Talanta 63 (2004) 139–148

Glass and polymeric membrane electrodes for the measurement

of pH in milk and cheese
Praveen Upreti a , Lloyd E. Metzger a , Philippe Bühlmann b,∗
a Department of Food Science and Nutrition, University of Minnesota, 1334 Eckles Avenue, St. Paul, MN 55108, USA
b Department of Chemistry, University of Minnesota, 207 Pleasant St. SE, Minneapolis, MN 55455, USA

Received 12 September 2003; received in revised form 5 December 2003; accepted 10 December 2003

Abstract

While there is a considerable interest in the food industry in determining various analytes using ion-selective electrodes (ISEs), only few
reports describe their use for direct measurements in food. In this study, the suitability of glass electrodes and ionophore-based solvent
polymeric ISEs for the determination of pH in Process cheese, Cheddar cheese and milk was investigated. The liquid junction potential
between a 3 M KCl bridge electrolyte and diluted as well as undiluted Process cheese was found to be negligible. Reference electrodes with
ceramic plug and sleeve-type junctions performed well, although precautions needed to be taken to prevent plugging at the junctions. While
the protein rennet casein posed no problems in pH measurements, the extraction of neutral lipophilic compounds or hydrophobic peptides into
solvent polymeric membranes was evident, resulting in some loss of selectivity for monovalent cations upon exposure to cheese. However,
it was found that ISEs based on tridodecylamine (R3 N) as ionophore and o-nitrophenyl octyl ether (oNPOE) as plasticizer can be used to
accurately measure the pH of milk and, after desensitization of the electrodes in a cheese emulsion, of diluted Process cheese. Since pH
measurements with a glass electrode showed that emulsions of cheese moderately diluted to a cheese content of 70% have the same pH as
undiluted cheeses, it is possible to determine the pH in cheese with ionophore-based ISEs. R3 N membranes also performed well in undiluted
milk.
© 2004 Elsevier B.V. All rights reserved.

Keywords: pH measurement; Ionophore; Ion-selective electrode; Cheese; Milk; Dairy products; Glass electrode

1. Introduction At the outset of this study, it appeared very likely that

Various polymer membrane-based ion-selective elec-
trodes (ISEs) have been developed [1–3]. The membranes
of the majority of these ISEs contain lipophilic complex-
ing agents that selectively bind to their target ions. Such
complexing agents are referred to as ionophores or carriers
[4]. The large number of ionophore-based ISEs developed
over the last 37 years made it possible to use ISEs in var-
ious types of routine analyses [3,4]. We are interested in
exploring applications of ionophore-based ISEs to on-line
measurements in food processing and the direct analysis of
food products in the laboratory, since the mineral compo-
sition is used in nutritional labeling of final food products
and in controlling food manufacturing processes.
∗ Corresponding author. Tel.: +1-612-624-1431;
fax: +1-612-626-7541.
E-mail address: buhlmann@chem.umn.edu (P. Bühlmann).
lipophilic food components from food samples might be ex-
tracted into polymer membranes ISEs and severely affect
measurements, as was similarly reported for urine analy-
sis [5]. However, there was no published literature regard-
ing this possibility and its implications on the performance
of polymer membrane ISE for applications in the dairy
industry. Therefore, the severity of the possible problem
had to be assessed. Presumably, problems encountered in
the direct analysis of milk and cheese using any polymer
membrane-based ISEs are similar. We chose to measure the
activity of the hydrogen ion (pH) rather than that of another
ion since this parameter can be measured independently
with the glass electrode. Methods to measure the activity of
many other ions—rather than the concentration of free ions
or their total concentration—are not available as readily. In
this paper, we describe the application of polymer-based
and glass pH electrodes for the analysis of pH in cheese
and milk.

