with this procedure (Table 11) were weighed into four beakers containing 50 ml of water, O.lNHC1,O.lN "*OH, and 0.1N NaOH, respectively. Five milliliters of the stock TPABr solution were added to each beaker. The observed results are tabulated in Table 11. Of the anions listed, it appears that all interference can be eliminated by pH control with the exception of iodide, permanganate, thiocyanate, and perrhenate ions. While a qualitative determination of this type rules out the possibility of simultaneous precipitation by anions other than the four listed, it does not preclude the interference in the form of coprecipitation or absorption. The qualitative check for interference of the chlorate ion was negative; however, in obtaining the data it was observed that the solubility of TPAClO3 was fairly low, as a precipitate formed in the area where the TPABr was added, and immediately redissolved upon stirring the solution. Because the specific application for which this procedure was developed involved a C104--C103- mixture, as well as the fact that chlorates are frequently associated with perchlorates, it was decided to check the quantitative effect of chlorate in this procedure. In checking the quantitative recovery of perchlorate in the presence of chlorate ion, high results are obtained, probably caused by coprecipitation of TPAC103. As indicated in Table I, the chlorate interference can be minimized by control of reagent concentration, pH, and total volume of solution. The interference caused by chlorate can also be eliminated by its reduction to chloride with NaHS03. A five-fold or greater excess of NaHS03, with respect to chlorate, was added to the mixture of chlorate and perchlorate, which had been made slightly acidic with 2-3
drops of concentrated HCl. The amount of excess NaHS03 relative to the precipitation of TPAClOd is not an important factor, as the reducing agent and the reduction products have no effect upon the determination. The solution is allowed to sit for at least 15 minutes before the addition of TPABr, with the remainder of the procedure being carried out as previously described. Quantitative separation is achieved for chlorate/perchlorate ratios as high as 100: 1 as shown in Table
I. SUMMARY
Tetrapentylammonium salts are proposed as analytical reagents for the determination of amounts of perchlorate ranging from 2 to 50 mg. Precipitation conditions are described and a qualitative determination of anions which may interfere with the procedure is presented. Of the anions studied, iodide, permanganate, thiocyanate, and perrhenate precipitate as tetrapentylammonium salts under the same conditions as does perchlorate. Perchlorate is quantitatively determined in the presence of equivalent amounts of chlorate. Prior reduction of chlorate with NaHS03 allows quantitative separation of mixtures with chlorate/perchlorate ratios as high as 100 :1. Excess NaHS03 does not interfere in the procedure. RECEIVED for review December 21, 1967. Accepted February 1, 1968. This work was supported by the United States Atomic Energy Commission. The contents of this paper are taken in part from a doctoral thesis to be submitted at the University of New Mexico.
Silver Electrode Potentials in Alkali Nitrate and Nitrite Melts L. G . Boxall and K . E. Johnson Diuision of Natural Sciences, Uniuersity of Saskatchewan, Regina, Saskatchewan, Canada
THETHEORETICAL emf series for chlorides, fluorides, bromides, iodides, and oxides at various temperatures have been calculated ( I ) and practical series have been developed for chlorides ( 2 , 3 ) ,sulphates ( I , 4 ) , and carbonates (4). The relationships between emf series in oxyanion systems may be expressed with reference to the oxide series on an Ellingham diagram (1, 5). The theoretical results give the following series of anions in order of increasing oxidizing power: silicate < borate < phosphate < carbonate < sulphate < nitrate. Alternatively, one may compare the free energies of formation per mole of metal for a series of salts of that metal. So far only the carbonate-sulfate relation has been measured. Bartlett and Johnson ( 4 ) found the Ag(1) in C032-/Ag electrode potential to be 0.253 V more negative than the Ag(1) in SOd2-/Ag electrode potential at 550" C. The correlation (1) H. E. Bartlett and K. E. Johnson, Can. J . Chem., 44, 2119 (1966). (2) . . H. A. Laitinen and J. A. Plambeck. J . Am. Chem. SOC..87. 1202 (1965). (3) D. L. Hill, J. Perano, and R. A. Osteryoung, J . Electrochem. Soc., 107, 698 (1960). (4) H. E. Bartlett and K. E. Johnson, Zbid., 114, 457 (1967). (5) K. E. Johnson, Proceedings of the Third International Symposium on High Temperature Technology, Asilomar, California, Sept. 1967, Butterworth, London, in press. I
.
