Precise Acid Standardization by Coulometric ... - ACS Publications

---

Precise Acid Standardization by Coulometric Titration Using Simplified Equipment EDGAR 1. ECKFELDT and

E. W. SHAFFER,

JR.

Research and Development Center, leeds and Northrup Co., North Wales, Pa.

b Performance characteristics of coulometric titration equipment of a commercially available kind were studied b y carrying out precise coulometric assays on potassium hydrogen phthalate and constant boiling hydrochloric acid. The cells investigated were of the shield-tube type in which separation of anode and cathode is accomplished by restricted flow of electrolyte. Tests were made on the conventional cell arrangement and on a modified version involving counterflow separation. From results obtained it is concluded that such cells and the associated coulometric instrument are capable of high measurement accuracy.

F

OR COMPELLING FUNDAMENTAL REASONS Tutundzic (17) has rec-

ommended that the coulomb be introduced by international agreement as the ultimate standard for all volumetric work, in place of various chemical standards. This is a sound proposal because the Faraday principle allows the total amount of electrochemical change in a cell to be known as accurately as the flow of electrons can be measured in the external circuit. A virtue of the method lies in the fact that results are referred to physical values-e.g., volt, ohm, second-which can be maintained accurately in any laboratory. From their definitive study Taylor and Smith (16) concluded that coulometric acidimetric titrations can be made with a precision and accuracy equal to, or exceeding that, obtainable by the most careful conventional methods of analysis. That paper and later ones also emanating from the National Bureau of Standards (12, 13) are of unusual significance in directing attention to the analytical capabilities of coulometry. Unfortunately, the work a t the Bureau has not encouraged widespread use of coulometry for standardizing purposes. Their apparatus appears too complex and their procedures seem too tedious to gain general acceptance for the method. Their work presumably was intended to support the Bureau’s standard sample program, and not to popularize coulometric standardization as a method for general laboratory use. 1534

ANALYTICAL CHEMISTRY

It would appear that few laboratories, if any, are applying coulometry today to establish titer values of standard solutions. This seems regrettable in view of the advantages of the method. h’ot only does the method offer improved margins of reliability, but also significant savings in time and skill can be gained by judicious introduction of coulometry into standardizing operations. These advantages would pertain in the case of laboratories where much standardizing work is done, and also in the individual, nonroutine case, where a worker may spend time in exploring and validating a conventional procedure. Taylor and Smith (16) have said that in coulometry reliable results are less dependent upon the manipulative skill of the analyst than are classical methods. Eckfeldt and Kuczynski (6) have compared the operational steps in standardizing solutions conventionally and by coulometry and have concluded that significantly fewer and simpler steps are involved in coulometry. Even with conventional procedures using NBS standard samples, more time and operator skill are needed than in corresponding coulometric procedures. There has been serious concern regarding the reliability of conventional standardizing methods as routinely carried out in industrial laboratories throughout the country. Committee E-15 of the American Society for Testing and Materials has pursued the subject and issued a tentative standard ( I ) , a purpose of which is t o improve standardizing results. That publication presents results obtained from extensive interlaboratory testing, showing that only moderate levels of precision can be expected in a number of common standardizations, Certain obstacles have hindered the adoption of coulometry. Formerly, the lack of instrumentation ruled out the coulometric approach. With the advent of modern electronics the picture has changed, and today coulombs can be easily and accurately measured. Another deterrent is the question of cell performance. For standardizing purposes, cells must operate with 100% efficiency, and this means that separation of anodic and cathodic action must

be effective, a t the same time avoiding loss of sample material. Most workers in coulometry have used shield tubes with restricted solution flow through porous frits to achieve these objectives. Taylor and coworkers (12, 13, 16) use a four-compartment cell, with restrained diffusion between compartments, and conclude a titration with a series of rinsing operations of the intermediate compartments. Their work implies that excellent cell performance cannot be achieved by the simpler shield-tube approach. A primary purpose of the present paper is to challenge this impression. The paper summarizes information obtained from an intensive study of cells of the shield-tube type. Careful coulometric measurements of acid content were made on two of the sample materials measured by Taylor and Smith ( I C ) , thereby allowing a critical comparison t o be made with their values and other data, including the precise titrimetric study done by Bates and Wichers (2) involving single-crystal benzoic acid. I n addition, numerous subsidiary tests were carried out on cells and other parts of the equipment to determine the significance of possible errors and interferences. The present study should help in evaluating the performance capabilities of simplified coulometric equipment, and may encourage application of the method. EQUIPMENT

Measurement of Coulombs and Mass. The arrangement of electrical equipment used in the work is schematically shown in Figure 1. A Leeds and Northrup 7960 coulometric analyzer furnished the titration current and gave the coulomb reading (in chemical microequivalents) from measurement of the time during which a constant predetermined current was delivered to the cell ( 5 ) . The synchronous timing motor of the 7960 instrument can be optionally driven by the power line or by an external frequency standard. A Model 2005 frequency standard manufactured by American Time Products, Inc., was used in the present work to avoid uncertainties of the power line frequency. The overall circuit was checked to ensure that there were no stray currents to vitiate the coulometric readings.

vT

ll5-Volt, 60-Cycle Pover Line

I

f

I

I

I

7597

Storage Battery

7

7553

7308 Standard Coll

(Type K-3) Potontiomet*r

I Galvanometer

I Potentiometric Circuit A L - - - - - - - -----------

1

I I

I

I

, I J

p%E--kpy-iI I

Potentiometer

I Potentiometric Circuit B L

I

I I I

I I

500-ML GLASS FLASK, ESERVOIR FOR BRIDGE TUBE SOLUTION

Spring-Open Type Svitch

____________--_--_-_--

Figure 1.

