APRIL 1983 The Analyst Vol. 108 No. 1285 Kinetic Static and Stirring Errors of Liquid Junction Reference Electrodes* Donald P. Brezinskit Corning Glass Works Sullivan Research Park Corning N Y 14831 USA Response characteristics of reference electrodes were determined by subjecting them to large changes in ionic strength and transference. Commercial elec-trodes often gave surprisingly poor performance exhibiting slow inaccurate and stirring-dependent responses. Slow response is caused primarily by diffusional entrapment of previously measured solutions within the junction. Large offsets at pH extremes or a t low ionic strength are caused by storage of no-flow electrodes in standard buffers junction charge and improper junction geometry. Stirring potentials are caused by shifts in offset error with local changes in concentration a t the junction surface.Adequate outward flow or diffusion of junction electrolyte serves to suppress these anomalies. Reference problems are often undetectable in the standard pH buffers typically used for calibration. Keywords Reference electrode ; liquid junction ; offset error; kinetics ; stirring potential Variation of the junction potential of the reference electrode is a recognised source of error in ion-selective electrode potentiometry. The junction potential is generally attributed to ionic interdiffusion at a direct (liquid - liquid) interface between the measured sample and the junc-tion electrolyte. Guggenheim1p2 showed that such diffusion potentials can be rendered minimal and reproducible by using a concentrated equitransferent junction electrolyte (typically 4 M potassium chloride solution) at a free diffusion interface with cylindrical sym-metry.The significance given to the residual junction potential under these conditions is different in the American (NBS) and British (BSI) pH scales. The multi-standard NBS sys-tem regards the residual junction potential as an error and the different primary standards are restricted to a range over which junction potential variation is rather negligible (<1 mV).S The single-standard BSI system incorporates the variation in junction potential into the defined pH v a l ~ e . ~ ~ ~ Whatever the theoretical interpretation of the junction potential the significance of practical measurements is based on the presumption that non-negligible junction potentials will be reproducible and equal to those yielded by a standard (definitive) junction.However reference errors (departures from standard junction potentials) seem likely as one proceeds from the optimised solutions and methodologies defining the pH scale into the much broader domain of practical pH measurement. The usual calibration standards being formulated with the intent of minimising junction error are generally moderate with respect to pH buffer value ionic strength and transference; however it is common practice to measure solutions for which these parameters are extreme (e.g. de-ionised water acid plating baths, colloidal soils) Also commercial reference electrodes employ a wide variety of junction materials geometries and electrolytes and as a group bear little resemblance to each other or the definitive junction.Small but significant errors (<3 mV) have been observed with commercial reference electrodes in standard pH buff ers.5 Much larger deviations might be expected in non-standard solutions ; accordingly Illingworth6 recently reported concentration-dependent errors averaging 0.2 pH per decade in a sampling of porous ceramic junction refer-ence electrodes. Although there have been numerous studies pertaining to the magnitude and * Presented a t the 32nd Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Atlantic City NJ March 9th 1981. t Address for correspondence 54798 CR653 Paw Paw MI 49079 USA. 42 426 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol.108 stability of liquid junction potentials @er se,7-lo the kinetics of practical reference electrodes have received scant attention. In particular their contribution to measurement drift and stirring dependence has not been elucidated. Finally reference electrodes with a permanent, gelled electrolyte have become popular primarily because they eliminate electrolyte flow and the consequent need for re-filling. Although this innovation is convenient and removes an unnecessary constraint on junction design namely control of flow its impact on performance remains uncertain. The purpose of the present study was to determine the general response characteristics of practical reference electrodes and how they might be improved. A rigorous test procol was established and applied to a representative selection of commercial reference electrodes.The level of performance obtained was surprisingly poor. Virtually all of the reference electrodes exhibited large transient errors (slow response) under certain conditions. More importantly, some electrodes exhibited large static (after-equilibrium) errors that would invalidate many measurements of practical importance. Finally some electrodes showed substantial shifts in potential when agitated in low ionic strength solutions. This “stirring error” made the correct pH value uncertain and made it difficult to titrate solutions to a desired pH. These various errors are explained and some prototype reference junctions with fast ideal response are presented in support of the theory.Although discussed in the context of pH measurement, these findings also apply to other electroanalytical uses of reference electrodes. Theoretical Definition of Correct Response The diffusion potential being a non-equilibrium phenomenon generally depends on inter-facial configuration. Guggenheiml distinguished three types of definite junction continuous mixture constrained diffusion and free diffusion. These junctions agreed to within 1 mV in the system he studied (0.1-3.5 M potassium chloride solution - 0.1 M hydrochloric acid) but a stable reproducible potential was not obtained unless the interfaces had cylindrical symmetry (all chemical and electrical gradients parallel to a straight line). The diffusion potential for a continuous mixture interface can be computed from the Hender-son equation,ll which can be expressed as where xi is the signed valence of the ith ionic species pi is its relative mobility (velocity/force) and C and C:’ are its molar concentrations in solutions I and 11.Table I gives Henderson TABLE I CALCULATED LIQUID JUNCTION POTENTIALS ( Uinside - Uoutside/mV) FOR INTERFACES BETWEEN DISSIMILAR ELECTROLYTES* Inside 4M l M 1M 10-4 M 4111 1M Outside KC1 KC1 HC1 HC1 equitransferentt equitransferentt 4 M KCl . . 0 -0.7 - 14 -4.6 -0.5 -0.9 1 M KCl . . 0.7 0 - 27 - 4.0 - 0.2 - 0.5 1 M HCl . . 14 27 0 152 14 27 M HC1 . . 4.6 4.0 - 152 0 0.0 0.0 l o 4 M KCl . . 6.1 4.5 - 169 - 27 0.0 0.0 M NaOH 4.9 4.2 - 160 - 33 0.0 0.0 1 M NaCl . - 1.3 - 4.4 - 31 - 43 - 2.2 - 4.9 1 ~ N a 0 H - 8.8 - 19 -33 -132 -9.6 - 20 10-7 M H+ OH- .. 7.8 7.2 -262 -110 0.0 0.0 * From equation (l) using Table I1 mobility values. t With ionic conductivity comparable to KCl p+ = p- = 75. 