Determination of Cadmium by Flame Photometry

Wild cherry. 66.8. 67.5. -0.7. 0.43 the sirup blank add 1 ml. of the sirup dilution. Mix all solutions by mild agitation. Measure theabsorbance of the...
3 downloads 0 Views 665KB Size
100% glucose after hydrolysis. Table 111.

Sirup Acacia

Assay of U.S.P. XV Sirups

Av. found 82 6

Cocoa

Citric acid Ferrous sulfate Glycprrhiza Ipecac Orange Simple Tolu balsam Kild cherry

63 1 80 8 82 2 61 5 70 6 81 7 83 6 81 0 66 8

Sucrose, yo Actual 80 0 60.0 82.4 82 5 63 8 70 5 82 0 85 0 82 0 67 5

the Firup blank add 1 ml. of the sirup dilution. N i x all solutions by mild agitation. Measure the absorbance of these solutions in 1-em. cells a t 400 mg in a suitable spectrophotometer, using the reagent blank as the reference solution. Calculations. Mg, of sucrose per nil. of sirup (after

hydrolysis) = ahs. sirup diln. - abs. sirup diln. bla& X abs. std. glucose (av.) 1.9 100 0.2 X - X - X sirup diln. factor 1 5

lfg. of glucose per ml. of sirup (before

hydrolysis)

=

abs. sirup diln. - abs. sirup diln. blank X abs. std. glucose (av.) 0.2 X sirup diln. factor DISCUSSION

The enzymatic color reaction is stopped by the addition of hydrochloric acid, so the timing intervals of

Dif. +2 6 1-3.1 -1 6

-0 3 -2 3 +o 1 -0 3

-1 4

-1 0 -0 7

Table

I1 shon s their results and that variaStd. Dev., 5 Replicates 0.95 0 47 0 88 0 24 0 72 0 81 1 63 0 64 0 30 0 43

the addition of samples and acid should be accurately controlled. The daily standard glucose absorbance value (400 mp, 1-cm. cells, Beckman Model DU spectrophotometer) from the average of three determinations per day, for 20 days, was 0.423 and the standard deviation was 10.016. The enzyme reagents for the above study were prepared from the same raw materials, but were made every 2 days. If a greater accuracy than =t4% is desired, standard solutions should be assayed with the samples. A sequence of glucose concentrations was assayrd simultaneously and linearity was observed in concentrations of 0.1 to 0.4 mg. per mi. of glucose, as shown in Table I. The precision and accuracy of the hydrolysis procedure were established by six replicates on a sucrose sample, yielding a n average of 99.1% and a standard deviation of 10.56%. TWO analysts assayed mixtures of glucose and sucrose containing a theoretical

tions in glucose or sucrose ratios do not alter the accuracy of the assay of glucose after hydrolysis. The U.S.P. sirups mere prepared as directed from reagent sucrose. Fresh cherrics and raspberries were not available, so cherry and raspberry sirups n-ere omitted. Liquid glucose, ha1 ing an indefinite amount of glucose, x i s omitted from the cocoa sirup. Five complete rrplicate assays were made on the initial dilution of each sirup and the results condensed into Table 111. The average differences of actual and found sucrose for the ten sirups are about -0.2% and the average coefficient of variation (S/Z x 100) for the five replicates is 10.93%. LITERATURE CITED

(1) Comer, J. P., AKAL. CHmr. 28, 1748 ( 1956).

(2) Coulthard, C. E., et al., Biochem. J . 39, 24 (1945). 13) Free. A. H.. et al.. Clin. Chem. 3. 163 \

,

\

,

(1957). (4) Keilin, D., Hartree, E. F., Biochem. J . 50, 331 (1952). ( 5 ) Keston, A. S., Abstracts of Papers, p. 31c, 129th Meeting, ACS, Dallas, Tex., April 1956. (6) Saifer. A . Gerstenfeld. S.. J . Lab. Clin. M e d . 51, 448 (1958j. ( 7 ) Theorell, Hugo, Arkiv. Kemi.Nzneral. Geol. 2, 1 (1942). (8) “United States Pharmacopeia,” 15th Revision, p. 711, Mack Publishing Co., Easton, Pa., 1955. RECEIVED for review May 8, 1958. Accepted September 2, 1958.