I I

_I

A Figure 2.

Special cell with counter-flow separation

Block diagram of electric circuit

(Except where noted, model numbers refer to L&N equipment)

The current value was measured continually during a run using instrumentation as shown by potentiometric circuit A i of Figure 1, in which the 4025-B-S resistor was usually a 20-ohm unit, but in some measurements a 15.844-ohm unit was used. Current deviation of + O . O O l ~ or ~ less could be readily observed by galvanometer deflections. The coulometric analyzer by its manner of operation produces a d.c. current having an appreciable ax. component. The integrating capability of the 2430-11 galvanometer (period of 3.4 secondi) was investigated using for compari on an L&N 2285 galvanometer of 9.8-second period and was found to be satisfactory within the limit of error of the current measurement. Potentiometric circuit B of Figure 1 was tested because of the apparent advantage it had in not requiring high accuracy in the potentiometric instrument. I n this circuit the 4025-B-S resistor of 15.844 ohms was designed to give a n IR-drop voltage equal to standard cell voltage. However, the circuit introduced a measurement error of about 0.05'%, presumably arising from nonequal characteristics of the standard cell to the fomard and reverse current components. Accordingly, circuit B is not recommended for measurements of high accuracy. The coulomb measurement was based on standards of time, voltage, and re-

sistance, which were carefully checked for fundamental value and stability. The frequency standard was checked three times using time signals from Radio Station C l X of the Canadian National Observatory and was found to be slightly fast by about 0.0005%. (The signal from that station was much stronger in our particular location than that from the NBS station.) The synchronous motor and digital readout of the 7960 instrument were used as a clock in making the comparisons. The start-stop error of the motor and counter, measured with a Berkeley 5500 counter and timer (operating from an intermediate mercury-wetted relay), averaged 0.017 second. This error amounted to only 0.011 peq. per startstop on the high current range and one tenth of this on the intermediate range. I n performing a titration only a single start-stop was used with the high current. T o bring the solution to a precise end point the instrument was switched to the intermediate current range, and no more than 10 start-stops were ever needed on that range to complete the titration. Hence the total start-stop error of a titration never exceeded about 0.02 peq. I n a run of 2000 peq. this error amounted t o only 0.0017c and was partially compensated by the slight frequency error. The standard of voltage was maintained using three different L&K 7308 standard cells, a t least one of which

was calibrated every several months by the L&N Standards Laboratory. That Laboratory also furnished calibration data on the 7553 potentiometer used. The same Laboratory checked each of the 4025-B-S resistors twice during the work. The 7960 instrument produced a titration current of much better constancy than the advertised =tO.O5%. The advertised tolerance assumes maximum change in both line voltage and cell resistance, a situation that was not encountered in the present work and which is rarely, if ever, encountered in practice. Furthermore, every 5 to 20 minutes or so the current was observed, and deviations arising from gross changes in line voltage were corrected by adjusting the Variac transformer. The actual current thus rarely deviated by more than 0.002% from the desired value. The overall limit of error of the coulomb measurement in the runs (including time errors) was probably no greater than ~ k 0 . 0 0 4 7 ~ . The standard of mass was a 1-gram weight of Class h l type, checked by the Sational Bureau of Standards to Class M specifications in 1952 and 1956 and to Class S specifications in 1960. Using this standard, the set of balance weights was calibrated twice during the work, special care being given the calibration of the 1-gram weight of the set. Titration Cells. The cells for t h e coulometric titrations mere made u p largely of components from a n L & N 7961 coulometric titration cell kit. I n all cases t h e titration vessel was VOL. 37, NO. 12, NOVEMBER 1965

1535

a 122043 beaker provided with a 067513 cover and 127182 retaining ring. T h e titrating electrode was a 117137 smooth platinum wire helix. The cell outlet was provided with a spray trap comprising a 5-mm. diameter glass tube, 31/2inches long, having two 10-mm. diameter bulbs blown in its midregion. The secondary or counter electrode was made by winding a 0.081inch diameter pure silver wire into a 24-turn helix, 9/32 of an inch in diameter and 2'/8 inches long. I n most of the runs the secondary electrode was used conventionally, inside a 067514 shield tube, but in one series of runs it was located in a special arrangement as shown in Figure 2. End points in the titrations were determined electrometrically, using an L&h' 7664 p H meter with glass electrode. The bridge tube of Figure 2 was constructed of glass from a standard 067514 shield tube by adding a ground joint at the top and a bent side tube terminating at the bottom in a frit of medium porosity, as shown. When the equipment was in operation, a solution of potassium chloride (saturated at 25' C.) filled the bridge tube and partially filled the secondary electrode beaker, The solution flowed at the normal slow rate of a few milliliters per hour into the titration chamber through the original shield-tube frit, and a t a faster rate of about 3 ml. per