of the nitrate and nitrite systems through the use of a silver electrode is the main concern of this paper but some studies with a nitrate-nitrite solvent mixture and the effect of changing the cation in the nitrate are also reported. The silver electrode has been used extensively in molten nitrates as a reference electrode ( I , 6-8). The Ag(1) has been introduced into the melt by either adding silver nitrate directly ( I ) or by anodizing a silver electrode, a process which has beenshown to be 100% current efficient (6-8). The emf's of silver nitrate concentration cells show ideal Nernst behavior up to 0.5 mol % in (NaK)N03 eutectic (9, IO), and in NaN03 (6). Calandra and Arvia (ZI) have used a silver electrode in an equimolar NaN03-KN03 mixture as a reference electrode for studies in NaN02. The reference electrode was connected (6) H. E. Bartlett and K. E. Johnson, J . Electrochem. Soc., 114, 64 (1967). (7) L. E. Topol, R. A. Osteryoung, and J. H. Christie, ANAL. CHEM., 37,970 (1965). (8) M. Blander, F. F. Blankenship, and R. F. Newton, J . Phys. Chem., 63, 1259 (1959). (9) S. N. Flengas and E. K. Rideal, Proc. Roy. SOC.(London), Ser. A , 233, 443 (1956). (10) C. M. Gordon, 2.Physik. Chem. (Leipzig),28, 302 (1899). (11) A. J. Calandra and A. J. Arvia, Electrochim. Acta, 12, 95 (1967). VOL. 40, NO. 4, APRIL 1968
831
Table I. Standard Cell Potentials at 309
=t2" C for
the Cell: Ag/Ag(I) (0.030rn),NaN03//NaN03 (100 NaNOdx), Ag(I)/Ag X is the mole per cent of NaNOz in the working electrode compartment X,mole per cent
Expt. slope for Nernst plot, V Standard deviation from theoretical slope of 0.1160 V E" measured, V Calculated liquid junction potential EJ,V E" chemical (E" measured -EJ)
10
25
50
73
100
0.1256
0.1224
0.1113
0.1140
0.1163
O.oo00
0.0023 -0.0550
0.0026 -0.1229
0.0011 -0.1739
0.0008 -0.2067
0.0033 -0.2455
O.oo00 O.oo00
-0.0017 -0.0533
-0.0042 -0.1187
-0.0086 -0.1653
-0.0117 -0.1950
-0.0160 -0.2295
0 0.1160 o.ooo1
to the nitrite melt by a special Luggin-Haber capillary. Calandra and Arvia reported that a silver electrode in NaNOz was tried but proved to be unstable. However, the Ag(1) in a Ag(1) in NaNOz/Ag electrode was coulometrically generated with 100% current efficiency by Bartlett and Johnson (6). They reported that the silver electrode in NaNOz was stable and obeyed the Nernst equation. EXPERIMENTAL
The NaN03 and KNO3, Fisher Scientific Co. reagent grade, were used without any further purification. The NaN02, Analar reagent grade, was recrystallized twice from water to remove organic impurities. All the chemicals were dried at 120" C for 24 hours and then stored in a vacuum dessicator over PzOsuntil needed. The electrolytic cell was similar in design to that used by Bartlett and Johnson (6). The temperature of the melt could be maintained to &2" C. All the electrode compartments were placed in the main melt 2 hours before the first electrical measurement for thermal equalization. The electrical measurements were made with a Cropico, Type P3 potentiometer with a maximum error of 0.02% and the potential was followed between measurements with a Beckman 10-inch millivolt recorder connected in parallel with the potentiometer. The potentiometric recorder's input impedance was 200 kn off the null point. A Sargent coulometric current source Model IV, was used to generate the silver (I) concentration in situ in preference to the addition of salt. Coaxial cable was used throughout the electrical system to eliminate interference. All the glassware was cleaned with concentrated nitric acid, rinsed with deionized water and dried at 120" C before use. Each melt was heated to the desired working temperature and maintained there for 24 hours. It was then purged with dried oxygen-free nitrogen for 24 hours. After the purging was complete the electrolytic cell was kept stoppered tightly at all times to protect the melt from the atmosphere. The amount of the melt in the electrode compartment was determined by weighing before and after dissolving