0039-9140/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.talanta.2003.12.020
140 P. Upreti et al. / Talanta 63 (2004) 139–148
The pH is an important control parameter for natural as
well as Process cheese manufacturers. Natural cheese, such
as Cheddar cheese, is manufactured by concentrating the
lipid and protein portion of milk using acids and/or coagulat-
ing enzymes and then ripening it for various lengths of time
to get the desired flavor and texture. The pH is measured dur-
ing cheese manufacturing because it is an indicator of acid
development by lactic acid bacteria added to the milk. With
the advent of mechanized, automated cheese-manufacturing
systems, on-line measurements of pH are desired. The pH
of natural cheese affects proteolysis, growth of non-starter
lactic acid bacteria, the amount of calcium bound to pro-
teins, all of which influence the flavor and texture of the final
product [6,7]. Process cheese, on the other hand, is manu-
factured by blending natural cheeses with emulsifying salts
(disodium phosphate, trisodium citrate) and heating to pro-
duce an emulsion. The pH of Process cheese can affect its
functional properties, such as its viscosity, meltability, hard-
ness, and stretchability. For instance, an increase in pH from
5.2 to 6.7 decreases the viscosity and meltability of Process
cheese [8].
Glass electrodes are commonly used in the dairy laborato-
ries for pH measurements. The majority of these electrodes
are combination electrodes, i.e. the reference and measur-
ing electrodes are combined in a single body. Their sensing
membranes consist of glass and make them highly selective
to H+ . However, they are brittle, which is a major concern
for on-line measurements. Fouling of glass electrodes due
to protein and lipid adsorption on the electrode surface has
been reported [9,10] and proper removal of adsorbed pro-
teins has been recommended [10,11]. Importantly, only little
information about the general guidelines of proper handling
of glass electrodes to measure pH in cheese has been pub-
lished [10]. Moreover, issues related to accuracy, precision,
and response time specific to cheese measurements have not
been reported.
Recently, non-glass electrodes have also found applica-
tion in the food industry. In general, these electrodes have
a pH-sensitive field-effect transistor (pH-FET) housed in an
epoxy or other plastic body, making them less brittle than
glass electrodes. Since the surface of a pH-FET is hydrated
silicon nitride, and is, therefore, very similar surface as the
glass electrode, pH-FETs and glass electrodes can be ex-
pected to exhibit very similar properties when used for mea-
surements in cheese. The use of quinhydrone electrodes has
also been suggested for measuring pH in cheese, especially
in cheese varieties that are low in moisture [10]. Because
their use involves lengthy sample preparation and is not
based on a direct measurement, their usage in dairy labora-
tories is very limited.
Polymeric membrane ISEs were also used for food analy-
sis. Their application to measure sodium, calcium, chloride,
fluoride, iodide, and nitrate in foods was described [12,13].
In particular, they were used to measure nitrate [14], chloride
[15], and fluoride [16] in cheeses. The sample preparation
typically involved a high dilution of cheese (20–50 times)
followed by a series of further sample preparation steps.
It appears that the cheese samples were diluted to avoid
problems with fouling of electrodes and drifts during mea-
surements [14,17,18], but few details on actual difficulties
encountered with cheese samples were reported. Instead,
the literature focuses on the successful sample preparation.
For example, Kaneda et al. [14] attributed the problems
faced in measurements of the nitrate activity in cheese to
the high sample concentrations of lipids, proteins and other
interfering ions such as, halides, inorganic anions, bicarbon-
ates, and organic carboxylates. To overcome complications
from such interferents, they subjected samples to dilution,
homogenization, heating and cooling (to melt and solidify
the lipids, facilitating their elimination by separation), cen-
trifugation (to separate lipids and proteins), treatment with
silver resin (to remove chloride ions) and aluminum resin
(to remove proteins and citrates), and filtration. To the best
of our knowledge, reports on direct measurements of H+ or
any other ions in cheeses using polymeric membranes have
not been reported.
The importance of pH for cheese makers, the problems
encountered during measurements, and the scarcity of rele-
vant literature made a systematic study of pH measurement
with polymer-based ISEs and glass electrodes desirable. We
chose Process and Cheddar cheese as representative cheeses,
and two H+ ionophores, tridodecylamine (R3 N) and octade-
cyl isonicotinate (ETH 1778), as representative ionophores.
R3 N has an extremely high selectivity for H+ but is not suit-
able for measurements at low pH, where Donnan failure may
occur [19]. In contrast, the less basic ETH 1778 has a lower
H+ -selectivity but is suitable for measurements in acidic so-
lutions [3,20–28]. Because the response characteristics of
polymeric membrane ISEs depend also upon the plasticizers
and membrane matrices [29], two plasticizers (o-nitrophenyl
octyl ether, oNPOE, and bis(2-ethylhexyl) sebacate, DOS)
and two membrane matrices (poly[vinyl chloride] and sil-
icone rubber) were evaluated. Simultaneous measurements
were made using polymer-based and glass electrodes, and
then compared with one another.
2. Experimental
2.1. Reagents
Reagents of the highest commercially available grade
were used. Deionized and charcoal-treated water (18.2 M-
cm specific resistance) obtained with a Milli-Q PLUS
reagent-grade water system (Millipore, Bedford, MA, USA)
was used for all sample solutions. Potassium tetrakis(p-
chlorophenyl)borate (KTpClPB), sodium tetrakis[3,5-bis(tri-
fluoromethyl)phenyl]borate (NaTFPB), and o-nitrophenyl
octyl ether (oNPOE) were purchased from Dojindo Lab-
oratories (Kumamoto, Japan); bis(2-ethylhexyl) seba-
cate (DOS) and poly(vinyl chloride) (PVC) from Fluka
(Buchs, Switzerland); tridodecylamine (R3 N) and octadecyl
 P. Upreti et al. / Talanta 63 (2004) 139–148
isonicotinate (ETH 1778) from Sigma-Aldrich (St. Louis,
MO, USA). The Dow Corning® 3140 (silicone rubber)
sealants was a gift from Dow Corning (Midland, MI, USA).
2.2. ISE membranes
Five types of polymeric membranes were evaluated. They
will be referred hereafter as R3 N/DOS, ETH 1778/oN-
POE, R3 N/oNPOE, ETH 1778/DOS, and silicone rubber
membranes. ETH 1778/oNPOE and ETH 1778/DOS mem-
branes consisted of 3.00 mg ETH 1778, 3.00 mg KTpClPB
(76 mol% relative to the ionophore), 100 mg PVC, and
200 mg plasticizer (oNPOE or DOS). Silicone rubber mem-
brane contained 20 mg R3 N, 6.00 mg NaTFPB (18 mol%
relative to the ionophore), 40 mg DOS, and 334 mg com-
mercial silicone rubber sealant, which consisted of >60%
hydroxy-terminated dimethyl siloxane, 10–30% trimethy-
lated silica, and 5–10% methyltrimethoxysilane. Curing
of silicone membranes occurs at room temperature upon
exposure to humidity.
Circles of ∼7 mm diameter were cut from the prepared
master membranes and mounted on Phillips-type electrode
bodies (Glasbläserei Möller, Zürich, Switzerland) equipped
with tubular glass assemblies incorporating an internal
Ag/AgCl reference electrode (Möller).
2.3. emf measurements
Double-junction sleeve-type Ag/AgCl reference elec-
trodes (Mettler Toledo, Switzerland) and ion meters (Law-
son Labs Inc., Malvern, PA, USA) were used for all emf
measurements. The cell assembly for the potentiometric
measurements was as follows:
Ag|AgCl|3 M KCl, AgCl sat.||outer filling solution||
sample|membrane|inner filling solution|AgCl|Ag
The outer filling solution for the reference electrode was
3 M KCl (unless noted otherwise). Membranes with R3 N as
ionophore were used with an inner filling solution contain-
ing 1 mM NaCl, 10 mM Na2 HPO4 , and 10 mM NaH2 PO4 ,
and were conditioned in water for at least 24 h. About 48 h
conditioning was required for silicone rubber membranes in
order to obtain Nernstian responses. Because of the limited
sodium ion discrimination provided by ETH 1778, mem-
branes with this ionophore were used in combination with an
inner filling solution containing 1 mM LiH2 PO4 and 1 mM
NaCl, and were conditioned in 10 mM LiH2 PO4 (pH 5.4).
Response curves of ionophore-based ISEs were obtained
by measuring the emf with ionophore-based ISEs and cali-
brated glass electrodes in water to which aliquots of 1 mM
H2 SO4 were added. The response curves shown in Fig. 7
are corrected for the difference of the liquid junction poten-
tials at the reference electrode in calibration and in sulfate
solutions (≤1 mV) using the Henderson approximation [30].
The response slopes (mV/decade ± 95% confidence limits)
141
for the membranes were as follows: R3 N/DOS, 56.0 (±1.0);
ETH 1778/oNPOE, 53.3 (±2.3); R3 N/oNPOE, 56.6 (±1.1);
ETH 1778/DOS, 52.9 (±2.6); silicone rubber, 50.6 (±3.2).
The pH half cell glass electrodes were of the Inlab® type
(Mettler-Toledo Process Analytical Inc., Wilmington, MA,
USA). A commercial combination type glass electrode (Type
6157, Denki Kagaku Keiki, Tokyo, Japan) was used only
for comparison for the experiment described in Section 3.1.
The glass electrodes were calibrated with pH 7.00 and pH
4.00 calibrating buffer (Ricca Chemical, Arlington, TX, or
MCB Manufacturing Chemicals, Gibbstown, NJ), certified
traceable to NIST standard reference material, and exhibited
response slopes of 57.5 ± 0.5 mV per decade. Glass elec-
trodes were thoroughly cleaned with water and 95% ethanol
between measurements in cheese samples. The emf values
obtained with cheese samples were corrected for the differ-
ence in the liquid junction potential between the reference
electrode and cheese samples (as determined in this study),
and the liquid junction between the reference electrode and
calibration buffers, as calculated with the Henderson approx-
imation [30]. The correction corresponded to a difference
in pH of 0.02. Measurements with cells comprising two pH
half cell glass electrodes in series suffered from excessive
noise pickup unless the connection between one pH half cell
and the input to the voltage follower op-amp was tied to
ground, either directly or through a resistor.
Resistance measurements were performed using the
method of potential reduction by a known shunt [31,32].
Selectivity coefficients were determined using the fixed in-
terference method (FIM) [33–35]. Activity coefficients were
calculated according to a two-parameter Debye–Hückel
approximation [36].
2.4. Preparation of rennet casein, lipids, and cheese
suspensions and emulsions
Process cheese has the unique property of forming stable
emulsions in water and producing macroscopically homoge-
nous mixtures. In contrast, blending and diluting of Cheddar
cheese produces suspensions and not emulsions. All suspen-
sions and emulsions were prepared by mixing appropriate
amounts of cheese and water using an electric blender. In the
case of Process cheese, the suspension was then heated in
a microwave oven to ∼37–39 ◦ C to form an emulsion. The
mixture was left to equilibrate at room temperature prior to
measurements.
A mixture of water and edible rennet casein (30 mesh;
New Zealand Dairy Board, Wellington, NZ) was used as
a representation of proteins in cheese. Rennet casein is
not completely soluble, but stirring during measurements
kept it suspended. Lipids were extracted from Process
cheese by suspending shredded Process cheese in petroleum
ether:diethyl ether (50:50, v/v) and then decanting the or-
ganic phase with the dissolved lipids into an aluminum pan
and evaporating the solvents.
142 P. Upreti et al. / Talanta 63 (2004) 139–148

