CuS04,5H20, as given in Figure 5 . The digital recorder weight losses n-ere in good agreement with those obtained from the X - Y recorder curve. For a 25-mg. weight change, the accuracy arid precision of the digital recorder were about *0.57,. The temperature was recorded to an accuracy of about =t1% and a precision of about +0.25%. Because of the 5-second difference in printing times, the weight value will not coincide nith the exact value a t the printed temperature. For a furnace heating rate of 5" C. per minute, the temperature difference bet\\-een the weight values is about 0.4' C. For most cases, this can be neglected or, if desired. a correction factor can be applied. For the size of the samples employed and the weight range of the X axis on the X-Y recorder, the digital recorder is probably more accurate than the former. I n the former type of recorder, the weight axis can be read only t o h 0 . 2 mg., which for a 25-mg. weight change would involve an accuracy to about +170.
ACKNOWLEDGMENT
The helpful assistance of Karner Iiendall is gratefully acknodedged.
56.5 MG.
LITERATURE CITED 16.0CURVE 16.1 DIGITAL
100
200 TEMP.
Figure 5.
t
20.6CURVE 20.5 DIGITAL
300
400
'C.
Thermogram of CuSO4.-
5H20 5' C. per minute furnace heating rate and air furnace atmosphere X.Y. Recorder values Digital. Digital recorder values
This instrument has been in daily use in this laboratory for the past six months.
(1) DuvaI, C., "Inorganic Thermogravi-
metric Analmis." Elsevier, Houston,
Tex., 1953.
( 2 ) Freeman, E. S.,Carroll, B J Phys Chem. 62, 394 (1958). (3) Gordon, S., Campbell, C., L Z ~ ~ ~ . CHEM.32, 271R (1960). ( 4 ) Fen-in, S. Z., J . Chem. E d x . 39, -4975 (1962). ( 5 ) Soulen, J. R.,ANAL.CHERf 34, 136
(1962). (6) Wendlandt, W. W.,Zbzd., 30, 56 (1958). ( 7 ) Wendlandt, W. W.,J . Chem. Educ. 38,571 (1961). (8) Wendlandt, W. W., George, T. D., Horton, G. R., J . Znorg. Nucl. Chem. 17,273 (1961).
RECEIVEDfor review June 25, 1962. Accepted September 24, 1962. Work supported by the Directorate of Chemical Sciences, Sir Force Office of Scientific Research, through Contract AF-AFOSR 23-63.
Precise Automatic Spectrophotometric Analysis in the Low Parts per Billion Range R. D. BRITT, Jr. Savannah River laboratory, E. 1. du Pant de Nemours &
b The sensitivity and precision of spectrophotometric methods were improved by means of automatic instrumentation. Quantitative analyses were automated for the determination in water of chloride, nitrate, nitrite, ferrous, ferric, and ammonium ions with a sensitivity of a few parts per billion. The precision of these methods was 2 to 3% at a concentration of 10 p.p.b. The automated methods are rapid and applicable to continuous in-line analysis as well as batch sample analysis.
P
and accurate ionic analysis of the heavy water (DzO) that is used as a moderator in the Savannah River Plant reactors is required to determine the rate of corrosion of reactor components, the fate of nitric acid used in p H control, and the mechanism of formation of radioactive impurities. The ions of interest are present in the concentration range of 1 to 100 p.p.b. Routine manual methods of analysis are time-consuming, lack the required sensitivity and precision, and are susceptible to trace contamination durHECISE
1728
ANALYTICAL CHEMISTRY
Co., Aiken, S. C.
ing sampling or analysis. The concentration of samples by ion exchange or evaporation into the range where reliable results could be obtained by conventional methods was unsuited for this work. Consideration was given to extending the range of spectrographic, coulometric, potentiometric, isotope dilution, and spectrophotometric techniques. Of these, only the spectrophotometric method appeared to be applicable for the determination of both cations and anions at these low levels. Another reason for this choice mas the availability of a commercial automatic colorimeter, the AutoAnalyzer, manufactured by the Technicon Instruments Corp., Chauncey, N.Y. The AutoAnalyzer is widely used for clinical and biochemical analyses (1, 3, 9)as well as for the determination of silica in boiler water ( I ) , nitrate in fertilizer ( I ) , and phosphate in detergents (6). KO information has appeared in the literature on the use of this instrument for analysis in the range of parts per billion, but the design and method of operation are such that a high degree of reproducibility and sensitivity appeared possible.
