2549
Anal. Chem. 1985, 57,2549-2552
Determination of Calcium in Nodular Cast Iron by X-ray Fluorescence Spectrometry, Calibrated by a-Particle Activation Analysis Carlo Vandecasteele,*' Filip Alluyn, Jacques Dewaele, and Richard Dams Institute for Nuclear Sciences, Rijksuniversiteit Gent, Proeftuinstraat 86, B-9000 Gent, Belgium
A method for the determlnatlon of caiclum in cast iron at the 1-40 pg/g level by wavelength .dispersive X-ray fluorescence spectrometry (XRF) on the Ca K a line was developed. A chromlum tube was used for the excitation and a LIF(200) crystal for the dlffractlon. To establlsh a callbratlon graph a number of nodular cast Iron samples with calcium concentrations between 1 and 40 pg/g were analyzed by a particle actlvatlon analysis using the "Ca(a,p)"Sc reaction. QSc was separated from the matrix by cation exchange In a mixture of hydrofluoric and nitric acids. The calibratlon graph Is linear and allows determination of calcium by XRF with a precision of 3-10% In the 10-40 pg/g range.
During the production of nodular cast iron, certain postinoculanb are added to the melt to obtain a maximum number of nodules and a minimum amount of carbides. Postinoculation with calcium bearing ferrosilicon is successful, although the precise role of calcium in the nucleation mechanism is not clear. According to the theory of Lux (1) carbides with a salt-type structure are formed in the melt and the graphite grows epitaxially on the carbon layers of the calcium carbide crystals. The theory is however not generally accepted, since the oxides and sulfides of calcium are more stable than the carbides. In fact, some investigators (2-4) showed by means of scanning electron microscopy that sulfides of calcium, magnesium, aluminum, cerium, and strontium are present in the center of the nodules and act as nuclei. Calcium may also have a negative effect (5, 6) even in minute quantities, since i t promotes the formation of chunk graphite in heavy ductile iron castings. In practice, the amount of calcium to be added should be optimized. It is generally below 0.0170,but depends on the pouring temperature, on the size of the casting, and on the chemical composition of the raw materials. Thus, for a better control of the production process, an accurate determination of calcium in nodular cast iron at the microgram per gram level is necessary. Several papers describe the determination of calcium in colosteel by atomic absorption spectrophotometry (7-10), rimetry (111, or flame photometry (12) often preceeded by solvent extraction or electrolysis a t a mercury cathode to remove the main components (Fe, Cr, Ni). These procedures require a time-consuming destruction of the sample and often difficulties with blanks or interferences arise. In most cases, the standard deviation is high or is not given at all. F u et al. (13) determine calcium in steel by electrothermal atomic emission spectrometry a t concentrations below 40 pg/g with a relative standard deviation of 10% or lower. Few papers, however, describe the determination of calcium in cast iron. Kuemmel and Karl (14) and Goto et al. (15)used flame photometry. Atsuya and Goto (16) determine calcium 'Senior Research Associate of the Belgian National Fund for Scientific Research.
Table I. Experimental Conditions for XRF X-ray tube voltage current diaphragm collimator diffraction crystal detector discriminator window peak angle background angle 1 background angle 2
-
Ca Ka, 3.69 keV Cr 60 kV 40 mA
28 mm coarse (0.40") LiF(2OO) flow counter (90% Ar, 10% CHI) 300-700 mV 113' 10' 111' 50' 115O 50'
by plasma jet spectrometry. However, in addition to the disadvantages mentioned above, the methods were not applied to (16) or showed insufficient precision (14) in the concentration range of interest (1-40 Fg/g). We chose wavelength dispersive X-ray fluorescence spectrometry (XRF) on the Ca Kcu line (3.69 keV), because it is purely instrumental and fast and sample preparation is relatively easy. Since certified reference materials for calcium in nodular cast iron are not available, the calcium concentration was determined by charged particle activation analysis (CPAA) in some cast iron samples, to establish a calibration graph for XRF. CPAA has the advantage that contamination of the sample surface can be removed by chemical etching after the irradiation and that standardization can be obtained from pure calcium compounds with known stoichiometry (17). Of course, charged particle accelerators are not available to most industrial laboratories practicing XRF. Therefore, we intend to make available cast iron analyzed for calcium to interested laboratories.