Table 1. Effect of Distillation Rate on the Composition of Constant-Boiling Hydrochloric Acid

(At pressures in different runs ranging from 752 to 760 mm. Hg) Hydrochloric acid content" Distillation rate, Observed value X 100 Literature value ml. per min. 7 100.36

(

7 3.6 1.8 2.0

100,26 100.11 100.03 100.03

a Analysis by coulometry with only moderate precision (10.0570).

Table II.

Constant-Boiling Hydrochloric Acid Samples

Sample No. 1 No. 2 Average pressure During collection of sample fraction, mm. Hg, corrected 753.2 752.7 Composition Air wt. of constant-boiling distillate that contains 1 mole vacuum wt. of HC1, conventional data, grams 180.047 180.037 Present investiga179 958 179.947 tion, grams

1536

ANALYTICAL CHEMISTRY

minute into the secondary electrode beaker through the side tube frit. The reservoir automatically replenished the bridge tube solution and maintained a constant head. SAMPLE MATERIALS

Potassium , Hydrogen Phthalate. T h e potassium hydrogen phthalate material used in the work was a sample of Standard 84-d from t h e National Bureau of Standards. Although t h e original certificate for 84-d material stated a n assay figure of 100.04%, careful comparative titrations by Bates and Wichers (2) involving single crystal benzoic acid as reference standard indicated that the 84-d material had a purity of 99.987% (or 99.988% if a somewhat questionable correction factor is introduced). Taylor and Smith (16) coulometrically found a value of 99.977y0 which was later revised (16) to 99.989% through adjustment for a then more recent value of the faraday, 96,490.0 coulombs per gram equivalent, and for a revised value of the equivalent weight of potassium hydrogen phthalate of 204.228 grams per equivalent. This material was prepared for use, as recommended by the Bureau, by gentle crushing and oven-drying for 2 hours at 120' C. Constant-Boiling Hydrochloric Acid. Constant-boiling hydrochloric acid was prepared as described by Hillebrand, Lundell, Bright, and Hoffman ( g ) , except that in the distillation equipment a 50-cm. condenser was used instead of one of 75 cm. The shorter condenser provided adequate cooling under all encountered conditions of distillation. I n preliminary work i t was found that if the distillation rate was as high as 8 or 10 ml. per minute, as recommended, the sample composition deviated significantly from the accepted value. (The samples were rapidly analyzed by coulometry, precision for this purpose being relaxed to about +0.05%.) The hydrochloric acid content a t low distillation rates was reproducible and slightly lower in value than a t high rates, as shown in Table I. Other workers have found distillation rate to have the reverse effect ( 8 ) or t o have no effect (3). I n the present work the high results a t fast distillation rates are explained by inefficiency of the distillation process, inasmuch as the starting acid had a somewhat greater acid content than the azeotropic composition. (If the starting mixture had been of lower concentration than the azeotrope, the effect of distillation rate would presumably have been the reverse of that observed.) These observations support the findings of others who question the reproducibility of constant-boiling hydrochloric acid as commonly prepared (11). Two samples of constant-boiling mixture were carefully and independently prepared for coulometric analysis. Each of these was the product of redistillation, The procedure was to combine fractions from several first distillations, all done at nearly the same pressure, to make

the starting material for a redistillation operation, carried out at nearly the same pressure as in the first distillations. I n all instances the distillation rate did not exceed 2 ml. per minute, and the atmospheric pressure changed by not more than 2 mm. during the entire course of each redistillation. The pressures recorded in Table I1 represent the critical collection period of the redistillation. The conventional assay values of Table I1 were obtained by interpolating the accepted data (8, 9) for the particular pressures involved. PROCEDURE