-79 sleeve-type reference electrodes with 1.5, 2.0, or 3.0 M KCl

as outer filling solution were placed in a sample of diluted
-79.25 Process cheese and the emf was measured over 2 h (to ex-
emf (mV)

clude the possibility of a slow response or memory effects).

-79.5 Between KCl solutions of different concentrations, the ref-
erence electrodes were cleaned, refilled with the respective
-79.75 concentrations of KCl (1.5, 2, or 3 M), and left for equili-
bration in the solutions of the corresponding KCl concentra-
-80 tions for a minimum of 3 h. After the cheese experiments,
0 0.5 1 1.5 2 2.5 all the reference electrodes were cleaned, refilled with 3 M
(a) Csat/C KCl and equilibrated overnight. These electrodes were then
placed in a 3 M KCl solution and the emf was determined to
1.5
account for (minor) electrode-to-electrode differences. The
1 potentials measured in cheese samples by the three refer-
ence electrodes were adjusted for the electrode-to-electrode
0.5 differences, and the corrected values were used to deter-
emf (mV)

mine liquid junction potentials (Fig. 1b). Our results show

0
-0.5
-1
-1.5
0 20 40 60 80 100
(b) % Process cheese (w/w)
Fig. 1. (a) Liquid junction potential for 100% Process cheese as determined
by the method of Bjerrum. (b) Liquid junction potentials at different
dilutions of Process cheese with respective standard deviations represented
by vertical bars.
3. Results and discussion
3.1. Liquid junction potential
Part of the electrochemical cell, the reference electrode
is very important in determining the emf of the cell. The
liquid junction potential between the sample and the bridge
electrolyte of the reference electrode can be affected by the
concentration of the chemical species in the sample solu-
tion, its pH, viscosity, and the presence of suspended parti-
cles [12]. Recognizing that liquid junction potentials might
be a possible source of systematic errors in direct measure-
ments of pH in cheese, we determined the liquid junction
potential between the reference electrode and Process cheese
samples using a method suggested by Bjerrum [37]. This
method is based on the assumption that the liquid junction
potential is inversely proportional to the concentration of the
KCl solution in the salt bridge contacting the sample and,
therefore, vanishes at infinite concentration. Thus, the liquid
junction potential can be determined by extrapolation of the
experimentally observed potentials for finite concentrations
of KCl to infinite concentration. This approach is illustrated
in Fig. 1a for undiluted Process cheese as the sample.
In order to determine the value of the liquid junction po-
tential at different levels of dilution of Process cheese (ratios
cheese:water from 20:80 to 100:0), three double-junction
that there is no significant liquid junction potential at any
level of dilution of Process cheese. The averages of the ex-
perimentally determined liquid junction potentials are all so
close to 0 mV that the error bars in Fig. 1b may appear com-
paratively large. However, since the typical error of a pH
measurement in a routine laboratory is of the order of 0.02
pH units, which corresponds to 1.2 mV, the errors depicted
in Fig. 1b are indeed well within an acceptable range.
Liquid junctions for the reference electrodes can be of dif-
ferent kinds (for example, ceramic or fiber junctions, ground
glass sleeves, double junctions) [12,37]. Because of lower
manufacturing costs, resistance to chemical degradation, and
ease in incorporation, the most commonly used junctions
in commercial combination pH electrodes are porous ce-
ramic junctions. The ceramic plug typically has a diame-
ter of ∼1 mm and is fused into the shaft of a combination
electrode. The outer filling solution can continuously flow
through that plug, for example at 0.01 ml day−1 mm−2 of
cross-section [12,37]. However, ceramic plug junctions are
often prone to blocking [12,37], affecting the outflow of
outer filling solution and, in turn, the liquid junction poten-
tial. Therefore, the performance of a combination electrode
with a ceramic plug junction was compared with that of a
sleeve-type junction electrode by monitoring the pH of 20%
Process cheese (w/w) over 20 h (Fig. 2). For this purpose, the
pH sensitive glass membrane of a commercial combination
electrode was bypassed by identifying the relevant leads and
measuring the emf of a pH half cell glass electrode relative
to the reference half cell of this combination electrode. Con-
currently, the emf of the pH half cell was also measured rel-
ative to a sleeve-type reference electrode and relative to the
pH sensitive half cell of the combination electrode with the
glass membrane. Interestingly, Fig. 2 shows no significant
difference in the performance of the ceramic plug junction as
compared to the sleeve-type junction. The somewhat larger
amount of noise for the ceramic plug junction is likely due
to the higher resistance of this junction, possibly increased
by partial clogging. Therefore, precautions should still be
taken to avoid blocking of ceramic plug junctions. Notably,
 P. Upreti et al. / Talanta 63 (2004) 139–148
80
60
40
emf (mV)
20
0
-20
-40
-60
-80
0 5 10 15
Time (h)
Fig. 2. Response in 20% Cheddar cheese as measured with a pH half cell
glass electrode relative to reference electrodes with a sleeve-type liquid
junction (1) and a ceramic-plug type liquid junction of a combination
electrode (2). The response of a pH half cell relative to the pH half cell
of the combination electrode is also shown (3).
Fig. 2 shows that both the phase boundary potential at the
two glass membranes and the liquid junction potential at the
reference electrode remain amazingly stable over a period
as long as 20 h. This is particularly remarkable when con-
sidering that 20% cheese emulsions contain such high levels
of proteins and lipids that they totally lacks transparency.
3.2. Measurements in cheese using polymeric
membrane electrodes
The results described above demonstrate that with proper
handling there are no drifts of the liquid junction poten-
tial due to cheese, and that the liquid junction potential is
negligible. Therefore, experiments were started with poly-
meric membrane electrodes containing the ionophores R3 N
or ETH 1778. Membranes with R3 N and DOS as plasti-
cizer as well as membranes with ETH 1778 and oNPOE as
plasticizer were reported previously for pH measurements
[20–28]. ISEs with R3 N/DOS membranes exhibit Nerns-
tian responses and an excellent selectivity in the pH range
of 4.5–11.0. For measurements in blood serum, their per-
formance was found comparable to that of glass electrodes
[9,38–40]. On the other hand, ISEs with ETH 1778/oNPOE
membranes were used under acidic conditions (such as in
the gastrointestinal tract) and were recommended for the
pH range from 8 to 0 [22]. Therefore, we expected that the
ETH 1778/oNPOE ISEs might be better suited for cheese
measurements (pH 4.8–5.4) than R3 N/DOS ISEs. However,
since it is well known that not only the ionophore but also
the type of plasticizer affects the membrane performance
[29] and, in particular, that oNPOE has been reported to
provide a wider measuring range than DOS [29], it became
necessary to distinguish the effects of ionophore and plas-
ticizer. Therefore, ETH 1778/DOS and R3 N/oNPOE mem-
branes were also prepared.
In order to investigate if any components of cheese in-
