Flame Spectrophotometric Determination of Iron in Siliceous Materials JOHN A. DEAN and J. C. BURGER, JR. Department o f Chemistry, University o f Tennessee, Knoxville, Tenn.
This investigation describes the application of the Beckman Model DU flame spectrophotometer to the rapid determination of iron in siliceous materials such as alumina refractories and limestone. The arc emission line at 386.0 mp was used to measure the iron radiation. Cobalt was incorporated in the samples in order that the cobalt 387.1-mp arc line could serve as an internal standard. An oxyacetylene flame w-as the excitation source. The radiative interference of many elements normally associated with iron in siliceous samples could be corrected by using an internal standard. Kone of the alkalies or alkaline earths offered interference when present in the amounts usually encountered, and large concentrations of aluminum can be tolerated. Magnesium offers serious interference, and when present, the separation of iron as the hydrous oxide is necessary. The optimum concentration range for iron is between 25 and 200 p.p.m. The standard deviation from the mean of replicate samples is approximately 3%. The method has been successfully applied to limestone and several varieties of alumina refractories.
T
HIS investigation describes the application of the flame spectrophotometer to the rapid, routine determination of iron in siliceous materials such as alumina refractories, glasses, and limestone. The method should be of particular interest to those laboratories which are desirous of a rapid method for the analysis of iron in these types of materials wherein the iron content embraces the concentration range of many colorimetric procedures. Determination of iron by flame spectrographic methods has been discussed by Mitchell (12) and McClelland and Whalley (11), and has been applied to agricultural materials, water, and blood by Griggs (Y), and Griggs, Johnstin, and Elledge (8). These workers used photographic recording in conjunction with a spectrograph. Gerber, Ishler, and Borker (6) have reported orally on the use of the Beckman flame spectrophotometer for the determination of iron. Therefore, a method employing the modern flame spectrophotometers would be desirable in addition to the restudy of some of the variables encountered when using the newer instruments and the integral atomizer-burners. EXPERIMENTAL
Apparatus. -4 Beckman Model DU spectrophotometer with Model 9220 flame attachment and photomultiplier unit was used. An all metal atomizer-burner unit, supplied with the flame attachment, was used as the excitation source. The gases chosen were oxygen and acetylene, largely because of availability, although the slightly higher excitation energy available from the oxyacetylene flame, as compared with an oxyhydrogen flame, was a prominent consideration. Reagents. A standard solution of iron, 1.00 ml. equivalent to 10.0 mg. of iron, was prepared by dissolving 10.0 grams of iron wire of known purity in 115 ml. of 6N hydrochloric acid and diluting to 1 liter with demineralized water. Weaker standard solutions were prepared by appropriate dilution with demineralized water. Sufficient hydrochloric acid should be added to the standard solutions to adjust the hydrogen ion concentration to approximately 0.3N. A standard solution of cobalt, 1.00 ml. equivalent to 14.73 mg. of cobalt, was obtained from the Burrell Corp., Pittsburgh, Pa.
