1164
Anal. Chem. 1981, 53, 1164-1170
Silver/Silver Chloride Electrodes Coated with Cellulose Acetate for the Elimination of Bromide and Uric Acid Interferences James R. Sandlfer Research Laboratories, Eastman Kodak Company, Rochester, New York 14650
Overcoatlng a sllver/sllver chlorlde electrode with cellulose acetate reduces the sensitlvlty of the electrode to bromlde and urlc acid In serum analysis. Apparently, the overcoat decreases the rate at which bromide attacks the sllver chloride by controlling Its flux. The electrode Is lnsensltlve to bromide as long as It remains In the chlorlde form. Interference by uric acid Is reduced because the cellulose acetate controls the rate of reactlon between urlc acld and sllver chlorlde In direct contact. Thls may be due to the acidic nature of cellulose acetate In conjunctlon wlth Its proteln screening abillty. Unless the solutlons are distlnctty basic, the electrodes must be preconditioned for use below about 20 mM sodium chlorlde. Otherwlse, more than 7 rnln will be required before a stable potentlal can be measured. Response tlmes are typlcally less than 1 mln for precondltloned electrodes. The electrodes can be precondltloned by soaking them for 5 mln In 100 mM sodlum chlorlde at any t h e up to at least several days before use. Slow response to dllute solutions has been attributed to an Ion-exchange reactlon which must precede diffusion wlthln the cellulose acetate membrane.
During the developmentof a chloride ion selective electrode for clinical analysis, traces of bromide and uric acid were found to limit severely the usefulness of silver/silver chloride electrodes for the analysis of serum. Bromide attacks the silver chloride and converts it to a mixed phase of AgCl,Br, which is sensitive not only to chloride but also to bromide ( I ) . Uric acid apparently reacts with silver chloride to form silver urate/silver chloride mixed crystals in much the same way. Overcoating the silver/silver chloride electrode with cellulose acetate effectively eliminatesthe severities of interferences by these agents (2). The various mechanisms involved were characterized by using potentiometry, conductometric and potentiometric titrations, and radiotracer techniques; the results of those studies are reported here.
THEORY Equation 1 is often used to describe the potential response of an ion-selective electrode to the analyte in the presence of interfering species (3). E is the measured potential, Eo is the
potential of the cell at unit activity of the analyte in the absence of interferences, R is the gas constant, T i s the temperature, F is the Faraday constant, and is the selectivity coefficient of the analyte relative to the interferent. Activities and valences are represented by a and n, respectively, subscript i designates the analyte, and subscript j designates interfering ions. Serum chloride concentrations vary between 98 and 108 mM in normal serum, whereas uric acid occurs between 2.0 and 7.1 mg/100 mL (0.12-0.42 mM), and bromide could be as high as -0.2 mM. These values were taken from Tietz (4).
The ionic strengths of different samples of serum are so nearly equal that the activities in eq 1 can be replaced by concentrations. The activity coefficient is then incorporated into
Eo. NO value for
IzpOtc~-,uric acid was found in the literature. has a theoretical value of 420, however, which can be calculated from the ratio of solubility product constants of AgCl and AgBr, according to Buck (1). Biases due to bromide could therefore be as high as 80%. Equations pertaining to diffusion in a plane sheet (5) were needed to interpret the ratiotracer data reported here. Flux of species through a membrane under steady-state conditions is given by eq 2, where C1 and C2 are the concentrations of kPotCl-pr-
Flux = P(C1- C J / l
(2)
species in aqueous solutions on either side of the membrane, 1 is the thickness of the membrane, and P is the permeability constant. It is assumed that
P=KD
(3) where D is the diffusion coefficient of the species and K is its partition coefficient (concentration inside the membrane/concentration of the solution). Equation 2 describes permeability under steady-state conditions. Equation 4 is more general and applies even before steady state is achieved, provided Fick's law is obeyed. C
is the concentration of species within the membrane a t a distance x from an interface with distilled water. The other side of the membrane is bathed by a solution of concentration C,. r is the diffusional time constant P/D. The total quantity of species Q, which will have diffused through the membrane up until the time t is given by eq 5.
Q, = y[t - 7/6 - 27
(-l)ne_,2*t,, r2m=i n2
C
If a species is allowed to enter the membrane from one side, but is not allowed to enter or leave it at the other side, then the membrane will fill up according to eq 6. Q, is the total
amount of species in the membrane at time t and Qm= KlC,.
