Anal. Chem. 2001, 73, 4262-4267
Quantification of the Concentration of Ionic Impurities in Polymeric Sensing Membranes with the Segmented Sandwich Technique Yu Qin and Eric Bakker*
Department of Chemistry, Auburn University, Auburn, Alabama 36849
Ionic impurities in solvent polymeric membranes have been the main reason early ion-selective electrodes (ISEs) without added ion exchanger exhibited a functional potentiometric response. Today, knowledge of such impurities and their elimination becomes important in view of designing ISEs with low detection limits and voltammetric ion-selective electrodes, to increase operational lifetime, and to optimize sensing selectivity. Here, a potentiometric segmented sandwich membrane technique is used to study the amount and properties of impurities in situ directly in the membrane phase. This technique can be used with a number of ionophores and with different membrane matrixes and does not require specialized equipment. The concentration of anionic impurities in PVC-DOS (bis(2-ethylhexyl) sebacate) and PVC-NPOE (o-nitrophenyloctyl ether) was found to be on the order of 100 µmol/kg with membranes containing weakly binding potassium ionophores. The concentration of cationexchanger sites of carboxylated PVC-NPOE was found to be higher (1.62 mmol/kg). Addition of the neutral lipophilic salt ETH 500 to both membrane segments had only a marginal effect on the results for PVC-NPOE membranes but had a large impact on the observed membrane potential for PVC-DOS membranes. Theory explains that the addition of such salt to membranes where ion pairing is predominant is essential for the accurate assessment of ionic impurities with this technique. Ion-selective electrodes (ISEs) are widely used in assessing ionic and neutral analytes. Carrier-based ion selective membranes are made of a lipophilic ionophore; a polymer matrix, such as poly(vinyl chloride); and a plasticizer.1 For neutral carrier-based ionselective electrodes, so-called ionic sites are added to the membrane to guarantee a permselective behavior of the membrane.2 From today’s perspective, it seems surprising that early ISEs were not used with such sites, but they still gave functional responses. This was possible because of the presence of anionic impurities in the membrane.3 Indeed, a rigorous purification of all sensing components in PVC-free membranes yielded flat, (1) Bakker, E.; Bu ¨ hlmann, P.; Pretsch, E. Chem. Rev. 1997, 97, 3083. (2) Morf, W. E.; Kahr, G.; Simon, W. Anal. Lett. 1974, 7, 9. (3) Horvai, G.; Buck, R. P.; Graf, E.; Pungor, E.; Toth, K. Anal. Chem. 1986, 58, 2735.
4262 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001
concentration-independent response curves,4,5 as theoretically expected.6 Today, the control or elimination of ionic impurities is important in view of optimizing membrane selectivity,7 understanding basic response principles,8 dramatically lowering the detection limit,9 and designing voltammetric ion-selective electrodes.10 Indeed, it is known that membrane selectivity is not only a function of the stability constant of the ion-ionophore complexes and the free energy of transfer of each competing ion, but is often dependent on the concentration of the active sensing components.11 The selectivity of monovalent over divalent ions, for example, can be maximized by minimizing the concentration of ion exchanger in the membrane.7 Currently, membrane impurities are the limit of such an optimization. Detection limits can be lowered dramatically beyond the traditional micromolar concentration levels by using modified measuring and conditioning protocols, different inner filling solutions, and membranes of optimized composition.9 It is desirable to use membranes that contain very low concentrations of ion exchanger, because this reduces transmembrane fluxes. Again, the concentration of impurities imposes a high limit in this respect. Indeed, lowering the concentration of active membrane components goes currently at the expense of selectivity.9 Further, a number of ionophores do not behave as classical neutral ligands, but can act as charged carriers,8 mixed mode carriers, can form ion pairs, and can be involved in a number of equilibrium processes that complicate the ion-selective electrode response.12 To elucidate the mechanism of such ionophores, different concentrations and charge types of added lipophilic ion exchanger are added to the membrane, and the electrode slope and selectivity are evaluated. It would be highly desirable to perform such