A weaker standard solution, 1.00 ml. equivalent to 2.00 mg. of cobalt, was prepared by transferring 67.9 ml. of the foregoing solution to a 500-ml. volumetric flask, and diluting to the mark with demineralized water. The latter solution was employed for all samples composited with cobalt added as an internal standard. Demineralized water, used exclusively in preparing all solutions, was prepared by passing ordinary distilled water through a bed of Amberlite MB-3 resin. Flame Spectrophotometer Settings. The instrument settings used to measure the iron luminosity were as follows: Sensitivity control, turns from clockwise limit Selector switch Phototube resistor, megohms Slit, mm. Acetylene, Ib. per square inch Oxygen pounds per square inch
9 (burner rated at 10
pounds per square inch by manufacturer) Different aspirator-burners, even though of similar construction, do not necessarily reproduce these tabulated luminosities when employing these operating conditions. Differences in diameters of the oxygen and capillary orifices, and tiny obstructions in or around the oxygen orifice, affect not only the flow of oxygen, but also the rate of aspiration of the solution under eyamination. These factors alter the flame temperature, and consequently, affect both the flame background and the iron luminescence. The burner should be cleaned frequently to remove the carbon deposits which tend to accumulate. Flame Spectra of Iron. The flame spectra of iron were determined by aspirating an aqueous solution containing 100 p.p.m. of iron into the flame. Figure 1 gives the major emission lines of iron that are found in the outer flame mantle in the region suitable for the flame photometric determination of iron. For quantitative analysis an iron line is needed which possesses sufficient luminosity to enable iron to be determined in refractory type materials wherein the iron content ranges between 0 and 200 p.p.m. Although this study revealed numerous iron lines in this region, only four emission lines possessed sufficient intensity to be considered suitable for quantitative analysis. Table I gives the wave length of each of these four lines, together with the relative intensities of each. The 374-mp peak is actually an unresolved grouping consisting of three closely spaced lines for iron 374.56, 374.59, and 374.8 mp. Such an unresolved grouping is unsuited for quantitative work. Of all the emission lines the 372.0- and the 386.0-mp lines were chosen initially for investigation because they were more sensitive. Study of the 372.0-mp line was soon abandoned u hen this line was found to suffer severe interference from strong magnesium
Table I.
Major Flame Emission Lines of Iron
MP
Relative Intensity of Line, P.P.M. of Irona
372.0 373.7 374.7 386.0
1.0 1.8 2.5 1.3
Wave-Length Line Peak,
a
1052
5 to 6 0 1 22 0 050 4 5
Expressed as minimum quantity detectable for a slit width of 0.050 mm.
V O L U M E 27, NO. 7, J U L Y 1 9 5 5
1053
band systems occurring between 36T to 373 nip. The 372.0-mp iron line suffers a further limitation in that it is a resonance linethat is, its emission arises from an electron falling back to t h e lowest energy level (ground state) from the next upper electron level. A resonance line is more prone to enhibit self-absorption than are other spectral lines of the same element. For these reasons. subsequent efforts were confined to the 386.0-m~line. Also noticeable in Figure 1 are several very weak iron emission lines. For slit widths of 0.050 mm , approximately 16 discrete iron enission lines can be resolved from 335 to 395 mp ( 0 ) . The remainder of the background radiation in the vicinity of the iron lines is essentially continuous and is attributable to the continuous spectrum of the carbon monoxide flame and the superposition of innumerable feeble iron lines.
was adjusted such that, on final dilution, the concentration n-ould be approximately 0.3M. The luminosity of the background, H , a t a point near the base of the analysis line, was subtracted from the total luminosity reading of the individual standards, L, at the line peak (386.0 mp) t o give net relative luminosities. The background reading should be taken a t the base of the 386.0mp iron line on the long nave-length side. This wave length minimizes the overlap of any magnesium band remnants and also serves as a suitable background point for the cobalt 387.1-mp line when an internal standard is added to the sample. The net relative luminosities were plotted against the conrentration of iron in the standard solutions to give the calibration curve, Figure 2.
OFTIMU31 INSTRUMENT SETTINGS
Fuel and Oxygen. The pressures chosen were 4.5 pounds per square inch of acetylene and 9 pounds per square inch of oxygen for a relatively wide orifice burner, and 5 pounds per square inch of acetylene and 18 pounds per square inch of oxygen for a burner with a narrow orifice. These burners XTete rated a t 10 and 19 pounds per square inch of oxygen pressure, respectively, by the manufacturer. The gas pressures were arrived at by varying the oxygen and acetylene presmres independently. Although higher acetylene-oxygen ratios gave larger luminosity readings, the background also increased proportionately and there was little over-all gain in sensitivity. These pressures were used throughout the remainder of the investigation because the ensuing flame permitted iron to be determined with the desired sensitivity. Slit Width. The slit width used was 0.050 mm. At this slit width the effective line width of the iron 386.0-mp line and the cobalt 387.1-~npline was sufficiently narrorT to enable both lines to be completely resolved from each other when cobalt was added to the sample solution as an internal standard. The cobalt 387.1-mp line is shown as the dashed curve in Figure 1. Cowan and Dieke ( 3 )have pointed out that slit widths of 0.030 mm. or larger should be used in cases where there is a possibility of selfabsorption. However, no perceptible self-absorption was noticed for iron for concentrations up to 600 p.p.m. CALIHRATIO\
CURVES
L minus H Method. Standards were prepared by appropriate dilution of the standard stock solution of iron with demineralized water. The hydrochloric acid concentration in these solutions
;I-
0 80
/
/
Figure 2.