terfere with pH measurements performed using ionophore-
143
1
emf (ionophore) - emf (glass)
2
2
3
3
1 4
50 mV
20
0 100 200 300 400 500 600
Time (s)
Fig. 3. Addition of 20% Process cheese aliquots (indicated by arrows)
to a phosphate buffer of the same pH: to correct for changes in the pH,
the emf as observed with a glass electrode was subtracted from the emf
as measured with the membrane electrodes. (1) ETH 1778/oNPOE; (2)
ETH 1778/DOS; (3) R3 N/oNPOE; (4) R3 N/DOS. Responses are separated
vertically to enhance clarity. (Outer filling solution of reference electrode:
1 M lithium acetate.)
based ISEs, the pH of 20% cheese emulsions (w/w) was
measured using a pre-calibrated glass electrode. A 10 mM
lithium phosphate buffer with the same pH (e.g. 5.42) was
then prepared. A reference electrode, half-cell glass elec-
trode, and two ionophore-based ISEs (R3 N/DOS and ETH
1778/oNPOE) were placed into this buffer, and aliquots
of a 20% cheese emulsion (w/w) were added. Despite the
matching of the pH of the added cheese and the buffer
solutions, some changes of the pH (up to 5–30 mV for the
highest cheese concentrations) were found upon addition of
the cheese, possibly related to the release of calcium ions
from casein and subsequent binding to buffer phosphate. To
account for this effect, the change in the emf as observed
with the glass electrode was subtracted from the emf values
measured with the ionophore-based ISEs.
Fig. 3 shows the thus corrected potential versus time pro-
files, each arrow indicating the addition of a further aliquot
of a 20% emulsion of Process cheese. Clearly, the R3 N/DOS
and R3 N/oNPOE membranes did not respond to any inter-
ferents that may be present in cheese. In contrast, the re-
sponse curves for ETH 1778/DOS and ETH 1778/oNPOE
membranes showed a distinct step for every aliquot of added
cheese emulsion. After addition of 100 ml of cheese emul-
sion, the corrected emf had risen by about 50 mV. Since
changes in pH are already taken into account by subtraction
of the emf as measured with a glass electrode, these steps
correspond to responses to some type of interferent.
Interestingly, between the additions of individual aliquots,
the corrected emf of the ETH 1778-based membranes re-
mained stable. Because gradual extraction of interferents of
low concentration limited by mass transfer from the sample
into the sensing membrane would be expected to result in
slow responses, the responses observed here are indicative
of one or several interfering cations occurring in relatively
144 P. Upreti et al. / Talanta 63 (2004) 139–148
1
2 2
emf (mV)
3
50 mV
0 500 1000 1500 2000 2500 3000 3500
Time (s)
Fig. 4. Response of membrane electrodes in 5% rennet casein for 1 h. To
correct for changes in the pH, the emf as observed with a glass electrode
was subtracted from the emf as measured with the membrane electrodes.
(1) ETH 1778/DOS; (2) ETH 1778/oNPOE; (3) glass. Responses are
separated vertically to enhance clarity. (Outer filling solution of reference
electrode: 1 M lithium acetate.)
high concentrations (this interference possibly enhanced by
extraction of electrically neutral lipids or peptides into the
membranes).
To explore how representative Process cheese is for cheese
in general, we also tested Cheddar cheese. Since Cheddar
cheese forms suspensions but no emulsions, experiments
analogous to those for Process cheese were performed with
20% Cheddar cheese suspensions (w/w). Again, membranes
based on R3 N showed neither steps nor drifts in the cor-
rected potential upon addition of cheese aliquots, while the
responses of ETH 1778 membranes exhibited steps with the
addition of every aliquot. The steps added up to 12 mV after
addition of 100 ml of cheese suspension (data not shown).
Further experiments were conducted to explore the rea-
sons behind the interferences from cheese as observed with
ETH 1778 membranes. To explore whether the interferences
from cheese are a direct response to proteins or lipids (the
two main constituents of cheese), ISEs with ETH 1778 mem-
branes were inserted into lithium phosphate buffer of pH
5.4, and aliquots of 5% rennet casein or cheese lipids were
added. The electrodes showed no significant responses to
5% rennet casein and were free of any significant drift when
observed for 1 h (Fig. 4).
Rennet casein is the major fraction of protein in young (or
freshly manufactured) cheeses. While an interference from
rennet casein due to its extraction into membranes was not
apparent in our experiments, this does not preclude any effect
of proteins. Indeed, proteins are broken down into smaller
peptides during ripening of cheese. This protein breakdown
may produce lower molecular weight hydrophobic peptides
that may be extracted into polymer membranes and cause
interferences.
When aliquots of pure cheese lipids were added to the
phosphate buffer, there was typically no initial indication of
interference. However, after an apparently random amount
emf (ionophore) - emf (glass)
1
3
70 mV
0 100 200 300 400 500 600
Time (s)
Fig. 5. Noisy response of membrane electrodes on addition of aliquots
of cheese fat to lithium phosphate buffer (indicated by arrows), possibly
due to the resistive oil film on the surface of these electrodes. (1) ETH
1778/oNPOE; (2) glass; (3) ETH 1778/DOS. (Outer filling solution of
reference electrode: 1 M lithium acetate.)
of time, the responses became very noisy. Two typical re-
sponses are shown in Fig. 5. As can be seen from the re-
sponse curve of the glass electrode, clogging of the liquid
junction is not responsible for the high noise level. A pos-
sible explanation may be the formation of a resistive oil
film on the surface of these membrane electrodes, which
would be consistent with the observation that the noise level
returned to a much lower level after the electrodes were
washed with water. In addition, the noisy responses observed
on addition of aliquots of lipids were very different from
the low-noise stepwise responses that were observed when
aliquots of cheese were added.
Consequently, the interference from cheese as observed
with ETH 1778 membranes does not appear to be primarily
a response to the major protein rennet casein or a compo-
nent of the lipid fraction of cheese. Moreover, considering
the stepwise and quick nature of the response, as discussed
above, the experimental data suggest that the interference
arises from cations occurring in cheese in relatively high
concentrations. A possible candidate is Na+ , which is the
most common inorganic cation in Process cheese (∼3%
on a weight basis). Moreover, membranes based on ETH
1778 are well known to have a lower discrimination to Na+
pot
(KH+ ,Na+ −5.6) than membranes based on the ionophore
pot
R3 N (KH+ ,Na+ −10.4) [3,28]. However, calculations on
the basis of KH+ ,Na+ and the activity of Na+ in the sam-
pot
ples indicate that the observed responses to cheese are not
only from Na+ interference. Other inorganic and also or-
ganic cations present in cheese may contribute as well. The
possibility of extraction of hydrophobic peptides into the
membranes cannot be neglected.
Another ion that is present at high levels (∼0.03 M
or 1.5%, w/w) in cheese is lactate. This anion is formed
by the decomposition of lactose by lactic acid bacteria.
Co-ion interference could be a possible concern for the
 P. Upreti et al. / Talanta 63 (2004) 139–148 145