fundamental experiments with membrane materials that have no ion-exchange properties by themselves. Ionic impurities also define the lifetime criterion of ion(4) Yajima, S.; Tohda, K.; Buhlmann, P.; Umezawa, Y. Anal. Chem. 1997, 69, 1919. (5) Buhlmann, P.; Yajima, S.; Tohda, K.; Umezawa, Y. Electrochim. Acta 1995, 40, 3021. (6) Karpfen, F. M.; Randles, J. E. B. Trans. Faraday Soc. 1953, 49, 823. (7) Na¨gele, M.; Pretsch, E. Mikrochim. Acta 1995, 121, 269. (8) Bakker, E.; Malinowska, E.; Schiller, R. D.; Meyerhoff, M. E. Talanta 1994, 41, 881. (9) Ceresa, A.; Bakker, E.; Hattendorf, B.; Gu ¨ nther, D.; Pretsch, E. Anal. Chem. 2001, 72, 343. (10) Jadhav, S.; Bakker, E. Anal. Chem. 2001, 73, 80. (11) Eugster, R.; Gehrig, P. M.; Morf, W. E.; Spichiger, U. E.; Simon, W. Anal. Chem. 1991, 63, 2285. (12) Amemiya, S.; Bu ¨ hlmann, P.; Umezawa, Y. Anal. Chem. 1998, 70, 445. 10.1021/ac0104126 CCC: $20.00
© 2001 American Chemical Society Published on Web 08/03/2001
selective electrodes. It is known that severe selectivity breakdown occurs when the ionophore concentration falls below the concentration of ion exchanger in the membrane.13 Even voltammetric ion sensors would greatly benefit from pure membrane materials without intrinsic ion-exchange properties. In that case, an externally applied potential forces the extraction of sample ions into the membrane phase. Spontaneous extraction without applied potential, due to ion-exchange properties of the membrane or coextraction effects, interferes with the desired response mechanism.10 The nature of ionic impurities in PVC membranes has been reported.14 The anionic impurities are assumed to originate mostly in the PVC, with a concentration between 0.05 and 0.6 mmol/ kg,15 although it is known that the plasticizer and even the active membrane components may contribute to this concentration.16 Unfortunately, an efficient way to measure these impurities has long been lacking. Nagele and Pretsch reported on the measurement of ionic impurities in the membranes on the basis of determining the selectivity coefficient for ions of different valences as a function of the concentration of ion exchanger added to the membrane.7 This interesting approach involves the measurement of the selectivity of a series of membranes, and is time-consuming. Moreover, only well-behaved model ionophores can be used with this technique. Recently, a new sandwich membrane method was evaluated and shown to yield very useful information about the binding properties of ionophores in ion-selective membranes.17,18 It involves the fusion of two membrane segments of known composition. The membrane potential directly indicates the concentration ratio of uncomplexed ions in both membrane segments. Here, this method is extended to the experimental study of the concentration of anionic impurities in the membrane.
now assumed that the membrane contains an ionophore, L, that forms stable 1:1 complexes IL+. The complex formation constant is inserted into eq 1 to obtain
QST ) [S-] + [QS] + [ILS]
(7)
THEORY This section evaluates the expected influence of added inert lipophilic electrolyte to the assessment of ionic impurities with the segmented sandwich technique. As established recently, the membrane potential, EM, of a concentration-polarized sandwich membrane is described as follows, if both aqueous solutions contacting the membrane are identical.17
QST ) [Q+] + [QS] + [QR]
(8)
+ RT ) [IL ] + [ILS] + [ILR]
(9)
RT ) [R ] + [ILR] + [QR]
(10)
EM )
+ RT [I ]′′ ln + F [I ]′
(1)
It is assumed that the membrane is equilibrated with a monovalent sample cation I+ at both sides, and that the activity coefficients in both membrane segments are identical. The symbols [I+]′ and [I+]′′ denote the uncomplexed ion concentrations of I+ in the organic phase boundary of each segment contacting the sample (′) and inner electrolyte (′′), respectively. The symbols R, T, and F have their established meanings. It is (13) Dinten, O.; Spichiger, U. E.; Chaniotakis, N.; Gehrig, P.; Rusterholz, B.; Morf, W. E.; Simon, W. Anal. Chem. 1991, 63, 596. (14) van den Berg, A.; van der Wal, P. D.; Skowronska-Ptasinska, M.; Sudholter, E. J. R.; Reinhoudt, D. N. Anal. Chem. 1987, 59, 2827. (15) Lindner, E.; Gra`f, E.; Nigreisz, Z.; To´th, K.; Pungor, E.; Buck, R. P. Anal. Chem. 1988, 60, 295. (16) Bu ¨ hlmann, P.; Yajima, S.; Tohda, K.; Umezawa, K.; Nishizawa, S.; Umezawa, Y. Electroanalysis 1995, 7, 811. (17) Mi, Y.; Bakker, E. Anal. Chem. 1999, 71, 5279. (18) Qin, Y.; Mi, Y.; Bakker, E. Anal. Chim. Acta 2000, 421, 207.