/I
I
I
Calibration curve, L minus H method
Enaniiuation of the calibration curve discloses the fact that the straight-line relationship could be extrapolated through the zero ppint. Theoretically, then, very low concentrations of iron could be determined if excitation conditions could be adjusted accordingly, or if sufficient instrumental amplification were available. However, the experimental results for amounts of iron less than 25 p.p.m. were not reproducible sufficiently for accurate quantitative work. Consequently, the recommended sample, or aliquot portion thereof, should be chosen so that it contains at least 25 p.p.m. of iron. The instrument was calibrated by a series of standards run before and after analysis. Internal Standard Method. The second type of calibration curve was obtained by the method of internal standardization as proposed by Cholak and Hubbard ( 8 ) . I n this method intensity ratios are measured-that is, the ratio of the net relative luminosity of the analysis line to that of an element added to the sample in constant amount, the latter being the internal standard. Such a pair of lines are referred to as an analysis pair. T h e advantage in using an internal standard is that it affords compensation for many factors, most of which are very difficult to control in practice. These include variations in flame temperature, fluctuations in fuel or oxygen pressure, and interference from other elements. Ahrens (1) has listed several factors which must be considered when choosing an internal standard:
It should be an element which has a negligibly small concentration in the analysis specimen. Internal standard and analysis lines should have similar excitation potentials. The internal standard line should be free from self-absorption Analysis and internal standard lines should be roughly the same wave length.
: 370
375
WAVE
Figure 1.
-----
380
385
390
LENGTH, mr
Flame emission spectra of iron
Cobalt 387.1-mp l i n e for 400 p . p . m . present Concentration, 100 p.p.m. iron a n d 0.050-mm. slit width
The cobalt 387.1-mp line satisfies all of these conditions and was used in this investigation. KO self-absorption was noticed
ANALYTICAL CHEMISTRY
1054 for either the iron 386.0-mp line or the cobalt 387.1-mp line. This may have resulted from use of slit widths of the order suggested by Cowan and Dieke ( 3 ) for minimizing this trouble. The calibration curve was obtained by the analysis of a series of standard iron solutions, each containing a fixed amount of cobalt. The appropriate background readings were subtracted from the individual iron 386.0-mp and cobalt 387.1-mp luminosities to obtain net relative luminosities. Then, the ratio of the net relative luminosity of the iron line to that of the cobalt line was plotted versus the respective iron concentration on log-log graph paper. A straight-line calibration curve results. Working curves showed a slight irregular drift which made it necessary to prepare a fresh calibration curve for each series of samples. Influence of Acids (including anions). The effect of various anions was determined, particularly those which might be introduced during sample dissolution. A series of several concentrations of each acid, or of the anions as their ammonium salts, was prepared with 100 p.p.m. of iron present in each of the samples. The results are tabulated in Table 11. Bcetate and perchlorate ions were found to be particularly detrimental in the determination of iron. The results obtained for hydrochloric acid are in agreement with the spectrographic results reported by Ells ( 4 )