EXPERIMENTAL SECTION Two silver/silver chloride electrodes overcoated with 5 g/m2 of cellulose acetate (37.6% acetyl) were used in this work. One of the electrodes could not be stripped of its overcoat so its uncoated counterpart was used for comparative studies. The other overcoated electrode could be stripped. The cellulose acetate was later changed to 2 g/m2 (39.4%acetyl), and a third electrode, so formulated,was also used in some of the experiments. A small piece of membrane could be stripped from this electrode and one other, overcoated with 3 g/m2of cellulose acetate (39.4% acetyl),
0003-2700/81/0353-1164$01.25/00 1981 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 53, NO. 8, JULY 1981
for use in the measurement of transference numbers. All of these electrodes had been manufactured as continuous sheets several inches wide (2)to a high degree of reproducibility. An overcoated silver/silver bromide electrode was also needed for some of the experiments. It was prepared by converting a bare silver chloride electrode to silver bromide in a 0.1 M potassium bromide solution and overcoating it with 5 g/m2 of celldose acetate (37.6%acetyl),from a 5% acetone solution. Part of the converted electrode was left uncoated and served as a reference. A similar cellulose acetate membrane was also coated onto a glass plate such that its surface density was 20 g/m2. The free acid content of cellulose acetate (37.6% acetyl), measured by potentiometric titration in tetrahydrofuran with 0.1 M tetramethylguanidine,, was 0.007 mequivlg. Potentiometric measurements were made with an Orion Model 801 pH meter in the millivolt mode and recorded on a Fisher Recordal Series 5000 recorder. To measure the severity of uric acid interference, we used a cell assembly consisting of a Beckman 39047 general-purpose cation glass electrode connected to the high-impedanceterminal of the pH meter and a bare silver/silver chloride electrode connelcted to the reference terminal. A piece of flat silver/silver chloride electrode was sandwiched between two Plexiglas plates. One plate had been drilled through, thus forming a cavity with the bare electrodeat the bottom. The cavity was completelyfilled when a 700-pL sample was introduced along with the bulb of the glass electrode. A Teflon tape spacer prevented the sample from spreading between the Plexiglas and the silver/silver chloride electrode. Samples were made by spiking a human serum pool having a uric acid concentration of 4.6 mg/dL (deteImined by Technicon Autoanalyzer Method N-13b) with uric acid in 5 mg/dL increments. The serum pool had a chloride concentration of 104 mM and a pH of 7.93. Dependence of response time upon concentration, pH, and preconditioning was studied by using the cell: bare electrode/ solution/coated electrode. Solutions consisted of 10-pL drops of saline containing various concentrations of sodium chloride and/or sodium bromide. These were sandwiched between pieces of flat electrode about 3 cm wide and 8 cm long. Only about 1 cm2 of the total area of either electrode in the sandwich was contacted by solution, however. A spacer kept the electrodesfrom contacting each other. pH dependencewas studied with samples of 20 mM sodium chloridle constituted with standard calibration buffers at pH 4, 7, and 10 and the electrode overcoated with unstrippable cellulose acetate. In comparative experiments, electrodes were preconditioned by bathing them 5 min in 100 mM sodium chloride, wiping them dry, and allowing them to stand overnight before use. Ionic strength effects were investigated by including 1 M sodium nitrate in some of the solutions. Other electrodes were soaked in distilled water for various periods of time immediately prior to use. Cellulose acetate membranes were used to separate the compartments of a two-chambered cell. One of the chamberswas fiied with distilled water and the other with a saline solution. The flux of chloride or bromide ions from the saline solution through the membrane and into the other chamber, which contained only distilled water initially, was monitored with a Beckman general-purpose cation glass electrode referenced against either a silver/silver chloride or a silver/silver bromide anodized wire electrode. Permeability constants of chloride and bromide ions were determined with 2 M solutions of the appropriate sodium halide and the cellulose acetate membrane (5 g/m2)stripped from one of the electrodes. The exposed surface area of the membrane was 0.713 cm2,and the volume of each solulion was 25 mL. The variation of potential ovler -30 min was used to determine the flux in eq 2. P was then determined from C1 = 2 M and Cz = 0 with 1 = 3.9 X lo4 cm. I was estimated by dividing the coverage (5 g/m2)by the density of cellulose acetate [1.3according to Barker and Thomas ( S ) ] . Permeability constants were also determined below 2 M by using radiotracer techniques with 36Cl-and 82Brto detect the small quantities of chlorideand bromide which could penetrate the 20-g/m2 cast membrane. Chloride transference numbers for differtent membranes were determined in the same two-chamberedcell using saline solutions of 10 or 100 mM sodium chloride. For these measurements the glass electrode was moved over to the “saline solution” side of
1165
Table I. Permeability Coefficients for j6C1- and 82Brpermeability composition of - constant b solutiona v1‘*Br- method 2 M halide 1.7 3.8 P 10 mM C10.50 R 30 mM Cl0.87 R 100 mM C10.69 R 100 mM C1- + 1.9 R 0.4 mM Bra C , in eq 2. C , = 0. b x 109 in units of cm2/s. potentiometric, R = radiotracer.