Determination of Cadmium by Flame Photometry PAUL T. GILBERT, Jr.‘ Nuclear Engineering and Manufacturing Deparfmenf, Norfh American Aviation, Inc., Downey, Calif.

A , survey of the excitation of cadmium in various flames has shown that the air-hydrogen flame i s particularly advantageous for the determinaiion of cadmium a t 326.1 mp. A sensitivity of 0.1 p.p.m. and a precision to 0.1% have been demonstrated with a Beckman flame photometer with multiplier phototube. Interferences of aluminum, iron, copper, and calcium with cadmium in the air-hydrogen flame, and of aluminum and iron in the oxyhydrogen flame have been measured. The working curve for cadmium i s nearly linear below 500 p.p.m. To circumvent interferences the method of successive dilutions i s used. The possibility of avoiding aluminum inter-

110

ANALYTICAL CHEMISTRY

ference b y use of oxyhydrogen or b y flame determination of aluminum i s also considered.

c

inrestigated by Lundegirdh ( I O ) in the air-acetylene flame; he mentioned a detection limit of 200 p.p.ni. a t 326.1 mp. The inner cone of air-acetylene emits also the resonance line 228.8 mp (8,9). Rusanov and Alekseeva (12) determinrd cadmium in ores semiquantitatively with a visual air-acetylene flame photometer. Rusanov (11) subsequently determined cadmium in solution by photographic photometry with the air-acetylene flame, and in mineral powders with the oxy- m x u c h f TWS



acetylcnc flame. The oxygen-coal gas flame is said to excite cadmium only n-eakly (15). The oxygen-methane flame is also poor for this metal (7); with a Beckman DU spectrophotometer and rnultiplicr phototube a detection limit of about 50 p.p.m. might be expected a t 326.1 nip. Egerton and Rudrakanchana ( 2 ) studied the flame spectrum of dimethyl cadmium, and Alekseeva and Mandel’shtam ( I ) examined the emission of cadmium in the inner and outer cones of the air-acetylene flame, but 1 Present address, Beckman Instruments, Inc., Fullerton, Calif. * Xow Atomics International, Canoga Park, Calif.

without applying t'he results to analysis. Russell, Shelton, and Kalsh (13)demonstrated a sensitivity of 0.1 p.p.m. for cadmium a t 228.8 mp in their absorption flame photometer, using air-coal gas. Straub, Bauer. and Cooke ( 2 4 ) recently adapted the Becknian oxyhydrogen flame spectrophotometer with photomultiplier for operation with a sparkin-flame source. They reported a detection limit as low as 0.02 p.p.in. for t'he cadmium spark line 226.5 Inp; iiieaii deviations could be kept d o x n to 2y0. They observed severe interference effects from lead, vanadium, and potassium. Cadniiuni has been studied in these laboratories in several flames. The detection limit of 2 p . p m . at, 326.1 mp reported in (4) refers to air-hydrogen, and the limit of 40 p.p.ni. a t 228.8 mp refers to oxyhydrogen, both with the Beckman DIT instrument. with photomultiplier. Subsequent work by Watanabc (16) and the author yielded the revised data sholvn in Table I. Data for the oxygen-cyanogen flame were obtained with a n experimental burner ( 6 ) . Among the air-hydrogen, oxyhydrogen, and oxyacetylene flames, yielded by one and the same burner in the sanic instrument, air-hJ-drogen gave the highest absolute emission intensit'y for cadmium a t 326.1 nip. I n addition, this flame has by far the lowest background emission, and is the coolest (about 22.50" K. when spraying vvater). The 326.1-nip line shows lit'tle selfabsorption, whereas the resonance line 228.8 nip is strongly self-absorbed, giving a parabolic n-orking curve. Among flame sources commercially available, the air-hydrogen flame in thtt Beckman flame photometer gives the best sensitivity for cadmium. INSTRUMENT