EXPERIMENTAL SECTION Preparation of the Cast Iron Samples. Nine different cast iron samples were produced by WTCM (Wetenschappelijk en Technisch Centrum van de Metaalverwerkende Nijverheid (Scientificand Technical Center of the Metallurgical Industry)). The melts were prepared in a medium frequency induction furnance, using a charge of S 100 grade Sore1 pig iron and commercial Fe-Si (70% Si), Fe-S (33% S),and Fe-Mn (75% Mn) alloys. At a temperature of about 1400 "C, calcium was added to the melt as a Ca-Si(33% Ca) alloy. Cylindrical bars of 60 mm diameter and 300 mm length were cast. Before shake-out,the castings were allowed to cool to room temperature. All nine cast iron bars thus prepared had a typical composition of ca. 3.8% C, ca. 2.2% Si, ca. 0.7% Mn, and 0 40Ca(a,pn)42mS~ 17.3 40Ca(d,a)38K 40Ca(a,p)43S~ 3.8
Q’
radionuclide formed
Table IV. Irradiation Conditions sample radiation emitted
tIl2
4 2 m S ~ 61.6 s p’, y 4ZrnSC 38K 7.63 min p’, y 43Sc 3.89 h p’, y (372.8 keV)
Table 111. Nuclear Interferences nuclear reaction 40Ca(a,p)43S~ 40K(a,n)43Sc 41K( a ,2n)43Sc 45Sc(a,a2n)43Sc
threshold energy, MeV
isotopic abundance,
3.8 3.3 14.4 22.9
96.94 0.012 6.7
%
100
polishing treatment, were minimized by rotating the sample. Sample preparation takes ca. 5 min per sample. A Philips PW 1400 spectrometer was used. The experimental conditions are summarized in Table I. A fixed 28-mm diaphragm was selected to minimize scattered radiation, since with a 35-mm diaphragm intense contributions to the Ca K a line were found, originating from the sample holder and the mask. Since no significant second-order interference was found, a wide discriminator window (300-700 mV) was set. Net intensities were calculated by linear interpolation of the background values taken a t angles before and after the peak, respectively (Table I). Each intensity measurement lasted 200 s, so that the total measuring time was 10 min per sample. Charged Particle Activation Analysis. Nuclear Reactions. Nuclear reactions for the determination of calcium should preferentially start from @Ca(96.94% isotopic abundance). Table I1 gives the nuclear reactions on @Cathat yield radionuclides with reaction has an half-life longer than 1 min. The 40Ca(a,p)43S~ the advantage that the 3.89 h half-life product allows a postirradiation radiochemical separation. Table I11 lists the nuclear interferences. Owing to the low isotopic abundance (0.012%) of 40K,interference of the 40K(a,n)43S~ reaction is negligible. The 41K(a,2n)43Scand 45S~(a,a2n)43S~ reactions do not occur for incident energies below 14.4 MeV. Irradiation and Chemical Etching. Calcium carbonate powder was used as a standard. The samples and the standards were irradiated in vacuum with a particles extracted from the CGRMeV 520 cyclotron at Ghent University. Copper foil was placed before the sample and the standards to serve as a beam intensity monitor. Between the monitor foil and the sample a 1.56 mg/cm2 aluminum foil was placed to stop recoil nuclei. Aluminum foils with different thicknesses were also placed before the calcium carbonate standards. Table IV summarizes the irradiation conditions and gives information on beam intensity monitoring. After the irradiation, the samples were etched for 17 s in 6 M nitric acid at room temperature to remove a 4.1-7.2 mg/cm2 surface layer. After being etched, the samples were rinsed with water and acetone and dried in a stream of hot air. Chemical Separation. From the Ge(Li) y-ray spectrum of a cast iron sample irradiated with a particles, it appears that the 372.8-keV peak of 43Scis superimposed on the Compton continuum of a number of radionuclides produced from the matrix and coincides with the 372.9-keV peak of W u . Although the interference of elCu can be corrected for by using other peaks of %u, a more accurate y spectrometric measurement of 43Scis possible after chemical separation from the matrix activities, e.g., 57C0produced by the 54Fe(a,p)57Coreaction, 57Niproduced by the 54Fe(a,n)57Nireaction, 61Cu produced by the 6sNi(a,p)61Cu reaction, and 66Gaproduced by the 63Cu(a,n)66Gareaction. Complete dissolution of the sample, including graphite and carbides, requires approximately 5 h of refluxing in a mixture of concentrated nitric and perchloric acids. The iron matrix can however also readily be dissolved in 6 M nitric acid under heating,