P=
the cell while the silver/silver chloride electrode remained in the distilled water. The saline solution with its glass electrode now served as a reference. The concentration of the solution on this side of the cell had no effect upon the measured transference number. Concentrated sodium chloride was added stepwise to the “distilled water” side of the cell, thus varying its concentration from nominally zero to 120 mM. Transference numbers were then calculated from linear portions of the potential/log concentration curves. Radiotracer measurements were performed on the overcoated electrode which could be stripped of its cellulose acetate. Solutions of 0.1 M sodium chloride labeled with Wl- were spiked either with 0.5 or 1.0 mM sodium bromide labeled with in one set of experiments or with 0.5 mM uric acid labeled with 14C in another. Differentregions of the electrode were spotted with 50-pL drops of these various solutions for known periods of time, and the drops were then blotted up. In this way the sorption of species by the overcoated electrode was followed as a function of time. After each determination the cellulose acetate was stripped from the surface of the electrode and counted separately to determine the amount of species sorbed by the membrane. The bare electrode was then counted to determine the amount of species sorbed onto the electrode itself. Sorptions of these species by the electrode, stripped of its overcoat before spotting, were also determined, and these measurements serve as controls. Two aqueous solutions of uric acid (concentration 3.3 mg/dL) were titrated conductometricallywith 0.1 M silver nitrate at pH 6.15 and 8.15. pH was controlled by the addition of 0.1 M sodium hydroxide. Reagents were added in 2.5 pM aliquots, and several minutes were allowed after each addition to establish equilibrium. A combination pH electrode was used to monitor pH. In similar experiments at lower pH, a silver wire was used to monitor pAg. These measurements were particularly valuable below pH 4.
-
RESULTS AND DISCUSSION Bromide. The theoretical value of the selectivity coefficient, k P o t ~ l - , ~ r - is , 420 (I), predicting a bias of about 80% in the analysis of serum. Usually biases no greater than -20% were seen. This difference was due to the fact that potentials were normally recorded after 3 min of exposure to the sample-before bromide could substantially convert the silver chloride to mixed crystals and therefore exert the full influence of its interference (7,8).Disposable, one-use, analytical devices fashioned from these electrodes take full advantage of this effect (9). Data will be presented elsewhere to show that the cellulose acetate overcoat further reduces the bias to insignificance. It might be assumed that cellulose acetate minimizes bromide interference by simply blocking this ion while allowing chloride to pass. The permeability constants listed in Table I clearly indicate that this is not the case. The permeability of bromide is more than twice that of chloride a t 2 M sodium halide concentration. This 211 ratio is consistent with a t least one literature value (IO), and the permeability constant of chloride has also been confirmed (11). Results obtained a t lower concentrations by radiotracer techniques are roughly half those found by potentiometry a t the higher concentration and with a thinner membrane. The
1160
ANALYTICAL CHEMISTRY, VOL. 53, NO. 8, JULY 1981
LO mM NaBr
‘0
2
4
6
8
10
Qm I2
~
14
16
18
20
0
Time, min
06t
2
4
6
A 5 m M NaBr
8
10
6
8
IO
12
14
I6
I8
x)
Flgure 3. Sorptlon of bromide by the electrode of Figure 2 with its membrane prevlously strlpped off. Solutions are 100 mM in sodium chloride with the indicated concentrations of sodium bromide.
12
I4
16
18
M
Time, min
Flgure 2. Sorption of bromide by the electrode overcoated with strippable cellulose acetate (5 g/m2, 37.6% acetyl): (0)total sorption; (X) sorption by the silver chloride layer. Solutlons are 100 mM In sodium chloride with the indlcated concentrations of sodium bromide. ratio of bromide to chloride permeabilities, however, is virtually the same in the two experiments, suggesting a discrepancy in thickness or a difference in preparation between the two membranes. Radiotracer determinations of the sorption of %C1-(Figure 1) and of 82Br-(Figures 2 and 3) provide an explanation for the decrease in bias due to bromide. Notice in Figure 1 that the membrane fills with chloride after about 3 min (Q,) and that the electrode is very nearly at equilibrium with respect to chloride. The potential generated by the electrode will reflect the solution composition because, at thermodynamic equilibrium, the electrochemical potentials of all chloride ions are equal, regardless of location in the cell. Therefore, the difference in electrical potential between chloride in the membrane and chloride in the solution will be exactly compensated by the difference in electrical potential between chloride in the silver chloride layer and in the membrane. Figure 1 shows that chloride exchanges slowly into the bare silver chloride (control). This is consistent with the arguments presented elsewhere (7,8) concerning the bulk, as opposed to surface, nature of this reaction. It should not affect the time response of the potential at zero current, however, since exchange of s5Cl-for 36Cl-results in no net reaction and the interface is at thermodynamic equilibrium. The curve labeled ‘‘Q; in Figure 1 indicates the total buildup of %Cl-on the surface of the silver chloride under the membrane and should obey eq 5 at short times, represented by the linear region shown on the curve, provided Fick’s law is obeyed. Equation 5 allows no provision for the membrane to fill up, however, which obviously occurs, since both Q, and Q, approach constant values. Use of the equation is therefore restricted to the linear region. Even then, the parameters will be only approximate. The slope of the curve near the in-
-