A Beckman D U spectrophotometer was used 11-ith a ;\lode1 9200 flame attachment and a standard oxyacetylene atomizer-burner (catalog KO. 4030). An earlicjr version of the present Model 4300 photomultiplier attachnirnt was supplied by the manufacturer. This had a continuously variable control for the dynode voltage. Extra batteries n ere added to increase the amplification to the point where shot-effect noise a t the photo-cathode became comparable with flame flicker, to attain maviinal sensitivit?. The time constant was about 0.25 second, and over-all flicker of the galvanometer needle \\-as equivaof the reading. By lent to about +l7. watching the needle for wreral seconds. the slide-wire could be balanced n i t h a repeatability of 0.1 to 0.2% (eupressed as mean deviation). I n using the oxyacetylene burner n itli air and hydrogen, the best air pressure \vas found to be about the same as that specified for oxygen by the nianufacturer. The best hydrogen pressure was such as to yield a flnnie nhose tip barely protruded through the hole

in the roof of the burner housing. The concave mirror in the housing was focused for the best combination of steadiness and luminosity a t a point in the flame above the center of the primary entrance beam, but slightly below the point of maximal cadniium emission. The nave length dial mas set to give peak response for the 326.1-mp line. The burner n a s used also n i t h oxyhydrogen. The specified oxygen pressure \\as used, nhile the hydrogen pressure n as somen hat greater than that for the air-hydrogrn flame. Because the background emission in OKYhydrogen is about 10 times as great as in air-hydrogen, the niirror and hydrogen pressure 11ere adjusted to provide a maximal cadmium-background ratio. This was found nhen the hydrogen pressure n a s 10 to 15YG above that maximizing the background emission, with the mirror focused upon a point in the flame somewhat above that a t which the total emission from 100 p.p.m. of cadmium a t 326.1 mp was maximal. Hoivever, n hen aluminum \vas being studicd, the optima1 adjustments nere nearer those mnuimizing the background. The selector snitch n-as kept a t 0.1, n hile t h r sensitivity knob and slit nere adjustrd to get the best reproducibility. The sensitivity setting ranged from fully clockwise to the midpoint (five turns). K i t h air-hydrogen, the slit nidth ranged from 0.08 mm. (at higher cadniium concentrations) to 0.16 nim. (at low concentrations). JJ7ith ouyhydrogm, the slit width was 0.06 to 0.10 mni. A fume duct over the burner remoi-ed the toxic cadmium vapors.

Table I. Detection Limits for Cadmium in Various Flames

[Beckman DU spectrophotometer xith 1P28 photomultiplier. Concentration in p.p,m. to give net response equal t o 1% of background (and at least 1 galvanometer scale division) for optimal instrumental adjustments] Concentration, P.P& 328.1 mp Wave length 228 8 326 1 (acetone m,u inp solvent) Air-hydrogen 100 0 5 Oxyhydrogen 10 5 2 Oxyacetylene 50 10 2 Oxygencyanogen 0 5 25 Table II.

Interference Effects for 100 P.P.M. Cadmium

Background Interference,

Radiation Interference,

Element 70 70 Air-hydrogen. Slit 0.15 mm. 5 -3 Ca A1

n i

-5

Fe

0.0 1.5 0.2 $0.2 cu Oxyhydrogen. Slit 0.10 mm. -0 5 Fe 20 Al= 1.7 -1.5 Cd and A1 a t 1000 p.p.m. for background interference, 500 p.p.m. for radiation interference, but calculated to 100 p.p.m. 5

The final reading on each sample was thus the mean of four separate means of three rapid readings each.

MATERIALS AND PROCEDURE

The samples nere small cylinders of aluminum plated with about 50 mg. of cadmium. Thrse were treated n ith 2 ml. of concentrated nitric acid, and the acid and rinsings were diluted to 50 ml. with ii-ater. Varying amounts of aluminum, perhaps 5 or 10 mg., dissolved along n i t h the cadmium. Standards were prepared from cadmium metal dissolved in nitric acid and cadmium oxide dissolved in hydrochloric acid. For handling the samples and standards, the 5 m l . sample cups, beakers, pipets, and stirring rods were coated with the water-repellent, Desicote ( 5 ) , which greatly facilitated the work. Precise dilutions were made with Desicoted pipets calibrated for complete discharge. I n taking sets of readings, three or four cadmium-containing solutions and a nater blank nere compared a t a time. For each sample, four rapid consecutive readings were obtained, the cup being lowered from the capillary for a fraction of a second between readings. Usually the first reading of the set Ivas lorn owing to bubble entrapment, and v a s ignored. It required 20 to 30 seconds to obtain a n average (triplicate) reading in this way. The set of samples \vas run through four complete cycles of such readings, requiring about 10 minutes altogether.