58 mV = 1 pH unit at 20 “C. No corrections have been made for activity or incomplete dissociation April 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 427 values for interfaces between certain key electrolytes. A study of Table I and equation (1) reveals that a junction potential arises when the net charge diffusion tendency ZzipiCi, changes across the interface. The junction potential between a concentrated and a very dilute electrolyte is determined primarily by the transference characteristics of the concentrated electrolyte. The potential is large if the concentrated electrolyte is very heterotransferent (e.g.hydrochloric acid) but small if it is fairly equitransferent (e.g. potassium chloride solu-tion). In the latter instance the magnitude of the residual junction potential depends logarithmically on the conductance (Cz:piCi) of the dilute electrolyte but not on its trans-ference characteristics. This gives a simple explanation for Picknett's observation12 that the residual junction potentials for different dilute electrolytes are nearly identical at similar levels of specific conductance. Thus the relative merits of different junction electrolytes are largely indicated by their concentration (self-diff usion) potentials [Table 11 (B)]. Although 4 M potassium chloride solution approaches the ideal of a concentrated equitransferent electrolyte and yields small junction potentials against standard buffers appreciable junction potentials may occur between it and concentrated heterotransferent samples (Table I) ; hence such samples cannot be meaningfully measured unless the junction potential is standardised and reproducible.The steady-state response of a capillary junction electrode with 4 M potassium chloride solution is used herein as the defining standard. The ideal practical reference would be one that quickly and reproducibly establishes the same potential relative to the capillary junction electrode in all solutions. A fixed offset is not regarded as error as it can be attributed to a difference in half-cell potential between the tested and stand-ard electrode and would normally be nulled by calibration of the ion meter.i i Avariable offset is indicative of electrode error. TABLE I1 DIFFUSIONAL CHARACTERISTICS OF IONS AND SALTS (A) Limiting ionic conductances13/ (B) Self-diffusion (concentration) potentials/mV* mho cm2 equiv-l (25 "C) H+ . . . . 349.8 K+ . . . . 73.5 NH,+ . . . . 73.5 Na+ . . . . 50.1 Li+ . . . . 38.7 NO,- . . 71.4 c1- . . . . 76.3 SO,2- . . . . 80.0 OH- . . 198.6 KC1 . . 1.1 KNO - 0.9 NH,NO - 0.9 HC1 . . . . -38.0 NaOH . . 35.0 KOH . . 27.2 Li,SO . . 1.4 LiCl . . . . 19.4 NaCl . . . . 12.3 * Table 11 (B) values are shifts in potential in passing from very dilute solution to one which is 10-fold more concentrated. Calculated from equation ( l ) using relative mobilities (p) determined by dividing Table 11 (A) conductances by unsigned valences ( I z I ) .Analysis of Errors Caused by Junction Charge Let (T represent the concentration of fixed ionic charge which for simplicity is assumed to be uniformly distributed within the junction void volume. This charge must be balanced by mobile counter ions of opposite sign. Thus C- = C+ + CT where C- and C+ are the concentra-tions of negative and positive mobile ions. The flux f of a given ionic species is the sum of diffusional electrophoretic and bulk-flow contributions so f = -DdC/dx + CpexE - Cv, where p is the ionic mobility D = pkT is the diffusion coefficient e is the unit charge z is the ionic valence E = -dU/dx is the electric field x is the distance into the junction and v is the flow velocity out of the junction.Local electroneutrality requires the anion and cation fluxes to be essentially equal. Assuming a 1 1 equitransferent electrolyte setting ff = f- and solving for E one obtains E = - F . z ( G + D . ~ ) RT 1 da v , 428 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vole 108 where C = (C+ + C-)/2 is the mean ionic concentration. Substituting equation (2) for E in the flux expressionf = (f+ + f-)/2 one obtains Finally a steady-state profile (aC/at = 0) implies that the flux is everywhere the same (af/ax = 0) but deep within the junction where C = 4 M>U the flux is due solely to outward flow transport s o f = -CJv where C is the concentration of the undiluted junction electrolyte. If da/dx = 0 and v = 0 equation (2) yields E = 0; thus the potential is constant along a uniformly charged unflowing junction in spite of concentration gradients and diffusional transport of an equitransferent electrolyte.Solving equation (3) under these conditions yields a linear concentration gradient. At the surface of the junction however there is an abrupt change in the ratio of free anions and cations so a Donnan-type potential is required to balance charge transport across the interface. Let Co and C denote the mean ionic concentrations just inside and outside the junction surface. Equating the diffusional and electrophoretic fluxes for each species which become large and must cancel throughout the abrupt interface solving for -E and integrat-ing one obtains the boundary potential AU = U, - U = (RT/F)ln(C;/C,) = (RT/F)ln(C,/Ct).Solving the simultaneous equations C$/C = C,/C; and C$ + a = C;, one obtains C$ = 4 C i + (0/2)~ - 4 2 and C; = 2/Ci + (012)~ + 4 2 . Thus the change in potential upon entering the junction is w(RT/F)(o/2Cl) for C > la1 ; -(u/ I aI)(RT/F)ln( 101 /Cl) for C >>. U. This shift in potential has the same polarity as the space charge. Equation (4) is a single-boundary equivalent of the Teorell - Meyer - Sievers membrane potential.14s15 If the concentration C at the junction outer surface is comparable to or smaller than a the shift in potential is large At the inner surface where C = 4 M > O there is no shift in potential. With flow through the junction streaming potentials are predicted. Equation (2) indicates that streaming fields will be inversely proportional to electrolyte concentration.Although such fields are likely to be negligible in the 4 M potassium chloride filled-portion of the junction, they might become appreciable near the end of the junction where C is reduced by diffusional exchange with the solution. Exact analysis of this situation would require solving non-linear equation (3) followed by numerical integration of E to yield the flow potential. However, dimensional analysis indicates that the steady-state flow potential due to diffusional exchange is independent of the flow velocity if junction charge is uniform. Equation (4) with f = -CJv and da2/dx = 0 is invariant under the parameter change v' = pv where j5 is the factor by which flow velocity is changed so the effect of increasing the outward flow velocity is to sharpen the concentration and field profiles and increase the field strength all by the same factor.Thus the potential at corresponding positions is unchanged because the higher field is exactly compensated by shorter distance. If the flow velocity is high enough to confine diffusional exchange to the outer portion of the junction the steady-state end-streaming potential is essentially independent of flow velocity. For example if the flow is increased the end-streaming potential should increase abruptly then decay back to its original level as a shorter steady-state profile is established. A rough approximation for the flow potential in a uniformly charged junction can be derived by assuming that fixed charge does not significantly alter the steady-state exponential con-centration profile obtained in a flowing uncharged junction (Fig.1). In this instance, C = CJ + (C1-CJ)e-Dz'D and integration of equation (2) yields This unbalanced shift in potential would yield a static error. I U(X) = - JEdx = (RT/F)(xflU/2DCj + (o/2C,)h([CJ + (C1 - Cj)e-Dz'D]/Cl>) 0 At large x the second logarithmic term in the brackets approaches a constant value while the first term corresponding to the streaming potential of the 4 M potassium chloride-fille April 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 429 portion continues to increase linearly with distance. The first term can simply be ignored if the streaming potential is negligible when C (k 4 M potassium chloride solution) fills the whole junction.Hence the end-streaming potential should change logarithmically with concentration of the exterior electrolyte with a Nernstian slope that is reduced by the factor 0/2CJ. As a/2CJ will typically be several orders of magni-tude below unity the contribution of the external solution to the streaming potential should be negligible even when C is comparable to (T and gives a significant static error. Then AU = ((T/2CJ) (RT/F)ln(C,/C,). 