EM )
+ RT [IL ]′′ [L]′ ln F [IL+]′ [L]′′
(2)
The membrane segment facing the inner filling solution (′′) may now also contain a known concentration of lipophilic ion exchanger, RT , as well as an optional excess of inert lipophilic electrolyte QS, which may dissociate into the ions Q+ and S-. In analogy to previous work,19 the ion pair formation constants for the formation of QS, QR, ILS, and ILR can be written as follows:
KQS ) KQR ) KILS ) KILR )
[QS] [Q+][S-] [QR] [Q+][R-] [ILS] [IL+][S-] [ILR] [IL+][R-]
(3)
(4)
(5)
(6)
In addition, the following mass balances hold for this membrane segment, where QST is the total concentration of inert lipophilic electrolyte.
For simplicity, the double primes (′′) are here omitted. Each associated species concentration [QS], [ILS], [QR], and [ILR] is written as a function of the dissociated ions by rearranging eqs 3-6. These relationships are inserted into eqs 7-10 to yield
QST ) [S-] + KQS[Q+][S-] + KILS[IL+][S-]
(11)
QST ) [Q+] + KQS[Q+][S-] + KQR[Q+][R-]
(12)
+ + + RT ) [IL ] + KILS[IL ][S ] + KILR[IL ][R ]
(13)
+ + RT ) [R ] + KILR[IL ][R ] + KQR[Q ][R ]
(14)
Three of the unknowns, [Q+], [S-], and [R-], can now be eliminated by combining eqs 11-14. It is now assumed that all four ion pair formation constants are approximately identical, and each is written as K. Both the cation and anion of the added inert (19) Rosatzin, T.; Bakker, E.; Suzuki, K.; Simon, W. Anal. Chim. Acta 1993, 280, 197.
Analytical Chemistry, Vol. 73, No. 17, September 1, 2001
4263
electrolyte are shielded, bulky ions, as is the complexed ionophore. Under these circumstances, this assumption appears to be justified, because all of the ion pairs will be relatively weak and nonspecific.17 Elimination of the three mentioned unknowns, reintroducing the double prime (′′) to denote that this result is valid for the segment contacting the inner filling solution, and solving the result for [IL+]′′ gives
[IL+]′′ )
-RT + RT x1 + KQST + 4KRT
2K(QST + RT)
(15)
The concentration of uncomplexed ionophore in this segment is given by [L]′′ ) LT - RTh , where LT is the total concentration of ionophore (see eq 2). The membrane segment contacting the sample is assumed to behave identically, but no added ion exchanger is present. Consequently, the ion-exchanger properties are given by the anionic impurities of the membrane, which are here denoted with the symbol XT . Substitution of the symbol RT by XT , therefore, gives the respective equation for that segment as
[IL+]′ )
-XT + XT x1 + KQST + 4KXT
2K(QST + XT)
(16)
As above, the concentration of uncomplexed ionophore in that segment is given by [L]′ ) LT - XT . These relationships, including eqs 15 and 16, are now inserted into eq 2 to predict the membrane potential as a function of the concentration of ion exchanger in both membrane segments.