flection point yields P = 0.64 X lo4 cm2/s. The value of Q, as t 03 yields K = 0.09. These two parameters can be used to calculate a diffusion coefficient D of 7 X cmz/s. According to eq 5, the intercept of the line should occur at 7/6 = 0.06 min. It actually occurs at -0.3 min, implying a smaller value of D (since r = P/O)during establishment of the steady-state concentration gradient through the membrane than actually exists at steady state. These observations are consistent with the results of Lonsdale, Merten, and Riley (II), who report K = 0.046,0.077, P = 1.2, 1.7 X cm2/s, and D = (4.3 f 2.0) X cm2/s. They determined D by fitting sorption data to eq 6. Notice that the steady-state value of D (P/K)would be about an order of magnitude larger than the dynamically determined value. Their solution concentration was 0.5 wt % sodium chloride. In contrast to chloride, the data shown in Figure 2 indicate that virtually no bromide is retained by cellulose acetate (Q, = 0). However, the amount of bromide which is actually sorbed (exchanged) into the silver chloride is only about 2% (after 3 min) of the amount which exchanges into the bare control (Figure 3). From eq 5, P = 3 X lo4 and 2.7 X cm2/s at 0.5 and 1.0 mM bromide, respectively. The intercept point on the time axis predicts D = 0.5 X cm2/s at both concentrations, so that K would be 6.1 at 0.5 mM and 5.4 at 1.0 mM. The permeability constants are reasonably close to the values listed in Table I; however, the values of K are about 2 orders of magnitude larger than expected and predict substantial amounts of bromide in the membrane (0.11 and 0.24 pg/cm2 at 0.5 and 1.0 mM, respectively). This quantity of bromide was not detected, indicating that, as with chloride, the diffusion coefficient of bromide is much larger at steady state than it is during the achievement of steady state. It appears therefore that the diffusions of chloride and of bromide are decidedly non-Fickian. The data are more consistent with a model in which these ions move through the membrane as “fronts” of constant velocity until their steady-state concentration profiles are established. Such situations occur when the diffusion coefficients are concentration dependent such that diffusion is more rapid at higher concentrations or when diffusion is preceded by chemical reaction (5). The surface concentration of bromide indicated in Figure 2 after 3 min is not enough to cause a large potential variation and therefore a large bias. Since the silver halide layer remains predominantly in the chloride form, it will be quite insensitive to bromide (7,8). Furthermore, the exchange of bromide must be rapid in order to maintain the linear relationship shown in Figure 2. This would effectively maintain its surface concentration near zero and further reduce sensitivity. This interpretation, illustrated schematically in Figure 4, is in keeping with the theory of Hulanicki and Lewenstam (12), who attribute “lower than expected” selectivity coefficients
h
0
4
Time, min
Flgure 1. Sorption of chloride by the electrode overcoated with strlppable cellulose acetate (5 g/m2, 37.6% acetyl), solutlon concentration 100 mM sodium chloride: top curve, control-uncoated electrode; middle curve, sorption by the silver chlorlde layer; bottom curve, sorption by the membrane.
/
2
ANALYTICAL CHEMISTRY, VOL. 53, NO. 8, JULY 1981
Dlstance
Figure 4. Schematic reprt~entation(not to sciale) of the concentration profiles of Chloride and bromide ions when the cell potential is mea-
sured. to diffusion effects. They predict that the selectivity coefficient equals the ratio of diffusion coefficients rather than the ratio of solubility product constants. The ratio of permeability coefficients (Table I) is about 2. That would predict, at most, 0.4% bias due to bromide in human serum. IZPOtC1-lB,- was less than 5 in other experimients with overcoated electrodes. Uric Acid. Potentiometric data for uric acid interference agreed with eq 1, within the clinical range, provided kPOtCl-luricacid = 1.4 mMJ(mg/dL) or 24 in dimensionlessunits. This selectivity coefficient predicts a bias of -7% in serum analysis. Measurementsmade with sera adjusted to lower pHs show that the bias is pII dependent and becomes insignificant below pH -6, provided potentials are irecorded 3 min after exposure (13). Extensive evaluations of the electrodes overcoated with cellulose acetate indicate negligible sensitivity to uric acid in sera containing the highest clinically significant levels. Conductometric titrations gave insighit into the mechanism of interference by uric acid. When the pH was maintained at 8.15, two end points were seen. The fiist was stoichiometric, providing a concentration only 3% below that determined by Technicon Autoanalyzer Method N-13b. The second occurred about 15% earlier than expected. Up to the first end point, pH could be controlled by addition of hydroxide, after each silver nitrate addition, equal to the concentration of added silver ions. Beyond the first end point, pH control was extremely difficult owing to sluggishness with which equilibrium was reestablished afteir each addition of silver nitrate. Only about one-third as much hydroxide as silver ions was needed, however. Beyond the second end point, pH control became unmanageable. Conductivities due siolely to uric acid and the products of its reactions were calculated by subtracting the contributions of sodium and nitrate i~onsfrom the measured values. These adjusted conductivitiehi remained constant up to the first end point, decreased until the second, and remained constant again thereafter. The equiv,dent conductanoe of urate was determined to be 47 at 1mM sodium urate, pll8.1. Similar resulk were obtained a t pH 6.15 except that the break was not as sharp at the first end point and the equivalent conductance of urate was about 11, reflecting the higher degree of protonation. The titration was not pursued all the way to the second end point. The data are consistent with a simple ion-exchange reaction in which silver ions displace protons from the urate monoanion up to the first end pomt.