the melt out of the frit. The Ag(1) was determined gravimetrically as silver chloride. The nitrite was determined by adding an excess of ceric sulphate solution to the aqueous solution from the frit and then back titrating with ferrous sulphate using ferrous-phenanthroline as the indicator. The weight of the solvent was corrected in all the calculations for the amount of silver in the melt. All the plots were fitted using a least squares program on an IBM 1130 computer. This consisted of computing the best intercept for a line with a given slope and the standard deviations for the experimental points from this line. A silver concentration cell in (Na, K)NOI eutectic was prepared to check the experimental system. The experimental 832
0
ANALYTICAL CHEMISTRY
- x),
Nernst plot had a slope of 0.1034 i 0.0001 V which is identical to the theoretical value at the experimental temperatures, 245' C. Analysis of the electrode compartments for Ag(1) showed 100% current efficiency in the coulometric generation of the Ag(1). Four sintered glass frits of fine porosity were filled with NaNOz and placed in a NaN03 melt to check for diffusion through the frit. The frits were removed after varying lengths of time in the melt and analyzed for nitrite content. The loss of nitrite after 4 hours was 1 % and after 16 hours 4x. All the cell potentials are expressed in terms of a common reference electrode of molality 0.0300 in Ag(1). It was assumed that the activity coefficient for the Ag(1) ion was only a function of the Ag(1) concentration and not of the solvent composition. This assumption enables the measured standard cell potential to be determined by the extrapolation of the Nernst plot to a Ag(1) concentration of 0.0300m in the working electrode compartment. The activity coefficients and concentrations in the log term of the Nernst equation cancel out and the experimental potential becomes the standard cell potential. RESULTS
The results from a set of experimental runs at 309 +2" C in which the reference electrode compartment contained NaN03 as the solvent are summarized in Table I. The solvent in the working electrode compartment varied in composition from 0.0 mol % to 100 mol % N a N 0 2and the Ag(1) concentration varied from a molality of 0.001 to 0.015. The Nernst plot for the combined data from three trial runs in which the working electrode compartment contained pure NaNOz had a standard error of 3.3 mV. The experimental slope of 0.1163 V compares well with the theoretical slope of 0.1160 V at 309" C. The measured standard cell potential is -0.2455 V for the cell: Ag/Ag(I) in NaNOs//Ag(I) in NaNOZ/Ag A plot of the experimental standard cell potential for the silver nitrate-nitrite cell us. the mole per cent of NaNOz in the working electrode compartment is presented in Figure 1. T o study the cation effect on the cell potential the solvent was changed from pure NaN03 to a 1 :1 mole mixture of N a N 0 3 and KN03. The results for the different cells (A and B) at 309 f2" C are summarized in Table 11. I n Table I1 cell (A B), the result of the combination of the two cells A and B, is the same as the pure nitrate-nitrite experimental cell. The combined E" equals -0.2376 V which differs from the directly determined value by -0.0081 V or 3.5%. The Ag(1) in NaNOn/Ag half cell was found to be stable
+
OA
0
I
Log(Molr X NO NO^) OB I2
1
1
I
1.6
20
1
Table 11. Silver Concentration Cell with Different Cell Solvents at 309 * 2 O C NaKN03 represents a 1:l mole mixture of NaN03-KN03
e-
W
-
-0.2
I
-0.3b
I
I
6 0 40 Mol@% NoN02
20
I
80
I
I OP
Figure 1. Dependence of chemical standard cell potential on mole per cent ( X ) of NaN02 in the electrochemical cell: Ag/Ag(I) in NaNOa//Ag(I) in XNaNOz+(loo- X)NaN03/Ag The solid points are experimental values and the open points in line 2 were obtained from line 1. Line 1 Potential DS. mole per cent NaN02 Line 2 - Potential DS. log,o (mole per cent NaNO2)
Cell
A
B
A+B
Cell solvents
NaNOa// NaKNOa
NaKN03// NaNOz
NaN03// NaNO?
