Spectrographic Limit of Detection of Phosphorus, Titanium, and Zirconium in the Direct Current A r c DANIEL P. NORMAN
AND
WILLIAM W. A. JOHNSON
N e w England Spectrochemical Laboratories, West Medway, Mass.
The spectrograms were taken in the first order of a 3-meter grating spectrograph, dispersion 5.6 d. per mm., on Eastnian 33 plates. The plates were developed for 5 minutes in Eastman D11 developer a t 18" C. When photographing the wave-length range above 3200 A., the overlapping higher orders of the gratin;: were absorbed by a Corning No. 738 glass filter placed before the slit of the spectrograph. A slit width of 15 microns was used. An enlarged image of the arc was formed on the slit of the spectrograph with a quartz lens and was so adjusted that the image of the cathode fell just off the edge of the Hartmann slide delimiting the slit length. The standard samples issued by the National Bureau of Standards offered a convenient source of analyzed samples of a wide range of composition and were used in these studies. 1Iany of the older Bureau of Standards standards contain trace amounts of elements which are not listed in the certificates of analysis.
The spectrographic limits of detection of phosphorus, titanium, and zirconium have been investigated in a series of National Bureau of Standards standard samples. The samples were weighed directly into a cored graphite electrode which was used as the cathode of a conventional direct current arc operated a t 1 5 amperes. The spectra were observed in the first order of a 3-meter grating spectrograph, per mm., b y a jumping-plate technique. dispersion 5.6 X2553.28 proved to b e the most satisfactory spectrum line for the detection of phosphorus. When the limit of detection of an element was defined as the smallest amount which could always b e detected with certainty, the limit of detection of phosphorus ranged from 0.05 to 0.80 microgram in nine different standards containing from 0.0013 to 0.59% phosphorus. x3371.45 was the most satisfactory for the detection of titanium. The limit of detection of titanium ranged from 0.04 to 4 micrograms in twelve samples containing from 0.0018 to 0.23% titanium. The optimum line for the detection of zirconium proved to b e X3391.98 of the ionized atom because the high temperature needed to volatilize zirconium favors the excitation of the ionized at the cost of the neutral atom. The limit of detection of zirconium was found to vary from 0.5 to 4 micrograms in six samples containing from 0.0037 to 0.19% zirconium. N o correlation was observed between the limit of detection and the percentage of an element in a sample.
A.
The results tabulated below represent the smallest amount of the element that could always be detected with certainty in the given sample. One half of these amounts was frequently detected; one tenth of these amounts was occasionally detected. Samples a s small as 0.2 mg. were arced, and there is no guarantee that such small samples are representative, even when they are National Bureau of Standards standards. I n certain samples (noted in the tables) it was frequently possible to detect one tenth, or occasionally one twentieth, of the amount listed. I t was assumed that these variations were due to inhomogeneities in the samples hut in every case the value listed in the tables represents the amount of the element which could always be detected with certainty. The values listed may therefore well be regarded as upper limits and not as the smallest amount which may be detected under exceptionally favorable conditions.
WHILE
carrying out numerous qualitative spectrographic determinations, in continuation of earlier work, the authors felt a decided lack of reliable information concerning the variation of the limit of detection of the various elements in a variety of matrices. Accordingly, the spectrographic limits of detection of several elements have been systematically determined in a number of different samples. The results for potassium have heen reported (9), the results for phosphorus, titanium, and zirconium :Ire reported in this paper, and the results for other elements will be presented at a later date.
PHOSPHORUS
Considerable confusion exists in the literature concerning the spectrographic sensitivity of phosphorus. Milligan and France ( 8 ) have discussed the situation and presented an adequate bibliography. Kiess, in a survey of the spectrum of phosphorus ( 5 ) shon.ed that . . the $timate rays of phosphorus . . ., lying a t 1774. 1782, and 1787 A,, . . . fall outside the range of ordinary spectrographs and cannot be relied on for makingospectrochemical analyses. The group of four lines near 2550 A. chosen by de Gramont must continue to serve this purpose. . . . They originate in one of the metastable states of the atom and must be designated as penultimate lines. . Harrison ( 2 ) gives the following wave lengths and intensities for these penultimate lines: ji.