standard
a-particle energy, MeV 17 17 intensity, FA 3-4 0.1-0.2 irradiation time, min 1-2 30 monitor foil, mg/cm2 Cu (6.9) Cu (6.9) 6sCu(a,n)66Ga nuclear reaction induced activity, t I j z 9.5 h E,, keV 1039.4 additional A1 foils 4.15 mg/cm2 (series 1) 1.56 mg/cm2 5.72 mg/cm2 (series 2) 7.28 mg/cm2 (series 3) 13.5-14.2 13.5-14.4 effective incident energy, MeV residual range, mg/cm2 34-37 but a residue of graphite and carbides remains undissolved. This procedure was preferred since it is much faster. Since part of the 43Scis in the residue, this is filtered off and measured separately. The matrix activity of the residue is relatively low, allowing an easy determination of the 372.8-keV peak. It must however be corrected for the contribution of “Cu. To separate scandium from radioisotopes of cobalt, nickel, copper, and gallium, cation exchange on Dowex 50W-X8 in 1 M hydrofluoric acid-0.1 M nitric acid was applied. By use of 46Sc (tip = 83.6 days) tracer it was found that about 97% of the scandium is eluted from a Dowex 50W-X8 column (2.3 cm internal diameter, 12 cm height of resin bed) with 100 mL of 1 M hydrofluoric acid-0.1 M nitric acid. Tracer experiments with 65Ni (tip = 2.5 h) showed that no nickel is eluted with 100 mL of this eluate. Since scandium is not quantitatively recovered, &Sctracer is added before the separation to allow the determination of the chemical yield and a subsequent correction. In a Ge(Li) y-ray spectrum of the eluate for an irradiated cast iron sample, the Compton continuum at the 372.8-keV peak is 2 orders of magnitude lower than that in the spectrum obtained without separation, with no interference from 61Cu. The experimental procedure is as follows: Dissolve the sample under heating in 25 mL of 6 M nitric acid. Add 5 mL of a 46Sc tracer solution and 2.5 mg of scandium(II1) oxide carrier. Filter off the residue on a filter crucible (G2) and wash with 6 M nitric acid. Evaporate to dryness and dissolve in 10 mL of 1 M hydrofluoric acid-0.1 M nitric acid under heating. Put the solution on top of a column (2.3 cm internal diameter, 30 cm height), filled to a height of 12 cm with Dowex 50W-X8 cation exchanger. Elute with 1 M hydrofluoric a c i d 4 1 M nitric acid and collect 100 mL of eluate. The procedure takes 4-4.5 h. Measurements. Five milliliters of &Sctracer solution is diluted to 100 mL with 1 M hydrofluoric acid-0.1 M nitric acid and measured for 90 min with a Ge(Li) detector. The residue of the dissolution is measured with Ge(Li) detector for 1h, 2-4 h after the irradiation, and the eluate is measured 6-8 h after the irradiation. The copper beam intensity monitor foils are measured for 10 min, 23-25 h after the irradiation a t a distance of 15 cm from the detector, which is covered with a 1 cm thick lead absorber to reduce the counting efficiency by a factor of 51. The standards are measured for 10 min, 2-4 h after the irradiation, either dissolved in 100 mL of 1 M hydrochloric acid or placed in a filter crucible. The beam intensity monitor foils for the standards are measured for 10 min, 5-6 h after the irradiation. Calculations. By comparison of the 889.3-keV 46Scpeak area in the spectra of the tracer solution and of the eluate, the yield of each chemical separation is determined. In the residue both %c and 61Cuoccur. From a 61Cuspectrum, obtained with a-irradiated nickel, the ratio of the area of the 372.9-keV peak to the sum of the areas of the 283.0- and 656.0-keV peaks was determined under the experimental conditions used. This ratio (0.0861 & 0.0022) is used to correct for the contribution of 61Cuto the 372.8-keV peak in the residue. Since the measuring geometries of eluate and residue are different, two calibration curves are established. The first one corresponds to the eluate and is obtained by measuring the calcium carbonate standards, dissolved in 1 M hydrochloric acid. The second one corresponds to the residue and is obtained by