Measurement Technique. The atmosphere was excluded from t h e cell during titrations and pretreatment operations by closing unused openings of the cell cover and flowing nitrogen gas a t a rate of 50 to 100 ml. per minute through the space above the solution. Under these conditions gas interchange with the stirred solution took place rather rapidly. (Gas was never bubbled through the solution itself.) Cylinder nitrogen of prepurified grade was treated by passing it first through a silica tube containing copper and cupric oxide heated to about 900' C. It was next passed through scrubbers containing, respectively, a sulfuric acid solution, a sodium hydroxide solution, and then two containing mater. The gas line was of glass, closely buttjoined using short lengths of plastic tubing. (In preliminary work, exposed lengths of one kind of plastic introduced a slight amount of acid impurity into the cell.) The shield tube solution was a saturated potassium chloride solution, made from reagent grade salt (Baker's Analyzed) in distilled water. A fresh portion of solution was used in each run, and its initial p H was adjusted in the range 6.8 to 7.2. The secondary electrode was cleaned free of silver chloride prior to making a run. Some runs were carried out with the shieldtube solution stirred, by passing nitrogen gas through a 3-mm. polyethylene line immersed to the bottom of t h e tube. The cell solution (where titration occurred) consisted of 1 gram of special potassium chloride (Baker and Adamson's Code 2632, biological grade) dissolved in 50 ml. of water, redistilled from alkaline permanganate, the concentration being held to + l % . A fresh solution was used for each run. I n preparing for a titration, the cell solution was initially acidified with hydrochloric acid to a p H of 4.5 to 5.0 and while being stirred was allowed t o remain a t that pH for 15 or 20 minutes with nitrogen gas flowing to expel1 carbon dioxide. The solution was then coulometrically neutralized to p H 7.00 + 0.10 and was again allowed to stand for about 20 minutes during which the p H was observed. If the solution p H drifted or if the sensitivity was low, measures were taken t o correct the situation. Sample materials were added as described under the particular determinations,

Table 111. Coulometric Assays of 84-d Potassium Hydrogen Phthalate Group No. 2

Shieldtube Shieldsolution tube not solution agitated agitated

KO.1

Assay,

70

Nos.

1& 2

combined

Assay, c7

/t

100.001

99.997 99.999 99.997

No. of runs 6 4 Rel. std. 0.003 0.002 dev., 70 Mean, 70 99.999 99.999 I-alues from KBS Taylor and Smith, coulometric, 7, 99.989 Bates and Richers, titrimetric, 9, 99.987 Mean, 7c 99.988 Difference (present, v a h e - NBS mean) Yo +0.011

RESULTS

The results of the coulometric determinations on potassium hydrogen phthalate are summarized in Table 111, and those on constant-boiling hydrochloric acid, in Table IV. The calculations used a faraday value of 96,490.0 coulombs per gram equivalent ( 4 ) and 1957 atomic weights (18). I n making the calculations the best figures for the values of the working standards of time, resistance, voltage, and mass were used, pertinent calibration and correction factors being applied, including that for air buoyancy in the case of the phthalate weighings. For comparative purposes, precise assay figures obtained by other workers are included in the tables and where applicable these are based on values of the faraday and atomic weights corresponding to the constants used in the present calculations. Figures for the last column of Table IV were obtained by dividing 100 times the presently accepted “air weight” value (8, 9) by the coulometrically obtained “air weight” value. Going to carbon-12 atomic weights (10) and the corresponding, recent value of the faraday (14) makes only slight change in the respective coulometric assay figures, increasing them by 0.003$7&

+

99: 994 99.998 100.003 99.999 100.002 99.998

... ... , . .

’

identical starting materials were used in all instances, the p H of the inflection point should have stayed the same from one run t o the next. The inflection point was carefully established from differential plots obtained by graphing ApEIApH against [pH (ApH/2)] for two runs. The values so obtained closely agreed with each other and were averaged to arrive at the stated figure. Hydrochloric Acid Determinations. Coulometric measurements were made on t h e two samples of constantboiling acid (1 and 2 of Table 11) using t h e conventional cell arrangement. On t h e second sample, measurements were also made using t h e special cell arrangement of Figure 2. The sample solution for analysis was prepared by diluting the constantboiling acid about tenfold with water. The operation was carefully done on a weight basis, using a 50-ml. weight buret as solution container. The amount of acid solution added to the cell was measured with the same weight buret. Weight burets were handled by special technique t o improve precision. I n addition to using finger cots, the outer surface of a buret before a weighing was lightly stroked with a pliable metal brush connected to ground to avoid electrostatic forces during the weighing operation. Variations arising from evaDoration loss from the outlet tip (eve; though covered by a cap in the usual way) were reduced by a procedure involved drawing the solution back from the tip. I n coulometric titration of hydrochloric acid, the final end point was at p H 7.00.

10 0.003 99.999

I n making a coulometric titration, 99% or more of the reaction was carried out in one continuous stretch using the high current. After recording the reading, the instrument was set to zero and switched to the intermediate current range, and the solution was brought to the end point by a series of small coulometric increments. Prior to each of the final two or three increments the cell contents were shaken to wash down the inside surfaces. The gas flow was stopped temporarily in the shake-down operation. The final shake-down was done vigorously enough t o cause some solution to enter the bulbs of the cell trap, solution then being allowed to drain back into the cell. The cumulative coulometric reading on the intermediate range was added t o the previous figure to obtain the total microequivalent reading for the run. Phthalate Determinations. The conventional cell arrangement was used for t h e phthalate determinations. Samples almost exactly one gram in amount were weighed t o a limit of A tared platerror of *0.003%. inum weighing boat was used with t h e specially calibrated 1-gram piece as t h e sole balance weight. T h e cell beaker with solution was removed from the cell assembly, the phthalate was added to the solution, the beaker was returned t o the cell assembly, and the weighing boat was reweighed. The titration was terminated at p H 8.65, which corresponded to the inflection point of the titration curve ( 2 ) . Inasmuch as the same amounts of

DISCUSSION

Subsidiary Tests On Cells. Details Of Cell operation were investigated t o detect and eliminate possible cell errors. Precautions as described b y

Table IV.