2o 0
t
365
370
I
375
I
I
380
385
390
LENGTH, rnp
Flame band spectra of magnesium
Arrows indicate peak wave l e n g t h s of iron 386.0-mp a n d cobalt 387.1-mp l i n e s Concentration, 500 p . p . m . m a g n e s i u m a n d 0.050-mm. slit width
Table 11. Effect of Anions as Their Acids on Flame Photometric Determination of Irona Anion Concn., Added Acetate b Chloride
.v
1.0 2.0 3.0
1.0 2.0
3.0 Fluoride S i t ra t e Perchlorate Phosphate
Sulfate b
b C
1. Elements xhich offer no interference. These include boron, sodium, titanium, vanadium, and the ammonium ion. 2. Elements exhibiting general background radiation. This occurs with many of the elements when present in relatively high concentrations. This type of interference vias compensated in both methods of measurement employed. 3. Coincidences, or near coincidences, such as the overlap of the magnesium band system centered about 383 mp with the iron 386-mp line, as shown in Figure 3. Manganese exhibits a similar type of interference owing t o its band spectra. 4. Interference with combustion processes was observed for aluminum, lithium, and zinc. However, through use of the internal standard procedure, good r:sults were obtained for iron in the presence of these elements. a. General interference a t either the background or the peak of the iron or cobalt lines is observed for calcium, barium, and potassium when these elements are present in amounts exceeding 50 p.p.m. iMETHOD OF PROCESSING SAWPLES
I
WAVE
Figure 3.
using a spray chamber type atomizer. Ells found no interference up t o 2 M solutions of the acid but an error of 475 for 3M hydrochloric acid solutions. When acids must be introduced, nitric or hydrochloric are recommended so long as their maximum concentration does not exceed 2.5M. Influence of Other Elements. A major part of the experimental work vias concerned with determining radiation interferences caused by the various elements normally associated with iron in alumina refractories, glasses, and limestone. For each element tested a series of solutions was prepared containing several known concentrations of the suspected interference and generally 100 p,p.m, of iron. Readings were taken a t the peak of the iron line, 386.0 mp, a t the background minimum between the iron and cobalt peaks, 386.5 mp, and at the cobalt line peak, 387.1 mp. The amount of iron found in each sample was then calculated by two methods: the L minus H method and the internal standard method. Analyses were not repeated for the internal standard method when the L minus H method proved satisfactory. Table I11 gives the results obtained by each method. From these studies the elements can be grouped into five categories:
50 C 100 500 1.0 2.0 3.0 1.0 2.0 3.0
l0OC 500 1000 5000 1.0 2.0 3.0
100 p.p.m. of iron present in all cases. Values obtained by t h e internal standard method. Concentration in p.p.m.
Iron Found, P.P.M. a t 386.0 mp 103 109 108 100 99 99 99 101 101 100 101 100
98 9G 92 104 98 97 101 100 101 96
