ANALYTICAL CHEMISTRY
952 the spectrographic examinations of the platinum metals used in this study. LITERATURE CITED
(1) Ayres, G. €I., and Berg. E. IT., Ax.4~.CHEM.,24, 465 (1952).
(2) Currah, J. E., hIcBryde, W. A. E., Cruikshank, A. J., and Beamish. F. E., IND. ENG.CHEX, ANAL.ED.,18, 120 (1946). (3) Hayes, J. R., and Chandlee, G. C., Ibid., 14, 491 (1942). (4) Hillebrand, W.F.. and Lnndell, G. E. F.. “hpplied Inorganic Analysis.” p. 223. S e w I-ork, John n’iley &- Sons, 1929.
( 5 ) Holrer, H., Z . anal. C h o n . . 95, 392 (1933). ( 6 ) Lehner, V., and Kao. C . H . . .I. P h y s . Chern.. 30, 126 (1926). ! i ) Voter, R. C., Bank?. C. \-., and Diehl, Harvey. ANL. CHEJI..20,
652 (1948).
( 8 ) Wunder, H., and Thuiinger. V,, Cheni. Z@.. 2, 550 11912). Toe. J. H.. and Overholser. L. G.. J . Am. Chem. S o c . , 61, 2058 (1939). . 20 (1951). (10) Toung, R. S.,d ~ d g s t 76, R E C E I V E for D review S o r e i n h e r i:i, 1951.
Determination of Alkali Metals in Insoluble Silicates by Ion-Exchange Chromatography RICIIARD C. SWEET, WILLIAM RIEMAN 111, AND JOHN BEUIiENK4MP School of Chemistry, Rutgers University, Xew Brimswick, X. J .
Recent adtances in ion-exchange chromatography indicate that the length: and tedious classical methods for the determination of lithium, sodium, and potassium in insoluble silicates can be advantageous15 replaced b? this technique. The application of ion-exchange chromatography to the determination of these alkalies has yielded a simpler. faster method without sacrifice of accuracy. The effect of altering the elution conditions on the parameters of the elution equation. and hence on the separation, is discussed. With minor modifications, the recommended procedure should be applicable to determination of alkali metals in other types of samples such as soap, biological material, and salt mixtures.
A
PPLICATIOS of ion-exchange chromatography to the determination of lithium, sodium, and potassium in insoluble silicates has produced a simpler, faster method than the lengthy and t,edioua classical methods, without sacrifice of accuracy. APPARATUS AND REAGENTS
Colloidal Dowex 50 was used as the ion-exchange rwin. Five different hatches of this product were designated by t’he nianufact,urer as lot,s 50572, L-211-32-1, L-211-32-2, L-211-32-3, and I,-211-32-4 but are designated in this paper as resins 1 , 2, 3, 4, and 3. respectively. The resins were generally prepared for use as previously described (5). For some of the work, the resins xere scparated into fractions by sieving. The ion-exchange columns had somelvhat different dimensions, but were essentially similar to those previously described (5). Capillary tips of appropriate dimensions were attached to the lower ends of the columns t.o regulat,e the flow rate. Reagent-grade chemicals were used throughout the a-ork. ANALYTICAL PROCEDURES
Determination of Sodium and Potassium. PROCEDI-RE .\. Samples that contained no lithium were analyzed by the following method. About 1 gram of sample was weighed into a platinum dish and moistened with a few drops of 0.7 M hydrochloric acid, and 2 n11. of 3 -1f sulfuric acid and 5 ml. of 6 Jf hydrochloric acid were added. Then 10 ml. of 48% hydrofluoric acid were added while the mixture was stirred wlth a platinum rod. The mixture was cvaporated in a Hillebrand evaporator without boiling to fumes sulfuric acid. After cooling, 5 ml. of 6 M hydrochloric acid \vere added. If any undecomposed silicate remained, the heatment with hydrofluoric acid was repeated. The solution was ;tgain cautiously evaporated. This time the evaporation was carried to dryness, and the full heat, of the RIeker burner was Ltpplied to the evaporator for 15 or 20 minutes. The cool residue was moistened with wat,er and warmed with j ml. of 6 M hydrochloric acid. The solution \vas transferred to :t 100-ml. beaker and treated with 3 ml. of 3% hydrogen peroxide to oxidize the ferrous iron. The mixture was warmed with the I,eaker covered until effervescence ceased. The solut,ion was then evaporated just to dryness under a heat lamp.