100% C0i”t I I I I Inside I (4 M KCI) I I I - Y (Flow) I solution I 0% coefi \ Distance from external surface Fig. 1. Exponential concentration profile in junction having outward flow. Total transport = flow + diffusion. dC/dt = vaC/ax + Da2C/ax2 = 0 a t steady state. Hence Cext(x) = Coexte-vz‘D for outside species and Ci*t(x) = Coint (1 -e-vpr’D) for inside species.Experimental Kinetic Test Reference electrodes were transferred between beakers containing about 150 ml of 1 M potassium chloride solution 1 M hydrochloric acid and 10-4 M hydrochloric acid. Potentials were measured versus sealed calomel electrodes left permanently in solution. These “station-ary” references provided stable potentials against which transient responses could be measured. The junctions of the stationary references were covered with perforated silicone-rubber caps plugged with glass-wool so that their potentials would not be affected by stirring. This allowed the tested electrode to be agitated selectively to check for stirring potentials. The transfer procedure was started by first equilibrating the electrode for at least 5 min in the 1 M potassium chloride solution.The electrode was then quickly rinsed in several aliquots of distilled water transferred to the next test solution and vigorously agitated for 5 s. After 4 mins the electrode was again vigorously agitated to indicate the shift due to stirring. The potential was recorded for an additional 1 min without stirring. Finally the electrode was rinsed and transferred into the next solution. Data were taken for all possible transfers between solutions. Particular care was taken in rinsing the electrode prior to its transfer into M hydrochloric acid which was extremely susceptible to carry-over from the 1 M solutions. Static Test Various commercial electrodes were initially evaluated by transferring them together for 15 min in 1 M potassium chloride solution 15 min in 1 M hydrochloric acid and then 30 min in 10-3~ hydrochloric acid.Differences in potential relative to a 1 mm bore glass capillary junction reference were noted at the end of each immersion period when fairly steady values had been attained. The capillary electrode (Fig. 2) was employed by flushing the capillary junction with fresh 4 M potassium chloride solution from the “purge” syringe rinsing the electrode with water and blotting dry immersing it into the solution and slowly pulling the “draw” syringe to move the solution interface into the capillary bore. Static errors were later determined from the dynamic test by noting the differences between the final potentials attained in each solution.However the stationary references slowl 430 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol. 108 drifted relative to each other over several days of immersion. To compensate for this drift, the solution potentials were measured periodically by use of the capillary junction reference, and the observed shifts in potential were subtracted from the experimental data. Finally it was found that a porous ceramic junction used with sufficient outward flux of pure 4 M potassium chloride solution could accurately replace the capillary junction. Unless otherwise specified experiments with ceramic junctions employed a silica-based sintered ceramic of 15% porosity 1 mm diameter and about 5 mm length clad in glass and ground flat at the exterior surface.Cell potentials were measured at 20 "C using a Corning Model 130 pH meter connected to a strip-chart recorder. Corning pH standard solutions were used 0.05 M potassium biphthalate pH 4.00; 0.05 M potassium dihydrogen orthophosphate - sodium hydroxide pH 7.00; and 0.05 M potassium carbonate - potassium tetraborate - potassium hydroxide pH 10.00, " Pu rg e " syringe // Capped stationary reference requiring calibration I Solution I Fig. 2. Capillary junction reference electrode used for deter-mining offsets between stationary references. A calomel internal electrode was used in some experiments. Sulphate-free (uncharged) agarose HSIF grade was obtained from Litex Corp. Celgard 3501 microporous polypropylene film was obtained from Celanese Corporation Greer SC USA, and Goretex fabric from W.L. Gore & Associates Inc. Elkton MD USA. Results and Discussion Because the liquid-junction potential should be determined primarily by the sample's ionic conductance and transference in theory only four types of solutions are needed to test ade-quately the performance of a reference electrode. These are concentrated equitransferent (e.g. 1 M potassium chloride solution) ; concentrated heterotransferent (e.g. 1 M hydrochloric acid) ; dilute equitransferent (e.g. M potassium chloride solution) ; and dilute hetero-transferent (e.g. lo4 M hydrochloric acid). Only one of the dilute solutions need be used as its composition should have little bearing on kinetic effects during transfers to and from the strong solutions. Dilute hydrochloric acid was selected as preferable to dilute potassium chloride solution for detecting steady-state errors with potassium chloride-filled electrodes ; a ca.4 M potassium chloride - 10-4 M potassium chloride interface being a single-species gradient would be insensitive to improper geometry. Fig. 3 shows a typical response to the kinetic test. A very large and persistent transient error occurred upon transfer from 1 M to The initial error was roughly -2 pH units and equilibrium was not reached even after 5 min. The other transfers M hydrochloric acid A@&? 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 431 resulted in smaller but significant transients. In the last transfer from 1 M potassium chloride solution to lo-* M hydrochloric acid the electrode “remembered” its previous exposure to 1 M hydrochloric acid and gave a fairly stable reading that was in error by about -0.6 pH unit.This type of response was characteristic of most commercial reference electrodes particularly those without junction flow (gel electrodes). For the sake of simple comparison the difference between millivolt readings at 20 s and at 5 min was taken as an Table I11 gives kinetic data for a variety of commercial electrodes. 1 M 110-41 1 M KCI 1 M HCI 10-4M HCI 1 M HCI I 1 M KCI I 10-4M HCI v j o 0 al c .- 4- - -50 > E \ - (II .- c. f -100 4-II al U 5 -150 al u1 - 1 -2 1 I 1 1 0 5 10 15 20 25 30 35 40 Time/min Typical response of no-flow reference electrode with 4 M potassium chloride junction electrolyte (5 mm long ceramic junction silver chloride saturated 4 M potassium chloride electrolyte gelled with 2% agarose stored in 4 M potassium chloride solution).Fig. 3. The arrows indicate times of brief agitation. index of “transient error,” and this difference was converted to equivalent pH measurement errors by dividing by +58 mV pH-1. Transient errors going from hydrochloric acid to other solutions were negative in polarity and were moderately large upon transfer to 1 M potassium TABLE I11 REFERENCE ELECTRODE KINETIC RESPONSE (pH, - pH 300 s) ‘l‘he tabulated values represent pH measurement errors at 20 s. combination electrodes. All references except A D2 and G were in Sealed no-flow electrodes are denoted Half-cells were Ag - AgCl except as noted. “gel” ; the others were re-fillable flow-type.A1 A2 A3 B C D1 D2 E F G H Electrode . . . . . . . . From 1 M KCI From 1 M HCI Type Ceramic Asbestos fibre calomel (blown clear) Ceramic gel Ceramic gel (at ground input) Polymer cloth gel Ceramic annulus Sintered PTFE body gel Annular glass channel Ceramic : No agitation 5 s vigorous agitation 20 s vigorous