membrane-based electrodes [19], and hence experiments Table 2

were conducted to study the possibility of such interfer- Response times (t90% ) of membrane electrodes as compared to glass
electrodes
ence for our membranes. For this purpose, R3 N membrane
electrodes were placed in a 0.1 M sodium phosphate buffer Electrode Response time Response time (s)
(s) in buffer in 20% cheese
(pH = 6.0), and 200 ml of 1 M sodium lactate (pH = 7.0)
Glass 8.4 57.8
was added in several aliquots while recording the emf (data
R3 N/oNPOE 11.4 57.5
not shown). However, no emf response and, therefore, no R3 N/DOS 12.0 54.3
co-ion interference from lactate was observed.

3.3. Effect of cheese on selectivity of R3 N membranes

Because membranes with R3 N as ionophores appeared
suitable for measurements in diluted cheese irrespective of
the plasticizer, the effect of cheese on the selectivity and re-
sponse time of these ISEs was investigated. It was shown
previously that the exposure of ionophore-based membranes
to neutral lipids occurring in urine deteriorates the potentio-
metric selectivity due to extraction of those lipids into the
sensing membranes [5]. To determine if the selectivity for
R3 N membranes similarly deteriorates on prolonged expo-
sure to cheese, selectivities for monovalent cations (Na+ ,
K+ , and Li+ ) were measured before and after 48 h of expo-
sure of R3 N membranes to a 20 and 100% Process cheese
emulsion. As Table 1 shows, there were some changes in
the selectivities for these cations. In particular, the selec-
tivity coefficients for Na+ and K+ rose by up to two units
upon exposure to cheese. To explore if lipids are the cause
of this selectivity loss, membranes were exposed to a lipid
extract for ∼24 h. A similar decrease in selectivity to cations
as observed on exposure of electrodes to 100% cheese was
obtained. Moreover, unlike in any of the other long-term
exposure experiments, there was a significant increase in
the membrane resistance of up to an order of magnitude
upon exposure to cheese. For example, the resistance of a
R3 N/oNPOE membrane rose from 0.3 to 3.2 M.
These results show that the selectivity of R3 N membranes
is remarkably robust but not immune to exposure to cheese.
Continuous exposure of these membranes to cheese does
result in slight loss of selectivity, which appears to be due
to extraction of neutral compounds such as lipids or hy-
drophobic peptides into the sensing membranes (see also
Section 3.6).
3.4. Effect of cheese on response time of
R3 N membranes
The response time is another important characteristic of
ISEs and is often reported as t90% [41], i.e. the length of
time between the instant at which the polymer-based ISE
and a reference electrode are brought into contact with a
new sample solution and the instant at which the potential of
the cell has reached 90% of the final value. Response times,
t90% , for ISEs and glass electrode were compared by plac-
ing glass and membrane electrodes along with a reference
electrode in lithium phosphate buffer, and adding drops of
100 mM H2 SO4 in three discrete steps. In addition, response
times were also measured in 20% Process cheese solution.
No significant differences between the response times and
response behaviors of glass, R3 N/DOS and R3 N/oNPOE
membrane electrodes (Table 2) were observed. Apparently,
the response times were limited neither for the buffer sam-
ples nor the cheese emulsions by properties of the electrodes,
but rather by the time it took to change the pH in the bulk
of the samples.