The residue was dissolved in 5 to 10 ml. of 0.7 .\I hydrochloric acid and heated to boiling. Then cadmium oxide \vas added in small portions until an excess of the purple-brown powder remained. Ferric and aluminum hydroxides were precipit,ated a t this point. The solution should he boiled continuously during this step, and any lumps of cadmium oxide should be broken ivith a st,irring rod. After cooling, the mixture ~ v diluted a ~ t,o 50 nil. in a volumetric flask. It was then filtered through a dry filter paper, and a 25-ml. aliquot was taken aft8erthe first fen milliliters had been discarded. This solution was evaporated to 5 to 10 nil. and transferrod to a column (SA em. X 4.6 sq. em.) of unsieved resin 1 . Then O . i O J I hydrochloric acid was passed t,hrough the column a t a flow rate of 1.2 em. per minute. The first 280 ml. of eluate, which contained all the cadmium of the aliquot and no other niet,allic ion, were discarded. The next 150 ml. of eluate contained all the sodium uncont,aminated by other met.als. The next’ 10 ml. contained only hydrochloric acid. Then there followed a fraction of 260 ml. containing all the potassium uncontaminated by other metals. The fractions containing the sodium and potassium were evaporated and t,itrated as previously described (3). X blank correction, that never exceeded 0.002 me., was applied to t8hesodium The blank for potassium was 0.000 me. The passage of 0.70 M hydrochloric acid through the column was continued until all the met,allic ions were removed, Magnesium was quantitat,ivelycontained in a fraction of 480 ml. that was separated from t,he potassium by 40 ml. of pure hydrochloric acid. Calcium followed the magnesium, probably with some cross contamination; hut this point, was not invcstigated. Overnight, passage of the acid through the column sufficed to prepare the column for another determination. Determination of Lithium, Sodium, and Potassium. PROCEDURE B. When lithium and sodium were present in the sample, the foregoing elution conditions failed to separate them quantitatively. I n this case, a column (37 cm. X 2.4 sq. em.) of resin 1, sieved t,hrough 120-mesh, \vas used. The eluant, 0.70 M hydrochloric acid, was passed a t a flow rate of 0.6 cm. per minute. The cadmium was isolated in the first 130 ml. of eluate, lithium in the next 50 ml., sodium in the next 60 ml., and potassium in 100 ml., following a fraction of 90 ml. of pure hydrochloric acid. The decomposition and preliminary treatment of the sample, the evaporation and titration of the isolated alkali-metal halides,
953
V O L U M E 2 4 , NO. 6, J U N E 1 9 5 2 and the regeneration of the column were performed exactly as in the absence of lithium. Elution Graphs. In order t o study the behavior of the several cations under various elution conditions and to develop efficient procedures for their separation, over a hundred elutions were studied in detail.
Small fractions of e1uat.e (usually ti to 8 ml.) were t'aken with t,he aid of a siphon pipet, ( 4 , 8) and analyzed by appropriate methods. -1Beckman flame photomcter was used to determine the alkali metals in these fractions, Magnesium and cadmium were titrated with Versene ( 2 ; 9) in a solution buffered at pH 10.1 n-it,h ammonia and ammonium chloride. Qualitative tests only were run for mercuric! beryllium, zinc, lead, manganese, ferric, and aluminum ions. RESULTS
The results obtained in the analysis of known salt mixtures by Procedure A are ahon-n in Table I. Each mixture contained 0.38 niilljmole of aluminum, 0.87 millimole of calcium, 0.85 niilliinolr of magnesium, and 0.01 milliniole of iron in addition to the indicated quantities of sodium and potassium.
Table I.
