where e represents an electron and s indicates the silver is metallic. The silver perchlorate is present in the solution to reduce to insignificant proportions the already meager spontaneous dissolution of silver into perchloric acid. The work establishing these desirable conditions was done at the National Bureau of Standards (NBS), by Craig et al [4] in preparation for his determination [5] of the Faraday.

In his paper [5], Craig postulated a set of possible reac tions which involved the chemical entities Ag*, CIO, OH, H*, H2O, O2, and so forth, which are known to be present in the solutions where the reaction of eq (2) takes place. From thermodynamic arguments he shows that the reactions of the constituents of the solutions either follow eq (2) namely at a potential of −0.799 V with respect to a normal hydrogen electrode, or have reduction potentials of at least 0.4 V lower. Hence, unless the overpotential of the dissolution of silver is high, the contribution from reactions other than the desired reaction is negligible, i.e. less than 0.1 ppm. The arguments, which our experiment demonstrates to be correct, are presented in Craig's paper and need not be repeated here. We shall take the view that if the overpotential of dissolution has measurable effect, the values of the electrochemical equivalent of the samples should vary with current density or dissolution rate. No such variation is found, as will be seen.

The formation of Ag2O and water from silver and aqueous hydroxyl ion during the reaction:

ously, as it comes to the surface of the silver anode it rea with perchloric acid. Sixteen parts by weight of oxyg would carry into solution 216 parts of silver, bypassing reaction of eq (2). Methods for eliminating oxide described in another section.

The reverse of eq (2) governs the cathodic recovery of t silver which falls away from the anode during dissoluti Details of the recovery are given below.

3. Materials

Perchloric Acid. The perchloric acid used in the cou meter was commercially available "70 percent" reage grade, which the manufacturer labeled as "distilled Vycor" and "99.9999 percent pure." The material receiv no further treatment. It exhibited no detectable attack silver over the period of an experiment, provided a sm amount of silver perchlorate was present.

Silver Oxide: The solution used as supporting electroly in which silver (within experimental error) is insoluble, is mixture of perchloric acid and silver perchlorate, prepar by adding silver oxide, Ag20, to 20 percent by weight p chloric acid solution. The silver oxide was of reagent grad purchased from a commercial chemical supply house, a was subjected to no further purification. Any solid silv from this material remaining in the perchloric acid-silv perchlorate solution was removed by filtration through so silver powder on a fine fritted glass funnel before the sol tion was introduced in the coulometer.

Water: The water used in making up the solutions w what is commonly called "conductivity" grade. The wat was distilled in a stainless steel boiler from water which ha been passed through an ion-exchange resin for pri removal of ionic impurities. In the boiler water was a sma amount of pyrophosphoric acid to hold back ammonia. Th steam so generated was passed through a scrubber | remove CO2 although no effort was made to guard the wat against reentry of CO2 after cooling. The water was co densed, kept in tin-lined storage tanks and fed to a spigot: tin-lined pipe. The result was a water which exhibited resistivity of 1.2 M 2-cm at air equilibrium. No attempt w made to remove equilibrium carbon dioxide.

Silver powder: As mentioned above, the perchloric-aci silver perchorate solution was filtered through silver powd on fritted glass. During the eight runs of the final expe ment we used a silver powder, advertised as "five nines pure, which was obtained from a manufacturer of high-pu ity metals. This was a precaution. However, during most the preliminary practice runs we used a silver powd obtained from a chemical supply house. No significant di ferences could be ascribed to this substitution of the hig purity material.

ver sample: The silver samples were part of a large lot hased by the National Bureau of Standards for certifin and issue as a Standard Reference Material. From ot, most of which was issued as a high temperature -pressure standard, were selected a few rods which characterized by residual resistivity ratio, and set for research purposes at NBS. The samples furnished by the Office of Standard Reference Materials of NBS from Rod 55 of the lot used for Standard Reference rial 748 Silver Vapor Pressure Standard. Rod 55 was ed into 45 cm lengths labeled by letter sequentially in irection of the draw. These lengths were cut into segs 50 mm long numbered again in the direction of the . The pieces furnished to us were labeled thus 55: A-7, D-7, E-7, F-7, G-Y2, G-Y3. The last two pieces were the end of section G.

e preparation of the silver sample by the manufacturer oprietary, but NBS was informed by the manufacturer migh purity silver was subjected to dissolution and elecsis, and was then melted in vacuum and formed. The ting rod is polycrystalline. No further purification was pted other than an effort to remove oxygen by heating high vacuum, a process described elsewhere in the

T.