INTERFERENCES

Table I1 summarizes the results of such interference studies as seemed pertinent. Background interference is defined here as the net interferent reading expressed as a percentage of the net reading on an equal concentration of cadmium, at 326.1 mp and for the slit width indicated. The net reading equals the total reading on the pure solution minus that on a water blank. I n a typical DU spectrophotometer, the background interference as thus expressed 0.03), will be proportional to ( S where S is the slit n i d t h in mm. The increment 0.03 represents band n idth widening due to aberrations, etc. The assumption is made that cadmium shon s no continuous radiation in the vicinity of the 326.1-mp line; observation showed that such radiation is negligible compared m-ith the monochromatic radiation, in air-hydrogen and oxyhydrogen. The background interference, regarded as a coefficient (per cent contribution by interferent, per unit interferent and cadmium each), will be virtually independent of the concentrations of cadmium and interferent, a t least u p to about 1000 p.p.m.

+

VOL. 31, NO. 1, JANUARY 1959

111

0 5

I

2 AL

Figure 1, cadmium

5 CD

I

2

5

10

RATIO

Depression of cadmium emission at 326.1 ml.c in airshydrogen, at 100 p.p.m. cadmium and various concentrations of aluminum

Radiation interference is expressed in Table I1 as the percentage of enhancement (positive) or quenching (negative) of the net monochromatic cadmium emission by an equal concentration of interferent. Radiation interference will be independent of instrument settings, but it is not likely to follow any rule as regards its variation with the concentrations of cadmium and interferent. Figure 1 shows this in the case of aluminum in air-hydrogen. The quenching effect of aluminum on 100 p.p.m. cadmium is proportional to the aluminum concentration up to perhaps 300 p.p.m., beyond which it changes less rapidly. Radiation and background interferences act additively. Table I1 shows that for cadmium in oxyhydrogen the radiation interference of aluminum will exactly balance the background interference a t a slit width of about 0.08 mm., when 500 p.p.m. of each metal is present. The nct interference of aluminum is then zero under these conditions, and may be expected to be zero for lower concentrations of aluminum] but not necessarily for other concentrations of cadmium. The interference of copper with cadmium is slight. But there are two strong copper lines, at 324.8 and 327.4 mp, each 1.3 mp distant from the cadmium line. To avoid spectral overlap, the slit width must be kept below 0.23 mm . The joint interference of calcium and aluminum with cadmium was studied in the hope that the calcium might reduce the interference of aluminum with cadmium. But calcium quenched the cadmium emission apparently independently of the presence of aluminum. Among other ions, fluoride had

112

k

Radiation interference of aluminum with

ANALYTICAL CHEMISTRY

Table 111.

1

1

1/12

i/B

1 1/4

Figure 2.

1

I/3

I

1/2 DILUTION

I 3/4

I I

I

Method of successive dilutions

Apporent concentration of cadmium in original sample, as determined at various dilutions and extrapolated to inflnite dilution

Logarithmic Slope of Working Curve

(Air-hydrogen, 326.1 mp) Concentration Range, P.P.M. 10,800-1080 2,300- 580 2.200- 550 1; 100- 540 580- 290 560- 270 540- 270 290- 145 275- 137 270- 136

Slope 0.75 0.89 0.91 0.93 0.98 0.98 0.98 0,995 0.995 0.985

little effect on the cadmium emission, while free hydrochloric acid slightly depressed the emission. WORKING CURVE

A careful intercomparison of the standards a t various dilutions established agreement within O . l % between the two principal standards, But because of their different anions, discrepancies appeared a t higher concentrations. Thus, if the chloride standard was compared against the nitrate standard as reference, it appeared t o be 2.5% low a t a concentration of 2300 p.p.m. of cadmium, but only 0.3% low when diluted to 600 p.p.m. Similarly] another chloride standard containing ammonium chloride appeared to be about 0.4% low when diluted to 500 p.p.m. These are really interference effects, and affect the shape of the working curve; but because there is no a priori reason for preferring one anion over another, the choice of a reference solution is arbitrary. Because the