1167
No effort was made to verify the structure of the proposed silver urate anion. It is consistent, however, with the results of crystallographic investigations of 4-hydroxy-6-methyl1,3,3a,7-tetrazaindene ( I d ) , which is considered a model compound. Since the pKA of uric acid is 5.75, according to Dryhurst et al. (15), the silver complex formed in this ionexchange reaction is expected to be negatively charged and capable of complexing an additional silver ion after the first end point, thus causing the observed decrease in the conductivity of uric acid related products. This reaction would not be pH dependent since no proton exchange is involved. Below the first end point the solution is slightly discolored (yellowish brown) but becomes distinctly yellow after the first end point. The color deepens with further addition of silver nitrate and turns orange after the second end point. On standing, the solution turns dark brown and becomes colloidal, indicative of oxidation by excess silver ions. Even solutions which contain silver nitrate a t concentrations intermediate between the first and second end points eventually show evidence of oxidation on prolonged standing (overnight). The small pH changes observed after the first end point may be caused by the oxidation. Below the first end point, color and pH are stable for a t least 24 h. To adjust solution pH, we added nitric acid or sodium hydroxide to solutions which contained equal 1mM concentrations of silver nitrate and sodium urate. pH and pAg were measured after each adjustment. Below pH "4, uncomplexed uric acid is expected to be fully protonated. Material-balance equations can then be used to calculate the concentrations of silver urate (AgHU) and of its conjugate base (AgU-) at the different pHs. From these calculations, the pKA of AgHU wm found to be 3.25 f 0.05, and the ion-exchange constant for the reaction Ag+ H2U -+AgHU H+
+
+
where H2U designates uric acid, was determined to be 22 f 2. The ion-exchange constant for the reaction Ag+ HU- + AgU- H+
+
+
can be calculated from these constants and the PKAof uric acid. A value of nearly 7000 is obtained. Simple stoichiometric calculations using this ion-exchange constant and the solubility-product constant of silver chloride reveal that nearly all of the uric acid will react with silver chloride at pH 8, even in the presence of 100 mM sodium chloride. At pH 3, only about 1% will react. These considerationsexplain the alleviation of interference by uric acid at the lower pHs. However, the severity of the interference at higher pHs cannot be attributed solely to chloride released by reaction of uric acid with silver chloride. The released chloride could never be greater than the amount of uric acid initially present. Exclusive of diffusional effects (12),the selectivity coefficient could not be greater than one. For larger values of the selectively coefficient, a silver complex of uric acid must be incorporated into the matrix of the silver chloride crystal. The solid-state activity of silver chloride would then be reduced below unity. According to Buck ( I ) , its activity will be
aAgc1 = (acr/&p~C')aAg+ (7) is the solubility-product constant of silver chloride. The solid-state activity of silver urate in the proposed mixed crystal would be ~ A ~ H= U( a H u - / K ~ p ~ ~ " ) a A ~ + (8) where KspkHuis the solubility-product constant of silver urate. uHU- is related to the total activity of uric acid in solution according to eq 9. uHzU is the total, or formal, activity of uric QHU- = KAQH,U/(QH+ + KAYH~U/YHU-) (9)
1188
ANALYTICAL CHEMISTRY, VOL. 53, NO. 8, JULY 1981
> 300
I. $2
E 200 aJ
2
100
0
1
2
3
4
5
6
7
Time, mln
Flgure 8. Potential/time responses of the electrode overcoated with unstrippable cellulose acetate (5 g/m2, 37.6% acetyl) to different concentrations of sodium chloride. Time, rnin
; 400
Flgure 5. Sorption of uric acid by the electrode overcoated with sMppable cellulose acetate (5 g/m2, 37.6% acetyl): top curve, sorption by the membrane-curve calculated from eq 6, truncated after the first term (parameters given in the text.); mlddle curve, controluncoated electrode: bottom curve, sorption by the silver halide layer. Solutlons are 100 mM in sodium chloride with 0.5 mM 14C-labeleduric acid.
acid and UH+ is the activity of protons. YH?U and YHU- are activity coefficients of uric acid and its conjugate base, respectively. Equations 7, 8, and 9 can be combined, subject to the condition that the solid-state activities of silver chloride and silver urate must sum to unity in the mixed crystal. An expression for the activity of silver ions results.