Figure 1. Calculated membrane potential change for a segmented sandwich experiment as a function of the concentration of ionic impurity in one membrane segment. Top: No inert electrolyte is added to any of the segments. Bottom: 20 mmol/kg of inert electrolyte is contained in both segments. Only the segment facing the inner electrolyte contains 5 mmol/kg of added ion exchanger NaTFPB. Curves were calculated using eq 17 and vary with various indicated logarithmic ion pair formation constants only in the top plot (see text).
(18)
plot). The electrode slope for the experiment without added ion exchanger is expected to vary greatly as a function of the indicated logarithmic ion pair formation constant in the membrane phase. Only in the case of negligible ion pair formation, an apparently Nernstian response slope is calculated. Because the ion pair formation constant is not a priori known, such a measurement will give large errors in the determination of the impurity concentration. On the other hand, the bottom plot indicates that the addition of 0.020 mol/kg of a lipophilic electrolyte to both membrane segments yields Nernstian response slopes as a function of the unknown impurity concentration, independent of the extent of ion pair formation in the membrane. The effect of the ion pair formation constant variation on the precision of the measurements was also independently calculated by assuming that KQS ) KILS and KQR ) KILR ) 104 but with KQS being either 10-fold smaller or larger than KQR. For membranes containing an excess (0.02 mol/kg) inert electrolyte, the resulting XT values varied by ∼1%. With no added inert electrolyte, the variation was up to 56%; also for this reason, the addition of inert electrolyte appears to be very important if ion pairing is dominant.
This is more clearly illustrated in Figure 1, where the expected membrane potential change as a function of the impurity concentration XT is plotted on the basis of eq 17. The ionophore is assumed to be in large excess (LT . RT > XT ), RT is 0.0050 mol/kg, and QST is either zero (top plot) or 0.020 mol/kg (bottom
EXPERIMENTAL SECTION Reagents. Sodium tetrakis[3,4-bis(trifluoromethyl)phenyl]borate (NaTFPB), bis(2-ethylhexyl)sebacate (DOS), o-nitrophenyloctyl ether (NPOE), high-molecular-weight poly(vinyl chloride) (PVC), poly(vinyl chloride) carboxylate (PVC-COO-), and tetrahydrofuran (THF) were purchased in Selectophore or puriss
EM ) RT -RT + RT x1 + KQST + 4KRT (QST + XT ) LT - XT ln F - (QS + R-) L - RT T T T -XT + XT x1 + KQST + 4KXT
(17) The effect of a high concentration of added inert electrolyte is expected to be 2-fold. First, it will provide a mutual, bulky counterion to the ion exchangers in both membrane segments. This should make the ion pair formation constants between each ion exchanger and its counterion more similar than without added electrolyte. Second, it should give a Nernstian response slope as a function of the unknown impurity concentration, XT , independent of the magnitude of the ion pair formation constant. Indeed, for QST . RT , QST . XT , and KQST . 1, eq 17 readily simplifies to RT RT LT - XT EM ) ln F X L - RT
T
T
4264 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001
quality from Fluka Chemical Corp. (Milwaukee, WI). Ionophores and all salts, including tridodecylmethylammonium tetrakis(4chlorophenyl)borate (ETH 500), were purchased from Fluka. Aqueous solutions were prepared by dissolving the appropriate salts in Nanopure purified water. Electrode Setup and EMF Measurements. Ion-selective electrode membranes were cast by weighing out the ionophore (20 mmol/kg, if used), the inert lipophilic salt ETH 500 (5 or 10 mmol/kg, if used), and NaTFPB (5 mmol/kg, if used), together with PVC or PVC-COO- and the plasticizer DOS or o-NPOE (1:2 by weight) to give a total cocktail mass of 140 mg, and by dissolving the mixture in 1.5 mL of THF and pouring it into a glass ring (2.2 cm i.d.) affixed with rubber bands onto a glass microscope slide. The solvent THF was allowed to evaporate overnight. The parent membranes were then removed from the glass and conditioned overnight in 0.01 M KCl solutions. A series of 6-mm-diameter disks were then cut with a cork borer from the parent membrane. For each measurement, one disk was incorporated into a single chosen IS-561 Philips electrode body (Moeller, Zurich, Switzerland). The body was carefully selected to give a potential value that deviated less than 2 mV for membrane disks with and without ionophore, but was conditioned in the same electrolyte. All membrane electrode potential measurements were performed at laboratory ambient temperature in unstirred salt solutions (identical to the conditioning and inner filling solution) versus a Ag/AgCl reference electrode with a 1 M LiOAc bridge electrolyte. The sandwich membrane was made by fusing two individual membranes together with pressure from a metal spatula immediately after blotting them individually dry with tissue paper. The obtained sandwich membrane was visibly checked for air bubbles before mounting in the same electrode