agitation Plastic sleeve (D. J.) : 10% KNOs filled 4 M KCI filled 0.01 M KCI filled 0.01 M KC1 filled, agitated Pt fibre thallium amalgam ? O I M I M -10-4; KCI HCI HC1 0.00 0.11 0.00 0.17 0.00 0.01 0.00 0.00 0.04 -0.06 0.00 0.13 -0.29 0.00 -0.10 0.33 0.00 0.11 0.01 0.01 0.12 -0.24 0.00 0.00 -0.01 0.00 0.22 0.07 0.00 0.24 0.27 0.04 0.00 0.04 0.14 0.00 0.07 0.00 0.00 -1.2 2.4 0.00 -1.3 2.4 0.00 0.00 0.01 To 1 M KCI - 0.11 - 0.16 - 0.01 - 0.07 - 0.10 0.19 - 0.11 -0.24 0.34 -0.46 - 0.14 - 0.08 0.89 0.83 - 0.06 I M HC1 0.03 0.00 0.05 0.02 - 0.03 0.02 0.02 0.02 0.04 - 0.01 -0.06 0.00 - 0.19 - 0.05 0.05 -10-4 M HCI -0.81 - 0.4 -1.3 -1.6 1.6 -0.74 - 2.0 -0.77 - 1.8 - 2.0 - 2.0 -1.9 -1.6 - 1.5 - 4.4 - 3.8 - 1.9 From lo-‘ M HCl r - p T o ~ M I M -10-4~ KCI HCI HC1 -0.01 0.14 0.10 0.00 0.01 0.04 -0.06 - - 0.01 0.03 0.00 - 0.04 - 0.08 0.01 0.00 0.28 -0.03 0.31 0.04 -0.24 0.08 0.15 -0.03 0.45 0.03 0.15 0.00 0.38 0.09 0.60 0.57 0.00 -0.05 0.13 0.02 -0.00 0.10 0.00 -0.09 0.09 -0.40 -0.34 0.21 -0.33 0.00 0.06 0.0 432 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol.108 chloride solution and extremely large upon transfer to M hydrochloric acid. Transfers to 1 M hydrochloric acid generally had positive transient errors. The larger transient error was associated with prior soaking in the weaker electrolyte. Rinsing with water and re-immersion in the same solutions gave fairly small transients. Data for electrode F showed that vigorous agitation had little effect on the transient. Thus the transient is not due to carry-over on the surface of the electrode nor does it depend much on mass transport within the external solution. The processes associated with the transient must take place largely within the electrode itself.Table I11 data for the double-junction electrode (G) indicate little difference in kinetic response between 4 M potassium chloride and 10% potassium nitrate filling solutions. However the use of 0.01 M potassium chloride solution as the junction electrolyte resulted in a very large increase in several of the transients; concentrated filling solutions gave a much faster response. Again stirring had negligible effect on response time. The polarity of transients almost always corresponded to lags rather than overshoots in pH response. The few exceptions included transfers of electrodes B and D1 from 1 M potassium chloride solution to lov4 M hydrochloric acid. These would give overshoot if the 1 M potassium chloride solution had a pH above 4. Static error is the error in potential that persists after the electrode has essentially stopped drifting.Columns 1 4 in Table IV show millivolt data for the static test (simultaneous Electrode A1 A3 . . B c D1 ,. D2 E F TABLE IV REFERENCE ELECTRODE STATIC RESPONSE Electrode potential/mV* 1 M KC1 (15 min) - 44.9 - 28.1 -37.5 - 53.6 - 39.0 -44.5 - 46.0 -49.9 1 M HC1 (15 min) -42.5 - 13.3 - 24.6 - 44.4 - 23.5 - 20.5 -41.2 - 50.2 - M HCl (15 min) -49.2 - 25.0 -7.0 - 75.3 - 14.7 -63.1 -41.9 - 54.9 - M HC1 (30 min)? -45.5 - 22.5 - 13.6 -65.1 - 17.9 - 55.5 -42.2 - 55.2 pH measurement discrepancy r m From 1 M KC1 1 M KCl 1 M HC1 To 1 M HC1 10-3 M HClt M HClt 0.04 -0.01 - 0.05 0.25 0.10 -0.16 0.22 0.41 0.19 0.16 -0.20 - 0.36 0.27 0.36 0.10 0.41 -0.19 -0.60 0.08 0.07 - 0.02 - 0.01 -0.09 - 0.09 * mV relative to calomel capillary reference.t 30-min value; re-agitated vigorously a t 15 min. measurement with all electrodes together in the same beaker). In columns 5-7 the shifts in potential relative to the capillary reference upon changing solutions have been converted to equivalent pH units. Suppose parallel measurements of pH are made using identical pH electrodes but one measurement uses the capillary junction reference and the other uses the indicated reference. After meter calibration to achieve agreement in the solution marked “From,” the electrodes are transferred to the solution marked “To.” The tabulated values are the measurement discrepancies that would be observed in the second solution.These often amount to several tenths of a pH unit. As calibration buffers more closely resemble potassium chloride in transference and as electrode non-idealities are probably minimised in strong potassium chloride electrolytes the discrep-ancies measured relative to potassium chloride (listed in the first two columns) probably reflect the errors that would be experienced in actual use. It was noted that the shifts in potential in column 2 were opposite for two different types of references made by the same manufacturer (D1 and D2) so the results obtained with these electrodes in 0.001 M hydrochloric acid should disagree by over 0.5 pH unit. To confirm a discrepancy pH measurements were performed to compare results obtained with the two references using the same pH electrode.In 1 M potassium chloride containing a trace amount of hydrochloric acid the measurements agreed closely as would be expected if junction anomalies are suppressed by the 1 M potassium chloride solution. In 1 M hydrochloric acid the readings differed by 0.2 pH unit and in M hydrochloric acid the stable readings differed by a full pH unit. The significance of these shifts is the following. These data are shown in Table V April 1983 C I Q 7.00 -0 + 6.99 -6-98 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES TABLE V Reference and pH electrode both exposed Reference 1 in baffle, pH sensor 15-cm reference exposed electrolyte head 1 -cm electrolyte head 1 - ---- - c - I I I I Stir Stir Stir Stir w w Vigorous stirring I I 0 5 10 15 20 433 DEPENDENCE OF MEASURED pH ON REFERENCE ELECTRODE Reference Solution electrode* mV PH 1 M KC1 trace of HC1 D1 210.8 3.49 1 M KC1 trace of HC1 D2 214.2 3.43 1 M HC1 D1 395.4 0.41 1 M HC1 D2 384.2 0.60 M HC1 D2 225.2 3.24 10-4 M HC1 D1 162.8 4.28 * pH electrode pH section of combination electrode D1.Refer-Reference D2 AgCl ence D1 AgCl ceramic-annulus junction. porous polymer body non-refillable. “Stirring potentials” are generated by the reference electrode and normally are a problem only at low ionic strength. In Fig. 4 for example the pH indicated by a glass pH electrode and a ceramic-junction reference electrode in pH 7 standard buffer (0.05 M phosphate) was shifted only slightly (ca.0.013 pH unit) by stirring. This shift originated at the reference electrode as it was completely eliminated by placing the reference electrode within a protective baffle (a plastic syringe body immersed in the solution). Fig. 5(a) shows the much larger stirring effect that occurred when the same electrodes were used to adjust de-ionised water to ca. pH 10 with sodium hydroxide solution. When the solution was stirred as required to distribute titrant the measurement shifted by about -0.2 pH unit. When stirring was stopped the indicated pH slowly drifted back to its unstirred value. This stirring effect made it difficult to adjust the solution to the desired final pH reading and moreover it was unclear which pH value was correct stirred or unstirred.Again shifting the reference to the unstirred baffle restored the quiescent pH value whereas placing the pH electrode within the baffle had no effect. The slowness of recovery when stirring was stopped was due primarily to continued movement of the solution. Recovery from longitudinal agitation of the electrode which did not disturb the bulk solution was much faster [Fig. 5 ( b ) ] and suggested that the intrinsic decay time for the stirring potential was in the order of a few seconds BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol. 108 5 U ; 9.8 .- 0 - 9.7 9.6 434 10.0 I TI Q 9.9 c, . 