Expt. slope for Nernst plot, v Standard deviation from theoretical slope of 0.1160, v E" measured, V Calculated liquid junction potential E J , V E" chemical (E" measured - E J )
0.1102
0.1210
O.OOO9 -0.0161
0.0018 -0.2380
-0.2541
+0.0153
-0.0318
-0.0165
-0.0314
-0.2062
-0.2376
-
provided the temperature remained below 320" C and the silver (I) concentration was not greater than 0 . 0 1 5 ~ ~ Once . decomposition started the reaction became fairly rapid with NOz being given off and a gray precipitate being formed. The precipitate was analyzed and found t o be silver metal. DISCUSSION The liquid junction potentials Ej for the cells were calculated according t o Klemm (12) by using specific conductances (13). The relative mobilities between two ions were assumed t o remain constant going from a pure salt t o a mixture. This has been shown to be valid for the NaN03-AgN03 (14) and the NaNOZ (15) systems. All the liquid junction potentials were calculated for 300" C and some of the data had t o be interpolated graphically t o obtain an estimate at this temperature. The melts were assumed to behave ideally so that molar concentrations could be used instead of activities. The calculated EJ were subtracted from the experimental emf's t o evaluate the chemical cell potentials. Because of the indeterminate error in the Ej values the experimental error in the chemical cell potentials can only be estimated. The standard deviation was used as an estimate of the experimental error in the cell potentials. The k 2 " C fluctuation was neglected because it represents an error of only 1k0.0004V in the Nernst slope. This temperature error is considerably smaller than most of the standard deviations. The chemical standard cell potential for the cell: Ag/Ag(I) in NaN03//Ag(I) in NaNOz/Ag is -0.230 10.003 V at 309" C. The E" expressed as AGO, the change in Gibbs free energy, is 5.30 &0.07 kcal/mole-l for the reaction: (12) A. Klemm in "Molten Salt Chemistry," M. Blander, Ed., Interscience. New York. 1964,. _uu _ 554-559. (13) Ibid., pp 564-578. (14) J. Bvrne. H. Flemina. and F. E. W. Wetmore. Can. J . Chem.. 30, 92i (1952). (15) H. Bloom, I. W. Knaggs, J. J. Molloy, and D. Welch, Trans. Faraday Soc., 49, 1458 (1953). '
I ,
~
NaN03
+ AgNOz
NaNOZ
+ AgN03.
(1)
The following information is taken from papers by Johnson (16)and Triaca and Arvia (19,respectively, at a temperature
of 309" C: NaN03
-
AgN03 +
+ 'iZOzAGO = 12.96 kcal/mole Ag + NO2 + l/zO~AGO = 4.92 kcal/mole NaNOz
(2) (3)
The former value is an estimated value (4) whereas the latter is calculated from thermodynamic tables. No estimate for the error in Equations 2 and 3 was given in the literature so a value of +O.l kcal/mole was used in the calculations. Using these two equations in conjunction with the equation for the experimental nitrate-nitrite cell, AGO can be estimated for the decomposition of AgNO2at 309' C. AgNOz -+ NO2
+ Ag AGO = -2.7
*0.3 kcal/mole
(4)
Such a small negative value was anticipated because the cell temperature was only 170" C above the decomposition temperature. A cation effect was observed going from pure N a N 0 3 to a 50 mol mixture of N a N 0 3 and KN08. The E" measured for the cell : Ag/Ag(I) in NaN03//Ag(I) in NaKN03/Ag was -16.1 mV. The calculated EJ is +15.3 mV which gives a true standard cell potential of -31.4 +0.9 mV. This indicates that the silver ion prefers the K N 0 3 environment to that of NaN03. I n order to introduce a silver ion into the melt a Na+ or K + ion must be displaced to maintain electroneutrality. Using the Quasilattice model (18) the hole produced by the sodium ion is smaller than that required by the silver whereas it is the other way around for the potassium (16) K. E. Johnson, presented at the 131st Meeting of the Electrochemical Society, Dallas, May 1967. (17) W. E. Triaca and A. J. Arvia, Electrochim. Acta, 9,919 (1964). (18) M. Blander, Ed., "Molten Salt Chemistry," Interscience, New York, 1964, pp 127-238. VOL. 40, NO. 4, APRIL 1968
833
ion. The cation effect would disappear as the temperature increases and the breakdown of the lattice becomes more prevalent. The equation from 10 to 80 mol NaNOz ( x ) in the E" os. mole per cent NaNOz in N a N 0 3 plot is:
Eo = +0.165 - 0.170 (log X ) 10 5
x
_< 80
small simple ratio may indicate some underlying relationship which is not apparent at the present or it may just be fortuitous. RECEIVED for review November 30, 1967. Accepted January 29,1968. This work is based in part on the B. A. Honors Thesis of L.G.B. (University of Saskatchewan, 1967). Support also received from the Defense Research Board of Canada under Grant No. 5412-07
(5)
It is interesting to note that the slope 0.170 has the same value as 2.303 RTiotF at 309" C with a = 213. The fact that a is a