method of extrapolation to infinite dilution was employed (see below), these discrepancies were of no consequence. The cadmium line in air-hydrogen shows a nearly straight working curve. I n logarithmic coordinates it has a slope of unity a t low concentrations. At higher concentrations, the slope (log emission plotted us. log concentration) slowly decreases owing to self-absorption. This logarithmic slope was measured with three different standard solutions, each a t various dilutions (Table 111). These data may be expected to vary with the burner and its adjustments, and have only relative significance so long as the actual composition of the flame is not known. During an interval xhen the capillary was partially clogged and delivering less sample to the flame, the logarithmic slopes rose slightly above those listed in the table. In finding the cadmium concentration of an unknown sample by comparison with a standard] the logarithmic slope was allowed for mathematically. Graphical interpolation takes more time for equal precision, and requires a very large graph. A few measurements in oxyhydrogen showed that the logarithmic slopes for cadmium a t 326.1 mp are equal t o those for air-hydrogen at the same concentrations. Earlier work had shown that oxyacetylene likewise gives the same logarithmic slopes, a t least below 2500 p.p.m. By contrast, the logarithmic slope for the 228.8-mp line in the hotter flames is about 0.5 (implying a parabolic n-orking curve) except a t the lowest concentrations. Aside from considerations of sensitivity, the analytical precision available with the 228.8-mp line would therefore be only about half that available with the 326.1-mp line.

METtIOD

OF SUCCESSIVE DILUTIONS

I n determining cadmium in an unknown solution by comparing its emission with that of a standard, the usual difficulty arises from the unknown magnitude of interference by concomitant elements. I n the present case, it was not easy to allow for the varying concentrations of aluminum, and three expedients seemed feasible: to compare solutions a t increasing dilutions and extrapolate to infinite dilution; to use the oxyhydrogen flame under conditions such that the net aluminum interference was zero; and to determine the aluminum by its o v n flame emission, and allow for it, Table IV shows the method of successive dilutions. The principle of the method rests upon the fact that many interferences, expressed as a percentage of the emission being measured, diminish with decreasing concentration, a t least a t sufficiently low concentrations. This is not true for simple background interference, but the background interference of aluminum in air-hydrogen (Table 11) is so small that it can be wholly neglected in the samples under consideration. When necessary, however, background interference can be compensated for by taking readings on either side of the line, instead of using a blank. In Table IV, three unknown samples, X, Y , and Z, are compared with the nitrate standard, A . The concentration of standard A was 2198 p.p.m. The divisors show the dilution factors: thus A/4 refers to standard A diluted fourfold, containing 549.5 p.p.m. The readings on standard, samples, and blank (water) in air-hydrogen a t 326.1 mfi are listed. I n the first comparison, because the cadmium concentrations in the samples n-ere entirely unknown, they were first computed by linear extrapolation and then corrected for the known logarithmic slope, a. I n the second comparison, the standard was diluted by a further factor of 3, while each sample was diluted by that integral factor which would match it to A/12 most closely. The readings were then made with greater care. Finally, each solution was diluted with an equal volume of water for a third comparison. Each individual reading, after subtraction of the blank, was converted to an apparent concentration, calculated back to the undiluted sample by means of the dilution factors. The logarithmic slope could be taken as 1.00 in the second and third comparisons. If the apparent cadmium concentrations of the original sample, as found a t the successive dilutions, are plotted against the reciprocal dilution factor, the true cadmium concentration should be found by extrapolating to infinite dilution (zero concentration). Such ex-

trapolations are shown for the three samples in Figure 2. Under the reasonable hypothesis that all interferences (here due chiefly to aluminum and perhaps nitric acid) exert a quenching (or enhancing) effect upon cadmium which is (percentagewise) independent of the cadmium concentration and proportional to the Concentration of interferents, the plots of Figure 2 should be straight lines. This seems to be accurately true for two of the samples; for the third, the departures amount to only *0.3%. Successive dilutions can be carried to the point where the permissible error becomes equal to the detection limit. In the present case, one further twofold dilution would have been feasible.

net effect of aluminum on cadmium in the oxyhydrogen flame is nil. This suggests the possibility of obtaining sufficiently accurate results from a single set of readings. Unfortunately, the much higher flame background in oxyhydrogen reduces the prpcision a t low concentrations. The undiluted samples of Table IV would be sufficiently concentrated for accurate comparison in oxyhydrogen. The background and radiation interferences of aluminum would be expected to be about 1% each, for a slit n-idth of 0.08 mm., and would compensate each other nithin the experimental error. Time v a s not available to test these conclusions; moreover, there may be other interferences that would remain uncompensated.