INSTRUMENT DESCRiPTlON
The basic Autohnalyzer unit has been discussed (1, 9) and only a description of the components pertinent to parts per billion analysis and the principles of operation is given here. The instrument used consists of a proportioning pump, mixing coils, a filtration unit, a heating bath, and a colorimeter equipped with a 50-mm. cell. The colorimeter output is connected to a range expander and a stripchart recorder. The filtration unit is an AutoAnalyzer dialyzer with a 0.4micron pore size membrane filter for aqueous solutions (Schleicher and Schuell Type A, coarse), substituted for the dialysis membrane. Filtration is accomplished by pumping the sample stream a t a rate of 7.8 cc. per minute. along with air segmentation a t a rate of 0.6 cc. per minute, into the upper plate of the filtration unit. The outlet from the upper plate is connected t o a tube in the pump which has a flow rate of 0.8 cc. per minute. The difference in flow rates forces the sample into the bottom plate a t a rate of 7.6 cc. per minute. The air segmentations serve to wipe the membrane and prevent plugging. These filters mere used for several months mithout plugging.
C Direction of Pumoina
cclrn
0.60
I
I
A I
0 4 P Membrone Fiiter
Sample
7.80
Air
0.42
0.60 I . 60
0.42 0.80
Mixing Coii Proportioning
21W sodium acetate in 2111 acetic acid. Hydroxylamine sulfate, 0.5M. Ammonia Method. HYPOCHLOROUS ACID. Saturate 500 ml. of water with chlorine gas a t room temperature. Dilute 2 to 1 as needed. Do not filter. The stock solution can be kept for several days if refrigerated. Store in a glass bottle. ALKALINE PHENOL.Dissolve 7.2 grams of sodium hydroxide in 100 ml. of water. Cool in an ice bath and add dropwise with stirring into 300 ml. of cold water that contains 16.7 grams of phenol. Refrigerate until needed. ANALYTICAL METHODS
Heating Both
4 8 0 rnp
I
lox
I
t
Discord
Figure 1.
Chloride flow diagram
For the Savannah River Plant, the basic AutoAnalyzer was modified and mounted in a ventilated cart provided with a number of sensing elements which automatically turn off the instrument in the case of tubing failure or overflow of the waste collection system. A moderator f l o ~control box with an overflow sampler prevents reagents that are hazardous to reactor operation from being inadvertently pumped into the reactor. A programming unit allows automatic switching between reactor streams and standards for unattended operation. Analyses can also be performed on batch samples, and gloved ports are provided for manipulation of the samples inside the vented hood. The AutoAnalyzer can be converted from one analytical method to another in approximately 15 minutes.