4. Apparatus

hen one electrolyzes pure silver anodically into a soluof perchloric acid, small quantities of finely divided silver fall away from the anode. In the experiments of [4] this finely divided material was recovered, filtered weighed. It was counted as part of the final weight of node. The weighing inevitably required the collection he transfer of this anodic residue to a filter before the weighing. There exists the possibility of loss of materiing transfer. Furthermore the large scatter induced by aring of a sintered-glass filter crucible would greatly > the random experimental error.

secure an accurate value for the mass of the residue, it lecided that the residue must never leave the anode artment of the coulometer and must be treated and zed in the compartment itself. The method fixed upon he filtration of all anodic electrolyte solution through a ed glass filter embedded in the anode compartment the dissolution of the finely divided solid silver into acid, the evaporation of the nitric acid to dryness leavlver nitrate and finally the analysis for silver by cond potential coulometry. The charge consumed in the peration was added to the total anodic charge in the ant-current part of the experiment in calculating the ochemical equivalent of the silver sample. A detailed nt of this procedure is presented in reference [6].

To effect the retention of the residue in the anode compartment for direct potentiostatic analysis within the compartment, the coulometer was fashioned as shown in figure 1. The weighed silver sample was attached to an adjustable fixture which set the immersion depth of the sample in the solution of perchloric acid-silver perchlorate. The anode compartment consisted of a beaker in the side of which was a fine fritted glass disc leading to a side-arm which curved upward to one part of a spherical joint. The spherical joint was set at about the same height over the bottom of the beaker as the beaker lip, as shown in figure 1, to ensure no loss of liquid from a full beaker. The second part of the spherical joint was attached to a siphon with stopcock-not shown on top for filling which had, at the far end, a medium-fritted glass disk. The glass frit was placed at this point to prevent solid silver, which might bear an electrical charge, from migrating from the cathode compartment. The first siphon led to a large beaker which was connected to another large beaker, the cathode compartment, by a second siphon. The cathode compartment contained a large platinum mesh electrode, with an area of about 90 cm2. The resistance of this cell, cathode to anode, was about 20 9.

Several precautions were taken to minimize leakage resistance and to protect the coulometer from contamination by air-borne pollutants. The whole assembly reposed upon a sheet of 6 mm-thick teflon which in turn rested on a 12 mm sheet of polymethyl methacrylate. Passing through the teflon and anchored into the methacrylate were rods (12 mm diameter), also of polymethyl methacrylate. One rod of aluminum was also passed through the teflon sheet and was anchored into the methacrylate. The plastic rods were posts for attachments of metal claw clamps which supported the siphons. Metal was substituted for plastic in the case of the post which supported the silver anode and its mechanism of adjustment of immersion. This last measure was required to lend rigidity to the anode holder and anode. Vibration of these parts affected the current control. The resistance of the supporting parts, from the metal rod to any other part of the teflon or methacrylate sheet or rods, with acid solution wiped onto the teflon was of the order 1013 Q.

To the metal post was attached a new metal clamp of the adjustable claw type with plastic covered fingers. The clamp held the anode immersion assembly, figure 2. The assembly consisted of a glass tube, 5 mm diameter, through the length of which passed a brass rod, 3 mm in diameter, threaded at one end. The tube was held upright by the claw clamp, and a nylon nut on the end of the adjustable brass rod served to set the height of the rod. Soldered to the lower unthreaded end of the rod was a piece of brass 20 by 75 mm. The brass rod was soldered about 25 mm from the end of this last piece. This 20 x 75 mm piece was large enough to receive the clamps from the positive lead and from the sample. The motion of this brass assembly was kept linear (ver

e whole glass assembly, the supporting rods, teflon, methacrylate sheets were housed in a methacrylate box hinged doors to allow loading of the coulometer and stment of the parts during the determination.

he electrical connections required for the dissolution of silver anode at constant current are shown in figure 3. current controller, a Tinsley type 5390,3 provided lowe, low-drift, direct current. The current was ripple-free. re beginning the experiment, constant current of the red level was maintained in a substitute load, Ra, whose dance was matched closely to the impedance of the coulometer. The precise value of the current was by comparing the voltage drop across a standard tor (R,) with the voltage of a standard unsaturated cell Slight adjustments in the controller settings were until the difference in voltage as read on D, was zero V). D, is a Leeds and Northrup 9829-D high-imped10 MQ) linear amplifier which was used as a nulltor. A reversing switch was used to eliminate the effect ermal emf's.

the beginning of the experiment, mecury-wetted relays taneously transferred current from the substitute tor to the coulometer and started an electronic timer. relays were de-activated to end the dissolution. uring the experiment, the current was continuously itored by D,. The high impedance and small voltage

and names are used throughout this paper only for purposes of identification. Such use imer endorsement by the National Bureau of Standards nor assurance that the equipment is

difference across the inputs of D, allowed continuous monitoring with negligible (<0.1 ppm) uncertainty.