Determination of Sodium and Potassium in Salt 3Iixtures
Xixture so. 1
2
Sodium, Taken i 201 i 010
Me. Found 7 218 i 005
Potassium, Me. Found Taken 0 414 0 425 0 463 0 450
=k 0.017, potassium oside by this method. Perchloric acid niay he substituted for sulfuric acid, but the latter has three advantages: (1) There is no hazard involved in its use. (2) Sulfuric acid precipitates any lead in the sample; this is important because the elution procedure does not separate lend quantitatively from potassium. (3) Sulfuric acid precipitates any barium in the sample; this is desirable because barium is held tenaciously by the resin and requires an abnormally long passage of h7cIi-ochloric acid to remove it. Sample 128 was found to contain 16.85 0,04YGsodium oxide and 1.09 i 0.02% potassium oxide \Then perchloric acid vas substituted for sulfuric acid. Ferric and aluminum ions presented a problem in the development of the method. They are retained by the resin so tenaciously that overnight passage of 0.7 31 hydrochloric acid doer: not remove them. If they are not, removed complete1~-betn-een successive elutions, they contaminate the alkali-metal fraction?, giving brown residues and high results. .\Tore concentrated h>.drochloric acid removes these ions more rapidly but also cause,= a shrinkage and settling of the resin; then the passage of 0.7 III acid t,hrough the resin causes an expansion that is liable to shatter the glass tube. The passage of diammonium citrate through tho column removes the trivalent ions rapidly, but not until the resin is first converted to the ammonium form. The time spent in this conversion and the time required t,o reconvert the resin to the hydrogen form make the upe of the citrate inefficient.
+
Table 111.
The results obtained with Procedure -4in the analrsis of Samples from the Sational Bureau of Standards are given in Table 11. Kach entry in the last four columns is the mean of a t least two, usually four, determinations. Attempts t o analyze dolomite and dolomitic limestone b> this procedure lvere not successful because potassium is not separated from large quantities of magnesium under these elution conditions.
Table 11. Determination of Sodium and Potassium in Standard Samples Bureau of Standards, % SO.
128 80 43
si
89 102 29
1
Sature Soda-lime glass Soda-lime glass Boron glass Opal glass Lead glass Silica brick Iron ore Argillaceous limestone
0.99 0.04 0.16 3 25 8.40 0.29 0.61 1.15
Mixture SO.
1 2
Lithiii, Me. ____Found Taken 0.656 0.664
0 659 0.6iO
Yodiririi, Ale. ________
Taken 0 863 0.640
Table 1V.
No. 77 97
Found 0.856 0.641
__
Determination of Lithium, Sodium, and Potassium in Standard Samples
Bureau of Standards, vG Liz0 Kan0 IinO 0.86 2.11 0.22 0.12 0.54
__
.\lean 1 ~ 2 0 dev. 0.23 0.04 0.12 0 01
Ion Exchanpe. .\lean Sag0 dev. 0.04 0.03 0.05 0.01
K?O 1.99 0 33
.\lean dev.
004 0 04
I o n Exchange Cr,
________ Mean Mean Sag0
16 83 16 65 4 16 8.48 5 io 0 06 0 45 0 33
Determination of Lithium and Sodiittii in Salt Mixtures
16.88 16.67 4.31
8.50
5.57 0.00 0.51 0.30
dev. 0.05 0.04 0.02 0.05 0 01 0.00 0 03 0 03
K20
1 08
0 01 0 14 3 33 8 34 0 34 0 52 1 15
dev. 0.03 0 01 0 02 0.04 0.03 0 02 0 04 0 04
The application of Procedure B (without the preliminary steps) to known mixtures of sodium chloride and lithium sulfatc. yielded the results given in Table 111. \Vhen Procedure B was applied to samples from the Sational 13ureau of Standards, the results of Table IV were obtained. The bureau values for lithium and sodium in sample 97, a flint clay, are based on the report of one analyst in 1941. I n the original certificate (1931), lithium was included with sodium. Like\\ ise, rubidium and cesium were not determined by the bureau but, if present, were included with the postassium. The dame is true of sample 7 7 , a burnt refractory (6). DISCUS SIOY
Comments on Procedures. Other methods of decomposing the sample were also investigated. The use of hydrofluoric acid alone failed to give complete decomposition in many cases. The method of Flaschka ( 5 ) requires more time than the recommended procedure but is satisfactory in other respects. Sample 128 was found to contain 16.93 0.05'70 sodium oxide and 1.09