DETERMJNATION IN OXYHYDROGEN

DETERMINATION OF ALUMINUM Earlier work (4, 6) had demonstrated detection limits of about 2 p.p.m. for aluminum in aqueous solutions in oxyhydrogen with equipment of the kind used here. Individual burners vary

Even on a routine basis, analysis by successive dilutions in air-hydrogen would require 15 minutes per sample for good precision. As mentioned earlier such conditions may be found that the Table IV.

Readings (slit 0.1, sensitivity 0) (sequence horizontal)

Analysis of Samples by Successive Dilutions Sample Y Z x-A/4 64.0 62.8 62.6 63.0 62.8 62.6 61

Apparent p.p.m. Cd ( a = 1.00)

Av. p.p.m. Cd Mean dev., p.p.m. Av. p.p.m. Cd corrected for true a

Readings (slit 0.25, sensitivity 1)

A/12 87.9 88.1 88.2 88.0 88.1

Apparent p.p.m. Cd

Av. p.p.m. Cd Mean dev., p.p.m. Readings

(slit 0.25, sensitivity[3)

A/24 81.5 81.0 81.5 81.5 81.4 81.5 81.5

77.3 77.5

119 119

94.2 94.5

77.1 77.1 76. 669 685 681 683 692 682 6

...

93.5

690 x/4 86.8 86.8 87.2 86 9 87.2 720 718 722 720 723 72 1 2

X/8 81.0 81.0

... ...

Apparent p.p,m. Cd

80.9 81 3 80 9 725 732 725 730 724

Av. p.p.m. Cd

i27

Mean dev., p.p.m.

3

... 116 1046 1066 1073

...

93 822 841 831 854

1062 10

837 10

1116 Y/6 90.2 90.3 90.6 90.3 90.4 1136 1134 1138 1136 1136 1136 1 Y/12 83.7

860 Z/5 84.9 84.8 85.2 85.2 85.1 875 870 875 877 875 875 1 z/10 79.4

83.4 ... 83.5 83.6

... 79.3 79.3 79.3

...

...

...

1146 1140 1144 1143

877 876 878 876

1143 2

877 1

VOL. 31, NO. 1, JANUARY 1959

Water 3.1

3.3

Water 20.9 20.9 20.9 20.9

...

Water 30.3

... 30. I 30.1

...

29.9

113

aidely in their sensitivity for aluminurn, and with the present burner the detection limit for the aluminum oxide band peaking a t about 485.5 mp (heads a t 484.2 and 486.6 mp) was 10 p.p.m. The lines a t 394.4 and 396.2 mp were still weaker, I\ hile the air-hydrogen flame showed no useful aluminum emission. B u t the band a t 485.5 mp was not very selective, rising only about 50% above the aluminum continuum in this region. The aluminum could be determined selectively by taking readings on the peak a t 485.5 and off the peak a t -178.0 mh, the effective detection limit being then 40 p.p.ni. The logarithmic slope for aluminum a t 485.5 mp in oxyhydrogen was about 0.97 between 1250 and 2500 p.p.m. Cadmium a t 1000 p.p.m. exercised no noticeable background or radiation interference on the aluminum at this wave length. T w o of the samples showed the aluminum peak Tvith a n intensity corresponding to about 200 p.p.m. of aluminum. I n air-hydrogen this concentration would be expected, from Figure 1, to depress the cadmium emission by about 1% at 100 p.p.m. of cadmium. Sample X shows nearly this much depression, but in samples Y and 2 the depression is less. Actually, in the absence of other interferents, the slopes of the lines in Figure 2 should be fairly accurate measures of the aluminum concentration. B u t other interferences may be a t work; the samples showed a continuum in the blue greater than that which could be accounted for by the cadmium and aluminum present.

Hence the separate determination of aluminum will not suffice for precise correction of the results unless all other interferences are under control. By using hexone as solvent, Eshelman et al. (3) attained the very good detection limit of 0.3 p.p.m. for aluminum with a n instrument of the present type, in the ” _ oxyhydrogen flame. SENSITIVITY AND PRECISION

Mean deviations for the cadmium results are shown also in Table IV. For the carefully conducted second and third coinparisons, the mean deviations average 0.2%. The probable error of each mean therefore amounts to about 0.1%. As this prrciqion mis attained a t concentrations as loiv as 100 p,p.m., i t follows that the probable error of reading in the averaged result can be kept to 0.1 p.p.m. below 100 p.p.m. The detection limit as commonly defined (Table I) is 0.5 p.p.m. ACKNOWLEDGMENT

’‘

This project was carried out on beOf R’ F’ Smith’ of the and standards. The author is grateful for the opportunity thus afforded.