Since gassing at the platinum cathode of the coulometer causes fluctuations in the current, this condition must be avoided. Gassing occurs when the concentration of Ag* at the cathode becomes so depleted that H must undergo reduction in order to maintain a constant current. As a remedy, a small amount of Ag2O was added to the cathode compartment at intervals of about 12 h. The oxide. dissolved, keeping the catholyte sufficiently rich in silver ion.

The current, I, maintained in the coulometer, is simply the voltage, V, of the standard cell divided by the resistance, R,, of the standard resistor. If current flows in the coulometer for an interval of time, t, then the total charge (Q) passed during silver dissolution is

Equation (3) only holds provided transient effects may be neglected. The following examination shows that this neglect is justified.

An oscilloscope was used to demonstrate that the timer and coulometer are activated and deactivated simultaneously. The minimum duration of any of the dissolution experiments reported in this paper is 13 × 103 s. Hence to achieve an accuracy of 0.1 ppm in Q, the switching must be simultaneous to 1 ms. This rather modest criterion was easily met.

Switching transients in the current controller must also be negligible. Experiment has shown that matching the

impedance of the substitute resistor to that of the coulometer will insure a negligible transient when the current is switched to the coulometer. As the impedance mismatch is made worse, the transient increases. Switching the current back to the substitute resistor does not produce any measurable transient in the coulometer. In the worst case encountered in these experiments, the effect of transients as determined by oscilloscope measurements is approximately 0.1 ppm in Q.

There is one last transient effect which deserves mention. It is assumed in eq (3) that all current which flows is "Faradaic" although in reality a small portion is not. A small fraction of the total current goes into charging a double layer capacitance at the surface of the anode. The rest of the current is due to transport of ions across that capacitance. The capacity of the double layer is thought to be 100 μF/cm2 and is charged within the first millisecond of current flow [7]. The surface area of our anodes was 10 cm2. Oscilloscope measurements show that our anode voltage jumped by 8 mV within the first millisecond as compared to the voltage of the quiescent silver electrode. Although most of this voltage rise is probably ohmic and not due to charging of the double layer, we will assume a worst case. Hence the amount of current which goes to charging the double layer is 10 μC. In the experiments we have performed, this amounts to less than 0.01 ppm in Q.

With transient corrections to eq (3) negligible, it only remains to explain how V, R and t were measured.

A standard unsaturated cell (S.C. mentioned above) was used to monitor current constancy during the course of silver dissolution. This unsaturated cell was maintained in a thermostated enclosure with four standard saturated cells. A cable was installed from the standard cells to the NBS standard cell laboratory. By means of this cable, our saturated cells were compared at weekly intervals with standard cells whose emfs were known in terms of the NBS as-maintained unit of voltage (defined in terms of the ac Josephson effect) to 0.1 ppm. It was, therefore, never necessary to move the cell enclosure in order to calibrate the standard cells.

We determined the emf of our unsaturated cell by means of the following procedure. A voltage divider was constructed in such a way that an adjustable emf of approximately 900 μV would be produced. This is the approximate difference in emf between saturated and unsaturated standard cells. The divider consisted of precision resistors, known to ±0.01 percent, and a mercury cell. The resistors were calibrated before and after the series of Faraday measurements reported herein. Prior to a measurement, the emf of the mercury cell was measured to better than 0.01 percent, using a calibrated Leeds and Northrup type K-5 potentiometer.

By means of appropriate switches, the emf of the unsaturated cell could be compared to that of a saturated

cell plus the divider output voltage. The variable resist was adjusted until the emfs were equal as determined by th high-impedance null detector previously described. Kna edge of the divider output voltage and saturated cell permitted calibration of the unsaturated cell. Thermal e in the divider were measured under typical operating co tions and found to be negligible, i.e. less than 0.1 μV.

Immediately preceding the start of a Faraday determ tion, the unsaturated standard cell was compared with ea saturated cell using the procedure outlined above. I measurements were repeated immediately after the diss tion of the silver anode was stopped. In this way, the uns urated cell emf was known to within 0.1 ppm during: experiment. No significant change in unsaturated cell from beginning to end of a Faraday determination was detected (fig. 4).