Therefore the ferric and alunlinum ions are removed from the solution of the sample before this is put into the column. Several reagents were tried before cadmium oxide was selected as the most sat'isfactory. Magnesium oxide was not dependable because sufficient magnesium ion often dissolved to interfere with the determination of potassium, Although calcium oxide has been used successfully elsenhere ( 7 ) ,the reagent-grade cheniical often contains enough sodium to introduce a large and rather erratic blank. Barium hydroxide has the same disadvantage. Mercuric oxide failed to precipitate the iron and alun~inunl. Cadmium oxide is a very sat ctory reagent for this purpose. It precipitates the ferric and aluminum ions fairly rapidly, and the dissolved cadmium does not contaminate the alkali-metal frat.tions of t,heeluate. Cadmium hydroxide n-ould be a9 good except that the excess is less readily observed. Elution Behavior of Various Metals. t-nder elution conditions similar to those of Procedure A, the follo4-ing metals a w quantitatively separated from each other and appear in the eluate in the order given: mercuric, cadmium, sodium, potassium, magnesium. The early appearance of the bivalent, ions of mercury and cadmium is due to the formation of chloride compleses. Beryllium appears in approximately the same fraction as potassium. Lead appears between potassium and magnesium, not quantitatively separated from either. Zinc and manganese come after potassium, quantitatively separated from it but not from magnesium. Lithium appears betffeen cadmium and sodium, incompletely separated from the latter. The more effecvti\.e
ANALYTICAL CHEMISTRY
954 elution conditions of Procedure B serve to separate lithium and sodium quantitatively. Effect of Elution Conditions on Elution Graphs. Most of the elution graphs follow approximately Equation 3 of (S), and the effect of changing the elution conditions can be described most concisely by stating the effect on the parameters of this equation. An increase in column length causes a proportional increase in the number of theoretical plates, p , the interstitial volume, V, and the volume of eluate, Cm,a t the peak of the graph, provided that the resin has uniform properties. Thus taller, narrower curves and better separations of two adjacent cations are obtained. Equation 1
centration in the interstitial solutions is no longer negligible in comparison with the concentration of the eluant, C does not remain constant during the elution. Then Equation 3 of ( 3 ) fails to describe, even approximately, the course of the elution. For example, the elution conditions of Procedure A give excellent separations of sodium, potassium, and magnesium when 2 millimoles of each are taken. However, when 5 millimoles of magnesium were taken, the magnesium graph spread in both directions and overlapped the potassium curve. When 10 millimoles of magnesium were taken, the spreading of the magnesium graph was greatly aggravated. Furthermore, the potassium graph was pushed to the left, so that the separation of sodium from potassium was unsatisfactory.
ELUTION GRAPH
is useful in calculating the minimum length of column that will quantitatively separate two given cations when their C values (distribution ratios) and the relationship between p and column height are known (IO). Here the two values of U' represent the Z values between which 99.9% of the given cation will be eluted. The number 3.09 is found in tables of probability integrals as the value of the abscissa corresponding to 99.9% of the integral between plus and minus infinity. The desired value of p is that value for which the larger U' of the cation eluted first will be equal to the smaller U' of the second cation.
32
-
t?
24 28
I I
20
5
0 16
4
40 20
I
I
I
I 400 OF
I
60
I
I
I
I
I 500
I
i
I
70
l
90
80 OF
l
l
l
100
l I10
l
120
ELUATE
Elution of Cadmium
197 CM/MIN
i
VOLUME
-
Figure 2. RATE.
