic absorption. However,the efficiency of reactions 1 and 3 is too low to account for the overall reaction which requires a much larger amount of energy to be converted to kinetic energy. On the other hand, since reactions 4 and 5 consume more energy than do 1 and 3, the efficiency of the former may become larger than that of the latter. This may account for the delay time of the emission which differs from that of the atomic absorption signal. The low efficiency of reactions 1 and 3 is also the reason that the emission lines appear only at the leading edge of the discharge ( I ) ;once the molecules are dissociated at the leading edge of the discharge, they are not excited again along
ANALYTICAL CHEMISTRY, VOL. 49, NO. 2, FEBRUARY 1977
269
the remainder of the discharge tube. Reactions 4 and 5 also account for the steep slopes of the calibration curves of many elements and the enhancement effect of potassium chloride on spectral intensity. When a sample solution is dried on the filament, the solute crystallizes and is deposited on the surface of the filament. As the concentration of the salts in the sample increases, the crystals grow larger, and the fraction of salt directly in contact with the filament decreases. In turn, this increases the fraction of salt that volatilizes as molecular vapor with the result that emission intensity increases rapidly because of reactions 4, 5, and 2. If no extraneous salt is added to the sample, for metals whose salts are decomposed easily, the slope of the calibration curve will be steep. In the presence of potassium chloride, its crystals will include the metal salts and prevent both the dissociation of the chloride molecule and the formation of the refractory oxides as the sample is volatilized , from the filament. Potassium chloride is not always efficient either in enhancing spectral intensity or changing the slope of the calibration curve, probably due to fractional volatilization of salts from the filament. In this case, a suitable salt should be chosen as an additive, e.g., calcium chloride when barium or strontium is to be determined, as described earlier. Addition of potassium chloride may strongly affect the excitation conditions of the helium plasma through the population of metastable helium atoms and the electron temperature. The fact that the position of the’ emission maximum in the discharge tube shifts toward the cavity ( I )
reflects these effects of potassium chloride. Since the intensity of helium lines and the electron temperature increase as the distance from the cavity decreases, the variation of the excitation conditions due to the addition of potassium chloride may be partly compensated for by the shift in the position of the emission maximum. The proposed mechanism described here fits the observed data. Other factors, such as changing impedance of the microwave cavity upon introduction of KC1, could also enter into the excitation mechanism and the enhancements observed. However, because of the transient nature of the sample introduction process, optimization of the cavity tuning is not practical.
LITERATURE CITED (1) H. Kawaguchi and B. L. Vallee, Anal. Chem., 47, 1029 (1975). (2) D. S. Auld, H. Kawaguchi, D. M. Livingston, and 8. L. Vallee, Proc. Natl. Acad.
Sci. USA, 71,2091 (1974). (3)H. Kawaguchi and D. S.Auld, Clin. Chem. ( Winston-Salem, N.C.), 21,591 (1975). (4) K. W. Busch and T. J. Vickers, Spectrochim. Acta, Part 6, 28, 85 (1973). (5)P. Brassem and F. J. M. J. Maessen, Spectrochim. Acta, Part 6, 29, 203 (1974). (6)R. Avni and J. D. Winefordner, Spectrochim. Acta, Part B, 30, 281 (1975).
(7) H. Kawaguchi, M. Hasegawa, and A. Mizuike, Spectrochim. Acta, Part 6,
27, 205 (1972). (8)Juan Ramirez-Mfinoz, “Atomic-Absorption Spectroscopy”, Elsevier Publishing Company, New York, 1968.p 271. (9)A. Ando, K. Fuwa, and B. L. Vallee, Anal. Chem., 42,818 (1970).
RECEIVEDfor review September 3,1976. Accepted November 11,1976. Work supported by Grant GM-15003 from the National Institutes of Health of the Department of Health, Education, and Welfare.
Determination of Impurities in Gases by Atmospheric Pressure Ionization Mass Spectrometry Hideki Kambara” and lchiro Kanomata Central Research Laboratory, Hitachi, Ltd., Kokubunji, Tokyo, Japan
An Atmospheric Pressure Ionization (API) mass spectrometric technique in which primary ions are produced by corona discharge at atmospheric pressure is successfully applied to detection of small inorganic molecules in highly purifled gases. The Ions, which result from a complex series of ionlzatlonreactions, are continuously supplied through two small apertures (0.1 mm and 0.2 mm) into the mass analyzing region. A total A is obtained at the mass analyzer Ion current of 1 X entrance sllt by using a two-stage differential pumping method. A collision-Induced dissociation method is employed to determine the impuritles. As a result, many impurltles are detected sensitively, Le., NO, COP, H20, and O2 in nitrogen gas and CH4, H20, and C2Hs in oxygen gas. The concentrations of NO in nitrogen gas and CH4 in oxygen gas are determined to be 250 ppb and 1.8 ppm, respectively.
The analysis of trace quantities of compounds has become increasingly important in various fields in addition to organic chemistry. For example, highly purified nitrogen gas is used as the standard for optical fluorescence analysis of air pollution. A t times, trace quantities of impurities in the standard gas have a critical influence on investigation results. It has been noted that high purity nitrogen gas produces a back270
ground interference ( I ) . However, the lack of an adequate analytical instrument has prevented comprehensive studies of this problem to date. The need for improved gas analysis techniques has also become vital in semiconductor technology. This is because it is considered that impurity concentrations should be kept below lo-’ mol parts. In addition, it is necessary to control and monitor process gas quality (2). This requires the detection of inorganic as well as organic compounds. One of the most effective gas analysis techniques involves a combination of mass spectrometry and gas chromatography. It can detect organic compounds as small as 10 pg. The sensitivity of a mass spectrometer depends on the background spectra, the number of ions reaching the ion collector, and the ionization rate of nondissociated samples. Therefore, there are many other difficulties involved in detecting small molecules than with large organic compounds. One is that, in the low mass region, many kinds of background ions can be observed which disturb the detection of the impurity trace quantities. Another is that these molecules cannot be concentrated with any available interface, e.g., jet separator, between the gas chromatograph and the mass spectrometer. No practical solution to these problems has been reported yet. Recently Homing developed a new mass spectrometric system, the so-called “Atmospheric Pressure Ionization Mass
ANALYTICAL CHEMISTRY, VOL. 49, NO. 2, FEBRUARY 1977