The electrode ultimately responds to silver ions. Therefore substitution of eq 10 into the Nernst equation
E = EaAg+/Ag
+ 59 log U A ~ +
(11)
where Eo,+/, is the standard reduction potential of the silver couple, results in an expression for the potential of the electrode relative to the normal hydrogen electrode. By comparison with eq 1, kP’’tCl-,.luricscid is given by eq 12. According
to this expression, the selectivity coefficient can be as high as the ratio of solubility-product constants at high pH but becomes smaller as the pH is lowered, consistent with experimental evidence. Radiotracer data obtained with 14C-labeleduric acid (Figure 5) confirm that uric acid sorbs to the surface of the bare silver/silver chloride electrode (“control” in the figure). However, after 3 min 90% of the uric acid imbibed by an overcoated electrode remains within the cellulose acetate (Qm). Only 10% adsorbs to the surface of the silver chloride (Q8). Furthermore, the surface concentration is only one-sixth of the corresponding amount using the bare control. The presence of a large quantity of free acid (0.007 mequiv/g -9 mM) in cellulose acetate was determined by potentiometric titration, as mentioned before. The pH of serum should therefore be decreased to a low value within the cellulose acetate membrane by this free acid. Exclusion of proteins by cellulose acetate might further aid acidification, since proteins have considerable buffer capacity (16). Consistent with the arguments presented above, the observed decrease in interference by uric acid is then readily related to the pH dependence of the selectivity coefficient.
r
Unbuffered
2
IO0
0
1
2
3
4
5
6
7
Time, min
Flgure 7. Effects of pH on the potentiil/time response of the electrcde overcoated with unstrippablecellulose acetate (5 g/m2, 37.6 % acetyl). Solutions are 20 mM in sodium chloride, pH shown beside each curve.
Interpretation of the radiotracer data using eq 6 yields the parameters D = 2.9 X cm2/s and K = 2. A plot of In (1 - Qm/Qm)vs. time predicts an intercept at -0.21, equal to the theoretical value (In 8/r2)predicted by eq 6. Fickian diffusion is therefore indicated. The curve labeled “Q,” in Figure 5 was calculated from eq 6 using the parameters reported. Use of eq 6 is justified in this case even though it was assumed in its derivation that the membrane would be completely fiied in the limit that t m and that no reaction could occur at the membrane/silver chloride interface that would remove species at that side. The fact that the surface concentration of uric acid (QJ is quite large clearly indicates that these assumptions are not completely valid. However, more extensive interpretation of the data indicates that the membrane is at least 93% full and that the heterogeneous rate constant (based on macroscopic surface area) for surface sorption cannot be greater than 1.4 X lo-’ cm/s. A heterogeneous rate constant of 17 X lo-’ cm/s can be calculated from the data corresponding to the uncoated control. Departure from the ideal nature of the equation is therefore not too serious. The data indicate that cellulose acetate protects the surface of the silver chloride from attack by uric acid, not by preventing the uric acid from reaching the surface but by controlling the kinetics of its sorption (and subsequent reaction) once there. The apparent heterogeneous rate constant is at least an order of magnitude slower in the cellulose acetate than in water. Response Time. Cellulose acetate overcoats effectively eliminate interferences by bromide and uric acid. They have the disadvantage, however, that they cause slow responses to chloride solutions at concentrations below 100 mM, as shown in Figure 6. The effect is most dramatic at about 20 mM. Furthermore, the response time (defined graphically in Figure 6) is quite pH dependent, as shown in Figure 7. The overcoated electrode responds most rapidly to basic solutions but slowly to acidic ones. Response to an unbuffered solution is extremely slow, further implying that pH is decreased within the membrane. It seems reasonable that fixed negative charges (carboxyl groups) (10,17) are responsible for the concentration and pH dependences shown in Figures 6 and 7. Chloride ion transference numbers through various membranes are quite small
-
ANALYTICAL CHEMISTRY, VOL. 53, NO. 8, JULY 1981
Table 11. Chloride Ion Transference Numbers through Cellulose Acetate Membranes iransference surface break no. below cellulose acetatea densityb pointc break pointd 37.6 39.4 39.4
5 3 2
25 13 17
0.26 0.08 0.08
a % acetyl. g/mz. mM-Potential/log concentration curves are linear above and below this concentration but have different slopes. d Transference numbers above the break point are near 0.5 in all cases.
Table 111. Comparison o f Response Tim@with and without the Presence of 1 M NaNO, response time cuncn a with without 10 40 80 120 a
mM NaCl.
6.0 2.0 1.5 1.5
13.2 4.0 1.5 1.0
Minutcas-as defined in Figure 6.