For siliceous minerals and glasses, weigh samples containing 0.9 to 7.1 mg. of ferric oxide into platinum crucibles. Moisten with demineralized water, then add 10 ml. of 48% hydrofluoric acid and 0.5 ml. of 3 6 s sulfuric acid. Heat to fumes of sulfuric acid. Cool and then add an additional 5 ml. of hydrofluoric acid and evaporate to dryness. Ignite carefully until fumes of ~u1furic acid cease to be evolved. Cool and add 10 ml. of 62Vhydrochloric acid. Warm until the sample dissolves. If any residue remains, repeat this treatment. Transfer the entire sample to a 25-m1. volumetric flask. I d d 5.00 ml. of the standard cobalt solution and dilute to the mark with demineralized m-ater. Aspirate the samples and measure the luminosities a t 386.0, 386.5, and 387.1 mp. Read the amount of iron resent from the calibration curve. d e n considerable amounts of magnesium, calcium, barium, or potassium are either k n o m or suspected to be present in the sample, generally amounts exceeding 50 parts per 1000 parts of sample, a preliminary separation of iron is necessary. To the solution obtained following the dissolution of the sample, add a feTr drops of 30% hydrogen peroxide, and then add a considerable excess of filtered, concentrated ammonia. Stir the solution thoroughly and heat to boiling. Filter off the hydrous ferric oxide and wash it with hot water. Dissolve it in 10 ml. of 1.1hydrochloric acid and wash the filter with demineralized water. Collect the filtrate and washings in a 25-ml. volumetric flask and proceed as described. DISCUSSION
Among the usual constituents of siliceous materials several elements offer serious interference to the flame spectrophotometric determination of iron. Large amounts of all the alkaline earths and the alkalies, with the exception of sodium, enhance the flame emission of iron. The cause perhaps is the altering of temperature of the oxyacetylene flame by these elements. It has been shown ( 5 ) that the population of an electronic state of low energy is determined mainly by thermal processes around a flame temperature of 2000" K. Most of the elements exhibit their full
1055
V O L U M E 2 7 , NO. 7, J U L Y 1 9 5 5 ~
~
~~
Table 111. Influence of Cations on Flame Photometric Determination of Iron (100 p.p.ni. of ’ iron present in all cases) Iron Found, P.P.11. Internal L - H standard Concn., method method P.P.M.
Cation Tested
100 500 1000 5000 100 500 1000 5000 100 500 1000 5000 100 500 1000 2000 100 500 1000 5000 100 500 1000 100 500
104 99 98 84 100 100 100 100 138 141 122 90
101 102 102 97
96 ..
97 91 89 88
Manganese
100 300
104 112
Potassium
!OO
118 116 98 102
Aluminuni
Ammonium
Barium
Boron
Calcium
Lithium Magnesium
000 1000 5000 500 1000 5000
Sodium
100 500 1000 5000 50 100 500 100 500 1000
Strontium
Vanadium Zinc
Table IV.
.. .. .. .. 93 93 96 100
106 108 104 116 10.5 85 93 110
100 99 92
..
.. ..
..
*
92 90
.. ..
102 106 103 101 106 105 off scale 99 103 101
,.
..
..
..
..
92 92 88
99 100 99
Analysis of Bureau of Standards Samples Samples
Compn. Argillaceous limestone l a 29 C a , 7 Si, 2 Al. 1 bIg
3 0 . 7 analyzed
7
Iron with Found with Associated Stand- Associated Standa r d Deviation‘ ard Deviation Certified % 70 1.14b 1 . 1 8 =t0 . 0 4
+ 0.001
Borosilicate glass 93 38 Si, 4 B , 3 N a , 1 A1 Burnt refractory 76 26 Si, 20 Al, 1 Ti, 0.5 lfg
6
0.053 zt 0 , 0 0 3
6
1.66 i0.06
1 . 7 6 zt 0 . 0 6
Burnt refractory 78 37 AI, 10 Si, 3 K , 2 Ti, 0 . 5 RIg
5
0 . 5 5 =k 0 . 0 8
0.60
Silica brick 102 4 4 Si, 1 Al, 1 C a
5
0 . 4 6 =k 0.03
0 . 4 8 3= 0 . 0 2
a b
-1 Standard deviation = if2(10)2where n - 1 Certificate of analysis missing.
D
0.052
+ 0.04
= deviation from mean.