body with the NaTFPB-free segment facing the sample solution. The potential was recorded as the average of the last minute of measurement of a 5-min measurement period. The elapsed time between sandwich fusion and exposure to electrolyte was normally less than 1 min. The potential of such sandwich membranes remains free from diffusion-induced potential drifts for ∼20 min.17 Standard deviations were obtained based on the measurements of sets of at least three replicate membranes disks that were made from the same parent membrane. RESULTS AND DISCUSSION The nature of anionic impurities in poly(vinyl chloride) has been studied spectroscopically14 and more recently by time-offlight mass spectrometry.20 Today, mostly sulfate groups are considered to act as anionic sites in such membranes. The concentration of anionic impurities has also been determined with radiotracers21 and by atomic absorption.15 It was claimed that not all anionic species are accessible for ion-exchange purposes. Recently, Nagele and Pretsch measured the impurities on the basis of selectivity changes of ion-selective membranes.7 The concentration of the anionic impurities in the PVC membranes was determined to be 0.063 ( 0.0016 mmol/kg. The segmented sandwich membrane technique was reported to be a simple method to investigate the binding properties in (20) Ye, Q.; Horvai, G.; Toth, A.; Bertoti, I.; Botreau, M.; Duc, T. M. Anal. Chem. 1998, 70, 4241. (21) Thoma, A. P.; Viviani-Nauer, A.; Arvanitis, S.; Morf, W. E.; Simon, W. Anal. Chem. 1977, 49, 1567.
the membrane.17,18 A concentration-polarized membrane is fabricated by fusing together two different membrane segments. Because the membrane potential is measured immediately after fusion, it is assumed that the inner- and outer-phase boundaries of this membrane are essentially identical to the initial separate membranes. This method has been used mainly to determine ionophore stability constants in the membrane by using one segment with and one without ionophore but with otherwise identical compositions.18 Studies were also performed by altering the concentration of anionic sites in one segment only.17 It was found that PVC-DOS membranes show half-Nernstian slopes. This was explained by the strong extent of ion pair formation in the nonpolar PVC-DOS environment, which indeed would induce half-Nernstian response slopes of the membrane. This slope becomes Nernstian again if an excess of a neutral lipophilic salt is present in the membrane. On the basis of these early experiments, the suitability of the segmented sandwich membrane technique to determine the effective concentration of anionic impurities in PVC membranes was evaluated. Initial trials focused on performing experiments with one segment containing a known concentration of a lipophilic cation exchanger and the other containing a blank membrane. Unfortunately, such experiments were unreliable, because potential signals were neither stable nor reproducible (data not shown). This is perhaps due to the leaching of these impurities into the sample solution.22 To circumvent this potential problem, both membrane segments were doped with a potassium-selective ionophore. It is known that the addition of an ionophore renders the cation exchanger more lipophilic because of the drastic concentration decrease of the counterion (in this case, potassium) that shifts the partitioning equilibrium in direction of the organic phase.22 On the other hand, the ionophore must not lead to substantial coextraction under the chosen experimental conditions. It is well-established that the extent of coextraction of sample electrolyte increases with increasing stability constant of the ionophore and decreasing concentration of ion exchanger in the membrane.2 For this reason, weakly binding potassium ionophores18 (see Figure 2) were chosen in this study. According to recently published work,18 BME-44 has logarithmic stability constants of 7.84 and 10.04 in PVC-DOS and PVC-NPOE, respectively, and K+-2 shows corresponding values of 6.88 and 10.22. Both ionophores are assumed to form simple 1:1 complexes with potassium. The values indicate a roughly 2 orders of magnitude weaker complexation of these ionophores than valinomycin.18 The risk of biasing the experiments by coextraction processes is, therefore, reduced. This was independently evaluated here by recording potentiometric response functions for BME-44 membranes without added lipophilic ion exchanger. A Nernstian response slope was, indeed, observed for PVC-DOS and PVCNPOE membranes for sample concentrations up to 0.1 M KCl, confirming that coextraction of electrolyte is negligible here. It appears to be most challenging to assess the concentration of anionic impurities in nonpolar PVC-DOS membranes, in which ion pairing is predominant. Sandwich membranes were constructed in which only one segment contained an added 5 mmol/ kg of cation exchanger NaTFPB. Otherwise, both compositions of the segments were identical and contained a potassium-selective (22) Bakker, E.; Pretsch, E. Anal. Chim. Acta 1995, 309, 7.