9.8 C -9.7 Agitate Agitate - -- I 1 I I --a)pH and reference Reference in electrodes both baffle; pH Both exposed I exposed I exposed -,Hard Ti me/m in Stir A 1 15 I 0 I I I 0 20 40 60 Time/s Fig.5. Large stirring potentials observed at low ionic strength. Same electrodes as in Fig. 4. (a) Solution stirred by magnetic bar; (b) longitudinal agitation of reference electrode. Explanation of the Various Errors The observed transient errors can be explained largely by diffusional entrapment of the previously measured solution within the physical junction. During electrode immersion, 4 M potassium chloride within the outer region of the junction is diffusionally exchanged for external solution species. When the reference is rinsed and transferred to the next solution, this exchanged layer intervenes between the 4 M potassium chloride junction electrolyte and the new solution. This layer disturbs the junction potential because diffusion potentials are not a transitive property.That is given solutions A B and C the diffusion potential at the interface A/C is not the sum of the potentials at the interfaces A/B and B/C. This is seen clearly from Table VI which compares the diffusion potentials for “sandwiched” and “direct” interfaces. The calculated differences in diffusion potential are in general agreement with the magnitudes of the observed transients. to 1 M hydrochloric acid where the observed transient is much smaller than predicted by the model. This might be expected as the residual concentration of hydrochloric acid and potassium chloride at the junction surface at time of transfer is undoubtedly much higher than 10-4 M, and would substantially reduce the diffusion potential against 1 M hydrochloric acid.The sole exception is the transfer from TABLE VI CALCULATED AND OBSERVED TRANSIENTS FOR “SANDWICH” INTERFACES Type of transfer A I 7 Type of interface 1 M KCI -+ 1 M HCl/mV 1 M HCI -D 10-4 M HCl/mV M HCI -+ 1 M HCl/mV 1 M HCI -P 1 M KCl/mV (A) Sandwich Inner ~ M K C I / ~ M K C I -1 ~ M K C I / ~ M H C I 14 ~MKCI/~O-~MHCI 5 ~ M K C I / ~ M H C I 14 Outer . . . . 1 M KCI/1 M HCl 27 1 M HCl/lO-4 M HCl - 152 M HC1/1 M HC1152 1 M HC1/1 M KCI - 27 Sum 28 - 138 157 -13 (B) Direct . . . . 4 M KCI/1 M HCI 14 4 M KCl/10-4 MHCI 5 4 M KCI/1 M HCI 14 4 M KC1/1 M HCI I (C) Difference (transient expected fromA -B) . . (Fig. 2) . . . . (D) Observed transient 14 -- 20 143 143 -130 30 -14 18 The potential at interface A/B is generally negligible compared with that at B/C if A is 4 M potassium chloride solution so in effect the layer of previously measured solution temporarily replaces 4 M potassium chloride solution as the junction electrolyte.Error kinetics are deter-mined by the time required to disperse this layer and obtain a direct interface. Under no-flow conditions the kinetics are determined by the duration of prior immersion with the thicknes April 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 435 of the junction being the limiting factor at the steady state. A faster response can be obtained by using a thinner junction. The diffusional relaxation time for a junction of thickness X is approximately T = X2/?r2D where D is the diffusion coefficient.Taking D = 2 x cm2 s-l for a typical electrolyte the relaxation time for a 0.5-cm junction would be roughly 30 min. This could be reduced to 20 s by reducing the effective junction thickness to 0.06 cm. This presumes that the junction permeability is low enough to keep the potassium chloride - sample interdiffusion profile from extending substantially beyond the physical junction. Thin-membrane junctions were made from Goretex fabric (microporous PTFE on a nylon backing, PTFE side out) and Celgard 3501 (microporous polypropylene 25 pm x 0.25 mm diameter, 1.8 kQ). As expected these no-flow junctions yielded very fast responses; their transient errors were respectively only -0.05 and -0.02 pH unit at 20 s in the difficult transfer from 1 to However these membranes proved impractical for general use owing to problems with inconsistent wetting, excessive porosity organic fouling etc.Outward flow establishes a steady-state profile in which the concentration of external ionic species decreases exponentially from the junction surface while the junction electrolyte (potassium chloride solution) shows a comparable increase (Fig. 1). The relaxation distance for such profiles is D/v where v is the flow velocity and D is the diffusion coefficient. When the junc-tion is transferred into a new solution the concentration of old solution species a t the junction surface should decay exponentially with a relaxation time of D/v2. To achieve a decay constant of 20 s the required velocity is v = (D/i)0-5m 10-3 cm s-l a very modest value.With a typical junction ceramic (1 mm diameter 15% porosity) this corresponds to a flow-rate of only 4 p1 h-1. However the junctions of conventional silver - silver chloride electrodes rapidly become clogged with precipitated silver chloride so the response becomes slow owing to lack of adequate flow.lG High flux with low total flow can be obtained by restricting the flow to a small aperture. Fig. 6 shows an “annular-pore” junction formed by inserting a silica fibre through a hole in 125-pm poly(viny1idine fluoride) film. Even small flow-rates (ca. 2 pl h-l) yielded very fast responses. The annular shape of the pore helped to prevent clogging by particulates. How-ever practical problems with the junction included a requirement for a positive flow and for a separate restrictor to control the flow-rate.The geometry of the interface between the two liquids is not well defined at the pore and depends on flow-rate; unstable results were obtained if a positive flow was not maintained. M hydrochloric acid and below 0.01 pH unit in all other transfers. Outward flow of junction electrolyte is another way of attaining a fast response. f - Quartz fibre PVF Film 125’ pm _F_ Outside solution Fig. 6. Annular-pore reference junction. With an outward flow-rate of 1.6 p1 h-l the transient error at 20 s was below 0.01 pH in all solution transfers. Junction resistance in 4 M potassium chloride solution was 2.6 kn. For the sake of discussion the observed “static” errors can be separated into two classes: very slow transient errors with discrepancies at 15 or 30 min which would eventually dis-appear with prolonged soaking and true static errors which would persist no matter how long the soaking.Some of the observed static errors probably belong to the first class (particu-larly those in 1 M hydrochloric acid all of which have the same polarity as the transient errors). The diffusional relaxation time for a 1-cm junction is roughly 1.4 h. If such an electrode is routinely stored in pH 7 buffer standard (0.05 M phosphate) between measurements as is usual practice then pH 7 buffer not 4 M potassium chloride solution effectively serves as the Some of the tested references (C D1) had junctions longer than 1 cm 436 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol. 108 junction electrolyte.Response problems would not necessarily be evident during calibration in various NBS-type standard buffers as these are fairly similar in transference and ionic strength. Upon insertion into the solution to be measured a fairly stable diffusion potential corresponding to the phosphate - sample interface would rapidly be established at the outer surface of the junction ; however with very dilute or non-equitransferent samples this potential may differ considerably from the correct value obtained against 4 M potassium chloride solution. To determine the possible magnitude of such errors a junction using pH 7 buffer was evalu-ated. An electrode body with ceramic junction was filled with gelled pH 7 buffer (0.05 M phosphate 2% agarose) and inner contact was made via a miniature 4 M potassium chloride solution bridge.The observed static error was 0 mV in 1 M potassium chloride solution +48 mV in 1 M hydrochloric acid and -14 mV in The observed value in 1 M hydrochloric acid corresponded well to the calculated value (54 mV) but the static error in M hydrochloric acid was smaller than expected (-44 mV). In distilled water an initial static error of about -65 mV was observed, Results for this electrode are shown in Fig. 7. M hydrochloric acid. J pH7 I M 1 M 1 0 - 4 ~ 1 M I M 1 0 - 4 ~ buffer KCI HCI HCI HCI KCI HCI C .