Table 1
Change in selectivities of polymer membrane ISEs before and after exposure of membranes to Process cheese and lipids
Cation Initial After 48 h exposure to After exposure to lipids

20% Process cheese 100% Process cheese

(a) R3 N/DOS
pot
KH+ ,Na+ −11.25 −10.51 −9.56 −10.07
pot
KH+ ,K+ −10.70 −9.54 −8.86 n.d.
pot
KH+ ,Li+ −10.93 −10.67 −10.19 >−7.0

(b) R3 N/oNPOE
pot
KH+ ,Na+ <−12.83 −11.51 −10.43 −8.83
pot
KH+ ,K+ −12.18 −10.39 −8.86 −8.23
pot
KH+ ,Li+ <−11.72 <−11.72 −11.12 −9.95

(c) Silicone rubber

pot
KH+ ,Na+ −11.8 −11.5
pot
KH+ ,K+ −11.5 −8.9
pot
KH+ ,Li+ −11.6 −11.2
146 P. Upreti et al. / Talanta 63 (2004) 139–148

3.5. Silicone membranes 6.00

5.86

In view of the small but nevertheless significant selec- 5.80

Process cheese
tivity losses of PVC-based membranes upon exposure to 5.59
5.60
cheese, we explored the suitability of alternative membrane 5.49 5.50 5.48 5.48
matrices. Since silicone rubber membranes were previously
pH

reported for measurements of bodily fluids with significant
protein and lipid components [3,4], PVC-free silicone mem-
branes were considered as alternatives.
Silicone rubber matrices have been optimized for elec-
troanalytical analysis with the addition of higher amounts
of lipophilic ionic additives and an appropriate plasticizer
[42]. Membranes without plasticizer were reported to ex-
hibit super-Nernstian responses for pH < 7 [42]. As Table 1
shows, the silicone membranes exhibited selectivities inter-
mediate between PVC-based R3 N/DOS and R3 N/oNPOE
membranes. Since these membranes contain a small amount
of DOS (10%, w/w) to reduce their electrical resistance, it
is not surprising that they do not exhibit much better se-
lectivities than corresponding PVC/DOS membranes, which
contain the same plasticizer, albeit in a much higher con-
centration. Membranes with Nernstian response were then
exposed to 20% Process cheese emulsion for 48 h and resis-
tances and selectivities were measured before and after expo-
sure. After exposure to cheese, the discrimination of potas-
pot
sium deteriorated from KH+ ,K+ −11.5 to −8.9, making sil-
icone membranes slightly less selective than the PVC-based
R3 N/DOS and R3 N/oNPOE membranes. These results show
that the plasticized silicone membranes behave similarly as
PVC membranes.
3.6. R3 N membrane electrodes for pH measurements
in cheese
Since the above experiments indicated a more severe loss
in selectivity upon continued exposure of R3 N membranes to
100% than to 20% Process cheese, and also because blended
but undiluted Process cheese samples have a high viscos-
ity and are somewhat more difficult to handle when using
conventional electrode bodies, it became of interest to know
whether moderate dilution of cheese affects its pH. Although
buffering properties of cheese due to the presence of pro-
teins, phosphates, and citrates has been well established, it
has not been reported up to what level a cheese sample can be
5.40 5.34
5.20 Cheddar cheese
5.08
5.01
4.98 4.99
5.00
4.80
0 20 40 60 80 100
% Cheese (w/w)
Fig. 6. Effect of dilution on pH of cheese diluted with water.
trodes to the same 20% Process cheese emulsion, there was
a continuous decrease in the emf as measured with the same
R3 N membrane electrode, whereas the pH as measured with
a glass electrode remained perfectly constant. To explore
this further, a similar experiment was conducted in which
R3 N membrane electrodes were also exposed repeatedly to
the same cheese emulsion, but before every exposure, a new
response curve was measured. It was found that there was
a consistent shift of the response curve after every exposure
to the cheese sample (Fig. 7). Evidently, this drift cannot be
explained by interfering ions.
A possible explanation for this drift seemed to be the
extraction of neutral compounds such as lipids or hydropho-
bic peptides from cheese samples into R3 N membranes,
leading to an interaction of those compounds with the
non-protonated R3 N ionophore and, concomitantly, an in-
crease in the concentration of the free H+ concentration in
the membrane. If this hypothesis is true, more accurate pH
determinations are expected when a R3 N membrane is con-
ditioned by complete equilibration with a cheese sample,
because this is expected to result in the presence of an equi-
librium concentration of those neutral compounds in the
sensor membrane during the measurement of the response
320
1
300
2
emf (mV)