Ferric nitrate, 0.25M in 3M nitric acid. Nitrate-Nitrite Method. Sodium hydroxide, 0 . 5 M . Copper sulfate stock solution, 1.6 x 10-3~. Hydrazine stock solution, 0.025M. Reduction mixture, prepared fresh daily by adding 65 ml. of copper sulfate stock to 1 liter of hydrazine stock solution. Acetone-water, 1to 2 parts by volume. Sulfanilic acid, 0.013M in 0.72M HCl. Sodium acetate, 1.9Af. 1-Naphthylamine, 0.02111 in 531 acetic acid. Iron Method. 2,4,6-Tripyridyl-striazine. Dissolve 0.3 gram in 1 ml. of concentrated hydrochloric acid and dilute to 1 liter. Sodium acetate-acetic acid buffer,
Methods were automated for the determination of chloride (4), nitrate (6), nitrite ( 7 ) , ferrous and ferric (2), and ammonium (8) ions. These methods mere selected after a survey of the literature to determine which methods were the most selective and sensitive. The rate of analysis of batch samples is approximately ten samples per hour. Chloride Method. Dioxane is added to the mercuric thiocyanate to eliminate the interference caused by peroxide in the moderator. '\STithout dioxane, as little as 1 p.p.m. of peroxide is a serious interference. A special mixing cell with a magnetic stirrer \?vas required to obtain mixing, because of the presence of dioxane. The sensitivity of the chloride method is 5 p.p.b. and the relative standard deviation is =t3% a t 10 p.p.b, The pumping diagram is shown in Figure 1. Nitrate-Nitrite Methods. hfaximum sensitivity is not obtained in this method because short time delays are used to give a fast wash, and the reduction, diazotization, and coupling steps are not brought to completion. The automated method is seniitive to
REAGENTS A N D CHEMICALS
I n the preparation of reagents and standards, reagents of analytical grade purity and deionized water were used. All glassware was selected from new stock and was carefully cleaned and kept separate from other laboratory glassware. For development work, standards were made up in light water, but for the analysis of heavy water samples, standards were prepared in deionized heavy water. The use of heavy water standards was necessary to prevent interfaces in the flow cell due to the density difference between the light and heavy water. Deionized water was prepared by passing the material through an Amberlite h9B-1 analytical grade ion exchange column. The final resistivity of light and heavy water was greater than 10 and 30 megohms. respectively. All reagents were made in deionized light water and filtered through a sintered-glass crucible (fine porosity). Chloride Method. Mercuric thiocyanate, saturated in aqueous 10% p-dioxane.
f- Direction of Pumping Sample
7.80
NaOH
0.37
Air
0.80
Acetone
0.36
A i r Toke-Off
1.20
Discard
Figure 2.
Nitrate-nitrite flow diagram VOL. 34, NO. 13, DECEMBER 1962
1729
5 p.p.b. of nitrate and has a reproducibility of k 2 p.p.b. Samples are analyzed for nitrite alone by omitting the hydrazine and acetone. Nitrite sensitivity is 1 p.p.b. with a reproducibility of ~ t 0 . 3p.p.b. The pumping diagram is shown in Figure 2. Tenposition end fittings were used in this manifold. Iron Method. Total ferrous and ferric iron are measured in this method; b u t by omitting the hydroxylamine, ferrous iron can be determined alone. Insoluble forms of iron do not interfere. The method is sensitive to 1p.p.b. of ionic iron, and has a relative standard deviation of =!=2% a t 10 p.p.b. The pumping diagram is shown in Figure 3. Ammonia Method. This method has a sensitivity of 1 p.p.b. and a relative standard deviation of +2.47, a t 10 p.p.b. The pumping diagram is shown in Figure 4. Interferences. The only impurities present in the moderator in relatively high concentrations are nitrate (1 p.p.m.), carbonate (20 p.p.m.), deuterium peroxide (15 p.p.m.), dissolved oxygen (3 p.p.m.), and ion exchange resin degradation products. These substances did not interfere in any of the methods except the nitrate method, up to the concentrations listed above. In the nitrate method 1 p.p.m. of peroxide is equivalent to 3 p.p.b. of nitrate. Turbidity was removed as a potential source of interference in all methods, except the nitrate-nitrite method, by passing the sample stream through the filtration unit. This unit was removed from the nitrate-nitrite method because of an interfering material that was leached from the membrane.