Coulometric Analysis of Constant-Boiling Hydrochloric Acid Samples .4ir wt. Coulometric Sample per (vac. wt.) assay comparison with Hillebrand air wt., HC1 mole of “21, grams et al., 7‘ grams found, peq. Sample No. 1 (dilution factor 0.089578) 179 962 100 047 Series A 7 7072 3836 37 179 927 100 065 Conventional 4 7914 2385 41 179 973 100 041 cell 7 6063 3785 90 6 7 4 4

1733 4899 5405 3947

3073 3728 2260 2187

04 08 20 37

hIean Rel. std. dev., Yo

Sample Xo, 2 (dilution factor 0.113741) Series B 3,2296 2041.28 Conventional 4.4992 2843.63 cell 5.1743 3270.63

179 179 179 179 179 0

949 967 953 074 958 009

179.955 179.962 179,944 5.1456 3252.51 179,943 3.4824 2201.32 179.934 Mean 179.948 0.006 Rel. std dev., Yc 179,920 3.5286 2230.69 Series C 179.965 4.1493 2622.43 Counterflow 3413.46 179.919 5.3995 cell of Figure 2 3.2057 2026.13 179.959 179.963 2.8785 1819.28 179.947 10.9973 6951.17 Mean 179.946 0.012 Rel. std. dev., yo 179.947 Series B and C groups taken together, mean All three groups taken together, mean

100 100 100 100 100 0

054 045 052 042 049 009

100.046 100.042 100,052 100.052 100,057 100.050 0.006 100,065 100.040 100,066 100.043 100.041 100.050 100.051 0.012 100.050

VOL. 37, NO. 12, NOVEMBER 1965

1537

Taylor and Smith (16) were taken into consideration, and one or more factors not mentioned by them were found to be important. SAMPLEDISPERSEMENT. The possible loss of trace quantities of sample material is a matter for concern. I n acid determination, gas evolution occurs at the electrode, causing gas bubbles to burst through the solution surface and to deposit droplets of solution, and hence a small amount of sample material, on the surfaces of the cell enclosure above the solution level. I n concluding a run the shake-down operation was therefore important to ensure that all of the exposed surfaces were thoroughly rinsed by solution. I n all of the runs at least a small amount of sample material was recovered, and in a few instances the recovery amounted to almost 0.1% of the total sample quantity. I n no instance, however, was any effect observed in rinsing the cell outlet trap, thus indicating that sample dispersement was restricted to the interior of the cell. I n the hydrochloric acid work a sensitive test on the effluent nitrogen gas was performed, with negative results, to see if traces of acid were being lost by mechanical carryover or as a result of liquid-vapor equilibrium. pH DRIFTS. As Taylor and Smith have pointed out a useful technique for studying cell operation is to observe the p H stability of the solution a t equivalence prior to and after a titration. At 7 pH the unbuffered solution in the hydrochloric acid work was very sensitive to traces of acid or base, and in preliminary work both upscale and downscale p H drifts occurred in a most perplexing way. Possible causes were considered and tested by removing one cell element a t a time and observing the consequent effect on the pH drift. Results of that study are summarized in the immediately subsequent paragraphs.

smooth-wire electrode gave greatly improved pH stability, it nevertheless was responsible for a very slight amount of p H drifting, possibly caused by microimperfections in its surface. The magnitude of the effect, however, was negligible with respect to total quantities titrated, and this electrode therefore served quite satisfactorily. CELL MATERIALS. Tests showed that the Teflon encapsulated stirring bar and the glass and reference electrodes did not cause end-point drifting. The dual-glass junction of the L&Y 1199-31 reference electrode is of favorable design and would not be expected to introduce sorption errors. Its leakage rate is very low. The solution did not normally come in contact with the cell cover, but the polyethylene construction of this and associated parts caused no difficulties. Simplicity of cell design, its closed construction, and the inertness of the parts all contributed to cleanliness and protected against errors from sample loss or contamination. SHIELD-TUBEFRIT. I n principle a shield-tube cell depends on making electrical connection between anolyte and catholyte by means of a stream of electrolyte solution flowing toward the titration zone through a porous flow restrictor (a frit) acting as a separator. Sample material must not enter the frit, else i t is lost from titration. Loss of sample across or within the frit might occur as a result of various mechanisms including diffusion, electrical migration, and electro-osmotic solution flow. Diffusion and electrical migration are both slow processes, and the bulk flow of solution through the frit caused by hydrostatic heat difference can be made sufficiently large to overcome any tendency for sample loss to occur by these mechanisms. I n the present work electrical migration of sample ions was toward the titrating electrode. Hence PLATINUM TITRATIXG ELECTRODE. this factor acted in a way to prevent I n preliminary work a platinum titrating sample loss. Tests were carried out to learn to what extent in the course of a electrode of composite construction was normal titration hydrochloric acid used. T h a t electrode was fabricated in passed through or was sorbed by the the form of a cylindrical cage, using platinum strips spot-welded together. frit. h miniature glass electrode (L&N 124138) was placed in the solution inside With that electrode, stable p H readings a shield tube, with the pH bulb resting in unbuffered solutions were hard to on the frit. The shield-tube solution obtain; the p H sometimes slowly drifted for 3 or 4 hours or longer, and was unbuffered and hence highly minor titration errors were introduced. susceptible to traces of acid. No The trouble was finally traced to sorpevidence of acid transfer through the tion and desorption of substances in the frit was noted. Moreover, the frit fissures of the electrode surface. The after exposure to acid solution in a sorbed substances might have been titration did not show a slow release of sample material, products of electrolysis, acid to delay attainment of a stable end or residues of strong cleaning agents point. The last observation indicates that partial permeation of the frit by previously used on the electrode. Although only very small amounts of acid, as might have occurred in regions material were involved, the condition of minimal solution flow, w&s not a serious matter. was intolerable for precise work. T o Electrc-osmotic solution flow can be correct the situation the electrode was kept small in magnitude by using an changed to the smooth-wire design previously mentioned. Although the electrolyte of high conductivity, where1538