complement of arc emission lines in the inner cone of an ovyacetylene flame wherein the flame temperature is much higher. This would certainly be true for the readily excited elements among the alkalies and alkaline earths. The maximum temperature for an oxyacetylene flame is not attained immediately above the inner cone, but several millimeters above the tip of the inner cone. Thus, the persistence of free atoms in a highly evcited state is probably the cause of the delay in reaching the normal temperature characteristic of the flame mantle surrounding the inner cone. Combustion processes proceed mainly by bimolecular reactions, but the recombination of free atoms is thought to require three-body collisions (IO),and would therefore be less
rapid. The normal flame temperature of the outer mantle cannot be attained until after the completion of the large heat relrase due to the recombination of free atoms formed in the reaction zone of the inner cone. If the recombination occurs with certain atoms-for example, calcium-then it might result in a form of chemiluminescence and result in a higher temperature in the outer mantle and consequently an enhancement of the iron luminescence. The internal standard method circumvents this type of interference and enables iron to be determined in the presence of the normal amounts of the alkalies and alkaline earths as would be encountered in alumina refractories and clays, and glasses. Seldom do these elements exceed 30 t o 100 parts per 1000parts of sample. Aluminum is also a serious interference and does not exhibit any spectral lines or bands of its own in the vicinity of the iron and cobalt lines. When it is present in an oxyacetylene flame, aluminum diminishes the intensity of iron and cobalt lines. The decrease in the intensity of the iron and cobalt spectral lines, as a result of the introduction of aluminum into the flame, is presumably due to the absorption of part of the energy by the aluminum atoms, similar to the effect of aluminum on the flame spectra of the alkaline earths ( I S ) . However, the attenuation affects the two lines equally, so that satisfactory results are obtained for iron in the presence of large amounts of aluminum when the internal standard method is employed. Amounts of magnesium in excess of 50 p.p.m. proved to be a serious interference regardless of which measurement method was employed. The cause is the overlap of the vibrational bands on the long wave-length side of the magnesium 383-mp band system with the base of the iron 386.0-mp line. The overlap is more pronounced for the nearer iron line than for the internal standard cobalt line which precludes use of the internal standard principle. When magnesium is encountered in samples, it is necessary to carry out a preliminary separation of iron as the hydrous oxide or to incorporate an equivalent amount of magnesium in the standard samples. Of particular interest is the fact that phosphorus offers no interference, a t least in concentrations up to 5000 p.p.m. Table IV summarizes the results obtained on Bureau of Standards samples. The reproducibility of the flame photometric results was very good. The standard deviation from the mean of replicate samples was approximately 3%. LITERATURE CITED
(1) Ahrens, L. H., “Spectrochemical .4nalysis,” Addison Kesley Press, Cambridge, Mass., 1950.
( 2 ) Cholak, J., and Hubbard, D. LI., ISD. ENG. CHEM..A s ~ L . ED., 16, 726 (1944). (3) Cowan, R.D., and Dieke. G. H., Revs. M o d e r n Phys., 20, 418 (1948). (4) Ells, V. R., J . O p t . SOC.Amer., 31, 531 (1941). ( 5 ) Gaydon, 8 . G., and Wolfhard, H. G., Proc. Roy. SOC.(LOndo7L), A205, 118 (1951). (6) Gerber, C. R., Ishler, K. H., and Borker, E.. Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, 1951. (7) Griggs, 11.h.,Science,89, 134 (1939). (8) Griggs, 2.1. A., Johnstin, R., and Elledge, B. E., IND.ESG. CHEM.,ASAL.ED., 13, 99 (1941). (9) King, W. H., Jr., and Priestley, W., Jr.. Am. SOC.Testing 1Iaterials, Tech. Pub. 116, 100 (1951). (10) Laidler, K. J., and Shuler, K. E., Cheni. Reus., 48, 154 (1951). (11) McClelland, J. -4.C., and Whalley. H. K., J . SOC.Chem. Ind. (London),60,288 (1941). (12) Mitchell. R. L., Ibid.,55, 267-9T 11936). (13) lIitchell, R. L., and Robertson, I. M.,Ibid.,55, 269T (1936).
RECEIVED for review J a n u a r y 21, 1955. Accepted March 30, 1955. Contribution 145 from t h e Department of Chemistry, University of Tennessee. Abstracted in part f r o m a thesis submitted b y J. C . Burger, Jr., t o t h e Graduate School of t h e University of Tennessee in partial fulfillment of t h e requirements f o r t h e degree of master of science, 1954. Work supported in part by the Atomic Energy Commission under contract No. AT-(40-1)-1337.