f
Cc/m 1.60
7.80
I . 20 0.60
0.80 0.80 0.80 Prbportioning
I I
610 m p
+
Discard
Figure 4. Ammonia flow diagram
DISCUSSION
The high precision and sensitivity which were obtained with the AutoAnalyzer can be attributed to several factors inherent in its design and method of operation. All of the operations of spectrophotometry are performed in a closed system and contamination from outside sources is virtually eliminated. Freedom from contamination is a critical factor in parts per billion analysis. Human errors, especially those involved in volumetric pipetting, are reduced to an absolute minimum. Another very important factor is that the standards and samples are in the system for exactly the same length of time and are both a t the same point in the color develop-
cc/m
r---i 4
I
/
0 . 4 ~ Membrane Filter
; v i n i t 0
Sample
7.80
Air
1.20
TPTZ
0.32
,
Inl !VI
112 min
Air Sodium Acetate Air Toke-Off
Mixing
Mixing
I m Heating Bath
n
610 rnp
Ois'cord
Figure 3. 1730
1
I " ,
1
Mixing Coil
ANALYTICAL CHEMISTRY
Direction o f Pumping
Iron flow diagram
0.60
o'60
0.80
ment stage when they enter the flow cell. Chemical reactions do not need to be brought to completion. Because of the improvement in precision, range expansion can be used to obtain greater sensitivity. With a tenfold range expansion, full scale on the recorder is 90 to 1 0 0 ~ otransmittance (T)instead of 0 to 100% T, thus allowing a tenfold increase in readability a t the high % T values. As an example of the improvement in sensitivity and precision that is typical of these automated methods, we found the relative standard deviation of the chloride method, when performed manually, to be *17% (n = 8) a t a level of 50 p.p.b. This can be compared with *3% a t 10 p.p.b. by the automated method. It should be possible to automate many spectrophotometric procedures, but several limitations of the AutoAnalyzer would cause difficulties in some cases. The colorimeter employs filters to isolate the required wavelength and the practical range is limited to the visible region of the spectrum. Complete mixing and the absence of gassing are necessary to obtain stable curves a t the high range expansion used in trace analysis. I n an attempt to automate a method for the determination of nitrate that employed sulfuric, phosphoric, and acetic acids, the reaction between these reagents produced minute gas bubbles that made it impossible to obtain smooth curves. Many reagents, such as concentrated mineral acids and organic compounds, have densities greatly different from the matrix stream and thorough mixing is required to produce a homogeneous solution. The large volumes required for mixing cause poor washout between samples and greatly reduce the number of samples that can
be analyzed. In addition, many of these concentrated reagents attack and destroy the pumping tubes and special tubes or techniques must be used. In general, the Autodnalyzer was found to be a very reliable instrument that requires little maintenance. It has made possible the rapid and continuous determination of trace impurities in heavy water with a sensitivity and precision that could not be obtained by other means.
LITERATURE CITED
(1) Ann. h'. Y . Acad. Sci. 87, ,4rt. 2, 609-961 (1960). (2) Collins, P. F., Diehl, H., Smith, G. F., ANAL.CHEM.31,1862-7 (1959). (3) Ferrari, A., Russo-Alesi, F. M., Kelly, J. M., Ibid., 31, 1710-17 (1959). (4) Iwasaki, I., Utsumi, S., Hagino, K., Ozawa, T., Bull. Chem. SOC.Japan 29,860-4 (1956). (5) Lundgren, D. P., AYAL.CHm. 32, 824-8 (1960). 16) ProchBzkovB, L.. Z. anal. Chein. 167. 264-60 (1959).' '
(7),Rider, B. F., Mellon, M. G., IXD. h w . CHEM., ~ A L ED. . 18, 96-9 (1946). ( S j Scheurer, P. G., Smith, F., -4s.4~. CHEY.27,1616-18 (1955). (9) Skems, L. T., .4m. J . Clin. Pathol. 28, ll-J('l957). RECEIVEDfor review March 9, 1962. Accepted October 1, 1962. Fifth Conference on Analytical Chemistry in Nuclear Reactor Technology, Gatlinburg, Tenn., October 1961. Information developed during work under contract .4T(07-2)-1 with the U. S. Atomic Energy Commission.
Improved Components for the X-Ray Emission Analysis of the Light Elements E. W. BALIS, L. B. BRONK, H. G. PFEIFFER, W. W. WELBON,' E. H. WINSLOW, and P. D. ZEMANY2 Research laboratory, General Electric Co., Schenectady,
b Low specific count rates have been decisive factors in the slow development of methods for the determination of the third period elements b y x-ray emission spectrography. Two instrumental changes-use of a chromium target tube and thin counter windowscan improve the count rates to the point where analyses are feasible for the light elements down to and including sodium. Data for the determination of chlorine from 10 p.p.m. to 25%, for aluminum from 60 p.p.m. to 2.570, and for sodium from 1 to 25% are given.