280
diluted without changing its pH. Therefore, we determined 3
the pH of cheese samples at various levels of dilution with 260
4
water using a pH glass electrode. Our results illustrate that
240
the pH of Process and Cheddar cheese remains unchanged
upon diluting up to 30% (w/w) (i.e. cheese:water = 70:30 220
on a weight basis; see Fig. 6). These results show that the
200
pH in pure Cheddar and Process cheese can be determined
4.25 4.75 5.25 5.75 6.25
upon blending and moderate dilution with water.
pH
The pH of diluted cheese samples was then measured
with R3 N membranes. In preliminary measurements, it was Fig. 7. Shift of the response curve for R3 N/oNPOE on repetitive exposure
observed that on repeated exposure of R3 N membrane elec- to 20% Process cheese.
 P. Upreti et al. / Talanta 63 (2004) 139–148
Table 3
Comparison of pH of milk and diluted cheese samples as measured with
glass and polymeric membrane electrodes
Electrode Milk 20% Process 70% Process
cheese cheese
Glass 6.71 (±0.004) 5.93 (±0.013) 5.62 (±0.015)
R3 N/DOS 6.76 (±0.019) 5.68 (±0.023) 5.63 (±0.265)
R3 N/oNPOE 6.70 (±0.011) 5.92 (±0.085) 5.77 (±0.129)
Silicone rubber 6.13 (±0.017) n.d. n.d.
Standard deviations indicated in parentheses.
curves. In view of this, we desensitized R3 N membrane
electrodes overnight (∼8 h) in a 20% Process cheese emul-
sion. Response curves were then made again before each of
several exposures to a 20% Process cheese emulsion. The
pH values thus observed show that there was no significant
difference between the pH determined with R3 N/oNPOE
electrodes and the pH measured with a glass electrode
(Table 3). The pH of the 20% Process cheese emulsion was
found to be 5.93 ± 0.013 as determined with the glass elec-
trode and 5.92 ± 0.085 as determined with the R3 N/oNPOE
electrodes. Not surprisingly, the pH response curves this
time overlapped completely (data not shown) and lacked
the drift observed with R3 N membranes that had not been
desensitized (as shown in Fig. 7). An apparently similar
effect of beneficial desensitization was previously reported
for the use of calcium ISEs in surfactant solutions [43].
Since the pH of 70% Process cheese and undiluted cheese
are the same, we similarly measured the pH in 70% Process
cheese using desensitized R3 N membranes (Table 3). The
standard deviation for the pH of a 70% Process cheese emul-
sion was found to be higher than for the 20% Process cheese
emulsion, presumably indicating a larger interference from
those neutral interferents.
3.7. pH measurements in milk
The promising results with R3 N membranes and diluted
cheese led to the question of whether R3 N membranes can
be used to measure the pH in milk. When cheese is made
utilizing automated-equipment, online monitoring of the
pH of milk in the cheese vat is important and desired by
cheese-makers as an indicator of acid development. The pH
changes in milk during the initial steps of cheese-making
can affect the composition and functional properties of the
final product (cheese) based on its influence on equilibria
between complexed and free forms of inorganic cations and
anions [7,44,45].
To study the feasibility of pH measurements in milk,
Grade A pasteurized, homogenized, Vitamin D-enriched
milk was obtained from a local supermarket and the pH was
measured five times directly with glass and ionophore-based
electrodes (Table 3). The resulting pH value as measured
with a R3 N/oNPOE membrane was 6.70 ± 0.011, which
is in excellent agreement with the value of 6.71 ± 0.004
as measured with a glass electrode (Table 3). Results with
147
a R3 N/DOS ISE were nearly as good, while the silicone
membrane electrodes performed poorly (Table 3). These
results indicate the suitability of PVC-based ISEs for mea-
surements of pH and possibly other ions in milk.
4. Conclusions
Whether glass or polymeric membrane electrodes are
more suited for clinical samples has been debated, but few
explicit statements can be found in the literature. While it
is clear that a high impedance and fragility are disadvan-
tages of glass electrodes, protein adsorption to glass and
polymeric membranes has never been compared in detail.
Considering recent advances in the understanding of poly-
meric membrane ISEs [46], it is evident that the adsorption
of a thick layer of proteins onto an electrode may affect pH
measurements either by increasing the response time due
to slow diffusion of the sample to the sensor surface, or by
producing a local pH that differs from the pH in the bulk of
the sample. Whether these effects are more pronounced for
glass or polymer membrane electrodes has been debated.
For example, Simon et al. reported that the performance of
a glass catheter electrode deteriorated rapidly, “probably as
a result of protein adsorption”, while a membrane electrode
operated for several hours without deterioration [9]. On the
other hand, Meyerhoff commented that protein adsorption
is less pronounced in the case of glass [47].
This study demonstrates that glass electrodes perform
well for measurement of pH in milk and cheeses. It shows
that liquid junction potentials at a reference electrode with
a 3 M KCl bridge electrolyte and Cheddar/Process cheese
samples are negligible if a double-junction sleeve-type elec-
trode is used. While the protein rennet casein does not cause
any drifts or serious changes in the response times of sol-
vent polymeric membranes, over time there appears to be
some extraction of electrically neutral compounds, such as
lipids or hydrophobic peptides, into polymeric membranes.
It is ironical that problems associated with the extraction of
cheese components into ISE membranes can be overcome
precisely by allowing this extraction to occur. After desen-
sitizing the membranes in a cheese emulsion, accurate pH