ANALYTICAL CHEMISTRY, VOL. 57, NO. 13, NOVEMBER 1985
b , ,
,
,
,
I
,
I
2551
t l
I
I
I
SAMPLE
A: -30%
-8
lool H
SLANl
L
Carborundum belt grinding
u
Silicon carbide disk grinding
Diamond paste polishing
300
Flgure 1. Part of the 28 scan of a blank sample and of a nodular cast iron sample: (a) excitation, tungsten tube; diffraction, LiF (200); (b) excitation, chromium tube; diffraction, LiF (200); (c) excitation, chromium tube; diffraction, PET.
measuring the calcium carbonate standards, homogeneously distributed over a filter crucible. The calibration curve gives asis-’ (1 - e-Xtha)-las a function of the energy. For the incident energy of the sample the corresponding asis-’ (1- e-htlmjS)-l ib obtained from the calibration curve and used to calculate the concentration by
where cx and cs are the calcium concentrations in the sample and the standard, ax and as are the count rates at the end of the irradiation in the 372.8-keV peak of 43Sc,for the sample and the standard, ix and is are the beam intensities deduced from the BGGa activities in the monitor foils, X is the decay constant of 43Sc,th3 and tirrBare the irradiation times, Rx(E) and &(E) are the aparticle ranges at energy E for the sample and the standard, EI is the effective incident energy, and ET is the threshold energy for the %a(~x,p)~~Sc reaction. Rx(E) and R,(E) were calculated by numerical integration from the stopping powers given by Ziegler (18). The correction factor [ R ~ ( E-I &(ET)] ) / [Rx(EI)- Rx(ET)] amounts to 0.072. The calcium concentrations corresponding to the 43Sc activity in the residue and in the eluate are calculated separately. Their sum gives the calcium concentration of the sample.
RESULTS AND DISCUSSION X-ray Fluorescence Spectrometry. Excitation Tube. For the excitation of Ca Ka,a chromium tube (Cr Ka = 5.41 k e y Cr KP = 5.95 keV) is more suited than a tungsten tube, as appears from Figure la,b. The Ca K a net intensity ( I p ) obtained with the chromium tube is 2.9 times higher. In addition, the background ( I B ) , which largely originates from scattering of the continuum radiation on the iron, is 20% lower, resulting in a 3.2 times higher Ip/IB112 with the chromium tube. Diffraction Crystal. Figure lb,c gives parts of the spectra, corresponding to the same energy range, of a “blank” cast iron (H7-1,O.g wg/g of calcium) and of a nodular cast iron sample (ca. 30 hg/g of calcium) obtained with LiF(2OO) and PET diffraction crystals. As a result of the better dispersion, diffraction with LiF(2OO) allows a mote accurate background correction. The lower background of the diffracted spectrum and the good reflection of the Ca K a radiation are further advantages of the LiF(200) crystal. The Ca K a radiation in the spectrum of the “blank” is largely due to calcium impurities in the X-ray tubes. Surface Treatment. Figure 2 shows that the Ca K a intensity decreases when the samples are ground with silicon carbide papers or with diamond pastes of decreasing grain size. Below 100 wm (=150 grit) a constant intensity is observed. Homogeneity of the Castings. Initially, 6 mm thick plate castings measuring 150 X 200 mm were poured. Investigation by X R F of the calcium concentration in the upper and lower
t 0 ’
I
I
3011
200
I 100
1I
I
0 Grain size l p )
Flgure 2. Ca K a intensity for a nodular cast iron sample, after different polishing steps.