ANALYTICAL CHEMISTRY

by important flow-influencing variables (the electm field gradient and the zeta potenthi) are diminished. Certain electrolytes show less electro-osmotic flow effect with glass frits than do others. The potassium chloride solution of the present work was favorable in this respect. The electro-osmotic flow effect was tested under normal head conditions by observing the net solution flow through a frit, with and without cell current passing. Flow of electricity did not significantly change the solution flow rate, thus indicating that the electroosmotic effect was not a serious factor. SHIELD-TUBE IMPURITIES. In a shield tube cell, the electrolyte flowing through the frit must not be allowed to carry with it into the titrating zone any reactive or interfering substances, such as might come from a solution ingredient, the secondary electrode reaction, or foreign matter trapped in the frit. Ironically, trouble with dirty frits often comes from the use of cleaning agents, as Taylor and Smith have pointed out. Frits must be very thoroughly rinsed after cleaning them with strong reagents. I n the present work the shield tubes were only seldom treated with strong reagents (followed by thorough rinsing). Usually, they were kept in good condition by letting them stand between runs in a frequently changed supply of shield-tube solution. As might be expected, the flow of the neutral shield-tube solution through a well-kept frit, into the cell solution, did not cause any pH drifting. When a cell is in operation, neither the titrating nor secondary electrode must be allowed to touch the frit. If this occurs the action of the shield tube may drastically change, and interfering effects may thereby be introduced into the titration. The 7961 cell design guards against this occurrence. The conventional shield tube arrangement when used as described was effective and practical for runs of a t least 5000-peq. extent, but a word of caution is in order. If the potassium chloride concentration or the anode area is made substantially less than indicated, there is a possibility that the shield-tube solution might start to become acidic at a microequivalent figure less than that stated, through discharge of hydroxyl ions. Objectionable acidity would consequently be carried into the titration. When conditions were as recommended, however, measurements with a miniature p H electrode showed that the solution stayed neutral. Nitrogen gas stirring of the shield tube solution proved to be unnecessary. The secondary electrode introduced a small amount of dissolved silver chloride into the shield-tube solution, and accordingly a trace of silver chloride found

its way through the frit and into the titration zone. The presence of silver chloride in the titration solution is objectionable, because electrode-reduction of silver ions will compete with the desired reduction of hydrogen ions, and a titration error may thereby be introduced. The flow of solution through the frit must be held to a moderate rate to minimize the introduction of silver chloride. The flow rate through a frit favorably tended to diminish as the frit was used, presumably as a result of slight deposition of silver chloride in the region of the frit where the emerging solution encountered a lesser chloride ion concentration. Accordingly in the later runs (some of the phthalate and all of the hydrochloric acid runs) the flow rate was satisfactorily of the order of 0.3 ml. per hour or less. Even after extended shield-tube usage no difficulties were encountered from excessive electrical or solution-flow resistance. The magnitude of the silver error was investigated during the phthalate runs by making analytical tests. After a period of use the amount of silver on the cathode was determined by dissolving the silver in nitric acid, precipitating it as silver chloride, and estimating the silver chloride nephelometrically using known silver nitrate concentrations as standards. Although in the first measurement silver mas found equivalent to a +O.OOS% error, the observed amounts exclusive of this initial high value averaged +0.004%. COUNTERFLOW SEPARATION. Because the results obtained in the Series d and B runs on hydrochloric acid (with the conventional shield-tube arrangement) were perplexingly high, it was decided to take drastic measures to avoid secondary electrode effects, particularly the silver error. Impurities from the action of the secondary electrode can be eliminated altogether by employing the principle of counterflow separation. I n this technique the anode and cathode chambers are connected bp a bridge of electrolyte in which flow of electrolyte occurs in two directions-toward the titrating electrode, and toward the secondary electrode. Although the principle can be carried out using a, small shield tube inserted inside a larger one, the equipment illustrated in Figure 2 affords more positive flow of the bridge tube electrolyte. The effectiveness of the Figure 2 equipment was proved in a preliminary test. An indicator was mixed with the bridge tube solution. The secondary electrode solution was then acidified by adding some hydrochloric acid to it. A small portion of the acidified solution was cautiously sucked through the porous frit and into the bridge tube until the line of color demarcation stood 6 or 8 nim. above the frit. Normal cell