- U C .-L' I Solution composition I Distillea water pH7 p H 4 pH7 pH 10 p H 7 > E buffer buffer buffer buffer buffer 1 (p==106 ~2 cm) m 0, .-0) 0 -1.3 mV Q) W U U g t U g o --65 mV -50 -- 0 L E z Ip I From pH 7 -100 I x f f e r -- t 9 H -501 d c 0 1 d - :L -100 2 0 5 From pH 7 buffer 1 M HCI Y-l 10 15 20 25 Fig.7. Anomalous response of no-flow junction saturated with pH 7 standard buffer. Ceramic double-junction reference filled with 0.05 M phosphate buffer gelled with 2% agarose. The arrows indicate times of brief agitation A@ril 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 437 which held steady for a while before drifting to more negative values. This initial error is probably caused by the unbalanced transference of phosphate buffer. Although the electrode response was very slow in certain transfers it was surprisingly fast in the transfers from to 1 M hydrochloric acid and then to 1 M potassium chloride solution; however the potentials obtained were erroneous.Finally the electrode was tested for performance in pH 4 7 and 10 buffer standards. Static errors relative to pH 7 buffer were below 8 mV and there was no observable kinetic or stirring error. During two-point calibration the electrode would appear to perform perfectly giving no indication that there would be problems in other solutions. Even with three-point calibration slope and offset adjustment would reduce the discrepancies at pH 4 7 and 10 to only *0.02 pH unit, which is likely to be ignored. Errors of the above type may be expected in routine use of no-flow gel electrodes and clogged flow-type electrodes with long junctions. The second class of static errors are those persisting at the steady state.Such errors are evident in Table IV; the long-term offsets for electrodes B and D1 in M hydrochloric acid (relative to 1 M potassium chloride solution) were substantially positive even though the short-term transients were negative in polarity. A likely explanation is the presence of fixed positive charge within the junction pores as demonstrated below. Large shifts in potential were observed with cessation of flow in the “annular-pore” junction of Fig. 6. This indicates non-standard interfacial geometry as another possible source of static error. Junction geometry may be more critical in solutions of low ionic strength where inter-diffusion can drastically change the concentration of the sample while having negligible effect on the junction electrolyte.Thus when 4 M potassium chloride solution flows out of an iso-lated pore into a dilute electrolyte a broad continuous-mixture type of interface results but when the dilute electrolyte flows inward into the potassium chloride solution interdiffusion causes the concentration interface to collapse to the immediate vicinity of the pore resulting in an interface with constrained and likely spherical diffusion. According to Guggenheim’s criteria,l the former interface should be very stable and the latter unstable. This may explain the empirical observation7 that outward flow is required for stability of leak-type junctions in dilute electrolyte. Similarly in this study the 4 M potassium chloride open-capillary junction was immediately stable to better than 1 mV after insertion into 1 M potassium chloride solution and 1 M hydrochloric acid but tended to drift slowly by a few millivolts after insertion into In contrast flat-surfaced ceramic junctions with adequate positive flow stabilised quickly in M hydrochloric acid probably because of greater control over the geometry and mixing pattern at the liquid interface.Other possible sources of static error include the introduction of extrinsic charge or ion specificity by clogging of junctions with solution precipitates (e.g. Ag,S and Hg,S in sulphide-containing solutions) and also the redox sensitivity of electronically conductive junction materials (e.g. platinum) when clogged ; however these possibilities were not investigated in detail. The origin of stirring errors is less evident.A seemingly attractive hypothesis is that the stirring potential is an electrokinetic effect analogous to a streaming potential and arises from physical displacement of a layer of charged solution. However such an explanation is untenable for several reasons. Firstly it is hard to envisage how fluid motion parallel to the junction surface could induce a substantial electrokinetic potential perpendicular to the surface. Also when the solution outside and within the junction is the same (e.g. loA4 M hydrochloric acid throughout) the stirring potential disappears (see below). This would not be expected if the stirring potential was due to a surface electrokinetic effect. Further it seems particularly unlikely that the potential is due to sweeping of counter ions away from a charged exterior surface as the counter ionic cloud is centred roughly 0.03 pm from the surface at M ionic strength (and decreases as I - l I 2 ) while the thickness of the “stagnant” Nernst diffusion layer exceeds 10 pm for even vigorous stirring17 Finally the diffusional relaxation time correspond-ing to a 0.03 pm cloud thickness is below 1 ps and voltages induced by current pulses applied through reference electrodes gave observed decay times below 200 ps suggesting that electro-kinetic effects should dissipate several orders of magnitude faster than the relaxation time observed for the stirring effects about 3-6 s in Fig.5 ( b ) . In contrast a relaxation time of 3-6 s corresponds to diffusion into a layer of roughly 120-170 pm which falls in the thickness range of Nernst diffusion layers in moderately stirred to unstirred solution (50-500 pm).Therefore the stirring potential is more likely associated with disturbance of free-diffusion Thus a fast response is no guarantee of accuracy. M hydrochloric acid 438 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol. 108 profiles extending beyond the junction into the adjacent convection-free layer of solution. When the external solution is stirred constrained diffusion is imposed at the external junction surface whereas free diffusion may propagate beyond the external junction surface in unstirred solution (Fig. 8). Thus stirring could affect the junction potential by altering the interfacial geometry or by reducing the ionic strength at the junction surface which would increase the boundary potential due to junction charge [equation (4)].Both mechanisms are supported by experiment. Large changes in potential (3040 mV) were observed upon agitation of very slow-flowing pore electrodes where interfacial geometry was the only likely factor. Con-versely stirring potentials were also observed with 4 M potassium chloride solution-filled ceramic junctions agitated in M potassium chloride solution where the single-species interface precludes geometric effects. Uniform CI Inside Physical Outside solution I junction I solution Fig. 8. Effect of stirring on junction concentration profile. The boundary layer thicknesses and con-centrations have been greatly exaggerated for clarity. Stirring changes the outer profile from free to con-strained diffusion and decreases the concentration at the junction surface.With conventional reference electrodes the boundary-potential explanation of stirring error seems plausible because diffusion should yield significant concentrations of potassium chloride