Table V. Calcium in Cast Iron Determined by a Particle Activation Analysis material
results, y g / g
std deP
mean
std devb
LD-1
13.16 12.74 13.93 14.07 0.90 19.33 14.12 14.83 19.02 16.81 13.31 13.25 17.58 18.54 37.9 39.0 43.1 34.8 9.98 9.97 24.31 23.53
0.84 0.72 0.73 0.77 0.28 0.78 0.53
13.48
0.63
16.1
2.8
17.9
1.6
13.28
0.04
18.06
0.68
38.48
0.81
38.95
5.87
9.98
0.01
23.92
0.55
H7-1 H8-1 H8-2 H9-1 H9-2 H9-3 Hll-3 H16-1 H16-2
0.61
0.76 0.61 0.57 0.51 0.57 0.72 1.0 1.2 1.5 1.3 0.82 0.80 1.55 1.13
Statistically expected standard deviation calculated from counting statistics on the measurement of 43Sc(residueand eluate) and on the measurement of esGa in the monitor foil. Experimental standard deviation. part of such plates showed differences up to 20%, indicating segregation of the calcium during solidification. For this reason, cylindrical 60 mm diameter bars were prepared in sand moulds. T o investigate their longitudinal homogeneity, 10 samples taken over a distance of 150 mm were analyzed. The standard deviation was comparable to the statistical uncertainty, namely, 1.5%. I n diametrical direction a standard deviation of 3.2% was found, which can still be considered as satisfactory. Activation Analysis. The 43Sc activity in the residue corresponds to 5-29% of the total calcium concentration. Table V summarizes the results. The material LD-1 was analyzed to test the a activation method. Four analyses yielded a relative standard deviation of 4.7%. For H8-1, H8-2, and H l l - 3 , the experimental standard deviation was significantly higher than the one expected from counting statistics.. Calibration Graph. In Figure 3 calcium concentrations for the materials H7-1 to H16-2 are plotted as a function of the Ca K a intensity. I t appears that the calibration graph is linear, but does not pass through the origin, probably due t o the presence of calcium impurities in the X-ray tube. If
2552
Anal. Chem. 1985, 57,2552-2555
by the weighted least squares method. On the basis of Currie's definition (19) a detection limit of 0.7 pg/g calcium is calculated for measuring times of 200 s at the position of the peak and at each background position.
ACKNOWLEDGMENT Grateful acknowledgement is made to Ir. Lietaert (WTCM) for preparing the samples and to J. Hoste for the interest taken in this research. Registry No. Ca, 7440-70-2; cast iron, 11097-15-7.
LITERATURE CITED
1 , 100d m200 0 I'
I
I
I
I
300
I
LOO
I
I
A
500 600 XRF l C o u n t s l s l Ca KN
Figure 3. Calibration graph.