TECHNIQUE
The technique used in these investigations has previously been described in detail ( 3 ) . Weighed samples were arced i n a direct current arc operated a t 15 amperes, with a 250-volt input. The voltage across the electrodes varied from 20 to 50 volts, depending on the sample being arced. The interelectrode gap was maintained manually a t 3 mm. as the electrodes burned away, with the aid of an enlarged image formed on a target by ai1 ausiliary lens. The electrodes were O.6-cni. (0.25-inch) diameter spectrographic graphite electrodes manufactured by the Dow C'hemical Co. I n the lower (cathode) electrode a crater 5 nini. wide and 4 mm. deep was drilled by a special fixture. Tlle upper electrode was pointed in a pencil sharpener reserved exclusively for this purpose. For each sample a fresh pair of electrodes was introduced in the arc stand and preburned for 40 s e o n d s to volatilize any trace of impurities introduced in preparing the electrodes. Then a 40 second spectrogram was taken to record any impurities intrinsically present in the electrodes and the cathode was removed and cooled rapidly in a metal Ilock. The cooled cathode was transferred to a weighing block and weighed, the sample was added, and the loaded electrode was weighed again and replaced in the arc stand. The spectrograph shutter was opened, the arc was started by touching the upper electrode to the lower electrode and immediately separating the two, and the spectrograph plate holder was rackgd down every 10 seconds until the sample was completely consumed.
."
Ionization
x
Intensities
2534.01 2535.65 2553.28 2554.93
25 50 40 30
Class I I
I I
Of these lines A2535.65, the most intense, is frequently masked by X2535.60 of iron (intensity 1000). Of the remaining lines X2553.28 is the most intense and is not masked even in ferrou3 samples. The limit of detection of phosphorus was determined a t this wave length in nine samples (Table I). As a general rule the phosphorus lines did not appear on the plates until the third or fourth spectra, and then persisted for several spectra more. Most common phosphorus compounds have rather low boiling points and phosphorus would normally be expected to appear in the first few spectra and not to persist 233
INDUSTRIAL A N D ENGINEERING CHEMISTRY
234
Table
1.
Spectrographic Limit of Detection of Phosphorus at
hP5 5 3 3 8
N.B.S. Standard
No.
Description
P
88
Dolomite Fluorspar Soda lime glass Refined silicon Flint clay Columbium steelO Lead barium glass Carwheel iron" Phosphor bronze'
0.001 0.002 0.004 0.008 0.03 0.035 0.10 0.31 0.59
79 128 57 97 123a 89 122 6% Upper limits. tive (see text).
5007.21 4999.51 4991.07 4981.73 4536.05 4535.92 4535.58 4534.78 4533.24 4305.92 3653.50 3642.68 3635.46 3377.54d 3371.45 3341.88
bands and is a$ost as sensitive as A3371, but it is seriously masked by Fe 3341.90A. in ferrous samples. A4305 was nearly as sensitive as X4999 but proved to be more sensitive to changes in matrix than to changes in the-titanium cooncentration. The group of five lines between 4533 A. and 4536 A. may very well prove to be a sensitive criterion for titanium where a low dispersion spectrograph is used. The authors' experience has been that doublets are sensitive and easily recognizable criteria on low dispersion. This group is not masked by any iron lines. Tabled11 lists the observed limit of detection of titanium at 3371.45 A. in twelve Yational Bureau of Standards standards.
The smallest samples used were not entirely representa-
TBble h
Limit of Detection Microoram 0.43 0.56 0.32 0.28 0.77 0.05 0.80 0.31 0.52
Vol. 17, No. 4
ZIRCONIUM
II. Strongest A r c Lines of Titanium
Harrison 200 200 200 300
Intensity Pearse and Exner and Gaydon Haschek 20 2 20 .. 20 3
?
I 15
'4 100 150 300 _.. 500 300 200 20 and 15 100 100
.. .. .. ..
.. 6 8
20 20 15 15u 15
Kayner 9 10 9 9 6R 8R 8R 9R 10R 10 10R
10
10
..
Ionizatioii Class 1
1OR 9R 8 9R 10r
I
1
I I I I I I I I I I I I I
See note in text about low dispersion spectrographs.