operation was then established by turning on the current. The line of demarcation between the acid and neutral solutions slowly fell, and after several minutes no trace of acid could be seen inside the bridge tube. It is believed that this test conclusively demonstrates that a product from the secondary electrode chamber (silver chloride, for example) could not have found its way into the titration chamber when the equipment was in operation. (It is necessary to use a solution of uniform temperature and concentration inside the bridge tube, otherwise density differences may lead to convective currents that can upset the intended flow pattern.) IMPURITIES OF SOLUTIOXS.An ingredient used to make up a solution can conceivably be a source of an impurity that introduces error through competitive reaction a t the working electrode. I n the acid titrations, for example, interference could have been caused by reduction of a heavy metal ion or an ion such as Fe+3 going to Fe+*. (In acidbase titrations certain kinds of electrooxidizable or reducible impurities need not cause concern if the reaction with the impurity occurs with the equivalent generation or consumption of base or acid. Many reactions involving free or combined oxygen are of this type.) An interfering impurity is of less concern in the shield-tube solution than in the titration solution, because only a small volume of shield-tube solution enters the titration chamber during a run. Commercially available potassium chloride of especially pure grade was used for making up the cell solution. Calculations based on the impurity limits quoted for this reagent indicated that any impurity error should have been negligible. However, an experimental way was open to test for impurity error. If interfering impurities were present, the cell solution should have improved u-ith usage a3 the impurities became used up. The hydrochloric acid runs were divided into two groups according to sample size, as shown in Table V. The fact that large and small samples showed identical assay values is strong evidence that there was no error from solution impurities, This conclusion applies equally to the phthalate work, because the solution ingredients were the same in both cases. With regard to possible contamination from the 1199-31 calomel reference electrode, calculation shows that the amount of mercurous compound entering the titration solution is many orders of magnitude below a significant level. END-POINT UNCERTAINTY. The technique used to establish the stoichiometric point of the coulometric phthalate titration could have been in error by enough to cause as much as 0.005%

Table V. Comparison of Results on Large and Small Hydrochloric Acid Samples

Av . sample No. of size, runs peq.

Coulometric wsay comparison,a

Sample type 7. Small 9 2200 100.050 samples Large 9 3800 100.050 samples a As expressed in Table ITr,

error. I n the hydrochloric acid work, however, the end point was very sharp and did not introduce significant uncertainty. Comments on Phthalate Results. As evident in Table 111, the phthalate determinations are in excellent agreement with the NBS average on the same sample material. The difference is only 0.011%. Nevertheless, the precision of our work and attention given to details may permit comment. The discrepancy is in the wrong direction to be explained b y sample loss; indeed, the cell study revealed no mechanism for sample loss. The positive discrepancy can be explained by a competing electrode reaction. From cell tests, reaction with the potassium chloride solution substrate appears unlikely, except with respect to silver chloride. Nephthelometric tests indicated, however, that the silver error probably did not exceed 0.004%, which is less than half the total discrepancy (see remark regarding silver error under “Comments on Hydrochloric Acid Results”). The remaining difference of 0.00770 may be within overall experimental error. The high assay value obtained on phthalate nevertheless raises a question, especially when it is considered in conjunction with a revised figure for the Taylor and Smith coulometric measurement of benzoic acid (16). The Taylor and Smith work, as well as our own, was done on certain sample materials as used by Bates and Wichers (2), hence allowing secondary comparison of coulometric results with a presumed 100% pure acidimetric standard, namely single crystal benzoic acid. Using a newer value of the faraday, 96,490.0, and an equivalent weight based on 1957 atomic weights, we recalculated the Taylor and Smith b-Series value for benzoic acid and obtained a revised assay figure of lOO.OlO%, which stands 0.014% greater than the Bates and Wichers value for the same material. Going to the latest value of the faraday, 96,487.0 ( I d ) , makes the situation worse. The Taylor and Smith coulometric measurement of benzoic acid now becomes 1 0 0 . 0 1 3 ~ ~ , VOL. 37, NO. 12, NOVEMBER 1965