Analytical Chemistry, Vol. 73, No. 17, September 1, 2001
4265
Figure 2. Chemical structures of the two electrically neutral potassium ionophores used to evaluate the concentration of ionic impurities in ion-selective electrode membranes. Table 1. Experimental Membrane Potentials and Corresponding Concentrations of Anionic Impurities XT Determined by the Segmented Sandwich Membrane Technique ionophore
ETH 500, mmol/kg
∆EMF, mV
BME-44 BME-44 BME-44 K+-2 K+-2
Plasticizer: DOS 0.00 98 ( 4 10.0 (both sides) 106 ( 2 5.0 (one side)a 146 ( 2 0.00 109 ( 1 5.0 (one side) 150 ( 5
BME-44 BME-44 BME-44 K+-2 K+-2
Plasticizer: NPOE 0.00 100 ( 1 10.0 (both sides) 96 ( 2 5.0 (one side) 98 ( 1 0.00 111 ( 1 5.0 (one side) 112 ( 2
a
XT , µmol/kg
103 ( 4
130 ( 3 152 ( 7 141 ( 3 84 ( 2 81 ( 4
In the membrane segment without added ion exchanger RT.
ionophore. The results of these experiments are shown in Table 1. According to theory (see Figure 1), it is very difficult to predict the exact relationship between observed membrane potential and concentration of ionic impurities if ion pairing is dominant. To evaluate this issue qualitatively, a number of experiments with and without the inert lipophilic electrolyte ETH 500 were performed. The addition of ETH 500 to the NaTFPB-free side shows a very significant effect on the observed membrane potential, increasing it by ∼50 mV. This was observed with each of the two potassium ionophores. This suggests that ETH 500 strongly affects the extent of ion association in these nonpolar membranes. On the other hand, theory predicts that the addition of inert electrolyte ETH 500 to both membrane segments levels the different extents of ion dissociation in the membrane, and a Nernstian relationship between membrane potential and ion-exchanger concentration is expected (see Figure 1 bottom and eq 18). Table 1, row 2, shows the data for the experiment with both segments containing ETH 500. This data set appears to be more reliable than that without added ETH 500 and was used to calculate the concentration of ionic impurities (see Table 1). The notion that the addition of an inert lipophilic salt such as ETH 500 induces near-Nernstian response slopes as a function of the concentration of added ion exchanger was directly tested experimentally. Note that earlier experiments were performed on ionophore-free membranes, with similar results.17 Figure 3 shows 4266 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001
Figure 3. Experimental PVC-DOS sandwich membrane potential as a function of added lipophilic cation exchanger NaTFPB in one membrane side. A Nernstian response slope is expected, since both segments contain an excess of inert lipophilic electrolyte ETH 500.