for any great length of time. The observed behavior of the phosphorus lines suggests that phosphorus is present in the sample as high-boiling phosphides, or is converted to such phosphides in the arc, and that these phosphides gradually decompose with the vaporization of phosphrus. TITANIUM
There is little agreement among investigators as to which titanium line is most satisfactory for the detection of small amounts of titanium. Much of the disagreement arises from the fact that the early work was carried out with spark and low amperage arc sources. Meggers (6) on theoretical grounds gives X4981.73 as the ultimate line of titanium, and numerous investigators have confirmed this fact empirically. Unfortunately this line is masked by a heavy carbon band in the graphite arc and cannot be used. When it is important to detect very minute traces of titanium this line can be observed in a copper arc. Harrison (B), Pearse and Gaydon (IO), Exner and Haschek (I), and Kayser (4)have listed the lines given in Table I1 as the strongest arc lines of titanium. The table lists the wave lengths of the linea as given by Harrison, and the intensities given by Harrison, Pearse and Gaydon, Exner and Haschek, and Kayser, in that order. Each of these investigators has used a m e r e n t intensity scale for his measurements: Harrison's scale Pearse and Gaydon's scale Exner and Haschek's scale Kayser's scale
0 to 0 to 0 to 0 to
10,000 10 1000 10
Of these lines, A4981 and A4991 are masked by carbon, and the 3600 A. trio are in the middle of a carbon band system. A4999 is about twice as sensitive as X5007, provided the plates are technically perfect-i.e., of excellent resolving power, free from fuzziness and b a c k g r o u n d - o t h e l s e they are of comparable sensitivity. A3371 proved to be at least twice as sensitive as A4999 and was therefore selected for this study. A3341 is more favorably placed for observation with respect to the carbon
Table IV lists the wave lengths of the lines in the spectrum of zirconium reported to be the strongest in the arc. The introductory remarks about intensity scales made for Table I1 apply with equal force to Table IV. The high boiling point of zirconium and of its carbide (both above 5000" C., 11) makes the behavior of this element in the arc somewhat anomalous. By the time a temperature sufficient to volatilize the refractory zirconium-bearing fraction of a sample has been attained in the arc the excitation is so high that the principal lines of theoneutral element are scarcely, if a t all, in evidence. The 3391 A. line of ionized zirconium can be detected consistently when 2 micrograms or less of zirconium are present in the arc, no matter in what matrix the zirconium is, present, whereas the appearance of the group of lines a t 4700 A. is very erratic and is a function more of the matrix in which the zirconium occurs than of the amount of zirconium present. All the lines lying a t wave lengths greater than 3438 A. are involved in carbon bands, and 3438 A. itself is frequently too heavily enmeshed in the wings of a carbon band to be satisfactory. The authors' observations indicate that A4772 is only one fourth as sensitive as A4739. Table -V lists the observed limit of detection of zirconium a t 3391.98 A. in six Riational Bureau of Standards standard samples. The restriction of the data for zirconium to samples rich in silicon was set by the lack of availability of National Bureau of
TaMe
111.
Spectrographic Limit of Detection of Titanium at
h3311.45
N.B.S. Standard No 79 123a
88 89 122 128 93 112 107 la 57 61 (1
Ti
Fluorspar Columbium steel Dolomite Lead barium glass Carwheel caat iron Soda lime glass Boroailicste %icon ~ a r b i 8 ? ~ NI-Cr-Mo cast iron Argillaceous limestone Refined silicon' Ferrovanadiuma
% 0.002 0.002 0.003 0.006 0.009 0.010 0.016 0.025 0.037 0.096 0.10 0.23
Limit of Detection biicrograma 0.04 0.08 0.08 0.16 0.10 0.08 0.85 0.15 0.30 0.07 4 1.2
Upper limits.
Table h
4772.31 4739.48 4710.08 4688.45 4687.80. 3601.19 3572.47 3547.68 3519.61 3496.21 3438.23 3391.9S0 0
Description
IV. Strongest A r c Lines of Zirconium
Harrison 100 100 60 50 125 400 60 200 100 100 250 300
Intensity Pearse and Exner and Gaydon Haschek 10 10 10
.. .... .... ..1 .. .2. 3 5
Lines recommended by Meggers (6, 7).
10
15
,...
..
10
..
10
Kayser 8 10 10 8 10 6 10 7 8 10 10 10
Ionization Clsss I I I I I I
I1 I I
I1 I1 I1
April, 1945
ANALYTICAL EDITION
235 LITERATURE CITED
Table
V.
Spectrographic Limit of Detection of Zirconium a t A3391 - 9 8
N.B.S. Standard No.
Description
Zr
89 93 57 112 78 97
Lead barium glass Borosilicate glass Refined silicon Silicon carbide Burnt refractory Flint clay
0.004 0.0096 0.025 0.027 0.089 0.18
%
Limit of Detection Microgtama 0.5 0.8 1.2 1.0 1.2 4
Standards samples containing analyzed amounts of zirconium and low silicon. Yet in even so narrow a range of matrices the limit of detection of zirconium varies by a factor of eight. .4 statistical analysis of the limits of detection of phosphorus, titanium, and zirconium failed to show any significant correlation between the limit of detectioli of a n element and the percentage of the element in the sample analyzed.