at the exterior surface of even fairly thick junctions. For example the surface potassium chloride concentration of a no-flow 1 cm long 10% porosity junction should be about 0.004 M in rapidly stirred solution (0.01-cm Nernst diffusion layer) and about 0.02 M in unstirred solu-tion (0.05-cm Nernst layer). Ironically the stirring error should disappear with very thick junctions leaving only a large static error With junctions of moderate thickness the static error should reach a limiting value in very dilute solutions and not be nearly as great as when the external solution is drawn into the junction by negative flow.Finally both static and stirring errors should be greatly suppressed with a positive flow of electrolyte through thick junctions. For example a 0.001 cm s-l positive flow would yield potassium chloride transport equivalent to a 0.2-cm (=D/lO%v) junction. The junction-charge theory of static and stirring error is confirmed in Fig. 9 which shows the changes in potential of a ceramic-junction reference electrode as solution stirring and junc-tion electrolyte flow are varied. The observed behaviour is as expected for a fixed negative charge in the junction. Confirmatory features include the following a large offset error but no stirring potential after dilute external electrolyte has been drawn deep into the junction (a i k 1); a minimum offset with stirring potential when potassium chloride solution is delivered to the junction surface by outward diffusion or flow of junction electrolyte (b-f m) ; an inverse relationship between stirring potential and outward flow-rate (b c n) ; the absence of a streaming potential with continuous outward flow which keeps the junction filled with 4 M potassium chloride solution (n); the presence of a streaming potential after prolonged inward flow which fills part of the junction with dilute electrolyte.(h j) ; and the absence of an “end-streaming” potential due to diffusional exchange (p).Also as predicted the polarity o April 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 439 the stirring error is the same as that of the static offset; and the polarity of the streaming potential which depends on direction of flow matches that of the static offset when flow is outward.According to the junction-charge theory clogging of junction pores by extraneous agents could aggravate static and stirring errors. Partially clogged pores have high surface to volume ratios which should increase the ionic-strength threshold for errors due to surface charge. Also the clogging agent might contribute surface charge of its own. For example Fig. 10 shows the reference response of a flow-type ceramic silver - silver chloride combination elec-trode which though having a clogged junction was considered functional by its users. Tests in pH 4 7 and 10 buffers indicated only a slow response and moderate relative offsets.How-ever when the electrode was transferred from pH 7 buffer to M hydrochloric acid the electrode showed very large positive static and stirring errors indicating the presence of posi-tive charge in the junction. On transfer to 1 M hydrochloric acid the potential decreased overall as the charge effects were suppressed whereas normally the initial shift is positive owing to transient diffusion potentials. M hydrochloric acid the potential increased to an even higher offset error (exceeding + 100 mV). Normally a negative diffusional transient would be observed with this transfer. The increased offset error in M hydrochloric acid after exposure to 1 M hydrochloric acid is probably due to H+ adsorption to materials within the junction.This H+ adsorption phenomenon is also evident to a lesser extent in earlier data with unclogged ceramic junctions. In Figs. 3 and 7, for example the stirring potentials are initially positive after transfer from 1 to M hydro-chloric acid but are negative after prolonged exposure to dilute solution. This change When the electrode was returned again to 0 head after drawingext. soh. deep into junction 0 > E -10 Fig. 9. 40cm zcm ocm -10 ocm -1ocm- head cm KCI Junction electrolyte (4 M KCI) pressure head I 10cmhead l c m ocm ocm Streaming potential develops as dilute M) electrol9e fills junction (h) Stirring potential is absent as solution is same within and outsidc junction (i) M KCI at junction outer surface causes large static error (9) 0 10 20 Ti me/mi n 30 40 Offset diminishes Stirring potential, stir error returns residual offset error as KCI concentration increases at surface are inversely related to positive flow-rate Flow-proportional streaming pote-"-' (plus fixed offsc., when M KCI fills stirrin! 7 potent cm Stir 10 cm Continuous stirring Brief intervals of +llOcm head -1 briefly stopped Qualitative confirmation of space-charge theory of static and stirring errors.Ceramic junction in M potassium chloride solution ueYsu.s similar capped no-flow junction 440 BREZINSKI KINETIC STATIC AND STIRRING Analyst Vol. 108 probably reflects Hf desorption from pore walls with net surface charge changing from positive to negative.Thus some of the kinetic behaviour of reference electrodes e.g. anomalously slow recovery from stirring is caused by slow adsorption or desorption of charge within the junction. Boundary potentials due to junction charge seem a likely explanation for the errors observed by Illingworth.6 A highly charged junction characterised by partial clogging or very hetero-geneous pore size would act as a Nernstian concentration sensor shunted by a liquid-junction leak yielding the observed sub-Nernstian logarithmic dependence on salt concentration. Accordingly he found that errors were diminished considerably when potassium chloride electrolyte was allowed to seep out and evaporate on the junction surface prior to electrode use. Tim e/m i n Solution composition 10-4 M 1 M 10-4 M 1 M \ HCI HCI HCI KCI PH J iuffer 0 5 10 15 20 25 Time/mi n Fig.10. Anomalous response of reference electrode with protein-clogged junction. The electrode was a silver - silver chloride combination with a ceramic junction used for 6 months, including storage in pH 7 buffer and measurement of protein solutions. The junction resistance was about 10 kR. The arrows indicate times of brief agitation. Conclusion Overall the performance of commercial reference electrodes was found to be surprisingly poor considering the very widespread application and presumed accuracy of pH measurement and the relative lack of attention to reference electrode performance characteristics and problems. The seemingly “passive” role of reference electrodes has probably resulted in many of their performance aspects being overlooked.Also reference problems tend to be suppressed in standard buffers where accuracy is usually checked. Therefore users may see no reason to disbelieve erroneous readings obtained in non-standard environments that have to be taken at face value. The observed errors apart from belying the significance typically displayed by pH meters, are large enough to be of practical consequence-they often correspond to many-fold differences in H+ activity. pH monotoring and control are key aspects in chemical processing and environ-mental protection and these fields frequently require measurement at extreme pH or low ionic strength. Problems may arise even under moderate conditions; some enzyme buffers prepared using the Fig.10 electrode were off by 0.5 pH unit April 1983 ERRORS OF LIQUID JUNCTION REFERENCE ELECTRODES 441 In view of these large errors the use of special equitransferent solutions for low ionic strength measurements seems of dubious benefit. An exactly equitransferent solution in place of 4 M potassium chloride solution would eliminate a roughly 6 mV shift in liquid junction potential between