below -20 mM sodium chloride, as shown in Table 11. Above 20 mM the membranes more readily allow chloride to pass, as evidenced by the increase in transference number. Notice in Figure 6 that the most dramatic increase in response time also occurs when the concentration is reduced to 20 mM. Heyde, Peters, and Anderson (17)found that the isoelectric point of Eastman 4644 cellulose acetate (39.8% acetyl) occurs a t pH 3 f 0.2. They estimate the fixed site concentration to be 2.7 mequiv/L of membrane. The 9 mequiv/L of membrane free acid found by titration will therefore protonate all of the carboxylate sites. Then when sodium chloride diffuses into the membrane, sodium ions can be exchanged for protons. Protons will tend to diffuse toward the surface of the electrode and also toward the bathing solution. Chloride ions will accompany the protons in order to maintain electroneutrality. Since aqueous diffusion coefficients are 3 to 4 orders of magnitude larger than coefficients of diffusion through cellulose acetate, it follows that protons and chloride ions will be preferentially swept away from the surface of the electrode and back out into the solution. Normal diffusion cannot occur until the ion-exchange process is complete and the flux of protons ceases. At pHsi above 3 (the isoelectric point) the protons can combine with hydroxide ions, thus freeing the chloride to diffuse through the membrane to the surface of the electrode. This mechanism becomes more efficient as the pH is raised and the response time of the electrode decreases, consistent with Figure '7. According to this argument, pretreatment of the overcoated electrodes with a saline solution should drive the ion-exchange reaction to completion and allow normal diffusion to occur. Indeed, electrodes which had been exposed to 100 mM sodium chloride solutions for 5 nnin and then wiped dry showed rapid (< 1min) and Nernstian responses to solutions down to 20 mM sodium chloride even after storage in the open air for several days. Potentials were Nernstian in the clinical range even in the presence of 1mM sodium bromide. When used for the analysis of serum, the electrodes are not preconditioned in any way. More dilute samples can be analyzed if the electrodes are soaked briefly in 100 mhlI sodium chloride; however, no incubation studies were performed to determine the long-term stability of electrodes so preconditioned. If chloride is electrostatically rejected from the membrane, then response time ought to be quite dependent upon ionic strength. Table 111 shows a comparison of response times
-
1169
Table IV. Effect That Presoaking of Overcoated Electrode Has on Its Response Time time of response response time of timeb soakinga timeb soakinga 0
3 4
2.5 2.7 2.5
1 2
2.2 2.1
Minutes-as defined in In distilled water (minutes). Figure 6. Exposure is to 60 mM sodium chloride. a
+ C
300
,
4ooL
0
1
2
3
4
5
6
7
8
Time, rnin
Figure 8. PotentiaVtimeresponses of the electrode overcoated with cellulose acetate (2 glm', 39.4 % acetyl) to different concentratlons of sodium bromide.
10-
0. 0
5 6 1
2
3
4
5
6
7
Time, mln
Flgure g. PotentiWtime responses of a silver/silver bromide electrode overcoated with cellulose acetate (5 g/m'), 37.6% acetyl) to various solutions. Solution compositions: (1) 1 mM NaBr, (2) 100 mM NaCI, (3) 2 mM NaBr 100 mM NaCI, (4) 1 mM NaBr 100 mM NaCI, (5) repeat of (4), (6) repeat of (2).
+
+
(defined in Figure 6) obtained with solutions with and without 1 M sodium nitrate. Notice that response is definitely enhanced by the medium with higher ionic strength, but the concentration dependence is still evident. An effect specifically involving chloride is implicated, and the qualitative electrostatic argument presented above may be too simplistic. Water imbibition and rate of swelling have little to do with the response time of the electrode, as shown in Table IV. Pieces of the electrode overcoated with unstrippable cellulose acetate (37.6% acetyl) were soaked in distilled water for different periods of time immediately before exposure to 60 mM sodium chloride. Overcoated electrodes respond to solutions of sodium bromide with concentration dependence similar to that seen with sodium chloride (Figure 8). The data suggest the possibility that interference by bromide is reduced because it diffuses through the membrane more slowly than chloride
1170
Anal. Chem. 1981, 53, 1170-1175
due to its lower concentration. The permeability constant listed in Table I for 82Br-in the presence of 100 mM sodium chloride indicates that this is probably not true. Experiments with the cellulose acetate overcoated silver/silver bromide electrode (Figure 9) further suggest the incorrectness of this interpretation. The electrode responds slowly and irreproducibly to sodium chloride solutions which contain no bromide (curves 2 and 6), but rapidly and reproducibly (relatively speaking) to sodium chloride solutions which contain 1or 2 mM sodium bromide (curves 3,4, and 5). In the absence of sodium chloride, 1mM sodium bromide was not sensed even after 7 min (curve 1). It is to be expected, then, that even though the results shown in Figure 8 may reveal slow response to pure sodium bromide solutions, responses to low concentrations of bromide in the presence of 100 mM sodium chloride are quite fast.
ACKNOWLEDGMENT The help of C. Battaglia, D. Secord, S. Kim, and various members of Kodak service labs is acknowledged. The author is especially grateful to R. Miller and his associates, who provided the radiotracer data at the requests of C. Battaglia and D. Secord. The author also gratefully acknowledges the technical assistances of N. Cody and L. Smith.