T
HE present work is aimed at the extension of practical x-ray emission spectrogr~~phy toward longer wavelengths. I I a n y useful analytical methods can be developed if high enough spccific count rates for the third period elements :%re obtained with practical equipment. The value of the chromium target tube and of thin counter windows for this purpose is demonstrated. Dnta for the determination of chlorine. aluminum, and sodium are presented to illustrate the value of improved equipment in the extension of the x-ray method t o lighter elements. The details of sample prepardion, surface conditions, matrix cffects, and other matters of crucial importance in particular methods are not emphasized. The lovi qpecific count rates observed for the third period elements are associ1 Present address, X-Ray Department, General Electric Co., St. Petersburg,
Fla.
2 Present address, Knolls Atomic Power Laboratory, General Electric Co., Schenectady, N.Y.
N. Y.
ated with insufficient excitation and high absorption in the spectrometer of long wavelength x-rays. -4further problem of finding crystals with good diffracting properties and sufficiently great d spacings has been solved well enough t o make determinations of the third period elements feasible. Both the excitation and absorption problems were recognized by Birks ( I ) , who obtsined a specific count rate adequate for the determination of sulfur in oil down t o 0.0570. Birks used a chromium target tube, a n evacuated spectrometer, and a special window on his Geiger counter. He obtained rates of only a few counts per second (c.P.s.) for Si and A1, much too low for practical determinations of these elements. The excitation problem can be stated in general terms: -in x-ray tube with a target of a lighter element will have intense characteristic radiation a t longer wavelengths and should excite the light elements in samples more efficiently (I, 3'). On the other hand, absorption of x-rays in the beam path is calculable within the limits set by the inaccuracies in the reported values of the absorption coefficients. The absorption in the beam path is important in x-ray emission work with the third period elements, C1 (KcY, 4.73 A,) to Na (KcY, 11.9 A . ) . Calculations of absorption effects (2) show that at 1 A. the absorptions by air and by windows in the beam path are small, while a t 10 A. air at 1 atm. absorbs all but a trace of the x-rays and the common windows absorb a large fraction. Either a helium path or a vacuum spectrograph should be satisfactory for the third period elements. For example, the count rate for aluminum metal (KcY,8.34 il.) in a
Sorelco vacuum spectrograph should increase about 15Y0 mhen the vacuum path is used instead of helium. There remains, however, the problem of absorption by the counter tube windows. EXPERIMENTAL
Equipment. T h e basic instruments were a Norelco vacuum x-ray spectrograph and a G E X R D 5 . Each spectrograph was equipped with the G E X R D 5 high voltage power component, the G E SPG KO.2 detector, and the Hamner 11'302 amplifier and pulse height discriminator. The Norelco vacuum spectrograph had a n integral gas flow counter tube and preamplifier and took a Philips FA-60 tungsten target tube. With the X R D 5 , a G E SPG No. 4 gas flow counter tube, a Hamner K354 preamplifier, and hfachlett AEG 50s tungsten and chromium target tubes were used. The tungsten tubes had 40-mil beryllium windows; the chromium tube had a 10-mil window. P-10 gas was used in both counters. Special, thin windows for the gas flow counter tube used with the X R D 5 were made by the following procedure. A clean microscope slide was dipped into a solution of 1 gram of Formvar powder in 400 grams of ethylene dichloride and was allowed to drain dry. Then a rectangle was cut through the Formvar film on the slide with a razor blade. The slide was slowly immersed in water, so that the Formvar film peeled off, and floated on the surface. Kext the counter tube mask was immersed in the water, slowly brought up under the film, and removed with the film in place, adhering t o the mask. -4 thin layer of aluminum, near 0.1 micron, was applied on the inside surface of the Formvar b y evaporation. Finally, with the mask attached to the VOL. 34, NO. 13, DECEMBER 1962
0
1731