measurements with R3 N/oNPOE membranes in diluted Pro-
cess cheese and milk can be made. However, the standard
deviation is larger than for measurements with a glass elec-
trode, in particular at high cheese concentrations.
While further advances in membrane technology are re-
quired for direct pH measurements with ionophore-based
ISEs in undiluted cheese, the necessary improvements
may be comparatively minor in nature. This is particularly
promising not only because it may make the replacement
of the fragile pH glass electrode with an ionophore-based
electrode possible, but also because it opens a venue to
measurements for other ions in dairy and other food prod-
ucts using ionophore-based ISEs, for which, typically, other
sensors do not exist. For example, the direct measurement
148 P. Upreti et al. / Talanta 63 (2004) 139–148
of free calcium ions in milk and cheese would be of con-
siderable interest for cheese manufacturing. Approaches
to measure calcium in dairy products and to exclude
lipids from ionophore-based membranes by use of novel
membrane matrices are presently being explored in our
laboratory.
Acknowledgements
We thank Dow Corning Corporation, Midland, MI for
the gift of silicone rubber sealants and Rohit Kapoor, De-
partment of Food Science and Nutrition, University of
Minnesota, for assistance in obtaining the cheese samples.
Thomas Lawson of Lawson Labs is gratefully acknowl-
edged for timely and very helpful assistance. We greatly
appreciate the help of Lynne Johnsrud in the preparation of
this manuscript.
References
[1] E. Bakker, P. Bühlmann, E. Pretsch, Electroanalysis 11 (1999) 915.
[2] Y. Umezawa, CRC Handbook of Ion-Selective Electrodes: Selectivity
Coefficients. CRC Press, Boca Raton, FL, 1990.
[3] P. Bühlmann, E. Pretsch, E. Bakker, Chem. Rev. 98 (1998) 1593.
[4] E. Bakker, P. Bühlmann, E. Pretsch, Chem. Rev. 97 (1997) 3083.
[5] P. Bühlmann, M. Hayakawa, T. Ohshiro, S. Amemiya, Y. Umezawa,
Anal. Chem. 73 (2001) 3199.
[6] N.F. Olson, S. Gunasekaran, D.D. Bogenrief, Netherlands Milk Dairy
J. 50 (1996) 279.
[7] L.E. Metzger, D.M. Barbano, P.S. Kindstedt, J. Dairy Sci. 84 (2001)
1357.
[8] S. Marchesseau, E. Gastaldi, A. Lagaude, J.-L. Cuq, J. Dairy Sci.
80 (1997) 1483.
[9] W. Simon, D. Ammann, P. Anker, U. Oesch, D.M. Band, Ann. N.Y.
Acad. Sci. 428 (1984) 279.
[10] N.F. Olson, Marschall Italian and Specialty Cheese Seminars, 1975,
p. 15.
[11] A.H.J. Maas, H.F. Weisberg, R.W. Burnett, O. Müller-Plathe, P.D.
Wimberley, W.G. Zijlstra, R.A. Durst, O. Siggard-Andersen, J. Clin.
Chem. Clin. Biochem. 25 (1987) 281.
[12] L.V. Coveney, Scientific and Technical Surveys No. 118, Leather-
head Food R.A., British Food Manufacturing Industries Research
Association, 1980.
[13] J. Comer, in: R.D. King (Ed.), Developments in Food Analysis
Techniques, vol. 1, 1978, pp. 197–222.
[14] Y. Kaneda, T. Kanamori, M. Iwaida, J. Hyg. Chem. 23 (1977) 301.
[15] E. Ayranci, Ö. Balci, Die Nahrung 37 (1993) 395.
[16] M.A. González, A.J. Aller, R. Pardo, E. Barrado, L. Deban, Elec-
troanalysis 3 (1991) 439.
[17] G.S. Craven, M.C. Griffith, Austr. J. Dairy Technol. 32 (1977) 75.
[18] A.W. Randell, P.M. Linklater, Austr. J. Dairy Technol. 27 (1972) 51.
[19] P. Bühlmann, S. Amemiya, S. Yajima, Y. Umezawa, Anal. Chem.
70 (1998) 4291.
[20] P. Schulthess, Y. Shijo, H.V. Pham, E. Pretsch, D. Ammann, W.
Simon, Anal. Chim. Acta 131 (1981) 111.
[21] D. Ammann, F. Lanter, R.A. Steiner, P. Schulthess, Y. Shijo, W.
Simon, Anal. Chem. 53 (1981) 2267.
[22] O. Dinten, U.E. Spichiger, N. Chaniotakis, P. Gehrig, B. Rusterholz,
W.E. Morf, W. Simon, Anal. Chem. 63 (1991) 596.
[23] E. Lindner, V.V. Cosofret, R.P. Buck, T.A. Johnson, B.R. Ash, M.R.
Neuman, W.J. Kao, J.M. Anderson, Electroanalysis 7 (1995) 864.
[24] D. Liu, M.E. Meyerhoff, H.D. Goldberg, R.B. Brown, Anal. Chim.
Acta 274 (1993) 37.
[25] P. Anker, D. Ammann, W. Simon, Mikrochim. Acta I (1983) 237.
[26] E. Lindner, V.V. Cosofret, S. Ufer, R.P. Buck, R.P. Kusy, R.B. Ash,
H.T. Nagle, J. Chem. Soc., Faraday Trans. 89 (1993) 361.
[27] C. Espadas-Torre, E. Bakker, S. Barker, M.E. Meyerhoff, Anal. Chem.
68 (1996) 1623.
[28] U. Oesch, Z. Brzózka, A. Xu, W. Simon, Med. Biol. Eng. Comput.
25 (1987) 414.
[29] E. Bakker, A. Xu, E. Pretsch, Anal. Chim. Acta 295 (1994) 253.
[30] W.E. Morf, The Principles of Ion-Selective Electrodes and of Mem-
brane Transport, Elsevier, New York, 1981.
[31] U. Oesch, W. Simon, Anal. Chem. 52 (1980) 692.
[32] D. Ammann, E. Pretsch, W. Simon, E. Lindner, A. Bezegh, E.
Pungor, Anal. Chim. Acta 171 (1985) 119.
[33] E. Bakker, E. Pretsch, P. Bühlmann, Anal. Chem. 72 (2000) 1127.
[34] G.G. Guilbault, R.A. Durst, M.S. Frant, H. Freiser, E.H. Hansen, T.S.
Light, E. Pungor, G. Rechnitz, N.M. Rice, T.J. Rohm, W. Simon,
J.D.R. Thomas, Pure Appl. Chem. 48 (1976) 127.
[35] R.P. Buck, E. Lindner, Pure Appl. Chem. 66 (1995) 2527.
[36] P.C. Meier, Anal. Chim. Acta 136 (1982) 363.
[37] H. Galster, pH Measurement: Fundamentals, Methods, Applications,
Instrumentation, VCH Publishers, Inc., New York, 1991, pp. 33, 90,
91.
[38] A. Mostert, P. Anker, H.-B. Jenny, U. Oesch, W.E. Morf, D. Ammann,
W. Simon, Mikrochim. Acta I (1985) 33.
[39] N. Oyama, T. Hirokawa, S. Yamaguchi, N. Ushizawa, T. Shimomura,
Anal. Chem. 59 (1987) 258.
[40] T. Bührer, P. Gehrig, W. Simon, Anal. Sci. 4 (1988) 547.
[41] G.G. Guilbault, IUPAC Inf. Bull. 1 (1978) 69.
[42] E. Malinowska, V. Oklejas, R.W. Hower, R.B. Brown, M.E. Meyer-
hoff, Sens. Actuators B33 (1996) 161.
[43] R.A. Llenado, Anal. Chem. 47 (1975) 2243.
[44] L.E. Metzger, D.M. Barbano, M.A. Rudan, P.S. Kindstedt, J. Dairy
Sci. 83 (2000) 648.
[45] L.E. Metzger, D.M. Barbano, P.S. Kindstedt, M.R. Guo, J. Dairy
Sci. 84 (2001) 1348.
[46] E. Bakker, P. Bühlmann, E. Pretsch, Talanta, in press.
[47] M. Meyerhoff, Clin. Chem. 36 (1990) 1567.