(1) Exner and Haschek, “Die Spektren der Elernente bei nor-
malen Druck”, Leipeig, Franc Deuticke, 1912. (2) Harrison, G. R., “M.I.T. Wavelength Tablea”, New York. John Wiley & Sons, 1939. (3) Johnson, W. W. A., and Norman, D. P., Astrophys. J., 97,46 (1943). (4) Kayser, H.,“Tabelle der Hauptlinien der Linienspektra”, Berlin, Julius Springer, 1926. ( 5 ) Kiees, C. C., J. R . S . Y ~ TNatl. & BUT.Standards, 8,393(1932). (6) Meggers, W. F.,J. Optical Soe. Am., 31,39 (1941). (7) Ibid., 31, 605 (1941). (8) Milligan, W. E.,and France, W. D., IND.ENQ.CHBM.,ANAL. ED., 13,24 (1941). (9) Norman, D. P., and Johnson, W. W. A.,Ibid., 15,152 (1943). (10) Pearse, R. W. B., and Gaydon, A. G., “Identification of Molecular Spectra”, Appendix, New York. John Wiley & Sons, 1941. (1 1) Richardson, D., “Proceedings of Fifth Spectroscopy Conferences”, p. 64,New York, John Wiley & Sons, 1938. PBEEENTED before the Division of Analytical and Micro Chemistry at the CEEMICAL SOCIETY,New York. N.Y. 108th Meeting of the AMERICAN
Extinction Coefficients of‘ Spectrophotometric Standards
As
Determined with the Beckman Spectrophotometer 1. M. VANDENBELT, JEAN FORSYTH, AND ANN GARRETT Research Laboratories, Parke, Davis & Co., Detroit, Mich.
The Beckman spectrophotometer has been applied to the determination of the extinction coefficients of several absorption standards. Two of these, anthraquinone and salicylaldehyde in ethyl alcohol, have been carefully evaluated for the first time. Series of data on these standards indicate that optimum extinction coefficients were obtained with instrument densities of 0.5 to 1.9. The extinction coefficients obtained for potassium nitrate in water, potassium chromate in 0.05 N potassium hydroxide, and a vitamin A ester in ethyl alcohol, are in good agreement with values obtained with other types of spectrophotometric instruments.
THE
increased
u8e
of spectrophotometric techniques of all
kinds has increased the importance of spectrophotometer calibration, especially through the medium of an absorption
standard. This ensures correct adjustment of the instrument, and permits comparison with the values of other workers. Standardization is especially important in collaborative assays, such as those of vitamin A in fish liver oils. In assays conducted recently by the United States Pharmacopoeia (17) and Vitamin Oil Producers Institute (18)committees, variations of 10 to 20% in extinction coefficients were not uncommon, even among groups using the same type of instrument. (Extinction coefficient is used in this paper to designate both molecular e, which is equal to the density, log Zo/Z, divided by the molar concentration, and which is the density divided by the per cent concentration. These are for cell lengths of 1 cm.) The variations were also great among specialized or abridged instruments, thus emphasizing the need for suitable standudization. As a n absorption standard, potassium chromate in 0.05 N potassium hydroxide has long been used. Both it and potassium nitrate have the desirable properties of purity, accessibility, and stability. However, the absorption maxima of these compounds are not at the wave length of the absorption maximum of vitamin A; therefore, they are not suitable for the calibration of specialized instruments in vitamin A determination. As a result, Morton (10) suggests anthraquinone and salicylaldehyde for this purpose. He points out t h a t these compounds have absorption
maxima a t very ,nearly the desired xave length, but adds that their precise e values have not been determined. To make these determinations, and to compare values from the Beckman spectrophotometer (6) with those already existing for well-known standards, this study was undertaken. A natural distilled vitamin A ester has also been examined. This preparation has the characteristic ultraviolet absorption of vitamin A, and was used because its value was relatively well established. APPARATUS
SPECTROPHOTOMETER. The instrument used was a Beckman Model DU, serial No. D-355. All readings were made using the hydrogen dischar e tube and power unit supplied by the National Technical Laboratories. New hydrogen tubes were inserted a t such intervals that the slit widths for each series remained constant. The sensitivity knob waa set at 8 turns from the clockwise limit. This position falls within the range recommended by the manufacturer and made possible a convenient slit width of 0.50 mm. a t the wave len h of the vitamin A maximum. The wave-length scale was Cali rated with a mercury vapor arc spectrum. ABSORPTION CELLS. A pair of cemented fused quartz cella matched and calibrated by the National Technical Laboratories were set aside for use in this investigation. They were cleaned periodically in warm soap solution. Between determinations, they were rinsed in turn with distilledwater, methanol, and chloroform. In the case of alcoholic solutions, chloroform alone waa used as a wash. Each cell was wiped carefully with a soft cloth before use. T o obtain one density value (log Za/I), readings through &11 absorption maximum were made with 1 mp incrementa. The c e e were then interchanged and the peak was read through again. The maximum densities in each case, corrected for cell dimensions, were averaged to give one density value which, when divided by the concentration, gave the extinction coefficient. The cell interchange served principally as a check on the physical state of the cells, since the values obtained with either cell as the blank were not measurably different. The volume effect of change in cell temperature was investigated. Readings were ordinarily within 1 of a normal room temperature of 25’ C., but there remained the possibility of a slight heating effect due to the hydrogen discharge tube. Thermometer tests indicated that if readings were taken within 0.5 houz after the hydrogen tube was turned on, a rise of not more than 1
f