-
I 300
60
0
t)
0 -
I
g
0 53 C M J M I N
RATE
-
MILLILITERS
-
I
80
O
12
0
=x
50
:: - 8
J
HEIGHT = 38 5 CM FLOW RATE * 0 55 CM /MIN
! NA
-
FLOW
-
Y\
i
MMOL
0 -
FLOW
>
t
CURVE A
2
OF
OF CADMIUM 36 C M Y 2 4 SO C M OF RESIN I iTHROUGH 120-MESH1
I
I
I 600
I
I
E L U A T E , ML
Figure 1. Effect of Flow Rate on Elution Curves Elution of 2.0 millimoles each of sodium and potassium. Crosssection area 4.0 sq. o m . Eluant 0.70 iM HCI. Resin 1, unsieved
The application of Equation 1 may be illustrated with some data obtained under the conditions of Procedure A: p / H = 4.90, V / H = 2.39, CL, = 2.39, C S = ~ 3.60 where H is the length of the column in centimeters. If the value of 1200 is assigned to p and the foregoing data are substituted in Equation 1, we find that U;,= 1540 or 1890 = 1890 or 2320
This means that a column of unsieved resin 1 consisting of 1200 plates (245 cm.) should be capable of giving a quantitative separation of sodium from lithium. The flow rate can be varied from 0.0 to 0.7 cm. per minute without significant effect on the elution graphs. This is in approximate agreement with the behavior (11 ) of spherical particles of Dowex 50. Further increases in flow rate shift the curves to the right and cause a decrease in p (lower and broader curves). These points are illustrated in Figure 1. The faster flow rates also tend to aggravate tailing, especially of the cations with the larger C values. The quantity of cation taken for elution can be varied within certain limits without significant change in p or C, the effect of the variation being chiefly a change in the height of the graph. However, if so large a quantity of a cation is taken that its con-
S o significant effect on the values of p and C was observed when the cross-section area was varied from 2.3 to 4.7 sq. cm. However, V and hence U , are directly proportional to the crosssection area. An increase in cross-section area may be utilized in order to accommodate an increase in the quantity of cation. Such an increase, however, has the disadvantage of giving larger volumes of eluate. The effect of the concentration of the eluant has been discussed
(5). Nonuniformity of Resin and Effect of Particle Size. ?he separation of lithium from sodium would require a column length of 245 cm. of unsieved resin 1 if all the other conditions were the same as in Procedure A. Since a column of this length would involve time-consuming elutions and awkwardly large volumes of eluate, a search was made for samples of Colloidal Dowex 50 that would give a more efficient separation-Le., larger value of p / H . Elution curves were obtained when mixtures of 1 millimole each of sodium and potassium chlorides were eluted through columns (40 cm. X 2.4 sq. cm.) with 0.70 A!f hydrochloric acid at 0.6 cm. per minute. From each graph, p was evaluated. Five unsieved resins and various sieved fractions of each were studied in this manner. The sieving was done by agitating the resin on the sieve with a spray of tap water until no more particles ran through the sieve. The results are given in Table V. Each entry for p / H is the mean obtained for sodium and potassium. Blanks in the table indicate that an insufficient quantity of that fraction size was obtained to fill the tube to a length of 40 cm. The data for the capacities were furnished by the manufacturer.
Table V. KO.
Capacity, Me./G.
Comparison of Resins
Unsieved
No. of Theoretical Plates per Cm. 20-40 40-80 80-120 Through 120
V O L U M E 2 4 , N O . 6, J U N E 1 9 5 2 It is clear from Table V that the finer-grained fractions of any resin are more efficient than the coarser fractions, and that the unsieved resins differ considerably from each other. Fractions sieved between the same limits from various resins are also different. There is (as one would expect) a correlation between the capacity of a resin and its efficiency in separations. This is particularly noticeable in the 80- to 120-mesh fractions, where resin 1 is best and 5 poorest. The poor behavior of unsieved resin 1 is probably due to the fact that its average particle size is larger than that of the others. The fact that only resin 1yielded enough of the 20-40 fraction to give a 40-cm. column supports this hypothesis. On the other hand, the poor behavior of the 40- to 80-mesh fraction of this resin is anomalous. Because resin 1 was distinctly superior to the others in regard to the SO- to 120-mesh fraction, it was believed that the fraction of this resin that passed through the 120-mesh sieve would be the most efficient for the separation of lithium from sodium. In order to get sufficient of this fraction for an elution, the 80- to 120mesh fraction was forced through the 120-mesh sieve by gentle pressure with the thumb. This fraction was then used in Procedure B and gave quantitative separations of lithium from sodium. The similar fraction of resin 2 gave incomplete separations of lithium from sodium under the same conditions. It is unfortunate that various batches of Colloidal Dowex 50 differ from each other. Because of this fact, quantitative separations of the alkali metals performed with a different batch of Colloidal Dowex 50 may, on the one hand, require a longer column or a smaller flow rate or, on the other hand, permit a shorter column or larger flow rate than those specified in this paper. In either case, the fractions of eluate within which any one cation will be contained will probably differ from those given in this paper. If a flame photometer is not available for the analysis of small fractions of the eluate, a simple flame test with a platinum wire and Bunsen burner suffices to indicate the beginning and end of the fraction containing any given alkali metal. Spherical Dowex 50 varies less from batch to batch than the colloidal product ( I ) , but it is less efficient for chromatographic separations. Elution Graphs of Cadmium. Figure 2 ie a typical elution graph of cadmium for the conditions of Procedure B. The unusual feature of this graph is the occurrence of two peaks. Even under the lesa effective conditions of Procedure -4,the graphs of cadmium always exhibited two peaks, but with an intervening minimum that did not extend to zero concentration. Time was
955 not available for investigating the cause of this unexpected phenomenon. I t may be due to a separation of the isotopes of cadmium. SUMMARY AND CONCLUSIONS
Methods are described for the determination of lithium, sodium, and potassium in insoluble silicates by ion-exchange chromatography. Once the behavior of any particular ion-exchange column has been ascertained, these methods require much lee9 time than the classical methods. With two columns in operation, an analyst can perform duplicate determinations of the three alkali metals in an insoluble silicate in one day and still have considerable free time during evaporation and elution. It is helieved that the ion-exchange methods are a t least as accurate as the classical methods. Flame-photometric methods are less time-consuming than ion-exchange procedures, but are less accurate unless the sample contains only a very small percentage of alkali metal. The effect of variations in the elution conditions upon the parameters of the elution equation has been discussed. Colloidal Dowex 50 exhibits appreciably different properties from batch to batch. The fractions with finer particles are more efficient for chromatographic separations. ACKNOWLEDGMENT
The authors express their gratitude to the Office of Saval Research for financial support of this investigation and to IT. C. Bauman for providing some of the samples of Colloidal Dowes 50. LITERATURE CITED
(1) Bauman, W.C., private communication. (2) Bersworth Chemical Co., Framingham, Mass., Tech. Bull. 2
(1951). (3) Beukenkamp, J., and Rieman, W., ANAL.CHEM., 22, 582 (1950). (4) Brown, W. E., doctor’s thesis, Rutgere University, 1950. (5) Flaschkrt, H., 2.anal. Chem., 129,326 (1949). (6) Hoffman, J. I., private communication. (7) Loblowitz, W., private communication. (8) Rieman, W., and Lindenbaum, S., ANAL.CHEM., in press. (9) Schwarzenbach, G., and Biedermann, W..H e h . Chun. Acta, 31,678 (1948). (10) Sweet, R. C., doctor’s thesis, Rutgers University, 1951. (11) Tompkins, E. R., Harris, D. H., and Khym, J. S . , J . A m . Chem. SOC.,71,2504 (1949). RECEIVED for review January 22, 1952. Accepted April 19, 1952.
Separation and Identification of Polymethylol Phenols by Paper Chromatography J . H. FREEMAN, Westinghouse Research Laboratories, East Pittsburgh, Pa.
T
HE precise identification of the products formed xhen phenol condenses with formaldehyde has long remained a difficult and fundamental problem for both the theoretical and the practical resin chemist. The initial products of the condensation in both acid and alkaline medium are generally conceded to be the hydroxybenzyl alcohols, commonly called methylolphenols (8, 9, 13, 22, 23). For obvious reasons most workers in this field have dealt almost exclusively with methylol derivatives of phenols, in which the reactivity is considerably reduced by substituents in one or more of the three reactive positions of the phenolic nucleus. By this means the number of possible methylol products and higher condensates is drastically curtailed. Similarly, the niethylol
analysis of these compounds and analogous resins has, until non-, been restricted to determination of total average methylol content aithout regard for the number or position of the methylol groups on the nucleus. If are to follow unequivocally the Qtepwise processes involved in resin formation, it is essential that trireactive phenol, which forms the basis for the majority of commercial resins, he studied, and that the initial condensation products of this substance with formaldehyde be individually identified and determined. Such an analysis has not hitherto been feasible. Until very recently only the ortho and para monomethylols of phenol were known. The existence of the 2,4-dimethyloland 2,4,6-trimethylolphenols\vas demonstrated by isolation of