LITERATURE CITED (1) Buck, R. P. Anal. Chem. 1968, 40, 1432. (2) Kim, S. H.,U S . Patent 4 199411, April 22, 1980. (3) Koryta, J. “Ion Selective Electrodes”; Cambridge University Press: New York, 1975; p 64. (4) Tietz, N. W. “Fundamentals of Clinical Chemistry”; W. B. Saunders Co.: Philadelphia, PA, 1 9 7 0 pp 936-938, 956. (5) Crank, J. “The Mathematics of Diffusion”; Ciarendon Press: Oxford, England, 1975; Chapters 4, 9, and 13. (6) Barker, R. E., Jr.; Thomas, C. R. J . Appl. Phys. 1964, 35, 87. (7) Rhodes, R. K.; Buck, R. P. Anal. Chlm. Acta 1980, 173, 67. (8) Sandifer, J. R. Anal. Chem. 1981, 53, 312. (9) Curme, H. G.; Babaoglu, K.; Babb, B. E.;Battagiia, C. J.; Beavers, D. J.; Bogdanowicz, M. J.; Chang, J. C.; Daniel, D. S.; Kim, S. H., et ai., Abstract 262, Clln. Chem. 1979, 25, 1115. (10) Lonsdale, H. K. In “Desalination by Reverse Osmosis”; Merten, U., Ed.; MIT Press: Cambridge, MA, 1966; Chapter 4. (1 1) Lonsdale, H. K.; Merten, U.; Riley, R. L. J. Appl. Pdym. Scl. 1965, 9 , 1341. (12) Hulanicki, A.; Lewenstam, A. Talanta 1977, 24, 171. (13) Battagiia, C. J., Eastman Kcdak Company, Rochester, NY, unpublished results, 1977. (14) Smith, D. L.; Luss, H. R. Photogr. Sci. Eng. 1976, 20, 184. (15) Wrona, M. Z.; Owens, J. L.; Dryhurst, G. J. flectroanal. Chem. 1979, 705, 295. (16) Siggaard-Anderson, 0. “The Acid-Base Status of the Blood”, 4th ed.; Williams and Wiikens: Baltimore. MD, 1974; p 42. (17) Heyde, M. E.;Peters, C. R.; Anderson, J. E. J. Colloid Interface Sci. 1975, 50. 467.
RECEIVED for review October 10,1980. Accepted April 6,1981.
Comparison of X-ray Photoelectron Spectroscopy and Cyclic Voltammetry for the Determination of Polymeric Film Thickness of Ruthenium Vinylbipyridine and Vinylferrocene Deposited on Electrodes M. UmaEa,‘ P. Denisevich, D. R. Rolison,* S. Nakahama,’ and Royce W. Murray* Kenan Laboratories of Chemistry, University of North Carolina, Chapel Hill, North Carolina 275 14
Fllm thlcknesses for poly[tris(4-methyl-4’-vinyl-2,2’-blpyrldlne)ruthenlum(II)]2*(X’-)2 and poly(vlny1ferrocene) deposited on Pt electrodes by electrochemically and plasma Initiated polymerlzatlon, by spontaneous adsorptlon, and by photochemlcal depositlon were obtained from angular dlstributlons of Pt 4f substrate photoelectrons In X-ray photoelectron spectroscopy (XPS) and from cyclic voltammetrlc measurement of the quantity of electrochemically reactive polymer. The fllm thlcknesses determined by the two approaches agreed for two of the six fllms, two others differed due to apparent dlfferences between bulk and fllm denshies, and two others dlffered because of nonunlformltles in film thickness.
Our laboratory has developed chemistry and procedures for coating metal and semiconductor electrode materials with thin films of polymers (1-11) which contain as structural members electron donors and acceptors such as ferrocene, 2,2’-bipyridine complexes of iron and ruthenium, and pyridinepentachloroiridate. These “redox polymers” and those devised by others Present address: A l l i e d Chemical Corp., Solvay, NY 13209. 2Present address: N a v a l Research Laboratory, Washington, DC
20375. Present address: T o k y o I n s t i t u t e Meguro-ku, T o k y o 152, Japan.
of Technology, 0-Okayama,
as thin films on electrodes (12-54) undergo facile electron exchange reactions with the electrode which oxidize or reduce the equivalent of many monomolecular layers of electroactive sites. The mechanism and kinetic aspects of such reactions are of interest as are redox polymer applications to electrocatalysis (18,19,29,30,36,37,40-43,48,53),photochemistry (26, 43, 46-49, 51-54), and rectifying interfaces (8). Much more is known presently about electrochemical properties of redox polymer films than about the physical arrangement of the polymer films. Notably, overall thickness and imperfections such as thin or thick patches and pinholes have important electrochemical implications but are difficult to evaluate electrochemically (24). This paper reports measurements of thickness ( d ) of redox polymer films, based on the angular distribution of photoelectron emission intensity in X-ray photoelectron spectroscopy (XPS). Theory, equipment, and applications of angular distribution XPS have been described by Fadley and others (55-62). Intensity measurements as a function of angle have been carried out on inorganic films (55-59) and an organic polymer example (61). This report is the first dealing with metal complex or organometallic films.
RELEVANT THEORY Essential geometrical features of the angular distribution XPS experiment are shown in Figure 1. The film thickness measurement can be based on substrate photoelectrons whose
0003-2700/81/0353-1170$01.25/0 0 1981 Amerlcan Chemlcal Society