determined membrane potentials for segmented sandwich experiments in which one side contains the ionophore K+-2, ETH 500, and varying amounts of NaTFPB, and the other side contains equivalent concentrations of K+-2 and ETH 500 but no added ion exchanger. The figure indicates a 68 mV slope for this experiment in PVC-DOS membranes, with standard deviations (n ) 3) shown as error bars. Clearly, the slope for this experiment is quite close to Nernstian, and it appears that the theory (eq 18) is adequate for the purpose of predicting concentration changes in the membrane in the presence of an inert lipophilic electrolyte. On the basis of the experiments described above, one would suggest that the more polar PVC-NPOE is a simpler matrix to assess the concentration of anionic impurities. Indeed, if all salts behave as if they were fully dissociated, a segmented sandwich experiment may always be analyzed on the basis of eq 18, which utilizes a Nernstian electrode slope. The addition of a neutral lipophilic salt should then ideally have no influence on the assessed concentration of anionic impurities, because it would not dramatically alter the concentration of the already dissociated ions (see Figure 1). Table 1 shows observed membrane potentials for segmented PVC-NPOE membranes with one side containing ionophore only and the other, ionophore and NaTFPB. The same experiments were performed with the neutral lipophilic electrolyte ETH 500 in just the NaTFPB-free or all of the membrane
segments. As shown in Table 1, the membrane potentials that were found are nearly identical to the corresponding experiments without added electrolyte, which stands in strong contrast to the experiments performed with PVC-DOS. This suggests that the assumptions used here are appropriate for PVC-NPOE. The potentials are equal within ∼11 mV for both ionophore-containing systems, which translates into effective impurity concentrations of ca. 0.084 and 0.130 mmol/kg. Because standard deviations were on the order of 2 mV for both sets of experiments, the difference indicates a systematic deviation. It could mean that the ionophores themselves are not perfectly pure and contribute to the concentration of anionic impurities found in the membranes, because potential differences were also found with PVC-DOS (see Table 1). The impurity concentrations determined here are within the range reported earlier.7 The value found in PVC-NPOE is also very similar to that found in PVC-DOS. It suggests that the protocol introduced here yields appropriate concentration values, even if the polarity of the membrane changes significantly. It also indicates that most of the impurities found in plasticized PVC do originate in the polymer, not in the plasticizer.23 An additional source of error might be any undesired ion-exchanger properties of the added lipophilic salt. The values shown for the measurement with NPOE (Table 1), in which the concentration of ETH 500 was systematically increased, are however, not significantly different. These data indicate that this possible issue is not too problematic here. As a final example of this technique, sandwich membrane experiments were performed with carboxylated PVC plasticized with NPOE. Originally, carboxylated PVC was synthesized in order to improve the adhesion properties of the membranes on electrode surfaces.24 It was reported that the extent of dissociation of these COOH groups varies according to the exact membrane composition.24 With the technique outlined here, the anionic (23) Watanabe, M.; Toko, K.; Sato, K.; Kina, K.; Takahashi, Y.; Iiyama, S. Sens. Mater. 1998, 10, 103. (24) Cosofret, V. V.; Buck, R. P.; Erdosy, M. Anal. Chem. 1994, 66, 3592.
impurities in PVC-COOH was determined with the ionophore K+-2 as 1.62 ( 0.09 mmol/kg (∆EMF, 36 ( 2 mV). The membrane was contacted with 0.01 M KCl at neutral pH and contained no added ETH 500, because the experiment was performed in NPOE as the plasticizer. The results support earlier notions24 that the concentration of the anionic sites in PVC-COOH is much higher than in unmodified PVC. It should be kept in mind that these carboxylic groups may deprotonate to various extents as a function of the membrane and sample composition on the basis of dissociation and ion extraction equilibria. Different experimental conditions will, therefore, likely yield different apparent ionexchanger values. CONCLUSIONS The anionic impurities in PVC membranes can be conveniently measured by the sandwich membrane technique. The concentrations found agree very well with earlier findings with other techniques. This technique gives insight on the extent of dissociation of the ion exchanger in the membrane, since it is effectively a measure of the free ion activity in the membrane. The addition of the inert lipophilic salt ETH 500 to both membrane segments levels the extent of dissociation by providing a large concentration of a common counterion to both membrane segments. With this protocol, the effective concentration of ionic impurities in PVC membranes can be adequately assessed. No significant difference was found by altering the polarity of the plasticizer. ACKNOWLEDGMENT The authors thank the National Institutes of Health (GM58589 and GM59716) and ACS-PRF for financial support of this work.
Received for review April 10, 2001. Accepted June 29, 2001. AC0104126
Analytical Chemistry, Vol. 73, No. 17, September 1, 2001
4267