calibration buffers and pure water (lo-’ M). However much larger static errors were observed with some conventional junctions at ionic strengths as high as M. Also equitransferent filling solutions are often less concentrated than 4 M and a reduction in electrolyte concentration should result in a roughly proportionate increase in static and stirring error due to junction charge [equation (4)].As discussed above and by Picknett,l2 the residual junction potential for 4 M potassium chloride solution in very dilute solutions should be a predictable logarithmic function of solution conductance. Conductance measurement and compensation particularly if accomplished directly by the ion meter are a possible alternative that could yield greater accuracy than replacing 4 M potassium chloride solution with a more equitransferent but less concentrated electrolyte. Colloidal suspension effects were studied by Jenny et aZ.,18 who observed differences as large as 5 pH units between the pH measured in a sediment of electrically charged particles such as clays and ion-exchange resin beads and the pH measured in the supernatant fluid.This pH discrepancy was attributed to a large junction potential between the reference electrode and the sediment. However their conclusion appears dubious in the light of the present findings. Provided that the concentration of the potassium chloride junction electrolyte is high at the junction surface potassium chloride should diffuse a distance into the sediment suppressing the boundary potential at the sediment - junction interface in accord with equation (4). Further the gradient of junction potassium chloride diffusing from the surface to the interior of a uniformly packed sediment should generate no additional potential [equation (2) with do/dx ZI = 01. Thus the over-all junction potential should be relatively low. On the other hand equation (4) predicts a much larger shift in potential a t the boundary between the charged sediment and the dilute supernatant.Owing to this potential difference hydrogen ions should be partitioned unequally across the boundary yielding equal electrochemical potentials U + (RT/F)ln aH+ (sensed by the pH electrode) but unequal chemical potentials and true pHs. This view is supported in a separate study.lg The origin of actual junction potentials is considerably more complex than has generally been assumed and it is clearly inadequate to view the reference junction in practical terms as a mere “leak.” However much present-day technology and practice seem to be based on an inadequate understanding of the requirements for satisfactory junction performance. To prevent static and stirring errors the porous junction material should exhibit a low charge to volume ratio at low ionic strength even after exposure to pH extremes.Boundary potentials due to residual junction charge should be further reduced by an adequate concentration of potassium chloride delivered to the junction surface by diffusion or flow. Delivery of ample potassium chloride is also especially critical to accurate pH measurement in charged sediments and suspensions. Fast response requires either adequate outward flow or very thin junctions. Most commercial electrodes particularly those designed for process applications do not meet these criteria. Also the extent to which junction performance is rapidly degraded by silver chloride clogging has not been appreciated (vix. “flow-type” silver - silver chloride electrodes).In view of the low toxicity and excellent thermal stability of the silver - silver chloride electrode the clogging problem is worth preventing and electrode designs with pure 4 M potassium chloride electrolyte are discussed elsewhere.l6 Electrode performance claims are usually based on tests in standard buffers which are largely ineffective in differentiating be-tween reference electrodes. Tests similar to those described could provide the basis for developing improved electrodes and more meaningful performance specifications. Improvements in user technique and awareness are also suggested. Thick junctions without flow should not be stored in standard buffers or be used for critical or continuous measurement. It is a common practice to store electrodes in very dilute solutions to “condition” them for low ionic strength measurements.This approach is completely wrong as virtually all of the junction anomalies are increased by reduction of ionic strength within the junction. Instead, reference electrodes should be stored in 4 M potassium chloride solution which suppresses anomalies and minimises junction clogging. The prevalent recommendation for continuous stirring during ion-selective electrode measurements may need qualification because stirring generally (but not always)16 increases the reference error at low ionic strength. In blood-gas and other electroanalytical instrumentation the reference junction is sometime 442 BREZINSKI located in a recess or downstream and is connected to the measured sample by a segment of running buffer.This configuration could constitute a sandwiched (double) interface which can cause static errors and should be avoided. Likewise an intervening particle-free layer between a measured colloid and the concentrated junction electrolyte must be avoided. The above recommendations are necessarily general because the requirements for optimum performance will vary with the electrode type and application. Specific procedural recom-mendations must be determined by appropriate user - manufacturer testing. A practical fast-responding no-flow junction based on the thin-membrane approach seems feasible but is subject to stringent material requirements. In addition to strength chemical durability and immunity to fouling the membrane material must have low porosity while maintaining a low charge to volume ratio.Further fast response in such a junction seems incompatible with accuracy in highly charged colloids because a low potassium chloride con-centration at the exterior junction surface would give rise to a colloidal boundary potential, whereas response will be slow if the interdiffusion profile extends substantially beyond the physical junction. Fortunately nearly ideal general-purpose response seems attainable with conventional flow-type reference electrodes. It is hoped that this study will contribute to improved accuracy in electroanalytical methodologies. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. References Guggenheim E. A. J Am. Chem. SOC. 1930 52 1315. Guggenheim E. A. J . Phys. Chem. 1930 34 1758. Bates R. G. “Determination of pH,” Second Edition Wiley Toronto 1973 p. 85. Covington A. K. Anal. Chim. Acta 1981 127 1. Bates R. G. CRC Crit. Rev. Anal. Chem. 1981 10 247. Illingworth J. A. Biochem. J . 1981 195 259. Covington A. K. in Durst R. A. Editor “Ion Selective Electrodes,” NBS Special Publication 314, Planck M. Ann. Phys. 1890 39 161. Bass L. Trans. Faraday SOC. 1964 60 1656. Hafemann D. R. J . Phys. Chem. 1965 69 4226. MacInnes D. A. “Principles of Electrochemistry,” Reinhold New York 1939 p. 232. Picknett R. G. Trans. Faraday SOC. 1968 64 1059. Dean J. A. Editor “Lange’s Handbook of Chemistry,” Eleventh Edition McGraw-Hill New York, Teorell T. Proc. SOC. Exp. Biol. Med. 1935 33 282. Meyer K. H. and Sievers G. F. Helv. Chim. Acta 1936 19 649. Brezinski D. P. Anal. Chim. Acta 1982 134 247. Bockris J. O’M. and Reddy A. K. N. “Modern Electrochemistry,” Plenum Press New York 1970, Volume I pp. 199 and 729; Volume 11 p. 1058. Jenny H. Nielson T. R. Coleman N. T. and Williams D. E. Science 1950 112 164. Brezinski D. P. Talanta 1983 in the press. National Bureau of Standards Washington DC 1969 Chapter 4 (general review). 1973 p. 6-30. Received July 5th 1982 Accepted September 24th 198