LITERATURE CITED

(1) illekseeva, V. G., Mandel’shtam, S. L., Zhur. Eksp. i Tekh. Fzz. 17, 765 (1947). ( 2 ) Egerton, &4.,Rudrakanchana, S., proc. fioy. sot. (London) & 2 5 , 427 (1954).

(3) Eshelman, H. C., Dean, J. A,, Menis, O., Rains, T. C., “Extraction and Flame

Spectrophotometric Determination of Aluminum,” Pittsburgh Conference on hnalytical Chemistry and Applied Spectroscopy, March 3, 1958. (4) Gilbert, P. T., Jr., I n d . Lab. 3, 41 (August 1952). (5) Gilbert, P. T., Jr., Sczence 114, 637 f 1951\. (6j Gilbert, P. T., Jr., “Study in Oxycyanogen Flame Photometry,” Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, >larch 3, 1958. ( 7 ) Gilbert, P. T., Jr., Hawes, R. C., Beckman, A. O., . ~ X A L . C H E M .22, i72 (1950). (8) LundegLrdh, H., Lantbruks-Hoyskol. Ann. 3, 49 (1936). (9) LundegLrdh, H., MetaElwirtschajt 17, 1222 (1938). (lO)-L°Lrdh, H., “QuanFhtive Spektralanalyse der Elemente, G. Fischer, Jena, Part I, 1929: Part 11, 1934. (11) Rusanov, A. K,, B I L I I . sei, U.R.S.S., Sdr. phys. 4, 195 (1940). (12) Rusanov, A. K., Alekseeva, IT. M., Z a v o d s k a V Lab. 7,963 (1938). (13) Russell, B. J., Shelton, J. P., Walsh, A,, Spectrochzm. Scta 8 , 317 (1957). (14) Straub, W.-4., Bauer, S. H., Cooke, W.D., “Spark in Flame Spectroscopy,” Pittsburgh Conference on Analytical Chemistry and dpplied Spectroscopl-, March 6, 1988. (15) Strouts, C. R. Y., Gilfillan, J. H;: \Jrilson, H. N., “Analytical Chemistry, ~ ~ 0 2, 1 . pp. 865-6, Oxford Univ. Press, h-ew York, 1955. (16) Watanabe, H., Beckman Instruments, Inc., Fullerton, Calif., unpublished work. RECEIVEDfor review Akug‘Jst 29, 1957. Accepted August 12, 1958. Project supported by United States Atomic Energy Commission.

Selective Reactivity in Gas-Liquid Chromatography Determination of 2-Bromobutane and 1-Bromo-2-methylpropane W. E. HARRIS and W. H. McFADDEN2 Chemisfry and Mefallurgy Division, Atomic Energy of Canada, Itd., Chalk River, Ont., Canada

F A method is described for the gas chromatographic determination of 2bromobutane and 1 -bromo-2-methylpropane in mixtures of alkyl bromides. A short reaction tube packed with silver nitrate on firebrick completely removes 2-brornobutane a t room temperature but does not remove 1 -bromo-2-methylpropane. The silver nitrate reactor is placed at the leading end of a gas-liquid chromatographic column. An analysis i s performed with and without the reactant and the amount of each component determined by difference. The reaction is specific for secondary and tertiary alkyl bro114

ANALYTICAL CHEMISTRY

mides. The effects of temperature, length of reaction zone, and other variables are discussed.

G

AS-LIQUID chromatography is use-

ful for the separation and identification of mixtures of organic liquids. Although the technique has been used for only 7 years ( 2 ) , considerable literature has been published demonstrating its versatility and outlining the basic principles. It is possible to separate azeotropic mixtures and many compounds with similar chemical and physical properties. However, some compounds having nearly identical boiling points and

structural properties have not yet been separated even by elaborate gas chromatographic techniques. In this laboratory, analysis of the products of hot atom reactions in alkyl bromides has made the separation of 2-bromobutane and 1-bromo-2-methylpropane desirable. Extensive experimentation by conventional methods failed to give a useful separation. How1 Present address, Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada. 2 Present address, Shell Development Co., Emeryville, Calif.