Meaning of the Term "Separation Factor" SIR: Faulty terminology can lead to misconceptions and confusion. The meaning of the term separation factor, used in different senses by analytical (and other) chemists, requires examination. Which meaning has a claim to validity? It seems reasonable to apply a simple criterion in deciding whether a term has been soundly named: Can the promise of the designation be fulfilled? Can the term called separation factor allow the extent of the separation of substances to be formulated or calculated? The mensurable aspects of the separation of two constituents, A and B, of a sample are quantitatively described by two ratios. One of these gives the fraction of A (the constituent to be isolated in as pure a form as possible) recovered at the end of the separation; the other ratio shows the extent of the separation of A and B:
Recovery factor for A
=
RA = QA/(QA)o,
where Q A is the isolated quantity of A and ( Q J O is the original quantity (an analogous expression can be written for RE). If desired, this ratio can be multiplied by 100 to furnish the percentage recovery. Separation factor
= SE/A =
QB,/(QB)o
QA
~
-
( Q A ~
(QA)o ~
x -Q B
( Q ~ o
QA
= RE -
RA
The separation factor thus defined is simply the factor (multiplicand) giving the change in the ratio of B to A . (We have a choice in expressing the ratio of the two constituents as A / B or B / A ; the latter is preferable because it makes A the basis of reference.) This definition of separation factor was proposed as early as 1950. In a quantitative separation, R A GZ 1 so that then S E / A RE. If A is recovered quantitatively in a separation, the separation factor as here defined is the factor by which the undesired constituent B is reduced. Decontamination factor, used by radiochemists, is the reciprocal of S B , A . Some chemists who (for reasons to be mentioned later) wish to avoid applying separation factor to R E / R A= SEIA,propose to call this quantity the enrichment factor (for A ) . However, depletion factor (for B ) would seem more appropriate, because A remains essentially constant in a quantitative separation. Actually there is no valid reason for designating SelA anything but separation factor; that is what it is. In separations based on distribution equilibria-for example, in immiscible solvent extraction-SE,A can be calculated if the 834
ANALYTICAL CHEMISTRY
distribution coefficients DA and DE and experimental variables such as phase volumes and the number of equilibrium stages are known. To take a simple example, if A and B are extracted to different extents into an immiscible organic solvent from an aqueous solution, the separation factor after n extractions, in which the ratio of organic and aqueous phase volumes is Vo/VRI0= r, is given by
If the original ratio of B to A is multiplied by the above quantity, the ratio of B to A in the combined organic extracts is obtained. S B / A is indubitably a separation factor. Note that factor is used in its exact sense as a multiplicand, not merely in the sense of an element contributing to a result. The separation factor can be written for any separation in which Re and R Acan be formulated. Some chemists balk at accepting separation factor as the designation of &IRA, not because any fault can be found with this definition, but because separation factor is used by them for another quantity, namely the quotient of distribution coefficients DA/DE(DA > DB). ( D A / D Eis used with special reference to liquid-liquid extraction, but analogous expressions can be written for other distributions.) Is the quotient DAi DE (= C Y )a separation factor? If so, CY will allow us to formulate and calculate, when values of the requisite experimental variables are specified, the extent of separation of A and B. But this it cannot do. To see that CY is not a separation factor, we need go no further than to note that a does not provide the individual values of DA and DE required for the formulation or calculation of the extent of separation of A and B. In general, CY is not equal to RA/RB (the reciprocal of S B ! A ) . As far as separability is concerned, CY is indeterminate except when it has the value 1 (neglecting the trivial case DA/DE = a , when Dg = 0). Specifying the value of the product DA X DE, in addition to CY, is equivalent to stating DAand DE. These strictural comments are not directed at CY as such, only at the practice of terming it a separation factor. T o be sure, no separation is possible when CY = 1, and large values of CY may betoken easy separability, but there is no exact or necessary relation between CY and separability. Under certain experimental conditions, CY = 5 might represent a better separation than CY = 10. The same value of CY may represent vastly different separabilities. Obviously it is not immaterial