1539

or 0.017% higher than the Bates and Wichers figure of 99.996% for the same material. Our coulometric assay of 84-d phthalate now becomes 100.002%, which is 0.015% higher than the Bates and Wichers figure of 99.987% for the same sample material. The Taylor and Smith coulometric assay of 84-d phthalate now becomes 99.992%, or 0.005% higher than the corresponding Bates and Wichers figure. [The recalculated S B S coulometric assay figures given here have recently been confirmed ( 1 5 ~ ) .The use of carbon-12 atomic weights (IO) does not significantly change the picture.] The three coulometric assays average 0.01270 higher than the Bates and Wichers values, as referred to their acidimetric reference standard, single crystal benzoic acid. This discrepancy would appear to lie outside of error limits, and hence the matter deserves future attention. It should be understood, of course, that practical application of coulometry need not wait for resolution of these minor differences. Comments on Hydrochloric Acid Results. As evident in Table IV, t h e hydrochloric acid results are very consistent, although precision is poorer t h a n with phthalate because of t h e weight buret limitation. Almost exactly t h e same assay comparison percentages were obtained in Series A, B, and C. The agreement between Series A and B shows t h a t the two independently prepared constant-boiling samples in this work were highly reproducible in composition. Agreement between Series B and C, with no possibility of silver error being present in the Series C work, shows that there was negligible silver error in the Series B runs. The slight difference betw-een the averages of the B and C runs is in the wrong direction to be explained by silver error in the Series B work. The table indicates, however, that the prese n t coulometric values are 0.05% greater than the presently accepted concentration data for constant-boiling hydrochloric acid ( 8 , 9 ) . , The work on phthalate shows t h a t cells of the type used in the present work are reliable in coulometric acid measurement. The phthalate measurements agree with NBS results within about O.Ol%, which is a small difference compared with the o.0570 discrepancy in the acse of hydrochloric acid. There are no apparent reasons to expect significant

1540

ANALYTICAL CHEMISTRY

competing reactions or other interferences in the hydrochloric acid measurement. Hence it is inferred that the values obtained in the present work on hydrochloric acid are true assay figures for the constant-boiling mixture. Accordingly, it is concluded that the presently accepted assay for this acidimetric standard (8, 9) is low by 0.05%. Going to the latest value of the faraday, 96,487.0 (14) increases our value by about 0.003%, thereby slightly widening the discrepancy. The above conclusion is reached despite the fact that Taylor and Smith (16) found good agreement between their coulometric assay and the usually accepted data on constant-boiling hydrochloric acid. The composition of the misture can deviate appreciably from a reproducible value unless special care is taken in its preparation. The slow distillation rate used in the present work is recommended for its preparation. The same shield tubes and cell arrangement were used in the Series A and 13 runs and in the earlier runs on phthalate. The evidence of no silver error in the Series A and B runs on hydrochloric acid strongly indicates that the silver error in the phthalate work, at least in the later of these runs where the flow rate of shield tube solution had diminished, was of negligible or very small magnitude.

gained by using for sample measurement the semiautomatic precision pipet described elsewhere ( 7 ) .

CONCLUSIONS

(9) Hillebrand, W.F., Lundell, G. E. F., Bright, H. A,, Hoffman, J. I., “Applied Inorganic Analysis,” 2nd ed., p. 181, Riley, New York, 1953. (10) “IUPAC Revises Atomic Weight Values,” C‘hem. Eng. ,Yews 39 (47), 42 (1961). (11) Liebhafsky, H. A., Pfeiffer, G. H., Balis, E. W., ANAL. CHEW 23, 1531 (1951). (12) Narinenko, G., Taylor, J. K., J . Res. ~Yat.Bur. Std. A67, 31 (1963). (13) Ibid., p. 453. (14) IYat. Bur. Std. ( U . S.L Tech. News Bull. 47, 175 (1963). (15) Taylor, J. K., KBS, Washington, D. C., private communication, 1960. (15a) Zbid., 1965. (16) Taylor, J. K., Smith, S. W., J . Res. Nut. Bur. Std. A63, 153 (1959). (17) Tutundzic, P. S., Anal. Chim. Acta 8 , 182 (1953). (18) Wichers, E., J . Am. Chem. SOC.80, 4121 (1958).

ACKNOWLEDGMENT

The authors express appreciation t o their coworker, A. J. Williams, Jr., for helpful suggestions in proving the integrating capability of the galvanometer in the current-measuring circuit, and to John K. Taylor of the National Bureau of Standards for supplying the 84-d sample of potassium hydrogen phthalate and for entering into helpful discussions. LITERATURE CITED

(1) American Society for Testing and

RIaterials, “Preparation, Standardiza-

tion, and Storage of Standard Solutions

for Chemical Analysis,” Tentative Standard E200-64T (1964). (2) Bates, R. G., Wichers, E., J . Res. Nat. Bur. Std. 59, 9 (1957). (3) Bonner, W. D., Titus, A. C., J . Am. Chem. SOC.52, 633 (1930). (4) Craig, I>. E., Hoffman, J. I., Law, C. A.. Hamer, W.J., J . Res. Nat. Bur. Std. A64, 381 (1960). ( 5 ) Eckfeldt, E. L., (to Leeds & Sorthrup Co.), 1:. S. Patent 2,832,734 (1958). (6)