it is assumed that the intensity can be determined with a negligible standard deviation and that the standard deviation on the calcium concentration determined by charged particle activation analysis i s the same for all measuring points, the best estimate of the linear functional relation between the Ca Ka intensity ( x ) and the calcium concentration (y) is y = -8.27
+ 0.0872~
(2)
The result of 43.1 pg/g for H l l - 3 was not considered, since the chance for such a large deviation to occur was less than 2%. The standard deviation of the regression coefficient is 0.0033 pg s/g. The 95% confidence interval for the calcium concentration deduced from the Ca K a intensity by means of eq 2 is f10.8%, *3.9%, f3.9%, and f4.6% a t 10,20, 30, and 40 pg/g, respectively. The 95% confidence limits for a given Ca Ka intensity are given by the dotted lines in Figure 3. Similar results are obtained when the straight line is fitted
(1) Lux, B. Mod. Cast. 1964, 4 6 , 222-232. (2) Muzumdar, K. M.; Wallace, J. F. AFS Trans. 1973, 81, 412-423. (3) Lalich, M. J.; Hitchings, J. R. AFS Trans. 1978, 8 4 , 653-664. (4) Jacobs, M. H.; Law, T. J.; Melford, D. A,; Stowell, M. J. Met. Techno/. (London) 1976, 3 (March),98-108. (5) Church, N. L.; Schelling, R. D. AFS Trans. 1970, 76, 5-8. (6) Karsay, S.I.; Campomanes, E. AFS Trans. 1970, 78, 85-92. (7) Scholes, P. H. Analyst (London) 1988, 93, 197-210. (8) Taylor, M. L.; Beicher, C. B. Anal. Chim. Acta 1969, 4 5 , 219-226. (9) Headrldge, J. B.; Richardson, J. Ana/yst (London) 1969, 94, 968-975. (IO) Samsoni, 2. Microchim. Acta 1978, 11, 177-190. (11) Yakovlev, P. Y.; Zhukova, M. P. Zavod. Lab. 1970, 3 6 , 1169-1173. (12) Sobkowska, A.; Basinska, M. Microchim. Acta 1975, 11, 227-234. (13) Fu, B.; Ottaway, J. M.; Marshall, J.; Llttlejohn, D. Anal. Chim. Acta 1984, 161, 265-273. (14) Kuemmel, D. F.; Karl, H. L. Anal. Chem. 1954, 26, 386-391. (15) Goto, H.; Ikeda, S.; Klmura, J. J . Jpn. Inst. Metals 1958, 22,
185-187.
(16) Atsuya, I.; Goto, H. Specfrochim. Acta 1971, 26,359-367. (17) Vandecasteele, C.; Strljckmans, K. J . Radioanal. Chem. 1980, 5 7 , 121-136. (18) Ziegler, J. F. "Helium, Stopping Powers and Ranges in all Elemental Matter"; Pergamon: New York, 1977. (19) Currie, L. A. Anal. Chem. 1968, 4 0 , 588-593.
RECEIVED for review April 5, 1985. Accepted June 20, 1985. The investigation is part of a research programme sponsored by IWONL (Instituut voor Wetenschappelijk Onderzoek in Nijverheid en Landbouw). Financial support was received from the IIKW (Interuniversitair Instituut voor Kernwetenschappen) and the NFWO (Nationaal Fonds voor Wetenschappelijk Onderzoek).
Chemiluminescence Method for Direct Determination of Sulfur Dioxide in Ambient Air Danian Zhang,' Yasuaki Maeda,* and Makoto Munemori Laboratory of Environmental Chemistry, College of Engineering, University of Osaka Prefecture, Mom-umemachi, Sakai 591, Japan
Sulfur dloxlde enhances the chemllumlnescence reaction of lumlnol with NO,. The enhanced signal is proportlonal to SO, concentration at a flxed NO, conoentratlon. Based on thls gas/llquld phase chemllumlnescence reactlon, a rapid and sensltlve method for the determlnatlon of SO, Is proposed. A 05% response is obtained wlthln 2 mln. Relatlve standard devlatlons for 10 ppb and 1 ppb of SO, are 0.9% and l o % , respectlvaly, and the detectlon llmlt Is approximately 0.3 ppb. By the present method, real tlme determlnatlon of SO, In amblent alr can be made.
'On leave from East China Institute of Chemial Technology, Department of Environmental Engineering, Meirong Rd, Shanghai, People's Republic of China.
Sulfur dioxide, one of the major air pollutants, is usually determined by the pararosaniline method after absorption of SO2 from air in a solution of potassium tetrachloromercurate (1)or by the conductometric method after absorption of SOz from air in a dilute sulfuric acid solution containing a small amount of hydrogen peroxide (this method is recommended in Japan as a standard method for the determination of SO2 in ambient air). These methods give only the average concentration of SOz during the sampling time (30-60 min) and besides they have following drawbacks: in the former, the toxic mercury compound causes environmental problems, and in the latter, any acid or alkaline gas interferes. Sulfur dioxide in the gas phase can be directly determined by a flame photometric method based on the chemilurnineseence reaction
0003-2700/85/0357-2552$01.50/00 1985 American Chemlcal Society